Housing Tax Policy, A Series: Part 18 - I was kind of wrong about the distributional effects of the housing shortage

I had previously written a couple of posts about the fact that lower compensation share hasn't been going so much to corporate operating profits as much as it has been going to home owners.  The housing shortage means that rents are rising and imputed returns to home ownership are rising, in absolute terms and relative to mortgage payments.  So, home owning households have de facto real incomes that are rising faster than reported real incomes. (Imputed rent is not counted as income.)  I had assumed that this was a purely regressive income transfer, since home owners tend to have higher incomes.

But, over the last few posts, I have discovered that real estate is actually a higher proportion of low income families' assets and incomes than of high income families, even accounting for the lower home ownership rates.  I am sure that much of the reason for this is that many of these families are retired households that own their homes, so that they have low incomes and low or no debt, and a house that is large relative to their income.  (Well, I say I am sure, but I am less sure of a lot of things after digging through this data.)

Thinking about this, I began to wonder what the distributive effects of the housing shortage really were.  And, as has been typical in this series, the answer is more stark than my newfound intuition suspected.

The first graph shows the level of net imputed rent (before interest expense) earned, per household, by income quintile. This is the average for all families (renters and owners).  This amounts to about 6% of income.  I have used the data from the Survey of Consumer Finances and data from BEA table 7.12 on rental income.  I don't think this imputed rent is counted in any typical measure of household income.  In the BEA construction of Gross Domestic Income, it is reported under Operating Surplus (capital income) as Rental Income.  This should figure into our discourse regarding the behavior of incomes over the past 35 years, because imputed rent has represented an ever-increasing portion of national income.

So, as we can see in the first graph, most rental income is claimed by high income families, in absolute terms.  But this relationship reverses when we look at imputed rental income as a proportion of family income.  The next three graphs are in terms of proportion of income.

Note that, since low income homeowners tend to be much less leveraged than high income homeowners, we have the interesting finding that net rent before interest tends to flow to low income families and interest expenses tend to flow from higher income families.  So, net rental income, after interest expenses, flows overwhelmingly to low income families!  I have been assuming that the recent gains in capital income going to homeowners had been mostly a high income phenomenon, but I was wrong.  Relative to compensation and corporate sources of income, Rental Income is weighted to low income households.

The Net Imputed Rent after Interest Expense/Income is roughly equivalent to the BEA measure of  Share of GDI as Rental Income to Persons with Capital Consumption Adjustment.  My measure runs slightly higher than the BEA measure, because mine is as a percentage of household income while the BEA measure is as a percentage of GDI.  In the SCF data, 1989 appears to be a bit of an outlier, for reasons I haven't dug into, so I will use 1992 as my baseline, so as not to inflate the effects I am describing here.

In 1992, imputed rent after interest was 1.3% of all incomes.  By 2013, it was up to 4.2%.  Please note, this has nothing to do with home prices.  The BEA uses rates in the rental market to estimate the expected rental rates of owner-occupied homes.  In fact, I believe growth in rents as a portion of GDP and growth in net rental income to home owners largely comes from two sources.  (1) Tax preferences to owning, implemented since the 1980's, which have increased housing demand for owner-occupiers, and (2) the collapse of the mortgage credit market after 2007, which has decreased housing supply.  Both of these factors benefit owners at the expense of renters, in terms of income.  (The credit market collapse harms owners, in terms of capital gains and losses.)

As I have argued previously, the BEA measure of rental income is somewhat incoherent, because it subtracts a nominal capital expense (interest expense) from a real capital income (rental income).  In order to correct for this, we need to only subtract the real interest expense from rental income in order to arrive at net rental income for the homeowner.  The inflation portion of the mortgage interest expense is really a pre-arranged savings plan, when considered in real terms.

This next graph shows the effect of imputed rental income on the true income of homeowners (renters are not included in this graph), using real interest expense.*  For all home owning families, this has increased from 4.8% of income in 1992 to 6.3% of income in 2013.  Note that the gains in rental income per household rise in the graphs above in 2001 and 2004, but remain fairly stable in those years in this graph that only includes homeowners.  That is because the growth in rental income in those years was being claimed by new home owners, as the rate of ownership was increasing.  The ownership rate has been falling since the 2004 survey, so the rental income per household is increasing both in terms of the average family and the average home owner.

But the overwhelming take-away from this graph is that changes in this source of income are much more significant for low income families than for high income families.  In 1992, net rental income for owner households in the lowest quintile amounted to 22.3% of reported income, and rose to 28.1% by 2013.  For the median family, it was 6.9% in 1992 and 8.9% in 2013.  For the top 10%, it was 3.3% in 1992 and 4.6% in 2013.

This makes sense because rent claims a smaller portion of income as income rises, and a large portion of families in the lowest income quintile are young renters or retired owners.  The retired owners established home equity over their working lives specifically for this purpose - to lower their retirement cost of living and to hedge against shelter inflation.

So, what if we add this imputed rent income to measured incomes, and track real incomes over time?  The first graph here is indexed to 1992.  We see a jump of about 20% among all income groups, except the top 10%, which sees a jump of about 60%.  The effect of housing is overwhelmed by what I presume is mostly a product of the technology boom.

But, in the next graph, indexed to 2001, we see real incomes which have been stagnant for 12 years because of two significant economic contractions.  Here, all income groups have moved sideways, and here, the best performing group is the lowest quintile.  This should match our intuition, because households in the lowest income quintile have low labor force participation, so they are less affected by economic upheavals.  The incomes of the top 10% have begun to climb again, as recovery takes hold after the 2008 crisis.  And the middle incomes have moved down in real terms.

And, during this period, we see that imputed rent from homeownership takes on more importance, especially among the low income households.  Rent, or implied rent, is an unusually large relative expense for both renters and owners in the lowest income group, so here we can see that perhaps the most significant income effect of the housing bust has been a passive transfer of aggregate income from low income renters to low income owners.  (Well, the most significant effect has been the subsequent crisis that dropped incomes across households.)


PS:  Here is an old post on median household incomes.  The reduction in incomes since 2001 appears to come entirely from the decline in the number of earners per household, which has also been reflected in falling labor force participation that is largely the product of aging.  As Scott Sumner points out, average wages have been growing during this period.





* There are some assumption I have to use in this case.  I have to assume that within each quintile the mean income for home owners is the same as the mean income for renters.  I have a mean income figure for renters and owners for all families, but I don't have one within quintiles.  While there might be some difference, I don't think it would change the outcomes significantly.  I also have assumed that home prices are relatively efficient until 2010, so that implied rental income can be deduced by using families' reported value of their primary residences.  Survey data does tend to overestimate these values, so I did adjust for this by comparing the total survey values to the reported value of owned residences in the Federal Reserve Flow of Funds report.  In 2010 and 2013, home prices are not efficient, so the inflation premium of mortgages that is implied by comparing real returns on homes to nominal effective mortgage rates becomes unreliable.  For these years, I manually set the inflation premium at 1.5%.