The main purpose of this research is to estimate the impacts that a set of urban amenities and environmental conditions have on property values in Bogotá, Colombia through the hedonic price method. Research in this direction is important because such an estimation could be used to approximate the nonmarketable benefits from infrastructure investments and policies aimed to improve air quality conditions. These non marketable benefits are usually not taken into account in traditional cost-benefit analysis. Furthermore, given the transformation of Bogotá during the last decade and the availability of geo-coded information of urban amenities and housing values, the city could be considered as a suitable “laboratory” for the hedonic technique in a developing country.

This research has four objectives:

The first is to build a data set that relates property values with structural characteristics and urban amenities such as distance to closest urban parks, canals, main roads, and commuting time to closest Central Business Districts (CBD). GIS developments in this study also entailed the use of nonparametric techniques such as the Inverse Distance Weighted (IDW) interpolation to estimate average pollution levels in the urban area of Bogotá. GIS application was shown to be a useful tool for both exploratory data analysis and data base construction with distance based variables.

The second objective is to use the hedonic technique in conjunction with simple GIS analysis to explore policy relevant questions in a local context such as (a) what are the benefits of improving accessibility to a CBD for properties located “nearby” rapid transit system lines?, (b) what the benefits for those properties located far away are? (c) How can we approximate monetary benefits from improving air quality conditions assuming, for example, a reduction of particulate matter (PM10) levels to those below the local environmental authority's standard? Results suggests that policies aimed to improve urban amenities such as ambient quality or accessibility to employment centers could bring as a result monetary benefits that may be capitalized in property values.

The third is to explore spatial econometric approaches to estimation of the hedonic price function. Econometric specification such as the Spatial-Lag, Spatial Autoregressive Error and Spatial Components Error are explored. The use of spatial econometric techniques is motivated by different spatial processes of the regression's disturbance that may affect parameter estimates. Also, it is recognized that spatial dependence in both property values and unobserved variables may be present due to intrinsic features of housing markets. Using non spatial techniques for a spatial research problem may be inappropriate when there is evidence of spatial effects in the data. OLS and spatial models all showed the same sign and significant coefficients for air quality and accessibility to employment centers, the variables of interest of this study. Lagrange Multiplier tests suggest that a spatial process characterized by autocorrelated errors is the most likely in this data. This suggests that omitted variables that vary spatially and that are captured in the error term may have a significant effect on apartment rent values.

The fourth objective is to use alternative techniques to OLS when asymmetric random error is hypothesized. Two processes that may affect the housing price formation are motivated for the use of stochastic frontier models' techniques within the context of the hedonic pricing method: “speculative overpricing” and omitted variables. In this specification the usual zero conditional mean assumption is relaxed. Instead, a systematic composed error is assumed. This error is characterized by the usual zero mean and constant variance idiosyncratic part and an asymmetric error component. Further the variance of the asymmetric error component is assumed to vary with pollution levels. A heteroskedastic frontier model with variance of the asymmetric error as a function of pollution is estimated. Results for sales do not support the “speculative over pricing” hypothesis. Results for rental prices suggest the presence of an asymmetric random error whose variance decreases with increased pollution levels. This result would be consistent with an “omitted variable” process that may characterize the housing price formation in this market.