Is this time different? I’ve often characterized our approach to the financial markets as a value-conscious, historically-informed, evidence-driven discipline. In recent years, we’ve often been asked whether the world has changed in a way that makes historical evidence an inadequate guide to investing.
Our own narrative in the half-cycle since 2009 certainly invites that question. While a good part of that was a self-inflicted outcome related to my fiduciary stress-testing inclinations in 2009, and we’ve adapted far more than is likely to be obvious until the present market cycle is complete (see Setting the Record Straight to understand the those challenges and how we’ve addressed them), the broader question remains – do historical regularities no longer apply?
In Probably Approximately Correct, Leslie Valiant describes the conditions required by both living organisms and artificial intelligence “ecorithms” in order to learn and successfully use induction from observations drawn from the environment. In order to reliably learn from inference, two assumptions are required.
First, we can’t expect those lessons or generalizations to be useful in contexts that are fundamentally different from the environment that produced the observations used for learning. In other words, lessons learned in one world may be poor guides in a far different world. This, of course, has been the perennial argument of speculators during every bubble in history. But we should be careful about over-using that argument. The “invariance assumption” doesn’t require that the world cannot change – it only requires that there are some regularities that remain true. Second, some useful criterion or regularity must in fact be detectable. That is, there must be some “learnable regularity” that can actually be inferred from the evidence – a random world with no relationships between cause and effect is not an environment that will produce useful or predictive generalizations.
In some cases, those learnable regularities can be derived on the basis of clear theoretical relationships that describe how the world works with reasonable accuracy.
For example, every long-term security is fundamentally a claim on a very long-duration stream of cash flows that can be expected to be delivered into the hands of investors over time. For a given stream of expected cash flows and a given current price, we can quickly estimate the long-term rate of return that the security can be expected to achieve (assuming the cash flows are delivered as expected). Likewise, for a given stream of expected cash flows and a “required” long-term rate of return, we can calculate the current price that would be consistent with that long-term rate of return. The failure to understand the inverse relationship between current prices and future returns is why investors frequently argue that rich equity valuations are “justified” by low interest rates, without understanding that they are really saying that dismal future equity returns are perfectly acceptable.
We also observe the very regular tendency for profit margins to increase during economic expansions (presently corporate profits are close to 11% of GDP), and to contract during softer periods. Corporate profits as a share of GDP have always retreated to less than 5.5% in every economic cycle on record, even in recent decades. Since stocks are most reliably priced on the basis of long-term cash flows, and not simply Wall Street’s estimate of next year’s earnings, we find that valuation measures that are either relatively insensitive to profit margin swings, or that correct for their variation over the economic cycle, are much better correlated with actual subsequent market returns than measures such as price/forward operating earnings that don’t do so.
Our valuation concerns don’t rely on any requirement for earnings or profit margins to turn down in the near term. Valuations are a long-term proposition that link the price being paid today to a stream of cash flows that, for the S&P 500, have an effective duration of about 50 years. In evaluating whether “this time is different,” it should be understood that current valuations are “justified” only if 1) the wide historical cyclicality of profits over the economic cycle has been eliminated, 2) the average level of profit margins over the next fivedecades will be permanently elevated at nearly twice the historical norm, 3) the strong historical advantage of smoothed or margin-adjusted valuation measures over single-year price/earnings measures has vanished, and 4) zero interest rate policies will persist not just for 3 or 4 more years, but for decades while economic growth proceeds at historically normal rates nonetheless. Believe all of that if you wish. Without permanent changes in the way the world works, on valuation measures that are best correlated with actual subsequent market returns, stocks are wickedly overvalued here.
