A Mental Model on Housing Stock and Flow

I often hear people say something like this: “household formation should be 1.5mm per year and new construction are running 1mm per year, therefore we’re facing a housing shortfall”. 

The obvious flaw with this statement is that it ignores existing inventories and focuses completely on trends and “flow”. Put another way, it assumes inventory is already in balance. The logical questions are then:  how about the existing inventory?  How do you know we didn’t overbuilt so much during the last cycle that there are still still excess home supply? 

Intuitively, I’d lay out household formation and housing starts (“flow”) against total number of households and housing units (“stock”) like this:

Assessing whether we’re overbuilding is then a 3 step process: 1) estimate the number of households, 2) estimate number of housing units that can be occupied, 3) compare the two numbers; if there’s a housing unit shortfall then that’s the number of units we need to build. 

In the example above, I started with 2013 number of households. An estimate 750k of household formation for 2014E gets me to 2014E households of ~115mm.

Next step is take the housing unit numbers and figure out how many of those are actually available to live in? This is where subjective judgment comes in. The economy is not perfectly efficient, so at any given time, there’s a healthy amount of vacant units in transition (it takes some time going from a rental listing to actually renting out the unit, a unit could be sold but the buyer has not moved in yet…etc.) These are units that are not available, so I’ll take those out. I also remove a normalized amount of second homes and “held for market – other” units from the stock. For 2013, I estimated ~12.2% of housing units are “normalized vacants”, and removed a corresponding 16mm unit from stock. This resulted in an estimated 116.6mm housing units that are actually available to be occupied.

Finally, compare 2014E households of 115.4mm vs 116.6 of available units at year end 2013 and it’s clear that we still have excess inventory. I expect this excess inventory to decrease only slightly at the end of 2014 because household formation barely exceeds net unit adds. 

Alternatively, I have also seen analyst keeping track of the stock of vacancies, and map out the difference between demand (household formations), and supply (housing starts, demolitions…etc.) as a burn rate against excess vacancy. This is slightly more elegant but should get you similar results. Either way, the point is you have to take into account existing stock (whether in terms of available units as I did above, or as inventory of excess vacancies), rather than just compare housing starts against household formation.

* As a note, the headline reported homeowner and rental vacancies can be misleading as they both understate total vacancy. The way these numbers are reported: if census can’t categorize if a vacant unit is for rent or own, that unit will be left out of the data. i.e these numbers exclude units that are held off market or seasonal vacant. For this reason I focus on total vacancy for a big picture (construction activities). Then only drill down to homeowner vs rental vacancy when I evaluate rent vs own type of decisions.