high crime, or because of managers’ inherent preferences for areas that are personally
familiar to them. Future work detailing the size and sources of firms’ consideration
sets when making location decisions, and improved description of the actual decision
process used to choose an element from the consideration set, would be useful.
Milton Friedman’s as-if methodology argues that it is acceptable to use an incor-
rect model of consumers and firms’ decision processes as long as it predicts behavior
accurately. As-if models may be misguided, though, for purposes of designing policies
to modify behavior. As every student of statistics learns, one can always add param-
eters to achieve arbitrarily good levels of fit with respect to a particular data set. But
fit is not prediction. Prediction requires generalization out-of-sample, often beyond
the current parameter range. Insofar as policy makers working on urban development
wish to achieve outcomes significantly outside the status-quo range of behaviors, they
will require empirically grounded behavioral models with sufficient psychological re-
alism to achieve accurate descriptions of firms and consumers’ decision processes.
27