This paper has a two-fold contribution. First, it examines the importance of accounting for (in)efficiency in the estimation of primal production function on the input elasticities, technical change, and calculation of returns to scale. Second, it applies a variant of the rolling regression technique to identify time-varying input elasticities, technical change, and return to scale. Empirical application to the Asian agriculture sector using Food and Agricultural Organization data from 1961-2005 indicates returns to scale are underestimated by the traditional pooled and panel models. Further, the time-varying estimates of input elasticities, technical change, and returns to scale indicate variations with each additional year of information.