Several observers have taken solace from the jump in housing starts; however others have noted the implications of the decline in housing permit applications are not so rosy.
Figure 1 shows the relationship between housing permits, starts and sales. As has been noted by CR among others, permits (blue) are a leading indicator for starts (red) and starts (red) are leading indicators for completions (not graphed) and sales (green). [Correction 8pm 3/22].
Figure 1: Housing permits, starts (1 unit) and sales (1 unit), in thousands, SAAR. Source: St. Louis Fed FRED II.
How does this tie in with the general economy? There are obviously many links, but I want to focus here (again) on the link between residential and (the larger component of GDP) nonresidential investment, depicted in Figure 2.
Figure 2: Log nonresidential fixed investment (red), equipment and software investment (blue) and residential investment (green), SAAR. Source: NIPA release of 28 February 2007.
As noted here, and repeatedly at CR as well (see here too), residential investment leads nonresidential investment. This bodes ill, given the severe dropoff in residential investment. In Figure 3, the q/q annualized growth rates for these series are displayed for the last eight years.
Figure 3: Quarter on Quarter growth rates in nonresidential fixed investment (red), equipment and software investment (blue) and residential investment (green), SAAR, calculated using log differences. Source: NIPA release of 28 February 2007 and author’s calculations.
To quantify the maximal correlations in these leads and lags, I calculated the cross-correlogram between changes in (logged) nonresidential and residential investment over the 1967-06 period. Figure 4 shows that the maximal correlation between nonresidential investment growth and residential investment growth is at lag 4 (i.e., residential investment leads nonresidential by about a year).
Figure 4: Cross-correlogram between Quarter on Quarter growth rates in nonresidential fixed investment and residential investment, SAAR, calculated using log differences. Source: NIPA release of 28 February 2007 and author’s calculations.
[Late addition: the set of bars on the left are the correlations between (the change in the log of) nonresidential investment and (the change in the log of) residential investment, at lags 0, 1, 2, ..., 8; the set of bars just to the right of the first set are the correlations between (the change in the log of) nonresidential investment and (the change in the log of) residential investment, at leads 0, 1, 2, ..., 8. The dashed lines indicate statistically significance; hence all the correlations from 0 to 6 lags are significantly positive.]
All the foregoing is suggestive of a decline in nonresidential investment in the next few quarters (and as shown in the graphs, growth in 06Q4 was near zero).
[Late addition: 8:15pm Pacific, 3/22]
Suppose the correlations that held over the 1967-2006 period hold into the future. Using the following regression:
0.106(dlRt-2) + 0.018(dlRt-3) +
0.162(dlRt-4) + ut
Where R2 = 0.29, n=155, DW = 1.305, dlNR is the first difference of log NonResidential investment, and dlR is the first difference of log Residential investment, SAAR, Ch.2000$.
To forecast dynamically nonresidential investment leads to the following figure. In this back of the envelope simulation, I’ve assumed that 2007 residential investment stays at 2006Q4 levels (i.e., we’ve “bottomed out”), and then rises at 4% per quarter in 2008 — mimicking what occurred in the recovery from the 1990-91 recession.
Figure 5: Predicted level of nonresidential fixed investment (red) using the indicated regression, and assuming levels of residential investment indicated in text. Gray shaded area is projection period. Source: NIPA release of 28 February 2007 and author’s calculations.
This is not really a model simulation, but merely an extrapolation using historical correlations. Nonresidential investment continues its moderate decline. Of course, I’ve assumed a pretty benign outcome for residential investment. Time will tell if I was an optimist or a pessimist.