Impact of apartheid on economic growth: implications and empirical evidence from South Africa.
In 1948, the National Party won the national elections in South Africa and implemented legislation enforcing a policy of white domination and racial separation known as 'apartheid' (separateness). (1) After decades of protests by the African National Congress followed by a widespread international outcry against apartheid, the first non-racial elections in 1994 marked the end of apartheid.
As noted by Moll (1991), several analysts such as Steyn (1970); Houghton (1978); Gervasi (1970); and Innes (1984) have claimed that the apartheid economy of South Africa was a success story in terms of economic growth until the 1970s. According to Houghton (1978), 'The South African economy with that of Japan probably had the highest growth rate in the world at that time'. (2) Innes (1984) describes South Africa's growth during the 1960s as 'exceptionally high by international standards'. (3) However, Moll points out that these claims are 'theoretically weak and empirically dubious' (1991, p. 273). He notes that such assertions were used to support the claim that apartheid was a 'unique system of social control designed to boost economic growth and industrialization in South Africa, thereby facilitating what has been termed a post-1948 "apartheid boom"'. (4)
The constraints on economic growth in South Africa, imposed by the system of apartheid, have been discussed by several authors, including Lipton (1985); Moll (1991); Jones and Muller (1992); Lundahl (1992); Lowenberg (1997); Lowenberg and Kaempfer (1998); and Liu and Saal (2001). Moll (1991) compares South Africa's GDP growth rate with those of 20 middle-income developing countries and finds that South Africa's growth record was consistently mediocre within this group of countries. Some attempts have been made to empirically quantify the economic impacts of apartheid. Thus, for example, Iyengar and Porter (1988), using a multisector model of a small open economy based on 1980 data, find that the removal of all apartheid restrictions would have increased real GDP by 6%-9%, depending on how fast blacks were assumed to be able to move into skilled jobs. Becker and Pollard (1990) show that the inefficient apartheid system, particularly the import substitution strategy, condemned South Africa's productivity and growth performance during the apartheid era to a lower trajectory than it would otherwise have been able to attain. (5) However, there is no study that examines, retrospectively, the effects of apartheid policies in the framework of a growth model. Our purpose is to fill this gap in the literature.
The section 'The Empirical Framework' presents the theoretical framework in which apartheid affects growth. The following section, 'The Data', describes our panel data set from 1961 to 2000 for South Africa and our choice of peer country groups. 'Empirical Results' presents the results of our panel estimation using system generalized method of moments (GMM) and explains the empirical strategy employed in comparing South Africa's actual performance with its predicted performance based on the panel regression coefficients of two sets of peer country groups. The fifth section, that is, 'Comparison of South Africa across Countries', examines the effect of apartheid policies on South Africa's economy and compares key macroeconomic variables across South Africa and similar countries. The last section concludes with policy implications for countries adopting economic policies similar to those of the apartheid regime in South Africa.
THE EMPIRICAL FRAMEWORK
In the basic neoclassical framework, apartheid and apartheid-like statist institutions can affect real GDP growth per capita by lowering the rate of investment in human and physical capital. We incorporate this consequence of apartheid into the basic neoclassical growth model using the framework of conditional convergence in Barro (1991, 2000).
Dy = F(y, y*) (1)
where Dy is the growth of per capita output, y is the initial level of per capita output, and y* is the long-run level of per capita output. In the neoclassical model, the economy's growth rate, Dy, varies inversely with its level of development, y, for a given value of y*.
We hypothesize that in South Africa during apartheid, and in apartheid-like statist regimes, there is discrimination in labor markets that leads to a reduction of profitability to investors and lower rates of return on investment. This makes savers and investors reluctant to undertake long-term projects and thereby lowers the rate of investment in physical capital. Apartheid education policies lead to low rates of investment in human capital of black workers. Consequently, the economy falls to a lower level of physical and human capital in equilibrium and hence to a lower real income per capita in the long-run equilibrium, y*. The transition to the steady state may take several decades, so a lower level of equilibrium GDP per capita relative to the current value will imply slower growth for a long time.
We employ a panel setup of the growth model based on the above framework, which has the conventional form:
ln [y.sub.it] - ln [y.sub.it-1] = 0 ln [y.sub.it-1] + [k.summation over (j=1)] [[PHI].sub.j] ln [X.sup.j.sub.it] + [[eta].sub.i] + [v.sub.it] (2)
where the dependent variable, In [y.sub.it]--ln [y.sub.it-1], is the growth rate of per capita output and In [y.sub.-1] corresponds to the initial level of per capita output. Since the dynamic convergence toward the steady state is valid for shorter-time intervals as well, the panel data setup allows us, 'after controlling for the individual country effects, to integrate this process of convergence occurring over several consecutive time intervals', as pointed out by Islam (1995, p. 1337).
The explanatory variables in the vector X are Inv, real gross domestic investment as a share of real GDP; Gov, real government consumption expenditure as a share of real GDP; Inflation, a measure of CPI inflation; Education, average years of schooling in the population over 15 years of age; Openness, defined as exports plus imports divided by real GDP; and Initial, the initial level of real GDP per capita for each of the time periods. (6) The unobservable individual country effects are accounted for through [[eta].sub.i]. The term [v.sub.it], the transitory error term that varies across countries and time periods, has a mean of zero and is serially uncorrelated. The notation i = 1 to n denotes the country; t = 2, ..., T denotes the time period. For ease of estimation, we write equation 2 above equivalently as: (7)
ln [y.sub.it] = [zeta] ln [y.sub.it-1] + [k.summation over (j=1)] [[PHI].sub.j] ln [X.sup.j.sub.it] + [[eta].sub.i] + [v.sub.it] (3)
where [theta] in equation 2 is equal to [zeta]--1.
We estimate equation 3 using two sets of peer countries. First, we consider a panel of a broad set of 29 countries conceptually similar to South Africa. Twenty of these countries were considered by Moll (1991) to be approximately comparable to post-war South Africa in terms of their populations and their low initial productivity levels, but they differed in terms of availability of natural and human resources and economic structures. These developing countries considered by Moll, which were roughly comparable to post-war South Africa, ranked in descending order of 1960 real GDP per capita in international prices, are Austria, Venezuela, Italy, Argentina, Spain, Japan, Chile, South Africa, Greece, Mexico, Portugal, Peru, Sri Lanka, Colombia, Turkey, Malaysia, Ghana, Algeria, Brazil, and Nigeria. Because South Africa is currently classified by the World Bank as an upper middle-income country, we expand the list of countries considered by Moll to include such upper middle-income countries as Botswana, Mauritius, and Uruguay. As some of the countries included by Moll are currently classified as high-income or low middle-income, for the sake of consistency we expand the list to include the high-income countries, such as Korea, Singapore, and Trinidad and Tobago, and the low middle-income countries, Philippines, El Salvador, Thailand, and Tunisia. The countries that were excluded from the sample of upper middle-income countries in the World Bank list were either too small to be compared meaningfully with South Africa, did not have sufficient data, or were former communist countries. (8) Our sample of countries, listed in descending order of 1950 real GDP per capita, is Argentina (6,941.88), Mauritius (5,724.33), Uruguay (5,514.58), Austria (4,853.76), Venezuela (4,809.32), Italy (4217.14), Chile (4398.89), South Africa (4,138.02), Trinidad and Tobago (3,420.79), Costa Rica (3,179.48), Spain (2,928.18), Greece (2,880.18), Mexico (2,709.32), E1 Salvador (2,613.82), Colombia (2,484.28), Peru (2,456.64), Portugal (2,367.15), Japan (2,187.63), Brazil (1,801.99), Turkey (1520.32), Philippines (1,377.12), Thailand (1,109), and Sri Lanka (864.42). Also included, listed in descending order of 1960 real GDP per capita, are Singapore (4,219.14), Algeria (3,843.10), Tunisia (2,102.64), Malaysia (1,829.30), South Korea (1,458.3), Botswana (1,168), and Ghana (411.86). (9)
We also consider a second list of peer countries and include only 15 of the 29 countries mentioned above that had per capita real GDP within 25% of South Africa's either in 1950 or in 1960 as per data availability. This group includes countries comparable to South Africa in terms of initial real GDP per capita. Listed in descending order of 1960 real GDP per capita, these countries are Austria, Italy, Trinidad and Tobago, Uruguay, Venezuela, Chile, South Africa, Spain, Costa Rica, Japan, Singapore, Greece, Algeria, Mexico, Portugal, and Mauritius. With this criterion of choosing similar countries, the question becomes why South Africa lagged behind countries that started at approximately the same initial point.
Since data on inflation are available only from 1961 and our focus is on South Africa's economic performance during apartheid, our estimation period is from 1961 to 2000. The variables are measured as averages of 5-year periods: 1961-1965; 1966-1970; 1971-1975; 1976-1980; 1981-1985; 1986-1990; 1991-1995; and 1996-2000. The dependent variable is Dy, the growth rate of real GDP per capita; the explanatory variables are Inv, Gov, Inflation, Education, Openness, and Initial, as defined in the previous section. Real GDP per capita, openness, and the shares of investment and government expenditure in GDP are obtained from the Penn World Tables, version 6.2, and are in constant international 2000 prices. Average years of schooling are from Barro and Lee (2000), and consumer price index (CPI) inflation is from the World Bank's World Development Indicators.
Estimation of the Solow growth model for peer countries
Estimation of equation 3 is reported in Table 1 below for the two panels of peer countries--the 29-country peer group and the 15-country peer group--discussed in the previous section. South Africa is not included in the estimation.
Estimations of growth models such as in equation 3 suffer from measurement errors in the explanatory variables and from endogeneity of regressors. Bond et al. (2001) point out that under such circumstances, the GMM approach has the potential for obtaining consistent parameter estimates. The GMM dynamic panel estimators are specifically designed to address the econometric problems resulting from unobserved country specific effects and joint endogeneity of explanatory variables, and are therefore well suited to estimate growth models as noted in Levine et al. (2000) and Bond et al. (2001).
Assuming that the error term is not serially correlated and that the lagged levels of the explanatory variables are weakly exogenous, the difference GMM estimators uses the lagged levels of the explanatory variables as instruments under the additional moment conditions that the lagged levels of the explanatory and the dependent variables are each not correlated with the change in the error term. Blundell and Bond (1998) propose the use of the system GMM method. Compared to the difference GMM, the system GMM may have superior finite sample properties (Blundell and Bond, 1998; Bond et al., 2001), and it is preferable to use the system GMM estimator in empirical growth work. The system GMM estimator combines the difference estimator with an estimator in levels in order to minimize potential biases. The equation in levels uses the lagged differences of the explanatory variables as instruments under the conditions that there is no serial correlation in the error terms and that the differences in the explanatory variables and the error terms are not correlated.
The validity of the instruments determines whether the system GMM estimator is consistent or not. Two tests that address this issue are the Sargan test of overidentifying restrictions and a test for second order serial correlation in the error term. The test for overidentifying restrictions has the null that the instruments are exogenous and the overidentifying restrictions are valid. The null hypothesis for the serial correlation test is that there is zero autocorrelation in the first differenced errors, that is, no second order serial correlation. The values reported for the Sargan test and the serial correlation test in Table I below are p values for the null hypothesis of valid specification. Since these p values are high, we do not reject the null hypothesis of valid specification.
The coefficient of initial real GDP per capita is reparameterized in accordance with the formulation in equation 3 and becomes negative and significant in both cases reported in Table i, consistent with the concept of conditional convergence, indicating that the growth rate of per capita real GDP depends negatively on the initial real GDP per capita after controlling for the other variables that affect growth. The share of government consumption expenditure has a negative effect on growth (Barro, 2000). Inflation has a negative and statistically significant, although small, effect on growth in both cases. Both real gross domestic investment as a share of GDP and average years of schooling have a positive and statistically significant effect on growth in both specifications.
A problem faced in empirical estimations of growth models, as noted by Sala-i-Martin and Doppelhofer (2004), is that growth theories are not explicit enough about which explanatory variables belong in the 'true' regression. Levine and Renelt (1992) observe that the most robust determinant of growth is the ratio of investment to GDP. Sala-i-Martin and Doppelhofer (2004) examine the robustness of 67 explanatory variables in cross-country growth regressions. Their strongest evidence is for the following three variables: log of initial GDP per capita, education, and the relative price of investment (included instead of the investment share of GDP). The share of government expenditure also ranked high among the explanatory variables. Our empirical results presented in Table 1 are in line with these authors.
Comparison of South Africa's actual performance with predicted performance based on peer country coefficients
We predict the level of South Africa's real GDP per capita over 5-year intervals from 1961 to 2000 under the two scenarios--one using the coefficients of the 15-country peer group and the other using the coefficients of the 29-country peer group--reported in Table 1 above. Let [[beta].sup.peer] denote the coefficients obtained from estimating the growth model with system GMM for the peer country groups, [Z.sup.SAF.sub.t] denote the explanatory variables for the case of South denote the actual GDP per capita of South Africa. (10) Table 2 Africa, and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] below reports the difference in actual GDP per capita relative to that predicted for South Africa as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)
Using either the 15-country or 29-country peer group, we note that South Africa's GDP per capita exceeds what is predicted for the 1961-1965 period, providing support for Moll's (1991) observation that South Africa's favorable resource position concealed its inherent economic weaknesses in the early decades of apartheid. Thereafter, South Africa's GDP per capita fell short of what was predicted until the end of apartheid in 1994 in the case of the 29-country peer group and until the end of the sample period in 2000 in the case of the 15-country peer group. Apartheid ended officially in 1994, but the burdens imposed by the apartheid legacy lowered growth performance in the late apartheid years and in the post-apartheid era as well. These effects of apartheid have been analyzed by Dollery (1994); Fedderke (2002); Jones (2002); Rodrik (2008); Edwards and Lawrence (2008); Hausmann and Klinger (2008b); and Frankel et al. (2008). The null hypothesis that the mean predicted value of South Africa's real GDP per capita is less than or equal to the mean actual value of South Africa's real GDP per capita is rejected with p values of 0.01 and 0.053 over the 1966-2000 period for the 1S-country peer group and over the 1971-1995 period for the 29-country peer group, respectively. We attribute this shortfall of actual GDP per capita in South Africa relative to predicted GDP per capita to apartheid, and we discuss why this seems plausible in the 'Comparison of South Africa across Countries'.
The separation of the output gap of South Africa with respect to peer countries: lower input levels versus lower productivity We note that actual GDP per capita in South Africa can be written as the sum of what is predicted for South Africa on the basis of peer country coefficients and an unexplained portion as indicated in Table 2 and again in equation 5 below. Actual GDP per capita in the peer country sample can analogously be written as the sum of what is predicted for the peer country group on the basis of peer country coefficients and an unexplained portion as indicated in equation 6 below. Therefore,
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
where [Z.sup.SAF.sub.t] denotes the magnitudes of the explanatory variables for South Africa and [[bar.Z].sup.peer.sub.t] denotes the mean values of the explanatory variables for the peer country group for each time period t.
Subtracting equation 6 from equation 5, we can write:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)
Equation 7 implies that the actual difference in GDP per capita between South Africa and the peer countries can be broken down into the first component (the predicted difference) that can be attributed to the differences in the magnitudes of the explanatory variables (between South Africa and the average of the peer group) and the residual unexplained component that does not arise from a difference in the quantity of inputs. Table 3 below reports these results for the 29-country peer group and the 15-country peer group.
In the case of the 15-country peer group, South Africa's GDP per capita consistently fell short of peer average values of GDP per capita for the entire period 1961-2000 (column (i)), but in the case of the 29-country peer group, South Africa's per capita GDP fell short of the peer average only from 1986 onwards (column (iv)). This difference reflects the composition of the two peer groups, with the broader 29-country peer group including several low-income to low middle-income countries--Sri Lanka, Ghana, Thailand, El Salvador, Philippines, and Tunisia.
Results with the 15-country peer group in column (ii) of Table 3 also show that South Africa's per capita GDP is predicted to be lower than the peer average for the entire sample period. This predicted shortfall in South Africa's GDP that is derived from the first term in equation 7 above can be attributed (in the case of the 15-country peer group) to the explanatory variables of interest: low initial real GDP, low share of investment in GDP, high share of government consumption expenditure, and low level of schooling, relative to the 15-country peer group average. These variables explain 51.34%, 38.73%, 10.10%, and 4.24%, respectively, of the predicted shortfall over the 1961-2000 period. (11) Over this period, South Africa's investment share of GDP was 10 percentage points lower, share of government consumption spending was 8 percentage points higher, and average years of schooling was lower by 1 year, compared to the 15-country peer group average values of these variables.
In the case of the 29-country peer group, South Africa's real GDP per capita is predicted to be lower than peer country averages from 1981 onwards, based on the first term in equation 7 and as shown in column (v) of Table 3 above. Over this period, low initial GDP, high share of government spending, low share of investment, and low level of schooling contributed 40.5 %, 15 %, 44.08%, and 2.3%, respectively, to the predicted shortfall of South Africa's real GDP per capita relative to the 29 country peer group average. (12)
Columns (iii) and (vi) of Table 3 indicate that, with both peer groups, there is a positive unexplained portion of growth in South Africa relative to the peer sample for the 1961-1970 period. As mentioned earlier (in the discussion of Table 2 results), this finding is likely due to a resource advantage that South Africa enjoyed relative to peer countries in the early decades of apartheid. In the period 1981-1985, the unexplained portion of the output differential is again positive in the case of both peer groups, probably because this was a period in which the dollar price of gold reached a peak. (13) South Africa was at that time one of the largest producers of gold in the world, and South Africa has always derived a large share of its export earnings from gold. Other than these periods in which South Africa's unique advantage in natural resources contributed to its improved relative performance, the unexplained portion is negative in the remaining years of the sample, 1971-1980 and 1986-2000, implying an inefficient use or lower productivity of inputs, primarily physical and human capital, in South Africa.
COMPARISON OF SOUTH AFRICA ACROSS COUNTRIES
In this section, we explain why it is plausible that apartheid policies in South Africa may have led to low shares of investment, low levels of education, and high shares of government consumption expenditure. We also explain why it is plausible that the unexplained shortfall in output in South Africa relative to peers could have arisen because of inefficiencies in the use of physical and human capital and from apartheid driven macroeconomic policy that was conducive to recessions.
South Africa during apartheid
Lowenberg (1997) documents the negative effects on growth arising from a number of specific apartheid policies. Thus, for example, apartheid labor market policies, such as job reservation and influx control, created a migrant labor system that resulted in severe shortages of both skilled and unskilled labor in the manufacturing sector, high costs of training and turnover of labor, and wasteful misallocations of scarce skills due to the limited mobility of labor between urban areas. Costly protectionist trade policies were implemented from an early date as a means of promoting the employment of white workers in preference to cheaper black labor. Firms were granted tariff protection explicitly as a reward for hiring whites and as compensation for the costs of discriminating against blacks. Import protection combined with export subsidies provided the basis for an import substitution industrialization policy which benefited white workers in the capital-intensive manufacturing sector but raised the cost structure of domestic industry and the prices of consumer goods, as well as generating current account deficits due to rising demand for imported capital goods and capital-intensive intermediate inputs. Subsequent extension of the import substitution policy to capital and intermediate goods led to higher domestic input prices, falling total factor productivity in manufacturing and a slowdown in the growth of manufacturing output and manufactured exports, (14) further exacerbating the current account deficit and dependence on foreign capital inflows.
Apartheid policy required separate educational systems for blacks and whites, with black education deliberately designed to be inferior in order to insure job preferences for whites (Lipton, 1985, p. 24). The result was a continuing shortage of skilled labor, especially after the growth of the industrial sector exhausted the pool of skilled whites. In order to stem the integration of blacks into white urban society, the apartheid regime adopted a policy of decentralizing manufacturing industry to selected 'growth points' on the borders of the black reserves or homelands. Voluntary relocation of firms to these border areas was encouraged by tax relief, subsidies, and concessionary loans, whereas a more coercive approach was taken by legislation in 1967, which gave the government the power to directly control the ratio of black to white workers employed by firms choosing to remain in designated white urban areas. However, the costs of this decentralization strategy were significant. In addition to shortages of skilled labor in the border areas, and the bureaucratic red tape associated with applying for permits to employ blacks in the urban areas, the decentralization policy encouraged mechanization of manufacturing industries which reduced job opportunities for unskilled blacks at the same time that the high import content of capital goods placed further strain on the balance of payments. Dependence on short-term foreign capital inflows to finance current account deficits became gradually more problematic as the international political environment grew less favorable to South Africa's racial politics, whereas the declining competitiveness of the country's inefficient, tariff-protected manufacturing sector continued to hamper the growth of exports. (15) To address the resulting depletion of foreign exchange reserves, the South African government took action to curtail imports by devaluing the currency and deflating the domestic economy. Such policy-induced recessions became a frequent attribute of South Africa's apartheid landscape, comprising a significant constraint on growth.
Apartheid South Africa was caught in a vicious cycle in which recessionary conditions, often deliberately induced in response to a deteriorating balance of payments situation, caused political unrest in the volatile urban townships, which in turn led to capital flight and consequent further worsening of the balance of payments. Lowenberg (1997) argues that the costs of continuing with apartheid policies came to be perceived as intolerable by the increasingly highly skilled and capital-owning white electorate, whose interests now lay in integrating the black population fully into a modern and competitive industrial economy. With the waning strength of its pro-apartheid constituency, the National Party government was driven to dismantle the system and negotiate its own abdication of power. (16)
Comparison of South Africa with similar countries
South Africa had the structures and institutions necessary for productivity gains to be made and transmitted across the economy (Moll, 1991). A plentiful supply of low-wage black labor was available which should have encouraged the growth of labor intensive manufacturing exports that could have resulted in high GDP growth by harnessing South Africa's comparative advantage (Lowenberg, 1997). Countries that had real GDP per capita very close to that of South Africa's in 1950 are indicated in Figure 1 below. It can be seen that in 1950, Italy, Uruguay, and Venezuela had higher per capita GDP than that of South Africa. Costa Rica, Mexico, Portugal, and Greece had lower per capita GDP than that of South Africa. Whereas Italy, Greece, and Portugal grew conspicuously faster than South Africa, the growth of the other countries, especially Venezuela, was not impressive. Venezuela's poor performance is partially attributable to the decline in 'democratic governance and political cooperation' (Monaldi and Penfold, 2006, p. 34). (17)
High inflation and political instability contributed to the poor economic performance of several countries in our sample. While Uruguay experienced average inflation rates of around 79% in the 1986-1990 period, Venezuela experienced average inflation rates of 45% over 1991-1995, and Mexico experienced an average inflation rate of 69 % in the 1980s, South Africa's inflation rate at its highest was an average of 14.66% in the 1980s. (18) South Africa had a political system that was tense but relatively stable. Compared with the Algerian civil war and violent political fluctuations in Argentina, Chile, El Salvador, Ghana, Peru, and Turkey, South Africa appeared politically stable, although such perceived stability was undoubtedly the result of high levels of repression. (19)
[FIGURE 1 OMITTED]
Other countries in the broader 29-country sample that faced economic deterioration at around the same time as South Africa had either one or a combination of the following unfavorable factors: a debt crisis in the cases of Argentina, Mexico, and Peru; high inflation rates in the cases of Argentina (787.01%) and Peru (1223.5 %); collapse of oil prices in the mid-1980s in the cases of the oil rich countries--Algeria, Mexico, Venezuela, and Trinidad and Tobago; and/or high levels of political instability in the cases of Thailand, El Salvador, Philippines, and Chile. In the case of Peru, there was a significant and prolonged growth collapse that began in the 1970s, caused by a decline in export earnings due to the fall in international prices (Hausmann and Klinger, 2008a). The slow growth experienced by Brazil in the 1980s was attributable to a rise in government consumption as a share of GDP and a sharp increase in the relative price of investment since 1980 (Adrogue et al., 2007).
The poor comparative performance of the South African economy is particularly striking when considered in the context of the enormous advantages enjoyed by South Africa relative to other middle-income developing countries (Lowenberg and Kaempfer, 1998, p. 220). Unlike many of those countries, South Africa possessed skilled managers and technicians, technological knowledge, excellent transport and communications infrastructure by developing-country standards, a modern financial and monetary system inherited from Britain, and, before 1960, easy access to world markets through membership in the Commonwealth. In this light, South Africa's undistinguished comparative performance seems decidedly poor (Moll, 1991, p. 280). Moreover, Moll (1991) points out that economic growth in South Africa after 1948 was only fractionally faster than it had been before that date, despite highly favorable economic conditions. The historical record indicates that, by the 1970s, apartheid bad begun to impose substantial costs in terms of forfeited growth opportunities. When compared to similar countries, all of the abovementioned evidence suggests that the apartheid economy of South Africa failed to achieve its growth potential.
Low investment as a factor contributing to slow growth
Our panel data results indicate the importance of investment in explaining growth. Evidence that low levels of investment contributed to low growth rates in South Africa is provided by Kantor (1993); Liu and Saal (2001); and Coulibaly (2009). Liu and Saal (2001) investigate the sources of structural changes in output growth in South Africa's economy over the 1975-1993 period using a decomposition method within the framework of an input-output model. They find that before 1981 overall growth in output was 'multicomponent driven,' with private consumption, government consumption, investment, export components, and import substitution, ali contributing positively to economic growth. However, the economic stagnation that characterizes the 1981-1993 period can, according to these authors, be traced to a single contributing factor--the collapse of investment demand. Investment made a -31.9 % contribution to economic growth in South Africa over the 1975-1993 period. In contrast, for every other country in their analysis, investment contributed positively to growth over the 1975-1993 period.
The role of apartheid in leading to a low investment ratio (20)
The low investment levels in South Africa relative to other countries in the sample can, in turn, be ascribed to some extent to the decrease in foreign capital inflows due to the economic uncertainties of apartheid. Before World War II, South Africa had no difficulty financing its occasional current account deficits with inflows of capital from abroad (Jones and Muller, 1992, pp. 221-222). But, South Africa subsequently became a chronic net debtor country partly due to import substitution industrialization policies that generated current account deficits throughout the post-World War II period (Davis, 1994, pp. 25-27). Between 1946 and 1974, South Africa's capital inflows mostly took the form of foreign direct investment. However, during 1975-1984, the bulk of the capital inflows switched from direct investment in the private sector to short-term loans to the public and banking sectors (Mohr, 1994, p. 51). Although rising indebtedness and a shift to short-term borrowing were part of a worldwide trend and were not exceptional to South Africa (Lipton, 1989, p. 343), a fall in expected profitability as a consequence of apartheid undoubtedly hastened the move toward increased reliance on short-term borrowing in South Africa. (21) The situation became critical in 1985 when increasing political unrest caused foreign lenders to withdraw credit (Lundahl, 1992, p. 325; Kantor 1993, p. 13; Davis, 1995, p. 180). With low domestic saving, the net capital outflows of the 1980s resulted in lower levels of domestic investment (Kantor, 1993, p. 12), a decrease in the growth rate of the capital stock, and consequent decrease in the growth of output (Coulibaly, 2009).
The foregoing arguments are supported by the following data from the World Bank's World Development Indicators: in 1970 foreign direct investment net inflows were 1.87 % of GDP for South Africa and this ratio was higher than that for the Latin American and Caribbean countries (1.01%) as well as for the group of upper middle-income countries (1.05 %). Starting from 1977, this ratio became negative in South Africa (-0.31%), but was 0.63 % and 0.57 % for the Latin American and upper middle-income countries, respectively. In 1992, while this ratio was 0.00% in South Africa, it was 1.15 % in Latin America and the Caribbean countries and 0.982 % in the group of upper middle-income countries.
Additionally, there were rising defense and security costs of the apartheid state. These were driven by growing black political unrest and an insurgency on the country's borders, and were financed by higher taxes, particularly on skilled wages and corporate profits, contributing to the high cost of capital, and reducing the share of investment in GDP even further.
Another possibility is that the effects of apartheid on investment worked through a Beckerian 'taste for discrimination'. Becker (1957) attributes racial wage differentials in a competitive labor market to a willingness on the part of prejudiced white capital owners to forgo profits in order to indulge their preference for hiring more expensive white workers over less expensive black labor. According to this view, discrimination against black workers results in a reduced rate of return to white-owned capital, but the pecuniary loss to white capital owners is offset by a so-called 'discrimination coefficient,' or subjective premium added to the black wage rate. In the South African case, the discrimination coefficient reflected a taste for discrimination on the part of the government and its political supporters, with the costs in terms of reduced profitability imposed on employers. Evidence for this view can be found in the perpetual complaints and opposition that white capital owners directed at the apartheid system. Employers in the manufacturing sector, in particular, disliked the migrant labor system created by influx control legislation, complaining about its adverse effects on the stability and morale of their workers (Lipton, 1985, p. 150). Employer groups, such as the Associated Chambers of Commerce (Assocom), the Federated Chamber of Industries (FCI), the Steel and Engineering Industries Federation (SEIFSA), and the Afrikaanse Handelsinstituut (AHI) lobbied vigorously against influx control and apartheid restrictions in the labor market (Lowenberg and Kaempfer, 1998, pp. 198-99). (22) As Moll (1991, p. 290) points out, 'Apartheid restricted capitalist firms in important ways'. Employers argued that influx control policies hampered their flexibility in reorganizing and changing their workforces to accommodate rapidly changing technological requirements (Lipton, 1985, pp. 140-41). (23) White capital owners also balked at the rising defense and policing costs of apartheid, not only because of the growing tax burden but also on the grounds that the deteriorating security situation and internal unrest were deterring foreign investment and contributing to the high cost of capital (Lowenberg and Kaempfer, 1998, p. 213). A plausible argument can therefore be made that apartheid policies, by depressing profitability, helped to bring about a low rate of investment, both domestic and foreign.
Other factors that caused low growth in South Africa
South Africa had lower levels of educational attainment compared with most other countries in our sample, for example, Mexico and Sri Lanka, both of which had considerably lower per capita CDP than South Africa in the 1950s and 1960s. In 1960, the average years of schooling in the total population over 15 years of age was 4.28 in South Africa, with the numbers for Mexico and Sri Lanka being 2.75 and 3.93, respectively. In 1995, these numbers were 6.03 for South Africa, but 6.96 and 6.45 for Mexico and Sri Lanka, respectively (Barro and Lee, 2000). South Africa also had one of the highest shares of government expenditure relative to GDP. An enormous administrative infrastructure was required to maintain the apartheid system. Apartheid necessitated separate bureaucracies to regulate economic and social activities in white, black, Indian, and Colored (mixed-race) communities, which resulted in considerable duplication of functions. For example, each ethnic group had its own national department of education, in addition to regional and homeland education departments (Lowenberg and Kaempfer, 1998, p. 213). (24)
CONCLUSIONS AND POLICY IMPLICATIONS
The economic and social impacts of apartheid have been discussed widely, but there is little empirical work that examines the economic impact of apartheid in the framework of growth models. We predict South Africa's real GDP per capita using peer country coefficients for two cases - a peer group of 15 countries that had initial starting GDP per capita within 25% of South Africa's, and a broader peer group of 29 countries that includes these 15 countries as well as several low- and middle-income countries. Our choice of peer groups is reflected in the results insofar as causing the actual and predicted values of South Africa's per capita GDP to each fall short of the respective peer averages much later in the sample period in the case of the broader 29-country peer group than in the case of the 15-country peer group. In either case, the shortfall of South Africa's predicted per capita GDP relative to peer country averages is due to differences in the magnitudes of the explanatory variables for South Africa and the peer countries. In particular, South Africa had lower levels of inputs of physical and human capital and a higher share of government consumption expenditure in GDP than peer country averages for these variables.
Regardless of the peer group considered, the GDP per capita predicted on the basis of peer country coefficients exceeds South Africa's actual GDP from 1970 onwards. For both peer groups, the portion of the difference between actual GDP per capita of South Africa and the peer group average that cannot be predicted is positive from 1961 to 1970 on account of a possible resource advantage of South Africa, but negative thereafter except for the 1981-1985 period that saw a peak in gold prices. When viewed in the context of literature analyzing the effects of South Africa's apartheid policies, it appears plausible that the unexplained negative component reflects the inefficiencies and low productivity of inputs associated with apartheid in South Africa.
Although it might have been instructive to measure directly the effects of cross-regional or time-series variations in apartheid on investment and growth, data difficulties preclude such tests. Apartheid was a national government policy and there is little evidence of regional variations of major apartheid regulations. Variations in apartheid over time can be captured most successfully by data on labor market institutions, which are unfortunately not available for the entire sample period of our study.
The low level of investment in physical capital in South Africa was due not only to shrinking foreign direct investment as a result of the political uncertainty caused by the apartheid regime, but also to a Beckerian 'taste for discrimination' on the part of the regime that lowered the rate of return to capital. The negative impact of apartheid on economic growth in South Africa remains an important question even after the end of apartheid. Although apartheid was unique to South Africa, many other countries today are afflicted by discriminatory political systems. Ethnic discrimination and affirmative action policies have been documented in Sudan, several countries in Latin America, the former republics of the Soviet Union, and in Eastern Europe (Wyzan, 1990). There are also many countries that are characterized by ethnic polarization, such as Bosnia and Herzegovina and Serbia and Montenegro. Authors such as Bridgman (2008) and Montalvo and Reynal-Querol (2005) have pointed out that ethnic polarization directly lowers income and negatively affects economic development by reducing investment and increasing government consumption.
Furthermore, a growing literature documents the importance of institutions, such as political and economic freedom, the rule of law and private property rights, in determining economic growth performance (Acemoglu et al., 2001, 2002; La Porta et al., 2002; Levine, 2005; Easterly, 2006; La Porta et al., 2008). Differences in legal and constitutional systems across developing countries have been shown to have an important effect on economic outcomes. There is much consensus that democratic political institutions are instrumental in sustaining those economic institutions that encourage growth. (25) Our study of South Africa under apartheid serves to further underline these conclusions by showing how the absence of democratic institutions, in countries with autocratic regimes, such as Zimbabwe and Myanmar, ultimately imposes a substantial cost in terms of forgone economic growth.
We thank an anonymous referee, the editor of the journal, and Andy Gill for valuable comments and suggestions. Any errors or omissions are our responsibility.
Acemoglu, D, Johnson, SH and Robinson, J. 2001: The colonial origins of comparative development: An empirical investigation. American Economic Review 91: 1369-1401.
Acemoglu, D, Johnson, SH and Robinson, J. 2002: Reversal of fortune: Geography and institutions in the making of the modern world income distribution. Quarterly Journal of Economics 117: 1231-1294.
Acemoglu, D, Johnson, SH and Robinson, J. 2004: Institutions as the fundamental cause of long-run growth Working paper 10481, National Bureau of Economic Research.
Adrogue, R, Cerisola, MD and Gels, RG. 2007: Brazil's long,-term growth performance--trying, to explain the puzzle Working paper 06/282, International Monetary Fund.
Barro, RJ. 1991: Economic growth in a cross-section of countries. Quarterly Journal of Economics 106(2): 407-443.
Barro, RJ. 2000: Inequality and growth in a panel of countries. Journal of Economic Growth 5: 5-32.
Barro, RJ and Lee, J. 1994: Data set for a panel of 138 countries, http://www.nber.org/pub/barro.lee.
Barro, RJ and Lee, J. 2000: International data on educational attainment: Updates and implications CID working paper 42.
Becker, CM and Pollard, PS. 1990: The vulnerability of South Africa to economic sanctions: An input-output analysis. In: Becker, CM., and Bell, T. et al., (eds). The Impact of Sanctions on South Africa. Investor Responsibility Research Center: Washington, DC.
Becker, GS. 1957: The economics of discrimination. Economic Research Studies, University of Chicago Press: Chicago.
Blundell, R and Bond, S. 1998: Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87(1): 115-143.
Bond, SR, Hoeffler, A and Temple, J. 2001: GMM estimation of empirical growth models. Working paper 2001-W21, University of Oxford.
Bridgman, B. 2008: Why are ethnically divided countries poor? Journal of Macroeconomics 30(1): 1-18.
Clague, C, Keefer, P, Knack, S and Olson Jr, M. 1999: Contract-intensive money: Contract enforcement, property rights, and economic performance. Journal of Economic Growth 4: 185-211.
Coulibaly, B. 2009: Effects of financial autarky and integration: The case of the South Africa embargo. Journal of International Money and Finance 28 (April): 454-478.
Coulibaly, B and Logan, TD. 2009: South Africa's post-apartheid two-step: Social demands versus macro stability. American Economic Review, 99(May): 275-281.
Davis, GA. 1994: South African managed trade policy: The wasting of a mineral endowment. Praeger Publishers: Westport, CT.
Davis, J. 1995: Sanctions and apartheid: The economic challenge to discrimination. In Cortright, D Lopez, G (eds). Economic Sanctions: Panacea or Peace building in a Post-Cold War World. Westview Press: Boulder, CO, 173-184.
Dollery, BE. 1994: Economic growth and redistribution in South Africa. Development Southern Africa 11(2): 199-204.
Doucouliagos, H and Ulubasoglu, MA. 2008: Democracy and economic growth: A meta-analysis. American Journal of Political Science 52: 61-83.
Easterly, W. 2006: The white man's burden: Why the west's efforts to aid the rest have done so much ill and so little good. Penguin: New York. Economist. 1995: September 9 46.
Edwards, L and Lawrence, RZ. 2008: South African trade policy matters: Trade performance and trade policy. Economics of Transition 16: 585-608.
Fedderke, JW. 2002: The structure of growth in tile South African economy: Factor accumulation and total factor productivity growth, 1970-1997. South African Journal of Economics 70: 611-646.
Feng, Y. 2003: Democracy, governance, and economic performance: Theory and evidence. MIT Press: Cambridge, MA.
Frankel, JA, Smit, B and Sturzenegger, F. 2008: South Africa: Macroeconomic challenges after a decade of success. Economics of Transition 16: 639-677.
Gallup, J, Sachs, LJ and Mellinger, AD. 1998: Geography and economic growth. Paper prepared for the Annual Bank Conference on Development Economics. Washington, DC.
Gervasi, S. 1970: Industrialisation, foreign capital, and forced labour in South Africa. New York: Monthly Review Press.
Hausmann, R and Klinger, B. 2008a: Growth diagnostics in Peru. CID working paper 181: 1-42.
Hausmann, R and Klinger, B. 2008b: South Africa's export predicament. Economics of Transition 16: 609-637.
Helpman, E. 2004: The mystery of economic growth. Belknap Press: Cambridge, MA.
Heston, A, Summers, R and Aten, B. 2006: Penn World Table Version 6. Center for International Comparisons at the University of Pennsylvania (CICUP), September.
Houghton, H. 1978: The South African economy. Oxford University Press: Cape Town.
Innes, D. 1984: Anglo American and the rise of modern South Africa. New York: Monthly Review Press and London: Heinemann.
Islam, N. 1995: Growth empirics: A panel data approach. Quarterly Journal of Economics 110: 1127-1170.
Iyengar, M and Porter, RC. 1988: South Africa without apartheid: Estimates from general equilibrium simulations. Michigan Research on Economic Development paper 123.
Jones, S (ed). 2002: The decline of the South African economy. Edward Elgar: Northampton, MA.
Jones, S and Muller, A. 1992: The South African economy, 1910-90. St. Martin's Press: New York.
Kantor, B. 1993: The future of the large South African corporation: an analysis of economic structure and performance. Working paper, School of Economics, University of Cape Town.
Knack, S and Keefer, P. 1995: Institutions and economic performance: Cross-country tests using alternative institutional measures. Economics and Politics 7: 207-227.
Knight, M, Loayza, N and Villanueva, D. 1993: Testing the neoclassical theory of economic growth. elano. Staff Papers-International Monetary Fund 40(3): 512-541.
La Porta, R, Lopez-de-Silanes, F and Shleifer, A. 2008: The economic consequences of legal origins. Journal of Economic Literature 46: 285-332.
La Porta, R, Lopez-de-Silanes, F, Pop-Eleches, C and Shleifer, A. 2002: The guarantees of freedom Working Paper 8759, National Bureau of Economic Research.
Levine, R. 2005: Law, endowments and property rights. Journal of Economic Perspectives 19: 61-88.
Levine, R and Renelt, D. 1992: A sensitivity analysis of cross-country growth regressions. The American Economic Review 82 (4): 942.
Levine, R, Loayza, N and Beck, T. 2000: Financial intermediation and growth: Causality and causes. Journal of Monetary Economics Elsevier 46(1/August): 31-77.
Lipton, M. 1985: Capitalism and apartheid: South Africa, 1910-84. Rowman and Allanheld: Totowa, NJ.
Lipton, M. 1989: The challenge of sanctions. South African Journal of Economics 57: 336-361.
Liu, A and Saai, DS. 2001: Structural change in apartheid-era South Africa: 1975-93. Economic Systems Research 13(3): 235-257.
Lowenberg, AD. 1997: Why South Africa's apartheid economy failed. Contemporary Economic Policy 15 (July): 62-72.
Lowenberg, AD and Kaempfer, WH. 1998: The origins and demise oy South African apartheid: A public choice analysis. University of Michigan Press: Ann Arbor, MI.
Lundahl, M. 1992: Apartheid in theory and practice: An economic analysis. Westview Press: Boulder, CO.
Mijiyawa, AG. 2008: Sustained economic growth: Do institutions matter, and which one prevails. Cato Journal 28: 385-420.
Mohr, PI. 1994: Restructuring the South African economy: Some pertinent issues. Journal for Studies in Economics and Econometrics 18: 41-54.
Moll, T. 1991: Did the apartheid economy 'fail'? Journal of Southern African Studies, 17(2): 271-291.
Monaldi, F and Penfold, M. 2006: The collapse of democratic governance: Political institutions and economic decline in Venezuela, http://www.cid.harvard.edu/events/papers/0604caf/ Monaldi_Penfold.pdf.
Montalvo, G and Reynal-Querol, M. 2005: Ethnic diversity and economic development. Journal of Development Economics 76(2): 293-323.
Olson Jr, M. 1996: Big bills left on the sidewalk: Why some nations are rich, and others poor. Journal of Economic Perspectives 10: 3-24.
Rodrik, D. 2000a: Institutions for high-quality growth: what they are and how to acquire them NBER working paper 7540, National Bureau of Economic Research: Cambridge, MA.
Rodrik, D. 2000b: Participatory politics, social cooperation, and economic stability. American Economic Review 90: 140-144.
Rodrik, D. 2008: Understanding South Africa's economic puzzles. Economics of Transition 16(September): 769-797.
Rodriguez, F. 2006: Understanding the evidence on Venezuelan economic growth. Canadian Journal of Development Studies 27(4): 503-529.
Rosendorff, BP. 1996: The transition to democracy in South Africa: Trade, contest and sanctions Working paper, Department of Economics, University of Southern California.
Sala-i-Martin, X and Doppelhofer, G. 2004: Determinants of long-term growth: A Bayesian averaging of classical estimates (BACE) approach. American Economic Review 94(4): 813-835.
Steyn, GJJF. 1970: Some aspects of the development of the domestic and foreign trade of the republic of South Africa during the post-war period. South African Statistics: Pretoria. World Development Indicators. 2007: The World Bank. World Development Indicators: Washington, DC.
Wyzan, ML (ed). 1990: The political economy of ethnic discrimination and affirmative action: A comparative perspective. Praeger: New York.
(1) For an overview of the evolution of apartheid policy and its antecedents, see Lowenberg and Kaempfer (1998, pp. 33-42).
(2) Quoted in Moll (1991, p. 271).
(3) Quoted in Moll (1991, p. 272).
(4) Moll, 1991, p. 272.
(5) The latter is equivalent to what Barro (2000, p. 10) refers to as the long run or target level of output.
(6) All variables except for inflation are measured in logarithms.
(7) This is equivalent to equation 16 on p. 15 of Bond et al. (2001).
(8) In a similar exercise aimed at investigating South Africa's more recent performance, Coulibaly and Logan (2009, p. 276n) compare South Africa against peer countries based on the World Bank's classification.
(9) These data are in constant international 2000 prices, from the Penn World Tables, version 6.2, Heston et al. (2006). Figures for Chile and Greece are for 1961, and Botswana for 1970. Nigeria was excluded from the sample of countries considered by Moll (1991) because of insufficient data.
(10) The Solow growth model is based on the Cobb-Douglas production function and is therefore estimated in logarithms (for example, Knight, Loayza, and Villanueva, 1993). In order to obtain specifications 4 and 5, we obtain the log inverse of the dependent variable.
(11) The contribution of the inflation rate to decreasing the predicted shortfall over the 1961-2000 period was 4.06% (or -4.06%) since South Africa had lower inflation rates relative to the 15-country peer average.
(12) The inflation rate decreased the predicted shortfall since South Africa had lower inflation rates relative to the 29-country peer average.
(13) From http://goldprice.org/30-year-gold-price-history.html.
(14) Edwards and Lawrence (2008) argue that this pattern of protection, particularly detrimental to exports of manufactured goods, seriously impeded total export growth. When global commodity markets were weak, these protectionist policies had the effect of significantly constraining GDP growth and diminishing the beneficial impacts of the sinking rand in the late 1980s.
(15) Poor export performance has continued to plague the South African economy, even after the dismantling of apartheid. Thus, Rodrik (2008) shows that weakness of export-oriented manufacturing has deprived South Africa of growth opportunities and job creation at the relatively low end of the skill distribution, contributing to the current high unemployment rate among blacks.
(16) Certainly political pressures from abroad played a role in the demise of apartheid, particularly the withdrawal of long-term private capital by overseas investors, culminating in the debt crisis of 1985, which was sparked by the refusal of foreign banks to extend further short-term credit to South Africa. The fall of the Berlin Wall in 1989 was also a major catalyst for political reform in South Africa (Lowenberg and Kaempfer, 1998, p. 217).
(17) However, Rodriguez (2006) argues that in terms of non-oil growth and total factor productivity numbers, Venezuela's economic performance does not deviate significantly from that of other Latin American countries.
(18) CPI inflation rates are from the World Bank's World Development Indicators.
(19) Barro (1991) and Gallup et al. (1998) find a strong negative relationship between political instability and economic growth. The average political instability indexes over the 1960-1985 period for the countries in our sample were Algeria, 0.04; Argentina, 0.42; Austria, 0.00; Botswana, 0.00; Brazil, 0.04; Chile, 0.10; Colombia, 0.01; Costa Rica, 0.00; El Salvador, 0.41; Ghana, 0.17; Greece, 0.09; Italy, 0.03; Japan, 0.00; Korea, 0.16; Malaysia, 0.02; Mauritius, 0.00; Mexico, 0.00; Peru, 0.08; Philippines, 0.23; Portugal, 0.12; Singapore, 0.00; South Africa, 0.03; Spain, 0.06; Sri Lanka, 0.04; Thailand, 0.14; Tunisia, 0.00; Turkey, 0.11; Uruguay, 0.04; and Venezuela, 0.17. These numbers are from Barro and Lee (1994), whose political instability index, Pinstab, is calculated as follows: Pinstab = 0.5 (number of assassinations per million population) + 0.5 (number of revolutions per year).
(20) Some of the material in this section is adapted from Lowenberg and Kaempfer (1998, pp. 209-211).
(21) See Rosendorff (1996, p. 15n).
(22) The FCI and Assocom, for example, opposed color bar policies from their very inception, arguing that these laws raised the costs of South African manufacturing by limiting the supply of black labor and creating administrative hurdles to hiring black workers. Especially unpopular among employers were the labor bureaus that administered the influx control system. These bureaus enforced cumbersome procedures for hiring workers, and firms often complained that the bureaus were ineffective in securing adequate supplies of suitable workers (Lowenberg and Kaempfer, 1998, pp. 198-199).
(23) As early as the 1940s, SEIFSA argued in favor of phasing out the migrant labor system, and by 1976 even the conservative AHI supported stabilization of the black labor force (Lipton, 1985, p. 159). In 1976, Assocom also called for an end to migrancy, for land to be released for black family housing and for blacks to be granted secure titles to their property (Lipton, 1985, p. 159).
(24) There were, in fact, no fewer than 19 separate education departments under the apartheid regime (Economist (1995, p. 46) September 9).
(25) See Knack and Keefer (1995); Olson (1996); Clague et al. (1999); Rodrik (2000a, 2000b); Feng (2003); Acemoglu et al. (2004); Doucouliagos and Ulubasoglu (2008); and Mijiyawa (2008). Helpman (2004) provides a review of research on the growth effects of institutions, including property rights protection, legal systems, customs, and political systems.
RADHA BHATTACHARYA  & ANTON D LOWENBERG 
 Economics, California State University at Fullerton, SGMH 3393, Fullerton, CA, 92834, USA.
 Economics, California State University at Northridge, 18111 Nordhoff Street, Northridge, CA 91330-8374, USA.
Table 1: Estimation of the Solow growth model with system GMM Explanatory variables 15-Country 29-Country peer group peer group Log (initial GDP) 0.819 ** 0.872 ** Log (open) 0.015 -0.008 Log (investment/GDP) 0.132 ** 0.169 ** Log (government spending/GDP) -0.07 0.134 ** Log (education) 0.113 ** 0.079 * Inflation -0.0012 ** -0.0001 ** Sargan test 0.47 0.64 Serial correlation test 0.36 0.26 Notes: Dependent variable is Log of real GDP per capita. GMM type instruments for differenced equation: third and deeper lags of all explanatory variables. GMM type instruments for level equation: second tagged difference of each explanatory variable. The values reported for the Sargan test and the serial correlation test are p values for the null hypothesis of valid specification. Significance at the Levels of 5% and 10% are denoted by ** and *, respectively. Table 2: Comparison of actual real GDP per capita in South Africa with predicted real GDP per capita for South Africa Year South Africa Based on 15-country peer group actual (i) South Africa Gap actual-- predicted (ii) predicted (iii) 1961-1965 5,503.28 5,435.42 67.86 1966-1970 6,200.85 6,270.56 -69.71 1971-1975 6,700.20 7,102.32 -402.12 1976-1980 7,222.03 7,308.28 -86.24 1981-1985 7,686.44 7,715.20 -28.77 1986-1990 7,566.06 7,785.40 -219.34 1991-1995 7,348.45 7,709.28 -360.82 1996-2000 7,810.78 7,875.05 -64.27 Year Based on 29-country peer group South Africa Gap actual-- predicted (iv) predicted (v) 1961-1965 5,230.81 272.46 1966-1970 6,148.68 52.16 1971-1975 7,025.21 -325.01 1976-1980 7,247.82 -25.78 1981-1985 7,698.62 -12.20 1986-1990 7,651.76 -85.7 1991-1995 7,486.51 -138.06 1996-2000 7,615.55 195.22 Table 3: Comparison of the actual difference in South Africa's GDP per capita and peer country GDP per capita with the predicted difference Year Based on 15-country peer group Actual Predicted Unexplained difference (i) difference (ii) difference (iii) 1961-1965 -228.58 -766.74 538.16 1966-1970 -799.37 -819.13 19.76 1971-1975 -1886.29 -1159.62 -726.67 1976-1980 -2794.53 -2656.48 -138.05 1981-1985 -2862.61 -3541.68 679.07 1986-1990 -3955.54 -3929.91 -25.63 1991-1995 -5803.85 -5108.28 -695.57 1996-2000 -6974.97 -6832.08 -142.88 Year Based on 29-country peer group Actual Predicted Unexplained difference (iv) difference (v) difference (vi) 1961-1965 1212.10 680.95 531.15 1966-1970 1165.94 1007.57 158.36 1971-1975 527.07 1010.79 -483.72 1976-1980 38.95 40.67 -1.72 1981-1985 129.97 -389.79 519.77 1986-1990 -751.65 -609.21 -142.44 1991-1995 -2219.83 -1814.35 -405.47 1996-2000 -3016.58 -2996.11 -20.46 Notes: Actual difference=South Africa's real GDP per capita-peer average real GDP per capita. Predicted difference=South Africa's predicted real GDP per capita-peer average predicted real GDP per capita using mean values of peer explanatory variables. Unexplained difference=actual difference--predicted difference.