Foreclosure starts on loans made between 2004 and 2007 in Sioux Falls, South Dakota, are examined using a combination of local and Home Mortgage Disclosure Act (HMDA) data, expanding on previous research restricted to a single year’s loans using an alternative meth-odology. Specifically, foreclosed loans are identified from local lis pendens records, and loans are matched with corresponding records in the HMDA database to provide borrower and loan characteristics for further analysis. The research employs limited dependent variable models to assess the effect of loan, demographic, and neighborhood characteristics on the likelihood of foreclosure. Key loan characteristics such as subprime interest rate, loan-to-value ratio, and additional loans at the time of the mortgage were found to have a significant effect on foreclosure starts. Minority borrowers were also found to have a significantly higher probability of foreclosure.