Methodology: Negative Equity

For the last four years, Zillow has been calculating a measure of the percentage of homes with a mortgage that are in negative equity (or “underwater” on their mortgage). Like others who try to estimate this piece of data which is incredibly important to understanding the housing market, we had to make certain assumptions in the calculation. Having only the value of the loan at the time of origination (and, often, not all of the details on the loan such as interest rate or whether it was a fixed or adjustable mortgage), we were forced to estimate how much of the principal mortgage balance might have been paid off since origination. And for home equity lines of credit, we were also forced to make assumptions about how much of the credit line had actually been utilized by the homeowner.

What we’d really prefer to have when computing negative equity is the current outstanding mortgage balance for each homeowner, where this amount includes the first and second mortgages and any home equity loans or lines of credit taken out either at purchase time or subsequent to purchase. When creating our new approach to estimating negative equity, we set out to do just that. We partnered with TransUnion, a global leader in credit and information management, to obtain the actual current outstanding mortgage balance for homeowners and we paired this information with the home value estimates of more than 100 million homes created by Zillow.

Home values
We begin by obtaining a set of unique addresses for all homes in our database, the name of the current owner of the home as registered with the county tax assessor and an estimate of each home’s market value (Zestimate). We include all single-family, condominium and cooperative properties found in the Zillow database. Zillow home value estimates are created three times per week so we obtain the estimate for each home on the last day before the end of a calendar quarter (e.g., March, June, September, December).

Outstanding mortgage balance
This file of addresses, owner names and home values is sent to TransUnion. On the TransUnion side, these properties are matched to their consumer credit records on the basis of address and owner name. The match rate is highly dependent on the region with an overall average of about 80 percent.

At this point, a homeowner’s equity is computed as the difference between the estimated home value and the sum of all outstanding mortgage debt (including lines of credit) associated with the home. If the homeowner’s equity is negative, then the owner owes more on the mortgage(s) than the home is worth. After equity is computed and all personally identifiable information (“PII”) is removed from individual records, the data are returned to Zillow.

In 2012 Q1 we had roughly 83 million homes in our negative equity dataset, of which roughly 80 percent were matched to a homeowner’s financial data. Overall, this data covered over 800 metros, 2,100 counties, and 22,200 ZIP codes across the nation.

Processing the Data
Once the data are returned to Zillow, we aggregate it to the ZIP Code, county, metro, state and national levels. We then integrate Census Bureau data to each level of geography to ensure that we have a representative sample of homes for each geographic unit. If not, we do not report data for that geographic unit. In those geographies for which we do have a representative sample of homes, we scale up the absolute number of observations to equal the number of homes and mortgages as reported by the Census. For example, if we have 8,000 homes with a mortgage in a given ZIP Code whereas the Census reports 8,900 mortgaged homes, we scale up our estimates of mortgaged homes to equal the Census estimate (while holding the negative equity percentage constant).

We observe data on both estimated home value and current outstanding mortgage debt on approximately 35 million homes across the United States. This data is adjusted at the county level using Census data on owner-occupied homes with a mortgage to estimate a full footprint of approximately 50 million homes. This estimated footprint of 50 million homes compares to a total of approximately 53 million owner-occupied homes with a mortgage overall in the United States. The difference between these two numbers (50 million versus 53 million) reflects the number of homes in areas where our data coverage is not sufficient for estimation purposes. In short, our cumulative number of underwater mortgages (~16 million) should be compared to a base of approximately 50 million mortgages.

Handling renters and investors
One concern in pairing data tied to homes with data tied to individuals is that we minimize the number of homes occupied by either renters or investors/landlords in our final data. Rental homes are problematic because, in the event that the home has a mortgage, the mailing address for the mortgage (which TransUnion has) is likely that of the landlords’ home. This would mean that the address of the property supplied by Zillow would not match to any mortgage debt at that address in the TransUnion data, making it appear that the home does not have a mortgage and is owned free and clear.

The homes of investors/landlords are problematic because there may be multiple mortgages tied to the same address (e.g., the mortgage for the investor’s own home plus those for any properties owned by the investor). This results in a home appearing to have much more mortgage debt than it actually does.

To minimize these two issues, we apply a filter to both the data provided to TransUnion and the data returned by TransUnion. In the first filter, we eliminate from the Zillow data any homes for which the tax assessment bill is delivered to another address (homes of potential renters). We also eliminate from the data any homes to which the tax assessment bill for another home is sent (homes of potential investors).

Once we receive the data back from TransUnion, we limit the dataset to exclude observations that have more than four mortgage trade lines (first and second mortgages, home equity loans or home equity lines of credit). From our analysis, addresses with more than four mortgage trade lines are much more likely to be investors.

Our first negative equity report using this new methodology can be found here and online here.