Home » 2015 Categories » 2015 Technical Presentation

Category Archives: 2015 Technical Presentation

Technical Review by Ibe Alozie

Assessing the Economic Impact of Sports Facilities on Property Values: A Spatial Hedonic Approach By: Xia Feng and Brad R. Humphreys[1]

I. Research Question

Brad Humphreys, Dennis Coates and many other urban economists have conducted research in the field of sports arenas and urban development. However, most research has focused on identifying and analyzing tangible, economic benefits of sports arenas on cities. Differentiating itself from prior research on the intangible benefits of sports arenas on cities, Xia Feng and Brad Humphreys’ paper proposes a spatial hedonic model that estimates the intangible benefits of two sports facilities in Columbus, Ohio on residential property values.

This discussion of the benefits of sports stadiums stems from the willingness of cities and towns to subsidize construction of expensive sports stadiums. As the rise in the size of these subsidies has coincided with the boom in the construction of new stadiums, urban economists conducted research on the costs and benefits of construction of new stadiums and arenas. Proponents of these subsidies posit income increases, job creation and multiplier effects (due to new spending) as tangible, positive impacts of building new sports stadiums. However, contrary to the aforementioned claims, made mostly by consulting firms (usually hired by the respective sports franchises), the findings from years of economic research have shown no positive impact of building new stadiums on cities. In fact, econometric evidence has shown that professional sports facilities can have little effect to net negative effects on the local economy.

Regardless of these well-respected and well-supported research projects, cities continue to subsidize the construction of sports stadiums. The continuation of this policy decision, which research finds in general to be neither cost-effective for cities nor beneficial to cities, forces consideration of intangible benefits. Few papers have empirically estimated the intangible benefits, such as the increased civic pride, increased city attractiveness or increased cultural benefits, of building sports stadiums. A couple papers have examined the impact of sports facilities on property values with varying results, and this study adds to the literature by providing new evidence based on data from different locations and different sports. Most importantly, this study does not ignore spatial effects. Spatial autocorrelation is the correlation among values of a single variable due to their close locational positions on a two-dimensional (2-D) surface. Spatial autocorrelation could have caused biased estimates and model misspecification in the few earlier models on the subject of stadium presence’s impact on housing prices

II. Theoretical Background

Because of the difficulty of measuring “intangible benefits or costs”, Feng and Humphreys assume that the presence of a stadium would be viewed as an intangible characteristic and the presence of a sports stadium would be capitalized in housing prices. Housing prices tend to be spatially correlated due to common neighborhood characteristics.

Feng and Humphreys use an adaptation of the spatial lag hedonic model:

(I − ρW y) −1 = I + ρW + ρ 2W2 + . . .

This model links each observation of the dependent variable to all observations of the explanatory variables through a spatial multiplier.

Using transactions data, containing observations on 9,504 single-family housing units, for the year 2000, Feng and Humphreys analyze the values of residential housing around Nationwide Arena and Crew Stadium in Columbus. The data set includes housing and neighborhood characteristics such as lot size, school quality, environmental quality and number of fireplaces.

To account for aspects of the model that were not incorporated into the adapted spatial lag hedonic model, certain modifications were made to the model. To account for the presence of Ohio Stadium, dummy variables were created. To control for the effects of businesses on housing values, Feng and Humphreys controlled for the number of commercial establishments in each zip code, which allowed the business-related variables to capture some of the effects of business location on residential property values.

III. Empirical Model

Known as a spatial weighting matrix, this symmetric matrix is used to define the locations for which the values of the random variables are correlated, and the rows in the weights matrix are standardized. The features of both housing markets and individual housing data make the definition of the spatial weights matrix W especially important. The aforementioned matrices specify “neighborhood sets”, and these neighborhood sets capture spatial interaction. Feng and Humphreys use GeoDa to specify the neighborhoods and to define the spatial weights matrix, and begin by using four different spatial weights to create the matrices. Next, Feng and Humphreys use the log-log form of the hedonic housing price with the appropriate spatial lags to best estimate the parameters.

IV. Results and Discussion

The results of the research of Feng and Humphreys suggest that the presence of sports facilities in Columbus have a significant positive distance-decaying effect on surrounding house values. For Nationwide Arena, at the average, all else equal, for each 1% decrease in the distance to the arena is associated with a 0.175% increase in the price of the average house. In dollar terms, a 1% decrease in distance from each house to the arena, on average, increases the price of an average house by $222. The primary variable used to evaluate the effects of sports facilities on surrounding housing values is the distance between each house and the sports facility, and analysis of this parameter shows that the presence of sports facilities has positive effects (though they diminish with distance) on housing values. Importantly, Feng and Humphreys also show that prior OLS models, which did not account for spatial autocorrelation, overestimated the distance parameters, and did not correct for heteroskedasticity when present.

V. Extensions

This paper elevated the credibility of the larger economic argument by finding the general importance of factoring spatial autocorrelation into property value modeling.   With regard to policy decisions, professional sports facilities generate intangible benefits in the local economy, and cities do have a rational economic argument to lodge in support of provision of subsidies to sports stadiums. While the costs of public support rarely exceed the cost of public funding for the stadiums directly, the subsequent rise in property values can set the foundation for more substantial growth in adjacent areas, and give the city’s business community the confidence necessary to invest. Feng and Humphreys offered a more precise method of analyzing costs and benefits, and show that there are positive effects (contrary to most research) of building sports facilities at least in this one example. This paper offers answers, and poses new questions. What other benefits can be discovered? How close can economists make it to quantifying the efficient subsidy level for stadiums and arenas?


[1] Humphreys, Brad & Feng, Xia. “Assessing the Economic Impact of Sports Facilities on Property Values: A Spatial Hedonic Approach.” LASE/NAASE Working Paper Series 8.12 (2008): 1-20. Web. 25 March 2015.

Mortgage Lending in Chicago and Los Angeles by Li Ding

In “Mortgage Lending in Chicago and Los Angeles: a paired-testing study of the pre-application process”, Ross et al. (2008) used paired testing to measure discrimination against African-American and Hispanic homebuyers in the mortgage lending process. Many studies have provided evidence that minority buyers are less likely to receive mortgage loans than white buyers and, if successful, receive less favorable loan amounts and terms. There is debate, however, on how much of this outcome can be attributed to discrimination. Due to differences in creditworthiness, it is not typically straightforward to isolate the effects of differences in racial and ethnic treatment. Most work done on the topic of race in lending has used HMDA data which does not contain many important lender and loan attributes such as credit history and lending ratios.

Using data from a recent paired test study of discrimination in lending, Ross et al. examine the effects of race and ethnicity on mortgage lending. Using paired testing, two individuals, one white and one minority, separately pose as homebuyers with equal qualifications for borrowing. Both members of the pair ask about the availability and terms for the same home mortgage loan. Since the two borrowers are constructed to be equal in every regard other than race or ethnicity, differences in the responses received by the two can provide direct evidence for differing treatment of minorities. It should be noted that this methodology will only focus on the first part of the lending process, the pre-application stage (which involves a loan officer that can observe the race of the applicant) rather than the approval stage (with an underwriter who typically does not).

Paired Testing Methodology

The study included approximately 250 paired tests of a representative sample of mortgage lenders in Los Angeles and Chicago. Testers posed as first-time homebuyers with limited assets making general requests for information from lenders about their mortgage loan options. The testers were given profiles that qualified them for loans targeted towards A- credit quality borrowers in their respective housing markets. Each tester was assigned sufficient income to purchase a median-priced home in the area (with a 30 year fixed-rate loan and 5% down payment) and randomly assigned one or two minor credit issues, mostly late payments. Each pair was given almost identical financial and household characteristics with the minority in the pair receiving slightly better qualifications. These pairs, it should be noted, were not permanent—a tester could be paired with multiple partners if more than one partner was available that also generally matched in gender, age, and appearance.

Table 1 below provides data on the lending institutions in the study. The study looked only at lenders that reported under the Home Mortgage Disclosure act, accepted at least 90 loan applications in 1998, and had reasonably located offices for a first-time homebuyer. 67 lenders in Los Angeles and 106 lenders in Chicago qualified under these criteria, and in order to draw a market representative sample, lenders were selected (with replacement) with a probability of selection based on loan volume. This provided 35 lenders for black-white testing in Los Angeles, 34 for Hispanic-Anglo in LA, and so on as indicated in the table.

Li Ding - Table 1

The basic testing protocol involved five steps:

  1. Obtain an appointment – testers called to arrange in person visits with lenders
  2. Make the initial request – testers requested help in figuring out a price range of housing they could afford and an estimated loan amount that they would qualify for
  3. Exchange personal/financial information – testers provided all requested information on income, debts, assets, credit history, etc.
  4. Record information on recommendations – testers noted suggested home price range, estimated loan amount, and financing options recommended
  5. End the visit – testers thanked the lender and allowed them to suggest follow-up contact

The testers then completed a test report form which allowed the study to gather information on the following six questions:

  1. Did the testers receive the information they requested about loan amounts and house prices they could afford
  2. How much were testers told they could afford to borrow and/or buy?
  3. How many specific products were discussed with the tester?
  4. How much “coaching”, such as offers of advice on paying down debts, down payment assistance, or a prequalification letter, did testers receive to help them qualify for a loan?
  5. Did testers receive follow-up calls from lenders?
  6. Were testers encouraged to consider FHA loans as an option?

 Statistical Analysis Methodology

The paired tests each generate a series of treatments t for the white and minority testers, designated as Wit and Mit respectively. An incidence measure i is derived by comparing the experiences of the two testers and classifying the test as majority favored, equal treatment, or minority favored. For loan amounts or house prices, a test is considered favored one way or the other if a tester receives an estimate that is 5% higher than their counterpart. Gross majority favored treatment is defined as the fraction of tests classified as majority favored, and likewise for gross minority favored treatment. The net measure of adverse treatment, Nt is then defined as

Li Ding - Equation 1

which is gross majority favored treatment minus gross minority favored treatment. Probability (Pr) in this case is solely a measure of sample frequency. Also, a severity measure for a treatment is defined as the difference in the treatment experienced by the two testers

Li Ding - Equation 2

where the expected value (E) is captured by the sample mean of the difference of the two series of treatments. These two measures, Nt and St, are commonly used estimates of systematic discrimination towards minorities. Statistical tests are performed on these two variables to determine if they differ significantly from zero using a two sided test. While it would be very unlikely to find unfavorable treatment for whites based on past studies, the authors decided to use the two-sided test as it was more conservative.

To address the potential issue of bias arising from using the normal distribution for small sample sizes, the authors use Fisher’s exact (permutation) tests, writing the null hypothesis for Nt as

Li Ding - Equation 3

For Sthe null hypothesis is

Li Ding - Equation 4

Empirical Results

Table 5 below summarizes the patterns of findings. Significant differences between the white favored and minority favored are indicated, with * representing significance at the 5% level and ** for the 1% level. The last row of the table shows that in Chicago, Hispanics and blacks received significant differential treatment from whites in three and four of the six categories, respectively. For both minority groups in Chicago, this leads to a rejection of the null hypothesis of equal treatment for whites and minorities at the 0.01 level. In Los Angeles, the data taken as a whole is consistent with the null hypothesis of equal treatment.

Li Ding - Table 2

In summary, the paper finds strong evidence of adverse treatment of Hispanics and blacks compared to whites in Chicago in the pre-application stages of the mortgage lending process. In the study, Hispanics were quoted lower loan amounts and house prices, were given less information about products, and received less coaching. African Americans were provided less information, received information about fewer products, received less coaching, and were less likely to experience follow-up contact. Los Angeles, on the other hand, showed no statistically significant differences in overall treatment of its white and minority borrowers. While minorities received worse treatment in some specific categories, this was not indicative of an overall pattern in LA.

Discriminatory treatment at this early stage in the mortgage lending process, though subtle,can have effects on the rest of the mortgage application. Minority homeseekers may be discouraged from applying for a mortgage due to their treatment by a lender, either abandoning their search completely or applying through the costlier subprime mortgage market instead. Also, loan officers provide more support and information to white applicants in certain circumstances which gives them a better chance of acceptance than a similarly qualified minority applicant.

Federal law, through the Equal Credit Opportunity Act (ECOA) and Fair Housing Act (FHA), forbids credit discrimination and real-estate related discrimination. The results from this study show that discrimination in these aspects is an unfortunate reality for minorities seeking home mortgage loans. Further study could be done on the reasons behind the different levels of discrimination found in Chicago and Los Angeles in the study. This research could then be used to help implement policies and effect change on a broader scale to help fight against unfair lending treatments and practices.

Referenced Paper

Stephen Ross, Margery Austin Turner, Erin Godfrey, and Robin Smith, 2008, “Mortgage lending in Chicago and Los Angeles: a paired-testing study of the pre-application process,” Journal of Urban Economics 63: 902-919.


Tables 3 and 4 below provide information on the proportions of each test that were favored for white or minority testers.

Li Ding - Table 3

Li Ding - Table 4