Abstract:
This paper examines the selection of data source and econometric technique for studies of banking efficiency using translog cost functions. We examine the use of Seemingly Unrelated Regression estimation for a cost function, as against estimation using Ordinary Least Squares. Choice of cost data to feed to the estimation is also important, and we find that use of wage and interest data may sometimes be superior to cost data inferred from bank accounting information. Lastly, we discuss filtering of data, where some observations may contain erroneous or noisy data.
