PERI Living Wage Study

Addendum to Comments of Dr. David Neumark
Visiting Fellow, Public Policy Institute of California
and Professor of Economics, Michigan State University

[Addressed to Dr. Robert Pollin, PERI]

Dear Bob:

I have read through the draft of your living wage report. I commend you for taking seriously my suggestion to ask employers how they might respond to a living wage increase. As I noted, the ideal study is one that can examine the consequences of actual living wage increases. But as Santa Monica contemplates their own ordinance, and one that is quite different from those other cities have implemented, this is not possible, and employers' responses, treated cautiously, are the best a researcher can hope to do. With that said, I have numerous comments on the report, detailed below. I first focus on major themes and conclusions.

Price and employment effects for restaurants

While I commend the fact that the employer survey was done, I am surprised at the conclusions you have drawn based on the survey evidence. Turning first to restaurants, on p. 65 you report that among restaurant owners, 58 percent said they were very likely to raise prices in response to a living wage, and another 14 percent said they were somewhat likely to do so. As I see it, you take this as rather convincing evidence that prices would rise. In a standard economic model, this would imply that demand for the product would fall, as would employment in the industries producing that product, although we don't know by how much. That seems a safe conclusion.

But you then go on to construct arguments why this shouldn't matter much. You suggest that the price changes that might occur are unlikely to matter ("unlikely to significantly alter a restaurant's relative competitive position"), especially at the high end. This, however, is nothing but speculation. You can only argue that there won't be employment effects if you know the price elasticity of demand for the product, which you don't. I can believe the price elasticity is low, but I don't know what it is.

What is particularly surprising, however, is that you have evidence on possible employment responses, from your survey. On p. 82, you report that among restaurant managers, 40 percent said they were very likely to lay off workers (an additional 9 percent said they were somewhat likely to do so), while 44 percent said they were unlikely to do so. This sounds like pretty clear evidence that employers at least expect that a living wage would reduce employment, as nearly one-half of them indicate that they would be at least somewhat likely to lay off workers, and the other half presumably would not respond. Surprisingly, though, you conclude that "these results send no clear message in terms of ... layoffs" (p. 82). 1 don't get it. To predict employment losses, would we have to see all employers indicating they would reduce employment? Of course not. If Santa Monica implemented a policy forbidding restaurant employees from washing their hands, and as a result customers at 40 percent of restaurants become sick, and customers at 60 percent of restaurants did not, would you conclude that the evidence on the policy effects is mixed?

Among hotels, the evidence of likely price increases is even more clear, as 100 percent indicated that they were very likely to raise prices. Clearly, then, if demand is sensitive to price, this could entail some serious cutbacks in demand and employment in the hotel industry. Here, though, you present the unusual argument that not only is demand insensitive to price, but that the quantity of hotel rooms sold (measured by the occupancy rate) is positively associated with price. How is this conclusion reached? Your evidence, essentially, is that over time prices and quantity have been going up in Santa Monica. But there is an obvious issue of identification that you ignore. You assume that in the graphs in Figure 5.1, the data are tracing out the demand curve for hotel rooms, when in fact it is entirely possible (and reasonable) to assume that what's going on is shifts in the demand curve against a supply curve, so you are tracing out the supply curve. As we all learn in econometrics, without a story about what is-causing exogenous shifts in supply or demand curves, we can't learn anything about slopes of demand and supply curves from observations on prices and quantities.

Based on this evidence, such as it is, you discuss why demand may respond positively to price, appealing to arguments about perceptions of the quality of service being driven by price (pp. 62-3). Even if accept this story, it is an explanation of cross-sectional variation in pricing. But it is less compelling as an explanation of the time-series data. First, it seems a stretch to argue that the same hotel can raise its price, and thereby signal higher quality and raise demand. Second, if higher prices elicit higher demand, why did hotels not raise prices earlier? If demand has been stable (which is necessary to learn anything about the demand curve from your figures), failure to do this on the part of hotels is throwing away money. But if, instead, demand curves have been shifting out, then in the earlier years hotels couldn't have raised prices without facing a decline in demand. The bottom line is that in the absence of exogenous shifts that provide the required identifying information, economists usually assume that rising prices and quantities over time imply rising demand, and you have no evidence here to the contrary.

There is some discussion of supply response on pp. 63-65, but I don't really see how this addresses the issue. You seem to argue that supply has not increased, but this is based on correlations between occupancy rates. and the rate of growth of supply. It doesn't establish that supply isn't growing. Furthermore, I suspect the issue of supply and occupancy rates is more complicated. Even with a fixed supply of rooms, one would expect hotels to increase prices if demand shifted out raising occupancy rates.

Again, though, we can get beyond the price effects and look at the survey evidence on expected employment effects. Here, you report that in response to a living wage, among hotels "nearly half thought they were likely to lay off workers, 71 percent [said] they would hire fewer workers in the future, .…" To this rather strong evidence, though, you apply the same conclusion as regards restaurants, that "these results send no clear message ...." What would have constituted a clear message of likely layoffs in these data, if this does not?

In chapter 7, you provide descriptive information on low-wage workers, in an attempt to describe who might be affected by the living wage ordinance. On p. 110, you state that "Claims are often made that a high proportion of workers in low-wage jobs are teenagers. The figures here [Table 7.4] make clear that such claims are erroneous, at least with respect to the Los Angeles labor market." This is important, because low-wage teens are much more likely to be in higher-income families than low-wage adults, so if living wages target low-wage workers, many of the benefits may accrue to families that are not poor and perhaps not even low income. This is the basis for the conclusion shared by nearly all researchers who study minimum wage types of laws that these are very crude instruments for targeting low-income families. (Even Card and Krueger conclude that "The minimum wage is evidently a "blunt instrument" for redistributing income to the poorest families" (1994, p. 285).

I must admit that your statements about teenagers stunned me. First, it's not accurate to say that "claims are made," when there is in fact lots of published empirical work indicating that teenagers are vastly over-represented among low-wage workers (none of which is cited here). Second, I looked at your figures indicating that teenagers constitute 4.2 percent of workers earning between $5.75 and $10.75 in puzzlement, as in 1997 teenagers constituted 5.1 percent of all employed workers (Handbook of U.S. Labor Statistics, Second Edition, p. 15). The only way for teenagers be a smaller proportion of the low-wage segment than of the total would be for them to earn higher wages than the rest of the workforce. Or something is very different in Los Angeles, with far fewer employed teenagers than elsewhere.

To see what was going on here, I went back to the CPS files that you use. The March Annual Demographic Files are problematic for this purpose, because they ask about the previous year. This makes it harder to construct a wage, and even a bit complicated to decide whom to classify as a teenager. In particular, anyone who is 20 at the time of the survey was 19 during part of the previous year (and some 21 year-olds may have been as well). If I take include the 20 year-olds, and look at those with any earnings in the previous year, I find that 6.81 percent of those earning between $5.75 and $10.75 were teenagers. If I include the 21 year-olds, this figure rises to 10.92 percent. I could also take account of the fact that during 1998, the year covered by the data, the minimum wage was $5.15 for part of the year. In the $5.15 to $10.75 range, 7.34 percent were teenagers if I include 20 year-olds, and 11.49 percent if I include 21 year-olds. Finally, these percentages are higher if I impose a lower cutoff on the wage range. So one reason your estimates are low is because you apparently don't include any 20 or 21 year-olds, and use too high a minimum wage for 1998. In addition, though, for some reason you have dropped all individuals who worked less than 250 hours in the previous year. This, of course, disproportionately drops teenagers, especially the youngest and lowest-wage ones. To check this, I verified that if I impose this restriction, I can nearly replicate your numbers.

Given the difficulties with using the Annual Demographic Files, I also looked at the Outgoing Rotation Group files for 1998, which are simpler to use, since they ask about work and wages in the survey month. Using the earner study weights, I looked at 16-70 year-olds living in Los Angeles. In these data, 8.75 percent of workers earning between $5.75 and $10.75 are teenagers (aged 16-19), and 14.04 percent of workers earning between $5.75 and $7.10 are teenagers. These numbers are much higher than the ones you report, and more in line with existing estimates.

This, of course, may help explain the differences you find between your CPS sample your Santa Monica sample, where the proportion of teenagers in the affected wage range is 14.4 percent (p. 115).

1. I don't understand how you relate the findings of DiNardo, Fortin, and Lemieux to the "ripple" effect on wages. This paper isn't really about wage effects above the minimum wage, but about how changes in the minimum wage (and the spike in the wage distribution that it creates) affect inequality. One can glean something about ripple effects in their paper from the non-parametric plots, but it's hard to see. (I think my NBER paper with Schweitzer and Wascher is much more informative about this particular question, which is not their focus.) You argue that "If a, decline in the minimum wage produces an increase in wage inequality, it suggests that an increase in the minimum wage would generate a reduction in wage inequality -- i.e., minimum wage increases produce wage compression due to a relatively weak ripple effect" (p. 46) I don't follow the argument here. Whether or not there is a ripple effect, decreasing the minimum will increase inequality, and increasing it will reduce inequality, unless for some reason the whole distribution shifts with the minimum, which it doesn't.

2. Returning to the survey information on prices increases among hotels, on p. 58 you note that the high fraction indicating likely price increases "is fully consistent with the pricing behavior of Santa Monica hotels over the past 13 years .." What does this mean? Presumably, the survey tells us about price increases caused by the living wage, ie., those above and beyond what would normally have occurred. Otherwise the survey is flawed.

3. I'm disturbed by the frequent reference to a higher wage mandated by the living wage ordinance raising morale, service, etc., in such a way that employers and their workers are not hurt. An example is on p. 66, where you write "Moreover, in paying a living wage to its employees, restaurant owners may find that both service quality and customers' perceptions of that quality could rise." Similar examples are given on p. 70, where you note that there are some employers who pay high wages and have lower turnover, absenteeism, etc. If employers could do better by raising wages, why don't they do this without being compelled to by the government? This question is never answered. Put differently, presumably those that gain from this pay higher wages and reap the gains, and vice versa. There is simply no reason to believe that the gains that some firms get from paying higher wages would be replicated if other firms were also forced to pay higher wages.

There are some cases in economics where individuals do not take actions because they are individually irrational, even though they might be better off if everyone does so. A possible example is maternity leave. If one employer offers very generous leave benefits in the hope that this will let him or her retain key employees, but other employers don't follow suit, then that employer may attract precisely the women employees who want to use these benefits, and end up paying very high costs. But if all employers are forced to do so, there is no longer any selection of women more likely to use these benefits into particular firms, and this "adverse selection" won't occur. Similar examples arise with respect to many forms of insurance (of which maternity leave is one). But in the case of living wages, this argument doesn't apply. There is no reason why an employer wouldn't reap these gains if he or she raised wages in isolation, and indeed some of the effects (such as higher morale from paying higher relative wages) would be stronger, rather than weaker.

Your claim to answer this question on p. 71, where you write: "[I]f firms could benefit through paying a higher minimum wage, why haven't they already voluntarily been paying the higher wage? One answer, noted above, is that some significant minority of firms do pay substantially higher wages than their competitors and still succeed in the market." That, of course, is not an answer. The question is why some firms don't pay higher wages, and you answer it by saying that some firms don't pay higher wages, although some do.

4. On p. 76, you state that "employment changes in the tiny covered sector ... will not be discernible throughout the 1.3 million-person uncovered market." I don't see the relevance of this comment. A small group of workers is potentially affected by the living wages, and we are interested in the outcomes for this group. The minimum wage literature is potentially informative about the effects of exogenous wage increases, although for much smaller price increases.

5. I was struck by the responses to the relocation question, in which 15 percent of restaurants indicate that they are very likely or somewhat likely to relocate as a consequence of the living wage ordinance. Again, though, you draw conclusions that aren't supported by the data. You write: "[T]he figure is high enough to support the notion that at least some of the covered restaurants would pause before dismissing the idea of relocation. At the same time, the fact that the majority of the restaurants say they are not likely to relocate is consistent with the idea that, after pausing, the covered restaurants would attempt other adjustment measures ... before choosing to relocate." Sorry, but I don't see any information in your survey about first considering a decision, then pausing and reconsidering, and then changing a decision. I only see evidence that many restaurants probably wouldn't relocate, and some might. The straightforward interpretation of this evidence is simply that some restaurants might relocate out of the covered area.

6. On p. 87, you summarize the survey evidence as saying "with the hotels and restaurants, firms are likely to lay off workers only if they are unable to absorb the cost increases through raising prices and productivity or reducing profit margins slightly." I don't believe this conclusion is supported by the evidence. The survey simply wasn't structured in such a way as to draw such inferences, although one could perhaps learn something by looking at interrelationships among the possible responses reported by employers.

7. I'm surprised by one thing reported in Table 7.6. In the first column you look at workers earning between $5.75 and $7.40 per hour. There are 2.1 earners per family, yet these low wage workers contribute about one-half of family earnings. That implies an extraordinary degree of assortative mating on wages across families, with low-wage women married to low-wage men, or sharing their households with low-wage teenagers. That may, in fact, be the case, but this is worth looking into further. Another simple tabulation to check on its plausibility is to show the fraction of earners with wages in various wage categories who have other workers in their families in various wage categories. The same issue arises in Table 7.13, where there are 1.9 workers per family but the lowest-wage workers supply more than one-half of family income. That seems impossible, unless there are a lot of workers below the minimum wage. Of course, family income could be badly reported. (Imagine how well teenagers know their families' incomes.)

8. Something appears to be wrong on p. 113. You report that over one-half of families of workers in the lowest age range are poor or near-poor (which means they are below 1.5 times the poverty line). But then you state that more than one-third are below 160 percent of the poverty line. Either this fraction is a lot more than one-third (over one-half), or something is misstated here.

9. With reference to this same paragraph (p. 113), you don't draw the clear conclusion here, that many families that would benefit from the living wage are neither poor nor near-poor. That is why other policies that explicitly target the poor (and don't make silly distinctions between whether one works in the Coastal Zone, for a city contractor, or for an employer of a particular size) make much more sense as anti-poverty policies.

10. On p. 119, you make the claim that commuting is "absorbing a significant portion" of earnings of covered workers. Unless I missed something, I don't see this claim substantiated in the report.

11. Your comments in Appendix 1 on my paper with Scott Adams are misdirected. You make it appear that our whole empirical strategy hinges on identifying which workers are covered and which are not, which is obviously impossible to do with the CPS. In fact, most of our analysis does not even try to do this, nor does it take any stand on "ripple" effects. We simply identify cities that passed living wage ordinances, and compare their changes in wages, hours, poverty, etc., with those in other cities that didn't pass these ordinances.

We only use the covered/uncovered classification to attempt to assess our findings. The paper states, "As further verification that we are truly detecting effects of living wage ordinances, we next attempt to estimate wage effects for covered and uncovered workers." We indeed find that effects appear stronger for those more likely to be covered, which is all we ever claim.

Indeed, we are well aware that we are overstating coverage. First, we show results including and then excluding cities with "business assistance" living wage ordinances. Second, we state that "Some living wage ordinances are not explicit about what types of workers are covered. For many localities, we had to make strong assumptions concerning the types of industries in which covered individuals work. Table 4 shows that we chose the broadest definitions of potential coverage, so as not to exclude those that are potentially affected. Actual coverage rates should be much lower than those we report."

Along the same lines, despite what you claim, we make no assumptions about ripple effects. Rather, the statement you quote summarizes the empirical findings.

I think one can criticize our research as lacking information on covered workers or employers. But what it does is estimate reduced form policy effects. The impacts we find suggest that far more workers are affected than are indicated by many coverage estimates in the existing living wage studies, which certainly raises some questions. But it is incorrect to claim that our estimates hinge on any sort of coverage estimates.

I hope you find these comments helpful in revising your report.

Sincerely,

David Neumark

This page was last updated on 10/25/06.