Explaining Production Heterogeneity By Contextual Environments: Two-Stage DEA Application to Technical Change Measurement
One of the most important objectives in eciency analysis is to investigate the rela- tionships between production decisions and their contextual environments like geographical regions, production time periods, modes of production, or policies and regulations. Using the measurement of technical change as a template, the study presents a general framework to better understand varying production decisions under dierent time periods by showing how such production heterogeneity can be attributable to the dierences in time-specic technological frontiers at industry level and the dierences in the prevalence of technical ineciency at producer level. In DEA, a leading non/semi-parametric frontier estimation method, these dierences can be analyzed through decomposing Malmquist produc- tivity index (MPI) into technical change (TC) and technical eciency change (TEC) respectively. The decomposition approach falls into the non-Hicks-neutral TC estimation as the mean distance measures among time-specic frontiers, which is generally less restrictive than the Hicks-neutral TC estimation as an intertemporal-shift component of the frontier specication under xed substitution patterns across time periods. The method is more generally applicable to the comparisons between any two dierent contextual environments, including before and after a policy intervention, by which a sample can be partitioned. To make the existing method more empirically accessible and appealing, the study pro- poses a regression-based MPI decomposition that overcomes its limitations, or the need of balanced panel data and the lack of control for potentially confounding non-production factors. The proposed methodology is demonstrated with an empirical application using data from the Schedule F Tax returns of 62 dairy farmers in Maryland during 1995-2009. For conventional, connement dairy operations, the preliminary results under preferred specications show a 26.4%/decade expansion in technological fron- tier, accompanied by a 6.3%/decade decline in the mean technical eciency levels (i.e. increases in the prevalance of technical ineciencies). The indicators for farm ownership and o-farm income are associated with a 4.5% increase and a 5.8% decrease in technical eciency respectively. Higher sea- sonal rainfalls and temperatures, except for winter rainfall and summer temperature, are associated with larger technical feasibility in a given year.