do minimum wage hikes raise us long term unemployment evidence using state minimum wage rates

14
Regional Studies, Vol. 33.8, pp. 713± 726 Do Minimum Wage Hikes Raise US Long Term Unemployment? Evidence Using State Minimum Wage Rates 1 MARK D. PARTRIDGE* and JAMIE S. PARTRIDGE² *Department of Economics, Stewart Hall, St Cloud State University, St Cloud, MN 56301-4498, USA ² Department of Management, St John’s University/College of St Benedict, St Joseph, MN 56374, USA (Received July 1998; in revised form March 1999) P ARTRIDGE M. D. and P ARTRIDGE J. S. (1999) Do minimum wage hikes raise US long term unemployment? Evidence using state minimum wage rates, Reg. Studies 33, 713± 726. Several recent studies have challenged the conventional notion that raising the minimum wage reduces employment. This study considers a related but relatively unexplored issue by examining the minimum wage’s in¯ uence on long durations of unemployment. By considering long term unemployment rates, this study extends the previous minimum wage literature by examining the persistence of minimum wage eVects. The empirical analysis considers state data from the latter 1980s, a unique period when many states raised their minimum wage above the federal level. The results suggest that a greater minimum wage increases long term unemployment rates. Further evidence indicates that increased minimum wage coverage also raises long term unemployment rates. Subsequent analysis yielded similar patterns for other aggregate labour market measures. Thus, state and federal policy makers should weigh these potential costs in deciding whether to increase minimum wage rates in the future. Long term unemployment Minimum wage Minimum wage coverage Unemployment P ARTRIDGE M. D. et P ARTRIDGE J. S. (1999) Les retom- P ARTRIDGE M. D. und P ARTRIDGE J. S. (1999) FuÈ hren be es des augmentations du salaire minimum sur le choà mage Anhebungen von MindestloÈ hnen in den Vereinigten Staaten de longue dure e? Des preuves provenant des salaires mini- zu langfristigem Ansteigen der Arbeitslosigkeit? Beweise, die mum aux Etats-Unis, Reg. Studies 33, 713± 726. Des e tudes sich auf staatliche Mindestlohnraten stuÈ tzen, Reg. Studies 33, re centes ont mis en question l’ide e recË ue qu’une augmenta- 713± 726. Verschiedene, kuÈrzlich veroÈ Ventlichte Unter- tion du salaire minimum entraõ à neune re duction de l’emploi. suchungen stellten die herkoÈ mmliche Meinung in Frage, daû Cette e tude-ci cherche a Á conside rer un sujet connexe mais ein Anheben der MindestloÈ hne zur Herabsetzung der Zahl relativement inexplore ; autrement dit, les retombe es du salaire der BeschaÈ ftigten fuÈ hrt. Dieser Aufsatz zieht eine damit minimum sur le choà mage de longue dure e. En examinant verbundene, doch relativ selten untersuchte Frage in les taux de choà mage de longue dure e, cette e tude e largit la Betracht, indem er den Ein¯ uû des Mindestlohnes auf lang documentation ante rieure qui porte sur le salaire minimum anhaltende Arbeitslosigkeit pruÈ ft. Mit Hilfe der langfristigen en conside rant la persistance des retombe es du salaire mini- Arbeitslosigkeitsraten erweitert die Studie durch Unter- mum. On analyse des donne es qui proviennent de la ® n des suchung der anhaltenden Auswirkungen von MindestloÈ hnen anne es1980, unee poqueexceptionnelle ouÁ bon nombre des die Literatur uÈ ber MindestloÈ hne. Die empirische Analyse e tats ont augmente leur salaire minimum au-dessus du niveau zieht oYzielle Datan der spaÈ ten achtziger Jahre heran, einer fe de ral.Les re sultats laissent supposer qu’un renche rissement einmaligen Periode, insoweit als viele Staaten ihre Min- du salaire minimum entraõà ne une hausse des taux de choà mage destloÈ hne uÈ ber die auf Bundesebene gezahltenanhoben. Die de longue dure e. Des preuves supple mentaires laissent voir Ergebnisse legen nahe, daû ein hoÈ herer Mindestlohn zum qu’un accroissement de la proportion de smicards ameÁ ne Ansteigen der langfristige Arbeitslosenraten fuÈ hrt. Weitere aussiaÁ une hausse des taux de choà mage de longue dure e. Beweise lassen erkennen, daû vermehrte Mindestlohn- Une analyse ulte rieure a fourni des re sultats comparables deckung auch langfristige Arbeitslosigkeitsraten anhebt. Die pour ce qui est des autres mesures globales du marche du anschlieû ende Analyse ergabaÈ hnliche Muster fuÈ r andere, den travail. Ainsi, les de cideurs, et au niveau de l’e tat et sur le Gesamtarbeitsmarkt umfassende Maû nahmen. Auf staatlicher plan fe de ral, devraient e valuer ces couà ts potentiels au moment wie auf bundesstaatlicher Ebene sollten politische Entschei- ouÁ ils de cident si, oui ou non, il faudra augmenter les taux dungen deshalb diese potentiellen Kosten mit in Betracht de salaire minimum. ziehen, wenn sie zukuÈnftige Anhebungen von Min- destloÈ hnenerwaÈ gen. Choà mage de longue dure e Salaire minimum Proportion de smicards ChoÃmage Langfristige Arbeitslosigkeit Mindestlohn Mindestlohndeckung Arbeitslosigkeit 0034-3404/99/080713-14 ©1999 Regional Studies Association

Upload: stampit

Post on 12-Nov-2014

214 views

Category:

Documents


0 download

DESCRIPTION

Do minimum wage hikes raise US long term unemployment? Evidence using state minimum wage rates, Reg. Studies 33 , 713-726. Several recent studies have challenged the conventional notion that raising the minimum wage reduces employment. This study considers a related but relatively unexplored issue by examining the minimum wage's influence on long durations of unemployment. By considering long term unemployment rates, this study extends the previous minimum wage literature by examining the persistence of minimum wage effects. The empirical analysis considers state data from the latter 1980s, a unique period when many states raised their minimum wage above the federal level. The results suggest that a greater minimum wage increases long term unemployment rates. Further evidence indicates that increased minimum wage coverage also raises long term unemployment rates. Subsequent analysis yielded similar patterns for other aggregate labour market measures. Thus, state and federal policy makers should weigh these potential costs in deciding whether to increase minimum wage rates in the future.

TRANSCRIPT

Page 1: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Regional Studies, Vol. 33.8, pp. 713± 726

Do Minimum Wage Hikes Raise US Long TermUnemployment? Evidence Using State Minimum

Wage Rates1

MARK D. PARTRIDGE* and JAMIE S. PARTRIDGE ²*Department of Economics, Stewart Hall, St Cloud State University, St Cloud, MN 56301-4498, USA

² Department of Management, St John’s University/College of St Benedict, St Joseph, MN 56374, USA

(Received July 1998; in revised form March 1999)

PARTRIDGE M. D. and PARTRIDGE J. S. (1999) Do minimum wage hikes raise US long term unemployment? Evidenceusing state minimum wage rates, Reg. Studies 33, 713± 726. Several recent studies have challenged the conventional notion that

raising the minimum wage reduces employment. This study considers a related but relatively unexplored issue by examining

the minimum wage’s in¯ uence on long durations of unemployment. By considering long term unemployment rates, this studyextends the previous minimum wage literature by examining the persistence of minimum wage eVects. The empirical analysis

considers state data from the latter 1980s, a unique period when many states raised their minimum wage above the federal level.

The results suggest that a greater minimum wage increases long term unemployment rates. Further evidence indicates thatincreased minimum wage coverage also raises long term unemployment rates. Subsequent analysis yielded similar patterns for

other aggregate labour market measures. Thus, state and federal policy makers should weigh these potential costs in deciding

whether to increase minimum wage rates in the future.

Long term unemployment Minimum wage Minimum wage coverage Unemployment

PARTRIDGE M. D. et PARTRIDGE J. S. (1999) Les retom- PARTRIDGE M. D. und P ARTRIDGE J. S. (1999) FuÈ hrenbe es des augmentations du salaire minimum sur le choà mage Anhebungen von MindestloÈ hnen in den Vereinigten Staaten

de longue dure e? Des preuves provenant des salaires mini- zu langfristigem Ansteigen der Arbeitslosigkeit? Beweise, die

mum aux Etats-Unis, Reg. Studies 33, 713± 726. Des e tudes sich auf staatliche Mindestlohnraten stuÈ tzen, Reg. Studies 33,re centes ont mis en question l’ide e recË ue qu’une augmenta- 713± 726. Verschiedene, kuÈ rzlich veroÈ Ventlichte Unter-

tion du salaire minimum entraõà ne une re duction de l’emploi. suchungen stellten die herkoÈ mmliche Meinung in Frage, daû

Cette e tude-ci cherche aÁ conside rer un sujet connexe mais ein Anheben der MindestloÈ hne zur Herabsetzung der Zahlrelativement inexplore ; autrement dit, les retombe es du salaire der BeschaÈ ftigten fuÈ hrt. Dieser Aufsatz zieht eine damit

minimum sur le choà mage de longue dure e. En examinant verbundene, doch relativ selten untersuchte Frage in

les taux de choà mage de longue dure e, cette e tude e largit la Betracht, indem er den Ein¯ uû des Mindestlohnes auf langdocumentation ante rieure qui porte sur le salaire minimum anhaltende Arbeitslosigkeit pruÈ ft. Mit Hilfe der langfristigen

en conside rant la persistance des retombe es du salaire mini- Arbeitslosigkeitsraten erweitert die Studie durch Unter-

mum. On analyse des donne es qui proviennent de la ® n des suchung der anhaltenden Auswirkungen von MindestloÈ hnenanne es 1980, une e poque exceptionnelle ouÁ bon nombre des die Literatur uÈ ber MindestloÈ hne. Die empirische Analyse

e tats ont augmente leur salaire minimum au-dessus du niveau zieht oYzielle Datan der spaÈ ten achtziger Jahre heran, einer

fe de ral. Les re sultats laissent supposer qu’un renche rissement einmaligen Periode, insoweit als viele Staaten ihre Min-du salaire minimum entraõà ne une hausse des taux de choà mage destloÈ hne uÈ ber die auf Bundesebene gezahlten anhoben. Die

de longue dure e. Des preuves supple mentaires laissent voir Ergebnisse legen nahe, daû ein hoÈ herer Mindestlohn zum

qu’un accroissement de la proportion de smicards ameÁ ne Ansteigen der langfristige Arbeitslosenraten fuÈ hrt. Weitereaussi aÁ une hausse des taux de choà mage de longue dure e. Beweise lassen erkennen, daû vermehrte Mindestlohn-

Une analyse ulte rieure a fourni des re sultats comparables deckung auch langfristige Arbeitslosigkeitsraten anhebt. Die

pour ce qui est des autres mesures globales du marche du anschlieû ende Analyse ergab aÈ hnliche Muster fuÈ r andere, dentravail. Ainsi, les de cideurs, et au niveau de l’e tat et sur le Gesamtarbeitsmarkt umfassende Maû nahmen. Auf staatlicher

plan fe de ral, devraient e valuer ces couà ts potentiels au moment wie auf bundesstaatlicher Ebene sollten politische Entschei-

ouÁ ils de cident si, oui ou non, il faudra augmenter les taux dungen deshalb diese potentiellen Kosten mit in Betrachtde salaire minimum. ziehen, wenn sie zukuÈ nftige Anhebungen von Min-

destloÈ hnen erwaÈ gen.

Choà mage de longue dure e Salaire minimumProportion de smicards Choà mage Langfristige Arbeitslosigkeit Mindestlohn

Mindestlohndeckung Arbeitslosigkeit

0034-3404/99/080713-14 ©1999 Regional Studies Association

Page 2: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

714 Mark D. Partridge and Jamie S. Partridge

INT ROD UCT I ON exceeded 20%, where the 20% ® gure was topped in

only one other year since World War Two (1983)(PARTRIDGE and R ICKMAN, 1998). Longer durationsSince its original enactment in the 1930s, federal mini-

mum wage increases have been viewed by many as an of unemployment likely reinforced the seemingly high

levels of job anxiety felt by workers during the 1990seVective avenue to redistribute income and reduce

poverty (FREEMAN, 1996) as well as to provide a (see AARONSON and SULLIVAN, 1998, for more

details of this anxiety). By comparison, the share ofliving wage’ (KUTTNER , 1997). Given that in¯ ationerodes the purchasing power of a ® xed minimum wage, long term unemployed in total unemployment was less

than 10% in the latter 1980s. This raises an interestingevery few years there are calls for further increases.

With US income inequality rising since the 1970s, the question: did the 1990± 91 increase in the federal mini-

mum wage contribute to longer durations of un-policy implications of minimum wage hikes are even

more signi® cant. In fact, many economists have argued employment? In addition, the onset of federal welfare

reform means that states will carry a greater share ofthat the declining purchasing power of the minimumwage since the 1970s can account for about 30% of the income redistribution burden. One implication is

that state policy makers may feel pressure to boost theirthe ensuing increase in earnings dispersion (FORTIN

and LEMIEUX, 1997). minimum wage to lift poor families out of poverty.

Given that many chronic welfare recipients can also beYet, economists have traditionally cautioned that

raising the wage ¯ oor has deleterious consequences viewed as outsiders, the relationship between long term

unemployment and state minimum wage rates mayincluding additional unemployment, reduced generaltraining and oVsetting reductions in fringe bene® ts. shed light on this topic as well.

In what follows, the next section sketches someBROWN et al.’s, 1982, survey of the literature indicated

that a 10% increase in the minimum wage would on simple minimum wage models. The third section dis-

cusses empirical implementation and the fourth sectionaverage reduce teen employment by 1± 3%, raise teen

unemployment rates by 0± 0´75%, with smaller eVects discusses the results. In addition, section ® ve will con-sider the in¯ uence of state minimum wage rates onfor adults. Such negative eVects temper the desired

redistribution. None the less, US economists rarely, if aggregate unemployment rates. This additional analysis

will examine the robustness of the long term un-at all, discuss minimum wage eVects on unemployment

duration. employment rate results and will be of interest in their

own light. The ® nal section presents some concludingRegardless, as FREEMAN, 1995, notes, it is possible

that minimum wage legislation can create a class of thoughts.permanent outsiders who cannot obtain employment,

even though they would be willing to work at a rateT HE MI NI MUM WAG E A ND T H E

below the minimum wage. In fact, S IEBERT, 1997,L A B OUR MA RKE T

contends that high minimum wage rates are a key

factor behind high levels of structural and long term The need to better understand all minimum wageunemployment in Europe. Yet, the possibility that eVects is highlighted by a possible breakdown in theminimum wage hikes increase long term unemployment consensus among economists regarding the minimumis a relatively unexplored topic for the US. However, wage during the 1990s. Foremost is a series of cross-CLARK and SUMMERS, 1990a, ® nd that most jobless- sectional and case studies by CARD and KRUEGER

ness in the US is composed of a hard core group of (e.g. 1992, 1994, 1995) which found that the 1990± 91unemployed who are jobless for more than six months, increase in the federal minimum wage had virtually noeven though the typical person exits unemployment impact on employment. Similarly, D ICKENS et al.,rather quickly (less than three months). It seems plaus- 1999, and MACHIN and M ANNING, 1994, found noible that members of this hard core group would be negative minimum wage eVects for UK employment.potentially susceptible to job loss due to minimum However, recent ® ndings that minimum wage hikeswage hikes, while ® nding a new job would be more have very little eVect on employment have not gonediYcult.2 Therefore, this study will investigate the unchallenged. For example, using a variety of data setsimpact of state minimum wage laws on the prevalence and techniques, NEUMARK and WASCHER, 1992,of long term unemployment (de® ned as greater than 1994; DEERE et al., 1995; K IM and TAYLOR, 1995;26 weeks). CURRIE and FALLICK , 1996; and PARTRIDGE and

One reason for the lack of emphasis on US long PARTRIDGE , 1999, have found that increases in theterm unemployment is that is has been viewed as a minimum wage reduce employment, while PART-

relatively small problem. However, this may be chang- RIDGE and PARTRIDGE , 1998, found that theying. The growing importance of long term unemploy- increase teen unemployment rates.ment in the US can be seen by noting that, between Given the con¯ icting minimum wage ® ndings in the

1992± 94, the (annual) share of unemployed workers literature, it is not surprising that there are numerous

models of how minimum wage legislation aVects thewho had sought employment for more than 26 weeks

Page 3: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 715

labour market (e.g. see M INCER , 1976; BROWN et al., superior to examining changes in unemployment rates

(or vice versa), suggesting a need to consider all labour1982; CARD and KRUEGER , 1995). It is not possible

to do a complete survey, but it is useful to outline some market indicators.

Standard labour market theory also suggests that aof the major themes in these models.The standard minimum wage model is the basic greater minimum wage can increase long term un-

employment. That is, if less skilled workers with a lowsupply and demand model. In its simplest form, impos-

ing a wage ¯ oor above the equilibrium wage reduces value of marginal product (with realistic reservation

wages) cannot be oVered a wage below the minimumthe quantity of labour demanded while inducing new

entrants into the labour market. Two outcomes are an wage by employers, their probability of receiving a joboVer will diminish and their duration of unemploymentincrease in the unemployment rate and a reduction in

employment. However, real world complexities will increase. Using 1988 as a representative year for a

stylistic example, the median duration of unemploy-complicate the analysis. First, minimum wage coverage

is incomplete. A greater minimum wage may induce ment was 5 9 weeks, roughly corresponding to a cons-

tant 0´11 weekly hazard rate for exiting unemployment.workers from the uncovered sector to queue for jobs in

the covered sector, while previously employed workers Suppose that a higher minimum wage slightly reducesthis to 0 10. This modest change would increase thefrom the covered sector may be forced to ® nd work in

the uncovered sector. Moreover, some workers may be probability by one-third that an individual’s unemploy-

ment spell would last at least 26 weeks, which isso discouraged by diminished employment chances that

they exit the labour force (and hence are not counted consistent with long term unemployment rates being

more cyclically responsive than short term unemploy-as oYcially unemployed). Further complicating the

analysis is that ® rms may reallocate their shares of full- ment rates (CLARK and SUMMERS, 1990a). Moreover,if there is negative duration dependence where theand part-time labour (RESSLER et al., 1996) such that

employment changes do not accurately re¯ ect actual hazard rate declines over time, a minimum wage

increase would further raise the chance of a longchanges in total labour inputs (WELCH , 1995). Finally,

® rms may substitute more skilled labour for less skilled unemployment duration.3 Also, THOMAS, 1996, ® nds

that the length of unemployment spells is negativelylabour, which further confounds the eVects on totalemployment. related to the previous job’s wages. This suggests that

low skilled, low wage workers are even more susceptibleThe monopsony labour market model has recently

received attention in the minimum wage debate to longer unemployment durations. Since these lower

skilled workers are the very people minimum wage(CARD and KRUEGER, 1995). In this case, if an

employer is (labour) supply constrained, minimum hikes are hoped to bene® t, greater unemployment

duration seems especially counter to the policy goals.wage increases can yield employment gains. Card andKrueger contend that information imperfections and A key advantage of considering the long term un-

employed is that it seems much less likely that a potentialmobility costs make the monopsony model applicable

in low wage labour markets. However, many econom- worker induced into the labour market, by say a 10%

minimum wage increase, would remain in the labourists question the general applicability of the monopsony

model in the typical labour market. force for more than six months without a job (i.e.

continuously searching for work for such a small gain).Therefore, the employment outcome of increasingthe minimum wage is hard to predict. Standard models That is, any increased long term unemployment should

be almost entirely concentrated among those whogenerally imply a reduction in the number of labour

hours, but not necessarily total employment, while a would have been in the labour force without a mini-

mum wage hike.monopsony model can imply the opposite. The relative

size and substitution elasticities in the covered and Despite the above discussion, the minimum wage

may have very little in¯ uence on long term unemploy-uncovered sector further confound employment pre-dictions. Changes in total employment are also compli- ment if industries that employ a disproportionate share

of minimum wage workers have almost no unemploy-cated by changes in the prevalence of part-time

employment and average hours of work as well as ment to begin with. However, using 1988 Bureau of

Labor Statistics data, this does not appear to be thechanges in skill intensities. Likewise, the quantity of

labour supplied may increase if new workers are case. The overall 1988 unemployment rate was 5 5%.Yet, in retail, where over one-half of minimum wageinduced by the greater wages, or decrease if workers

become discouraged and exit the labour force. Thus, workers are concentrated (SMITH and V AVRICHEK,

1992), the unemployment rate was 6 1%. It was anunemployment rates may not necessarily increase.

However, if workers from the uncovered sector queue even higher 8 9% for eating and drinking establishment

workers, which is where retail sector minimum wagefor employment in the covered sector, these search

costs add to the social waste of increasing the minimum workers are especially concentrated. Relatively highunemployment rates are found in other industrieswage (M INCER, 1976). One implication is that, in

determining the impact of minimum wage laws, it is where the minimum wage is relatively important:

apparel, 8 1%; food and kindred industries, 8´2%; andnot clear whether examining employment changes is

Page 4: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

716 Mark D. Partridge and Jamie S. Partridge

private household services, 6 8%. Hence, unemploy- (e.g. see CLARK and SUMMERS, 1990a, regarding the

relationship between non-employment and unemploy-ment rates are relatively high in industries where the

minimum wage plays a key role, suggesting that mini- ment spells). Another feature of our long term un-

employment rate measure is that, to be considered longmum wage increases can, in turn, extend unemploy-ment durations. Conversely, older displaced workers term unemployed, a jobless worker must continuously

search for work for six months, which means that wewould appear unlikely to be in¯ uenced by minimum

wage increases. Yet, even here, because displaced are focusing on individuals who are the most attached

to the labour force. Thus, a rise in our long termworkers earn about 25% less on average up to ® ve years

after displacement ( JACOBSON et al., 1993), minimum measure suggests that at least a portion of the labourforce (who are quite attached) has experienced ratherwage rates may in¯ uence displaced workers who are in

the tail of the wage loss distribution.4 signi® cant increases in search costs and other costs. In

later analysis, we will also consider more aggregateSince we will examine variations in minimum wage

laws caused by state legislation, not federal legislation, unemployment, employment/population, and labour

force participation rates to examine the robustness ofthere are other conceptual considerations when making

predictions. Foremost, there is little incentive for ® rms the long term unemployment results.Regardless of whether the competitive model or theor households to relocate or migrate in response to

federal minimum wage changes because there are no monopsony model primarily applies, the unemploy-

ment rate is determined by the interaction of labourcross-state variations. However, state minimum wage

changes can create incentives for cross-state ® rm and supply and labour demand in each state. Our empirical

speci® cation will re¯ ect a reduced form model ofhousehold relocation, as well as aVect the likelihood of

® rm births and deaths. One implication is that mini- these relationships. Factors that increase labour demandshould reduce unemployment rates, while factorsmum wage responses driven by state legislation may be

larger than if we were examining the responses due to associated with an increase in labour supply should

increase the unemployment rate. Although there arefederal changes. Another implication is that potential

® rm and household movements (or ® rm births and modest diVerences, the empirical speci® cation is gener-

ally consistent with NEUMARK and WASCHER ’s,deaths) may occur after a lag, suggesting that stateminimum wage responses will be more drawn out. 1992, 1994, and D EERE et al.’s, 1995, state-level exam-

ination of minimum wage hikes and it is consistent with

PARTRIDGE and R ICKMAN ’s, 1995, 1997, analysis ofE MP IRICA L MOD E L

state-level unemployment.

The data set covers the 48 contiguous states over theOur empirical methodology will be similar to

NEUMARK and WASCHER ’s, 1992, use of cross-state 1984± 89 period. Data from the late 1980s has sub-stantial advantages in sorting out the impact of mini-diVerences in minimum wage levels and coverage, while

correcting for many of the empirical criticisms levelled mum wage hikes. For example, WELCH and

CUNNINGHAM, 1978, p. 144, noted that high rates ofby CARD et al., 1994; and CARD and KRUEGER,

1995. Yet, the major diVerence between this study and federal minimum wage coverage along with the rela-

tively high federal minimum wage rate empiricallymost other minimum wage studies (including Neumark

and Wascher) is that this study does not consider swamped any variation in state minimum wage laws in1970. This changed in the late 1980s when many statesteen employment. That is, by considering long term

unemployment rates, this study extends the minimum raised their minimum wage in response to the federal

minimum wage remaining unchanged between 1981wage literature.

For the purposes of this study, long term unemploy- and 1990. The result is that only one contiguous state

had a minimum wage above the federal level in 1984ment is de® ned as being unemployed for more than

26 weeks. Hence, the long term unemployment rate is (Connecticut), while this increased to 13 states by 1989(along with Alaska, District of Columbia and Hawaii).de® ned as the number of individuals unemployed for

more than 26 weeks divided by the total labour force as This yields both time-series and cross-sectional vari-

ation in minimum wage levels. However, in April 1990,de® ned by the Current Population Survey (CPS).5 One

trait of CPS data is that unemployment rates are derived the federal minimum wage increased from $3.35 to

$3.80, greatly reducing cross-sectional variation.6 Like-from ongoing spells of joblessness. For example, someof the short term unemployed may eventually become wise, the 1984± 89 period was a period of national

economic expansion, while 1990 represented thelong term unemployed or already would be long term

unemployed if they had continuously remained in the beginning of a national recession, which can confound

the estimates.labour force. Another implication is that the long term

unemployed may disproportionately exit the labour The variables are described below while the data

sources are fully detailed in Table 1. Note that oneforce by discontinuing their job search. These examplestend to mean that potential increases in long durations advantage of the data is that they are based on annual

averages. Conversely, Neumark and Wascher’s dataof non-employment due to minimum wage laws would

be understated by our long term unemployment measure were based only on May CPS estimates, where small

Page 5: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 717

sample sizes for less populated states can introduce et al. contended that instead of a Kaitz index, it is

better to use the minimum wage rate itself or thevariable measurement error, which may have signi® -

cantly aVected their results (CARD et al., 1994). fraction of workers aVected by a minimum wage

increase.7The basic speci® cation for state s in year t is:Hence, the log of the maximum of the state or

LOUTCOMEst5 bLMINWAGEst federal minimum wage is used as the minimum wage

measure.8 Other advantages of directly using the mini-1 aLCOVERAGEst 1 pLWAGEst

(1)

mum wage include easier interpretation and less

measurement error. Many researchers only include the1 d INDMIX EMP GROWTHstcontemporaneous minimum wage rate because mini-

mum wage jobs are perceived to have high turnover1 F AGEst 1 ss 1 s t 1 est

rates (especially for teens ± see BROWN et al., 1982).where: LOUTCOME is the log of the labour market However, we include the lag minimum wage rate foroutcome of interest; LMINWAGE is a vector of log several reasons. First, it may take time for employers tominimum wage variables; LCOVERAGE is the log substitute more skilled for low skilled labour or substi-share of the state’s non-supervisorial labour force tute capital for low skilled labour (NEUMARK andcovered by federal minimum wage legislation; LWAGE WASCHER, 1992, 1994). For example, BROWN,is the log average hourly manufacturing production 1995, and HAMERMESH , 1995, contend that CARD

worker wage; INDMIX EMP GROWTH is the state’s and KRUEGER, 1995, did not ® nd these negative inputemployment growth rate if all its industries grew at the substitution eVects on employment because they likelynational growth rate; AGE is a vector of demographic take more than a year to develop, which is longer thanage groups; ss is the state ® xed eVect; s is the year Card and Krueger allowed.9 Second, as noted earlier,® xed eVect; e is the residual term. b, a, p, d , and F it may take additional time to realize the adverse state-are coeYcient vectors. The state ® xed eVects control level ® rm and household relocation eVects after a statefor unmeasured factors that in¯ uence each state’s labour minimum wage increase (including fewer ® rm birthsmarket. One primary factor accounted for by the state and more ® rm deaths).® xed eVects is the possibility that states which raised Third, because we are using incomplete unemploy-their minimum wage were systematically diVerent than ment spells, the previous year’s minimum wage maythe other states. The year ® xed eVects account for play a role in extending the duration of unemploymentnational economic and demographic factors that have and increasing the current year’s long term unemploy-a common eVect across all states. Following DEERE ment rate. That is, if a worker loses his/her job, itet al., 1995, the base speci® cation will be in double- would take at least six months of unemployment tolog form for the variables of interest (see below). aVect the long term unemployment rate, greatly

The proper minimum wage measure is in some reducing the contemporaneous minimum wage eVect.dispute. In this regard, NEUMARK and WASCHER, Moreover, note that the minimum wage is measured1992, used a state-level Kaitz index. The Kaitz index is as an annual average (see note 8). States often increasedthe product of the share of the state’s non-supervisorial their minimum wage rate eVective after 30 June. Thisworkforce covered by the federal minimum wage with means that almost all the long term unemployment ratethe maximum of the federal or the state minimum eVects would be delayed into the next year.wage rate divided by the average hourly wage. This Generally, the competitive model suggests positivemeasure has the advantage of adjusting for variation in minimum wage coeYcients, while a (supply con-state minimum wage coverage. However, one assump- strained) monopsony model would generally predicttion behind the Kaitz index is that a 10% increase in negative (or zero) coeYcients. However, the eVectsminimum wage coverage has the identical eVect of a just described suggest that the lagged minimum wage10% increase in the nominal minimum wage rate as coeYcient will more likely be positive than the con-well as the exact opposite eVect of a 10% increase in temporaneous minimum wage eVect. Further, as dis-the average wage. NEUMARK and WASCHER, 1994, cussed in more detail below, NEUMARK andacknowledge that there is no basis for this restriction, WASCHER, 1992, 1994; and PARTRIDGE and PART-

although intuitively the minimum wage should be less RIDGE , 1998, 1999, ® nd results that are consistentbinding in high wage states or in states with lower with state policy makers considering current economiccoverage. Yet, CARD et al., 1994, found that Neumark conditions in their minimum wage rate decision. Spe-and Wascher’s Kaitz index was negatively related to teen ci® cally, state policy makers may be more likely towage rates, suggesting that it is a poor wage measure. increase the minimum wage when they believe goodCard et al., argued that this problem was caused when economic conditions in the near term will oVset anystates that had more economic growth also had higher adverse minimum wage eVects.10 For example, in theoverall average wage rates and greater teen employ- latter 1980s, state minimum wage hikes were concen-

ment. This can create a spurious negative correlation trated in the Northeast and along the Paci® c Coast, or

states that greatly bene® ted from the Reagan defencebetween the Kaitz index and teen employment. Card

Page 6: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

718 Mark D. Partridge and Jamie S. Partridge

build up. In fact, this cyclical eVect suggests that the state male employment rate (KUTTNER, 1997). That

is, Deere et al. attributed the eVects of the 1990± 91contemporaneous minimum wage coeYcient may be

negative, further suggesting that our results understate federal minimum wage increase to year dummy

coeYcients for 1990, 1991 and 1992. Yet, if the (lower-the minimum wage in¯ uence on long term unemploy-ment rates. skilled) demographic groups examined by Deere et al.

have greater cyclical variations in their employmentMinimum wage coverage is incomplete due to

exemptions for industry and ® rm size (which leads to rates, the male employment ratio may inadequately

account for the eVects of the 1990± 91 recessioncross-state variation). BROWN et al., 1982; and

BROWN, 1988, strongly argue for coverage’s inclusion (KUTTNER, 1997). Hence, their 1990± 92 year indi-cators may be capturing cyclical eVects and overstatingbecause of its importance in the standard model,

although CARD and KRUEGER, 1995, downplay its the negative impacts of the minimum wage hike.

Finding suitable cyclical controls is diYcult becauseimportance. Therefore, the log of federal coverage of

minimum wage laws at the state level is also included unemployment and cyclical conditions are jointly

determined. Thus, as a labour demand shifter, we usein the speci® cation. It would be optimal to also include

binding minimum wage coverage rates due to state the (two-digit level) private sector, non-farm industrymix employment growth rate from shift-share analysislaws, but such data are unavailable for our sample

period. None the less, our state-level coverage data (PARTRIDGE and R ICKMAN, 1995). The industry

mix employment growth rate re¯ ects how much thesigni® cantly improve on the coverage data used by

NEUMARK and WASCHER, 1992, 1994. Due to data state’s employment would grow if all its industries grew

at their respective national growth rates, i.e. it measuresavailability at the time of Neumark and Wascher’s study,

coverage data were unavailable for many years of their whether the state has a mix of fast or slow growingindustries. Given that national industry growth ratessample, most notably 1987 to 1989. However, we have

obtained coverage estimates for every year of our are determined primarily at the national level, industry

mix employment growth should not be endogenous.study.11 This is fortunate because the US Department of

Labor ceased publishing state-level estimates of federal In fact, BARTIK , 1991, and BLANCHARD and KATZ,

1992, used it as an exogenous instrument in theirminimum wage coverage after 1990.The natural log of the average production worker regional economic analysis. Indeed, PARTRIDGE and

R ICKMAN, 1995, 1997, found that industry mixwage is included in the model. As indicated when

discussing the Kaitz index, a minimum wage should be employment growth is strongly related to state un-

employment rates, where its coeYcient re¯ ects anyless of a constraint in states with a relatively high

wage structure (suggesting a negative average wage multiplier eVects from national demand shifts. More-

over, among other things, the state ® xed eVects controlrelationship with long term unemployment). More-over, CARD and KRUEGER, 1995, noted that the for persistent growth diVerences across states, such as

rapid growth in the sunbelt. Finally, the year ® xedprice of a substitute should be included in the model,

where they also used average production worker hourly eVects account for national cyclical (or demographic)

trends common across all states.wage (p. 185, p. 199). None the less, the production

worker wage is likely to be endogenously related to The potential labour supply of groups that dispropor-

tionately compose minimum wage earners is includedthe state’s cyclical conditions (FREEMAN, 1982), sug-gesting that a reduced form model that omits produc- in the model. Minimum wage earners are primarily

teens, young adults and older workers (over 55)tion worker wages would provide more accurate

estimates (where the state ® xed eVects would capture (M ELLOR, 1987; SMITH and VAVRICHEK , 1992).

Thus the model includes the share of the state’s popula-persistent wage level eVects). Below, we will also use

the average private sector hourly wage to consider the tion between 15 and 19 years old; 20 and 24 years old;

55 and 64 years old; and 65 and older.robustness of our wage measure choice, although thismeasure is probably even more endogenous than pro-

duction worker wages.E MP IRICA L RE S ULT S

Estimating minimum wage eVects can be confoun-

ded if the state’s cyclical conditions are not adequately Table 1 presents the (unweighted) descriptive statistics

used in the empirical model. For example, the averagecontrolled for (CARD et al., 1994). DEERE et al., 1995,criticized CARD ’s, 1992, cross-state analysis for not state long term unemployment rate was 0 91%, the

average total unemployment rate was 6 3%, while 85%accounting for low wage Sunbelt states (where the

federal minimum wage was most binding) growing of non-supervisorial employees were covered by federal

minimum wage legislation.faster than the national average in 1990 and 1991. This

can make it appear as though the increased federal In 1989, 13 of the contiguous states had a minimum

wage above the federal rate of $3.35. In these 13 statesminimum wage had no negative employment eVectson these states.12 Similarly, the analysis by DEERE et al., (not shown), on average only 0 33% (s.d.5 0 12) of the

labour force was unemployed for more than 26 weeks,1995, of 1985± 92 state-level employment was criticized

for only controlling for cyclical eVects with the overall while the total average unemployment rate was 4 4%

Page 7: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 719

Table 1. Descriptive statistics for the 48 contiguous states, model shown in equation (1), while columns (2)± (6)

contain alternative speci® cations.1984± 891

In column (1), the contemporaneous eVect fromMean (Std dev)

raising the minimum wage is negative and statisticallyLong term total unemployment rate 0 91 signi® cant, but this is more than oVset by the positive(GPEU) (0 69) and statistically signi® cant lag eVect.13 As describedTotal unemployment rate 6 3

above, one possible reason for this dynamic pattern is(GPEU) (2 1)that states which raised their minimum wage did soMale unemployment rate 6 2

(GPEU) (2 2) because they were more likely to experience higherFemale unemployment rate 6 5 economic growth. Such a tendency would result in a(GPEU) (2 1) negative (contemporaneous) relationship between theTotal employment/population 62 0

minimum wage and the long term unemployment rate,(GPEU) (4 6)where standard eVects would occur after a lag. Also,Total labour-force/population 66 1

(GPEU) (3 8) since the long term unemployed must be out of workMinimum wage2 3 39 for at least six months, negative minimum wage eVects(MLR) (0 12) would be further delayed. This would mean that ourMinimum wage coverage ratio 0 85

contemporaneous estimate understates any unemploy-(MWMH) (0 05)ment increases that result from raising the state mini-Production worker hourly wage 9 66

(EE) (1 24) mum wage. Conversely, this pattern could be consistentAverage private sector hourly wage 9 21 with CARD and KRUEGER ’s, 1995, monopsony(GPEU, WE) (1 28) model with mobility costs and search frictions playingIndustry mix employment growth 0 030

a strong role in the short run (D ICKENS et al., 1999)(BEA) (0 012)but with the standard model dominating over time.Share 15± 19 0 077

(USDC) (0 005) Summing the minimum wage coeYcients results inShare 20± 24 0 079 an elasticity point estimate of 1´81. Yet, the F-test(USDC) (0 007) reported at the bottom of Table 2 shows that we canShare 55± 64 0 089

reject the null hypothesis that the sum of the two(USDC) (0 008)minimum wage coeYcients equals zero at only theShare 65 1 0 125

(USDC) (0 018) 17% level (p 5 0 166), suggesting that caution shouldN 288 be exercised in interpreting the results. Two possible

causes of this imprecision are that state-level long termNotes: 1. The sources are listed in parentheses: BEA, Bureau ofunemployment duration is measured with error in theEconomic Analysis; USDC, US Department of Commerce,

Bureau of the Census, Current Population Survey Data; CPS and endogeneity of average wages.MWMH, US Department of Labor, Minimum Wage and Using the estimated minimum wage elasticity ofMaximum Hours Under the Fair Labor Standards Act; EE, US

1´81, a 27% increase in the minimum wage (whichDepartment of Labor, Employment and Earnings, States and

equals the federal increase between 1990 and 1991)Areas; GPEU, US Department of Labor, Geographical Pro® lewould increase the average state long term unemploy-of Employment and Unemployment; MLR, US Department

of Labor, Monthly Labor Review, January issues; WE, US ment rate to about 1´35% when evaluated at theDepartment of Labor, Wage and Earnings, Annual Averages. 1984± 89 sample mean long term unemployment rate2. The minimum wage is the maximum of the state and

of 0´91% (or by about one-half ). For comparison, thefederal minimum wage rates.

1989 national long term unemployment rate was only

0´52% of the labour force, while the average 1992± 94

long term unemployment rate averaged about 1 38%.(s.d.5 0 8). For the other 35 states the respective un-

These results suggest that the federal minimum wageemployment rates were 0 57% (s.d. 5 0 32) and 5 4%

increase of 1990± 91 contributed to the increase in long(s.d.5 1 36). This pattern is in accord with Card and

term unemployment in the ® rst half of the 1990s, butKrueger’s claim that modestly raising the minimum

note that these estimates are measured with somewage has little if any negative consequences on labour

imprecision. Yet, bear in mind that national levelmarket opportunities. None the less, as noted above,

responses are likely smaller than responses at the statethere could be other factors behind this relationship,

level.for which we turn to the regression analysis to sort out

Regarding minimum wage coverage, the results inthe causal eVects.

column (1) indicate that its coeYcient is positive and

statistically signi® cant at the 7% level. In fact, the

magnitude of the coverage estimate is rather large,Long term unemployment

suggesting that CARD and KRUEGER, 1995, were tooquick to ignore coverage’s eVect. A ® ve percentageTable 2 contains several alternative regression speci® ca-

tions using the log of the long term unemployment point increase in minimum wage federal coverage

would raise the average long term unemployment raterate as the dependent variable. Column (1) re¯ ects the

Page 8: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

720 Mark D. Partridge and Jamie S. Partridge

Table 2. Log long term unemployment rate regression results (absolute value of t-statistics)1

(1) (2) (3) (4) (5) (6)

Log(minimum wage) 2 2 93 2 2 99 2 0´57 2 3 04 2 2 89 2 2 90

(3 45) (3 65) (0 68) (3 19) (3 39) (3´35)

Log(minimum wage)t2 1 4 74 4 79 5 72 4 65 5 02

(2 69) (2 88) (3 09) (2 62) (2´79)

Log(minimum wage coverage) 3 42 5 55 3´08 2 22 3 16

(1 84) (2 79) (1 64) (1 20) (1 80)

Log(minimum wage coverage)t2 1 0 66

(0 57)

Log(production hourly wage) 2 1 52 2 2 35 1´70 2 1 44 2 1 60

(1 51) (1 95) (2 29) (1 41) (1´50)

Log (production hourly wage)t 2 1 3 59 3 00 3 59 3 28

(3 59) (2 73) (3 59) (3´15)

Industry mix employment growth 2 15 07 2 15 81 2 15 21 2 16 33 2 14 91 2 15 54

(2 62) (2 22) (2 81) (2 85) (2 56) (2´71)

Share 15± 19 43 27 54 56 49 20 44 02 40 83

(2 63) (3 16) (3 06) (2 65) (2´41)

Share 20± 24 2 88 04 2 77 30 2 74 68 2 86 86 2 92 13

(5 58) (4 95) (5 02) (5 32) (5´95)

Share 55± 64 2 34 26 2 23 17 2 23 89 2 34 69 2 26 27

(1 72) (1 14) (1 21) (1 75) (1´32)

Share 65 1 2 38 11 2 42 57 2 41 91 2 38 28 2 36 56

(2 46) (2 67) (2 70) (2 47) (2´37)

Year ® xed eVects Y Y Y Y Y Y

State ® xed eVects Y Y Y Y Y Y

R2 0 89 0 87 0´88 0 89 0 89 0 89

F-stat: bMWt 1 bMW(t2 1) 5 02 1 93 1 83 4 66 1 81 2 67

(p5 0 166) (p5 0´177) (p5 0 032) (p5 0 180) (p 5 0 104)

F-stat: Kaitz restriction3 2´15 4 16

(p 5 0 119) (p5 0 017)

F-stat: Lag Kaitz restriction3 6 90

(0 001)

Autocorrelation q 4 0 114 0 181 0´122 0 140 0 113 0 114

N 288 288 288 288 288 288

Minimum wage coverage elasticity 3 42 5 55 3´08 2 22 3 82 n.a.

Minimum wage elasticity 1 81 1 80 2 0´57 2 68 1 76 2 12

Notes: 1. The t-statistics use the White correction for heteroscedasticity. Long term unemployment is de® ned as an unemployment duration

longer than 26 weeks.

2. The F-statistic for the null hypothesis that the sum of the minimum wage rate coeYcients equals zero.

3. The Kaitz restriction is that the coeYcient on the minimum wage variable equals both the coeYcient on the coverage variable and

the negative of the coeYcient on the average wage variable.

4 The ® rst-order autocorrelation of the within-state residuals.

to 1 07% when evaluated at the sample mean of 0´91% pattern suggests that the lagged minimum wage variable

plays an integral role in detecting the conventional(or by over one-sixth). This again supports the argu-

ment that minimum wage laws can increase long term unemployment outcome (as suggested by NEUMARK

and WASCHER, 1992, 1994).unemployment.By not including the age controls, the model in One of the reasons that the average production

worker wage is included in the base model is that it iscolumn (2) examines the sensitivity of the ® ndings to

accounting for the supply side controls. The results a measure of the degree to which the minimum wage

is a binding constraint in the state’s labour market. Yet,indicate that the coverage variable’s magnitude and

statistical signi® cance improved in this case (1% level), this likely introduces bias when production workerwages are endogenous. By omitting production workerwhile the minimum wage results are substantially

unchanged. wages, the model in column (4) re¯ ects a reduced form

speci® cation. These results suggest that the sum of theThe model in column (3) shows the results when

the lagged minimum wage variable is omitted from the minimum wage coeYcients is signi® cant at about the

3% level with the minimum wage elasticity rising tobase model in column (1). In this case, the contempor-

aneous minimum wage coeYcient is negative and 2´68. However, the coverage variable is no longersigni® cant.statistically insigni® cant, which is not surprising given

that it takes at least six months before long term In further sensitivity analysis (not shown), speci® ca-

tions similar to those in columns (3) and (4) wereunemployment rates would be aVected. Foremost, this

Page 9: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 721

estimated. In this case, the contemporaneous minimum ing the impacts of the minimum wage and, thus,

they did not include it in their model. Despite thewage variable and the production worker wage variables

were omitted from the model. Also, to focus just on signi® cance of the coverage variable, to examine this

possibility, column (6) reports results from the modelthe minimum wage rate, the coverage variable was alsoomitted (which had little in¯ uence on the results). In which omits coverage from the speci® cation in column

(1). The minimum wage elasticity is slightly larger inthis case, the lagged minimum wage elasticity was 2´01

with a t-statistic of 1 44. Likewise, in another model, this case and the joint signi® cance of the sum of

the minimum wage coeYcients is slightly improved1990 data were included. The lagged minimum wage

elasticity equalled 2 08 and the t-statistic equalled 1´85. (p 5 0 104 vs. p 5 0 166).Table 3 highlights additional sensitivity regressionsHowever, because the federal minimum wage increase

took eVect in April 1990 and a recession began in that were estimated to examine the robustness of the

model shown in column (1) of Table 2. Row (1) is a1990, these results should be cautiously interpreted.

Finally, a corresponding linear model (not log) was two-stage least squares model (2SLS) treating produc-

tion worker wages as endogenous.14 Row (2) adds theestimated (without 1990 data). The minimum wage

elasticity equalled 2 62 with a t-statistic of 2 36. Overall, lag of the industry mix employment coeYcient toaccount for dynamic economic eVects correlated withthis sensitivity analysis suggests that the positive lagged

minimum wage response is robust and not an artefact long term unemployment. Row (3) adds the overall

employment/population ratio as an additional cyclicalof including the contemporaneous minimum wage

variable. control to examine whether adding a clearly endogen-

ous cyclical variable to the model confounds the results.The speci® cation in column (5) adds the lag of

minimum wage coverage to the model in column (1). Finally, row (4) adds the log average private sectorhourly wage in place of the average production workerThe lagged coverage variable’s t-statistic is only 0´57,

supporting our contention that there is very little lagged hourly wage.

The consistent pattern in Table 3 tends to aYrmcoverage eVect. Conversely, the contemporaneous cov-

erage coeYcient is little changed from before as well the ® ndings in Table 2. First, the contemporaneous

minimum wage eVect remains negative and generallyas the two minimum wage coeYcients. The bottom ofTable 2 shows the results of an F-test for separately statistically signi® cant. The lag minimum wage co-

eYcient is positive and generally statistically signi® cant,imposing the Kaitz restriction contemporaneously and

with a lag (on the minimum wage, coverage and where the magnitude of the lag coeYcient more than

oVsets the contemporaneous eVect. The consistentmanufacturing wage coeYcients). The null hypothesis

can be rejected at the 5% level in both cases (p 5 0 017; ® nding that the lag eVect is larger than the contempor-

aneous eVect lends additional support to this pattern.p 5 0´001) suggesting that the Kaitz restriction is inap-propriate. In both the contemporaneous and lagged Finally, the coverage elasticity ranges from 2 0 to 4 0.

Speci® cally, the results in the ® rst two rows of Table 3cases (not shown), the Kaitz coverage adjusted mini-

mum wage coeYcient would have been negative and are little changed from column (1) of Table 2. As

expected, adding the employment/population ratiostatistically insigni® cant (t-statistics of 2 0´54 and

2 0´29, respectively). Thus, imposing the Kaitz restric- (row (3)) aVects the ® ndings, where the minimum

wage eVect is diminished, but this is oVset by thetion would have given the impression of a negative(but insigni® cant) relationship between the minimum coverage eVect becoming more in¯ uential. Substituting

the average private sector hourly wage also lessens thewage and long term unemployment.

CARD and KRUEGER, 1995, argued that including minimum wage eVects in row (4). However, this could

be caused by the average private sector wage suVeringfederal minimum wage coverage adds little to estimat-

Table 3. Alternative log long term unemployment rate regression models (absolute value of t-statistics)1

(4)

(1) (2) (3) F-test

Log(minimum Log(minimum Minimum wage H0 :BMW 1 (5)

Model wage) waget2 1) elasticity BMW(t 2 1)5 0 Log coverage

(1) 2SLS, production worker wage endogenous 2 3 22 5´63 2 41 2 63 3 20

(1 57) (2 85) (p5 0 106) (1´76)

(2) Add lag industry mix employment growth 2 2 95 4´80 1 85 1 98 3 45

(3 45) (2 70) (p5 0 161) (1´83)

(3) Add log (total employment/population) 2 1 80 2´24 0 44 0 15 3 74

(2 31) (1 48) (p5 0 700) (2´84)

(4) Use log average private sector hourly wage 2 2 80 3´83 1 03 0 63 2 41

(3 35) (2 10) (p5 0 429) (1´38)

Note: 1. The models are the same as used in column (1) of Table 2 with the changes indicated in the row. Long term unemployment is

de® ned as an unemployment duration longer than 26 weeks.

Page 10: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

722 Mark D. Partridge and Jamie S. Partridge

from endogeneity and measurement error, suggesting Aggregate unemployment rates

that those results are possibly spurious.15

Given the fairly large long term unemploymentThe results in Tables 2 and 3 appear to consistently

response to changes in the minimum wage, minimumsuggest that greater minimum wage coverage increases

wage changes should have a perceptible impact onlong term unemployment rates. Likewise, they indicate

aggregate unemployment rates, especially if other formsthat the magnitude of the lagged minimum wage

of unemployment also increase. For example, D EEREcoeYcient more than oVsets the negative contemporan-

et al., 1995, provide evidence of large disemploymenteous minimum wage coeYcient (where both are eVects for at risk’ population subgroups, although it isusually statistically signi® cant at the 5% level). Yet, the reasonable to expect a smaller response for aggregatesum of the minimum wage coeYcients is in many cases unemployment than long term unemployment. Hence,only statistically signi® cant at the 10 to 20% level. to examine this possibility, as well as to examine theRegardless, the reduced form model without produc- robustness of the patterns found above, Table 4 reportstion worker wages (column (4) of Table 2) suggests a regression equations using three diVerent aggregatesigni® cant minimum wage response at the 5% level, state unemployment rates as the dependent variable.where there are reasons to more heavily weigh those The basic model is unchanged with the exception thatresults. the average manufacturing wage is omitted from these

As indicated above, state-level long term unemploy- speci® cations since it is quite likely that it is endogenousment rates are very diYcult to estimate precisely by when considering the aggregate labour market (i.e. itthe Bureau of Labor Statistics, which increases the is a reduced form model).noise to signal ratio in the regressions. Thus, before In Table 4, columns (1)± (3) respectively reportmaking any ® rm judgements, it would seem prudent regression results using the total, male and femaleto examine the minimum wage’s eVect on more aggre- unemployment rates as the dependent variable. Gener-gate unemployment, employment/population and ally, the aggregate results follow the same pattern aslabour-force participation rates, which are likely to be before ± the contemporaneous minimum wage is

negative and the lagged coeYcient is positive (wheremeasured more accurately.

Table 4. Aggregate labour market regression results (absolute value of t-statistics)1

(1) (2) (3) (4)

Log total Log male Log female Log employment/ (5)

unemployment rate unemployment rate unemployment rate population Log LF/population

Log(minimum wage) 2 1 49 2 1´52 2 1 45 0 13 0 054

(3 93) (3 30) (3 28) (1 86) (0 88)

Log(minimum wage)t2 1 2 88 3´19 2 42 2 0 34 2 0 166

(4 33) (4 12) (3 24) (3 06) (1 78)

Log(minimum wage coverage) 2 07 2´29 1 91 0 11 0 21

(2 97) (2 48) (2 99) (1 04) (2 55)

Industry mix employment growth 2 7 35 2 6´95 2 7 38 0 49 0 06

(3 16) (2 69) (2 70) (1 51) (0 19)

Share 15± 19 14 28 12´40 18 32 0 06 2 0 52

(2 06) (1 56) (2 31) (0 05) (0 47)

Share 20± 24 2 31 69 2 24´90 2 39 43 5 44 2 93

(5 50) (3 67) (6 27) (5 74) (3 78)

Share 55± 64 2 21 09 2 21´93 2 19 25 0 26 2 1 13

(2 91) (2 77) (2 25) (0 23) (1 29)

Share 65 1 2 28 87 2 22´62 2 36 98 2 79 2 0 05

(4 52) (2 85) (5 79) (3 25) (0 08)

Year ® xed eVects Y Y Y Y Y

State ® xed eVects Y Y Y Y Y

R2 0 92 0´91 0 91 0 97 0 97

F-stat: bMWt 1 bMW(t2 1) 5 02 8 54 8´77 3 55 9 03 3 76

(p5 0 004) (p 5 0 003) (p5 0 061) (p5 0 003) (p 5 0´054)

Autocorrelation q 3 0 129 0´077 0 065 0 239 0 223

N 288 288 288 288 288

Minimum wage coverage elasticity 2 07 2´29 1 91 0 11 0 21

Minimum wage elasticity 1 39 1´67 0 97 2 0 21 2 0 11

Notes: 1. The t-statistics use the White correction for heteroscedasticity. Columns (1)± (5) respectively have regression results for the total

unemployment rate, male unemployment rate, female unemployment rate, total employment/population percentage, and total labour

force/population percentage. The dependent variables are measured in natural log.

2. The F-statistic for the null hypothesis that the sum of the minimum wage rate coeYcients equals zero.

3. The ® rst-order autocorrelation of the within-state residuals.

Page 11: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 723

both are statistically signi® cant at the 1% level). The that, after a lag, people will exit the labour force as

they become discouraged by the negative minimumF-statistics at the bottom of Table 4 indicate that the

null hypothesis that the sum of the minimum wage wage eVects.

The minimum wage results in columns (4) andcoeYcients equals zero can be rejected at the 1% levelfor total unemployment and male unemployment and (5) are consistent with these expectations. The initial

response of the employment rate is positive, but this isat about the 6% level for female unemployment. There-

fore, the aggregate unemployment results are estimated more than oVset by a negative lagged minimum wage

response. Likewise, the labour force initially expands,more precisely than before. The minimum wage elasti-

city equals 1´39 for total unemployment, indicating but this is overwhelmed by people exiting the labourforce with a lag. The overall employment/populationthat a 10% increase in the minimum wage would raise

the average state’s unemployment rate about 0 9% minimum wage elasticity is 2 0 21 after two years

where the corresponding participation elasticity ispercentage points (to 7 2% on average). As expected,

all three aggregate minimum wage elasticity estimates 2 0 11. The joint F-statistics show that the overall

employment response is statistically signi® cant at theare smaller than the corresponding estimate for long

term unemployment (in column (1), Table 2). 0´3% level and the overall participation response issigni® cant at the 6% level. Together, columns (4) andMinimum wage coverage is positively related to all

three aggregate unemployment rates, where the eVect is (5) suggest that the aggregate unemployment rate rises

after a lag because employment falls faster than thestatistically signi® cant at about the 1% level. Again, the

estimated coverage elasticity is smaller for the aggregate labour force in response to a minimum wage hike. In

fact, using a derivation in BROWN et al., 1982, themeasures than for the long term unemployment rate.

As noted earlier, the ® nding of a favourable’, con- results in columns (4) and (5) suggest that a 10%state minimum wage increase will lift the aggregatetemporaneous, state-level, minimum wage response

that is more than oVset by an `unfavourable’ , lagged, unemployment rate by 0 9%, the same as in column

(1). Finally, the results in columns (4) and (5) suggestminimum wage response is not limited to this study.

For example, NEUMARK and WASCHER, 1992, 1994, that minimum wage coverage lifts aggregate unemploy-

ment by lifting the labour supply.found this pattern for teen (and 16 to 24 year old)employment/population ratios. PARTRIDGE and Overall, the aggregate results follow the same

dynamic pattern as the long term unemploymentPARTRIDGE, 1998, also found this same pattern for

teen unemployment rates. This was the case when results, further suggesting that the long term unemploy-

ment response is not a statistical artefact. In addition,they directly examined teen unemployment rates or

indirectly when they considered teen employment/ the precision of the aggregate coeYcients increases our

con® dence in the long term unemployment patterns.population rates and labour force participation rates.Finally, PARTRIDGE and PARTRIDGE , 1999, found Thus, it may be possible that the long term unemploy-

ment estimates were aVected by measurement error.the same dynamic pattern for retail and non-farm state-

level employment. As an additional characteristic of Nevertheless, beyond exploring the robustness of the

previous results, the aggregate ® ndings also suggest athese studies, note that unemployment, employment/

population and labour-force participation rates are from strong response to state-level minimum wage increases.

Given that these results are based on variations in statethe household CPS and retail and non-farm employ-ment are from establishment surveys, suggesting that minimum wage rates, these ® ndings are not surprising.

That is, diVerences in state minimum wage rates canthe dynamic minimum wage response is not an artefact

of a single survey. generate cross-sectional variations in ® rm births and

® rm deaths as well as shifts in state migration patterns.

Conversely, ® rms or individuals cannot relocate toAggregate employment/population and labour-force participa-

avoid federal minimum wage requirements, implying ation results

smaller federal minimum wage response.

To further investigate the robustness of the results,

columns (4) and (5) report regression results using theCONCL US ION

log employment/population ratio and the log labour

force participation rate as the dependent variables. If This study examined the impact of state-level minimumwage rates on unemployment rates, focusing on thethe above explanations are correct, the employment

rate should be positively related to the contemporan- long term unemployed over the 1984± 89 period. The

emphasis on long term unemployment extends mini-eous minimum wage rate due to strong unmeasured

economic conditions at the time of passage. This posi- mum wage research by considering how unemploy-

ment durations for attached low skilled workers aretive eVect would then be overwhelmed by a negative

lagged minimum wage response. Likewise, the initial aVected. By examining the 1984± 89 period, we tookadvantage of the large cross-sectional variation thatfavourable economic conditions along with a higher

minimum wage should induce more people into the occurred when many states raised their minimum wage

above the federal minimum rate of $3.35.labour force. However, the above explanations imply

Page 12: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

724 Mark D. Partridge and Jamie S. Partridge

Association International Meetings in BuValo, NewWe consistently found that the minimum wage isYork, and at a Federal Reserve Bank of Chicago seminar.positively related to long term unemployment rates

2. Given their low skill levels, minimum wage rates areafter a lag. However, it was often the case that theespecially thought to price teen workers out of the jobsum of the contemporaneous and lag minimum wagemarket. Yet, teens are also associated with even briefercoeYcients was of only modest statistical signi® cancespells of unemployment than adults. Regardless, CLARK

(i.e. at the 10± 20% level). Yet, the reduced form resultsand SUMMERS, 1990a, 1990b, ® nd that there is also a

without the production worker wages suggested ahard core group of teens that suVer long durations of

statistically signi® cant minimum wage result at the 5% unemployment, which composes about one-half of teenlevel. Regarding minimum wage coverage, we found unemployment.that greater coverage increases long term unemploy- 3. A typical explanation for negative duration dependencement, where the statistical signi® cance was stronger is a stigma eVect where long durations of unemploymentthan for the minimum wage rate results. One possible are a negative signal to potential employers of the worker’sfactor that in¯ uenced the minimum wage rate results productivity. For more details on duration dependence,

see VAN DEN BERG and VAN OURS, 1996.was measurement error in estimating the long term4. Among demographic groups, teens (especially minorityunemployment rates. In fact, further analysis using

teens) are thought to be most in¯ uenced by minimumaggregate labour market measures yielded resultswage rates (SMITH and VAVRICHEK , 1992). Besidesstrongly consistent with long term unemployment pat-teens, other demographic groups could be prone toterns. This increased our con® dence in the long term(long) unemployment spells as a result of minimum wageunemployment results.hikes. For example, the 1988 unemployment rate for

These results suggest that policy makers at the federalthose between 20 and 24 years old was 8´7%. NEUMARK

and state level should weigh additional total unemploy-and WASCHER, 1992, ® nd that this age group is about

ment, as well as the possibility of longer durations as adversely aVected by minimum wage hikes as teens.of unemployment, in contemplating future minimum Moreover, the 20± 24 age group’s share of the labourwage increases. In particular, state policy makers should force in 1988 was about four-® fths larger than the shareconsider the prospect of ® rm and household relocation for the 16± 19 age group. Likewise, high school drop-in their decision making. Likewise, in an era of welfare outs over the age of 25 had an unemployment rate thatreform, these results suggest that some low-skilled was about four percentage points above the overall

unemployment rate in 1992 (the earliest year available).welfare recipients may experience long job searches,Drop-outs are another demographic group that aresuggesting that minimum wage hikes may run counterstrongly aVected by minimum wage laws.to the work requirement goals of welfare reform. One

5. Although longer periods of duration are of interest inpossible policy alternative is expanding the earnedEuropean nations, the CPS measure of 26 weeks wasincome tax credit. To be sure, this does not mean thatchosen due to data availability at the state level. Yet,policy makers should forego minimum wage increases,given that a very small share are unemployed for morejust that they should fully weigh the costs and bene® ts.than one year and that state unemployment insurance

The ® ndings tend to support the standard predictionbene® ts typically expire after 26 weeks, this duration is

that there are negative consequences from raising theof great practical importance in the US.

minimum wage, at least at the state level. Foremost, 6. After the federal minimum wage increased to $4.25 anthis was the case even after adjusting for the empirical hour in April 1991, only three contiguous states had aconcerns raised by CARD et al., 1994; and CARD and minimum rate above the federal rate. In fact, the stateKRUEGER , 1995. None the less, more research should minimum wage changes of the latter 1980s created whatbe conducted to explore if these results apply to other CARD, 1992, termed remarkable’ variation in minimumperiods besides the 1980s. However, a challenge facing wage levels that forms an unique natural experiment

into the eVects of minimum wage legislation which hassuch research is that, unlike the late 1980s, there hasnot been replicated before or since. In contrast, nationalbeen considerably less cross-state variation in minimumtime-series studies are hampered by the constraint thatwage rates in the 1990s. Another complication is thatthe federal minimum wage varies infrequently, creatingthe Federal Government no long publishes state-levelsigni® cantly less time-series variation.minimum wage coverage estimates.

7. One reason for the double-log form in equation (1) is

that the Kaitz restriction is a special case. Speci® cally,

the Kaitz restriction requires that b 5 a and b5 2 p,

which will be statistically tested below. To see this, noteAcknowledgements ± The authors thank Dan Aaronson,that the Kaitz variable is COVERAGE*MINWAGE/Gary Hunt, Dan Rickman and Dan Sullivan for their helpfulAVG WAGE. Taking the natural log results incomments with this study.LCOVERAGE 1 LMINWAGE2 LAVG WAGE, where

L re¯ ects the natural log.

8. If a state changed their minimum wage in mid-year, aNOT E S

weighted annual average of their minimum wage was

used. State minimum wage rate information was found1. An earlier version of this paper was presented at the

November 1997 North American Regional Science in various January issues of the US Department of Labor,

Page 13: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

Do Minimum Wage Hikes Raise US Long Term Unemployment? 725

difference as their empirical speci ® cation. This may be aMonthly Labor Review, in which state labour law changes

are summarized. misspeci® cation when there are state ® xed eVects in stategrowth rates, such as persistent growth in the Sunbelt. A9. As shown in CARD and K REUGER , 1995, p. 218, it is

unnecessary to de¯ ate wages by the price de¯ ator in the related concern is that there are only 51 observations

using this approach.current speci® cation (e.g. by the CPI). That is, the sumof the real minimum wage for state s and year t and the 13. It is possible that the two-year lag of the minimum wage

rate may aVect long term unemployment. However,year indicator variable is: blog(MINWAGEst/CPIt) 1 s *t .

This can be rewritten as blog(MINWAGEst)2 when we included the two-year lag in the model (notshown), it was not statistically signi® cant at even theblog(CPIt) 1 s *t . Since CPIt does not vary across states,

this can be rearranged as: blog(MINWAGEst) 1 s t , where 20% level while the other two minimum wage rate

coeYcients were each statistically signi® cant at the 1%s t 5 blog(CPIt) 1 s *t . Similar reasoning applies to theaverage wage term. level.

14. Two exogenous instruments were used in the ® rst-stage10. Despite our best eVorts, it will be very diYcult to fully

control for state economic conditions related to passage model. The ® rst is the real wage mix of the state, whichis the hypothetical wage rate in the state if all the state’sof a minimum wage increase because law makers rely

on a host of unmeasurable factors including constituent industries paid their respective national average wage

(PARTRIDGE and R ICKMAN, 1995, 1997). The realeconomic con® dence. Conversely, NEUMARK andWASCHER, 1992, tried to instrument for the timing of wage mix is a measure of whether the state has a mix of

high or low paying industries. The second instrument isminimum wage hikes, but their results were unaVected.

Regardless, it is very diYcult to ® nd an exogenous the percentage of the state’s non-farm employment inhigh-tech manufacturing industries.instrument related to the passage of minimum wage

increases that is also unrelated to the economic condi- 15. The production worker hourly wage is a direct hourly

wage estimate. Conversely, we indirectly derived thetions of the state and region.11. We expect that coverage will not have a lagged eVect, average private sector hourly wage by taking average

private sector weekly earnings divided by average weeklyalthough this assumption will be tested. With the excep-

tion of the expansion of coverage to state and local hours. Both weekly earnings and (especially) hours arelikely endogenously related to long term unemployment.government in 1985, there was no change in the indus-

tries covered by federal minimum wage legislation in Also, this wage measure is quite similar to K IM and

TAYLOR ’s, 1995, wage measure, which was criticizedthe late 1980s, suggesting that there should be littleadjustment lag for employers. for measurement error (CARD and KRUEGER, 1995;

KENNAN, 1995).12. CARD and K RUEGER, 1995, used the 1990± 92 ® rst-

RE F E RE NCE S

AARONSON D. and SULLIVAN D. G. (1998) The decline in job security in the 1990s: displacement, anxiety, and their eVect

on wage growth, Federal Reserve Bank of Chicago (mimeo).BARTIK T. J. (1991) Who Bene® ts from State and Local Economic Development Policies? Upjohn Institute, Kalamazoo, MI.

BLANCHARD O. J. and K ATZ L. F. (1992) Regional evolutions, Brookings Pap. Econ. Activity 1, 1± 75.

BROWN C. (1988) Minimum wage laws: are they overrated? J. Econ. Persp. 2, 133± 45.BROWN C. (1995) Review symposium: myth and measurement: the new economics of the minimum wage, Ind. Lab. Rel.

Rev. 48, 828± 30.

BROWN, C., G ILROY C. and K OHEN A. (1982) The eVect of the minimum wage on employment and unemployment,J. Econ. Lit. 20, 487± 528.

CARD D. (1992) Using regional variation in wages to measure the eVects of the federal minimum wage, Ind. Lab. Rel. Rev. 46,

22± 37.CARD D., K ATZ F. and KRUEGER A. B. (1994) Comments on David Neumark and William Wascher `Employment eVects

of minimum and subminimum wages: panel data on state minimum wage laws’, Ind. Lab. Rel. Rev. 47, 487± 96.

CARD D. and KRUEGER A. B. (1992) The eVect of the minimum wage on the fast food industry, Ind. Lab. Rel. Rev. 46, 6± 21.CARD D. and KRUEGER A. B. (1994) Minimum wages and employment: a case study of the fast-food industry in New Jersey

and Pennsylvania, Am. Econ. Rev. 84, 772± 93.

CARD D. and KRUEGER A. B. (1995) Myth and Measurement: The New Economics of the Minimum Wage. Princeton UniversityPress, Princeton, NJ.

CLARK K. B. and SUMMERS L. H. (1990a) Labor market dynamics and unemployment: a reconsideration, in SUMMERS L.

(Ed) Understanding Unemployment. MIT Press, Cambridge, MA.CLARK K. B. and S UMMERS L. H. (1990b) The dynamics of youth unemployment, in SUMMERS L. H. (Ed) Understanding

Unemployment. MIT Press, Cambridge, MA.

CURRIE J. and FALLICK B. C. (1996) The minimum wage and the employment of youth: evidence from the NYSY, J. Hum.Resources 31, 404± 28.

DEERE D., MURPHY K. M. and WELCH F. (1995) Employment and the 1990± 1991 minimum wage hike, Am. Econ. Rev.

85, 232± 37.D ICKENS R., MACHIN S. and MANNING A. (1999) The eVects of minimum wages on employment: theory and evidence,

J. Lab. Econ. 17, 1± 22.

Page 14: Do Minimum Wage Hikes Raise US Long Term Unemployment Evidence Using State Minimum Wage Rates

726 Mark D. Partridge and Jamie S. Partridge

FORTIN N. M. and LEMIEUX T. (1997) Institutional changes and rising wage inequality: is there a linkage?, J. Econ. Perspectives

11, 97± 116.FREEMAN R. B. (1982) Economic determinants of geographic and individual variation in the labor market position of young

persons, in FREEMAN R. B. and W ISE D. A. (Eds) The Youth Labor Market Problem: Its Nature, Causes, and Consequences.

University of Chicago Press, Chicago.FREEMAN R. B. (1995) Review symposium: myth and measurement: the new economics of the minimum wage, Ind. Lab.

Rel. Rev. 48, 830± 34.

FREEMAN R. B. (1986) The minimum wage as a redistributive tool, Econ. J. 106, 639± 49.JACOBSON L. R., LALONDE R. and SULLIVAN D. G. (1993) Earnings losses of displaced workers, Am. Econ. Rev. 83,

685± 709.

HAMERMESH D. S. (1995) Review symposium: myth and measurement: the new economics of the minimum wage, Ind. Lab.Rel. Rev. 48, 835± 38.

KENNAN J. (1995) The elusive eVects of minimum wages, J. Econ. Lit. 33, 1,949± 65.

K IM T. and TAYLOR L. J. (1995) The employment eVect in retail trade of California’s minimum wage increase, J. Bus. &Econ. Statist. 13, 175± 82.

KUTTNER R. (1997) So much for the minimum wage scare, Business Week, 21 July, p. 19.

MACHIN S. and MANNING A. (1994) The eVects of minimum wages on wage dispersion and employment: evidence fromthe UK wage councils, Ind. Lab. Rel. Rev. 47, 319± 29.

MELLOR E. F. (1987) Workers at the minimum wage or less: who they are and the jobs that they hold, Monthly Lab. Rev., July,

pp. 34± 38.M INCER J. (1976) Unemployment eVects of minimum wages, J. Pol. Econ. 84, S87± S104.

NEUMARK D. and WASCHER W. (1992) Employment eVects of minimum and subminimum wages: panel data on state

minimum wage laws, Ind. Lab. Rel. Rev. 46, 55± 81.NEUMARK D. and WASCHER W. (1994) Employment eVects of minimum and subminimum wages: reply to Card, Katz, and

Krueger, Ind. Lab. Rel. Rev. 47, 497± 512.

NEUMARK D. and WASCHER W. (1995) Minimum wage eVects on employment and school enrollment, J. Bus. & Econ.Statist. 13, 199± 206.

PARTRIDGE M. D. and PARTRIDGE J. S. (1998) Are teen unemployment rates in¯ uenced by state minimum wage laws?,

Growth & Change, 29, 350± 82.PARTRIDGE M. D. and PARTRIDGE J. S. (1999) Do minimum wage hikes reduce employment: state-level evidence from the

low-wage retail sector, J. Lab. Res. 20, 393± 414.

PARTRIDGE M. D. and R ICKMAN D. S. (1995) DiVerences in state unemployment rates: the role of labor and product marketstructural shifts, Southern Econ. J. 62, 89± 106.

PARTRIDGE M. D. and R ICKMAN D. S. (1997) The dispersion of US state unemployment rates: the role of market and non-

market equilibrium factors, Reg. Studies 31, 593± 606.PARTRIDGE M. D. and R ICKMAN D. S. (1998) Regional diVerences in chronic long-term unemployment, Quart. Rev. Econ.

& Finance 38, 193± 215.

RESSLER R. W., WATSON J. K. and M IXON F. G. (1996) Full wages, part-time employment and the minimum wage,Appl. Econ. 28, 1,415± 19.

S IEBERT H. (1997) Labor market rigidities: at the root of unemployment in Europe, J. Econ. Perspectives 11, 55± 74.

SMITH R. E. and VAVRICHEK B. (1992) The wage mobility of minimum wage workers, Ind. Lab. Rel. Rev. 46, 82± 88.THOMAS J. M. (1996) An empirical model of sectoral movements by unemployed workers, J. Lab. Econ. 14, 126± 53.

VAN DEN BERG G. J. and VAN OURS J. C. (1996) Unemployment dynamics and duration dependence, J. Lab. Econ. 14,

100± 25.WELCH F. R. (1995) Review symposium: myth and measurement: the new economics of the minimum wage, Ind. Lab. Rel.

Rev. 48, 842± 49.

WELCH F. R. and CUNNINGHAM J. (1978) EVects of minimum wages on the level and age composition of youth employment,

Rev. Econ. & Statist. 60, 140± 45.WELLINGTON A. J. (1991) The eVects of the minimum wage on the employment status of youth, J. Hum. Resources 26, 27± 46.