An investigation of the time between mortgage default and foreclosure

Abstract

This paper evaluates the speed at which lender's enter the mortgage foreclosure process, interpreting slower rates as evidence of forbearance. A telephone survey was conducted to evaluate the presence of lender discrimination in defaults that lead to foreclosure in New Orleans, Louisiana between 1985 and 1990. These data are used to estimate the independent effects of race and neighbourhood characteristics on the extent of lender assistance during the foreclosure process. The analysis does not reveal discrimination based on race of borrower or racial composition of the neighbourhood. Instead, greater time in default is granted loans with lower interest rates and loans with outstanding balances below the mean value of housing in the surrounding neighbourhood.

Keywords:

• race discrimination
• mortgage default and foreclosure
• lender assistance
• forbearance

Notes

Sampling only defaulted loans that end up in foreclosure may distort the importance of loan‐to‐value ratio in borrower calculations about how to resolve a defaulted mortgage. A critical difference between a sample of defaults that includes loans that are reinstated after workout and a sample restricted to defaults that result in foreclosure is that, regardless of borrower intention or lender strategy, any loan workout procedure requires the borrower to eventually pay the outstanding loan balance. Borrowers without employment or significant assets are not equally able as more affluent borrowers to take advantage of loan workout alternatives to foreclosure, regardless of the LTV ratio of the property. This reduces the option of extremely poor borrowers to avoid foreclosure, even if the mortgage has a low LTV ratio. This may not be a significant bias because less than 20 per cent of survey respondents were unemployed at the time of default (see Table 4). Regardless, a potential sample bias toward defaults by poor borrowers makes it important to reiterate that the analysis investigates potential discrimination in the administration of defaulted loans that end up in foreclosure, not potential discrimination in the administration of all defaulted loans.

Recent research supports the attribution of generally high reliability to retrospective survey responses (Giele & Elder, 1998). However, several issues complicate the use of retrospective data. Reliability of retrospective survey data appears to depend most on the issue queried (e.g. illegal drug use) rather than on the respondent interviewed. The contention is that survey respondents will not forget if a lender offered them forbearance in the mortgage default process prior to losing a home to foreclosure. This is particularly true because of the personal loss and potential embarrassment associated with losing property to foreclosure. The emotional nature of the experience etches the details into memory and increases reliability of retrospective responses to survey questions about the experience. Inaccurate recall of dates when particular past events occurred creates the problem of forward telescoping (Golub et al., 2000). The problem of forward telescoping does not affect data collected for this paper because important dates were coded from court records, not from survey respondents' memories. The foreclosure process begins with filing in Civil District Court of the Petition For Executory Process which, under Louisiana law, can usually be filed only after the mortgage is 90 days in arrears. Subsequent court filings may include a Notice of Seizure of Property, Notice of Appraisal, Notice of Demand, Notice of Foreclosure Sale, and a Sheriff's deed issued after the foreclosure sale. The dependent variable in the analysis, number of days delinquent before start of foreclosure action, was calculated from dates contained on the Petition for Executory Process. The petition contains the date of filing and the last date the mortgage was paid up (almost always a specific day, but if only the month was listed, the last day of that month was coded). Number of days delinquent before foreclosure is the difference between the day after the last date the mortgage was paid up and the date the petition was filed. Other dates taken from official court filings and used in the analysis include date of loan origination, outstanding loan balance at time of foreclosure filing and interest rate of original loan.

Multicollinearity could be a problem in these models were the correlation between any pair of independent variables to exceed 0.31. This was not the case in the models (see Appendix D), thus multicollinearity does not pose a problem for the interpretation of the coefficients.

Appraisals are not required for foreclosure proceedings in Louisiana. Since only 25 per cent of the survey respondents had property appraisals at the time of foreclosure, a property‐specific loan to value ratio (LTV) could not be constructed for the regression analysis. Instead, the value of a foreclosed property was measured as the 1990 mean value of owner occupied housing in the block group of that foreclosed property. This permits calculation of a ratio of loan balance of foreclosed property to the average value of owner occupied property in the neighbourhood where it is located. The weakness of this measure is that it is not a traditional loan‐to‐value ratio, which limits direct examination of economic theories of the default process. The specification used in this analysis may better capture the decision frame of lenders than borrowers because it compares amount owed on defaulted property with property values in the surrounding neighbourhood. This taps the importance of location in the foreclosure process but does not measure the borrower decision calculus so important in the option theory of foreclosure.

Models were run which included a variable that measures the difference between the interest rate at the time the loan was originated and the prevailing interest rate at default. This variable is not a significant predictor of days in default before foreclosure and the models that included this variable were much weaker than those reported in Table 6. Additional models were run which included a variable to measure the erosion of borrower equity between the date of default and the date of foreclosure (see Springer & Waller, 1993). This variable is not a significant predictor of time in default before foreclosure and is not included in the models reported in Table 6. Another model included change in percentage of black in the block group from 1980 to 1990, which was not a significant predictor and was not included in any models reported in Table 6.