Development statistics and correlations: A comment on Hicks and Streeten

Abstract

We agree with many of the substantive conclusions reached by Hicks and Streeten in their paper on ‘Indicators of development: the search for a basic needs yardstick’ (World Development, Vol. 7, 1979, pp. 567-580). The present note is concerned with their use of correlations between selected social and economic variables to reach certain conclusions about the measurement of development, and with the large and puzzling discrepancies between some of the COIrelations given by Hicks and Streeten (writing from the World Bank) and correlations even for the same variables for the same year obtained from the Research Data Bank of Development Indicators of the United Nations Research Institute for Social Development (UNRISD). We report here the results of an enquiry made in an attempt to understand the reasons for the discrepancies. Using cross-national data for seven selected social variables (listed in Table l), Hicks and Streeten conclude that in general there is only a ‘modest’ correlation of these variables with GNP per capita: the average r2 = 0.50.’ Moreover, when the data are ‘disaggregated’ into separate figures for developed and developing countries, the average r2 is only 0.25 (r = 0.49) for developing countries and 0.18 (r = 0.38) for developed countries. Hicks and Streeten note that the relationship of the social variables to GNP per capita tends to be non-linear, and that a better overall correlation would be achieved by using a non-linear function (such as a semi-log function for relating life expectancy to GNP per cupifa).2 But they add: ‘A non-linear function would obscure the fact that the correlation exists, however, only among the middle-income countries.’ Therefore, they conclude, the GNP per head should not be used as an indicator of social development. 1. DEVELOPMENTAL CORRELATIONS AND MIDDLE-INCOME COUNTRIES