New FHFA Released Zip-Code Housing Price Indices

Facebooktwittergoogle_pluslinkedinmailFacebooktwittergoogle_pluslinkedinmail

FHFA 1 - Main PhotoFHFA 1 - Main Photo 2

Highlights from the Recent FHFA working paper:
Local House Price Dynamics: New Indices and Stylized Facts

In a recently released working paper[1] the Federal Housing Finance Agency (FHFA) has introduced new, experimental ZIP code level price indices.[2] The creation of the data set is useful for examining sub-city trends in housing prices, and in particular, for providing observations against which predictions of academic theory can be tested. The authors have not only developed the price indices, but also provided examples of how the data might be used to test academic theories, such as the standard urban model.

Conclusions from the new data

Based on the new data, the authors find support for the following:

  • In large cities, “appreciation rates and volatility between 1990 and 2015 are higher near the central business district (CBD) than in the suburbs.”
  • The data show a “steepening house price gradient[3] in large cities, suggesting that the demand for locations near the center-city is, in fact, increasing.”
  • The data are “consistent with a relatively high elasticity of housing supply in suburbs and in smaller cities.”
  • The “results are also broadly consistent with the Land Leverage Hypothesis (LLH)… [which] predicts higher house price appreciation and volatility in high land leverage areas.[4]

Characteristics of the new data

In relation to existing sources
Most currently available local data is at the city-level at relatively high frequency (monthly or quarterly). While high frequency is great for illuminating changing market conditions, it creates a gap in knowledge at the sub-city level. The ZIP code price indices address that gap by providing more granular local data (ZIP code) on a less frequent basis (annually).

Source of the microdata
The ZIP code House Price Indices (HPIs), like other FHFA HPIs, use a weighted repeat sales methodology in which successive home sales of the same property are observed so that constant-quality price changes can be measured. The ZIP code HPIs are based on the “all-transactions” dataset, meaning that the new measures are estimated using sales prices and appraisal values (from refinance transactions) for properties with mortgages guaranteed by Fannie Mae and Freddie Mac.[5] Data are filtered to remove sales pairs that may indicate a change in quality, namely property “flips”—transactions within 12 months of one another—and transactions with absolute annual average appreciation rates greater than +/-40 percent[6].

Coverage of the indices
Of the roughly 43,000 ZIP codes in the US, nearly 18,000 ZIP codes have a house price index for some period of time and roughly 13,500 have 2015 data.

Limitations of the data
Given relatively small sample sizes, producing a set of quarterly or monthly measures for such small geographic areas was fraught with complications. Indeed, even given the focus on annual measures, for some five-digit ZIP codes, small sample sizes can make the five-digit ZIP code indices highly volatile.

Caution for data users
Users of the new indices should note that, although many of the data series are extremely volatile (particularly in early years), no data substitution or data “smoothing” procedures have been implemented. Decisions on whether and how to make adjustments for extreme volatility have been left to the user.

Highlight of some detailed findings:

Appreciation rates are faster near the city center
In the images below, we see the authors’ estimates of average annual appreciation rates by ZIP code for four major metropolitan areas. The concentration of dark blue near each city’s center shows the higher average annual rates of home price appreciation in these areas.

Additional uses of the data:
Top Appreciation ZIP codes from 1990 to 2015
The ZIP codes with the highest level of price appreciation from 1990 to 2015 are scattered across the country with at least one ZIP code from each region represented in the top 25.

FHFA 2 Zip Codes

Top Appreciation ZIP codes from 2000 to 2015
The ZIP codes with the highest level of price appreciation from 2000 to 2015 are also scattered across the country. Again, at least one ZIP code from each region is represented in the top 25.

FHFA 3 Zip Codes second

Appreciation in Philadelphia, PA ZIP codes from 2000 to 2015
Within a city, there can be a high degree of variation in home price change by ZIP codes, especially over time. This is depicted visually for a handful of areas above, but it is worth thinking about what these differences can mean in dollars. Looking at the average rate of price appreciation from 2000 to 2015 in the ZIP codes HPIs in Philadelphia, PA, we find a low of 3.0 percent (19142) to a high of 9.4 percent (19125). Home values in both areas were quite similar in 1999 according to the 2000 census; the price in 19125 was $44,000 while the price in 19142 was $43,700. The FHFA ZIP code repeat sales price index would imply a price of $170,050 in 19125 and a price of $68,200 in 19142. By comparison, the most recent American Community Survey estimate in these areas showed a median price of $161,500 in 19125 and $82,000 in 19142—a narrower difference in appreciation rates than seen in the FHFA data.

[1] Bogin, A., Doerner, W. and Larson, W. (2016). Local House Price Dynamics: New Indices and Stylized Facts. Federal Housing Finance Agency, Working Paper 16-01. The working paper is accessible at http://www.fhfa.gov/papers/wp1601.aspx.

[2] ZIP code level data can be found here: http://www.fhfa.gov/DataTools/Downloads/Documents/HPI/HPI_AT_ZIP5.xlsx
and the visualization here:
http://www.fhfa.gov/DataTools/Tools/Pages/HPI-ZIP5-Map.aspx.

[3] The house price gradient describes housing prices in relation to distance from the central business district, and derives from the monocentric model of urban form first proposed by Alonso (1964), Muth (1969), and Mills (1967). Since the central business district in this model is the source of jobs and commerce to which workers must commute, rents (and house prices) are expected to be higher as distance to the central business district is smaller.

[4] Bogin, A., Doerner, W. and Larson, W. (2016). Local House Price Dynamics: New Indices and Stylized Facts. Federal Housing Finance Agency, Working Paper 16-01. See pages 1-2. http://www.fhfa.gov/PolicyProgramsResearch/Research/PaperDocuments/wp1601.pdf

[5] The benefit of the all transaction index is that it yields many more observations. Some drawbacks include that not all of the observed prices are “market” transactions with a buyer and a seller. Additionally, Fannie and Freddie data by definition excludes Federal Housing Agency (FHA) mortgages (many first-time home buyers) and non-conventional mortgages (jumbo).

[6] In its officially released HPIs, the FHFA also implements a small number of supplemental data screens which have not been applied to this data.