Improving estimates of economic parameters by use of ridge regression with production function applications

Ridge regression is a promising alternative to deletion of relevant variables for alleviating multicollinearity and can provide smaller mean square error estimates than unbiased methods such as OLS. However, ridge estimates can also be unreliable and misleading under certain conditions. To avoid erroneous conclusions from ridge regression, some prior knowledge about the true regression coefficients is helpful. A theorem on expected bias implies that ridge regression will give much better results for some economic models, such as certain production functions, than for others because of smaller expected bias.