GMO letter for the first quarter ended March 31, 2016; titled, “Keeping The Faith”
The past five years have been challenging for long-term value-based asset allocation. We do not believe this constitutes a paradigm shift, dooming such strategies in the future. The basic driver for long-term value working historically has been the excessive volatility of asset prices relative to their underlying fundamental cash flows, and recent history does not show any evidence of that changing. Outperforming the markets given that pattern requires either betting that the excessive swings will reverse over time or accurately predicting what those excessive swings will be. The former strategy amounts to long-term value-based investing, while the latter requires outpredicting others as to both what surprises will hit the markets and how the markets will react to them. Our strong preference is to focus on long-term value, despite the inevitable periods of tough performance that strategy will entail.
JPMorgan’s Erdoes From Delivering Alpha: How The Pandemic Is Triggering Deglobalization
The more connected the world has become, the easier it has become for investors to look outside their home country to find alpha. However, the pandemic has resulted in a shift toward deglobalization. In an interview on CNBC's Delivering Alpha, JPMorgan Asset and Wealth Management CEO Mary Callahan Erdoes, Alibaba Executive Vice Chairman Joseph Tsai Read More
It’s no secret that the last half decade has been a rough one for value-based asset allocation. With central bankers pushing interest rates down to unimagined lows, ongoing disappointment from the emerging markets that have looked cheaper than the rest of the world, and the continuing outperformance from the U.S. stock market and growth stocks generally, it’s enough to cause even committed long-term value investors to question their faith. Over the past several years, we at GMO have questioned a lot of things, including assumptions that we had held without much question for decades, but we have not wavered in our belief that taking the long-term view in investing is the right path and that in the long run no factor is as important to investment returns as valuations. In this letter, I’m going to talk about some of the reasons we continue to believe this so strongly.
Perhaps the first point to make on why we continue to stick to our beliefs is that this is far from the first period in which the patience of long-term value managers has been tested. A decade ago, we were all told that the great moderation had changed the rules of investing, making it safe to invest in risky assets without the margin of safety that used to be required. Less than a decade before that, we were all told that the internet had changed the rules even more profoundly, making anyone who was prepared to put money into boring REITs or TIPS in return for paltry mid to high single-digit real returns a fool for foregoing the hugely greater potential returns from investing in the loss-making companies that were someday soon to become massively profitable. As GMO’s portfolios were positioned for the opposite in both cases, we got plenty of complaints from clients that we just didn’t get it. But the periodic struggles of long-term value investors far pre-date GMO’s founding. You can hear the frustration evident in John Maynard Keynes’s quote from the General Theory back in 1936: “It is the long-term investor, he who most promotes the public interest, who will in practice come in for most criticism.”1 I can’t tell you exactly what was going on in Keynes’s head when he wrote that, but to me it has all the hallmarks of someone who has just come back from a particularly trying investment committee meeting.
So this is certainly not the first period that has tested the faith of long-term value investors, but the fact that a style of investing has seen problems before is far from a guarantee that it will succeed in the end. While our faith is helped by the fact that this is not our first experience with misbehaving markets, the reason for our belief comes much more from a systematic study of history and the fundamental drivers of asset returns. The evidence is clear that asset prices are much more volatile than can be justified by the underlying fundamentals. This is the basic driver of the long-term returns to valuebased asset allocation, and recent history, as painful as it has been for some of our bets, shows every bit as much excess volatility as the more distant past did. A world in which value-based mean reversion will not work in the long run is a low volatility world in which asset prices do not deviate from the slow-moving fundamentals that power financial markets. It is extremely hard for us to justify the last several years of market behavior through that lens, which leaves us confident that our strategy will work in the end.
Value has always been a risky strategy, particularly for those trying to run an investment business. The drivers of mean reversion are not hugely powerful at any given time, meaning asset prices and even the underlying fundamentals can move in unexpected ways for disappointingly long periods. It is a little glib to say that without this risk, it would be difficult for asset prices to get meaningfully out of line in the first place, but the reality is that the only way you can get really exciting opportunities for mean reversion is to have misvalued assets become even more misvalued before they revert to fair value. This is the catch-22 of value-driven investing. Your best opportunities will almost always come just at the time your clients are least interested in hearing from you, and might possibly come at the times when you are most likely to be doubting yourself.
GMO – Asset prices are too volatile
So, why do we believe that asset prices are more volatile than they should be? Robert Shiller, the Nobel Prize winning Yale economist who is the source of so much common sense wisdom on financial markets, did a simple but powerful test of this almost 30 years ago in a paper for Science magazine.2 Shiller noted that while we cannot know the future with any degree of certainty, we have no such limitation when it comes to the past. He looked at U.S. stock market prices and dividends back to the 19th century and came up with a “clairvoyant” fair value estimate for the market based on the actual dividends that were paid over the next 50 years. This analysis, which is reproduced and updated in Exhibit 1, made the following important point.
The volatility of U.S. stocks since 1881 has been a little over 17% per year. The volatility of the underlying fair value of the market has been a little over 1% per year. Well over 90% of the volatility of the stock market cannot be explained as a rational response to the changing value of the stream of dividends it embodies. This means that the volatility is due to some combination of changing discount rates applied to those cash flows, and changes to expectations of future dividends that turned out to be incorrect. It is difficult to determine exactly which has been the driver at any given time, but there doesn’t seem to be a lot of evidence for changing discount rates having been a major force. Even in the most extreme overvaluation in U.S. stock market history, the 1999-2000 internet bubble, none of the investors we heard explaining why the stock market was rational to have risen to such giddy heights explained it on the basis that future returns should be lower than history.3
It’s somewhat harder to dismiss the changing discount rate argument as a partial explanation for today’s high equity valuations. Unlike 1999, bond yields today are strikingly low, and so if stocks should be priced by demanding a risk premium off of a risk-free rate (an idea that my colleague James Montier, for one, rejects4), we could understand why stock valuations might be high today relative to history.
But if a changing discount rate had historically been a significant driver of stock market volatility, we would expect to see some explanatory power for bond rates in explaining changes in stock prices. This connection, at least historically, is almost completely absent, as Exhibit 2 shows.
If this chart looks like a complete mess, that’s because it is. The R2 of changes in nominal bond yields to explain stock market returns is 0.02, and for changes in real bond yields it is 0.01, although because the beta on the factor has the wrong sign (higher real bond yields imply positive stock market returns), it’s hard to even give credit for that 1% explanatory power. In either case, we cannot explain any meaningful part of the excess volatility of the stock market historically as coming from the piece of the discount rate that can be observed.5
GMO – Predicting surprises is hard
The fact that markets are more volatile than they need to be certainly doesn’t mean you can’t try to predict and profit from those “unnecessary” moves. But how do you predict them? To a very significant degree, markets in the short term are driven by “news” – events that would have to be predicted ahead of time to properly guess what the markets will do in the future. This task is complicated by the fact that even given an accurate prediction of news or surprises, which for this purpose are synonymous, predicting the actual response of the market is surprisingly tricky, because the connections between data and events and stock market performance are often unintuitive.
One point to recognize is that predicting news may be possible, but it certainly cannot be done by the median participant, because if the median participant predicted it, it wasn’t news in the first place. But the market consensus is generally dreadfully unimaginative in any event. To take a relevant example, probably the most important single piece of macro information that investors would like to know ahead of time is the timing of recessions. Exhibit 3 shows the historical ability of economists to predict recessions in the form of predictions of the following year’s GDP growth.
Economists have entirely failed to predict any of the recessions we have had in the time since consensus forecasts were available, and this may actually be an unfairly easy test. In principle, economists could have successfully predicted recessions and had the information be useless for investors because the market had already priced in the probability. But every single recession – and indeed pretty much every surge and dip in GDP growth over the past 40 years – has been poorly predicted by Wall Street economists.
GMO – Figuring out what to do with surprise
But let’s assume that you could actually predict surprise, and in a piece of data widely known to move markets. For this purpose, I’m going to use U.S. payroll data. We get new payroll data every month and we have a record of the consensus forecast going into each release since 1997. So we can see the impact of properly predicting the surprise in payroll data going back almost 20 years. Over a day, the impact is clear. A positive payroll surprise leads to an average return of 0.24% for the day, while a negative surprise gives an average return of only 0.02%. This seems perfectly reasonable, but what is peculiar is what happens over a longer period. We might readily expect that the bulk of the excess return to the data would come out in the day it is announced, but the real world is actually not even that kind. Exhibit 4 shows the one-day, one-month, six-month, one-year, and three-year returns. You can see that for a day, the impact of a surprise on payrolls is material. As the time horizon lengthens, the “importance” of the surprise disappears quickly and then, oddly, actually shifts sign! This is a pretty striking result. The market does indeed rise when we get a positive surprise on payrolls, but as time passes, the news turns out not to actually matter for the true fair value of the stock market, or at least not in the manner traders might have originally assumed.
It’s a little hard to explain the one-month and six-month data consistently – why would the impact of surprise be negative over the first month and then positive over the next five? But the pattern over one to three years certainly suggests that, all else equal, you’d rather own the stock market in the few years after a disappointment in employment than after a positive surprise. Does this mean that investors should sell when the payroll data surprises to the upside? Probably not. It’s not a particularly huge effect and there are many possible reasons why it could have wound up looking odd. But it does mean that if you are planning to make money by predicting payroll data better than the other guy, you’d better be quick about it, because there is no actual evidence that strong payroll data actually increases the value of the stock market in any lasting way.
See full PDF below.