I have
previously looked at home values, on a city-by-city basis, with the idea of comparing price levels to rent inflation. If a city has had persistently high rent inflation, then an expectation of future inflation should cause home prices in that city to rise. This is based on the basic financial formula for the present value of a perpetual cash flow:
For houses, without constraints to supply, we would expect rent to roughly grow with inflation over the long term, and we can estimate net cash flows after depreciation, property taxes, and expenses, so that net cash flows reflect a perpetual stream of rents on a home with a stable value. So, this equation simplifies to Price= Net Rent (after expenses and depreciation) / Real Long Term Interest Rates. But, in several cities, political constraints to new building have developed, such that there is persistent rent inflation, above the general rate of inflation, so there should be a premium in those cities on real estate prices which is related to that growth rate adjustment in the denominator.
If Net Rent/Price is less than the normal discount rate, that implies expected rent inflation. Low Rent/Price is the inverse of high Price/Rent.
So, anyway, I did that calculation in the earlier post, but I have been thinking about the discount rate. I had used 30 year mortgage rates as a proxy for the housing discount rate. But, I have decided the most of the temporal changes in the mortgage spread are related to cyclical and secular changes in expected pre-payments, so that it actually doesn't make that great of a proxy for required market returns to home equity.

Here is the graph of what the breakeven expected excess rent inflation looked like using the mortgage proxy. But, the expected inflation from the 1990s seemed too high, and I have come to the conclusion that that is because mortgage rates of 7-8% created more pre-payment risk, causing the mortgage rate spread to rise in a way that shouldn't affect returns to home equity.

Maybe, the best proxy is simply long term treasury bonds. There is a 30 year TIPS (real, inflation adjusted) bond that goes back to 1998, and I have inferred it back to 1995, since that has been the benchmark date I have been using as the beginning of the current problem. And, as I have shown in several graphs, for the short time we have had TIPS bonds, long term real treasury yields and the aggregate average implied yields on homes moved together well, as we should expect (until disequilibrium happened in 2007).

It is possible that for owner-occupiers, there is not a 1-to-1 mathematical reaction of price to long-term rates, as there would be with bonds. One reason would be tax benefits to home ownership, the largest of which is the non-taxability of imputed rents. Landlords must pay taxes on their net rental income, but homeowners do not, because there is no cash transaction. This benefit would be proportional to the rental income - that is to say, proportional to the implied real yield. In addition, the tax deferral and tax exemption benefits of capital gains would also be a benefit to homeowners. This benefit would be proportional to the inflation premium portion of nominal long term interest rates.
Here is a graph comparing four time series. The orange line is excess Trailing One Year CPI shelter inflation above Core Inflation. In an unconstrained market, this should have a mean near zero. The green line is the rent inflation required to justify home prices if we assume a required real yield on homes that is unresponsive to interest rate changes. It is fixed at 3.3%, which is the 65 year average. The blue lines assume that 1995 aggregate home prices require no expected excess rent inflation. The light blue line shows the expected excess rent inflation that would have been required to justify home prices over time if home yields track long term real treasury yields. The dark blue line assumes that owner-occupiers require a yield equal to 80% of long term real treasury yields, reflecting tax advantages of owning, plus a constant 1.4% yield premium, reflecting things like liquidity and non-diversification. The net effect is to make home prices less responsive to changes in the real yield.
I think actual market behavior in equilibrium, is probably similar to the dark blue line. One could play around with the specifications, but the range of expected rent inflation that reasonable specifications would suggest, on the national aggregate level, even at the top of the market in 2005, is within the range of rent inflation that has been persistent.
In the previous posts I had reproduced graphs for individual cities. I am not going to reproduce those here. Relative rent expectations between cities would be the same.
But, thinking through this, I think I have turned away from the intuition I was following way back when I started this series of posts. When I started the series, I had kind of figured out that most of the change in home prices was due to the rise in actual rents plus the sharp decline in real long term interest rates. I suspected that the remaining portion of the rise in home values was due to increased demand because low interest rates lowered the barrier to ownership, increasing demand and lowering the "alpha" that comes from limited access to an asset class. (Here's an
old post with a visualization.) And, I suspected that tax benefits that accrue only to owner-occupiers added to the destabilization of the bust by creating a price gap between owner-occupiers and potential diversified buyers that might have been able to fund homes outside the retail mortgage market.
Where I thought I had a novel approach was in identifying the low long term interest rates with tight monetary policy instead of loose monetary policy. Monetary policy is generally thought to work through short term interest rates. Long term rates, especially those that would effect values on a perpetuity like a home, are a different matter, and the secular trend of lower long term real rates may be mostly unrelated to monetary policy. In addition, a long term secular trend of tight monetary policy since the 1970s means that the inflation premium embedded in long term nominal rates is also low, because inflation expectations are low. So, I thought, "Ah, ha!" persistently tight money caused demand to push home prices up because it lowered long term interest rates mainly through lower inflation, but also possibly by lowering real rates. This led to creative lending policies, because the monthly payment was very manageable, and households with little savings could trade an easier down payment for a slightly higher monthly payment. In a high inflation, high interest rate environment, the monthly payment is the constraining factor, so previously, during loose monetary eras, like the 1970s, households weren't enticed into that trade-off.
But, I have become convinced that this demand-side factor is small. Some reasons:
1) From the early 1960s to the late 1970s, we moved from a low inflation context with moderate real interest rates to a high inflation context with low real interest rates, and Price/Rent levels rose dramatically. Unfortunately, I don't know if there are good enough proxies for very long term real interest rates to do a precise quantitative analysis, but it seems clear that regarding home prices relative to rents, intrinsic value (net rent/market value) dominated over demand-side constraints (unaffordable mortgages).
2) Viewed a certain way, point #1 can be missed. There is a decent correlation between home prices and mortgage levels, but the correlation could go either way. Since there is some correlation between nominal interest rates and real interest rates, it is easy enough to imagine that the causation runs from mortgage rates, through buyer liquidity, to higher demand, and finally to high prices. But, the causation could run from low real rates to higher intrinsic values to higher market prices to higher mortgage levels. Point #1 seems to be strong evidence in favor of the second direction of causation - the efficiency story, not the liquidity story. Since 2006, we have been running a sort of natural experiment. We have put strong limitations on mortgage credit. The characteristics of typical approved borrowers have shifted well above normal levels; homeownership has declined sharply; mortgages outstanding have been declining or level for a decade; the private securitization market which had grown to 20% of the mortgage market has all but disappeared.
I think this has been a sharp enough dislocation upon the way our home buying markets normally operate that it has pushed home values below an efficient level. Homes now do provide "alpha" as an investment, relative to the values they have had compared to real long term interest rates in the past. And, yet, the power of the intrinsic value of future rent payments, discounted at current (low) rates, is pulling home prices up, even relative to rents. The current drag on mortgage availability is unprecedented, it should be much stronger now than the effect of marginal changes in nominal mortgage rates. If prices can push toward efficiency in this context, the effect of changing nominal rates in normal markets must be quite weak.
In this graph, we can see how growth in mortgages and changes in Price/Rent ratios tended to move together, appearing to suggest a correlation, before 2007. But, since we have decoupled mortgage availability from mortgage rates, this correlation has broken down. Price/Rent is rising now, even though mortgages outstanding have not been. The correlation between mortgage availability and home prices was spurious. Low real rates cause high home values, which leads to rising home prices and, when we allow it, rising mortgage levels.
 |
Change in Mortgages Outstanding & P/R: Source |
Some might argue that the Fed has been accommodative and that QE money has been propping up the housing market. I would agree, to an extent, that QE probably helped markets overcome the stagnant mortgage market. But, it seems that QE didn't work through the expansion of mortgage credit, and if it worked through broader avenues, we would have seen broader signs of inflation. Also, if Fed policy has been accommodative and targeted at housing, it is strange that housing starts have been so low - and they have been incredibly low for a long time. I would argue that a model of the housing market that suggests home prices are high right now (which seems to be a popular sentiment) would have to be a model that gave little weight to demand-side factors like credit availability. Yet, it seems to me that most observers are claiming that home prices are high right now for demand-side reasons. The efficiency explanation seems more reasonable to me, considering these various peculiar pieces of data.
3) The tax benefits of homeownership (mostly related to the non-taxability of imputed rent and deferred or exempt capital gains) are proportional to inflation rates. So, when nominal rates are low, the effect of taxes on intrinsic values will be less and when nominal rates are high, the tax effect will make intrinsic values higher. This should mitigate any demand-related effects in the other direction. For instance, consider that creditors had to pay taxes on the interest income in the late 1970s, even though most of that income was simply a reflection of inflation, but home buyers did not have to pay taxes on the imputed rental income that was rising by more than 10% per year. From an intrinsic value perspective, this would be a strong draw into home ownership, and provides a further explanation of high home prices in the face of disruptively high interest rates due to inflation.
4) We should not forget that
characteristics of
homeowners during the 2000s boom were stable. The evidence suggests that homebuyers were not utilizing credit markets during the period of expansion to increase their demand for housing, in terms of rental value.
5) During the boom of the 2000s, increased intrinsic values from changes in real interest rates and existing rent levels explained 2/3 of nominal home price increases. The city-specific data on Price/Rent levels suggest that most, if not all, of the remaining price increases were related to expected persistent rent inflation. Of the 130% in average price increases from the mid 1990s to 2006, very little, or none, of that price increase requires a demand explanation.