Declining foreclosure neighborhood effects over time

Abstract

According to previous studies, residential foreclosures reduce the value of neighboring residential units and the initial negative effects decay over time and space. This study attempts to investigate the temporal path of the initial effects by following cohorts of single-family housing distressed sales (foreclosures and real estate owned sales) over time. A hedonic model estimation of single-family housing sales in Saint Louis County, Missouri, produced larger marginal impacts for new distressed sales in the year 2000 compared with the marginal impact of new distressed sales in 2007, that is, the marginal impact of new distressed sales is declining in at least one housing market. This result holds true for the distressed sale neighborhood impact, the effect of distress on the same unit's future sales price, and the discount on a distressed unit's current “liquidation sale” price.

Keywords:

• foreclosure
• housing
• prices

Notes

¹Scholars have also detailed the costs to local governments as they respond to housing vacancy issues; one analysis of these types of local expenses based on the experience of Minneapolis, MN, estimated that each foreclosure cost, on average, $27,000 to local government (Moreno 1995). A more recent study on foreclosures in Chicago, suggests that local governments spend up to $30,000 dealing with foreclosed properties involving personnel from as many as a dozen city neighborhoods (Apgar and Duda 2005). These scholars have generally explored the impact of various types of housing/community development-related investments on residential property prices. For example, scholars have found strong evidence that investments in new housing positively impacts sales values (Simons, Quercia, and Maric 1998; Ding, Simons, and Baku 2000; Ding and Knapp 2003; Ellen et al., 2001; Lee, Culhane, and Wachter 1999); additionally, some studies have shown a similar impact with the rehabilitation of existing housing (Lee, Culhane, and Wachter 1999; Ding, Simons, and Baku 2000), with some variation based on the type and scale of the project.

²Pennington-Cross (2006) did not use a hedonic model but rather compared the price appreciation of foreclosed properties to the average price appreciation in the MSA, across MSAs, as measured by the Office of Federal Housing Enterprise Oversight.

³Homeowners are encouraged to sell quickly in order to avoid Institutional sellers (Pennington-Cross 2006), are under a unique set of incentives, and may be willing to sell at a discount due to regulatory opportunity costs (Downs 1992; Cory and Zinn 1994). Both regulatory capital requirements and credit agencies encourage a small balance of nonperforming assets (Downs 1992).

⁴The source of bargaining power may not come from asymmetric information. Turnbull and Sirmans (1993) and Watkins (1998) found no statistical evidence of weak bargaining power for out-of-town and first-time buyers.

⁵See Goetzmann and Peng (2002), Fisher et al. (2003), and Lin and Vandell (2007).

⁶In a strong housing market, properties typically sell within five months, and within 10 months in a weak market.

⁷Results of all control variables (Σβ _aA) are presented and discussed in Appendix A.

⁸Immergluck and Smith (2006a) and Lin, Rosenblatt, and Yao (2009) also include foreclosure counts but they do not indicate if the observed sale is included in the count of past foreclosures.

⁹In contrast, Immergluck and Smith (2006a) used two spatial-temporal rings.

¹⁰The time was calculated as follows: (Count of months since January 2000)/12.

¹¹The Saint Louis County Assessor's office excludes from arms-length housing sales all foreclosures, liquidations, transfers, and “quit-claim” deeds. According to the County Assessor's offices the foreclosure category includes all foreclosures, liquidations, real estate owned sales (REOs), and some sales during the default stage, but not estate sales and other sales with unusual amounts of personal property. Separate attempts by the author and the County Planner to separate units that sold during foreclosure and REO sales were unsuccessful, as automated techniques required many arbitrary assumption about what identifies a foreclosure, and reading through each deed was impractical.

¹²Immergluck and Smith (2006a) do not report their variables' descriptive statistics, so it is difficult to compare my results to theirs. One would expect population density to be a major contributor to the difference in foreclosure results. However, it is not clear that the average population density in their dataset is much different from mine. According to the 2000 Census, Saint Louis had 2,000 people per square mile. In Illinois, the population densities are as follows, Cook County 5,700, DuPage County 2,700, Lake County 1,400, Kane County 800, and Will County 600.

¹³Alternative weights of 15 and 20 nearest neighbors were also tested using a Lagrangian Multiplier test (Anselin 1988) but the 10 weight produced the most significant results. The use of other weights did not change the results in any significant way.

¹⁴See Anselin (1988) for a detailed taxonomy of spatial models.

¹⁵Virtually all parametric functional forms used for hedonic models are nested in the quadratic Box-Cox form as presented in Halvorsen and Pollakowski's (1981) seminal paper; however, for brevity, the three most common forms are referred to directly.

¹⁶A wealth of housing studies has indicated that a quadratic form in the age variable is appropriate. While all housing units depreciate, relatively older units are more likely to receive reinvestment.