Macroeconomic Consequences of the Demographic Transition Ronald Lee UC Berkeley July 9, 2008 Talk prepared for Rand Summer Institute Research supported by NIA R37 AG025247 Thanks to Andy mason and NTA country teams Data from Plan • • • • Demographic transition Dependency ratios and support ratios Savings rates and capital intensification Human capital Data from I. The Demographic Transition • A classic illustration: The transition in India, 1890-2100. • Mixture of historical estimates, UN projections, and simulation based on fitted variations with time. Data from Data from Pre fertility decline; child dependency ratio rises During fertility decline, child dependency ratio declines Population aging: old age dep ratio rises Data from The total dep ratio rises, falls, then rises again, ending up where it started. The changes in the total dependency ratio are transitory. Data from But there is a big permanent change: At start, many children and few elderly. At end, few children and Many elderly. Data from Comments on simulation • Assumed TFR stabilized at 2.1; but often has declined below replacement. • Assumed e0 stopped rising at 80, but many countries already above this. • Some countries experienced important baby booms and busts which distort this classic shape. • Many countries now have declining populations and declining working age pops. Data from II. The economic life cycle: • Age profiles of consumption and labor income • Use estimates from the National Transfer Accounts project, or NTA. • Consumption patterns are quite similar for Third World countries in Asia and Latin America. • Consumption in Industrial populations looks different. Data from Per Capita Consumption and Labor Income A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 Includes self employment, wages,unpaid family labor, & fringe benefits. Averages 0’s and both male and female. 600 500 400 Labor Income Consumption 300 200 100 0 0 20 40 60 Age Data from An-Chi Tung Includes both private expends and in-kind public transfers (health, education, long term care) 80 Per Capita Consumption and Labor Income A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 600 Labor Income 500 Flat cons age profile in adult years reflects extended family sharing. Quite different than most industrial nations. 400 Consumption 300 200 100 0 0 20 40 60 Age Data from An-Chi Tung 80 Per Capita Consumption and Labor Income A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 600 Labor Income 500 400 Consumption 300 200 100 0 0 20 Large deficits 40 and at young old ages.Age Data from An-Chi Tung 60 80 A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 Per Capita Consumption and Labor Income 600 500 Reallocations from surplus to deficit 400 ages required. Consumption 300 200 Labor Income 100 0 0 20 40 60 Age Data from An-Chi Tung 80 A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 Other income comes from assets, foreign loans, and remittances from abroad—its not all labor income. Per Capita Consumption and Labor Income 600 500 400 Consumption 300 200 Labor Income 100 0 0 20 40 60 Age Data from An-Chi Tung 80 Per Capita Consumption and Labor Income A Typical Asian Economic Lifecycle: National Transfer Accounts estimates for Taiwan, 1998 600 Labor Income Asset income is partic Impt for old age 500 400 Consumption 300 200 100 0 0 20 40 60 Age Data from An-Chi Tung 80 Components of US Consumption, 2003 Unlike Taiwan and other Third World, in US cons rises strongly with age. True in other industrial too. Dollars (US, 2000) 40000 Public Health Private Edu Public Edu Private Health Private Durables 20000 Later I will measure HK investment Private Other As sum of pub and priv spending on hlth and educ as shown here. Public Other 0 0 10 20 30 40 Age Data from 50 60 70 80 90 • Levels of age profiles change fast with economic development. • Shapes of age profiles change slowly, • Are broadly similar across countries at very different levels of development. III. Dependency and Support • Concern about pop aging is mostly about old age dependency. • Sharpest concerns for age-sensitive public sector programs – pensions – health care – Long term care • But should place these in broader context – Full range of public programs – Private consumption • Use shape of estimated profile I just showed. Support Ratios • Effective labor is weighted sum of pop using labor income age profile. • Effective consumers is similar. • Ratio of effective labor to effective consumers is the “Support Ratio”. • Other things equal, consumption per effective consumer is proportional to the support ratio. P x yl x Effective Workers 0 Support Ratio Effective Consumers P x c x 0 1 Support Ratio for China, 1950-2100, Based on UN population projections and average LDC age profiles from NTA Effective Producers Per Consumer Population aging 0.9 0.8 0.7 First Dividend 0.6 0.5 1950 200 7 1970 1990 2010 2030 Year 2050 2070 2090 1 Support Ratios for Five Less Developed Countries, 1950-2100, Based on UN population projections and average LDC age profiles from NTA Effective Producers Per Consumer S. Korea China India 0.9 0.8 0.7 Brazil Niger 0.6 0.5 1950 200 2008 7 1970 1990 2010 2030 Year 2050 2070 2090 1 Support Ratios for Five Less Developed Countries, 1950-2100, Based on UN population projections and average LDC age profiles from NTA Effective Producers Per Consumer S. Korea China India 0.9 0.8 0.7 Brazil Niger 0.6 0.5 1950 2050/08 Rate %/yr 1970 Niger 1.20 0.43 1990 S. Korea 0.78 200 7 -0.59 2010 China 0.86 -0.35 2030 Year 2050 India 1.09 0.22 2070 Brazil 0.96 -0.09 2090 Support Ratios for Five More Developed Countries, 1950-2100, based on UN long term population projections and the NTA age profile for the US. Spain Effective Producers Per Consumer 0.8 US 0.7 Germany Italy 0.6 Italy, Low Fert. Spain, Low Fert. Japan 0.5 1950 1970 1990 2010 2030 Year 2050 2070 2090 Support Ratios for Five More Developed Countries, 1950-2100, based on UN long term population projections and the NTA age profile for the US. Spain Effective Producers Per Consumer 0.8 US 0.7 Germany Italy 0.6 Japan 0.5 1950 1970 2050/08 Rate %/yr US 0.91 -0.2 1990 Spain Italy 2030 2050 0.72 0.75 Year -0.8 -0.7 2010 Japan Germany 2070 2090 0.75 0.81 -0.7 -0.5 Italy, Low Fert. Spain, Low Fert. Proportionate Changes in the Support Ratio from 2007 to 2050 for Selected MDC and LDC Proportionate Changes in the Support Ratio from 2007 to 2050 for Selected MDC and LDC Proportionate Change in Support Ratio from 2007 to 2050 0.3 0.2 0.1 0 US Spain Italy Japan Germany Niger -0.1 -0.2 -0.3 -0.4 Country S. Korea China India Brazil Not written in stone. Many policy possibilities: • Change the age profile of labor income – – – – Later retirement Earlier entry into labor force Higher female labor force participation Reform seniority system • Change the age profile of consumption – In many industrial nations, the elderly consume much more than younger adults. – Makes population aging more costly – Role of public transfer policy: pensions, health care, long term care • Change the demographic trends: immig, fert IV. Further on Interage Flows of Income • Comparison of Japan and Indonesia Per capita consumption or labor income Per capita consumption and labor income by age for Indonesia and Japan 1,000,000 • Differences in consumption Indonesia, 2002 800,000 600,000 400,000 200,000 0 20 40 60 80 100 Per capita consumption or labor income in Yen Age – Education in Japan – Rising consumption in old age in Japan 500,000 Japan, 2004 400,000 300,000 200,000 100,000 0 20 40 60 80 100 Age Data from Maliki (Indonesia) and H. Ogawa (Japan) Aggregate Life Cycle Deficit for Indonesia (2005) in Rupiah Aggregated Consumption Labor Income 40,000 30,000 20,000 10,000 - 0 20 40 60 80 100 (10,000) (20,000) (30,000) Age Aggregate Life Cycle Deficit for Japan (2004) in Yen Aggregated Consumption - Labor Income 5,000 4,000 3,000 2,000 1,000 (1,000) 0 20 40 60 80 100 (2,000) (3,000) (4,000) (5,000) Age Data from Maliki (Indonesia) and H. Ogawa (Japan) Aggregate Life Cycle Deficit for Indonesia (2005) in Rupiah Aggregated Consumption Labor Income 40,000 30,000 20,000 10,000 - 0 20 40 60 80 100 (10,000) (20,000) (30,000) Age Aggregate Life Cycle Deficit for Japan (2004) in Yen Aggregated Consumption - Labor Income 5,000 4,000 3,000 2,000 1,000 (1,000) 0 20 40 60 80 100 • Green arrows show transfers from surplus of prime working years. • Red arrows show asset income consumed by elderly out of earlier savings. (2,000) (3,000) (4,000) (5,000) Age Data from Maliki (Indonesia) and H. Ogawa (Japan) 100 Population weighted average age Ac Per capita consumption or labor income Ac=30 Ayl=39 Indonesia, 2002 1,000,000 800,000 600,000 400,000 200,000 0 20 40 60 80 100 Per capita consumption or labor income in Yen Age Ac=45 Ayl=45 500,000 Japan, 2004 400,000 xPop x c x 0 100 0 Pop x c x • In Indonesia, average unit of income is earned at 39 and consumed at 30 • Travels 9 years down the age scale. • In Japan, it is earned and consumed at nearly the same age. 300,000 200,000 100,000 0 20 40 60 80 100 Age Data from Maliki (Indonesia) and H. Ogawa (Japan) Average Consumption-Earning Gap by Average Age of Population 4 Austria, 2000 2 Slovenia, 2004 0 Japan, 2004 Uruguay, 1994 US, 2003 Av Age Gap -2 -4 France, 2001 Sweden, 2003 13 yrs S. Korea, 2000 Costa Rica, 2004 Thailand, 2004Taiwan, 2003 -6 -8 Chile, 1997 Indonesia, 1999 India, 1999 -10 Philippines, 1999 -12 20 25 30 35 Ave Age of Population Data from NTA Country Teams 40 45 How much of the difference in age gaps is due to the shapes of the age profiles? Data from NTA Country Teams Average Age of Labor Income Average Age of Labor Income and Consumption with Population Held Constant (stationary, e0=75) 46 India, 1999 45 Japan, 2004 US, 2003 Philippines, 1999 Chile,Sweden, 1997 2003 Indonesia, 1999 44 43 Thailand, 2004 42 France, 2001 Taiwan, 2003 Costa Rica, 2004 Uruguay, 1994 S. Korea, 2000 41 Slovenia, 2004 40 Austria, 2000 39 Ac = Ayl 38 38 39 40 41 42 Average Age of Consumption Data from NTA Country Teams 43 44 45 Average Age of Labor Income Average Age of Labor Income and Consumption with Population Age Distr. Constant (stationary, e0=75) High cons (health care) and work when old. 46 India, 1999 Low age of cons due to heavy spending on education. 45 44 43 Japan, 2004 Philippines, 1999 Chile,Sweden, 1997 2003 Indonesia, 1999 Thailand, 2004 42 France, 2001 Taiwan, 2003 US, 2003 S. Korea, 2000 Costa Rica, 2004 Uruguay, 1994 41 Slovenia, 2004 40 Austria, 2000 39 Generous old age support. Ac = Ayl 38 38 39 40 Very young Ayl due 41 to early 42 start, 43 early retirement. Average Age of Consumption Data from NTA Country Teams 44 45 Average Age of Labor Income Average Age of Labor Income and Consumption with Population Held Constant (stationary, e0=75) Difference in average ages is the distance above (-) or below (+) the diagonal. 46 45 India, 1999 Japan, 2004 US, 2003 Philippines, 1999 Chile,Sweden, 1997 2003 Indonesia, 1999 44 43 Thailand, 2004 42 France, 2001 Taiwan, 2003 Costa Rica, 2004 Uruguay, 1994 S. Korea, 2000 41 Slovenia, 2004 40 Austria, 2000 39 Ac = Ayl 38 38 39 40 41 42 Average Age of Consumption Data from NTA Country Teams 43 44 45 Average Age of Labor Income Average Age of Labor Income and Consumption with Population Held Constant (stationary, e0=75) 46 India, 1999 45 Japan, 2004 US, 2003 Philippines, 1999 Chile,Sweden, 1997 2003 Indonesia, 1999 44 Indonesia -3.3 yrs 43 Thailand, 2004 42 France, 2001 Taiwan, 2003 Costa Rica, 2004 Uruguay, 1994 S. Korea, 2000 41 Austria +2.2 yrs Slovenia, 2004 40 Austria, 2000 39 Ac = Ayl 38 38 39 40 41 42 Average Age of Consumption Data from NTA Country Teams 43 44 45 • total range in age gap was 13 years • range due to differences in profiles is 5.5 years. • So both population age distribution and shapes of age profiles help determine gap. V. Wealth and the age gap: the golden rule case • Demographic and economic steady state • Saving and capital such as to maximize per capita consumption. • r=n+g Now suppose babies had to go into debt to feed themselves…. • At the start of life, c(x)>yl(x); dependency. • Suppose we keep a notional account of debt and credit over the life cycle, discounted to age 0. • Credit gained (or lost) at age x is: e-rx l(x) [yl(x)-c(x)] where r is interest rate, l(x) is survival from 0 to age x. • Cumulated up to age x, we get W(x): W x e l x yl x c x dx x 0 rx • Now find the average level of per capita in the whole population, call it W • W = pop(x)*W(x)/totpop W 0 e l x W x dx rx 0 e l x dx rx The average wealth per capita in the population may be pos or neg The Willis result W = c(Ac – Ay) , where c = per capita cons – If Ac>Ay then indivs need to hold onto some output for later consumption, so wealth, W, is on average positive in the population. – If Ac<Ay then indivs consume before they produce, and must go into debt on average, so W is negative. • Alternatively: W/c = Ac – Ayl • So wealth relative to consumption is roughly proportional to Ac - Ayl • Given comparative analysis of Ac-Ayl, suggests that demand for wealth rises over the demographic transition. • Why? – Older people hold more wealth; in old population, there are more of them. – Longer life means workers need to accumulate more wealth for longer old age. – Lower fertility means adults consume more and need to save more to maintain in old age. VI. The role of intergenerational transfers • We just considered the wealth needed to achieve consumption targets. • Wealth can be held in two forms: – Transfer wealth (expected future transfers received minus expected future transfers made) – Assets or Capital NTA data on shares of old age support from different sources • Asset income (land, equities, interest, etc.) • Family transfers (not including bequests at death) • Public transfers (Pay As You Go pensions, health care, and long term care) • Triangle graph shows shares, not levels, so must add to 100%. • Bequests not included; just old age cons. Old-age Reallocation System, Selected Countries. Familial transfers equally important in Thailand, Korea, and Taiwan (36-40%). Net public transfers to elderly are zero in Thailand; about 25% in Taiwan and Korea. 100 0 25 Net familial transfers near zero in US, CR, and J. Large public transfers in CR and J 75 Public transfers (%) Thailand US 50 Asset-based (%) 50 Korea Taiw an Costa Rica Japan 75 25 100 0 100 75 50 Fam ily Transfers (%) Diagram from Andy Mason 25 0 Old-age Reallocation System, Selected Countries. Public transfers: Thailand none, Japan and Costa Rica around 70% US, Korea, Taiwan, middling 100 0 25 75 Public transfers (%) Thailand US 50 Asset-based (%) 50 Korea Taiw an Costa Rica Japan 75 25 100 0 100 75 50 Fam ily Transfers (%) Diagram from Andy Mason 25 0 Old-age Reallocation System, Selected Countries. 100 Reliance on assets : Japan, Taiwan, C.R. are low; Thailand high; US middling 0 25 75 Public transfers (%) Thailand US 50 Asset-based (%) 50 Korea Taiw an Costa Rica Japan 75 25 100 0 100 75 50 Fam ily Transfers (%) Diagram from Andy Mason 25 0 VII. Demographic Transition and Capital Accumulation • Changing dependency gets most attention for ec dev and pop aging. • Changes in capital accumulation may be more important. Calculating the demand for wealth and capital over the demographic transition • Based on different theoretical models, approaches. • Model with Social Planner maximizing discounted social welfare function with full foresight. • Model with individuals saving and consuming over their life cycles to maximize their life time utility. Here take a different approach – no optimization--emphasizes institutional setting • Assume – share of old age consumption supported by asset income stays constant over time. – altruistic sharing maintains the shape of the cross sectional consumption age profile. – Demography is known in advance. • Can solve recursively for unique growth path and asset holdings. Two scenarios: high level of transfers to elderly (65%); or low level (35%) • Other assumptions – Productivity growth raises income age profile by 2% per year. – Open economy; rate of return on assets is 3%. • Aggregate saving is calculated to maintain asset share of old age consumption support. • Results will be shown relative to a 2% growth trajectory from prod gr. Simulated Saving Rate, ASEAN (S.E. Asian countries), 1950-2050 0.25 Net Saving Rate 0.2 Low IG Transfers 0.15 0.1 0.05 High IG Transfers 0 1940 1960 1980 2000 2020 From Mason, Lee and Lee (2008) 2040 2060 Simulated Assets/Labor Income, ASEAN Assets/Labor Income . 8 6 Ratio of assets to labor income rises greatly in any case, but 3 or 4 times as much with low IG transfers. Low IG Transfers 4 High IG Transfers 2 0 1940 1960 1980 2000 2020 From Mason, Lee and Lee (2008) 2040 2060 Consumption Index (1950=100) . Simulated Consumption, ASEAN 160 Low IG Transfers 140 High IG Transfers 120 100 With low IG transfers, saving is higher from 1990 to 2020, reducing consumption. 80 60 1940 Thereafter, it is higher. 1960 1980 2000 2020 From Mason, Lee and Lee (2008) 2040 2060 These sorts of results are qualitatively like those from optimization approaches • • • Timing of swings differs Level of savings rates differs Capital/labor income ratios differ Big picture is the same: 1. The demographic transition leads to a major increase in capital per worker. 2. The greater the role of transfers to the elderly, the smaller is the increase in capital intensity. 3. Eventually consumption rises with lower transfers, but initially it is lower. 4. Population aging leads to a decline in savings rates but an increase in capital intensity. VIII. Human capital and the demographic transition • Measure public and private expenditures on health and education at each age. – Sum these for health ages 0-18 – Sum for education ages 0-26 – Gives synthetic cohort HK investment per child • Construct ratio of HK to average yl(x), ages 30-49. • Plot log of HK/w against log of TFR. Figure 1. Per Child HK Spending (Public and Private) vs. Fertility ln(HK per Child/Av Lab Inc 30-49) 2.00 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 0.00 0.20 0.40 0.60 0.80 ln(TFR) Data from NTA country teams 1.00 1.20 1.40 Figure 1. Per Child HK Spending (Public and Private) vs. Fertility ln(HK per Child/Av Lab Inc 30-49) 2.00 Twn 1.80 1.60 1.40 Swd Jpn Slv Hng Aust Fr Kor Brz US Mex Fin 1.20 Thai 1.00 Chl CR Urg 0.80 Phil 0.60 Indonesia 0.40 0.20 0.00 0.00 India 0.20 0.40 0.60 0.80 ln(TFR) Data from NTA country teams 1.00 1.20 1.40 Figure 1. Per Child HK Spending (Public and Private) vs. Fertility ln(HK per Child/Av Lab Inc 30-49) 2.00 1.80 1.60 1.40 1.20 1.00 0.80 0.60 y = -1.05*x + 1.92 R2 = 0.62 0.40 0.20 0.00 0.00 0.20 0.40 0.60 0.80 ln(TFR) Data from NTA country teams 1.00 1.20 1.40 Now calculate total HK spending on all children • Multiply TFR times HK per child, and plot its log against log(TFR). ln(TFR X Per Child HK Spending/ Av Lab Inc 30-49) 2.50 Figure 5. Total Expenditures Per Woman for All Children's HK vs. Fertility for 18 NTA countries (log scale) 2.00 1.50 1.00 Roughly a horizontal cloud, 6.8 years of labor income are invested negative. inperhaps total HKslightly on average. 0.50 0.00 0.00 1/12 of lifetime labor income for a couple. 0.20 0.40 0.60 0.80 ln(TFR) Data from NTA country teams 1.00 1.20 1.40 Association is non-causal • We don’t know whether fertility decline causes rising HK investments per child. • Desire to make bigger HK investments causes fertility decline. • Some other factor like rising income causes both fertility and HK changes. • Here is one theory about a causal path from income growth to other changes. In some models the HK growth causes income growth. The standard Quantity-Quality model • Assume that the share of total labor income spent on HK is fixed, consistent with scatter plot. • Draw budget constraints for differing levels of income. • Quantity and quality interact multiplicatively in the budget constraint, both with positive income elasticities for constant price. The Standard Model: Rising Income Leads to Choice of Lower Fertility and Higher HK Investment per Child Human Capital Investment per child 7 Yn=6 6 5 4 Yn=4 3 2 1 Yn=1 0 1 2 3 4 5 Number of children 6 7 8 The Standard Model: Rising Income Leads to Choice of Lower Fertility and Higher HK Investment per Child Human Capital Investment per child 7 With same data, plot ln(HK/w) instead of HK, against ln(TFR) instead of n. Yn=6 6 B 5 The budget lines collapse onto a single straight line. 4 Yn=4 3 2 C 1 Yn=1 A 0 1 2 3 4 5 Number of children 6 7 8 Figure: The transformed budget constraint showing different quantity-quality choices. d is HK expenditure ln(d) expressed in years of work at rate w Ln(pqq/w) B D Quite similar to empirical scatter Intercept of scatter indicates years of work expended on HK is 6.8. Share of lifetime labor income is 1/12. Ln(n) Slope (elasticity) = -1 C A • So our scatter plot shows a common transformed budget constraint with different fertility-HK choices. • Differing incomes is one possible cause. • Many others. Production and Human capital • Human capital (HK) – Portion of wage, W(t), workers invest in their children is inversely related to their fertility, F(t) – Human capital of workers one period later is – HK(t+1) = h(F(t)) W(t) • Wage (W) – Wage is increasing in human capital – W(t) = g(HK(t)) Baseline Specifications HK t 1 W t 12 F t W t HK t .33 Other sources of variation in fertility/HK choice • Pref for HK: Rate of return to HK; survival rates; consumption value of HK. • Price of HK due to medical technology, transportation improvements, etc. • Price of number: family allowances, fines for second child, changing access to effective contraceptives • Cultural influences on varying share of income allocated to total HK expenditures and on number. Model—basic structure • Take fertility variations as given, trace out consequences for HK, w, consumption. • 3 generations: children, workers, retirees; usual accounting identities. • No saving or physical capital. • HK drives wage growth; wage growth drives HK growth. (Lee and Mason 2008) Figure 6. Macro Indicators: Baseline Results Value (percent of year 0) Boom 130.0 (demoraphic dividend) 120.0 110.0 100.0 Support ratio C/ EA 90.0 Fertility bust, but 80.0 consumption 70.0 remains high Fertility recovers: modest effect on C/EA 60.0 0 1 2 3 4 5 6 Period Bottom line: Low fertility leads to higher consumption. Human capital investment has moderated the impact of fertility swings on standards of living. From Lee and Mason (2008) Figure 6. Macro Indicators: Baseline Results Value (percent of year 0) 130.0 120.0 110.0 100.0 Support ratio C/ EA 90.0 80.0 70.0 During first dividend phase, consumption does not rise as much as support ratio. 60.0 0 2 4 in HK.5 The1 difference is 3 invested 6 That is why ih Period later periods, consumption is proportionately higher than the support ratio. From Lee and Mason (2008) Conclusions for changes over the transition • Support ratios change over demographic transition; ending where started, roughly. – Importance in long view may be exaggerated. – In shorter view, we are in painful payback phase. • Bigger effect is on capital intensity – These raise productivity per worker – Raise wealth and asset income • Increased human capital results from low fertility—so closely related to aging: same cause for both. – Raises productivity. • However, increased demand for wealth can be met either by increased asset holdings or through increased transfer wealth. • Major role for policy and institutions at every point; nothing inevitable.