Estimation of the value-added/intermediate input substitution elasticities consistent with the GTAP data

Abstract

Elasticities are often a combination of expert decisions and literature estimates— many of which are outdated. Previous efforts have focused on estimating the most commonly used elasticities in economic models (e.g., the Armington elasticity of trade); however, several elasticities still have little empirical basis. The elasticity of substitution between intermediate inputs and value-added is one example, but this elasticity is quite important as it governs producers’ production regimes across sectors and regions reflecting their level of efficiency. We examine and estimate this elasticity for one of the most widely used CGE models (parameter ESUBT in the GTAP model), using the latest five datasets available (2004, 2007, 2011, 2014, and 2017) in the version 11 GTAP database. Our work finds that the default value of zero in GTAP does not reflect the behavior implied by the data. Using our estimates, we propose a set of new values for the short run (about one year), two medium runs (three years and six years) and the long run (i.e., infinite time horizon). We demonstrate the importance of our new estimates using a scenario from the EU Farm to Fork policy where we find that using the estimated elasticities leads to much milder market and welfare impacts, and that these effects are further dampened as the time horizon of the simulation increases.