Estimating Future Stock Returns, September 2019 Update

As I have been writing about this topic for 3+ years, I want you to understand what my greatest doubt is regarding this theory. I can’t predict the ratio of the value of equities to the value of all assets well. Thus, each quarter when I write, I have a surprise, because the ratio changes in ways that I can’t explain.

Is the difference that big? No, but I would like to be able to track the differences better, so that I can have a better explanation for what is going on.

The variable of the ratio of the value of equities to the value of all assets is composed of two separate variables:

  • The value of equities
  • The value of everything else

To understand the ratio, I needed to understand how the two underlying variables behaved. I had two hypotheses:

  1. The value of equities would track the S&P 500.
  2. The value of everything else would track the returns on T-bills and 10-year T-notes.

My first hypothesis was not falsified. The regression that I ran of increases in the value of equities against the returns on the S&P 500 had an intercept near zero, and a beta coefficient near one. The joint hypothesis test did not falsify the idea of the intercept being zero and the slope coefficient being one. The R-squared was 83%.

The second hypothesis validated the idea that T-bill returns were correlated with the growth of all other assets, but not 10-year T-notes. The sign on the T-notes was negative and insignificant. The upshot of the regression was that all other assets grew at roughly (5% + 0.5 * T-bill yield) per year. The R-squared was 27%.

Now over the last 10 years, the value of all equities was drawn down by 0.82%/quarter. So, with the growth in “all other assets” and the diminution of the value of all equities net of unrealized capital gains, the ratio variable has always tended to decline versus a forecast that naively applies the appreciation of the S&P 500 to the ratio.

So, now I have two “facts:”

  • The value of equities appreciates at the rate of the S&P 500 less 0.82%/quarter. On net, people take money out of stocks.
  • The value of everything else appreciates at the rate of (1.25% + half the T-bill rate) per quarter. It’s as if half of the money is in the bank, earning nothing, and the other half is in T-bills. On net people add money to their non-risk assets.

As such, I have a good idea as to how the ratio should track over time. It’s not perfect. There will be surprises. But it should minimize the differences period to period.

With the run in the markets since September 30th, the estimated return on the S&P 500 over the next ten years has fallen from 3.92%/year to 2.97%/year. Expected return levels like that are in the lowest 5% of returns.

That said, return regimes often last longer than they should, and result in larger mis-valuations than should ordinarily occur. Part of the misvaluation here is that interest rates are low, and there is no alternative to stocks.

My view is that cash will not lose. Bonds will lose a little. Stocks may lose a lot. I am not yet trading out of stocks, but I am getting closer to doing so.

Be wary. Valuations are elevated. Reduce equity exposure as you get the opportunity. If valuations rise further, I will reduce my own stock holdings. I only write about this quarterly, so this is my warning to you. Consider selling some of your stocks.

9 thoughts on “Estimating Future Stock Returns, September 2019 Update

    1. Not really. I do three things.

      1) I inform them of what I think, and offer them the opportunity to make changes to lower risk strategies.

      2) I raise and lower a limited amount of cash for the stock strategy.

      3) I raise and lower a limited amount of bonds for the balanced strategy.

      #1 is meant to do the heavy lifting. I used to hire managers for equity strategies, so I don’t want too much latitude for market timing in either the stock or balanced strategies.

  1. It looks like the most recent Z.1 report data is as of Q3 (it’s not clear to me whether that means 7/1/2019 or 10/1/2019)? Either way, how does your model estimate returns as of December 2019 without updated data from the Z.1 report? Is there a source for this data on a monthly basis?

    1. After some thought, I suppose the article is the answer. 🙂

      In lieu of the Z.1 report for months after 7/2019 (or 10/2019?), are you using the estimated aggregate investor equity allocation (described in the article) to forecast the 10 year return?

    2. Sorry for all the posts!

      I think I found the answer in your post here:

      https://alephblog.com/2017/02/14/two-questions-on-returns/

      ?Making adjustments for time elapsed from the end of the quarter is important, because the estimate is stale by 70-165 days or so. I treat it like a 10-year zero coupon bond and look at the return since the end of the quarter. I could be more exact than this, adjusting for the exact period and dividends, but the surprise from the unknown change in investor behavior which is larger than any of the adjustment simplifications. I take the return since the end of the last reported quarter and divide by ten, and subtract it from my ten year return estimate. Simple, understandable, and usable, particularly when the adjustment only has to wait for 3 more months to be refreshed.?

      P.S. – what throws me off on the FRED data is that the csv output lists the 2019Q3 data as 7/1/2019, but I think it?s actually as of the end of Q3. I?d like to know for sure though.

    1. It’s a very simple paper, and it doesn’t add that much to what is already out there. He sort of hints at an idea that I have proposed, and talked to a professor at Towson State about, but it would require access to the CRSP database, and for me it is too expensive, or I don’t have the time to work with the professor.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Theme: Overlay by Kaira