Tuesday, September 29, 2015

Housing, A Series: Part 64 - Price/Rent measures have had an upward bias

Bill McBride shares a recent quote from San Francisco Fed President John Williams:
I am starting to see signs of imbalances emerge in the form of high asset prices, especially in real estate, and that trips the alert system...But I am conscious that today, the house price-to-rent ratio is where it was in 2003, and house prices are rapidly rising. I don’t think we’re at a tipping point yet—but I am looking at the path we’re on and looking out for potential potholes.

On Price-to-Rent Ratios and Inflation Measures

Source
Williams refers to a rising price/rent level.  Bill McBride has this graph of Price/Rent.  This uses Case-Shiller indexes for prices and CPI Owner-Equivalent Rent inflation to adjust rent.  This is reasonable, since both of these measures attempt to measure the change in prices and rents for individual homes.  However, both of these measures are capable of drifting over long periods if there are any measurement biases.

My preferred measure for Price/Rent is to use the measure from the Federal Reserve's Financial Accounts report for the Market Value of Owner Occupied Real Estate for Price and the measure from BEA Table 7.12 for Imputed Rent of Owner Occupiers.  These should represent more of an independent estimate in each period of the aggregate Price/Rent level.  The Case-Shiller index is a value weighted index, and the aggregate Fed/BEA numbers should act like a value weighted index.  The difference between these measures should basically point to measurement drift in the measure that McBride, and presumably Williams, are using.  And, the difference is substantial.

Since 1995, the Case-Shiller/OER ratio has diverged about 23% from the Fed/BEA ratio.  That suggests that either Case-Shiller home prices have been overstated or rent inflation has been understated by more than 1% per year over the past 20 years.  This is surprising, because even as measured, rent inflation has been persistently high throughout this period.  I have used this persistent inflation - centered in the problem metropolitan areas (NY, SF, etc.) - to argue that there has been a persistent housing shortage.

I think what we are seeing here is a form of substitution effect.  There isn't a measurement error, per se, in either the OE rent or the Case-Shiller price measure.  But, what we have been seeing is rent inflation in the problem cities driving up prices with stagnant housing stock.  The prices there are rising due to both the rising rents themselves and due to rising Price/Rent levels, which we can see are especially strong in the Case-Shiller 10 cities, relative to the national level.  So, marginal new homes have to be built outside those cities, and those homes have lower Price/Rent ratios.

Even if rent inflation and price changes of each individual house are accurately tracked over time, the fact that marginal new housing stock tends to have a lower Price/Rent will mean that we have overestimated the price of the home of the typical household.  Measures of real and nominal housing expenditures (rent and imputed rent) would not be biased by this substitution.  And, the Case-Shiller measure of home prices would not be biased regarding the homes.  But, the substitution does mean that Case-Shiller does overstate the cost of homeownership for the typical household over time.


I think this bias in Case-Shiller may be somewhat inevitable, because persistently high rent inflation and the related inflation of Price/Rent will tend to only be sustainable in a supply-constrained environment.  In an unencumbered market, new building would be attracted by the high Price/Rent level.  If high Price/Rent persists, then it seems likely that marginal new building will have to occur in lower Price/Rent locations.

Source
So, while it is accurate to say that the typical house that existed in 2003 has a similar Price/Rent today that it did then, the typical household lives in a house with a Price/Rent ratio similar to 1999.  This is because many households have traded down to homes with lower Price/Rent ratios.  In this Price/Rent graph that goes back to 1950, we can see that aggregate Price/Rent ratios are well within normal long term ranges.  Considering the present very low long term real interest rates, Price/Rent is very conservative.  This is why I claim that home prices are too low.  In the places where we allow homes to be built, homeownership is very affordable.  High income households are bidding up the price of real estate in San Francisco and New York City, but that is a negative supply issue.

In a narrative sense, for every household that remained in a condo in San Jose whose imputed rent rose from $4,000 to $4,200, and whose price rose from $1 million to $1.1 million, there was another family who moved to from San Jose to Phoenix and bought a home with imputed rent of $3,800 that sold for $800,000.  This bias in the Case-Shiller indexes created a false sense of an over-heated housing market in the 2000s in two distinct ways.  Regarding rent, from 1995 to 2006, households spent a stable level of their incomes on housing (rent), about 18%.  But their real housing expenditures over that time declined by about 13%.  Since Case-Shiller tracks homes, not households, it does not reflect this shift to less valuable housing.

Regarding Price/Rent, as mentioned above, there appears to be a bias since 1995 of about 23%.  Together, this means that the typical house, tracked by Case-Shiller, overstates the change in the price of the typical household's home since 1995 by about 39%.  The average household lives in a home with lower imputed rent and a lower Price/Rent ratio because they have been downshifting on their housing consumption.  This was especially the case during the boom.  The divergence between the Case-Shiller Price/Rent ratio and the Fed/BEA ratio mostly happened between 2002 and 2008.  As with so many factors I have considered in this series, this is basically the opposite of what everyone thinks happened during that time.

Source
This chart, from Bill McBride is accurate, that the typical home price, in real terms, is about 40% higher than in the 1990s.  But, since households have been adding marginal new housing stock which has tended to be lower in value - mostly due to being in less valuable locations - the average household lives in a house with a price, in real terms, approximately the same as the average household's home in the 1990s, similar to what we see in the Financial Accounts/BEA measure of Price/Rent above.

5 comments:

  1. Interesting post. The Fed seems to want to repeat its error, and try to pop a bubble. Housing prices inevitably rise in recoveries, even if we had good government---it takes time to build a house or condo tower.

    Toss in bad government, single-family detached zoning, anti-manufactured housing rules, rules again trailers, and you get even slower market responses to demand.

    Then add in that the US tax code more or less forces a successful (financially) person to buy a home, and then housing becomes an investment.

    The Fed needs to forget about housing inflation, unless it is to advocate the elimination of city zoning against dense housing (a legislative effort outside its ken, obviously).

    I see no solution for housing inflation, as measured or in fact. The libertarians go mute when single-family housing zoning becomes the topic. The Supreme Court upheld city zoning in 1926, and evidently everyone is happy with it.

    Everyone thinks their own neighborhood is special----from Brooklyn to Newport Beach----and needs conservation.

    The Fed should just accept 3% inflation and go from there.

    ReplyDelete
    Replies
    1. Hear! Hear!

      I think this idea that liquidity is somehow causing home price appreciation but not consumption inflation has everything backwards. I keep meaning to do a post about it. I keep seeing papers that claim financial expansion leads to slower growth or recessions. And, they seem to always treat low long term interest rates as a result of loose monetary policy. But, it looks to me like the causation is the other way around. Low long term interest rates lead to higher home prices and thus mortgage expansion and also presage slower growth.

      We are reacting to higher home prices with policies that lead to lower real growth, when we should be doing the opposite.

      Delete
  2. I am beginning to thinks the Fed should set a ceiling at 4% unemployment, and let the chips fall where they may after that.

    Inflation too high? How about the Fed's response: "It will take structural reforms to bring inflation down. Cut SSDA disability rolls, and the VA's also. Get cities to zone in some density. Inflation is not our problem."

    ReplyDelete
  3. Great post.

    I've been recently pondering the relationship between rents & values; the demand for each is often compartmentalized, however, I think you're correct in that they are substitutes.

    On a technical note, what is your opinion on the validity/reliability of ACS data for gross/contract rents and home values? I know you prefer Case-Shiller & Fed. data...just curious if there are viable alternatives?

    I like how ACS data provides different geographic lenses (tracts, block groups) and considerable demographic & socioeconomic information as well.

    ReplyDelete
    Replies
    1. The ACS data is great, and I use it as much as I can. The reason I don't use it more than I do is because it only goes back to 2005. What I wouldn't give for ACS data back to 1995.

      Delete