The implications of uncertainty for costbenefit analysis remain controversial. One issue raised in the evaluation of welfare compensation tests has been the appropriate use of observable economic behavior in the measurement of agents' willingnessto-pay (or willingness-to-receive). In the context of measure, this question has received considerable attention in the analysis of the welfare impact of price changes under certainty (e.g., Willig 1976; Chipman and Moore 1980; Hausman 1981; McKenzie and Pearce 1982). Moreover, the introduction of uncertainty has raised a number of questions about the appropriateness of alternative welfare measures (e.g., Schmalensee 1972, 1975; Graham 1981). For example, Graham (1981) has argued that there is little basis to recommend either option price or expected for a role of prominence in welfare analysis. The issue rather is whether or not there exists a distribution of individual payments (or receipts) that maximizes aggregate willingness-to-pay and passes the Pareto improvement test, i.e., that leaves no individual worse off and makes at least one individual better off (Graham 1981). But, in the implementation of the Pareto improvement test, how is the aggregate willingness-topay supposed to be measured empirically? More specifically, what is the influence of uncertainty and its temporal resolution on the validity of agent surplus (the welfare triangle given by the area between a compensated choice function and its corresponding price) as a welfare measure? Some progress has been made in this direction by Chavas, Bishop, and Segerson (1986) in the context of option price. However, as shown by Graham (1981), while option price is the appropriate measure of benefits in situations involving similar individuals and collective risk, it may not be the appropriate welfare measure in more general situations. Therefore, there is a need to refine our understanding of how agent surplus can be used as a welfare measure in the general context of a potential Pareto improvement test under temporal uncertainty. The objective of this paper is to investigate how temporal uncertainty can affect the measurement of willingness-to-pay in general welfare compensation tests. Temporal uncertainty corresponds to the situation where new information becomes available over time, which resolves at least part of current uncertainty facing economic agents. This new information can influence future individual decisions as well as individual willingness-to-pay. Particular attention will be given here to the case where income compensation is conditional on the new information that will become available in the future and results in a more efficient allocation of risk among individuals. This paper focuses on competitive agents (e.g., households or firms) for which welfare changes can be expressed in terms of price changes. This includes the analysis of the welfare effects of various government pricing policies. This also includes the evaluation of public projects based on the travel cost method which, under a weak complementarity condition, consists in using price changes of related private goods as an indirect means of measuring the value of a public project (Maler 1974). The analysis is presented in the context of a general two-period model where the agents are uncertain about the state of the world. As suggested by Graham (1981), it is shown that the option price provides a lower bound on