In many empirical studies, theoretical models are tested using cross sectional data where the dependent variable is observable only at one specific point in time. However, even simple models can often refer to just one of the possible outcomes of an irreversible conditional choice made sometimes in the past. This can cause serious: (i) truncation problems due to the unobservable future choice; (ii) sample selection of the eligible unit whose choice is observable; and (iii) censoring of the subset of outcomes. Moreover, the current level of the dependent variables could be the consequence rather than the determinant of the status of the unit, generating endogeneity and misspecification.
This paper demonstrates how a simple multinomial selection rule could account for all these problems and how the cross-sectional nature can be combined with the time specification when the dependent variable is observable only at one specific point in time. The validity of this approach is demonstrated over a cross section of firms, using the firm's investment decision conditional on having in the past used at least one unit of the new technology and not having completed the replacement process.
|