Nondisclosure laws and other factors have hindered the production and release of the median home prices in many areas across the country. This paper attempts to fill the gaps and develops a simple approach to estimating median prices for a geographically complete set of areas. The methodology begins by aggregating mortgage-level data from Fannie Mae, Freddie Mac, the Federal Housing Administration, and First American CoreLogic. Redundant observations are removed from the aggregate data and median home prices are calculated from the resulting dataset. Using county recorder data obtained for several hundred counties across the United States, the methodology then estimates the relationship between the calculated median prices and medians determined from the recorder data. The paper illustrates how this relationship can then be used to extrapolate median home values for areas for which recorder data are unavailable. Because one potential application of this approach would be to support the setting of conforming loan limits in high-cost areas, which are calculated as a function of median prices, actual 2009 high-cost-area conforming loan limits are then compared to the loan limits that would have been in effect had the approach been used to set the limits. The paper also briefly discusses the application of the basic methodology to estimate a national average home price, a statistic important for the setting of the national conforming loan limit.

Although house price medians are regularly produced for some of the largest metropolitan areas in the United States, no single source provides a geographically complete set of medians for the entire country on a regular basis. The Federal Housing Finance Agency (FHFA) produces house price indexes (HPI) that reflect price changes for virtually every county in the United States but price level information (such as average prices or median prices) currently is not released. The data sample used by FHFA to construct its indexes is constrained to homes financed with conforming mortgages. This constraint could introduce bias into any median or average price measure that might be produced.2 To be sure, broader theoretical problems plague any summary measure of home values that might be constructed by FHFA or any other entity. For example, sample selection bias can be a significant problem if measurement of values only occurs when a specific event happens. When transaction prices (e.g. sales prices) are used to measure median home values, for instance, that metric can be biased if homes that transact in any given period are not representative of the underlying housing stock.