Examination of Alternative Heteroscedastic Error Structures Using Experimental Data

Impacts of alternative specifications for heteroscedastic error structures are examined by estimating various production functions for corn in Central Texas. Production- and profit-maximizing levels of inputs and the shape of the profit equation obtained from models not corrected for heteroscedasticity differed from those obtained from models corrected for heteroscedasticity. Using the profit-maximizing input levels for each production function gave essentially the same estimated yield and profit, regardless of the specification for heteroscedasticity employed. Differences of up to one-quarter to one-third are noted, however, in the amount of profit-maximizing levels of inputs used, depending on the heteroscedasticity correction.