|Efficient Markets Today
John H. Cochrane
The theory of efficient markets has been an astonishingly powerful and successful organizing principle for empirical work in finance. Under its banner, finance grew from a collection of anecdotes to a science. It sparked the revolution in financial markets that we now enjoy.
But this is not a dead field, endlessly reveling in 35-year old triumphs. Our understanding of financial markets continues to change in fundamental ways. A second revolution is underway, and a third is beginning. Though less heralded, each will change our understanding of financial markets as much as the efficient markets revolution did.
At its heart, the efficient markets theory says that prices should equal expected discounted cashflows. In the early days, the “expected” part took center stage: Researchers focused on the “efficient” incorporation of information into prices. Since the early 1980s, however, our focus has been much more on the “discounted” part.
Empirical discoveries forced this shift. Pretty much any time we see information, we find that it is quickly reflected in market prices. But prices also move a lot (over time and across securities) when there is no cashflow information. Every expected-return anomaly turns out to correspond to a source of common movement. Even the momentum stocks rise or fall together. These are just the sort of observations that suggest variation in discount rates, i.e. expected returns, or risk premia.
Our empirical understanding of discount rates has changed completely from the simple random walk - CAPM view of the 1970s. We used to think expected returns are constant over time, and values change when expected earnings change. Now we think that all variation in market valuation ratios corresponds to changing discount rates, and none to changing forecasts of earnings or dividends. We used to think that if interest rates are higher at long maturities or in other countries, interest rates would rise or exchange rates would depreciate to offset the lure. Now we think that all such interest rate variation corresponds to changing risk premia, and none to expected changes in rates. We used to think that all variation in average returns across assets corresponded to variation in one, market, beta. Now, market betas explain none of the cross-sectional variation in risk premia. In each case the “all” and “none” have reversed.
And we’re not done. The number of apparently priced “factors” is growing explosively. And, worse, more and more research is quietly abandoning betas altogether in favor of characteristic adjustments. We have a lot yet to learn about risk premia.
This focus on discount rate variation, and this dramatic change in our empirical understanding of discount rates, changes deeply how we think about markets. Let me give you a few examples.
Macroeconomics and finance
If we want to understand markets, if we want to assess whether markets are “rational” or not, we can no longer simply look at prices and news. We have to understand whether the risk premia we see in asset markets are correctly linked to macroeconomic events.
It’s not hard to intuit what those events might be: in addition to fearing market falls, investors fear that a stock might fall when jobs disappear or private businesses fail, when prospective returns are poor, when house values fall, or when financial markets dry up, illiquid assets are hard to sell and credit becomes hard to get. It’s not at all surprising that factors such as these turn out to be important. The surprise is that the simple CAPM worked so well for so long!
It has taken much more work to come up with useful measures of these feared macroeconomic events. However, this search is slowly bearing fruit. As I look across the broad brush of empirical work, something like “recession” and “financial distress” factors are at work, exactly as we might suppose.
It’s easy to get discouraged by the slow pace and seeming imprecision of this work. But every known anomaly comes down to some pattern of expected returns and betas. The only measure of “rationality,” is whether the risk premia we see in asset markets are correctly linked to macroeconomic events. Gene’s “joint hypothesis” theorem, now perhaps the most important part of his efficient markets paper, proves that any other discussion is meaningless.
Where will this go? I think we will end up with two kinds of models: There will be finance models in which ad-hoc-looking “mimicking portfolios,” such as size, value, and momentum, account for a larger cross-section of assets and puzzles. There will be separate economic models that explain why those factors get risk premia. We will use the first kind of models for day-to-day non-structural risk adjustment and to summarize the information in many assets. But we will never have an economic understanding of markets or their efficiency, nor will we be able to answer structural questions like “how does stabilization of the economy affect discount rates?” without the second kind of model. In fact, this is where we always were: The CAPM was always silent on why the market equity premium was so high, so never really provided an economic explanation of risk premia. CAPM nostalgia is misplaced.
Alphas and betas
The “alpha and beta” framework used to evaluate active management has lost its meaning in the discount-rate world.
Recall the good old days. There was one source of systematic risk: beta. “Alpha” if there was any, reflected “inefficiencies”, “skill” in finding stocks whose prices did not yet incorporate the manager’s information. Investors understood this, had decided how much beta they wanted, and could then intelligently benchmark managers, manage risk, and chase alpha claims.
Suppose that we do a modern evaluation of a modern hedge fund, and we replicate its performance with the market, size, value, and momentum portfolios, a put option-writing strategy, and mechanical strategies that invest in currencies following interest differentials, in corporate bonds following credit spreads, and in treasuries following yield and liquidity spreads. (This is only the beginning of the exploding list of “factors” we now routinely use!) “Aha,” we say, “we have explained your ‘alpha’ by ‘systematic risk’ of ‘mechanical portfolios.’”
“Wait a minute,” the hedge fund manager responds. “None of my investors has ever thought through how much put option or carry-trade risk-exposure they want. They’re sitting on the market index. These premia are alpha to my investors. Knowing which factors work, how to run these regressions, and how to trade these assets is my skill.”
Our manager has a point. If we want to call what he has “information”, it is information about discount rates of groups of securities (and their covariance matrix!) rather than information about cashflows, and it is information that is incorporated in prices.
The “style” choice is now the main, hard, and only portfolio problem. For the most current strategies, there is no alpha and beta, there is just beta you know about, and beta you don’t understand yet, and no clear separation between the two. The alpha/beta, style/selection, active/passive, skill/index paradigm has become meaningless.
Arbitrage vs. risk sharing; “marginal” vs. average investors; segmented markets.
The process of arriving at equilibrium is fundamentally different in the discount-rate world.
If a piece of information is not correctly incorporated in market prices, we only need a few arbitrageurs or “marginal investors” to trade. They don’t have to take very large positions or bear much risk. In fact, the no-trade theorem studies the puzzle that in theory their private information should be revealed in prices with no trading at all!
However, if some “systematic” factor (momentum, carry trade, put option writing) has an unwarranted risk premium, the only cure is for that risk to be more widely shared. The average investor must change his demands. This is much harder, so markets can maintain “segmented” risk premia for a long time, even while trading within each market quickly removes any informational “inefficiencies.”
This insight helps to explain why many discount-rate anomalies are not quickly “arbitraged away.” It also explains why tests for informational inefficiencies were such a quick and resounding success in the 1970s, but understanding discount rate variation across different markets and strategies is so elusive; why we find endless new premia, each associated with a common source of variance, but each bearing strained relations to overall, average-investor, macroeconomic risk. And changes in such premia can help us to understand a lot of otherwise puzzling price volatility, as we observed last summer.
How is this risk sharing going to happen? If, say, the risk premium on credit-default swaps becomes bigger than it should be, how are you and I going to increase our demands? The only realistic hope is that our investment is mediated by a high-tech institutional investor that can quickly spot, process, and take on systematic risks on our behalf. I have just described a hedge fund. As a Chicago Economist I am delighted to find some use for a rapidly growing institution that evidently meets some market test, rather than just continue to deplore active management.
The discount-rate world will force a sea-change in how we do all sorts of things in finance.
Standard cost-of-capital calculations use the CAPM and a steady 6% market premium. These need to be rewritten, recognizing multiple pricing factors, and varying risk premia over time. The results should be more realistic too, for example telling firms to invest more when values are high and the market premium is low.
A change in prices driven by a discount rate change does not say anything about how close the firm is to bankruptcy, contrary to the standard models of corporate bond spreads and capital structure (debt vs. equity). This observation can address the puzzle that firms do not rebalance capital structures when the stock market goes down.
In asset pricing, I think discount factor variation means we will end up studying stock prices and cash flow streams rather than one-period returns. Really, why is the covariance of tomorrow’s price with the market price (beta) the exogenous variable that explains average returns? Two-period thinking works in an iid, cashflow-driven, CAPM world, but no longer. Tractable present value models are emerging that include realistic risk premia. These models will change our treatment of stocks just as deeply as the Cox-Ingersoll-Ross model changed our treatment of bonds and the Black-Scholes-Merton model changed our treatment of options.
The next revolution.
Though this one is not half over, I think I spy the next revolution on the horizon.
Our current theory produces nothing like the enormous trading volume we see, because it has no room for information trading. Now, I see increasing evidence that prices and information-based trading are deeply interrelated.
Let me show you a picture. The left-hand panel shows price indices through the late “bubble,” and the right-hand panel shows volume. The price explosion in Nasdaq tech stocks neatly coincided with a volume explosion. This is a pervasive pattern: every “bubble” has coincided with a trading frenzy, and an obvious and asymmetric flow of information, whether about tulips or transistors.
All current theories used to analyze these prices ignore volume. This is just as true of “behavioral” models, e.g. based on excessive investor optimism as it is of “rational” theories, e.g. based on the option value of technology.
Well, perhaps the volume is a coincidence. Or perhaps prices really are set exactly as they would be with no volume, and volume follows. But, with such a striking and pervasive correlation in front of us, perhaps there is some causality going the other way. Or, perhaps the explosion of information-based trading volume should be at the center of our account of these prices – which means that something other than conventional cash flow expectations and discount rate measures is at work!
There is a similar stunning correlation of orderflow with price changes, indicating the “price discovery” process in which informed agents trade while the market learns their information, and an new literature that finds first-order effects of liquidity – only of interest to high-frequency traders – on prices.
I’m not sure how we will understand information-based trading and its price effects. We don’t have good economic models yet. It seems a stretch to me to claim that there is no such model -- that the mere existence of the NYSE proves once and for all that people are “irrational,” whatever that means. But once again, the facts are pushing us to think harder about it. And the result will nicely tie up the loose end of “market efficiency”: we should understand how information becomes reflected in prices.
In sum, we haven’t been sitting around rehashing efficient-market glory for the last 35 years. Asset pricing is not a dead field in a mop-up phase. The discount rate revolution is only half over, and its implications, for academia and for practice, are only beginning to be appreciated. The revolution of understanding information-based trading, and its effects on prices, has only barely begun.
Efficient markets –the proposition that information is quickly reflected in prices -- are with us as always, but as Darwin lies over the modern explosion of genetic research, not as a biblical, final statement of the last grain of truth, or even as a focus of much useful debate. That’s good. A science moves on. And that’s why it’s such exiting time to be working I this field, and especially to be working at Chicago.