Economic growth and social indicators: An exploratory analysis

Abstract

What are the connections between social and political conditions and economic growth? This paper explores the uses and misuses of statistical analysis of cross-national data in addressing this question. It shows that social, political and economic indicators are linked by “webs of association”. Such webs of association suggest the possibility of distinct groupings of social indicators with differentiated impacts on economic growth. But such correlations also make it difficult to disentangle causal relationships, especially when theorizing is weak, data are badly behaved, and the number of observations is small. Although under such conditions statistical techniques can help preclude premature generalizations, they are easily overinterpreted. Nonetheless, data analysis can help identify countries that seem exceptions to the general patterns, where careful case studies may be especially valuable.

Notes

1. Of course, endogenous growth theory has offered its own explanation for divergence (see for instance Romer 1986, 1990, 1994, Lucas 1988, Grossman and Helpman 1994, Mankiw 1995; in a different vein, see Nelson and Wright 1992).

2. Which postulates a link from economic development to democratization (see the discussion in Lipset 1959, Cutright 1963, Bollen 1979, Bilson 1982, Bollen and Jackman 1985, Sirowny and Inkeles 1990, Diamond 1992, Helliwell 1992, Burkhart and Lewis-Beck 1994).

3. Since democracies at low levels of development may be subject to populist pressure for redistributive policies (see Weede 1983, Marsh 1979, Landes 1990, Barro 1994).

4. See Venieris and Gupta 1986, Alesina and Tabellini 1989, Ozler and Tabellini 1991, Londregan and Poole 1990, Barro 1991, Alesina and Perotti 1993, Olson 1993, Knack and Keefer 1995, De Haan and Siermann 1996. Though note that disruption of rent-seeking activities may have a positive impact on growth (Olson 1982), and the possibility of simultaneity between growth and instability has also been mooted (Olson 1963, Londregan and Poole 1990).

5. For a fuller discussion of these issues see Klitgaard and Fedderke (1995).

6. A number of ratings for the political, civil and “economic” rights the citizens of different countries enjoy have been devised. The ratings are subjective, and their reliability in the strict statistical sense is unknown. Nevertheless when one compares several of the available indices of rights, rank correlations of the indices of political rights range between 0.64 and 0.91, with a median of more than 0.7 even over different years. The metaphor of a latent trait suggests itself.

7. On the other hand, the strength of the rank correlations is not particularly strong, ranging from 0.20 to 0.40 for the 1960–1985 average growth rate. For the strongest correlation (that for the Gastil POLRIGHT measure), a standard deviation improvement in rights is equivalent to the substantial difference in average rights between El Salvador and Italy, and yet would be associated with an increase in the annual growth rate of 0.7%. Moreover, the growth rate is not significantly correlated with a wide range of measures of political and economic discrimination, a direct measure of corruption, voter turnouts in election, and measures of assassinations, riots, government crises and strikes. If at all, separatist pressures and political party fractionation are only weakly associated with economic growth.

8. Higher levels of real per capita GDP are associated with greater rights, a lower level of regime-threatening political instability, a higher level of efficiency in public institutions, higher levels of political party fractionation, with lower levels and intensities of separatist pressures, and with higher levels of non-regime-threatening political instability.

9. Measured by revolutions, coups, and constitutional changes.

10. Measured by the average number of politically motivated strikes and riots per annum.

11. Note particularly the absence of significant correlations with income inequality (in contrast to Alesina and Perroti 1993, Persson and Tabellini 1994, Clarke 1995), and neither individualism nor hierarchical steepness confirm plausible a priori expectations of significant associations with growth.

12. Though here too care must be exercised. The negative correlation with reported rapes is driven by the influence of two outliers – quite apart from the fact that the rapes measure is concentrated on developed countries, and may simply reflect superior data collection in more developed societies.

13. A one standard deviation increase in ethno-linguistic fractionation, equivalent to the substantial increases associated with a move from Kuwait to Sri Lanka, Korea (the most homogenous country in the sample) to Mexico or Panama, or from Malawi to India or South Africa, would result in a 0.7% decrease in the average annual growth rate.

14. For homicide rates it is worth noting that a strong threshold effect is present. The negative correlation between the level of per capita GDP and homicide rates reflects high homicide rates for countries with low per capita GDP, while mid to high real per capita GDP countries do not demonstrate a big difference in homicide rates.

15. The political rights measures are not entirely identical, and some differences between the measures do emerge: only the Bollen and Humana indices show significant association of higher rights with higher voter turnout in elections, with the 1960 measure of political discrimination (with higher discrimination being associated with lower rights), income and land inequality (lower inequality under better rights), individualism and hierarchical steepness of society (better rights associated with greater individualism, and less hierarchical steepness), and suicide rates (higher suicide rates under better rights). Finally, only the Bollen indices suggest that better rights are associated with a lower need for certainty, and only the Gastil measures suggest that higher rights are associated with a higher score on the masculinity index.

16. A standard deviation improvement in the Gastil POLRIGHT measure (equivalent to a shift from El Salvador to Italy), entails a decrease in the average number of annual revolutions for 1960 to 1985 from a level experienced by Panama to that of Switzerland, and an efficiency improvement in the judiciary from Dominican Republic to French levels.

17. The only exception being the strikes measure, for which better rights are associated with higher levels of strike activity, though the strength of the association is low, ranging from 0.21 to 0.36.

18. For ethno-linguistic fractionation the association is not particularly strong, and a standard deviation improvement on the Gastil POLRIGHT measure is associated with the relatively minor decrease in fractionation from that of the Netherlands to that of Korea (the most homogenous country in the sample).

19. A standard deviation reduction in the revolutions (regime-threatening) measure is associated with the strong improvement in judiciary efficiency from Algerian to Swiss levels. By contrast a standard deviation decrease in the political riots and assassinations (non-regime-threatening) measures are associated with less dramatic improvements in judicial efficiency from German to Austrian, or Nicaraguan to Panamanian levels respectively.

20. Countries with high levels of regime-threatening instability are likely to be less individualistic, and have greater hierarchical steepness on the Hofstede measures.

21. Again the strength of the rank correlations of political instability to the ethnic-fractionation index is not high, however.

22. Higher levels of regime-threatening political instability are associated with higher homicide rates and youth unemployment, but lower suicide rates, reported rapes and drug related crimes.

23. Similar patterns emerged in Klitgaard and Fedderke (1994, 1995). The earlier work found two independent factors that in turn correlate highly with an array of cultural, political, social, and economic variables – though not with all. The first factor loaded on poor rights, and fundamental instability indicators, while the second loaded on protest instability indicators.

24. The simultaneity issue has received some attention in the literature (see Scully 1988, Londregan and Poole 1990, Helliwell 1992, Barro 1994).

25. GR denotes the growth rate of real per capita GDP, POLRIGHT the Gastil rights measure, INV the investment rate, HUMKAP a human capital proxy, GOV government consumption expenditure, GPOP the population growth rate, REVCOUP the composite of revolutions and coups, and CONSTCH the constitutional change measure.

26. Though as Friedman (1992) and Quah (1993) point out, the coefficient of starting GDP does not provide a legitimate test for conditional convergence.

27. A standard deviation change in political rights is associated with a 0.21 standard deviation change in growth.

28. For full sensitivity tests see Fedderke and Klitgaard (1996).

29. Though a number of standard proxies of human capital are problematic. For instance, primary school enrolment rates suffer from strong ceiling effects at low levels of per capita GDP, and neither secondary nor primary school enrolment rates control for the quality of human capital.

30. Even in the presence of regressors which rendered political rights insignificant in other studies.

Additional information

Notes on contributors

Johannes Fedderke

Johannes Fedderke is Professor of Economics and Director of Economic Research Southern Africa at the University of Cape Town.

Robert Klitgaard

Robert Klitgaard is the President of Claremont Graduate University, where he is also a University Professor. He has been at Harvard, Yale, the University of Natal, and the University of Karachi.