What information is captured in home prices? Clearly prices should reflect characteristics such as square footage, build quality, and the number of bedrooms. Economists also believe that house prices reflect local characteristics, such as school and air quality, presence of open space, and crime rates. Traditionally, researchers employ hedonic models, where the marginal willingness to pay of these characteristics is obtained by running a linear regression of housing and neighborhood characteristics on the log of house price. These models have been used to study the value of a myriad of topics, from pollution and crime rates, to views of windmills and presence of nearby methamphetamine labs. As with many methods of analysis, hedonic models are subject to numerous assumptions and caveats, many of which are often ignored.This dissertation explores several complexities of and proposes new means of employing house prices in economic analysis. The first chapter asks a question that has received little study: why do buyers pay different prices for the same house? In most studies utilizing house transactions, the researcher does not know who the buyers and sellers are, and thus implicitly assumes that specific types of individuals have no effect on home prices. The chapter measures the effect of experience: the relative number of transactions the buyer and seller have taken part in over a given period of time. First, I develop a two-sided real estate search model that incorporates information costs, search costs, and Nash bargaining power. I test the implications of this model using repeat-sales housing data on 113,272 transactions from 1998-2006 in two large metropolitan regions of Ohio. The main results show that more experienced buyers purchase properties at a discount, experienced sellers sell at a premium, and that the magnitude of these differences varies depending on the relative and absolute levels of buyer and seller experience and geographic location. On average, being more experienced than the other party leads to a 4% better price, although it appears that experience is more important to buyers than sellers. The results are found to be robust to different specifications, including varying the time interval and geographic area of study.
The second chapter proposes an entirely new means of utilizing housing characteristics. Rather than measuring how house characteristics impact prices, I use these characteristics and prices as a proxy for unobserved demographic characteristics of thousands of families to answer an important yet unanswered question in urban economics: Do people mitigate their commuting costs by selecting a more fuel efficient or larger, more comfortable vehicle, and what are the implications for the monocentric city model? In this paper I empirically test for evidence of vehicle choice due to distance to the central business district or commuting time. Utilizing a unique data set of every registered vehicle in Franklin County, Ohio, I match vehicles with parcel sales transactions of single family homes to obtain a proxy for demographic characteristics. Utilizing nearest neighbor matching, I find that otherwise identical individuals with longer commutes select less fuel efficient, larger vehicles, in the order of 05.-5% change to fuel economy and vehicle size. Much of this difference is due to selection of vehicle class. I also investigate the probability of selecting different vehicle classes by employing multinomial logit and probit models, and find that both distance and demographics effect the selection between different types of vehicles. I also observe how distance and demographic characteristics are associated with the fuel efficiency or size within a particular vehicle class. I conclude that discomfort and disutility of commuting outweigh gasoline costs, and consider the implications for urban spatial structure.
The final chapter employs home prices in a more traditional setting, namely the study of an environmental characteristic. The goal is to observe whether and how harmful algal blooms are capitalized in home prices in several Ohio counties along Lake Erie. Harmful algal blooms have increased in severity and frequency, reducing the amenity value of the lake, posing health risks, and even threatening the drinking water supply. The main contribution is the use of satellite readings of chlorophyll-a to measure the level of algae, and the use of a repeat sales model to address the problems of omitted variable bias due to unobserved emitter sites. I find that a reduction of 1µg/L of chlorophyll-a yields a 2-6% increase in home prices, though the effect decays rapidly by distance and depends on the current level of algae. Secchi disk depth, while previous found to be significant in the literature, is statistically insignificant when utilizing repeat sales, and I find noticeable evidence of omitted variable bias when not controlling for emitter effects.