Home Values, Employment, and Permit Data Issues and Limitations

A lot of the analysis I do is based on home value data derived from either the S&P/Case-Shiller House Price Indexes or OFHEO's Housing Price Indexes. I also use the Bureau of Labor Statistics Local Area Employment data and the Commerce Department's C-40 Residential Permit Reports for supply and demand trends. These data sets are useful because they drill down to metropolitan areas (there is wide variation in performance between markets). Also, they are published frequently (monthly for S&P Case-Shiller, BLS, and Commerce Department data, quarterly for OFHEO). However, like all data there are issues with each data set.

Although the S&P/Case-Shiller and OFHEO indexes use a similar approach to estimating value changes, there are important differences which are explained Andrew Leventis (an OFHEO economist) here. The S&P/Case-Shiller data has less complete geographic coverage, so data may not be available for some markets, and the aggregate index may be skewed by the smaller set of markets covered. The OFHEO index relies on data from Fannie Mae and Freddie Mac loans which means homes using jumbo loan financing are not represented. If, for example, prices on high priced homes are falling faster than more affordable homes and the OFHEO sample is overweighted with affordable homes the OFHEO index will not fall as fast as an index including the higher priced homes. When possible I try to use both indexes and keep in mind their potential biases.

The primary issue with the BLS employment data is explained in this post from The Big Picture. The BLS attempts to estimate employment changes from new businesses just starting up and from businesses which have terminated using a birth/death model. The consensus is the model tends to understate job losses during a slowdown and understate job gains at the start of a recovery. Again, the best approach seems to be an awareness of the potential bias.

Finally, the permit data issues relate to potential reporting and sampling errors. The Commerce Department summarizes the issues here.

None of these data sets seem fatally flawed, especially if the data is looked at over a period of time and averaged over a number of periods to miminize the impact of errors or distortions in any one period.