Composite competitiveness indicators with endogenous versus predetermined weights: An application to the World Economic Forum's global competitiveness index

The purpose of this paper is to examine how the ranking of countries based on the World Economic Forum's (WEF') competitiveness index is changed when the underlying primitive data dimensions of this composite index are aggregated using weights that are endogenously determined for each country, instead of aggregated using the WEF's fixed set of weights applied to all countries.

The paper presents a method based on data envelopment analysis to determine weights for aggregating the underlying primitive data dimensions of any composite indicator. The approach determines endogenously the “best” weights a given observational unit (e.g. country) on the basis of its revealed performance on each primitive sub‐dimension underlying a composite index. The ranking of countries based on the values of a composite competitiveness index that uses the proposed endogenous weight method is then compared to the ranking based on the WEF's competitiveness index for the year 2006. The rankings are then compared and assessed to determine if the observed difference in the rankings are statistically significant.

A comparison of the ranking of countries on the basis of the value of each index reveals that countries do undergo a change in their competitiveness rank when endogenous weights are used. The results suggest the WEF's competitiveness index, which uses the same fixed weights applied to every country (or group of countries), creates a bias that favors countries that score high on the “technology” sub‐dimension of the index.

The study presents an alternative to the current practice of using a fixed set of weights applied uniformly to the basic unit of analysis. The method serves as a starting‐point for further research on the biases created by different weighting schemes to construct a composite indicator that aggregates primitive data, with the resulting composite index values then used to rank entities.

The method to determine endogenously the weights to be applied to each unit of analysis when constructing a composite indicator is novel and has wide applicability to the general issue of comparing performance across different units of analysis based on a composite index of performance (i.e. benchmarking).