Abstract
Choice models are a key tool of empirically oriented rational choice researchers. However, most researchers do not have the information they need to adequately test their hypotheses and rely solely on individual characteristics of the choosers. The lack of information on the options and constraints parents face may lead researchers to conclude that choices are not rational rather than to question whether they have the appropriate data. This issue is particularly important for researchers interested in modeling child-care choice, since the characteristics of care are more amenable to policy intervention than the characteristics of persons. This paper estimates models of choice of child-care arrangement that include measures of price, quality and availability obtained directly from parents in the National Child Care Survey of 1990. This paper compares alternative estimates of the associations of these characteristics with mode choice. Adding care characteristics to models with individual characteristics improves the explanation of parental child-care choices. Using parent reports of the characteristics of arrangements they used and those they did not use produces results more consistent with expectations than estimates based on predicted values. The results also suggest that having parent reports on the characteristics of alternatives not used improves the precision of estimates of the associations between prices and child-care mode choices. Correcting for selectivity does not appear to be a reasonable substitute when such information is lacking.
