Stefano Tarantola, Elena Giachin Ricca

Composite indicators of well-being: the relative importance of weights

Composite indicators are becoming mainstream tools for benchmarking and policy making. The standard approach to build composite indicators consists in combining an underlying set of indicators Xi through a series of treatments and algebraic manipulations. The result of this process is a set of composite scores Yj for the various elements (eg universities, regions, cities, etc.). Quite often, a weighted average of the Xi s – through a set of weights wi, is used to obtain Y. We call these weights subjective as they are customarily assigned subjectively by the developers.
Composite indicators can also be obtained by estimating micro-econometric models from given empirical data for each individual i, belonging to country j, at time t: .
Here life satisfaction data are used as the dependent variable. Zjt are well-being explanatory variables at macro level. Xijt are other explanatory variables at individual level, eg socio-demographic factors. is an error term and the coefficients a, ß, and ? are estimated. The well-being composite indicator for country j at time t, is defined as , where ? are the estimated weights. We call these weights objective because they are obtained using a statistical estimation procedure from given observations.
We compare the standard approach with that based on objective weights and test the robustness of the country composite scores in terms of the various sources of uncertainty (weights plus others). The robustness analysis can have relevant policy implications as it helps us to know whether the two approaches provide similar policy conclusions.