How wrong can you be? Implications of incorrect utility function specification for welfare measurement in choice experiments

Abstract

Despite the vital role of the utility function in welfare measurement, the implications of working with incorrect utility specifications have been largely neglected in the choice experiments literature. This paper addresses the importance of specification with a special emphasis on the effects of mistaken assumptions about the marginal utility of income. Monte Carlo experiments were conducted using different functional forms of utility to generate simulated choices. Multi-Nomial Logit and Mixed Logit models were then estimated on these choices under correct and incorrect assumptions about the true, underlying utility function. Estimated willingness to pay measures from these choice modelling results are then compared with the equivalent measures directly calculated from the true utility specifications. Results show that for the parameter values and functional forms considered, a continuous-quadratic or a discrete-linear attribute specification is a good option regardless of the true effects the attribute has on utility.We also find that mistaken assumptions about preferences over costs magnify attribute mis-specification effects.

Incorrectly accounting for preference heterogeneity in choice experiments: Implications for welfare measurement

“…However, little attention has been paid to examine the effects on welfare estimates of mistaken assumptions about the nature of preference heterogeneity. As pointed out by Torres et al (2011), the interest in analysing the bias and efficiency of welfare estimates within non-market valuation has been mainly centered on investigating, through MC analysis, issues such as the i) specification of the recreation demand function in travel cost models, ii) WTP elicitation in the contingent valuation approach, and iii) experimental design under different utility specifications in choice experiments (Kling, 1987;1988;1989;Adamowicz et al, 1989). Investigations of the effects of decisions over appropriate nesting structures in multiple site recreation demand models represent a related area of concern (Herriges and Kling, 1997;Kling and Thomson, 1996), which also makes use of MC analysis.…”

“…To our knowledge, only Torres et al (2011) attempt to examine the errors from mistaking the way of explaining heterogeneity in CEs. In particular, and with a focus on different attribute specifications, they analyze the effects on welfare estimates from i) correctly assuming RPL taste heterogeneity but mistaking parameter distributional assumptions, and ii) incorrectly assuming RPL taste heterogeneity when it is driven by the scale factor.…”

Do we care about sustainability? An analysis of time sensitivity of social preferences under environmental time-persistent effects

“…The model is a random parameter logit (RPL) to take individual-specific preferences into account, which requires assuming that parameters are random. In our model, only the cost parameter has been specified as random and it has been assigned a lognormal distribution to constrain the coefficient to have the same sign over all individuals (Torres et al 2011). Taking into account that individuals make their choices of their preferred alternatives in the face of risks of impacts' occurrence on 'specialist' bird species, the utility function specification used for estimation purposes, which is illustrated in Equation 2, follows the expected utility (EU) theory approach.…”

A new approach to specify and estimate non-normally mixed multinomial probit models

“…3 In the context of continuous mixing distributions, the normal distribution has been used quite extensively in the past. However, several studies (see, for example, Amador et al, 2005, Train and Sonnier, 2005, Hensher et al, 2005, Fosgerau, 2005, Greene et al, 2006, Balcombe et al, 2009, and Torres et al, 2011 have underscored the potentially serious misspecification consequences (in terms of theoretical considerations, data fit, as well as trade-off evaluations) of using the normal distribution. In particular, the symmetric nature of the normal distribution, when combined with mean values that may not be too far away from zero, implies that a significant fraction of individuals may have an unexpected sign on variables (such as a 3 Note here that discrete distributions may also be used for the mixing.…”