Buffett’s Lucky Bet

Pilotless planes will routinely fly passengers by 2030.

Machine intelligence will have passed the “Turing Test” by 2029.

These are two examples of Long Bets sited in Berkshire Hathaway’s annual letter to shareholders.

Of course, the bet that received the most attention was the one that Mr. Buffett placed himself.  In 2005, Warren wagered $1,000,000 that no investment pro could select a set of hedge funds that would outperform an S&P index fund over ten years.

Ted Seides, co-manager of Protege Partners, stepped up to the challenge in 2008. And with one year left on the clock, his odds look pretty grim.

Through nine years, the compounded annual increase for the Vanguard S&P Index Fund was 7.1%. Mr. Seides’ five picks fell far short of this benchmark, gaining an average of only 2.2% annually through 2016 (click for larger image below).

Buffett laid out his reasons for the bet in a statement that was posted on the Long Bets website when the bet commenced. Here are his assertions at the time:

In other words: If Group A (active investors) and Group B (do-nothing investors) comprise the total investing universe, and B is destined to achieve average results before costs, so, too, must A. Whichever group has the lower costs will win.

In short, friction costs weigh on performance. Fees generate friction. Taxes generate friction. And excessive trading generates excessive friction. The more friction, the lower the returns.

In our annual letter to investors, we discussed another source of friction often ignored by passive investors, summed up in the chart below.

Our recent letter rehashed the argument we originally made in a post titled Blind Capital – passive investors typically look brilliant in the late stages of a bull market.

To be clear, we agree with the majority of issues highlighted by Mr. Buffett concerning the failures of active management. But we don’t blindly agree with every statement that comes out of Omaha.

In this case, I’m concerned that investors are being led to put too much emphasis on the outcome, and not enough on the judgment.

This is an important distinction that often goes overlooked. As stated in our annual letter:

Only a focus on the inputs can determine the long run success of the outputs. We must replace the question “Did it have a good outcome?” with the more difficult question, “Was it a good decision?” Was our judgment correct? Was it reasonable given the information available at the time?

Nine years in, it does appear that Mr. Buffett’s bet will have a good outcome. But was it a good decision?

I believe his judgment, in this particular case, was questionable. And the outcome, perhaps a bit lucky. Let me explain why.

Let’s start with the compound annual gain in the market over the last nine years. Mr. Buffett states that the 7.1% return for the index was “a return that could easily prove typical for the stock market over time.”

This is a fair statement (although it depends on one’s definition of “typical”).  Historically, the stock market has generated about 6.5% annual real returns.  Adding 2.5% for inflation gets us to about 9% on average, so it’s fair to say 7.1% is “typical” – i.e. it’s in the same ball park as the market’s average historical returns.

However, the fact that returns since 2008 have been about average, tells us little about what we should have expected in January 2008, based on the information available at the time.

The market’s average historical performance provides a good base rate for forecasting expected returns (which is why most analysts rarely deviate from predictions of mid-single-digit returns).  But a good forecast shouldn’t take the base rate as a given. A good forecast will use the base rate as a starting point and adjust as needed. In this case, we would suggest a big adjustment was needed.

In the same annual letter to shareholders, Mr. Buffett states that “prospective returns fall as assets increase.”

We would add, and I’m willing to wager that he’d agree, prospective returns also fall as prices increase.” All else equal, paying a higher price for a given set of cash flows lowers prospective returns.

Vanguard’s S&P Index Fund is not an exception to this rule. The fund is simply a collection of businesses whose value is determined by a future stream of cash flows. So the price paid for the fund relative to those cash flows is the primary driver of future performance.

Over time, investors can expect to earn a “typical” return from an S&P index fund, only if they buy the fund at a typical price. That price has historically averaged about 16x normalized earnings.

At the start of 2008, investors were paying about 26x normalized earnings for the S&P.  So a prudent forecast, in my opinion, should have ratcheted down expected returns significantly.

For example, on GMO’s numbers, US equities were priced to deliver real returns of roughly zero or 2.4% nominal in early 2008.
Yet, through 2016, the S&P compounded at a 7.1% annual rate over the past nine years.  That’s “typical” for the stock market over time, but far from typical given our starting point.

Here’s a look at what the S&P has done from various starting points historically. A proper forecast should have adjusted the base rate accordingly.

Most investors will look at this data, compare it to the actual and historical results, and conclude that these pessimistic forecasts were simply wrong. How could they be that far off base?

The actual return was pretty close to the market’s “typical” return, so obviously, a “typical” forecast was a more logical forecast. It was certainly more accurate. Right?

Yes, the outcome was more accurate. But that doesn’t necessarily make it a smart bet. This is a slippery slope.

This is the patsy at the table overestimating the probability that the next card will win the hand, betting big, and winning.  But winning doesn’t make a foolish bet wise.

In this case, winning may have been lucky. History clearly shows that actual returns are lower than average returns when investors overpay for the market. Take another look at the chart above and the steady decline in returns associated with higher starting prices.

Returns over the past nine years have been anything but typical. In fact, recent returns for the S&P have been severely inflated. An example may help demonstrate why.

Let’s assume that in January 2008, we believed earnings would grow at 6% annually, which is about in line with long-term historical earnings growth. That would put a $1 of 2008 earnings at about $1.60 today. Throw in the dividend earned along the way, and one might expect a “typical” return for the stock market.

Yet, to actually earn this “typical” return, one would also have had to assume that the market, then trading at 26x normalized earnings, would continue to trade at this historically extreme valuation. The fact that it actually did, does not make this a smart bet, in our opinion.

Mean reversion is perhaps the most powerful force in finance. Yet timing is always uncertain. Extremes can become more extreme before becoming more normal. But on average, valuations have tended to mean revert in about seven years (it is not a coincidence that this is the same time horizon for GMO asset class forecasts).

We did see extreme mean reversion following valuation extremes in early 2008. The market’s normalized earnings multiple plummeted from its peak around 26x to a low below 14x in little more than twelve months. Yet the rebound back to historical extremes occurred just as rapidly.

Since Buffett wagered “the bet,” the multiple on the S&P has stretched from 26x normalized earnings in 2008 to an even more extreme 29x earnings today!

The multiple on the S&P averaged nearly 23x earnings over the past nine years versus the long-term average of about 16x. There are very few nine-year periods over the past century where valuations have remained this high. Making this implicit assumption nine years ago would appear to us a very low probability bet. Yet, the bet paid off.

A bet that pays once tells us nothing about the probability of winning. Unfortunately, there is no way to run this experiment over and over to know how the odds might have played out given a large enough sample size. But the data we do have shows very few, if any, periods in history where valuations have remained this elevated for this long.

Valuations have tended to revert to normal over time. Consequently, a higher probability bet ought to have assumed that valuations trended toward average over this forecast period.

If this were the case, the actual results would have looked much different. Assuming growth of 6% annually, earnings would have grown about 60% from $1 to $1.60 over the past nine years. But all of this earnings growth would have been wiped out by mean reversion in the market multiple. A move from 26x to 16x would have trimmed 40% from today’s value, leaving the market roughly flat for the past nine years: $1 x (1+60%) x (1-40%) ~ $1.

I’m over simplifying here, but you get the point.

Buffett may have gotten lucky. It’s impossible to know, but it’s worth examining rather than taking The Oracle at his word.

This says nothing of the unacceptably low 2.2% returns generated by Mr. Seides’ collection of overpaid hedge fund managers.

Historically, betting on the index has been a smart bet. After transactions costs and fees, most managers underperform.

Yet in January 2008, we would have assumed that a pool of active managers would have outperformed the market, given the starting point. We would have been wrong. We’d be willing to double down today.

Our logic is outlined above. It is also presented in our annual letter. I believe the logic is sound. I believe the outcome of Mr. Buffett’s bet has been greatly distorted by greatly distorted market prices.

Mr. Buffett should know this as well.

He may have gotten lucky.