In the previous two chapters, we discussed ways of integrating first-order information to get exact or approximate measures of utility (and welfare) for large changes. When the associated variables are traded in markets, we can recover the associated measures from sufficiently rich demand data. However, it is highly implausible that we can collect demand information detailed enough to perform the required numerical integrations. Moreover, since we are primarily concerned with allocating collective (hence, unmarketed) goods, market demand information alone (no matter how detailed) will not be sufficient unless we can generalize the spanning procedures of Chapter 13. This chapter discusses ways of recovering (or in most cases estimating) required information from observable data and using resulting measures to evaluate and compare alternative large projects.

We first look at methods for recovering surplus or willingness to pay for collective goods from sufficiently rich information on market demand. However, since “full information” is virtually never available, we need to look at ways of estimating or otherwise learning about utility. A standard estimating procedure involves imposing parametric restrictions on the form of utility (or demand) functions and estimating the parameters econometrically. We explain and evaluate this method in Section 16.2. Alternatively, we might try to learn about utility by designing a planning mechanism that induces agents to reveal their true surplus. Of course, we know from our discussions in Chapter 5 that the potential for solving broad classes of economic problems with a single planning mechanism is quite limited.