|
Modeling Spatial Spillover Effects From Rental to Owner Housing: The Case of Seattle Author: Julia Koschinsky Dissertation School: University of Illinois-Urbana-Champaign Pages: 172 Publication Date: May 2008 Availability: Access Number: 10838 Abstract: The successful research question under which conditions subsidized housing generates spillover effects for values of neighboring single-family homes is extended in two ways: 1) By incorporating unsubsidized rental housing in the context of single- and multifamily zoning, and 2) by exploring in how far spatial dimensions of the data, research design, and methodology affect the reliability and precision of impact results from state-of-the-art adjusted interrupted time series/difference-indifference (AITS-DID) models. Although these models are widely applied, little systemic research exists that identifies and assesses potentially remaining sources of bias. A sensitivity analysis assesses the spatial heterogeneity of the AITS-DID results from the OLS estimator and compares them to results from locally weighted regressions and cross-sectoral spatial lag and error regression models (with spatial regimes). A comprehensive dataset was compiled for the city of Seattle, Washington (1987-97), including 52,142 sales records, 302 subsidized project-based and 1,156 tenant-based sites, 369 unsubsidized multifamily apartments, and aerial images. The main substantive finding is that, if rental spillover effects exist, they are generally associated with neighborhood upgrading of lower-value micro-areas. However, this result is very context-sensitive: It only holds for areas in single-family zones, near wealthier neighborhoods, with either higher or lower racial diversity, and without prior subsidy exposure. In contrast, none of the subsidized or unsubsidized rental sites impact prices in multifamily zones. Larger concentrations of vouchers are the only rentals with net negative spillovers - in both low-income and high-income areas. Plausible explanations are Seattle's comprehensive dispersion policies and gentrification (with a possible net loss of low-income rental housing). The sensitivity analysis suggests that, at least for the case of Seattle, the AITS-DID model is very sensitive to which years are pooled and which spatial subsets are taken. The fact that the pre-post relationship is not found to be robust in the case of Seattle is cause for concern since it is at the heart of the analysis. Spatial fixed effects did not fully control for spatially correlated sales prices as expected (controlling for spatial autocorrelation tends to primarily make a difference for impact significant at a 0.05 level although there are important exceptions). |
|
|
Last updated: Wednesday, March 06, 2013 HUD USER Helpdesk P.O. Box 23268 Washington, DC 20026-3268 Toll Free: 1-800-245-2691 Fax: 1-202-708-9981 Email: oup@oup.org |
 |
Privacy Statement |
     |
