88221 MAY 2014 • Number 144 IS AGING BAD FOR THE ECONOMY? MAYBE Harun Onder and Pierre Pestieau The world’s aging population is expected to shape the future of economies across the globe. Without behavioral adapta- tion by current and future generations, this demographic transformation is likely to slow down economic growth. How- ever, aging will also induce behavioral adjustments in savings and labor force participation. Will these adjustments be large enough to reverse the negative effects of demographic change? The answer depends on a number of conditions. This note suggests that determining whether aging is driven by an increase in longevity or by a decrease in fertility is impor- tant for understanding the size and direction of these demographic effects. Moreover, the type of unfunded social security system that exists in the economy could also influence the net effect of aging. Changing demographics in modern times are continuing to reshape both the scale and age composition of the world’s break records. First, it was the remarkable speed of popula- population in coming decades, albeit by different degrees tion growth. According to United Nations (UN) calculations, across countries. the world population nearly doubled in the last 40 years, in- Table 1 shows that the “big jump” in the median age has creasing from about 3.7 billion in 1970 to 6.9 billion in 2010. already occurred in high-income countries (HICs). In 2010, Next, it will be about slowing down. Total population is ex- the median age in HICs was 9.5 years older than in 1970, and pected to grow by only 40 percent within the next 40 years, is expected to increase an additional 5.9 years by 2050. In reaching about 9.6 billion by 2050. comparison, middle- (MICs) and low-income countries In addition to slowing population growth, the median (LICs) are expected to have such jumps within the next 40 age of Earth’s inhabitants will continue to get older, a phe- years. However, these changes will happen in different ways nomenon known as demographic aging. Table 1 shows that due to variations in fertility and longevity trends. the median age has increased by 7.1 years since 1970, and it First, compared to the last 40 years, the reduction in fer- is expected to increase by another 7.8 years to reach 37 by tility rate will not be as dramatic in MICs, while it will con- 2050. tinue to decrease significantly in LICs. Second, the increase in To put these two trends into perspective, it is impor- life expectancy at birth will slow down in both groups. How- tant to remember that aging and population growth are ever, the increase in life expectancy at the age of 60, a measure mainly driven by two factors. The first factor is longevity. of longevity, will slow down in MICs and accelerate in LICs Mortality rates for each age group have been declining (see last column of table 1). Between 1970 and 2010, the in- steadily for many decades. On average, people now tend to crease in life expectancy at the age of 60 was 3.4 years in MICs live longer. The second factor is fertility. The average num- and 2.5 years in LICs. These will change to 2.6 and 3.4 years, ber of births per woman has also declined sharply in the last respectively, by 2050. several decades. These trends in increasing longevity and What effects will these changes have on different econo- decreasing fertility rates are expected to continue, and will mies? There is no obvious answer to this question concerning 1 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Table 1. Demographic Trends by Income Groups Life expectancy Median age Total fertility rate Life expectancy at birth at age 60 1970 2010 2050 1970 2010 2050 1970 2010 2050 1970 2010 2050 Low-income countries 18.0 20.0 27.5 6.7 4.6 2.8 45.1 57.5 69.7 14.3 16.8 20.2 Middle-income countries 19.5 28.1 37.7 5.7 2.5 2.1 54.4 67.2 73.9 15.2 18.6 21.2 High-income countries 29.6 39.1 45.0 2.6 1.7 1.8 69.2 77.8 83.7 17.7 22.8 26.2 World 22.1 29.2 37.0 5.0 2.6 2.2 55.5 66.5 73.9 16.2 19.8 21.9 Source: Onder, Hernandez, and Usmani (2014), based on UN population projections. such a multi-dimensional process. A number of consequenc- Other Things Being Equal, Aging Is Bad for es will clearly be unfavorable. For example, with aging comes the Economy an increase in the old-age dependency ratio, or in other words, A simplistic way of demonstrating the economic impact of ag- an increase in the number of elderly individuals per young ing is to use a counterfactual analysis. This is, in essence, an person. This, in turn, leads to imbalances in unfunded (pay- accounting exercise that studies the relative size of demo- as-you-go [PAYG]) pension budgets. However, some of the graphic shifts in the absence of behavioral adjustments. In the consequences could be positive: for instance, a slowdown in population growth could relieve some of the pressures on first step, historical growth rates in gross domestic product limited environmental resources. (GDP) are decomposed into three components: labor produc- There are some consequences for which there is no con- tivity growth, population growth, and growth in labor force sensus on whether the outcomes would be beneficial or not. participation rate (box 1). Next, the last two components are A particularly important one is the effect of aging on capital replaced with projections of future demographic changes, accumulation and economic growth. The conventional wis- while the first component is held constant. A comparison of dom is that aging would lead to slower growth, because the this new “counterfactual” GDP series with the original shows old save less, which translates into higher interest rates, lower how a reduction in population and labor force participation investments, and lower labor productivity. In contrast, others would have affected GDP growth in the absence of a change in think aging may actually increase economic growth because productivity. people will adapt by saving more and working longer (The Figure 1 shows the results of such an exercise for 167 Economist 2014). Bloom, Canning, and Sevilla (2001) discuss countries from all income categories. On average, had the different approaches regarding the link between economic world already experienced the future demographic transition growth and demographic change. Overall, this debate has not within the last 40 years, the average GDP growth rate would yet been concluded. have fallen by about 1.2 percentage points annually. To put Dedry, Onder and Pestieau (forthcoming) investigate this into perspective, a reduction of this size in the annual this last effect by separating the incidences of increased lon- growth rate would translate into an almost 40 percent de- gevity and reduced fertility on capital accumulation. Results crease in countries’ GDPs by the end of 40 years. The reduc- show that the type of aging is important in determining the tion in the GDP growth rate reflects both the slowdown in outcomes. This is mainly because changes in fertility and lon- population growth and the increase in old age dependency gevity rates provide different incentives for individuals to save ratios. In comparison, the GDP per capita growth rates con- more or work more. These differences could be more pro- trol for the change in the size of population (see also box 1). nounced depending on the type of social security system in a As a result, growth rates would be reduced by only 0.4 per- given economy. centage points annually. This note starts with an investigation of the implications These results, however, are not the same across countries. of the demographic transition without a discussion of the Whereas all HICs (except Estonia) would have experienced a changes in incentives or associated behavioral adjustments, reduction in growth rates, many LICs and some MICs would focusing on accounting effects only, and then introduces the have benefitted from the prospective dynamics of their demo- behavioral effects as discussed in Dedry, Onder and Pestieau graphics. This contrast mainly arises from the “differences in (forthcoming). differences.” Remember that the counterfactual exercise basi- 2 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Box Box1. Counterfactual Exercise 1. Counterfactual Box 1. Counterfactual Exercise Exercise This counterfactual Box 1. Counterfactual Exercise analysis provides a rule-of-thumb method for illustrating the implications of a forthcoming demographic This counterfactual This counterfactual analysis analysis provides a rule-of-thumb method for illustrating the the implications of of a a forthcoming demographic change on GDP growth. The provides first step a rule-of-thumb involves reorganizing method the for GDP illustrating per working-age implications population to explicitly demographic forthcoming identify the change This counterfactual change on Box on GDP GDP 1.analysis growth. Counterfactual growth. provides The The first first a Exercise step rule-of-thumb step involves involves reorganizing method for illustrating reorganizing the the GDP GDP per per the working-age implications working-age population of a forthcoming population to to explicitly explicitly demographic identify the identify thechange population on GDP population growth. Thesize and age structures. population size size firstand and stepage age involves structures. structures. reorganizing the GDP per working-age population to explicitly identify the population size and This counterfactual analysis provides a rule-of-thumb method for illustrating the implications of a forthcoming demographic age structures. Let and be t, respectively. change onthe GDP GDP growth. (measured The first in step purchasing involves reorganizing power parity the terms)GDP per and total population working-age population in year to explicitly identify the Then, the Let Let Let Yt and and following andPequation population t be be be the the the GDP GDP GDP presents size and (measured age (measured (measured an identity structures. inin in that purchasing purchasing purchasing must hold power power power for all parity parity parity values: terms) terms) and terms) and total population total total populationin population in inyear yearyeart t, ,t, respectively. respectively. respectively. Then, Then, Then, the thethe fol- following following lowing equation equation equationan presents presents presents identity an anthat identity identity must that that hold must must for hold hold all values: for for all values:all values: Let and be the GDP (measured in purchasing power parity / and total population in year t, respectively. Then, the terms) / following equation presents an identity that must hold for = / all = values: = / / / / where shows the working-age population. By using a log transformation = and rearranging, the above identity becomes: where Ltwhere where shows the shows shows working-age the the working-age working-age population. population. population. By using By By a using using log transformation a log a log transformation / transformation and rearranging, and and rearranging, rearranging, the above the above the above identity identity identity becomes: becomes: becomes: where shows the working-age population. ln( By / using ) ) a = ln( log / transformation ) + ) + ln(and ln( / / ) rearranging, ) the above identity becomes: ln( ln( / = ln( / / ) = ln( / ) + ln( / ) Whereas the left-hand side denotes the income per ln( capita / ) =in ln( year / t), the + ln( first / term ) on the right-hand side shows the output per Whereas Whereas thethe the left-hand left-hand side side denotes denotes the the income income per per capita capita in year t t, , tthe first term on the right-hand side shows the output per Whereas working-age left-hand person, side and thedenotes second the term income shows per the sharein capita in of year year working-age the , the first first term group term on inon thetotal right-hand the right-hand population. side The shows side shows first theandoutput the per per output second working-age working-age working-age person,Whereas person, andthe left-hand person, the and and second the the side seconddenotes second term term term shows the income shows showsthe per share the the capita share share of in year working-age of of working-age t, the first working-age group term group group on total in thein in total right-hand total population. population. population. side shows The The the output The first first first and per second and and secondsecond terms on terms on the right-hand side of second this equation canthe beshare interpreted as average labor productivity and employment ratio, terms terms the right-hand on onworking-age side the the of right-hand person, right-hand this equation side and side the can of of bethis this equation interpretedterm shows equation can can as be be average interpreted interpreted oflabor working-age as as productivity average average group in andlabortotal labor productivity population. productivity employment The ratio, and and employment firstrespectively, and second under employment ratio, ratio, respectively, terms under on the a set of assumptions. right-hand side of this equation These include can be interpreted the absence as average of unemployment, labor productivity full labor and employment force participation ratio, in a set respectively, respectively, of assumptions. These under underinclude a a set set the of of absenceassumptions. assumptions. of unemployment, These These include include full the the labor absence absence force of of participationunemployment, unemployment, in full full labor working-age labor cohorts, force force participation participation and no laborin in force working-age respectively, cohorts, and under no ano labor set force of assumptions. participation These include among the the elderly absence and ofand children. unemployment, full labor force participation in working-age working-age participation among cohorts, cohorts, the elderly and and andno labor labor children. force force participation participation among among the the elderly elderly and children. children. working-age cohorts, and no labor force participation among the elderly and children. Finally, Finally, the the population population can can be beremoved removed from from the the left-hand left-hand side side to to focus focus on on GDP GDP growth. growth. Finally, Finally,Finally, the the populationpopulation the population can be can be canremoved removed be removed from from from the the the left-hand left-hand left-hand side side sideto to tofocus focus focus onon on GDP GDP GDP growth. growth. growth. ln( ))= ln( / )) + ln( / )) + ln( )) ln( ) = =ln( ln( / )+ + ln( / ) ln( + ln( + ) ln( ln( ) = ln( / ) / +ln( ln( / / ) + ln( ) Iterating Iterating this this identityidentity for for one one period period and and subtracting subtracting gives gives usus the the growth growth rates rates in ineach each component: component: Iterating this Iterating this identity Iterating identity for this identityfor one one period for period and one period and subtracting and subtracting gives subtracting gives us gives us the us the growth the growth rates growth rates rates in in in each each each component: component: component: +1 +1 �� +1 +1 ≡����������� ln( ≡ +1 ln( ) �����������− ln( ) − ln()= ) = �ln � � ln �+1 �− �− − ln ln �� �� + + ��ln ln�� +1 +1 �� − − ln ln �ln ����� [ln( + ������������� [ln( + ������������� [ln( +1 +1 ) − ln( )− ln( )] )] ≡ ln( ����������� +1 ) ) − +1 ln( ) ) = � ln � +1 � ln � �� + � ln � +1 � − � �� + + ������������� [ln( ) )− ln( ln( )] +1 +1 ≡ ln( ����������� +1 +1 − ln( ℎ = ������������� � ln ������������� � �������������+1 +1 � − ln � �� + ������������� � ln ������������� ������������� � +1 +1 � − ln � �� ������������� +1 +1 − ℎ )] ℎ ℎ ������������� +1 +1 ℎ ℎ ������������� +1 +1 ℎ ℎ ℎ ℎ ℎ ℎ ℎ ℎ ℎ ℎ The The first first The termtermfirst in interm square square square inbrackets brackets brackets onon the on the the right-hand right-hand right-hand side side side shows shows shows thegrowth the the growth growth inin in per per per worker worker worker output, thus output, output, the the thus thus growth the in labor growth growth inin labor labor produc- The The first first term term in in productivity. square square brackets brackets The second on on term shows the the theright-hand right-hand side side growth in group shows shows working-age the the group growth growth in in to populationper per worker worker ratio, thus output, output, the change thus thus the in thethe growth growth dependency in in labor labor tivity. The second term shows the growth in working-age to population ratio, thus the change productivity. The second term shows the growth in working-age group to population ratio, thus the change in the dependency in the dependency ratio. Finally, productivity. The ratio. Finally,second term the third term shows shows the growth population in working-age in working-age group growth. group toto population population ratio, ratio, thus thus the change in the change in the the dependency the thirdproductivity. term shows The second population term growth.shows the ratio. Finally, the third term shows population growth.growth dependency ratio. ratio. Finally, Finally, the third the third term shows term shows population population growth. In the spirit In of Bloom, the spirit ofCanning Bloom, and Fink Canning and Fink growth. (2011), the second (2011), stage the second of of stage thetheexperiment experiment uses theabove uses the above decomposition decomposition to to generate In the counterfactual spirit GDP ofof generate Bloom, growth Canning rates. counterfactual Specifically, GDP and growth Fink the(2011), rates.(2011), average theannual Specifically, second the stage employment average of the annual experiment ratio employment ratio uses growth andthe growth above population and decomposition growth population growthrates tobetween In the spirit In the spirit Bloom, of Bloom, Canning Canning and Fink and Fink (2011), thethe second second stage stage of the of the experiment experiment uses uses the above the above decomposition decomposition to to 2010 and generate 2050 counterfactual are rates plugged between 2010GDP into and growth the 2050 are rates. equationplugged Specifically, along intowith thethe the equation average actual alongvalues withannual the of employment output actual per values ratio ofworker output growth growth per workerand from population the growth from the growth period 1970–2010. generate generate counterfactual counterfactual GDP GDP growth growth rates. rates. Specifically, Specifically, the the average average annual annual employment employment ratio ratio growth growth and and population population growth growth between ratesshow The results period what the and 2010 1970–2010. 2050 average are plugged results The annual show what growthinto theaverage the rate ofequation GDPannual along would with growth be theof rate between actual GDP 1970 values and be would 2010 hadper output ofbetween 1970 the worker and 2010 countrygrowth had in from thealready the discussion rates between between rates the 2010 2010 and and 2050 2050 are are plugged plugged into into the the equation equation along along with with the the actual actual values of output per worker growth from the period experienced country 1970–2010.in discussion forthcoming The already results show demographic experienced what the change. the forthcoming average annualdemographic growth change. rate of wouldof GDPvalues output be between worker per 1970 and 2010 from growth the had the period period 1970–2010. 1970–2010. The The results results show show what what the the average average annual annual growth growth rate rate of of GDP GDP would would be be between between 1970 1970 and and 2010 2010 had had the the country Source: Onder, in discussion Hernandez, Source: and Usmani Onder, already 2014.experienced Hernandez, and Usmani the forthcoming demographic change. 2014. country in country discussion already in discussion already experienced experienced the forthcoming demographic the forthcoming demographic change. change. Source: Onder, Hernandez, and Usmani 2014. Source: Onder, Hernandez, Source: Onder, Hernandez, and Usmani 2014. and Usmani 2014. cally compares future demographic changes with the past. The first important behavioral adjustment can occur via Those with favorable demographic dynamics in the coming savings. As people live longer lives after retirement, or as their decades, or those who already completed the demographic retirement incomes change due to demographic transition, transition to a new stable level, would have relatively smaller their incentives to save or dissave will also change. Changes in differences between the actual and counterfactual growth savings will, in turn, affect the productivity of labor by chang- rates. In comparison, those in the midst of the transition ing the capital available per worker. would exhibit large impacts because the demographic chang- The second important behavioral adjustment could oc- es in the future could be substantial compared to changes in cur in decisions regarding labor force participation. A longer the past. life span could induce individuals to extend their active work life if retirement is not mandatory. Thus, the size of the labor But, Other Things Will Not Be Equal force may not shrink as much as the demographic simulations The counterfactual analysis is useful for demonstrating the predict. relative magnitude of future demographic changes. However, By holding the age of retirement and labor productivity in practice, aging affects the economy through other channels growth constant, the counterfactual exercise above shuts as well. Individuals can change their behavior to cope with a down these behavioral channels. Therefore, aging and the new economic reality. Overall, this adaptation can significant- slowdown in population growth always bring about negative ly influence the impact of aging on economic growth. consequences in the accounting exercise. Therefore, the next 3 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Figure 1. What Would Demographic Dynamics from 2010–50 Imply for Growth Rates in 1970–2010? a. GDP b. GDP per capita 1 1 Implied change in growth rate Implied change in growth rate 0 0 -1 -1 -2 -2 -3 -3 -4 -4 -5 0 5 10 -5 0 5 10 Actual Actual HIC MIC LIC HIC MIC LIC Mean deviation from zero: -1.181 Mean deviation from zero: -0.442 Mean absolute deviation from zero: 1.257 Mean absolute deviation from zero: 0.636 Source: Onder, Hernandez, and Usmani (2014), based on Penn World Tables 8.0; the United Nations World Population Prospects: The 2012 Revision, http://esa.un.org/wpp/; and World Bank staff calculations. obvious question is: can behavioral adjustments overcome the emerges in the way that they respond to demographic negative accounting effects? changes. Under the DC system, contributions are fixed, and any reduction in the number of contributors leads to a Types of Aging and Social Security reduction in benefits per beneficiary. In comparison, un- Systems Matter der the DB system, the benefits are fixed; hence, any loss Dedry, Onder, and Pestieau (forthcoming) contribute to the due to a reduction in the number of contributors is com- discussions on behavioral adaptations by showing that adjust- pensated by increasing the contribution rate. The DA sys- ments may differ under different types of aging and social se- tem is similar to the DB system—the only difference is that curity systems. Their analysis uses a two-period overlapping benefits are paid in lump sum under the DB, whereas un- generations (OLG) model with endogenous retirement deci- der the DA they are paid in annuities. Therefore, the DB sions. The model separates the factors that change the scale system is not affected by changes in the life span, whereas and age composition of the population, for example, decreas- total benefits under the DA system increase if the individ- es in fertility and increases in longevity. Savings are motivated ual lives longer. by financing consumption in life after retirement. Therefore, To see how different pension systems and types of aging changes in fertility and longevity have different impacts on may influence the impact of aging on capital accumulation, it optimal savings. is useful to begin with a benchmark case. Behavioral adjustments can affect capital per worker and Benchmark case with no pension system or work in old age labor productivity through two channels. First, individuals The well-known standard Diamond model can be used as an prefer to smooth consumption over their lifetime. Therefore, appropriate benchmark. In this model, individuals live two when needed, they increase the savings to finance their con- periods in the absence of social security and work after retire- sumption in old age. Higher savings will then increase the ment. Table 2 summarizes the results. Whether longevity in- capital per worker, thus increasing labor productivity. Sec- creases or fertility decreases, aging increases the capital accu- ond, individuals can increase the length of their active work mulation in the standard case. life, especially when life spans increase. However, in contrast When aging is driven by a decrease in fertility, capital to savings, higher labor force participation decreases produc- per worker increases because savings do not change, but the tivity, mainly because there are more workers for a given level number of workers decreases compared to the case without of capital stock. any change in fertility rate. When an increase in longevity is The study also compares three unfunded pension sys- the reason for aging, the increase in capital per worker is tems: defined contribution (DC), defined benefit (DB), driven by an increase in per person savings. This adjustment and defined annuity (DA). In all cases, total contributions is a result of consumption-smoothing motives; for example, are collected in the form of payroll taxes and fully distrib- individuals need more savings to support themselves if they uted as benefits. Therefore, the pension system is finan- live longer. Thus, with higher savings comes more rapid cap- cially balanced. The main difference of these systems ital accumulation. 4 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Table 2. Aging’s Impact on Capital per Worker under Different Unfunded Pension Systems Standard case Defined contribution Defined benefits Defined annuities Mandatory early retirement Decrease in fertility >0 >0 ≷0 ≷0 Increase in longevity >0 ≷0 ≷0 ≷0 Optimal retirement Decrease in fertility >0 >0 ≷0 ≷0 Increase in longevity >0 ≷0 ≷0 ≷0 Source: Dedry, Onder, and Pestieau forthcoming. An unfunded pension system with defined contributions In the context of increased longevity, the outcomes are In this case, pension benefits adjust to satisfy the fiscal balance identical for both DC and DB systems. This is mainly because when fertility rate changes. Changes in longevity, on the other the change in longevity has no effect on pension balances, be- hand, do not create imbalances in the DC case. cause the pension lump sum payment is invariant to the To see the impact of a reduction in fertility, remember length of the second period. Thus, there is no adjustment via that there are two demographic factors that are relevant here. pension contributions or benefits channel. First, the number of pension contributors decreases. This An unfunded pension system with defined annuities translates into a decrease in pension benefits. In response to Under a DA pension system, changes in both longevity and this reduction in old-age income, individuals save more. Sec- fertility affect the contribution ratios (box 2). Therefore, a de- ond, the number of workers decreases. Therefore, with the crease in the fertility rate or an increase in longevity that is number of workers shrinking and savings simultaneously in- more than the endogenous change in optimal retirement age creasing, the fertility rate decrease has a positive impact on lead to higher contribution rates to rebalance the pension capital accumulation. fund. In contrast, the positive result from the benchmark case The impact of a decrease in the fertility rate on capital weakens, and is possibly reversed, when there is an increase in accumulation with defined annuities is identical to the DB longevity and a PAYG pension system with DCs and the pos- case. The net effect is ambiguous, and it is determined by two sibility of work in old age are introduced. Remember, increas- opposing forces: a reduction in number of workers, which in- ing longevity has multiple effects on decision makers. First, creases per worker capital, and a reduction in per person sav- other things being equal, a longer life span induces higher sav- ings, which decreases it. In the end, country-specific charac- ings via the consumption-smoothing channel. Second, as life teristics will determine which effect dominates. spans increase, individuals would choose to work longer if re- The effects of an increase in longevity are complex in a tirement age is not regulated. However, this second adjust- DA system. In contrast to other social security systems, an in- ment could weaken or even reverse the need to save compared crease in life span increases second-period income. With man- to the benchmark case. Therefore, the net effect of an increase datory retirement, this is solely because each individual re- in longevity on capital per worker is ambiguous under a DC ceives the annuity for a longer period. When individuals can pension system. choose the retirement age, however, a longer work life will An unfunded pension system with defined benefits optimally contribute to second-period income via wage earn- Unlike DC pension systems, when the fertility rate changes ings as well. The increase in old-age income, in turn, reduces under a DB system, pension contributions adjust to satisfy the young generation’s incentives to save. However, savings the pension system balance. Changes in longevity, however, may still increase if the annuities are small compared to first- do not create imbalances, as in the previous case. period income, resulting from the need to finance a longer life When the fertility rate decreases, contribution rates in- span. Thus, the net effect on savings is indeterminate, as is the crease to reestablish the balances in the social security system. net effect on capital per worker. This reduces first-period income. As a result, individual sav- Simulations and More Technical Results ings decrease to allocate adjustment costs across the periods as implied by the consumption-smoothing motive. At the same There are two results that are interesting from more special- time, the number of workers decreases with the reduction in ized perspectives. First, in the numerical simulations per- fertility rate. Overall, the net effect on capital per worker is formed by Dedry, Onder, and Pestieau, the net effect of both ambiguous because lower savings have a negative impact and decreasing fertility and increasing longevity on capital accu- fewer workers have a positive impact on capital per worker. mulation and utility remained positive in all ambiguous cas- 5 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Box 2. Modeling the Effects of Demographic Change on Capital Accumulation Dedry, Onder, and Pestieau (forthcoming) use a standard two-period overlapping generation model with endogenous retirement to investigate the impact of demographic changes on capital accumulation and growth. An individual belonging to generation lives in two periods: and + 1. The first period of the individual’s life has a unitary length, while the second one has a length ≤ 1, where reflects variable longevity. In the first period, the individual works and earns , which is devoted to the first period’s consumption, , saving , and pension contribution . In the second period, the individual works an amount of time +1 ≤ ≤ 1 and earns +1 +1 . This earning, the proceeds of saving +1 , and the PAYG pension finances second-period consumption, +1 . The model assumes that working in the second period +1 implies a monetary disutility 2 (+1 , ), with > 0, 2 > 0 for the existence of a unique solution, and < 0, which reflects the idea that an increase in longevity fosters later retirement. Note that, for simplicity, the earnings in the second period of life are not taxed. Intuitively, the end of the first period can be interpreted as the statutory age of retirement. Dedry, Onder, and Pestieau (forthcoming) also abstract from modeling the funded social security system by assuming it is identical to the standard savings. Thus, the pension contribution parameter measures the relative size of the unfunded pensions. In other words, = 0 implies that the whole pension system is funded. Denoting by (. ) the utility function for consumption or and the lifetime utility, the problem of an individual of generation is: +1 +1 ++1 +−(+1 , ) max, = ( − − ) + � �, where is time discount factor and the link between pension contributions () and benefits () is defined by the particular pension system as follows: : (1 + ) = ̅ : ̅(1 + ) = : � ( − ) = (1 + ), where is the population growth rate and denotes the annuities under the defined annuity system. Production side of the economy is characterized by a Cobb-Douglas production function with constant returns to scale: 1− = ( , ) = , and where is the stock of capital, is productivity parameter, and is the labor force. Capital accumulates according to the rule: +1 = , where is the size of generation . The analysis proceeds by solving for the equilibrium saving and retirement-age decisions for individuals under different social security systems. Impacts of changing fertility and longevity are analyzed separately. Source: Dedry, Onder, and Pestieau forthcoming. es. However, these are all long-term effects which appear in behavioral adjustments, the results are mainly driven by ac- the steady state. A dynamic simulation exercise revealed that, counting effects (for example, less workers and more depen- in the short term, a transitory decrease in utility is likely when dents per worker in the economy), and are therefore necessar- aging is driven by a reduction in fertility and a DC pension ily negative. system is in place. Second, the analysis assumes underinvest- However, once behavioral adjustments are considered, ment in all cases, for example, the equilibrium is on the good this picture may change substantially. First, workers could be- side of the “golden rule,” thus, more savings are good. Interest- come more productive due to higher savings and, thus, in- ingly, and again in all cases, mandatory retirement produced creased capital accumulation. Second, individuals may choose more desirable outcomes than optimal retirement, despite to work longer, thus, the labor force could be larger than the the distortions it imposed on the economy. These outcomes levels used by demographic projections with constant ages of are mainly due to the fact that forced retirement induces retirement. workers to save more than they would save if they were freely This note highlight the importance of considering pen- choosing their own retirement age. sion system characteristics and types of aging when assess- ing the extent to which behavioral adjustments may en- Conclusion hance or reverse the negative results suggested by the This analysis shows that the net effect of demographic aging accounting effects. Results show that savings, thus capital on economies may differ across countries. In the absence of accumulation, could increase when aging is mainly driven 6 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise by decreasing fertility rates and the pension system is char- About the Authors acterized by defined contributions. This effect is weaker in Harun Onder is an Economist in the Economic Policy and Debt DB and DA systems. In comparison, when aging is driven by Unit (PRMED) of the World Bank. Pierre Pestieau is Professor of increasing longevity, the effects on savings are ambiguous in Public Economics at the Université de Liège and a member of the all types of unfunded pension systems. However, if regula- Center for Operations Research and Econometrics at the Univer- tions allow, labor force participation among the older popu- sité Catholique de Louvain. lation could increase. This is especially true in DC and DB systems. References Considering the important differences in aging styles Bloom, D. E., D. Canning, and J. Sevilla. 2001. “Economic Growth and social security systems across countries, and particularly and the Demographic Transition.” NBER Working Papers between developing and advanced economies, studies that in- 8685, Cambridge, MA. vestigate the economic impacts of aging would do well to take Bloom, D. E., D. Canning, and G. Fink. 2011. “Implications of Population Aging for Economic Growth.” PGDA Working these differences into consideration. Papers 64, Harvard University, Cambridge, MA. Dedry, A., H. Onder, and P. Pestieau. forthcoming. “Aging, Social Acknowledgments Security Design, and Capital Accumulation: CORE Discussion Papers.” Université Catholique de Louvain, Center for Opera- The authors are grateful to Jeffrey D. Lewis, Otaviano Canu- tions Research and Econometrics. to, Leonardo Hernandez, Apurva Sanghi, William G. Battaile, Onder, H., F. L. Hernandez, and F. Usmani. 2014. “Aging Grace- Stefanie Sieber, Julia Burr Oliver, and Martha Mary Schmitz fully? Macroeconomic and Fiscal Implications of Demographic Trends in the Modern World.” Mimeo. for valuable comments and suggestions. The Economist. 2014. “Global Ageing: A Billion Shades of Grey.” April 24. The Economic Premise note series is intended to summarize good practices and key policy findings on topics related to economic policy. They are produced by the Poverty Reduction and Economic Management (PREM) Network Vice-Presidency of the World Bank. The views expressed here are those of the authors and do not necessarily reflect those of the World Bank. The notes are available at: www.worldbank.org/economicpremise. 7 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise