Tuesday, April 28, 2015

Housing Tax Policy, A Series: Part 30 - Secular Stagnation, a Housing Bubble, and Loose Monetary Policy can't all be true.

This graph shows the long term decline in real housing expenditures.  Contrary to popular narratives about the housing boom, there is no indication here of overbuilding.  On the contrary, as a proportion of total consumption, households have been bidding up the price of a dwindling housing stock.

This topic is a little bit tricky because of the position of the owner-occupier as both a consumer and a supplier.  It is tempting to think of rising home prices as a product of high demand.  But, if we think about this carefully, the housing "bubble" narrative is actually saying that households were being induced into being housing suppliers.  When a household takes on a new mortgage and buys a new home, they are increasing supply by investing capital into a property which will provide additional real supply over time, as measured in inflation-adjusted rent.  It is not possible to say whether they are increasing demand.  The best window into housing demand that we have comes from the measures of nominal and real rent and imputed rent.

A context with stable nominal expenditures and declining real expenditures suggests a relatively stable demand and a shortage of supply.  If we truly had an oversupply of housing, then real and nominal housing expenditures would have looked more like this:

Maybe the BEA messed up the imputations of owner-occupier rent, and nominal housing expenses were actually going up.  As a double check on that, here is an estimate of housing expenses with no imputations.  This is an estimate of cash expenses for all households, including tenant rent and owner-occupier gross interest expense, direct expenses, subsidies and transfers.  This was actually declining during the boom.

The trend in cash expenses is much flatter if we only include the real portion of mortgage interest.  As I have described before, the inflation premium portion of mortgage interest is really a purchase of home equity by the borrower from the lender, when considered in real terms.  So, much of the decline in cash expenses is the product of the decline in expected inflation, which reduced nominal mortgage payments.  That is a kind of imputation, so I have not included that adjustment here.  But, even if I did, it basically gets the housing expenses trend just back up to a flat line.

Without a rising level of nominal housing expenditures, we can't explain the housing boom as a combination of both rising demand and rising supply.  But, what if the measure of rent inflation is somehow wrong because of the difficulties of estimating rent imputations?  What if housing expenditures looked like my counterfactual graph, above, with level nominal spending, and an oversupply of homes which meant that real housing consumption was actually increasing?  That seems like the outcome we would have to see if there was a "bubble" in home construction and home ownership, but no increase in nominal home expenditures.

First, here is a graph of core inflation over time, separated between Shelter inflation and non-Shelter Core inflation.  (I have used a static weight for Shelter Inflation equal to 40% of core, which I think is a good approximation of the shelter weight over time.)  First, without talking about counterfactuals, just looking at Core minus Shelter Inflation over time, we can see that shelter inflation has been high since 1995.  Excluding the shelter inflation, core minus shelter inflation has moved between 1% and 2% for 20 years.  I believe that this reflects supply constraints, and has been pushing monetary policy too low, to the extent that the little inflation we have seen has been interpreted as a demand phenomenon.

But, what if we apply our housing counterfactual?  What if real housing consumption has been rising, and this has been mismeasured as inflation?  In that case, Shelter inflation should have been running below Core Inflation.  In this graph, I have adjusted Shelter inflation so that it is slightly below Core inflation instead of slightly above it, which is the inflation level that would correspond to our housing oversupply counterfactual.  In the counterfactual, Core Inflation has been running between 0% and 1.5% since 1997.

So, unless someone can unearth data that points to rising nominal housing expenditures, relative to total consumption, it seems to me that there are two possible narratives to choose from:

1) There has been a decades long shortage of housing, combined with relatively tight monetary policy that has contained non-shelter inflation generally below the 2% target rate.


2) There was a housing bubble, which created an oversupply of housing in the late 1990s and early 2000s.  The flood of money into the housing market created an increase in the real housing stock, relative to incomes.  This means that RGDP has been running at much higher levels than previously thought and inflation has been lower than previously thought.

If we believe the data, narrative number 1 is the correct narrative.  But, if we are going to believe narrative number 2, then the narrative includes neither secular stagnation nor loose monetary policy before 2006.  There is an awful lot of data that would have to be massaged in order to tell a story that includes an oversupply of housing, secular stagnation, and loose monetary policy.  That's the story I keep seeing in the papers.  The data tells the opposite story.  (In fact, secular stagnation just might be coming from the constraints we have placed on housing supply.)


  1. As always, thought-provoking. What about sf of housing per capita? Useless? We may have houses where we do not need them, in the Midwest.

    1. That might give some meaning, but a studio in Manhattan and a 2 level castle in flyover country could both have the same value - and both be homes that are built in a functional economy. So, I think there is limited value in thinking about it that way. Local limits to building in the high cost metro areas definitely skews housing to less valuable areas, though.

  2. Kevin, sorry to be clueless, but I'm finding this hard to follow because I'm not sure what you mean by "real rent" or "real housing expenditure". I can imagine at least three ways to define "real rent":

    1) nominal rent divided by some general-purpose deflator, such as the GDP deflator or the PCE deflator or the CPI deflator;

    2) nominal rent divided by a shelter-specific deflator, e.g., build an estimate of shelter inflation and divide by that;

    3) rent measured in some non-dollar way, such as square feet; i.e., real rent expenditure is increasing if people are renting more square feet per household.

    Which is it? Does it matter? (It strikes me that #2 and #3 might be the same thing expressed in different units, so it wouldn't matter if you picked #2 or #3, but #1 could be much higher than #2 if, say, supply constraints were forcing renters to bid up rents of the existing housing stock.)


    1. Good questions, Ken.

      It's number 2.

      The BEA keeps data on both real and nominal expenditures, so the first graphs are directly from those measures. Shelter-specific inflation would reflect the difference between those measures for shelter-related spending. I am being a little loose here, because I think changing consumption patterns and inflation levels might lead to some distortion of real spending as I am showing it, especially as you move back in time. But, the scale of any distortions shouldn't change the main consumption patterns I am talking about.

      The best way to look at it, I think, would be to imagine an average household, as renters. For simplicity, let's say their nominal and real income remains the same. Rent inflation is the change in the rent they are paying on their house (in a zero income growth environment). A real increase would mean that they moved to a more expensive house (maybe larger, maybe closer to downtown).

      In a world with changing incomes, we can imagine a second step, that when their real income increases, they move to a slightly better house (larger, better location, etc.).

      For the past 30 years, the average household has been paying the same rent (relative to their income), but has been moving to a slightly smaller (or otherwise less valuable) house each year.

    2. Thanks, Kevin. I think I'm following.