Version 1.0 (April 24, 2019)
A couple of weeks ago, I got ensnarled in one of these debates on Facebook that do not go anywhere; it was triggered by the Australian Labor Party’s recent Living Wage policy proposal and the related discussion about the merits of minimum wages, and there specifically whether increases in minimum wages have negative employment effects and even more specifically whether such detrimental employment effects hit those at the low end of the wage distribution. These debates tie into other current debates like the one about lacking wages growth about which even the RBA is now concerned; see also Fig 17.17 here, or the one about wage theft which even the current government — not known for its charitable inclinations — says it wants to address, or the one about growing inequality which, as it affects aggregate demand, has to be a growing concern for any economist worth her or his money.
The debate on Facebook I referred to at the outset did not go anywhere because it turned out that the conceptual and empirical issues of the effects of minimum wages on employment are quite different from what most people have learned in basic Econ 101. And, unfortunately, about 99 percent of those currently engaging in the debate about the effects of minimum wages have not gone beyond that level of understanding, even if they claim economics expertise. So, I bowed out of that debate at some point and read up on it. Here is what I understand the current state of the science to be. I am open to suggestions on where I might get it wrong. Feel free to comment away.
The traditional understanding results from what is (still) being taught in most principles courses: that the demand curve for any good, or factor of production, is downward sloping, i.e.. the higher the price (e.g., of labor) the less of it will be demanded. That given, a simple graph establishes that equilibrium price and quantity are determined by the interaction of demand and supply. If a binding floor is established (i.e., if a price is set by some law or ordinance above the equilibrium price), quantity demanded will be constrained and less of it will be employed. Ignoring for now the important question whether labour should be considered just another factor of production or whether there is more to it – a point for another overdue debate –, the fact that less labor might be employed is not necessarily a bad thing, a point that many people do not seem to grasp. It all depends on the “elasticity of demand” which is a measure that relates the responsiveness of the quantity demanded to price changes. If the price increases (brought about by a law or ordinance that sets a price above the equilibrium price) are large but negative employment effects not quite as much, a case can possibly be made for some such law or ordinance.
So far so good and probably uncontroversial even among most economists and journalists who write on economic issues. It is also fairly uncontroversial that long-run measures of responsiveness of employment to price changes (“elasticities”) tend to be larger than short-term measures and that these elasticities might be different for different locations on the demand curve and for that matter different demand curves. What exactly the appropriate elasticity estimate is for various contexts is arguably more controversial. On balance, economists are sceptical about market interventions of this kind, and often for good reasons (e.g., the well documented consequences of rent-controls although even here a more critical attitude is emerging).
In come Card & Krueger (1994) with a study published in one of the top journals of economics, and a provocatively titled book a year later with one of the top presses, in which they seem to be able to show, relying on an intriguing natural experiment that inspired hundreds of studies and contributed to the credibility revolution in economics, that raising the minimum wage has no adverse effect on employment. Wow.
Not surprisingly, this study was heavily contested.
The debate continues to the day and I will below discuss some key contributions to the debate and also some of the kerfuffle surrounding the recent Seattle minimum wage experiment which is ongoing.
As you will see there are many moving parts, conceptually and empirically, confounding these debates. My list of things to keep in mind follows; below I will refer to them as “caveats”.
Caveat 1: The simple Econ 101 model of the effects of a minimum wage (or price floor) is flawed because it does not take into account dynamic effects. For example, in a growing economy (such as the Seattle market has been for a while), it is important to study a minimum wage effect against the counterfactual which is not easy to establish. Among the ways labor market researchers have approached the topic is the “synthetic approach” of identifying districts outside of an urban center that are, however, in their sum well matched in observable characteristics. Another strategy has been natural experiments across state borders when one state changed the minimum wage and the other did not.
Caveat 2: Any minimum wage increase (especially when drastic) will not only have, possibly, employment effects for those whose wages go up but for those that have wages above the price floor. It is unclear conceptually how some such “ripple effect” could pan out: Will those workers that earn more (as apparently a considerable number of workers in the restaurant industry do), suddenly put in less effort, and/or will those that have their wages increased provide more? What are the results of the long-term adjustments processes of what is the relative efficiency wage between those with lesser skills and those with more?
Caveat 3: An important issue is temporary job-seekers (e.g., students on their holidays or teenage workers to earn some change, or other part-time workers, etc.) and to what extent adverse employment efects will disproportionately fall on them. In his simulation exercise, MaCurdy makes the important point that, even if one accepts that there are no adverse employment effects, those benefitting from an increase in minimum wages are not necessarily low-income families. In fact according to his data “low-wage families are typically not low-income families.” (p. 534) And, “The increased earnings received by the poorest families are only marginally higher than those of the wealthiest. One in four families in the top fifth of the income distribution has low-wage worker which is the same share as in the bottom fifth.” (pp. 534-5) The problem is that while “fewer than one in four low-income families benefit from a minimum wage increase of the sort adopted in 1996, all low-income families pay for this increase through higher prices, rendering three in four low-income families as net losers.” (p. 535) Clearly the incidence of price increases is an important consideration, and needs to be controlled for.
Caveat 4: Relatedly, there is the issue of the gig-economy which tends to put pressure on wages in the low-skill sector and allows price floors for labor to be circumvented by making participants into “entrepreneurs” of sorts. This seems currently a completely understudied area.
Caveat 5: Another important issue is whether wage increases are implemented locally, regionally, or on the national level. Minimum wages on the national and even the regional level are presumably less easily circumvented than those on the local level, although it is probably a mistake to underestimate the attractiveness of big urban centers.
Caveat 6: An important issue is anticipation effects which can pollute supposed quasi RCTs. This explicitly motivated the study by Bell & Machin (2018).
Caveat 7: If the simple Econ 101 model of the effects of a minimum wage (or price floor) is flawed because it does not take into account dynamic effects, then the question has to be asked which alternative model could explain the results. Prominent competing labour market theories are those of monopsonistic, or oligopolistic, competition with search costs, or efficiency wages. Belman & Wolfson (2014) have a useful discussion of such models in graphical form. See also this or this.
Caveat 8: As for the abatement of carbon emissions there are typical several other policy options and therefore it is important to keep in mind that there may be other policy tools such as earned income tax credits that would address the policy goal of a minimum, or living, wage more effective- and efficiently. In this context the important question arises who ultimately pays for minimum wages. Assuming no adverse employment effects, the question is whether employers manage to push increased labor costs through to consumers, or whether they have to take some hit to their profits. This seems to be another poorly understudied area. The available evidence (see MaCurdy 2015 and Draca et al. 2011), Bell & Machin 2018, and this very interesting paper by Harasztosi & Lindner 2017) is inconclusive. More about these papers below.
Caveat 9: Debates about minimum or living wages should not be backward looking. They ought to happen in the context of an ever increasing automatization and inequality (in Australia surely in terms of wealth and quite possibly also in terms of income).
Caveat 10: Many of the empirical findings that we have are from The United States. Naturally, we should not take for granted that the findings (to the extent that we might agree on them) translate to the Australian context. There is, fortunately, an interesting literature documenting the effects of the (1999) introduction of minimum wages in England (including at least one meta-study), and an emerging literature documenting the effects of the (2015) introduction of a living wage in England (Bell & Machin (2018). There will soon be yet another emerging literature that draws on the minimum wage experiment that Germany launched in 2015. And importantly, there is the very interesting paper by Harasztosi & Lindner (2017) on the Hungarian experience with a dramatic increase of 60 percent in the minimum wage in 2001.
Back now to the debate and some key contributions and also some of the recent kerfuffle about the Seattle experiment which are ongoing.
Card & Krueger (1994, 1995a, 1995b, 2000, 2016, chapter 1, 2017) studied the fast-food industry in New Jersey and Pennsylvania. The former increased the minimum wage in April 1992. The original Card & Krueger 1994 findings suggested – to many economists’ surprise — that there was no employment effect of that increase in the minimum wage. This finding was contested by Neumark & Wascher (2000) who argued that the Card & Krueger data – because they were collected through telephone surveys – were unreliable. Drawing on payroll administrative data, they argued furthermore that the New Jersey minimum-wage increase led to a decline in fast-food employment. Neumark has emerged as arguably the most influential critic of Card & Krueger and maintains, for example in this recent IZA primer, that ”A great deal of evidence indicates that the wage gains from minimum wage increases are offset, for some workers, by fewer jobs. Furthermore the evidence on distributional effects, though limited, does not point to favourable outcomes from minimum wage hikes, although some groups may benefit.” In their response to Neumark & Wascher, Card & Krueger questioned the representativeness of the Neumark & Wascher data and instead used two different kinds of longitudinal and repeated cross-sectional administrative data reported to the BLS. They conclude that “the increase in the New Jersey minimum wage in April 1992 had little or no systematic effect on total fast-food employment in the state, although there may have been individual restaurants where employment rose or fell in response to the higher minimum wage.” (p. 1398) In chapter 1 of the 2016 version of their book Card & Krueger re-iterate this finding and argue “This book presents a new body of evidence showing that recent minimum-wage increases have not had the negative employment effects predicted by the textbook model. Some of the new evidence points toward a positive effect of the minimum wage on employment; most shows no effect at all. Moreover, a reanalysis of previous minimum wage studies finds little support for the prediction that minimum wages reduce employment.” (Chapter 1, page 1)
Building on a meta-analysis that Card & Krueger did in the same year they published their book, Doucouliagos & Stanley (2009) provided a meta-regression analysis of minimum wage research drawing on 64 (!) minimum wage studies. Importantly, they controlled for publication selection bias and they find that, when selection effects are filtered out, “no evidence of a meaningful adverse selection effect” (p. 422) (thus confirming the earlier results of Card & Krueger). Specifically, “In the minimum-wage literature, the magnitude of the publication selection is large or larger, on average, then the underlying reported estimate. … Even under generous assumptions about what might constitute ‘best practice’ in this area of research, little or no evidence of an adverse effect remains in the empirical research record, one the effects of publication selection are removed.” (p. 423) In passing, Doucouliagos & Stanley, who have no recognizable dog in this fight (as much of their work involves meta-analyses of various areas), suggest that the “subjective narrative review” provided by Neumark & Wascher (2007) that covers the same literature (but comes to a very different conclusion) is wanting because it does not control for publication bias. A more recent meta-analysis by Belman & Wolfson (2014) of minimum wage studies published between 2000 and 2013 also confirms the Card & Krueger results, finding a median elasticity of employment or hours with respect to the minimum wage of between -0.05 and -0.03, not controlling, however, for publication bias.
Dube and his colleagues (including Michael Reich, one of the protagonists of the Seattle minimum wage study controversy) provided yet more evidence in support of the results claimed by Card & Krueger. Inspired by their identification strategy, they compare restaurant and retail employment in contiguous countries across state borders, and there in particular segments with minimum wage differences over a 17-year period. The border discontinuity design has attractive features spelled out in Dube (2017) where one can also find a succinct summary of his later work, a good discussion of related work, and of alternative approaches. Acknowledging that “the topic of employment effect of minimum wages remains controversial, with sometimes conflicting evidence” (p. 820), in his own work he finds employment elasticity close to zero (also for teens), and “even when we considered the longer-terms effects (e.g., four or five years out), we found employment estimates to be fairly small.” (p. 820) Reflecting on his recent work with Lester and Reich (2016), he notes “We found a striking pattern … This trifecta of results – strong positive wage effect, small employment effect, and strong negative turnover effect – is a signature of a model with search frictions … “ (p. 821) None of these papers, however, controls for the effects of minimum wages on prices (recall MaCurdy’s results) and/or profits, something which seems necessary if one wants to understand the net effect of a minimum wage increase.
Additional empirical evidence in favour from the UK, where a national minimum wage was introduced in 1999, and a national “living wage” in 2015 (by a conservative government, no less), is reviewed in Lemieux (2017) and Card & Krueger (2017); summarizes Lemieux, “By moving from a situation of no minimum wage to one with a large and differentiated (by age) minimum wage, the United Kingdom was an ideal laboratory … . A clear consensus in the British literature supports that the new minimum wage had, and continues to have, no effects on employment. … On balance, it appears that the evidence accumulated since 1995 has, if anything, reinforced Card and Krieger’s conclusion of no (or modest) employment effects of the minimum wage.” (p. 824) While Lemieux mentions the results by Machin, Manning, & Rahman (2003) … he does not mention Leonard, Doucouliagis, & Stanley (2013) which is puzzling.
Card & Krueger point out that these new data allow also inferences about how the wage distribution is affected. “Relative to the literature on the employment effects of minimum wages, there are fewer recent studies of the distributional impacts. …Exploiting the remarkable history of minimum wage legislation in the United Kingdom, Dickens et al. (2012) concluded that the introduction of the national minimum wage had a strong effect on the lower tail of British wages, pushing up the wages of workers as high as the 35th percentile in the overall wage distribution.” (p. 829) That is a remarkable result that, if true, speaks to Caveat 2 that I formulated above and it suggests that restricting attention to those directly affected by wage increases is a problematic strategy.
As stated by Dube, “the topic of employment effect of minimum wages remains controversial, with sometimes conflicting evidence”. (p. 820) Another prominent contrarian, apart from Neumark, is Jonathan Meer. Meer & West (2016), who estimate a large negative effect of minimum wages on aggregate employment, is discussed in Dube (2017) who suggests – based on his empirical work – that these putative job losses happen higher in the wage distribution, “raising questions about the causal import of their estimates”. (p. 820) Meer (2018) summarizes his take on the literature, including his own work and that of the Jardim et al. (2017) study of the Seattle minimum wage experiment. Says he, riffing on an even more partisan earlier assessment of the literature (Meer 2017): “Following the minimum wage increase [to $13 per hour in 2016 [January] from $11 in 2015 [April] and $9.32 in 2014], total payroll for low-wage workers actually fell by an average of $125 per month: those workers for whom the increase was supposed to help were actually receiving fewer dollars on average after the minimum wage increase than before. Unsurprisingly, the study came in for immediate criticism from minimum wage advocates, but its methodological approach is sound and most of the critiques are groundless.” (p. 6) These are strong claims and they would be more credible if his literature review would be less partisan (e.g., no word about the meta-studies by Doucouliagos & Stanley 2009 and Belman & Wolfson 2014, also no word really about the English experiment which is of interest for the simple reason that minimum wage is differentiated by age, thus having the potential to address one of Meer’s concerns).
Since the Jardim et al. (2017) paper, the same group of researchers has published a follow-up study (Jardim et al. [October] 2018) that assesses the impact of the first and second min wage increases in 2015 and 2016, using again longitudinal workforce data (“employment trajectories of thousands of individual employees engaged in low-wage work immediately before each increase”, p. 4) collected by the state of Washington’s Employment Security Department. The new analysis follows the same identification strategy as the earlier study (of trying to compare Seattle low-wage workers with matched controls from outlying Washington State) but comes to a different conclusion: “While these workers experienced a modest reduction in hours worked, on net their pretax earnings increased an average of around $10 a week” (p. 4, see also p. 25) However, the bulk of these gains went to more experienced workers, with less experienced workers being about as well of as before. As Jardim et al. (2018) admit, “The findings contrast with our earlier work, which showed that the total amount paid to workers in low-wage jobs in Seattle declined after the second minimum wage increase in 2016.” (p. 25) I am tempted at this point to paraphrase Meer (2017), If your immediate reaction to this study is to dismiss it, “it is time to admit your views cannot be swayed by science. They might as well be religion.” (p.7) Say it as it is, Jonathan.
But, seriously, it is important to understand that any one single case (study) is only so telling and often a work in progress. The case of Seattle is so unique (see also this excellent recent story about Amazon) that the caveat in Jardim et al. (2018) not to generalize this finding to state and federal policy changes seems warranted.
Looking at all the evidence that is currently out there (and that I paraded above), I conclude that the balance of the evidence seems to provide considerable support in favour of policy recommendations that contradict the standard Econ 101 narrative, even if we accept that labor should be considered just any old factor of production.
Local circumstances have, of course, to be part of any attempt to implement minimum wages and/or living wages. It is also worthwhile remembering Alan Krueger’s admonition from 2015 that “$15 an hour is beyond international experience, and could well be counterproductive. Although some high-wage cities and states could probably absorb a $15-an-hour minimum wage with little or no job loss, it is far from clear that the same could be said for every state, city and town in the United States.” (Krueger 2015) Krueger, however, issued that caveat before the results in Harasztosi & Lindner (2017) started to circulate which are remarkable indeed. These authors analyse a very large (~60% in real terms) minimum wage increase in Hungary in 2001. Among the remarkable results are that this very large increase displaced four years out only 1 out of 10 minimum wage workers while those that held on to their job experienced a 50% wage increase. Importantly, while firms predictably responded to the increase by trying to substitute away from labor to capital, they showed that “around 80% of the wage increase paid by consumers of goods produced by minimum wage workers and only 20% was paid by firm owners.” (abstract) Not surprisingly, there is considerable heterogeneity to be found in firms’ ability to pass through the increased minimum wage. Interestingly the result that forced increases in labor cost can be for the most part pushed through to consumers contradicts the results reported for the English minimum and living wages experiences. Draca et al (2011), based on the 1999 minimum wage increases, find that they find their way directly into profit reductions. Bell & Machin (2018), based on the 2015 minimum wage increase, find that it finds its way directly into lower firm value, a story consistent with Draca et al.’s but not with MaCurdy or the story Harasztosi & Lindner (2017) tell. Hmmmh.
The implications of the currently available evidence for Australia are worth thinking through carefully. Taking into account purchasing power parity which probably means that a wage of around 15 Aussie dollars seems a safe proposition in particular if age – adjusted following the English model. Importantly, there may be better policy interventions out there such as earned income credit and other active labor market programs. Card et al. (2018) have recently reviewed the available options.
It seems very desirable to think through these issues in non-partisan matter (i.e., have what we know assessed through forms of adversarial collaborations by knowledgeable people representing key stakeholders in the debate) and with a longer-term perspective that takes into account the nature of work in the future. Not being a labor market researcher, one of the frustrations I experienced when reading up on this literature is the often very partisan assessment of the findings out there. For example, if you are an opponent of minimum wages and selectively point at evidence to support your stand, you lose credibility right away when not mentioning the meta-studies out there that do exist (and that so far support the Card & Krueger findings).
I appreciate, without implicating, Stepan Jurajda’s and Jonathan Meer’s pointers towards relevant literature and Stepan’s and my colleague Gigi Foster’s feedback on an earlier draft. Needless to say, they are not to blame for errors in fact or opinion.
Consider following me on Twitter at @aortmannphd