Friday, May 29, 2015

Housing Tax Policy, A Series: Part 35 - The effect of limitations to building in the coastal cities

I've been using data from the BLS and the American Community Survey (ACS) to try to gauge the relative effect of limits to building in the major cities.  I've been using the 10 cities from the Case-Shiller index as the representative basket (hereafter, "CS10").  This isn't a perfect match for the problem.  Denver, Chicago, and Las Vegas don't exhibit excessively abnormal rent inflation and price levels.  Boston and Miami are worse.  But, New York City, Washington, DC, San Diego, Los Angeles, and San Francisco are in a class by themselves, and they are all on the list of 10.  The Case-Shiller 10 is a pretty good proxy for the cities behind the housing supply problem.  (I don't have full CPI data for Las Vegas and Washington, DC, so some of this analysis is limited to data from the other 8 cities.)

Shelter Inflation reflects a supply problem

As I have looked at housing data and monetary policy over the past several months, I have come to the conclusion that there are two general components of inflation over the past 20 to 30 years.  There is a Shelter component, which is generally high because of supply side issues, and there is Core Inflation minus Shelter, which I think we can use as a proxy for the demand side (monetary policy).  As a rule of thumb, I treat Rent inflation above the Core minus Shelter level as a supply issue.  (I welcome feedback regarding any technical criticism anyone has about this treatment, but please read the series of housing posts first, if you have conceptual criticisms.)

In the mid-1980's, Shelter inflation briefly rose, but this was limited to owner-occupied properties.  I suspect that this largely reflects tax arbitrage in single family homes by homeowners due to the new importance of the mortgage interest tax deduction in the mid-1980's.  Single family homes for rent are subjected to a less favorable tax treatment, so owner-equivalent rent for owners, which reflects pre-tax expenditures, rose.  But, after the brief rise in owner imputed rent, shelter inflation followed core inflation until the mid-1990s.  Since the mid-1990s, shelter inflation has been consistently high (except for the brief post-2008 shock), and it has been highest for renters.

We can see in the Fred graph that since the mid-1980s, multi-unit housing has had very limited growth during recovery periods.  In fact, I have come to believe that this was one of the important factors at play in the housing boom of the 2000s.

Note that total housing starts were not unusually high in the 2000s.  But, limitations to multi-unit building, which are largely occupied by renters, caused rent inflation to rise, and drove households into single family homes.  So, total new units was fairly normal, but all the extra growth had to come through new single family homes.

Home price appreciation was concentrated among the CS10 cities, and I have argued that this reflected rents that were rising especially sharply in those cities.  ACS data, which only covers 2005-2013, gives us some additional detail on this picture, and helps to estimate the scale of the BLS inflation data that goes back to 1982 for many of these cities.  Keep in mind that these data sets cover entire metropolitan areas, including the suburbs, so in some ways, the data showing housing supply problems in these cities understates the problems that are specific to the city centers and to limits on multi-unit housing.

We can see from the first graph, above, that rent inflation was worst during the housing boom, and was especially bad in the 2006-2008 period, after home prices and new building leveled off and began to fall.  During this time, national vacancy rates were somewhat elevated, but they remained low in these high-rent coastal cities.

I think, when we put all of these things together, what we see happening is limited multi-unit building in our metropolitan core cities driving up rents there, pushing households out to the suburbs.  Sticky prices and other frictions in the housing market prevented single family housing growth from fully meeting demand in real time, so rents continued to climb during the housing boom.  Then, beginning in 2006, the Fed began to pull back on the money supply and the mortgage market dried up. Housing supply was already struggling to meet demand, but this new limit to single family home construction completely undermined the market.  Now housing starts collapsed and rent inflation accelerated.

Notice in the Fred graph above that multi-unit housing starts and single unit housing starts had always moved roughly in sync.  But, notice how single unit starts collapsed in early 2006, but multi-unit starts continued apace until the summer of 2008.  Also, note that home vacancies rose in 2006 (even while new building collapsed) but rental vacancies remained level until 2009 (even while building continued).  The money supply and the mortgage market collapse are what led to the housing collapse.  There was still demand for housing, but there was no cash or credit for it any more.

Now, rental vacancies are at new lows, and multi-unit starts have recovered back to pre-recession levels.  But, it's like there is a cap at about 400,000 units per year.  And, in the 20 years during which it looks like there is a 400,000 unit lid on multi-unit housing, we have had persistently high rent inflation.  And this supply problem is coming from our coastal cities.

Rent Inflation among the Case-Shiller 10 and the rest of the country

"Core minus Rent" (CmR) inflation from 1995 to 2013 averaged an estimated 1.8%.  Rent inflation for all households since 1995, in all the areas outside the CS10, has averaged 2.4% - that is 0.6% higher than CmR inflation, suggesting a small housing supply issue.

But, for these 8 cities on average, while CmR inflation was similar to national CmR inflation, at about 1.9%, Rent inflation for all households was 3.2%, and Rent inflation for renters in these cities averaged 3.6%.

The Case-Shiller 10 cities account for 21% of the US population and 31% of residential real estate, by value, but during this period, they were responsible for 55% of the nationwide excess owner-occupier rent and 62% of the nationwide excess renter rent (above the CmR inflation level).*

In the decade before 1995, rent inflation in these cities was no higher than core inflation.  There was a previous period of relatively high shelter inflation, in the 1970s and early 1980s, but I don't have city-specific data for that period.

These cities create a sort of triple-whammy.  Renter inflation has risen more than owner-occupier inflation, these cities have a higher proportion of renters, and rent in these cities tends to claim a higher proportion of household income.

There isn't particularly a reason why these cities need to have perpetual rent inflation.  High density cities aren't that expensive (HT: OB), and many large US cities have grown without unusual rent inflation.

Using rent inflation figures and median rent expenses for the 8 cities with BLS data, I have constructed an estimate of median incomes in a counterfactual world where rent inflation was in line with Core minus Rent inflation since 1995.

Nationwide, the median home owner would have real income 3.0% higher if we did not have this scarcity in housing.  In non-CS10 areas, median owner real income would be 2.2% higher, and in CS10 cities, it would be 5.3% higher.  Owners do earn this income back, to the extent that they have equity in their homes.  But, in terms of asset values, this income is reflected in unusual capital gains captured by existing home owners.

For renters, the median national renting household would see 7.5% higher real income.  The non-CS10 median renter would have 4.9% higher income and the CS10 median renter would have 12.6% higher income.

This loss of real income is basically paid as capital income to real estate owners, as returns on their unusual capital gains.  Homes in the CS10 cities have cumulative excess rent inflation since 1995 of 27%.  Homes in non-CS10 areas have cumulative excess rent inflation since 1995 of 10%.  The national average is 16%.  In 2013, gross rent paid and imputed for housing amounted to 10.5% of GDI.  So, I would estimate that real estate owners are capturing about 1.4% of GDI from the capital gains they have accumulated due to this imposed scarcity.  This is split roughly in half.  About 0.7% goes to property owners in the CS10 cities and about 0.7% goes to property owners in the rest of the country.

The Effect of Housing Scarcity by Income

I can also use real estate and income data from the Survey of Consumer Finances, the CPI rent inflation data, and BEA data on rent expenses to estimate the effect across incomes.  By this measure, for home owners, housing scarcity claims 1.7% of mean family income, ranging from 1.3% of the top decile to 2.9% of the bottom quintile.

Here the effect on income distribution is muted, because home ownership is not that low in the lowest quintile (37% in 2013), and the lowest quintile has the highest proportion of equity among all income levels (reported at 80% in 2013).  So, most low income home owners are probably older households who have pocketed the capital gains from these 20 years of housing scarcity.

Please give me a link in the comments if anyone knows where I can get data on household income and rent for renting households, by income quintile.  The ACS has some data, but the data I have found is not broken out like the SCF data where I can calculate mean levels of expenditures as a proportion of income.

But, if the median renting household spends more than 30% of their income on rent, the poorest households must reach something around half of their incomes.  That means that the renters who make up 63% of the lowest income quintile probably have real incomes reduced by around 10% because of these housing issues, and renters in the CS10 cities - especially New York, Washington DC, and the California metro areas - might see real income cuts of around 20%.  And that is for entire metro areas.  In the core cities, where the limits on multi-unit housing are most severe, real incomes for the poorest households could have taken an even larger hit, although, for this group, the effect of public subsidies, rent controls, etc., must be complex.

Since high income households spend less on housing and capture some of the gains of ownership, they will only see minor declines in lifetime incomes from this problem.  This could be a major factor in the apparent stagnation of lower incomes.  Actually, I think that this spread in income growth between income quintiles would be in addition to measured increases in income inequality, because these household-specific costs would not be captured by average national inflation statistics.

*This is based on stable nominal expenditures at 2013 levels, so it doesn't include any compounding effects from cities with persistently higher inflation.

No comments:

Post a Comment