Why Americans Work So Much
Edward C. Prescott
7
that the average labor supply (excluding the two outlier
observations) is close to the actual value for the other
12 observations.
Actual and Predicted Labor Supplies
Table 2 reports the actual and predicted labor supplies
for the G-7 countries in 1993–96 and 1970–74. For the
1993–96 period, the predicted values are surprisingly
close to the actual values with the average difference
being only 1.14 hours per week. I say that this number
is surprisingly small because this analysis abstracts from
labor market policies and demographics which have con-
sequences for aggregate labor supply and because there
are significant errors in measuring the labor input.
The important observation is that the low labor sup-
plies in Germany, France, and Italy are due to high tax
rates. If someone in these countries works more and
produces 100 additional euros of output, that individual
gets to consume only 40 euros of additional consumption
and pays directly or indirectly 60 euros in taxes.
In the 1970–74 period, it is clear for Italy that some
factor other than taxes depressed labor supply. This
period was one of political instability in Italy, and quite
possibly cartelization policies reduced equilibrium labor
supply as in the Cole and Ohanian (2002) model of the
U.S. economy in the 1935–39 period. The overly high
prediction for labor supply for Japan in the 1970–74
period may in significant part be the result of my util-
ity function having too little curvature with respect to
leisure, and as a result, the theory overpredicts when
the effective tax rate on labor income is low. Another
possible reason for the overprediction may be a measure-
ment error. The 1970–74 Japanese labor supply statistics
are based on establishment surveys only because at that
time household surveys were not conducted. In Japan
the household survey gives a much higher estimate of
hours worked in the period when both household- and
establishment-based estimates are available. In the other
Table 2
Actual and Predicted Labor Supply
In Selected Countries in 1993–96 and 1970–74
Labor Supply*
Differences
Prediction Factors
(Predicted Consumption/
Period Country Actual Predicted Less Actual) Tax Rate
�
Output (c/y)
1993–96 Germany
19.3 19.5
.2
.59
.74
France
17.5 19.5
2.0
.59
.74
Italy
16.5 18.8
2.3
.64
.69
Canada
22.9 21.3 –1.6
.52
.77
United Kingdom 22.8 22.8 0
.44
.83
Japan
27.0 29.0
2.0
.37
.68
United States 25.9 24.6 –1.3
.40
.81
1970–74 Germany
24.6 24.6
0
.52
.66
France
24.4 25.4
1.0 .49
.66
Italy
19.2 28.3
9.1
.41
.66
Canada
22.2 25.6
3.4
.44
.72
United Kingdom 25.9 24.0 –1.9
.45
.77
Japan
29.8 35.8
6.0
.25
.60
United States 23.5 26.4
2.9
.40
.74
*Labor supply is measured in hours worked per person aged 15–64 per week.
Sources: See Appendix.
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
8
countries household surveys are used to estimate labor
supply.
An important observation is that when European
and U.S. tax rates were comparable, European and U.S.
labor supplies were comparable. At the aggregate level,
where idiosyncratic factors are averaged out, people are
remarkably similar across countries. This is true not
only for the G-7 countries, but for Chile and Mexico as
shown by Bergoeing et al. (2002) and for Argentina as
shown by Kydland and Zarazaga (2002). Apparently,
idiosyncratic preference differences average out and
result in the stand-in household having almost identical
preferences across countries.
I am surprised that virtually all the large differences
between the U.S. labor supply and those of Germany
and France are due to differences in tax systems. I ex-
pected institutional constraints on the operation of labor
markets and the nature of the unemployment benefit
system to be of major importance. They do appear to be
important in Italy in the 1970–74 period.
Changes in U.S. Labor Supply
An interesting feature of the data is that U.S. labor in-
creased by 10 percent between 1970–74 and 1993–96,
yet the marginal tax rate on labor remained at 0.40. The
fact that all the increase in labor supply was by married
women and not by males or by single females suggests
that the appropriate marginal tax rate may have fallen
with the flattening of the income tax rate schedule asso-
ciated with the tax reforms of the 1980s, in particular, the
1986 tax reform (McGrattan and Rogerson 1998). The
U.S. Department of the Treasury (1974, 1996) lists the
number of married households’ tax returns by adjusted
gross income categories as well as reports the income
tax schedule. These data show that the marginal tax rate
for large changes in income such as those that would
occur from moving from a one-earner household to a
two-earner household was significantly higher in 1972
than it was in 1994.
Households switching from having one wage earner
to having two probably faced lower marginal tax rates
in the 1993–96 period than in the 1970–74 period, even
though the Feenberg-Coutts marginal income tax rates
are the same. This possibility is illustrated in Table 3 in
the example of a two-person household. In the early pe-
riod, if the working individual in the household increases
hours worked by a small amount, the marginal income
tax on the additional labor income is 20 percent, which is
the Feenberg-Coutts estimate for that period. However,
if the household doubles its labor supply by switching
from a one-earner to a two-earner household, the mar-
ginal income tax rate on the additional labor income is
40 percent for the numerical example in Table 3.
The situation is very different in 1993–96 when the
household has two earners. Small changes in labor
supply in this case are still subject to a 20 percent tax
rate as in the 1970–74 period, which is what the Feen-
berg-Coutts method finds for that period. However,
the marginal income tax on the labor income associ-
ated with switching from a one-earner to a two-earner
household is only 20 percent, not 40 percent as it was
in the 1970–74 period.
This issue of the effect of the nature of the income
tax schedule on labor supply for households with two
potential wage earners warrants more attention. Feld-
stein (1995) examines the consequences of the 1986 Tax
Reform Act using a U.S. Treasury Department panel of
more than 4,000 tax forms and finds micro evidence
consistent with this hypothesis. It is further supported
in the Feldstein and Feenberg (1993) analysis of the
Clinton Tax Plan.
Some macro evidence is provided by what happened
after the 1998 Spanish tax reform that flattened the Span-
ish income tax schedule in much the same way that the
1986 U.S. tax reform flattened the U.S. tax schedule.
Subsequently, Spanish labor supply increased by 12 per-
cent and tax revenue by a few percent. If the change in
the factor that converts the average income tax rate to a
marginal tax rate were the same in the United States and
Table 3
How a Flatter Income Tax Schedule
Affected U.S. Households With
Two Potential Wage Earners: An Example
Hypothetical Amounts
Assumed Rate
Number of of Income Tax
Earners in Labor
Period Household Income Taxes Average Marginal
Before
Tax Reform 1 10 1.3 13.0% 20.0%
(1970–74) 2 20 5.3 26.5 40.0
After
Tax Reform 1 10 1.5 10.0 20.0
(1993–96) 2 20 2.6 13.0 20.0