Issues in technical efficiency measurement with an application to the domestic airline industry
Semenick, Ila Margariet
Doctor of Philosophy
This dissertation consists of three essays with technical efficiency as the focus. The first essay uniquely integrates three literatures: technical efficiency, comovement and convergence. Using a panel data set of eleven domestic airlines followed quarterly from 1970 to 1990, time series technical efficiency scores are determined using the linear programming approaches of Data Envelopment Analysis and Free Disposable Hull and the parametric approach of Stochastic Frontiers. Interesting patterns, such as the dramatic decline in Eastern's efficiency prior to its demise, emerge. The market structure theory stating that an increasingly competitive environment fosters efficiency, is examined by testing these scores for comovement and convergence. Of those technical efficiency series exhibiting unit roots, a majority of firm pairs do not reject the null hypothesis of long run comovement (i.e. cointegration). Furthermore, tests of convergence support the hypothesis that firm performance is becoming less disperse over time--firms are becoming more alike as efficiency advances diffuse through the industry. The second essay derives a semi-parametric efficient estimator which allows firm specific effects to be orthogonal to certain regressors and correlated with other regressors. The multi-product nature of the airline industry is modelled and the effects are allowed to be correlated with the long run product mix. Efficiency gains are found to be as high as 64% over the first-step consistent within estimator. The final essay presents evidence linking two measures of firm performance: technical efficiency and stock market returns. Innovations in linear programming technical efficiencies and industry adjusted returns are found to exhibit significant correlation only in the two months following each quarter-end. This period corresponds to the time frame during which data necessary for determining the scores becomes available. Further, using the linear programming scores to identify firms which improve or decline in performance, an investment strategy, where the former is bought and the latter is sold short, is adopted. This strategy produces significantly different returns between the two portfolios over the same two month period. Thus, strong support is found for the hypothesis that a company's stock market performance is linked to its resource utilization ability.