December 21, 2007
As you can imagine, CBO prides itself on its analytical capabilities -- and its analysis is (justifiably, in my opinion!) widely respected. From time to time, though, criticisms of our analysis are raised. I won't respond to most of these criticisms, but in some cases, it may be worth examining the issues raised in a critique of CBO's analysis -- and that's what this post does.
A publication from a consulting firm claims that the discussion of housing price-rental ratios in recent CBO testimony was misleading. (For an entry in the Wall Street Journal's economics blog about the critique, see here.) Those claims, though, rest on a number of misunderstandings, both of the CBO testimony and of the underlying data.
With respect to the testimony, CBO used a figure -- showing the ratio of house prices to rents -- as a short-hand indicator to support the idea that housing prices were high relative to underlying fundamentals, particularly in 2005-2006, and that there was a general expectation that housing prices would be weak in the future. The figure was not used to say that the ratio of housing prices would necessarily fall back to the average of earlier periods in the near future. Here is the section of CBO's most recent testimony that referenced the figure:
"The outlook for home prices is highly uncertain, but it seems likely that house prices and, consequently, housing wealth will continue to fall next year. The inventory of unsold homes stands at high levels, which will place continued downward pressure on house prices in many regions of the country. Moreover, the ratio of housing prices to rents still seems very high relative to its history (Figure 4). To be sure, homebuyers expectations of home prices may deviate from long-term fundamentals for extended periods of time, and the pricerental ratio may therefore not provide a reliable guide to potential changes in prices over relatively short periods of time."
The section then references a 2004 paper from the New York Federal Reserve that discusses the pros and cons of using the ratio to determine whether the economy was in the midst of a housing price boom at that time.
With regard to the data used in the chart, some measurement issues have indeed been raised with regard to both the numerator (the OFHEO price index for houses) and the denominator (the PCE price index for owner-occupied non-farm dwellings, which is essentially the same as the owners' equivalent rent index in the CPI-U). The indexes used by CBO, however, are both widely accepted as useful indicators of what they purport to measure despite their limitations.
The critique of CBO's presentation centers on the owners' equivalent rent measure, so I will focus on that index. (For a brief explanation of the PCE rental measure, see here.) Although researchers have raised concerns about the index, little hard evidence exists about the magnitude of any problems associated with it -- and it remains widely used by analysts and government agencies. As some examples, the Bureau of Labor Statistics (BLS) stands behind the index, the Bureau of Economic Analysis uses it as a price index for a large category of personal consumption expenditures, and it accounts for about 12 percent of the core PCE price index that the Federal Reserve has indicated plays an important factor in its policy decisions.
The critique suggests that the rent index is somehow distorted because "when capital gains are high, a low rent is adequate to make the profits from renting a house equal to the interest that could be earned by selling it and investing the proceeds." That's not a distortion, though: it is simply another way of saying that when house prices are rising a lot, we should expect to see a large discrepancy between house prices and rents, as the figure in CBO's testimony illustrates.
Although the rental measure is widely used and thought to be a valuable indicator, researchers have raised some concerns about it. The most serious issue appears to be associated with its sample.
- The current BLS sample is based on 33,000 units and some cities are not included at all. (Data from cities that are sampled within a region are ascribed to similar-sized cities in the region.) Some researchers are concerned about the sample at the top end of the distribution. For example, the 2004 paper mentioned above found that a rent measure constructed for the 1990s using data from the American Housing Survey moved in concert with the BLS owners' equivalent rent measure for the first half of the 1990s, but diverged during the second half. The paper's summary noted that the differences may have occurred because the BLS sample of rental units is thin at the upper portion of the distribution. For the top-most portion of the distribution of home values, there are often few equivalent rental units in the same neighborhood.
- The BLS, though, argues that the sampling limitations do not undermine the usefulness of the index.
- A more significant concern may be that the sample was set in 1999 -- and there has been no updating of the sample since then, so it may no longer be representative of the mix of renters. BLS is concerned about this, but has not received funding to create panels and rotate the sample. (Virtually all other samples in the CPI are continuously updated.)
The key point, again, is that CBO used widely accepted indexes, which are known to have limitations but are still viewed as valid indicators, as an illustration of how housing prices appear to have diverged from underlying fundamentals. Even then, the testimony did not suggest that the ratio of housing prices to rents was necessarily a reliable predictor of future price movements in the near term.