There's been a lot of interest in housing so I thought I'd present some data. Before doing so I want to emphasize that these are correlations, I am not going to try and sort out causality with such a simple graphical approach.
The graphs below are the I, investment, in the familiar Y = C + I + G + NX function. Investment, I, has three parts, inventories, residential fixed investment, and non-residential fixed investment. I am going to just look at two of the components of investment, residential and non-residential fixed investment. First, here are the the two series as they come "out of the box":
However, these data can be deceptive because a decline of, say, 100 is a much larger percentage change at the beginning of the sample as compared to the end. Taking logs of the two series overcomes this problem. Here are the logged values of residential and non-residential fixed investment:
Note the relative stability of the residential component in recent years as compared to earlier years.
Another way to look at these data is to examine the percentage of total fixed investment accounted for by each of the two components. The following graph shows the declining share of residential fixed investment over time, though in recent years there appears to have been an increase departing from the past trend. It is too early to tell if this is a change in the trend itself, or a drawn out version of past cycles:
The next graph shows exactly the same information, but adds the NBER recession dates. There is a general, though by no means a certain, association between declines in the share accounted for by the non-residential component of housing and economic slowdowns:
How does this look for the individual components? Here's the log of non-residential investment along with the NBER recession dates:
As you can see, there is a fairly close association between recessions and declines in non-residential investment. The next graph shows the log of residential investment:
One characterization of these data is that the declines seem to be associated with the non-residential component more so than the residential component. But in the recovery, the period just after the shaded regions, there is a housing boom in every case (interestingly, the one slightly questionable episode is the most recent). There is one case around 1968 where the residential construction boom is not associated with a recovery, but in general the strongest movements in residential fixed construction are just after the troughs of cycles. Recessions are associated with the non-residential component while recoveries are associated with the residential component.
The 1991 recession is a good one to look at to make this point. The residential construction slowdown started years before the recession, as early as 1985 according to the last graph. If this caused the recession, there was a substantial lag involved. However, the slowdown in non-residential construction shown in the second to last graph seems much more proximate to the recession. After the trough, the recovery is associated with a strong increase in residential construction, but the decline in business investment continues beyond the trough in the cycle. The 2001 recession looks similar, a fall in non-residential investment is associated with the decline in output, and the fall in non-residential investment continues past the trough of the cycle. The recovery is associated with strong residential investment, though in this case the growth is not as fast as in previous recoveries.