A Multilevel Property Hedonic Approach to Valuing Parks and Open Space

1. A Multilevel Property Hedonic Approach to Valuing Parks and Open Space Treg Christopher Dissertation Seminar Oct. 15, 2009

2. Outline • Ecosystem Goods and Services and Valuation • Property Hedonic Model • Spatial Issues • Multilevel Modeling Methods • Research Results • Baltimore City Parks • Baltimore County Open Space

3. Benefits of Parks and Open Spaces • Protects water quality • Provides wildlife habitat • Improves physical & mental health • Educational opportunities • Flood regulation • Offsets urban heat sink

4. Valuing the Environment • If we can assign a quantitative monetary value on ecosystem goods and services, we can: • justify the use if public funds for restoration or protection projects; • make objective decision on the most efficient use of limited resources; • through cost-benefit analysis, set priorities for programs, policies or actions.

5. Valuation Methods • Stated Preference: • Uses surveys to ask people what they would be willing to pay. • How much would you be willing to pay to improve water quality in a nearby lake? • Revealed Preference: • Uses market prices to estimate how much people are willing to pay. • How much does proximity to a park contribute to the value of a property?

6. Types of Revealed Preferences • Market price method • Estimates economic value based on the price of marketable goods. • Productivity method • Estimates economic value based on the cost of producing marketable goods. • Hedonic price method • Estimates economic value based on effects on the price of other marketable goods. • Travel cost method • Estimates economic value of a site based on how much people are willing to pay to travel to visit the site.

7. Property Hedonic Approach • Assumes that a home buyer implicitly prices characteristics embodied in the property: • Structural (house size, age, # of baths, lot size) • Neighborhood (ethnicity, income, unemployment, crime) • Locational/Environmental (access/proximity to parks, schools,shopping) • To live in a location with higher levels of an amenity • Households pay higher house prices • Cost differences reflect households’ implicit valuation

8. Property Hedonic Approach • Hedonic prices are identified through a comparison of similar goods that differ for the quality of one characteristic Proximity to a park Proximity to a park Distance 100m 500m

9. Property Hedonics & Environmental Externalities • Negative externality • air pollution (Batalhone 2002) • proximity to hazardous waste sites (McCluskey 2003) • Positive externality • open space (Bastian 2002; Bolitzer 2000; Geoghegan 1997; Irwin 2002) • Improvements to water or air quality (Cho 2006; Kim 2003; Leggett 2000)

10. Previous Studies of Parks • Positive value to preserving most types of open space in urban-suburban areas • Close proximity to parks in urban areas has a significant positive impact on home value but depends on context: • Neighborhood Effects: • Parks in urban areas and more densely populated suburbs tend to show greater benefits than those in more sparsely-populated areas • Park Characteristics: • Natural, wooded areas have greater beneficial impacts • Larger parks have greater beneficial impacts • High crime in the areas surrounding parks will reduce the value of parks

11. Property Hedonic Limitations • Need large data sets and detailed information on all aspects that affect prices • We assume that buyers choose houses that maximize their utility • requires information that the buyer may not perceive • Assuming a single housing market for all consumers • Only measures “use” values of the environment • recreation and aesthetics • Numerous statistical assumptions

12. The Pitfalls of Spatial Analyses • Spatial dependencies (autocorrelation) • Data from location near to each other are more likely to be similar than data from location remote from each other • Causes bias in coefficients and standard errors • Heterogeneity of relationships (Non-stationarity) • Phenomena is not distributed evenly in space • Scale and the Modifiable areal unit problem (MAUP) • Results may depend on the areal unit used • Block Groups vs. Tracts (scale issue) • Block Groups vs. ‘Neighborhoods’ or Zipcodes (zoning)

13. Testing for SA:Moran’s Index • positive when attributes of nearby objects are more similar than expected • 0 when arrangements are random • negative when attributes of nearby objects are less similar than expected Close in space Repulsion of attributes Attributes independent of location Close in space Affinity of attributes

14. Spatial Non-Stationarity • Global models (e.g. OLS): Assume relationships are stationary and as such are location independent • Non-Stationarity: a different relationship in different parts of the study region • Local models: spatial decompositions of global models, the results of local models are location dependent

15. Spatial Scales • Conclusions about processes and relationships determined at one scale should not expected to be similar at other scales • Atomistic Fallacy • Inferences about broad scale/group level/aggregated data and relationships are based on small scale/individual level analyses • Ecological Fallacy • Inferences of fine-scale/individual level data and relationships are based on broad scale or aggregated data

16. “Tree scale”: Spatial relationship betw/ trees (the regular distribution) determined by competition for light, water, nutrients which prevents the trees from growing to closely to one another Negative association Small/ Fine “Stand or Community Scale”: Spatial relationship determined by same species having common needs (light, water nutrients) which are heterogenous across space SCALE “Forest Scale”: Spatial relationship determined by disturbances such forest fires, pest & diseases Large/Broad

17. Baltimore Data • Property data • MD PropertyView: Structural characteristics & Sale info • Sales between 1998-2002, $ converted to Yr2000 • Property records removed if: • Sale price <$50,000 • House size <500ft2 • Lot size <500ft2 • Number of bathrooms > 6 • Not townhouse or individual residence • Approx. 14,000 records • Census data • Block groups from 2000 Census • over 400 groups with at least 5 property sales

18. Baltimore Data

19. Non-Spatial Statistical Assumptions • Linearity - the relationships between the predictors and the outcome variable should be linear • Normality - the errors should be normally distributed • Homoscedasticity- the variance of errors should be stable across space • Model specification - the model should be properly specified (including all relevant variables, excluding irrelevant variables ) • Multicollinearity – independent variables should not be highly correlated with each other

20. Multilevel Models • Early history: Educational research • effects of context such as classroom and schools on individual scholastic achievement (Goldstein 1993; Raudenbush 1991). • Individuals that are members of a group cannot be modeled as independent observations • Statistical violation leads to erroneously small Standard Errors • spurious statistical significance of coefficients • Examine potential Non-stationarity • Avoiding the atomistic and ecological fallacies • Data is modeled at the appropriate scale • Allowing cross-level interactions between individual level factors and group level (contextual) factors.

21. y1 r12 u2 0 γ0 u1 Do Groups Matter? House Price Null Multilevel Model OLS Model $1M+ y1 r1 02 $110,000 01 $50,000

22. Null/Baseline/Unconditional Model Price for any individual (i) within Block Group (j) is a function of the group mean and the individual-level error term • Combined model Each Block Group mean is a function of the grand mean and a group-level error term Individual level variance Group level variance

23. Intraclass Correlation Coefficient (ICC) • The ICC is the proportion of variance in property price between block groups to the total variance Individual level variance: σ2 Τ00/(Τ00 + σ2 ) 65.5% of the variance in property price is between block groups Group level variance: Τ00 0.17189 / (0.17189+0.0904)=0.655

24. Level 1 Model Fixed intercept, Fixed slope: OLS Model House Price House Size

25. Level 1 Model Random intercept, Fixed slope: Multilevel, L1 Model House Price House Size

26. Spatial Error Autocorrelation • Moran’s Index OLS (SPSS): Multilevel (HLM): Moran's Index: 0.227255 p-value: 0.000000 Moran's Index: 0.014214 p-value: 0.115993

27. Modeling Random Effects Random intercept, Random slope: Multilevel, L1 Model Random intercept, Fixed slope: Multilevel, L1 Model House Price House Size

28. Why Allow Slopes to Vary? 1. Helps reduce problem of heteroscedasticity from level 1 residual Fixed Slope Random Slope Test of homogeneity of level-1 variance ---------------------------------------- Chi-square statistic = 2291.81500 Number of degrees of freedom = 404 P-value = 0.000 Test of homogeneity of level-1 variance ---------------------------------------- Chi-square statistic = 2095.47234 Number of degrees of freedom = 387 P-value = 0.000

29. Why Allow Slopes to Vary? 2. Mapping non-stationarity

30. Level 2 Model: “Means as Outcomes”(MAO) The mean price for each group is an outcome to be predicted by group characteristics • Combined Equation

31. Census Variables & Spatial Autocorrelation

32. Compositional Effects of Structure • Price is determined by a property’s structural attributes & the structural attributes of neighboring houses Individual house attributes: Age, House Size, Lot Size Neighborhood averages of individual attributes • House Size and Age are significant at the neighborhood level Conclusions about processes and relationships determined at one scale should not expected to be similar at other scales

33. Cross-Level Interactions House Size Results MedHsInc w/ Cross-level interaction Cross-level interpretation: Not only does house size have a positive effect on price but the slope increases with increasing neighborhood income

34. Valuing Parks in the City of Baltimore • Include “official” city parks & other, undeveloped open spaces • n=81 • use parcels boundaries to verify and edit park boundaries • only parks > 2ha

35. Distance to Parks Euclidean Network Combined

36. Result of Park Coefficient Interpretation: a 0.003% (0.015% in OLS) decrease in price with every 1% increase in distance • Why Non-Significant in HLM Models vs OLS? • Spatial Autocorrelation? • Non-stationarity?

37. Spatial Non-Stationarity of Park Coef.

38. Explaining Non-Stationarity of Park Coef. • Interaction variables • Structural: House Size, Age, Lot Size • Park: Crime, Size of Park (ha), % Open • Neighborhood (cross-level): MedHsInc, %Unemployment, Crime, PopDensity • Results: Lack of significant interaction • What if Block Groups are divided?

39. Conclusions: Valuing Parks in the City of Baltimore • The effect of park proximity on property price varies across space • Previous studies using Global Models are unable to examine such variation • For neighborhoods that do positively value proximity to parks: • larger and more open parks tend to be more highly valued • neighborhoods with high population densities tend to place higher value on park proximity • Unknown interactions for neighborhoods that negatively value proximity to parks

40. Valuing Open Space in Baltimore County • Proportion of different land uses that surround a home: • commercial, residential, open space • Are their differences between open space types? • Difference in value between different spatial scales?

41. Valuing Open Space in Baltimore County • Open space types: • Private, Conserved • easements and riparian buffers • Private, Developable • “Public” • publicly accessible • golf courses, cemeteries, school fields, campuses • About 75% wooded and 25% open: grass or farms

42. Land Use at Multiple Scales • Do the effects of proportions of open space and other land uses vary with the scale at which they are examined? • Block group, 1km, • 500m,100m 1 km 500m 100m

43. Results

44. Conclusions • Private, Conserved > Private, Developable > Public • Preference for visual amenities over direct use (recreation) • Preference for absence of negative externalities vs presence of positive amenities such as recreation • Results do change with scale

45. Future Research • Include a 3rd level • group block groups by tracts, PRIZM (consumer behavior) classes, school districts • Cross-classified design • use a non-nested grouping where properties “belong” to a socio-economic group (neighborhood) and the park to which they are closest Property Block Group Tract Park Block Groups Park

46. Multilevel Weaknesses • Need for bounded (discrete) L2 groups may result in artificial boundaries being formed whereas processes and associations may be more diffuse • Block Group (Census units) boundaries are created to minimize the heterogeneity within groups • Requires large datasets (min 30 groups & 5 observations/group) • Shrinkage of unreliable estimates of random effects is towards the grand mean rather than neighboring groups

47. Multilevel Valuation of ES • Valuation of Ecosystem Service is dependent on context • supply of services & demand of consumers • Context is scale dependent • variables should be measured at appropriate scale(s) • interactions can occur across scales

48. Thanks for Listening! Questions?

49. References • Acharya, G., and Bennett, L.L. 2001. Valuing open space and land-use patterns in urban watersheds. Journal of Real Estate Finance and Economics 22(2): 221–237. • Allen, T.F.H., and Hoekstra, T.W. 1992. Toward a unified ecology. University of Columbia Press, NY, NY. • Allen, T.F.H., and Starr, T.B. 1982. Hierarchy: Perspectives for ecological complexity. The University of Chicago Press, Chicago. • Anderson, S.T., and West, S.E. 2006. Open space, residential property values, and spatial context. Reg. Sci. Urban Econ. 36: 773–789. • Anselin, L. 1988. Spatial econometrics : methods and models. Kluwer Academic Publishers London. • Anthon, S., Thorsen, B.J., and Helles, F. 2005. Urban-fringe afforestation projects and taxable hedonic values. Urban Forestry and Urban Greening 3: 79–91. • Assessment, M.E. 2003. Ecosystems and human well-being: A framework for assessment. Report of the Conceptual Framework Working Group of the Millennium Ecosystem Assessment. • Bastian, C.T., McLeod, D.M., Germino, M.J., Reiners, W.A., Blasko, B.J. 2002. Environmental amenities and agricultural land values: a hedonic model using geographic information systems data. Ecol. Econ. 40: 337-349. • Beland, F., Birch, S., and Stoddart, G. 2002. Unemployment and health: Contextual-level influences on the production of health in populations Social Science and Medicine 55(11): 2033-2052. • Benson, E.D., Hansen, J.L., Schwartz, A.L., and Smersh, G.T. 1998. Pricing residential amenities: The value of a view. Journal of Real Estate Finance and Economics 16(1): 55-73. • Bhat, C.R. 2000. A multi-level cross-classified model for discrete response variables. Transp. Res. Pt. B-Methodol. 34: 567-582. • Bin, S., and Polasky, S. 2004. Effects of flood hazards on property values: Evidence before and after Hurricane Floyd. Land Econ. 80: 490-500. • Bolitzer, B., and Netusil, N.R. 2000. The impact of open spaces on property values in Portland, Oregon. J. Environ. Manage. 59: 185-193. • Borjas, G.J. 1998. To Ghetto or Not to Ghetto: Ethnicity and Residential Segregation. Journal of Urban Economics 44: 228-253. • Bourassa, S.C., Hoesli, M., and Peng, V.S. 2003. Do housing submarkets really matter? Journal of Housing Economics 12: 12–28. • Bourassa, S.C., Hoesli, M., and Sun, J. 2004. What's in a view? Environ. Plan. A 36(8): 1427-1450. • Box, G.E.P., and Cox, D.R. 1964. An analysis of transformations. Journalof the Royal Statistical Society, Series B (Methodological) 26: 211–252. • Brasington, D.M. 1999. Which measures of school quality does the housing market value? The Journal of Real Estate Research 18(3): 395-413. • Brown, K.H., and Uyar, B. 2004. A hierarchical linear model approach for assessing the effects of house and neighborhood characteristics on housing prices. Journal of Real Estate Practice and Education 7(1): 15-23. • Browning, C.R., Feinberg, S.L., and Dietz, R.D. 2004. The paradox of social organization: Networks, collective efficacy, and violent crime in urban neighborhoods. Social Forces 83(2): 503-534. • Carson, R.T., Mitchell, R.C., Hanemann, W.M., and Kopp, R.J. 1999. A Contingent Valuation study of lost passive-use values resulting from the Exxon Valdez oil spill. Report to the Attorney General of the State of Alaska. • Clapp, J.M., A., N., and Ross, S.L. 2007. Which school attributes matter? The influence of school district performance and demographic composition on property values. Journal of Urban Economics 63: 451-466. • Clapp, J.M., and Giaccotto, C. 1998. Residential hedonic models: A rational expectations approach to age effects. Journal of Urban Economics 44(3): 415-437. • Cliff, A., and Ord, J.K. 1981. Spatial processes: Models and applications. Pion, London. • Costanza, R., Grasso, M., Hannon, B., Limburg, K., Naeem, S., O'Neill, R.V., Paruelo, J., Raskin, R.G., Sutton, P., Van den Belt, M., D'Arge, R., De Groot, R., and Farber, S. 1997. The value of the world's ecosystem services and natural capital. Nature 387(6630): 253-260. • Coulson, N.E. 1991. Really useful tests of the mono-centric model. Land Econ. 67(3): 299-307.

50. References • Cropper, M.L., Deck, L.B., and McConnell, K.E. 1988. On the choice of functional form for hedonic price functions. The Review of Economics and Statistics 70: 668–675. • Day, B., Bateman, I., and Lake, I. 2004. Nonlinearity in hedonic price equations: An estimation strategy using model-based clustering. University of East Anglia: Centre for Social and Economic Research on the Global Environment (CSERGE). • de Groot, R.S., Wilson, M.A., and Boumans, R.M.J. 2002. A typology for the classification, description and valuation of ecosystem functions, goods and services. Ecol. Econ. 41(3): 393-408. • Deaton, B.J., and Hoehn, J.P. 2004. Hedonic analysis of hazardous waste sites in the presence of other urban disamenities. Environ. Sci. Policy 7: 499–508. • Dehring, C., and Dunse, N. 2006. Housing density and the effect of proximity to public open space in Aberdeen, Scotland. Real Estate Economics 34(4): 553 - 566. • Des Rosiers, F., Thériault, M., and Villeneuve, P. 2000. Sorting out access and neighborhood factors in hedonic price modeling. Journal of Property Investment and Finance 18(3): 291-315. • Diez-Roux, A.V. 2002. A glossary for multilevel analysis. Journal of Epidemiology and Community Health 56: 588–594. • Diez-Roux, A.V., Link, B.G., and Northridge, M.E. 2000. A multilevel analysis of income inequality and cardiovascular disease risk factors. Social Science & Medicine 50: 673-687. • Downes, T.A., and Zabel, J.E. 2002. The impact of school characteristics on house prices: Chicago 1987–1991. Journal of Urban Economics 52: 1-25. • Ekeland, I., Heckman, J.J., and Nesheim, L. 2002. Identifying hedonic models. American Economic Review 92(2): 304-309. • Fone, D.L., and Dunstan, F. 2006. Mental health, places and people: A multilevel analysis of economic inactivity and social deprivation. Health & Place 12: 332-344. • Fotheringham, A.S., Brunsdon, C., and Charlton, M.E. 2002. Geographically Weighted Regression: The analysis of spatially varying relationships. Wiley, Chichester. • Freeman, A.M. 2003. The measurement of environmental and resource values: Theory and methods. Resources for the Futures, Washington, D.C. • Garrod, G.D., and Willis, K.G. 1992. Valuing goods' characteristics: an application of the hedonic price method to environmental attributes. J. Environ. Manage. 34(1): 59-76. • Gayer, T., Hamilton, J.T., and Viscusi, W.K. 2000. Private values of risk tradeoffs at Superfund sites: Housing market evidence on learning about risk. The Review of Economics and Statistics 82(3): 439-451. • Gelfand, A.E., Banerjeeb, S., Sirmans, C.F., Tu, Y., and Ong, S.E. 2007. Multilevel modeling using spatial processes: Application to the Singapore housing market. Comput. Stat. Data Anal. 51: 3567 – 3579. • Geoghegan, J. 2002. The value of open spaces in residential land use. Land Use Pol. 19: 91-98. • Geoghegan, J., Pritchard, L., Ogneva-Himmelberger, Y., Chowdhury, R.R., Sanderson, S., and Turner, B.L. 1998. Socializing the pixel and pixelizing the social in land-use and land-cover-change. In People and pixels. Edited by D. Liverman, Moran, E., Rindfuss, R. and Stern, P. National Academy Press, Washington, DC.