Indexing of Technical Change in Aggregated Data

Abstract

The Baltagi–Griffin general index of technical change for panel data has earlier been applied to aggregated data via the use of period dummy variables. Period dummies force modeling into estimation of the latent level of technology through choice of dummy structure. Period dummies also do not exploit the full information set because the order of observations within periods is ignored. To resolve these problems, I suggest estimating the empirical equation for all possible structures of the dummy variables. The average over the different dummy coefficient estimates provides an index of technical change. More generally, the method estimates a general, model-free trend in linear models. I demonstrate the method with both simulated and real data.