In global growth regressions, the initial health and – less robustly – edu- cation status of the population are often strong predictors of subsequent growth. In table1.1we showed that, over the entire 1960–2000 period, the gap between African life expectancies and literacy rates and those in other developing regions did not widen as rapidly as did the gap in incomes.
Figure 1.7provides an alternative perspective on the same phenomenon.
We plot the estimated coefficients ˆγt,γˆt+1, . . . ,γˆ2000, along with their 2-standard-deviation bounds, from OLS regressions of the form
dit =α+βlnyit +
2000
j=t
γj(SSAi·yeart),
wheredis the life expectancy rate at birth, or the adult illiteracy rate, for countryiin yeart, andy(as before) is PPP-adjusted real incomeper capita.
The estimated coefficients provide a measure of year-by-year gaps in aver- age human development between countries in SSA and other developing regions, controlling for the gap readily attributable to differences in real
22Bloom and Williamson (1998) emphasize the “demographic dividend” that accrued to East Asian countries over this period as a result of a falling ratio of dependents to workers.
−20−15−10−50
1960 70 80 90 2000
Year
Life expectancy
−100102030
1970 80 90 2000
Year
Adult illiteracy rate
Figure 1.7 Life expectancy and adult illiteracy, SSA/year interactions and 2SD bounds, 1960–2000
Note: Dotted lines show+/−2 standard errors around the estimated coefficients.
GDPper capita. At the time of political independence for much of SSA, life expectancy rates were fifteen years below those of countries with similar incomes globally, and adult illiteracy rates were 25 percentage points higher.
By the early 1990s, African illiteracy rates were statistically indistinguish- able from those of similar-income countries. Life expectancy rates had also converged steadily to global income-adjusted norms for much of the period but, in contrast with literacy rates, they show a marked slowdown starting in the mid-1980s and a striking reversal with the onset of HIV/AIDS in the 1990s.
While figure1.7documents a reasonably rapid convergence of African human development measures to income-adjusted global norms, the adjust- ment for income is crucial. We have already seen that, on average, African incomes stagnated over the period, diverging sharply from incomes in other regions. Figure1.7simply confirms that human development measures did better: they continued to advance at a slow pace, in the face of very limited improvements in overall living standards.23
23Easterly (1999) argues that in global samples, HDIs display a broad tendency to improve that is not tightly tied to differences in national growth performance.
The growth impact of education has been the subject of a celebrated conundrum in growth econometrics. Microeconomic evidence suggests that the private returns to education are substantial, and growth theory routinely imputes a social return to human capital investment that is at least as great as the private return. But growth researchers have had an extraordinarily diffi- cult time finding statistically significant and economically plausible impacts of educational variables in global growth regressions (Pritchett2001). In our own framework paper (O’Connell and Ndulu2001), measures of educa- tional attainment and enrollment performed very poorly in conditional OLS regressions incorporating demographic measures and life expectancy rates, and their limited availability dramatically reduced the size of the African sub-sample.
More recent research has begun to reconcile the microeconomic and growth evidence via better measurement of educational attainment and greater care in the treatment of collinearity and endogeneity. Using an improved dataset on educational attainment, Cohen and Soto (2001) and Soto (2002) uncover statistically significant impacts of human capital invest- ment on growth that are in the range of 7 to 10 percent.24Others have found significant impacts once thresholds are passed; Barro (1999), for example, finds that school attainment at the secondary and higher levels for males aged twenty-five and over has a positive effect on the subsequent rate of economic growth. The estimated impact for this category is such that an additional year of schooling raises the growth rate impact by 0.7 percent per year, a very large effect indeed for slow growers. This impact is mediated predominantly via improved capabilities to absorb technological advances.
Barro’s results are consistent with those of Borensztein, De Gregorio, and Lee (1998), who find that the productivity advantages latent in for- eign direct investment (FDI) are subject to human-capital threshold effects.
These authors find that the growth contribution of FDI exceeds that of domestic investment only when the host country’s average secondary-school attainment exceeds 0.52 years (for the male population of working age).
This level is far above that of the majority of African countries. Consistent with these results, the vast bulk of FDI into Africa flows into the mineral and energy sectors. Within Africa, Lumbila (2005) finds a similar threshold effect using secondary enrollment rather than attainment rates: returns to FDI are significantly higher in countries with secondary enrollment rates exceeding 25 percent.
24There is some evidence that human-capital accumulation induces accumulation of additional physical capital. Taking this indirect effect into account, these authors estimate the total long-run impact of an additional average year of education on incomeper capita at between 12 and 16 percent.