c ti p liticomparative politics - faculty of...

C ti P litiComparative Politics

Topic 4Topic 4

Goals of the Lecture

1 Political institutions in comparative perspective1. Political institutions in comparative perspective.

Autocracy vs. Democracy.

2. Electoral rules.

Models of Electoral Accountability with multiple districts: Plurality vs Proportional Models of Electoral Accountability with multiple districts: Plurality vs. Proportional Representation.

3. Regime Type (Form of Government): Presidential vs. Parliamentary Regimes.3. Regime Type (Form of Government): Presidential vs. Parliamentary Regimes.

Comparative Politics and Models of Regime Types.

Political Economy - Trebbi2

Persson and Tabellini Ch. 8, 10

General Overview

So far we have investigated the interaction between economic agents (firms) and theSo far we have investigated the interaction between economic agents (firms) and the government within a mature democratic system. Political economy of rich countries.

Now we start widening our horizon a bit.

We look at different characteristics of political systems and what are the consequences of political institutions (e.g. democracy vs. autocracy).

Note: In future lectures we will also look at economic institutions (e.g. rule of law,

Political Economy - Trebbi3 Political Economy - Trebbi3 Political Economy - Trebbi3

f ( g fexpropriation risk, strength of the courts).

On the Differences Among Political Regimes

1 Democracies are not all alike They differ in a large number of constitutional features1. Democracies are not all alike. They differ in a large number of constitutional features (we have seen some differences already in Topic 1 when we compared Canada and US). We will see what are the implications of these features (e.g. for fiscal policy).

Persson and Tabellini (2003) The Economic Effects of Constitutions.

2. Many countries in the world are not even democracies. Some countries are de iureautocracies (e.g. Qatar), while other countries are democracies just on paper (e.g. North Korea). Until 1991 autocracies ruled the majority of countries (according to POLITY IV scores) and today they still rule the majority of world population.

Either majority does not rule *or* the minority is under the tyranny of the majority.

A l d R bi (2006) E i O i i f Di t t hi d D

Political Economy - Trebbi4 Political Economy - Trebbi4 Political Economy - Trebbi4 Political Economy - Trebbi4

Acemoglu and Robinson (2006) Economic Origins of Dictatorship and Democracy

Polity IV Democracy Time Series

Source: Polity Project Monty G Marshall and Keith Jaggers Principal Investigators George Mason University and Colorado State University; Ted


Source: Polity Project. Monty G. Marshall and Keith Jaggers, Principal Investigators, George Mason University and Colorado State University; Ted Robert Gurr, Founder, University of Maryland.

Polity IV Democracy Time Series


Source: Polity Project. Monty G. Marshall and Keith Jaggers, Principal Investigators, George Mason University and Colorado State University; Ted Robert Gurr, Founder, University of Maryland.

Political Institutions: State Instability and Failure of Authority

Political Economy - Trebbi7Source: Polity Project. Monty G. Marshall and Keith Jaggers, Principal Investigators, George Mason University and Colorado State University; Ted Robert Gurr, Founder, University of Maryland.

Democracy, Autocracy, and Economic Development

Why do we care?

1. Political rights are valuable per se. Sen (1999). Democratic regimes protect them.

2 As economists we are also interested at the role of political institutions in the process of2. As economists we are also interested at the role of political institutions in the process of development.

Democracies tend to be richer than autocracies. All OECD countries are democracies.

But does democracy foster economic development? Does economic development cause political development (the modernization hypothesis of Lipset, 1959, Barro, JPE 1999)?

The following evidence is from Acemoglu, Johnson, Robinson and Yared (Income and


Democracy? AER 2008).

Democracies have higher income per capita levels


But Changes in Democracy do not correlate with Changes in Income


Income does not grow with Democracy.


Existing work does not establish causal effects of income on democracy nor vice versaExisting work does not establish causal effects of income on democracy, nor vice versa.

Acemoglu et al. (AER 2008) show that:

a. Within-country analysis (e.g. country-fixed effects analysis where unobserved country-specific characteristics that are constant over time are controlled for).

b. Instrumental Variable approaches (using past savings rates or predicted income constructs predicted income for each country using a trade share-weighted average income of other countries.)

Both yield no strong result of a role of economic development in fostering political development.


Evidence on the reverse channel of causation is equally scant.


However Acemoglu et al. (AER 2008) find that:

“[…there is a] 500-year relationship between changes in income and democracy[…]. This pattern is consistent with the hypothesis that the positive cross-sectional relationship between income and democracy today is the result of societies embarking on divergent development paths at certain critical junctures during the past 500 years”

Of course this is a difficult proposition to prove because 500 years ago most countries were poor and autocratic, so really the interesting variation in the data remains the cross-sectional one in the 1990s. [xi,1990-xi,1500 = xi,1990-x]

However they present empirical validation of this hypothesis by showing that certain t i ll b k d i diff t d l t th t i l j t 500countries really embarked in a different development path at crucial junctures 500 years

ago (they show correlations with constraints on the authority or religion or population density in 1500s as possible proxies for divergent paths).

Acemoglu and Robinson on Why Nations Fail (2012).


Acemoglu and Robinson on Why Nations Fail (2012).

The Devil is in the Details (1)

Persson and Tabellini (AER P&P 2006): Within-country analysis too blunt a tool.

They report three facts on the interaction between economic development and democratizations employing yearly data on 150 countries (about 120 regime changes) over the period 1960-2000. Add other 120 changes when backdating to 1850’s.

1. Democratizations and economic liberalizations can co-occur, but the sequence of reforms is key.“Countries liberalizing their economy before extending political rights do better.”

See also Giavazzi and Tabellini (JME 2005) and Rodrik and Wacziarg (AER P&P 2005)

“Countries where economic liberalization preceded democracy: South Korea, Taiwan, Chile and Mexico. The opposite sequence took place in countries as Argentina, Brazil, the Philippines and Bangladesh.”

Think about a country that becomes a democracy before property rights are well-defined. Usually, what is going to happen is that the majority will try to expropriate or redistribute part of the wealth with dire consequences on investment decisions and on economic growth


part of the wealth, with dire consequences on investment decisions and on economic growth.


2. Democracies differ.

“Specific democratic institutions influence the fiscal and trade policies implemented after democratization, which may explain why presidential democracy leads to faster growth than parliamentary democracy.”

We will see soon that different constitutional details produce large changes in policy making incentives. Persson and Tabellini (2003).

This affects the incentives of politicians (Aghion, Alesina and Trebbi, QJE 2004) and leads to potential growth-diminishing policies.



3. Expectations of political reforms may foster economic growth even if further democratization attempts faildemocratization attempts fail.

“Taking expectations of regime change into account helps identify a stronger growth effect of democracy.”

Expecting a reform produces effects even before the reform is implemented, as investmentdecisions are forward-looking. Anticipation may attenuate the estimated effect of the change in political institutions.

See Przeworski Alvarez Cheibub and Limongi Democracy and Development: PoliticalSee Przeworski, Alvarez, Cheibub and Limongi. Democracy and Development: Political Institutions and Well-Being in the World 1950-1900.

Political Economy - Trebbi15 Political Economy - Trebbi15 Political Economy - Trebbi15 Political Economy - Trebbi15 Political Economy - Trebbi15

But how do you explain this? Economically Successful Autocracies (above 80th percentile of average growth rates - All world, pooled)

Political Economy - Trebbi16 Political Economy - Trebbi16 Political Economy - Trebbi16 Political Economy - Trebbi16Source: Besley and Kudamatsu (2007).

Policies in Democracies vs. Autocracies: More doubts.

Important questionImportant question.

Numerous scholars point at democratic political institutions as more efficient (Persson and Tabellini 2003, Rodrik and Wacziarg 2005), while others deny any difference., g ), y y

Mulligan, and Sala-i-Martin, Gil (JEP 2004) emphasize how autocracies differ from democracies only in policies restricting political competition and the monopoly of force (military and police spending for instance or censorship and freedom of the press regulation), but not in terms of economic and social policies, once we control for the level of income.

See also Glaeser et al. (JOEG 2004)


Policies in Democracies vs. Autocracies

Political Economy - Trebbi18Source: Mulligan, Sala-i-Martin, Gil (2004).

Democracies and Autocracies do not seem to differ in spending policy.


Democracies and Autocracies do not seem to differ in tax policy [Tax revenues are higher


Democracies and Autocracies do not seem to differ in tax policy. [Tax revenues are higher in autocracies just to make up for the higher military spending in the next table.]


Democracies and A t iAutocracies

differ in policies affecting Public

Office Competition.


Making Autocracy Work

Several arguments put forward of why we observe relative success in autocracies.

Besley and Kudamatsu (2007) show how we can think about the inner working of autocracies.

They start from the empirical regularity that we have discussed so far: Not all autocracies are economic disasters. Some dictatorships are fairly successful economically (Singapore, South Korea between 1960 and 1988, etc.)

How?

It is interesting to see how they succeed. A model of “selectorate”, an elite group to which the leader is accountable and that can “kick him out of office”.


Idea similar to the contestable markets hypothesis from Mulligan and Tsui (2007).

Making Autocracy Work

Besley and Kudamatsu (2007) : Successful autocracies are the ones were the elite and theBesley and Kudamatsu (2007) : Successful autocracies are the ones were the elite and the leader are not too interlinked. The selectorate has to survive without the leader.

The majority of coups actually arises within the elite (i.e. the selectorate).j y p y ( )

Efficient autocracies tend to have higher turnover.

Even single-party systems tend to have political competition. Just of a more subtle form.


How Is Power Shared in Africa? An Introduction

Francois, Rainer and Trebbi (2012). Data fully explained in Rainer and Trebbi (2011).

Growing theoretical literature analyzes the organization & evolution of ruling elites in autocratic & institutionally weak polities.

(e.g. Acemoglu & Robinson 2003, Bueno de Mesquita et al. 2003) ( g g , q )

• But due to paucity of data systematic empirical research is scarce.

In this paper we study how political power is shared among ethnic elites in African countries.

We collect data on the ethnicity of all government ministers between 1960 2004 We collect data on the ethnicity of all government ministers between 1960-2004 in 15 countries & assess the degree of inclusiveness/ proportionality of African governments.

Shed light on the inner workings of autocracies.


Parliamentary democracies as a benchmark

To have a benchmark: Parallel between the ethnic politics in African countries & partisan politics in Parliamentary democracies.

In particular, we compare the outcomes of political bargaining among p , p p g g gAfrican ethnic elites to the outcomes of legislative bargaining in parliamentary democracies.

For parliamentary democracies: Ansolabehere, Snyder, Strauss and Ting (05) [henceforth, ASST]


Main Findings

Much discussion in the literature on how ethnic spoils system works in Africa. Assumption: Winning ethnic group takes it all.

False. Not a One-Man Show.

Proportionality: Ethnic group’s size systematically predicts cabinet seat shares almost 1:1.1:1.

Leadership premia: The Ethnic group of the leader commands a positive seat premium, but it is comparable to formateur premia in Parliamentary democracies.

Linguistic proximity between ethnic groups does not play any role in ministerial allocation.

Coalition Size: Winning coalition size is 80% of the population It is 50% inCoalition Size: Winning coalition size is 80% of the population. It is 50% in Parliamentary democracies. (i.e. Rare that an ethnic group with a large base goes unrepresented within an African Cabinet).


Main Implications

1. African Political failures are evident. We are not challenging such evidence (i.e. no l i Af i i i l i th S d )claim Africa is more inclusive than Sweden).

2. But much discussion (& evidence) about political failure in Africa & the role of ethnic fractionalization.

3. How does ethnicity matter for political failure in Africa? Is it between-ethnicities conflict or within-ethnicity conflict that matters for political failure?

4. It’s not Haves vs. Have-Not’s: Too few groups are left out from sharing the pie for this to be the story. Our indirect evidence does not support a pure “between-ethnicities friction” explanation.

5. Then, the failure must also come from within-group frictions (i.e. within each group, the represented elite vs. the unrepresented rank & file).


Table 7: Inclusiveness in Parliamentary Democracies and Africa. Population

Country

Share of Voters Not Represented in

Government, 1960-2004 Average Country

Share of EthnicitiesNot Represented in

Government, 1960-2004 Average

Australia 53 23 Benin 28 23Australia 53.23 Benin 28.23Austria 39.48 Cameroon 18.35 Belgium 41.05 Cote d'Ivoire 13.93 Denmark 59.34 Dem. Rep. of Congo 28.16 Finland 40 38 Gabon 13 72Finland 40.38 Gabon 13.72Germany 45.67 Ghana 29.84 Iceland 41.54 Guinea 7.54 Ireland 52.99 Kenya 8.93 Italy 49 08 Liberia 50 37Italy 49.08 Liberia 50.37Luxembourg 41.19 Nigeria 12.12 Netherlands 42.51 Rep. of Congo 11.12 Norway 60.39 Sierra Leone 15.92 Portugal 64 90 Tanzania 42 87Portugal 64.90 Tanzania 42.87Sweden 56.43 Togo 31.94 Uganda 27.59 Mean 49 16

22 71Mean 49.16 22.71

T-stat for Difference in Means = 6.48; degrees of freedom = 27; p-value= 0.0000


Inclusiveness in Parliamentary Democracies and Africa

As expected, the size of government coalitions in parliamentary democracies is around 50 percent.

But in African countries it is surprisingly high – close to 80 percent!

Thus, African governments are surprisingly inclusive and very few ethnic groups are ever left outside.

Furthermore: i) the levels of inclusiveness in African countries are quite stable over time; ii) do not differ much between periods of autocracy & democracydemocracy.


Figure 2: Population Share of Ethnicities Not Represented in Government, African Sample, 1960-2004

5010

0Benin Cameroon CongoKinshasa Cote d'Ivoire

Gov

ernm

ent

050

100

Gabon Ghana Guinea Kenya

esen

ted

in G

050

100

Liberia Nigeria Republic of Congo Sierra Leone

s N

ot R

epre

05

00

1960m1 1980m1 2000m1

Tanzania Togo Uganda

of E

thni

citie

050

1

1960m1 1980m1 2000m1 1960m1 1980m1 2000m1 1960m1 1980m1 2000m1

Pop

. Sha

re

YearP YearGraphs by Country


Figure 1: Share of Voters Not Represented in Government, Parliamentary Democracies, 1960-2004

5010

0Australia Austria Belgium Denmark

rnm

ent

050

100

Finland Germany Iceland Ireland

ted

in G

over

050

100

Italy Luxembourg Netherlands Norway

ot R

epre

sent

05

100

1960m1 1980m1 2000m1 1960m1 1980m1 2000m1

Portugal Sweden

of V

oter

s N

o0

501

1960m1 1980m1 2000m1 1960m1 1980m1 2000m1

Shar

e o

YearYearGraphs by Country


Table 9a: Level of Inclusiveness in Cabinet Seats Allocation and Level of Democracy in Africa, 1960-2004. Within-Country Evidence.y , y

Pop. Share of Not Represented

Pop. Share of Not Represented Ethnicities

(E cl ding PolRepresentedEthnicities

(1)

(Excluding Pol. Transitions)

(2) Polity2 -0.0215 -0.0013y [0.1771] [0.1671] Cabinet Duration 0.0553 0.0623 [0.1128] [0.1353] Leader Duration -0.0399 0.0457 [0.0922] [0.1106] R2 0.80 0.83 N 637 588

The share of the sample for which Polity2 ≥ 0 is 13.66%.All specifications include country and year fixed effects and a linear trend by country. Cabinet and leader duration indicate years since the leader was installed and years from the last occurrence of a simultaneous replacement of more than half the cabinet members, respectively. St d d l t d t th t l l i b k tStandard errors clustered at the country level in brackets.* p<0.05; ** p<0.01


Country-Level Analysis: Proportionality

F Af i iFor African countries: g = ethnic group, pg/P = g’s share of population, yg = g’s share of government posts.

For Parliamentary democracies: g = political party, pg/P = share of seats in parliament, yg = share of government

posts.p

Definition 1: A perfectly proportional government is one for which pg/P = yg, "g.

Many ways of defining disproportionality (Gallagher, 91).


Table 8: Disproportionality in Parliamentary Democracies and Africa.

Country LSq Disproportionality

1960-2004 Average Country LSq Disproportionality

1960-2004 Average Australia 40.13 Benin 16.58 A t i 26 61 C 11 59Austria 26.61 Cameroon 11.59Belgium 28.71 Cote d'Ivoire 13.47 Denmark 41.93 Dem. Rep. Congo 12.97 Finland 27.90 Gabon 15.64 Germany 35.55 Ghana 16.38Iceland 31.44 Guinea 16.59 Ireland 40.87 Kenya 10.94 Italy 32.72 Liberia 37.99 Luxembourg 24.10 Nigeria 14.22 Netherlands 24.09 Rep. of Congo 19.61 Norway 41.40 Sierra Leone 17.02 Portugal 40.33 Tanzania 16.06gSweden 39.79 Togo 17.43 Uganda 14.25 Mean 33.97

16.72Mean 33.97 16.72

T-stat for Difference in Means = 7.07; degrees of freedom = 27; p-value = 0.0000


Figure 5: Least Squares Disproportionality in Cabinet Allocation, African Sample, 1960-2004

3060

Benin Cameroon CongoKinshasa Cote d'Ivoire0

3060

Gabon Ghana Guinea Kenya

ality

03

60

Liberia Nigeria Republic of Congo Sierra Leone

spro

porti

ona

030

0

1960 1980 2000

Tanzania Togo Uganda

LSq

Di

030

60

1960 1980 2000 1960 1980 2000 1960 1980 2000

YearGraphs by Country


Figure 4: Least Squares Disproportionality in Cabinet Allocation, Parliamentary Democracies, 1960-2004

3060

Australia Austria Belgium Denmark0

3060

Finland Germany Iceland Ireland

nalit

y0

30

60

Italy Luxembourg Netherlands Norway

Dis

prop

ortio

n0

300

1960 1980 2000 1960 1980 2000

Portugal Sweden

LSq

D0

3060

1960 1980 2000 1960 1980 2000

YearYearGraphs by Country


Proportionality in Parliamentary Democracies & Africa

African countries appear significantly more proportional than parliamentary democracies (largely because they are more inclusive).

High degree of proportionality in Africa suggests that the size of ethnic groups is g g p p y gg g pan important predictor of their share of seats in the government (more on this later).

As with inclusiveness, the levels of proportionality in African countries are quite stable over time and do not differ much between periods of autocracy and democracy. y


Table 9b: Level of Disproportionality in Cabinet Seats Allocation and Level of Democracy in Africa 1960-2004 Within-Country EvidenceDemocracy in Africa, 1960-2004. Within-Country Evidence.

All Posts

Least Squares Disproportionality

(1)

LH Disproportionality

(2)

Rae Disproportionality

(3)

Sainte-Laguë Disproportionality

(4)Posts (1) (2) (3) (4)Polity2 -0.0409 -0.1240 -0.0148 -1.2646 [0.0545] [0.1136] [0.0136] [1.2288] Cabinet 0 0412 0 0347 0 0092 1 8428Cabinet 0.0412 0.0347 0.0092 1.8428Duration [0.1105] [0.1079] [0.0167] [2.8227] Leader -0.0363 -0.0555 -0.0086 -0.7221 Duration [0 0645] [0 0890] [0 0138] [0 8858]Duration [0.0645] [0.0890] [0.0138] [0.8858] R2 0.75 0.77 0.80 0.81 N 637 637 637 637

The share of the sample for which Polity2 ≥ 0 is 13.66%. All specifications include country and year fixed effects and a linear trend by country. Cabinet and leader duration indicate years since the leader was installed and years from the last occurrence of a simultaneous replacement of more than half the cabinet members, respectively. Standard errors clustered at the country level in brackets. * p<0.05; ** p<0.01


Preliminary Conclusions on Power in African Dictatorships

Overall, our results reject the idea that African autocracies are run as “one-man shows” by a single leader and his ethnic groupman shows by a single leader and his ethnic group.

Instead, they suggest that African autocrats are often involved in strategic bargaining with different ethnic elites.g g

Our findings also indicate that interethnic rivalry for cabinet positions may be less intense than commonly thought. y g

Poor accountability of the ministers to their own ethnic constituencies may potentially be more of a problem.


Comparative Constitutional Analysis in Democracies

We now zoom in on the inner workings of democraciesWe now zoom in on the inner workings of democracies.

One reason is that they are easier to understand than autocracies.

We will start from an analysis of the main channel of political representation: elections.

Crucial to the definition of electoral incentives of politicians is the electoral ruleCrucial to the definition of electoral incentives of politicians is the electoral rule.

We will then move to the analysis of the executive, and in particular we will focus on the constraints on the executive as determined by the form of government of the country.constraints on the executive as determined by the form of government of the country.


Electoral Rules

It is the procedure through which voters’ preferences are translated into political representation. p

In a representative democracy elected politicians form the legislature and influence or determine policy.

Political scientists tend to focus on the effects of electoral rules on political outcomes. Example: the Party structure. Duverger’s Law states that plurality electoral rules tend to induce a polarized party system (i.e. a two-party system, like the UK). Fujiwara (2012).

Political economists tend to focus on the effects of electoral rules on economic outcomes. E l Fi l P li P d T b lli i (2003 2004) h h PR t hExample: Fiscal Policy. Persson and Tabellini (2003, 2004) show how PR systems have larger governments and more spending than plurality systems. Milesi-Ferretti, Perottiand Rostagno (QJE 2002) show that geographic transfers are higher in plurality systems. Persson, Tabellini and Trebbi (2003) show how PR systems tend to have


y ( ) yhigher levels of political rents.

World Distribution of Electoral Systems


Source: The International Institute for Democracy and Electoral Assistance (International IDEA) 2005. Lower chamber or single chamber rule.

Classifications (source: IDEA)

List Proportional Representation (List PR)Under a List Proportional Representation (List PR) system each party or grouping presents a list of candidates for a multi-member electoral district, the voters vote for a party, and parties receive seats in proportion to their overall share of the vote In some (closed list) systems the winning candidatesin proportion to their overall share of the vote. In some (closed list) systems the winning candidates are taken from the lists in order of their position on the lists. If the lists are ‘open’ or ‘free’ the voters can influence the order of the candidates by marking individual preferences.Examples: Brazil, Finland, Italy, Netherlands, Israel.

First Past The Post (FPTP)First Past The Post is the simplest form of plurality/majority electoral system. The winning candidate is the one who gains more votes than any other candidate, even if this is not an absolute majority of valid votes. The system uses single-member districts and the voters vote for candidates rather than political partiespolitical parties.Examples: Canada, India, UK, USA.

Two-Round System (TRS)The Two-Round System is a plurality/majority system in which a second election is held if noThe Two-Round System is a plurality/majority system in which a second election is held if no candidate or party achieves a given level of votes, most commonly an absolute majority (50 per cent plus one), in the first election round. A Two-Round System may take a majority-plurality form–more than two candidates contest the second round and the one wins the highest number of votes in the second round is elected, regardless of whether they have won an absolute majority–or a majority run-off form–only the top two candidates in the first round contest the second round


off form only the top two candidates in the first round contest the second round. Examples: Egypt, France.

Classifications (source: IDEA)

Parallel SystemsA Parallel System is a mixed system in which the choices expressed by the voters are used to elect representatives through two different systems–one List PR system and (usually) one plurality/majority system–but where no account is taken of the seats allocated under the first systemplurality/majority system but where no account is taken of the seats allocated under the first system in calculating the results in the second system.Examples: Japan, South Korea, Pakistan.

No Direct Elections (N)No Direct Elections (N) Examples: China, Saudi Arabia.

Block Vote (BV)Block Vote is a plurality/majority system used in multi-member districts Electors have as many votesBlock Vote is a plurality/majority system used in multi member districts. Electors have as many votes as there are candidates to be elected. The candidates with the highest vote totals win the seats. Usually voters vote for candidates rather than parties and in most systems may use as many, or as few, of their votes as they wish.Examples: Lebanon, Mauritius

Mixed Member Proportional System (MMP)Mixed Member Proportional is a mixed system in which the choices expressed by the voters are used to elect representatives through two different systems–one List PR system and (usually) one plurality/majority system–where the List PR system compensates for the disproportionality in the


plurality/majority system where the List PR system compensates for the disproportionality in the results from the plurality/majority system.Examples: Germany, Mexico, New Zealand.

Electoral System Families


Source: The International Institute for Democracy and Electoral Assistance (International IDEA)

Electoral rules are not constant over time

Example 1: New Zealand changed its electoral rule from First-Past-The-Post to a mixed system including proportional representation (PR) elements.

Example 2: In 1994 Japan moved from a Single Non-Transferable Voting (SNTV) system to a Parallel system with both plurality and PR elements. It was first employed in the 1996 electionselections.

Example 3: In 1993 Italy changed its electoral rule from pure list PR to a parallel system with 75% of the seats allocated by plurality rule and 25% by PR. In 2005 it reverted y p y yback to pure PR right before the elections (Incumbent government was fearing larger losses under the FPTP than under PR). See Persson and Tabellini (2004).

Example 4: Strategic Manipulation of electoral rules. Alesina and Glaeser (2004) discuss how PR was abandoned in the post-war US in order to exclude socialists and communists. Southern US Cities immediately after the 1965 Voting Rights Act moved from Single-District elections to At-Large electoral rules in order to disenfranchise the

Political Economy - Trebbi45 Political Economy - Trebbi4545

from Single-District elections to At-Large electoral rules in order to disenfranchise the newly registered African-American voters (Trebbi, Aghion, and Alesina, QJE 2008).

Systematic Features

The design of electoral rules involves some systematic features.

An important feature of an electoral rule is representation. An electoral rule has to translate voters’ preferences into an elected body of representatives which mirrors them as closely as possible.

Second, accountability is key. An electoral rule has to make politicians accountable to voters.

See Handbook of Political Economy chapter by Persson and Tabellini.


Systematic Features

Are representation and accountability in conflict?Are representation and accountability in conflict?

Think of a single district electing one politician by majority rule.

Assume candidate 1 is voted by 51% of the population (constituency A).

Who’s going to represent the remaining 49% (constituency B)?Who s going to represent the remaining 49% (constituency B)?

Nobody.

On the other hand, an incumbent elected by such a narrow margin will be very responsive to voters. If he or she disappoints 1% of constituency A (and at the next election they vote for the other candidate to punish you), candidate 1 is out.


p y )

Systematic Features

Now make it a multi-member district with PR 100 membersNow make it a multi member district with PR. 100 members.

Each constituency (A and B) is going to elect different politicians (51 and 49 respectively). So representation of B is higher than under plurality (before was 0). p g p y ( )

But who’s going to discipline the politicians? If you disappoint 1% of your group, you are just going to lose only 1 member, not the whole lot as under plurality rule. [If you follow the parallelism you will notice that under majority rule it would be equivalent to losing all 51 members].

Note: If you are not convinced, here’s another way to look at the difference in representation between a plurality system and proportional representation: Under plurality rule a party can control the legislature with just 25% of the popular vote (50% of votes in 50% of the districts). Under PR a party can control the legislature only with 50% of the


50% of the districts). Under PR a party can control the legislature only with 50% of the popular vote.

Systematic Features

Think harder Is the trade off between representation and accountability always so obvious?Think harder. Is the trade off between representation and accountability always so obvious?

What happens to accountability in a plurality system when a politician knows that a majority of voters is going to be voting for her anyway?g g g y y

See Myerson (1993). Bandwagon effect of plurality rule. Suppose a politician is a “bundle of policies”, some of them welfare-diminishing (e.g. the politician is corrupt). You may knowingly tolerate corrupt politicians in order to have some policy favorable to you g y p p p y yimplemented for sure (say, you care about ideology).

“The idea is that voters vote strategically, and may vote for the dishonest but ideologically preferred candidate if they expect all other voters with the same ideology to do the same. S it hi t th h t did t i k i i th i t t did t th th id fSwitching to the honest candidate risks giving the victory to a candidate on the other side of the ideological scale.” This is not a problem under PR where majority premia are small.

Example: Ideological voters in Tuscany (60 years voting for the same leftist party).


Example: Ideological voters in Tuscany (60 years voting for the same leftist party).

Additional (Important) Features

Electoral Formula: It translates votes into seats. Within PR, Mixed, and FPTP different formulae determine specific families of electoral

rules (e g plurality differentiates into FPTP TRS AV etc ) Sometimes this term alsorules (e.g. plurality differentiates into FPTP, TRS, AV, etc.). Sometimes this term also refers to other fine details within an electoral rule (D'Hondt method, Sainte-Lague method for allocation in PR etc.).

District Magnitude: It determines how many seats a political geographic unit (i.e. the district) controls.

Example: In the US 1 representative is elected from 1 congressional district for the House (district magnitude is 1). However, for the Senate magnitude is 2 (2 Senators from each State). In Israel 120 representatives are elected from 1 district (which is the country itself). Lower district magnitude implies more individual accountability.

Ballot Structure: How citizens cast their preferences. Say, they can pick their candidate out of a list (open list) or can just select the party list ranking (closed list).

All these features do not necessarily co-vary in blocks We lose at lot of information when


All these features do not necessarily co vary in blocks. We lose at lot of information when discussing an electoral rule as a bundle without addressing the details.

Evidence on Accountability

Since politicians tend to respond more to individual incentives than to collective incentives (Holmström ,1982, incentives in teams), systems that have party-list PR tend also to ( , , ), y p yhave politicians without much discipline.

In Persson, Tabellini and Trebbi (2003) we show that in cross-section countries with party-list ballots tend to have higher political corruption and rent seeking levels (controlling for a large set of country characteristics, including GDP per Capita, colonial and legal origin).

Estimates imply that plurality rule could reduce political corruption, by making politicians directly accountable to their constituency, by as much as 20%. (This is a large effect: Twice the coefficient of being a Latin American Country) .Twice the coefficient of being a Latin American Country) .

Persson et al. (2003) also shows that countries with large district magnitude (many representatives per district) tend to have higher political corruption and rent seeking


p p ) g p p glevels

Evidence on Representation

When asked about the 2009 emergency bailout of financial institutions, several US congressmen opposing the bill cited as a reason the fact that bill did not have an g pp gimmediate beneficial effect on their constituency. See Mian, Sufi, and Trebbi (2009).

This should not come as a surprise, since FPTP systems like the US tend to pull in the direction of narrowly targeted programs affecting very precise geographic regions (the congressional districts). There are many examples of government spending programs producing local benefits at diffused costs (e.g. Farm bill, Indian Gambling and Gaming Provisions)Provisions).

Such programs are what we usually call pork-barrel programs (spending programs motivated by electoral motives). Bickers and Stein (1995).by electoral motives). Bickers and Stein (1995).

Also, in plurality systems parties have an incentive in targeting swing districts and pivotal districts and not those where the party is a sure winner.


p y

Evidence on Representation (2)

On the other hand multimember districts and PR systems tend to rely on broader groups ofOn the other hand, multimember districts and PR systems tend to rely on broader groups of voters.

Consider the case of the Netherlands with a single multi-member PR district. g

There is no particular reason why a Dutch party should target voters in a region relative to another if all it matters is the national vote share.

This is why PR systems will tend to prefer universalistic government programs (such as social security, pension programs or welfare programs) as opposed to localized targeted redistribution.

We will see theoretical models of this in the following lectures (Persson and Tabellini, 2000)


2000).

Evidence on Representation (3)

Plurality systems should distort their fiscal policy towards more targeted programs and PRPlurality systems should distort their fiscal policy towards more targeted programs and PR systems towards more universalistic programs.

This is last prediction is verified by Milesi-Ferretti, Perotti, and Rostagno. (2002) and p y , , g ( )Persson and Tabellini (2003). PR systems spend on average 2-3% of GDP more in social security and welfare than plurality systems.

The fact that in PR systems you have to please more voters than the pivotal set implies that government spending should be generally larger in PR systems relative to plurality systems.

Lizzeri and Persico (2001) also show that plurality rule may induce a systematic under-provision of public goods as opposed to localized pork-barrel projects.


Additional Effects of Electoral Rules on Representation

So far I have focused on differences in terms of fiscal policy, but there are also obvious effects of electoral rules on political outcomes.p

Weak parties in plurality. Localized incentives tend to weaken parties vis-à-vis the individual representative because of often the politician’s narrow interest will conflict with the interest of the majority of the other members of the party.

Plurality rules tend to under-represent small parties and converge towards polarized party systems. This is because you need overcome large thresholds in at least one district small parties tend to disappear. Think about the green party and the third party in the US. Except if they have a very strong geographic component, (e.g. Bloc Quebecois). This over time reduces the number of parties much more than under PR (DuvergerThis over time reduces the number of parties much more than under PR (Duverger 1954, Lijpahart 1990). Taagepera “Predicting Party Sizes: The Logic of Simple Electoral Systems” (2007).

PR tends to produce a proliferation of parties (in fact, certain formulae require thresholds for


representation to limit such fragmentation.)

Additional Effects on Representation

A corollary of this is that often PR systems will require coalitionsA corollary of this is that often PR systems will require coalitions.

Intra-bargaining within coalitions and common-pool problems usually lead to overspending.

Particularly, the bargaining within a coalition will induce overspending if each party is a veto player (say because it can destabilize the coalition). Veto players will be able to extract rents in the forms of inefficient programs. For an analysis of veto players see Tsebelis (2002).

Overspending usually drives up budget deficits. Alesina and Perotti (1995) discuss the evidence on how PR systems and coalition government tend to be more likely to run larger deficits.

Oth id f K t l d P tti (1999) d ith li ti US


Other evidence comes from Kontopoulos and Perotti (1999) and, with an application US cities spending, Baqir (JPE 2002).

Form of Government

We will now briefly discuss another important institutional feature. Form of government.


Time Series (Democracies only)

Political Economy - Trebbi58Source: Joseph Colomer.

Form of Government

Broadly, speaking we will restrict ourselves to two forms of government: Presidential versus Parliamentary Systems.

The main differences will revolve around the type of constraints faced by the executive branch.

This is the really fundamental difference across political regimes.


Electoral Constraints of the President

An presidential regime is characterized by a directly and popularly elected presidentAn presidential regime is characterized by a directly and popularly elected president.

Hence under a presidential regime it is possible to decouple the electoral constraints of the executive from those of the legislative.g

Furthermore, no stable majority in the legislative is needed to support the executive.

Instead within a parliamentary regime the executive is expression of the legislative. The majority party (say, under Westminster-type regimes) or a coalition of parties selects a prime minister in charge of policymaking.

In a parliamentary regime the executive is accountable to voters only indirectly, through the legislative.


Accountability through Motion of Confidence

The lack of direct electoral accountability in parliamentary regimes requires an alternative form of checks and balances: The confidence (nonconfidence) motion.

A prime minister is subject to maintaining the confidence (majority of support) of the legislative chambers.

The executive loses power if the legislative loses power.

Thi h t iThis has two main consequences:

1. Makes the electoral term endogenous to the policy (decreasing the insulation of the executive under parliamentary regimes)executive under parliamentary regimes).

2. Allows for policy adjustment during the legislature but also increases legislative cohesion (motivated by the incentives not to lose valuable agenda-setting power by thecohesion (motivated by the incentives not to lose valuable agenda setting power by the governing coalition).


Trade-offs in the Form of Government

Another typical characteristic of presidential regimes is strength of legislative proposal and the veto power over legislation. In general, presidential regimes have stronger executives. Shugart and Carey (1992).

This induces a strong trade-off between ability of presidential government to implement reform through insulation from the legislative and reduction of accountability.

Alesina and Rosenthal (1995) show how the US midterm elections may play a role in reducing such constraints through changes in Congressional strength (the divided

t h th i )government hypothesis).


Stability the Form of Government

There is a heated debate in the comparative politics literature on the stability of form ofThere is a heated debate in the comparative politics literature on the stability of form of governments. Particularly, the fragility of presidential regimes to autocratic rule.

Lijphart (1999); Linz (1978); Linz and Stepan (1978); Shugart and Carey (1992) all discuss jp ( ); ( ); p ( ); g y ( )the virtues of parliamentary regimes in terms of higher stability.

It is the winner-takes-all component of presidentialism that most likely makes it prone to be the form of government of choice of autocratic regimes. Also the strong concentration of legislative powers in the hands of the executive branch. Parliamentary systems more consensual by design.

Strong evidence in West Sub-Saharan Africa starting from the 1970’s: Changes towards strong presidentialism from hybrid systems (like France: a president and a prime minister, sometimes of different parties) and parallel curbing of political rights. Also


minister, sometimes of different parties) and parallel curbing of political rights. Also think of the evolution of Russia’s semi-presidential system.

Form of Government in Local Governments

Parliamentarism is also a feature of sub-national political systemsParliamentarism is also a feature of sub national political systems.

For instance cities are often managed by a council-manager form of government as opposed to a strong mayor system. g y y

This institution was developed first in the US South during the progressive era.

In a council-manager form of government the council hires a city manager than then implements the policy and serves at the will of the council.

Strong mayors are instead directly elected by the voters in parallel to the city council.


Some Theoretical Formalization


Electoral Accountability and Electoral Rules

Persson and Tabellini (2000, ch. 9.2)

Consider a political system with one incumbent politician and N identically sized groups of voters, indexed by J.

V i J ili f i J d l bli dVoters in group J get utility from consumption c J and a general public good g

U J = c J + H(g)

where H has standard properties (H’>0, H”<0). Let us assume that voters consume all disposable income so:

UJ = y – τ + f J + H(g) (1)

Wh i i i di t t d f J d t ti l t f t


Where y is income, τ indicates taxes, and f J denotes a nonnegative lump sum transfer to members of group J.

The Politician: Public goods and Rents

As in Topic 1 the politician (i.e. the government) employs tax revenues to produce the public good, but can also appropriate part of the revenues as private rents r.

The production of public good g entails a cost of transforming private goods into public goods of θ, a random variable whose realization is common knowledge.

Then, the government budget constraint is:

θg = Nτ – r - f (2)

where f = J f J .



Notice that we are enriching our setting.

We are introducing potential conflict among voters over redistributive transfers f J .

In a second I will also introduce details on the electoral rule and check what are the consequences in terms of political accountability and redistribution.


The Politician: Public goods and Rents (cont.)

As in Topic 1 the politician enjoys (exogenous) rents from being in office R. The politician can obtain R only if elected, which happens with reelection probability p.

Preferences of politicians are then given by:

γr + pR (3)

where γ < 1 reflects the fact that politicians face some transaction cost in extracting private rents.


Timing of the game

Sequential structure:

1. State of θ is realized.

2. Voters pick a retrospective voting rule p

3. Incumbent chooses the policy vector [{f J}, g, r, τ]

4. Elections are held.


Recall: Our solution without transfers (Topic 1)

Voters coordinate on the same retrospective voting strategy:p = 1 if U(g(θ),r(θ)) > k(θ) (4)

0 th i0 otherwisewhere k(θ) indicates the voter’s reservation utility.

At t (3) th liti i ith t l thi tti * 0 * d t * NAt stage (3) the politician either steals everything, setting g* = 0, τ* = y and rents r* = Ny, orpleases the voters and earns reelection. The incumbent does so at the minimum cost, so it provides them with their reservation utility k(θ) to obtain p = 1.

The politician will be pleasing the voters if her utility is higher doing so γr+R ≥ γNyand voters will require the above condition to hold with equality r* = Max[0, Ny - R/γ]

Voters will optimally set their reservation utility tok*(θ) = y - (θg*(θ)+r*)/N + H(g*(θ))

where the cost of producing the optimal amount of public good is θg*(θ) and taxes are


where the cost of producing the optimal amount of public good is θg (θ) and taxes are τ*=(θg*(θ)+r*)/N.

Case 1: Single-District Elections

The incumbent runs for office against an identical opponent.

Proportional representation in a single-district, two party system.

Needs only a minimum winning coalitions of N/2 groups of voters to win.

Equilibrium breaks down immediately.

Any intuition why?y y



Suppose the incumbent is facing a decision rule like (4).

She will play the groups of voters one against the other and get them to set their reservation utility so low, she basically fully expropriates all voters.

Suppose in fact at stage (3) the politician sets τ = y and sets positive {f J} for N/2 voters so to maintain a majority satisfied at k*(θ).

This is feasible and the politician can increase rents this way. Proof: For the lucky majority, f J + H(g*(θ)) = k*(θ) = y - (θg*(θ)+r*)/N + H(g*(θ)), which makes

t t l t N f J N/2 θ *(θ)new total rents Ny - f J N/2 - θg*(θ).New rents are always larger than r*, under our original assumption that to begin with r* left

enough revenue for optimal public good level g(θ) in every state θ, θg*(θ) ≤ R/γ. [Actually the politician could even underprovide the public good (below g*) get even larger


[Actually, the politician could even underprovide the public good (below g ), get even larger rents and still make it. So this is a fortiori valid]


But now notice that everybody is taxed, while transfers reach only a lucky (minimum winning) majority.

This strategy cannot be an equilibrium for those left out of the minimum winning coalition.

They will bid down their reservation utility below k*(θ) in order to be included in the minimum winning coalition.

A race to the bottom.



In fact, the new equilibrium of this game requires an additional condition.

If group J chooses reservation utility k*J(θ) , then it must be a best response to k*I(θ) for all I∫ J.

Groups play Nash among each other (while voters still cooperate within the group).



The new equilibrium must satisfy:

1. Voters must not be so demanding the incumbent is better off foregoing reappointment in office.

2. The equilibrium policy must be optimal for the incumbent and she will please only a majority of the voters.

3. No group of voters can benefit from an unilateral change in its reservation utility, given the reservation utility of the other groups.


Case 1 Equilibrium

The new equilibrium must satisfy for at least a minimum winning coalition N/2

y – τ + f J + H(g) ¥ k*J(θ)

This makes new total rents for the incumbent equal to tax revenues less government spending

N τ - f - θg(θ),

which she will maximize for {f J}, τ and g

subject to τ § y, 0§ f, y – τ + f J + H(g) ¥ k*J(θ) for N/2 voters.


Solution: Taxes

In order to maximize rents, the incumbent will fully expropriate voters.

Taxes will be at the highest possible level in equilibrium. τ ** = y

To see why consider that voters are pitted against each other and cannot block total expropriation.

Since taxes and transfers are perfect substitutes in voters’ preferences, the incumbent can always increase taxes on everybody by D τ and redistribute to only N/2 groups through transfers.

Thi th t l d d t ill bid d hi ti tilit b ti hi hThis means that any excluded voter will bid down his reservation utility by accepting higher and higher taxes in order to be included in the minimum winning coalition.


Solution: Public Good

What is the level of g that the incumbent will choose?

Will she be able to set g = 0? [Note: Given the incumbent receives no utility from g, 0 is what she would pick, were she

unconstrained by inequality (4)].

Let us start from our socially optimal level of public good provision. The optimal level of gis pin down by the Samuelson criterion NH′(g)=θ (i.e. the sum of the marginal utilities f t h ld b l t th i l i l t f *)of agents should be equal to the social marginal cost of g*)



Suppose the incumbent deviates and lowers the public good level by a tiny Dg.

She would release an amount of revenues equal to θDg.

She would also reduce utility of all voters by H(g) - H(g - Dg) = Dgÿ H′(g) = Dgÿ θ/N.

Compensating N/2 voters for their loss will cost the incumbent N/2 ÿ Dg ÿ θ/N = θDg/2p g g gwhich leaves additional rents θDg/2 > 0.

So the incumbent will start lowering g below the optimal level for sure. But when is she


going to stop?


The marginal utility of g is decreasing (H is concave), so it becomes increasingly costly to compensate voters for their loss of public good.

Suppose now the supply of g has decreased all the way down to the point where NH′(g)=2θ .

S h i b l h bli d l l i b i DSuppose the incumbent lowers the public good level again by a tiny Dg.

She would release an amount of revenues equal to θDg.

She would also reduce utility of all voters by H(g) - H(g - Dg) = Dg ÿ H′(g) = Dg ÿ 2θ/N.

i / f h i l ill h i b / θ/ θ hi hCompensating N/2 voters for their loss will cost the incumbent N/2 ÿ Dg ÿ 2θ/N = θDg which leaves 0 additional rents.

S th i b t ill t l b l thi i t b ti t ill b


So the incumbent will not lower g below this point because compensating voters will became too expensive below this point!


Such deviations will be no longer profitable.

The new equilibrium level of g** is at the level satisfying:

NH′(g)=2θ .


Solution: Transfers

Since all voters must play best response, the voters receiving transfers must not be better off than those receiving some transfer.

In fact, they all must receive the same reservation utility k*J(θ). But since all receive utility from the public good and are taxed equally, then transfer cannot be different depending on J being in the majority or not.

In equilibrium f J = 0 for all J.

This makes new total rents for the incumbent equal to tax revenues less government spending r**= N y θg**


spending r = N y - θg

Case 1: Equilibrium Discussion

Notice that the voters (any half of them) can still kick the politician out of office so she cannot just set g = 0.

This is interesting because even if transfers have played voters one against the other, yet the indivisibility of the public good allows voters to set their reservation utility contingent on a measure of aggregate performance g.

S t bilit i b t it i l th i th h di t ib ti t fSome accountability remains, but it is lower than in the case where redistributive transfers were not allowed.

Notice that in equilibrium transfers are not employed but the very treat of using themNotice that in equilibrium transfers are not employed, but the very treat of using them strongly decreases accountability of the incumbent vis-à-vis voters.


Case 1: What would happen if X voters could coordinate?

So far we have assumed groups of voters could not coordinate.

I will now briefly extend the model to allow for a subset of the groups to form a coordinated group of size X.

You can alternatively consider this an extension with groups of heterogeneous sizes.

We will consider two possibilities:

a. X > N/2b. X < N/2

We will see how group size and political power will relate to each other.


Persson and Tabellini (2000, ch. 9.5 – problem 3)

X > N/2 voters coordinate

A candidate needs at least half the votes, so to win the incumbent needs the support of group X > N/2. Group X is necessary for reelection and not fungible.

Since group X can coordinate, it can demand its maximum payoff given the incumbent’s participation constraint.

Recall that the incumbent’s participation constraint was found by considering that the politician has two strategies at stage (3): St l thi tti * 0 * d t * N P liti i ’ tilit i thia. Steal everything setting g* = 0, τ* = y and rents r* = Ny. Politician’s utility in this case will be γNy

b. Please the voters and earn reelection.

The politician will be pleasing group X if her utility (3) is higher doing so γr+R ≥ γNyTo minimize rents paid to the incumbent, for any θ, group X will require r* = Max[0, Ny -

R/γ]


R/γ]

X > N/2 voters coordinate

Given r* group X will maximize its utility:

max{J œX f J + XH(g)}

subject to θg = Ny – r* – J œX f J

which is solved for [{f J}, g] using FOC:

XH′(g*(θ))=θ

f J(θ)= (Ny – r* – θg*(θ))/X for all JœX


X < N/2 voters coordinate

A candidate needs at least half the votes, but now to win the incumbent does not need the support of group X < N/2 necessarily. Group X isn’t necessary for reelection and is fungible with any other group summing up to N/2.

Even if group X can coordinate, the competition with other groups of voters will bring us back to a situation where the incumbent can pit groups against each other.

Case 1 results will apply to this case and taxes will be the maximum, transfers will be zero, and public goods will be underprovidedand public goods will be underprovided.


Case 1: What would happen if X voters could coordinate?

The political power of an electorally pivotal group X > N/2 is such that the incumbent is kept to her minimal rents.

However, in this model there is a sharp discontinuity when we consider the political power of an electorally nonpivotal group X < N/2

In the latter case the incumbent extracts her maximal rents!


Case 2: Multiple-District Elections

The incumbent runs for office against an identical opponent.

M multiple districts with plurality rule in each district, two party system.

Each district is identical. In each district all groups are represented in amount N/M . So total number of groups is still N.

In order to control the legislative are necessary M/2 districts to win.


Case 2: Example

J = 1 J = 2

N = 3

J = 1 J = 3 J = 1 J = 2

SINGLE- DISTRICT PR

J = 3J = 1 J = 2 J = 1 J = 3 J = 1 J = 2J = 3

M = 2

MULTIPLE- DISTRICT FPTP

D = 1 D = 2


Case 2: Example

SINGLE DISTRICT PR: WHAT MATTERS ARE AT LARGE SHARES

J = 1 J = 2 J = 3

SINGLE- DISTRICT PR: WHAT MATTERS ARE AT-LARGE SHARES

MULTIPLE- DISTRICT FPTP: WHAT MATTERS ARE SHARES WITHIN DISTRICTS

D = 2D = 1



In order to control the legislative, M/2 districts are necessary.

But to win a district is necessary to convince only ½*N/M voters.

This implies that the total number of voters necessary to win the election is no longer N/2

The total number of voters necessary to win is M/2* N/2M = N/4.



This clearly does not bode well for the voter groups, that now are going to be pitted against each other with an even worst constraint since each group is split across M different districts.

The new equilibrium must satisfy for at least a minimum winning coalition N/4

y – τ + f J + H(g) ¥ k*J(θ)

What changes relative to Case 1? Public good provision will be even lower!

Let’s show it.



Suppose the supply of g is at the point where NH′(g)=2θ . This pinned down the equilibrium amount of g in Case 1.

Suppose the incumbent lowers the public good level again by a tiny Dg.

Sh ld l f l θDShe would release an amount of revenues equal to θDg.

She would also reduce utility of all voters by H(g) - H(g - Dg) = Dg ÿ H′(g) = Dg ÿ 2θ/N.

Compensating N/4 voters for their loss will cost the incumbent N/4 ÿ Dg ÿ 2θ/N = θDg/2which leaves θDg/2 >0 additional rents.

So the incumbent will now have a profitable deviation relative to Case 1.

Th ilib i t f i C 2 ill b i b NH′( ) 4θ


The new equilibrium amount of g in Case 2 will be given by NH′(g)=4θ.

Solution: Taxes and Rents

The equilibrium solution is identical to Case 1.

Since the new equilibrium amount of g *** is given by NH′(g)=4θ, this implies a lower level f bli d h d C 1 i H′( ) i d i (H( ) i )of public good than under Case 1 since H′(g) is decreasing (H(g) is concave).

Gi h h i i f ll i ( ) i id i l hGiven that the incentives to fully expropriate voters (τ = y ) remain identical to the proportional representation case, this also implies higher rents for the politician (r = N y - θg ***).


Electoral Competition and Electoral Rules

Persson and Tabellini (2000, ch. 8.1-8.4)

So far we have consider the issue of electoral accountability and how electoral rules under a single-district (PR) versus multiple-district plurality electoral rule can change the rents of politicians and the provision of public goodsof politicians and the provision of public goods.

We now consider the issue of how two opportunistic and rent seeking political candidatesWe now consider the issue of how two opportunistic and rent-seeking political candidates will interact and compete with each other.


Electoral Competition and Electoral Rules

Consider a political system with two competing candidates (A and B) running for election d N 3 id ti ll i d f t i d d b Jand N=3 identically sized groups of voters, indexed by J.

Voters in group J get utility from government policy in the form of consumption c J and a general public good ggeneral public good g

u J = c J + H(g)

where H has standard properties (H’>0, H”<0). Let us assume that voters consume all disposable income so:

uJ = 1 – τ + f J + H(g) (5)

where y = 1 is income τ indicates taxes and f J denotes a nonnegative lump sum transfer to


where y = 1 is income, τ indicates taxes, and f denotes a nonnegative lump sum transfer to members of group J.


As in our analysis of electoral accountability, politicians (i.e. the government) can employ tax revenues to produce the public good, but can also appropriate part of the revenues as private rents r.

The production of public good g entails a cost of transforming private goods into public goods equal to 1. (q = 1. No uncertainty about it.)

Th t b d t t i t iThe government budget constraint is:

g = 3τ – r - f (6)

where f = J f J .


Timing of the game


1. The two candidates A and B commit to policy platforms qA = [{fAJ}, gA, rA, τA] and

qB = [{fBJ}, gB, rB, τB] respectively, conditional on ex ante electoral preferences. They

act simultaneously and do not cooperate.

2. Elections are held.

3. The winning policy vector is implemented.


The Candidate

As in Topic 1 the politician enjoys exogenous rents from being in office R and endogenous rents r.

In this model a candidate can obtain R and r only if elected, which happens with (endogenous) election probability p.

Preferences of politicians are then given by the expected value of victory:

E(v) = p(γr + R) (7)

where γ < 1 reflects the fact that politicians face some transaction cost in extracting private rents.


Electing a Candidate

We assume that the election outcome is uncertain at the moment of deciding about policy (there is electoral uncertainty at stage 1) Uncertainty about voters’ preferences(there is electoral uncertainty at stage 1). Uncertainty about voters preferences.

In equation (5) we have assumed that voters are identical with respect to preferences for policy. In order to introduce uncertainty we add that voters are heterogeneous withpolicy. In order to introduce uncertainty we add that voters are heterogeneous with respect to preferences for politicians for ideological reasons.

Define U J(q) as the indirect utility obtained by replacing equation (6) into (5). (q) y y p g q ( ) ( )

Voter i in group J votes for candidate A if:

U J(qA) > U J(qB) + (d + sJ,i) (8)

where (d + sJ,i)§ 0 or ¥0 indicates the voter’s ideological preference for candidate B.


( ) g pNotice that it depends on a common value d and an idiosyncratic component sJ,i .


We assume that the common value d has a uniform distribution on [-1/(2z), 1/(2z)].Notice that the higher z, the lower the variance of d .

Uncertainty about d resolves at stage 2, right before elections are held.

Think about d as an aggregate popularity shock (some last-minute electoral scandal, like Spain on 3/11/2003).

Th idi i J i i diff f h J 1 2 3 d i if hThe idiosyncratic component sJ,i is different for each group J=1,2,3 and is uniform over the interval [-1/(2wJ)+ sJ, 1/(2wJ)+ sJ]. Candidates know these group-specific distributions when they pick policies.

Notice that the group-specific mean s J indicates the average ideology of the group.

Also groups differ in their ideological homogeneity The higher wJ the lower the group-


Also groups differ in their ideological homogeneity. The higher w , the lower the group-specific variance of sJ,i.


Assumption 1: Groups are ranked based on their average ideology s J :1 < 2 < 3s 1 < s 2 < s 3

Let s 2 = 0 as a normalization.

Assumption 2: Assume group 2 has the highest density (it’s the most homogenous):w2 > w1 , w3

Finally, for analytical convenience (but negligible for the gist of our results) assume also that s 1w1 + s 3w3 = 0


Ideological preferences for B

w2

w1

More neutral votersô

w3

sJ,is 1 s 30Lik l t f A Lik l t f B


òLikely to favor A Likely to favor Bô

Note: Probabilistic Voting

We have assumed probabilistic voting.

In a unidimensional policy space Downs’ (1957) traditional political competition modelIn a unidimensional policy space Downs (1957) traditional political competition model shows that two candidates (who can, by assumption, commit to specific platforms) converge to the same platform and both candidates select the policy preferred by the pivotal voter.

In a multidimensional policy space if no policy dominates any other policy, cycling can occur and we can end up with no equilibria (i.e. given some policy choice by the adversary a candidate can always rearrange some policy dimensions to capture aadversary, a candidate can always rearrange some policy dimensions to capture a winning coalition of voters and win). The function linking policy choice and electoral results is (very) discontinuous.

d id hi di i i i f h did i i bIn order to avoid this discontinuity some uncertainty from the candidates’ viewpoint about the mapping from policy choice to electoral results is added. Probabilistic voting indicates a class of models characterized by uncertainty about electoral outcomes. Electoral support become a smooth function of the policy platform and Nash equilibria


ecto a suppo t beco e a s oot u ct o o t e po cy p at o a d Nas equ b ausually exist.

Swing Voters for Candidate B in the Three Districts

w2

U 1(qA) - U 1(qB) - d = s 1,i

w1

U (qA) U (qB) d s

U 2(qA) - U 2(qB) - d = s 2,i

U 3(qA) - U 3(qB) - d = s 3,i

From eq. (8)ôYou can identify who the swing voters in J are.

w3

sJ,is 1 s 30


Who votes for A?

w2

U 1(qA) - U 1(qB) - d = - d = s 1,i

Assume both candidates present the same policy platform qA= qB.

w1

U (qA) U (qB) d d s

U 2(qA) - U 2(qB) - d = - d = s 2,i

U 3(qA) - U 3(qB) - d = - d = s 3,i

All voters in J=2 vote Bô

w3

All i 3

sJ,is 1 s 30

òAll voters in J=3 vote BOnly voters in J=1 with sJ,iôbelow -d vote A.


- d

What happens if the average ideological preference for B declines?

w2

U 1(qA) - U 1(qB) - d = - d = s 1,i

Assume both candidates present the same policy platform qA= qB.

Only voters in J=2 with s2,iôbelow d’ vote A

w1

U (qA) U (qB) d d s

U 2(qA) - U 2(qB) - d = - d = s 2,i

U 3(qA) - U 3(qB) - d = - d = s 3,i

below -d vote A.

w3

ò Only voters in J=3 with s3,i below -d’ vote A.

sJ,is 1 s 30

Only voters in J=1 with s1,iôbelow -d’ vote A.

ò Only voters in J 3 with s below d vote A.


- d 'More people vote for A.

Candidate’s Strategies

Assume candidates A and B start from the same policy platforms.

If candidate A decides to offer a policy with higher public good provision g she will increase her electoral prospects in J=1,2,3.

All swing voters move right.

This is a symmetric effect across all districts for any realization of d.y y

U 1(q’A) - U 1(qB) - d > - d = s 1,i

U 2(q’A) - U 2(qB) - d > - d = s 2,i

U 3(q’A) - U 3(qB) - d > - d = s 3,i


Q: What happens if candidate A decides to lower her rents r? What if she raises t?

Candidate’s Strategies

Assume candidates A and B start from the same policy platforms.

If candidate A decides to offer a policy with higher targeted transfers to group 1 at the expense of group 3 she will increase her electoral prospects in J=1, while she will decrease her support in J=3.

Swing voter in 1 moves right, swing voter in 3 moves left.

This is an asymmetric effect across districts for any realization of d.

U 1(q”A) - U 1(qB) - d > - d = s 1,i

U 2(qA) - U 2(qB) - d = - d = s 2,i

U 3(q”A) - U 3(qB) - d < - d = s 3,i


Who votes for A?

Vote share of candidate/party A in district J is:

p J,A = wJ [(UJ(qA) - U J(qB) - d) - (-1/(2wJ)+ sJ )]

1/2+ J [U J( ) U J( ) d J ] (9)= 1/2+wJ [U J(qA) - U J(qB) - d - sJ ] (9)

Obviously the vote share of candidate/party B in district J is 1- p J ,A



Let’s now check the consequences of different electoral rules.

Proportional representation in a single-district, two party system.

Seats allocated in perfect proportion of votes.

Needs only a minimum winning coalitions of 1/2 voters to win 1/2 seats and the right to set the policy vector q.

Q: What is the probability that candidate A sets the policy under single-district PR?



The probability that candidate A is elected and sets the policy is:

p A = Pr[1/3*J J,A¥1/2]

And using (9) and the uniformity assumption on we get:

A / * { / J J( ) J( ) J } /p A = Pr[1/3*J {1/2+wJ [U J(qA) - U J(qB) - d - sJ ]} ¥1/2]

= Pr[1/3*J {wJ [U J(qA) - U J(qB) - sJ ]} ¥ wd]

where w = 1/3*J wJ and p A, recalling that J wJsJ = 0, further simplifies to:

p A = 1/2 + z/(3w)*J {wJ [U J(qA) - U J(qB)]} (10)


Case 1: Equilibrium

The solution of this problem entails realizing that the problems for candidate A and B are symmetric.

p B =1 - p A = 1/2 + z/(3w)*J {wJ [U J(qB) - U J(qA)]}

This is the specific result of the assumption J wJsJ = 0 . That assumption is theThis is the specific result of the assumption J w s 0 . That assumption is the combination of group 2 being symmetric around 0 and that group 1 and 3 symmetrically balancing each other. To see this last point consider that if s 1w1 + s 3w3

= 0 the number of voters in group 1 below zero w1 [0 - (-1/(2w1)+ s1 )] is the same as h b f i 3 b 3[(1/(2 3) 3 ) 0]the number of voters in group 3 above zero w3[(1/(2w3)+ s3 )-0].

This symmetry is thus not general, but depends on specific assumptions for the three groups positionspositions.

Nonetheless, in this symmetric case it will not be a surprise then that the two candidates choose the same policy in equilibrium


choose the same policy in equilibrium.

Case 1: Equilibrium

We can find the equilibrium policy by considering the policy maximizing the expected value f i tof victory:

E(vA) = p A(γr + R) (7)

with respect to qA, taking qB as given, and subject to:uJ = 1 – τ + f J + H(g) J=1,2,3 (5)

g = 3τ – r – J f J (6)

p A = 1/2 + z/(3w)*J {wJ [U J(qA) - U J(qB)]} (10)τ § 1τ § 1


Case 1: Equilibrium

By substituting in equations (5), (6), and (10)

E(vA) =

(1/2 + z/(3w)*J {wJ [1 – (f A1+f A

2 +f A3 +gA+rA)/3 + f AJ + H(gA) - U J(qB)]})*(γrA + R)

with respect to [{f J} g r τ ] and taking q as givenwith respect to [{fA }, gA, rA, τA] and taking qB as given.

Finally, the first order conditions will have to be evaluated at the point where qA = qB.


Case 1: FOC

The first order conditions with respect to government spending and rents are:

∑ E(vA)/∑ g A = ∑ pA/∑ g A *(γrA + R) = z/3w*J {wJ [ – 1/3 + H’(gA)]}*(γrA + R) = 0 ( ) g A p g A (γ A ) J { [ (gA)]} (γ A )

∑ E(vA)/∑ rA = ∑ pA/∑ rA *(γrA + R)+ pA*γ = z/3w*(- w)*(γrA + R) + γpA = 0 ( ) A p A (γ A ) pA γ ( ) (γ A ) γpA


Case 1: FOC

The first order conditions with respect to transfers are:∑ E(vA)/∑ f A

1 = ∑ pA/∑ f A1 *(γrA + R) = z/3w*(w1 - w)*(γrA + R) = 0

∑ E(vA)/∑ f A2 = ∑ pA/∑ f A

2 *(γrA + R) = z/3w*(w2 - w)*(γrA + R) = 0

∑ E(vA)/∑ f A3 = ∑ pA/∑ f A

3 *(γrA + R) = z/3w*(w3 - w)*(γrA + R) = 0 ( ) f A p f A (γ A ) ( ) (γ A )

Notice that these three first order conditions cannot hold with equality, inducing corner solutions. In particular the net marginal gain from transferring money to group 2 will

always be positive and the highest of the three since w2 – 1/3*J wJ = w2 - w > 0 (recall Assumption 2).

This implies: ( ) 1 d ll f i b i d i f 2 ( h i ) h(a) 1 dollar of revenues is better invested in a transfer to group 2 (the swing group) than to

any other group, hence f A2 >0, f A

1 = f A3 = 0;

(b) Since w2 > w it also implies that 1 dollar of revenues has a marginal electoral benefit (i.e. redistributing transfers to group 2) larger than its marginal electoral cost (i e increasing


redistributing transfers to group 2) larger than its marginal electoral cost (i.e. increasing taxes on all). This means τA = 1.

Case 1: Equilibrium Conditions

How do we find the optimal levels of public goods and rents?

Equalize the net marginal gain from one extra unit of public good (increases the probability of winning) to the net marginal gain from transferring money to group 2(increases the probability of winning).

Equalize the net marginal gain from one extra unit of rents (decreases probability of winning but increases gains from being in office) to the net marginal gain from transferring money to group 2 (increases the probability of winning).y g p ( p y f g)

∑ E(vA)/∑ g A = ∑ E(vA)/∑ f A2 fl J {wJ H’(gA)} = w2 (11)

∑ E(vA)/∑ r A = ∑ E(vA)/∑ f A2 fl γpA = z/3w*(w2)*(γrA + R) (12)



Equalizing the net marginal gain from one extra unit of public good to the net marginal gain f t f i t 2 i th ilib i t f bli dfrom transferring money to group 2 gives you the equilibrium amount of public good provided g*

A. Simplify equation (11) to see it:

H’(gA) = w2/3w > 1/3

The last inequality implies that the public good is underprovided relative to the socially l l l d d b 3H’( ) 1optimal level determined by 3H’(g) = 1.



Equalizing the net marginal gain from one extra unit of rents to the net marginal gain from t f i t 2 ( ti 12) itransferring money to group 2 (equation 12) gives:

γ(1/2 + z/(3w)*J {wJ [1–(f A2+gA+rA)/3 + f AJ + H(gA)-U J(qB)]})= z/3w*(w2)*(γrA + R)

which together with the government budget constraint (6) and g*A delivers fA

2* and r*A

Even easier: You can plug in the equilibrium condition pA =pB =1/2 into eq. (12) and check that rents can be positive in equilibrium:

γ*1/2 = z/3w*(w2)*(γrA + R)with rA ¥ 0.


Case 1 Discussion

A Nash equilibrium is determined by equal policy choice by both candidates.

Hence we have fully characterized the political equilibrium.

Result 1: Political competition does not necessarily reduce political rents to 0 This isResult 1: Political competition does not necessarily reduce political rents to 0. This is because the two candidates are not perfect substitutes and preferences about who is in power are idiosyncratic (recall there’s a component of utility that comes from ideology).

Result 2: The larger the density of swing voters w2, the lower the level of equilibrium rents r*

A.

Result 3: The higher the variance in electoral results (i.e. the lower z), the higher the level of equilibrium rents r*

A, since the expected vote share becomes less sensitive to policy when there is more electoral uncertainty (so candidates will bias the policy in their


when there is more electoral uncertainty (so candidates will bias the policy in their favor).

Case 2: Multiple-District (FPTP) Elections

Again we will compare the results under a single-district PR system with a multiple-district FPTP system.

3 (one seat) electoral districts with plurality rule in each district, two party system.

Each district is identical and coincides with each group J (Persson and Tabellini (2000) also deal with the case of less than perfect overlap).

In order to control the legislative 2 districts are necessary.

Think about a parliamentary regime like the UK with two main parties running in each district.


Case 2: Multiple-District (FPTP) Elections

Under this electoral rule the existence of equilibrium is not guaranteed without further assumptions.

Indeed you could have cycles where candidate A courts any two districts at the expense of the remaining one. Given that strategy, the opponent could deviate and buy either one of the two districts supporting A plus the district left out under A’s platform.

Y l thi li bl b ki di t i t 1 d 3 f h f 2 0You solve this cycling problem by making districts 1 and 3 far away enough from s 2 = 0, so that it is not convenient for a candidate that is really ideologically disliked in those districts (candidate B in district 1 and candidate A in district 3) to pay for their support.

Assumption 3: The ideological biases at the extremes are large, s 1 << 0 << s 3 .

Under Assumption 3 the equilibrium is such that A and B announce equal policies and all the


Under Assumption 3 the equilibrium is such that A and B announce equal policies and all the competition takes place in district 2 only.


The probability that candidate A is elected is:

p A = Pr[ 2,A¥1/2]

And using (9) and the uniformity assumption on we get:

p A = Pr[{1/2+w2 [U 2(qA) - U 2(qB) - d - s2 ]} ¥1/2]A B

= Pr[{w2 [U 2(qA) - U 2(qB) - s2 ]} ¥ w2d]

p A further simplifies to:p pp A = 1/2 + z*[U 2(qA) - U 2(qB)]} (13)

All hinges on what happens in district 2.


g pp

Case 2: Equilibrium

As for Case 1 the solution of this problem entails realizing that the problems for candidate Aand B are symmetric.

We can find the equilibrium policy by considering the policy maximizing the expected value of victory:

E(vA) = p A(γr + R) (7)

with respect to q taking q as given and subject to:with respect to qA, taking qB as given, and subject to:uJ = 1 – τ + f J + H(g) J=1,2,3 (5)

g = 3τ – r – J f J (6)p A = 1/2 + z*[U 2(qA) - U 2(qB)] (13)

τ § 1



You can follow exactly the same steps as for Case 1.

Equalize the net marginal gain from one extra unit of public good (increases the probability of winning) to the net marginal gain from transferring money to group 2(increases the probability of winning).

Equalize the net marginal gain from one extra unit of rents (decreases probability of winning but increases gains from being in office) to the net marginal gain from transferring money to group 2 (increases the probability of winning).y g p ( p y f g)

∑ E(vA)/∑ g A = ∑ E(vA)/∑ f A2 fl w2 H’(gA) = w2 (14)

∑ E(vA)/∑ r A = ∑ E(vA)/∑ f A2 fl γpA = z*(γrA + R) (15)



Equalizing the net marginal gain from one extra unit of public good to the net marginal gain f t f i t 2 i th ilib i t f bli dfrom transferring money to group 2 gives you the equilibrium amount of public good provided g**

A. Simplify equation (14) to see it:

H’(gA) = 1 > w2/3w > 1/3

Result: The first inequality implies that the public good is underprovided relative to the l (C 1) l l d d b H’( ) 2/3proportional representation (Case 1) level determined by H’(g) = w2/3w.

The last inequality implies that the public good is underprovided relative to the socially optimal level determined by H’( ) 1/3optimal level determined by H (g) = 1/3.



Equalizing the net marginal gain from one extra unit of rents to the net marginal gain from t f i t 2 ( ti 15) itransferring money to group 2 (equation 15) gives:

γ(1/2 + z*[1–(f A2+gA+rA)/3 + f A2 + H(gA)-U 2(qB)]= z*(γrA + R)

which together with the government budget constraint (6) and g**A delivers fA

2** and r**A

E Y l h l b d 1/2 d h k hEven easier: You can plug in the equilibrium condition pA =pB =1/2 and check that rents can be positive in equilibrium:

*1/2 *( + R)γ*1/2 = z*(γrA + R)with rA ¥ 0.

N ti th t hi h t k did t l t t hi h t i FPTP l ti


Notice that higher rents make candidates lose votes at a higher rate in FPTP elections.


Compare the Case 2 (multi-district FPTP) condition for optimal rents:γ*1/2 = z*(γrA + R)

with the Case 1 (single-district PR) condition:

γ*1/2 = z*w2/3w*(γrA + R)

Higher rents make candidates lose votes at a higher rate in FPTP elections.

The equilibrium level of rents are lower with multiple-district FPTP than with single-district PR.

l l i i i iff b i h iElectoral competition is stiffer because it concentrates on the most responsive voters.


Discussion of Electoral Competition

Plurality elections concentrate electoral competition in key marginal districts and inducePlurality elections concentrate electoral competition in key marginal districts and induce more targeted redistribution (you can show that transfers to district 2 are higher –hint: use the budget constraint) and lower provision of public goods than proportional representation single-district systems.

Since voters in marginal districts also can be more responsive to economic benefits, then electoral competition is stronger in majoritarian systems and rents are lower.


More Empirical Evidence on Electoral Rules

Persson Tabellini and Trebbi (2003): Lower political rents/corruption in FPTP regimesPersson, Tabellini, and Trebbi (2003): Lower political rents/corruption in FPTP regimes. Kunikova and Rose-Ackerman (2005) show that closed list PR are particularly detrimental in terms of corruption (this seems also related to career concerns of politicians, a topic we have not addressed explicitly).

Milesi-Ferretti, Perotti and Rostagno (2002): More redistributive programs and higher government spending/GDP in PR systems in a panel of OECD countries.

Baqir (1999) however finds no relation between size of US municipal government and single-district (at-large) vs. multiple-district (ward) electoral rules.


Form of Government

Persson and Tabellini (2000) ch 10Persson and Tabellini (2000) ch.10

We now shift gear and focus on types of political regimes.

We will focus in particular on the behavior of the legislature under different regimes.

I will first set up a benchmark “Simple Legislature” modelI will first set up a benchmark Simple Legislature model.

Then I will twist the agenda-setting components of the model in order to decouple “ways and means” (i.e. taxes) vs. “appropriations” (i.e. spending) in the budgetary process.and means (i.e. taxes) vs. appropriations (i.e. spending) in the budgetary process.

Analogies in presidential-congressional & parliamentary regimes.


Simple Legislature

Consider a political system with N=3 identically sized groups (of unit mass) of voters indexed by J.

Each group J coincides with a district. Each district is represented by an incumbent legislator l=1,2,3.

Voters in group J get utility from government policy in the form of consumption c J and a general public good g

u J = c J + H(g)u c + H(g)

where H has standard properties (H’>0, H”<0). Let us assume that voters consume all disposable income so:p

uJ = y – τ + f J + H(g)

where y is income, τ indicates taxes, and f J denotes a nonnegative lump sum transfer to


w e e y s co e, τ d cates ta es, a d f de otes a o egat ve u p su t a s e tomembers of group J.

Legislators and Rents

Once again, legislators (i.e. the government) can employ tax revenues to produce the public good, but can also appropriate part of the revenues as private rents r = l r l.

Note: r l indicates the rents appropriated by legislator l.pp p y g

The production of public good g entails a cost of transforming private goods into public goods equal to 1. (q = 1. No uncertainty about it.)

The government budget constraint is:

g = 3τ – r - f

where f = f J


where f J f .

The Politician: Public goods and Rents (cont.)

As in Topic 1, each incumbent politician enjoys (exogenous) rents from being in office R. The politician can obtain R only if elected, which happens with reelection probability pl.

Preferences of politicians are then given by:

l lγrl + plR (16)

where γ < 1 reflects the fact that politicians face some transaction cost in extracting private trents.


Timing of the game


1. Nature selects at random among the legislators an agenda setter a.

2. Voters set their reelection voting strategies setting pl contingent on their utility level.

3. The agenda setter a proposes a policy vector q = [{f J}, g, {rl}, τ] conditional on ex ante electoral strategy pl .

4. The three legislators vote on the proposed policy vector q. If a majority supports the proposal, q is implemented; otherwise a default policy q0 = [{f J =0}, g=0, {r0>0}, τ=r0] is implemented.

5 P liti i f id ti l h ll El ti h ld


5. Politicians face an identical challenger. Elections are held.

Voters’ Strategies

As usual we restrict our attention to equilibria in which voters in each group J coordinate on the same retrospective voting strategy:p g gy

pl = 1 iff U(q) > kJ (17)0 otherwise

where kJ indicates the voter’s reservation utility and J = l.

Voters play Nash against all other districts.

Legislators have an interest in increasing the welfare of their district if they wish to beLegislators have an interest in increasing the welfare of their district if they wish to be reelected.

Note: This voting rule is ex post optimal only under our assumption about politicians facing


g p p y p p gidentical challenger. The timing would matter in a probabilistic voting model.

Equilibrium Conditions

Let me indicate the equilibrium policy vector of this simple legislature (L) game as qL

An equilibrium of this game has to satisfy the following three conditions:

1. For any vector of reservation utilities {kl}, at least one legislator l ∫ a must weakly prefer policy qL to policy q0.

2. For any vector of reservation utilities {kl}, legislator l = a must strictly prefer policy qL

to any policy obtaining a majority in the legislature.

3. The reservation utilities {kl} are optimal taking as given the reservation utilities in other districts and taking into account the policy implemented.


Solution: Transfers to l ∫ a

The districts that are not the agenda setter cannot obtain positive transfers.

The intuition is that a minimum winning coalition of two districts is necessary: The agenda setter plus another legislator (no matter from where).

The agenda setter will join forces with the district/legislator that is the cheapest to buy (Bertrand competition again).

This will push down the requests the two legislators/districts l ∫ a will be able to see satisfied.

In particular, excluding collusive behavior by assumption, Bertrand competition will push transfers outside a to zero.


f L,l = 0 for l ∫ a

Solution: Rents to l ∫ a

In case of failure of the proposal qL to pass legislators who are not the agenda setter canIn case of failure of the proposal q to pass, legislators who are not the agenda setter can garner reservation utility γr0

Consider for instance the legislator in coalition with the agenda setter. His choice is between g gsupporting the proposal and get γr + R or voting against the proposal and get γr0

So the minimal amount of rents that the agenda setter is going to allow her allies is

γrL,m + R = γr0

orrL,m = r0 - R/γ

which we assume negative (the default option is not as good as staying in office without rents). Hence, no rents accrue to l ∫ a. (This is without impact on our results other that


you do not have to compensate allies to vote for the proposal.)

a’s Rents

a has a decision to make:

a. Steal everything setting g* = f *= 0 and total rents r* = 3y. However, a has to convince another legislator to support this “Leviathan” proposal. Since both politicians will not

0be reelected, the minimum level of rents to allocate to the ally has to be r0. Rents for ain this case are 3y - r0

b E l ti If i i i t l ti t th L a +Rb. Earn reelection. If a is aiming at reelection, a gets then γrL,a +R

So the incentive compatibility constraint for a is:L a +R (3 0)γrL,a +R ¥ γ(3y - r0)

orrL,a ¥ 3y -R/γ -r0

V t i di t i t ill k th liti i d h k b ki th i lit bi d


Voters in district a will keep the politician under check by making the inequality bind.

a Voters’ Problem

Notice that rents to legislators l ∫ a and transfers to districts J ∫ a are allocated by politicianNotice that rents to legislators l ∫ a and transfers to districts J ∫ a are allocated by politician a, which is in turn held accountable by voters in J = a.

Ultimately the voters will be able to set a reelection strategy such that the policy proposed y gy p y p pby the agenda setter satisfies:

Max{y – τ L + f L,a + H(g L)}

subject to:g L = 3τ L – r L,a – f L,a

rL,a ¥ 3y -R/γ -r0

where I’ve already replaced the results that there are no rents to legislators l ∫ a and no transfers to districts J ∫ a


a Voters’ Solution

Voters have all the incentive to tax all other groups to the maximum provide some publicVoters have all the incentive to tax all other groups to the maximum, provide some public good, and redistribute to themselves, so tL = y

Voters in district a will keep the politician under check by making the inequality on rents p p y g q ybind:

rL,a = 3y -R/γ -r0

Transfers can be obtained from the government budget constraint: f L,a = R/g + r0 – gL

Voters will set the provision of the public good to the point where their private marginal utility H’(gL) is equal to their private marginal cost (1):

H’( L) 1


H’(gL)=1

a Voters’ Solution

Remark 1: Voters in a will set the provision of the public good to the point where theirRemark 1: Voters in a will set the provision of the public good to the point where their private marginal utility H’(gL) is equal to their private marginal cost (1):

H’(gL)=1(g )

The benefit to the other two districts is not internalized & the public good is underprovided.

Notice, however, that f L,a = R/g + r0 – gL ¥ 0. So we cannot have a too-large a gL. Assume this holds.

Remark 2: The reservation utility of the voters in l ∫ a is met at H(gL). That is, the public good savages some utility from the less powerful districts, by linking their welfare to the welfare of the most powerful group a.


Summary of the Equilibrium

The unique subgame – perfect equilibrium of this game satisfies:rL = rL,a = 3y – R/g -r0r r 3y R/g r

rL,l = max[0, r0 - R/γ]=0 for l ∫ a

tL = yt y

H’(gL)=1

f L,a = R/g + r0 – gL

f L,l = 0 for l ∫ a

k L,a = H(gL)+ f L,a

k L,l = H(gL) for l ∫ a


All politicians are reelected.

Discussion

Notice that since all the groups are playing Nash districts with l ∫ a really suffer fromNotice that since all the groups are playing Nash, districts with l ∫ a really suffer from Bertrand competition as in the case of electoral rules and political accountability.

However, voters in the district represented by the agenda setter a will be still able to , p y gdiscipline their politician, since staying in office in such a valuable position is worthwhile.

Voters in the district represented by the agenda setter a will be also benefit from the position of power of their own representative in the form of positive transfers. They will expropriate resources from the two other districts.

Finally, in the “simple legislature” public goods are underprovided relative to the optimum . Differently from transfers, g benefits all districts and not just voters in district a. However, we have shown that in equilibrium group a will only consider its private


However, we have shown that in equilibrium group a will only consider its private benefit and cost (i.e. Only 1/3 of the social benefits are internalized.)

Presidential-Congressional Regime

Presidential-Congressional regimes can be linked to our model of the simple legislature.

Consider how the proposal power on taxes and the allocation of spending may be split between to two different legislators or branches of government. Think of it as the President presenting the budget and Congress proposing the ways and means to implement it.

This model will present some of the interesting feature of our separation of powers discussion. Just now we will be also have a (more realistic) segmented polity of N =3 districts.

Of th th f t h t i i id ti l t Thi d l d tOf course, there are other features characterizing presidential systems. This model does not encompass all of them.

We are mostly going to capture the gist of the typical dispersion of proposal power typical of


We are mostly going to capture the gist of the typical dispersion of proposal power typical of presidential regimes.

Timing of the game


1. Nature selects at random among the legislators an agenda setter for g, ag, (say, the president) and an agenda setter for τ, aτ, (say, the ways and means committee).

2 Voters set their reelection voting strategies setting pl contingent on their utility level2. Voters set their reelection voting strategies setting p contingent on their utility level and depending on the status of their legislator.

3. The agenda setter aτ proposes a tax rate τ. 4. The three legislators vote on τ. If the proposal fails, then default tax rate τ0>0.4. The three legislators vote on τ. If the proposal fails, then default tax rate τ 0. 5. The agenda setter ag proposes a budget q = [{f J}, g, {rl}] conditional on ex ante

electoral strategy pl and τ.6. The three legislators vote on the proposed policy vector q. If a majority supports the g p p p y q j y pp

proposal, q is implemented, otherwise a default policy q0 = [{f J =0}, g=0, {r0>0}] is implemented.



Solution: Transfers to l ∫ ag

Solve backward. At stage 5, the districts that are not the ag agenda setter cannot obtain positive transfers.p

The intuition is that a minimum winning coalition of two districts is necessary: The agenda setter plus another legislator (no matter from which district).

The agenda setter will join forces with the district/legislator that is the cheapest to buy.

This will push down the requests the two legislators/districts l ∫ ag will be able to see satisfied.

In particular, excluding collusive behavior by assumption, Bertrand competition will push transfers outside ag to zero:

Political Economy - Trebbi152 Political Economy - Trebbi152 Political Economy - Trebbi152 Political Economy - Trebbi152 Political Economy - Trebbi152 Political Economy - Trebbi152

f C,l = 0 for l ∫ ag

Solution: ag’s Rents

ag has a decision to make:

a. Steal everything setting g* = f *= 0 and total rents r* = 3t. However ag has to convince another legislator to support the proposal. Since both politicians will not be reelected, the minimum level of rents to allocate to the ally has to be r0the minimum level of rents to allocate to the ally has to be rRents for ag in this case are 3t - r0

b. Earn reelection. If ag is aiming at reelection, ag gets then γrC,ag +Rb. Earn reelection. If a is aiming at reelection, a gets then γr +R

So the incentive compatibility constraint isγrC,ag +R ¥ γ(3t - r0)γr +R ¥ γ(3t r )

orrC,ag ¥ 3t -R/γ -r0

Voters in district ag will keep the politician under check by making the inequality bind (rC,ag


Voters in district a will keep the politician under check by making the inequality bind (r=max[0,3t -R/γ -r0]).

Solution: Rents to l ∫ ag

In case of failure of the proposal qC to pass legislators who are not the agenda setter canIn case of failure of the proposal q to pass, legislators who are not the agenda setter can garner reservation utility γr0

Consider for instance the legislator in coalition with the agenda setter. His choice is between g gsupporting the proposal and get γr + R or voting against the proposal and get γr0

So the minimal amount of rents that the agenda setter is going to allow her allies is

γrC,m + R = γr0

orrC,m = r0 - R/γ

which we assume negative, so no rents accrue to l ∫ ag. (This is without impact on our


g ( presults other that you do not have to compensate allies to vote for the proposal.)

ag Voters’ Decision on g

Voters in ag will still set the provision of the public good to the point where their private marginal utility H’(gC) is equal to their private marginal cost (1):

H’(gC)=1

Recall, any dollar spent on g does not get to them in the form of transfers., y p g g

Notice, however, that f C,ag = R/g + r0 – gC ¥ 0. So, as before, we cannot have a too-large a gC. Assume this condition holds.


Taxation Stage (3)

Presidential regimes have this additional stage where the ways and means are determined d th ll i f th b d t i tand the overall size of the budget is set.

The voters supporting the tax policy agenda setter aτ know that at stage (5) no transfers accrue to any l ∫ agaccrue to any l ∫ ag

At stage (5) indeed the legislator ag will have no incentives to enforce any agreement between them (assume a one-shot game) and no rents will be shared eitherbetween them (assume a one shot game) and no rents will be shared either.

Interestingly, we can make again the point we made in the separation of powers’ analysis: Agenda setter aτ is not the residual claimant over tax revenues (the residual claimant is g (ag ) and aτ has no incentive to raise taxes beyond the level necessary to strictly provide g*C.


This is good for voters because they have full alignment with legislator aτ .

Taxation Stage (3)

In the presidential-congressional regime tax rates can be minimal:

τ*C = g*C/3

which implies r*C,ag = 0 and f *C,ag = 0 is an equilibrium.

This equilibrium is supported by cutoff rules of the type:

k C,l = y - g*C/3 + H(g*C) for all l

Note: Voters in ag are better off accepting no transfers than making an incompatible demand triggering a deviation by their politician with g=f=0 and maximal rents. This follows from y - g*C/3 +H(g*C) > y - g*C/3 which is what district ag would get in case any incompatible demand was not met and taxes were already set at τ*C = g*C/3 (note


any incompatible demand was not met and taxes were already set at τ = g /3 (note all tax revenues would go to Leviathan legislators not to transfers).

Equilibrium discussion

Persson and Tabellini (2000) show that this equilibrium is not unique. [We have just described the best equilibrium for voters in district aτ .]

There are also equilibria where rents are still zero, but voters in district aτ accept to be taxed above g*C/3 and ag voters get positive transfers:

τC œ [g*C/3, (R/g + r0)/3][g ( g ) ]and

f C,ag œ [0, R/g + r0 – g*C].Note: rC,ag =max[0,3t -R/γ -r0] implies that taxes above (R/g + r0)/3 automatically go into [ γ ] p ( g ) y g

rents to ag. Also voters in aτ are better off accepting taxes above g*C/3 than making an incompatible demand triggering a deviation by their politician with g=f=0 and maximal rents.

Since cutoff rules are set simultaneously at stage (2) we cannot rule out these equilibria (there are many ways in which voters in ag and aτ can agree on a zero-rents platform).

However the general result is that rents and taxes are lower under a presidential


However, the general result is that rents and taxes are lower under a presidential-congressional regime and public goods are still heavily underprovided.

Parliamentary Regime

We now present a model which clearly identifies the incentives for coalition-building thatWe now present a model which clearly identifies the incentives for coalition building that are present within parliamentary regimes.

Consider how the proposal power on taxes and the allocation of spending may be split p p p p g y pbetween to two different legislators, but now allow the two proponents cooperate too a certain extent, weakening the separation of powers

Think of aτ proposer of the tax rate τ and the expenditure agenda setter ag as ministers of the same cabinet.


Timing of the game


1. Nature selects at random among the legislators an agenda setter for g, ag, (say, the president) and an agenda setter for τ, aτ, (say, the ways and means committee).

2 Voters set their reelection voting strategies setting pl contingent on their utility level2. Voters set their reelection voting strategies setting p contingent on their utility level and depending on the status of their legislator.

3. The agenda setter aτ proposes a tax rate τ. 4. The agenda setter ag proposes a budget q = [{f J}, g, {rl}] conditional on ex ante4. The agenda setter a proposes a budget q [{f }, g, {r }] conditional on ex ante

electoral strategy pl and τ.5. The two agenda setters have veto power on the other’s member proposal. If there is no

veto, τ and the proposed policy vector q are implemented. If the proposal fails (no confidence motion), a default policy q0 = [{f J,0 =0}, g=g0, {r’/3>0}, τ0=g0+r’] is implemented.



Default Policy Characteristics: Technical Assumption

Persson and Tabellini have to be a bit untidy here to simplify their life and set a default policy vector for the government crisis in the following way:

{f J 0 0}{f J,0 =0}, g= g0, such that H’(g0)=1 {r’/3>0}, with r’ = 3y – (R/g + r0)

0 0 ’τ0= g0+r’

These values match the expected payoffs of a “caretaker government” subgame (Example: Government crisis triggers a technocrat government to rule ad interim Italy 1994)Government crisis triggers a technocrat government to rule ad interim. Italy 1994).

The idea is to make “breaking the coalition” and losing their power very costly for the two agenda settersagenda setters.

The caretaker subgame is the following: (1) One legislator is picked at random to be the caretaker; (2) Voters reformulate electoral strategies; (3) The caretaker makes the entire


ca eta e ; ( ) Vote s e o u ate e ecto a st ateg es; (3) e ca eta e a es t e e t ebudget proposal; (4) Congress votes for the proposal.

Parliamentary Regime: Intuition

Politicians will be jointly deciding the budget so the two agenda setters will clearly colludePoliticians will be jointly deciding the budget, so the two agenda setters will clearly collude more.

Now both legislators are residual claimants on tax revenues by setting their own rents.g y g

Voters will see their strength reduced vis-à-vis politicians relative to the presidential-g p pcongressional regime.

However, mutual veto power gives bargaining power to all members of the coalition, so more of the social benefits from public goods will be internalized.


Parliamentary Regime: Rents

aτ and ag have a joint decision to make:

a. Steal everything setting g* = f *= 0 and total rents r* = 3y. Since both politicians will not be reelected, total utility in this case is γ3y

b. Earn reelection. If aτ and ag are aiming at reelection, each a gets then γrP,a +R

So the joint incentive compatibility constraint isγrP +2R ¥ γ3y

orrP ¥ 3y -2R/γ

Voters will keep the two politicians under check by making the inequality on total rents bind. Of course rents to the remaining legislator will be zero.


You can show that rL ¥ rP ¥ rC

Parliamentary Regime: Taxes

aτ and ag are both residual claimants on tax revenues and since high taxes maximize the amount of resources available for rents they like to raise taxes.

On top of that, voters in their districts will also be happy to fully expropriate voters outside the governing coalition and get part of the resources back in the form of transfers and public good provision.

I ilib i th d tt τ t t P d g tIn equilibrium the agenda setter aτ proposes a tax rate τP = y and ag accepts.

This implies that τL = τP ¥ τC


Solution Sketch

Redistribution will favor the majority.

However, the fact that now two legislators enter the cabinet and have both veto power implies that the provision of the public good will be higher, as the optimal decision will entail comparing the marginal cost of increasing the public good provision and to the sum of the marginal benefits of both aτ and ag voters. More of the social benefit is internalized.

To solve this problem one has to maximize the utility of voters in district aτ given the utility of voters in district ag (or vice versa).


Solution Sketch

It is possible to show that if {f J,P >0}, J = aτ, ag , g is such that 2H’(g)=1

But since once again there’s simultaneity in the process of setting reservation utilities by theBut since once again there s simultaneity in the process of setting reservation utilities by the voters welfare can be split among the two groups of voters in many different ways, other equilibria exist with {f J,P ¥0}, J = aτ, ag , g is such that 1/2§ H’(g)<1.

This equilibrium is supported by cutoff rules of the type:

k P,J = f*P,J + H(g*P) for J = aτ

and k P,J = f*P,J + H(g*P) for J = ag

and


k P,J = H(g*P) for J ∫ aτ, ag

Institutional Trade-off: Low rents vs. Public goods under-provision

Presidential-Congressional regimes rely on the stronger separation of power betweenPresidential Congressional regimes rely on the stronger separation of power between branches of government to curb on legislator’s abuse. Presidential regimes exploit the conflict of interest among legislators to the voter’s benefit.

Parliamentary regimes allow for collusion among different members of the governing coalition, which brings rent-seeking and taxation to higher levels relative to presidential regimes.

However, parliamentary regimes allow for more provision of public goods, by encompassing coalitions including representatives from a larger spectrum of the electorate and hence internalizing more of the benefits of the public goods (the rulingelectorate and hence internalizing more of the benefits of the public goods (the ruling majority being larger).


Further Discussion on Form of Government

Controlling for income per capital and a set of other standard cross-country covariatesControlling for income per capital and a set of other standard cross country covariates, Presidential-Congressional regimes tend to have a 10% lower public spending to GDP relative to Parliamentary regimes (Persson and Tabellini, 1999).

No clear statistical evidence on corruption.

If you combine parliamentary regimes and PR systems you also see more broadly oriented public spending programs relative to targeted programs (form of government alone is not significant).


Democracies and Autocracies

Our attention now shifts to model of democratic transitionsOur attention now shifts to model of democratic transitions.

Many interesting theories and models of political transitions (Lipset (1959); PrzeworskiMany interesting theories and models of political transitions (Lipset (1959); Przeworski (1991); Linz and Stepan (1996); Myerson (2007, 2008); Mulligan and Tsui (2007), etc.)

Here I will focus on Acemoglu and Robinson (AER 2001) A Theory of Political Transitions.

This is in many respects the workhorse model of their 2006 book and it encompasses the model of Acemoglu and Robinson (QJE 2000) on the extension of the democratic franchise in Western Europe and many subsequent papers.


The Model

Two groups of agents: The Poor and the Rich (the Elite)Two groups of agents: The Poor and the Rich (the Elite).

Two political states: Democracy (median voter is Poor and controls policy) and Autocracy (the Elite controls policy).( p y)

During an Autocracy the Poor can mount a revolution (storming the Bastille) and the Elite can establish a democracy.

During a Democracy the Elite can mount a coup,

Income is stochastic and the opportunity cost of coups and revolutions changes with income.

No commitment to future level of taxation is possible.


p

Setup

Infinite horizon Discrete timeInfinite horizon. Discrete time.

Measure 1 of agents. A fraction >1/2 agents are Poor (p), the rest is Rich (r).

There is an unique consumption good y and a unique productive asset with total stock h.

At the beginning of time t=0 the poor owns hp and the rich hr > hpAt the beginning of time t 0 the poor owns hp and the rich h > hp.

To parameterize income inequality in this economy we use the parameter q < such that the share of capital owned by the poor is less than proportional to their size hp= hq /the share of capital owned by the poor is less than proportional to their size h hq /and hr = h*(1- q )/(1-).

Higher q implies lower income inequality (and vice versa).


g p q y ( )

Setup

Income accrues from production: yi = A *hi for i =p r where A indicates aggregateIncome accrues from production: yi,t At h for i p, r where At indicates aggregate productivity, which follows a stochastic process:

At = Ah = 1 with prob. 1-st pAt = Al = a with prob. s

where a < 1 and indicates a period of recession and s < 1/2 so recession a relatively rare. p y

All agents maximize expected discounted consumption :

Et j j Ci,t+j for i = p, r


Setup

Disposable income is given by after-tax income plus (not-group-specific) transfers:Disposable income is given by after tax income plus (not group specific) transfers:

(1-tt)*Athi +ft

Also assume that there is a cost in raising taxes equal to c(tt)*Ath with c’(0)=0, c’>0, c”>0, c’(1)=¶ .

The government budget constraint is given by:

ft = tt*Athp + tt*Athr - c(tt)*Athft tt Ath tt Ath c(tt) Ath = tt - c(tt))*Ath


Revolution

A revolution can be attempted in any nondemocratic period after t = 0. If attempted, it always succeeds.

After a revolution the share of capital owned by the poor becomes more proportional to their size hp / with q < . p tells how high the returns to the revolution are.

The revolution destroys (1 - m) > 0 fraction of resources in the period in which happens. m tells how cheap the revolution is (the higher m, the less the poor lose from revolution).

For the poor the return in the period of the revolution is mAthp / and the per period return ever after is Athp /.

Note: the discounted net present value of a revolution is Wp(R)=(1 - s + sa)*hp /p ( ) ( )

For the rich the revolution is very costly: they lose everything forever. Wr(R)=0. This is the disciplining device through which the poor can obtain redistribution or


democratization. The rich will always try to avoid it.

Coup d’etát

A coup can be attempted in any democratic period after t = 1A coup can be attempted in any democratic period after t 1.

If attempted, the coup always succeeds.

The coup destroys (1 - f) > 0 fraction of resources in the period in which it happens. f tells how cheap the coup is (the higher f , the less the rich loses from coup).

For the poor and for the rich the return in the period of the coup is fAthi.


Timing of the game


1. Nature determines aggregate productivity At .2. If there has been a revolution in the past, the poor receive their income, consume, and

the period ends If the state is democracy the poor picks a tax rate τ If the state isthe period ends. If the state is democracy the poor picks a tax rate τt. If the state is autocracy the rich picks a tax rate τt.

3. In a nondemocratic regime the rich decide whether to extend the electoral franchise to the poor. In a democratic regime the rich decide whether to stage a coup. Whoever is in p g g ppower now can fix a new tax rate τt for the period.

4. In a nondemocratic regime the poor decide whether to initiate a revolution. If there is a revolution the poor share the remaining income in the economy. If there is no

l ti th t t t i l t drevolution the tax rate τt gets implemented.5. All receive their income, consume, and the period ends.

N h d k l i i di l f ( d i

Political Economy - Trebbi176 Political Economy - Trebbi176 Political Economy - Trebbi176 Political Economy - Trebbi176 Political Economy - Trebbi176 Political Economy - Trebbi176 Political Economy - Trebbi176

Note: the poor cannot undertake a revolution immediately after a coup (you need to start in a nondemocratic regime).

Equilibrium

Two players: Poor and Rich representative agentsTwo players: Poor and Rich representative agents.

Focus on (pure strategy) Markov Perfect Equilibria: Strategies depend only on the current state and the prior actions taken within the same period. [It’s a standard way for p p [ yexcluding history-dependence of strategies and other complications.]

For a given level of productivity A, there are three possible states S: (A, D) = poor in power (Democracy)(A, E) = Elite in power (autocracy)(A, R) = Revolution (an absorbing state)


Transitions

(A, E)

(A, E) (A, D)

(A, D)

(A, R)

( , )

(A, E)

Revolution Coup d’etát

Democratization


Democratization

Strategies

For the rich:r(S| τp) = { τr} (S| τ ) { , , τ }

, in state (A,E) extend the franchise () or not () in state (A D) given τp stage a coup () or not (), in state (A,D), given τp, stage a coup () or not ()τr in state (A,E) or in state (A,D) after a coup () , fix the tax rate by the rich.

For the poor:For the poor:p(S| ,τr) = {, τp}

in state (A E) initiate the revolution () or not (), in state (A,E) initiate the revolution () or not ()τp in state (A,D) fix the tax rate by the poor.


Equilibrium

A pure strategy Markov Perfect Equilibrium is a strategy combination r*(S| τp), p*(S| ,τr) such that these strategies r*, p* are best-responses to each other for all possible states.

See paper for a more formal definition.


Analysis: Preliminaries

Some preliminaries firstSome preliminaries first.

What is the optimal tax rate tm the poor would set, absent the risk of a coup?

Simply maximize the per-period consumption of the poor:

Max{(t )*A hp + t - c(t ))*A h}Max{( tt) Athp + tt - c(tt)) Ath}

which implies c’(tm))=(l-q)/lc (t ))=(l-q)/l

using the fact that hp= hq / . So the higher the inequality (the lower q), the higher the taxes. Notice also that in the absence of political change taxes would be constant.


Notice also that in the absence of political change taxes would be constant.

Analysis: Preliminaries

Define d i(q) A the net amount of redistribution that a person of group i receives in state ADefine d (q) At the net amount of redistribution that a person of group i receives in state Atwhen the tax rate is tm.

So, d i(q) Atf m tm *Athi, ( ) t f t

and from the budget constraint transfers are f m t m - c(t m))*Ath

This implies net transfers to the poor: d r(q) < 0 < d p(q)


Assumptions

Assume revolutions and coups are not worthwhile in periods of economic expansionAssume revolutions and coups are not worthwhile in periods of economic expansion (i.e. At = Ah = 1).

Assumption 1: Sufficient condition for which coups are not profitable in good timesAssumption 1: Sufficient condition for which coups are not profitable in good times.

The cost of a coup for a rich agent in normal times (the direct loss from the coup minus taxes paid = (1-)hr +d r(q) ) is always larger than the taxes tm avoided forever (-(1 – s p ( ) ( ) ) y g ( (+ sa)*d r(q)*)

That is:(1-)(1-)hr > -(1- s(1-a)) d r(q).


Assumptions

Assumption 2: Sufficient condition for which revolutions are not profitable in good times.

In state (At,E) the value of a revolution isVp(At, R)= m*Athp / + Wp(R)

The value of never undertaking a revolution and hence never receiving any transfer from the rich from there after is:

Vp(At, E)= Athp + (1 –s + sa)*hpOf course this value is a lower bound of the utility under autocracy for the poor (because

occasionally there could be redistributive taxation – the rich could tax themselves and give to the poor).

So we assume that for At = Ah = 1, Vp(1, E) > Vp(At, R)

With this assumption we know there is never going to be any revolution in good times And


With this assumption we know there is never going to be any revolution in good times. And this also implies the rich will never redistribute to the poor in good times.

Analysis

We need to derive some intuitive value functions in the different states of the world.

What is the value of being in a democracy during good times (At = 1) for agents i =p, r?

V i(1 D) = hi + d i(q)+ W i(D)V (1, D) h + d (q)+ W (D)

where we make use of the fact that there is never going to be a coup in good times (hence the net transfers are d i(q)*1) and the continuation value from next period on of being in ( ) ) p gstate D is:

W i(D) = (1-s)*V i(1, D) + s *V i(a, D)

which depends on the state of the economy next period (could be a boom or a recession).


Analysis

What is the value of being in a democracy during bad times for agents i =p, r, V i(a, D)?

Now, here the situation is tricky for the poor. If they redistribute too much, they can trigger a coup. So they may decide to keep taxes low and transfer less to themselves in bad times just to avoid a coup by the elite.

Call this tax rate (if feasible) td < tm.

S d t th thSuppose td prevents the coup, then

V i(a, D)= v i(a, D| td)= a(hi + i(q ,td))+ W i(D)

where net transfers under the threat of a coup are i(q ,td))At f d td *Athi .Notice that the continuation value is still W i(D) which tells us that if the next period the

poor have good times they will increase taxes back up to tm (at that point this can’t


poor have good times they will increase taxes back up to tm (at that point this can t trigger a coup because times are good). The poor cannot commit to keep taxes low.

Analysis

Reducing taxes in a democracy may not be enough to prevent a coup though. Let’s see what is the decision by the rich about this:y

V r(a, D)= max { V r(a, E)= + (1-vr(a, D| td)}

where the continuation value of a coup () in state (a, D) is:

V i(a, E)= ahi + W i(E)( , ) ( )

which depends on the fact that the rich will be able to set taxes to zero right after the coup (recall there cannot be revolution immediately after a coup, only in the following period).

The continuation value from next period on of being in state E is:


W i(E) = (1-s)*V i(1, E) + s *V i(a, E)

Analysis

What happens when the Elite is in power?In a boom the rich will set taxes to zero, since there cannot be a revolution by Assumption 2,

so for agents i =p r:so for agents i =p, r:

V i(1, E) = hi + W i(E)

In a recession the rich will have several options :

1 They can democratize ( )1. They can democratize ()2. They can decide not to democratize () but they can raise taxes from 0 to te to appease

the poor and avoid a revolution ().3 A revolution may occur ()3. A revolution may occur ().

Since we start from an autocracy, if either or then you would never observe a democracy. Since we use V i(a, E) in calculating a deviation from democracy in what


democracy. Since we use V (a, E) in calculating a deviation from democracy in what follows along the equilibrium path let us focus on case 1 ().

Analysis

So if then for agents i =p, r:

V i(a, E) = a(hi + d i(q))+ W i(D)

which depends on the fact that the poor will set tm taxes (recall there cannot be a coup immediately after a democratization, only the following period, so the poor will pick the best tax rate for them) and the continuation value from next period on of being in state D is what we derived earlier.

Finally allow for:Assumption 3: Assume revolutions are worse than democracies, so democratizations can

help preventing revolutions.help preventing revolutions.Vp(a, R) < Vp(a, D)

(Excludes case 3 in previous slide)This completes the derivation of the value functions Let’s now look at the properties of the


This completes the derivation of the value functions. Let s now look at the properties of the equilibrium.

Coup Constraint

In state (a D) the elite would prefer not to stage a coup if it is too expensive or:In state (a, D) the elite would prefer not to stage a coup if it is too expensive, or: V r(a, E) < vr(a, D| td)

That is by replacing the expressions in the previous slides:That is, by replacing the expressions in the previous slides:ahr + Wr(E) < a(hr + r(q ,td))+ Wr(D)

or more intuitivelyor, more intuitively,

Wr(E) Wr(D)) -a r(q ,td) < ahr (1- ) (18)Capturing power & reducing taxes from td to 0 < Cost of the coupCapturing power & reducing taxes from t to 0 < Cost of the coup .

Note: If a is large (recession not too deep) the coup is more expensive and less likely.


Consolidated Democracy

In state (a, D) the elite will never stage a coup for levels of low enough, such that:

Wr(E) Wr(D)) -a r(q ,tm) < ahr (1- ) (18)

l lor, more clearly,Wr(E) Wr(D)) -a r(q ) < ahr (1- ) .

Capturing power & reducing taxes even when taxes at the maximum tm < Cost of the coup

Substituting the value functions ,you get a threshold , a, s):

) (hr r(q )) + r(q ))/(( hr), a, s) sa(hr r(q )) + s r(q ))/((sahr)

For coups never occur.


Consolidated Democracy

For coups never occur, so increasing decreases the range in which coups might occur:occur:

1. ∑, a, s)∑more equal societies are easier to consolidate (lower need to tax the rich).rich).

2. ∑, a, s)∑ a less severe recession make consolidation easier (by increasing the opportunity cost of a coup).pp y f p)

3. ∑, a, s)∑ s more frequent recessions make consolidation easier (increasing the frequency at which the rich pay lower taxes makes democracy less costly to the rich).


Semi-Consolidated Democracy

In state (a, D) the elite always stages a coup for levels of high enough, such that:

Wr(E) Wr(D)) r(q 0) > hr (1 ) (18)Wr(E) Wr(D)) -a r(q ,0) > ahr (1- ) (18)or, more clearly,

Wr(E) Wr(D)) - 0 > ahr (1- ) .C t i & d i t h t t th i i d 0Capturing power & reducing taxes even when taxes are at the minimum td =0

> Cost of the coup

S b tit ti th l f ti t th h ld ) F lSubstituting the value functions you get a threshold , a, s). For coups always occur during recessions:

, a, s) sahr + s r(q ))/((sahr)

For the democracy is semi-consolidated, that is, in order to prevent a coup, during recessions the poor lowers taxes to a level td tm

Note: During booms taxes go back up to tm Even if the country remains democratic the


Note: During booms taxes go back up to t . Even if the country remains democratic the threat of a coup influences tax policy!

Revolution Constraint

In state (a E) the poor would prefer not to start a revolution if it is not worth it or:In state (a, E) the poor would prefer not to start a revolution if it is not worth it, or: V p(a, R) < vp(a, E| te)

where the rich may wish to avoid revolution by conceding some redistribution te

That is:m*ahp / + Wp(R) < a(hp +dp(q ,te))+ Wp(E) (19)

where d i(q ,te))af e te*ahi and means:

Capturing power through the Revolution < Value of living in an autocracy for the poorCapturing power through the Revolution < Value of living in an autocracy for the poor

Note: The elite may have the opportunity of avoiding revolutions by just increasing taxes and redistribute during recessions. However, even giving te =tm may not be enough to


and redistribute during recessions. However, even giving t t may not be enough to satisfy (19). In that case they will need to democratize.

Democratizations

In state (a, E) the poor would always start a revolution if : V p(a, R) > vp(a, E| tm)

where the rich tries to avoid revolution by conceding maximum redistribution tm

That is:m*ahp / + Wp(R) > a(hp +p(q ))+ Wp(E) (20)

Substituting the value functions you get a threshold m, a, s). For m > m a revolution is always attractive during recessions even at maximum redistribution tm :

) (h ( )) h h /(( h)m, a, s) sa(hp+p(q ))+s hp s+ash /((ah)

In this case the rich can only democratize to avoid a revolution.


Democratizations

For m < m the autocracy can prevent the revolution by redistributing resources during recessions, that is, in order to prevent a revolution during recessions the rich increases taxes (on themselves) to a level te tm .

During booms taxes go back down to 0.

Note: Even if the country remains autocratic the threat of a revolution influences tax policy!Note: Even if the country remains autocratic the threat of a revolution influences tax policy!


Democratizations

For m > m a revolution is always attractive during recessions even at maximum di t ib ti th i h d ti d i th i t id l ti (bredistribution, so the rich democratize during the recession to avoid a revolution (by

Assumption 3).

1 ∑ m a s)∑more equal societies are less likely democratize and the poor are more1. ∑ m, a, s)∑more equal societies are less likely democratize and the poor are more likely to just be happy with redistribution (autocracy is not that costly the poor).

2. ∑ m, a, s)∑ a less severe recessions make societies less likely to democratize and the poor are more likely to just be happy with redistribution (by increasing the p y j ppy ( y gopportunity cost of a revolution).

3. ∑ m, a, s)∑ s more frequent recessions make societies less likely to democratize and the poor are more likely to just be happy with redistribution (increasing the frequency

h h h h h h k l l h )at which the rich pay higher taxes makes autocracy less costly to the poor). Frequency of recessions acts as a commitment to redistribution.


Equilibrium (under Assumptions 1-3): Proposition

1. If m < m, a, s), then society remains nondemocratic forever. Intuition: a revolution can always be bought off by the elite and the system remains an autocracy.y g ff y y y

2. If m > m, a, s) and if , a, s), then society democratizes the first time the state is (a, E) (at the first recession) and it remains a consolidated democracy forever. Intuition: The revolution threat forces democratization and then coups are too costly to stage even when taxes are at the maximum tm.

3. If m > m, a, s) and if , a, s)<, a, s) , then society democratizes the first time the state is (a, E) (at the first recession) and it remains a semi-consolidated democracy forever. Intuition: The revolution threat forces democratization and then coups are not too costly to stage, so taxes have to be lowered in bad times.too costly to stage, so taxes have to be lowered in bad times.

4. If m > m, a, s) and if , a, s) , then society becomes an unconsolidated democracy the first time the state is (a, E) (at the first recession) and then at every


y ( ) ( ) yrecession it continuously switches regimes. Intuition: Democratizations follow coups.

Empirical Implications

1E

1 a 1 1 a1 a

Case 1ôOUTPUT

ô REGIME TYPED

1 1 a 1 1 a1 a

0 0 te 0 0 te0 te ô FISCAL POLICY

ôOUTPUT

DE

1 1 1 11

Case 20 0 tm tm tm tm tm tm

ô REGIME TYPE

ô FISCAL POLICY

1

DE

1 a 1 1 a1 aCase 3

0 0 tm tm tm tm td td

ôOUTPUT

ô REGIME TYPE

ô FISCAL POLICY

1

DE

1 a 1 1 a1 a

Case 4

0 0 t t t t t t

ôOUTPUT

ô REGIME TYPE


D0 0 tm tm tm tm 0 tm ô FISCAL POLICY

Examples

1. Society remains nondemocratic forever. Several examples were the poor are bought off, like Singapore, Saudi Arabia.g p ,

2. OECD countries. Extension of the democratic franchise to Western societies.

3. Countries that, albeit formally democratic, the threat of coups still constraints redistribution. Russia (? )

4. African and Latin American countries where continuous alternating between regime types continue and are often triggered by economic downturns (Chile, Argentina).


Notes

1. Societies with high asset inequality (low q) are more likely to have both more coups and revolutions. Hence they are more likely to be in case 4 and switch back and forth between democracy and autocracybetween democracy and autocracy.

2. In our model the richer the country (the higher h) does not imply the more democratic the country Both coup and revolution constraints are unaffected by the level of h perthe country. Both coup and revolution constraints are unaffected by the level of h per se.

3. Higher inequality increases fiscal policy variability within each case of equilibrium.3. Higher inequality increases fiscal policy variability within each case of equilibrium.

4. Societies where distortions c(.) are lower (less convex) will have higher taxes and less likely to consolidate (the rich fearing higher taxes will have more incentives to stage y ( g g gcoups). In a sense distortions are a good commitment device against expropriation of the elite.

h h l d l d ( d


Note: The paper has also a very interesting second section on consolidation (not covered in the final exam). Read it.

Conclusion on Comparative Politics

Wide cross-country and within country variation in political institutionsWide cross country and within country variation in political institutions.

Analysis of democracies vs autocraciesAnalysis of democracies vs. autocracies.

Within democracies: Many Constitutional features of relevanceWithin democracies: Many Constitutional features of relevance.

Electoral rules and Form of Government in particular are interesting among the otherElectoral rules and Form of Government in particular are interesting, among the other things, for their impact on fiscal policy.


c ti p liticomparative politics - faculty of...

Documents