Democratization and the dynamics of income distribution in low- and middle-income countries

Abstract

Redistribution of income in developing countries for the sake of human development and economic growth is a major, but under-studied challenge. Recently released high-quality data on the dynamics of income distribution in developing countries allow us to test whether democratization has helped the median voter and her class allies to improve their income share, as predicted by the median-voter hypothesis (MVH). Using decade sharegains of the third and first (poorest) quintiles as dependent variables, and controlling for prominent political and economic features of developing countries, the MVH is found to be an inaccurate guide to the effects of democratization. The MVH underestimates the consequences of power concentration and the effects of hegemony, and assumes too glibly that the median income earner is also an influential voter. The decisive voter in developing countries lies in the richest quintile that gains most from the introduction of competitive elections, fiscal redistribution, and from economic liberalization in general. Redistribution does not flow automatically from the introduction of democracy and economic liberalization, but must be targeted specifically.

Notes

1. By adding measures of asset inequality and income/expenditure distribution to standard growth regressions, a number of authors in the early to mid-1990s came to the conclusion that growth is indeed retarded by higher levels of income and land inequality (notably Alesina and Rodrik, 1994; Persson and Tabellini, 1994; Aghion and Garcia-Penalosa, 1999). As Benabou (1996, p. 13) in his detailed overview of the literature in the mid-1990s concludes: ‘These regressions, run over a variety of datasets and periods with many different measures of income distribution, deliver a consistent message: initial inequality is detrimental to long-run growth’. Some subsequent cross-country regression studies have confirmed Benabou's conclusion, but it has also been challenged (see Barro, 1996; Barro, 2000; Forbes, 2000; Quah, 2001; Banerjee and Duflo, 2000; Li and Zou, 1998). However, these contradictory findings refer mostly to developed countries. When focusing on poorer countries alone, Deininger and Squire (1998) find that growth is significantly depressed by income inequality (see also Rudra, 2002) and in particular by high levels of initial land inequality in non-democracies. See also the discussion in Bruno et al. (1998) and Kentor (2001). For general introductions to these and related debates, see Kanbur (2000) and Ray (1998, pp. 169–248).

2. The MVH is based on the work of Downs (1957) and Meltzer and Richards (1981). For discussions and applications of the median voter hypothesis, see Evans (2004), Mueller (2003), Congleton (2002), and Borck (2003).

3. Bearse, Glomm and Janeba (2000).

4. See Doorenspleet (2000) for an analysis of this phase of global democratization.

5. ‘Global Poverty Monitoring Database.’ http://www.worldbank.org/research/povmonitor/For more information on the methodology used by the GPM, see Adams (2002) and Chen and Ravallion (2000).

6. I rely on the Deininger and Squire (1996, and 1998) and Knowles' (2001) criteria: use only households surveys; the population covered must be representative of the population as a whole; and the measure of income or expenditure must include income from self-employment, non-wage earnings, and non-monetary income. A further quality criterion, relying on expenditure surveys only (see also Nel, 2003) cannot be applied here because most surveys for Latin American countries are income based.

7. There are no comparable data for any of the ‘developing’ countries of Central Asia or Eastern Europe.

8. For a recent discussion see Knack (2004).

9. There is considerable debate about the relative merits of democracy measures (Reich, 2002; Vanhanen, 2003; Przeworski et al., 2000). Although this was not the explicit or only purpose of designers of the POLITY IV dataset, their DEMOC measure is useful to identify gradations of electoral democracies.

10. See also Robinson (1996); Diamond (1999); O'Donnell (1996); Diamond, Linz and Lipset (1989).

11. For students of comparative politics, the MVH explains why it is not easy to democratize a highly unequal society in which wealth is not mobile. Regime choice and redistributive struggles go together (Boix, 2003). See also Burkhart (1997, p. 148).

12. See Congleton, (2002); Mueller (2003); Milanovic (2000). Nelson (1999) challenges assumption (c), but does so only for developed countries.

13. See O'Donnell (1999).

14. There is some evidence that the relationship between initial levels of inequality and redistribution is non-linear (Benabou, 2000; de Mello and Tiongson, 2003). Note, however, that Benabou (and de Mello and Tiongson, 2003) use fiscal transfers as a measure of redistribution, while I am here interested only in the sharegain of the median voter and the poor. I will not assume, therefore, that non-linearity applies to equation (1).

15. It is likely that the distribution of income would have an effect on level or depth of democracy (see Burkhart, 1997; Muller, 1988). However, I am here interested in simultaneous changes in two dynamic variables, namely democratization and redistribution.

16. t refers to the period 1986–95 (t variables register changes over this ten-year period, or are ten-year averages, except when stipulated otherwise).

17. Headcounts of the share of the population living below the widely accepted World Bank standard of US$32.74 per person per month) indicate that on average around 20 per cent of the population in developing countries falls below the poverty line. I assume, therefore, that the income share of the lowest 20 per cent on the income scale in any particular country coincides with the income share of the poor.

18. The Freedom House civil liberties measure is scored from 1 to 7, with a score of seven indicating the worst record. It is thus more accurate to treat this rather as a measure of the repression of civil liberties. The civil liberties looked at are: freedom of speech and belief, association, and political organization; freedom from unjustified imprisonment; the rule of law; protection of private property (see also Knack, 2004).

19. According to Boix (2003, pp. 25–26), elites in these countries agree to democratization when the cost of moving their wealth is lower than the cost of the future tax rate.

20. In terms of Boix's assumptions it is no coincidence that the hegemony of the neo-liberal ideological position coincides with an upsurge in democratic experiments in developing countries: the mobility options opened by the liberalization of the world economy have altered the risk perceptions of wealthy elites. For an interpretation of these developments that place them in a global perspective that highlights the hegemonic role of the USA, see Robinson (1996).

21. In many developing countries this ideological hegemony was institutionalized in the 1980s and 1990s in the form of structural adjustment programmes promoted by the international financial institutions. Although these programmes did not reduce government spending in toto, certain sectors with distributive potential did suffer significantly: education, infrastructure and agriculture in Africa; agriculture and health in Asia; education and infrastructure in Latin America (Fan and Rao, 2003).

22. See the discussion of the ‘mixed-blessings’ of financial openness by Rodrik (1999).

23. For the salience of ethnicity in Africa see Easterly and Levine (1997).

24. There are insufficient data for a large section of my sample to complement this redistribution-in-kind indicator with a measure of redistribution-in-cash, such as social welfare spending.

25. Although inequality and democratization could be regarded as endogenous variables to equation (1), I use a reduced form of the equation by lagging these variables. Inequality is initial inequality, while democratization is the degree of democratization preceding the final observation of income distribution.

26. A surprising finding is that ethnic, religious, and linguistic fractionalization is strongly and positively correlated with the sharegain of the third and first quintiles, but I expect that this is a spurious relationship. In any case, given the absence of any supporting evidence for the MVH, the hypothesized collective action problems supposedly caused by ethnic, linguistic and religious fractionalization become irrelevant.

27. Beck, Clarke, Groff, Keefer and Walsh (2001).

28. See also Castro-Leal, Dayton, Demery, and Mehra (1999).

29. For a general theory of the rent-seeking behaviour of political elites, see Verdier and Ades (1996).

30. Note the finding by Alderson and Nielsen (1999) that increasing levels of FDI do increase inequality, but that at high levels of FDI ‘dependence’ inequality declines, suggesting a more complex relationship than is assumed by both neoliberals and their critics.

31. Economic growth at t+1 is positively related to the sharegain of the third quintile at t (r=.188), and negatively with the deepening of poverty at t (r=−.264).