In this study, stochastic frontier analysis with a time-varying panel data model was applied to analyse short- and long-run cost functions; the short- and long-run cost efficiencies of rice farms in Taiwan from 1980 to 2008 were measured by treating five size classifications in 15 counties in the form of 75 cohorts. The results reveal the presence of economies of scale and artificial inefficiency in these farms. Further, an analysis of the short- and long-run efficiencies of all the cohorts by area, size and year dummy variables indicates that even the areas where agriculture is well developed and topographic conditions are suitable have better long-run cost efficiencies, and large-scale farmland have lower long-run but higher short-run efficiencies. Due to the inefficient use of farmland in the long run, farmers do not have the incentive to expand production scale; they are unable to enjoy the economies of scale, which shows minimum cost. Finally, a random-effect estimation of panel analysis confirms that improvement in the infrastructure development of farmland and the functioning of the rental market of farmland contributes towards increasing both short- and long-run cost efficiencies.
