Monday, February 2, 2015

Housing Tax Policy, A Series: Part 5 - Mortgages ARE savings

I've written about how the Mian & Sufi style narrative about debt fueled instability gets it all backwards.  Before, I have noted that increases in mortgage levels were not a sign of unsustainable consumption.  They were an artifact of middle class Americans deferring consumption by investing in real estate.  When a household buys its first home, with say a 20% down payment, they are making a significant saving decision.  In the process, for each dollar they save, $4 of debt are created.  But that debt isn't due to consumption.  It is an artifact of our conventions around real estate ownership.  It's due to saving.

But, in thinking about returns to real estate, I've realized that their error is even worse.  Mortgages aren't just an artifact of saving.  They, in part, are savings.

We think about these things in nominal terms, so we think of the homeowner as capturing capital income on imputed rent.  If he has a mortgage, we think of the claim on that income being split between equity and debt holders.  This split is represented by the interest payment.  So, as I framed it in the previous post, we start with the total return to the home, and then we subtract the interest payment to find the residual return to ownership for the homeowner.

We can see from the graph that there is some distortion in this way of thinking, because the returns to the homeowner are too volatile by this measure.  The error made by this framing becomes more clear if we think in real terms instead of nominal terms.  The returns in the graph are real returns, because homes are an inflation-protected security, with the inflation adjustment coming through rent inflation.

To think of home returns in real terms, we need to think of the mortgage interest rate as having two parts - the real rate and the inflation premium.  The interest payment, then, is divided between a payment for the real cost of credit and a payment for inflation.

So, the real total return is still the total of net rental income to the owner and total interest expenses.  But, to find the true real returns to the owner, we only subtract the real portion of the interest expense.

The inflation portion of the interest expense simply reflects the change in the value of money and assets over time.  Inflation will cause the home to increase in nominal value, which is a capital gain for the owner, and it reduces the real value of the mortgage repayment, and the mortgage provider takes the inflation premium payment in cash.

In national accounts, this is treated as interest income for the lender and interest expense for the homeowner.  But, thinking about it in real terms, that is not an accurate portrayal of the transaction.  In effect, the homeowner has committed to a saving plan, where each year he agrees to purchase a portion of the equity in the house, equal to the expected rate of inflation x the outstanding mortgage, from the lender.  So, this is more accurately portrayed as capital income earned by the homeowner, which is then saved by the homeowner through the purchase of additional home equity, in real terms.

It's important to be clear about the difference in expected earnings and realized earnings.  Clearly this income isn't earned in such a clean, stable fashion.  But, that is a product of the different exposures the lender and the homeowner are taking as the market prices of the mortgage and home fluctuate.  This is about expected earnings and income, not experienced capital gains and losses.

Note that with the current tax treatment of owner-occupied real estate, mortgage financed home ownership has the advantages of both a regular IRA and a Roth IRA.  The contributions (interest payments) are tax deductible and the distributions (realized capital gains) are mostly untaxed.

The Public Policy Implications

Before I move on to look at how this effects the reported numbers, let's think about the public policy implications of the "unsustainable debt" narrative.  It takes as its motivating fact the sharp rise in household debt.

Note when those rises in debt levels occur.  The mortgage tax deduction advantage was created in 1986 and the capital gains tax advantage was strengthened in 1996.  What a coincidence, huh?

But, the narrative interprets this as an unsustainable attempt by low-to-middle income households to make up for stagnant incomes.  These bar graphs show some numbers, by income level (from 2004).  Keep in mind that 35% of households rent.  So looking at the debt levels of these households, by income, we see that essentially the entire mortgage tax benefit goes to the top 50%.  And the bottom half of households hold a very small portion of this debt.  The debt that adds up and makes scary looking graphs like this one comes from upper income households.

I have included the homeownership rate in the line graph, also.  Notice that the tax incentives for mortgages and home ownership don't seem to correlate with higher homeownership.  Homeownership rates did stop declining when the mortgage deduction advantage was instituted in 1986, but it settled at the same level it had been in the 1960s.  And home prices and real estate values also didn't move up after 1986.  Then, homeownership levels increased coincidentally with the new CRA in 1994.  But, note that the homeownership rate had already moved from 64% to 67% before mortgage levels started to increase in the 2000s.  (Don't get me wrong.  I think increased access to homeownership from initiatives like the CRA might be a good thing.)  So, it appears that the increases in mortgage utilization and real estate investment that were related to these tax policies did not result in new households getting access to homeownership.  They simply provided new tax benefits to the households that already owned homes.

And what is the "unsustainable debt/income stagnation" narrative proposal for fixing their misdiagnosed problem?  The proposed solution is to raise marginal tax rates on high income households.

Now, what effect do you think higher tax rates will have on the supposed debt problem?  There will be more incentive for these households to move assets into tax advantaged investments - like homes and mortgages!  So, upper middle class houses will buy larger homes and take out larger mortgages.  And in the New York Times, we'll read articles on the homes of the rich, and how obnoxiously extravagant they are.  And next to those articles will be articles about indebted households and how they just keep getting further and further underwater.  "The system is rigged!", they'll say.  And how could you deny it?  And, I suppose the solution then will be to raise taxes even higher?

The Numbers

First, I am going to compare effective nominal mortgage rates (BEA table 7.12, line 159, divided by Household Home Mortgages Level from Flow of Funds report) to the total return on owner-occupied homes (rental income + net mortgage interest expense / market value of homes).  I will assume that the required real return on the home is equal to the required real return on the mortgage.  Obviously, there are many factors that differ between these two assets, but I think that in the aggregate, it is a decent proxy.  Correcting the 30 year mortgage rate for inflation gives a similar result.

First, look at this relationship.  Until 2007, the inflation premium implied by the difference between mortgage rates and total returns to homes was a decent reflection of inflation expectations.  Based on this long-term relationship, returns to homes were close to their reasonable value (and, thus, so were prices).  It is the relationship since 2007 that is completely off-track.  Starting in 2007, home prices collapsed due to the liquidity crisis, so returns to homes, as a proportion of home prices, have been way too high.  Clearly, it is home prices in 2008-2015 that are out of order, at a much greater scale than home prices of 2005 could have been.

Now, by allocating returns to debt between real returns and the inflation payment (which is a de facto equity purchase by the owner) we can use this information to re-allocate imputed returns on the home between the owner and the lender.  This looks much more reasonable than Net Rent/Price did in the first graph.  We can see that most of the dip in net Rental Income in the 1970s and 1980s was the result of homeowners capturing very large real gains in equity ownership from their lenders because of the high inflation rates of the time.  Then, after 1986, the real gains in equity came from the higher Loan to Value levels, as households took advantage of the tax subsidies.  (At 2% inflation, a household with 40% LTV would only expect to gain 0.8% of equity through their mortgage arrangement.  But, a household with 50% LTV would gain 1%.)

The current proportions are the result of a disequilibrium in the housing market.  Until 2007, I believe that these proportional returns reflect relatively efficient prices.  But, since 2007, home prices have collapsed due to the inoperable mortgage credit market, so homes have been earning higher returns than mortgages have, distorting the share of rental income split between owner and lender.  I have previously taken a stab at real returns to home ownership, with a more simple approach, which also found excess returns to homeowners in the current market.  When I first began thinking about this, I had thought that homeowners earned excess returns in general because of the limited access to home ownership that results from limited access to mortgage credit.  As I work through the data, and think about the details, I am coming to a more detailed understanding of what might be happening here.  I will be getting into those issues in later posts.

So, homeowners are currently capturing equity in their homes through nominal expected capital gains.  It's just not showing up in the graph as a cash transfer to the lender because they are getting it for free.  If the housing market were functional right now, home returns would be similar to those of the mid-2000's. Homes would be earning a little under 3% real returns.  (The lower returns would generally be a function of homes having higher nominal prices than they currently do.)  But, there has been a lot of deleveraging since the mid-2000's, so the owner's portion of the real return to the home would be higher and the lender's would be smaller now than they were in the mid-2000s.  And owners would be making positive payments to the lender for the inflation premium.


  1. Very interesting stuff with regard to transfers from renters to homewoners.

    But who captures those gains. As someone who bought a house in 2009, am I a net beneficiary of these transfers? Or did I pay a I higher price for the home that I would have absent the transfers, so that I, as a recent buyer, am no better or worse off than I would have been otherwise?

    1. Interesting question, Michael. My first instinct is to say, "Obviously, it's to the owner's benefit." Imagine a home fully mortgaged 100%. You are paying 4% on a mortgage, and you are earning 4% on the net rent plus 2% on rent inflation. You're getting a net return with no capital.

      But, your question does suggest a risk-adjusted way of looking at this. Mortgages seem to have always paid about the same spread as home ownership - about 2% above nominal treasuries. It seems strange to me how little variation in mortgage interest rates there is between a mortgage with LTV of 50% vs. one with LTV of 80% or 90%. But, maybe before 2007 the expected foreclosure risk of a home starting at 90% LTV and one at 50% LTV were seen as similar. Default rates in the early years would be very low and default risks in the out years, even with a high beginning LTV, would be low, because LTV would quickly decline as the home's nominal value rose. So, why wouldn't mortgages tend to have a lower rate, since they don't have as much exposure to the volatility of ownership value? I suspect that though the spreads on the returns are similar, the return premium on mortgages is more a result of prepayment risks, etc.

      So, I suppose that you could look at what has happened and say that mortgages have similar premiums to what they have always had, but there has been a large increase in the perceived volatility of ownership. So, (1) lenders require lower LTV's and (2) homeowners require higher returns to account for the variance.

      This would suggest risk premiums that would be more persistent than the credit constraint issues that I have blamed for the high implied returns to homes. That would mean that home prices won't continue to recover as easily as my credit story would lead us to believe.

      But, the strange thing here is that there wasn't a collapse in rental income. Equities are more volatile than bonds because their income source is more volatile. So, by my narrative, the housing crash came out of the liquidity and credit collapse. Lower home prices caused the higher implied required returns, not the other way around. In other fixed income securities, the higher required returns would cause the lower price.

      So, I don't think the risk adjusted story holds here. It's a demand side story. Now, because we've just been through this mess, there might be some extra return premium that the market will require in homes because the demand problem caused the price volatility. So, maybe the causality starts running both ways. But, even if that's the case, I don't think the premium would high or persistent enough to keep home prices from returning to their previous levels.

      In either case, the sign to watch for would be a renewed growth in bank real estate credit. Because, if required returns do go back to previous lower levels, that implies less expected volatility, which means that the banks will be comfortable with rising LTV's, and the available set of potential mortgage credit will expand.

      I would expect there to be a 10-15% governor on home price rates of increase, just do to frictions in the real estate market. The possible stories I've outlined above would probably readjust quickly enough to allow that rate of home price appreciation (an annual decline in the implied required return of well less than 1%, I suspect, though I'm too lazy right now to check the math.)

      Anyway, thanks for the question. I hadn't really thought of the issue that way before thinking of an answer.