The use of Maximum Entropy to estimate input-output coefficients from regional farm accounting data

This paper proposes the use of the Generalised Maximum Entropy (GME) method to estimate input-output coefficients, which reflect the unobserved alloca tion of farm input accounting costs to the various outputs produced. The GM E method uses Shannon's information criterion as a basis for estimation. Th e performance of the GME method is compared with three other estimation tec hniques: Ordinary Least Squares (OLS), Bayesian estimation, and Linear Prog ramming (LP). The various methods are applied to accounting data from a sam ple of beef-dairy farms in Brittany, France. The analysis shows that the GM E method offers an interesting alternative to "traditional" estimation meth ods. In contrast with the latter, though, the GME method is suitable to han dle easily the problems of singularity, constrained estimation, and zero-ob servations. Moreover, due to its flexibility, transparency and relative eas e of implementation, the GME method is of great value to practitioners. How ever the sensitivity of the GME estimates with aspect to the design of the prior information set needs to be investigated further.