Brad Delong claims that equity premium in the US is about 7 1/2% because capital markets have become deranged by mistrust of corporations following the financial crisis:
If you look at asset prices now, you confront the minus two percent real return on the debt of sovereigns that possess exorbitant privilege with what Justin Lahart of the Wall Street Journal tells me is now a 5.5% real earnings yield on the U.S. stock market as a whole. That 7.5% per year equity premium is a major derangement of asset prices. It makes it very difficult for us to use our standard tools to think about what good policy would be…
His point is much more about appropriate macro policy than about the outlook for the equity market, but I found his claim of a 7 1/2% equity premium jarring, given that my very rough proxy of the premium is 1 1/2%.
I don’t have a strong view on the market here, beyond believing that medium to longer-term returns are likely to be very subpar. But just as an accounting exercise, what might explain these 600 basis points?
First, he measures the earnings yield for the overall market and reports that it is 5 1/2%. I see both trailing and Shillerized operating earnings in the S&P at about $ 100 per share *, which means that the market currently delivers an earnings yield of just under 4 1/2%. So that gets the equity premium down from about 7 1/2% to 6 1/2%.
Second, for some reason he uses a blend of real borrowing rates for countries with “exorbitant privilege”, rather than an estimated real yield on the 10-year Treasury. I see the 10-year US yield at 2.6%, which translates to a real yield of just over 1%, given my inflation expectation proxy of just under 1.6%. (This is roughly consistent with TIPS, once we take account of the fact that I am measuring inflation with the deflator, rather than the CPI.) So that gets the equity premium to just under 3 1/2%.
Finally, and this is the controversial bit — which is why I call my proxy “rough” — I adjust the equity premium for the deviation of potential growth from its all-sample average, which results in chopping off almost another 2 percentage points. The justification for this is simple, if not fully convincing, even to me. If we are going to take on board the lower real borrowing rate, then ought we not also take on board the lower potential growth rate that it may primarily reflect?
If you believe that the earnings yield is equal to the expected real return to equities by virtue of an arbitrage relationship, then the answer to my rhetorical question is no. But within that perspective, earnings are DEFINED as that which generates the result, rather than measured directly. Not wanting to fall for a tautology, I prefer to take on the lower growth rate directly. But I concede — indeed insist — that this is the source of some uncertainty.
Another approach, which is probably actually better, is to solve for the equity premium within a multi-stage dividend discount model, as Goldman does. If I recall correctly, that approach generates a premium with roughly the same shape as my rough proxy, although with a narrower range, presumably because the forecasts implicit in the approach are correlated to some extent with market prices. If you have access to their work ,you might want to cross check that against my own metric.
For now, I don’t think asset markets are priced as though they are deranged by pessimism and mistrust, at least here in the US.
* The Shiller CAPE uses a ten year-moving average of inflation-adjusted reported earnings. When calculating “Shillerized” earnings, I prefer (following Jesse Livermore at Philosophical Economics) to use operating earnings, which currently net out the effects of the large write-downs during the financial crisis. So I apply Shiller’s smoothing algorithm to operating rather than reported earnings. However, I need to make two adjustments, which are somewhat offsetting. First, I discount the operating earnings series by 10% because there seems to be a recurring upward bias of that amount, even when there are no major write downs. Second, I center the Shillerized figure by multiplying it by the trend real earnings growth rate compounded over five years. I want the measure to be smoothed for the cycle, but to line up with the period for which it is being observed.
JPM Core PCE estimate
I don’t buy Jamie Dimon’s claim that he does it all for America. But I do think that JPM economists do a pretty good job of relating the individual price detail in the CPI and PPI reports into an estimate of the core PCE deflator for the same month.
This month they say the core PCE deflator will have been up just 0.03% in November. The CPI was up just 0.1% and much of that was rents, which are overweighted in the CPI. In the chart below, I update the right pane with the JPM estimate for November, although the left panel is shown only to October. Watching this inflation renormalization, which is what I hope it is, is like watching paint dry.
I don’t know why bond yields are down today. Out of it. It is not this. The PPI and CPI were out a while ago and I got this JPM estimate pretty belatedly.