Estimating firm technical efficiency using alternative frontier function approaches: an application to farrow to finish hog production units

The purpose of this study was to estimate the relative technical efficiency of individual farrow-to-finish swine production units and to analyze the relationship between firm technical efficiency and selected swine production practices. Study areas included 11 states in the North Central region and seven states in the Southeast region of the U.S. hog enterprise production data were obtained from a cross-sectional cost of production study conducted by the Economic Research Service, U.S.D.A. in 1981. Information available for 216 farrow-to-finish swine units in the North Central region and 339 farrow-to-finish swine units in the Southeast region was used to derive the empirical data on annual output and resource use of each swine unit. Technical efficiency was estimated for each firm using alternative frontier production function approaches. The frontier function approaches used were (1) the "Farrell" linear programming approach to estimating multifactor productivity measures of technical, scale, and input congestion efficiency, and (2) a statistical composed error frontier approach to measuring technical efficiency relative to a random stochastic frontier. The resulting estimates of technical efficiency were used as dependent variables in explanatory regression models which related technical efficiency among firms to specified production characteristics. Production characteristics specified as independent variables were sow herd production intensity, level of confinement, type of management,type of farm business organization, type of manure handling practices, and type of feed and feed processing practices. Linear programming measures of technical efficiency permitted the derivation of estimates of scale efficiency and congestion efficiency for each hog unit. Mean technical efficiency for hog units in each sample was higher as the frontier function was altered to allow variable returns to scale and weak input disposability in the frontier relation-ship. Results showed scale and congestion inefficiency to be minor compared to technical inefficiency among swine units in each regional sample. The statistical, stochastic frontier approach gave results showing that technical efficiency was higher for larger size hog units and the estimates of the parameters of the frontier indicated the elasticity of production was less than one for each factor and for all factors collectively. Results derived from regression models using the deterministic measure of technical efficiency showed very little explanatory power and in many instances the signs, magnitude of the coefficients obtained for explanatory variables were contrary to expectations. Estimates from the regression models using the stochastic frontier indicated a moderate degree of explanation for the variation in technical among firms and estimated coefficients were more plausible. The existence of technical inefficiency in the hog production process seems quite well documented, but the determinants of hog farm technical efficiency are still unclear based on the results of this study.