The efficiency of the cross-entropy method when estimating the technical coefficients of input–output tables

Abstract

Updating or estimating regional input–output tables is a challenging task addressed with non-survey methods. These can be classified into two groups: location quotient (LQ) methods and constrained matrix-balancing methods. This paper focuses on the second group and, specifically, on the performance of the cross-entropy method (CE). The most important finding is that the RAS method slightly outperforms the CE method on average, but its efficiency varies greatly from country to country. On the contrary, the performance of the CE method is more stable over countries and time. More interestingly, a fair implementation of the CE method boosted by the Flegg location quotient (FLQ) method outperforming the competing CE approach in terms of accuracy.