There has been a long-standing relationship between home Net Rent / Price levels and cyclically adjusted real long-term bond and mortgage yields. There are several ways to adjust nominal bond yields for inflation. In this version of the graph, I have adjusted the mortgage yield with Rent Inflation instead of Core CPI Inflation. Mortgage rates move in a fairly tight range with 10 year treasuries. Here, the persistently high rent inflation we have seen since the 1980s pulls the real mortgage rate below the real 10 year treasury rate (plus a 2% premium). If I used the same inflation adjustment, they would be very close throughout this period.
The yield adjusted by Rent Inflation is the more appropriate adjustment, if we are looking at the non-arbitrage yield level here, because persistently high Rent Inflation acts as an income boost to home ownership, pushing the equilibrium real yield of home ownership down to compensate.
I have dismissed the lack of a tight relationship in the 1980s by averring that there was a high premium on nominal bonds at the time for inflation uncertainty. This premium was very high until about 1985, then there appears to be a small premium that slowly fades until the early 1990s. (The premium in the late 1980s could also reflect the larger benefit of the mortgage tax deduction in a high inflation environment, which would reduce the required pre-tax return on homes.)
So, maybe the recent divergence in relative returns reflects an uncertainty premium in homes, resulting from the recent volatility in home prices. I think this is a theoretically sound point, and if we had been in an equilibrium environment, I would agree that this effect could be in force. But, I don't think it is the constraining factor at work, for the following reasons:
1) The uncertainty in bonds in the 1980s was a direct result of changing real returns. A bond bought in the 1970s, with a coupon rate of 8% (3% real and 5% inflation) was now being paid off with inflated dollars, so that the effective 8% payments were 10% inflation and -2% real. There was a loss in the real value of the cash flows. This is what caused the inflation uncertainty premium to rise and the market value of those bonds to fall.
But, for homes after 2007, there was no income shock. A $200,000 home receiving $6,000 in imputed net rental income in 2005 has continued receiving about $6,000 in imputed rent (adjusted for inflation), with just a temporary lull in rent inflation that would have a very small effect on housing income.
So, the shock in 1980s bonds came from the changing real value of coupon payments while the shock in 2000s homes was purely a price phenomenon. You could say that the changing yield on homes reflected a liquidity shock. But, then, that is my point. A liquidity shock could be mitigated by reintroducing liquidity. There might be some residual liquidity uncertainty premium, but that will be much smaller than the premium created by the lack of liquidity itself.
3) Even in the case of 1980s bonds, there was still convergence over the following decade, as uncertainty receded. In the last graph, here, we can see that volatility in bond market values declined around the same time as the uncertainty premium in the 1980s, sharply at first in the first half of the decade, then slowly for the last half of the decade. And, we do see a similar pattern in home price volatility in the recent period, with higher volatility for about 5 years, which now looks like it is slowly declining. So, if the uncertainty premium has been the binding factor, it should be aging out of market expectations as homes continue to re-establish non-extreme price behavior.
In any case, returns to homes should be converging with bond returns so that, if we re-establish non-arbitrage pricing, bond yields, with the typical risk adjustments, at the top of this cycle should approach or pass imputed home returns.