Is the Millennium Development Goal on Poverty Still Achievable? The Role of Institutions, Finance and Openness

We are grateful to T. Elhaut, Director, PI, IFAD, for his support and guidance throughout this study. We are also grateful for the incisive comments of two anonymous reviewers. The views expressed are, however, our personal views and not necessarily of the organizations to which we are affiliated. Valuable research assistance by Valentina Camaleonte and Sundeep Vaid is greatly appreciated.

Abstract

Drawing upon new World Bank poverty data, the analysis examines the feasibility of attaining the Millennium Development Goal of halving extreme poverty (MDG1) when the interrelationships between finance, institutions, trade liberalization, growth and poverty are taken into account. The authors' econometric results suggest a slowing down of poverty reduction in the more recent years since 2000. They also confirm: the role of better institutions in income growth, poverty reduction, trade openness and financial development; the role of financial development in economic growth; and the positive effect of capital liberalization on financial development. Simulations for different regions show that MDG1 is attainable in most regions if the historical growth rate is maintained over 2006–15. However, improvements in institutional quality are crucial for halving extreme poverty in sub-Saharan Africa.

Notes

We are grateful to T. Elhaut, Director, PI, IFAD, for his support and guidance throughout this study. We are also grateful for the incisive comments of two anonymous reviewers. The views expressed are, however, our personal views and not necessarily of the organizations to which we are affiliated. Valuable research assistance by Valentina Camaleonte and Sundeep Vaid is greatly appreciated.

¹ See, for example, Besley & Burgess (2003) and Gaiha (2003).

² Estimates of extreme poverty are based on the poverty lines of $1.08 a day (1993 PPPs) or $1.25 a day (2005 PPPs). Although another indicator of poverty—the poverty gap ratio—is included in the MDGs along with non-monetary indicators (e.g. under-five mortality rate, maternal mortality rate); we have confined our analysis to the head-count ratio for two reasons: larger coverage of countries; and easier data availability.

³ The US dollar value in 1993 of the new poverty line of $1.25 a day is $0.92 a day—15% lower than the poverty line of $1.08 a day at 1993 PPP. Or, to put it differently, updating the old 1993 line for inflation in the USA, a poverty line of $1.45 a day in 2005 is obtained (Chen & Ravallion, 2008).

⁴ Use of annual panel data would be better, but we have pooled the data for these periods as international poverty estimates are available only for a few or more years for each country and the estimation of highly unbalanced panel data tends to produce biased estimates.

⁵ This measure is constructed from four binary dummy variables that codify restrictions on cross-border financial transactions that are reported in the IMF's Annual Report on Exchange Arrangements and Exchange Restrictions. Chinn and Ito reverse these binary variables—so that they are equal to unity when capital account restrictions are non-existent—and derive the first principal component, which is their summary measure (Baltagi et al., 2009, p. 288).

⁶ Estimates of the Gini of land distribution (for different years during the 1970s and 1980s) were taken from Deininger & Squire (1998).

⁷ The data are available from http://iresearch.worldbank.org/PovcalNet/povcalSvy.html (accessed 27 December 2009).

⁸ The full data are available at: http://info.worldbank.org/governance/wgi/index.asp

⁹ We have also used Lane & Milesi-Ferretti's (2007) index of financial globalization, defined as the share of a country's foreign assets and liabilities in GDP, and obtained similar results. See Baltagi et al. (2009) for a detailed discussion of the pros and cons of financial openness indices. The Chinn & Ito (2006) index is available at: http://www.ssc.wisc.edu/ ∼ mchinn/research.html

¹⁰ We used the Gini coefficient of income distribution and obtained similar results. The results will be furnished on request.

¹¹ An important issue here is whether the same specification is justified for different institutional quality indicators. This is contentious as it could be argued that voice and accountability may have little to do with investment decisions and openness, while the remaining indicators (e.g. the rule of law, control of corruption, political stability) may be closely linked to them and consequently to income levels; but, given the limited knowledge of how institutions evolve and interact with growth and the distribution of gains, we have opted for the same specification as a first approximation.

¹² The physical isolation index is based on the proportion of a country's population that lives less than 100 km from a coast (McArthur & Sachs, 2002). So the higher the proportion of coastal population in a country, the less isolated it is.

¹³ We have computed results with the poverty gap ratio as well. As similar results were obtained, these are not discussed here to avoid making the present paper unwieldy.

¹⁴ As suggested by Greene (2000), the order and rank conditions for identification are met trivially for the model with many predetermined or exogenous variables. The order condition is that the number of exogenous variables excluded from a particular equation is at least as large as the number of endogenous variables on the right-hand side of that equation, while the rank condition is a sufficient condition for the uniqueness of the solution. Given that we have assigned at least one exogenous or predetermined variable included as an identifier for every equation (e.g. European settlers' mortality rate for the institution equation), both order and rank conditions are met for all the equations in the system (see Greene, 2000, for details). However, caution is needed to interpret the results in case the instrument is not statistically significant in the equation.

¹⁵ We also tried Baltagi's (1981) EC (error correction) 2SLS random-effects estimator and obtained similar coefficient estimates. See chapter 7 of Baltagi (2005) for technical details of 2SLS and its variants applied to panel data.

¹⁶ The income equation could be estimated by dynamic panel data models to address its persistence, for example, as in Guariglia & Poncet (2008). We have, however, applied the dynamic panel data model only to finance.

¹⁷ As noted earlier, the higher the proportion of the coastal population, the greater the trade openness.

¹⁸ We have tried the specification where the MLD index of income inequality is instrumented by the land Gini, that is, six equations are estimated by 3SLS for 18 countries for which the data are available. It is noted that the land Gini has a positive and significant effect. An important change in the results, however, is that the MLD index ceases to be significant in the poverty equation in most cases. That this change is a consequence of the small number of countries cannot be ruled out. We observed a broadly similar pattern of results to the previous case. Details can be furnished on request.

¹⁹ There are other ways of deriving the poverty elasticity of growth. For example, Ravallion (1998) takes into account the fact that the elasticity depends on time series changes in inequality. See Heltberg (2002) for a review of the literature.

²⁰ We have obtained broadly similar results by using static panel data model with the same specification.