In this paper, the authors seek to evaluate the effects of the cross subsidy-stratification system for public utility services on housing prices in Bogotá, Colombia. In addition, they use their estimates to assess the extent to which the distortions in housing prices associated with the subsidy-stratification system affect final subsidy beneficiaries. Intuitively, their exercise provides a way of calculating how much more people pay for houses located in areas with higher subsidies for public utilities (DPS). Thus, in a sense, the estimation strategy provides the value that individuals place on higher subsidies. Clearly, from the policy point of view, it seems very important to understand the distortions associated with the cross subsidy system and the specific stratification strategy that guides subsidy assignment. From both these perspectives, this paper is an important contribution.
The basic estimation strategy is to estimate a relatively standard hedonic price model in which public utility subsidies received or paid by a given dwelling have an effect on the price of the house, controlling for a variety of observed characteristics of the house and the neighborhood. The key issue dealt with in the paper is the endogeneity of the subsidy in such an equation. In other words, estimating the causal effect of DPS subsidies on housing prices is difficult because houses that receive high subsidies are located in lower socioeconomic strata (that is, poorer neighborhoods) and thus are associated with lower prices. Similarly, houses that receive low subsidies or pay contributions are located in higher socioeconomic strata (that is, richer neighborhoods) and thus are associated with higher prices. Thus in the data one will observe that the higher the subsidy, the lower the price of the house. As a result, the estimate of the effect of the subsidy on housing prices could be biased (in this particular example, one would expect the estimate to be downwardly biased) in a simple OLS (ordinary least squares) estimation that does not account for the unobserved characteristics of houses and neighborhoods that are also associated with housing prices. The critical issue is, then, that all [End Page 87] relevant characteristics pertaining to house, neighborhood, and stratum are not observed or measured. Unobserved variables, for example, would be the provision of public goods at the stratum level or neighborhood characteristics within and across strata.
To deal with this issue, the authors implement a type of regression discontinuity design (RDD) by taking advantage of the stratification system by which households are assigned to one of six socioeconomic strata that are then used to target differential subsidies for public utility services. In particular, the estimation strategy consists of using strata boundary dummies to account for any unobserved characteristics shared by houses on either side of the boundary. Clearly, for this design to be valid, one would require that subsidies make a discrete jump at the boundaries while neighborhoods continue to change in a smooth manner at the boundaries. Intuitively, the authors claim that while there is significant heterogeneity across strata and, as expected, homogeneity within stratum the location of the exact boundaries can be arbitrary to a great extent because a very large number of dwellings were assigned to very few groups (six in total). Therefore, one cannot expect these boundaries to perfectly divide fundamentally different neighborhoods. Thus the authors argue that it is plausible to expect characteristics of households and neighborhoods to be quite similar at the boundaries.
However, one must be aware of the fact that the key assumption of the regression discontinuity design would be plausible in cases in which boundaries are determined according to characteristics uncorrelated with the independent variables in the hedonic price equation (and, in particular, the subsidy, which is the variable of interest) and the unobserved error term. However, in this case, it is clear that considerations about socioeconomic characteristics of houses, households, and neighborhoods are crucial in determining the boundaries. The results reported in the paper indicate that, in fact, even for houses located very close to the boundary, at least 50 percent of the observed characteristics of dwellings are statistically significantly different on both sides of the boundary. These results might suggest that the stratification...