The validity of parametric link functions in discrete choice models of food purchases: empirical evidence from Greek microdata

Parametric discrete choice models are used in agricultural economics to correct outcome equations for selectivity bias. This note assesses empirically the distributional assumptions underlying these models using the HH-test and microdata on food purchases by Greek households. High rates of rejection are encountered, especially for the probit model, suggesting that the computational cost involved in testing parametric link functions is justified.