Abstract
This article examines the influence of macroeconomic factors on personal income distribution in developing countries using a parametric modelling approach. The technique is based on the selection and estimation of a theoretical parametric model (a Dagum distribution) which fits accurately to the empirical income distributions of the countries examined. The parameters of the model specifically related to inequality are subsequently used as dependent variables in econometric models in order to examine the impact that certain macroeconomic variables (GDP growth, inflation, employment and real interest rates) have on inequality. The results reveal that GDP growth, employment rate and real interest rate are the macroeconomic factors with greater impact in shaping personal income distribution in developing countries.
Acknowledgements
This study benefits from funding support of the projects CSO2011-29943-C03-02 of the Ministerio de Educación, Cultura y Deporte and ECO2012-32178 of the Ministerio de Economía y Competitividad of Spain and of the Instituto de Estudios Fiscales. We thank MariaDolores de Prada for her comments on an earlier version.
Notes
1 A review of these studies can be found in Parker (Citation1999).
2 On the contrary, there is less ambiguity in the relationship between growth and poverty, an area where a large number of studies have been conducted (for a review, see Chen and Ravallion, Citation2010).
3 Garcia and Prieto (Citation2011) provide an exhaustive analysis of the influence of the parameters of the Dagum model on different measures of income distribution.
4 See more details in Dagum (Citation1977).
5 One alternative procedure would be to estimate two different equations, one for each parameter. Yet, this kind of formulation would allow introducing the variance of the parameter estimators, but not the covariance between them.
6 On the contrary, the models proposed by Thurow (Citation1970) and Salem and Mount (Citation1974) did not incorporate the sampling variability of the income distribution parameters estimators.
7 The details of the survey of each country (including name, year, sampling size and geographical coverage) can be consulted in the data section of ‘PovcalNet’ (http://iresearch.worldbank.org/PovcalNet).
8 2005 was a year of widespread economic expansion (according to International Monetary Fund estimates, the growth of the world economy was 4.8% and that of emerging and developing countries was 7.5%).
9 In order to evaluate the robustness of the results, we have estimated the models for two different subsamples of countries classified according to their growth rates. The hypothesis of equal coefficients could not be rejected. The results of the tests are available from the authors on request.
10 These results are very similar to OLS estimations (detailed results are available from authors on request). Note that GLS and OLS estimators would coincide if the sample variability of the Dagum parameter estimates were not introduced in the SUR system. Under our assumptions, the OLS estimator is consistent and the FGLS estimator also, provided that the variance–covariance matrix of the joint errors has been estimated correctly.
11 For example, in the case of the relationship between the Gini index and employment, knowing that the Gini index is a function of the parameters a and p , and as a and p are unknown, we will substitute them with the FGLS estimation of the SUR model. By substituting the values of the employment rate in the expression of the Gini index and maintaining the rest of the variables constant, we can examine how the variations in the employment rate affect the Gini index.