Monday, July 18, 2011

Has the housing industry boom-and-bust been driven by an even *bigger* bubble-and-bust in household creation? 

Out of curiosity, a little while back I graphed for myself the historical data on home construction to see how far trends have gone compared to past historical experience. Well, the numbers show a big drop recently, no surprise! One might conclude that such a depressed level of construction, by the power of regression to the mean, implies a correspondingly big increase is in the cards for the future, sooner or later.

But when trying to be more specific about how big a turnaround is likely to occur and when, the problem is: how low is today's low? Compared to *what*? With no context or causative theory indicating where home construction should be today in a "normal" economy, as opposed to in a Great Recession, there is no way to say what an industry recovery might look like. Speculating about it is just guessing.

Then I came upon an article about investing in the home building industry (buy cheap!) that mentioned a strong correlation between household formation and home construction. That's logical and obvious enough -- new households need new homes. But due to a lack of easy to find good data on the relation, I went back to the Census and graphed its historical data on household formation too.

Putting together the two data lines gives an interesting picture that I haven't seen discussed anywhere before.

In short: The housing boom-bust has been following an even *bigger* rocketing up and collapse in household formation.

Household creation and housing units started, thousands, 1964-2010.

The main picture. The single-year numbers are very volatile so I've used a three-year moving average.

From 1960 through 2010 new housing unit construction averaged 1.16 times growth in the number of households. That makes sense, as new housing units are required both to provide homes for new households and to replace old housing that is abandoned.

In the chart the line for new housing units is generally above that for households created, as it should be. But look at what happens in 2003 -- household creation starts rocketing up to an all-time high, well *above* housing starts. Housing starts follow upward, get very high, but never as high as household creation. Then household creation plunges to by far its all-time low. The infamous collapse in housing starts is only following behind it.

The single year numbers are even worse than the moving average:

[] Household formation is down 78% since 2007, 75% from the average for 1998-2007.

[] Home starts are down "only" 57% since 2007, 67% from the average for 1998-2007.

New homes actually have been *overbuilt* during the last three years, compared to the historical 1.16 new starts-to-household growth ratio: during 2008 to 2010 the ratio has been 1.34 to 1 ... in 2010 it was 1.64 to 1.

A few other pictures of the same data may provide some perspective:

Household creation and home unit starts, three-year moving averages, versus their 15-year moving average trend lines.

The scale of the recent household creation run-up is more clearly visible when plotted against trend.

The peak of household creation in 2003 was more than twice as far above the trendline than any other cyclical peak since 1960. The fall so far is 50% further below trendline than any prior fall.

This is simply an unprecedentedly huge swing up and down.

In contrast, the height of the peak in home construction is in the same ballbark as three prior ones -- and only 55% as far above trend as is the peak in household creation.

Note: The trendlines are 15-year averages -- the average of the given year year plus the seven before and seven after, until 2000.

For 2001 and later: (1) the household creation trendline number is figured by multiplying the growth in total US population by the historical household creation/population ratio, and (2) the housing unit start trendline equals the household creation trendline number times 1.15.

Household creation peaked in the 1970s  as the Baby Boomers came of age.  But at all times the trend line for home unit starts has been above that for household creation.

There was a general slowing in US population growth during 2000 to 2010 compared to during 1990 to 2000, as population grew only 10% during the later period compared to 13% during the former. This causes somewhat declining trendlines estimated for 2001 to 2010 -- yet still the household creation and home starts lines plunge far below them.

The mid -2000s rocketing-up of household creation over home starts is clearly aberrational.

"Accumulated Overhang" of household creation.

The rate of household formation has exploded and collapsed in the last decade. But will it continue collapsing? We might ask, as people do about the presumed overhang of homes built during the same period, whether the "excess stock" has been worked off yet. If the accumulated number of households created during recent years is still over the number that would be created by the long-term trend, we might think the plunge will continue.

Trying to use these numbers is to guestimate this is shaky business, so take the try at it that follows with a big tumbler of salt.

By counting the households created each year/over under the trendline, keeping a running tally, then dividing the cumulative amount by the current trendline number for annual household creation, one can get a number for "excess" household creation in terms of years.

But the challenge is picking the right year to start counting from. It makes a big difference.

The best way to measure an economic cycle is from peak-to-peak or trough-to-trough, and since we are very near or at the bottom of this cycle now (we hope!) the idea is to measure from a past bottom.

Census data on household creation goes back to the 1940s, the first year for which the trendline method used here is available is 1954 -- and as it happens 1954 was the bottom year for household creation for more than a decade. So charting the data from there...

By this quick-and-dirty method it looks like the "excess" of households over trend has finally evaporated, with the number of household finally nosing below trend, by 0.35 of a year of trend household creation.

But maybe not -- the small difference between starting in 1954, 1955, 1956, etc., makes a big difference. Here's the result as measured from the years 1954 to 1963.

1954 -0.35 years
1955  0.08
1956  0.20
1957  0.16
1958  0.30
1959  0.41
1960  0.37
1961 -0.04
1962  0.15
1963 -0.02

So it may well be that the household count has not yet gotten down to trend level after all.

Looking ahead.

There are a lot of people about opining "the economy won't be fixed until the home building market gets fixed". But all the above implies the reverse is true.

Whatever may have caused the boom in household creation to begin with, there's no doubt that a bad economy puts the brakes on it. In a Great Recession we'd expect the household creation rate to fall well below trend even if there hadn't been a boom in it to go bust. That suggests the fall in the household creation can be expected to continue for a significant while yet after getting down to trend.

General economic recovery requires a recovery in the home building industry ... Recovery in the home building industry requires recovery in household creation to increase the demand for housing ... Recovery in household creation requires general economic recovery ...

I'm going to go think about something else.