Transcripts of the monetary policymaking body of the Federal Reserve from 2002–2008.

Hardly a day goes by without another anecdote-laden article in the press claiming that the U.S. is experiencing a housing bubble that will soon burst, with disastrous consequences for the economy. Indeed, housing market activity has been quite robust for some time now, with starts and sales of single- family homes reaching all time highs in recent months and home prices rising rapidly, particularly along the east and west coasts of the country. But such activity could be the result of solid fundamentals underlying the housing market. After all, both nominal and real long-term interest rates have declined substantially over the last decade. Productivity growth has been surprisingly strong since the mid-1990s, producing rapid real income growth primarily for those in the upper half of the income distribution. And the large baby-boom generation has entered its peak earning years and appears to have strong preferences for large homes loaded with amenities.

One of the conditions of an asset bubble is that the price of the asset has risen well above what is consistent with underlying fundamentals. In the current debate, two measures of relative value have been applied to single-family home prices— price relative to income and rent relative to price. In the comments that follow I will concentrate on the price-to-income measure, but my conclusions apply equally to the rent-to-price measure.

In most analyses of the ratio of home price to income, the measure of home price that is used is the repeat-sales home price index published by the Office of Federal Housing Enterprise Oversight (exhibit 1). This index tracks changes in the average price of homes purchased (or mortgages refinanced) with loans purchased by Fannie Mae and Freddie Mac, or conforming conventional loans. Therefore, it excludes cash sales as well as purchases or refinancings financed with FHA [Federal Housing Administration], VA [Veterans Affairs], and jumbo conventional mortgages. It is called a repeat-sales index because it is derived by observing the sales prices—or appraised values, in the case of refinancings—of properties at specific addresses at two or more points in time. Finally, it is a transactions-based index in that it reflects the prices of homes that are sold (or refinanced) rather than the entire universe of single-family homes.

A lesser known home price index is the constant-quality new home price index published by the Bureau of the Census (exhibit 2). This index is based on a sample of new homes sold, regardless of how the sale was financed. Hedonic methods are employed to hold the physical and locational characteristics constant over time. This index is part of the Census Bureau statistical program through which the single- family residential investment deflator of the national income and product accounts is derived. As shown in exhibit 3, the increase in prices indicated by these two indexes is quite different. For example, over the four years from 2000 to 2004, the OFHEO index increased at a compound annual rate of 8.2 percent, while the constant-quality index increased at a 5.4 percent annual rate. As shown in exhibit 4, the current ratio of price over median family income derived from these two indexes is vastly different. If the OFHEO index is giving an accurate picture of what is happening to home prices, I think one could say with some confidence that prices have been bid up to unsustainable levels. However, if the constant-quality index is a better reflection of reality, home prices actually look somewhat low relative to median family income, particularly compared to the late 1970s. I believe the constant- quality index provides a more accurate indication of what is happening to the price of a typical single-family home. In contrast, the OFHEO index is subject to upward biases that accumulate over time and distort ratios such as price-to-income and income-to-rents.

To help us understand the biases in the OFHEO index, exhibit 5 presents the distribution by value of all single-family homes in the U.S. in 2003, with the specific values at the 25th, 50th, 75th, and 80th percentiles. 4 The median value in 2003 was $150,000 with the distribution skewed toward the right. The value at the 25th percentile was $90,000 while the value at the 75th percentile was $250,000. We do not know with certainty where the OFHEO index falls on this distribution, as it is an index rather than a series of values. But we can be reasonably certain that it lies somewhere between the average price of all existing single-family homes sold and the average price of homes purchased with conventional loans. That means the OFHEO index is a closer reflection of what is happening at the 75th percentile rather than the 50th percentile. Moreover, it is very likely that over time the point on that distribution represented by the OFHEO index has been drifting to the right. One cause of this rightward drift is what I call transaction bias. As shown in exhibit 6, the American Housing Survey (AHS) data suggest that both appreciation rates and turnover rates increase as one moves out the home value distribution. For example, from 1997 to 2003 the compound annual rate of appreciation at the 25th percentile was 4.5 percent, increasing to 8.7 percent at the 80th percentile. Corresponding average turnover rates for the period from 1997 to 2003 were 5.9 percent and 7.4 percent. That means, of course, that the average rate of appreciation of the units that turn over is higher than the average rate of appreciation of the entire distribution. While the amount of bias in any one year is likely to be small, it does cumulate over time and becomes quite important when one is comparing levels versus income or rents.

Another potential upward bias in the OFHEO index is that while it is a repeat- sales index, there is evidence to suggest that it is not a constant-quality index. In addition to the strong pace of new housing starts, another aspect of the housing boom of the past decade has been a significant increase additions and alterations to the existing housing stock. Exhibit 7 presents in the top panel the ratio of the OFHEO index over the constant-quality index plotted over the period from 1977 to the present. In the lower panel are plotted real improvements per unit of housing stock per year over the same period. Over the past decade real improvements per unit have increased about 25 percent, which appears to be associated with a further increase in the ratio of the OFHEO index over the constant-quality home price index. Research suggests that higher-income households have a higher income elasticity of demand for improvements to their primary residences, suggesting that this source of upward bias is likely to be more pronounced in the right half of the distribution of all single-family homes. 5

Another way of looking at the issue of home prices over income is to go back to the AHS data and see what is happening at various points on the distribution of all single-family homes. This information is presented in exhibit 8. At the 25th percentile the ratio of home price to income has been relatively stable, while it has increased sharply at the 75th and 80th percentiles, reminiscent of the price-to­ income ratio computed with the OFHEO index. Let me pause just a moment and emphasize that I am comparing home prices at a percentile with the incomes of the people who live in those homes at the same percentile. This chart is likely the equivalent of the finding that the increase of home prices has been most pronounced in areas of the country where home prices were already relatively high due in part to relatively inelastic supply. It is likely that in these areas of the country, where land values are high, the inclination to make substantial improvements to existing properties is the greatest.

Clearly, not everyone agrees that the constant-quality new home price index provides an accurate indication of what is happening to the price of a typical single- family home. For example, it has been argued that most new construction takes place at the fringe of metropolitan areas where land prices may not be rising as fast as is intra-marginal land. There are several counter arguments. First, the theory leading to the conclusion that intra-marginal land values increase at a faster rate than land at the fringe is based on a theory of the development of a metropolitan area with fairly restrictive assumptions. Modern metropolitan areas have multiple commercial/employment centers. Many households have preferences for rural or suburban residences over urban residences. Restrictions on development to counter suburban sprawl have reportedly resulted in sharp increases in prices of parcels of land suitable for home building. Finally, I would like to note that the increases in land prices implicit in the constant-quality home price index, shown in exhibit 9, are substantial, particularly in the Northeast and the West. These estimates were derived assuming that land represents 50 percent of the value of the total property and that the prices of the other inputs increase at the same rate in all regions of the country.

In closing, I would like to comment on one other aspect of the housing bubble issue that has received substantial attention. Earlier this year a major real estate related trade association released the results of a survey indicating that a significant percentage of single-family home sales were of investment properties and second homes. This was widely interpreted as evidence of speculative buying of rental properties, another feature of a housing bubble. Again, I believe that such reports should be viewed skeptically. According to the AHS, in 2003 single-family investment properties, defined as homes renter-occupied or for rent, represented 14.2 percent of all single-family homes while second homes represented another 4.7 percent. Therefore, we should not be surprised that such properties represent about 20 percent of the sales that take place at any point in time. Moreover, the American Housing Survey data indicate that single-family investment properties have been declining as a share of all single-family homes for some time and declined in absolute numbers from 2001 to 2003. A principal reason the rental vacancy rate for single-family homes has risen in recent years is that the number of renter-occupied single-family homes has declined as people switch from renting to owning. So if a lot of people are buying rental properties, it must also be the case that a lot are selling as well. That concludes my report. Thank you.

Keyboard shortcuts

j previous speech k next speech