Global Approximations to Cost and Production Functions using Artificial Neural Networks

The estimation of cost and production functions in economics usually relies on standard specifications which are less that satisfactory in numerous situations. However, instead of fitting the data with a pre-specified model, Artificial Neural Networks (ANNs) let the data itself serve as evidence to support the model's estimation of the underlying process. In this context, the proposed approach combines the strengths of economics, statistics and machine learning research and the paper proposes a global approximation to arbitrary cost and production functions, respectively, given by ANNs. Suggestions on implementation are proposed. All relevant measures such as Returns to Scale (RTS) and Total Factor Productivity (TFP) may be computed routinely.