It has recently been claimed that economic growth depends on social development, but not vice-versa. This paper reconsider the nature of such links. It is argued that those claims find only weak support in the data and that the theoretical framework is too fragile to justify a satisfactory economic interpretation of the estimates. In a paper recently published in World Development, Newman and Thomson analyzed the causal relationship between economic development and social development. The former was measured in terms of GDP per capita at constant 1975 dollars, adjusted for purchasing power parity (RGDP); the latter was proxied by Morris' Physical Quality of Life Index (PQLI), which takes into account literacy, infant motility, and life expectancy at age one. In particular, the paper attempted to identify the existence of a "trickle-down effect" (whereby social development would lead to economic growth) and/or of a "trickle-up effect (whereby economic growth would make welfare' improvement easier and faster). This paper raises two sets of problems for the interpretation of the regression coefficients: one about the methodology used by the authors, the other about the interpretation of the estimates. From what has been argues, it follows that during 1960-80: (a) High economic growth occurred in countries characterized by initially low economic development, although the absolute increase was higher, the higher the level of RGDP at the beginning of the period, in fact, a larger share of the absolute increase in income was independent of its initial level; (b) The impact of social development on economic growth, although statistically significant, was quite marginal; (c) Fast social development took place in countries characterized by low initial welfare. This was because the absolute increase in social development was claimed to be independent of initial welfare and other variables. Omitted-variable misspecification cannot be ruled out. All this cannot be considered evidence in favor of a trickle-up, nor against the trickle-down, hypotheses. Setting aside all issues about the reliability of the proposed "causality test," Newman and Thomson's data just show that over 1960-80 rich countries were probably in a better position to take advantage of the large increase in world demand, whereas poorer countries grew faster primarily because of their low initial GDP levels. The fact that the level of social development is statistically significant might in fact point out that the RGPD variable is what Newman and Thomson refer to as an unsatisfactory "economic indicator." The evidence presented by Newman and Thomson also does not necessarily undermine the trickle-down hypothesis. For example, if foreign aid did play a favorable role in social development, economic development may well become significant. In other words, omitted variable misspecification must be ruled out before any conclusion is reached in this respect. A Reply By Barbara A. Newman and Randall J. Thomson Enrico Colombatto's critique of out paper raised several issues to which we wish to respond. His problems with the study seem to fall into three major categories: his discussion of the "causal" properties of the model; his recalculation of out results; and his discussion of possible specification error in the model. We address each of these points and respond to several other of his observations. Our paper "economic Growth and Social Development: a Longitudinal Analysis of Causal Priority" was conceived as an attempt to examine the relationship between economic growth and social welfare development in less-developed countries. We believe that our results are interesting, both to theoreticians and to policy makers, in that they provide support for a "trickle-up" approach to development, and do not provide any support for the more widely held "trickle-down" model. We are not claiming to have found the definitive answer to the causal priority question, but we hope that our study will provoke further research in the area, and persuade researchers to reconsider their assumptions of temporal ordering of these two variables in cross-sectional designs. In light of this, we welcome any additional discussion or comments which will help clarify the issues involved.
Bibliography: 1991 1441-1444 World Development, Vol. 19, No. 10
