Demographic Influences on Economic Resiliency: Revisiting the Developing Country Growth Collapse of the 1970’s and 1980’s. Brant Liddle btliddle@alum.mit.edu ABSTRACT This paper bridges two related, but up to now, unconnected literatures: economic growth stability and population-economic growth. The paper differs from previous population-economic growth analyses by focusing on instability of economic growth in developing countries. This study contributes to a previous paper on the developing country growth collapse by adding important demographic variables. The paper provides an explanation for “new” negative correlations of population and economic growth: because 1960s were a relatively smooth time for economic growth, youth dependency did not seem important; however, during turbulent 1970s and 1980s, countries with falling dependency burdens weathered economic shocks better. Keywords: instability of economic growth, population growth, demographic transition, developing countries Published in Journal of International Development (2011), Vol. 23, pp. 476-492.

2

1. Introduction An important purpose of this paper is to bridge two highly related, but up to now,

rather unconnected literatures: the economic growth stability literature and the population and

economic growth literature. Essentially, a left-hand-side variable from the former literature is

combined with some of the right-hand-side variables of the latter. By providing a link between

these two literatures, the paper also seeks to shed light on two emerging puzzles from these

literatures: (1) why did some countries’ growth collapse sometime during the 1970’s, and

others’ did not; and (2) why did population growth have a negligible-to-no effect on economic

growth during the 1960s and 1970s, but have a negative impact in the 1980s.

This paper follows a research strategy recommended by Pritchett (2000): to analyze

the determinants of changes in growth rates. Specifically, the paper examines how population

may impact countries’ abilities to withstand external shocks, and thus help explain the

volatility of developing countries’ growth, particularly in the last two decades. This paper is

most like Rodrik (1999), who argued that the presence of domestic social conflict and the

institutions in place to deal with those conflicts impact how countries react to external shocks.

He examined the change in per capita GDP growth over two periods 1960-1975 and 1975-

1989, and various measures of shocks, internal division, and institutions of conflict

management. He found that existence of social consensus and working conflict management

institutions lead to greater persistence of economic growth. The social conflict-management

institutions hypothesis was determined to be robust when considering alternative explanations

for the growth collapse. Indeed, explanations like “more open trade regimes avoided trouble,”

“large public sectors were worse-hit,” or “high indebtedness in the 1970s led to trouble”

added little explanatory power.

Since the Rodrik paper a nascent literature has developed on growth volatility and

democracy (e.g., Quinn and Wolley 2001; Mobarak 2005; Yang 2008), which typically finds

that democracy lowers volatility. However, Yang adds the qualification that democracy’s

3

impact on volatility is most evident in ethnically heterogenous countries. This literature

focuses on the standard deviation of the GDP per capita growth rate over certain intervals, and

because of its interest in democracy, looks at developed and developing countries; but beyond

ethnic heterogeneity, it does not consider demography. By contrast, the current paper is

concerned with the developing country growth collapse of the 1970s and 1980s, as opposed to

simply the movement of growth rates around their mean.1

The only population variable Rodrik (or the democracy and GDP volatility literature

has) used, however, is a measure of ethnolinguistic fractionalization, which comes from

Taylor and Hudson (1972). This variable is constructed from a 1960 USSR study, and

measures the probability that two randomly selected persons from a given country will not

belong to the same ethnolinguistic group. The rationale for including this variable, used by

others in growth analyses (e.g., Mauro 1995; Easterly and Levine 1997; Kenny 1999),

typically is that countries with more divided societies will either have greater internal conflict

or more difficulty implementing certain policies. But the study reported here indicates that

another population variable—youth dependency—has a significant bearing on how external

shocks impact economic growth. Yet, the population-economic growth literature has focused

on youth dependency’s impact on the level or magnitude of economic growth, not focused as

this paper is on the determinants of changes in or stability of economic growth, and the basic

finding of that literature can be summarized as: youth dependency had an insignificant impact

on economic growth in the 1960s and 1970s, but had a “negative, statistically significant, and

large” impact in the 1980s (Kelley and Schmidt 2001).

2. Background

Much of the early work on population growth and per capita income growth found

little relationship between the two. Indeed, Kelley and Schmidt (1995) claimed the most

1 Mobarak (2005), who uses as a dependent variable the standard deviation of the GDP growth rate, does recognize the significance of negative GDP growth: “… variability of growth around 0 % is more detrimental to

4

important finding of the literature up to that point “. . . is the failure, in more than a dozen

studies using cross-country data, to unearth a statistically significant association between

growth rates of population and of per capita output.” This earlier work typically examined

only aggregate population growth rates.

More recent work has either decomposed population growth into fertility and mortality

components or considered the growth rates of important sub-populations (e.g., school-aged,

working-aged, retired). These studies (Barlow 1994; Bloom and Freeman 1988; Brander and

Dowrick 1994; Kelley and Schmidt 1995; Crenshaw et al. 1997; Bloom and Williamson 1998;

Bloom et al. 2000) found a more complex role for population. Specifically, they found

(depending on their exact explanatory variables) either that growth of the working-aged

population is good for economic growth while growth of young, dependant population is not;

or that increases in fertility or births have an immediate, negative impact, although the

resulting eventual increase in the economically active population will have a delayed, positive

impact. Thus, the demographic transition—the change from high to low rates of mortality and

fertility—produces a “demographic gift” (Bloom and Williamson 1998), “demographic

windfall” (Crenshaw et al. 1997), or “window of opportunity” (Barlow 1994) through the

combined increase in the labor force and in savings that this more favorable age structure

causes. Bloom and Williamson (1998) and Bloom et al. (2000) argued that this demographic

dividend is responsible for as much as one-third of East Asia’s “economic miracle.” (The

following paragraphs will show that East Asia’s most miraculous accomplishment was

sustaining their economic growth over such a long period.) Some studies that break variables

down according to time periods in their analyses (Bloom and Freeman 1988; Brander and

Dowrick 1994; Kelley and Schmidt 1995) found a stronger negative correlation between

population growth and economic growth in the 1980s. This “new” negative correlation has not

development than variability around 4 %.”

5

been fully explained (although Kelly and Schmidt 1995, unlike the others, do offer an

explanation).

The developing country growth collapse—i.e., that growth in developing countries

was much lower in the second half of the 1970s and in the 1980s and 1990s than in the 1960s

or in the first part of the 1970s—is now well documented. Easterly (2001) calculated that the

median per capita income growth rate in developing countries fell from 2.5 percent over 1960-

79 to 0.0 percent over 1980-98. Pritchett (2000) calculated a single “break point” (or year) in

each country’s growth path from 1960 to 1992 (or to the most recent year with data). For the

87 developing countries in his sample, only 24 had growth rates of less than 1.5 percent per

year prior to their break year. However, after their break year only 26 countries had growth

rates over 1.5 percent, and 32 had negative growth rates post-break point. In a similar study,

Ben-David and Papell (1998) calculated specific year break points in growth that are

statistically significant for countries between 1950 and 1990. For the total of 37 developing

countries in their sample, 21 developing countries had negative growth after their break point.

(For the countries with negative post-break point growth, those break years ranged from 1972-

1983.)

For the sample of 94 developing countries considered here, the average per capita

GDP growth rate fell from 2.5 percent over 1960-75 to 0.7 percent during 1975-1992. Table 1

shows the regional breakdown of per capita GDP growth over the two periods. This table

demonstrates two points: (1) the 1960s were a period of widespread prosperity; and (2) the

developing country growth collapse of the 1970s and 1980s was a largely regional

phenomenon. It was essentially avoided in Asia (only Papua New Guinea had negative GDP

growth in the second period). Although only one (Cyprus) of the 11 North African, Middle

East, and Mediterranean countries had a higher rate of growth from 1975 to 1990 than from

1960 to 1975, only two countries (Iran and Iraq) had negative growth in the second period.

6

Meanwhile, nearly half of the Sub-Saharan African sample followed positive growth in the

first period with negative growth in the second, and only eight countries did better in the

second period (and for two of those, “better” meant a slower rate of contraction). In some

ways the collapse was even more dramatic in Latin America and Caribbean. Over the first

period, only Haiti had negative GDP growth; however, over the second period an additional

nine countries experienced negative growth, and only two countries (Chile and Uruguay) had

higher rates of positive growth over the second period (Haiti’s second period growth was less

negative than its first). Yet, the final two lines of the table indicate that many countries did

rebound in the 1990s from the growth collapse.

Table 1

This collapse of growth is not, however, easy to explain by changes in what are widely

regarded as determinants of growth. According to Easterly (2001), indicators like educational

enrollment, infrastructure, life expectancy, fertility, inequality, and real exchange

overvaluation showed improvement over the 1980s and 1990s, while black market premiums,

inflation, and trade openness failed to show the kind of deterioration that could explain the

growth decline. Indeed, Easterly argued that, if models of growth using these indicators and

policies were correct, developing country growth rates should have been significantly higher

during the 1980s and 1990s than in the previous decades. Furthermore, external shocks (like

wars, the oil crises, or the economic slowdown in the developed countries) can only explain

some of the variation in growth during the 1970s and 1980s (Easterly et al. 1993; Rodrik

1999). For example, many of the high-growth East Asian countries experienced external

shocks at least as strong as those encountered in Latin America. Rodrik (1999) argued that

variation in investment cannot drive the variation in growth rates over short horizons, since

investment rates are persistent over time—investment in one period is very strongly correlated

with investment in the subsequent period. Thus, it seems GDP growth, particularly in

7

developing countries, became much harder to achieve in the period beginning sometime

during the 1970s.

This study differs from previous population-economic growth analyses in several

important ways. First, rather than try to explain economic growth by assuming it is persistent

(which the previously discussed economic growth literature showed it is not), this paper

focuses on explaining the important finding of the volatility/instability of economic growth in

developing countries. Some of the recent population and economic growth literature (Mason

2005; Bloom and Canning 2008) did recognize that economic growth in the 1970s and 1980s

was considerably different in Latin America than in Asia, and Bloom et al. (2007) remarked

that economic growth slowed down over 1980-2000 in a way that is not easily explained by

country-specific factors. But this recent population and economic growth literature has not

focused per se on growth volatility.

Second, this paper suggests a new theory on how population may matter to economic

growth (the old ways being through the size of the labor force and dependents’ impact on

investment and savings). That is, it hypothesizes that countries with large working populations

relative to dependent ones may be more resilient to external shocks.

There are two possible reasons why lower dependency ratios may improve the stability

of economic growth. First, the relationship between lower dependency ratios and economic

growth stability could exist through the increased economic policy flexibility that having

fewer dependants (and, thus, fewer “sticky” programs) may afford. This argument is similar to

the basic population-economic growth argument, which states that lower dependency ratios

reduce necessary government consumption and free resources for investment. Second, high

dependency ratios (particularly growing ones) may contribute to higher levels of domestic

social conflict. Choucri and North (1975) and Easterlin (1978) have found a similar

relationship between population growth and conflict.

8

Accordingly, again by utilizing the recent findings on the instability of developing

country growth, growth rates are calculated based on theoretically meaningful periods—rather

than calculating a single rate spanning the entire available data set or using culturally

appealing break points (i.e., the decades), as all previously mentioned population-economic

growth studies did. Focusing on economic growth stability and examining growth before and

after the documented developing country growth collapse reveal an explanation of the “new”

population-economic correlation discussed above. In addition, like Bloom and Williamson

(1998), Bloom et al. (2000), and Crenshaw et al. (1997), the more theoretically appealing age-

specific growth rates are used to explain that correlation, rather than birth and death rates.

Specifically, this study uses the rate of change of the youth dependency ratio, which equals

the difference between the growth rates of the working aged population and of the young,

dependant population.

3. Methodology and Data

Following Rodrik (1999), the sample is divided into two periods, 1960-1975 and

1975-1990. (Actually, we use as the final year in the second period a year ranging from 1987

to 1992, depending on the availability of the GDP data. The specific countries used are listed

in the appendix.) There are several reasons to support this disaggregation. Pritchett (2000)

calculated a mean break year for developing countries of 1977 and a median of 1978. In

addition, Rodrik (1999) ran his regressions by dividing the sample into two periods using both

1975 and Pritchett’s individual country break years, and got very similar results.

The dependant variable is, as in Rodrik (1999), the growth differential between 1960

to 1975 and 1975 to 1992. All rates of change used in this paper are defined as the natural log

difference between the later data point and the earlier one, divided by the number of

intervening years. Rates of change are expressed in percentage terms. The GDP data is the

chain-linked index of real GDP per capita measured in 1985 purchasing power parity dollars

9

from the Penn World Tables Mark 6.1. All population data comes from the World Bank

Development Indicators 2000 (with the one exception of Taiwan, for which the data comes

from the Monthly Bulletin of Statistics of the Republic of China).

The independent variables in the base regression are: regional dummies for Latin

America and the Caribbean, Asia, and Sub-Saharan Africa; the natural log of per-capita GDP

in 1975; lagged GDP growth (growth over 1960-1975); the youth dependency ratio (the

population ages 0-14 over the population ages 15-64) in 1975; and the rate of change in the

youth dependency ratio from 1970 to 1980. The countries included in the regional dummies

conform to the standard World Bank definitions. To get a more complete, dynamic picture of

population this paper considers the change in youth dependency around the mid-point as well

as the level of youth dependency at the mid-point. Including the change in dependency from

1970 to 1980 should not create problems of endogeneity since much of the change from 1975

to 1980 has already been put into motion by changes in population from 1970 to 1975 (and

earlier), a period before the mid-point growth collapse. However, population changes that

occur over the second period may impact economic performance over that same period; yet, in

the absence of famine or economically induced war or migration, the economic performance

during the second period can cause only minimal population changes over that same period.

For example, even if economic factors led to a drop in fertility in 1977, that drop would have

only a small impact on youth dependency in 1980 since the 1977 fertility drop would have no

impact at all on the size of the 3- to 14-aged cohorts. The lagged GDP growth term is the

average per-year GDP growth from 1960 to 1975. In addition, I, like Rodrik, include

ethnolinguistic fractionalization, and a measure of government effectiveness/quality.

All the independent variables used are either constant throughout both periods or taken

from averages around or just before the mid-point, unless a substantial number of countries

would be lost from the sample because of a lack of data for them. For example, the index for

10

government effectiveness/quality used here comes from a set of aggregate measures by

Kaufmann, Kraay, and Zoido-Lobaton (1999). They compiled six different measures of

governance, each measure made up of data from various sources, and assigned countries

values ranging from –2.5 to 2.5 for each measure for which a country has some data. The six

aggregate measures are: voice and accountability; political instability and violence;

government effectiveness; regulatory burden; rule of law; and graft. For each country, the

average over their various measures (not every country has a value for each of the six

different measures) and over the two years for which they report data (1998 and 2001) is

calculated. We use their measures and take an average over the various measures so that each

country for which there is relevant population and GDP data can be included in the

regressions. Measures of government effectiveness for a period around 1975 only are difficult

to find; a popular government quality index (but containing fewer countries)—which was

used, for example, in Easterly and Levine (1997) and Easterly and Kraay (2000), from the

Institutional Investor Risk Ratings, collected over 1960-1998—correlates highly (0.7) with the

measure used here. In addition, Rodrik (1999) argued that measures like Freedom House’s

democracy index are very stable over time (the correlation coefficient across decades is 0.9).

Table 2 describes most of the variables used and lists their source.

Table 2

4. Results: a first cut

Before discussing the regression results, it is instructive to look at the data. Figure 1

indicates the first puzzle mentioned at the top of the paper, viz., many countries had very

different growth experiences in the period 1960-1975 than in 1975-1992. Furthermore, growth

in the earlier period seems to have very little relationship to growth in the later period. The

trend line in Figure 1 is closer to the x-axis than the 45-degree line, and the trend line’s R2 is

very small. Figures 2 and 3 provide some indication of the second puzzle—the changed

11

relationship between population and economic growth in more recent years. Figure 2 shows

little relationship between the average annual per capita GDP growth during 1960-1975 and

the annual rate of change in the youth dependency ratio over that same period. However,

Figure 3 shows a much stronger, negative relationship (R2 of 0.3) during 1975-1990; that is,

greater economic growth is correlated with a larger decline in youth dependency.

Figures 1-3

The result of the base regression (Regression 1) is shown in Table 3. The independent

variables have the expected signs and are, in general, highly statistically significant. Most

important, ethnolinguistic fractionalization, quality of government, and the change in youth

dependency are all significant and impact as would be expected; i.e., greater social division

leads to greater susceptibility to external economic shocks (or lower growth in the later

period), and a better quality of government and lower dependency burden lead to higher

growth. The level of youth dependency in 1975 had the expected negative sign, although it

was not significant at a standard level (its t-statistic was significant at 0.86 percent).

Table 3

Comparing the above result with the results from the two most similar of Rodrik’s

regressions (reported in Table 4, columns 5 and 6 of Rodrik 1999), the magnitude, sign, and

significance of most coefficients here are quite similar to or the same as those in Rodrik’s

regressions. One difference is that my substituting the population dependency variables for

Rodrik’s measure of terms of trade shocks (a variable not significant in his regressions) seems

to improve the explanatory power—an adjusted R2 of 0.70 versus 0.62 and 0.54 in Rodrik’s

study. It should be noted that Rodrik’s measure of government is somewhat different; the

other variables, however, are essentially the same.

Regression 2 (also in Table 3), examines whether trade impacts GDP growth stability.

That regression includes variables that measure openness to trade, the extent to which

12

countries’ trade relies on commodities, and a measure of countries’ exposure to terms of trade

shocks. Openness to trade is defined as total trade—imports plus exports—in percent of GDP

averaged over 1975-1990 (data from Global Development Finance and World Development

Indicators). To determine whether a country’s openness to trade made the country susceptible

to shocks over 1975-1990, it might seem reasonable to gather data over an earlier period, like

1970-1974 (which is what Rodrik did in his analysis of trade). We use the 1975-1990 data

because using an earlier period would mean “loosing” countries from the data set (only Iraq is

lost with the longer, more recent period), and this trade openness variable when calculated

using the two different periods (1970-1974 and 1975-1990) is highly correlated (0.94).

Another openness-to-trade variable, based on averages over 1973-1977, had 15 fewer

observations, but produced similar results (the correlation coefficient between this variable

and the trade variable used here is 0.97).

Also, this regression contains two dummy variables that measure the extent to which a

country’s trade depends on fuel (mostly oil) or on nonfuel primary products (SITC 0, 1, 2, 4,

and 68), respectively. These two variables have values of one if the export category described

above accounts for 50 percent or more of the country’s total exports in the period 1988-1992

(data from World Development Report 1995). Lastly, a terms of trade shock variable was

calculated as the growth rate of export prices minus the growth rate of import prices over the

period 1970-1974 (from Barro and Lee). A similarly defined trade shock variable using data

from the entire decade had 19 fewer observations, but led to essentially the same outcome.

It seems reasonable, a priori, that countries heavily reliant on commodities would be

susceptible to shocks (particularly of the energy induced variety). From Regression 2, it is

clear that neither openness to trade nor susceptibility to trade shocks had anything to do with

the growth collapse (Rodrik 1999, similarly found three different trade variables statistically

insignificant); however, the structure of exports (a variable not used by Rodrik) does have an

13

impact. Countries heavily reliant on primary (nonfuel) exports will be more likely to

experience a growth collapse.

5. Further investigation

5.1 Government policy and national savings and investment

The first two regressions shown in Table 4 examine whether high debt (a serious

problem in many Latin American countries) accumulated just prior to, or at the beginning of,

the shocks, or whether high levels of government consumption, contributed to the growth

collapse (Rodrik rejected the high debt hypothesis for his sample). The two new variables are

external debt as a percentage of GDP averaged over 1973-1977 (data from Global

Development Finance and World Development Indicators), and the average ratio of real

government consumption expenditure to real GDP over 1970-1974 (data from the Penn World

Tables). From Regression 3, it appears neither of these variables are important; however, the

inclusion of debt variable has caused the change in youth dependency to lose its significance.

This diminished significance results from the loss of 19 data points caused by the inclusion of

the debt variable. When Regression 3 was run on the same 75 countries but without the debt

variable, youth dependency change was still insignificant (results not shown). Regression 4

shows that, when the debt variable is removed from the regression on the full sample, change

in youth dependency is again statistically significant. In addition, there is some evidence that

countries with high shares of government consumption had a more difficult time avoiding the

growth collapse, since this consumption variable is now significant at the 90 percent level.

Table 4

One of the main reasons a drop in fertility and the consequent lowering of youth

dependency is hypothesized to lead to economic growth is the increase in national savings and

investment such population changes are supposed to induce. The second two regressions in

Table 4 attempt to control for this aspect of population change. In Regression 5 both the

14

average investment as share of GDP from 1973 to 1977 and the natural log of the average

total investment from 1973 to 1977 are included (data for both variables from the Penn World

Tables), as well as government consumption as a share of GDP. In Regression 6 the average

gross domestic savings as share of GDP from 1973 to 1977 (data from the World Bank

Development Indicators) is substituted for the investment share variable.

The results of both regressions show no indication that higher savings or investment

helped countries avoid the growth collapse, and imply that the lowering of youth

dependency’s positive impact on growth stability must come through different avenues. (Also,

the government consumption variable is no longer statistically significant.) At first, the lack of

significance for savings and investment on growth stability may be surprising; however,

Easterly and Levine (2001) have found that economic growth comes from “something else”

besides factor accumulation (namely, improvements in labor productivity and technology).

And, as mentioned earlier, investment, unlike economic growth, tends to be persistent over

time (Rodrik 1999). Finally, a number of geographic and cultural dummy variables, often used

in growth analyses, like European language, British or French legal system, landlocked,

island, or tropical location, were all statistically insignificant (results not shown).

5.2 Population aging

There is some concern that population aging in areas like East Asia may produce a

drag on economic growth (Bloom et al. 2003),2 although Bloom and Williamson (1998)

earlier found an insignificant relationship between aged dependency and economic growth.

Table 5 presents results from regressions that consider aged dependency too. Regression 7 is

the base regression with two aged dependency terms added: the aged dependency ratio

(population 65 years and over divided by the population ages 15-64) in 1975, and the rate of

2 Indeed, much of the recent population and economic growth literature focuses on aging’s impact on the economy; see for example some of the papers found in Prskawetz, A., Bloom, D.E., and Lutz, W. (Eds.) (2008). Population Aging, Human Capital Accumulation, and Productivity Growth. Supplement to Population and Development Review and at http://www.hsph.harvard.edu/pgda/seminars/workshops.htm.

15

change in the aged dependency ratio from 1970 to 1980. The change in aged dependency is

insignificant, but the level of aged dependency in 1975 is statistically significant and positive;

whereas, the level of youth dependency in 1975 is now highly insignificant (a considerably

lower t-statistic than in previous regressions). Regression 8 shows the reduced form

relationships with the insignificant variables (level of youth dependency and aged dependency

change) removed.

It is not clear why the level of aged dependency would be important to economic

growth volatility. None of these developing countries have very advanced age structures (the

highest aged dependency ratio in 1975 is Barbados’ at 0.18, compared to Japan’s 2005 ratio of

0.32); thus, the dependency ratio may be a proxy for having an older workforce, which could

better prepare countries to weather economic shocks. A number of studies (see Skirbekk 2003

for a survey) have found that individual productivity has an inverted-U shaped relationship

with age, the peak productivity occurring during ages 35-44 (more on human capital in the

following sub-section). This explanation could be elucidated by adding to the model a detailed

break-down of working age cohorts. However, Bloom et al. (2004) found that adding detailed

working-age structure had little impact on explaining economic growth rates.

Another explanation of a role for aged dependency involves what Mason and Lee

(2008) called the “second demographic dividend.” The first demographic dividend arises

when working-age people have fewer children; however, when those workers retire, there will

be fewer working-age people to support them. Yet, under appropriate policies, according to

the second demographic dividend, that older demographic structure can “… both raise capital

per worker other things equal, and additionally create a powerful incentive for individuals to

accumulate assets to provide for old age” (Mason and Lee 2008). To test the importance of

aging and investment the total investment term (insignificant in Regressions 5 and 6) is

16

reconsidered. Now total investment is marginally significant (Regression 9), and an

interaction term of aged dependency and total investment is highly significant and positive

(Regression 10). None of the other investment/savings variables (from the regressions shown

in Table 4) were significant—either when added separately with aged dependency or as an

interaction term with aged dependency (results not shown).

Table 5

5.3 Human capital influences

Lastly, we control for human capital in examining population change’s impact on

economic growth stability. First, a number of variables (all taken from Barro and Lee 1993)

are added that measure both school enrollment ratios and years of schooling at or near the

mid-point (of 1975). However, none of these human capital variables are statistically

significant (results not shown). Second, we explore how prior growth in the youth population

might be good for economic growth by considering youth population’s interaction with

human capital. It seems reasonable that if countries with growing young populations invest in

education, they would reap benefits later on. To test this theory, a number of interaction terms

consisting of combinations of 1975 primary and secondary school enrollment ratios and

earlier periods of young (ages 0-14) population change (1965-1975 and 1970-1980) are

created and applied.

None of these interaction terms entered the regressions as statistically significant—

indeed, they all had t-statistics of one or less (results not shown). One explanation for this

(perhaps) surprising result is none of my interaction terms fully represents the benefit of a

growing, more educated population (perhaps, because the dates used to calculate the terms do

not accurately capture the timing of these human capital effects). Another possibility is that,

as with the level of aged dependency discussed above, a further disaggregation of age

structure is needed to fully capture the effects of changes in human capital and aging. Indeed,

17

Crespo Cuaresma and Lutz (2007) found that the growth in the human capital of younger

adults (aged 20-39) added more to the explanatory power of typical economic growth models

than the growth in the human capital of older adults. Still another possibility is that, although

increasing human capital is certainly important for achieving economic growth, it may have

little impact on economic growth stability. A last explanation is data quality, either because a

number of countries were dropped from the regressions because of lack of school enrollment

data, or because the simple enrollment or average years of education data do not completely

measure human capital’s full contribution to economic growth.

6. Conclusions

The paper makes a contribution by combining findings in two related, but until now,

rather unconnected literatures. Specifically, the paper uses demographic variables to help

explain the instability of economic growth in developing countries, and in doing so, advances

a new hypothesis on how population may impact economic growth.

Pritchett (2000) demonstrates the importance of considering economic growth

instability—particularly for developing countries. Rodrik (1999) shows that growth instability

(or lack of it in the Asian case) is the defining characteristic of the Asian “miracle.” Bloom

and Williamson (1998) and Bloom et al. (2000) illustrate that the demographic transition

(from high to low rates of mortality/fertility) was an important part of the performance of the

high-growth Asian countries. This paper pulls those findings together by demonstrating that

the demographic transition (or lack of it) is valuable in explaining the stability of economic

growth (or lack of it) in developing countries, even after accounting for institutional factors

and other measures of social division proved relevant by Rodrik (1999). Furthermore, this

relationship between growth stability and demographic change holds true for countries other

than the high-growth Asian ones.

18

The importance in these regressions of both government quality and youth dependency

supports Bloom et al.’s (2003) contention that good polices and institutions are a precondition

for enjoying the “demographic dividend.” Also, combining (i) the finding (discussed earlier in

the paper) that economic growth (for whatever reasons) became considerably more difficult

beginning in the latter half of the 1970s with (ii) this paper’s theory of population’s impact on

weathering economic shocks can help explain what Kelly and Schmidt (1995) call “the new

negative correlations.” In other words, because the 1960s were a relatively trouble-free time

for economic growth, population dependency’s full importance was not apparent; however,

during the more turbulent 1970s and 1980s countries with lower (and falling) dependency

burdens weathered the economic shocks better.

It appears rates of change in, rather than levels of, youth dependency is most

important. There was, however, a positive association between the level of aged dependency

and growth stability; such an association is explained perhaps because an interaction between

aging and investment exists (i.e., the second demographic dividend), or perhaps because

productivity increases with age until middle-age. Yet, interaction terms attempting to capture

the relationship among human capital (or school enrollment rates), growth in the youth

population, and economic growth were all statistically insignificant. Although important for

economic growth, high human capital may not provide much insurance against economic

shocks and growth collapse. Alternatively, as with achieving a greater understanding of the

level of aged dependency’s impact, further disaggregation of working-age cohorts may prove

fruitful in understanding more fully human capital’s impact on growth stability.

19

References

Barlow, R. (1994). Population Growth and Economic Growth: Some More Correlations. Population and Development Review, 20(1), 153-165. Barro, R. and Lee, J. (1993). International Comparisons of Educational Attainment. NBER Working Paper no. 4349. Ben-David, D. and Papell, D.H. (1998). Slowdowns and Meltdowns: Postwar Growth Evidence from Seventy-four Countries. The Review of Economics and Statistics, 80 (4), 561-571. Bloom, D.E. and Freeman, R.B. (1988). Economic development and the timing of components of population growth. Journal of Policy Modeling, 10(1), 57-81. Bloom, D. E. and Williamson, J.G. (1998). Demographic Transitions and Economic Miracles in Emerging Asia. The World Bank Economic Review, 12 (3), 419-455. Bloom, D.E., Canning, D. and Malaney, P.N. (2000). Population Dynamics and Economic Growth in Asia. Population and Development Review, 26, Supplement: 257-290. Bloom, D.E., Canning, D. and Sevilla, J. (2003). The Demographic Dividend: A New Perspective on the Economic Consequences of Population Change. Santa Monica, C.A.: RAND. Bloom, D.E., Canning, D. and Sevilla, J. (2004). The effect of health on economic growth: a production function approach. World Development, 32 (1), 1-13. Bloom, D.E., Canning, D., Fink, G. and Finlay, J.E. (2007). Does age structure forecast economic growth? International Journal of Forecasting. 23, 569-585. Bloom, D.E. and Canning, D. (2008). Global demographic change: Dimensions and economic significance. Population and Development Review, 34 (supplement), 17-51. Brander, J.A. and Dowrick, S. (1994). The role of fertility and population in economic growth: empirical results from aggregate cross-national data. Journal of Population Economics, 7(1), 1-25. Choucri, N. and North, R.C. (1975). Nations in Conflict: National Growth and International Violence. San Francisco: Freeman. Crenshaw EM, Ameen AZ, Christenson M. (1997). Population dynamics and economic development: Age-specific population growth rates and economic growth in developing countries, 1965 to 1990. American Sociological Review, 62 (6), 974-984. Crespo Cuaresma, J. and Lutz, W. (2007). Human capital, age structure and economic growth: Evidence from a new dataset. Interim Report IR-07-011. Laxenburg, Austria: International Institute for Applied Systems Analysis.

20

Easterlin, R. (1978). What Will 1984 Be Like? Socioeconomic Implications of Recent Twists in Age Structure. Demography, 15 (4), 397-421. Easterly, W., Kremer, M., Pritchett, L, and Summers, L.H. (1993). Good Policy or Good Luck: Country Growth Performance and Temporary Shocks. Journal of Monetary Economics 32, 459-483. Easterly, W. and Levine, R. (1997). Africa’s Growth Tragedy: Policies and Ethnic Divisions. The Quarterly Journal of Economics, 112 (4), 1203-1250. Easterly, W. and Levine, R. (2001). It’s Not Factor Accumulation: Stylized Facts and Growth Models. The World Bank Economic Review, 15 (2), 177-219. Easterly, W. and Kraay, A. (2000). Small States, Small Problems? World Development, 28(11), 2013-2027. Easterly, W. (2001). The Lost Decades: Developing Countries’ Stagnation in Spite of Policy Reform 1980-1998. Journal of Economic Growth, 6, 135-157. Kaufmann, D., Kraay, A. and Zoido-Lobaton, P. (1999). Governance Matters. World Bank Policy Research Department Working Paper No. 2196. Kelley, A. C. and Schmidt, R.M. (1995). Aggregate Population and Economic Growth Correlations: The Role of the Components of Demographic Change. Demography, 32, 4, 543-555. Kelley, A. C. and Schmidt, R.M. (2001). Economic and Demographic Change: A Synthesis of Models, Findings, and Perspectives. In Birdsall, N., Kelley, A.C., and Sinding, S.W. (Eds.), Population Matters: Demographic Change, Economic Growth, and Poverty in the Developing World. (pp. 67-105). Oxford: Oxford University Press. Kenny, C. (1999).Why aren't countries rich? Journal of Development Studies, 35(5), 26-47. Mason, A. (2005). Demographic Transition and Demographic Dividends in Developed and Developing Countries. Population Division, Department of Economic and Social Affairs, United Nations. UN/POP/PD/2005/2. Mason, A. and Lee, R. (2008). Reform and support systems for the elderly in developing countries: capturing the second demographic dividend. Genus, LXII (2), 11-35. Mauro, P. (1995). Corruption and Growth. The Quarterly Journal of Economics, 110, 3, 681-712. Mobarak, A. (2005). Democracy, volatility, and economic development. The Review of Economics and Statistics, 87(2), 348-361. Pritchett, L. (2000). Understanding Patterns of Economic Growth: Searching for Hills among Plateaus, Mountains, and Plains. The World Bank Economic Review, 14, 2, 221-250.

21

Quinn, D. and Woolley, J. (2001). Democracy and national economic performance: The preference for stability. American Journal of Political Science, 45 (3), 634-657. Rodrik, D. (1999). Where Did All the Growth Go? External Shocks, Social Conflict, and Growth Collapses. Journal of Economic Growth, 4, 385-412. Skirbekk,V. (2003). Age and individual productivity: A literature survey. MPIDR Working Paper 2003-028. Max Planck Institute for Demographic Research, Rostock. Germany. Talyor, C. L. and Hudson, M.C. (1972). World Handbook of Political and Social Indicators. 2d ed. New Haven: Yale University Press. Yang, B. (2008). Does democracy lower growth volatility? A dynamic panel analysis. Journal of Macroeconomics, 30, 562-574.

22

Table 1. Regional Breakdown of Developing Country GDP Growth Over Different Periods All countries (94) SSA (41) LAC (25) Asia (17) Positive Growth 60-75 85 33 24 17 Positive Growth 75-90 56 16 15 16 Positive Growth 60-75 & Negative Growth 75-90

31 19 9 1

Positive Growth in both 60-75 & 75-90 53 13 15 16 Negative Growth in both 60-75 & 75-90 6 5 1 0 Growth 75-90 > Growth 60-75 25 8 3 13 Growth 75-2000 > Growth 75-90 79 34 22 13 Positive Growth 75-2000 & Negative Growth 75-90

31 20 9 1

Table 2. Descriptions and Sources of Independent Variables Description Source Obs. Mean Std. Dev. Natural log of GDP per capita in 1985 PPP$

Penn World Tables Mark 5.6 & 6.1

94 7.37 0.79

Average rate of GDP per capita growth in 1985 PPP$ from 1960 to 1975

Ibid 94 2.42 2.11

Average investment as share of GDP from 1973 to 1977

Ibid 94 0.15 0.088

Average ratio of real government "consumption" expenditure to real GDP from 1970 to 1974.

Ibid 94 0.18 0.077

Natural log of the average total investment from 1973 to 1977.

Ibid 94 20.87 1.97

Youth dependency ratio (population ages 0-14 over population 15-64) in 1975

World Bank Development Indicators

94 0.81 0.14

Average rate of change in youth dependency ratio from 1970 to 1980

Ibid 94 -0.73 1.43

Aged dependency ratio (population aged 65 and older over population 15-64) in 1975

Ibid 94 0.070 0.025

Average rate of change in aged dependency ratio from 1970 to 1980

Ibid 94 0.21 1.41

Average gross domestic savings as share of GDP from 1973 to 1977

Ibid 83 0.17 0.15

Probability two randomly selected persons do not belong to same ethnolinguistic group

Talyor and Hudson 94 0.44 0.28

Index of government quality/ effectiveness averaged over 6 different measures taken at 1998 and 2001

Kaufmann et al. 94 -0.23 0.66

Average of imports plus exports as a share of GDP from 1975-1990

Global Development Finance and World Development Indicators

93 0.044 0.020

Average external debt as share of GDP from 1973 to 1977

Ibid 75 0.28 0.18

Average growth rate of export prices minus growth rate of import prices from 1970 to 1974.

Barro and Lee 94 0.027 0.10

Indicator (0, 1) for the export of nonfuel primary products

World Development Report

94 0.38 0.49

Indicator (0,1) for export of fuels (mainly oil)

Ibid 94 0.096 0.30

23

Table 3. Initial growth regressions, including effect of trade. Dependent Variable: Per Capita GDP Growth 1975-1992 Minus Per Capita GDP Growth 1960-1975 Variable 1 2 Constant 14.74*** 14.92*** (5.32) (4.59) Dummy for sub-Saharan Africa -1.11** -0.92 (-2.05) (-1.19) Dummy for Latin America & Caribbean -1.56*** -1.40** (-2.82) (-2.37) Dummy for Asia 1.11* 1.06 (1.82) (1.54) Log of GDP per capita, 1975 -1.51*** -1.52*** (-4.69) (-4.18) GDP per capita growth over 1960-1975 -0.86*** -0.81*** (-9.68) (-7.41) Ethnic fractionalization -2.47*** -2.38** (-3.22) (-2.57) Average government effectiveness measures, 1998-2001 1.29*** 1.08*** (3.95) (2.63) Youth dependency ratio, 1975 -1.89 -1.89 (-1.49) (-1.26) Change in youth dependency ratio over 1970-1980 -0.49** -0.53*** (-2.51) (-2.66) Trade openness -0.0031 (-0.70) Terms of trade shock, 1970-1974 -1.05 (-0.39) Indicator for oil exporter 0.64 (0.83) Indicator for primary products exporter -0.84* (-1.79) Observations 94 85 Adjusted R2 0.70 0.67 Notes: White heteroskedasticity consistent t-statistics are reported in parentheses. Levels of statistical significance indicated by asterisks: *** 1 %, ** 5 %, * 10 %.

24

Table 4. Regressions considering government consumption, external debt, national savings, and investment. Dependent Variable: Per Capita GDP Growth 1975-1992 Minus Per Capita GDP Growth 1960-1975 Variable 3 4 5 6 Constant 13.58*** 17.43*** 16.56*** 16.32*** (3.62) (5.92) (5.03) (4.04) Dummy for sub-Saharan Africa -0.43 -0.86 -0.76 -0.92 (-0.59) (-1.39) (-1.16) (-1.35) Dummy for Latin America & Caribbean -1.05* -1.37** -1.34** -1.55** (-1.76) (-2.37) (-2.29) (-2.52) Dummy for Asia 1.76*** 1.21** 1.13* 0.85 (2.71) (2.07) (1.86) (1.43) Log of GDP per capita, 1975 -1.35 *** -1.73*** -1.78*** -1.63*** (-3.33) (-5.67) (-5.47) (-3.94) GDP per capita growth over 1960-1975 -0.89*** -0.88*** -0.89*** -0.92*** (-7.47) (-8.85) (-8.49) (-8.39) Ethnic fractionalization -2.47*** -2.30*** -2.36*** -2.36*** (-3.02) (-3.17) (-3.11) (-2.94) Average government effectiveness measures, 1998-2001

1.11** 1.32*** 1.33*** 1.07**

(2.57) (4.02) (3.98) (2.42) Youth dependency ratio, 1975 -0.71 -1.78 -1.88 -2.35 (-0.45) (-1.40) (-1.41) (-1.63) Change in youth dependency ratio over 1970-1980

-0.28 -0.39** -0.41** -0.40*

(-1.26) (-2.03) (-2.08) (-1.88) Indicator for primary products exporter -0.95** -0.94** -0.92** -0.96** (-1.96) (-2.33) (-2.25) (-2.12) Government consumption as share of GDP, averaged 1970-1974

-4.80 -5.49* -5.14 -4.63

(-1.46) (-1.73) (-1.59) (-1.23) External debt as share of GDP, averaged 1973-1977

-1.01

(-0.80) Log total investment, averaged 1973-1977 0.06 0.037 (0.62) (0.34) Investment as share of GDP, averaged 1973-1977

0.03

(0.02) Gross domestic savings as share of GDP, averaged 1973-1977

0.73

(0.61) Observations 75 94 94 83 Adjusted R2 0.72 0.74 0.73 0.71 Notes: White heteroskedasticity consistent t-statistics are reported in parentheses. Levels of statistical significance indicated by asterisks: *** 1 %, ** 5 %, * 10 %.

25

Table 5. Regressions considering aged dependency and investment. Dependent Variable: Per Capita GDP Growth 1975-1992 Minus Per Capita GDP Growth 1960-1975 Variable 7 8 9 10 Constant 10.82*** 12.01*** 9.30*** 12.37*** (3.45) (6.34) (3.87) (6.49) Dummy for sub-Saharan Africa -0.22 -0.36 -0.05 -0.20 (-0.35) (-0.63) (-0.08) (-0.33) Dummy for Latin America & Caribbean -1.06** -1.08** -1.08** -1.08** (-2.09) (-2.20) (-2.20) (-2.24) Dummy for Asia 1.90*** 1.75*** 1.60*** 1.69*** (3.04) (3.36) (3.24) (3.30) Log of GDP per capita, 1975 -1.55*** -1.60*** -1.74*** -1.69*** (-5.60) (-5.80) (-6.07) (-5.97) GDP per capita growth over 1960-1975 -0.85*** -0.84*** -0.86*** -0.84*** (-9.26) (-8.77) (-9.09) (-8.94) Ethnic fractionalization -2.33*** -2.30*** -2.43*** -2.33*** (-3.18) (-3.11) (-3.17) (-2.33) Average government effectiveness measures, 1998-2001

1.20*** 1.16*** 1.21*** 1.15***

(4.06) (3.87) (3.94) (3.89) Youth dependency ratio, 1975 0.72 (0.40) Change in youth dependency ratio over 1970-1980 -0.47** -0.46** -0.49*** -0.49*** (-2.46) (-2.48) (-2.64) (-2.58) Aged dependency ratio, 1975 24.01** 21.37*** 23.65*** (2.42) (3.16) (3.41) Change in aged dependency ratio over 1970-1980 -0.04 (-0.31) Indicator for primary products exporter -0.88** -0.87** -0.79** -0.84** (-2.24) (-2.24) (-2.14) (-2.22) Log total investment, averaged 1973-1977 0.17* (1.77) Aged dependency ratio (1975) x log total investment (average 1973-1977)

1.18***

(3.38) Observations 94 94 94 94 Adjusted R2 0.74 0.74 0.75 0.75 Notes: White heteroskedasticity consistent t-statistics are reported in parentheses. Levels of statistical significance indicated by asterisks: *** 1 %, ** 5 %, * 10 %.

26

Figure 1: The relationship between average, yearly per-capita GDP growth over the periods 1960-1975 and 1975-1992 for a sample of 94 developing countries. The trend-line, trend-line equation, and its R2 are shown.

y = 0.319x - 0.1725R2 = 0.0688-6

-4

-2

0

2

4

6

8

-4 -2 0 2 4 6 8

1960-1975

1975

-199

2

27

Figure 2: The relationship between average, yearly per-capita GDP growth over 1960-1975 and average, yearly change in the youth dependency ratio (the population ages 0-14 over population 15-64) over the same period for the same 94 developing countries shown in Figure 1. The trend-line, trend-line equation, and its R2 are shown.

Figure 3: The relationship between average, yearly per-capita GDP growth over 1975-1992 and average, yearly change in the youth dependency ratio over 1975-1990 for the same 94 developing countries shown in Figures 1 and 2. The trend-line, trend-line equation, and its R2 are shown.

y = -0.2628x - 0.6042R2 = 0.299

-3

-2

-1

0

1

2

-5 -3 -1 1 3 5 7

GDP growth

chan

ge in

you

th d

epen

denc

y

y = -0.1481x + 0.3942R2 = 0.1298

-3

-2

-1

0

1

2

-5 -3 -1 1 3 5 7

GDP growth

chan

ge in

you

th d

epen

denc

y

28

Appendix Countries used in study ALGERIA GUINEA NIGERIA ANGOLA GUINEA-BISSAU PAKISTAN ARGENTINA GUYANA PANAMA BANGLADESH HAITI PAPUA NEW GUINEA BARBADOS HONDURAS PARAGUAY BENIN HONG KONG PERU BOLIVIA INDIA PHILIPPINES BOTSWANA INDONESIA PUERTO RICO BRAZIL IRAN RWANDA BURKINA FASO IRAQ SENEGAL BURUNDI JAMAICA SIERRA LEONE CAMEROON JORDAN SINGAPORE CAPE VERDE KENYA SOMALIA CENTRAL AFRICAN REPUBLIC KOREA, REPUBLIC OF SOUTH AFRICA CHAD LESOTHO SRI LANKA CHILE LIBERIA SURINAME CHINA MADAGASCAR SWAZILAND COLOMBIA MALAWI SYRIAN ARAB REPUBLIC CONGO MALAYSIA TAIWAN, CHINA COSTA RICA MALI TANZANIA COTE D'IVOIRE MALTA THAILAND CYPRUS MAURITANIA TOGO DOMINICAN REPUBLIC MAURITIUS TRINIDAD AND TOBAGO ECUADOR MEXICO TUNISIA EGYPT MOROCCO TURKEY EL SALVADOR MOZAMBIQUE UGANDA ETHIOPIA MYANMAR URUGUAY FIJI NAMIBIA VENEZUELA GABON NEPAL ZAIRE (Congo, Dem. Rep.) GAMBIA, THE NICARAGUA ZAMBIA GHANA NIGER ZIMBABWE GUATEMALA