Who governs or how they govern: Testing the impact of democracy, ideology and globalization on the well being of the poor

Abstract

This paper examines the effects of regime type, government ideology and economic globalization on poverty in low- and middle-income countries around the world. We use panel regression to estimate the effect of these explanatory variables on two different response variables: national poverty gap (104 countries from 1981 to 2005) and child mortality rate (132 countries from 1976 to 2005). We find consistent and significant results for the interactive effect of democracy and government ideology: strong leftist power under a democratic regime is associated with a reduction in both the poverty gap and the child mortality rate. Democracy, on its own, is associated with a lower child mortality rate, but has no effect on the poverty gap. Leftist power under a non-democratic regime is associated with an increase in both poverty measures. Trade reduces both measures of poverty. Foreign direct investment has a weak and positive effect on the poverty gap. From examining factors that influence the welfare of poor people in less developed countries, we conclude that who governs is as important as how they

Keywords:

• Democracy
• Ideology
• Political economy
• Globalization
• Poverty
• Domestic institutions

Acknowledgements

The authors thank Alma B. Calderon, Geoffrey Garrett, Kerim Can Kavakli, Edward Leamer, Dong-wook Lee, Lisa Piergallini, Ronald Rogowski, George Tsebelis, and three anonymous reviewers for helpful comments and assistance with this paper. All errors remain our own

Notes

1 Pribble et al. (2009) find, in Latin America at least, that the strength of leftist parties is a significant and negative predictor of national poverty level. Huber and Stephens (2012) also find that the strength of leftist parties is strongly associated with lower inequality and poverty in Latin America.

2 Since the child mortality rate data extends back to 1970, we report changes across this full data set. Due to data limitations in our other variables, regressions are based on 1976–2005 data. WDI reports annual data for child mortality rate, filling each observation for five years with the same values. As such, Fig. 2 reports 5 year averages of child mortality rate.

3 Our main results are also robust to the use of a conservative measure of government ideology data that treats all the controversial measures as missing. This measure treats government ideology as missing: (1) when the ideology of the coalition government is used to compute government ideology instead of all of the government parties, (2) when the ideological position of any of the government parties is unavailable (e.g., small religious or regional parties with less than two seats), and (3) when the values for the ideology of any of the government parties have huge discrepancies among main data resources (e.g. left vs. right). The result tables are available upon request.

4 In the dataset, religious, regional, and personalist parties are treated missing unless their ideological positions for state control of the economy are available in the resources. Religious parties are heterogeneous and weakly organized in LDCs as compared to those in advanced democracies (Mainwaring & Scully, 1995).

5 Measuring government ideology as three categories (left, center, and right) is crude as it does not reflect the differences between, for instance, center-left and extreme-left parties. However, data on LDCs is very limited for cross-country studies (e.g. Huber & Inglehart, 1995). Ha’s (2012) government ideology data is highly correlated with the data generated by two popular expert surveys: Castles and Mair’s (1984) continuous party ideology measure for advanced industrial countries and Coppedge’s (1997) five-category party ideology data (left, center-left, center, center-right, and right) for Latin American countries (r = 0.60 and r = 0.90 respectively) (Ha, 2012).

6 Female education may have a stronger effect on child mortality (McGuire, 2006; Pritchett & Summers, 1996). The results for child mortality rate are robust when we control for female education, measured by female secondary school enrollment rate.

7 The AR1 process models the error term as: y_it = α + βx_it + ε_it where ε_it = ρε_it−1 + u_it, u_it is assumed to be white noise (Plümper et al., 2005).

8 We also test the interactive effects by creating left-dummy and right-dummy variables (with Left or Right power ≥50%). We retest using Democracy-dummy, Left-dummy, Left-dummy × democracy-dummy, Right-dummy and Right-dummy × democracy-dummy. In these models, only Democracy-dummy × Left-dummy showed significant and negative impacts on premature death, suggesting that only leftist, democratic governments reduce poverty.

9 The WDI dataset does include access to drinking water, but observations are available only from year 1990. WHO/UNICEF provides yearly observations after 1990 and five-year observations prior to that year. Missing data for year t − 1 and t + 1 are imputed by available data in year t.

10 The main results are robust to the exclusion of democracy and ideology variables and to the inclusion of separate variables for education, health, and social security and welfare expenditures. Following Rudra (2008), we also use an instrumental variable approach (i.e., two-stage least squares) to test the effects of leftist power in government, democracy and globalization, and test for the presence of an endogenous relationship between social expenditures (% GDP) and child mortality rate. Our main results are still robust to the method, except that FDI inflow is significantly and negatively associated with child mortality rate. Please see Appendix D for the results.