Revisiting Methods for Estimating Interregional Input-Output Accounts: It's Not Just About Trade Flows
Ana Lúcia Marto Sargento,Michael L. Lahr,João Pedro Ferreira,Fernando de la Torre Cuevas
DOI: https://doi.org/10.1016/j.strueco.2024.04.004
IF: 5.059
2024-04-19
Structural Change and Economic Dynamics
Abstract:A basic underlying assumption in most of the research to date is that intermediate industry accounts of the economies in multiregional input-output (MRIO) tables exist and are accurate. In fact, if they exist at the subnational level, such accounts are, at best, roughly estimated and predicated on far less empirical information than is available for economies of nations. Moreover, intra- economy intermediate-industry flows are typically larger than the set of a region's commodity in- and out-flows. So, if intermediate industry flows in a set of MRIO accounts are noticeably mis-estimated, it follows that interregional trade coincidentally derived using them must be even more conspicuously in error. We hypothesize as more information is used to estimate MRIO accounts, the better the estimates should be. We start our experiment by consolidating 2019 FIGARO accounts of the 27 member states of the European Union, while maintaining sectoral detail, to produce a "national account". We then test several approaches to constructing MRIO tables. The approaches distribute interregional trade fully by receiving industry, as in FIGARO, as well as strictly in the form of a diagonalized matrix as if the commodity inflows are competitive imports. To do this, both a gravity model and RAS are applied to each approach. We then test to see how well the approaches estimate main features of FIGARO's MRIO accounts and detail a rather consistent ranking of the relative accuracy of them. We also find that the level of error inherent to the estimated MRIOs is markedly similar across approaches, particularly for multipliers. Further, relaxing interregional trade to a diagonalized matrix tends to add very little error. The approach that uses the least data is, however, markedly worse in replicating countries' direct requirements matrices and Leontief inverses, which suggests its use in a more-limited set of applications.
economics