Krugman has suggested that Ryan Avent and I look at the literature on labor supply elasticities. Or perhaps, as Ambrosini suggests, just the half of that literature with which he agrees:
Today he says that Sumner and Avant should read the literature on macro vs micro labor supply elasticities. Well, ok, he says they should only read half of that literature. Over the last decade, Prescott has been doing a lot of work showing that differences in taxes explain differences in employment and hours worked between Europe and the US. I think his Nobel speech was about this.
The micro people threw fits though because their estimates of the response of labor supply to tax changes is much less extreme than Prescott’s finding suggest. They basically find labor supply curves are vertical. This would mean that taxes simply can’t have an effect on labor supply.
For a while, these guys had me convinced because, in general, micro/labor types do a much better job of identification and I trust their estimates more than I trust macro estimates. More recently, however, macro people1 have been making the case that the “labor supply elasticity” estimated by the micro people is different from the “labor supply elasticity” the macro people estimate. The difference isn’t due to statistical methodology, we were just calling two different things the same thing.
Of course, its the macro elasticity that matters for tax policy, though. Prescott’s work (and not the paper that PK links to) is the place to go for understanding differences between Europe and the US. He says that difference is due to differences in tax and transfer policies.
I haven’t studied the recent literature in this field, although of course I was aware of the Prescott study. But if Ambrosini is right I feel much less ignorant than I feared. Back in the 1970s at Chicago we were taught that you needed to use income compensated supply and demand curves when examining tax policy. I just assumed that everyone knew that when government collected taxes, the money was injected back into the economy through various means. I guess this insight only recently reached certain segments of the profession. Now of course the progressives that favor high taxes can always argue that the government expenditures go to activities that are so worthless that society is impoverished. In that case people keep working hard for the same reason Bangladeshis work hard. But I’m going to assume they don’t want to make that argument.
This is from the Alesina, Glaeser and Sacerdote paper recommended by Krugman, which argues that institutional factors such as unions, and not taxes, explain the big difference in hours worked between the US and Europe:
In a recent, provocative paper Prescott (2004) argues that “virtually all of the large differences between U.S. labor supply and those of Germany and France are due to differences in tax systems.” Prescott calibrates a dynamic model of investment and labor supply; and shows that under what appear to be reasonable assumptions about parameter values, all of the difference between the U.S. and the major European countries can be explained by different marginal tax rates. Indeed the marginal income tax rate differences between the U.S. and Europe were much smaller in the 1970s, when labor supply differences were much smaller. Prescott’s view is supported by the statistical evidence of Davis and Henreksson (2004) who use a panel of richer countries and find large labor supply responses to higher tax rates.But Prescott’s argument that taxes explain U.S./Europe differences relies critically on assumptions that ensure an elasticity of labor supply that is hard to reconcile with most standard estimates of labor supply elasticities. In the case of male labor supply, we are not aware of any within-country estimates of labor supply elasticities that are even in the same ball park as those used in the Prescott’s calibration. For women, estimated labor supply elasticities are much closer to those used by Prescott (his assumptions still veer toward the upper limit of available estimates); however the reduction of hours worked is by no means a women-only phenomenon. Female labor force participation shows an increasing trend in most European countries since 1973 even though the U.S. shows an even faster growth of female participation; this is a case where the effect of marginal tax rates may indeed be very important. Prescott himself is well aware of this discrepancy between the traditionally estimated elasticities from “micro” evidence and the “macro” elasticity needed for his calibration exercise to work, but he offers little explanation of why the “micro” elasticities are wrong.This paper examines two different hypotheses for the mismatch between macro and micro labor supply estimates. The same hypotheses also offer us different theories of the differences in hours worked between the U.S. and Europe. First, we consider the possibility that the macro-estimates are right in this context and the micro-estimates are misleading. Micro-estimates may be statistically correct, but they are inappropriate because they consider only the direct impact of taxation. One indirect effect of taxation is the government transfers that it funds. These transfers create an income effect that might induce lower work hours. This is one reason why Prescott’s elasticities are so high, but there is little evidence suggesting income elasticities of the level that his work assumes.
My first reaction is that Ambrosini is right. They write as if Prescott’s income-compensated labor supply estimates were some sort of novel innovation. Again, the substitution effect is the only first order effect from tax changes. My second reaction is to wonder whether micro estimates are particularly useful for macro analysis. They talk about studies showing that labor supply doesn’t respond much to changes in unearned income. But here’s my problem with those sorts of micro estimates; labor supply is partly influenced by all sorts of societal norms, which are ultimately determined (in the long run) by incentives such as MTRs. Here are a few societal norms:
1. The standard 40 hour work week.
2. The standard retirement age of 65.
3. Attitudes toward working mothers.
4. The number of legal holidays.
5. Attitudes toward young people who are indefinitely unemployed and living on the dole.
I once lived in England, and was shocked at how casually young Englishmen I knew would accept long term unemployment. Now all of this might seem very unChicago-like. But I am not saying that tax rates don’t affect labor supply, rather I am saying that the effects are probably mediated through quite complex mechanisms, some of which involve collective choice. Go back to the unearned income example. If a factory worker at GM inherits money from his aunt, and starts earning interest income, how easily can he go to GM and ask to be shifted from a 40 hour work week to a 35 hour work week? And yet if every factory worker at GM had such good fortune, perhaps they’d ask their union to negotiate shorter hours.
If societal norms change slowly (partly because of “first-mover” problems), then I could imagine that countries with more collective bargaining might be able to more quickly move to the regulatory structure that is now optimal in light of higher tax rates. This might involve changes in the standard retirement age, the number of official holidays, and/or the number of weekly hours worked at which overtime pay kicks in. This doesn’t mean taxes are not ultimately responsible for these collective decisions, just that an econometric study might not identify taxes as the proximate cause of changes in labor supply.
But even if I am wrong and if the true causal factor is unions and other institutional structures, and not tax rates, does this really help the progressives’ argument regarding US fiscal policy? I don’t see how, unless they can identify low union membership countries with very high tax rates and very high levels of hours worked. But where are those examples?
At first glance it seems more “scientific'” to try to isolate the effects of each causal variable, and to try to figure out what happens if we raise taxes and hold everything else constant. But on closer inspection it isn’t clear that there is anything scientific about this approach. Indeed I can imagine four possible interpretations for the AGS findings, only one of which supports the progressive agenda:
1. The regressions suffer from an identification problem due to the social norms issue I mentioned, as well as other factors.
2. Taxes are caused by the institutional factors that are associated with the Western European economic model. And these are in turn caused by cultural differences, making it impossible to implement the economic model in the US.
3. Institutional factors cause hours worked, and culture does not prevent the US from adopting the institutional structure of Western Europe. But taxes are endogenous, and reflect the same forces that lead to things like strong unions.
4. Taxes and institutional structure are both exogenous, and both can be determined independently.
Case two we can’t adopt the European model for cultural reasons. Case three says we can adopt the European model, but it is all or nothing. Either we go whole hog into the European model that the progressives like, and suffer a big drop in GDP per capita, or we stick with our current model. In that case the European model as a whole doesn’t produce significantly more tax revenue (PPP) per capita than the US model. That’s the implication of Prescott’s work, and that would also be the implication of AGS’s study. Only in case 4 do we get the policy implication that high tax rates in the US would not significantly reduce hours worked.
Obviously it is possible to argue that we can have our cake and eat it too. It’s theoretically possible that we could have high tax rates and continue to have high levels of hours worked. But where are the successful examples of such a model? Where are the countries that have European tax rates and US levels of productivity and hours worked? If there are no such examples among the 200 countries in the world, then it is hardly “scientific” to claim we can choose such a model as casually as picking an item off a menu. The truly scientific approach would be to assume everything is endogenous.
As I said in my first post on this issue, I think it might make more sense to view the entire progressive policy regime as a single choice:
Why is per capita GDP in Western Europe so much lower than in the US? Mankiw seems to imply that high tax rates may be one of the reasons. I don’t know if that’s the answer, but if it’s not my hunch is that the factors that would explain the difference are other government policies that the left tends to favor (strong unions, higher minimum wages, more regulation, generous unemployment insurance, etc.)
Do we buy into the European model or not? If we do, let’s not kid ourselves that we can avoid a big hit to GDP/person.
In contrast, I do have an ideal real world model in mind: Singapore.
BTW, Ryan Avent criticized me for cherry-picking Singapore as an example of a rich, low tax economy. He pointed out that there are also rich metro areas in Western Europe, and I agree that that is a very good counter-argument. But take a look at the astounding data on hours worked that commenter “mbk” sent me:
Labor productivity, GDP per hour worked in k$:
That’s indeed because of k hours worked per employed person and year:
(note 1 Japan is at 1.79 contrary to popular perception)
(note 2 re: Singapore I wouldn’t be surprised if family business work were to not even fully enter into the true hours worked by young and old on evenings and weekends)
(note 3 yes this means Singaporeans work almost twice as many hours as Europeans, and I’d assume a higher % of the population is in the workforce as well)
(note 4 this squares well with my anecdotal perception that everyone in Singapore works all the time)
Singapore productivity may be a bit higher than estimated—his data uses a fairly conservative PPP adjustment. I have seen estimates where Singapore’s GDP per capita already exceeds the US. But the hours worked is the real eye-opener. Singaporeans, who face even lower taxes than Americans and Japanese, work far more hours each year. And mbk also suggests a higher percentage of people may be working. So even if one discounts the high income of Singapore as reflecting its status as a city-state, the hours worked provides strong evidence that the ultra-low taxes have an impact. I doubt whether people work anywhere near that long in Paris, Frankfort or Milan. And I’ll bet you’d also find extremely high hours worked in Hong Kong, the other developed country with tax rates far below US levels.
[BTW, I know that there are developing countries with low tax/GDP ratios, but as I argued elsewhere they are “developing” for a reason. They have all sorts of hidden implicit MTRs that discourage work effort.]
In another post I pointed out that Singapore was still growing faster than other developed countries. Mbk’s data suggests why—they still haven’t caught up to the developed world in per capita productivity. When they do, they will probably have extremely high per capita GDP. And Singapore already has universal health care, despite the low tax rates. One more piece of evidence that forced saving is a much more efficient way to finance social insurance than taxes.
I still have an open mind on the question of taxes and work effort. But I tend not to put too much weight on dueling econometric studies. Call me a luddite, but in my view it’s hard to beat the following:
1. Find the best natural experiments you can.
2. Get the appropriate descriptive statistics.
3. Carefully think through the implications of various theories.
I almost never see conclusions reached this way overturned with fancy pants econometric studies. But I often see the reverse. I see sloppy thinking about the implications of theory. I see people “disproving” the Quantity Theory by testing whether nominal income and money are cointegrated. I see people testing the AS/AD model using unit root tests. And I see people applying micro studies of labor supply elasticity to aggregate tax changes. None of these tests are theoretically appropriate.
It might be fun to discuss counterexamples in the comment section. What are the theoretical models that cannot be demonstrated with the judicious use of descriptive statistics? Are there any?
PS. Mankiw reached the same conclusion that I did:
There is a literature about how and why the European workforce differs from the American workforce—specifically, why Europeans enjoy spending more time at the café than Americans do and why we work harder than they do. There are many hypotheses out there. Olivier Blanchard says that it stems from cultural tastes (Blanchard, 2004): Europeans have more joie de vivre than Americans, and therefore, they want to enjoy their high productivity by spending more time enjoying leisure. My colleagues Alberto Alesina and Ed Glaeser say that it is the presence and scope of powerful labor unions in Europe that have negotiated shorter work weeks, more vacation days, and so on. But Ed Prescott tells us it is the high tax rates in Europe (Prescott, 2004), and I actually find this argument the most compelling.
PPS. It was a busy Easter so this post was rushed. I will post updates to correct the errors that people will inevitably find.