Transpn. Res.-B Vol. 21B, No. 5, pp. 339-357, 1987 0191-2615/87 $3.00+.00 Printed in CnP.atBritain. ~ 1987 Pm~m'no,aJournals Ltd. USING A HICKSIAN APPROACH TO COST-BENEFIT ANALYSIS IN DISCRETE CHOICE: AN EMPIRICAL ANALYSIS OF A TRANSPORTATION CORRIDOR SIMULATION MODELt TIMOTHY D. HAU Department of Economics and Centre of Urban Studies and Urban Planning, University of Hong Kong, Hong Kong (Received 3 October 1985; in revised form 20 August 1986) Abstract--The general equilibrium effects of alternative transportation policy proposals are analyzed using a multi-modal, benefit-cost model of demand and supply within a discrete choice framework. Using the expenditure function as an empirical construct yields Hicksian consumer's surplus measures, which are usually not directly observable. Despite the absence of data on the complete budget and only mode choice data, the procedure we use to prove the stochastic analog of Roy's Identity identifies the entire indirect utility function, and hence the explicit expenditure function. The model is applied to a corridor simulation model of Interstate 580 of the San Francisco Bay Area. Travel demand is calibrated using multinomial logit on a sample of work-trip commuters. De- tailed modal costs are estimated and entered as parameters of the demand model. The Scarf algorithm equilibrates the analytic disaggregated demand model and a parallel analytic supply model. Our fore- cast is based on a synthetic sample of households generated by an efficient program that uses census data. The benefits of policy alternatives, such as marginal (resource) cost pricing are computed and discussed. 1. INTRODUCTION Disaggregated, behavioral demand models have long been popular, especially in the transpor- tation literature. Interest in the application of conventional cost-benefit analysis to qualitative choice models for zero--one outcomes intensified with the appearance of Small and Rosen's paper (1981). In their important contribution, the expenditure function was used as a theoretical construct, as opposed to the empirical construct that is used here.~; We ask how much money a traveller is willing to pay for the benefit of having lower travel cost or time. The answer is given by the economic concept of Marshallian consumer's surplus [see Agnello (1977) and Jara-Diaz and Friesz (1982) for details]. Alternatively, we ask how much money one is willing to pay so that he/she is just as well off as before. The answer to this second question yields the Hicksian compensating variation--an exact consumer's surplus measure. Interest in various consumer's surplus measures for project evaluation were heightened more recently by Willig's (1976) path-breaking paper and other important papers such as Hausman's (1981). Yet in practice, the use of Marshallian consumer's surplus measures assumes the constancy of the marginal utility of income. A Hicksian approach to cost-benefit analysis not only circumvents the Marshallian path-dependency problem but produces theoretically correct mea- tAn earlier version of this paper was presented at the 59th Meeting of the Western Economic Association International, Las Vegas, June 24-28, 1984. Part of the research was done while the author was associated with the University of California, Davis. Faculty research grants obtained therein are gratefully acknowledged. ~The main difference between Small and Rosen (1981) and our approach here can be seen in the respective, implicit compensation objectives involved [see Hau (1985)]. Small and Rosen derive an expression for the amount required to compensate an individual for a change in price or quality given that the individual chooses to consume the affected good. Their compensation scheme is based on the change in actual utility relative to the base case whereas our compensation objective is based on the change in expected utility (to be discussed later). Hence, their ex post approach contrasts with the ex ante approach adopted here. Small (1983) obtains the difference between the compensated and Marshallian consumer's surpluses by using a modified version of the Slutsky equation. ZRB21: S-A 339