by Calculated Risk on 1/20/2008 12:45:00 PM

On Thursday, the Census Bureau reported housing starts were at the lowest level since the 1991 recession. Here are some more thoughts on starts:

Click on graph for larger image.

This graph shows total and one unit structure starts since 1968. Yes, I've marked the current probable recession on the graph!

Some people might argue that the graph isn't normalized by population or the number of households. That is correct, and it would appear with increasing population that the number of starts is pretty low right now. However, there is more to demographics than population growth; changes in household size can have a significant impact on the demand for housing. In fact, a majority of the demand in the '70s was driven by declining household size. (see this demographics and housing demand).

Here is a table from the above post that shows the number of housing units added due to population growth, and the number added due to changes in household size. In the '70s, there were more housing units added due to smaller households, than growth in the population. Conversely demand in the '90s was almost exclusively due to population growth, as household sizes stabilized.

Housing Added due to Population Growth and changes in Household Size
Decade	Due to Population Growth	Due to Change in Household Size	Total Housing Units Added
1940s	5.84 Million	2.86 Million	8.70 Million
1950s	9.11	3.08	12.19
1960s	8.10	2.27	10.37
1970s	9.05	10.65	19.70
1980s	9.13	4.77	13.90
1990s	13.47	0.13	13.60
2000s (through July '06)	7.63	2.04	9.67

Predicting population growth and changes in household size are both important in predicting the need for new housing units. With the high cost of homeownership, I wouldn't be surprised to see household sizes increase slightly over the next couple of years - lowering the demand for new housing units - at least until prices fall significantly.

Another driver for housing starts is the current housing supply. I've discussed this here: Housing Inventory and Rental Units. Currently there are about 1.6 to 1.7 million housing units in the U.S. above the normal level of inventory.

If housing units were transportable, and each unit was a perfect substitute for another, starts could fall to zero in 2008 and there would still be some excess inventory at the beginning of 2009!

Starts won't fall to zero in 2008 - units are not transportable, and they are not perfect substitutes - so we need to look at starts from a different angle to estimate how far starts will actually decline.


This graph breaks down housing starts by type and intent since 1974. This data is available quarterly at the Census Bureau. Note: this is the annual rate, not seasonally adjusted, Q4 2007 is estimated using monthly data.

We could use this data to make some estimates for 2008. First single family units built for sale is clearly dropping sharply. This was at about 550K annual rate in Q4, and has been as low as 400K in 1982 (300K quarterly). An estimate in the 400K range is probably reasonable for 2008.

Homes built for owners has been running over 300K per year for some time, although the level has fallen recently. A recession will probably cause a dip in owner built homes, however this is a housing area with favorable demographics (people in their 50s like to build their "dream home"). Perhaps 250K homes will be built for owners in 2008.

The next category is built for rent. This has been running just over 200K per year, and I expect the number to be at least that high in 2008.

The final category is condos. Condo starts are starting to decline too, but there is a fair amount of momentum in the larger projects, so a decline to 50K to 100K is probably reasonable.

This means something like 400K Built for Sale, 250K Owner built, 200K rental units, and 75K condo units will probably be started in 2008. That puts starts at 925K - higher than many other estimates.

If starts came in at 925K, this would actually be good news for the economy since residential investment wouldn't fall as far as some are forecasting, but this would mean very little of the excess inventory would be worked off in 2008 keeping pressure on housing prices and the homebuilders in 2009 and 2010.

A more pessimistic view would take starts to 800K in 2008: 350K built for sale, 200K owner built, 200K rental, 50K condos.