Why Americans Work So Much
Edward C. Prescott
5
labor income that is taken in the form of taxes, holding
investment, or equivalently savings, fixed. From the
household’s budget constraint,
(5)
�
�
�
�
=
+
+
h
c
c
1
.
Two first-order conditions are used to construct the
key equilibrium relation that is used to predict labor
supply. One is that the marginal rate of substitution
between leisure and consumption is equal to their price
ratio; that is,
(6)
�
�
(
)
/
(
) .
1
1
1
−
= −
h
c
w
The other is the profit-maximizing condition that the
wage equals the marginal product of labor; that is,
(7)
w
k h
y h
= −
= −
−
(
)
(
) / .
1
1
�
�
�
�
From equations (6) and (7), the key relation is obtained,
namely,
(8)
h
c
y
it
it
it
it
=
−
− +
−
1
1
1
�
�
�
�
.
This equilibrium relation clearly separates the inter-
temporal and intratemporal factors affecting labor sup-
ply. The intratemporal factor is captured by
1−
�
,which
distorts the relative prices of consumption and leisure
at a point in time. The c/y term captures intertemporal
factors. If, for example, the effective tax rate on labor
income is expected to be higher in the future, people
will choose a lower current value for c/y, and current
labor supply will be higher. The same is true if the cur-
rent capital stock is low relative to its balanced growth
path level. More formally, equilibrium c/y is a function
of the predictive probability distribution of future tax
rates and productivities and the current capital stock.
Knowing the value of this function and the current ef-
fective tax rate on labor income suffices for predicting
current labor income.
In focusing on the role of taxes in determining ag-
gregate labor supply, I am not implying that other factors
are unimportant. Cole and Ohanian (1999) and Chari,
Kehoe, and McGrattan (2003), using the discipline em-
ployed here, present strong evidence that other factors
were important in accounting for the low labor supply
in the United States in the 1930s. Similarly, Cole and
Ohanian (2002) present evidence that the low labor sup-
ply in the United Kingdom in the 1920s was due to other
factors, and Fisher and Hornstein (2002) find that labor
market distortions that increased the real wage signifi-
cantly above the competitive level were the major factor
in accounting for the huge decline in German output in
the 1928–32 period. In focusing on the role of marginal
tax rates on labor income, I want to determine what role,
if any, they play in accounting for the huge differences
in labor supplies across this relatively homogeneous set
of market economies at a point in time and in account-
ing for large changes in labor supplies over time across
these countries.
2
The theory abstracts from many features of reality
that affect labor supply, in particular, whether a married
household has one or two wage earners. This issue is
discussed briefly in the context of the change in the U.S.
labor supply in conjunction with the change in the nature
of the income tax schedule that occurred as a result of
the 1986 U.S. Tax Reform Act.
Estimating Tax Rates
The theory has the household paying the taxes. Conse-
quently, it is necessary to adjust the national income ac-
counts to be consistent with this theoretical framework.
The adjustment, which is a major one, is to treat indirect
taxes less subsidies as net taxes on final product. This
means removing net indirect taxes as a cost component
of GDP and reducing final product components.
In using SNA data to estimate tax rates and making
the distinction between prices facing producers and
consumers, I am following Mendoza, Razin, and Tesar
(1994). There are some important differences in the
approach with my estimated tax rates being in greater
part model-economy dependent. In what follows, the
capital letters are SNA statistics. I assume that two-thirds
of these indirect taxes net of subsidies fall directly on
private consumption expenditures and that the remaining
one-third is distributed evenly over private consump-
tion and private investment. Thus, net indirect taxes on
consumption,
IT ,
c
are
2
Three recent studies that address issues related to the ones considered in this
article are Davis and Henrekson 2003, Nickell 2003, and Olovsson 2003.
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
6
(9)
IT
C
C I
IT
c
=
+
+
2 3 1 3
/
/
where C is SNA private consumption expenditures, I is
SNA private investment, and
IT is net indirect taxes. The
motivation for this assignment of indirect taxes is that
most indirect taxes fall on consumption whether these
taxes are value-added taxes, sales taxes, excise taxes, or
property taxes. Some taxes, such as fuel taxes on diesel
fuel used by trucks that transport goods, property taxes
on office buildings, and sales taxes on equipment pur-
chases by businesses, fall on all forms of product.
The model economy’s consumption c and output y
are
(10)
c C G G
IT
mil
c
= + −
−
and
y GDP IT
=
−
where
G is public consumption,
G
mil
is military expen-
ditures, and GDP is gross domestic product.
My estimate of the consumption tax rate is
(11)
�
c
c
c
IT
C IT
=
−
.
There are two taxes on labor income, the income tax
with marginal rate
�
inc
and the social security tax with
marginal rate
�
ss
. My estimate of the social security tax
rate is simply
(12)
�
�
ss
GDP IT
=
−
−
Social Security Taxes
(1
) (
)
.
The denominator is labor income if labor is paid its
marginal product.
In some countries, some social security taxes are sav-
ings because benefits increase with income. But this is a
marginal tax rate. Often there are no additional benefits
to working an additional year. In the United States, the
marginal savings factor is tiny. First, when I use a 4 per-
cent discount rate and a 2 percent growth rate in the real
wage, which are numbers for the U.S. economy in the
twentieth century (McGrattan and Prescott 2003), the
present value of benefits is only one-quarter of the pres-
ent value of contributions. Second, the social security
benefit scheme is highly progressive. Third, benefits to
married couples typically go up little if both people work
rather than if only one works. Fourth, beginning in the
early 1990s, a significant part of social security benefits
is subject to income taxes for many people. Fifth, for
many older workers, their current-year taxable labor
income has little or no consequences for the retirement
benefits they receive.
Social security taxes are listed as an expenditure of
the household sector in the SNA. They include taxes
used to finance health care and unemployment pay-
ments, and not just taxes used to finance retirement
programs. These taxes are typically proportional taxes
on labor income, and they are treated as such in this
analysis. In the SNA, these taxes are treated as part
of compensation when they are paid by the employer,
which is typically the case.
The average, not marginal, income tax rate is
(13)
�
inc
GDP IT
=
−
−
Direct Taxes
Depreciation
.
Direct taxes are those paid by households and do not
include corporate income taxes. Like social security
taxes, they are listed as an expenditure of the household
sector in the SNA.
My estimate of the marginal labor income tax rate
is
(14)
�
�
�
h
ss
inc
=
+1 6
.
.
The most problematic number in my analysis is the 1.6
factor that reflects the fact that the marginal income tax
rates are higher than the average tax rates. I use 1.6 be-
cause it results in the marginal income tax rate obtained
using the Feenberg and Coutts (1993) methodology for
the United States in both the 1970–74 and 1993–96
periods. Feenberg and Coutts’ methodology uses a repre-
sentative sample of tax records to compute the marginal
tax rate on labor income by determining how much tax
revenue increases if every household’s labor income is
changed by 1 percent. The total change in tax receipts
divided by the total change in labor income is the Feen-
berg-Coutts estimate of the marginal income tax rate on
labor income. I will return to this point later.
Two parameters must be specified before formula (8)
can be used to predict labor supply. One is the capital
cost share parameter
�
in the production function. For
all the countries, in both periods this number is close to
the average of 0.3224, so
�
is set equal to this value.
The other parameter is the utility of leisure parame-
ter
�
. The value 1.54 for this parameter is chosen so