Wednesday, September 2, 2015

Forecasting the 2016 election economy: using housing permits toforecast the unemployment rate

D  - by New Deal democrat

One of the best economic indicators for the electoral success (or not) of the incumbent party in Presidential elections has been whether the unemployment rate during the 2nd and 3rd quarters of the election year is declining or not.

Among economic indicators, the unemployment rate is unusual.  In and after recessions, it is a lagging indicator.  At economic peaks, however, it is a leading indicator.  In other words, it is overly sensitive to the downside, turning negative before a recession, but remaining negative after the recession is over.

One good way to anticipate the unemployment rate is to look at long leading indicators.  Specifically in this post I am going to examine the relationship between building permits and the unemployment rate.  The interactive graph of housing permits and the unemployment rate can be found here .

To give you the easy overall picture, let's put up the graphs.  First, here are housing permits (blue, left scale) and the unemployment rate (red, inverted, right scale):

The median number of months by which the unemployment rate lags housing peaks is 12 months, and for troughs is 10 months.

That tells us that the unemployment rate should continue to trend lower through at least about next May.

The next two graphs are the YoY% change in permits (blue) vs. the YoY% change in the unemployment rate (red, inverted), for 1964-84, and then for 1985-2015:

That permits lead the unemployment rate is obvious .

Further, we can say with certainty as to the last 50 years that when housing permits have advanced by more than 10% YoY, the unemployment rate YoY has never turned negative.  Similarly, with one exception (December 1987 and January 1988), whenever permits have been negative by more than 20%, the unemployment rate has always turned negative.

Unfortunately, from December 2013 through April 2015, permits while positive were up by less than 10% YoY. Here the record is more mixed.  In the last 50 years, there have been similar slowdowns 8 times.  On 4 of those occasions, the unemployment rate never turned negative YoY. On the other 4, unemployments did turn negative.  Conversely, in the last 50 years, there have been 6 times that permits turned negative, but less so than 20%.  Of those occasions, the unemployment rate also turned negative YoY twice.  In the remaining 4 times, the unemployment rate YoY remained positive.

Turning back to the 4 occasions where a slowdown in the growth in housing permits led to a YoY increase in the unemployment rate, the lead times were 8, 13, 13, and 14 months on the way down and -1, 2, 6, and 14 months, respectively.

In other words, if the slowdown in permits that began in December 13 were to have led to a YoY increase in the unemployment rate, based on the admittedly limited sample we would expect that downturn to have already started, and a coin flip as to whether it would already be over.

Even where there has been a downturn in permits without a recession, the negative turn in the unemployment rate began between 2 and 11 months later, and ended from 1 to 12 months later.

Finally, the trough in the unemployment rate has followed the trough in permits (including both slowdowns and downturns) by a median of 5 months.  Since the YoY trough in permits occurred in April 2015, this means the YoY trough in the unemployment rate should occur this month (September) and almost certainly no later than April 2015.

Finally, here is a slightly different look: the YoY change in permits (blue, in 100,000s) vs. the YoY change in the unemployment rate (inverted, red):

Here is a close-up of the last 5 years:

While the same general leadling/lagging relationship is apparent, this measure highlights how unusual 2014 was.  Permits increased YoY generally all year, but only slightly, i.e., by less than 100,000.  In the last 50 years there have been only 2 similar periods, 1962, 1968 (somewhat), and 1986 (shown in the graphs below).

Typically, permits have merely "passed through" the range of 0 to +100,000 YoY on the way to a stronger expansion on the way up, or a recession on the way down, so the extended pause was anomalous.  

In 1963, the unemployment rate actually rose slightly (by less than 0.5%) in response to the poor housing market the year before. The lag between the beginning of the pause in permits and the rise in the unemployment rate was 9 months. The trough in permit growth was also 9 months after the onset.  The worst YoY unemployment rate comparison happened both 9 and 11 months after the onset of weakness in permits.  The lag at end of the pause was 16 months.  This is the "worst case" scenario for the incumbent party. If this were to happen now, it would suggest the worst YoY comparisons for the unemployment rate in the 2nd quarter of 2016, 

In 1987 and 1988, the unemployment rate gradually tailed off YoY from a decline of -1.0% to -0.5%.  Permits in 1968 were very gradually decelerating until they outright declined in 1969.   The unemployment rate in 1969 gradually rose into the 1970 recession.

The bottom line is:

1. We are currently in a softening of the YoY% change in the unemployment rate, that began in December 2014, forecast by the slowdown in building permits
2. Based on past incidents, we should be near the maximum point of that slowdown.
3. The softening in the YoY decline in the unemployment rate should pass by the 2nd quarter of 2016.
4. In the worst case scenario for the incumbent party, based on 1962-63, the unemployment rate might rise slightly over the next few months before resuming its decline by the end of this winter.
5. The most likely scenario is that, with some monthly variation, the unemployment rate should continue to decline through at least through next March, and probably through next May.  Beyond that is presently unknown.

This is good news albeit preliminarily for the democratic Presidential candidate.