On specifying double-hurdle models

The double-hurdle model (DHM hereafter) has been used in microeconometrics to analyse a wide range of individual and household commodity demand. Indeed, a search of ‘‘EconLit’’ (the American Economic Association’s electronic bibliography of economic literature) reveals 30 citations for ‘‘double hurdle’’ dating back to 1989, although the DHM literature dates initially from Cragg’s seminal paper [6] published in 1971. Important contributions to the DHM literature include Jones [16], in which the demand for cigarettes is modeled, and Blundell and Meghir [3], which was concerned with the labour supply of married women. Other fields in which the DHM has been applied include finance (e.g., Dionne et al. [9] examine credit-scoring) and sociology (e.g., Zorn [35] examines legislative response to court rulings). An interesting discussion on inferential uses of the DHM in the context of recreational activities appears in Shonkwiler and Shaw [28]. The DHM has been applied to infrequency-of-purchase contexts; e.g., Deaton and Irish [7] report on p-Tobit models of household consumption of tobacco, alcohol, and durable goods-the p-Tobit model is a special case of the DHM, see Maki and Nishiyama [24]. However, the majority of commodity demand applications of the DHM cater for situations in which individual preferences generate zero observations due to corner solutions as well as abstentions from consumption. Generally, the data used in most applications have been cross-sectional on individuals or households; however, recently Labeaga [19] has applied the DHM to estimate reduced form parameters in a structural model for tobacco demand using panel data.