1# RECORD 620cbd7217ca28bcea726920009e42496a17ea8b2# TITLE: The Impact of Brownfields on Residential Property Values in Cincinnati, Ohio: A Spatial Hedonic Approach3# YEAR: 2012.0 VENUE: Journal of Regional Analysis & Policy DOI: 10.68479/001c.79414# AUTHORS: Oana Mihaescu; Rainer Vom Hofe5# PAGES: 14 CANDIDATE TABLE PAGES (1-based): [9]6# ABSTRACT: This paper empirically assesses the impact brownfield sites have on market values of7 single-family residential properties. Using three different hedonic model specifications, Ordinary Least Squares and Spatial Autoregressive and Spatial Error Models, this study shows that properties located less than 1,000 feet from a brownfield site experience a significant depreciation in property values. As expected, spatial dependence among real estate property values is present, meaning that the two spatial hedonic model specifications are superior to the Ordinary Least Squares specification. The study contributes to the relevant real estate and urban economics literature by determining the negative impact brownfield sites have on the value of single-family residential properties in our study region in Cincinnati, Ohio, by explicitly accounting for the phenomenon of spatial dependence in the dependent variable and by addressing the declining influence of brownfield sites on property values in a nonlinear relationship8910===== PAGE 1 (score 10, numeric-paren 0, table-words 1, var-words 8) =====1112 13The Impact of Brownfields on Residential Property Values in 14Cincinnati, Ohio: A Spatial Hedonic Approach 15 16 17Oana Mihaescu and Rainer vom Hofe 18School of Planning, University of Cincinnati – USA 19 20 21 22Abstract. This paper empirically assesses the impact brownfield sites have on market values of 23single-family residential properties. Using three different hedonic model specifications, Ordi-24nary Least Squares and Spatial Autoregressive and Spatial Error Models, this study shows 25that properties located less than 1,000 feet from a brownfield site experience a significant de-26preciation in property values. As expected, spatial dependence among real estate property 27values is present, meaning that the two spatial hedonic model specifications are superior to 28the Ordinary Least Squares specification. The study contributes to the relevant real estate and 29urban economics literature by determining the negative impact brownfield sites have on the 30value of single-family residential properties in our study region in Cincinnati, Ohio, by explic-31itly accounting for the phenomenon of spatial dependence in the dependent variable and by 32addressing the declining influence of brownfield sites on property values in a nonlinear 33relationship. 34 35 36 371. Introduction 38 39Brownfield sites (or simply brownfields) are 40abandoned, idled, or underused real properties that 41are currently or were formerly employed for indus-42trial or commercial purposes and whose expansion, 43redevelopment, or reuse may be complicated by the 44presence or potential presence of a hazardous sub-45stance, pollutant, or contaminant (Braswell, 1999; 46Bacot and O'Dell, 2006; US EPA, 2009, among oth-47ers). Contaminated properties yield little revenue 48for local governments. They are unsafe for ecosys-49tems and human health and impair the neighbor-50hoods that contain them, both socially and aestheti-51cally. In addition, they negatively impact the hous-52ing prices of residential properties that are located in 53close proximity (Bromberg and Spiesman, 2006). 54It comes as no surprise that a large number of 55studies have strived to understand the benefits of 56brownfield redevelopment. Attoh-Okine and Gib-57bons (2001), Amekudzi et al. (2003), and Barnett 58(2006), for instance, focused on the revitalization of 59inner city neighborhoods through housing, jobs, and 60tax revenues resulting from brownfield redevelop-61ment. In an environmentally-oriented regional level 62study, Alberini et al. (2005) argued that redevelop-63ing brownfields helps to reduce the harmful effects 64of the site's groundwater, water, soil, and/or air pol-65lution on ecological systems as well as on human 66health1. Akinmoladun and Lewis (1998) and Attoh-67Okine and Gibbons (2001) emphasize the im-68portance of brownfield redevelopment for combat-69ing the negative environmental effects of urban 70sprawl. Amekudzi et al. (2003) studied the effects of 71infill development to address urban sprawl. Gist 72(1999), Amekudzi et al. (2003), and Bacot and O'Dell 73(2006) focused on the impacts that redevelopment of 74contaminated properties has on neighborhoods and 75their citizens. 76In this study, we use the spatial hedonic pricing 77framework to empirically estimate the impacts 78 791 See also Bartsch, 1997; Attoh-Okine and Gibbons, 2001; Bacot 80and O'Dell, 2006. 81JRAP 42(3): 223-236. © 2012 MCRSA. All rights reserved. 82 838485===== PAGE 9 (score 122, numeric-paren 86, table-words 10, var-words 16) =====8687Spatial Hedonic Analysis of Brownfield Impact on Property Values 88 89 231 90 91 92variables remained in the final model specifications. 93Among others, the total finished square footage of 94the house (HOUSESQFT) and the total number of 95rooms (TOTALROOMS) were highly correlated with 96the other structural variables and were therefore 97removed from the model. Though multicollinearity 98may not reduce the usefulness of the model as a 99whole, it does have severe consequences for the in-100terpretation of the marginal effects of individual re-101gression coefficients (Chatterjee and Hadi, 2006), for 102example, the effect of the distance of residential 103properties on nearby brownfield sites (DISTBR). 104The problem of ambiguous regression results has 105been avoided by removing some of the severely cor-106related explanatory variables, while keeping almost 107all of the important and relevant variables in the 108model. 109In addition, for reasons of parsimony the median 110household income by ZIP code (MEDHHINC) and 111the distance to the CBD (DISTCBD) variables were 112removed due to statistically insignificant parame-113ters. Of the remaining fifteen variables, all calculat-114ed coefficients have the expected sign, an indication 115that no severe multicollinearity problems are pre-116sent (see Table 4). Based on the Breusch-Pagan test, 117we reject the null hypothesis of homoskedastic error 118at the 99% level of confidence for both spatial mod-119els. This is a commonly observed and documented 120phenomenon of spatial regression models. 121 122 123Table 4. Estimation results for OLS, SAR, and SEM model specifications. 124 125 126 OLS 127 SAR 128 SEM 129Variable 130Coefficient 131p-value 132Coefficient 133p-value 134Coefficient 135p-value 136Constant 13710.51091** 138(0.000) 13910.42543** 140(0.000) 14110.55953** 142(0.000) 143AREALOT 1443.7870E-06** 145(0.003) 1463.3899E-06** 147(0.007) 1484.82604E-06** 149(0.001) 150FRONT 1518.7561E-06** 152(0.001) 1531.0600E-05** 154(0.000) 1557.42170E-06** 156(0.003) 157ATTIC 1580.00014 159(0.067) 1600.00013 161(0.098) 1620.00014 163(0.065) 164BEDROOMS 1650.08504** 166(0.000) 1670.08581** 168(0.000) 1690.07047** 170(0.000) 171FULLBATHS 1720.15120** 173(0.000) 1740.16015** 175(0.000) 1760.14032** 177(0.000) 178HALFBATHS 1790.10480** 180(0.002) 1810.11255** 182(0.000) 1830.10348** 184(0.001) 185AGE 186-0.00698** 187(0.000) 188-0.00672** 189(0.000) 190-0.00643** 191(0.000) 192GARAGE 1930.12720** 194(0.000) 1950.12730** 196(0.000) 1970.12189** 198(0.000) 199FIREPLACES 2000.23866** 201(0.000) 2020.23163** 203(0.000) 2040.20293** 205(0.000) 206CVPOOR 207-0.83248** 208(0.000) 209-0.80300** 210(0.000) 211-0.86721** 212(0.000) 213AIRCOND 2140.36702** 215(0.000) 2160.35357** 217(0.000) 2180.28278** 219(0.000) 220HEATING 2210.31162 222(0.062) 2230.26759 224(0.107) 2250.31981* 226(0.035) 227EDUCATION 2280.01346** 229(0.000) 2300.01360** 231(0.000) 2320.01346** 233(0.000) 234EXP-DISTBR 235-0.22867* 236(0.026) 237-0.22844* 238(0.025) 239-0.25361* 240(0.030) 241DISTHW 242-6.6047E-06* 243(0.017) 244-5.6457E-06* 245(0.041) 246-6.7431E-06* 247(0.029) 248ߩ 249 250 2510.01007** 252(0.001) 253 254 255ߣ 256 257 258 259 2600.32641** 261(0.000) 262** significant at the 99% confidence level; * significant at the 95% confidence level. 263 264The univariate global Moran's I test statistic of 2650.4919 is statistically significant at the 99% level 266(p-value = 0.001), which indicates the presence of 267spatial dependence in housing prices. This is ex-268pected, as the price of a residential property is not 269solely determined through an array of housing and 270neighborhood characteristics, but also depends on 271the prices of neighboring properties. As for the OLS 272estimators, they are biased and inconsistent in the 273presence of spatial correlation. Accordingly, the 274discussion of our regression results will focus on the 275spatial model specifications. The presence of spatial 276dependence is further supported as the additional 277spatial parameters ߩ = 0.01007 and ߣ = 0.32641 in the 278SAR and SEM specifications are both statistically 279significant at the 99% level (Table 4). The relatively 280low ߩ value further implies that, though highly sig-281nificant, the spatial spill-over effect of neighboring 282residential properties is relatively small. This find-283ing can be explained by the fact that residential 284properties around brownfield sites do not show 285large variations in their values. Our sample data 286indicate that 80% of the property values fall between 287$30,464 and $163,800, compared to $67,500 to 288$319,970 citywide. As a matter of fact, most proper-289ties included in our study lie in the Mill Creek 290291292===== PAGE 10 (score 33, numeric-paren 0, table-words 6, var-words 21) =====293294232 295 296Mihaescu and vom Hofe 297Corridor, an old and depressed industrial area 298known for its low property values. 299The fact that the SAR specification contains a 300spatial lag of the dependent variable requires an ad-301ditional step to allow interpretation of the coeffi-302cients from the model. We follow LeSage and Pace 303(2009), who suggest estimating the summary 304measures of the total impacts of each explanatory 305variable on property values as: 306 307ߚ்ൌሺ1 െߩሻିଵߚ 308(7) 309 310The estimated total impacts are shown in Table 5 311below. 312 313Table 5. Calculated total SAR impacts. 314 315Variable 316Estimated SAR 317coefficients 318Total SAR 319impacts 320Constant 32110.42543 32210.53148 323AREALOT 3243.3899E-06 3253.4244E-06 326FRONT 3271.0600E-05 3281.0708E-05 329ATTIC 3300.00013 3310.00014 332BEDROOMS 3330.08581 3340.08668 335FULLBATHS 3360.16015 3370.16178 338HALFBATHS 3390.11255 3400.11369 341AGE 342-0.00672 343-0.00679 344GARAGE 3450.12730 3460.12860 347FIREPLACES 3480.23163 3490.23399 350CVPOOR 351-0.80300 352-0.81117 353AIRCOND 3540.35357 3550.35717 356HEATING 3570.26759 3580.27031 359EDUCATION 3600.01360 3610.01374 362EXP-DISTBR 363-0.22844 364-0.23076 365DISTHW 366-5.6457E-06 367-5.7031E-06 368 369Most of the variables in each of the three model 370specifications are statistically significant at the 99% 371level of confidence. The exception is the attic square 372footage (ATTIC) variable, which is only significant at 373the 10% level in each of the three models. Further, 374the presence of a heating system (HEATING) ap-375pears to be significant at the 95% level in only the 376SEM model specification. In addition, both distance 377measures, i.e., distance to brownfield sites and 378highways, are significant at the 95% level as well. 379The interpretation of the spatial dependence in 380the SAR specification is a structural process, based 381on the assumption of omitted variables (LeSage and 382Pace, 2009). As such, the SAR approach does ac-383count to some extent for these unobservable latent 384influences, such as noise and crime issues and other 385excluded variables. The total SAR impacts in Table 3865 show that one additional bedroom adds 8.67% to 387the market value of a house, while a full bath and a 388half bath add 16.18% and 11.37%, respectively. Hav-389ing a central air conditioning system increases the 390property value by 35.72%, while a heating system 391adds 27.03% and a garage adds 12.86%. As 392expected, the age of the property and educational 393attainment level of the residents impact residential 394property prices to a lesser extent. An increase in the 395age of the house by one year causes a decrease in the 396property value by only 0.68%, and a 1% increase of 397the population aged 25 or older holding a bachelor's 398degree or higher by ZIP code increases the property 399values by 1.37%. Also of note is that properties in 400very poor and unlivable condition are devalued by 40181.12%. 402The total estimated impact with respect to the 403distance from a property to its closest brownfield 404is -0.2308. Relying on a non-linear distance function, 405as shown in Figure 2, means that the estimated 406decrease in 407property value 408with 409respect 410to 411the distance to the closest brownfield site needs to 412be 413calculated 414using 415the 416exponential 417form 418݂ሺ݀݅ݏݐܾݎሻൌ݁ݔሺିೞ್ೝ419ఱబబሻ%, where distbr denotes the 420distance to the closest brownfield in feet. 421Referring to Figure 2 we see that decreasing the 422maximum distance allowed to a brownfield site, i.e., 423from 1,000 to 0 feet in our study, is equivalent to a 424reduction 425of 426the 427exponential 428function 429of 430݁ݔሺିଵ/ହሻ to ݁ݔሺି/ହሻ, which is equivalent to a 431reduction from 0.1353 to 1.00 on the vertical axis. 432This translates into a maximum possible reduction 433in property value of ሺ1 െ0.1353ሻ∗23.08% or 43419.96%. In other words, given that the maximum 435change in the explanatory variable is ሺ1 െ0.1353ሻൌ4360.8647, the largest change possible in a residential 437property value is 19.96%. The relationship between 438selected changes in actual distances in feet to the 439closest brownfield site, the equivalent changes in the 440exponential function, and the reductions in property 441values are shown in Table 6 below. 442As indicated in Table 6, reducing the distance 443from 500 to 0 feet results in a decrease of residential 444property values of 14.59%. Analogously, reducing 445the distances from 1,000 to 500 feet, from 500 to 100 446feet, and from 100 to 0 feet result in property value 447reductions of 5.37%, 10.40%, and 4.18%, respectively. 448As indicated in Figure 2, the largest change in prop-449erty values occurs in close proximity to the brown-450field sites, while reductions in property values level 451off when moving beyond the 1,000 feet cut-off dis-452tance used in our study. For instance, reducing the 453