Moj včerajšnji komentar Vulgarizacija minimalne plače je naletel na zelo negativen odziv s strani predstavnikov GZS. Samo Hribar Milič, generalni direktor GZS, mi je celo očital “načrtno zavajanje javnega mnenja z enostranskim prikazom študij o minimalni plači“. Ta očitek je seveda malce smešen, kajti v komentarju se sklicujem na moj post iz septembra 2015 (Ali minimalna plača ubija delovna mesta? Hm, ne, študije tega ne potrjujejo), kjer sem na kratko predstavil 11 ključnih novejših analiz učinkov minimalne plače, od tega 3 meta analize vseh študij, medtem ko se Hribar Milič (in tudi Bojan Ivanc, vodja Analitike GZS) sklicujeta na eno samo študijo (meta analizo Neumarka, 2014), ki pa je v akademskih krogih sporna zaradi pristranskosti pri vključevanju / izključevanju analiz minimalne plače v njun izbran vzorec (glejte spodaj) (o pristranskosti Neumarka in njegovega kolega Wascherja sem sicer pisal že oktobra 2015; zelo učinkovito pa sta ju leta 2000 razkrinkala dva vodilna ekonomista s področja ekonomike trga dela D. Card & A. Krueger).
Nekako sem vedno živel v prepričanju, da ljudje sledijo aktivnim povezavam v posameznih člankih in si preberejo tudi originalne zapise, na katere se sklicujem. Vendar sta me včerajšnja komentarja Bojana Ivanca in Sama Hribarja Miliča boleče opomnila, da temu ni tako. Zato sem se odločil, da tukaj objavim kratek pregled najbolj znanih preglednih študij oziroma meta analiz vseh analiz o učinkih minimalne plače. Teh meta analiz je kar nekaj, najbolj pregledna in kompletna med sodobnimi meta analizami se mi zdi ta, ki jo je leta 2013 naredil John Schmitt (CEPR) “Why Does the Minimum Wage Have No Discernible Effect on Employment?“. Schmittova analiza je nevtralna, hkrati pa ima tudi izjemno dober povzetek, zakaj študije o minimalni plači ne uspejo najti negativnih učinkov na zaposlenost. Toplo priporočam, da si preberete celotno analizo, saj bom jaz predstavil samo uvodni del pregleda glavnih študij.
Preden pa začnem s Schmittovim pregledom, naj še enkrat opozorim, da se v zadnjih 3-4 desetletjih ni spremenil samo pogled ekonomistov na minimalno plačo (od sovražnega do nevtralnega ali celo pozitivnega), pač pa so tudi ključne mednarodne inštitucije (IMF, OECD, ILO in Svetovna banka) spremenile svoj pogled in leta 2012 v skupnem poročilu zapisale, da uradno določene minimalne plače na primerni ravni nimajo negativnega učinka na povpraševanje po delovni sili in da lahko celo povečajo participacijo v delovni sili (vključenost v aktivnost), sploh za tiste, ki so na meji ali bi se vključili na trg dela ali ne:
a statutory minimum wage set at an appropriate level may raise labour force participation at the margin, without adversely affecting demand, thus having a net positive impact especially for workers weakly attached to the labour market” (ILO, 2012).
In 1977, the Minimum Wage Study Commission (MWSC) undertook a review of the existing research on the minimum wage in the United States (and Canada), with a particular focus on the likely impact of indexing the minimum wage to inflation and providing a separate, lower, minimum for younger workers. Four years and $17 million later, the MWSC released a 250-page summary report and six additional volumes of related research papers.2In their independent summary of the research reviewed in the MWSC, Brown, Gilroy, and Kohen, three economists involved in producing the report, distinguished between employment effects on: teenagers (ages 16-19), where they concluded that a 10 percent increase in the minimum wage reduced teen employment, most plausibly, from between zero and 1.5 percent; young adults (ages 20-24), where they believed the employment impact is “negative and smaller than that for teenagers”; and adults, where the “direction of the effect…is uncertain in the empirical work as it is in the theory.” Their summary of the theoretical and empirical research through the late 1970s suggested that any “disemployment” effects of the minimum wage were small and almost exclusively limited to teenagers and possibly other younger workers.”
For a decade, the MWSC’s conclusions remained the dominant view in the economics profession. By the early 1990s, however, several researchers had begun to take a fresh look at the minimum wage. The principal innovations of what came to be known as “the new minimum wage research” were the use of “natural experiments” and cross-state variation in the “bite” of the minimum wage.
Card and Krueger (1994)
Without a doubt, the most influential of the studies using a natural experiment was David Card and Alan Krueger’s (1994) paper on the impact on fast-food employment of the 1992 increase in the New Jersey state minimum wage. In advance of the 1992 increase in the New Jersey state minimum wage, Card and Krueger conducted their own telephone survey of fast-food restaurants in New Jersey and neighboring Pennsylvania. They repeated the survey after the increase had gone into effect and then compared the change in employment in New Jersey’s restaurants (the minimum wage treatment group) with what happened in Pennsylvania (the control group). They found “no evidence that the rise in New Jersey’s minimum wage reduced employment at fast-food restaurants in the state.”
The “New Minimum Wage” research also emphasized research methods based on important differences in the “bite” of the federal minimum across the states. Any given increase in the federal minimum, the thinking went, should have more impact in low-wage states, where many workers would be eligible for an increase, than it would in high-wage states, where a smaller share of the workforce would be affected. Card, for example, divided the U.S. states into three groups–low-impact, medium-impact, and high-impact–according to the share of their teenage workforce that would be affected by the 1990 and 1991 increases in the federal minimum wage. His analysis concluded: “Comparisons of grouped and individual state data confirm that the rise in the minimum wage raised average teenage wages… On the other hand, there is no evidence that the rise in the minimum wage significantly lowered teenage employment rates…“
Card and Krueger’s book Myth and Measurement: The New Economics of the Minimum Wage is the best (though early) summary of these two strands of the “new minimum wage” research. Their detailed review of studies using a variety of methods and datasets to examine restaurant workers, retail employment, and teenagers, concludes: “The weight of this evidence suggests that it is very unlikely that the minimum wage has a large, negative employment effect.”
Myth and Measurement also inspired a considerable response from economists more critical of the minimum wage. David Neumark and William Wascher’s book Minimum Wages brings together much of this critique, with an emphasis on their own work. In Neumark and Wascher’s assessment, the most reliable recent research on the minimum wage has built on the earlier time-series analysis that informed the main conclusions of the MWSC. This new generation of time-series analysis typically applies modern econometric techniques to state-level data on teenagers (and sometimes less-educated workers). Neumark and Wascher’s conclusion is that “…the preponderance of evidence supports the view that minimum wages reduce the employment of low-wage workers.”
Doucouliagos and Stanley (2009)
Hristos Doucouliagos and T. D. Stanley (2009) conducted a meta-study of 64 minimum-wage studies published between 1972 and 2007 measuring the impact of minimum wages on teenage employment in the United States. When they graphed every employment estimate contained in these studies (over 1,000 in total), weighting each estimate by its statistical precision, they found that the most precise estimates were heavily clustered at or near zero employment effects (see Figure 1). Doucouliagos and Stanley’s results held through an extensive set of checks, including limiting the analysis to what study authors’ viewed as their best (usually of many) estimates of the employment impacts, controlling for possible correlation of estimates within eachstudy, and controlling for possible correlation of estimates by each author involved in multiple studies.Doucouliagos and Stanley concluded that their results “…corroborate [Card and Krueger’s] overall finding of an insignificant employment effect (both practically and statistically) from minimum-wage raises.”
In their view: “Two scenarios are consistent with this empirical research record. First, minimum wages may simply have no effect on employment… Second, minimum-wage effects might exist, but they may be too difficult to detect and/or are very small.”
Wolfson and Belman
Paul Wolfson and Dale Belman have carried out their own meta-analysis of the minimum wage, focusing on studies published only since 2000. They identified 27 minimum wage studies that produced the necessary elasticity estimates and corresponding standard errors, yielding 201 employment estimates in total. They then produced a range of meta-estimates, controlling for many features of the underlying studies, including the type of worker analyzed (teens or fast food workers), whether the study focused on the supply or the demand side of the labor market, who the authors of the study were, and other characteristics. The resulting estimates varied, but revealed no statistically significant negative employment effects of the minimum wage: “The largest in magnitude are … positive [and] statistically significant … Several are economically irrelevant though statistically significant and several others [are] slightly larger but … statistically insignificant.“
Neumark and Wascher (2006, 2007)
Meanwhile, Neumark and Wascher (2006, 2007) conducted a qualitative review of the research since the early 1990s on the employment effects of the minimum wage in the United States, other OECD countries, several Latin American countries, and Indonesia.15In their summary remarks, focusing on the U.S. experience, they note:
“What may be most striking to the reader who has managed to wade through our lengthy review of the new minimum wage research is the wide range of estimates of the effects of the minimum wage on employment, especially when compared to the review of the earlier literature by Brown et al. in 1982 [for the Minimum Wage Study Commission]. For example, few of the studies in the Brown et al. survey were outside of the consensus range of −.1 to −.3 for the elasticity of teenage employment with respect to the minimum wage. In contrast, even limiting the sample of studies to those focused on the effects of the minimum wage of teenagers in the United States, the range of studies comprising the new minimum wage research extends from well below −1 to well above zero.”
Based on their subjective weighting of the quality of the research and the reliability of the resulting estimates, Neumark and Wascher conclude:
“Although the wide range of estimates is striking, the oft-stated assertion that the new minimum wage research fails to support the traditional view that the minimum wage reduces the employment of low-wage workers is clearly incorrect. Indeed, in our view, the preponderance of the evidence points to disemployment effects.”
By their calculations, of the 33 studies “providing the most credible evidence; 28 (85 percent) … point to negative employment effects.”18The Neumark and Wascher review, however, is considerably more subjective and arguably less relevant to the United States than the two meta-studies discussed earlier. Only 52 of the 102 studies reviewed by Neumark and Wascher analyzed U.S. data. Of these, Neumark and Wascher designated 19 as “most credible,” five of which were their own studies.The Neumark and Wascher (2006) review also excludes several important papers that were not published until after the review was completed, including the important contributions of Arindrajit Dube, William Lester, and Michael Reich (2010) and Sylvia Allegretto, Dube, and Reich (2011) (to which we will return to below).
Wolfson and Belman
Wolfson and Belman (forthcoming) also produced an extensive qualitative review of minimum wage research since 2000, including a significant number of studies published too late for inclusion in Neumark and Wascher (2006, 2008). Of the studies they reviewed, 40 analyzed U.S. data. Fourteen of these found negative employment effects; thirteen found no effects; one found positive effects; and twelve, a mixture of negative, positive, and no effects. To sort out these conflicting findings, Wolfson and Belman appealed to their meta-study, which as noted earlier, concluded that there were no statistically and economically meaningful employment losses associated with the minimum wage.
Dube, Lester, and Reich (2010)
Probably the most important and influential paper written on the minimum wage in the last decade was Dube, Lester, and Reich (2010)’s study, which offered a comprehensive reappraisal of both the new minimum wage research and its critics. The study was built around a key methodological innovation, which essentially generalized Card and Krueger’s New Jersey study to make it nationally representative, and identified a significant weakness in much of the earlier minimum-wage research based on the analysis of state employment patterns, which had failed to control for regional differences in employment growth that were unrelated to the minimum wage.
The most convincing critique of Card and Krueger’s (1994, 2000) study of the increase in the New Jersey minimum wage (relative to Pennsylvania, where the minimum wage did not go up) was that it is difficult to generalize from a single case study. Even a perfect experiment will have random error that could affect the results in a single experiment. Imagine that the minimum wage had a small, but real, negative employment effect. Random errors will lead the results of separate tests to be distributed around this hypothetical negative employment effect, sometimes producing a larger disemployment effect than the “true” level, sometimes producing a smaller disemployment effect than what is “true”–even zero or positive measured disemployment effects. By this thinking, Card and Krueger’s experiment could have been perfectly executed, but still represent only one result from a distribution of possible outcomes. Absent other information, the best estimate of the true effect of the minimum wage would be Card and Krueger’s actual results, but we cannot convincingly rule out, based on that single case, that the effects were in truth larger or smaller than what was observed in the case of New Jersey in 1992.
In recognition of this problem, Dube, Lester and Reich (2010) essentially replicated Card and Krueger’s New Jersey-Pennsylvania experiment thousands of times, by comparing employment differences across contiguous U.S. counties with different levels of the minimum wage. The three economists carefully constructed a data set of restaurant employment in every quarter between 1990 and 2006 in the 1,381 counties in the United States for which data were available continuously over the full period.They also matched these employment data with the level of the federal or state minimum wage (whichever was higher) in the county in each quarter of each year in the sample. They then compared restaurant employment outcomes across a subset of 318 pairs of bordering counties where the prevailing minimum wage could differ, depending on the level of the federal and state minimum wage.
Their methodology effectively generalizes the Card and Krueger New Jersey-Pennsylvania study, but with several advantages. First, the much larger number of cases allowed Dube, Lester, and Reich to look at a much larger distribution of employment outcomes than was possible in the single case of the 1992 increase in the New Jersey minimum wage. Second, since they followed counties over a 16-year period, the researchers were also able to test for the possibility of longer-term effects. Finally, because the relative minimum wage varied across counties over time, the minimum wage in a particular county could, at different points in time, be lower, identical to, and higher than the minimum wage in its pair, providing substantially more experimental variation than in the New Jersey-Pennsylvania (and many similar) studies. Using this large sample of border counties, and these statistical advantages over earlier research, Dube, Lester, and Reich “…find strong earnings effects and no employment effects of minimum wage increases.”
Allegretto, Dube, and Reich (2011)
Sylvia Allegretto, Dube, and Reich (2011) applied the insights of Dube, Lester, and Reich (2010) to teen employment over the period 1990-2009. Their work made at least two important contributions to the policy debate. First, they analyzed teen employment, rather than industry employment, making their results more directly comparable to the bulk of earlier research on the minimum wage. Second, they included data covering the deep recession that ran from December 2007 through June 2009, allowing them to measure any possible interactions between the minimum wage and strong economic downturns.
Allegretto, Dube, and Reich analyzed data on teenagers taken from the Current Population Survey (CPS) for the years1990 through 2009.28Because the CPS sample is smaller than the QCEW data used in the county-analysis, Allegretto, Dube, and Reich instead tracked teen employment at the state level. When they produced standard statistical analyses of the kind used in much of the research since the mid-1990s on teen employment, the three economists found results similar to those found in that earlier research (a 10 percent increase in the minimum wage reduces teen employment slightly more than 1 percent). But, once they controlled for different regional trends, the estimated employment effects of the minimum wage disappeared, turning slightly positive, but not statistically significantly different from zero.
Allegretto, Dube, and Reich also investigated whether the impact ofthe minimum wage is greater in economic downturns. They “…do not find evidence that the effects are systematically different in periods of high versus low overall unemployment.“
… teh analiz in preglednih študij je še več, zato vas toplo vabim, da preberete Schmittovo analizo do konca in predvsem njegove sklepe, zakaj študije o minimalni plači ne najdejo značilnih negativnih učinkov.
Hudič pa je, da tudi zdaj, ko sem prek Schmittove pregledne študije zelo pregledno predstavil številne analize, ki v povprečju niso uspele najti negativnega učinka minimalne plače na zaposlenost, to ne bo prišlo do živega tistim, ki so bodisi ideološko obremenjeni ali prepričani v svoje apriorno prepričanje. Takšnim osebam nobene gore podatkov ali nasprotnih dokazov ne morejo priti do živega ali jih pripraviti do tega, da bi se zamislili glede svojih prepričanj in jih poskušali pretehtati. Pač pa se bodo obesili na en podatek v milijonu drugih, ki gre v prid njihovemu prepričanju.
Tukaj gre, kar se mene tiče, za preprosto intelektualno poštenost. Treba je biti odprte glave in če se podatki spremenijo, je pač treba spremeniti svoje mnenje (kot je že zdavnaj zapisal John Maynard Keynes). Ne samo, da ni sramotno spremeniti mnenja, kadar podatki kažejo nasprotno, pač pa je to intelektualna dolžnost. In je tudi stvar osebne časti.