Constrained optimisation applied to Input/Output Table regionalisation with scarce data: application to La Réunion (France)

Mardi | 2018-06-12
Salle des thèses 16h – 17h20


Local data scarcity may lead Input-Output (I/O) analysts to build aggregated subnational models. Based on the case of the French economy, we claim that local I/O tables should instead be as detailed as the national ones. Any local data – never mind how aggregated – is more thoroughly used to adapt existing detailed national tables, than in an oversimplified model of the local economy. Using a deterministic approach based on industry output disaggregation and constrained optimisation, we successfully apply these recommendations to the construction of I/O tables for the French overseas department of La Réunion. Refining these tables by using a time series of economic accounts lead to mixed results. The overall model fitness to local accounts improved significantly, but at the expense of some instabilities in the I/O coefficient estimates. We explore several plausible explanations for these instabilities, and suggest that more refined probabilistic or Artificial Intelligence methods could alleviate them.