Efficiency versus bias: the role of distributional parameters in count contingent behaviour models

one of the challenges facing many applications of nonmarket valuations is to find data with enough variation in the variable(s) of interest to estimate econometrically their effects on the quantity demanded. a solution to this problem was the introduction of stated preference surveys. These surveys can introduce variation into variables where there is no natural variation and, as a result, natural experiments are not possible. The problem of no or insufficient variation in naturally occurring data to estimate the effects of interest has led to a large literature on stated preference methods. among the methods developed, two can be linked directly to observed behaviour. unlike contingent valuation questions, these approaches key off of actual choices that individuals have made in the past or are contemplating in the future. Consequently, the consistency of these choices with the responses to stated preference questions can be examined. The two methods are those based upon random utility theory and those based upon demand theory. While the two can be linked theoretically in practice, one either adopts a random utility framework or a demand framework. The demand framework, adopted here, is frequently identified as a ‘contingent behaviour’ approach. The contingent behaviour method was proposed by Englin and Cameron (1996). Their paper suggests focusing on the number of trips an individual might make under different situations rather than how a single choice might vary (random utility model) under different situations. The advantage of the contingent behaviour model is that it includes both the intensive and the extensive margin while the random utility approach focuses solely on the intensive margin. as a result, the contingent behaviour approach can capture improvements with the extensive margin as well as quality reductions on the intensive margin. While considerable effort has been expended examining the functional form, parametric specifications and distributional assumptions used in contingent behaviour studies, no effort has been spent examining the role