The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Business Review Federal Reserve Bank o f Philadelphia March *April 1992 ISSN 0 0 0 7 -7 0 1 1 Business Review The BUSINESS REVIEW is published by the Department of Research six times a year. It is edited by Sarah Burke. Artwork is designed and produced by Dianne Hallowell under the direction of Ronald B. Williams. The views expressed here are not necessarily those of this Reserve Bank or of the Federal Reserve System. SUBSCRIPTIONS. Single-copy subscriptions for individuals are available without charge. Insti tutional subscribers may order up to 5 copies. BACK ISSUES. Back issues are available free of charge, but quantities are limited: educators may order up to 50 copies by submitting requests on institutional letterhead; other orders are limited to 1 copy per request. Microform copies are available for purchase from University Microfilms, 300 N. Zeeb Road, Ann Arbor, MI 48106. REPROD U CTIO N . Perm ission must be obtained to reprint portions of articles or whole articles. Permission to photocopy is unrestricted. Please send subscription orders, back orders, changes of address, and requests to reprint to Publications, Federal Reserve Bank of Philadelphia, Department o f Research and Statistics, Ten Independence Mall, Philadelphia, PA 19106-1574, or telephone (215) 574-6428. Please direct editorial communications to the same address, or telephone (215) 574-3805. 2 MARCH/ APRIL 1992 ARE REGIONAL PER CAPITA EARNINGS DIVERGING? Gerald A. Carlino For several decades, per capita earnings converged across regions. But since 1978, the trend has reversed, and regional per capita earnings are diverging. Does this divergence indicate a permanent change, or is it merely a response to temporary changes in certain regional factors? Re gional per capita earnings probably should vary, and the divergence of the last de cade probably signifies a short-run ad justment to a new long-run equilibrium. SAVING AND DEMOGRAPHICS: SOME INTERNATIONAL COMPARISONS Stephen A. Meyer During the decade of the 1980s, Ameri cans saved less than their counterparts in Japan and Germany. During the next 20 years, however, the gap most likely will narrow. Why? Demographic factors, such as the age composition of the popu lation and the proportion of the popula tion that is working or retired, explain some of the differences and provide a basis for projected changes in saving be havior in the U.S. and those in other coun tries. FEDERAL RESERVE BANK OF PHILADELPHIA Are Regional Per Capita Earnings Diverging? Gerald A. Carlino* In the last 10 years, per capita earnings have tended to diverge across U.S. regions after decades of gradual convergence. In 1990, for example, earnings in the states of the New England and Mideast regions were well above the U.S. average. In 1978, however, earnings in both regions had been close to the U.S. average. (See Definitions of Regions on page 12.) Mean while, the Plains' and Rocky Mountain's 1990 per capita earnings, which were slightly below the rest of the country in 1978, had fallen even farther behind the U.S. average. *Gerald A. Carlino is a Senior Economist and Research Adviser in the Regional and Urban Section of the Philadel phia Fed's Research Department. He is also an Adjunct Associate Professor at the University of Pennsylvania, Wharton Real Estate Center. Differences in region-specific factors, such as labor force participation ratios, industry mix, and amenities, result in differentials in real per capita earnings across regions in the long run. It may be that the widening gap in regional earnings after 1978 was caused by changes in these factors. If so, the gap would reflect a permanent adjustment toward a new long-run equilibrium. Alternatively, the widening gap may reflect the effects of powerful, but tempo rary, shocks to the national economy—energy and agricultural shocks, for example— that af fect regions differently. Does the widening gap in regional earnings indicate a fundamental change in regional economies, or is it just a temporary reversal of the trend toward equality that results from powerful, but transitory, economic shocks? 3 BUSINESS REVIEW MARCH/APRIL 1992 THE CONVERGENCE HYPOTHESIS On the eve of the Great Depression, per capita earnings differed widely across regions. In the New England, Mideast, Great Lakes, and Far West regions, per capita earnings were well above the national average. (See Relative Re gional Per Capita Earnings.) In the remaining areas, however, they ranged from 43 percent below average (as in the Southeast) to 8 percent below (as in the Rocky Mountain region).1 Economists have identified several reasons for the vast inequality of regional per capita earnings in the years before World War II. One reason is the relatively low level of agricultural prices, which depressed earnings in regions where agriculture was relatively important.2 In addition, national immigration policies virtu ally halted the influx of cheap labor after World War I, removing the constraints on wage in creases in industrial regions that had been employing most of that labor—mainly the Mid east and Great Lakes. ’ Earnings are by place of work. Wages and salaries, including tips, commissions, and in-kind receipts, account for the bulk of earnings. The other sources of earnings are proprietors' income and "other labor income," which con sists primarily of employers' contributions to private pen sions and group insurance plans. 2Edgar Hoover and Frank Giarrantani, An Introduction to Regional Economics (New York: Alfred A. Knopf, 1984), p. 335. Relative Regional Per Capita Earnings Percent of U.S. Average Percent 140 r 130 New England Mideast Great Lakes Plains Southeast Southwest Rocky Mountain Far West 60 DO Os CN ON t-H CO co ON H Digitized for 4 FRASER «o r- co O n co IT ) ON OS Os ON ON T-H rH T-H t-H T-H t -H t -H to VO ON vO On On NO On CO ON rrON 00 ON oo n ON On 00 On t-H t-H t-H t -H t -H T-H T-H t -H t -H r- IT ) O FEDERAL RESERVE BANK OF PHILADELPHIA Are Regional Per Capita Earnings Diverging? The period between 1929 and the late 1970s saw a pronounced trend toward equalization, or convergence, of regional per capita earnings. All of the low-income regions made substantial gains, while the high-income regions lost ground. A main source of the convergence during the period was the shift of labor from low-wage agricultural employment to higher paying nonagricultural jobs.3 This shift of em ployment largely occurred within regions. Reinforcing the intraregional shift in the work force was the increased ability of workers to move freely from region to region in search of the highest return. Continuing improve ments in communications and transportation technologies since 1929 have given labor and capital more mobility, narrowing the differ ences in regional per capita earnings. By the late 1970s it appeared as if regional per capita earnings might equalize. Should We Expect Per Capita Earnings to Be Equal Across Regions? In its crudest form, the issue of regional per capita earnings differ entials has been addressed in terms of nominal earnings, or earnings that have not been ad justed for differences in regional living costs. From the viewpoint of households, the possible advantages of working in a region with high nominal earnings depend partly on how expen sive it is to live there. Other things being equal, households should be indifferent between a region whose earnings and prices are at the national average and one whose living costs and earnings are, say, 10 percent above the average. Thus, households will choose a region on the basis of real earnings differentials—that is, earnings that have been adjusted for differ ences in living costs. Since workers can move freely from region to region, why should differentials in regional earnings persist once we have adjusted for the 3George H. Borts and Jerome L. Stein, Economic Growth in a Free Market (Columbia University Press, 1964). Gerald A. Carlino cost of living? For one thing, differences in the educational or skill level of the work force or in occupation or industry mix might lead to earn ings differentials, even after cost-of-living ad justments. For example, real earnings in a region may be higher if the dominant industries there are among those offering their workers higher real wages.4 To the extent that differ ences in industry mix influence real earnings, these differences in earnings will persist across regions. Studies that have controlled for the many factors that could affect real earnings find that real earnings are relatively higher in areas with a larger proportion of jobs in the mining, transportation, manufacturing, and government sectors.5 Another important factor behind the earn ings differentials across regions is the differ ences in amenities offered by the regions. Workers may trade off real earnings for ameni ties, accepting lower earnings in high-amenity regions and demanding higher earnings in lowamenity regions. For example, workers seem to care about environmental characteristics, such as the number of sunny days per year, the average annual rainfall, and nearness to large bodies of water. Since these sorts of environ mental amenities differ across regions, regional earnings could differ in equilibrium. Econo mists have found that part of the difference in earnings is due to just such a trade-off.6 Even if real earnings per worker were to equalize across regions, earnings per person 4See Gerald A. Carlino, "Do Regional Wages Differ?" this Business Review (July/ August 1986). 5See Richard P. Voith, "Compensating Variation in Wages and Rents," Journal o f Regional Science 31 (1991), pp. 127-45. Jennifer Roback, "Wages, Rents, and the Quality of Life," Journal o f Political Economy 90 (1982), pp. 1257-78. Studies have found that regional amenity differences tend to be reflected in both land values and wages. 5 BUSINESS REVIEW need not equalize. Real per capita earnings depend not only on how much a region's work ers earn, but on the number of workers relative to the region's total population. The propor tion of a region's population that is employed depends on 1) the fraction of its population that is old enough to work; 2) the percentage of its working-age population that chooses to work (the labor force participation rate); and 3) the proportion of its population that chooses to work and actually finds work. A study by the Federal Reserve Bank of Boston found that differences in the proportion of a region's popu lation that is employed were an important source of differences in regional per capita earnings.7 Moreover, some of these demo graphic factors, such as the relative age struc ture of a region, change slowly over time. If the factors that affect a region's real per capita earnings remain constant, an equilib rium differential exists between the region's real per capita earnings and the national aver age, and the region's relative per capita earn ings should approach that differential through time. (See Equilibrium Differentials in Regional Per Capita Earnings.) As long as the gaps in regional real per capita earnings reflect only such differences as participation ratios, indus try mix, and amenities, workers will not have an incentive to migrate from regions with rela tively low per capita earnings. If so, relative regional per capita earnings will have con verged and the remaining gaps in per capita earnings would reflect equilibrium differen tials.8 * 7Lynn E. Browne, "Shifting Regional Fortunes: The Wheel Turns," Federal Reserve Bank of Boston New England Eco nomic Review (M ay/June 1989). 8Although real per capita earnings may differ across regions, in equilibrium, wages and rents will adjust to ensure equalization of worker utility across regions. 6 MARCH/APRIL1992 Equilibrium Differentials in Regional Per Capita Earnings Differentials in regional per capita earn ings have a tendency to converge to an equi librium. To simplify, assume two regions: a frostbelt region and a sunbelt region. Sup pose that, in equilibrium, per capita earnings are $10,000 annually in both regions. Initially, suppose that per capita earnings in the frostbelt are well above $10,000, while per capita earn ings in the sunbelt are well below $10,000. This inequality in regional per capita earnings causes workers to migrate from the sunbelt to the frostbelt in search of higher per capita earnings. Over time, the migration of workers causes an expansion in the frostbelt's labor force and a decline in the sunbelt's labor force. The increased supply of labor causes per capita earnings in the frostbelt to fall. Similarly, the reduced labor force in the sunbelt causes its per capita earnings to rise. If the two regions were the same in all respects, migration from the sunbelt to the frostbelt would continue until per capita earnings are equal to $10,000 in both regions. But other things are not always equal across regions. For example, suppose workers place a $1000 value on the environmental character istics that the sunbelt offers. That is, workers will accept relatively lower per capita earn ings to live in this region. Similarly, workers must be compensated to live in the frostbelt. In equilibrium, per capita earnings are still $10,000 annually for the nation. But now, the equilibrium per capita earnings are $10,500 in the frostbelt and $9500 in the sunbelt. If the actual differential exceeds $1000, workers would continue to migrate from the sunbelt to the frostbelt until the actual differential in per capita earnings between these two regions is $1000. Once the difference has been reduced to $1000, there is no tendency for the differen tial to decline any further. This gap in regional per capita earnings reflects the equilibrium differential to which these two regions con verge over time. FEDERAL RESERVE BANK OF PHILADELPHIA Are Regional Per Capita Earnings Diverging? The narrowing differential in regional per capita earnings between 1929 and the late 1970s suggests that perhaps we were approaching such an equilibrium, but the widening of the differentials since then has raised doubts. Re cently, several studies have looked at the sources of the gaps in regional per capita earnings and at whether the recent widening is temporary.9 WHAT IS THE EVIDENCE? Studies on regional convergence have some times considered per capita income and some times per capita earnings, but both variables have exhibited the same pattern of conver gence and divergence since the late 1920s.1 In 0 the Boston Fed study, Lynn Browne reported that the main source of divergence in regional per capita income after 1978 was changes in earnings per capita, especially earnings in the more locally oriented industries. Changes in other forms of income, such as dividends, inter est, and rents, reinforced the basic earnings pattern. In general, shifts in industry mix played a minor role in the changes in relative earnings that occurred after 1978. Randall Eberts, in a Cleveland Fed study, built on Browne's work by looking at earnings of individual workers rather than at earnings that have been averaged across all of a region's 9The equilibrium differential we are discussing is taken to be characterized by a constant gap. In an endogenous growth framework, however, the gap in the equilibrium differential could be increasing through time. See Robert Lucas, Jr., "On the Mechanics of Economic Development," Journal of Monetary Economics 22 (July 1988), pp. 3-42. 10See Thomas D. Rowley, John M. Redman, and John Angle, "The Rapid Rise in State Per Capita Income Inequal ity in the 1980s: Sources and Prospects," U.S. Department of Agriculture, Economic Research Service (January 1991). Income includes interest, rent, and transfer payments, as well as earnings. Earnings, however, accounted for the largest portion of personal income. In 1987, for example, earnings accounted for more than two-thirds of personal income. Gerald A. Carlino residents or workers.1 He reported that a 1 narrowing of the regional differentials in earn ings of workers similar in terms of occupation, skill, and education accounted for much of the convergence during the 1970s. Similarly, a widening of the earnings differentials for simi lar workers accounted for much of the diver gence of regional earnings in the 1980s. Eberts speculated that tem porary shocks to the economy from the 1980 and 1981-82 recessions and from the fall in oil prices during the 1980s are probably responsible for this interruption in the long-term trend toward more equal earn ings across regions. Cross-Sectional and Time-Series Evidence. A growing body of research examines per capita earnings (income) convergence at both the na tional and international levels. These studies have used two approaches—cross sectional and time series. The cross-sectional approach looks at the average rate of convergence across regions, given initial differences. The timeseries approach looks at the long-term effects of economic shocks on a region's per capita earn ings (income) relative to the national average. In 1929, regional per capita earnings showed more inequality than could be readily explained by equilibrium differentials alone. For crosssectional convergence, regions having a rela tively low (high) level of earnings per capita should grow relatively quickly (slowly) through time. A study by the National Bureau of Eco nomic Research (NBER) looked for evidence of cross-sectional convergence in real per capita income levels by state for the 1930-88 period.1 2 11Randall W. Eberts, "Accounting for the Recent Diver gence in Regional Wage Differentials," Federal Reserve Bank of Cleveland Economic Review (Third Quarter 1989), pp. 14-26. 12Robert Barro and Xavier Sala i Martin, "Economic Growth and Convergence Across the United States," Work ing Paper 3419, National Bureau of Economic Research (August 1990). 7 MARCH/APRIL 1992 BUSINESS REVIEW The study examined the correlation between a state's growth rate of real per capita income over this roughly 60-year period and its level of real per capita income in 1930. Convergence implies a negative correlation between a state's real per capita income in 1930 and its growth rate during the 1930-88 period. The NBER study finds evidence of cross-sectional conver gence in state-level data.1 3 Using an approach similar to the one em ployed in the NBER study, we find evidence of cross-sectional convergence of regional per capita real earnings during the 1929-90 period. Specifically, for every $1000 that a region's real per capita earnings was above (below) the na tional average in 1929, the region's real growth rate of per capita earnings was lowered (raised) by 0.35 percentage point per year during the 1929-90 period.1 4 Although the cross-sectional approach is one way to examine per capita earnings conver gence, this technique may not provide conclu sive evidence on convergence. If shocks to the nation's economy have differential and longlasting effects across regions, these shocks may widen the dispersion in regional per capita 13Robert Barro has also examined convergence for a cross section consisting of 98 countries during the 1960-85 time period. He finds evidence of convergence when the cross-country regression of per capita output growth rates on initial per capita levels includes proxies for human capital development. See Robert Barro, "Economic Growth in a Cross Section of Countries," Quarterly Journal o f Econom ics 151 (May 1991), pp. 407-43. 14The result from the cross-sectional estimation is: g .= °r,i 3.0958 - 0.35R.n lO (14.3) (-5.27) R2 = 0.8222 where gr . represents compound average annual real growth rate in region i from 1929 to 1990 in real per capita earnings, and R.q represents the level of real per capita earnings in region i in 1929 in thousands of dollars. The numbers in parentheses are t-statistics. 8 earnings. Until recently, economists have viewed the effects of shocks to the economy as temporary, lasting one year or less. However, recent research reveals that economic shocks to the national economy tend to have highly per sistent and long-lasting effects. One recent study found that anywhere from 50 percent to 113 percent of an initial shock to real per capita GNP persists after four years. Even after 100 years, between 19 and 41 percent of a shock still persists.1 5 A shock is an event that causes the level of aggregate per capita income and earnings to deviate from trend. From 1930 through the early 1970s, changes in per capita income and earnings were generally caused by aggregate demand shocks, such as changes in fiscal policy or investment demand. But the shocks of the 1970s occurred largely on the supply side of the economy. One example of a supply shock is the oil price shock that doubled the real price of oil between 1979 and 1980 and reduced per capita income and earnings. Weather-related crop failures are another example of supply shocks that have had adverse effects on the national economy. Of course, some shocks—such as the oil shock in 1986 that lowered the relative price of oil—can have beneficial effects on per capita income and earnings. Although the 1986 energy shock had benefi cial effects on the national economy, it had adverse effects on some regions. Energy shocks influence per capita earnings differently for energy-producing regions than for energy-con suming regions. Per capita earnings would decline in the energy-producing regions and increase in the energy-consuming regions as a result of a 1986-type drop in energy prices. 15Francis Diebold and Glenn Rudebusch, "Long Memory and Persistence in Aggregate Output," Journal o f Monetary Economics 24 (1989), pp. 189-209. The upper bound (113 percent) of their estimate indicates that the initial effects of a shock may even be magnified. FEDERAL RESERVE BANK OF PHILADELPHIA Are Regional Per Capita Earnings Diverging? A study by the Federal Reserve Bank of St. Louis showed that per capita income in the energy-producing states fell from 95.4 percent of the U.S. average in 1978 to 86.8 percent by 1987.1 Per capita income in the farming states 6 of the West North Central region was also severely affected by the 1981-82 recession and the farm crisis during the first half of the 1980s.1 7 According to the St. Louis Fed study, average per capita income in the West North Central region declined from 97 percent of the U.S. average in 1978 to 93 percent in 1987. Other types of shocks that could have differential regional effects include increases in national defense-related expenditures and the intro duction of new technologies that favor some regions, such as the development of the hightech industries that favored New England dur ing the 1980s. Are the Effects of Economic Shocks on Relative Regional Earnings Temporary? Drawing on the evidence, we can examine the extent to which shocks to relative regional per capita earnings also have persistent effects. As we saw, per capita earnings tended to converge across regions until the late 1970s. We saw also that the introduction of new shocks, such as the oil shocks of the past two decades, affected regional per capita earnings differently. If the differential effects of these shocks are highly persistent, they may, for practical purposes, lead to a permanent widening in the dispersion in regional per capita earnings. Given the evidence that national shocks have differential effects across regions, the effects of these shocks on a region's per capita earnings must be tem 16See Cletus C. Coughlin and Thomas Mandelbaum, "Why Have State Per Capita Incomes Diverged Recently?" Federal Reserve Bank of St. Louis Review (September/Octo ber 1988). 17The West North Central region consists of Iowa, Kan sas, Minnesota, Missouri, North Dakota, Nebraska, and South Dakota. Gerald A. Carlino porary in order for regional per capita earnings to converge over time. Extending a Philadelphia Fed study by Gerald Carlino and Leonard Mills, we find that, over the entire 1929-90 period, 72 percent of a shock in the Mideast region persisted five years out and 38 percent persisted 10 years out.1 (See 8 How Persistent Are the Effects of Shocks?) Stated differently, if some event raises per capita earn ings in the Mideast region an additional $1 above the national average, five years later per capita earnings would be 72 cents above the national average because of that event. Ten years later, per capita earnings in the Mideast region would still be 38 cents above the national average. The persistence of shocks is less pronounced in the other regions. For example, each dollar shock leads to a 28-cent deviation in New England's relative per capita earnings five years out and to an 8-cent deviation 10 years out. For every region, at least 15 cents of a $1 shock remains five years out. The regional persistence of economic shocks is not what theory predicts. Workers can move freely from region to region in search of em ployment, which would tend to dampen the differential regional effects of any shock through 18Gerald Carlino and Leonard Mills, "Have Regional Per-Capita Incomes Converged?" Working Paper 91-18, Federal Reserve Bank of Philadelphia (1991). This article extends the analysis in the working paper by including data for three more years— 1988,1989, and 1990. We examine regional per capita earnings relative to national average per capita earnings to control for the common effect of national economic shocks across regions. A number of recent studies test for stochastic convergence (that is, the persistence of shocks to relative output) across countries. They generally find that shocks have highly persistent effects, a result inconsistent with cross-sectional convergence. See Andrew Bernard and Steven Durlauf, "Convergence of International Output Movements," Working Paper 3717, National Bu reau of Economic Research (May 1991), and Danny Quah, "International Patterns of Growth: I. Persistence in CrossCountry Disparities" (January 1990), mimeo. 9 MARCH/ APRIL 1992 BUSINESS REVIEW time. Are there other factors, then, that help explain the pattern of relative regional per capita earnings over the past half century? R esearchers have recently questioned whether the high degree of persistence found in time series of various measures of income or economic activity may be accounted for by a major event, such as an oil shock, that repre sents an unusually large departure from their previous trends.1 Could such a single major 9 disturbance account for the persistence that we find in shocks to relative regional incomes in the 1929-90 period? A number of studies have found that the dispersion in regional per capita incomes has increased since 1978.20 After con trolling for a single break in the convergence trend in 1978, we find that persistence is re duced for all regions. The effects of shocks are found to be most persistent in the Southwest region. For every dollar shock to the Southwest's per capita earnings, 33 cents remains after five years and 10 cents after 10 years. Among the other regions, no more than 6 percent of a shock remains 10 years out. In four regions—Great Lakes, P lain s, Sou th east, and Rocky Mountain—the effects of shocks are essentially gone within 10 years. In an earlier study on per capita income, Richard Easterlin reported that a break in the convergence trend occurred in 1946.2 Indeed, 1 inspection of the data on relative regional per 19Pierre Perron, "The Great Crash, the Oil Price Shock, capita earnings reveals that such a break may and the Unit Root Hypothesis," Econometrica 57 (November have occurred after World War II. After con1989), pp. 1361-401. trolling for such a break, we find a sub stantial reduction in the persistence of How Persistent Are the Effects of Shocks? shocks occurring (Estimated proportion of a shock to relative regional earnings during the 1929-90 that remains 5 and 10 years out) period. The results 1929- 1990 1929- 1990 1929 -1990 (Break in 1978)a (Break in 1946)b 5 vrs 10 vrs 5yrs 10 vrs 5yrs 10 vrs New England Mideast Great Lakes Plains Southeast Southwest Rocky Mtn. Far West 0.28 0.72 0.25 0.15 0.25 0.58 0.20 0.33 0.08 0.38 0.06 0.04 0.04 0.29 0.04 0.13 0.25 0.25 0.11 0.08 0.02 0.33 0.10 0.17 0.06 0.03 0.01 0.01 0.00 0.10 0.01 0.03 0.23 0.31 0.10 0.02 0.07 0.33 0.11 0.13 0.05 0.05 0.01 0.00 0.00 0.10 0.01 0.02 aControIs for a break in trend in 1978. bControls for a break in trend in 1946. Source: Extension of Gerald Carlino and Leonard Mills, "Have Regional Per-Capita Incomes Converged ?" W or king Paper 91-18, Federal Reserve Bank of Philadelphia (1991). 10 20See Coughlin and Mandelbaum (September/Octoberl988);Barro and Sala i Martin (August 1990); and Rowley, R ed m an, and A ngle (January 1991). 21Richard Easterlin, "In terregio nal D iffer ences in Per Capita In come, Population, and Total Incom e, 18401950," in Conference on Research in Income and W ealth, Trends in the American Economy in the N ineteenth Century 24 (1960), pp. 73-140. FEDERAL RESERVE BANK OF PHILADELPHIA Are Regional Per Capita Earnings Diverging? Gerald A. Carlino are similar to what we found when we incorpo gion was almost 1 percent per year below rated a break in 1978. Thus, after controlling for national average growth in the 1929-45 period, breaks in first 1978 and then 1946, we find that but equal to the national average on an average shocks during the 1929-90 period did not have annual basis during the 1946-90 period. highly persistent effects on relative regional per capita earnings. This finding is important, CONCLUSION Regional per capita earnings, which varied since even if the shocks that occurred over the past 60 years had increased the dispersion in substantially in 1929, continue to differ today. regional per capita earnings, the effect in most cases was not highly persistent. The time-series analysis confirms the usual convergence view that the initially poor re 22The Great Lakes region is an exception, since it con gions tended to catch up to rich ones over time. verged at a faster rate during the postwar period. The rate During the 1929-45 period, regions that had per of convergence in the New England region reverses from -1.7 percent per year in the 1929-45 period to slightly posi capita earnings above the national average in tive in the 1946-90 period This reversal in New England's 1929 grew less rapidly than regions with per rate of convergence is an anomaly, since per capita earnings capita earnings below the U.S. average. (See in the region were still above the national average in 1946. The Trend Rate of Convergence Slows After 1946.) Similarly, the rate of convergence in the Rocky Mountain For example, per capita earnings in the New region changes from 0.8 percent per year in the 1929-45 period to slightly negative in the 1946-90 period, even England region were above the national aver though per capita earnings were still below the national age in 1929, but the region had an annual average in 1946. In the New England and Rocky Mountain growth rate 1.7 percent per year below the regions, however, the rate of convergence in the 1946-90 national average growth during the 1929-45 period is not significantly different from zero. period. Similarly, per capita earn ings in the South The Trend Rate of Convergence Slows After 1946 west region were below the n a Per Capita Per Capita Rate of Rate of tional average in Convergence3 Earnings Convergence3 Earnings 1946-90 Relative to 1929-45 Relative to 1929, but the re U.S. Average U.S. Average gion had an an (percent) (percent) in 1929 in 1946 nual growth rate 2.2 percent above national average 0.2 ABOVE -1.7 New England ABOVE grow th during ABOVE Mideast ABOVE -0.9 0.0 the same period. ABOVE - 0.1 -0.4 ABOVE Great Lakes D uring the BELOW 0.9 BELOW Plains 0.0 postwar period, a 1.4 BELOW Southeast 0.6 BELOW slowdown in the Southwest 0.2 BELOW 2.2 BELOW rate of conver BELOW Rocky Mtn. BELOW 0.8 - 0.1 ABOVE - 0.2 ABOVE Far West - 0.2 gence is indicated for m ost re gions.22 For ex Estim ated average annual rate at which the ratio of regional per capita earnings relative to ample, growth in national per capita earnings changed. the Mideast re 11 BUSINESS REVIEW Should we expect per capita earnings to vary across regions? It appears so. Regional differ ences in labor force participation ratios, indus try mix, and amenities result in differentials in real per capita earnings across regions. In fact, the gap that had been narrowing through the late 1970s has widened since. Does the widening gap in regional per capita earnings after 1978 represent a reversal of the long-run trend toward convergence? Appar ently not. After allowing for the possibility that the equilibrium gap widened in the late 1970s, we find that what appears to be a divergence of regional per capita earnings may actually rep resent a short-run adjustment to a new longrun equilibrium. Once this adjustment has occurred, the gap, although wider, should re main stable provided that there are no further changes in the underlying equilibrium. More over, economic shocks that occurred during the entire 1929-90 period have not generally had highly persistent effects on a region's rela tive per capita earnings. Digitized for12 FRASER MARCH/APRIL 1992 Definitions of Regions New England Connecticut Maine Massachusetts New Hampshire Rhode Island Vermont Mideast Delaware District of Columbia Maryland New Jersey New York Pennsylvania Southeast Alabama Arkansas Florida Georgia Kentucky Louisiana Mississippi North Carolina South Carolina Tennessee Virginia West Virginia Southwest Arizona New Mexico Oklahoma Texas Great Lakes Illinois Indiana Michigan Ohio Wisconsin Far West California Nevada Oregon Washington Rocky Mountain Colorado Idaho Montana Utah Wyoming Plains Iowa Kansas Minnesota Missouri Nebraska North Dakota South Dakota FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons Stephen A. Meyer* The United States saved only a small share of its aggregate income during the 1980s—a much lower share than most other large, industrial countries saved. As shares of GNP, both house hold and government saving were smaller in the U.S. The differences in personal saving owe much to demographic factors—especially to differences in the age composition of countries' populations. The differences in government "■ Stephen A. Meyer is Vice President and Associate Di rector of Research at the Federal Reserve Bank of Philadel phia. He also is an Adjunct Professor of Finance at the Wharton School of the University of Pennsylvania. saving around the world reflect budget bal ances: surpluses in high-saving countries and deficits in low-saving countries. Differences in saving are important because saving is the source of funds required to finance investment in plant and equipment, structures, and housing. Persistently higher saving shares in both West Germany and Japan, for example, financed greater net investment relative to GNP than in the U.S. Over time, investment has increased the stock of productive capital and contributed to growing labor productivity and a rising standard of living in all three countries. But West Germany and Japan experienced a more rapid increase in labor productivity over 13 BUSINESS REVIEW MARCH/APRIL 1992 the past 40 years because they TABLE 1 saved a larger share of their ag Saving and Investment gregate incomes.1 Some analysts argue that the as a Share of GNP share of personal income saved in (Average for 1980s, in percent) the United States will rise strongly during the next 20 years, as more West and more of the baby-boom gen Japan U.S. Germany eration enters middle age.2 These Net National Saving 10.2 analysts anticipate that saving 3.0 18.1 rates in the U.S. will rise to levels Household 3.8 7.8 11.0 more like those in Germany and Business 1.7 1.0 2.8 Japan. Indeed, recent economic Government -2.5 1.3 4.3 research does suggest that pro jected demographic changes are Net Fixed Investment 5.1 8.0 15.4 likely to narrow the gap in saving rates between the U.S. and Ger many or Japan, but this result derives as much from falling saving rates abroad ing in West Germany, when measured as a as from a rising saving rate in the U.S. share of GNP, was more than three times as large as in the U.S. And in Japan the share was COMPARING U.S. SAVING IN THE 1980s six times that in the U.S.4 Among the compo TO GERMANY AND JAPAN nents of national saving, household saving was The net national saving rate in the United more than twice as big a share of GNP in West States was lower in the 1980s than in many Germany as in the U.S., while in Japan it was other industrial countries, especially West Ger many and Japan (Table l).3 The lower saving rate in the U.S. was accompanied by lower net investment relative to GNP. Net national sav- terly National Accounts, published regularly by the Organi 1From 1950 through 1989, labor productivity grew at an average rate slightly below 6 percent per year in Japan, at an average annual rate somewhat above 4 percent in Germany, and at a rate averaging just 1.75 percent per year in the U.S. For a technical discussion of the relationship between in vestment and productivity growth, and a careful examina tion of the data, see Wolff (1991). Productivity growth rates cited here are from Wolff (1991), updated with data for the 1980s from the OECD Economic Outlook, July 1991. 2See, for example, "Upbeat Generation," Barron's, Au gust 1,1988, pp. 15 and 30. 3The saving, investment, and aggregate income data cited in this article are those available as of November 1991. The data are taken from National Accounts and from Quar- Digitized for 14 FRASER zation for Economic Cooperation and Development (OECD) in Paris. These publications draw on each country's official national income and product accounts. Figures for net investment cited here include net fixed investment spend ing by governments—spending on capital equipment, build ings, and infrastructure such as roads, bridges, dams, water systems, and airports. Government investment is included in the OECD data for Germany and Japan. Investment spending by governments in the U.S. was estimated using other sources. 4Part of a country's saving is used to finance the replace ment of buildings and capital equipment that wear out or are abandoned each year. To the extent that saving and investment spending exceed replacement investment, they can contribute to growing labor productivity and a rising standard of living. The definition of saving that corre sponds to this concept of funds available to finance growth of the capital stock, or net investment, is net national saving. FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons nearly three times as big. The government sector in the U.S. ran budget deficits through out the 1980s, thus reducing national saving. In contrast, governments in West Germany and Japan added to national saving by running budget surpluses in most years. Net business saving (mostly retained earnings) differed much less among these countries.5 * Not only was net national saving a smaller share of GNP in the U.S. than in West Germany and Japan, but the gaps widened during the 1980s. Household saving declined relative to GNP in all three countries as the 1980s pro gressed. In Germany and Japan, government sector budget surpluses rose relative to GNP, offsetting the drop in personal saving (and in Germany business saving also rose strongly). But in the U.S., the government sector contin ued to run budget deficits, thus reinforcing the decline in personal saving relative to GNP. Differences in government saving largely reflect political decisions about government spending and taxes. Differences in household saving rates between the U.S., Germany, and Japan during the 1980s reflect demographic factors, in part. The way in which demographic factors affect household saving can best be understood by looking at some basic economic theory of saving behavior. THE FORWARD-LOOKING THEORY OF SAVING Economic theory has focused on five major reasons why people save: (1) to provide for 5Fumio Hayashi of the University of Pennsylvania ar gues that Japanese national income accounts understate replacement investment, and thus overstate net national saving and investment, so that the gap between Japanese and American saving and investment shares is smaller than indicated here. But Robert Dekle of Boston University and Lawrence Summers of the World Bank present other data suggesting that the gap really is as large as shown by the official statistics. See Hayashi (1989) and Dekle and Sum mers (1991). Stephen A. Meyer their retirement; (2) to leave a bequest; (3) to bridge temporary declines in their incomes; (4) to finance unanticipated expenditures such as medical bills; and (5) to finance purchases of durable goods such as furniture and automo biles. The theory of saving behavior is complex, and economists' understanding of saving be havior is still evolving. We can avoid many of the complications, but nonetheless gain a good deal of insight into the effect of demographic factors on saving behavior, by focusing on a somewhat simplified description of the basic forward-looking theory of family saving. Theory. The basic theory of saving begins by recognizing that real earnings usually are rela tively low early in people's careers, peak shortly before retirement, and then fall substantially after retirement. The theory proceeds with the idea that people prefer to spread their con sumption of goods and services evenly over their lives to the extent they can do so and, in particular, that people would rather not have their consumption fall sharply when they re tire. From these premises, the theory predicts that younger families actually will spend more than their incomes if they have ready access to credit or an inheritance, so their saving will be negative on average. If they do not have ready access to credit or an inheritance, younger families will save at most a small share of their incomes. Middle-aged families typically will save a larger share of their rising earnings as they prepare for retirement and accumulate an estate. Families headed by retired people typi cally will save little, if any, of their incomes, and in many cases they will dissave. This predicted pattern of changing personal saving rates over one's lifetime is known as life-cycle saving. Evidence. Surveys of consumer spending and finances in the U.S. yield results that are broadly consistent with the life-cycle pattern of earnings and saving discussed above. Average earnings do rise with age, and there is a broad peak in average earnings between ages 50 and 15 BUSINESS REVIEW 60. Very young families, those headed by people less than 25 years old, do tend to spend more than they earn (their savings are negative, on average), indicating that they are incurring debts or spending gifts and inheritances. Sur veys show that average earnings substantially exceed average spending for families in the peak-earning years. And many families do draw down their savings during retirement; families headed by people over 64 save very little of their incomes, on average. During the mid-1980s, the share of income saved by house holds headed by people between the ages of 45 and 64 averaged 6 to 8 percentage points higher than the share saved by households headed by people over the age of 64.6 Surveys of saving behavior in Canada and Japan yield similar results.7 All of these observations are broadly consistent with the predicted life-cycle pattern of saving derived from a simple forward-look ing theory. The simplest version of the life-cycle saving model, described above, does not explain all of what we observe about family saving. Many families in the peak-earning years save an ap preciably larger share of their incomes than do families whose heads are between 25 and 44, but some families save a small and relatively constant share of their incomes throughout their working years. Many older families draw down their savings, but others neither save nor dissave, and those with substantial wealth typi cally continue to save. These observations are 6Data cited here are based on Tables 3, 13, and 23 in "Consumer Expenditure Survey: Integrated Survey Data, 1984-86", BLS Bulletin 2333 (August 1989), published by the Bureau of Labor Statistics of the U.S. Department of Labor. The Survey of Consumer Finances, conducted by the Fed eral Reserve System, yields broadly similar conclusions; see Kennickell (1990). 7Saving behavior in Canada and Japan, as well as in the U.S., is examined in Bosworth, Burtless, and Sabelhaus (1991). 16 MARCH/APRIL 1992 broadly consistent with the life-cycle saving pattern predicted by more complex versions of the forward-looking theory, which incorporate precautionary saving, social security, borrow ing constraints, uncertainty about lifespans, and saving to accumulate an estate.8 Even in these more complicated models, age affects saving behavior. Demographic factors are not the only deter minants of household saving. People's saving also is affected by the tax treatment of saving and interest, by the structure of social security and pension systems, by the variability of in comes, by the extent to which people can insure against unanticipated expenditures or income reductions, by unanticipated changes in wealth, and by the strength or weakness of the economy, among other factors.9 There also is evidence 0 1 that household saving is affected by the size of government budget surpluses or deficits. De mographic factors do have important influ ences on household saving, however. THE EFFECT OF DEMOGRAPHICS ON HOUSEHOLD SAVING While there is vigorous debate among econo mists about how much household saving is generated by each of the five major reasons for saving, there is broad agreement that the age composition of a country's population can af- 8For a brief review of economists' knowledge of saving behavior, see Weil (1991). For a more detailed and extensive treatment of the roles of saving for retirement and for bequests, see Kotlikoff (1989). For a discussion of the role of precautionary saving and evidence of its importance, see Carroll (1991). For a focus on borrowing constraints, see Zeldes (1989). 9See Boskin and Lau (1988) for a careful discussion of the roles of these and other factors. 10A lively exchange between Franco Modigliani of MIT and Laurence Kotlikoff of Boston University summarizes much of the agreement and disagreement. See Modigliani (1988) and Kotlikoff (1988). FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons feet the share of income that is saved.1 The 0 "life-cycle saving" theory suggests that a high share of household income will be saved in a country that has a large fraction of its popula tion in the high-saving years from 45 to 64 and a small fraction in the low-saving or dissaving years up to 20 and beyond 64. Household saving would be a relatively large share of GNP as a result. Conversely, the theory predicts that household saving would be a smaller share of GNP in a country that has a small fraction of its population in the peak-saving years from 45 to 65 and a large fraction of its population in the low-saving or dissaving age groups. Empirical research largely bears out these expectations. The strongest demographic ef fect appears to result from an increase in the share of the population that is beyond retire ment age, accompanied by a decrease in the working-age share of the population. Research ers estimate that a 1-percentage-point increase in the ratio of the population over the age of 64 to the working-age population, holding con stant other factors that affect saving, has been associated with a reduction in the ratio of household saving to GNP by an amount in the range from 0.4 to 1.4 percentage points, for an average estimate of 0.9 percentage point. There also is evidence that an increase in the ratio of the under-20 population to the working-age population reduces household saving relative to GNP, but this effect appears smaller— perhaps half as big.*1 1 * 9 COMPARING U.S. DEMOGRAPHICS IN THE 1980s TO GERMANY AND JAPAN The share of the U.S. population in the highsaving age group from age 45 to 64 was appre ciably lower than in Germany and Japan in both 11For a summary of the results of these studies, see Table 9 in Heller (1989). The numbers presented here correspond to those presented by Heller, but here they are presented as shares of GNP rather than as shares of national income. Stephen A. Meyer FIGURE 1 Population Shares (percent) The share of the U.S. population in the highsaving age group fell in the 1980s, but will rise more strongly during the next 20 years than in Germany and Japan. The share of the U.S. population in the low-saving age groups will continue to fall during the next 20 years, but will rise in Germany and Japan. High saving group Q A ge 45-64 C/5 Low saving group ■ Age 0-19 I IAge 65+ Percent Percent 50 1980 1990 2000 2010 1990 2000 2010 Percent 50 1980 17 BUSINESS REVIEW 1980 and 1990 (Figure 1). And the share of the U.S. population in that high-saving age group declined during the 1980s, while it rose strongly in Germany and Japan. In addition, the total share of the U.S. population in the low-saving age groups was larger at both the beginning and the end of the 1980s than was the case in either Germany or Japan.1 2 On average during the 1980s, 19 percent of the U.S. population was in the high-saving group from age 45 to 64— about 4.5 percentage points lower than in Japan and nearly 5.5 per centage points lower than in Germany. The ratio of 65-and-over to working-age popula tions in the U.S. averaged roughly 4 percentage points lower than in Germany during the 1980s, but 4 percentage points higher than in Japan. The ratio of under-20 to working-age popula tions also was higher in the U.S., averaging 12 percentage points higher than in Germany and about 4 percentage points higher than in Japan. In total, an average of nearly 42 percent of the U.S. population fell into the two low-saving age groups during the 1980s, about 3 percentage points higher than in Germany and Japan. On balance, these demographic ratios imply somewhat higher personal saving relative to GNP in Germany than in the U.S. during the 1980s, and much higher household saving rela tive to GNP in Japan, as we in fact observed. Taken together, the demographic differences can account for roughly one-third of the gap in personal saving relative to GNP between Ger many and the U.S. during the 1980s and roughly two-thirds of the gap between Japan and the U.S.1 3 Demographic factors help explain not only why household saving in the U.S. was lower relative to GNP than in Germany and Japan but MARCH/APRIL1992 also why household saving shares declined during the 1980s. As the ratio of the 65-andover to working-age populations grew in the U.S. from 1980 to 1990, personal saving de clined relative to GNP (Table 2). Similarly, the sharp rise in the ratio of 65-and-over to work ing-age populations in Japan from 1980 to 1990 helps explain why personal saving declined relative to GNP in that country. Demographic factors, by themselves, do not entirely explain the decline in household sav ing during the 1980s.1 Personal saving in Ger 4 many did not rise relative to GNP, for example, even though the ratio of 65-and-over to work ing-age populations declined. Other factors that may have reduced household saving shares include expanded coverage by pension, social security, and medical insurance systems, large capital gains in equity and housing markets, and declining birthrates. While demographic factors are not the sole determinant of house hold saving behavior, the fact that saving by households was roughly twice as large relative to GNP in Germany as in the U.S. during the 1980s, and roughly three times as large in Ja pan, does reflect differences in the age compo sition of these countries' populations. PROJECTED DEMOGRAPHIC CHANGES AND FUTURE SAVING RATES Demographers project that the age composi tion of the U.S., German, and Japanese popula tions will change markedly during the next 20 years. Those changes have the potential to raise personal saving relative to GNP in the United States and to reduce it in Germany and Japan. Population Shares Will Change Markedly. Demographers project that the share of the U.S. population in the high-saving years from 45 to 64 will rise by half from 1990 to 2010, to nearly 28 percent from about 18.5 percent, as the baby- 12Population data and projections cited in this article are taken from United Nations (1982). 13Based on the average size of the estimated effects found by the studies cited in Table 9 of Heller (1989). 18 14See Kennickell (1990) and Boskin and Lau (1988) for evidence on this point. FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons Stephen A. Meyer next two decades have been bom already, and Ratio of 65-and-Over Population major industrial coun tries have reasonably to Working-Age Population accurate census data. and Household Saving As a result, demogra phers' projections of 1990 1980 1985 changes in the age United States: composition of those 21.4 19.9 20.3 65-and-over/working-age (%) countries' populations 3.1 3.3 5.0 household saving/GNP (%) over the next 20 years are likely to be fairly Germany: accurate. There is un 24.1 26.8 23.9 65-and-over/working-age (%) certainty about future 8.2 7.2 household saving/GNP (%) 8.3 birthrates and about how much average Japan: 16.9 19.0 65-and-over /working-age (%) 13.0 lifesp an s may in 10.7 9.4* 12.7 household saving/GNP (%) crease. Even so, popu lation projections for the next two decades are unlikely to be far T h is number for Japan is for 1989. Japanese saving data for 1990 were not yet off unless birthrates or published when this article went to press. death rates change dramatically—or un boom generation ages (Figure 1). The working- less migration occurs on a larger scale than age share of the population will grow as well. observed in recent decades. The possibility of The share of the U.S. population over the age of larger-than-usual immigration to Germany is 64 is projected to be roughly constant during quite real, considering ongoing developments the next 20 years, and the share under the age in Eastern Europe. Should large-scale migra tion occur, the working-age population in Ger of 20 is projected to decline. In Germany and Japan, in contrast, the share many might not shrink as projected. For the of the population in the high-saving age group U.S. and Japan, however, larger-than-usual is projected to change much less during the next immigration seems unlikely. 20 years. But demographers project that the If demographic projections prove correct, share of the population over the age of 64 will the ratio of 65-and-over to working-age popu rise by one-third in Germany, to nearly 21 lations will rise by 6.4 percentage points in percent. In Japan, the share of the population 65 Germany and by 9.5 percentage points in Japan or older will rise by two-thirds, to nearly 20 during the next 20 years, but will remain essen percent in 2010. While the share of the popula tially unchanged in the U.S. (Figure 2). During tion under the age of 20 is projected to shrink the same period, the ratio of under-20 to work about as much in each of those countries as in ing-age populations is projected to fall by about the U.S., the working-age share of the popula 7 percentage points in the U.S., but to rise tion also is projected to shrink in Germany and roughly 1 percentage point in Germany and to Japan, in contrast to the U.S. reverse course and begin rising in Japan. Clearly, People who will be older than 20 during the projected changes in the age composition of TABLE 2 19 MARCH/APRIL 1992 BUSINESS REVIEW FIGURE 2 Changes in Ratios of 65-and-over and Under-20 Populations to Working-Age Population The ratio of low-saving, 65-and-over popula tion to higher-saving, working-age popula tion in the U.S. will remain essentially un changed over the next 20 years, but will rise strongly in Germany and Japan. And the ratio of low-saving, under-20 population to higher saving, working-age population will continue to decline in the U.S., but not in Germany and Japan. Q 6 5 + to working-age C 0-19 to working-age Percent _____ £ -5- G -10 G -15 1980-1990 1990-2000 2000-2010 Percent -5 < - 10 - - 15 - 1980-1990 20 1990-2000 2000-2010 these three countries' populations have the potential to raise household saving in the U.S. and to lower it in Germany and Japan. Changes in Population Shares Will Affect Household Saving. There is widespread agree ment on this point among economists who have studied the issue, but there is disagreement on how large those changes are likely to be. Taking the average of earlier-cited estimates of the effects of changes in population ratios on house hold saving behavior, and multiplying that average by the projected changes in population ratios, one might conclude that demographic trends, by themselves, have the potential to raise the ratio of household saving to GNP in the U.S. by 3.5 to 4 percentage points by the year 2010, and to reduce the ratio by 6 to 6.5 percent age points in Germany and by 7 to 7.5 percent age points in Japan. Changes in saving shares are unlikely to be so large, however. Because the U.S. baby-boom generation is so large relative to previous generations, it is dif ficult to predict just how much baby boomers' incomes and saving will rise as they enter their peak-earning years. Their peak incomes may not exceed the incomes they earned earlier in their lives by as much as was the case for earlier generations simply because so many baby boomers are competing for jobs. Also, if labor productivity in the U.S. continues to grow as slowly as it did during the 1980s, baby boomers' incomes will not rise as rapidly over the re mainder of their working lives as was the case for earlier generations. Both of these possibili ties suggest that U.S. baby boomers' saving rates may not rise as much when they enter their peak-earning years as was the case for earlier generations. In Germany and Japan, though, the shrinking working-age populations may cause the incomes of those entering their peak-earning years during the next two dec ades to rise more rapidly than was the case for earlier generations. In addition, if lifespans continue to grow longer, those now at work in all three countries may postpone retirement FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons and may save more as they prepare for a longer retirement. Thus longer lifespans may raise aggregate personal saving. Although it is difficult to forecast just how much household saving behavior will change as a result of projected demographic changes, some recent economic research provides at least a rough idea. Taking into account pro jected changes in the ratio of 65-and-over to working-age populations and in the ratio of under-20 to working-age populations, and also taking into account the changes in real wages and in income distribution that seem likely to result, Paul R. Masson of the International Monetary Fund and Ralph W. Tryon of the Federal Reserve estimate that demographic shifts during the next 20 years have the poten tial to raise the ratio of household saving to GNP in the U.S. by roughly 2 to 2.5 percentage points by the year 2010. In contrast, they estimate that demographic shifts can lower the ratio of household saving to GNP by roughly 1.5 to 2 percentage points in Germany and by roughly 3.5 to 4 points in Japan.1 5 Changes in Population Shares May Affect Government Saving. Changing age distribu tions may affect government budget balances as well as private saving. As the number of people aged 65 and over in Germany and Japan grows rapidly during the next 20 years, their governments' spending on medical care, pen sions, social security systems, retirement hous ing, and other programs for the elderly is likely to grow rapidly too. On the other hand, spend ing on education may decline as the number of children shrinks. One careful study under taken by the OECD estimates that, on balance, projected demographic shifts would increase 15Masson and Tryon (1990) undertake a careful empiri cal investigation of the effects of projected demographic changes in these three countries. For studies focusing on the U.S., see Kennickell (1990) and Auerbach and Kotlikoff (1989). Stephen A. Meyer government spending in Germany by an amount equal to nearly 5.5 percent of GNP by 2010, and in Japan by 9.5 percent of GNP—if there are no changes in government pension or benefit pro grams.1 At the same time, the working-age 6 population in those two countries is projected to decline, making it more difficult for govern ments to raise additional revenues. Thus gov ernment saving in Germany and Japan may well shrink relative to GNP during the next two decades and could become negative, although governments are likely to offset at least part of the effect of demographic changes on their budgets. In the United States, by contrast, the number of people 65 and older is projected to grow slowly during the next 20 years. Over the same period, the working-age share of the popula tion is projected to rise. Thus in the U.S., the demand for government services for the eld erly is likely to grow less rapidly during the next two decades than it did during the previ ous two. The OECD study cited in the preced ing paragraph estimates that demographic shifts could reduce government spending in the U.S. by an amount roughly equal to 1.5 percent of GNP over the next 20 years. Over the same period, changing demographics are projected to generate large and growing surpluses in the U.S. Social Security System. Government dissaving in the United States could well shrink as a result. Over the next two decades, then, projected changes in the age distribution of populations are likely to lower personal saving relative to GNP in Germany and Japan and perhaps re- 16For a discussion of the implications of projected demo graphic changes for social policy and the demand for gov ernment services, see Ageing Populations: The Social Policy Implications, published by the OECD (Paris, 1988). Several other estimates of the likely size of changes in demand for government services are summarized in Heller (1989) and in Table 2 of Masson and Tryon (1990). 21 BUSINESS REVIEW MARCH/APRIL 1992 duce government saving as well. In the United States, by contrast, demographic trends are likely to increase personal saving relative to GNP—and perhaps raise government saving, too. Beyond 2010, however, the share of the U.S. population over 65 is projected to rise and the share in the high-saving, peak-earning years is projected to gradually decline as the babyboom generation begins to enter the retirement years. Thus the projected rise in national sav ing relative to GNP in the U.S. may prove temporary. Demographers project that the ratio of over-65 to working-age populations will continue to rise in Germany and Japan, although more gradually. The projected de cline in national saving relative to GNP in those two countries may prove longer lasting. SUMMARY From 1980 to 1990, both personal saving and net national saving were much smaller relative to aggregate income in the U.S. than in Ger many or Japan. The lower ratio of personal saving to GNP in the U.S. partly reflected the age composition of the three countries' popula tions: an appreciably smaller share of the U.S. population was in the peak-earning, highsaving years from age 45 to 64 than was the case in Germany and Japan; and the share of the U.S. population in the low-saving years before people enter the labor force and after they retire was larger than for the other two countries. In addition, the government sector in the U.S. ran budget deficits throughout the 1980s, thus re ducing net national saving, while the govern ment sectors in Germany and Japan ran budget surpluses. Looking ahead, demographic projections suggest that personal saving will rise relative to GNP in the U.S. during the next 20 years as the share of the U.S. population in the high-saving years rises by half and the share in the lowsaving age groups declines. Demographic pro jections also suggest that personal saving will fall relative to GNP in Germany and Japan during the same period, as the share of their populations that is over the age of 64 rises sharply and the working-age share of their populations shrinks. The projected changes in the age composi tion of the German and Japanese populations seem likely to reduce government saving as well as household saving. But in the U.S., demographic changes are likely to contribute to smaller budget deficits—and possibly to budget surpluses. Overall, projected demographic changes are likely to narrow the gap between high national saving relative to GNP in Germany and Japan and low national saving relative to GNP in the U.S. But demographic trends seem unlikely to raise U.S. saving rates to levels observed in Germany or Japan during the 1980s. The pro jected narrowing of saving gaps during the next 20 years will result as much from lower saving shares in Germany and Japan as from a higher saving share in the U.S. REFERENCES Auerbach, Alan J., and Laurence J. Kotlikoff. "Demographics, Fiscal Policy, and U.S. Saving in the 1980s and Beyond," Working Paper 3150, National Bureau of Economic Research (October 1989). Boskin, Michael J., and Lawrence J. Lau. "An Analysis of Postwar U.S. Consumption and Savings Behavior, Parts I and II," Working Papers 2605 and 2606, National Bureau of Economic Research (June 1988). 22 FEDERAL RESERVE BANK OF PHILADELPHIA Saving and Demographics: Some International Comparisons Stephen A. Meyer Bosworth, Barry, Gary Burtless, and John Sabelhaus. "The Decline in Saving: Evidence from Household Surveys," Brookings Papers on Economic Activity 1 (1991). Carroll, Christopher D. "Buffer Stock Saving and the Permanent Income Hypothesis," Working Paper 114, Division of Research and Statistics—Economic Activity Section, Board of Governors of the Federal Reserve System (February 1991). Dekle, Robert, and Lawrence Summers. "Japan's High Saving Rate Reaffirmed," Working Paper 3690, National Bureau of Economic Analysis (April 1991). Hayashi, Fumio. "Is Japan's Saving Rate High?" Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1989). Heller, Peter S. "Aging, Saving, and Pensions in the Group of Seven Countries: 1980-2025," Working Paper 89/13, International Monetary Fund (January 1989) Kennickell, Arthur B. "Demographics and Household Savings," Finance and Economics Discussion Series Paper 123, Board of Governors of the Federal Reserve System (May 1990). Kotlikoff, Laurence J. "Intergenerational Transfers and Saving," Journal of Economic Perspectives (Spring 1988). Kotlikoff, Laurence J., ed. What Determines Savings? (MIT Press, 1989). Masson, Paul R., and Ralph W. Tryon. "Macroeconomic Effects of Projected Population Aging in Industrial Countries," Working Paper 90/5, International Monetary Fund (January 1990). Modigliani, Franco. "The Role of Intergenerational Transfers and Life Cycle Saving in the Accumu lation of Wealth," Journal of Economic Perspectives (Spring 1988). United Nations. Demographic Indicators of Countries: Estimates and Projections as Assessed in 1980 (New York, 1982). Weil, David N. "What Determines Saving: A Review Essay," Journal of Monetary Economics (August 1991). Wolff, Edward N. "Capital Formation and Productivity Convergence Over the Long Term," American Economic Review (June 1991). Zeldes, Stephen P. "Consumption and Liquidity Constraints: An Empirical Investigation," Journal of Political Economy (April 1989). 23 FEDERAL RESERVE BANKOF PHILADELPHIA BU8INE8S REVIEW Ten Independence Mall, Philadelphia, PA 19106-1574