Home Price Expectations and Residential Investment

I have often wondered whether the very low level of Private Residential Investment (also known as building houses) in the USA is due to unusually low expectations of long term increases in home prices. Here, I add that I would guess that the change was and end of the belief that the relative prices of houses has a strong upward trend (so houses are good investments) which belief is not supported by Robert Shiller’s data in Irrational Exuberance.

Of course I knew that, to find out, I should google something like ( Shiller house price expectations ). I forget what I told google but google sent me to this pdf.

Which you must read, but from which I will steal data on expectated house price increases over the next 10 years and this graph

shiller

yes indeed, expected appreciation of house prices has dropped dramatically as one would expect given the bursting of the bubble. The question is does this explain the low level of residential investment ?

hinv

Note that Shiller thinks the relevant variable is the expected growth of home prices minus the mortgage interest rate. Unfortunately even he only has 10 observations of the expected growth of home prices over 10 years. It is easy to fit the ratio of private residential investment to GDP with this variable, but this may just be because it is easy to fit 10 observations.

shiller2

It is mildly interesting that the difference alone seems to explain the huge change in residential investment without any need to appeal to financial frictions.

The more interesting question would be whether the relationship between the variables in recent years is the same as it was before the great recession. Alternatively, it is possible that some factor other than low home price increases is needed to explain low US residential investment. The natural candidate would be unusually tight lending standards as an over-reaction to the disasters caused by the extraordinarly loose lending standards of the early 00s.

so I back-casted with the parameters of a 10 data point regression

Number of obs = 10
R-squared = 0.6312

exp10mortg Coef. t
resinvgdp .9967796 3.70
constant -1.19086 -0.97

conveniently the regression says that the expected growth of home prices minus the mortgage interest rate shifts one for one with residential investment as a percent of GDP.

The absurdly simple model (regression specification) implies that expected annualized 10 year home price appreciation is always equal to the mortgage interest rate plus the ratio of residential investment to GDP minus 1.2%.

Unfortunately, this implies absurdly high expected home price appreciation in the 1980s when mortgage interest rates were extremely high. In particular, there is one byte of information about home price expectations in the 80s in Case and Shiller’s seminal New England Economic Review article

(note how the profession is moving towards appreciation of the importance of their work — from New England Economic Review to the Brookings Papers on Economic Activity) .

shiller3

* standard errors.

I think the cities which are useful when discussing nationwide housing demand are Boston and Milwaukee where the average expected annualized home price growth over 10 years was 8%.

Mortgage interest rates were around 10% in 1988, so the difference was negative. Currently it is around zero. The regression suggests that housing investment should have been much lower than the current 3.1% when it was, in fact, 4.9%.

The specification suggested by Shiller’s graph just doesn’t fit the data. Residential invesment would have to respond much more to expected house price changes than to mortgage interest rates.

Not willing to let the data convince me that my story is wrong, I tried to think of arguments why mortgage interest rates wouldn’t matter so much. I have a couple. First people might buy when mortgage interest rates are very high and plan to refinance when they fall. The option to refinance exists and is valuable (and wasn’t restricted back in the 80s). Second their is the required return on the down payment (which in the US is traditionally about 10% of the purchase price). This is an oppurtunity cost and home buyers might believe that they can’t get returns as high (and variable) as mortgage interest rates. This argument is silly for two reasons, first the down payment is usually tiny compared to the mortgage and second because long term treasury rates are not that much lower than mortgage rates and they move up and down together. But insisting on the argument, I note that the mortgage is repaid — over the life of the buyer, the mortgage goes away but the exposure to capital gains and losses doesn’t. If someone inveted in a checking account (which is crazy) putting less weight on the mortgage rate than on expected house price increases would make sense.

Well that was feeble. While sure I have convinced no one, I consider long term expected home price appreciation minus long term expected consumer price inflation. Fortunately it is possible to cheat on this, because the long term expected CPI inflation was always almost exactly 2.5% during the 10 years with home price expectations data. So the regression is just a regression of housing investment on expected home price increases

shiller4

Again the fit looks nice. the back cast regression is

. reg exp10 resinvestpercentgdp

Number of obs = 10 R-squared = 0.7688

exp10 | Coef. t –
resinvestp~p | 1.408554 5.16
_cons | 2.456457 1.96

Average long term expected CPI inflation in the Livingston and Blue chip surveys was 4.3 to 4.35 % (different surveys in 1988) so 1.8 to 1.85 percent should be added if one wants to claim that what matters is expected home price inflation minus expected consumer price inflation.

update: Some arithmatic corrected below.

Residential investment was 4.9% of GDP so the fitted value for home price inflation (with no correction for changes in expected inflation) is about 2.5+1.4*4.9 = 9.2%

adding 4.3% – 2.5% = 1.8% gives 11% which is well within the range of expected home price inflation in the 1988 sample. The sample does include two cities chosen for their housing bubbles.

So if, for some reason, expected real mortgage interest rates don’t matter too much, then low expected home price appreciation could explain low housing investment with no need to appeal to unusually severe financial frictions.

That’s a big “if” and that’s where I (finally) stop typing.