Let me turn now to more specific comments on the housing sector. Private single-family housing starts peaked in January 2006 at an annual rate of 1.823 million units. Since that time, housing starts have continued to fall; in April 2008, they were only at an annual rate of 692 thousand units, roughly 38 percent of the previous peak value. As striking as these statistics appear, they are not unprecedented. To put the contraction in housing production in a longer-term perspective, it is useful to divide housing starts by the number of households. In February 2006, housing starts per household peaked at 1.65 percent. In March 2008, they had declined to 0.64 percent per household. A startling fall, to be sure, but it has happened before. In January 1973, housing starts per household peaked at 2.10 percent before declining to a trough of 0.94 percent in February 1975. The 1973-75 decline was similar in magnitude to our current situation, according to this metric. A similar phenomenon occurred later in the 1970s. In December 1977, housing starts per household were 2.03 percent before beginning to fall. They fell all the way to 0.66 percent in November 1981. These statistics remind us that housing can be a highly cyclical industry. They also illustrate that the current contraction has clear precedents. Indeed, it is perhaps comforting that the order of magnitude in the previous declines is not unlike today’s figures and that, when housing starts per household have fallen this low in the past, it was near a turning point.

Statistically, there are several ways to estimate the normal level of starts. A common method is to estimate a model of some sort. For our purposes, assume that single-family housing starts in any year is a function of three primary variables: interest rates, household income and demographics.(2) In 2006, a model of this sort projected that housing starts would total about 1.7 million units, about 12.5 percent more than the actual level of about 1.5 million units. Assume: 1) no change in the average level of interest rates this year (relative to 2006); 2) that real GDP increases by 3 percent this year; and 3) that the number of households increases by 1 percent. With these assumptions, the model predicts that single-family housing starts will total about 1.4 million units this year, which we can compare to the actual 1.5 million units in 2006. This projection for 2007, which would be a 2.5 percent decline from 2006’s average, appears to be at the high-end of most forecasters’ expectation, perhaps because the model just outlined makes no allowance for the overhang of excessive inventory at the beginning of this year. But, eventually, as the inventory is worked off, home-building activity should pick up substantially from the 2006 year-end level of about 1.2 million units at an annual rate.

[C]alculations about the sustainable level of housing starts based on demographic factors, such as population growth and household formations, suggest that starts may be closer to their trough than to their peak. Although such calculations are, in general, not particularly useful for near-term forecasting, they do suggest that any overbuilding in 2004 and 2005 was small enough to be worked off over coming quarters at close to the current level of housing starts.