High levels of government expenditure pervade all modern industrialized economies. In 2006, total government expenditures as a fraction of gross domestic product (GDP) averaged more than 40 percent in countries belonging to the Organization for Economic Cooperation and Development (OECD).¹ In several countries, including France, Italy, and Sweden, this ratio exceeded 50 percent. Although this fraction is lower in the United States than in most other advanced economies, even here it exceeded 35 percent. Government spending funds many different programs and activities, including entitlement programs such as Social Security and Medicare, social insurance programs such as unemployment insurance and disability insurance, and services such as healthcare, education, and national defense. An important public policy issue in all economies is to determine the scale at which these programs should operate. These permanent or long-run decisions about government policy play a significant role in defining the overall economic climate of the United States and thus have a large influence on the long-run economic well-being of Americans.

The United States faces important decisions about the size of its government spending programs. Pressure for change comes in part from the budgetary imbalances associated with changing demographics—the aging of the baby boomers and increasing life expectancies—as well as from the increasing relative cost of healthcare. Effective public policy decisions about the longer-run scale of various government programs require a careful assessment of both the costs and benefits associated with the size of these programs and how they are funded.

Something has changed in Europe relative to the United States that has led to these very different changes in the amount of work being done.

Central to understanding these costs is the fact that the revenues funding government programs are raised largely by taxes on labor. It follows that an unfortunate and unintended consequence of expanded government spending is a disincentive for individuals to work. Because labor is the dominant input used in producing output, less labor implies less output, which in turn means that overall consumption and living standards must decrease. Or, as some commentators put it, expanding the size of government programs shrinks the size of the pie that everyone must share.

But how large is this disincentive effect? Is it of first-order importance for policy makers, or can it safely be ignored? Many policy makers seem to believe that the effect is small and can therefore be ignored. This belief is seriously mistaken. The belief that disincentive effects are small tends to be based on studies using unreliable methods applied to U.S. data. I argue that the most valuable and reliable information about these disincentive effects is found outside of the United States.

To appreciate why it is important to look abroad for evidence of these effects, it is important to first understand why the United States is not a good source of evidence, and why many policy makers are drawing unwarranted conclusions from U.S. data. Because there have been changes in the scale and design of government spending in the United States over the last 50 or so years, in principle the historical data could tell us about the magnitude of the effects of labor taxes on hours of work. But certain realities serve to obscure these disincentive effects in the U.S. data. Central among them is that the changes in U.S. programs over the last 50 years have not been particularly large, and that many other changes have also affected the economy during this time. When many changes occur simultaneously, it is difficult to reliably determine the contribution of any one particular factor to the overall economic changes that we observe. Some would say that we have a situation where the signal-to-noise ratio is relatively small.

Many policy makers are drawing unwarranted conclusions from U.S. data.

Failure to appreciate this fact has led some observers to argue that higher taxes on labor do not lead people to work less. In the context of recent debates about the economy, several policy makers have observed that labor taxes increased during the 1990s, but that total hours of work in the U.S. economy nonetheless grew substantially. This is proposed as evidence that higher taxes do not create a disincentive for work. But this argument neglects to take into account that something else very substantial was also affecting the U.S. economy at this time. In particular, the 1990s were characterized by high investment levels, high productivity growth, and high output growth. (Some proclaimed the advent of a “New Economy.”) It is hardly surprising that total hours of work increased during a period featuring an investment boom and high productivity growth. No one suggests that the high rate of investment and high productivity growth were due to the increase in taxes; that is, they are thought to be the result of other factors. It follows that economic outcomes in general in this period, including hours worked, were probably dominated by these other factors. In short, this episode provides no information about the disincentive effects of higher labor taxes.

So why should we turn to the experiences of other countries for evidence? The most transparent source of information would come from a situation where large permanent changes in the scale of tax and transfer programs have occurred, where these changes were large compared to other changes in the economy over the same period, and where there has been sufficient time for individuals to respond to the changes. While the United States does not fit this ideal situation, it turns out that these three conditions are much more likely to be met in some other advanced economies.

Labor Taxes in the OECD

Table 1 shows the effective average tax rate on labor in several OECD countries in 1960, 1980, and 2000. In each case the value is the average for a five-year interval centered on the year in question.

The last row of this table is striking—it shows that in terms of averages, tax rates on labor increased by more than 16 percentage points over this time period, almost three times the increase that was observed in the United States. In fact, the United States was the only country to have an increase that was less than 10 percentage points. Some countries even had increases in excess of 20 percentage points. The table also shows that these increases have been sustained, since almost two-thirds of the overall increase takes place during the first 20-year period. It follows that other countries are much more likely to provide a cleaner look at the effects of higher tax rates on aggregate hours of work.

Hours Worked in the OECD

Table 2 shows hours worked in 1960, 1980, and 2000 for the same set of countries as Table 1.

Note: For 1980 and 2000, the value is the average over a five-year interval centered on 1980 and 2000, respectively. Because data is only available from 1960 onward for some countries, the 1960 value is an average over the period 1960–62.

These data are interesting in and of themselves, and so it is worthwhile to take some time to discuss the patterns that they exhibit. Let’s begin by looking at what happened to the average value for hours worked. What we see is that hours worked decreased dramatically from 1960 to 1980, followed by a small decrease thereafter. The overall drop in hours from the beginning to the end of the period is more than 18 percent. This is an enormous drop in hours worked. Consider by way of comparison the labor market fluctuations associated with the business cycle. Going from normal times to a fairly severe recession is usually associated with a drop in total hours worked of about 3 percent. The size of the average drop that we see across countries is more than six times as large. So this drop in hours worked over time is very dramatic.

The second striking pattern in these data is the dramatic differences in the overall change in hours worked across countries. At one extreme is the United States, which actually witnessed an increase of 10 percent between these two dates, and at the other extreme are Germany and France, with declines of more than 30 percent. If we contrast the differential changes between the United States and France, the difference is staggering—more than 45 percent.

While the United States is at one extreme, it is important to note that it is not an outlier. Canada also displayed substantial growth in hours of work, and Australia had only a very small decrease. Moreover, even among those countries that exhibited substantial decreases in hours worked, there is still a lot of variation. Switzerland, for example, had a decrease in hours worked of 17 percent, which is much less than what occurred in France, Germany, and Belgium.

We can now examine what the data say about the effect of changes in labor taxes on hours of work. It is instructive to begin with a look at what happened to the simple averages across countries for both labor taxes and hours worked. Between 1960 and 2000 the labor tax rate increases by 16.5 percentage points, and hours of work decrease by 18.7 percent. This suggests a strong negative effect of taxes on hours of work.

While it is interesting to look at the averages, a more powerful test is to examine the pattern of changes across countries. That is, to what extent do countries with larger increases in taxes also have larger decreases in hours worked? This first figure provides a graphical look at the relationship. It plots the data for changes in tax rates and percentage change in hours of work for each of the 15 countries. The picture shows a clear negative relationship: the correlation between the change in tax rate and the percentage change in hours of work is equal to –0.30.²

Cultural Differences

If we look at the data shown for the year 2000 in Table 2, we see that hours worked are lowest in the continental European economies of Belgium, France, Germany, and Spain, are at intermediate levels in the United Kingdom, Finland, and Sweden, and are the highest in the United States, Canada, Australia, Japan, and Switzerland. Many commentators are content to explain away these differences as due to cultural differences; the idea is that people from some countries either enjoy leisure more or focus less on work.

This view leads some people to argue that U.S. policy makers should not look to these other countries for information about how tax policies influence hours of work. They would argue that many Europeans work less than Americans not because of high taxes on labor and generous transfer systems, but instead because of systematically different preferences toward work and consumption across countries.

A closer look at the data suggests that this is not a very compelling explanation. In 1960, hours of work were actually higher in Germany, France, and Belgium than they were in Canada, the United States, and Australia. That is, 50 years ago the relative work levels of these countries were reversed. This evidence seems inconsistent with the view that Europeans work less because they either value leisure more or do not care so much about consumption.

Since that time, hours have increased somewhat in the United States, but decreased dramatically in most European countries. The question is why. Something has changed in Europe relative to the United States that has led to these very different changes in the amount of work being done. I conclude that the dominant force behind these very different changes is the relative changes in the rate at which labor is taxed.

Home versus Market Production

There is yet other evidence that higher taxes are largely responsible for the significant differences in hours worked across countries. In this section I consider some related evidence that has to do with cross-country differences in time devoted to household production.

Specifically, if taxes increase, this creates an incentive for individuals to do more things for themselves rather than purchase them in the market. The intuition is simple: if you are working to purchase something in the market, then higher taxes imply that you have to work more hours in order to earn enough money to make the purchase. Since time spent in home production is not taxed, higher taxes make it more economical to do things for oneself rather than purchasing services through the market.

It follows that, holding all else constant, we should expect to see more time devoted to home production in an economy with a higher tax rate on labor than in an economy with a lower tax rate. And if the effects of taxes are sizable, then these differences should also be sizable. Time allocated to home production has changed over time in the United States. Unfortunately, the data that would permit a cross-country comparison of time series changes in home production are not available. However, four recent papers look at available time-use data to provide a recent snapshot of how time devoted to market and home production differ across some countries.

Americans devote more time to market work and less time to home production than do Germans.

A common finding is that differences in market work are indeed significantly offset by differences in home production. Richard Freeman and Ronald Schettkat (2001) study time allocation data for married couples in Germany and the United States in the 1990s and find that Americans devote more time to market work and less time to home production than do Germans. The striking finding is about total time devoted to work (i.e., market work plus home production): it turns out that the two countries are virtually the same. This study also shows that the patterns of consumer expenditure differ in a corresponding fashion, i.e., Germans spend more time preparing meals at home and spend less money at eating establishments.

Freeman and Schettkat (2005) extend this analysis to a larger set of countries and report that, as of the early 1990s, time spent in home production in European countries is about 20 percent higher than in the United States. This implies that increased time in home production only partially offsets the decrease in time devoted to market work.

Using data from the recent Harmonized European Time Use Study, Kelly Ragan (2005) compares several European countries with the United States and finds that on average, individuals in Belgium, France, Germany, Italy, and the Netherlands devote between 15 and 20 percent more time to home production than do Americans.³

In another study of time-use data, Michael Burda, Daniel Hamermesh, and PhilippeWeil (2008) reach a similar conclusion based on information for Germany, Italy, the Netherlands, and the United States. In particular, they find that Europeans spend 15 to 20 percent more time in home production than do Americans.⁴

Related work has also been carried out by Steven Davis and Magnus Henrekson (2004). Consistent with the tax effects on home versus market production discussed above, they show that countries with higher marginal tax rates systematically have lower employment in those market activities for which there are good nonmarket substitutes. The magnitude of the estimated effects are large. An increase in taxes of one-quarter of one percent leads to a decrease equal to 2.4 percent in the employment share in the broad set of sectors that have good home-produced substitutes. They find that tax effects are most noticeable in precisely these sectors.

The Netherlands

The Netherlands turns out to be the only country in our sample that has experienced both a persistent increase in taxes and a persistent decrease in taxes. All of the other countries have persistent increases in taxes, and differ from one another only in the extent of the increase. My previous analysis has shown that those countries with the largest increases in taxes also tended to have the largest declines in hours of work. But there would be a stronger case for taxes as the dominant factor behind these decreases in hours if there were also evidence that decreases in taxes lead to increases in hours worked.

But while we have examples in the data that allow us to compare economies that have increased taxes by different amounts, the only instance of a persistent decline in tax rates is in the Netherlands. This next figure shows the trend behavior of the average labor tax rate for the Netherlands. We can see that the Netherlands has a large and sustained increase in taxes for roughly the first 25 years of the sample, followed by a smaller but still substantial and sustained decrease in tax rates over the final 20 years.

The third figure, below, presents the time series behavior of hours worked in the Netherlands. It shows a dramatic and sustained decrease over the first part of the sample, followed by a smaller but sustained increase in the later part of the period. The increase in hours begins about five years after the decrease in tax rates, but given that there is some noise in the tax rate measures this lag of five years should not be taken too literally; the actual maximum value of the tax rate occurred in 1983. The fact that there is some lag between the tax changes and the response in hours is consistent with ideas that we emphasized earlier, including the point that there needs to be sufficient time following a tax change for the effects to take place. In any case, the time series changes for taxes and hours of work in the Netherlands seem very persuasive evidence regarding the effect of taxes on hours of work: just as tax increases have a negative effect on hours worked, tax decreases seem to have a positive effect.

Data from countries other than the United States are likely to contain much better information on the effect of increases in labor taxes on hours worked. The reason for this is that labor taxes have increased much more in these other countries than in the United States. Whereas the changes in the United States are not large enough to dominate other effects during the same time period, in many other countries the tax changes are as much as three times larger. When we compare outcomes across countries, we find sharp evidence that greater increases in labor taxes lead to greater decreases in hours of work. These changes in other countries are not likely to be due to changes in other policy or institutional factors, such as employment protection or unionization, or to cultural differences. Consistent with the notion that the changes in hours worked are primarily due to changes in taxes, we find that in countries with higher labor tax rates, individuals devote less time to market work and more time to home production.

Richard Rogerson is the Rondthaler Professor of Economics and a Regents’ Professor at the W.P. Carey School of Business at Arizona State University. This essay is adapted from his new book, The Impact of Labor Taxes on Labor Supply: An International Perspective (AEI Press, 2010).

FURTHER READING: Andrew Biggs says “Spending, Not Tax Cuts, Is the Real Driver of the Fiscal Mess” and asks “Should We Raise Taxes on the Middle Class? We Already Are.” Alan Viard discusses “Keynes at the Border” in tax policy, while Michael Barone says the “Obama Economy Sends Americans to Their Mattresses” and Andreas Bergh and Magnus Henreckson describe “Lessons from the Swedish Welfare State.”

Image by Rob Green/Bergman Group.

Notes

1. The thirty member countries of the OECD include the richest countries in North America, Europe, and Asia and are typically identified as the most advanced economies in the world. All statistics in this paragraph come from the OECD.

2. Excluding Sweden, the correlation coefficient is –.50. The next chapter discusses why Sweden should be viewed differently.

3. Alesina, Glaeser, and Sacerdote (2005) present data from another source which challenges this conclusion. These authors note, however, their data set seems ill-suited to cross-country comparisons. The Harmonised European Time Use dataset used by Ragan (2005) was designed to specifically address the shortcomings mentioned by Alesina and others and hence seems more reliable.

4. In comparing countries using the 2003 data, it is important to recall the changes in the U.S. survey design mentioned in chapter 4: relative to earlier surveys in the U.S., the American Time Use Survey, initiated as part of the Current Population Survey, tends to generate larger amounts of time reported to child care. In the U.S. this results in an almost 50 percent increase in time devoted to child care relative to the 1985 time-use survey data.