My response to comments

on Jan 2, 2008 Real Time Economics Blog Entry “Houses, Rents, and Bubbles”

also available here (in full) at WSJ.com

 

Let me start by providing some other background information on why the rent-price ratio for housing – which is like the dividend yield for housing as an asset – is a useful metric.

 

In a world without uncertainty and with constant growth, the dividend yield of an asset has the simple expression of r-g, where r is the discount rate on future dividends and g is the growth rate of dividends.  Thus, if the dividend yield falls, either the discount rate r has fallen or the expected growth rate of dividends g has increased or both. 

 

Knowledge of the dividend yield to housing is intrinsically useful to studying the properties of housing the asset.  Others have also pointed this out.  The specific contribution of our paper was to use micro data from 4 decades of the Decennial Census of Housing (DCH) surveys to document the level and behavior of the series starting at 1960.

 

(By the way, the rent-price series we document is for the aggregate United States.  MSAs like as San Francisco will have systematically lower rent-price ratios).

 

Some have questioned the usefulness of the exercise itself.  A few have mentioned that they believe the best way to describe housing fundamentals is to study the ratio of house prices to incomes.  The nicest thing I can say is that I strongly advise against this strategy.  From a statistical point of view, at the MSA level, house prices and incomes are not cointegrated (see Josh Gallin’s 2006 paper in Real Estate Economics [working paper version]).  In terms of the economics, relatively “high” house prices can be easily supported in areas that expect relatively fast growth.  This is why we study the dividend yield in the first place. 

 

Others have acknowledged the usefulness of a rent-price ratio, but question whether the rent-price ratio we report is “correct.” 

 

One line of comments had the following theme:  Since rental homes are different goods than owned homes, it may not be useful to compare changes to a rent index (for rental units) with changes to a price index for owned units.  This criticism has some merit.  But, at the same time, our paper is not valueless.  If a homeowner wants to know what his or her house would rent for, this person would impute a rental value to his or her unit using market rents on like rented units.  This imputation procedure basically replicates what we do in the paper. 

 

So, the real question is whether homeowners receive more in “implicit” rents from their units than the market rental price for that unit.  (This is related to an idea in economics that when a person buys a product, the value of the product to the person is likely higher than the price paid!  This is the idea of consumer surplus.)  Of course, even if there is a gap between implicit and market rents, all that is required for our study to be helpful is that this gap be constant in percentage terms over time.

 

A second set of comments (see Richard Green’s Jan 1 blog) questioned specific differences in rental and owned units, and whether we account for these differences correctly.

 

One issue is whether we have enough hedonic variables to accurately derive imputed rents for owner occupied housing.  Specifically, the DCH (Decennial Census of Housing) does not include square footage as a hedonic, so our sequence of Census benchmarks may be off – assuming the square footage of owned units has increased at a more rapid rate than the square footage of rentals.  A possible alternative benchmarking procedure would involve data from the Annual and American Housing Surveys (AHS).  These data only extend back to the mid 1970s, and have their own set of problems.  For example, house values are top-coded at $350,000 in the AHS.

 

Regardless, our last DCH benchmark was in 2000.  To extrapolate the rent price ratio after 2000, we multiply rents by the BLS index for rent of primary residence and multiply prices by the CMHPI.  Both of these indexes are supposed to be constant-quality!  In other words, unit quality is held roughly constant since 2000 in our computations.  So, the fact that owned new housing units have continually increased in size since 2000 does not explain why our estimates of the rent-price ratio have rapidly declined since 2000.

 

A final set of comments asked whether we should have expected to see a decline in the rent price ratio due to (a) a decline in long-term real interest rates and (b) a change in the tax code making it easier to shield capital gains on housing from income taxes. 

 

The answer to both is “yes,” with caveats.

 

Start with point (a).  I’m being loose here, but between 2000 and 2003, the real 10-year Treasury fell by about 1 percentage point, from about 3 percent to about 2 percent.  If this is perceived to be a permanent decline, then the rent price ratio should have fallen by about 1 percentage point.  My best guess (see my 2006 working paper with Campbell et. al.) is that this is not a permanent decline.  If this isn’t a permanent decline, the impact this should have had on the rent price ratio is perhaps around 50-75 basis points.  This puts the new “expected” rent-price ratio at 4.25 percent rather than 5 percent.

 

Of course, the real 10-year Treasury is a benchmark yield for a riskless asset.  However, there is a risk premium paid to housing on top of the 10-year Treasury and risk-premia everywhere seem to have increased because of the subprime mess we are in.

 

I’m less certain about point (b).  Under the old laws, many homeowners could effectively never pay any capital gains tax on housing.  Also, the economic environment seems to fundamentally change every year, and that was true from 1960-1997 as well.

 

Finally, I’d like to raise two more points.

 

First, I want to explain why I am focused on land.  Land is the inelastically supplied component to housing.  Thus, scarcity of land is the only reason that house prices rise over time.  Think about DVD players.  The price of DVD players falls every year.  So, if housing were purely a manufactured good, we would expect the price of housing to fall every year as well.  The fact that the price of housing has increased is entirely due to the change in value placed on good locations.

 

Thus, the real question we face as a society is why is it that the price of 60 million urbanized acres is increasing at such a rapid rate when we have at least 1000 million acres of crop and forest land we could develop?  Is it due to simple supply restrictions in New York and San Francisco?  My guess is that it is not:  What is going on with land and housing reflects something fundamental about the way we engage in production.  It seems we need to focus our production in cities to exploit our comparative advantage of the production of services.  Thus, over the next 20 or 30 years, I expect land and housing to be a good investment.  It’s just the next three years that are in question.

 

That said, over the next few years house prices are going to fall.  The reasons are straightforward:  A large class of borrowers that had access to mortgage credit just two years ago can no longer get a mortgage; down payment requirements have increased; and the jumbo-conforming spread has widened.  This is terrible news for house prices and homeowners.  Any regulations or outcomes that make it harder or more expensive for households to obtain mortgage credit will put downward pressure on house prices.  The reason is that as mortgage credit becomes harder or more expensive to obtain, the number of households that can afford any given house falls.  So prices must fall to clear markets.