When Bad Things Don’t Happen
Shortly after the start of the war in Ukraine, when Russian stocks became un-investable and went to zero for US investors, Chinese stocks also saw a precipitous drop in price. At the time, I interpreted this as the market pricing in the possibility that the same thing could happen to them. If the Chinese government supplied weapons to Russia, it seemed likely that Chinese stocks would also become un-investable and go to zero for US investors.
In the end, that did not occur. China kept Russia sufficiently at arms length to avoid sanctions, and within four months, Chinese stocks were back to their original prices, more or less.
Looking back, there are no indications in company fundamentals that the potential for this bad outcome ever existed. Earnings and dividends paid out to investors were as expected. The risk was real, but the only evidence that it existed was in the movements of stock prices.
Explaining Stock Price Volatility
I bring this up because it is, in my mind, a straightforward counterexample to part of the argument in Robert Shiller’s 1981 paper that stock prices change too much to be estimates of intrinsic value using the dividend discount model. Shiller makes the claim that, if stock prices are “optimal forecasts” of future dividend payments, properly discounted back to present value, then stock prices should be less volatile than dividend payments, whereas in reality, they are more volatile.
We saw above, however, that this need not be the case. If a stock suddenly has a 50% chance of being a zero, then its price should drop by 50%. If later on, that possibility evaporates, then it should move back up to its original price. These changes in prices are rational. However, if the possibility of going to zero does not materialize, then the dividend payments may not demonstrate any volatility, even though the risk was real.
In short, the record of dividend payments will only reflect outcomes that did happen, whereas prices incorporate what could happen. Risk means “more things can happen than will happen”, so prices could, in some cases, be more volatile that dividend payments, rather than the opposite, as Shiller claimed.
Stationarity
The paper has a mathematical proof of his claim that stock price volatility, if it is the estimated present value of future dividends, must be less volatile than the dividends themselves. That proof makes some mathematical assumptions, at least one of which must be violated in the example above since the proof itself is peer-reviewed and almost certainly correct.
To my eye, it looks like the violated assumption is that the dividend payments are drawn from a probability distribution that is (in some sense) “stationary”, meaning that it does not change with time. The paper does assume that we learn more information over time, which can change our expectations about the likely amounts of future dividend payments, but the true distributions of those dividend payments do not change. They are not allowed to be described by, say, a normal distribution up until a certain time and then, after that, described by a 50% chance of being from a normal distribution and a 50% chance of being always zero.
An assumption of stationarity along these lines seems necessary to make the math tractable. As Shiller says, “some stationarity assumption is necessary if we are to proceed with any statistical analysis”. However, not everything in real life is mathematically tractable. Investing in stocks involves “unknown unknowns”, which none of us can price accurately. We have to wait until they become “known unknowns” to do that.
Pricing in a Recession
The example of Chinese stocks is, admittedly, exotic. Stocks rarely see the sudden arrival of a significant possibility of going to zero. However, I think most stocks do see a similar phenomenon on a regular basis, namely, when the market as a whole starts to price in the possibility of a recession — when we go from thinking that a recession will happen “at some point in the future” to thinking that one may happen in the next 12 months.
Since recessions are usually accompanied by drops of 30–50% in the stock market, I interpret a drop of 20% (a bear market) as the market pricing in a significant possibility of a near-term recession. If, as people like to say, the market has predicted “9 of the last 5 recessions”, then when a bear market occurs, there is a 5/9 chance that a recession is starting soon. If that recession comes with a 36% drop in stock prices, then we would price that in with a (5/9) x 36% = 20% drop right now. If no recession occurs after all, then just as in the example of Chinese stocks, we may not see any change in fundamentals even though the risk was real. The market is not being irrational by changing prices even though future fundamentals did not change.
That said, I do have a difficult time rationalizing the associated, 30–50% drops in stock prices when recessions occur. Even a couple years of completely lost earnings in a recession would not decrease the intrinsic value of most stocks by 30%. After all, more than 80% of the intrinsic value in a typical stock is from earnings more than five years in the future, well after the recession will have ended. And that is assuming earnings go to zero during the recession. In reality, while a recession would lead to a decrease in earnings for a typical company, it would not wipe them out completely. If we assume earnings drop by 50% for three years, we wouldn’t get anywhere close to a 30% drop in intrinsic value, so as I said, it seems difficult to justify a 30% price drop using discounted dividends or cash flows.
Pricing Growth vs Value Stocks
Fellow OSAM Research partner, Jesse Livermore, made a similar argument about the effects of interest rate changes on the intrinsic value of growth stocks. It is true that a higher risk-free rate (holding everything else constant) would decrease the the intrinsic value of growth stocks more than value stocks, so there was a basis for this widely-accepted explanation of the recent outperformance of value over growth. However, if you try to calculate the impacts of rates on growth vs value stocks using discounted cash flows, you should see that the difference is minor. It would not — as far as I can see — lead to the magnitude of price changes seen in the last year.
In this investor’s opinion, the market is very clever in the direction of price changes. Most price moves can be explained as pricing in real risks that have recently become apparent — in fact, the market often prices them in before they are apparent to most of us. On the other hand, the market seems hysterical in the magnitude of its price changes, typically moving prices up or down much more than can be justified by a reasonable calculation of intrinsic values, at least as far as I can see.
Conclusions
I find thinking about risks to be the most fruitful way of understanding most price moves. The risk of recession can explain overall drops in the market as well as relative outperformance from “safe” sectors like utilities and staples. Moves in individual sectors are also usually explained most easily by risks: oil & gas stocks can sell off over increased risk that more investors will see the sector as un-investable, cable companies can sell off over increased risk that more consumers will switch to streaming over linear TV, and so on.
An individual stock, a sector, or the market as a whole can see prices move due to actual events, but to me, most price moves seem to be in anticipation, thinking about risks of what might happen in the future. A stock can gap down on an earnings miss, but the magnitude of the drop is more usually consistent with worries about a permanent change in growth rates rather than a single dividend payment being smaller than expected.
Most of the time, it seems that these risks are not realized or, at least, are not as bad as feared, so the eventual changes in fundamentals are smaller than the price changes. To me, that is as it should be. Investors should price in the risk of potential bad outcomes, even when they are unlikely to occur, so it does not surprise me that prices are more volatile than fundamentals.
As I said above, however, when I try to analyze price changes using discounted cash flows, I am usually unable to justify the magnitude of those changes in terms of the risks that seem to explain their direction. So in the end, I have to say that I agree with Shiller. Even though I feel that there is nothing wrong with prices that are more volatile than dividends, I too am unable to find a way to justify typical price movements by changes in intrinsic value, as calculated using the dividend discount model. I may disagree with parts of his argument, but I can’t find a way around his conclusion.
