Welfare Indicators and GDP

The clamor for indicators of well-being after the recession of 2008 was predicated on the idea that Gross Domestic Product (GDP) was an insufficient measure of an economy’s overall performance. Though GDP measure was not designed to serve as a measure of welfare, it can be shown to be informative about well-being. This paper examines the relationship between a set of Organisation for Economic Co-operation and Development (OECD) well-being indicators and household income. Our data consists of 12 years of submissions of well-being indicators by 44 OECD countries and we have approximately 520 observations for each indicator.

We use entropy measures to assess dependency. We find that Mutual Information and Conditional entropy generally show dependency between household income and each of the 18 indicators. For example, the entropy of the subjective indicator “Feeling Safe at Night” is 4.02 bits (of information) and this value derives from the indicator values submitted by countries to the OECD, which reflects the views of their populations. The mutual information between household (HH) income and this indicator is 1.3 bits or 32% of the information about the indicator and the conditional entropy of 2.72 bits, 68% of the total entropy, is the uncertainty remaining. Whether 32% of the information is sufficient is determined by the intended user and should consider the acquisition cost of the remaining 68%. For context, the correlation in this case is 0.69, a moderately strong linear relationship between this indicator and HH income. The important difference between the two perspectives is that the entropy measures are nonnegative, nonlinear, and not constrained by the need to have matched pairs.