DEALIGNMENT DECADES ON:

PARTISANSHIP AND

PARTY SUPPORT IN

GREAT BRITAIN,

1979-1996

DISSERTATION

Presented to the Graduate Council of the

University of North Texas in Partial

Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

By

Karl Ka-yiu Ho, M.S.

Denton, Texas

December, 1996

37<?

AlQll /Vo. V v o y

DEALIGNMENT DECADES ON:

PARTISANSHIP AND

PARTY SUPPORT IN

GREAT BRITAIN,

1979-1996

DISSERTATION

Presented to the Graduate Council of the

University of North Texas in Partial

Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

By

Karl Ka-yiu Ho, M.S.

Denton, Texas

December, 1996

37<?

AlQll /Vo. V v o y

f/bt

Ho, Karl Ka-yiu, Dealignment Decades On: Partisanship and Party Support in

Great Britain. 1979-1996. Doctor of Philosophy (Political Science), December, 1996,217

pp., 27 tables, 17 illustrations, references, 103 titles.

This dissertation surveys electoral change in Great Britain during the period

between 1979 and 1996. It analyzes the long-term factors and the short-term dynamics

underlying the evolution of three aspects of the electorate: party identification, voting

intentions and party support in inter-election periods. Drawing on cross-sectional and

panel data from the British Election Studies and public opinion polls, I investigate the

impacts of long-term socialization and short-term perceptions on voters' political

decisions. I hypothesize that, over the last four elections, perceptual factors such as

evaluations of party leaders and issues, particularly economic concerns, emerged as the

major forces that account for the volatility in electoral behavior in Britain. Accordingly,

this study is divided into three sections: Part I probes into the evolution in party

identification across age cohorts and social classes as illustrated in trends in partisanship.

Part II focuses on changes in voting intentions as affected by perceptual factors and party

identification. Part III investigates the public's support for governing parties by analyzing

the dynamics of aggregate party support during inter-election periods.

TABLE OF CONTENTS

Page

LIST OF TABLES v

LIST OF ILLUSTRATIONS vi

Chapter

I. INTRODUCTION 1

Introduction Literature Review

The Class Voting Model The Issue Voting Model New Direction in Issue Voting

Organization of the Study Conclusion Endnotes

II. GENERATIONAL REPLACEMENT, SOCIAL CHANGE AND TREND OF PARTISANSHIP 19

Introduction Measuring Change in Partisanship Partisanship in the Post-dealignment decade Explaining Trend of Partisanship Conclusion: Differential Dealignment and Realignment of Degree Endnotes

III. PERCEPTION MATTERS - LEADER IMAGE, ECONOMIC EVALUATIONS AND ELECTORAL CHOICE 40

Introduction Short-term Forces Overlaying Long-term Factors The "Aged" Class Alignment Modeling Partisanship in the Post-Dealignment Era Conclusion Endnotes

m

IV. NEW DIRECTIONS IN ISSUE VOTING 67

Introduction Measuring Issue in British elections A Graphical Analysis A Statistical Analysis Modeling Multi-party Election Conclusion Endnotes

V. THE "PRESIDENTIALIZATION" OF PRIME MINISTERS: MODELING PARTY SUPPORT BETWEEN ELECTIONS 102

Introduction Leader Matters, Leader Matters Not? Three Prime Ministerial Models Modeling Leadership Effects on Party Support A Hyper-parameter Model on Prime Ministerial Approval

Analysis of Intervention Analysis of Time-varying parameter

Conclusion: The Presidentialization Thesis in Perspective Endnotes

6. CONCLUSION 130

Endnotes

APPENDIX 137

BIBLIOGRAPHY 210

IV

LIST OF TABLES

2.1 Voter Shares by Party and Voting Turnout in Great Britain, 1945-92 2.2 Strength of Party Identification in Britain, 1945-92 2.3 Mean Strength of Party Identification in Britain, 1964-1992 2.4 Party Identification in Britain, 1964-1992 2.5 Mean Strength of Party Identification by Age Cohort, 1987-92 2.6 Patterns of Turnover in Party Identification: Changes in Party Affiliation, 1987-92

Panel Survey 3.1 Party Identification and Social Class: 1964,1979-92 3.2 Changes in Social Class Structure, 1964-92 3.3 Alford Indices for Class Voting, 1964,1979-92 3.4 Mean Strength of Party Identification by Class and Party, 1964-92 3.5 Changes in Party Identifications and Partisan Strength Among Natural- and

Cross-Class Identifiers, 1987-92 3.6 Two-Stage Least Square Analysis of Party Identification in Great Britain, 1987-92

Panel 3.7 Standardized Coefficients of Two-stage Least Square Models of Party

Identification, 1987-92 Panel 4.1 Mean Issue Positions of Respondents and Mean Perceived Party Positions on Five

Issues in British General Election Survey, 1992 4.2 Binary Logit Model: Issue Voting for Political Parties in Great Britain, 1992 4.3 Encompassing Test, Proximity Model versus Directional Model on Issue Voting,

1987-92 4.4 Encompassing Test, Proximity Model versus Directional Model on Issue Voting,

1983-87 4.5 Different Impacts of Issue Variables on Political Parties 4.6 Multinomial Logit Model: Multi-party Vote Choice in Great Britain, 1987-92 4.7 Multinomial Probit Model: Multi-party Vote Choice in Great Britain, 1987-92 5.1 Dickey-Fuller Unit Root Tests for Stationarity of Conservative Party Support,

Prime Ministerial Approval and Subeject Economic Evalutions, 1979m7-1996m4 5.2 Error Correction Models of Conservative Party Support, 1979m7-1996m4 5.3 Models of Prime Ministerial Approval, 1979m8 to 1996m4 5.4 Encompassing Tests: The Effects of Four Economic Evaluation Variables on

Prime Ministerial Approval 5.5 Weak Exogeneity Test of Prime Ministerial Approval and Conservative Support 5.6 Structural state-space Models of Party Support in Great Britain, 1979m7-l 996m4 5.7 Time-varying Parameter Model of Leadership effect on Party Support

LIST OF ILLUSTRATIONS

2.1 Strength of Party Identification, 1964-92 2.2 Distribution of Party Identifiers, 1964-92 2.3a Partisan Strength of Conservative, 1964-92 2.3b Partisan Strength of Labour, 1964-92 2.3c Partisan Strength of Liberals, 1964-92 2.4 Mean Strength of Party Identification, 1964-92 2.5 Mean Strength of Party Identification by Age Cohort, 1987-92 4.1 Three-dimensional Plot of Proximity Model of Issue Voting 4.2 Three-dimensional Plot of Directionl Model of Issue Voting 4.3 Distribution of Views on Policy Alternatives in Britain, 1992 4.4 Party Vote Shares and Positions of Voters on Issues in Great Britain, 1992 5.1 Government Popularity and Prime Ministerial Approval in Great Britain, 1979-

1996 5.2 A Structural Break at the Currency Crisis: CUSUM Graph of Model Residuals 5.3a Prime Ministerial Approval and PMSAT Hyperparameter: Personal Expectation

Model 5.3b Prime Ministerial Approval and PMSAT Hyperparameter: Personal Retrospection

Model 5.3c Prime Ministerial Approval and PMSAT Hyperparameter: National Expectation

Model 5.4 Prime Ministerial Approval and PMSAT Hyperparameter: National Retrospection

Model

VI

CHAPTER I

INTRODUCTION

1.1 Introduction

The British polity has been widely applauded as the prototype of representative

democracy. Distinguished by its enduring political stability (Ranney 1956, Epstein

1980), the country has a primarily two-party system that is characterized by a stable

social structure and durable class-party ties. This party system is deemed as the

cornerstone of partisan equilibrium that contributes to the political stability the country

has enjoyed throughout the twentieth century (Converse 1969, Abramson 1992). During

the postwar decades prior to 1979, it was argued that the Conservative and Labour parties

acquired approximate balance in governance1 and public support rooted in salient class

cleavage. An enduring demographic structure operated to maintain a system that relied

on generational replacement as the primary motor of political change (Butler and Stokes

1969).

Toward the end of the 1970s, prolonged economic malaise subjected this

seemingly stable system to tremendous stress and strains. The winter of discontent in

1979 — a landmark in British history for workers' strikes, price hikes and high-standing

unemployment ~ precipitated the defeat of the Labour government led by James

Callaghan. The new Conservative prime minister, Margaret Thatcher, entered Downing

1

No. 10 with a portfolio filled with policies that replaced the then prevalent "Butskellite"

consensus. Beginning that same year, the Thatcherist government embarked on an era

that brought forth a series of political and social reforms that effected a "sea change" in

the British polity.

Mrs. Thatcher's espousal of a free market economy, as exemplified by her

privatization policies and reduction in direct taxation, successfully "revolutionized" the

political and social structure of the country (Jenkins 1988, King ed. 1993). The

government's reverse social engineering in housing policies, for instance, significantly

reshaped British society and yielded a considerable "political dividend" for the Tories

(Crewe 1993). These policies were typified by the council house sale program that

delivered home ownership to more affluent working class families. In 1950, only 29

percent of the population owned their homes; the figure soared to 67 percent in 1990.

The surge in the number of homeowners helped to undermine the social base of the

Labour party and resulted in a substantial swing to the Conservative party2 (Heath et al.

1991).

On the economic front, Mrs. Thatcher's industrial policies aimed at the trade

unions lessened the unionized workers' powers and influence. A series of employment

acts systematically removed the traditional immunities and prerogatives enjoyed by labor

unions in strikes and member recruitment. These policy instruments, coupled with the

rise in unemployment, helped slash union membership by 3 million between 1979 and

1991 (Crewe 1993). As a consequence, the problem of strikes, which had plagued the

3

country in the 1970s, was scaled down to insignificance throughout much of the 1980s

(Hyman 1992, Crewe 1991).

The trade unions' diminishing influence was not unexpected given the structural

changes in the British economic environment. Transformations of the economy from

manufacturing and large-scale plant productions to services and self-employed, small

businesses led to a decline in manual worker population. In 1950, manual workers made

up a majority (68 percent) of the working population; in the 1980s, the figure dropped to

48 percent. Corresponding to the decline, union membership shrank from 50 percent of

the workforce in 1979 to 36 percent in 1990 (Butler and Kavanagh 1992). It has been

argued that this contraction of the working class contributed to a movement toward

liberal values on government economic policies which has enormously benefited the

Conservatives at the expense of Labour (Heath, Jowell and Curtice 1985).

In the 1980s and 1990s, new political and economic developments, including the

rise of a new political party,3 the outbreak of foreign war,4 the growing salience of the

European Community and, in particular, the issue of a unified European currency, has

brought new challenges to the British political environment and the electorate. Perhaps

most importantly, as traditional social class attachments have diminished, British voters

have resorted to an issue-oriented approach when making political choice and expressing

political preferences (Harrop 1977; Alt 1979; Clarke, Stewart and Zuk 1986).

Even prior to 1980, the growing importance of issues among British voters, and

the concomittant decline of class voting, led to a period of electoral volatility of voting

4

behavior (Franklin 1985). The new voting pattern focusing on issues and short-term

factors continued in the 1980s. The British electorate was evolving from voting based on

social structural factors to calculating utilities corresponding to political and economic

evaluations. Three-quarters of the British voters, as survey evidence suggests, are no

longer anchored by partisan allegiance or family-class-party ties in the general elections

after 1979 (Rose and McAllister 1986). Key to this trend has been the general decline of

partisanship among the British voters. The erosion of partisan loyalties ~ as described by

the "dealignment of degree" in party identification — became an established pattern

among British voters in the 1980s (Clarke and Stewart 1984). The decline in partisan

ties, in both strength and durability, became manifest in the elections beginning in the

1970s (Rose 1974; Rose and McAllister 1986). Short-term factors such as

macroeconomic conditions (e.g., fluctuations in unemployment, interest rates and

inflation), party leader images and political shocks (e.g., international war and political

scandals) have usurped the predominance of social class cleavages. Economic factors,

specifically, voters' evaluations of the state of the national economy and their personal

financial situations, have emerged as major factors governing public support for political

parties and coloring the electorate's perception and support for political leaders (Lewis-

Beck 1988; Clarke etal. 1992).

In recognition of these changes, this study investigates three aspects of the

evolution of the British electorate: party identification, voting choice and support for

governing parties in the periods between general elections. As for the first two aspects,

5

the focus is on probing changes in partisanship and electoral choice since 1979 and

identifying factors that contribute to such variations. Specifically, it analyzes the effects

on party identification and voting choices of i) sociological (long-term) factors as

exemplified by such variables as occupational class, home ownership, union membership

and education as well as ii) psychological (short-term) factors such as perception of party

leader images, position on issues and evaluations of economic conditions. Drawing upon

individual-level data from the British National Election Studies (BES) surveys, I

investigate the relationship between voters' socio-demographic and psychological

attributes and political choices. After examining partisanship and electoral choice at the

individual level, this study utilizes time series data covering the 1979-1996 period to

examine the dynamics of governing party support at the aggregate level.

1.2 Literature Review

a. Electoral Research in Britain

Research on voting behavior and elections in Britain can be traced back to the

1940s when electoral studies were an enterprise of historians. The pathfinder of the

British General Elections Studies series, R. B. McCallum, pioneered by collecting

historical information on election campaigns and their impact on electoral outcomes. The

Nuffield series-named after their connection with Nuffield College, Oxford-had its

origins in 1945 when the focus was on examining relationships among issues, political

parties and leaders, coverage of mass media and ultimately, the election outcome. This

6

focus has been maintained by David Butler and his colleagues over the intervening

decades.

The quantitative aggregate level data gathered from the studies in the 1940s and

1950s concerned voter turnout and the distribution of votes and parliamentary seats on

the national and constituency level. These aggregate data provided a valuable picture of

the general public's political complexion and cross-constituency and inter-election

changes in collective voting outcome. However, analyses based on such data are

inherently limited in ability to account for the decision process of individual voters and

the effects of other intervening factors such as party allegiance, personal attributes, social

economic status and personal judgment of party leaders and issue concerns.

Awareness of the limitations of the aggregate level data prompted the introduction

of electoral surveys in the early 1950s. Using the sampling methodology developed in

the American election studies in the late 1940s, the first British election survey was

launched in 1950 studying the voting behavior of the Greenwich constituents during the

general election in the same year. The first of its kind, the survey interviewed 914

Greenwich citizens to investigate the relationship between political choice and a series of

sociological factors. The results were reported in How People Vote (Benney, Gray and

Pear 1956).

The Greenwich study ushered in a proliferation of a number of other electoral

surveys focusing on local elections or specific issues (Milne and Mackenzie 1954,1958;

Trenaman and McQuail 1961; see Denver and Hands 1992). Although election surveys

7

in the 1950s represented important advances in studying voting behavior, they were

confined to localities or several constituencies. It was not until 1963 when the first

national election survey was conducted. One year prior to the 1964 general election,

David Butler and Donald Stokes embarked on a national survey of voting behavior,

which has been followed by a series of similar studies in every subsequent general

election5 (Denver and Hands 1992).

Most of the electoral research in Britain was guided by two major voting models.

Both models emphasized long-term forces. The simple "class voting" model prevalent in

the 1950s and 1960s suggests that a voter's electoral choice is determined by long-term

social class factors. Its successor, the "social psychological" model, combines political

socialization theory with the Michigan party identification model, and argues that

electoral choice is determined by the long-term factor of social class, which, in turn,

determines partisan allegiance (Butler and Stokes 1969,1974; See also Denver and

Hands 1992). These partisan attachment tend to be durable forces in the set of

psychological factors affectiing electoral choice.

In their pioneering work on electoral changes in Great Britain, Political Change in

Britain: Forces Shaping Electoral Choice, Butler and Stokes analyze political change by

focusing on social class, party identification and electoral choices of the British voters.

The 1969 edition of this book, followed by a second edition in 1974, was based on data

collected in three previous national surveys (1963,1964 and 1966 with additional survey

data from 1970 in the second edition). The authors carefully segregate the short-term

8

changes from the long-term evolution of a relationship between social attributes and

political behavior. In their typology, changes can be categorized into three fundamental

kinds. The first of these involves the generational replacement that induces attitudinal

and behavioral differences among age cohorts. As Butler and Stokes argue, the

"inexorable processes of birth and maturation on one hand and aging and death on the

other" provide the sources of cumulative changes as the political complexion of the

electorate alters little by little every year as a result of replacement of cohorts (Butler and

Stokes 1969: pp. 4-5). The role of this long-term factor in effecting political change, in

the authors' views, is of fundamental significance as compared with other political

variables.

The second type of political change entails mutation in the elector's attachment to

political parties. Embedded in the endurance of party alignments, as the authors

articulate, is primarily the influence of social class cleavage. It is unlikely for a partisan

to convert once a party identity is developed and behind this political allegiance among

voters is the social class cleavage that has been bifurcating the British electorate since the

rise of the Labour Party early in this century. Given the stability of the class-party

cleavage, the demographic renewal of the electorate will therefore take years to yield

meaningful changes in aggregate political alignment.

The third type of electoral variability in Britain arises from the ongoing march of

political issues and political events that constitute short-term forces on electoral behavior.

Such stimuli, which the authors argue have a stronger impact on younger voters,

9

nevertheless do not affect the lasting party alignment in the long run. The image of a

party leader, the surge of a hotly-debated issue or a controversial fiscal policy are subject

to diminution of the "more enduring bases of party strength" and the "more basic

cleavages [that] cut through the electorate."

During the 1970s, the British polity experienced a period of electoral volatility as

the linkage between class and party began to attenuate. The dealignment of party support

evident in election survey findings gave rise to a new school of electoral studies

challenging the orthodoxy of the class voting model. A new generation of psephologists

began to cast doubt on the class voting model that resorted to demographic changes in

accounting for electoral change. Instead of seeking the source of political change in long-

term sociological factors, scholars embarked on a stream of research focusing on short-

term forces as exemplified by issue concerns and economic evaluations (Sarlvik and

Crew 1983, Franklin 1985, Whiteley 1986). This "issue voting" model argues that voters

are becoming more conscious of government policies and tend to base their vote on their

assessments of government performance. In this model, short-term forces provide the

primary sources of electoral change.

The new conventional wisdom of issue voting in the 1980s did not prevail without

controversy. In the mid-1980s, debate focused on the longstanding topic of class and

partisan dealignments. Social class cleavages, and, in turn, long-term partisan

attachments, as some commentators argue, were still "alive" and acting as major forces

determining electoral behavior (Heath, Jowell and Curtice 1985; Rose and McAllister

10

1986). Having resurrected the class factor, scholars like Heath and his colleagues

adopted a new "interactionist" approach to interpret political change in the 1980s (e.g.

Heath, Jowell and Curtice 1985). They argued that social class was still the most

prominent factor defining partisanship and voting choice. Class-based ideologies and

political principles provide cues for voters to filter information when choosing between

parties, while parties interact to help shape and redefine political support in the electorate

(Heath, Jowell and Curtice 1985: p. 174).

The protracted and sometimes heated debates on class voting versus issue voting

have yet to arrive at a unified model to explain relationship between long-term

socialization effects and short-term perceptions in affecting electoral change. The two

models are divided on the conceptualization and role of party identification and its

electoral affinity with class and voting choice (Denver and Hands 1992). Drawing on

individual- and aggregate-level data, this study endeavors to resolve the differences of the

two models by integrating long-term and short-term forces in an explanation to the

evolution of party support in the British electorate.

a. The Class Voting Model

During the postwar decades, partisanship and electoral choice in Great Britain

were largely a function of social cleavages and family/parental influence. For example,

the 1964 British election was a typical class-based group voting as a vast majority (up to

99 %) of working class electors voted for Labour (Franklin 1985). As Peter Pulzer

(1967) reckons, "(C)lass is the basis of British politics; all else is mere embellishment and

11

detail" (p. 98).

Empirical studies of elections after 1964 portrayed a different picture of the

British electorate. As Butler and Stokes note, the social mechanisms and class forces that

had driven the class voting pattern had "passed their peak" in influencing election results

(Butler and Stokes 1974). Despite controversies over the linkage among class, party

identification and vote, the decline of class voting has become widely accepted in recent

research on partisanship and electoral choice (Miller 1978; Crewe 1983; Crewe 1986;

Heath, Jowell and Curtice 1985). Subsequent studies have found that the dominant class

voting pattern in the 1960s gave way to the significance of psychological or perceptual

factors such as evaluation of leaders and issue preference in selection of government (Alt

1979; Franklin 1985).

b. The Issue Voting Model

The new calculus of electoral choice prevalent in the 1970s and the 1980s was

partially a consequence of cohort replacement and general social changes that were not

confined to the British environment (Abramson 1992). Parallel trends were observed

among most Western industrial democracies including Germany, the United States and

Canada (Franklin et al. 1992, Kornberg and Clarke 1992). During these two decades,

erosion of the class-party-vote linkage that connected social class cleavage with

perceptions of parties' issue priorities gave way to a much more flexible view of voters'

issue concerns and party-issue linkages (Franklin 1985). Specifically, the notion that the

working class prioritized employment as a primary concern over others in supporting

leftist parties no longer dominated models of how issues affected voting. Rather, it was

12

argued that new generations of voters perceived issue priorities from different

perspectives. Short-term factors such as movements in macroeconomic indicators, viz.

inflation, interest rate and unemployment, became significant indices for voters in making

electoral decisions. Economic evaluations—particularly, the "pocketbook" factor, or

concerns for personal financial situations—emerged as a highly salient issue that

preoccupied most voters (Sanders et al. 1987).

In the early 1990s, the prolonged economic malaise led the public to pay renewed

attention to an old issue: the ailing British economy. For voters in the 1992 election, the

problem of the economy was no longer bifurcated into individual aspects as "iconized"

by inflation and unemployment. The "valence-issue" election of 1992, which focused on

the widely-perceived national recession, reflected the fact that voters were concerned

about the future and how competent were the parties to manage the economy in the years

ahead. Parties strove to convince voters that they were able to handle matters better than

their opponents (Newton 1993). As political analysts suggest, it was such prospective

evaluations that prompted the electorate to make a last moment swing to from Labour to

the Conservatives in April 1992 (Sanders 1993, Newton 1993).

c. New Directions in Issue Voting: From Spatial to Directional Models?

Against the background of class dealignment in the 1970s, there rose a new

determinant of electoral choice: issue orientations. In this regard, earlier studies of issue

voting focused on the proximities of voters' positions on an issue and their perceived

positions of the parties. The closer the individual's issue position to that of a party, the

more favorably the voter would evaluate the party (Downs 1957; Davis, Hinich and

13

Ordeshook 1970; Enelow and Hinich 1984). Recent empirical research on issue voting in

the United States and Norway argues that this traditional spatial model fails to evaluate

the issue effect on voting decisions. Voters do not view issues in a straightforward

fashion as the model assumes. As Rabinonwitz and Macdonald articulate, the public does

not possess specific policy preference "along a set of issue continua" (1989). At best,

voters know the pros and cons or which side of an issue debate they favor. In light of

this, the authors propose a directional issue position variable, which captures both the

direction and intensity of a voter's stance on issues, as a better surrogate in modeling

issue effect on party identification and voting choice. To investigate the impact of issues

on political choice, this study replicates Rabinowitz and Macdonald's analysis in the

British context and compares the explanatory power of the directional model with the

traditional proximity model. By applying encompassing tests to determine which model

provides a more adequate explanation (Mizon and Richard 1986; Cuthbertson, Hall and

Taylor 1990), the study attempts to distinguish the model that best captures how issues

affect electoral choice.

1.3 Organization of the Study

This dissertation is composed of six chapters that consider the long-term and

short-term forces on party identification, voting choice and party support in inter-election

periods. Individual-level data used in Chapters II, III and IV are taken from the British

Election Studies surveys in the general elections of 1979,1983,1987 and 1992 and the

panel studies administered in the 1983-1987 and 1987-1992 period. Aggregate-level data

14

employed in Chapter V come from the Gallup public opinion poll and OECD economic

indicators. As an introduction, Chapter I has chronicled the social, economic and

political developments of Great Britain during the period between 1979 and 1992, paying

special attention to the factors that foster the structural transformations in British society.

It also provides a framework in guiding the subsequent analyses.

Chapter II to Chapter IV concentrate on investigating i) trends in party

identification of individual voters as reflected in the four BES election surveys from 1979

to 1992 and how partisanship fluctuates across generations; ii) social cleavage factors

such as social class characteristics (occupation, self-defined class and home and share

ownership), union membership that contribute to changes in party allegiance (Rose and

McAllister 1986, Franklin 1985); and iii) perceptual factors that color the attitudinal

disposition of political parties such as leader evaluations and issue evaluations (Clarke

and Stewart 1984, Clarke et al. 1992, Rabinowitz and Macdonald 1989). Analyses in

these chapters focus on the hypothesis that voters' party identification is a function of

both long-term and short-term factors. The former group of variables include a) objective

class attributes like occupational class, home ownership and union membership; b)

demographic attributes like age and education; and c) previous party identification.

Short-term forces comprise e) the voters' issue positions as referenced to those of the

political parties; and f) their evaluation of party leaders.

Specifically, Chapter II compares voters from different age cohorts on: a) effects

of generational replacement, b) lifecycle effects and c) effects of other socio-demographic

factors on the decline of partisanship. This chapter documents trends in partisanship in

15

terms of the strength of voters' attachment to the three national political parties, namely,

Conservative, Labour and the Social Democrats/Liberal Democrats. Chapter III studies

the determinants of the strength and direction of party identification. Contrary to the

classical "Michigan" model (e.g. Campbell et al., 1960), British studies conducted in the

1980s have suggested that party identification is conditioned by a voter's evaluations of

party leaders and issues (Clarke and Stewart 1984, Alt 1984). Accordingly, this chapter

examines the effects of these two variables by testing a series of multivariate models.

These models specify that voters' party identification in the 1992 election is a function of

their previous party identifications (1987), their evaluations of party leaders and his

positions on various issues. Thus, in addition to the leader image and the issue variables,

a lagged dependent variable-party identification in the previous election-is also included

in these party identification models. This inclusion of a lagged party identification

variable is consistent with arguments that voters' partisan attachments at time t are

driven, in part, by their previous (at time t-1) partisan attachments (Fiorina 1981, Clarke

and Stewart 1984).

In Chapter IV, the perspective turns to the voting behavior of individual voters.

To what extent does dealignment in partisanship affect the voting decision of individual

voters? How does the difference in issue positions between voters and parties affect

voting choice? How will the other factors affect the voter's decision at the ballot box?

To address these questions, this chapter focuses on an analysis of individual voting

behavior. The chapter investigates the hypothesis that perceptions of party leaders, issue

position proximities, and party identification are the principal determinants of voters'

16

electoral decisions. A multivariate model is developed to analyze the voting with

particular emphasis on assessing applicability of Rabinowitz and Macdonald's (1989)

directional voting model. As indicated above, I compare the two variants of the issue

voting model, namely, traditional proximity model and directional model using an

encompassing test as usually applied in contemporary econometric analyses.

To examine party support in the periods between general elections, Chapter V

employs time series analysis to follow the trajectory of public sentiments for political

parties since 1979. Using aggregate data collected from Gallup polls on party support, a

multivariate model is tested with special attention on three categories of variables6. They

are: 1) subjective evaluations of the economy; 2) satisfaction of the prime minister and 3)

political interventions such as the Falklands War of 1982 and the currency crisis of 1992.

Chapter VI summarizes the empirical evidence gathered from the preceding

chapters and reconsiders the evolution of party support in the British electorate since

1979.

1.4 Conclusion

After its narrow victory in 1992, the popularity of Conservative Party plummeted

to a historic low in 1995. In the same year, Conservative leader John Major reached a

nadir in prime ministerial approval, earning him both the titles of the "most popular and

most unpopular" leader in this century. Meanwhile, a revitalized Labour Party under the

leadership of Tony Blair has managed to garner tremendous amount of popular support.

While almost every sign augurs a Tory doomsday in the next general election, observers

17

of British politics are ready to recognize that the changeable nature of the British

electorate makes electoral forecasting a very "risky business". The following chapters

consider various aspects of this electorate and the political changes in the past two

decades that have produced this volatility that makes British politics less predictable and

perhaps more fascinating than it was before the "decade of dealignment".

18

Endnotes

1. Between 1945 and 1979, the Conservatives ruled for 17 years in 5 governments and Labour ruled for 18 years in 4 governments.

2. According to Heath et al. (1991), the swing accounts for 4.8 percent of votes from Labour to the Conservatives between 1964 and 1987.

3. Noteworthy is the Social Democratic Party established by four former Labour leaders in March 1981 and its subsequent alliance with the Liberal Party eight months later. The Alliance once achieved 51 percent of popular support, leading both major parties upon its launching in late 1981.

4. The Falklands War in 1982, for instance, was a landmark in British electoral history. Significantly, the South Atlantic battle boosted popularity of the governing Conservative party along with the electorate's economic expectation that precipitated the Tory's landslide victory in the following year (Sanders, Ward and Marsh 1987; Clarke, Mishler and Whiteley 1990).

5. The national election survey series starting from 1964 (one year after the 1963 study) includes a cross-sectional survey of general elections in 1964,1966,1970, February 1974, October 1974,1979,1983,1987 and 1992.

6. The time series sample used in Chapter VI extends from July 1979 to April 1996.

CHAPTER II

GENERATIONAL REPLACEMENT, SOCIAL CHANGE

AND TREND OF PARTISANSHIP

"Something is required simpler and more permanent, something which can be loved and trusted, and which can be recognised at successive elections as being the same thing that was loved and trusted before; and party is such a thing."

- Graham Wallas, Human Nature in Politics 1910

2.1 Introduction

Back in the dawn of this century, observers of British politics have recognized

that partisanship is "such a thing" that can be loved and trusted and, more importantly,

can perpetuate over time. Elections after elections, partisan allegiance signifies the

psychological attachment on which voters rely in making their political choices. Voting

results testify the linkage between party and the electorate, rendering psephologists

evidence to put their theories on ground. Indeed, Wallas was not alone in embracing the

perpetuality of "love and trust" to the political parties. Philip Converse, for instance,

believes that a voter's tie with his political party grows over time. In his work on partisan

stability, he argues that partisanship develops, strengthens and stabilizes over a course of

three generations (1969). In the first generation, as Converse theorizes, voters, who are

free from any parental influence on party choice, begin to choose and identify themselves

19

20

with political parties of their preference. During the second generation, most adults

inherit party identifications from their parents. The heritage goes on as party labels are to

be bequeathed to the third generation in spite of the generally weaker strength in party

loyalty. Nevertheless, voters of this generation are likely to reinforce their partisan

allegiance through voting and other political behavior. In addition, partisanship grows

with age and life experience (Converse 1969,1976; see also Abramson 1992).

Echoing Converse's partisan stability thesis, electoral studies prior to the 1970s in

Britain delegated a high place to the role of partisanship in analyzing voting behavior.

The British electorate was portrayed in these studies as a society of responsible electors:

party loyalists, with stable and strong partisan ties, and a high voting turnout. These

features, as Converse postulated, contribute to the maturity and stability of the political

system in Great Britain as Converse postulates. Election results and corresponding

research in the 1950s and 1960s provided supporting evidence to the partisan stability of

the British system featured with loyal partisans (Butler and Stokes 1969).

Subsequent research on elections in the 1970s, however, reported a rather

different scenario: partisanship of the two major parties, namely the Conservative and

Labour, eroded progressively while the electorate became increasingly volatile during

that decade (Crewe 1974; Crewe, Sarlvik and Alt 1977). Studies extending to the 1979

election went further and portrayed the 1970s as a "decade of dealignment," signifying

the general weakening of partisan ties across all social strata (Sarlvik and Crewe 1983).

In a 1974 article, for instance, Ivor Crewe first challenged the viability of the

21

partisan stability model suggested by Butler and Stokes in explaining electoral change.

He argued that, according to the Butler-Stokes model which was adapted from the

Michigan study, the enduring and self-reinforcing feature of partisanship should lead to

an increase in partisan strength over time. A stable electorate as projected from the

model is, however, anomalous to what has occurred in Britain in the 1970s: the gradual

fall in voting turnout, the persistent shrinkage of the major-party share of votes and

accelerating volatility of party support. These findings spelled out the failure of the

traditional party identification model in explaining political changes in Britain.

The trend of partisan decline became the focus of another article by Crewe and his

colleagues who introduced the term "partisan dealignment" to describe the process of a

general depreciation of party allegiance (Crewe, Sarlvik and Alt 1977). According to the

authors, partisan dealignment refers to the phenomenon in which "none of the major

occupational groups provides the same degree of solid and consistent support for one of

the two major parties as was the case in the earlier postwar era" (Sarlvik and Crewe 1983:

pp.332-333).

In his another work with Bo Sarlvik, Crewe reinforced his earlier findings of a

dramatic partisan decline among British voters with analysis extended to the 1979

election (1983). As Sarlvik and Crewe unfolded, the number of partisans who considered

themselves "very strongly" attached to a certain party fell from 40 percent of the

electorate in 1964 to less than one-fifth (19%) in 1979. The "decade of dealignment," as

Crew and Sarlvik note, witnessed the general depreciation of voters' bonds with parties,

22

particularly among lower class voters whose traditional attachments to the Labour party

became more vulnerable (Sarlvik and Crewe 1983).

What has caused the changes in political kinship between voters and parties?

Earlier studies in the 1960s postulate that generational replacement is the major force that

shapes partisanship of the British electorate. Among the most prominent is Butler and

Stokes' analysis of political changes in the 1960s and early 1970s (1969,1974). They

contend that among three kinds of changes that occurred to the British electorate, the

importance of demographic change is obvious. Although short-term fluctuations in

partisan swings and immediate issues and events precipitate transitional variations,

generational renewal of the electorate is cumulative and crucial. As new voters enter the

electorate replacing the old, it brings attitudinal changes to the political complexion of the

electorate. The evolutionary process of political conversion, despite its slow-ness, is the

major factor contributing to the changes in political alignment (Butler and Stokes 1969).

Apart from the generational effect, Clarke and Stewart (1984) note that two other

effects also account for the changes of the electorate over the course of time. The

extension of voting eligibility in 1970 to 18-year-old voters brought in new voters of the

younger generation who have generally weaker political party allegiance. As the authors

remark, this new generation of electors, which does not have the lengthy life experiences

of older generations, would not have sufficient time to be influenced strongly by the life-

cycle processes that strengthen their partisan ties (Clarke and Stewart 1984). In their

study, Clarke and Stewart report that these two factors have contributed to a clear

23

downward trend of partisanship among British voters between 1964 and 1983. Instead of

describing it as a large-scale realignment or dealignment, they tag the trend as a

"dealignment of degree" as they found the decline is a weakening in strength of party

identifications among partisans in both major parties. Moreover, the weakening of

partisanship is found to be rampant in almost all age cohorts (Clarke and Stewart 1984:

pp. 694-697).

In an attempt to extend the analysis of partisan decline in Britain to elections in

the 1980s and 1990s, this chapter assesses the trend of party identification over the 1979-

1992 period. The trend of partisanship across age cohorts is examined by using cross-

sectional and panel data from the British Election Study surveys. After studying the

general trends in party identification in the four general elections, I investigate the pattern

of partisan change among different generations/cohorts with comparisons extending back

to 1964.

2.2 Measuring Change in Partisanship

The decline in partisanship during the 1960s and the 1970s is described in earlier

studies as drastic and pervasive. Examination of the trend of partisan dealignment entails

the measurement of three types of changes:

1. Change in Partisan Intensity

This type of change refers to the mobility of partisanship in terms of partisan

strength with which a person's attachment to a political party is formed. Studies of these

24

types of changes constitute the majority of research in political change in Britain.

2. Change in Participation

Changes of this kind reflect the mobility of partisans in conversion between a

party identifier and a non-identifier. This kind of change may lead to withdrawal of

electors/partisans from the electorate and change in proportion or number of non-voters

and/or non-identifiers. This can be reflected in the decline of voter turnout or vote shares

of major parties as partisans cease to identify with these parties.

3. Change in Party Identity

The third type of partisan change alludes to the switching of partisans to other

opposing parties and shifting in partisan alignments.

Put in a continuum of partisanship, these three kinds of changes can be visualized

as movements along an imaginary scale with central point at zero representing non-

identifiers, right end strong partisans and left end strong partisans of other parties. The

first kind of change entails movements in either direction along the right-hand side of the

scale, excluding zero. The second type represents the straightforward movement to zero

and the third type is the movement in either direction across both arms of the scale across

zero as shown below1:

25

2

Strong Partisan of < > Strong Other Parties Non-identifiers Partisan

- 3 0 3

Most of the previous studies in partisanship have focused on measuring the first

type of change, i.e., movement along the right-hand side of the partisan scale. Two series

of measures were developed to examine this type of partisan change: the proportion of

"very strong partisans" (e.g. Crewe 1974, Crewe and Sarlvik 1983) and mean strength of

partisanship (e.g., Clarke and Stewart 1984, Abramson 1992).

This chapter aims to reassess earlier studies with primary focus on the 1979-1992

period. It concentrates on the change of partisanship denoted as the first type of changes

identified in previous discussion. Further scrutiny on changes of the other two types

forms the task of the following chapters.

In assessing partisanship among British voters, I adopt the conventional coding of

party identification as used in American election studies and previous British studies.

Data are collected from the British General Election surveys, in which a series of three

questions was asked regarding the respondents' paly affiliations and strength.2 Based on

the answers to the three questions, an ordinal measure of partisanship is derived from four

type of partisans:

1. very strong partisans (for those who answer "feel very strongly" toward the chosen

26

party);

2. fairly strong partisans (for those who answer "feel fairly strongly");

3. weak partisans (for those who answer "not very strongly" plus those who did not

identify themselves with a party but said "closer to one" and;

4. non-identifiers, i.e., persons who do not identify themselves with a party and also state

that they do not feel closer to a party. Each of the four categories is assigned a score of

party identification, which ranges from 0 (non-identifiers) to 3 (strong partisans).

2.3 Partisanship in the Post-dealignment Decade

Partisan decline in the 1970s has been described in previous studies as the major

feature of political change in Britain in the postwar decades. Using similar measures

from these studies, this chapter scrutinizes the alleged trend of dealignment and

demonstrates a different portrait of partisan dealignment during the 1979-1992 period in

contrast with earlier elections.

Table 2.1 chronicles the electoral changes in Britain since 1945. It delineates a

clear trend of decline in the Labour vote shares and, relatedly, the two major-party (Tory

and Labour) vote shares during the last three decades. Pertaining to the former, the figure

indicates that Labour suffered a remarkable plunge in the 1974 February election when

the Liberal Democrats, with other nationalist parties like Plaid Cymru and Scottish

Nationalist Party, surged to strip votes from the two major parties, fattening its share to

almost one-fifth of the total votes. This slight "realignment of degree" led to the slide of

27

two major parties' votes beginning that election. Despite the October election in the same

year returned Labour to office, the electoral loss was permanent for Labour. As a result,

the two major-party vote descended to a level of below 80 percent (except 1979) of the

total vote.

Electoral outcomes provide a rough sketch of voters' choice of governments and

the tide of partisan attachment to political parties. They are, nevertheless, inaccurate

indices to gauge the flow and ebb of party identification in the electorate. Table 2.2

presents a picture of partisan strength of all nine elections since 1964. Furnishing a

detailed breakdown of the direction and strength of partisan attachment by parties in each

election year, this table demonstrates a clear decrease of "very strong" partisans. It

indicates that, prior to the 1974 February election, more than 40 percent of the partisans

were "very strong" identifiers of their parties. This proportion has since been dropping

dramatically, leaving less than 30 percent of "very strong" partisan thereafter. The

downward trend continues and stabilizes at around 20 percent beginning in 1979 (see

Figure 2.1).

The remarkable plunge in the number of "very strong" partisans as displayed in

Figure 2.1, reflects a generalized weakening of partisan strength. It characterizes a less

partisan electorate in Britain since the 1970s. Although it does not necessarily refer to the

widespread departure of electors or contraction of the electorate, this "dealignment in

degree" is evident among partisans of the two major parties, particularly to the

disadvantage of Labour. Labour identifiers with "very strong" attachments, for instance,

28

account for 21 percent of the entire electorate in 1964. The figure dives to only 8 percent

in 1992, compared to Tories' 19 to 10 percent. Also, as depicted in Figure 2.2, the

proportion of Labour identifiers, regardless of intensity of identification, demonstrates a

steady diminution. In 1964,42 percent of the electors identify themselves as Labour

compared to 39 percent as Tory. In the 1979 election, the lead of Labour partisans

vanished as the Conservatives began to take over. This discrepancy between identifiers

of the two parties continues to widen from a gap of 2 percentage points in that year to 12

in 1992.

Across different groups of party identifiers, the partisan decline varies in

proportion and strength. In 1964, 50 percent of those who identified themselves with

Labour (or 21 percent of the electorate) were very strong partisans, which is the modal

group among others (see Figure 2.3b). The corresponding figures for the Tories and the

Liberals were 19 percent and 4 percent, respectively. The "very strong" category was the

mode for both major parties until February 1974 as the tide turned to the "fairly strong"

partisans. Note that, for both the major parties, the proportions of "very strong" partisans

decreased sharply in February 1974, after which the figures remain generally stable.

Using the second measure in examining partisan change, the mean strength of

party identification, a similar picture of decline in partisanship emerges.3 As illustrated

Figure 2.4, the average partisan strength registers a general downward trend. The overall

mean strength drops from 2.23 in 1964 to 1.76 in 1992, with the nadir being 1.74 in 1983.

Noteworthy is the slowdown of partisan decline beginning 1979 for both major parties.

in

29

Indeed, the general trend leveled as Mrs.Thatcher first government took office despite

that the depreciation of mean strength for Liberals continued.

The trend of partisanship over the last three decades manifests a general erosion in

both proportion of strong party identifiers and average strength of party support. The

analyses using both distribution of strong partisans and mean strength provided above

demonstrate two clear characteristics of the movement in partisanship during the twenty-

eight year period. The first is the differential depreciation of partisanship across different

groups of party identifiers. Consonant with previous studies, it is found that Labour

suffers most in the decades of dealignment as it loses as many as 10 percent of party

identifiers in the whole electorate (42 percent in 1964 slashed to 32 percent in 1992).4 In

contrast, the Tories increases slightly in size with corresponding figures of 39 percent to

44 percent.5 The fact that the proportion of Conservative identifiers remains generally

stable throughout the period in study (see Figure 2.3) as compared to that of Labour

identifiers, illustrates this "differential dealignment."

The second feature of the partisan trend in the 1964-1992 period is the

stabilization of partisan strength since 1979. Beginning with the first Thatcher

government that year, the trend of eroding partisan loyalties has slowed, while the mean

strength of partisanship for the two major parties remains generally constant in spite of

slight downward movements. Table 2.3 presents mean strengths of the three British

parties with inter-election differences. The sudden drop between the 1974 October and

1979 elections is manifest for all three parties, with Labour's loss being the greatest. The

30

level of partisan loyalty has stabilized since then with the exception of the Liberals, who

experienced a minor decline in 1992.

Table 2.4 presents the distribution of party identifiers, including those who did not

identify themselves with a party but only said closer to one. It portrays a more stable

electorate in the 1979-1992 period. The proportions of party identifiers in different

parties do not exhibit any sizable margins except a 6 percent drop of Labour partisans in

1983. Other than that, distribution of identifiers in all three parties remains relatively

stable during this period. Note that the demise of the Social Democratic Party, with its

Alliance with Liberals, delivers uneven blessings to the major parties in the 1992

election. Among all, the Tories benefited the most (4 percent of the electorate) while the

Liberal and Labour merely secure small gains.

In earlier studies, Crewe and other observers posited that the contraction of

combined Conservative and Labour shares of vote and partisans lent weight to the

partisan dealignment thesis, which depicted a smaller partisan electorate in the 1970s. As

analyzed above, the two-party vote share began to rebound in the 1979 election while the

figure fluctuated in the 1980s. Vote gains for the Alliance in 1983 and 1984 cut the

major parties' vote to as low as 70 percent (see Table 2.1). However, the figure

rebounded in subsequent elections and climbed back to a moderate level in the 1992

election. It is too early to conclude that it signifies an ultimate resurgence in the 1990s.

The data suggest, however, that the earlier erosion in major party support has ceased.

Pertaining to the change in proportions of major-party partisans, Sarlvik and

31

Crewe contend that the obvious downward trend of the percentage of major-party

identifiers is another evidence for the softening of electoral bases of both the

Conservative and Labour parties.6 Using their measure, it is found that the combined

percentage of two-party identifiers drops even lower in 1983 (63 % compared to 81 in

1964), given the breakaway of four prominent leaders from Labour to form the SDP in

1981. The proportion of Tories and Labours started to rise thereafter and it scored 76

percent in 1992. The moderate surge is evidenced in another analysis which takes into

account those who only think of themselves as "closer to one party" than as a party

identifiers (see Table 2.4).

2.4 Explaining the Trend of Partisanship

From the analyses using the mean strength of partisanship and proportion of

strong partisans, a clearer picture of partisanship trend in the 1980s and 1990s emerges:

the dealignment of party allegiance is in differential among various British parties and the

trend of decline has been stabilized beginning the 1979 general election. These findings

do not necessarily lead to a counter-argument to the dealignment thesis which has been

articulated in the 1970s and early 1980s. Partisanship in Britain has been demonstrated to

be a function of various effects. The cohort analysis literature proposes three kinds of

effects that account for a population's attitudinal variation over time (Glenn 1977).

Generational effects involve the influence of population turnover that is associated with

entrance and departure of generations with distinct values. The aging effect is derived

32

from the life-cycle of a person in forming and altering his/her attitudes as he/she ages.

Period effects refer to the influence of a certain period of time during which special

events or circumstances cause attitudinal transformation (Glenn 1977).

Previous studies on electoral changes in Britain suggest that generational effect

has been the major source of partisan change as older cohorts having stronger ties with

parties who have departed the electorate, being replaced by younger generations with

weaker party loyalties. Converse, in his well-debated theory of partisan stability,

suggests that partisanship grows over time as voters age. In testing this theory,

Abramson analyzes and compares eight British elections starting in 1964 in search for the

generational effect that moves partisan trends (1992). Contrary to what Converse posits,

he discovers that partisanship in Britain, instead of strengthening over time, declines in

the last three decades. Applying cohort analyses on the proportion of very strong

partisans and the mean strength of partisan loyalties, he suggests that the long-term

declivity is due to generational replacement and short-term forces.

Granted the elegance of his work, Abramson's study is vulnerable to two

methodological problems rampant in cohort analysis. For one, election surveys vary in

sample sizes as well as age distributions. Comparison between a cohort of 200

respondents in one election and another cohort of 50 respondents in the next election will

be invalidated by differences in sampling errors (Glenn 1977). This is not uncommon in

analysis focusing on comparison between old age groups that may vary considerably in

size in the next election owing to high mortality rate. The second drawback arises from

33

difficulties in disentangling different effects from each other. It is formidable, if not

impossible, to discern the impact of generational replacement from period effects when

cross-sectional surveys of two adjacent general elections are compared (Glenn 1977: pp.

48-54).

To circumvent the problems of incomparabilities using different election data, I

apply the cohort analysis on panel data that taps the responses of the same group of

respondents in two general elections. Although it does not illustrate the whole picture

across all generations in the last three decades, the panel study uncovers the inter-

generational and intra-cohort changes during the 1987-1992 period. The panel study of

birth cohorts excels in two aspects. Age effects can be distinguished without

considerable error, because the same group of respondents is observed over a fixed period

of time. Moreover, comparison across age cohorts will be more reliable given the

relatively small sampling error and similar size of standard deviations.

Table 2.5 documents the mean strength of partisanship across 16 age cohorts in

the 1987 and 1992 general elections. Variations across the first two columns reflect the

age group differences that demonstrate a positive relationship between partisanship and

age in both elections. It does not necessarily suggest that adults tend to grow in their

partisanship with age, but rather that cohorts born in different periods have different

strengths of partisanship. In other words, adults born in 1918 have a higher average mean

strength of party identification (namely, 2.05 in 1987) than those born in 1974

(correspondingly, 1.46). Reflecting the impact of different life experiences, it lends

34

weight to previous findings that older people are more likely to have a higher partisan

bond than younger voters (Butler and Stokes 1969,1974; Crewe, Sarlvik and Alt 1977).

Column 5 of the same table shows such differences, and column 4 shows the

change of partisanship in the same cohort after five years. Similar to previous studies, it

appears that younger voters tend to strengthen their partisan attachment after successive

elections until arriving at a stable level in the middle age (see, e.g., Clarke and Stewart

1984). Starting from the 47-54 age group, all cohorts display a weaker partisan strength

as compared with the same group in the last election, suggesting inter-cohort differences

of partisanship among middle to older age cohorts. Moreover, it is the same group of

adults who began to have an intra-cohort decline in partisanship compared to what they

registered five years earlier. In other words, as these groups of adults grow five years

older, they only become less partisan than in the previous election. All age cohorts older

than that group reveal a similar downward pattern except the two oldest groups.

Unlike Clarke and Stewart's findings in 1984, analysis focusing on the two

elections in 1987 and 1992 does not exhibit an overall direction of a weakening

partisanship in all age cohorts. However, it can be argued that the age effect that registers

a positive linkage between age and partisanship is attenuated in recent elections. This

phenomenon is summarized in Figure 2.5. Both series of mean strength of partisanship in

1987 and 1992 register an upward trend, indicating a positive relationship between age

and strength of partisan loyalties. This relationship does not imply that voters tend to

increase in their partisan strength as they age. Rather, it represents the generational or

35

cohort effect as experienced by different cohorts. Starting from the middle age (47-53)

group, the generational effect sets in as voters belonging to this or older cohorts exhibit a

generally weaker strength of partisanship in comparison with the same age group five

years earlier. On average, the mean strength difference is 0.1 among these nine age

groups, with the largest margin of 0.16 and 0.15 among the oldest group and the 63-70

group.

Based on the findings from the cohort analysis of partisan trends in the last two

elections, one cannot conclude that a general decline of partisanship is manifest in all age

groups. Again, this provides evidence supporting the proposition that dealignment in

differential in different parties and age groups.

The next question prompted by the argument is: how does the differential

dealignment affect electoral outcomes? Does the weakening of partisan tie, particularly

for Labour identifiers, suggest a higher mobility of votes switching between parties?

Table 2.6, which presents the inter-election turnover based on the 1987-1992 elections

panel survey, sets the stage for further inquiry on the question. As the table shows,

Conservatives maintain a more stable group of partisans than Labour. More than 86

percent of Conservatives who identified themselves in 1987, representing 35 percent of

the whole electorate, carry their Tory label to the next election in 1992. In contrast,

Labour keeps a similar proportion of stable partisans (83 percent), but it only represents

slightly more than a quarter of the electorate. The nine-point margin reflects the

difference in size of the stable electoral base of the two parties. Intriguingly, the

36

proportion of defectors of each party to opponents appears to be relatively minor (5 to 7

percent), in spite of the alleged decline in partisanship.7

2.5 Conclusion: Differential Dealignment and Realignment of Degree

In his partisan stability thesis first published in the late 1960s, Converse

postulated that partisanship tends to grow and strengthen with life and voting experiences

(1969). Along this line, the traditional party identification model advanced by Butler and

Stokes (1969) portrays an electorate with strong party loyalties rooted in social cleavage

and class-related interests. According to what this model predicts, demographic renewal

of the electorate and reinforcement of partisanship through life experiences provide the

major sources for political change. However, reassessment of the model in light of

political changes in the 1970s reveals the deficiency of Butler and Stokes' argument.

During that decade, the general decline of partisanship gainsays the Butler-Stokes model.

Instead of demonstrating a strengthening tie between parties and voters over time,

the last three decades witnessed an erosion of party loyalties among British voters.

Generational replacement does not contribute to a more stable electorate but, on the

contrary, less loyal partisans. The trend of decline found in the 1970s, as evidenced in

this study, did not sustain through the 1980s and early 1990s. A closer scrutiny extended

over a longer period-including most recent elections-illustrates that the dealignment of

partisanship in Britain is not ubiquitous, but time-bound and group-specific. As this

chapter exhibits, partisanship differentially declines among various parties and age

37

cohorts in the "post-dealignment" decade. Meanwhile, the partisan strength of the

general electorate has become stabilized at a relatively moderate level since 1979.

Compared to the studies in the 1970s and 1980s, analyses in the chapter paint a

"similar but somewhat different" picture of the British electors than their predecessors in

the last two decades: political decisions of partisans with a moderate psychological

attachment to the parties are not habitual practices based on strong partisan loyalty;

Labor's electoral base is weaker but not more likely to defect than the Tories one;

younger voters are likely to strengthen their partisanship with voting practice; middle age

partisans are more prone to lessen their allegiance to political parties than in the past.

The lack of a strong tie between parties and voters gives leeway for a more mobile

electorate in the 1990s. The decision of electors on choice of party and vote becomes

more open to influences of short-term forces, such as mass media coverage of political

events, evaluation of issues and image of political leaders (Clarke and Stewart 1984,

Crewe and Sarlvik 1983). Franklin (1985), in his work on decline of class voting,

attributes the changes in partisanship and voting decision to the attenuation of

sociological influences such as class structure and impact of inherited partisanship. He

proposes that the British electorate now is more likely to rely on individual evaluations of

issues and policies in determining their votes. Suggesting another candidate, Clarke and

Stewart (1984,1992) focus on the impact of party leaders in influencing public evaluation

of parties. To further investigate the factors affecting the trend of partisanship in the

post-dealignment era, analyses of these two proposed models constitute the major task of

38

the following chapter.

39

Endnotes

1. A similar scale can be created in measuring the changes in voting decisions with the zero point representing non-voters. Literature of studies of voting encompass analyses of voting turnout and ticket-splitting (Crewe, Fox and Alt 1977; Denver and Hands 1985), which are beyond the current purposes of this chapter.

2. Respondents were asked a series of questions in sequence: "Generally speaking, do you usually think of yourself as Conservative, Labour, or Liberal or what?" For those who chose a party, a second question will be asked: "Well, how strongly (chosen party) do you generally feel - very strongly, fairly strongly, or not very strongly?" Those who did not have a party preference were asked: "Well, do you generally feel a little closer to one of the parties than the others? Which party is that?"

3. This measure includes those respondents who did not choose a party outright but who indicate they are closer to one. These responses are coded 1 for weak partisans.

4. The party hit its lowest share in the electorate at 29 percent in 1983, which is a 13-point plunge.

5. Note that the figure is in proportion to the whole electorate, taking into account the portion of non-identifiers.

6. Sarlvik and Crewe's finding was based on the proportion of respondents who identified themselves with one of the parties. Those who refused to answer or only said "closer to one party" are excluded.

7. This suggests the triviality of inter-party mobility mentioned in an earlier part of the chapter as the third type of partisan change.

CHAPTER III

PERCEPTION MATTERS - LEADER IMAGE, ECONOMIC EVALUATIONS AND

ELECTORAL CHOICE

3.1 Introduction

The evidence we gathered on partisan change in Chapter II illustrates a new

pattern in the relationship between party and voters developed since the end of the 1970s.

The trend of partisanship during the 1979-1992 period exhibits a weakened but stabilized

partisan commitment among British electors. While the decline of partisanship varies in

degree between different parties and different age groups, explanations can hardly be

subsumed under one single factor of demographic replacement. The findings in the last

chapter demonstrate that the subtlety in the evolution of partisanship calls for more

detailed investigations into the long-term and short-term forces that drive political choice

and partisan loyalties. In this chapter, we continue to explore in a broader context these

forces such as social cleavage, issue concerns and leader image to investigate the

dynamics of the partisan tide in contemporary Britain.

Earlier studies on partisan change postulated that a strong class-party tie

connecting voters and parties is responsible for the divide in party support between the

middle and working- class voters. In defining class, Butler and Stokes employ a variety

of criteria, among which the authors claimed the role of occupation provided one of the

40

41

"best predictors" in characterizing class (1969:65-80). Using a six-scale social grade

measurement first employed in a 1963 study,1 they identified the lower three categories

of manual workers as working class and another three non-manual categories as the

middle class.

Based on their occupational class identifications, Butler and Stokes found that

social cleavage imposes a strong partisan self-image among British voters. This is

manifest in the enormous difference between the parties' shares of votes in each class.

Findings based on the British national survey in 1963 and election study in 1964, for

instance, demonstrate a clear class cleavage in party support. Seventy-nine percent of the

self-identified middle class described themselves as Tories while a comparable proportion

of working class respondents (72 percent) identified themselves as Labour. Based on this

evidence, Butler and Stokes posit that the "preeminent role (of class) can hardly be

questioned" (1969: 76). Accordingly, the close alignment of class identity and party label

in the 1960s contributed to strong and stable partisan commitments among British

citizens. In 1964, only about one-tenth of the voters describe themselves as not-very-

strong partisan while almost half of the electorate (43 percent) claim they have very

strong partisan ties with political parties (see Table 2.2 in Chapter II).

The sharp divide between the working- and middle-class in the 1960s is reflected

in the strong correlation between class and party identification. It accords well with Peter

Pulzer's widely-cited remark: "class is the basis of British party politics; all else is

embellishment and detail" (1967). In explaining the bifurcation of political support

42

between the two classes, Butler and Stokes identify three models that tap into the

conception of politics among voters and their psychological underpinnings of this class-

party linkage. In the first model, the authors suggest that opposing class interests

constitute the staple of divergence between middle-class and working-class voters. In

this class-conflict model, individuals conceive of politics as a "zero-sum game" in which

parties attained political payoffs for one class at the expense of the other. The belief that

parties represent different and exclusive class interests projects the perceived class

conflicts into a polarized partisan context in which parties are against each other as well

as the "other classes." The second model is a variant of the class-conflict model. Instead

of viewing politics as a zero-sum game, it relaxes the assumption of purely antagonistic

class interests and suggests a "positive sum game." Parties play a similar role in this

model in representing and optimizing interest ("simple representation") for their

constituent classes. An individual identifying with a certain party develops a positive

linkage with the party, but not necessarily a negative attitude toward the rival parties.

Other reasons than class interests can also be considered in party choice and vote

decisions. In the third model, Butler and Stokes seek explanation from the class culture

which has evolved in the British social milieu over time. Despite that perception of class

interests is integral in forming party allegiance, the normative structure in which the

class-party concept is embedded is pervasive in both the middle and working classes.

The norms developed within each class, along with the perceptual framework of self-

identity and class interests, provide cues that guide voters' political behavior. This

43

normative quality of class-party identification is nurtured by family and parental

influences. Then it is reinforced through interactions with neighbors and members of the

same class and political participation, like voting during adulthood (Butler and Stokes

1969:80-94, see also Miller 1978).

Based on the theoretical framework derived from the three models, the authors

construed the class-partisan bond as a result of long-term forces originating from the

social class milieu in which an individual is socialized. During the socialization process,

a political identity is inherited from parents, influenced by occupational self-image and

socioeconomic status, and reinforced through political activities and participation.

Among the two classes, Butler and Stokes found that the Labour working class has the

strongest appeal for class interest. In contrast, the Tories are perceived as more

concerned with interests that involve the whole nation. Class alignment, as termed in

their study, is not static over time. In the revised edition of Political Change in Britain.

Butler and Stokes observed five years later the "ageing" of the class-party link. As a

consequence of structural changes in the British economy and general betterment of

economic conditions for both classes, the decline of class alignment in the 1970s was

manifest. The expansion of the goods and services sector redefined the social

identification among workers who became reluctant to cling to the working-class label.

Embourgeoisement of the working class narrowed the gap between the two classes.

Affluence was blurring the line that divides blue-collar and white-collar workers. Class

alignment was being replaced by class dealignment that contributed to an increasingly

44

volatile electorate with a weakened class-party tie. To substantiate their argument, the

authors noted that the difference between the proportion of Conservative identifiers and

Labour identifiers among the working class dropped remarkably, from 44 percent in 1963

to 28 percent in 1970. The corresponding figure for the middle class is 50 and 40

percent, respectively (Butler and Stokes 1974).

Studies of class dealignment continued in the mid-1970s and 1980s, with Crewe,

Sarlvik and Alt (1977) providing a generational analysis of the erosion of class-party tie.

The authors demonstrate that the class basis of party support had been diminishing since

the 1960s. The impact of class on party choice became gradually "fainter" regardless of

the economic conditions. As discussed in the last chapter, the withering of class effects

on party identification and voting choice constitutes the thesis in the former two authors'

work on partisan dealignment in 1983. During the 1970s and early 1980s, the theme of

class dealignment became the orthodoxy in studies of political change in Britain.

However, this new conventional wisdom came under severe challenges after the 1983

general election. In 1985, Heath, Jowell and Curtice published findings from their study

based on the 1983 election. In their book, How Britain Votes, they articulated

skepticism regarding the previous studies that alleged a decline in class voting. Heath

and his colleagues argued that dividing the electorate into two classes was problematic

and it was incapable of capturing subtle changes of class-party linkages, particularly in

the light of the rise of new occupational groups like the self-employed and salariats.

Employing new statistical measures based on odd-ratios and log-linear analysis, the

45

authors contended that the class basis of politics, as best represented by relative class

voting, remained intact. The relationship between class and party was masked by the

decline of absolute class voting, which merely represented the temporary changes in

political support to each party. These changes in political support have been a function of

the short-term performance and policy effectiveness of the governing or opposition party.

They are, as Heath et al. argue, unrelated to the long-term linkage between parties' class-

interest appeal and the country's social-class structure.

The controversy concerning class dealignment turned white-hot following the

publication of Heath et al's book. Crewe and Dunleavy spearheaded the debate in

making counter arguments against the "resurrection" of class voting as promoted by

Heath and his colleagues. The key point in the debate concerns the class basis of

political choice in party identification and voting. Is there is a downward trend in the

influence of class on party, or are the observed fluctuations only "trendless"? Crewe

(1983) defended the conventional use of the Alford Index, a measure that accounts for

the class-party votes with the difference between same-class voting (working class) and

cross-class voting (middle class), in demonstrating the decline of class effect on voting.

The contraction of the working class, as he maintained, did indeed reflect the abatement

of class impact on political outcome. From another perspective, Dunleavy (1987)

critiqued methodologically the use of odd-ratio and log-linear analysis, which, he argued,

is extremely sensitive to small changes and leads to exaggerated findings Heath et al.

embrace.

46

3.2 Short-term Forces Overlay Long-term Factors

Despite the plethora of fierce arguments on the topic, it has been widely

recognized that the evolution of class structure in contemporary Britain has precipitated

profound changes in the linkage between class and voting, in absolute and relative terms

alike. As recent studies disclose, class no longer plays a predominant role in determining

party identification and electoral outcomes as in the 1960s. Clarke and Stewart, for

instance, examine the long-term and short-term forces that led to partisan dealignment in

the 1974-1983 period, and found the latter claims ample warrant in accounting for

variations in voters' partisan loyalties in recent elections (1984). The rise in significance

of short-term forces did not gainsay the part social class plays in informing the party and

voting choice, yet variables such as issue perceptions and party leader image now

command much attention in recent studies of the dynamics of political support (e.g.,

Whiteley 1983, Franklin 1985, Stewart and Clarke 1992). "Shorter-term forces have

overlaid longer-term trends," write Clarke and Stewart (1984:700).

Regarding the first variable, issue concerns have long been accorded a role in

accounting for party choice in Britain (Butler and Stokes, 1969,1974). Evaluations of

issues and government policies provide the essentials for the rational voter model as

postulated in liberal democratic theory. An ideal voter, as the model projects, is the one

who is well-informed, rational and utility-maximizing in his/her political decision

calculus, based on judgment of issues and policies as advanced by different parties.

Butler and Stokes, however, found the model wanting as they discovered that issue

47

factors have limited force, given the complexities of issues and inconsistence among

British voters in understanding and interpreting political issues (1969). As the authors

found, policy preferences and issue evaluations were, at best, "atomistic" and

unstructured.

Partisan decline in the 1970s was accompanied by a resurgence of issue concerns

as principal factors in predicting electoral choice. Sarlvik and Crewe contend on the

basis of their findings from the 1979 election study that voters' opinions on issues and

policies, together with the perceived positions taken by each party, furnish good

predictors of party choices and electoral decisions (1983). In his analysis of the decline

of class voting, Franklin (1985) concurs by demonstrating that issue evaluations have

displaced the role of class as voters become more sophisticated in their electoral choice.

Clarke and Stewart (1984) have taken into consideration the political and

economic contexts in the 1970s and 1980s in examining the role of issue evaluations.

They singled out the landmark example of the "Winter of Discontent" in 1979, which put

in front of voters a vivid illustration of prolonged economic malaise. A deteriorating

economy heavily laden with nationalized industries, the imminence of the influential

EEC, and sundry other salient issues kept voters from relying on simple class-party links

to determine allegiance to a party and, in turn, to cast a vote of party loyalty. Moreover,

the perception that economic hardships were linked to the aggravated strikes in the 1970s

prompted the electorate to reconsider class-interest-oriented issues. Indeed, the valence

issue of the decaying economy was so pervasive that it crosscut party lines. This

48

indirectly benefited the party with a less class-interest appeal, namely, the Conservatives.

As noted by Clarke and Stewart, the contextual factors of the 1970s laid the backdrop for

the rise of issue concerns as factors affecting party and voting choice.

Unlike issues, party leader image has achieved prominence in modeling party

identification only in recent research. Most analysts of British politics have been

reluctant to adopt the thesis that evaluations of party leaders play a major role in voters'

decisions. Butler and Stokes (1974: 244), for example, assigned less than a score of

pages in their 500-page book to analyzing the electoral impact of leaders. In a short

chapter close to the end of the book, the authors note that the factor "remains but one"

among others that determine variability in partisan strength, yet the pull of leadership

impact can be "easily outweighed" by other factors. Another researcher, Ivor Crewe

(1985), did not deny the influence of leader image. However, he merely attributes the

1983 election results to the negative impact of a particular Labour leader, Mr. Michael

Foot, than to suggesting general predictive power to leader image variables.

It is only recently that leadership effects have become the research focus in studies

of party choice and voting intentions. Clarke and Stewart (1984) note that mass media

coverage of election campaigns brings to the spotlight the party leaders' style and

personalities, which epitomize the capabilities of these candidates and~relatedly, their

parties--in running the country. This leadership effect is most notable during campaigns,

as research demonstrates leader image is closely linked to party image and has substantial

influence on party support and voting choice (Miller et al. 1990; see also Bean and

49

Mughan 1989). Empirical studies based on time series of government popularity also

reveal the close ties between a prime minister's approval and party support in voters'

minds (e.g., Clarke and Stewart 1995; Clarke, Stewart and Whiteley 1996). In another

study, Stewart and Clarke (1992) found that leader image is composed of two

dimensions, competence and responsiveness. Using confirmatory factor analysis, the

authors demonstrate that, in voters' eyes, these two dimensions are more closely

connected for opposition leaders than for the prime minister. In other words, voters are

more attentive and discriminating when evaluating the leaders of a governing party.

Research in the last few decades reveals that the dynamics of party support are

affected by long-term forces such as social-class identification and short-term factors

exemplified by issue evaluations and party leadership effects. In modeling party

identification in the contemporary context, the rest of the chapter continues to investigate

the changes in party identification by first reexamining the well-debated class

dealignment hypothesis in the 1979-1992 period with reference to the 1960s. Then, a

model of party identification including both a class variable and short-term forces is

tested.

3.3 The "Aged" Class Alignment

As noted in earlier studies, the 1970s witnessed the erosion of social-class

influence on party choice, owing to the structural change in British society and its

economy. As a result, voters are not subscribed to a "natural" partisanship as much as in

50

the 1960s. Survey data gathered at the time of the 1964 elections, for instance, showed

that among the two classes defined by manual/non-manual nature of occupations,2 more

than 57 percent of the electorate identified with their natural party, i.e., manual workers

with Labour and non-manuals with Conservative. A majority of middle-class voters

named themselves Tories (61 percent) in 1964, whereas the comparable proportion of

working-class respondents making a similar pledge to the Labour party was 55 percent.

This pattern of natural partisanship did not last. In 1979, the proportion of natural class-

party identifiers fell to 50 percent, which remained stable until the 1992 election (see

Table 3.1). This apparent slight drop, however, warrants closer investigation. Note that

the decrease was compensated by the parallel rise in non-manual workers' identifications

with the Conservative party. Taking the proportion of natural class identifiers separately,

it is noticeable that Labour suffers a substantial loss of support among manual workers.

In 1964, almost 40 percent of the whole electorate were members of the working class

and Labour identifiers. The figure dropped to slightly over 30 percent in 1979 and

descended further to 21 percent in both 1987 and 1992. In contrast, the middle-class

Tories gained more than 10 percentage points during the period, and comprised 29

percent of the electorate in 1992.

The decrease in Labour's natural-class identifiers produced a weakening in

absolute terms of the party's electoral base. However, it does not necessarily lead to a

corresponding increase for the Conservatives and a strengthened class basis for the

Tories. The argument necessitates two qualifications. The first regards changes in

51

distribution of occupational classes in the last three decades. Table 3.2a to 3.2c present

the changes of social-class structure measured in occupations, home-ownership and trade

union membership. All three measures reflect a reversal of working-class dominance,

which was widely acknowledged in the 1960s. To summarize, manual workers dropped

from 70 percent in 1964 to less than half in 1992. More than half of the British voters did

not own their homes in 1964; more than 70 percent do in 1992. Four out of ten

respondents claimed they were trade union members in the 1964 survey, less than a

quarter do so in 1992. These figures lend weight to the argument that the working class

has diminished markedly in size after twenty-eight years, and the middle class has grown

in reciprocal proportion. It follows naturally that the Conservative party is supplied with

an augmented basis for recruits and, in turn, a better chance of stronger support, provided

that the party's class linkage maintains. It is the stability of this class-party bond on

which the second condition for a stronger electoral basis for the Tories relies. Whether

there exists a consistent class-party alignment in both parties constitutes the prerequisite

in examining the broadly debated theme of class voting.

In previous studies of class alignment, a variety of measures have been developed

to gauge the impact of social class on party choice and vote decisions. Among them,

most widely used is the Alford index. This measure, which is derived from subtracting

the percentage of non-manual Labour (cross-class) identifiers from that of manual Labour

(natural-class) identifiers, is designed to assess the extent of class voting.3 It embodies

the assumption that "the deviation of either the manual or non-manual stratum from any

52

given level of vote (party identification) for one of the major parties is the important fact

relevant to an assessment of class voting (party identification)" (Alford 1976: 80). Table

3.3 displays the Alford indices for the four elections between 1979 and 1992, and a

comparison with the 1964 election. Including only the Conservative and Labour

identifiers, the Alford indices illustrate a decline from 42.4 percent in 1964 to 29.3

percent in 1979. The figure fluctuates within a narrow range thereafter, suggesting that

the relationship between social-class identities and party choice remains stable during the

post-1979 period. Indices based on the whole sample including Liberal identifiers and

non-identifiers yield similar results. They testify to the statement that class becomes a

less influential factor when contemporary voters consider identification with a political

party. Stabilization of "class dealignment" after 1979 suggests that the social factor of

class still plays a role, albeit weakened, in affecting political choice in contemporary

Britain.

Focusing on the 1979-1992 period, both the Conservative and Labour parties

exhibit apparently stable proportions of natural class followers within each class (on

average 52 percent and 45 percent, respectively). The strength of partisan allegiance

aside, about half of the electors in each class still align party labels with their social-class

identifications. Table 3.4 displays the mean strength for natural-party and cross-party

identifiers during the period with reference to 1964. The former group registers, in

general, a slightly higher partisan strength than the cross-class identifiers. Compared to

that of the 1964 election, partisan loyalties are generally weaker in the most recent four

53

elections. Note that the discrepancy between natural-class identifiers and cross-class

identifiers grows steadily wider in the 1980s and early 1990s. This sheds light on the fact

that the general erosion of partisan ties renders more discernible the natural-class

partisans' tie with party as compared with cross-class counterparts. To explore the

mutabilities of partisan intensity comparing these two groups, a closer analysis is in

order.

To test the hypothesis that party allegiance of cross-class identifiers is more

susceptible to change than natural-class identifiers, survey data drawn from the 1987-

1992 panel are examined. The partisanship and partisan loyalties of four categories of

party-identifiers—namely, natural-class identifiers including middle-class Tories and

working-class Labours and cross-class identifiers including middle-class Labours and

working-class Tories-are compared in Table 3.5. Consonant with the preceding findings,

natural-class partisans in the 1987 election are less likely to switch parties in 1992.

Among partisans identifying with their natural class, an average of 88 percent stayed with

their party label in the next election. In contrast, only 78 percent of cross-class identifiers

did so, with the variability most remarkable among working-class Tories. In accordance,

this group of party identifiers is found to be most volatile, among which less than half

will retain a stable party tie. Working-class conservatism has been cited as one of the

major factors responsible for the increasing volatility of the contemporary British

electorate (Crewe 1983, Franklin 1985). It is clear from this table that instability

characterizes this category of party identifiers.

54

Because the difference in change of partisan strength between the two groups of

party identifiers is modest, the argument that class factor is the main source of partisan

change has yet to be resolved. Bivariate correlations between partisan strength and the

three class measures, i.e., occupational class, home-ownership, and union membership,

all exhibit statistically insignificant relationships,4 rendering further scrutiny using more

sophisticated methods necessary.

Based on the descriptive analyses above, the impact of social class on party

identification and voters' strength of allegiance to the party remains unclear. Recent

research has sought to study the short-term forces in filling the vacuum vacated by the

decline of class voting (Franklin 1985, Clarke and Stewart 1984). In modeling voting

behavior and party strength, these studies suggested that issue evaluations and leadership

effects were better candidates to explain electors' choice (see also Sarlvik and Crewe

1983, Denver and Hands 1992). The analysis that follows employs multivariate

techniques to examine the influence of short-term forces on partisanship.

3.4 Modeling Partisanship in the post-Dealignment era

The rest of the chapter concentrates on developing a multivariate model of party

identification based on a panel study of the 1987 and 1992 elections. To measure a

voter's attachment to a political party, a partisan continuum was used in the preceding

chapter to explain changes in a voter's partisan tie. The following analysis adopts the

same measure and combines party identification with the strength of commitment to

55

operationalize a voter's party allegiance. The dependent variable, PARTYID, is an

ordinal variable that ranges from a value of +3, representing very strong identification to

a party, to -3, that denotes very strong identification to other parties. Zero is assigned to

non-identifiers.

In modeling the impacts of both long-term and short-term forces on party

identification,

four groups of independent variables are identified: demographics, class, issue

evaluations and party leader image. The first group of variables, which entails age and

educational level, is designed to study the age effects as analyzed in the last chapter and

the effect of increasing educational attainments of the British voting population. The

next three groups of variables merit more detailed explanations. The second group,

which represents the long-term social force of class structure, uses three measures to

gauge class identification based on occupation, trade union membership and home

ownership. These measures represent different dimensions of changes in contemporary

British society. Among them were the structural changes in the British economy that

were reflected in the proportions of blue-collar and white-collar workers. Occupation is

thus utilized to distinguish changes in social grades. The measure of occupational class

has been deployed to index social identities since the pioneering British election studies

in early 1960s. Empirical analyses based on this categorization scheme find consistent

results in defining the middle class and working class using the manual and non-manual

divide (Butler and Stokes 1969,1974; Crewe 1983). Despite that controversies arose on

56

reclassification of social grades based on different schemes of categorization (Heath et al.

1985), analysts maintained that the traditional manual/non-manual dichotomy derived

from the six-point scale provides a reliable device that mirrors changes in class structure

during the postwar decades.5

Not unrelatedly, the increase of white-collar workers redefined the role of trade

union as in the 1970s new middle-class occupations were unionized and developed new

class values. The trend of union membership parallels changes of union movements,

particularly in the aftermath of the 1979 strikes. As illustrated above, the proportion of

union members in the British population dropped steadily from 40 percent in 1964 to less

that 25 percent in 1992. Apart from numeric counts, union membership provides a scale

for assessing class voting with which one can test the influence of unions in socialization

of working-class values.

Regarding the third measure of social class, home ownership, recent research

furnishes abundant evidence of the impact of Mrs. Thatcher's housing policy in early

1980s (Dunleavy 1980, Franklin 1985, Rose and McAllister 1986, Heath et al. 1991).

Studies of the British housing policy under Thatcher's governments observe that the new

class of owner-occupiers exhibit economic interests and values that differ from those of

council tenants. A new identity developed from home ownership brought a different

value system to party and electoral choice (Franklin 1985). To investigate the effects of

the aforementioned class factors, dummy variables are created to represent the middle

class (1) and working class (0), union members (1) and non-union members (0), and

57

home owners (1) and non-home owners (0).

The next two independent variables involve a group of subjective variables that

garner extensive attention in recent research of British electoral studies. Evaluations of

party leaders, for one, has become a focal point in recent elections as mass media

increasingly emphasize coverage of political figures during election campaigns. Earlier

studies on political support suggest that voters' evaluations of party leaders are strongly

associated with party choice (Stewart and Clarke 1992). To operationalize the leader

effect, a summary measure is developed based upon a battery of questions designed to

examine the British public's perceptions of the political leaders in the 1979-1992 panel

survey.6 Respondents in the survey were asked questions regarding their feelings about

the political leaders on whether they are capable, caring, extreme or caring for one class.

The variable, LEADERS, is the sum of responses from these questions. The ordinal

variable ranges from +4 (best rating) to -4 (lowest rating).

The next variable concerns issue evaluations. Issue evaluations have been

components of models of party identification and voting since the early 1960s. However,

they were not accorded close attention until recently, when researchers reconsidered the

influence of issue concerns as the social-class factor declined in salience. The evolution

of issue agenda in the public mind is also responsible for the growth in importance of

issues that involve government policies. One of the major causes is the protracted

economic hardship the country has experienced during much of the postwar period. The

pervasiveness of this "valence issue" of economic well-being overshadowed other

58

polarized class-oriented position issues and emphasized the importance of an issue-based

instead of a class-based party support judgment (Franklin 1985).

When making party support decisions, issue judgments involve assessment of the

distance between an individual's issue position and his/her perception of where the

parties stand on that issue. According to the spatial theory first proposed by Anthony

Downs, voters choose the party whose policy position is closest to their own (1957). The

requisite for this spatial model is the existence of an issue space in the voters' minds on

which they position themselves with reference to perceived party positions. However,

other researchers have suggested different criteria to specify this issue space (see

Macdonald, Listhaug and Rabinowitz 1991). For present purposes, however, we focus on

the "perceived" distance that a voter has with reference to a political party's position on a

specific issue. Alternative conceptualization of issue space will be considered in the next

chapter that analyzes the voting models, ger se.

To operationalize a voter's projected issue distance from parties, an additive index

is constructed from the sum of the differences between the respondent's position and

his/her perceived party positions on five issues.7 The issues are defense, unemployment

versus inflation, tax and welfare, nationalization, and redistribution of wealth. Because

the principal interest in the current context is the overall issue distance between the voter

and different parties, the index provides information pertinent to measurement of the

effect of voters' issue evaluations on their party preferences.

Previous empirical research on American party identification models suggests that

59

partisan affiliation is a function of its lagged value, i.e., the party identification in the

previous election (Fiorina 1981, see also Alt 1984). Analysts using these models posit

that the "retrospective model" supplies information regarding past experiences with the

party summarized in the lagged dependent variable in predicting the contemporaneous

variable. In the current context, the model is as follows:

PARTYID, = a + b, PARTYIDt., + b2OCLASSt+ b3HOMOWNt + b4UMEMt +

b5EDUCt+b6AGEt+ b7LEADERt + bg PROXIMITYt + et (Eq 3.1)

where: PARTYID, = Party identification with strength at time t; PARTYIDt.j = Party identification with strength at time t-1; OCLASSt = Occupational class at time t; HOMOWN, = Home ownership dummy at time t; UMEMt = Union membership at time t; EDUC = Educational level at time t; AGE = Age at time; LEADER, = Evaluations of party leader at time t; PROXIMITYt

= Distance of issue positions between the voter and party at time t;

a = Constant; bj-bg = parameter coefficients; et = error term.

Including a lagged dependent variable may lead to inconsistent estimates given

the correlations between the variable (PARTYID^) and the disturbance term (e,)

(Gujarati 1988, Green 1990).8 To circumvent this problem, the two-stage least squares

method is employed with the use of instrumental variables (Greene 1990, Kennedy 1992).

In the first stage, ordinary least square analysis is applied with PARTYIDt_, as the

dependent variable. In this first stage model, the independent variables are the lagged

60

versions of those in the original model, i.e. age, education level, union membership, home

ownership, occupational class, leader evaluations and issue evaluations in 1987. The

fitted dependent variable-namely, the predicted party identification in 1987 ~ is utilized

as the instrumental variable for estimation in the second stage model.9

The 2SLS models as shown in Table 3.6 demonstrate the effects of various factors

on the party identification. Instead of modeling the change in partisan strength (i.e. the

intensity of support for a single party), I chose to analyze partisanship in the three parties

separately in order to compare the effects of different variables in different groups of

party identifiers. Because the dependent variable is designed to measure changes in

strength of party identification in both directions, it furnishes a better device to examine

both changes in strength of party loyalties and changes in party affiliation.

As expected, the lagged party identification is the single most powerful predictor

among others. This variable, as Fiorina notes, represents the "past political experiences

of the socializing agents" that provide ready guidance for political choices (1981: 76). As

derived from the estimated value in the first stage of regression, the lagged dependent

also serves as a proxy of the collectivity of various forces that formulate political identity

in the previous election. In all three parties, this variable scores statistical significance on

the most stringent level (p<0.01), with respectable partial effects (unstandardized B's =

.648, .749, .744 for Conservative, Labour and Liberal Democrats, respectively). Model

diagnostics suggest no clear sign of collinearity among the independent variables.10

In the Conservative model, two of the three class variables, home-ownership and

61

occupational class, fail to exhibit statistically significant effects. They indicate that

whether one owns his or her own home, or whether one has a blue-collar job, has no

statistically significant connection with party choice. Union members, however, are less

likely than others to identify with the Conservative party. Among Conservative

identifiers, less than 30 percent have union membership and the mean strength of these

unionized Tories is statistically lower than the rest by .69 (t-test=-6.35, pO.OOl, results

not in tables). Intriguingly, Conservative identifiers are the only group of partisans for

which the age factor is statistically significant, indicating that older people tend to have

stronger support for party. Asymmetry of the impact, however, suggests that it does not

necessarily define a partisan divide between Tories and non-Tories. Its influence, which

is confined to the Conservative identifiers, is, at best, from slight to modest (b=0.004).

Compared to its two counterparts, Labour is clearly a class party. All three class

variables—occupational class, home ownership and union membership—register

statistically significant effects for Labour identifiers, suggesting the Labour party is still

attracted to the working class, the housing estate tenants and union members. This

finding illuminates the fact that class, measured in all standards, has a definitive role for

the Labour party. Support for party, ceteris paribus, is partially contingent upon the

working class attributes among the Labour identifiers.

It is evident that the class variables have different effects for the three parties. In

contrast, both the issue and leader effects delineate statistically significant results among

all party identifiers. Regarding the former, the more favorably the leader is rated, the

62

more likely it is that a respondent will identify with his or her party with positive

proportional strength. For the issue evaluation variables, the significant results across

three models strongly suggest that the issue space exists in the public mind and that it

operates to provide a sense of perceptual or ideological distance between party and

voters. This is manifest in all party identifiers: the closer the ideological distance

between party and respondent, the stronger the attachment to the corresponding party.

In general, the Conservative model registers the best fit, accounting for 71 percent

of variance, as compared with 60 percent in the Labour model. In contrast, only 30

percent of the variance in the Liberal model is explained. Noteworthy, also, is the

relative importance of each class of variables in the three partisanship models. Table 3.7

reports the beta (standardized) coefficients of the independent variables in the

abovementioned models. Among all, issue proximities present is the most important

factor in affecting party identifications (except previous party affiliations). With regard

to the party leader effect, Prime Minister Mr. John Major was ascribed greater salience

compared to the other two leaders when party identification is considered. In other

words, Conservative identifiers are more likely to seek the judgment of the government

leader in deciding their strength of affiliation to the party (Beta=.14). Interestingly,

evaluations of party leaders are barely a function of political experience and seniority.

Mr. Kinnock, for example, has been the Labour leader running for the general election

since 1983, while both Mr. Major and Mr. Paddy Ashdown were entering their first

election in 1992. Yet, the mean score for Mr. Kinnock ranks last, at 4.38 on an eight-

63

point scale, lagging far behind Mr. Ashdown (6.35) and Mr. Major (6.0). Mr. Ashdown's

lead, however, was not a surprise. Since he assumed leadership of the reorganized t

Liberal Democrats from the ruins of the disbanded Alliance, the party became more

coherent with a better foundation at the local level. For his seriousness about policy and

personal image, the Liberal leader's popularity exceeded that of Mr. Kinnock and

paralleled that of the prime minister (Butler and Kavanagh 1992). This is consonant to

the results in the party identification models in which Kinnock's leader effect was

considered the weakest in comparison to that of his counterparts.

3.5 Conclusion

In this chapter, we have addressed the issue of class-party linkages in the context

of a post-dealignment Britain. A model of party identification taking into account the

long-term social forces and the emergence of short-term forces based on the instrumental

judgment of the voters is identified and tested. During the 1979-1992 period, the British

electorate has a generally weakened but stable attachment to political parties. Structural

changes in British society and the economy have provided sources for the electoral

metamorphosis. Oscillations in the trend of partisanship, however, have not been as

remarkable as in the 1970s. The influence of social cleavage, for instance, does not

entirely wither away with regard to affiliations to political parties. Among Labour

identifiers, traditional conditions for defining a "working-class" people still have validity

despite the significant contraction of the percentage of such people in the population. A

64

decline in class voting was evident in absolute terms. For non-Labour identifiers,

however, the class divide is too obfuscated for political identification. Instead, voters

from this category seek other cues when making party and voting choices. In this vein,

we proclaim the emergence of instrumentalism in the post-dealignment era: political

identity becomes a function of a voter's individualized judgments to issues and party

leader image. Evaluations based on available cues from the environment, particularly

through the transmission of mass media, are the major determinants on not just what

individuals identify but on how they identify with political parties. In the next chapter,

we will look closer at the consequence of the rise of instrumentalism and its impact on

voting behavior.

65

Endnotes

1. The scales are: I Higher managerial or professional; II Lower managerial or administrative; III Skilled or supervisory non-manual; IV Lower non-manual; V Skilled manual; VI Unskilled manual (Butler and Stokes 1969:70, see also Kahan, M.J., David Butler and Donald Stokes 1966. "On the Analytical Division of Social Class" British Journal of Sociology 17:127).

2. Ibid.

3. The index was originally designed for measuring impact of class on voting. I modified the index by substituting the voters' percentage with that of identifiers in examining class-party identification.

4. The results are as follows:

Partisan Strength 87 Partisan Strength 92

Occupational Class 0. .045 0. .022 (0, ,085) (0. .396)

Home Ownership -0. .046 -0. .065 (0, .072) (0. .009)

Union Membership -0, .018 -0. .029 (0, .486) (0, .258)

* p-values in parentheses

5. The arguments on the use of the traditional scale versus the new scheme proposed by Heath et al. were detailed in Political Studies (1986) vol. 34. Crewe, one of the defenders of the former measure, contended that the new scheme was unable to capture evolution in class structure as the changes to the two new classes, salariat and proletariat were too small to reflect the mutabilities of social cleavages.

6. Four questions were selected to measure the respondent's perception of party leader if they are capable, caring, extreme and caring for one class. The questions are: "Do you think (party leader) is extreme/good for one class/capable/caring ?"

7. The variable is derived from five sets of issue position variables with a scale ranging from 1 (agree) to -1 (disagree). Each set carries four questions asking respondent's own position and his perceived positions of the three parties on a certain issue. The formula from which the spatial variables derive is as follows:

66

PROXIMITY^ Eflti-Itik)2]172

where: PROXIMITY^ = Issue Spatial variable with reference to party k at time t In = Voter's own position on issue I at time t Itik = Voter's perceived position of party k on issue I at time t.

8. The correlation between the disturbance and the lagged dependent variable is estimated at a statistically significant .488.

9. The 2SLS model is estimated using the Hatanaka estimator. Hatanaka's efficient two-step estimator is required if a lagged dependent variable is included in the 2SLS model. A lagged version of the independent variables should be estimated in the first stage also. In computing the estimator of o2 in the model, e'e/N is used, not e'e/(N-K), i.e., there is no degrees of freedom correction (Green 1991).

10. Although some of the pairwise correlations among the independent variables are moderately large~the highest one being -0.62 between leader evaluations and issue evaluations in the Conservative model-both the tolerance (0.585) and Variance-Inflation Factors (1.708) give no evidence to a near-dependency (Fox 1991).

CHAPTER IV

NEW DIRECTIONS IN ISSUE VOTING

4.1 Introduction

In the previous chapter, we examined a model of party identification as a function

of class factors and other short-term forces such as individual evaluations of issues and

party leaders. Voters' judgments on issues and party leaders, as Chapters II and III

exposed, have emerged to be the major factors in affecting partisan change since 1979.

The decline of class voting in the 1960s and 1970s gave rise to an individualized decision

on party choice that calls for a "rational" assessment of available information. In this

chapter, we continue our inquiry along this line and reason that policy issues provide the

information for this rational decision-making process. In particular, we focus on the

impact of issues in terms of their "spatial" relation with the voter, which has become a

major topic in recent research on electoral behavior (Enelow and Hinich 1984). To start

with, we first perform a graphical simulation of issue effects, followed by a multinomial

discrete choice model that examines the voting behavior in the British multiple-party

context.

Analyses of the relationship between voters' opinions on issues and their voting

behavior were first conducted in American electoral studies. Among the pioneering

studies was Converse's work on the American public's belief system and its influence on

67

68

voting decisions (1964). He argued that the "issue" impact was volatile and unclear.

Voters' positions on policies and issues were not as stable as was suggested in the

traditional model of liberal democracy. His arguments place in doubt the existence of a

"real" attitude toward issues and policies among the general public. In parallel studies in

Britain, Butler and Stokes echoed what Converse contended in influence of issues on

voting outcome. They suggested that voters' issue orientations were subject to change

and were "remarkably unstructured and inconsistent over time" (1969: 341-358).

According to what they found in the 1963 and 1964 election surveys, a majority of voters

in the 1960s did not have clearly formulated issue positions and their knowledge about

even highly salient issues was limited.

Analyses of political issues used a heuristic image of an "informed spectator" in

making choices of policy alternatives embedded in his/her expectation of parties and

government. Based on this assumption, Butler and Stokes tested against the "informed

spectator" model using survey data collected from the 1963,1964 and 1966 panel

surveys. They found that the public's understanding of policy issues fell away "very

sharply" from how the ideal "rational voter" should behave. Not only are the electors

poorly informed, but also they seldom understand the meaning of issues and do not

possess consistent attitudinal orientations towards policy alternatives. Public opinions

towards nationalization issue, for instance, were so volatile that a majority of respondents

stepped across the line between supporting and opposing nationalized industries in one

year. In consecutive interviews administered in 1963 and 1964, the authors found only

69

39 percent of the respondents can provide an identical and definite position on the

nationalization issue (1969: 178). In the 1970s and 1980s, analysts reexamined the

influences of issues and found that voters were better informed than were their

predecessors. Franklin (1985) reported that, a decade after Butler and Stokes' study, only

a small portion of respondents (6 percent) were unable to identify two of the most

important problems facing the nation, as compared to 40 percent in the 1960s.

A different approach was adopted to investigate the linkage between voting and

issue positions. Instead of gauging the stability of voters' issue responses, studies of

issue voting focused on the voters' issue positions with reference to those of the political

parties. For instance, Sarlvik and Crewe (1983) used a series of questions asking

respondents' own views on certain issues and their perceptions of the stands of political

parties for each issue. A measure of issue positions was derived from comparing a

voter's own views with his perceived positions of the parties. The authors employed this

measure to examine the importance of issue "distances" on voting decision. Their

findings indicated that, once the individual's issue position is adequately represented, the

"rational" voter model has stronger support.

This "spatial" approach of studying voting or party choice was not novel. In the

1950s, Downs (1957) began to construct a spatial theory based on the voter's utility as a

function of his or her closeness of policy preference to the political party.1 The model

was further developed by rational choice theorists who analyzed the dynamic of

individual and party electoral decisions (e.g., Enelow and Hinich 1984, 1990). The

70

spatial voting theory hinges on the individual voter's position in a hypothetical issue

space. This position, which reflects the voter's ideal set of policy alternatives, is

compared with perceptions of where the political parties stand on these issues. Based on

this comparison, a voter makes his decisions in choosing a party that has issue positions

closest to his/her own in the issue space. Traditional spatial models of issue voting utilize

the linear distance between the voter's and political parties' issue positions in defining the

voting utility function (e.g., Davis, Hinich and Ordeshook 1970; Enelow and Hinich

1984). This model projects a negative relationship between the proximity of issue

positions and the utility returns. In other words, the closer the voter and the party in the

issue space, the higher the utility in supporting and voting for this party.

Recent research on issue voting challenges this conventional "proximity"

approach in specifying issue distance. Rabinowitz and Macdonald (1989) argue that the

proximity model is flawed in its basic assumption that voters possess specific, ordered

policy preferences along a set of issue continua. As they contend, issues operate in a

"dispositional" or diffuse manner that the vast majority of the voters can, at best, identify

which side (direction) of an issue debate they favor. As an alternative, they proposed a

"directional" model that defines issue positions in terms of two qualities: direction and

intensity. The former refers to the issue position the voter sides with in a dichotomous

view of that specific issue. The latter represents the magnitude of support a voter assigns

in evaluation of that issue. Rabinowitz and his coauthors operationalize the utility

function by a scalar product (multiplication) between the issue positions of voters and

71

parties as opposed to the linear Euclidean distance (subtraction) specified in the

traditional proximity model.2 In comparing the traditional proximity model with their

directional model, Rabinowitz and his colleagues found supporting evidence for the latter

in both the U.S. two-party system and the Norwegian multi-party system. According to

their findings, the directional theory provides a better "vehicle" in modeling issue voting

in "conflict- or cleavage-oriented politics" (Macdonald, Listhaug and Rabinowitz 1991).

Rabinowitz et al. 's directional theory does not end the debate on a spatial theory

of issue voting. More recent research on this topic disputes the claims made about the

superiority of the directional model in explaining electoral outcomes. Ivensen (1994), for

instance, applies both the directional and proximity model and a mixed model to the

elections in six Western European countries and finds statistically significant results in all

three models. In a recent article, Merrill (1995) compares the two models in both the

American and Norwegian systems. He discovers that his findings contradict what

Rabinowitz et al have articulated in their earlier studies. He argues that the predictability

of the directional model is dependent on the extent of information the voter possesses to

evaluate issues. Uninformed voters may more likely be influenced by direction, while

informed voters will tend to respond to proximity (Merrill 1995: 285). The author

presents a combined model, taking into account both the proximity and direction of

voters' issue positions, and finds better results than the previous models.

To date, a majority of the spatial models on issue voting have focused on elections

in the United States and nations on the European continent in contrasting issue influences

72

in the two-party and multi-party system. The impact of the two rival spatial models in the

British tri-party context remains largely unexplored.3 To fill this gap, this chapter

attempts to compare the two issue voting models in the British system. Drawing data

from the 1987-1992 panel survey, we evaluate the two rival models by using a

"parsimonious encompassing" principle (Mizon and Richard 1986). To address the issue

of model choice in the multi-party context in Britain, I test the different variants of the

discrete choice model to determine which provides the most accurate characterization of

the British voters' decisions in multi-dimensional issue space.

4.2 Measuring Issues in British Elections

Butler and Stokes (1969) outlined three criteria for an issue to exert influence on

voters:

a. the individual's well-formed attitude toward a salient issue itself;

b. the skewedness or balance between the two rival bodies of such attitude/opinion

on that issue; and

c. the association with parties of each side of the issue in the public mind.

The authors postulated that the theory of issue voting hinges on the assumption

that voters order their preferences for the parties according to their own issue stances on

the left-right dimension, as compared with the political parties' positions. However,

evidence obtained from surveys in the 1960s did not support what Butler and Stokes

embrace. They surveyed voters' views on seven issues and found a majority of the

73

respondents had no recognition of left-right concepts in interpreting these issues.4 They

concluded that an issue voting model built on the assumption of left-right recognition

"seemed virtually non-existent" (1969:214).

With a better measuring device, Sarlvik and Crewe (1983) studied issue voting in

the 1979 general election. They adopted similar requirements as proposed by Butler and

Stokes in 1969, among which they were most concerned with the balance or division of

the issue as perceived by the general public, i.e., whether there is a clear cleavage in

opinions that determines the impact of that issue on the voters' choice. Sarlvik and

Crewe employed a seven-point scale to gauge both voters' positions and their perceived

positions of the political parties. Voters' opinions were then measured in terms of the

distances between their own opinions and their perceptions of the parties' positions. In

contrast with Butler and Stokes, Sarlvik and Crewe's "distance" model produced "very

assuring" results: more than half of the variance in voting decision could be explained by

the issue variable. They also found that the issues which concerned the party's capability

in managing the economic problems like nationalization, unemployment and prices were

most important (1983:276).

As noted above, in a series of recent works on issue voting models, Rabinowitz

and his colleagues question the applicability of the conventional spatial model employed

by Sarlvik and Crewe. They are dubious the linear distance is an appropriate measure to

capture the "closeness" between voter and party from which political support is derived.

Consider a seven-point scale on support of a policy with a neutral point at zero and +3

74

and -3 at two ends. A voter who positions himself at -1 (fairly unfavorable with the

policy) has an equal distance with two parties standing at -3 and +1, respectively. The

conventional proximity model suggests the voter will support the two parties with equal

weights. In reality, however, it is more likely for the voter to choose the party at -3, given

its similar view in the same direction than crossing the line to support another party with

opposite opinion.

Rabinowitz et al. argue that the proximity model ignores two important elements

of issue support in the public mind: direction and intensity. They believe it is implausible

to assume that an individual voter has a well-ordered set of policy alternatives in a sharp

positional fashion. Instead, voters view issues from a perspective of a pro-con

dichotomy. Starting from the neutral point, voters choose one side of the opinion before

deciding the amount of support or intensity of support to their stance on the issue. In

presenting their "directional" model alternative to the traditional proximity model,

Rabinowitz et al. go beyond the "positional perspective" in perceiving the role of issues

and take into account these two elements.

Another key feature of the directional model is the neutral point of opinion in the

public mind. Rabinowitz et al. assign this point to zero not merely to refer to those who

have no opinion on the issue of concern, but also the group of voters to whom the issue

has no impact at all. In this vein, the authors attempt to consider two possibilities: first,

the issue has no impact if a voter is at zero on the scale or has no opinion; second, a

candidate with a neutral position on an issue receives no support from the effect of that

75

issue.

Rabinowitz et al. 's model raises both methodological and theoretical concerns

about issue voting. Regarding the former, they introduce the concept of a neutral point,

from which they develop a measurement for both the direction and intensity of opinion in

lieu of linear proximity. They used a scalar (multiplicative) product between the

positions of the voters and parties to capture the idea of party support. When the two are

in the same direction (sign), for instance, the measure will be positive, representing

favorable votes on that issue. Values with different signs, on the other hand, result in

negative support. Put in utilitarian terms, the higher value on either or both of the

positions, the more is the utility for the voter or support for the party. Substantively, the

directional theory espouses a model of advocacy in which voters are more concerned with

"Are you on my side? How will you work in my interest?" than "How close are your

positions to mine?" (Macdonald, Listhaug and Rabinowitz 1991).

Recent research on the nature of issue effects on voting is divided. Merrill (1995),

for example, is skeptical that the directional component is the best device to model

voters' utility. Empirical tests based on a mixture of directional and proximity

components find significant results on both in different electoral systems (Iversen 1994,

Merrill 1994, see also Merrill 1995). In some cases, like Piatt et al. 's (1992) model on

U.S. congressional roll-call votes, the proximity variable was found to be a better

predictor than the directional variable.

From the methodological perspective, critics also argue that previous research

76

relying on binomial models falls prey to the misspecification problem in modeling the

voting choice in the multi-party, multi-dimensional electoral context (Whitten and Palmer

1996). In "heightening the concern" of model selection in voting research, Whitten and

Palmer lament the inappropriate practice of "structuring electoral choice by grouping

parties along a single continuum" into a dichotomous choice between government and

opposition for ease of estimation (1996: 231-232). As the authors contend, traditional

binomial logit or probit models following this practice disallow the estimation of the

balance of electoral support in the multiparty system. Specifically, probabilities of choice

among different opposition parties that were improperly merged into one group, will be

ignored or inaccurately estimated. To circumvent this problem, they propose that a

multinomial logit model (MNL) be employed to estimate the polychotomous electoral

choice.

Latest empirical research on this methodological issue, however, contests the

choice of the multinomial logit model in analyzing multiparty elections. This method, as

Alvarez and Nagler note, is no better than successive applications of binomial logit, with

which the multinomial logit is recommended to replace (1996). The authors argue that

the major weakness of MNL is its assumption of Independence of Irrelevant Alternatives

(IIA), which implies that "the ratio of probability of choosing one party to the probability

of choosing a second party is unchanged for individual voters if a third party enters the

race" (Alvarez and Nagler 1996). To give an analogy in economics, from which this

model is derived, consider a choice of transportation modes between driving a car or

77

taking a red bus. The relative probability for choosing one of the two should be .5 given

an indifferent commuter. IIA assumes that this ratio will be retained if a blue bus enters

the choice set. This almost approaches impossibility, however, as the blue bus not only

provides a close substitute to the original bus but also the substitutibilities for the red bus

and the car differ.

In place of MNL, Alvarez and Nagler suggest that a multinomial probit model

(MNP), which does not impose IIA, should be employed in modeling multi-party

elections. Unrestricted by the unrealistic IIA assumption, the model also allows for both

choice-specific and individual-specific variables to be included as predictors of the

model, which is an advantage untenable in MNL. Choice-specific variables refer to the

characteristics related to the alternatives in the dependent variables (e.g., the issue

position of the party), whereas individual-specific variables are the characteristics

regarding the individual, such as age and education level of the voter. MNP, however,

also has its Achilles' heel: its high computational cost and its sensitivity to models with

large numbers of predictors. Due to its technical complications, some analysts deem the

method computationally "impractical" (see Whitten and Palmer 1996). Hence, caution

should be taken in estimating a parsimonious MNP model.5

In addressing the two aforementioned issues, I put the traditional proximity

model and the directional model in empirical tests using data from the British election

panel study in 1987-19926 and attempt to identify the multivariate model that best

represents the issue effect in the British election.

78

4.3 A Graphical Analysis

To formalize the impact of an issue on voting decision, I define voting or party

support as a function of voter and party issue positions in two forms7:

1. Proximity model

UP(S) = m -[£ (v - p)2]1/2 = m - [(vr p,)2+ ....+ (vk-pk)2]1/2 (4.1)

i=i

where S is the support a party garners from the issue impact, m is the maximum issue

effect or most support the party can derive from the issue factor; v is the voter's issue

position; and p is the party's issue position. This function utilizes the linear Euclidean

distance or the absolute value between voter and party positions in a k-dimensional issue

space. The neutral point is not at issue in this model; only the distance will be concerned.

Hence, it is not necessary to have a zero point. When the positions of the two are

equivalent, the party gains the most.

2. Directional model

Ud(S) = £ v ^ = v,pi+ ....+ v,j)k (4.2) i=i

In this model, the function follows a multiplicative form, where party acquires positive

support from voters on the same side of a positive-negative scale with zero as the neutral

point. Negative impact results when the two are on different sides. The model is

unbounded on both sides, indicating that support for a party mounts when either one

79

tends to approach the extreme. This, however, carries a condition as specified by

Rabinowitz et al. as the region of acceptability, i.e., location beyond the region or the

extreme positions produces penalties.

The two support functions are illustrated in Figure 4.1 aid Figure 4.2. The three-

dimensional graphs portray the interplay between the positions of voter and party on a

single issue and the support derived from the relationship between the two variables

under the two voting models. The left side panel of Figure 4.1 presents two proximity

models that depict the negative, linear relationship between party support and issue

distance. These two plots represent the scenarios in which party and voters have

congruent, opposite and extreme positions, respectively.8 The upper graph indicates that

voters have stronger support for party as they get closer to the party on issue positions. It

peaks at a point where the two have the same positions. Meanwhile, congruent positions

always generate maximal support. The lower left graph describes graphically four

situations when either one of the two has fixed, extreme positions. In those cases,

support will be solely a function of the other side's position. For example, if the party is

at 11, or the extreme right, voters' support for this party rises as they approach the right.

To further explore these scenarios, the right-side panel elaborates interaction between the

three variables. The upper right graph paints the support function when the party's

positions are fixed at 1, 3, 5 ,7,9 and 11, respectively. Except for the two extremes, the

support functions bounce back after reaching the peak. Note that the lines parallel with

each other, i.e., they have the same slopes. The lower right graph shows the contrary

80

cases in mirror images. From the main plot in the middle, it is noticeable that the support

for a party (z-axis) in the proximity model peaks at the diagonal between the x and y

axes, beyond which the surface declines, with the same slopes on both sides, toward the

ends of the two axes.

The directional model's curvilinear surface is contrasted with the quasi-pyramid

in the proximity model (Figure 4.2). The upper left graph portrays the parabolic

relationships when voters and parties are at congruent and opposite positions. This

reflects that the intensity of appeal is at play, i.e., the increment in support of or in

opposition to the party grows at an increasing rate with issue positions of both the voter

and the party. The stronger the appeal of either the voter or the party for an issue, the

higher the voting return thus resulted, provided that the two are on the same side.

Likewise, a negative effect kicks in, with an increasing rate, when the voter and the party

are at odds on the issue. Despite that the linear relationships at extreme positions (lower

left) appear the same as the proximity model, all four lines pass through the zero point,

indicating the existence of neutrality or the absence of impact. The two models also

differ in slopes. Graphs in the right side panel illustrate the scenarios in which either the

voter or the party has fixed positions. While the lines are all tangents to either the x=0

axis or y=0 axis, they have different slopes as contrary to the proximity models.

Comparing the proximity surface, the directional model as shown in Figure 4.2's middle

main plot demonstrates that the maximum support only occurs at the two ends when the

voter and the party are in congruent positions with highest intensity. Noteworthy is the

81

curvilinearity of the model, which suggests variable slopes instead of fixed slopes in the

proximity model. In this case, voting support is a function of both the voter and party

positions, rather than being independent of the latter, as the proximity model suggests.9

Having examined the properties of the two hypothetical models, we next fit them

into the real situation drawing data from the 1987-92 British election panel study. The

panel survey has a series of questions that were designed to gauge issue positions of

respondents with reference to their perceived positions of the political parties. The series

covers five policy issues that were considered most important in differentiating political

parties, particularly Conservative and Labour (Heath et al. 1985; Denver 1994)10. They

are: 1) unemployment and inflation; 2) taxation and government expenditure; 3)

nationalization; 4) integration with the European Community, and 5) equalization of

income11. For each issue question, interviewers would read two statements representing

two sides of opinions before asking the respondents to place their own positions on an

eleven-point scale (from A to K). For instance, on the issue of taxation and government

services, the respondents will be given the two statements like: "... we should put up

taxes a lot and spend much more on health and social services" and ".... we should cut

taxes a lot and spend much less on health and social services." After indicating their

own positions, the respondent would be asked to give their perceived positions of

different political parties. The distributions of opinions on these five issues are

demonstrated in Figure 4.3.

Displayed in the bar charts are the distributions of simplified versions of the issue

82

responses, which were recoded into a dichotomous scale for illustration purpose. With

+1 representing supporters and -1 the opponents, all the scales have a central point of zero

which refers to those who hold neutrality. As reflected in the figures, opinions on the two

major economic issues—namely unemployment/ inflation and taxation/government

expenditure—are overwhelmingly inclined toward the left side of the scale. The British

public tends to favor creating jobs and cutting taxes rather than keeping prices low and

increasing government spending. Although voters are divided on the two policy issues

on income equalization and integration of the European Community, opinions regarding

nationalization are somewhat balanced, given a substantial proportion of respondents who

put themselves at the neutral point (28 percent). In general, owing to the leftward skew

on economic issues, public opinion is tilted toward the liberal side on an overall left-right

issue scale (see the lower right chart in Figure 4.3). Figure 4.4 presents

percentages of the vote obtained by the three parties by voters' positions on the five

issues. The vote percentages on the y-axis are the proportion of support the parties

received from each group of voters having the same position on the issue. These curves

demonstrate the propensity of voters to support a party that shares their issue positions.

Generally, the closer the elector is to the right, the more likely he will vote for the

Conservative, and vice versa for the Labour. However, voters for the Liberal Democrats

do not show any issue effect, as indicated by the flat slope. Note that the slopes of the

vote share curve vary from one issue to the other, which suggests the different levels of

intensity. Also, most of the curves are not unbounded as to what the directional model

83

predicts. At the intercept of the Conservative and Labour curves is the point where the

voter at that position is willing to trade votes, i.e., the probability for voting either party is

the same. Most of the intersection is on the left of the scale. Interestingly, the intercept

for the general left-right curves is the mean score of the general public's overall issue

positions.

Summarizing voters' views on various issues, Table 4.1 gives the mean issue

position scores of the voters and perceived positions of the three parties. The original

scale was modified by subtracting 6 to give a scale ranging from -5 (most liberal) to 5

(most conservative). If the mean scores indicate the general intensity of concern on

issues, the British public is most concerned with the problem of joblessness, followed by

tax reduction.12 Besides the income equality issue, respondents have the widest gaps with

the governing Conservative party on these two issues. Comparing the three parties, the

public positions the Tories on the moderate right, the Liberal Democrats closer to the

middle and Labour on the left. However, judging from the graphical demonstration and

mean scores does not suffice to decipher how issues affect electoral choice. More

sophisticated statistical analyses are required.

4.4 A Statistical Analysis

In the previous chapter, we tested the party identification model using four groups

of predictors including demographic variables, class factors, evaluations of issues, and

evaluation of political leaders. Informed by previous voting studies (Fiorina 1981, Clarke

84

et al. 1992), I add the dependent variable in the previous model-party identification~to

analyze the voting function. In order to compare the predictabilities of the proximity and

directional variables, two parallel voting models will be estimated, each of which

employs one of the two spatial variables. Also, as recent electoral research has shed new

light on the geographic cleavages in electoral support, I include regional dummies to

examine the widely-cited North-South divide between Labour and Conservative

supporters (Curtice and Steed 1982, Johnston and Pattie 1989, see also Denver 1992).

Because of the qualitative nature of the dependent variable, I first employ a binary

logit model to test the voting intention in the following form (Pindyck and Rubinfeld

1991):

p I (4-3> u l+e'Zi

where Py is the probability for voter i for selecting party j and

Zi = b0 + bjAGEti + b2GENDERti + b3EDUCti + b4INCOMEti+ b5OCLASSti + b6UMEMti + b7HOMOWN„+ b8REGIONti+ b9PARTYIDtli + b10LEADERti

+bnPROXIMITYti (4.4a)

Zi = b0 + biAGEti + b2GENDERti + b3EDUCti + b4INCOMEti+ bsOCL ASSti+ b6UMEMti + b7HOMOWNti+ b8REGIONti+ b9PARTYIDtli + b]0LEADERti

+b11DIRECTIONti

(4.4b)

where:

85

AGEti - age of voter i at time t GENDER* - gender EDUCti - education level INCOME,! • income OCLASSti - occupational, manual/non-manual class UMEMti - Union membership HOMOWNti - home ownership REGIONti - regional dummies PARTYID,., - strength of party identification at time t-1 LEADER,; - evaluations of party leaders PROXIMITY,) - proximity variable DIRECTION^ - directional variable b0 - constant bj - parameter estimates

Table 4.2 presents the results of the binary logit estimates of the voting models,

drawing data from the 1987-1992 panel study. Lagged party identification is among the

best predictors and is statistically significant at the most stringent level (0.001) across all

three parties. With other factors, it helps contribute to a high percentage of correct

prediction in the three voting models (80.7 percent to 90.4 percent). The leader and issue

variables also achieve statistically significant results. Regarding the former, it is evident

that voters' evaluations of party leaders occupy integral positions in ballot behavior

among supporters for all parties. The prime minister's image consistently plays a

significant role, while the opposition party leaders vary on effects of images. For

instance, the leader of the Labour party, Mr. Kinnock, does not make as much difference

for Liberal Democratic voting as do images of Liberal Democratic and Conservative

leaders, Messrs. Ashdown and Major, respectively. Likewise, for Tory voting, Mr.

Ashdown is largely irrelevant. Pace leadership effect issues always matter in voting

86

decisions. Both the traditional proximity and the directional variables register

statistically significant results. Note that because the former is testing against the

hypothesis of inverse relationship between distance and support, negative signs indicate

correct predictions.

The remaining right-hand-side variables exhibit sporadic results. The income

variable is a divider between the Conservative and Labour supporters. Voters in

relatively wealthier groups tend to support the Tories rather than Labour. Income has no

influence on Liberal voters, however. Similar patterns can be found for the age variable:

younger voters tend to side with the opposition Labour, as compared to older supporters,

who tend to opt for the Tories. Not unexpectedly, controlling for party identification,

most of the class and regional variables become trivial. Regarding the latter, I test the

hypothesis that voters in the affluent south incline to elect the Conservative party as the

north supports Labour. Accordingly, the south is treated as the reference group in the

Conservative model, whereas the north serves the same role in the Labour and Liberal

models. As the Labour results demonstrate, southerners, as compared to northerners, are

less likely to vote Labour. Comparatively, while voters for the governing party are

insensitive to regional differences, Liberal Democrats draw support from the richer south.

These estimates provide a preliminary picture of issue influences on voting

choices. Comparing the issue variables, the traditional proximity models for the three

British parties predict almost identical results, as do the directional models. They report

different signs owing to the fact that the former group of models testifies the negative

87

relationship between issue distance and support, and the latter taps into a positive impact

of issue intensity. Both issue variables perform well and both score statistically

significant results at the 0.001 level.

I next examine the hypothesis that the directional issue variable serves as a better

predictor than proximity one, as claimed by Rabinowitz, et al. Comparing the logit

estimates in the voting model does not suffice to discern a winner, and goodness of fit

indices such as McKelvey's pseudo R square and percentage of correct prediction also

fail to tell the difference, given the limits of these two statistics. To achieve this goal, I

employ an encompassing test to determine which of these two rival models excels.

First proposed by the London School of Economics (LSE) econometricians, the

principle of encompassing furnishes a research strategy in determining the adequacy of

competing models (Hendry 1995). The essence of encompassing hinges on "a

standardized comparison of predictability of rival models embedded in a common

framework" (Hendry 1995: 502; see Granato 1992). It requires a given model to account

for the behavior or performance of other rival models (Charemza and Deadman 1992,

Mizon 1984). It follows that a larger model always encompasses a smaller, nested model

which is a subset of the initial nesting model (i.e., M, £ M2 if M2 <= M,)13. For competing

non-nested models, the encompassing test calls for a comparison of variance being

explained by a different subset of parameters (variance encompassing). For M, to

encompass M2, for example, it must be able to predict at least as well as M2 or explain the

variation of errors in the latter. A variance encompassing test, however, does not

88

necessarily lead to mutually exclusive results; it is possible that both models encompass

each other. In that case, we have to conduct an alternative test to determine a better

model, or concede that both models perform equally well.

Currently, there is not a formal test for encompassing in non-linear models. To

implement an encompassing test in the current context, I employ a log-likelihood ratio

(LR) test based on the result from the logit models to approximate the F-test in time-

series or linear regression models. The LR test tests the hypothesis that one of the two

rival, non-nested models has the ability to account for the behavior of the other (Mizon

and Richard 1986, see also Hendry 1995).14 Table 4.3 reports the results of the

encompassing tests in which the general issue effect is tested, followed by individual

issue effects.

The first column presents the LR test estimates with respect to the hypothesis that

the directional model encompasses proximity (null hypothesis directional does not

encompass proximity); the second testifies for the contrary. Entries on the first group of

rows are the chi-squares for the general issue effect, which is derived from the aggregate

of five individual issue variables. Following are the tests for effects of individual issues,

i.e. testifying to whether one model outperforms the other in certain issues.

Comparing the two columns, it is apparent that proximity is a clear winner in all

voting models. All the chi-squared statistics in the second column — which test the null

hypothesis that the proximity model does not encompass directional — are significant,

rejecting the null at the 0.05 level or better with most of coefficients reaching the most

89

stringent confidence level (99.9%). However, the result is conclusive in some of the

Labour models. In the general effect model, for instance, the chi-squares for both tests

are statistically significant, indicating the two models encompass each other. A similar

pattern is revealed in the individual issue test of taxation and government spending, in

which case both of the directional variables demonstrate a slightly stronger impact than

those of its opponent. Thus, despite that proximity registers better results in general, the

strong effect of the directional variable on the Labour voting model is too noticeable to be

neglected.

Are these results peculiar to voting in the 1992 election, or do they obtain in

earlier elections, as well? To answer this question, I employ data from the 1983-1987

British election panel study.15 The analyses are presented in Table 4.4. Intriguingly, the

1987 panel model reveals the same pattern, in which proximity dominates in all voting

models except the case of Labour. In the latter, both proximity and directional variables

(general) encompass each other. The directional influence even surpasses its rival in the

aggregate issue variable (X2=14.224, as compared to 7.567 in the proximity model),

indicating its strong impact on Labour voting. Note that when the issue of defense is

concerned, Labour voters appear to vote on a directional instead of a proximity basis, i.e.,

the former better represents the issue impact in the voter's mind. In accordance with this

finding are the consistently higher correlations between the directional variable and the

Labour vote. As shown in Table 4.5, the directional variable invariably outperforms the

proximity among Labour voters in the 1983,1987 and 1992 elections,16 suggesting that

90

Labour voters tend to respond to the party's issue appeal while supporters for the Tories

and Liberal Democrats favor the party that mirrors their own political views.

From the encompassing analysis, it is reasonable to conjecture that, in general, the

traditional proximity variable is a better representation of the issue impact on voting

choice. This embodiment of issue influence, however, is not static over time and is

strictly invariant across different parties and issues. Labour's anomaly suggests that

voters perceive the leftist party with a different expectation: besides the ideological

closeness, as epitomized by the proximity variable in the current analysis, they also look

up to the party for a role of "advocate" on alternative policies. Nevertheless, care should

be taken in interpreting this finding, as Labour's "directional support" does not

necessarily come from its position as the major opposition party, but from its being the

only leftist party in the British polity. In fact, supporters for another opposition party,

Liberal Democrats, consistently elect the candidates with closest political views, like their

Conservative counterparts.

Rabinowitz, et al. contend that, informed by the ideas of symbolic politics, the

mass-elite linkage is built on the policy guidance delivered by the political parties. In

providing political leadership, political parties derive support from voters who anticipate

electoral cues from party elites. Voters in this model prefer a candidate with the same

issue direction and with intense positions on issues, as well. From what we observe in

the present analysis in the British context, their proposition is contradicted. When

considering a ballot choice, British voters are not as concerned with intensity of issue

91

advocates as with ideological distance; support is based on the party with the closest

political views. Based on this finding, we continue to estimate the voting utility function

using a more sophisticated model that allows for the estimation of multi-party electoral

choice.

4.5 Modeling Multi-party Elections

Previous voting studies have relied on the binary discrete choice models to

estimate electoral choice between incumbent and opposition parties. In the multi-party

context, however, differentials in support among opposition parties would be ignored in

these binary models, because all the opposition parties were collapsed into one group as

opposed to the governing party. Critics in recent research have complained that the

practice was counter-intuitive and misinformed. Finding this method wanting, they claim

binomial estimations are seriously flawed when the multiple party choice is concerned.

In the latter context, more sophisticated models, which allow for estimation of

probabilities for electing different parties, should be employed. Variants of the

multinomial discrete choice models, as originated in econometric studies, have been

proposed to resolve the issue. Among the candidates are the multinomial logit and f

multinomial probit models. While both models relax the limit in the categories of the

dependent variable, each model has technical shortcomings. For the former, its implicit

assumption of Independence of Irrelevant Alternatives, which supposes the probabilities

across various choices being constant and independent, is unrealistic for the voting

92

decision model. The restriction in model parameterization that confines solely to

individual specific variables (characteristics of the individual voter) also poses

substantive difficulties for MNL to estimate probabilities that vary with the attributes of

the choice alternatives (i.e., the parties). On the other hand, despite its more flexible

assumptions, the highly-acclaimed multinomial probit model also presents technical

obstacles given its high computational cost and complexities in implementation.17

To circumvent these methodological hurdles, I apply both models successively in

a discriminative manner. For the initial stage, I estimate a multinomial logit model for

preliminary analysis. Based on the results from MNL models, I then estimate a more

parsimonious multinomial probit model built upon the significant findings from the first

model.

The Multinomial Logit model has the forms as follows:

® ki

Py - z z for j=1,2 (Labour, Conservative) (4.4) 1 + L t=oe id

P« for j=0 (Liberal) (4.5)

'J „ z 1"*" Zk=0e ki

where is the probability for voter i for selecting party j and

Zki = a0 + b£X t i (4.6)

in which a0 is a constant, bj's are logit coefficients and Xti denotes the individual-specific

93

variables including party identification, age, education level, occupational class, home-

ownership, union membership, region and evaluations of party leaders. Because the

multinomial logit model assumes the utility for voter i choosing party j to be a function of

an individual voter's aspects that vary across voters but remain constant across choices,

only individual-specific variables are employed (Greene 1990). In this context, Liberal is

selected as the reference group (0) in comparison with Labour (1) and Conservative (2).

Results of MNL models are presented in Table 4.6.

In comparison with the binomial logit estimates as presented in Table 4.2, the

MNL model gives very similar results in terms of statistical inferences. All estimates are

predicted in correct signs as in the binomial model. Note that the multinomial logit

model provides more sensible estimates by taking into consideration a comparison across

three party choices. In other words, controlling for the probability of selecting another

party, say, the Conservative in the present model, the effect of each predictor on Labour

is individually estimated with reference to the base category of Liberal Democrats instead

of a combined non-Labour group. For instance, as column 1 in Table 4.6 illustrates, age

and income are found to be influential factors to Labour voters as compared to Liberal

Democrats. It suggests that, controlling for Conservative votes, younger voters and lower

income groups are more likely to align with Labour than with Liberals. Similarly, with

regard to regional differences, Labour finds more support in all other regions than

Liberals. The leader image variables have brought to light more intriguing findings:

when a voter contemplates a choice between Labour and Liberal, he or she will not only

94

evaluate leaders of the two parties, but also the prime minister. In contrast, when

considering a vote between Conservative and Liberal, the evaluations of Kinnock become

irrelevant. For the rest of the variables, the Conservative model reflects largely the same

findings, except the less visible regional effect.

A caveat is in order in interpreting the party identification variables and issue

effect from the MNL model. As mentioned before, the specification of MNL allows only

individual-specific variables that are invariant across different categories in the dependent

variables. Party identification and issue proximity, however, are conditional upon the

attributes of the parties and vary from one party to another. Issue proximity between a

voter and a party, for instance, is contingent upon his issue position and his perception of

the party's stance on policies. This perceived distance will not be the same across the

three parties. Similarly, party identification has the comparable, albeit less

distinguishable, choice-dependent characteristic.18 That said, I include the two variables

in the MNL model with other individual-specific variables for two purposes: first, to

supply a priori information for further estimation and, second, to serve as yardsticks for

model comparison.

Despite its restrictive assumptions, MNL provides a feasible and easy-to-

implement modeling technique for preliminary analysis of a multiple-choice model. On

the basis of MNL estimates, we proceed to the next stage of the analysis that focuses on a

parsimonious multinomial probit model that better approximates the reality in the British

multi-party electoral context. The multinomial probit model has two advantages over its

95

logit opponent counterpart. First, with more flexible model specification,19 MNP is not

restricted by the unrealistic IIA assumption that limits the applicability of the model.

Second, as a result from the first point, MNP allows more room for a theory-guided

parameterization of the multiple choice model and provides the analysts with ample

freedom to include different types of variables, choice-specific and individual-specific

alike (Alvarez and Nagler 1996). The multinomial probit takes the same form as the

multinomial logit in (4.4), except its inclusion of a group of choice-specific variables and

its specification in the disturbance term:

Zki - a„ + b£X t i + Cj £W t i + Ej 1=0,1,2 (4.7)

[eoBj.eJ-NtO, S]

In this setup, X(i and Wti denote, respectively, the individual-specific variables and

choice-specific variables, whereas, e, is the error term in alternative / (Greene 1990), ao is

the constant and bf and ct are the probit estimates for the two groups of predictors.

Unlike MNL, the MNP model includes both systemic and stochastic components.

In addition to the predictors, the model also estimates the sigma matrix that contains the

variances and covariances of the k-1 disturbance terms. In the present context, the

variance-covariance matrix bears substantive meaning that pertains to the IIA

assumption. It admits correlations between the probabilities of choosing different options

in the dependent variable, i.e., the party choice. Specifically, a statistically significant

96

covariance term suggests rejection of the IIA or assertion of interdependence between

choices.

The results of the MNP models are illustrated in Table 4.7. Two models are

estimated focusing on the party identification, issue evaluations and leader effects20.

Starting values for estimation of MNP coefficients are derived from MNL estimates. As

shown in Table 4.7, the MNP model performs very well and all the variables are

predicted in correct directions. Most of the estimates achieve statistical significance at

the most stringent 0.001 level and accord well with the substantive theory. Notice that

both of the error covariance terms (o's) are statistically significant, which testifies to the

fact that the IIA assumption is violated. In other words, the probability of choosing one

of the three parties is dependent upon the choices of the other two.

Noteworthy is the substantive implication that the MNP model offers: it takes into

account both the individual attributes and choice-dependent variables, as exemplified by

the issue proximity variable. In the previous MNL model, all predictors are estimated as

if they are independent from the choices, viz. parties, which is an undue assumption for

certain variables. By definition, voters' issue distance does actually vary with respect to

different parties. In this sense, MNP excels its logit rival by providing a modeling

framework that better approximates the real world. Moreover, the model's liberation

from the IIA assumption allows for an estimation of the voting decision that incorporates

information from both the individual choosers and the choices. In contrast, MNL

presents a deterministic model in which voting decision is determined by individual

97

characteristics such as age, education level and socioeconomic status. In real life, a

political choice is a far cry from what this model projects. A voting decision is not

simply a utility formula dependent upon static input values as derived from the

predetermined personal attributes. Instead, ballot choice is an informed, rational and

interactive process that is dynamic in nature and is conditioned by both the systemic and

stochastic components. To serve this goal, the MNP model provides an appropriate

representation of this complicated process.

4.6 Conclusion

In this chapter, we utilize different methods to explore the subtlety of the electoral

choice process as conditioned by different groups of long-term and short-term factors.

Comparing rival models of issue voting, we discover that, in general, the traditional

proximity model provides a better representation of issue impact in the British context.

The directional model, which finds significant results in the United States and other

European countries, also presents applicability in Labour voting. The finding indicates

that the directional model performs better in a context where parties have strong appeals

and tend to take a role as advocate on alternate policies. We pay particular attention in

modeling the issue factors as informed by the proposition that voting decisions are a

product of a rational and educated evaluative process. This process is an interaction

between the choosers and the choices-in this case, the political parties-on the basis of

available political information. To this end, I test the hypothesis that issues provide an

98

important source of political information for voters to apply their calculus of utility when

a choice of government is considered. Although our focus has been on modeling the

issue effect on ballot choice, we also overcome methodological obstacles in examining

the electoral decision-making process, which is interactive and dynamic in nature. We

have found that the issue proximity--or the ideological distance between voters and party-

-in an MNP model provides a robust representation of the former feature. As our next

task, we proceed in the following chapter to investigate the dynamics of political support

over time in the interim between elections.

99

Endnotes

1. The spatial theory of voting has its very first origin from Hotelling's work on economic decision analysis (1929). This line of research was introduced into voting studies through the analysis by Black (1948,1958) and Downs (1957).

2. The Rabinowitz et al. directional model is in a form of: k

U(V) = E ....+ Vkpk 1=1

where vL vk represent the voter's positions and p]...pk represent the party's positions in a k-dimensional issue space.

The traditional proximity model is:

U(V) - - I V - p I = - [(v,- p„ J+. . . .+

See also endnote 7 in Chapter III.

3. Ivensen (1994) has applied the issue voting models on six European countries, including Britain. His model, however, employs the simple binary rankings of the proximity and intensity (i.e., 1 represents closer in distance and higher in intensity, 0 otherwise) instead of the original spatial variables.

4. The issues covered in the 1963 and 1964 surveys were nationalization, nuclear weapons, entry into Common Market, immigration, death penalty, influence of big business and importance of the Royal family.

5. For instance, an MNP model written in GAUSS code with 10 predictors took a Pentium-133 five hours to reach convergence. Another model with more predictors, however, failed to converge.

6.1 do not compare the issue effects in all the elections between 1979 and 1992 owing to the differences in issue questions in the four elections. Differences in scaling pose another difficulty for comparison.

7. Unlike the traditional approach, I define the support function instead of the utility function as in Rabinowitz et al. (1989,1991) and Merrill (1995). These two, however, are two sides of the same coin. The support function is utilized in the latter part of the chapter for a multivariate model.

8. Note that in a three-dimensional graph, the vertical axis z (the support variable) is the dependent variable, whereas the horizontal planes x and y are independent variables.

100

Hence, the line that connects the root of spikes on the floor plane does not refer to a relationship between x and y.

9. Under the proximity model, support bounces back in the same slope but in the opposite direction. Also, at the point where the two positions are equivalent (i.e., v=p) support reaches the maximum, which is a constant and independent of party position.

10. The British General Election Surveys also cover other issues such as racial equality, class conflict and abortion policy. These issue questions were not employed in the current study because they are merely designed for collecting respondents' own opinion without being referenced to the perceived positions of political parties.

11. Only five out of seven issue questions were selected for this study since responses from the other two questions, defense and standard of living, were collected from a separate subsample.

12. In this case, the position score reflects two qualities: direction and intensity of concern to the issue. While zero represents a point of no interest and no position, the two extremes signify the most favorable and unfavorable with highest interest.

13. The notation, is used to represent the encompassing relationship, which properties are reflexive, anti-symmetric but not necessarily transitive (Hendry, 1995).

14. The chi-squared statistic is derived from differencing the log likelihood ratios of the model and the joint model that contains the variables in both rival models.

15. Issue coverage in the 1983-1987 panel survey was slightly different from the 1987-1992 survey. It covers issues such as inflation/unemployment, taxation/welfare, defense and nationalization. Equalization of income and issue of European Community were not available in 1983.

16. Correlations between issue effects and voting choice in 1979 were not reported, as questions on issue distance are in a different format in the 1979 survey.

17. Because of its need to evaluate multiple integrals of the normal distribution in error covariances, MNP has long been considered as impractical and computationally inefficient. See Greene (1990) and Whitten and Palmer (1996).

18. The fact that the existing coding scheme for party identification (PID) is designed for binomial models renders the variable somewhat in between the choice-specific and individual-specific characterization. A Labour identifier with a party strength of 3, for instance, will have the score -3 by default for both his Conservative and Liberal PID. The variable falls within the choice-specific category, notwithstanding its limited variation.

101

19. Since MNP allows for the correlation among the error terms, it is free from the restriction of the IIA assumption that sabotages the applicability of MNL (Alvarez and Nagler 1996, see also Hausman and Wise 1978).

20. For the sake of parsimony and computational efficiency, the voting models were reparameterized to focus on the major key variables, namely, party identification, evaluations of party leaders and issue proximity. Other class variables and demographic variables were dropped due to statistically insignificant results in full models. While preceding analyses using other models such as 2SLS and MNL demonstrate that including these variables does not significantly contribute to model prediction, estimation of the full model with all ten predictors is computationally costly. See endnote 5.

CHAPTER V

THE "PRESIDENTIALIZATION" OF PRIME MINISTERS:

MODELING PARTY SUPPORT BETWEEN ELECTIONS

5.1 Introduction

To date, we have studied the factors that contribute to change in British voting

behavior at the individual level. Micro-level analyses in previous chapters furnish a

general foundation in understanding the "electoral evolution" among individual British

voters. We examine in these analyses the long-term social factors and short-term

perceptual factors that contribute to electoral change in the Thatcher-Major era. With

primary focus on the surveys administered before or after the general elections, the

preceding investigations provide snapshots of the nation's political landscape. However,

as noted in Chapter IV, voting behavior is dynamic in nature and variable over time. The

volatility in voting choice ought to be studied by following the temporal trajectory of

public support for political parties. To that end, the current chapter is devoted to the time

series analysis of party support at the aggregate, "macroscopic" level. In particular, we

pay special attention to the electoral impact of the British prime minister, who has

become the focus in recent studies of "presidentialized" elections in the Western

parliamentary systems (Mughan 1993). Employing Kalman filter and cointegration

modeling techniques, this chapter investigates the validity of the presidentialism thesis by

102

103

analyzing the dynamics of the impact of prime ministerial approval on governing party

support.

The organization of the chapter is as follows: part 1 identifies the three models of

party support with special focus on leadership effects; part 2 presents time series analyses

of governing party support based on the Gallup poll data drawn from the 1979-1996

period. The results of these analyses form the basis for a Kalman filter model in part 3,

which tests the structural stability of the relationship between prime ministerial approval

and governing party support. Part 4 concludes.

5.2 Leader Matters, Leader Matters Not? Three Prime Ministerial Models

Early studies of British elections assign little weight to the electoral effect of

leader images. Voting studies in the 1950s were almost unanimous regarding what party

leaders do in elections: they were "secondary components." In the face of the

preponderance of long-term forces such as social class, leaders played little role in

electoral choice. Their contribution, according to some analysts, is at best the

"embodiment of general political attitudes" (Trenaman and McQuail 1961). Similarly,

Butler and Stokes were reluctant to classify leader images as one of the major

determinants of electoral outcome (1969). The resurgence of the Tories in the 1960s, as

the authors argue, has no necessary tie to the general public's support for the then

Conservative prime minister, Mr. Heath. The authors maintain that the importance of

leader images cannot be discounted, but that it is only among the factors that affect

104

"transient shifts of party strength." In other words, only when there is a marked, huge

imbalance between the two leaders will such a difference be made. This effect will easily

be "outweighed by other issues and events" of public concern (p. 388). Under the

traditional "leaders-matter-not" model, leadership effect merely played in the

background.

Erosion of social-class influence in the 1970s has prompted recent scholarship to

call into question the "unimportance" of leader images. Empirical studies in the 1980s

and the 1990s reexamine the electoral impact of political leaders using more sophisticated

methods and find notably dissimilar results. Researchers noticed that the partisan

dealignment in the 1970s brought substantial changes to the role of political leaders in

guiding voting behavior (Bean and Mughan 1989, Graetz and McAllister 1987, Stewart

and Clarke 1992). As the long-term forces generated from stable class alignment began

to wither in the 1970s, electoral choice was open to individual volatility amenable to

short-term influences. This is fostered by the activities of the mass media and the

commercialization of electioneering. Stewart and Clarke (1992), for instance, argue that

the high visibility of political leaders on television during election campaigns renders the

leader image a salient factor in a voter's choice of government. Multivariate analyses in

the same study demonstrate that the leadership variable even outperforms other factors,

such as economic evaluations and issue concerns in the 1987 general election (pp. 466-

68). But not every analyst was convinced that the traditional model had been displaced

by this "leaders-do-matter" model. Heath and his associates, for instance, posit that party

105

leaders are negligible vis-a-vis other social and political sources of electoral change

(1991). Despite their long tenure as party leaders and prime ministers, for the authors of

this study, Margaret Thatcher and Harold Wilson might not have existed to effect change

in the British electorate.

The growing body of research in the 1990s finding evidence for the leadership

influences, however, renders the "importance" of leader image-particularly government

leaders-undeniable (Mughan 1993; Crewe and King 1994; Nadeau and Mendelsohn

1994; Clarke and Stewart 1995a, 1995b, 1995c; Nadeau, Niemi and Amato 1996 and

Jones and Hudson 1996). The notion that party leaders matter in affecting voting choice

has, in practice, become the new conventional wisdom in contemporary electoral research

(Nadeau, Niemi and Amato 1996). There persists, however, several methodological and

substantive issues that concern researchers in the field. One of the primary facets is to

distinguish the general public's approval of leaders from the support for their political

parties. As Clarke and Stewart have pointed out, correlations between the two series are

"too close for comfort" (1995a). This concern calls for empirical answers to two major

questions: Are the two variables readily distinguishable from each other? If so, is one

exogenous to the other? The authors identify prime ministerial approval as weakly

exogenous to governing party support. Moreover, the two series are cointegrated, i.e.,

they tend to "travel together" in the long run (p. 163). They also argue that the

replacement of Mrs. Thatcher with Mr. Major witnessed a temporary divergence of the

two series—a divergence that lends weight to the proposition that party support is

106

distinctive from leader evaluations.

With the empirical significance of the "leaders-do-matter" model being

established, the next question concerns the extent of influence of party leaders on election

outcome. Studies that take into consideration the campaign strategies of political parties

point to a trend of so-called "presidentialization" of leader effects. Representative of this

stream of research is Mughan's work, articulating the growing influence of party leaders

in the age of mass media. He argues that elections in parliamentary systems have become

more presidential in the sense that parties are more inclined to project their image through

their leader in mass media during election campaigns. Political parties seek exposure on

television and coverage in news reporting-particularly during the 1987 and 1992 British

general elections-by highlighting their leaders via commercials and personality-based

campaigns (Mughan 1993, see also Farrell and Wortmann 1987 and Bean and Mughan

1989). Mughan concludes from analyses based on the 1964-1992 election results that

voters' preferences for prime minister have become more influential in recent elections.

He also discovers the partisan dealignment in 1974 and the 1987 election were the onset

dates for this "leaders-matter-more" or presidentialization model of general elections.

While acknowledging the impact of party leader image on voters as substantial,

researchers drawing inferences from studies at different levels of analysis have yet to

arrive at unanimity on the leadership model. This is partially due to the fact that a

substantial portion of the empirical findings were based on micro-level (individual)

analyses, while others were derived from macro-level (aggregate) time series studies. In

107

this chapter, we focus on the aggregate level of analysis in an attempt to arrive at a

general model of leadership effect. In the following section, we proceed from the basic

model in which government support is a function of prime ministerial approval, economic

evaluations and political interventions. Employing time series data drawn from Gallup

polls covering the period of Conservative tenure since 19791 (July, 1979 to April, 1996),

we analyze the three models, paying particular attention to leadership effects on

governing party support.

5.3 Modeling Leadership Effect on Party Support

When examining the relationship between governing party support and approval

of prime minister, Clarke and Stewart (1995) counsel caution. The authors are concerned

with the widely accepted conception that the two measures are in fact "tapping the same

thing" (p. 156). A graphical plot of the two variables reveals that the authors' prudence is

not unwarranted. As Figure 5.1 demonstrates, the two series move very close to each

other. With a high correlation of 0.905, the only visible divergence occurs in a short

period between where Major replaced Thatcher as prime minister and the 1992 election.

Such a departure in this period, nevertheless, explains the fact that the two series are

individual, separate variables (Clarke and Stewart 1995: 156-57). Bearing in mind this

observation only provides intuition, statistical evidence is in order.

The fact that the two variables trend up and down closely together provides a

strong hint of the existence of what is designated as a cointegrating relationship. A group

108

of nonstationary variables will be regarded as cointegrated if a linear combination of

them constitutes a stationary series. Substantively, cointegration suggests a long-run

equilibrium relationship between two or more non-stationary series being maintained by a

mechanism that corrects the departure of any of the series from the common trend (error

correction mechanism). In other words, whenever there is an abrupt movement of either

series, its close tie with the other cointegrated series will subsequently offset such a

departure and adjust the movement of the series until the original equilibrium is

reestablished (Cuthbertson, Taylor and Hall 1992; see also Clarke and Stewart 1995 and

Ostrom and Smith 1990). This accords well with the trajectories of prime ministerial

approval and Conservative party support as illustrated in Figure 5.1.

The test for cointegration requires two conditions to be met: first, the series of

interests are integrated of the same order and, second, the linear combination of the

series has a lower order of integration. To determine if there exists a cointegrated

relationship between Conservative support and prime ministerial approval, a series of

Dickey-Fuller unit root tests is performed to demonstrate the stationarity of the two

variables (Dickey and Fuller 1979). As shown in Table 5.1, which presents the findings

of the tests, it is clear that both the prime ministerial satisfaction (PMS AT) and governing

party support (CONS) are non-stationary. That the differenced versions of the variables

exhibit stationarity suggests these series have unit roots and follow an 1(1) or first-order

autoregressive process (Cuthbertson, Taylor and Hall 1992, pp.129-32). The stationarity

(1(0)) of the linear combination of CONS and PMSAT, ZCONSPM2, provides evidence

109

that the two series are cointegrated with a common trend.

Modeling nonstationary variables often leads to a methodological dilemma.

While applying regression analysis to these series risks misleading inference and

"spurious regression" (Granger and Newbold 1974, Hendry 1995), first-differencing the

variables to achieve stationarity could hardly be spared from the criticism of "throwing

the baby out with the bath water" and forgoing potential long-run relationships

(paraphrased from Mills 1990; see also Beck 1992 and Hendry 1995). To circumvent this

problem, an error correction model is proposed to estimate the cointegrated series with

the "error correction mechanism" that captures the long-run equilibrium relationship

(Beck 1992, Hendry 1995, Clarke and Stewart 1995a). ECM also provides a setup for

testing the assumption that support for the prime minister is distinct from support for his

or her party. The test for weak exogeneity, as suggested in Charemza and Deadman, can

be performed within the error correction model that estimates both the long-term and

short-term elements (1992; see also Clarke and Stewart 1995a). The test for weak

exogeneity stipulates two conditions: first, the error correction mechanism variable

(thereafter ECM term) should be statistically insignificant in the model for the

hypothesized exogenous variable (prime minister approval); second, the residuals from

the model for the exogenous variable should be statistically insignificant in the model for

the endogenous variable.

The setup for the error correction model that examines cointegration between

prime ministerial satisfaction and party support is as follows3:

110

A CONSt= p0+ p, APMSATt+p2ECMt.l+p3ECONEVt+p4EVENTt+p5FALKt+

P6MAJORt+p7JOHNt+p8EL83t+ p9EL87t+ pi0EL92t+et (5.1)

where:

CONS = Conservative party support; PMSAT = prime ministerial approval; ECM = error

correction mechanism; ECONEV = subjective economic evaluations; EVENT = political

events; FALK = Falklands war; MAJOR = Major replaces Thatcher; JOHN = Major

reelected as party leader; EL83-EL87 = Election dummy for 1983,1987 and 1992; po =

Constant; p(= regression coefficients; et = error term; A = differencing operator

Informed by previous studies, the error correction model of Conservative party

support is composed of four major classes of variables: prime ministerial approval,

economic evaluations, political events and interventions and an error correction

mechanism (Clarke and Stewart 1995a, 1995b; Nadeau, Niemi and Amato 1996). We

hypothesize that political support for the British governing party is a function of support

for the prime minister, the public's subjective evaluations of economy, political shocks

and events that color the public's impression of government performance and, last but not

least, a long-term equilibrating force that balances prime ministerial support and

government popularity. Among them, the economic evaluation variables warrant further

elaboration. The "subjective economy" variables include four types of economic

evaluations, i.e., voters' national versus personal evaluations in forward-looking or

backward-looking perspectives. In the British case, Sanders and his colleagues have

I l l

argued that voters are guided by their personal economic expectations when making

voting decisions (e.g., Sanders et al. 1987,1990; Price and Sanders 1993). The personal

expectations model (hereafter PE model), according to the researchers, provides the best

predictor among the four subjective economy variables in forecasting the 1992

Conservative victory (Sanders 1993). Subsequent research in political economy,

however, finds that while personal economic expectation is an important factor, other

dimensions of the same variable like national retrospective evaluations, alongside such

other political intervention variables as the Falklands war also matter in voting choices

(e.g., Clarke, Mishler and Whiteley 1990).

To address the issue of subjective economy on electoral choice, I test the four

variants of the economic evaluation variable in the error correction model in comparing

Sanders et al. 's model with the other three rivals. The results are shown in Table 5.2.

The four models document the short-term relationship between public support for the

prime minister and her or his party, as well as the long-term profile of the country's

political economy in the past twenty-seven years. All models perform satisfactorily as

illustrated by the diagnostic tests on autocorrelation, functional form, model normality

and heteroscedasticity. Regarding the parameters, satisfaction with the prime minister,

for Mrs. Thatcher or Mr. Major regardless, is evidently a major factor for government

support, and such a long-term relationship is maintained by an error correction

mechanism (ECM) accounting for 27 to 28 percent in monthly adjustment. In other

words, a 10 percentage point drop in either series will, ceteris paribus, be re-equilibrated

112

by about three points in the next month and will vanish within five months.4 However,

the fact that the party has been plagued by a series of unfavorable interventions,

particularly under Mr. Major's tenure, renders any recovery ephemeral. The predicament

precipitated by the widely criticized poll tax pronounced in England in March 1990, for

example, had cost the party at least three points in popularity. The impact lingered on

and was eventually translated into more severe political damage for her leader, Mrs.

Thatcher. Mr. Major's replacement as party leader in November 1990 did actually regain

some Tory lost ground, inasmuch as he claimed in his reelection to premiership in 1995

(five and seven points respectively), other political events such as the currency crisis in

1992 and the continual loss in by-elections were yet more destructive than the momentary

rebounds.

Economic factors in the model reveal no less subtlety than the political

components. Conventional wisdom advises that economic evaluations play key roles in

determining government popularity (Sanders et al. 1987; Clarke, Mishler and Whiteley

1990; Lewis-Beck 1988), but these effects are not ubiquitous. Consonant with an earlier

study, ECM models demonstrate that not all the economic evaluation variables have

statistically significant effects on governing party support (see Clarke and Stewart

1995a). Among them, the PE model performs the best both in terms of model fit

(adjusted R2 = .571) and size of impact (.62). While the personal retrospection and

national prospection variables score marginal results (b=0.51, t=1.599; b=0.018, t=1.488,

respectively), national retrospection is far from achieving statistical significance (b=0.01,

113

t=0.726). Encompassing tests pitting the four together, however, yield ambiguous

results,5 suggesting that the personal expectation model may not be the best performer as

claimed in Sanders et al. This finding also prompts the argument that the underlying

influence of the latter three variables may operate through other forces, such as the

evaluations of the prime minister. To explore further evidence on the importance of

different dimensions of "subjective economy," a parallel set of prime ministerial

approval models is tested, with results shown in Table 5.3.

The same battery of political intervention variables is applied to the prime

ministerial satisfaction (PMSAT) models, in which the dependent variable and the

economic evaluations are in first-differenced form.6 Most political intervention variables

were statistically significant, with the effects of the variables that signify Major's

replacement of Thatcher and the currency crisis outperforming the others. The sizes of

the parameter estimates, which vary greatly from the party support model, suggest these

events' effects are of a much larger scale on people's judgment of the prime minister:

Major's succession as government leader registered a head start of at least 22 points

higher than his predecessor; the currency crisis, unfortunately, cost him a free fall of

about 18 points. The huge difference in effect size between the two models indicates that

these political events are filtered through the voters' judgment of the prime minister

before exerting influence upon the public's support for the governing party.

On the economic evaluation variables, a similar game rule applies. The public's

retrospective evaluation (NR) of the general economy is highly significant in the prime

114

ministerial model, while it is relatively trivial in the Conservative support model. For the

rest of the four, personal expectation is statistically significant with respectable size of

impact (b=. 114, t=2.685), as compared to personal retrospection (b=.76, t=l .512) and

national expectation (b=.053, t=2.785). Encompassing tests provide a clearer picture for

comparison: the NR model encompasses all, as the PE only does on personal

retrospection (see Table 5.4). This testifies to the notion that, when it comes to the

judgment of the head of the government, the public performs the evaluation in a different

manner than the way it does on the party. The records of the national economy become

the essential source of information. Yet, will the evaluations of the prime minister be

directly applied to public support for the party, or will the reverse be true, given the long-

term relationship as exposed in the error correction models? Apparently, the answer

seems obvious. Nevertheless, justification of the answer warrants an empirical test of

exogeneity between the two.

Following Charemza and Deadman (1992), the test is performed in the ECM

context, in which both series were treated as dependent variables in separate models with

the parallel set of regressors.7 The results are reported in Table 5.5. The estimates in

Model A and Model B demonstrate that neither the lagged Conservative support nor

ECM term has any statistically significant effect on prime ministerial support. The

residual of Model B is also unable to score significant results in Conservative support, in

which both PM support and ECM are statistically significant. Now, the verdict is clear:

support for the prime minister is exogenous to government popularity.

115

From the analyses above, we found that political intervention and economic

factors have different impacts on support for the prime minister and the governing party.

Although some issues and events have direct and strong impact on party (e.g., the

national elections), a myriad of others are attributed to the paramount leader of the

government. In other words, under certain circumstance, voters are more likely to hold

the prime minister than the party as a whole accountable for some policies, such as the

poll tax in 1990. Voters also interpret the economy in different terms. The prime

minister should be more responsible than the party for the past performance of the

national economy. When it comes to a choice of future government, the general public's

judgment will be based upon personal economic expectations.

Yet, does Mr. Major's government differ from his predecessor's in the way the

British public evaluates the economy and political support? Or does the relationship as

specified in the previous models remain stable and static over time? Applying the test

separately in the periods under the two governments does not render dissimilar results.8

However, testing the stability of the model over the long run calls for a more

sophisticated technique that allows for the estimation of the model coefficients that may

vary over time in a systematic manner. In the following section, we will explore the issue

employing a relatively new method that estimates time-varying "hyper-parameters" in the

political support model.

5.4 A Hyper-parameter Model on Prime Ministerial Approval

116

Above, we have built time series models based on the assumption of stable

structure, i.e., the model coefficients are primarily invariant over time. Such an

assumption is rather restricted, particularly in the face of an extended time period during

which structural breaks—such as a change of regime or a shock resulted from a political

event-are not aberrant occurrences. In fact, during the 1979-1996 Conservative period,

the British polity has experienced four general elections, two international wars, a change

of leadership and numerous political and economic crises. These interventions are all

potential sources of structural shifts that cause variations in the effect of the independent

variables. Most often, we tend to ignore the unobserved components that may potentially

drive the long-term movements of the series. In the following section, we relax the time-

invariant parameter assumption and analyze the party support series by taking into

account the long-term, stochastic trend components in a structural time series

specification.

Most of the time series models in political analysis take on a fixed coefficient-

adaptive model with the static form:

Yt = at + btXt + et (5.2)

where both a and P are assumed to be stable over time with a well-behaved error term e.

In the variable parameter adaptive model, we relax the restriction that the intercept

and coefficients are invariant over time. Interventions are also analyzed in various level

of effects. The model may be rewritten in a state-space form (SSF) as follows:

Yt =at + A,tWt + p , X t ++ et

117

in which

a t = at.j + et (a stochastic, time-varying trend)

X, = T A,,., + r|t (a stochastic, time-varying adaptive parameter)

Pt = T PM + c5t(" ") (5.3)

where X represents explanatory variables, W denotes interventions and e„ et, <5tand rit

are error terms independent of each other with zero means and constant variance. In this

context, the long-run trend component signifies the general direction in which the series

is moving. It may follow a random walk (a first order autoregressive process) or be at a

fixed level otherwise. Similarly, the time-varying parameters X and |3for the explanatory

variables and interventions represent the stochastic process with which the independent

variables exert various levels of effects over time. Below, we start with the analysis of

the former groups of variables-i.e., interventions-with particular emphasis on testing a

regime change at the time of the Thatcher-Major leadership transition, followed by the

study of time-varying parameters in the explanatory variables.

a. Analysis of Intervention

In earlier analyses, we identified major political events and interventions that

affect the popularity of the prime minister and the governing party. In the current

context, we further investigate these intervention variables or irregularities regarding

their impacts on structural changes in the level and trend of the party support series.9

From the previous analyses, we have learned that one of the most noticeable

118

interventions affecting Conservative support was the change of leadership in 1990, when

Mr. Major replaced Mrs. Thatcher as prime minister. The result was a spectacular boost

in prime ministerial approval (24 points) and a smaller, but hardly inconsequential, one

for the Tory party (7 points). Another visible move was, however, in another direction

two years later. The controversy over the European currency unification in September

1992 caused approval for the prime minister to nosedive to 23 percent, setting a free fall

record of 19 points, or 40 percent, in total support within one month (See Figure 5.1).

For the Conservatives, the plummet was of a smaller scale, but it continued to haunt the

party for the following months.

These pronounced shifts in the two series suggest the possibility that one or more

structural breaks have caused "regime changes" in British party support. The time series

analyses performed in the preceding section of this chapter only treated various events as

dummy variables, without taking into account the potential structural breaks.10 In fact,

modeling political shocks as transitory effects neglects potential structural impacts these

interventions could have on the party support series. To probe into this possibility, the

Kalman filter modeling technique is employed to reanalyze the currency crisis and the

Major succession variables in a structural, state-space form (SSF).11 Providing an

optimal updating scheme for analyzing unobserved components (i.e., long-run stochastic

trends) and time-varying parameters, the Kalman filter performs estimation of structural

time series models with a set of recursive equations (Cuthbertson, Hall and Taylor 1992).

Using this modeling technique, I reestimate ECM models in the state-space form, which

119

consists of unobserved components in the stochastic trend. The results of the structural

model are presented in Table 5.6.

The state-space models are parameterized as the preceding error correction models

except that only the personal economic expectation, which has been found to have a

direct and strong effect on governing party support, is employed for model parsimony. In

the first model, both the Major and currency crisis variables are included, but in different

treatments. Singled out from other explanatory variables, the two interventions are tested

in various forms. After controlling for the stochastic trend in the final state vector, the

Major variable is identified as an irregular or momentary pulse. All diagnostics,

including the normality and heteroscedasticity statistics, appear to be satisfactory,

reflecting a generally good fit of the model. Two R2s are reported, one of which is the

regular coefficient of determination while the other is Rd2 computed on the differences of

the variables.12 Both recommend the model.

Note that the final state vector, which is composed of the unobserved long-term

trend with stochastic level and slope, recognizes a general downward trend (-0.032, t=-

2.041) in the series across all three models. Intriguingly, when the two variables are

estimated separately, the unaccounted-for effects are reflected in the stochastic long-term

components.

The currency crisis variable is modeled as a structural break in the slope,13 which

takes the values 1,2,3... and so on after the occurrence of the intervention. Post-

intervention tests testify to the change of regime as precipitated by this crisis that broke

120

out in late 1992. The significant Chow test statistic (F63134=1.586, p=0.013) lends weight

to the suggestion of a regime change, which is further supported by the CUSUM plot

illustrated in Figure 5.3. The cumulative sum of residuals curve stretches well beyond

the 10 percent significant bound, testifying to the parameter instability in the model. The

staircase intervention registers statistical significance

(-0.029, t=-1.663), implying the intervention's steady, negative impact on the

Conservative party. The small size of the coefficient speaks of the net effect of the

influence of other variables. Nevertheless, the gross effect of the politically damaging

event, as estimated in another model, can be up to -0.473, or almost a half-point drop

every month.14

b. Analysis of Time-varying parameter

The Kalman filter models demonstrate that a structural break occurred not at the

change of leadership but during the current crisis under Major's government. As

evidenced in Table 5.6, the Major factor only posed as a pulse to the Conservative

popularity and the booster was carried away shortly by the far more devastating "Black

Wednesday" effect when Major decided to drop out of the European Monetary system on

September 16,1992. Substantively, none the less, it is not unreasonable to believe that

the Major effect is mediated via other variables that operate over time in a systematic

manner. This may, in turn, "masks" the impact of the leadership change and turns it into

a transitory shock in the time series model. Justification is in order. First, external

evidence derived from the previous chapters informs that the British electorate has

121

evolved from a class-oriented party loyalist to a "consumer" of government choice

relying on the readily available information to make individual voting decision.

Evaluations of the party leaders become the major source of information based upon

which the electoral judgment is made. Hence, there is compelling reason to infer that

any shock to the leadership variable will relay the effect to the party support. Second, the

importance of the leadership factor has been established in the ECM analysis, in which a

long-term equilibrium relationship between support for the prime minister and the

governing party is manifested. Yet, given the subtle relationship between the two series,

any effects that originate from the leadership variable (i.e. the Major effect in this case)

would very likely be absorbed in the "common trend" and would not be readily

distinguishable in regular time series models. Using Kalman filter, the following

analysis proceeds to identify the true impact of the new leadership on party support.

To test the hypothesis that the replacement of Major as prime minister exerts an

indirect impact on the government support, a time-varying parameter model of party

support is developed including the dependent variable and prime ministerial satisfaction

(PMSAT) in level form.15 The simple model containing only prime ministerial support

and currency crisis intervention manifests the time-varying feature of the PMSAT

parameter (see Table 5.7). The hyperparameter of PMSAT is statistically significant,

suggesting the stochastic nature of prime ministerial support on party support. To further

investigate the time-varying parameter of PMSAT, I reanalyze the party support model

using recursive least squares.16 The recursive least squares (RLS) approach tests the

122

stability of the parameters in a model by estimating the same model repeatedly, using

ever larger subsets of the sample data, in a bid to detect structural change. Four party

support models employing the variants of economic evaluation variables are tested and

the recursive coefficients for PMSAT are plotted in Figure 5.3a to Figure5.3d. The

original PMSAT series is superimposed on the graphs to illustrate the leadership effect.

All four of the graphs demonstrate an almost identical pattern: the

hyperparameters, which represent the time-varying effect of prime minister approval on

party support, come to an abrupt descent right at the point Major took over the office of

premiership. This is strongly suggestive of a structural break in the leadership effect on

government support. The most intriguing feature of the illustrations, however, is the

crossing point of the two series, where lies the crux of the delicate relationship between

Major's popularity and his impact on party support. Despite the very height of his scores

at inauguration, Major's influence was in effect falling way behind that of his

predecessor! As shown in Figure 5.3a, the PMSAT coefficient, which represents the

prime minister's influence on party support, stands at approximately 0.5 before the Major

government. The new Downing 10 occupant, however, registers a significantly smaller

score at 0.37, suggesting that the prime minister, as the new party leader, has lost at least

20 percent of leverage as compared to Mrs. Thatcher's influence on party support!

The erosion of prime ministerial effect reaches a lower level shortly after another

smaller break occurred around late 1992. This testifies to the previous finding that the

Black Wednesday posed as another structural break in level for Major's government.

123

However, the two structural shifts differ in the way they characterize the new regimes.

For the currency crisis, the downward shift in both level and slope (note the continuous

decline after 1992) depicts a general depreciation of public confidence in the governing

party's capability to manage a crisis. Even though the erosion of public support for the

party was not as pronounced as Mr. Major's free fall in popularity, the impact was

continual, leading the party to slide to its historic low of 15 points in January, 1995.

Comparatively, the leadership change effect is not as easily identifiable. Masked by the

highly visible surge in prime ministerial popularity, the effect is not discernible without

employing special modeling techniques. Using the Kalman filter and recursive least

square analysis helps us to unveil the hidden fact that Mr. Major's rise to leadership, in

fact, precipitated a downfall in the impact of prime ministerial approval on party support!

5.5 Conclusion: The Presidentialization Thesis in Perspective

In studying recent general elections, Mughan observes that the prime minister has

become increasingly "presidentialized" on the British political stage, and general

elections now feature highly "leader-centered" campaigns (1993). Portraits, personality

profiles and television commercials characterize modern elections in Great Britain, with

an overwhelming emphasis on the party leader. As a consequence, as the author

contends, the growing salience and electoral impact of party leaders has become

inevitable. The author, with his colleagues, however, stops short of providing more

substantive and empirical evidence to justify the assertive conclusion in his "leader-

124

matters-more" thesis.

In this chapter, we have analyzed British party support models with special

emphasis on leadership effects. We have identified the trend that drives the movement of

the government's popularity, embedded in which-as shown in the cointegration analysis-

-is a long-term equilibrium relationship between public satisfaction with the prime

ministers and the Conservative party. Generally downward moving, the two series both

experience structural shifts in 1990 and 1992, respectively. More intriguingly, we

discover that, behind the scene of Mr. Major's all time popularity records—the highest

score, the greatest climb and the steepest fall—was a decline of the impact of prime

ministerial approval on party support.

Overall, our findings sparsely lend any credence to the "presidentialization"

thesis. Instead of finding greater salience of party leaders, our analyses document the

contrary, that is, a trend of downward movement, in both level of popular support and

electoral influence, over the long haul. Revisiting the presidentialism thesis, however,

places new emphasis on the role the prime minister plays in the contemporary British

political arena.

As compared to his or her U.S. counterpart, the British government leader has

become no less "presidential" in the media, particularly during election campaigns. The

British prime minister, however, is in a different political environment. Faced with a

parliamentary system, the British leader finds himself or herself embedded in a different

accountability framework in which the people elect the party and the party elects its

125

leader. In the past, the prime minister could rely on the governing party as a "safety

zone" for neutralization of any political censure. The contemporary British polity barely

inherits such a "firewall" system. As the forgoing study demonstrates, the public holds

the prime minister accountable for government policies in a different manner than it does

the governing party. In the poll tax incident, for instance, the British voters dealt a major

blow to Mrs. Thatcher, holding her responsible for the widely detested policy. The

negative impact of this incident, however, lands on the Conservative party to a much

smaller extent. In the 1990s, the British party leader finds no safeguard in the party in the

way he or she did in the past. In this sense, the prime minister becomes more

"presidentialized" where the elected leader is directly under public scrutiny by the general

electorate.

On the other hand, our analysis testifies to a new asymmetry of the reward-

punishment model in the British context. Given the valence issue of the economy, the

electorate punishes or rewards in a different manner. For the prime minister, Key's

model holds for the leader as every record in the past, pocketbook and national alike,

counts (Key 1966). For the party, the public, however, looks forward in perspective

when a choice of future government is made. The widely-studied reward-punishment

asymmetry still operates, but in dissimilar configuration: punish the prime minister for

the dark past, but reward the party for the bright future (see Sanders 1993, Bloom and

Price 1975).

In sum, the chapter addresses the theory of presidentialization of the British prime

126

minister by analyzing the party support and leadership effects in a new methodological

framework. Besides finding evidence for the thesis, the study throws new insights into

the role of party leader in the British parliamentary system. The new trend of leadership

effect, along with the movement in partisan support, plays a major role in the

evolutionary process of the electorate in the post-dealignment Conservative era.

127

Endnotes

1. Data are drawn from the monthly Gallup series. The party support questions are:

a. "If there were a General Election tomorrow, which party would you support?" or b. "Which party would you be most inclined to vote for?" If the respondent answers "I

don't know" in a., the party support for the governing Conservative party will be calculated as the sum of the percentages of respondents selecting Conservative in a. or b..

The economic evaluations questions are:

a. Personal prospections: "How do you think the financial situation of your household will change over the next 12 months?"

b. Personal retrospections: "How does the financial situation of your household now compare with what it was 12 months ago?"

c. National expectations: "How do you think the general economic situation in this country will develop over the next 12 months?"

d. National retrospection: "How do you think the general economic situation in this country has changed over the last 12 months?"

Respondents were given a five-category scale ranging from "get a lot better" at one end to "get a lot worse" at the other. The neutral point is "stay the same." The four variables are constructed by subtracting the percentage of negative responses from that of the positive ones.

The prime ministerial approval variable is derived from the percentage of responses to the question: "Are you satisfied or dissatisfied with as prime minister?"

2. The variable is derived from the residual of the simple OLS model in which Tory support is regressed on prime ministerial approval.

3. An alternative setup for the exogeneity test is to switch prime ministerial approval (PMSAT) with Conservative party support (CONS) like:

A PMSATt= p0+ , p t ACOHtPPCM +p pCQNEV+p EyENTfP FALK + P6MAJORt+p7JOHNt+pgEL83t+ p9EL87t+ p10EL92t+et

Note that the ECM variable is the residual from the model that regresses PMSAT on CONS.

4. The rate of re-equilibration is:

128

Z*(l-ECM)n

where Z is the change in percentage points, ECM is the error correction mechanism and n the number of elapsed months.

5. The PE model only marginally encompasses the NR one, and tests with the other two produce inconclusive results.

6. Because the economic evaluation variables are stationary and unit root tests suggest no common trend exists between them and the dependent variable (Table 5.1), a regular OLS model is applied to the variables in first differences.

7. Informed by previous analysis, I only employ personal expectation as the economic evaluation variable, which suffices for the purpose of the exogeneity test.

8. Both the model estimates and encompassing test reveal very similar, but not identical, results.

9. In structural time series modeling, intervention effects can be analyzed in four cases taking into account the stochastic trend and seasonal components. While it can be treated as transitory effects (pulse) as in classical time series analysis, it can be a structural break or shift (step) in the level of the series or a staircase intervention or change in the slope of the trend. Interventions can also be modeled as seasonal change as most usually found in the financial series (Harvey 1989:397-98; Koopman et al. 1995). In the political context, we are mostly concerned with the structural break that an intervention can cause in a change of regime or a partisan dealignment.

10. More in-depth analysis of interventions can be conducted using the Box-Jenkins (ARIMA) approach, which adopts a different methodology than the structural time series analysis employed in this study.

11. Three statistical software packages-namely, STAMP 5.0, Reg-x 386 and Microfit 4.0--are employed to implement the Kalman filter models.

12. The Rd2 is the comparison of the Prediction error variance (PEV) with the variance of first differences, Ayt=yt - yul. It should be noted that Rd2 is not bound between 0 and 1. It can be negative, which indicates bad fit.

13. See note 9.

14. The structural break model, which focuses on the level for both the dependent variable and the currency crisis intervention, gauges the gross, permanent effect of the thus far most damaging event for the prime minister and the governing party alike.

129

15.The time-varying parameter model uses levels for analysis, as opposed to the first differenced form in the ECM context. The model can be written as follows:

Y|=XP, + E,

Pi=Pt-i+l1, (5.4)

where Y is party support, X is the data matrix consisting of prime ministerial support and interventions, P is the time-varying parameter and e and r\ are well-behaved error terms (Kalman 1960; see also Cuthbertson, Hall and Taylor 1992 and Price and Sanders 1993).

16. The recursive least square (RLS) analysis is implemented using Microfit 4.0. The RLS approach requires a special "trick" in estimating a model with dummy variables. Because the program is unable to start estimation if a series starts with a long range of zeros (which is quite common among dummy variables), the modeler is required to implant a small value (e.g., 0.01) at the start of the series to "fake" a "non-zero." The tiny number does not affect the estimation results.

CHAPTER VI

CONCLUSION

"If either party fails to recognize that the electorate has changed, then it will doom itself to a long period on the opposition benches in Parliament"

(Bo Sarlvik and Ivor Crewe, Decade of Dealignment, 1983: 338)

Thirteen years ago when Bo Sarlvik and Ivor Crewe wrote the realistic remark on

partisan decline of the British electorate, they set out to predict the fate of political parties

in the wake of partisanship decline in the 1970s. Students of British politics ought not be

surprised by the authors' prophetic wisdom in depicting the British polity. Indeed,

Sarlvik and Crewe's observation still has currency, even after thirteen years. The

electorate has changed, and it is still changing.

As the title speaks for itself, this dissertation resembles Sarlvik and Crewe's

exercise on analyzing political change in Great Britain after more than a decade. It

surveys the British polity since the beginning of Thatcher's first term in 1979 and

examines the electorate in three primary facets: partisanship, voting behavior, and party

support between elections. The question Sarlvik and Crewe studied in 1983 has not lost

its relevance in 1996: "Are we witnessing the erosion of the foundation of the

Conservative and Labour two-party system which has so decisively put its mark on most

of the post-war era? " (Sarlvik and Crewe 1983:441). The rise of the Alliance in 1981

130

131

seemingly validated the authors' concern, although its demise showed it did not have any

major enduring impact on the political system. Nonetheless, changes in the electorate did

have such effects.

As Chapter II has demonstrated, the erosion of party loyalty among voters in the

1970s weakened the foundation of the British party system that was built on the country's

long-existing class structure. The gradual dissolution of this class structure, resulting

from the long-term evolution of British society, redefines socioeconomic identities for the

traditional working and middle classes. Compared to the 1970s, we found more subtlety

in the trend of partisanship in subsequent decades. Though attenuated, strength of

partisanship has stabilized since the mid-1980s. Also, partisan dealignment is different in

degree among various groups and partisans. While decline of party allegiance has

affected both Conservatives and Labour, its largest effect has been to undermine the

electoral base of the latter party. The shrinkage of Labour's proportion of stable

identifiers is at the root of the party's electoral volatility. As a consequence, the

continuing dealignment has fostered a less partisan electorate and left elections more

susceptible to the inherently volatile short-term forces.

The general downward shift of party loyalties has reduced the long-term social

class factor to a secondary role in determining electoral outcome. As shown in Chapter

III, class-related variables like home-ownership, occupation and union membership all

have become less important vis-a-vis the short-term judgmental variables. Does this

indicate a movement toward a politically "classless" society? The answer is both no and

132

yes. The class influence has by no means disappeared, but it operates in a society where

class-interest, for most British voters, has waned considerably in both literal and electoral

terms. In this sense, Sarlvik and Crewe anticipated the redefinition of the principal

determinants of voting in the 1990s: immediate economic interest, instead of traditional

class membership, has increased in importance. As their analysis foreshadowed, the

British electorate becomes highly vulnerable to the volatility of short-term forces. A

telling fact in Chapter Ill's findings portrays a "similar but somewhat different" pattern.

In the party identification model for the 1980s, short-term factors are found to be more

important concerns for party identifiers. Intriguingly, re-paraphrasing Pulzer's widely-

cited remarks, issues positions and judgments of party leaders usurp the place of social

class to become the "basis" of British politics, displacing the latter variable to the rank of

"embellishment and detail"!1

British voting behavior has become more of a practice of rational, individualistic

decision-making than a habitual exercise of pledging allegiance to class and party.

Family heritage, peer pressure and environmental influences have all become subordinate

to individual judgments based on available information. Among all of the factors studied

in this work, I identify no single variable as influential as the class variable was claimed

to be in studies such as Pulzer's. At the least, nothing can be found as much like a

"basis" as the previously well-acclaimed long-term force of social class. Prudent

modelers of contemporary electoral politics ought to be skeptical regarding claims for the

existence of such a ubiquitous factor.

133

The investigation in Chapter III confirms the impact of party identification on

electoral choice. We also showed that party attachments are subject to other short-term

evaluative variables, the latter being functions of various objective measures such as

inflation, unemployment and political events. Chapters IV and V presented a host of

similar models interwoven by "inter-dependent" variables. Party support models, for

instance, are among the typical cases.

Asymmetry is another feature that characterizes partisanship and voting models.

The study in Chapter III, for instance, has shown that the social class factor wields power

on Labour while it exerts barely any influence on Conservative and Liberal/Alliance

supporters. Although council house tenancy or manual labor jobs can be defining

features of Labour identifiers, home-ownership or white-collar positions have no

statistically significant relationship with a Conservative or Liberal vote. These results

demonstrate that, at a minimum, more than half of the British voters have departed from

the traditional model of social-class voters who form their political choice on the basis of

class identity. Being a blue-collar worker no longer precludes a vote for the Tories;

likewise, one can be a Labour voter and yet still be his or her own landlord.

Personal attributes have yielded to more sophisticated causes. Individual

evaluations of issues, economy and party leadership have become the principal forces

driving electoral choice. Chapter IV's finding that British voters can well position

themselves on various issues when casting their vote presents a major anomaly to the

directional model offered by Rabinowitz and his colleagues. As the authors argue, voters

134

in the United States and other European countries cast their votes by identifying their

issue directions, or the pros and cons, as compared with the candidates. They are unable

to give exactly their own positions and those of the candidates (Macdonald et al. 1991,

Rabinowitz and Macdonald 1989). In contrast, as this chapter unveils, the British voters

made their electoral choices in a more sophisticated manner. Not only do they have

knowledge of more precise issue positions as compared with those of the candidates, but

more importantly, the voters effectively utilize this information to guide their electoral

choice.

As the issue evaluations variable becomes crucial to the electoral decision, the

findings on the two rival issue-voting models are revealing. Traditional voting models

are at stake. The issue priority model, which suggests most working class voters have a

high level of concern on unemployment as much as the middle class do on prices, has

become less appealing. As the aggravated unemployment and high-standing inflation

rates prevalent in the 1980s become history, British voters, Conservative or Labour alike,

are more conscious with the other dimensions of the valence issue of economy than

merely objective macroeconomic indices. These dimensions include individual

subjective evaluations of the national and personal economy as perceived retrospectively

and prospectively. Likewise, the reward-punishment model faces difficulties in

explaining the contemporary elections. From the analyzes presented in Chapter V, one

finds that voters will not simply punish the governing party for bad economy in

retrospect. The 1992 election presents a case for the new economic voting model: instead

135

of punishing the Tories, the forward-looking electorate chose to retain the governing

party for another term because most voters believed that the economic outlook would be

gloomy under Labour rule!

The finding in Chapter V that a voter's prospective evaluations of personal

financial situations is a major determinant of party support accords well with the new

conventional wisdom as articulated by Sanders and his colleagues (Sanders et al. 1987,

Sanders 1992). However, it does not diminish the significance of other dimensions of the

subjective economy. Retrospective and prospective evaluations of the national economy

also matter, particularly the former. Empirical evidence established in Chapter V

illustrates that the impact of economic evaluations on party support are not unmediated.

The judgment is mediated through evaluations of the prime minister before affecting the

support for party. The conclusion is similarly asymmetric, but distinct: dismal economic

record, punish the prime minister; rosy future, reward the party.

The examination of leadership effects in Chapter V provides new insights to the

contemporary party support models: while Mughan's presidentialization thesis is

affirmed in the Thatcher era, John Major is a different story. Recognizing the growing

influence of leaders in contemporary electoral politics, we discover that this time-bound

effect varies from one prime minister to another.

In sum, if the 1970s has been designated the decade of dealignment and the 1980s

the decade of conviction, the evidence presented in this dissertation argues that the 1990s

should clearly be called the decade of choice. Individualized evaluations prevail and

136

become the rudiments of voting decisions. When it comes to party support,

contemporary British voters have obviously rejected the famous Thatcherist slogan that

pronounces TINA (There Is No Alternative). Alternatives abound in elections of the

1990s. This is what democracy stands for, and how it evolves.

Endnotes

1. Pulzer wrote (1967): "Class is the basis of British politics; all else are embellishment and detail."

137

Appendix

Questions from the 1979,1983,1987 and 1992 British General Election Study survey and

the 1983-87 and 1987-92 panel study*

1979

1. Party Identification

"Generally speaking do you think of yourself as Conservative, Labour, Liberal, Social Democrat, or what?" (48, M129156)

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others? If yes, which Party?" (49, Ml 30157)

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (50,

M131158)

3. Age

"Would you say in what year were you born?" (105, M191279)

4. Education

"Education/Qualification?" (91, M000242-M000252)

5. Income

"What is the total income of your household?" (109, M212284) 1 Less than £3,999 2 £4,000-5,999 3 £6,000-7,999 4 £8,000-9,999

Question number and variable number are included respectively in parentheses following questions.

138

5 £10,000-11,999 6 £12,000-14,999 7 £15,000-17,999 8 £18,000-19,999 9 £20,000-22,999 10 £23,000-25,999 11 £26,000-28,999 12 £29,000-31,999 13 £32,000-34,999 14 £35,000-37,999 15 £38,000 or more 97 Refused 98 Don't know 99 Not answered

6. Union membership

"Are you a member of a trade union?" (61, M245190)

7. Home ownership

"Would you tell me whether your home is owned or rented?" (106A, M187A280)

8. Vote

"...which party did you vote for in the General Election?" (41, Ml 15147)

1983

1. Party Identification

"Generally speaking, do you think of the Liberal Party as being closer to the Conservative Party or closer to the Labour Party ?" (ql3a)

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others ? IF YES , Which Party ? [ IF " ALLIANCE " PROBE : Liberal or Social Democrat ? ]?" (ql3b)

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (ql3c)

3. Age

139

"What was your age last birthday?" (q56)

4. Education

"Where did you go to school, that is, the last school you attended, was i t . . . (READ OUT) in the U . K. ?" (q59b)

5. Union membership

"Are you now, or have you ever been, a member of a trade union ?" (q52c )

6. Home ownership

"Do you - your household - own or rent this (house / flat / accommodation) ? [ OWNS or CO - OWNS : include mortgages as ownership ] [ if rented : From whom ? ]? (q65b)

7. Income

"And which of the letters on this card represents the total income of your household from all sources, before tax?" (ql09a)

8. Vote

"Which Party did you vote for in the general election? (q9a)

9. Leader evaluations

"Would you say Mrs. Thatcher/Mr. Foot/Dr. Owen/Mr. Steel is good or bad at getting things done?" (V20A1-V20A4)

1 Good at getting things done 2 Bad at getting things done 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as extreme, or moderate?" (V20B1-V20B4)

1 extreme 2 moderate 3 Neither or both

140

8 Don't know 9 Not answered

"Would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel looks after one class, or, looks after all classes?" (V20C1-V20C4)

1 looks after one class 2 looks after all classes 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as capable of being a strong leader, or, not capable of being a strong leader?" (V20D1-V20D4)

1 capable of being a strong leader 2 not capable of being a strong leader 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as caring, or, uncaring?" (V20E1-V20E4)

1 caring 2 uncaring 3 Neither or both 8 Don't know 9 Not answered

"Would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel is likely to unite/divide nation?" (V20F1-V20F4)

1 unite nation 2 divide nation 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as likeable as person?" (V20G1-V20G4)

1 likeable as person

141

2 not likeable as person 3 Neither or both 8 Don't know 9 Not answered

10. Issue Evaluations

Unemployment and Inflation

Some people feel that getting people back to work should be the government's top priority. These people would put themselves in Box A. (POINT) Other people feel that keeping prices down should be the government's top priority. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (28a-28d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about unemployment and inflation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Taxation and Government Services

Some people feel that government should put up taxes a lot and spend much more on health and social services. These people would put themselves in Box A. (POINT) Other people feel that government should cut taxes a lot and spend much less on health and social services. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (29a-29d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about taxes and government spending.

142

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Nationalisation

Some people feel that government should nationalise many more private companies. These people would put themselves in Box A. (POINT) Other people feel that government should sell off many more nationalised industries. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (34a-34d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about nationalisation and privatisation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Redistribution

Some people feel that government should make much greater efforts to make people's incomes more equal. These people would put themselves in Box A. (POINT) Other people feel that government should be much less concerned about how equal people's incomes are. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT

143

RIGHT K-F) (35a-35d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

European Community

Some people feel that Britain should do all it can to unite fullv with the European Community. These people would put themselves in Box A. (POINT) Other people feel that Britain should do all it can to protect its independence from the European Community. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A3 9a-A3 9d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about the European Community.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

144

T,aw and order

Some people feel that protecting civil rights is more important than cutting crime. These people would put themselves in Box A. (POINT) Other people feel that putting crime is more important than protecting civil rights. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (39a-39d)

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Welfare

Some people feel that the poor in Britain are entitled to more help from government. These people would put themselves in Box A. (POINT) Other people feel that the poor in Rritain should get less help from government and do more to help themselves. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (35a-35d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of

145

the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

1987

1. Party Identification

"Generally speaking, do you think of yourself as Conservative Labour, Liberal, Social Democrat, (Nationalist / Plaid Cymru), or what ? [ IF " ALLIANCE " PROBE : Liberal or Social Democrat ?" (ql2a)

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others? If yes, which Party?" (ql2b)

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (ql2c)

3. Age

"What was your age last birthday?" (v58c)

4. Union membership

"Are you now a member of a trade union or staff association?" (49a)

"Have you ever been a member of a trade union or staff association?" (49b)

5. Home ownership

"Do you - your household - own or rent this (house/flat/accommodation)? (60ab)

6. Income "And which of the letters on this card represents the total income of your household from all sources, before tax?" (64)

7. Vote

"Which Party did you vote for in the general election? (8a)

146

8. Leader evaluations

"Would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel is good or bad at getting things done?" (V20A1-V20A4)

1 Good at getting things done 2 Bad at getting things done 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as extreme, or moderate?" (V20B1-V20B4)

1 extreme 2 moderate 3 Neither or both 8 Don't know 9 Not answered

"Would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel looks after one class, or, looks after all classes?" (V20C1-V20C4)

1 looks after one class 2 looks after all classes 3 Neither or both 8 Don't know 9 Not answered "Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as capable of being a strong leader, or, not capable of being a strong leader?" (V20D1-V20D4)

1 capable of being a strong leader 2 not capable of being a strong leader 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as caring, or, uncaring?" (V20E1-V20E4)

1 caring 2 uncaring 3 Neither or both

147

8 Don't know 9 Not answered

"Would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel is likely to unite/divide nation?" (V20F1-V20F4)

1 unite nation 2 divide nation 3 Neither or both 8 Don't know 9 Not answered

"Would you describe Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel as likeable as person?" (V20G1-V20G4)

1 likeable as person 2 not likeable as person 3 Neither or both 8 Don't know 9 Not answered

9. Issue Evaluations

Unemployment and Inflation

Some people feel that getting people back to work should be the government's top priority. These people would put themselves in Box A. (POINT) Other people feel that keeping prices down should be the government's top priority. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (28a-28d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about unemployment and inflation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

148

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Taxation and Government Services

Some people feel that government should put up taxes a lot and spend much more on health and social services. These people would put themselves in Box A. (POINT) Other people feel that government should cut taxes a lot and spend much less on health and social services. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (29a-29d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about taxes and government spending.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Nationalisation

Some people feel that government should nationalise many more private companies. These people would put themselves in Box A. (POINT) Other people feel that government should sell off many more nationalised industries. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (34a-34d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about nationalisation and privatisation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you

149

think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think conies closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Redistribution

Some people feel that government should make muct greater efforts to make people's incomes more equal. These people would put themselves in Box A. (POINT) Other people feel that government should be much less concerned about how equal people's incomes are. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (35a-35d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

Law and order

Some people feel that protecting civil rights is more important than cutting crime. These people would put themselves in Box A. (POINT) Other people feel that cutting crimp i« more important than protecting civil rightsThese people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (39a-39d)

150

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal/Alliance) And now please tick whichever box you think comes closest to the views of the SDP/Liberal/Alliance?

1992

1. Party Identification

"Generally speaking, do you think of yourself as Conservative, Labour, Liberal Democrat, or what?" (A6a, B13a)

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others? If yes, which Party?" (A6b, B13b)

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (A6c, B13c)

3. Vote

Which party did you vote for in the general election? (9a)

4. Age

"What was your age last birthday?" (915c)

5. Home ownership

151

"Does your household own or rent this accommodation?" (919b)

6. Occupational Class

"Social Class" (905m)

-2 Skp,nvr paid job had -1 Skpd,never had j ob 1 I professional 2 II managerial and technical 3 IIIN skilled non-manual 4 IIIM skilled manual 5 IV partly skilled 6 V unskilled 7 armed forces 8 inadequate description

7. Income

"And which of the letters on this card represents the total income of your household from all sources, before tax?" (921)

1 Less than 3,999 2 4,000- 5,999 3 6,000- 7,999 4 8,000- 9,999 5 10,000- 11,999 6 12,000- 14,999 7 15,000- 17,999 8 18,000- 19,999 9 20,000- 22,999 10 23,000- 25,999 11 26,000- 28,999 12 29,000- 31,999 13 32,000- 34,999 14 35,000- 37,999 15 38,000 or more 96 ? 97 Refused 98 Don't know 99 Not answered

152

8. Education

"Have you passed any of the examinations on this card?" (912a)

"And have you passed any of the exams or got any of the qualifications on this card?" (913a)

9. Union membership

"Are you now a member of a trade union or staff association?" (903c)

10. Leader evaluations

"Would you describe John Major/Neil Kinnock/Paddy Ashdown as extreme, or moderate?" (22a-c)

1 extreme 2 moderate 3 Neither or both 8 Don't know

"Would you say John Major/Neil Kinnock/Paddy Ashdown looks after one class, or, looks after all classes?" (23a-c)

1 looks after one class 2 looks after all classes 3 Neither or both 8 Don't know

"Would you describe John Major/Neil Kinnock/Paddy Ashdown as capable of being a strong leader, or, not capable of being a strong leader?" (24a-c)

1 capable of being a strong leader 2 not capable of being a strong leader 3 Neither or both 8 Don't know

"And on the whole would you describe John Major/Neil Kinnock/Paddy Ashdown as caring, or, uncaring?" (25a-c)

1 caring 2 uncaring 3 Neither or both

153

8 Don't know

11. Issue evaluations

Unemployment and Inflation

Some people feel that getting people back to work should be the government's top priority. These people would put themselves in Box A. (POINT) Other people feel that keeping prices down should be the government's top priority. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A35a-A35d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about unemployment and inflation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Taxation and Government Services

Some people feel that government should put up taxes a lot and spend much more on health and social services. These people would put themselves in Box A. (POINT) Other people feel that government should cut taxes a lot and spend much less on health and social services. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A3 6a-A3 6d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about taxes and government spending.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you

154

think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think conies closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Nationalisation and Privatisation

Some people feel that government should nationalise many more private companies. These people would put themselves in Box A. (POINT) Other people feel that government should sell off many more nationalised industries. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between. along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A37a-A37d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about nationalisation and privatisation.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Redistribution Some people feel that government should make muct greater efforts to make people's incomes more equal. These people would put themselves in Box A. (POINT) Other people feel that government should be much less concerned about how equal people's incomes are. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A3 8a-A3 8d)

(Respondent's own view)

155

In the first row of boxes, please tick whichever box comes closest to your own views about redistributing income.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

European Community Some people feel that Britain should do all it can to unite fully with the European Community. These people would put themselves in Box A. (POINT) Other people feel that Britain should do all it can to protect its independence from the European Community. These people would put themselves in Box K. (POINT) And other people have views somewhere in-between, along here (POINT LEFT A-F) or along here (POINT RIGHT K-F) (A39a-A39d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views about the European Community.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

1983-87 panel study

1. Party Identification

156

"Generally speaking do you think of yourself as Conservative, Labour, Liberal, Social Democrat, or what?" (83: 7a; 87: 10a))

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others? If yes, which Party?" (83 : 7b; 87: 10b))

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (83: 7c; 87:

10c)

3. Age

"What was your age last birthday?" (83: 33b; 87: 49b)

4. Home ownership "Do you - your household - own or rent this (house/flat/accommodation)?" (83: 35b; 87: 46b)

5. Union membership

"Are you now, or have you ever been, a member of a trade union?" (83: 32c; 87: 37c)

"Which party have you decided to vote for?" (83: 3a; 87: 31a)

7. Leader evaluations

"Which of the qualities on this card would you say Mrs. Thatcher/Mr. Kinnock/Dr. Owen/Mr. Steel has?" (83:13a-13d; 87: 18a-18d)

Caring Determined Shrewd Likeable as a person Tough Listens to reason Decisive Sticks to principles

NONE OF THESE Don't know

157

8. Issue evaluations

Defence

Please look at this card. People who are convinced that we should get rid of all nuclear weapons in Britain without delay will put a tick in the last box on the left (Letter P) -POINT TO BOX P AND STATEMENT - while those who are convinced that we should increase nuclear weapons in Britain without delay will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here -POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY (83: 28a-28d; 87: 33a-33d)

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal Alliance) And now please tick whichever box you think comes closest to the views of SDP/Liberal Alliance?

Unemployment and Inflation Scale

Please look at this card. People who are convinced that we should getting people back to work should be government's top priority will put a tick in the last box on the left (Letter P) - POINT TO BOX P AND STATEMENT - while those who are convinced that we should keeping prices down should be the government's top priority will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY (83: 29a-29d; 87: 34a-34d)

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views.

158

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think conies closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal Alliance) And now please tick whichever box you think comes closest to the views of SDP/Liberal Alliance?

Taxation and government services scale

Please look at this card. People who are convinced that we should put up taxes a lot and spend much more on health and social services will put a tick in the last box on the left (Letter P) - POINT TO BOX P AND STATEMENT - while those who are convinced that we should cut taxes a lot and spend much less on health and social services will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE -or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY (83: 30a-30d; 87: 35a-35d)

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal Alliance) And now please tick whichever box you think comes closest to the views of SDP/Liberal Alliance?

Nationalisation/privatisation scale

Please look at this card. People who are convinced that we should nationalise, many more private companies will put a tick in the last box on the left (Letter P) - POINT TO BOX P

159

AND STATEMENT - while those who are convinced that we should sell of many more nationalised industries will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY (83: 31a-31d; 87: 36a-36d)

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative)

First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(SDP/Liberal Alliance) And now please tick whichever box you think comes closest to the views of SDP/Liberal Alliance?

1987-92 panel

1. Party Identification

"Generally speaking do you think of yourself as Conservative, Labour, Liberal, Social Democrat, or what?" (87-92: vl2a)

(If none/don't know) "Do you generally think of yourself as a little closer to one of the Parties than the others? If yes, which Party?" (87-92: vl2b)

2. Strength of Party Identification

"Would you call yourself very strong , fairly strong or not very strong?" (87-92:

vl2c)

3. Age

"What was your age last birthday?" (87: v58c; 92: y322b)

4. Home ownership

160

"Do you - your household - own or rent this (house/flat/accommodation)?"

5. Union membership

"Are you now, or have you ever been, a member of a trade union?"

"Which party have you decided to vote for?"

6. Leader evaluations

"Which of the qualities on this card would you say Mrs. Thatcher/Mr. Kinnock/Mr. Ashdown?"

Caring Determined Shrewd Likeable as a person Tough Listens to reason Decisive Sticks to principles

NONE OF THESE Don't know

7. Issue evaluations

Defence Scale

Please look at this card. People who are convinced that we should get rid of all nuclear weapons in Britain without delay will put a tick in the last box on the left (Letter P) -POINT TO BOX P AND STATEMENT - while those who are convinced that we should increase nuclear weapons in Britain without delay will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here -POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY.

(Respondent's own view) In the first row of boxes, please tick whichever box comes closest to your own views.

161

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party?

(Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Unemployment and Inflation Scale

Please look at this card. People who are convinced that we should getting people back to work should be government's top priority will put a tick in the last box on the left (Letter P) - POINT TO BOX P AND STATEMENT - while those who are convinced that we should keeping prices down should be the government's top priority will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY.

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Taxation and government services scale

Please look at this card. People who are convinced that we should put up taxes a lot and spend much more on health and social services will put a tick in the last box on the left

162

(Letter P) - POINT TO BOX P AND STATEMENT - while those who are convinced that we should cut taxes a lot and spend much less on health and social services will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE -or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY.

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

Nationalisation/privatisation scale

Please look at this card. People who are convinced that we should nationalise many more private companies will put a tick in the last box on the left (Letter P) - POINT TO BOX P AND STATEMENT - while those who are convinced that we should sell of many more nationalised industries will put a tick in the last box on the right (Letter J) - POINT TO BOX J AND STATEMENT. So, as you can see, people who hold views that come somewhere between those two positions will tick a box somewhere along here - POINT TO BOXES FROM LEFT TO CENTRE - or somewhere along here - POINT TO BOXES FROM RIGHT TO CENTRE. RE-EXPLAIN IF NECESSARY.

(Respondent's own view)

In the first row of boxes, please tick whichever box comes closest to your own views.

(Conservative) First the Conservative Party. In the next row of boxes, please tick whichever box you think comes closest to the views of the Conservative Party? (Labour) Now in the next row please tick whichever box you think comes closest to the views of

163

the Labour Party?

(Liberal Democrats) And now please tick whichever box you think comes closest to the views of the Liberal Democrats?

4J Cfl

M U

5

I 53

164

CI en Ol H

in

as H

%

d •H flf 4J -rl U « •P <0 a>

o

d •rl

4J 3 0 d (N rH VO 00 H CM in 00 ch o a\ O H ro M h) • • • • • • • • • • • • • •

ic5 00 ch vo VO CO r- OS (7> in o o ro vo H 2 00 00 a\ OS as 00 00 00 r> 00 r- r-

0> CA a & •H W in CM VO as VO 00 OS 00 CM CM ro H vo in .M W • • • • • • • • • • . . . . 0 EH (N H o o o o o H ro ro ro ro CM ro > O

»d

a 1 fH H H CM in ro vo in o in ro (0 • • • • • • • • • • . . . .

I o O O o o o o H CM ro CM rH H CM >1 4J M 10 1 r"' rH in r- as CM in in ro ro 00 VO 00 a« • » • • • • • • • • • • . •

1 Cft as CM CM in H 00 r- 0\ 00 ro in CM N H H H H CM CM H A

CQ a> PQ| 1 ro H 00 00 H OS o rH CM O vo 00 M • • • • • • • . » • . . . (0 iHl 1 00 VO 00 i LD ro rt* ro o\ I"** o £ vp ro ro ro CM ro ro CO

0) •P A §1 1 " in o l> T# CT% 00 00 a\ ro CTi

g a 1 ro 00 <T\ as ro H vo r* in ro CM CM H

ro ro ro H •

CM

rH pd LD o rH m OS VO o G\ ro CM in in in in VO VO R- r- r* 00 00 cr*

<0 W OS as as as a\ as OS a\ o\ Ch OS cr» 0% as H >* rH rH tH TH rH TH iH rH rH rH rH H TH TH

<N as as H

»M O

G o •H •P O <D rH pa

fd u <U a aj o

CO -H 4J •H

PQ

a) E-t

CM o> o\ H

43 Di (d G (d >

CQ •H

<D Q

T3 a td

t j •H > cd Q

CQ

m a a> -H rH 4J 4J *H 3 (d m s

CQ d) a

3 o w

1 6 5

CI os i

i n

as

a -H 10 •P -H M «

d •rl

d o •rl •U 10 0

-H IM -H AJ a 0)

•d H

it S Pi

0 £

4J 01 a <d u -P CO

CN

CI ©

rH

• s H

9 H

§

H En H

£ W Q

>H

I

w o

B

§ W

e co

\D os H

1 o Eh

>h P* P

P*

1 Eh CO

H rtj EH Cn CO

O £ § Pi O pa P< > EH

CO

<T> CN rl ro h

^ in m

u> vo vo

H ^ H CN

w o w

<#> CN OS

vo vo c*

EH o Eh

g o H >h Eh Pi o < W 53 O > O H W Eh Eh H O CO g 53

W Q H >H O

h3 & >H P< o Eh H Pi S3 Eh < CO 04

W o 0

>< £ p Pi O Eh w Pi 2 > Eh 53 CO w « Eh CO

in in o ro H

^ H

^ <r» in

cn ro H CN

<A° o OS

OS

a 8

o H >h EH Pi 0 < £ O O H Pi W EH EH H O CO

l ! a

& P H >H 0

J 5 >H P^ O EH H Pi S3 < EH *3 CO P<

fe O 0

>H £ P Pi O EH

2 W Pi EH

2 > EH 53 CO W Pi EH CO

B H CO j*j < W W P< J CO H

CO >H

^ 0 s i O w § Eh C* 53 h O Os w W EH H 0 EH

Oi CN CO ro ^

<*>

OS 00

^ VO OJ

vo o ^ VD ro

o O CN CN CN

w o w t 8

•o <D 0 a

-H 4J a o o

166

n a\

w

o\ i—i

•H

X! JJ t n a a) n 4J

w

c i CM

a) rH

•9 H

g

H

•p a -H O M H >• tt EH O

< W « O > O

•H H & W EH EH

« H O CO 0 Eb! £

•H 4J W «* P O H * 55

• r t A fe <w Jx O -H H H V Pt rtj EH a < W CO <D Q*

-d H W

O O >1 •P w O M EH W M « 0 > EH

CO W

*w & O Eh

CO

in o ro ro ^ H

VD VD

h h h H H

H VD <N H H

W O W

<*» 00 00

<r» (N

i O

3 H

I H P"4 H

W o «

§ PQ Ptf H CO

g o

P4 O W g > o &

b< co

Eh §

a P4 H

£

0

s §

S B 03

§ H

s W W 04 ^ co w

co

W O

55 H W K O Eh

^ o ^ ro ^ H

vo vo ^

00 a \ 00

ch ^ CN

W O W

§ 1 H

8 H W H

% avl Eh r -O as H H

g o Eh

>4 04 0 W g

!*

H

o

Eh CO

W CO

1 & CO

00 VD ro ro

CO h ^

<*» I f ) 00

a\

o CO VD CN H

in

CTi O (N CN

CO O W g o Eh

<D

3 -ri

a o o

1 6 7

CI a\ in

a \

a •H <0 •U •H M «

a -H

a 0

•H 4J (0 O

•H

4J

a

• a H >f •P u « ft *W

0

XJ

0) a a) M 4) w

M

01 ®

r-4

•8 H

3 H

§

H (x< H

W P H

§ ft fa o

£ W

Eh CO

ro oo o\

i O Eh

>h p

s

I IH CO

I « IH ICO

§ § H CO

CD E H CO

pa fa ft hJ co w

CO

S fa J o

W W CD H

<T> O ro M H

00 00 ^

CN LH H rH

CT> Ch H

CO O W

<*> a <*» ro < CO H r l O r* EH ro ro H 00

O EH

O H >h

O EH OS CD 9

8

5

OJ (N < H S (N

U > 2 H fa Eh EH H O CO

S3

W Q

^ H >• CD o ^ in <T\ ro •J fe CM H ro

>h O EH H p$

< Eh 3 fa CO ft fa O CD

o\ >h ON (J\ H (T> H « « O H

Eh W CD > Eh & CO W

E-» CO

' CD S3 H CO

2 * W fa ft ptj CO W

CO H >H Eh J fa i-3 o > ^ (CE « g a

J CM ^4 w 5 c2 a 3 ^ W S3 CO o w 2 Eh 00 2 H S PQ PQ Eh O 0^ W ffi O 2 H O EH H CD P U 3 A Eh

s H

§

H fa

£ 1 ft

fa O

S 3 W

Eh CO

g

8

Jh ft CD

5 O

H O 55

CO

a & § H &

Cm CO

I P4 Eh CO

CO < fa W CO

fa O

W 5 5 H M 8 CD Eh

^ M (N ^ n H

H 0> VD

<#> 00 00

VD <N

n in in CN H

ro

o oo H

<J\ H

CO O W Eh O Eh

I <D (0 0 45 •U 17) C5 •rt TJ ;J H U a

168

s g on ra •H "H 4J» 4J k U m a

<0 Oi

0) 0) cm c a os 0 0 os u n H l 4J a 4J n w IO os CO u H 00 r* 00 V0 H in in r-

A O H CM O H o o o 0 > * a) •rt i • • • • • • • • •u a o o O O © O o o o 4J

• r t d I l l 1 1 i 0 (0 a) m a 4J u u

-H a> a> (Q M4 CM 01 os CO o in CO o • r t (0 CQ m O o © o CM o o o «W CQ

•rt <0 i • • • • • • • • • r t TJ a Q a o o o o o o o o 4J <D • r t

a 1 1 1 1 1 1 a 0 •o 0

A 0 TJ 0 0 -H •H -H 4J H rH 10 in CM H H 1 <D •P D a O O CM o HI O o O c U

<0 ® 0 i • • • • • • • t 0 <0 U HI a o o o o O o o o £ • r t O 1 1 1 1 1 1 0 m l

• r t U Hi o u 0 (0 8 a 4J H CI o CM in in HI H 8 «•

<D a) o iH H O H o o O 0 >1 •d H i • • • • • • • • u 4J H <D 0 o

1 0 1

O 1

o 1 O

I o i o o U (0

>1 a 0< 4J <D U cn 10 <0 in CO U> 00 CO 00 CO 10 C Oi A H CM o 00 V0 10 in (0 0

• r t • • • • • • • • • M 4J M-i CI CM CM H H H H H H 0 X! 4-> s

N A «3» a) V o CM co rH H V0 CO CO a a Oi XI co CM CM O OS CM OS a) 0 a (0 • • • • • • • • • M H a) *4 CM CM CM CM CM CM H H H 4J 0 M CQ c 4J CO a W 00 OS O rt* os in CO CM H 10 <0 a Ci co CO H O OS OS OS OS 00 (0 O • • • • • • • • • •H to 0) U OI CM CM CM CM H H H H 4J a> S3

rl M (0

> H co 10 Pi a) • U CO O O V0 os m 10 CQ cm Q) CM CM H o\ os 41 6 a • • • • • • • • • 0 0) © 0 CM CM CM rH H H H H H X! H •s

o a (0 a) 4J

H jQ 4J n 0 O n

O a> M u 10 O OS CO r- CM £ a (0 vo VO 1 r- 00 00 OS H • r t

a> <n os o\ Ol os OS OS OS OS i* rH H H H H H H HI H -it P

169

•a <D •a 3 i—i o K <D

m <D 3 HI 10 >

a •H to Q

•H

<N o>

VO Cft

<0 -U •H M ffl

a •H

a o -H •u 10 O -H «W "H •P a 'O H

•U

<0

A

CSI a)

•3 H

* a o •H •p (0 u

•HI

•H •P a a) TJ H > •u M (d P4

H Cl H • • •

in CO CO 11 CI H

O VO • • co in

i H ^ lO O f ) n * • • • • • • H CI H in H CI VO ^ M H

CN CO ^ CO H VO • • • • • • • O CM CN VO cm in f CO H

vo vo in • • •

H 0% CI ^ CO H

I s N o

vo in h CO ^ r l

C* f* o\ d co **

** o • • H in

vo vo • • vo

• •

m

a) > -H JJ <0 > M a) co

4J 10 M a

0) Q

a A o « u ^

o is

<0 10 to "H *rl <D 0) O H J 3 f j O H 4J O M O 55

00 CJ

co vo

co

CO a% oo co

vo co oo

CI m ci H

CI r* r*.

m <D

•H •u to 10 Oi

<D X! •p

<W o

a) §

0 •P

to O Q 0

CO 10

a <D >

Xi JJ

*W O

M a

-H

.P

I

0) 00 0 X 4J (0 a) »o 3 rH O a H

170

CM Ol a\

0 u

r* oo o\

4J U j* 0 U

0

S* *

a 0 •H V id u -H M-i •rl JJ a

<D

H

<0 04

*W 0

4J 01 a a> H 4J M

a as «

s

in CM

a> H

•3 H

a> jd 4J

0

<D 0 0

0 •p

• CM »o o\ O in CM in r- o rl a\ o\ H rl CM rl 1^ r* 00 E • o\ 00 r*- r* 00 00 r* l> 00 00 00 00 00 00 u

tt H • • • • • • • • • • • • • • • • a> CO 0 rl

• U 00 CM 00 00 © o in CM rt* o\ in r* CM 01 CM in

• 01 00 r- l> 00 00 00 01 00 00 00 01 00 00 CQ CO H # • • • • • • • • • • • • •

• • at

c

CO 00

• 01 > a *w H H H H in in H 00 VO CM rl 00 m O m 0\ vo rl

a) 3 I O O O o o O o o rl O o rl rl o o rl 9 Ol 0 •H CM • • • • • • • • • • • • • • • • 2 4 Q a\ O o O o o o o o O o o o O o o O 6

o a\ 1 i 1 1 1 1 1 i 1 1 1 1 a) • o H £ TJ

4J 0) c- •0 00 <M 3

•P • 0i 0 H M H CM CO cn H co in 00 H co o o m CM o\ U

0) 0 m I 1 H o o O o o o O o rl rl rl o o © M X bl •H CM • • • • • • • • • • • • • • • a 0 < 0 Q a» O o o o o o o O o o O O o o o •H

U a\ 1 i 1 i 1 1 1 1 £ <a H 4J <D

P 0 rl

a CM GO CM o u> o o in m cn cn 0\ vo (0 (0 o\ in VO r^ 00 00 00 01 00 00 CO 00 01 rl > > a) o\ • • • • • • • • • • • • • • • •

s rl H H rl HI HI H H H rl rl H rl H rl rl CI a> CQ a 0 •rl A CO

a VO Q\ H in 01 in CM VO r* in cn 00 cn in o 4J 10 (tf 00 m vo i- r* 00 01 o\ 0* 01 01 o\ o cn •H a) a% • • • • • • • • • • • • • • • • 00 S3 rl H H H rl H rl H rl rl rl rl rl rl rl CM CM a> S3

'd •

A 3 10 Cn rl 0)

in © cn 00 H VO 0\ CM in O cn 00 rl •H O •H CM CI cn cn in in VO vo 1^ F* r* 00 £ a 4J i 1 i 1 I 1 1 1 i 1 i 1 1 1 1 i H U 00 cn vo rl o\ CM F* o in CO cn vo rl rl 10 H CSI CM CO CO CO in in in vo vo r- 00 * Ol

171

§ A

ci o\ o\ H l t"* 00

a 0 •H 4J Id •H

«W

N 4J M <d 04

a <D

§

u

a o -H •P

a •rl *W •H JJ

G a) •d H

ir & a«

M a) >

0 g 3

*W 0

to a M

4J 4J (0 01

V0

CM

•s H

*

ST > u 3 CQ

— + + + +

r> oo o\

a •H

m M a> -H

-rl -P a

*0 -H

S •P u Id Oi

> PH H in o 0 id • oi 4J *0

0

o + — 1

+ i

o 1 i t I H i CO

as 1 • i • H 1 i ffi 1 i H 1 i o 1 i

s o 1 i ro o 1 i H • I 00 i 00

ro 1 • i • s 01 1 i ro J p 1 i

a CQ 1 i

f* ID CM • in ro

ro

o ro .

CM in

m h vr> o in • ko

ro in o H o

+ +

— +

o 0

01

H

I

o o

o o

CI

10

— + +

CM

ci

in ro

ci

o>

ro

to ci

10

*0

o

oo

H

+ —

10

CO

00

o in •

ci

ro o CM

10 o\ H ro

00 ro • 10 o

00 o t*» •

in

O i i

S3 • o H CQ l o

j*j < 1 •

M to 1 Pi a 1 CQ w

CQ |H a Ik«

O

H g CM

g H o\ w W a\ 0 IH H

o o

§ §

£ H

I w CQ §

u

o o

ci

o o

o o

a« Q CO N m H \

a

H 0 0 H U

M

<d M 0 •U V a>

<o 4J o •M

<D J*

C 0

•0 ©

n id

CO ®

tn id jj

a 01 o M <o p<

U <d

CQ 0) •H U 4J a H

V0 0\ H H I""**

OJ OO CM as CM CM H IN

1 7 2

CI a\ o\ H I

as T-a\

\o as

m m <0 H U

10 -H 0 0 W

A <0

A O

•H 4J (0 U •H

-H 4J A <D •D H *

id FTI

RO ®

H

•S H

U

W

8 H £

O H &

H «

O 3:

CO

A

W H! CO Q CO

§ 3 ^ S O U> (TK O H S3

H CO W CO

§ 3 & O

IS RL CN (N (NL

(N W) RL ^ F H OJ

as O CO ^

OJ VD H VO R* OJ H

RO C» O OO as

N ^ H ^ N H OJ ^

ID (N O OJ O

(N IN O H IN OJ H ^

as OO H RO H

OJ ^ RO IN VD H OJ

H H 00 U) OJ

H 00 0- ^ OJ H CO

CN O IN P-

RO OJ H OO VO OJ CO VO

H

OI H IN OJ

OO IN ^ A\ H OJ

V£> O- RO

CM

ra JH <L> •H 4H -H JJ

•H I

A

I§

H 00 H O CM CO O*

R*1 A* >

3 W

A 0 *H 4J U A) RH 0)

E£5 O NJ A)

A •H

03 •U G A) TF fi 0 ft CO A) M

RH RH CD .

TJ MH <U 0 TJ

3 01 RH B U u X <D A) -P

<U A •H FD

*3 w A) <D v -H A* •P RH U 3 <D O ft RH 05 *H O A)

A <U 4J IN 0 AJ

0 CO •U <U 01 CO (D 4J A) A •H A) MH O •H

4J A> G ft A)

TF * H

Oi a\ o\

173

VO as H

d)

4J U 3

4J CO

W W id

<0 •H O 0 CO

a •H

CO <1) 0) a

,u a CM

rO

a) H

•8 Eh

ci o\ a\

r oo as

a

a> ro * 00 M as 0 H £ rH o\ <0 g Ol a H

Is- ro <#> • • O

in ^ ^ in

o H

00 CM <*> • O

id n o ^ in h

in in <#> . . o

cr> o o ^ in h

^ U) e° . o

i> cm o vo ro H

l r* ro (A® C3 vo • . O 0 a* o cri O $3 H r- CM H rH 10 0) P £ 10 u 0 56 a> SB

M (0 u rH

CM 0 cd • SK

ro a rH cd «0 £

i fl a flj 0 s a

a. •H

tQ a> §

o

a)

8 « -Q CM

CM o\ Ol

C 00 a*

CO oo as

as

0\ H

vo a>

Ctt H <Ao • • O

h (N O CM r- H

in in <a° • • o

G\ O o CM r* H

VO <A<> • • o

^ in o ro vo h

OS H

CM

VO in ^

u <D

o u a) 0

<A<> o o H

6 0 «

•U

<N Ol o\

00 o\ H

CO 00 Ol

o\

a\ H

VO as

ro r* <a« • • o

vo ro o P- CM H

00 CM c\° . O

ro vo o !> (N H

ro <A<> • • o

CM l> O l> CM H

CM 00 <A® • • O

<7* O O VO ro H

00 CM o\° O

O O H

as in

U d)

a> S fi

o •H

B V

J§

174

!> OQ 10 <0 H U

U 0 m

a 0) o •H *0

H

tJ M 0

< CO

ro 0) r l

•8 IH

w CD ro ON CO • • • • •

p CO ro (N VO r-p in CM H ON

H hH ro CM S u •

ON r-ON a CM CM H ss ON H H in

H CO • • • • •

CO ID in vo ON vo ro ON

o 3 £ U

PA IX) H VO r^ ON

CO • • • • •

p CO LO O o CM 00

oo ON 0

2 U

H CO

in (N H CO

H VD 00 ro 00

w . 4 p p

ON

ON 0

H CO

ro o ON H ro ro h ro ^ oo

M CM H ON

oo h oo o ro

(N in

^ ON CNI CM

CO r- H o ro CM Pi *5 ro H ON ON

1 § • a ON R* ON W CM O 00 CM CM H CO • • • • •

P CO in o ON ON % P in H H 00

H w 00 in vo ro S U • •

ON 00 vo CM H ON o CM ro

H 55 ON o VO CM % H CO • • • • •

tn J*J CO in o vo o ro a P4 ro H ON

• H O A •P I S u

ON ro

CM ON CM CM

0) U 05 CQ <D -H 4J U 05 ft J-i <D

RCJ 4J o

o

4J

CO u <D

•H 4-1 •H 4J a <U TJ

a) >

3 w

a o •H -P O <u

XJ u nf a)

a •H

CQ <u tn oj 4J fl a) o a) 04

o o

CO in CM vo ro 00 CQ 04 *5 ro H ON CO O a 05 £ u H

U

W 00 ro CM a A CO • • • • • •H P CO H r- in ON ,G P VO H TH ON 4J H a 00 -H

^ £ U •

vo r- CM ON 0 ro (U •

H JZJ o VO CM ON O H CO • * • • t 05 <U & CO ro o in O ON T5 P4 ro in H ON CQ o a <u rH & u XX •H U

(d M G iH a XJ *H

(x) td + a J2J JZi CQ PQ A M Q O 3 H O 0 O J2J U a J a •K a

175

CI OS o\ H

l

vo as

« 04

• d a «$

m Dd <o

H u

*

a 0 -rl 4J id U

•H M-l •rl 4J a 0

*d H

N 4J M al

01

M-l 0

£ - p

tT) a a) M 4J t o

G 10 0) s

CO

a> pH *

w a P P

CM H 0 \ Ch

§ H CO

CO s a s u

3 u

w • 3 co 0 <9 p H 2

§ H CO

g 5 o &

w

P r o H oo as

V

5 U

§

H 03 6 CO

* a w i 4

P OS H i> 2 CT>

u

§ H CO M CO g a £ o

w J CO P CO

IS o

p H £

§ H CO & CO g a SB U

r o i n vo m co i n

H H H

VO CO r o oo o% i n

H H

vo r o as oo vo

H H H

r o oo vo 00 as vo

H H H

vo a \ co a \ oo vo

H H

r * oo r o co m vo

H H H

H 0 \ h

H H

i n CM M <y> o r *

CN U) I s

^ r o H

CM CM CM

r - cr> r o r o H

CM CM CM

m u

a) •H IW *H 4J a

o a) H TJ • •H

O CQ CQ aJ i—i o U1 w 0 u u XI

CM a H cd O 03

u

<D •H M-l •H U a a> TJ •HI COf OS CO

O «J • HI o u

iH A3 5-1 3 4J a a) JCJ

in 4.) o . d o a) a) 3s 4J a) x*

<3* o

4 3 u

£ 0) u 4-> CO

a cd a) 6 a

- H

(I) u G <u

a> u > > i * <1)

•H V <D <4-1 -P ^ O nJ m a -H > iH P4 a) Ti

u i a) 3 co a> <D

CO 0 <D CO 4-1 XS a XI XI (tf M-J EH 0 nJ •H rH - H u • 3 a p *

176

Cn a

£ X! 43 43 0* 0) a a 0) <D u M 43 43 CI CO w o* Ol >1 a h 4J id i u w r* id •H 00 Pi 43 C\ M H m P4 * a •d u a ® id "H Oi CTl a -h a\ a -P H o a •h a> a 43 u 0 (d h •H o JJ •H W Id *W 10 0 •h <d •H 43 H a a •rl o i 43 *d w a H CQ 0 •0 >1 M H 43 u u >1 (d *0 43 a« a fc

(d id a 04 -H I HI ta id 0 M di 3 a id id A fc u in ro a) rH JQ (d EH

® ^ H vo r-M • • • •

<d in as co r* <D (N H OJ H <D H ov ro VD <T\ •S ^ O 00 m jj in vo in w M (D h V£> 00 fO (S * ' " ' 0 VO O (N (N Kl H CNI <N (N 43 CO

id M © u) co m in JQ • • • • •H co vo vo ro a <D 43 •h in 03 <N C\ • H O • • H Pi OJ VO H VO

© in h h ^ § o o\ r* in CQ <T\ h f- 00

r-oo as

m m a> 3 •hi o u XI O aS Eh »J CQ CQ at to

M 3 O H rQ O (0 Eh w m w Ul as <d

v a i i a) a) O H H W TJ *d (0 T5 TJ H -H -H u s s

Tf a) tj 3 iH U a •H a) M «j CO 0) •H -P M cd Q*

U o •n <d B

a) a) xi 4J a) £ •P

d o

(Xl (TS i—i <—I r-i O O Oi Cn a a *H -H

X X M U o o IS IS

0} a) (d 4J a <d 0 5H a) 01 w to cd

a •H Xi 4J •H £ <D u 05 CO <D -H U 4-> d W

a> s cu

CI <J\

I r * oo o\

111

<a v •rl u m

4J m a) M

o

a •H a o •rl V (0 a

•H *W •H 4J a a) •a H

>1 4J H m a

0

a •rl DO J*

rH m

0)

s &

W

•P 09 <d a>

a> 03 •P w

1

CO

0) H

•8

CO 0

CO co r - CO VO H o 00 4J W o (N o (N as H co H nJ OJ o o rH H o H o O M co • • • • • • • • •

u o o o O O o O o o O B 4J <u a Q <u * * *

*H * * * * * rH O * * * * * * <15 •H in <N r - 00 H 00 as

M-4 H O CO KD i n (N as vo 0) M-l O o O CN o o o

XI <D • • • • • • • • •

•H O o o o O o o o o o u 1 1 1

, as <N CM KD 00 00 00 rH CN W 00 O CN 00 as r - CM rH W

H O O O o o o O O co • • • • • • • • •

o O o o o o o o O O o 00 r4

a> flj -r l J o

•H MH MH a> o u

a) > •H •P aj > M a) _ o

-H O <H U

<y o u

* * * * * * * * * # * * * * * * * as o H as i n as (N 00 as o H OJ i n 00 V£> r * o O H CO rH o O O

o o O O o O o O O

. as Ol 00 o in 00 OS W vo o H l> 00 vo co H o

H o O o o o o o o CO . • • • • • • • •

O o o o o o o o o

a) 4J O rH a H ,Q rtJ |H flj 4J

•H cq a U G H D

O O P > U ri! W

co co

3 o

3 O H

S

ft H CO

P4 | l j (T> H CO H a ^ w w ci Qi PQ -H W £ g g Q i s h

e g i< CO

a; W co ft J H

cx> r * 00

VO

in as

* * * ° H o * * * * * ° H o

* * * * * H co CO H o rH 00 CN •

VjD O H 00 H (N r - O rH a o o o rH V0 H o V

CU O o o o o O O O o

13

LO

o

o V cu * *

o V a *

* *

178 A to

•H •u •rl u m

d "H

CQ <D 3 m to H

> •H h

a 0

at a 0 0

•H 0 •P a •rt <D LTI vo 00 00 01 CN 10 U IT) CN CO o H CO 0 <D * • • • • •

Q* «W O o o o H o

>i -H 4J Q

00 co 0* O a\ •

(0 a CO o vo CO H in 01 H • • • • • • +

»J H CN o H H H U I 1 1 1 I l 0 a) <D 4J > V

•H a in © <D a> a\ <T> 00 VO i o n ro H vo 00 r$ M 0 • • • • • • B <D o rH OJ rH o H 0 A *W

H •H

s H OI H r* OI rH a)

<D 0) o Oi o\ 00 a\ rH 8 3 • • • • • • <d 8

M c* CO OI CM o OI a - a 1 1 1 1 1 1 CO d ® 3 O -p

a d (0 0 00 ID 00 vo 00 •H .M u CO 00 in H 00 0 a 9 • • • • • • Oi

«W 0* CM O1 ro o OI I •d H4 d a •rl <D 0 Q > ft «<• CO Oi ro <D (0 o co 00 Tt« rH <D O • • • • • • 0) rt U o H Oi H o HI

r1

*W M fd

0 •P d d

m a> 0* in o\ OI in 0 a CNI •d "ft 00 OI OI o 0 o \ d • • • • • • CO

•H o \ 0 01 H o H o H d 4J H Pi 1! 1 i I I 0 •H a -rH 01 * a> 4-> 0 >i rt 03 •rH Oi 0) d <D CO Oi

> 0 O d 0 0) M •H *H 0 ft 3 3 •p > •H m CQ <d U •P >1 d to i—i <D d -P fd H a 4-1 C/3 •H a)

0 d d *H d £ a •rl H •P o U 0 m 4J d -H 4J B <D a) 0 •P <u •P CO B S (I) d e (d •H 0 <d

HI cu d N TJ U H H 6 -H a) CO • r>1 a) rH rt d <u

rH 0 > (d (d d) •H (TJ to r-H 0 d 0) a) tJ) U

® © ft a 0 e 04 (d 4J rH a) e X •rH 0 0 u d •P d to CD X -P o u a) w TO a) m d rd cd d d > H 0 H £> EH & H w < *

179

W ^ r* vo on t*- on . in o oo n tn

CQ LD O O H O O

cod 4J •P O (d cd id <D h in H o M U XI vo vo o o 0 -H vo in o o Q P . . . . 1

CM l 01 Q

r l pa CN ON VO H «d X » oo in O ON u 4J co in o o h a> •H . . . .

X) *rl •H

X A Cd CM 0 o r» r - vo ON u XI H CM O VO ON Oi on in o o r l .

V l> o

h r - on cm in o o

in ON o

r - in cm on on CM r - H

. cm •

in cn vo CO H cm ^ ro

• H •

O CD r0 VO ON CM 00 H

• CM •

O ON CM o on ro CM CM CO

on CM CM

X5 00 ro in

ro CM

r -cm in o o

H H ON ON H 00 VO 00 O ro o 00 ro o O o O ON rH in

ON

00 • • • • • o ro in

ON o H o ON 00

XJ as X) rH o

ro 00 H H

oo cm ^ 1 r - on vo

cm ^ ro

U in r* cm oo in i vo h o

o o in

i

r -o

00 o ^ ^ in co o o o

X* O 00 H ro on H O

00 ro o

fd r-» ro

vo on CM

ro on ro i

oo • in o on oo

i i

a •H (0 4J •rl M CQ

•P id

U 0

a •H

a © -H 43 M <0 a<

<0 o

H 0 0*

u 0

m

ft "H 4J >

0) 3 on CO H

a> •d £

•U •H 0» 0 *

M (0 a

-H ft

XI (0 H

m a o

-H j j o <p M

•H

•H K o M a«

a) a a o u

> | •P

-rl

•rl X 0 M a<

W vo oo h U) ro U) . on in o h ro vo

w in o o cm o o

fd cm o r -

rQ on ro ro r* cm vo on in o o o o

I I I I

W vo oo r -. ro in o

CQ VO O O

id id XI

ro vo H ro vo cm o o

X* ro ro cm in in vo oo cm cm vo on r* in in o o o o

CO

XJ fd fd cd vo h r - vo cn vo

J3 o cm cm ro r l o

ON VO O O H O

H l

H in 00 o ro O in ON in ON CM in i l ON ON VO in in r - o H ro ON ^ CN 00 CN CN CM ro o o o o CM ro ON ^ CM . . . . • • • • CM l> ro

• h r -O H 00

u cd X5 (d cd (d cd CM o vo in o ro 00 00 CN ro

vo H o CM l i r* o in in vo CM i CM 00 ON o 00 ro CM O

1 CM

1 1 i 1 i 1

ro r - ro ro in vo in o in VO CM CN vo 1 ON 00 vo in vo r - ro CM co

CM ON CM CM CM ro o o o o H ro in cn • CM . . . . • • • • • VO o CN • . h r -

H o H 00 V cd X* (d cd <d (d • CM

o CN o ro vo CN 00 in ON ro CN o O in 1 ON 00 00 CN vo O 1 ro o o H H o ro CM CM

H I

X3 O H ON ON ON

ON in CM •

CM

U ON O ^ h ro H CM 00 O O O VO

cd cd in oo ^ CM VO CM

cd ON CM CM

s o ON CN I

H ON

I I I I l

a» 00 rH •H on a) Q* Xl H > •H CQ

a) cq x : U a j CQ CQ d)

*H fd U A d H (1) g

4J Q 0 u c: <D CQ 13 (h H -H % B t j a cd u <u - o a a cd j j >i a) 6 (d al a . « o xl XJ fd CQ rH CQ 4J -d 0 u w 0 <D 0 -H 4J 4J rH <D 4-J d u o) a o z id U g -H 0 ^ U T3 iH 0 0 cd cn a) a TJ h u ° d a) 0 0 -H (d O u o H w u o a D « CO S S ^ CO

•P <d

J* d u g

JH O O o a t j

• n a X! cd - h s w

CQ <

•P o a>

M4

w

a) P a a

o TJ

— 3 73 <D O CQ O J*

XI •H CQ

TJ <u •U u

•«H TJ dJ

CM

a)

•H J

>1 <D >

<u U) w o o

4J U <U

o u

CN ON ON

r* 00 ON

>1 a) > U 4-» 3 CQ co a)

j j in H 00 ro ro C CM ON O vo CM OJ rH <1* rH vo vo o CN ro vo vo o ro i r * 00 ON r - 00 VO 00 o ON o fd 13 vo O o CM o o CN CN CN CM CM in ro o O o o ro o in in g a) vo

• . . . . ro . . . . . . • • • H 00 m • fd rH .. . 00

rH ON 04 •H o H 00 cd

fd fd fd (d X5 fd id o fd O fi 4J CN ON in ON 00 ro l> in o r - CM rH in ro 0

a) 00 00 CM in H CM ON CM i r - H in r - CM in CM i •H a) 00 vo o o rH o O O vo o rH o in o CM H o 4J a 00 vo . ro • • . . . . « • 0 0 ro i

1 1 i i 1 0) rH W

05

vo ro 00 o ON 00 vo r^ ro in rH vo CN in H rH •

CM vo o ro ro vo vo vo i r - ON o 00 00 VO 00 ro fd r - o o CN o o CM CM CM CM CM vo ro o O o o in ON U II . . • • • . . ro . . . . . • • • • in ON H <D V

• rH ON

• S3 & 0 O

xl CQ

•H 4J •H !h PQ

6 O I Jh

M-l X*

<d o *h o

Tf • II

cd v .u fd p4 Q

* CI o \

I

r* oo o \ H

180

* 01 a

•rl 4J O >

<d a a H

G 0

H 0

•d 0 « H 10 a 0

•H 4-> 0 <D U CQ|

•HI a) Q CQ a CQ

H 01 n <U a> > > •H

(XI H © rH

•d rH 0 < S

IW 0

4J •d a) § 4J

-H 05 K 0 O <u

M u

Pi 0 5

*

CQ 4J .. <0 4J <D U H a)

MH tn 4-1 G w -H a a) CQ (0 CQ & CQ 6 H 0 0 rH a rd H a) ro CJ • <u

<«* O

® rH X) (0 H

PS H P

vu»

X o 04

P I

# X §

cu

\ x f

P4

O

ro co ^ cr» ^

a vo o

a> >

•H 4J > rH JH u cd <D 3 CQ 0 a) a •Q A

0 (0 •H u •3

(d a* <d 4J H (d cd (d H in (N (d P o\ iH in in vo C* rH o o CN as vo M-l vu» H vo

a X

• • •

vo r - vo H X r - <N 05 n H ro N o rH H

4J U <D

iw MH w

(D 3 CO CQ

TJ

* s Pi cu

Ui»

H p

H 00 rH O 00 O O

O ro O

01 a

*H T3 ti a) p

CO

4-)

s I a) > o 0 N X m

4J u a)

4-1 MH pa

a) 3

H P

yxj*

X o & P I

* XJ o p{

P I

vu»

PS H P

(U rQ 00 00

o

^ as (N

h H n [> vo

43 VO CN ^ 00 OS OJ

^ r -

H r- o

CQ H

rH td 3

nd a> -H a) > > >

•H -H -H 4J 'd 4-1 (d c P > rH H > rH ^ M (d u u (d a) 3 M 0) 3 u

cq o <u CQ 0 <U

a xi A a o m -H o (d -h u d J U J J

H 00

Lf>

O

O VI PU

I

o

ro vo i—i

• a) vo Tf II 0 . B

> >1

u 4J *•>—' •H

6 6 0 H •H TJ O X <u . 0 <u o u

VI Pi MH CU

MH i 1

0 X

<D o

<D a) • s Pi

»—>

Pi

Cn 00 • <*

<u OJ rH ' d 00 a)

•

rH o 0 H

4J II fd

II rH

rs > cd X! • a

o 0 d) s — ' •H JH 4-) (d H O

O 0) CQ O in a) . -H

•H o P }H VI 4J 04 i a 0) i PJ

* cd P

a •H CQ W

<d

§ •

O o a

CO a) 4J o a a)

181

* CI 0> I

00 as

a •rl V

£ ©

0 m m H

a o

© •0

£ H <0 a o -H 4J O a) u •H p

<o

a u © >

© to £

>1 •p •H 6 •H K 0

Of

to •U <a a) H

cn a •H n CQ <0 Pi

8 o a H

d o •H •P (U N •H r-H OS CJ O *H 4J as

J>

n © * 2

•H © -U h a ^ o (d o h —

Pi H P

vu* X §

Pi

* X O Pi Pi

vx/»

Pi H P

(0 (0 (0 oj in Ch oi a\ o> a\ in o

^ in

ro ^ cn O H ^ 00 10

H M H

0) > •H •P as > u <D CO a o u

pa

T5

oi H P

VXi»

X o Pi Pi

* X o Pi Pi

UJ»

Pi

as as cr>

in (N ko ro ON

CN

H o CO ^ > (T\ oi r- ro

XJ XI

<U > -H V as > u a) CQ

3 O

a xi o as *h u ^

as

a)

>1 4J

i o u

a as a> §

u 3 pa

4J u a) M-l iW pa

Q) 3 CQ CQ H

TJ •H > •H

Pi H p

UJ»

s Pi Pi

* X O PS Pi

yxs

Pi H P

as in vo

in CN

as as t> o o ^ m a\

as H

O KQ G\ H R-ro H ^

a) > •H 4J as > u a) CQ a xi o as a A

00

CO

in o

0 VI 01 i

0

in ro VO r—1

. <D VO TJ

5 1 o x>

§ ~ e 0 H 'H *9 ° x a) . o a) o m ^ VI Pi m a,

i M-l i o x m & 2 0) prf <l> .* p,

tD 00 •-® <N rH TJ oo <D tXl

, • TJ j

H o O H £

t« n aS . h

m > aJ ft x • a g u o

<D *— *H M V .. as h o (U

0 <1) CQ O M CQ a) • -H <u -H O p 4-> U vi

01 • •P a <D

a •H CQ CQ as

o o a

o a a)

I Pi TJ H

as p uj»

* r-oo ro oo a\

182

% 01 fl -H 4J O > ®

a <& H a 0 H a) •d 0 S Hi (d a 0 •H 4J O <D M col -H a) Q 3 CQ CQ CQ 3 H a u

3 > O fc

H rH •a rH 0 < S 4H s O jj •H <u

4J -H cd X 0" O a) M SH 04 U d s < ra 4J ••

10 .P ® u l-i <D iw 0> iw G w •H CO 0) (0 3 (0 CQ 9» CQ n H 0 0 rH a Cd w Jh a) rt* fl • a) o a) r-H A <d H

oi H Q yxp

X § 04

#

X O & 04

>xp Qi H Q

XI r* in vo in ro in ro cd as c

o CN ro

fl3 ^ vo cn ro (N r-ro cn OJ o ^ ro

<D > •H 4J as

i—i > <H td ^ M (d M <D 3 M <D cq o a) a a xi x> <p o (d *h o u a

Tl

Tf

UJ» X O 04

•K X o Pi 04 vxi» H Q

.Q o in

0\ H

t-4 -H O* XI H (d XI cd

a\ in fl Q o in r* H ro 0) o H in in o P ro H H . . CO X

• • •

H a\ X <N (N 4J o H H a OJ <u 04

^ H H vo rOH co o n H CM O

(U > *H JJ cd > u a) tn

cd •3

*

X o 04 Ui» 05 H Q

ro o oo vo in vo (Ti 00 (N (N O ^

a XI XI o a

M 3 O -a xi xi o cd -h • 1-3

cd u a;

GO

in o o VI P<

B O TJ a) (D

O

in ro vo rH » <D vo *0 II 0 . B >

>1 6 4J •rl B

H •H o X . 0 o u VI 04 p4

1 1 X XI o 01 04

*—*.

00 04 rH 00 a) • T3 o 0 H e II • rH > fd a O 0 "W •H -P H u O a) O • -H p

0} CQ cd & o o a

CQ

o * (d

•u o i a 0) tj H Q Ui*

183

•o ® I

•rl •P a o a

r* CO I ro 00

G •H 4J > © 3 a a H

a 0

*0 £

CO

10 ti o •rl •P O <D M -H Q m 3 a u >

'O £ J? •rl

•rl

0 u 0*

m •P a a> IH

a •rl a to (0 cu 8 u g w

Q) r l

•8 H

a o -H 4J fd N

05 a o -H fd

-P u <u iw iw w a) 3 CQ m H

aJ •g -H > -H

OS (d H o (d P ro CN ro VD r-

Ui» o vo X vo CN o H 04 04

*

X O 0* H H KD 04 O H O O O

Ui» • • •

o H H 05 H Q

a> > *H 4J fd > r~1 u M fd <u 3 CQ o <D a XJ 0 (d •H u •J

as 3 TJ •H > •H *3 a

os H Q

vu» X O P4 cu

fd fd ro H 00 vo O <J\ 00 H

r- H H ro

*

X O A 0* ro LO CTk 04 a\ CN o\ 00 H 00 UJ» • • *

o 00 o 04 p

a) > -H 4J fd > u 0) 03 a o u

m o o VI a*

in ro vo a)

u> T* < • o jj w •H

B B 0 H •H Tf O X <u . 0 a) o u u VI 04 Q*

i 1

0 X a) rQ o a) 04 a) • s 04 04

00 <u CN 1—1 TJ 00 a) TJ fl h O 0 •H H B CQ .p II CQ (d rH fd

JQ .Q (d -H a

X <D cd CQ a) -H M •P fl a)

(d & ti § o o •h a 4J Jh u a> i\ s ° s I c a) as TI H Q vu»

184

tt <D

•H 4J U (0 0«

<0 O

•H •P

0 Oi

a o m <i) H •S •H M 10 >

a) 3 w (a H

«W 0 CO 4-> O m

H

4J a 0) M © *W •H Q

a) H •3

10 a o

•H 4J o a> M

•H p

(nj OS ro H H R*-m in o o o o

ro co o in in in in in H

o o o

o o o os ro o ^ in H o o o

it -H X 0 14 a«

o I

r* vo OS 00 r - o

KQ H O H Ln u> (N VD in ro in

o O O o o o o

a •H 4-> >

CD > •H •P (TJ !> U «D

ro ICQ

CD O a cd

-H

cd

0) oo a & &

cd -H a A

<u > •H 4J S> U U <D 3

I> w o CD a J Q 0 \ O HJ H U ^

<u > -H -U> > u a)

<N to

0 1

<D O a cd

•H

o> fl XI X* OS O 05 -H H U

>1 •P M cd Oi

o m a) u o

HH

d o

•H 4J G

a •H a) u a a) JH a)

a) M-I . P M-I

a -H •H T* 01 o a j j

•H 4J o >

Tf O £ — , Cd *0

a) 4-> a

rH (D ,Q CQ (d a)

CQ <u

u 04 -p o a

•H iH cd >

a) 3 to a)

cd A CQ

a) a <u o £ *H 4J -P <D O

^ 0) rH

u a) CQ os a o CQ

M cd <u PM <u M

CQ

4J a

Cd <D • H

CQ L> a) - H

• H M-I t n u •P a w

a • H RH cd o CQ

CI o%

00 o\

185

a •rl <0

•rl u m

4J 10 © u o

a •H a> o -H 0 a)

£ 4J M (0 01 •H •P HI

a) •o £

•H Ol 0 t-J

10 1 a •rl V H I

vo

•S

H (0 M • H ro 0) w VO vo *Q • [> o -rt CA • •

J o o <D > •rl •U <0 > cd u vo H a> in 0\ (0 XI ro in ti • •

0 o o u i

oo H in co o ro ro vo o (N o o o o o o

£ o O ^ G\ (NJ co h in o ro o o o o o o

. VO 00 00 H pa r- in o H (0 vo o o CM VI w • • • •

0) o o o O XI -H J M H P 0 cd cd XI .Q m 00 (N O 10 H 00 CM H A H ro O O

o O O

in r-ro vo o o

XI 00 00 (J\ 00 O O

C 0 •H 10 O

•rl 4J G 0) •d H

VI (0 Pi

O •H .3 §•

& Q

M a)

6 a) G . a) tn a) a ts Q ^ O H W

a) > a) a

a o •H 4-> (d u 3

H r-CM ro CM

CM ^ in in

o o o

v v o% oo vo r* o\ vo CM in in 0 o o 1 i

H ro O in o\ ro CM (N CM

O O O

H VO 00 ^ ro ^ ro CM ro 0 o o 1 I

CQ CQ as ft i—I -H u a, xi

•H CQ r-f Xi U (0 CQ <D S iu-9 o *H -P ? cd O & 3 <D t> g U O O ffi

0\ CM H CM H ro H 00 00 VO in 00 VO ro CM CM in o o o o O O o o o o o o

U u cd (d (d ro vo vo OS in ro ro CM CM 00 00 in ro in ro vo 0\ o ro CM o o o O o o o o

i i i in CQ r- <u a\ o\

CM . CM 'd H <u 00 rH in *H

Ch CTi CM in vo o in i (d in oo in ro in vo 00 ro 4J (N CNI o o o o 1

<1) o o o o o o o o a 0

in o cd XI XJ A o (d <d (d ro CM o o 00 o vo vo it CM o H ro H 00 r- V o r- ro H H ro H 0*

H O H H O o o o i 1 t o

H O a •

u II <D V •a ifi 04 (0 0) TJ 1 0 N 0 XI

43 XJ 0 •H •H •*.

n rH H m •H <U O a X O 0 0 •H •

CQ nd •rl u J II m TJ a JJ rj 04 V a a (d (0 o M-l 0 xl (d CQ r—1 0 0 a) 0 0*

•H JJ rl <u 4-3 H 0 a Tf 01 U TJ rH 0 <0 C w Cn i 0) O -H cd o > fd •H CQ m 0 « E s £ 03 M S < H A Jz; (d

CSI a%

i

r -00 as

186

d •H <0 •P •H M «

4J 10

M o

G - H

<D O

- H

0 XS

a

a>

4J

£ 4J

<0 O .

•H 4J H

1

0) •d

i

•U •rl JQ 0 U to

(0 -H §

a -H •U

r l

£

•3

(0 H a)

A •rl J N >

-H •P 10

£ <d a a o

u

w

cn

o vo r* m ^ o

, Q

H <0

0 )

A -rl Hi \

M

0

• 8 a

<0 (0 ro ^ ^ v d

^ c n

H O

oo in c n r * i n a\ o o o

H CO O

ttl (0 <0 A (0 f0 r o 00 00 l> O 00 O oo i n 00 CT> CN r - rH

i n H CN CN CN r o

r-l o o O O O O O

(0 (0 H O

O I

( 0 as i > o

0 1

10 a u 0 <D

•H • d . 0 10 (0 a) a i J

-H M-l *W •H 0 •U a a © 0

4J • d •H

s H

(0 4-> N w rH a u (0 0 10 > u Oi w

6 o ?

M O O o a t j

• n a rU ( 0 - H CQ s « <

> 1 4J • g

- H

X o M 04

0 ) 3 a CQ H

r -H co cn o o CN

• <N

* C0 (0 * r o c^ CN 00 as r o 00 1

w H r o r o ro i n H CN as w CI o o o o O o 00

w • • • • • • • •

o o o o o o ro CN

T3 0 0 A

<D M • H •—l

CN CO © ©

tD O ^ £

X -H U 4J (0 B 4J a a) •U CQ

CQ a 0 o

o •H •U rri UJ (0

a> 4J CQ

CQ d) 0 -P M •P a) a>

rH •rH

<D C0 •M 4J •H l rC <u £ a

0 B 0 U i n

o

13 a) ii > V

•H ft U <D i

TJ 0

CQ M H 0 O U •

U II W V

04 nd U 1 C0

TJ .Q a (0 •«. 4-> H CO O

O 4J •

CQ II 3 V

,Q 0 04 &

1

* t0

H (0 •H to © V m -H d

187

§ •rl to 0«

4J to 0 Oi P i 3 T » ca a)

u N a V 0) to to m o i n r o <0 <D o j o cr> o j Oi 4-1

M-4 i n vo o \ Q) -H H H rH H > P I I I I

•rl 6 •P vo -P (d a\ CQ i> <n to to H -H a) m i fS 0 U fi Oi

o \ a) M-4 r - -H 0 cn to * *

H a> in m h i> S CO vo o o •P " . • • • •p| CQ rH 2

2

4

3

to a m i i i i m o a a -H -H 0 4J Q

*H (8 ' H 4J 3 M fll H O 4J (d CO >

W to O 0

-H

| U 0 •p a CQ O 9 O H H

— B P (D CO (CjJ 0 > 13 S3 n-O -H O S ! W OS 4J U Oi 04 04

a ^ VI ^ — 4J <D d -H - r» 0 a rQ -H P 3

CO r-l (d

to (d 3 9 "d k' rH <U H a o (d > H (d to > a) -H

P4 w > ^ b H 4J 04 - H O • «d to *i! u j j a) > i J> 0 -H O P4 a> o p4 H g a> CQ

M to ft (d o ft o o P i 3 -H a CQ to "H P« co to O to 4J Q < a) u <u <u

a) 4-> w 04

in © H ,Q m

r* as

r - m

CO o o

> CQ • h -pi o 4J CJ > oS -pi -H > S 4J to O a) a) a) CQ a o o

a)

5 A" to 3 p4 CO

<u > <u -H > j j •H U 4-) a) U ft a) CQ ft o CQ to to -U <D 0) Oi &

rl H rl H rd fd fd fd a a a a o o o o CQ CO -H -H to to -U 4-> <u <D fd fd 04 04 55 52;

ro r-in i

S co S 04 H

XI + O

JQ il CO S 3 O a M-l o

S «* 04 3 CO TJ § "H o cq u d) n &

0) > a)

in o

4J cd 4-> o o to 4J •H

LT) O

CQ 4J rH 3 CQ a) to a) 6 fd CQ

a) •H >1

a <u u 4J

43 .u •H £

CQ 4-> CQ <D H

TJ a <u to 4J O

CQ •H CQ d)

XI -P O S - TJ ^ CO (1)

CO -P PH • U

<N H3 i n3

3 o A -P •H

ft £

rH 3 a .P u a) ••

* m (D 0) 3 to i—I <d o > -P

rH CQ (d t—i u -H -H (d 4J &4 -H to * O

£ O O CQ •P CQ a> -P

188

1 VO 01 <J\ H L 00 n a\ r* o\

v u a % CO

b & PI

0 >

•rl V > *4

m A 0 U

O

m H A> "O

£ A O

•rl 4J O A> H u o U

O u u W

CNI in ©

r l

•8 H

a) v o s

as as T3 JQ as A as O u ,Q as vo in o

X X 00 vo 00 H CO CN in in

<J\ o 00 X X X H a\ O o\ vo Ch H H (N a M O 00 o in 0> r CO 1 OJ H in vo CN

1 1 H

as as O as A as o U as as as

X O vo 00 in O vo (N CN O O

00 H 00 X X H X O o\ in r- a\ H H CO RH cr\ RH R-- O as a\ vo CO RH r- r- CO o (N • * . • • . . • • • Oi in vo CO 1 H

OS (U 00 00 CO

& cr» o o FT O ^ (N

X IN o

as as U O CN CO O O o R* VO H CN O

vo X XIX

.U 4J

CQ a) rH rQ as •rl U as > U 0 4J 4J a o aS •rl 4J Tl CQ a) a u 0 ft a

a O •H 4J U as 0} -H H •P I as .p W ^

s s FT O <1 w

-P — a)

> <D -H > -P -H U 4J <D U 04 a) to ft o o CQ U U O 4J 4J m a) a) ft

AS ^ H rl w as as

£ C o o CQ CQ >H B a) <D ft ft

O <J < o W

CQ £ O •H -P as 3

U •rH £ O a

V AS XX AS 43 O AS O IN CO VO O CO R- O VO IN VO CN O VO VO H 00 00 O TP 00 H [> IN CN

CN RH IN VO CO I H

in vo in

IN CT> CN CO CN H IN R- CO O O O H VO H 00 O CN

CN

23.

CN

vo in

vo vo in

A O •H •P U a) u u O a

as •rl u A) W •P w

a) a) IN O IN CO 4J u

H R* !> 00 CO AS O rH rH H VO O TL

AS

(N • • • • • (U B H CN

22 CN rH

•rl 0 it

VO O 00 VO H H VO H R- VO O rH IN CN CR> O O CN

CN

25.

H H

O

as a) 4J <U I ^ <U M FI O MH

O

K*1 •P •H U

O as XI as u O x* as TS in 00 RH 00 a\ CN CO CN o H O in VO 00 H o CO m VO CO R— RH VO H RH in vo 00 CO vo in 00 vo in CO O H CN in o in • • • • • • • • • CN • • • • • o

CN in vo CN I H CN in H RH RH

•P <D — > •H a) 4J > U -rl <D 4J FT O CQ d) CQ ~ FT A

CQ O 0 -H

- U JJ P4 FT U

a) H r) H S S ™ O O rH •H -rH 0J "T* "J* ^ as as o 53 FC -H <3 <1 4->

as

a) TS

£

<u <u

TS TJ

•P CQ

CQ as <D T3 <D d)

V (J) CQ FT <D O

Q U » <U

• .p <u a) N W as

& CN CQ I II

•rl 0) ft X!

(N *-I OO a)

x! H O

JJ >i as

5 H

T3 M a) as u £ as

I—I W FT T3 a) A ^ aS

RO H M H ^ OO OO O\ M O <Y* <R> CR* RH -N rl rl rl (C (C £

a >1 0 •p

CQ •H •rH TJ CQ 4J 4J £ u a) •H a as >H •H 4J CQ a> rH O -P o •H > a <u (X| CQ <u U w «N •rl $H as rH U 4J }H rH

X <D rH CQ o as >1 (D X a) >i as Tf •H a a M as oi O o <D 0 •H CQ E-t a •H 4J as rH •rH rH 0 u <D 4J CQ JJ as 4J as »H

rH 0 u •H CQ •rl O B d) rH *r~» u rH •m I U rt u 4J 0 as 3 0 TJ Ss a) p 0 0) ft S o ft P w W

A A O O

C| -rl -rl O 4J I) •H O U JJ <D <U as ft ft •U CQ CQ U 0 O <U u u ft 4J -P X A) A) W

KN 4-> •H rH as £ U !§

O CQ V

0) ft CQ FT 0) X I 4J W

rH i—I i—| as as as A A A O O O CQ CQ -rl U M 4J a) a) as ft FT G

B u o

I I

W PD ft ft

XI o •H

- 4J IW H CQ O O rH O G AS • tn A

0 as O V -H -H FT TS 4J

O 1 IH A RJ ^

AS MH

189

s vo o\ Ol H

0 4J CQ

r-H 00 <U S TJ <n 0

o\ H

H cd > 0 U ft Pi < H at

-H

a> 4J to •H

•H

<D

•H U Ck

m 0

ra H a) ' d 0

CO •

in

<D H A cd H

p*

w ft

u o w w <u u tji <D 04

o CN o

o

o

as CM o

X X

,Q CO in

X o

cd ID r -

X X o

*d VD r -

X o X X

&

H x X X

<Tk rQ O cd O cd V cd

<Tk O 00 o VD H vo 00 o\ r-l as o O in o 00

00 O H CM t> CM in in as

H H 00 u> O ro 00 VD H l CN 1 H

I

A u cd a cd O cd ro 00 ro 0\ O u> o H H CN KD in o H CN i> in H H 0\ r - ro O r - as CM O

H H I1** VD H ro VD CM

<D a) > >

a) •H •H a) > 4-> •U >

CQ •H O U •H 4J <U <U 4J

0 O ft ft O CQ •H a) CQ CQ 0) G 4J ft 0 0 ft 0 cd CQ U CQ •H 3 O JJ 4J 0 4J rH }H <D a) U u cd ft as ft 0) > rH W i—1 rH rH rH M

cd cd cd cd M

JJ o a a a a rH fl •H 0 0 0 0 cd cd g CQ CQ -H -H £

o u 4J 0 ro CQ a Q) <u jd cd •H 0\ d 0 04 ft $3 S5 4J as 0 a <3 <3 < j <1 cd H a pa 55

CN

Ok as A o cd O cd A cd

Ok as 00 ro rH as in o in CN CTk 00 ro O o H CN o l> 00 o rH in CO t> in VD ro 00

H H 00 VD H 00 VD H i CN i H

l

in a\ JQ rQ Cd u id o cd

in a\ rH ro o o CN 00 r -ro as VD 00 in as as as as ro in O O H CO ro H H as

H H 1* 00 rH ro 00 VD H > CM i H

cd & CM

W OJ CO T3 co as (3 OS H

CN as as

U <D

Xi O 4J OJ

E-«

T3 <D V cd

i—I ft <D

0 * n cd S

CQ *d •H CQ

*H >H U

CQ 4J fl a) >

a> w

>. a> O

o> CM O

CM 00

in cm in as rO CO CM CM

o h ro

H O CM

CM

CM

^ H

oo in oo H O CM o in h

^ H

H O O CM cm in

OS

O a\ vo ^ H

in h

oo ro vd as io o co in

O H CM •

a\

ro o cm

in h

a o

•H •U g cd JH

r-H o a) Jh M o u

PS E

a o

r—I *H cd 4->

*H O u a <u 3 W

>1 4J •H o

*H 4J CQ cd TJ

<d >. <u

*H CQ rH 0 Cd Jh g a) *H 4-> O 0) 5 «

a) r—I T?

•U CQ (1) 0) 4J M

TJ a) g rH 0 •H T3 (d <D •U a) I k <u mh a 0 4H

a> t j o S

a o

•H 4J cd ft 4J CQ O O a) Jh ft - P X a) w Pi

0) v H o a) S t j

o a s

U -H a) 4-> ft cd CQ -U o a m a> 4J ft a) x « H

4J •rH O

•H 4J CQ

O CQ (d H dJ TJ

. a) <u 0 ^ ^ v tn to ft a) 0

Q Sh 1 a)

• 4-» t j (u aj

n w - cd

ui d 0 a1 cm

• CQ O I II V -H ft^ >1

U 4J 1 »H

ti rH O -H (d

- a) I H M O 0 «d z

O CQ ^ V 4J H ft CQ

a) ii 1 4J

<d cd cd (d S 8 S

,Q o •H

6 O

o o CQ CQ

•P M-I CQ

<u a) cd cd ft o< S2S Jz;

w « g w ft ft g a

O rH o a cd

• tn a o cd o V -H -H ft T3 V

U

cd <C m

<0 > 0 U S* Qi i f rH <0

-rl U 0 •P 10

-H

-H

I •rl U 04

a o

m 9

rH •s -H >

o *H 4J <0 3

HI (0 > H

0 •H

1 a o o H

u 3 0 fo

0

4J U 0)

H ®

CQ 4J m a> H

0) a

-H CO CO <0

0 o c w •«*

in a)

r l •s B

1 9 0

P$ Pi

w Pi

r—l r—i r—i r—r r—i r—i r—i i—i r—i r—i r—i r—i r—i r—i r—i ,—, r—i r—i o OJ OJ r - r - 00 VD OS ro R* R* (?> H OJ VD H H ro o O O OJ OJ OJ 00 ro o o o r - 00 00 0\ OS Ch o o O o o O H ro ro ro ro ro

0 o H H H H H

U—J u uJ. u-1 i_i I__J I__J u UJ, u-1 L—1 UJ. JL . *. .J, t n rH i n r - Is* 00 U.A OJ a\ as 00 00 H vxi» o ro r - as Ch o VD VD 00 00 00 CN OJ i n LO o o 04

vxi» r - i n H VD VD 00

o H VD rH H OJ H OJ i n r * OJ OJ H o ro OJ o o o H o <N CM as Pn ro i n o o O Pi 00 ro ro ro ro r -

• • • • • • • as as • • • • • • • . . i n ro CO CN OJ H • • H H H H H H H H H

r-oo

0 -H 4J a r—i r—l r—i r—i r~i r—i (U

W ro H H

P W 00 00 [> CQ P< r- (Ti c^ r-0 . • • # . .

•—• L_J t-1 i—< «—J i—J Ui» 00 OJ ID

d) ^0 00 00 H rH & [> ro ro r^ t>

P! 0^ ro ro 00 00 OJ rH P< ro OJ OJ OJ OJ 00 aJ • o o • * o fi i • • . 0 CQ U (U ft c^

00 CQ H 3 CQ

H (D >

En a 0

-rH 4J a5 4J U U •H tin <U a P CQ •H X •H 4J -P •P 4J 4J CQ W 4J CQ CQ CQ CQ CQ 03

cd <U <u (D 0) 0 rd rH 4J Eh Eh IH Eh Eh (X cd 03 I I l l B a S3 Eh hD < 0 0 4-> {2J O CQ CQ a u <D w Q) Eh Pi

r-oo 00

a o

•rl -P O a) P CQ o .p <u

a o

•rH •P (d 4J O a)

8 w

r—i r—» r—i r—i r—i r—. •H 1—1 r—i i—i r—-i r—i r—i o o O r - r - JJ H i n CTi i n i n VD

w o o O o o o (d ro o O ro ro ro

w o o o o o o 4J W o H H H H H P i • • • • • • O Pi « • • • .

—1 u •—J V—J i—J i__« a) t__i i—i i—i i—J i—-i i—i UJ» 00 H VD i n i n P i> OJ r - VD VD as

H 00 X OJ H H i n i n VD P4 00 O 00 o o H w Wl VD OI o as as ro 55 OJ H as ro

rH 55 H vo VD OJ

i n ro OJ CNI f * (d OJ H H H H OJ

r-oo

frl

o -H -P CQ

*H -P CQ CQ CO CQ CQ CQ as <y <u a) <u <u cd • • " • E-» E-* " • -

I IS

tn C

•H 4J 4J 4J JJ i ) CQ

Eh i

& Eh

s

<D Eh

& O U a w

a o

•H •P <d •P O <D P X W

l> 00

a -H -P CQ

•H •P

S1

•H .p -P -P -P -P CQ CQ CQ CQ CQ CQ CQ

rd a> a> <u <u a) <d •u H EH EH EH EH a CQ

4J CQ <1) Eh

i Eh & 3: b

g O o d w

191

§ •H U 0*

d o

to <D

H

•8 •rl fc (0 >

d 0

•rl ii <d 3

H 10 >

W

o •rl §

§ a

M

M 3 0 fa

4-1 0

CQ •P 0 a 0 0

•H «W •P M fd

-P a) U

A <U q X

HI w

Hi •P rH 00 fd a> a

0 •H

tn •P a fd

•rl J2J a CQ CQ (0 3 & CQ 6 1 O 0) 0 > a w a

0 •• r l •H

(0 -P ^ - S > a •a 0 Q) % u p

0» CQ d £ O 0 c V •P

rH Q) (0 &

•rl M rH

in 0) cd 4J a

© (0 0 H •rl •H A a -P 13 •rl fd H 2

r—i O

r--i H

i CM

r-—i m r—i in

r - i o

r—i o

r—i o H

r—i H

p—i H

o O O H H H o o o o O O o o O O O O P4 o o o o O O • • • • • • ft • • • • • •

l -J •—J L_J •——» i—J 1—J l—J 1—J l—J •—-J i—j r* in ro o o ro UJ» o CN in 00 00 vo

o H o o (N UJ»

r- H 00 CN CN H vo os H in in O H 00 vo in in 00 H H o § o in ro ro CN

ro ro ro CN CM VO CN vo vo ro ro H i i i H i i H

r -oo

fa

r*> oo

fa

U •H tn 4J a CQ *H

•H 4J -P 4J •P 4J CQ -P CQ CQ CQ CQ CQ CQ fd <u <D <u a) <D fd •P H EH EH E-» E-« ft W 1 l 1 1 1 £

525 EH £ b !< 0 4J J2J U CQ G <D W E-»

£| o

•H 4J U <N

r—1 r—1 .—, r - i r—i i—i P r—. r—i r—t r—1 r—1 r—1 00 O OI vo vo i> CQ VO H H <Ti OS ro 00 00 00 00 00 0 ro ro in vo vo in in in

4.) g OS 00 00 cr* OS

•—1 l—J l—J l—J 1—-1 UJ a) U—J L—J 1—1 L—J UJ k—1 CN 0 CN ro ro vo p:{ Ui» O OS OS H H OS <N OS OS 00 CN OS CN OS cr O in rH O ro ro in 1—1 CN CN ro VO H H H OS fd ft 00 rH H H vo

1 1 1

in in CN a 0 CQ

VO O

CN CN vo 0

vo 0 O

O

« o

•H -P V <D P CQ o u 4J <U

00

H

fa

U

4J CQ

•H -P fd -P CO

CQ a>

4J 4J -M 4J 4J CQ CQ CQ CQ CQ dJ 0) a) a) <u E-» E^

S5 EH t-D

. rH rH a) a)

1—1 •d 13 rH 0) 0 0 <U

r - TJ s S 00 0 0 H s a C s

0 0 fl •H *H a

H 0 -P 4J 0 •H O U •H 4J a) a) 4J

fa <d ft ft fd 4-> CQ CQ 4J U 0 0 U (U u u a) Qt 4J ft X <D a> X W w

01 a rH •—1 1—i rH

-H fd fd fd fd CQ a C a a CQ 0 0 0 0 fd CQ CQ •H *H & 4J 4J B <D a) fd fd 0 r\ ft ft J2J 55

a w

w # w ft ft g £

192

4J U

a §* CO ©

> -H •P <0 &

0 (0 a o o

*0 a

<o > o N

8 1

£

H (d

-H M a) •u a

•H a •H

§

-H M 04

«W 0

10 <D

5 •U -H

a ©

0 ) o £

M (d a) £

fd fd O u jQ A o a fd ro ro OJ as i> CM in (N in vo 00 o vo H OJ o OJ in VD H X OJ ro o o VO r- ro r- VO Ch VD as

i OJ in i OJ 1 H

ro g vo ° i n 0 1

OJ H as ro OJ r- o 00 H o o II H OJ H in o

• . • s OJ VD H 4J i—l rH CQ

<D a) V

(d

a r^ o o

ro r -

X o

in 5H • 0 •P

in CQ a CQ fd © a) •U rH Sh CQ

.Q D a (0 a) 0 04 u

4-)

U

§• CO

aji >

•H V fd . >

U <D w G o u

PQ

a) ^ rQ

u o — s -P MH

3 0

& rH S 4 ^ 3 3 CO *d

H CQ <D

U <D

A V V

2 04

< J

nd <D

rH -H fd

6 *§

a) rQ A fd TJ fd •Q fd 4J <u VD ro ro OS H as VD ro VD VD

CNI O r- H 00 H 1 4J

OJ VD vp OJ 00 VD OJ as r- VD CNI O as ro ro OJ <D MH •H H OJ m O H as ro o in OS

VD CNI

OJ ro H as o H a 0 l+H O •H

H rH 1 1 00 20

ro i 18

VD H OJ 18

in rH 0 4J to

o v a oj

Q

m fd a) t 3 a) (D M . * t n co

o

rQ rQ fd TJ fd A fd 0) a) X# r^ r^ i n i n VD i n 00 cr> ro OJ

.4

26

o 00 00 ro i n U M

OJ VD OJ o\ as i n VD OJ VD i n <Tl

.4

26

o o H o ro •s fd O o o X H i n OJ H OJ 00 av

.4

26

OJ ro i n H H i n o & OJ l 1 H H i

i 00

21

r o i 1

7

VD H OJ 18

i n H o to i ii v -H

P i A O

I a

u -H

>1 4J -H iH m

<U M H M O o <d 55

O V

•H fd a >1 p4 CQ 4J fd 0 4J a) II O s EH CQ •rH -H i -P <U CQ 73 4J 4J e O g

Oi T5 <u a fd in •H u U CQ u 0) CQ 4J a) 1—1 0 4-J -H 0 0 (d O •H o > O Q) 1X4 01 •- 4J IM U as fd }H a) w Ol -H u fd h to Oi rH O rH 04 4J u rH o o rH

CQ 04 a> rH CQ 0 fd <u o a fd rH TJ <U X N a) fd ' d •H u a 4J M • t n a fd d u fd O M U <u V 0 •H to o fd 0 a fd H a •H u fd rH •H rH 0 V -H *H 0 ro OJ i—i u a) }H to u fd fd u 0 4 ' d CQ 00 00 as 0 i—1 0 -H to •H o g <D o

cr» as as rH • n rH *m rH • n 1 u a }q 4J 1 rH C3 0) H H H fd fd 0 3 fd 0 £ 0) 3 0 a) rH Pu S 04 u S 04 ! 5 Q CO S3 W fd < MH

193

a) as as 3 s* vo

H ^ o <d •- • > CN CN i i i

1 V0 01 en H

i

S o \ i> o \

a -H (If 4J •H U «

© o

. u u a S1

CO

>1 4J H (If Oi

0

09 H Q)

I f

3 <D U (If P i (Q

<D •M (If 4J TO

4J a .Q A a> VO ro

•rl O ro O o -H • •

4-1 H o 4H l 1 0) 0 u

<u ro d H

rH ro IT)

05 • •

> CN CN 4J 4J a A A 0) a\ VO •H vo o o VO o •H • •

tw o o MH 1 1 a) 0 u a) H H

CO rH VO O aJ • •

> CNI CN 4J

4J £» £* a 00 CN <D VO ro •H o O • •

•H ii i H o 4-4 <w

' 1

<u 0 o

o as H r - i n VO CN 00 H rH en ro t> 00 VO i n i n r- CN i n 00 o O 00 i n CN H r^ H VO r-as VO CN H o ro CN CN i n CN H _

' r - VO i n as i n vo O 00 w r- CN t> 0 H 00 00 0 0 H O 00

rd nJ O O rd o A (tf U H H H 0 0 CN O as r** i n i n vo o ro 0 H 1 H

00 CO LO o r* H im vo <T> ro ro CN o VO as i n CN O

o 0 1

o CN o CN i

i n H i n O 1

vo ro o CN as VO as i n CN i n i n ro o CN CO ro vo H

ro rH 00 r - ro H vo ro i n X

o VO CN H o ro rsj CN CN ro H i i ro vo 0 00 i n O r -

H 0 VO 00 H CN 00 CN 00 i n VO 0 O H 00 00

rd rd X1 O rd JQ ni (d cd HI H H 0 O CN H 0 OS 00 00 H as rH ro cr> vo 0 r - 1 1 CN O 00 i n CN O r* vo H 00

CN o i n as 00 rH ro i n 00 X

O 0 1

o CN o OJ 1

i n H vo

i n o O ro O 00 r - ro CN ro o CN i n i n CN H O 0 vo vo VO i n H as vo ro HI r - vo vo

a VO CN H o ro CN CN CN ro H H i 1 1 O ro H i n H i n O as

O 00 VO VO CN H VO CN 00 0 VO 0 O H CN 00

cd rd A O rd rQ rQ (d rd <d O H H H 0 O CN ro 0 i n 00 vo o CN VD r - 0 r - H as 1 1 CN rH o i n rH O ro CTk vo 0 VO CN ro o vo 00 00 H ro 0 0

O o o CN o CN m H m vo 0

rH IM (0 0) U rH 4J 3 U ,Q V 0 rd V 4J -H d 3 0 0 u a) rd -H 4J > > 4J CO U

a) > rd V U

vo rd 0 CQ ^ 4J u *H H

i n co rd V 1 rH <u d rd 4J

Q> rH 0) Q4 rd CO —

hi rd > 0 rH A a CL) rH 0 S 2 df H J W X U E-i fe W < Cfl

a o

•H 4J <d 4J Q a> 04 x m

v -H 6 o d o u w

rd rH d t n 0) Jh M H

to d o

•H 4J O <D

rH W

rH X rd rd d 0 ro r - CN

•H 00 00 as rH 4J as as as rH rd H rH H 0 J25 p<

CN 00 cn

>1 rd S

Jh rd £

rd -H rH 4J M -H rH rH as o

CO 00 ID

r o o vo

vo

4J CQ a)

4J

T* a)

rH -H rd 4J I

<u d o

o V 0<

I

TJ

LO O

0 V Q*

1

o

H a ) a ) v o O

O i v o

u 0 O 4 J rH *—• V rd W N H ro 04

» — U H 04

EH •H •w •w l W a d d

0) T J •H *H 0 0 JQ V a ) CQ 4 J •H •H a u -H CQ V a) rd oi rd rd rd H

rH rH a u u V rH rH O <D 04 •H 0 cu a) <U O Q) <D r>i 4 J u 4 J u ro . Pi O CQ u •H CQ U H O

d 0 w rH 0 0 0 V U a) d rd u 0 0 ro 04 0 0 t j • Q) 0 0 rH M

04

* r i -m cd T J 4 J - U • M S : -' ' 0 IN 1 rd iS d •H 0 CU d d Q a o i S S u P CO 525 « rd rd a o i

rd

194

4J M 0 §• CO

>1 4J M 10 01

a o

4J o a>

M4 a)

Q* -H X! a M

•a <d 0)

H 0)

•d

£ <l>

U fl*

Oi

tJI a

•H

<0 >

i •H H

l >

in ©

H A (0

CO

CQ

CO CO w 04 o w

0 w PS

o H Eh CO H fc

W E-« W Q

s I

Eh

tf O a:

w

Eh CO

1 a

w §

H

3 I w P4 W Q

W H a H fa fa W O a

I w h! §

H

I

LT) LO ^ in n n o ro lo

n I s

H H I

CO O CO OJ h o

O H l>

O H O

fO nJ ns LD rp cn, ro o LO VO OI

o in in

o o o

X 04

• t h u

| § 3 CU CJ CQ

00

™ o

a H E h Cfl H

8 CO

00 in

01

as CN r-ro

in in

ro o

W oi KO OI OI t> 00 ro r* OI in H vo . • » H

O TH

Eh <>«i CO «

ro in

M K

cq x

04 fa

00

01 OI o\ CM

r-in H l

V to a)

4J

T*

a) H •H cd 4J I a) §

o V P<

in o

0 v ft

1

u

0 V Pi

1

JQ

o o

o V 04

195

Figure 2.1 Strength of Party identification, 1964-1992

e V

"53

5 08 ft. Urn © .a

00 ON r-Ui Urn rr t-~-ON

8 ON

r-> OS oo

Os oo Os

Election - Very Strong

- Fairly Strong

-Not Very Strong j

196

Figure 2.2 Distribution of Party Identifiers, 1964- 1992

s v T3

eu o V W>

1964 1966 1970 1974FEB 19740CT 1979

Election

1983 1987 1992 -Conservative i

-Labour j

- Liberal

197

Figure 2.3a Partisan Strength of Conservatives, 1964-1992

MMM

• VERY STRONG

• FAIRLY STRONG

• NOT VERY STRONG

Rgure 2.3b Partisan Strength of Labours, 1964-1992

mm iyiyivi M M

I? 10

Elections

• VERY STRONG

• FAIRLY STRONG

g NOT VERY STRONG

Rgure 2.3c Partisan Strength of Liberals, 1964-1992

Elections

8 8

• VERY STRONG

'•FAIRLY STRONG

' •NOTVERY STRONG!

198

Figure 2.4 Mean Strength of Party Identification by Parties, 1964 1992

C

I CO c © S

3.00

2.80 -

2.60 -

2.40

2.20 --

1.80 -

1.60

1.40

1.20

1.00

Con

Lab

Lib

'Genera!

1964 1966 1970 1974Feb 19740ct 1979 1983 1987 1992

Year

199

Figure 2.5 Mean Strength of Party Identification by Age Cohort, 1987-1992

2.4

22

g> 1.8 £ 55 e Z 1.6

1.4

\2

1 I

1 ! A

! ! I j i ! 1 ' : ! ! i I i / i i ; i ! ; i J / /

! | ^f / '

i 1 ! | ; i

J r ^ ^ ~ 1 ! ! I

V \ i i

!

j ; i : i

1

• M ' :

! i 1 : :

j j j 1

— — js if o t— r- r- oo ao T O r > i r * » G » A 3 0 r n o — r r ! 5

'1987 '

•1992 :

Age groups

200

_ £ ,\ 'A\ \s * V* \/

| l S t

I 15 M O O M

/ 1 = JS s

I

GO _c *-J3

£ <U 3 CO CO

OJ

T J

I

J £

eu

Q C T ^ C Q ^ t O V n ^ C - i C ^ / / "T- / \JV

Support.

*\«9«v

<D 13 o £ *

.1 tt 2 a .

o 4-*

o a* 13 c

.2 CO C <D

£

i <D £

JC H

Tt*

<u u. 3 OD

%Mppmv

201

H«8»"0^V J1

O)

C8 fi ©

* 5

CJ

2

vn <2.

00 . g '•Z3

£ <u 9 CO CO

<4-1

O 13 T3 O

03 C O o

' q <4-4 o

CO §

CO §

e

I <u <D

H

CN

2

tu

Ph 30

Back to Work/Keep Prices Down? Gks waited lyWTFACTCR

80

b

-1.00 .00

Tax Up and Spend Mare? Cases wa^ited by WTFiACTCR

t3

Oh

70

60

50'

40'

30'

20'

10'

-1.00 .00 1.00

IVtoe Effort to Equalise Income?

-1.00 .00 1.00

Unite Fully with ESC? Cte vei ted by W/CTCR (ises weighted by WTF^OOR

-1.00 .00 1.00

Nationalise or Privatise More? Cases weighted by WIF^CTOR

Ph 30

1.00

Left-Right Distribution of All Issues Gses WJ todby WIFACTCR

* -1 represents "No" or "Left/Liberal" answers while +1 "Yes" or "Right/Conservative" answers Figure 4.3 Distribution of Views on Policy Alternatives in Britain, 1992

Unemployment/Inflation European Community 203

Positions of Voter

L n ^ r O ( N H O H ( N r O ^ L n i I I l i

Positions of Voter

Equalization of Income Nationalization

40 ..

u n ^ f O c N j H O H c s j c o ^ i n I i I ! I

Positions of Voter Positions of Voter

Tax/Government Spending General Left/Right

0 tXi «S •M d 0 o u 0) Pi

>

w ro in Positions of Voter

Positions of Voter

•Conservative •Labour »Liberal

Figure 4.4 Party Vote Shares and Positions of Voters on Issues in Great Britain, 1992

Prime Ministerial Approval (%)

204

<D O

s I — -c

#o o <D a* CO 2 13 & 13 c

#o c3 2

<u 13 e -C 3 C3

J3 > a 2 § « > *c <u w S •a .2 H S <3 <u ! • = a. o-

5 §

6 U

I H

^ s

a, 5 •o ~ § .o

<—i '•*-* ca o > £ O Q* C &o a* o a. is <; «

- a! •e c« 3 a W o co • £« c « s z a>

Ut a-T3 cn trl 3 W) £

juspgjsoo XVS^J

Prime Ministerial Approval (%)

205

<D T3 O

a o c3 u <L> Ou X W 13 c 90 cS Z

aS I « & o & u .

• f l

> "C a> -2-S -I

•rs G h S — < « 52 S *§ *n a. a*

<u <U s 1 & a> Q* IS H < ^ -TJ ^ 1

J . I °5 "5 > B 2 w <u g- Q* O- X w C3 C3 *c g 0) O w

.22 cs

.S 2 <D s

o cn *ri a 3 W) £

lirapijjsoo XVSPVd

206

Prime Ministerial Approval (%)

o o CO

WO CN

o n

co CN

CN ON

CN

CN

CN OO . >

oo oo

oo VO oo to oo

oo co oo CN oo

oo

CO 10 ® o

<u <u s •= CO « J3 > a 2 5P &

« .2 > c £ £ S

s ? S I a. a.

<D 4—* <u s § & <u Q ,

ir

< on «

a, ^ s ® c3 o

a *s ca u C3 U > s o & c oo CX O Q- i 3 <; <u

-! OS ctf — •e ^ v 5 w o •25 2 .S 5 0) s "C a. J2 CO to <D Ui 3 &X)

JU3PUJ3oo XVSPVd

Prime Ministerial Approval (%)

207

<U T3 O

a

#o *+-> C3 4-* o <u a* X W c o CO

u 0-

& <d S

§ > H 2

- I

5 -8 i CD

<u +Z

| - i *<« 0J CO a ? •= a. a.

U <D 4—* aj s 1 & cL X

< s S £ 3 § c ca > o 03

CJ « OJ 9* cu Q- * < tu 13 "c3 'C G 8 o CO u» *5 ^ .5 Ou

<u S #c a< cs cn uS 2 3 00 E

JU3I3IJJ300 IVSPVd

208

O > M o o c a o ' £

CO

3

O ca

cs 3 !3 *33 <2

I <*» o -C Ou 2 a S D CO D U Jo

>% o c g 3 U <D

C3 ca 2 CQ 13 u. 5 o 2 w

00 <

CN *r!

3 £P E

VO ON ON

I ON R* ON

209

vo .ON

ON

<N ON

o ON

OO OO

VO OO

TF 00

CM 00

o 00

03 > 0 U4

Cu CX

<

*C 2 CO

G •««*

<D s £ 1 I I I

T : o G* Q*

G CO o >

*W 03 > U <U CO c o

U

#G *53 4m*

*C CQ tS 2 O c C3 >

2 Q-Q*

<

.Is #c <D •W CO

e *N A* T3 G Ctf

^2 3 Gu O

CL W G <U E s <U > O

A

TO

2 3 AP

E

BIBLIOGRAPHY

Abramson, Paul R. 1992. "Notes and Comments: Of Time and Partisan Instability in Britain." British Journal of Political Science 22:381-395.

Alt, James and K. Alec Chrystal. 1983. Political Economics. Berkeley: University of California Press.

Alt, James. 1984. "Dealignment and the Dynamics of Partisanship in Britain." In Paul Allen Beck, Russell Dalton, and Scott Flanagan, eds., Electoral Change in Advanced Industrial Societies. Princeton: Princeton University Press.

Alt, James. 1979. The Politics of Economic Decline. Cambridge: Cambridge University Press.

Alvarez, R. Michael, Shaun Bowler and Jonathan Nagler. 1996. "Issues, Economics and the Dynamics of Multi-Party Elections: The British 1987 General Election." Working paper archived at the Political Methodology Homepage site at http://wizard.ucr.edu/polmeth/working_papers96/alvar96.html/.

American Political Science Association. 1950. "Toward a More Responsible Two-Party System: A Report of the Committee on Political Parties." supplement of American Political Science Review 44.

Bean, Clive and Anthony Mughan. 1989. "Leadership Effects in Parliamentary Elections in Australia and Britain." American Political Science Review 83: 1165-80.

Beck, Nathaniel. 1992. "Comparing Dynamic Specifications: The Case of Presidential Approval." In Political Analysis, vol. 3, James A. Stimson. ed., Ann Arbor: University of Michigan Press.

Beck, Paul Allen. 1974. "A Socialization Theory of Partisan Realignment." In Richard G. Niemi, ed., The Politics of Future Citizens: New Dimensions in the Political Socialization of Children. San Francisco: Jossey-Bass.

1979. "The Electoral Cycle and Patterns of American Politics." British Journal of Political Science 9:129-56.

Benney, M., A.P. Gray, and R.H. Pear. 1956. How People Vote. London: Faber.

210

211

Birch, Anthony H. 1983. The British System of Government. 6th ed. London: George Allen & Unwin.

Black, D. 1948. "On the Rationale of Group Decision-making." Journal of Political Economy 56:23-34.

Box, G.E.P. and G.M. Jenkins. 1970. Time Series Analysis: Forecasting and Control. San Francisco: Holden-Day.

Brown, J.A. 1993. "The Major Effect: Changes in Party Leadership and Party Popularity." Parliamentary Affairs 1993: 549-64.

Budge, Ian and Dennis J. Farlie. 1983. Explaining and Predicting Elections: Issue Effects and Party Strategies in Twenty-Three Democracies. London: George Allen & Unwin.

Budge, Ian, Ivor Crewe and Dennis Farlie.eds. 1976. Party Identification and Beyond: Representatives of Voting and Party Competition. London: John Wiley & Sons.

Butler, David and Dennis Kavanagh. 1992. The British General Election of 1992. New York: St. Martin's.

Butler, David and Dennis Kavanagh. 1980. The British General Election of1979. London: Macmillan.

Butler, David and Donald Stokes. 1969. Political Change in Britain: Forces Shaping Electoral Choice. New York: St. Martin's Press.

Butler, David and Donald Stokes. 1974. Political Change in Britain: The Evolution of Electoral Choice. 2nd ed. London: Macmillan.

Campbell, Angus, Philip E. Converse, Warren E. Miller and Donald E. Stokes. 1960. The American Voter. New York: Wiley.

Chappell, Henry and William Keech. 1985. "A New View of Political Accountability for Economic Performance." American Political Science Review 79:10-27.

Charemza, W.W. and D.F. Deadman. 1992. New Directions in Econometric Practice. Alershot; England: Edward Elgar.

Clarke, Harold D., Euel W. Elliot, William Mishler, Marianne C. Stewart, Paul F. Whiteley and Gary Zuk. 1992. Controversies in Political Economy: Canada, Great Britain, the United States. Boulder, CO: Westview.

212

Clarke, Harold D., William Mishler, and Paul Whiteley. 1990. "Recapturing the Falklands: Models of Conservative Popularity, 1979-83." British Journal of Political Science 20:63-81.

Clarke, Harold D., Marianne C. Stewart. 1985. "Dealignment of Degree: Partisan Change in Great Britain, 1974-1983." Journal of Politics 46: 689-718.

Clarke, Harold D., Marianne C. Stewart and Gary Zuk. 1986. "Politics, Economics and Party Popularity in Britain, 1979-83." Electoral Studies 5:123-141.

Clarke, Harold D., Marianne C. Stewart and Gary Zuk, eds. 1989. Economic Decline and Political Change: Canada, Great Britain, the United States. Pittsburgh: University of Pittsburgh Press.

Clarke, Harold D. and Marianne C. Stewart. 1995a. "Economic Evaluations, Prime Ministerial Approval and Governing Party Support: Rival Models Reconsidered." British Journal of Political Science 24:597-622.

Clarke, Harold D. and Marianne C. Stewart. 1995b. "Economic Evaluations and Election Outcomes: An Analysis of Alternative Forecasting Models." In British Elections and Parties Yearbook 1994, D. Broughton, D. Farrell, D. Denver and C. Railings, eds.. London: Frank Cass.

Clarke, Harold D., Marianne C. Stewart and Paul Whiteley. 1995. "Can the Tories Win Again? The Dynamics of Conservative Support and the Next General Election." Paper presented at the Political Studies Association Annual Conference, University of York.

Converse, Philip E. 1969. "Of Time and Partisan Stability." Comparative Political Studies 2:139-71.

1976. The Dynamics of Party Support. Beverly Hills, CA: Sage.

Crewe, Ivor. 1974. "Do Butler and Stokes really explain political change in Britain?". European Journal of Political Research. 2:47-92.

Crewe, Ivor and David Denver, eds. 1985. Electoral Change in Western Democracies: Patterns and Sources of Electoral Volatility. New York: St. Martin's.

Crewe, Ivor, Pippa Norris, David Denver, and David Broughton, eds. 1991. British Parties and Elections Yearbook 1991. Hemel Hempstead, Hertfordshire: Harvester Wheatsheaf.

213

Crewe, Ivor and Anthony King. 1994. "Did Major Win? Did Kinnock Lose? Leadership Effect in the 1992 Election." In Labour's Last Chance? The 1992 Election and Beyond. Anthony Heath et a/., eds. Aldershot, Hants: Dartmouth.

Cuthbertson, Keith, Stephen G. Hall and Mark P. Taylor. 1990. Applied Econometric Techniques. Ann Arbor: University of Michigan Press.

Davis, Otto A., Melvin J. Hinich, and Peter C. Ordeshook. 1970. "An Exploratory Development of a Mathematical Model of the Electoral Process." American Political Science Review 64: 426-48.

Denver, David and Gordon Hands, eds. 1992. Issues and Controversies in British Electoral Behaviour. Hemel Hempstead, Hertfordshire: Harvester Wheatsheaf.

Denver, D., P. Norris, D. Broughton and C. Railings, eds. 1993. British Elections and Parties Yearbook 1993. Hemel Hempstead, Hertfordshire: Harvester Wheatsheaf.

Denver, David. 1994. Elections and Voting Behaviour in Britain, 2nd ed. Hemel Hempstead, Hertfordshire: Harvester Wheatsheaf.

Dickey, D.A. and C.T. Fuller. 1979. "Distribution of the Estimators for Autoregressive Series with a Unit Root." Journal of the American Statistical Association 74:427-31.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.

Dunleavy, Patrick and Christopher T. Husbands. 1985. British Democracy at the Crossroads: Voting and Party Competition in the 1980s. London: George Allen & Unwin.

Enelow, James M., and Melvin J. Hinich. 1984. The Spatial Theory of Voting: An Introduction. New York: Cambridge University Press.

ed. 1990. Advances of the Spatial Theory of Voting. Cambridge: Cambridge University Press.

Epstein, Leon D. 1980. "What Happened to the British Party Model?" American Political Science Review 74: 9-22.

Eulau, Heinz and Michael S. Lewis-Beck. eds. 1985. Economic Conditions and Electoral Outcomes: The United States and Western Europe. New York: Agathon Press.

Farrell, David M. and Martin Wortmann. 1987. "Party Strategies in the Electoral Market: Political Marketing in West Germany, Britain and Ireland." European Journal of

214

Political Research 15: 297-318.

Fiorina, Marris. 1981. Retrospective Voting in American National Elections. New Haven: Yale University Press.

Fox, John. 1991. Regression Diagnostics: An Introduction. Sage University Paper Series on Quantitative Applications in the Social Sciences 07-079. Newbury Park, CA: Sage.

Franklin, Mark. 1985. The Decline of Class Voting in Britain: Changes in the Basis of Electoral Choice, 1964-1983. Oxford: Clarendon Press.

Franklin, Mark, Tom Mackie, Henry Valen et al. 1992. Electoral Change: Responses to Evolving Social and Attitudinal Structures in Western Countries. Cambridge: Cambridge University Press.

Frey, Bruno and Friedrich Schneider. 1984. "Public Attitudes Toward Inflation and Unemployment and Their Influence on Government Behavior." In T. Willett. ed. Inflation and the Political Business Cycle. San Francisco: Pacific Institute.

Graetz, Brian and Ian McAllister. 1987. "Party Leaders and Election Outcomes in Britain, 1974-83." Comparative Political Studies 19: 484-507.

Harvey, Andrew C. 1989. Forecasting, Structural Time Series Models and the Kalman Filter. New York: Cambridge University Press.

Hausman, J. A. and D. A. Wise. 1978. "A Conditional Probit Model for Qualitative Choice: Discrete Decisions Recognizing Interdependence and Heterogeneous Preferences." Econometrica 46: 403-426.

Heath, Anthony, Roger Jowell and John Curtice. 1985. How Britain Votes. Oxford: Pergamon Press.

Heath, Anthony, Roger Jowell, John Curtice, Geoff Evans, Julia Field and Sharon Witherspoon. 1991. Understanding Political Change: The British Voter 1964-1987. Oxford: Pergamon Press.

Hibbs, Douglas A. 1982. "Economic Outcomes and Political Support for British Government Among Occupational Classes: A Dynamic Analysis." American Political Science Review 76:259-79.

Hotelling, H. 1929. "Stability in Competition." Economic Journal 39:41-57.

Hyman, Richard. 1992. "Trade Unions and Industrial Relations." In Peter Catterall. ed. 1992.

215

Contemporary Britain: An Annual Review 1992: 254-263.

Jenkins, Peter. 1988. Mrs. Thatcher's Revolution: The Ending of the Socialist Era. Cambridge, MA: Harvard University Press.

Johansen, Seren. 1991. "Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models." Econometrica 59:1551-1580,1991.

Jones, Philip and John Hudson. 1996. "The Quality of Political Leadership: A Case Study of John Major." British Journal of Political Science 26: 229-244.

Kavanagh, Dennis. 1987. Thatcherism and British Politics: The End of Consensus ? Oxford: Oxford University Press

King, Anthony, Ivor Crewe, David Denver, Kenneth Newton, Philip Norton, David Sanders and Patrick Seyd. 1993. Britain at the Polls 1992. Chatham, NJ: Chatham House.

Koopman, Siem Jan, Andrew C. Harvey, Jurgen A. Doornik and Neil Shephard. 1995. Stamp 5.0: Structural Time Series Analyser, Modeller and Predictor. London: Chapman & Hall.

Lazarsfeld, P., B. Berelson, H. and Gaudet, 1944. The People"s Choice. New York: Columbia University Press.

Macdonald, Stuart, Ola Listhaug and George Rabinowitz. 1991. "Issues and Party Support in Multiparty Systems." American Political Science Review 85:1107-1131.

McCallum, R. B. and A. Readman. 1947. The British General Election of 1945. Oxford: Publisher.

Miller, William L., Harold D. Clarke, Martin Harrop, Lawrence Leduc and Paul F. Whiteley. 1990. How Voters Change: The 1987 British Election Campaign in Perspective. Oxford: Clarendon Press.

Mills, Terence C. 1990. Time Series Techniques for Economists. Cambridge: Cambridge University Press.

Mizon, G. E. and Richard J. F. 1986. "The Encompassing Principle and its Application to Testing Non-nested Hypothesis." Econometrica 54: 657-678.

Mughan, Anthony. 1993. "Party Leaders and Presidentialism in the 1992 Election: A Post-war Perspective." In British Elections and Parties Yearbook 1993. Denver, D., P. Norris, D. Broughton and C. Railings, eds. Hemel Hempstead, Hertfordshire:

216

Harvester Wheatsheaf.

Nadeau, Richard, Richard G. Niemi and Timothy Amato. 1996. "Prospective and Comparative or Retrospective and Individual? Party Leaders and Party Support in Great Britain." British Journal of Political Science 19:211-236.

Nadeau, Richard and Matthew Mendelsohn. 1994. "Short-Term Popularity Boost Following Leadership Change in Great Britain." Electoral Studies 13: 222-228.

Norpoth, Helmut. 1992. Confidence Regained: Economics, Mrs. Thatcher, and the British Voter. Ann Arbor: University of Michigan Press.

Norpoth, Helmut, Michael S. Lewis-Beck and Jean-Dominique Lafay. 1991. Economics and Politics: The Calculus of Support. Ann Arbor: University of Michigan Press.

Ostrom, C.W. and R.M. Smith. 1990. "Cointegration and Error Correction in Multiple Time Series Analysis: Presidential Approval and the Quality-of-Life Equilibrium Hypothesis." Paper presented at the 1990 Political Methodology Summer Conference, St. Louis, MO: Washington University, July 12-21.

Perron, Pierre and Rene Garcia. 1995. "An Analysis of the Real Intereset Rate under Regime Shift." a Centre interuniversitaire de recherche en analyse des organisations manuscript.

Phillips, P.C.B. 1988. "Time Series Regression with a Unit Root." Econometrica 55: 277-301.

Phillips, P.C.B. "Regression Theory for Near-Integrated Time Series." Econometrica 56: 1021-1044.

Price, Simon and David Sanders. 1993. "Modeling Government Popularity in Postwar Britain: A Methodological Example." American Journal of Political Science 37:317-334.

Pulzer, Peter. 1967. Political Representation and Elections in Britain. London: Allen and Unwin.

Rabinowitz, George and Stuart Elaine Macdonald. 1989. "A Directional Theory of Issue Voting." American Political Science Review 83:93-121.

Ranney, Austin. 1956. The Doctrine of Responsible Party Government. Urbana: University of Illinois Press.

217

Rose, Richard. 1974. Electoral Behavior. New York: Free Press.

Rose, Richard, ed. 1976. Studies in British Politics: A Reader in Political Sociology. New York: St. Martin's.

Rose, Richard and Ian McAllister. 1986. Voters Begin to Choose: From Closed-Class to Open Election in Britain. London: Sage.

Sanders, David, Hugh Ward, and David Marsh (with Tony Fletcher). 1987. "Government Popularity and the Falklands War: A Reassessment." British Journal of Political Science 17:281-313.

Sanders, David, Hugh Ward, and David Marsh. 1990. "A Reply to Clarke, Mishler and Whiteley." British Journal of Political Science. 20: 83-90.

Sanders, David. 1991. "Government Popularity and the Next General Election." Political Quarterly 62:235-261.

1993. "Forecasting the 1992 British General Election Outcome: The Performance of an Economic Model." In David Denver et al. eds. British Elections and Parties Yearbook 1993. Hemel Hempstead, Hertfordshire: Harvester Wheatsheaf.

Sarlvik, Bo and Ivor Crewe. 1983. Decade of Dealignment: The Conservative Victory of 1979 and Electoral Trends in the 1970s. Cambridge: Cambridge University Press.

Sarlvik, Bo, Ivor Crewe and James Alt. 1977. "Partisan Dealignment in Britain, 1964-1974." British Journal of Political Science 7: 129-90.

SPSS, Inc. 1993. SPSS for Windows: Base System User's Guide, Release 6.0. Chicago: SPSS, Inc.

Stewart, Marianne and Harold Clarke. 1992. "The (Unimportance of Party Leaders: Leader Images and Party Choice in the 1987 British Election." Journal of Politics 54:447.

Theil, Henry. 1976. Econometrics. New York: John Wiley & Sons.

Trenaman, Joseph and Denis McQuail. 1961. Television and the Political Image. London: Methuen.

Wallas, Graham. 1910. Human Nature in Politics. London: Constable.

Whiteley, Paul. 1983. The Labour Party in Crisis. New York: Methuen.