Reserve Federal Bank of Minneapolis Quarterly Review Vol. 28, No. 1, July 2004, pp. 213
Why Do Americans Work So Much More Than Europeans?
Edward C. Prescott Senior Monetary Adviser Research Department Federal Reserve Bank of Minneapolis and W.P. Carey Chair Department of Economics Arizona State University
Abstract Americans now work 50 percent more than do the Germans, French, and Italians. This was not the case in the early 1970s, when the Western Europeans worked more than Americans. This article examines the role of taxes in accounting for the differences in labor supply across time and across countries; in particular, the effective marginal tax rate on labor income. The population of countries considered is the G-7 countries, which are major advanced industrial countries. The surprising nding is that this marginal tax rate accounts for the predominance of differences at points in time and the large change in relative labor supply over time.
The views expressed herein are those of the author and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
Why Do Americans Work So Much More Than Europeans?*
Edward C. Prescott
Senior Monetary Adviser Research Department Federal Reserve Bank of Minneapolis and W.P. Carey Chair Department of Economics Arizona State University
Americans, that is, residents of the United States, work much more than do Europeans. Using labor market statistics from the Organisation for Economic Co-operation and Development (OECD), I nd that Americans on a per person aged 1564 basis work in the market sector 50 percent more than do the French. This was not always the case. In the early 1970s, Americans allocated less time to the market than did the French. The comparisons between Americans and Germans or Italians are the same. Why are there such large differences in labor supply across these countries? Why did the relative labor supplies change so much over time? In this article, I determine the importance of tax rates in accounting for these differences in labor supply for the major advanced industrial countries and nd that tax rates alone account for most of them. This nding has important implications for policy, in particular, for nancing public retirement programs, such as U.S. Social Security. On the pessimistic side, one implication is that increasing tax rates will not solve the problem of these underfunded plans, because increasing tax rates will not increase revenue. On the optimistic side, the system can be reformed in a way that makes the young better off while honoring promises to the old. This can be accomplished by modifying the tax system so that when an individual works more and produces
2
more output, the individual gets to consume a larger fraction of the increased output. The major advanced industrial countries (the G-7 countries) are the European countries France, Germany, Italy, and the United Kingdom, plus Canada, Japan, and the United States. Comparable and sufciently good statistics for these countries are available to carry out this investigation. The data sources are the United Nations system of national accounts (SNA) statistics and the OECD labor market statistics and purchasing power parity gross domestic product (GDP) numbers.1 The periods considered are 197074 and 199396. The later period was chosen because it is the most recent period prior to the U.S. telecommunications/dot-com boom of the late 1990s, a period when the relative size of unmeasured
*This is the 2002 Erwin Plein Nemmers Prize in Economics lecture, presented April 21, 2003, at Northwestern University. The author thanks Sami Alpanda, Simona Cociuba, T. C. Tong, and Alexander Ueberfeldt for excellent research assistantship, as well as the participants at lectures at Berlin, the Bank of England, Industry Canada, Tokyo University, the University of Toulouse, and the University of Illinois. The nancial support of the National Science Foundation under SES 9986667 is also acknowledged. 1 For Italy, GDP is reduced by 20 percent because Italys GDP statistics include estimates of the underground untaxed economy. The theory is concerned with the above-ground taxed economy, and I want GDP for this sector. This is why I do not follow Maddison (1995, pp. 24150) and increase the OECD labor supply numbers by 16.0 percent in the 197074 period and 17.6 percent in the 199396 period.
Why Americans Work So Much
Edward C. Prescott
output was probably signicantly larger than normal and there may have been associated problems with the market hours statistics. The earlier period was selected because it is the earliest one for which sufciently good data are available to carry out the analysis. The relative numbers after 2000 are pretty much the same as they were in the pretechnology boom period 199396. I emphasize that my labor supply measure is hours worked per person aged 1564 in the taxed market sector. The two principal margins of work effort are hours actually worked by employees and the fraction of the working-age population that works. Paid vacations, sick leave, and holidays are hours of nonworking time. Time spent working in the underground economy or in the home sector is not counted. Other things equal, a country with more weeks of vacation and more holidays will have a lower labor supply in the sense that I am using the term. I focus only on that part of working time for which the resulting labor income is taxed. Table 1 reports the G-7 countries output, labor supply, and productivity statistics relative to the United States for 199396 and 197074. The important observation for the 199396 period is that labor supply (hours per person) is much higher in Japan and the United States than it is in Germany, France, and Italy. Canada and the United Kingdom are in the intermediate range. Another observation is that U.S. output per person is about 40 percent higher than in the European countries, with most of the differences in output accounted for by differences in hours worked per person and not by differences in productivity, that is, in output per hour worked. Indeed, the OECD statistics indicate that French productivity is 10 percent higher than U.S. productivity. In Japan, the output per person difference is accounted for by lower productivity and not by lower labor supply. Table 1 shows a very different picture in the 197074 period. The difference is not in output per person. Then, European output per person was about 70 percent of the U.S. level, as it was in 199396 and is today. However, the reason for the lower output in Europe is not fewer market hours worked, as is the case in the 199396 period, but rather lower output per hour. In 197074, Europeans worked more than Americans. The exception is Italy. What caused these changes in labor supply? Theory Used To account for differences in the labor supply, I use the standard theory used in quantitative studies of business cycles (Cooley 1995), of depressions (Cole and Ohanian
3
Table 1
Output, Labor Supply, and Productivity
In Selected Countries in 199396 and 197074 Relative to United States (U.S. = 100)
Period Country Output per Person* Hours Worked per Person* Output per Hour Worked
199396 Germany France Italy Canada United Kingdom Japan United States 197074 Germany France Italy Canada United Kingdom Japan United States
*These data are for persons aged 1564. Sources: See Appendix.
74 74 57 79 67 78 100 75 77 53 86 68 62 100
75 68 64 88 88 104 100 105 105 82 94 110 127 100
99 110 90 89 76 74 100 72 74 65 91 62 49 100
1999 and Kehoe and Prescott 2002), of public nance issues (Christiano and Eichenbaum 1992 and Baxter and King 1993), and of the stock market (McGrattan and Prescott 2000, 2003 and Boldrin, Christiano, and Fisher 2001). In focusing on labor supply, I am following Lucas and Rapping (1969), Lucas (1972), Kydland and Prescott (1982), Hansen (1985), and Auerbach and Kotlikoff (1987). This theory has a stand-in household that faces a labor-leisure decision and a consumption-savings decision. The preferences of this stand-in household are ordered by (1) E t ( log ct + log(100 ht ) ) . t =0
Variable c denotes consumption, and h denotes hours of labor supplied to the market sector per person per week. Time is indexed by t. The discount factor 0 < < 1
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
species the degree of patience, with a higher value indicating more patience for consumption and leisure. The parameter > 0 species the value of nonmarket productive time for the household. Given that on a per person basis a household has about 100 hours of productive time per week, nonmarket productive time is 100 h hours per week per working-age person in the household. Following the tradition in macroeconomics, this nonmarket productive time will be referred to as leisure even though much of it is time allocated to working in the nonmarket sector and in the underground market sector. The important thing for the analysis is that any production using this time is not taxed. In the model economy, the household owns the capital and rents it to the rm. This is an assumption of convenience because the ndings are identical if the rm owns the capital and the household owns the rm, or if the rm is partially debt nanced. The law of motion governing the capital stock is (2) kt +1 = (1 ) kt + xt
where k is the capital stock, x is investment, and is the depreciation rate. The theory also has a stand-in rm with a CobbDouglas production function, (3) yt = ct + xt + gt Ait kt ht1 .
Here y denotes output, c consumption, and g pure public consumption. The capital share parameter is 0 < < 1, and the total factor productivity parameter of country i at date t is Ait . I will not specify the process on{Ait} because it plays no role in the inference being drawn, except to implicitly restrict the process governing its evolution in a way that results in the existence of a competitive equilibrium. The households date t budget constraint is (4) (1 + c ) ct + (1 + x ) x t = (1 h ) wt ht + (1 k )(rt ) kt + kt + Tt where wt is the real wage rate, rt the rental price of capital, c the consumption tax rate, x the investment tax rate, h the marginal labor tax rate, k the capital income tax rate, and Tt transfers. I emphasize that the marginal and average labor income taxes will be very different. All tax revenue except for that used to nance the
4
pure public consumption is given back to the households either as transfer payments or in-kind. These transfers are lump sum, being independent of a households income. Most public expenditures are substitutes for private consumption in the G-7 countries. Here I will assume that they substitute on a one-to-one basis for private consumption with the exception of military expenditures. The goods and services in question consist mostly of publicly provided education, health care, protection services, and even judiciary services. My estimate of pure public consumption g is two times militarys share of employment times GDP. In having only one consumption good, I am following Christiano and Eichenbaum (1992). Rogerson (2003) nds that this one-consumption-good abstraction is not a good one for studying aggregate labor supply in the Scandinavian countries. One possible reason is that some publicly provided goods, such as child care for working parents, must be treated as a separate good. Often the receipt of this good is contingent on working, and this must be taken into account in the households constraint set. However, the one-consumption-good abstraction used in this study is a reasonable one for the set of countries considered. This is a far simpler tax system than the one employed in any of the G-7 countries. Introducing accelerated depreciation and investment tax credits would affect the price of the investment good relative to the consumption good, but would not alter the inference drawn in this article. Similarly, introducing a corporate sector, with dividends not taxed, as is generally the case in Europe, or taxed as ordinary income, as they are in the United States, would not alter any conclusion signicantly. For further details on these issues, see McGrattan and Prescott 2002. What is important here is the price of consumption relative to leisure, and it is determined by the consumption tax rate c and the marginal labor income tax rate h . The most important parameter that will enter the equilibrium relation that I use to predict the consequences of the tax system is the utility of leisure preference parameter , which measures the value of leisure relative to consumption. The capital cost share parameter also enters the relation, but is of less importance. The labor and consumption tax rates can be combined into a single tax rate ,which I call the effective marginal tax rate on labor income. It is the fraction of additional
Key Equilibrium Relation
Why Americans Work So Much
Edward C. Prescott
labor income that is taken in the form of taxes, holding investment, or equivalently savings, xed. From the households budget constraint, (5)
=
h +c . 1+c
Two rst-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 h) = (1 ) w. 1/ c
The other is the prot-maximizing condition that the wage equals the marginal product of labor; that is, (7) w = (1 )k h = (1 ) y / h.
From equations (6) and (7), the key relation is obtained, namely, (8) hit = 1 . c 1 + it yit 1 it
Kehoe, and McGrattan (2003), using the discipline employed 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 supply in the United Kingdom in the 1920s was due to other factors, and Fisher and Hornstein (2002) nd that labor market distortions that increased the real wage signicantly above the competitive level were the major factor in accounting for the huge decline in German output in the 192832 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 accounting 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 briey 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. Consequently, it is necessary to adjust the national income accounts 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 nal product. This means removing net indirect taxes as a cost component of GDP and reducing nal 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 consumption and private investment. Thus, net indirect taxes on consumption, ITc , 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.
This equilibrium relation clearly separates the intertemporal and intratemporal factors affecting labor supply. 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 current 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 effective tax rate on labor income sufces for predicting current labor income. In focusing on the role of taxes in determining aggregate labor supply, I am not implying that other factors are unimportant. Cole and Ohanian (1999) and Chari,
5
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
(9)
C IT ITc = 2 / 3 + 1 / 3 C+I
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 ofce buildings, and sales taxes on equipment purchases by businesses, fall on all forms of product. The model economys consumption c and output y are (10) c = C + G Gmil ITc and y = GDP IT
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 benets they receive. Social security taxes are listed as an expenditure of the household sector in the SNA. They include taxes used to nance health care and unemployment payments, and not just taxes used to nance 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 =
where G is public consumption, Gmil is military expenditures, and GDP is gross domestic product. My estimate of the consumption tax rate is (11)
Direct Taxes . GDP IT Depreciation
c =
ITc . C ITc
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)
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)
h = ss + 1.6 inc .
ss =
Social Security Taxes
(1 ) (GDP IT )
.
The denominator is labor income if labor is paid its marginal product. In some countries, some social security taxes are savings because benets increase with income. But this is a marginal tax rate. Often there are no additional benets to working an additional year. In the United States, the marginal savings factor is tiny. First, when I use a 4 percent 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 benets is only one-quarter of the present value of contributions. Second, the social security benet scheme is highly progressive. Third, benets to married couples typically go up little if both people work rather than if only one works. Fourth, beginning in the early 1990s, a signicant part of social security benets
6
The most problematic number in my analysis is the 1.6 factor that reects the fact that the marginal income tax rates are higher than the average tax rates. I use 1.6 because it results in the marginal income tax rate obtained using the Feenberg and Coutts (1993) methodology for the United States in both the 197074 and 199396 periods. Feenberg and Coutts methodology uses a representative sample of tax records to compute the marginal tax rate on labor income by determining how much tax revenue increases if every households labor income is changed by 1 percent. The total change in tax receipts divided by the total change in labor income is the Feenberg-Coutts estimate of the marginal income tax rate on labor income. I will return to this point later. Two parameters must be specied 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 parameter . The value 1.54 for this parameter is chosen so
Why Americans Work So Much
Edward C. Prescott
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 199396 and 197074. For the 199396 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 consequences for aggregate labor supply and because there are signicant errors in measuring the labor input. The important observation is that the low labor supplies 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 197074 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 193539 period. The overly high prediction for labor supply for Japan in the 197074 period may in signicant part be the result of my utility 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 measurement error. The 197074 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 199396 and 197074
Labor Supply*
Period Country Actual Predicted
(Predicted Less Actual)
Differences
Prediction Rate Factors
Tax Consumption/ Output (c/y)
199396 Germany France Italy Canada United Kingdom Japan United States 197074 Germany France Italy Canada United Kingdom Japan United States
19.3 17.5 16.5 22.9 22.8 27.0 25.9 24.6 24.4 19.2 22.2 25.9 29.8 23.5
19.5 19.5 18.8 21.3 22.8 29.0 24.6 24.6 25.4 28.3 25.6 24.0 35.8 26.4
.2 2.0 2.3 1.6 0 2.0 1.3 0 1.0 9.1 3.4 1.9 6.0 2.9
.59 .59 .64 .52 .44 .37 .40 .52 .49 .41 .44 .45 .25 .40
.74 .74 .69 .77 .83 .68 .81 .66 .66 .66 .72 .77 .60 .74
*Labor supply is measured in hours worked per person aged 1564 per week. Sources: See Appendix.
7
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
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 expected institutional constraints on the operation of labor markets and the nature of the unemployment benet system to be of major importance. They do appear to be important in Italy in the 197074 period. Changes in U.S. Labor Supply An interesting feature of the data is that U.S. labor increased by 10 percent between 197074 and 199396, 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 attening of the income tax rate schedule associated 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 signicantly 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 199396 period than in the 197074 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 period, 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,
8
if the household doubles its labor supply by switching from a one-earner to a two-earner household, the marginal income tax rate on the additional labor income is 40 percent for the numerical example in Table 3. The situation is very different in 199396 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 197074 period, which is what the Feenberg-Coutts method nds for that period. However, the marginal income tax on the labor income associated with switching from a one-earner to a two-earner household is only 20 percent, not 40 percent as it was in the 197074 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. Feldstein (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 nds 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 attened the Spanish income tax schedule in much the same way that the 1986 U.S. tax reform attened the U.S. tax schedule. Subsequently, Spanish labor supply increased by 12 percent 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 Labor Income Taxes Assumed Rate of Income Tax Average Marginal
Period
Number of Earners in Household
Before Tax Reform (197074) After Tax Reform (199396)
1 2 1 2
10 20 10 20
1.3 5.3 1.5 2.6
13.0% 26.5 10.0 13.0
20.0% 40.0 20.0 20.0
Why Americans Work So Much
Edward C. Prescott
Spain and sufciently large to increase U.S. labor supply by 10 percent, then the predicted increase in Spanish labor supply would be the observed 12 percent. More research is needed to determine whether the hypothesis that the attening of the tax schedule is the principal reason for the large increases in labor supply in both the United States and Spain after their tax reforms. The welfare gains from reducing the effective marginal tax rate on labor income in the high tax rate countries are large. The measure of welfare used is the standard one, namely, by what percentage consumption today and in all future periods must be increased in order that the households would be indifferent to the policy change in question. This measure is called the lifetime consumption equivalent measure. If France were to reduce its effective tax rate on labor income from 60 percent to the U.S. 40 percent rate, the welfare of the French people would increase by 19 percent in terms of lifetime consumption equivalents. This is a large number for a welfare gain. This measure of the welfare gain takes into consideration the reduction in leisure associated with the change in the tax system and the cost of accumulating capital associated with the higher balanced growth path. The reduction in leisure is from 81.2 hours a week to 75.8 hours, which is a 6.6 percent decline in leisure. I am surprised to nd that this large tax rate decrease did not lower tax revenues.3 The welfare gains if the United States reduced its marginal tax rate on labor income are smaller. If the tax rate is reduced from 40 percent to 30 percent, the gains in terms of lifetime consumption equivalents are 7 percent. Implications for Policy Tax system modications have implications for public retirement programs, such as U.S. Social Security. If labor supply is xed, a pay-as-you-go social security system cannot be converted to a fully funded system in a way that makes every generation better off. If, however, the labor supply is not xed, the transition can be made in a way that makes every generation better off. The only issue is how long the transition will take. Using the utility of leisure parameter, , obtained in the rst part of this article, I now explore this issue of how long such a transition will take. The model economy is modied is two respects. First, I follow Auerbach and Kotlikoff (1987) and use the overlapping generations structure rather than the innitely lived family structure employed earlier.4 In the modied
9
structure, the key relation used to forecast labor supply continues to hold. Second, the technology assumed has perfect substitution between capital and labor. The productivity of labor grows at the rate of 2 percent a year, which implies that the real wage will grow at 2 percent a year as it has on average throughout the twentieth century. The productivity of capital is constant and is such that the after-tax return is 4 percent. Alternatively, I could have assumed that capital income tax rates, which are not formally modeled, are adjusted to maintain a 4 percent return on capital if the capital/output ratio changes as a result of the reform. This 4 percent return is the after-tax real return that has prevailed in the United States in the 18802002 period (McGrattan and Prescott 2003). Having some dynastic families would also work in the direction of keeping the interest rate constant. I assume that an equal number of people begin their working career every year at age 22, they work for 41 years, and then they live an additional 19 years. This implies that they retire at 63, which is the average U.S. retirement age. They receive social security benets equal to 0.319 of the wage that prevailed when they were 66 beginning when they are 67 and continuing for 14 additional years. In fact, for the U.S. system, the wage base is the one that prevailed when an individual was 60 years old, so the replacement rate is approximately 36 percent. The effective tax rate on labor income is 40 percent, as it is in the United States, with 10 percent of this being a social security retirement tax. I use 10 percent rather than the U.S. 12.4 percent rate because some social security taxes are used to provide disability and survivors benets in the United States. The assumption of no population growth is not realistic and introduces two errors. These errors, however, are of opposite sign and offsetting, so my example is still valid for building quantitative economic intuition. One error is that the relative number of people with social security claims is smaller if population growth is positive. This reduces the initial implicit liabilities relative to GDP of the pay-as-you-go system. The other
3 Mendoza and Tesar (2002) also nd that revenue is maximized with a tax rate slightly above 50 percent. 4 See the July 1999 issue of the Review of Economic Dynamics, which is devoted entirely to studies of the U.S. Social Security system. These studies are much richer in detail than this one. But they do not use the utility function used in this study, and as a result, my results are different. Conesa and Garriga (2003) address the status quo problem in Social Security reform.
FEDERAL RESERVE BANK OF MINNEAPOLIS
QR
Table 4
Effects of a Shift to a Fully Funded Social Security System
Steady States in a Model With Each System
Capital System Output Output
Labor Supply
Consumption Per Person
Soc. Sec. Liabilities Net Output Welfare*
Effective Labor Tax Rate
Pay-As-You-Go Fully Funded
100 123
2.77 4.91
100 111
100 117
4.62 0
100 109
40.0% 27.05
*Welfare here is measured in lifetime consumption equivalents.
error is that with a growing population the pay-as-yougo system will have higher levels of benet payments associated with a given social security retirement tax. This increases the implicit liabilities of the current system. The pay-as-you-go system that I consider has the property that social security benets paid are equal to social security taxes collected. The model economys time period is a year. The steady state of a pay-as-you-go system and a fully funded system are reported in Table 4. With the fully funded system, steady-state labor supply is 11 percent higher, consumption 17 percent higher, and welfare in lifetime consumption equivalents 9 percent higher. The problem with just switching from the current pay-asyou-go system to a fully funded system is that the initial old would suffer. The following reform makes all better off. There are still better reforms than this one, in particular, plans that have tax rates that depend upon age at the time of reform.
Proposed Reform
social security transfers are left unchanged. Steady-state social security liabilities of the payas-you-go system are large: 4.62 times gross national income (GNI). With the reform, those aged 37 and younger choose the new system. The welfare gain to the 22-year-old at the time of the change exceeds 4 percent in lifetime consumption equivalents. Associated with the change, the annual capital/output ratio increases from 2.7 to 3.3, as seen in Table 5. This increase takes 45 years. Table 5 also shows that pension liabilities of the payas-you-go system are large: 2.30 times GNI. With the new system, the decline steadily becomes zero 35 years after the reform.
Some Equity Considerations
People are given the option to continue with the current system or to shift to a new system. With the new system, 8.7 percent of wage income is put into an individual account with the government that earns a 4 percent real return. Upon retirement, savings in this account are annuitized. Effectively, people have the option to have their tax rate on labor reduced from 40 percent to 31.3 percent and to save 8.7 percent of their labor income in a government retirement account or to continue with the current social security system. With the reform, non
In the model, all individuals earn the same wage when, in fact, some people earn higher wages than others. Given that earning a 4 percent after-tax real return is an attractive investment, equity considerations suggest an upper bound on contributions. Similarly, lower-income households should have the right to contribute more than 8.7 percent of their labor income. Still another consideration is how to deal with married couples. An equitable solution is that each party has an account and household contributions are split equally between the two accounts with the contribution limit discussed above applying to an individual account and not to a household account. Some will be so unfortunate that the amount in their account will be insufcient to provide for a minimal acceptable retirement. This sug10
Why Americans Work So Much
Edward C. Prescott
Table 5
Effects of a Shift to an Optimal Government Individual Retirement Account
For 60 Years After Reform, Assuming Workers Aged 1537 Years Choose the New Account
Soc. Sec. Liabilities Year Output
Capital Output
1 15 30 45 60
2.30 ...