Understanding Home Price Indices

A home price index is an economic measurement used to assess trends and the underlying strength of the housing market. This brief explainer article is meant to equip the reader with the tools necessary to understand strengths and biases of popular approaches to measuring home values at the macro level.

What’s a Home Price Index?

In the broadest possible sense, a home price index is an intended measurement of the overall value of the housing stock. Clearly, as more units of housing are added to the housing stock, the value of the aggregate stock of housing must increase. Therefore, the aggregate number tells us very little about how much individual homes are appreciating (or depreciating) in value. Therefore, virtually all home price indices try to capture some average or median home value, which by definition adjusts for the number of housing units. This approach sounds simple but it actually presents some annoying problems.

All Transaction Indices

We actually don’t know the exact market value of any home until it actually sells. No two homes are alike, and every unit of housing has a unique location. One popular approach to creating a home price index may be track the prices of all home sales over a given time interval and report the median or mean. This method is known as a all-transaction index. However, the homes actually put up for sale is not a random sample of all homes, so trying to extrapolate a population mean or median from it will entail bias.

In general, the bias is upwards, because higher-end homes tend to transact more often. Say a neighborhood has 10 homes. Eight of them are modest three-bed, two-bath homes, but two are significantly larger. If during a given period both large homes sell but the other homes do not trade, an all-transaction index would grossly overestimate home values for this hypothetical area.

Another bias created by an all-transaction index is that home sellers may be reluctant to sell their homes if they have doubts about the condition. Or similarly, many home sellers improve their homes in anticipation of a sale in order to fetch more money. This also creates upward bias in an all transaction index, because homes in worse condition which never go on the market are excluded.

Lastly, all transaction indices must contend with the bifurcation of the new versus existing home market. Fundamentally, newly constructed homes and existing homes are different product types and different statistical populations. The composition of new builds in terms of size and quality can change rapidly. Imagine a period where many condominiums were built followed a period where most construction was single family homes. Here, a index of new home sales is mostly measuring the composition in terms of size and quality of the homes being added to the housing stock, not an increase in home values per se. For this reason, most indices differentiate between new and existing home sales.

Repeat Sales Index

To address these biases and compositional effects, many indices use the repeat sales method. Rather than measure the median price of all homes that transact, the repeat sales index looks at the change in price since the last time a home sold. This controls for home size or quality, since the same homes are compared to themselves.

There are two main disadvantages of repeat sales indices. First, they are costlier and require more data to construct, and thus only exist for certain areas. Second, they trade one form of bias (composition) for another. Repeat sales indices must use homes that have transacted at least twice. This by itself is not an insurmountable problem since sales records often go back decades. However, the index must adjust for the time between sales, and most come up with a weighting system to determine how time between sales determines the impact on the overall index of a given data point. The intuition here is that two relatively close sales are likely to have data which is less “noisy,” and long time intervals between sales increase the probability that substantial improvements have been made to the property which affect value. The mathematics of the weighting system to account for time between sales is beyond the scope of this article, but it is not a perfect process and does not strip out all bias in the index.

Conclusion

Home price indices are useful tools but are not without biases. All transaction indices suffer from compositional effects because more valuable homes tend to sell more often. Repeat sales indices attempt to fix this issues by comparing homes with themselves over long time intervals. However, they are more expensive to construct and require more data. Lastly, the weighting system with regard to time between sales itself is imperfect and may create additional biases.