Chaos without change: Africa’s long-run growth, 1975-2005

This section examines long-run trends in per capita income growth for 44 African economies. Data on GDP per capita at 2000 international PPP prices are taken from the World Development Indicators and span the years 1975 to 2005.2 Our sample contains all SSA countries for which PPP GDP data exist. There are no GDP per capita PPP data for Liberia, San Tome and Principe and Somalia, therefore they are excluded from the analysis.3 The unweighted mean GDP per capita between 1975 and 2005 for the 44 countries in our sample was US$2,306. Mean GDP per capita using GDP-weighted data was US$1,702.4 Table 8A.1 presents descriptive statistics for income and growth at the country level.

Figure 8.1 presents the timepaths of unweighted and GDP-weighted per capita income growth rates.5 Although the trajectories of the unweighted and weighted series appear similar, their means and variances are significantly different. The region’s unweighted mean growth rate was 0.71 per cent and its standard deviation (SD) was 6.32 per cent. The mean and SD of the weighted data are -0.17 per cent and 1.7 per cent, respectively, reflecting the fact that Africa’s bigger economies grew more slowly than its smaller ones. Between 1975 and 2005, South Africa, which represents, on average, 42 per cent of the region’s GDP, grew in per capita terms by an mean of only 0.12 per cent a year; and Nigeria, the region’s second-largest economy, (13.50 per cent of GDP) grew by 0.28 per cent.

Gross domestic product per capita growth, 1975-2005

Figure 8.1 Gross domestic product per capita growth, 1975-2005: (a) unweighted data; (b) weighted data

Both the unweighted and weighted series show a positive trend beginning in the mid-1990s. In the period 1995-2005, unweighted mean GDP growth per capita was 1.81 per cent, more than twice the long-run mean. In order to test for statistically meaningful breaks in the mid-1990s, we ran recursive residual estimations and other stability tests. Figure 8.2 shows the recursive estimation for the growth series. There is statistical evidence that growth accelerated around 1995. Both the Chow breakpoint and forecast tests support the conclusion that a structural break in the income growth series occurred in the mid-1990s.6

Growth rates for individual countries were low and the coefficient of variation was high, indicating that growth was highly erratic (Table 8A.1). Figure 8.3 shows that African economies have by far the least predictable growth globally,

Stability test

Figure 8.2 Stability test: recursive residual estimation of growth rates Source: Authors’ computations.

Gross domestic product per capita growth

Figure 8.3 Gross domestic product per capita growth: means, standard deviation (SD) and coefficient of variation (CV) by region (weighted data), 1975-2005

Source: Authors’ computations.

Table 8.2 Decomposition of standard deviation of GDP per capita and growth 1975-2005



Standard deviation


Between countries

Within countries

GDP per capita





GDP per capita growth





Source: Authors’ computations.

Notes: Statistics calculated from panel data.

as measured by the coefficient of variation (CV). Countries with different levels of income (such as South Africa and Malawi), geographical locations (such as Mali and Senegal), resource endowments (such as Nigeria and Ethiopia) and long-run GDP per capita growth patterns (like Gabon, Niger, Madagascar and Swaziland) share a common characteristic, high growth volatility.

Table 8.2 decomposes the standard deviation of GDP per capita and its growth into within- and between-country components. Growth is highly unstable in individual countries; the ratio of the within-country SD to the total SD of growth rates is 94 per cent. The Comoros (-22.6), Ethiopia (18.4), Guinea-Bissau (-11.9), Malawi (24.6), Mauritania (34.6), Namibia (19.8), Nigeria (18.7) and South Africa (20.6) are notable for their extremely high volatility, even by regional standards.

Only three economies - Botswana (0.5), Cape Verde (0.8) and Mauritius (0.4) - have coefficients of variation of less than 1.0. These three economies are also notable for their high long-run growth rates, ranking second to fourth out of the sample in terms of their overall rate of per capita income growth, 1975-2005.7

Kernel densities of the distribution of per capita GDP growth rates at 10-year intervals are shown in Figure 8.4a-d. The growth acceleration of 1995-2005 is clearly visible in the rightward shift of the distribution.

The most striking change in the distribution over time, however, is the extent to which growth rates have converged (Figure 8.4a). The 1976 distribution is remarkably flat. Since then, there have been increasingly more acute peaks around the mean (Figure 8.4b-d).

The SD of growth rates dropped from 8.2 per cent in 1976 to 3.6 per cent in 2005. Two sets of outliers - high performers and economies in decline - also appear to be emerging in the 1995 and 2005 distributions (Figure 8.4c and d).

An important question with respect to long-run growth is whether it has been persistent. Figure 8.5 shows the results of regressing mean GDP per capita growth on growth in the first year of our series, 1975-1976:

where AY is the mean growth of country i and AY-6 is the growth rate of country i in 1976, the first year in our series. Not surprisingly, given the extreme variability

Density of gross domestic product per capita growth across countries

Figure 8.4 Density of gross domestic product per capita growth across countries: (a) 1976 compared with 2005; (b) 1976 compared with 1985; (c) 1985 compared with 1995; (d) 1995 compared with 2005

Source: Author’s computations.

of growth rates, there is no evidence of growth persistence. The coefficient of в is close to zero and insignificant (в = 0.061, t = 1.65).8 Growth in 1976 for the representative African country fails to predict mean growth in the subsequent 30 years.

We also stratify the data before and after 1995 to assess whether there was evidence of persistence during either of the two subperiods:

where AY76-94 is mean GDP per capita growth between 1976 and 1994, Alf5-05 is mean GDP per capita growth between 1995 and 2005, and AT,76 and AT,95 are the growth rate of country i in 1976 and 1991.

The results are shown in Figure 8.6a and b. The coefficients of в are 0.06 (1.38) and 0.21 (3.06), respectively, for the first and second periods, suggesting that

Gross domestic product (GDP) per capita growth as a function of initial conditions

Figure 8.5 Gross domestic product (GDP) per capita growth as a function of initial conditions

Source: Authors’ computations.

Mean growth as a function of initial conditions by time period Source

Figure 8.6 Mean growth as a function of initial conditions by time period Source: Author’s computations.

growth became more predictable from the mid-1990s on, a result that is in line with the kernel density exercises.9

As an additional check for persistence at the individual country level, we calculate the correlation coefficients of growth over time for individual countries. Statistically significant coefficients indicate that country growth rates follow predictable patterns. The very large majority of correlation coefficients before 1995 are not statistically significant, but about a third of the coefficients of 1995-2005

are significant (Arbache and Page 2008). This suggests that at the individual country level, growth was generally erratic, although there was increased persistence between 1995 and 2005.10

In sum for the region as a whole, and for the vast majority of African economies, growth from 1975 to 2005 has been both disappointing and volatile. Growth on average has accelerated and has shown a weak tendency to become more persistent over time, but for the individual African country, past growth helps very little to predict future growth.

< Prev   CONTENTS   Source   Next >