Dividends, Treasuries, Inflation, and the Cheapness of Stocks
Today I posted an article regarding the issue of earnings yields versus bond yields and whether that can point us towards an overvalued or undervalued market. For those interested in the subject, I really encourage a read of a paper that Alex Dumortier pointed me towards -- Fight the Fed Model by Cliff Asness.
I think Asness hits it on the head when he says that the Fed model is largely bunk -- at least in the sense that investors should automatically assume that earnings yields should line up with Treasury rates. But in the article I also hinted at the fact that I didn't necessarily agree with everything Asness suggested.
In particular, I don't buy his assertion that the present value equation shouldn't suggest to investors that stocks should carry somewhat higher multiples when bond yields fall.
So, first, why does he say this? His reasoning is that when inflation is low, that brings down future nominal earnings growth and he backs up that assertion by presenting a regression of nominal earnings growth on inflation, which shows inflation being significantly predictive of nominal earnings growth.
His assumption, then, is that bond yields are driven by inflation expectations, so when bond yields are low, inflation will be low. And so when we're working out our present value equation, when the risk free rate falls, it's a drop in prospective growth that balances it out, not a lower required rate of return (and therefore a higher valuation).
Now here's where my problem comes in. Asness points out that it's difficult to work with data on inflation expectations and that's why he used coincident nominal earnings growth and inflation data. But that's a problem because his leap of logic doesn't work as a result.
When future period inflation is regressed on current period bond yields, we see that there is actually little, if any, predictive power there. In other words, just because bond yields drop, it doesn't necessarily mean that inflation will actually be low, and therefore that growth expectations need to be pulled back to the full extent. And if we did use the resulting regression model to predict inflation from bond yields, we'd get inflation predictions that are more moderate (that is, higher lows and lower highs) than if we relied on bond yields directly. In other words, bond investors seem to overreact -- which I don’t think should be all that surprising.
As a result, I think that the actual functioning is probably somewhere between the pure Fed approach -- where valuation does all the moving -- and Asness' assertion -- where growth does all the moving. That is, we get some movement from valuation, and some movement from earnings expectations.
From the view of a corporate investment, let's say 3M has $500 million in free cash sitting around and let's say the company's only two choices are to invest in bonds or build a new factory. Initial returns expectations for the factory are 5% per year and bond yields are at 5%.
Now what happens if bond yields fall to 3%? At first glance the factory is now much more attractive. However, as Asness would point out, the fact that bond yields have fallen means that the bond market is saying that inflation is likely to be low. And if inflation is low, then it could be that the goods the factory produces won't sell for as high of a price and returns will be somewhat lower. But my counterpoint is that the bond market isn't always terribly prescient when predicting inflation and so returns may not fall as much as bond yields. So we could end up with bonds at 3% and prospective factory returns at 4% or 4.5%. The factory is still a better investment than the bonds, but not as good as it seemed if we didn't adjust returns.
We could put together a similar scenario if bonds jumped to 10% -- bonds would probably be the better way to go, but an increase in nominal expected returns on the factory would close the gap somewhat.
Would I stand up and defend the Fed model? No. But I would argue against Asness' assumption that bond yields / interest rates don't or shouldn't have an impact on equity valuations. Of course Asness is a pretty sharp mind (to say the least), so I figure it's quite possible that I've missed something here, so I'm happy to hear comments / thoughts / refutations.