More on the Housing Overbuild

Note: I will be traveling this week, thus posts will be very light / I will not be able to respond to comments in a timely manner.

Last week, EconomPic asked 'What Housing Overbuild?' showing that even during the bubble, the number of units built didn't appear to be extraordinary on a per capita basis. Reader dblwyo gave Calculated Risk a heads up on the post, which responded over the weekend with some great additional insight in the post Housing: Impact of Changes in Household Size:

If we look at a long term graph of housing starts, we notice that there were more starts at the peak in the '70s than during the recent housing bubble. If we plotted housing starts per capita, or per total households, the surge in housing starts during the last decade would not look extraordinary at all (ht Dave).

But it was extraordinary ...

He then details his reasoning as to why housing starts were extraordinary as compared to the 1970's.

The key is household formation.

Household formation is a function of changes in population, and also of changes in household size. During the '70s, the baby boomers started moving out of their parents' homes, and there was a dramatic decrease in the number of persons per household. And that lead to a huge demand for apartments (the surge in total starts).

In other words, it is not only the growth of the population that matters, but also the number of people within each house (i.e. if there were 4 individuals per house and now there are 3, you need more houses all else equal). As a result, there were too many units built this past cycle as the number of individuals in each home did not continue to contract at the same pace as it did during the 1970's.

Because of the changes in household size, the U.S. needed far more additional housing units in the '70s and '80s than in the '90s and '00s. If we could normalize the housing start chart by household formation, we would see that the last decade was indeed extraordinary!

The following chart summarizes a table he built using data from the Census showing the above phenomenon.



Using the household formation data from the Census, I was able to create the following chart, which attempts to show the normalized housing starts by household formation Calculated Risk details.

More specifically, the above chart shows the annual number of housing permits issued in each year divided by the number of households formed on average over a ten year period (i.e. the normalization part). This in theory removes some of the chicken or the egg issue as to whether:

• Households are formed, which drives building of new homes
• Building new homes adds supply, which allows new households to be formed

This revised analysis does remove the spike in permits per capita from the 1970's and shows housing builds were indeed well above the long term average of ~1.1 permits per new family* during this bubble period. But if you are going to label periods of home building that were 0.40 more homes above "ordinary" levels as extraordinary, I think you must label periods of home building 0.70 homes less than the "ordinary" level, extraordinary.

Which once again brings up my question as to the possibility that we'll go from an housing overbuild to a shortage within a 5-10 year period?

Outside of a huge reversal in the number of persons per household that may result from a prolonged economic slump, if the number of new homes remains near these historic low levels, I think it is possible.

Source: Census 1 / Census 2


* I figure the level is greater than 1 due to depreciation of existing stock (new homes to replace old housing) plus second homes.