«DiscoVeriNg HomelessNess Volume 13, Number 1 • 2011 U.S. Department of Housing and Urban Development | Office of Policy Development and Research ...»
2. Time to foreclosure. For those whose homes ended up being sold in foreclosure, we measured the time to foreclosure by calculating the gap between the foreclosure sale date and the bankruptcy filing date for those who filed for bankruptcy. For homeowners who did not file for bankruptcy, time to foreclosure is defined as the difference between the foreclosure sale date and the foreclosure initiation date.
3. Lender loss rate. We defined lenders’ losses as the difference between the mortgage plus arrearage outstanding and the sale price adjusted for inflation and house price growth. Lender loss rate is the ratio of the losses and mortgage outstanding. We used the sheriff’s sale price as the foreclosure sale price if the house was sold to a third party. If the sale was a lender buy-back, we use the price at which the lender subsequently sold the house to a third party. In a few cases, we observed a symbolic $10 sale price when the lender sold its repossessed house to another institution. We excluded these cases from our analysis of sale prices.
Exhibit 3 presents summary statistics of these three quantitative measures. About 28 percent of debtors lost their houses to foreclosure despite filing for bankruptcy. The foreclosure sale rate rose to 41 percent for those who had been delinquent on their mortgages for 12 months or more. We examined homeowners whose houses were listed as in foreclosure sale between August 1, 2001, and August 1, 2002, and who did not file for bankruptcy, although they are not directly comparable.
We found that 43 percent of this group lost their houses to foreclosure by October 2007. Most of the auction sales were lender buy-backs (that is, the sheriff’s auction did not generate a third-party sale) although most of the lender buy-back properties were sold to third parties within 1 year.
The time-to-sale measurement of bankruptcy filers’ homeownership experience captures the tension between borrowers and lenders in foreclosure sales. When the ultimate outcome is foreclosure sale, the longer a homeowner stays in his or her house without making proper payments, the more benefits the homeowner enjoys and the greater cost the lender bears. In our sample, the average time between bankruptcy filing and foreclosure sale is about 28 months. On balance, the longer a filer has been delinquent at the time of bankruptcy filing, the shorter the time to foreclosure. However, even for filers who were already 1 year delinquent on their mortgage payments at the time of filing and who, without filing for bankruptcy, would most likely be in foreclosure already, the average foreclosure sale did not occur for 26 months. This time to foreclosure is 10 months longer than the average foreclosure sale length, defined as the days between the first foreclosure notice and the eventual sale recorded by the Sheriff’s Office between 2001 and 2007.18 This finding is consistent with Capozza and Thomson (2006), who, starting in 2001, tracked a sample of seriously delinquent subprime mortgages for 8 months and found that loans that move from delinquency to bankruptcy ultimately resolve in foreclosure and eventual disposition of the real estate collateral (Real Estate Owned, or REO), but the time to get there can be quite extended. About 86 of the foreclosures occurred after the termination of the bankruptcy cases.
Foreclosed houses in our sample sold for, on average, $97,241 in 2001 dollars, less than the average $121,149 for which other foreclosed houses were sold between 2001 and 2007 in New Castle County. Third-party sales generated higher median prices but lower average prices than lender buy-backs. A longer time to sale is also associated with a lower sale price; the correlation coefficient of the gap between bankruptcy filing and foreclosure sale and the final foreclosure sale price to third parties adjusted for inflation and house price growth is -0.16.
To isolate the worst of the nonfiling homeowners to see if they look more like filing homeowners, we studied the foreclosure length at the 75th percentile for our nonfiling homeowners. The length of the foreclosure is 20 months, which is still one-half year shorter than the foreclosure length of our bankruptcy sample.
Finally, in nominal terms, the foreclosure sale price generally exceeded the estimates that owners made when they filed for bankruptcy because of the overall run-up in house prices during that period.19 After we adjusted for house price growth using the Office of Federal Housing Enterprise Oversight (OFHEO) house price index, the real sale price amounts to about 91 percent of what homeowners estimated when they filed for bankruptcy. Assuming a 20-percent foreclosure cost,20 we calculated lenders lost, on average, 28 percent of the face value of their debt, and the median loss rate is 31 percent. In absolute terms, the loss was $33,516 for the average house price and $23,156 for the median price.
Panels a, b, c, and d of exhibit 4 chart the distribution of the time between foreclosure sale and bankruptcy filing and between foreclosure sale and bankruptcy termination, as well as the ratio of sale price to estimated property value at the time of bankruptcy filing, adjusted for inflation and house price growth and adjusted for lender loss rate.
.02.02.01.01 0 10 20 30 40 50 −40 −20 0 20 40
.5.5 0.5 1 1.5 2 2.5 −1.5 −1 −.5 0.5 1 Grover, Smith, and Todd (2006) also found in their 2002 sample of mortgage foreclosures in Hennepin and Ramsey Counties, Minnesota, that the strong and appreciating housing market in the early 2000s had a positive effect on the sheriff’s foreclosure sale price. Contrary to our findings in the bankruptcy sample, however, they found that most foreclosed properties were sold for more than the outstanding amount due on the mortgage for the foreclosure sample.
Stark (1997) found that the costs amounted to 19.1 percent of the final judgment amount—the amount mortgage borrowers owed to lenders—in 1993 foreclosure sales cases and 18.43 percent of the final judgment in the 1994 sale cases.
Empirical Analysis and Policy Implications In this section, we analyze circumstances related to the loan, borrower, and lender that affect the probability that a certain homeownership result will occur. We then conduct a policy analysis using our empirical model.
Estimation Results In this subsection, we present the results of our estimation of the three measures of households’ homeownership experience after filing for bankruptcy.
Foreclosure Outcome The first-stage estimation is a Probit estimation of a foreclosure outcome; we estimated the probability that a filer’s house would be foreclosed on during our sample period. Our explanatory
variables included the following:
• Information on debtors’ housing situations at filing as characterized by whether the filer has been more than 1 year delinquent on his or her mortgage.
• Mortgage LTV.
• Whether the house is unaffordable (mortgage-debt-service ratio exceeds 50 percent).
• House tenure.
• Whether the filer was already in foreclosure at the time of bankruptcy filing and, if so, whether it was a tax lien foreclosure.
• Whether the mortgage was lent by a lender classified as subprime.
• Whether mortgage lenders are local.
• Household characteristics, such as previous bankruptcy experience, whether filer hired an attorney, job tenure, marital status, household size, and adverse events the filer may have experienced, such as a recent divorce or unemployment spell.
• Income and financial information, summarized by whether the ratio of medical debt to total debt in default exceeds 10 percent, debt in default as a portion of monthly income, assets relative to total debt, and mortgage arrearage relative to total debt in default.
We also included local economic information, such as filers’ income relative to local median income, filers’ house value relative to the local median house value, the percent of households living in poverty, and local housing maintenance costs (local median housing costs relative to the local median house value for mortgaged properties).
Finally, we introduced two additional variables to capture filers’ expectations regarding local house price and unemployment rate growth rate over the next 5 years at the time of filing. In particular, we used the house price index constructed for Delaware by OFHEO at the annual frequency and the household unemployment rate for New Castle County from the Bureau of Labor Statistics 126 Refereed Papers The Homeownership Experience of Households in Bankruptcy at the quarterly frequency and a simple (rolling) autoregressive forecast model with four lags as households’ forecast model. Thus, filers update their house price forecast annually and their unemployment rate forecast quarterly. Note that we adopted different forecasting frequencies for the two series to avoid perfect collinearity of the two variables. Households that filed for bankruptcy in 2002 expected a higher house price growth rate going forward than those that filed in 2001, which is consistent with ex post movements in house prices at that time. The expectation of unemployment rate growth has no clear quarterly pattern.
the filer has hired an attorney. Adverse events, such as unemployment, on the other hand, increase the likelihood of foreclosure significantly. Among financial and income variables, mortgage arrearage relative to total debt in default increases the probability of foreclosure. None of the regional economic variables matter, with the exception of housing costs: in areas with higher local housing costs—measured as the sum of mortgage payment, property tax, utilities, and insurance—it is more likely that the filer will lose his or her house to foreclosure.
Time to Sale Exhibit 6 reports our second-stage estimation results concerning the time between foreclosure sale and bankruptcy filing. We present results with and without the Heckman selection bias adjustment (Heckman, 1979), and they look similar for variables of significance. Explanatory variables at this stage include, at the time of bankruptcy filing—
• Whether the filer has been delinquent on his or her mortgage payment for more than a year.
• His or her mortgage LTV ratio.
• Whether his or her house is unaffordable, as defined by HUD.
• Whether the house was in foreclosure and, if so, whether it was a tax lien foreclosure.
• Whether the filer hired an attorney.
• Whether the lender is headquartered in Delaware, Pennsylvania, or Washington, D.C.
• Local ZIP Code-level economic information summarized by the filer’s income relative to the local median household income.
• The filer’s estimated house values relative to the median local house value, local poverty rate, and local housing costs as a percentage of median house value.
• The filer’s expectation of future house price and unemployment rate movement at the time of filing.
The other variables regarding filers’ income and financial information and demographics affect the time to sale only through their effect on the probability of the house being foreclosed.21 Among all these variables, the ones that explain the time to sale at the 5- or 10-percent statistical significance levels are—
• Whether the foreclosure was initiated by the county government because of tax delinquencies.
• Whether the mortgage is subprime.
• Local housing affordability as measured by loan median annual housing cost relative to median house value for houses with mortgages.
• The estimated house price growth rate over the next 5 years.
Tax lien foreclosures substantially lengthen the time between the bankruptcy filing and the foreclosure sale (by about 13 months). This is because tax lien foreclosure in Delaware is the only foreclosure that provides borrowers redemption rights for up to a year. The presence of redemption rights complicates the foreclosure process and discourages potential buyers from purchasing the property. It also takes longer to foreclose houses in areas with higher housing costs and houses with subprime mortgages. By contrast, foreclosed houses sell faster at times when households expect higher housing price growth.
Lender Loss Rate Exhibit 7 presents the second-stage estimation results regarding lenders’ loss rate. Remember that the loss rate equals the sum of the sales price adjusted for inflation, local house price appreciation, and transaction cost, minus mortgage outstanding, all as a proportion of mortgage outstanding.
Note also that, for lender buy-backs, we used the price at which the lender subsequently sold the property to a third party. In the event that the sale price was booked at a symbolic $10 or less, we excluded the observation from our estimation.
The explanatory variables in the second stage are exactly the same as those used to estimate time to sale. To reiterate, the variables included at bankruptcy filing are whether filers have been late on their mortgage payment for more than a year; mortgage LTV ratio; whether housing cost exceeds We exclude homeowners’ income and finance variables in our second-stage estimation to achieve identification. These household-specific variables affect the time to sale mainly through their influence on whether the house will be foreclosed in the first place.
50 percent of household income; whether the household was already in foreclosure and, if so, whether it was a tax lien foreclosure; whether the filer hired an attorney; the same lender information that was used in the first stage, including whether the lender is a subprime mortgage lender and whether the lender is headquartered in Delaware, Pennsylvania, or Washington, D.C.; lender and local ZIP Code-level economic information summarized by filers’ income relative to the local median household income; filers’ estimated house values relative to the median local house value;
the local poverty rate; and local housing costs as a percentage of median house value.