Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Standard view
Full view
of .
Look up keyword
Like this
0 of .
Results for:
No results containing your search query
P. 1
Time-Series Minimum-Wage Studies: A Meta-Analysis David Card;

Time-Series Minimum-Wage Studies: A Meta-Analysis David Card;

Ratings: (0)|Views: 97|Likes:
Published by Deliajrs

More info:

Published by: Deliajrs on Jan 30, 2010
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less





Time-Series Minimum-Wage Studies: A Meta-analysis
David Card; Alan B. Krueger
The American Economic Review
, Vol. 85, No. 2, Papers and Proceedings of the Hundredth andSeventh Annual Meeting of the American Economic Association Washington, DC, January 6-8,1995. (May, 1995), pp. 238-243.
The American Economic Review
is currently published by American Economic Association.Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/aea.html.Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community takeadvantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.http://www.jstor.orgMon Jul 2 13:54:34 2007
Time-Series Minimum-Wage Studies:
Meta-analysis 
One of the best-known predictions ofstandard economic theory is that an in-crease in the minimum wage will lower em-ployment of low-wage workers. The evi-dence that is frequently cited in support ofthis prediction is based on aggregate time-series studies. There have been over 30published time-series studies of the effectof the minimum wage on employment inthe United States since 1970. Most ofthese studies correlate the employment-to-population rate of teenagers with the rela-tive level of the minimum wage. In theirlandmark survey Charles Brown et al. (1982p. 508) summarized the time-series evidenceup to the early 1980's as follows:In summary, our survey indicates areduction of between one and threepercent in teenage employment as aresult of a 10 percent increase in thefederal minimum wage. We regardthe lower part of this range as mostplausible because this is what moststudies, which include the experienceof the 1970s and deal carefully withminimum-wage coverage, tend to find.More recent time-series studies (AllisonWellington, 1991; Jacob Klerman, 1992)have found an even smaller effect of theminimum wage.In contrast to the time-series literature, anumber of recent studies based on cross-sectional comparisons (Card, 1992a, b;Lawrence Katz and Krueger, 1992; Card
*Department of Economics, Princeton University,Princeton, NJ 08544, and U.S. Department of Labor,200 Constitution Avenue N.W., Washington, DC 20210,respectively. We thank Lisa Barrow for outstandingresearch assistance, andOrley Ashenfelter and CharlesBrown for comments. The views expressed in this pa-per do not reflect the position or opinion of the U.S.Department of Labor or the U.S. government.
and Krueger, 1994; Stephen Machin andAlan Manning, 1994) have estimated negli-gible or even marginally positive employ-ment effects of the minimum wage. Becauseof the central role that the time-series evi-dence has played in the minimum-wage lit-erature, and the apparent discrepancy be-tween the time-series evidence and thecross-sectional studies, it is important toassess the validity of the time-series esti-mates. In this paper we present a"meta-analysis" of the published time-series litera-ture. Our analysis builds on the observationthat more recent studies have access to manymore observations than earlier studies. Ba-sic sampling theory suggests that thereshould be a simple "inverse-square-root"relationship between the sample size andthe
ratio obtained in different studies. Wetest this prediction using the estimates fromthe time-series minimum-wage literature.Our findings are difficult to reconcile withthe hypothesis that the literature containsan unbiased sample of the coefficients and
ratios that would be expected given thesample sizes used in the different studies.Specifically, we find that the
ratios re-ported in different studies are
correlated with the underlying sample sizes.In addition, the estimated employment ef-fect of the minimum wage tends to be abouttwice its standard error, regardless of thesize of the standard error. These findingssuggest that the time-series literature mayhave been affected by a combination ofspecification searching and publication bias,leading to a tendency for statistically signif-icant results to be overrepresented in thepublished literature.
Publication Bias
It has long been argued that academicjournals have a tendency to publish papers
with "statistically significant" results (ColinBegg and Jesse Berlin, 1988; BradfordDe Long and Kevin Lang, 1992). Statisticalsignificance is usually judged by whether the
ratio for the main explanatory variable($exceeds 2 in absolute value. Because a sta-tistical analysis is deemed more decisive ifthe null hypothesis of zero is rejected (i.e.,if one can reject the hypothesis that the keyexplanatory variable has no effect), there isa natural tendency for reviewers and editorsto look more favorably on studies with sta-tistically significant results.Furthermore, in the case of the minimumwage, economists have a strong theoreticalpresumption that a rise in the minimumwage will lower employment. This presump-tion leads to two additional sources of pub-lication bias. First, reviewers and editorsmay be predisposed toward accepting pa-pers that show a "significant negative effect"of the minimum wage. Second, researchersmay use the criterion of a negative andsignificant employment effect as a guide inchoosing their empirical specifications. Re-searchers have a great deal of discretionover the control variables they include, thefunctional form they impose, the samplethey analyze, and the estimation techniquethey use. This discretion may inadvertentlylead to biases if particular choices are de-termined in part by whether they generatenegative and statistically significant employ-ment effects.Meta-analysis is the quantitative analysisof a body of studies. Meta-analytic tech-niques may be used to summarize a set ofrelated studies (Stephen Jarrell and
D.Stanley, 1990) or to evaluate the reliabilityof the findings in a statistical literature.Meta-analysis can also be used to test forpublication bias. For example, Begg andBerlin (1988) interpret the lack of associa-tion between sample size and statistical sig-nificance in clinical trials of cancer treat-ments as evidence of publication bias.In the context of time-series minimum-wage studies, a natural test for publicationbias arises from fact that more recent stud-ies use more data. The first of these studieswere conducted in the early 1970's, whenthe available time series were relativelyshort, typically going back only to 1954.More recent studies can use up to twice asmany observations as the early studies. Adoubling of the sample size should lowerthe standard error of the estimated employ-ment effect and raise the absolute
ratio byabout 40 percent if the additional data areindependent and the statistical model is sta-ble. More generally, the absolute value ofthe
ratio should vary proportionally withthe square root of the number of degrees offreedom, and a regression of the log of the
ratio on the log of the square root of thedegrees of freedom should yield a coeffi-cient of 1.Aggregate time-series data are unlikely tobe independent. In recognition of this fact,two-thirds of the time-series studies havecorrected their estimates for serial correla-tion. In principle, this adjusts for the depen-dence in the data. Studies that do not makethis adjustment are implicitly assuming thatthe data are independent, so the relation-ship between the
statistic and sample sizestill provides a valid test of publication bias.Furthermore, even with dependent observa-tions, one would expect the
ratio to in-crease with the sample size.What might prevent the
ratio from ris-ing with the sample size? One obvious pos-sibility is publication bias. If studies are onlypublished if they achieve a
ratio of
ormore, and if researchers choose their speci-fications in part to achieve statistically sig-nificant results, then the early studies maytend to have high
ratios despite theirsmall samples. Another possibility is thatstructural change may alter the statisticalmodel. In this case, the
ratio might rise orfall with sample size. If the effect of theminimum wage has become weaker overtime, then the
ratio could remain constantor even fall as the available sample sizeincreases.
To explore the possibility of publicationbias we related the
ratios found in differ-

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->