It’s even worse than Buffett thinks.

Here’s another article showing the efficiency of markets:

Warren Buffett made a friendly bet four years ago that funds that invest in hedge funds for their clients couldn’t beat the stock market over a decade. So far he’s winning.

The 81-year-old Buffett, who is chairman of the holding company Berkshire Hathaway Inc. (BRK/A), ended last year neck and neck with the Protégé funds as the Vanguard fund climbed by 2.1 percent and the Protégé hedge funds lost an estimated 3.75 percent.

The first two months of this year pushed the Vanguard fund ahead as stocks returned 9 percent, more than twice the gains of hedge funds.

If you do the math the performance gap is around 10%, even larger than four years of the 1.25% annual expense ratio for mutual funds that invest in hedge funds. They’d have been better off throwing darts.

But before Buffett gets too cocky, he might want to consider the final sentence of the article:

Berkshire Hathaway shares have slumped almost 17 percent since the end of 2007.

Ouch! Three years ago I argued that even if markets were perfectly efficient, they would look inefficient. That’s because for every 1,000,000 investors you’d expect one guy to outperform the market for 20 consecutive years. The masses would call that lucky guy a genius, even if he was just an average bloke from Nebraska.

I predicted that after the lucky guy became the richest investor on Wall Street, his returns would no longer look so impressive. So far I appear to be right. That doesn’t mean the EMH is precisely true (it’s not), but I’m having more and more trouble finding any useful anti-EMH models for investors.

And while we are at it, how about those bearish stock market comments by Nouriel Roubini in May 2009? And why wasn’t Robert Shiller screaming “buy” in March 2009?” One by one all the anti-EMH arguments that have been thrown at me seem to be melting away. The EMH is widely ridiculed, but will outlive all its critics.

PS.  In fairness to the other side, I did a post bashing Keynes as an investor about a year ago.  It now looks like his decisions managing the Cambridge University endowment were substantially better than those managing his personal investments.  Oddly, he wasn’t able to market time, rather he was a successful stock-picker.  One would expect a macroeconomist to win through market timing, if they had any advantage at all.

HT:  Cyril Morong


Tags:

 
 
 

47 Responses to “It’s even worse than Buffett thinks.”

  1. Gravatar of 123 123
    9. April 2012 at 05:50

    Why did the Berkshire Hathaway shares went down 17 percent since the end of 2007? Because Warren Buffet was wrong about macro.

    Previously he was boasting that value investors can ignore macro, but in his latest annual report he wrote “Wise monetary and fiscal policies play an important role in tempering recessions, but these tools don’t
    create households nor eliminate excess housing units.”

  2. Gravatar of JP Koning JP Koning
    9. April 2012 at 05:59

    Like the financial markets, the market for academic ideas is also perfectly efficient. No academic can consistently outperform over the long term. Take for instance, this blog’s somewhat mediocre habit of repeating the same EMH blog post each month. Meh. Underperforming. That doesn’t mean the EMH is precisely true (it’s not),but I’m having more and more trouble finding any useful anti-EMH models for academics.

  3. Gravatar of MP MP
    9. April 2012 at 06:06

    “Ouch! Three years ago I argued that even if markets were perfectly efficient, they would look inefficient.”

    Even if markets were inefficient, wouldn’t they look efficient if investment skill was incredibly rare? And from a practical standpoint, does it make a difference which is true if the implications are the same? You say EMH is approximately true therefore don’t try and beat the market. Buffett says investment skill is too rare, so don’t try and beat the market.

  4. Gravatar of Mike Mike
    9. April 2012 at 06:23

    Buffet’s been outperforming the market for 56 years, not 20. He opened his first investment fund in 1956.

  5. Gravatar of Jon Jon
    9. April 2012 at 06:55

    Largely what has happened is that Berkshires price to book ratio has declined. The market used to expect an excess return and no longer does.

    One common claim is that this is due to the absence of any clear succession, but regardless, buffet claimed that the book value reflected intrinsic enterprise value. If that is true, then projecting the last ten years gives a return over twenty to be about 9%.

  6. Gravatar of Cthorm Cthorm
    9. April 2012 at 07:09

    The benefits of hedge funds and conglomerates like Berkshire are pretty severely understated by these arguments. Leverage and large illiquid investments amplified their losses (valuation losses for BH. Revenue increased 50% from 2006-2011, but EPS fell from ~$7 to ~4. Railroad business has very high operating leverage, so EPS will outpace revenue growth once [if] NGDP picks up). Ask yourself what would have happened had the Fed been doing NGDPLT. Leverage and low-liquidity investments can boost returns, but they also increase sensitivity to systematic (i.e. monetary policy) risk. The Hedgefund/Berkshire/Equity Market comparison is a bit of a false dichotomy regarding the EMH as it is, because they all have very different risk/return expectations, and often try to exploit areas where market information is particularly incomplete. Absent monetary policy risk Hedge funds/Berkshire should outperform the broad equity market.

  7. Gravatar of John hall John hall
    9. April 2012 at 07:37

    1) Some of the best evidence that would be consistent with EMH is that log changes in individual equities and the market can be approximated by a random walk and that prices in liquid securities adjust very quickly changes in information. However, EMH can’t be tested due to the joint hypothesis problem.
    2) Buffet can’t still produce outsized returns because he is too big and well-known. If he began again with a $1mn portfolio, I would still bet that he could outperform the market over a twenty year period (if he stayed alive and working). After twenty years of 20% returns, he would only have $38mn, far from his current portfolio.
    3) Buffett’s counter to the lucky investor out of a 1mn every 20 years in the Superinvestors of Graham and Doddsville was that if you have several investors with the same investment philosophy and all have strong returns over time, then it is hard to say that it was luck that drove their returns.
    4) I never found that 1mn every 20 years argument very convincing. On a strictly mathematical basis, its obvious that only 50 percent of the money invested in every year can outperform the market. However, that doesn’t mean that it holds for specific investors since not everyone has the same initial endowment. Further, the only people who care about beating the market every year are braggarts and marketing people. After 20 years, the investor should care about the total return of their portfolio versus the total return of the index. If you underperform in a few years by a small amount when the market is up, then this isn’t really a big deal so long as you produce outsized excess returns in other years.
    5) The better argument is the empirical facts that most mutual funds after fees do not add significant alpha. The same can be said for your point about investing in a portfolio of hedge funds. An EMH world would show the same result, but the evidence of it doesn’t necessarily mean that EMH is true. Nor does it mean that investment management can’t add alpha. It is possible to find a handful of funds that have produced very strong returns and it is unlikely that these returns were due to luck (for instance, Renaissance’s Medallion fund). The hard part is finding these funds before they are successful and when they are willing to take money. To do so is likely just as hard as choosing a stock that will outperform over the next twenty years.

  8. Gravatar of Benjamin Cole Benjamin Cole
    9. April 2012 at 07:41

    Scott, On Keynes for Cambridge: As you point out, Keynes could have been lucky, or, in that position, given some market-advantage real inside information.

    The EMH is broadly and mostly true. Sure, people working fiendishly for years might devise profits in nooks and vortices of the huge current. The analogy falls apart, but those people make the current even more EMH.

    BTW, on Buffett: His company actually has made a lot of purchases of other whole companies, often in private transactions. This may not be stock-picking as much as finding good companies with so-so management, and bringing in fresh management and leverage of all kinds–financial, operational, connections.

    The returns should be higher for active hands-on management of companies, than in passive ownership through stock picking.

  9. Gravatar of MP MP
    9. April 2012 at 08:11

    “I’m having more and more trouble finding any useful anti-EMH models for investors.”

    This got me thinking about two things. First, about some models “value” investors have advocated using for the public equity markets. They say stick to smaller cap stocks. Or illiquid special situations. Or beaten up companies from which most buyers have fled. In general, they say look for areas where there is little competition because that’s where the excess returns are (indeed, as Berkshire has gotten bigger and bigger, Buffett has admitted that its returns will become worse and as he’s forced away from these more fertile sub-markets). This seems to me the same conclusions that a pro-EMH investor would arrive at. Saying you want to focus on markets with less competition is the same as saying you want to stay away from markets where EMH is more likely to be (approximately) true. In fact, these strategies are a big area of focus in the Columbia MBA program (where Buffett studied under Graham). So I think you are right that EMH will outlive its critics as it’s already put to practical use by them, whether they know it or not.

    This leads to a second point, which is that the “stock market” is not a single, homogenous market. Yet when people debate EMH, they’re often imprecise and miss that nuance. The market for Microsoft is not the same as the one for Northeast Community Bancorp. When an investor tries to debunk EMH he’ll usually point to a company like the latter ($70 million market cap), but there’s no reason that proves anything about the efficiency of Microsoft’s stock.

  10. Gravatar of Jason Jason
    9. April 2012 at 09:13

    Scott,

    Have you read Berk and Green? I wonder what your thoughts are about this paper and efficient markets?

    The basic idea is that the lack of persistence is exactly what you would expect if investors in mutual funds are knowledgeable Bayesians and managers have skill, or stock picking ability. The gains from trade go to the managers not the investors.

  11. Gravatar of ssumner ssumner
    9. April 2012 at 09:54

    123, From that quotation it looks like he’s wise to ignore macro.

    JP, Touche.

    MP, I agree.

    Mike, You said;

    “Buffet’s been outperforming the market for 56 years, not 20. He opened his first investment fund in 1956.”

    That fact would only be important if he’d outperformed the market in all 56 years. But he hasn’t.

    Jon, I’m not sure I followed your argument.

    Cthorm, I accept that these may be false comparisons. I’m just reporting all the arguments that were thrown at me; “Hedge funds do great, Warren Buffett is smarter than the market, etc.”

    John Hall, I agree with most of what you have to say, but don’t see it as refuting my claim. I don’t believe the EMH is literally true, just that anti-EMH models are not very useful, and the EMH is useful.

    You said;

    “3) Buffett’s counter to the lucky investor out of a 1mn every 20 years in the Superinvestors of Graham and Doddsville was that if you have several investors with the same investment philosophy and all have strong returns over time, then it is hard to say that it was luck that drove their returns.”

    Maybe, but not if their investment periods overlapped.

    Ben, Good point about the returns to hands-on management.

    MP, Good points.

    Jason, I don’t see how that could explain mean reversion in returns, but I haven’t read the paper. How do they explain the fact that the skilled manangers aren’t able to consistently beat the market?

  12. Gravatar of Jason Jason
    9. April 2012 at 10:03

    Just think about it in very simple terms. Lets say there is free flow of money into mutual funds but variable fees. If the mutual fund industry itself is efficient, then it must be true that fees on a fund rise to the point where investors do not expect excess returns above the market. Otherwise, money would continue to flow in to the fund. In other words, the gains from trade from skill always have to go to the people with the skill, not investors.

    How do fees actually rise to do this? Well the fee is constant but the ability of the manager displays diminishing returns. Managers do something called “closet indexing”. The basic idea is that if you have very little money, you can earn much higher returns because with a lot of money your trades have a big impact on market prices. As money flows in, your opportunities to invest that money diminish and your fund starts to look more like the index.

    The result is that enough funds flow in to eliminate any persistence in returns.

  13. Gravatar of StatsGuy StatsGuy
    9. April 2012 at 10:34

    “That’s because for every 1,000,000 investors you’d expect one guy to outperform the market for 20 consecutive years. The masses would call that lucky guy a genius, even if he was just an average bloke from Nebraska.”

    Actually, that’s a statistical falsehood. You are assuming no serial correlation.

    _systematically_ outperforming is the definition of “skill”

    _randomly_ outperforming can include long-term bets (or strategies, or simple cognitive worldviews) that create correlation across years

    Thus, the odds of outperforming the market over 20 years are greater than 1 in a million. Likewise, the odds of underperforming the market over 20 years are also greater than 1 in a million.

    As a general aside, one area of hedge fund activity that has been performing is private equity – the reason for outperformance, however, has largely to do with private information and limited access to the market. As the market broadens, returns will decline.

  14. Gravatar of John hall John hall
    9. April 2012 at 10:38

    I’m not sure what you mean by anti-EMH models.

    For instance, is a pairs trading model an anti-EMH model? You can’t predict the price of either security, but there is evidence that for some types of securities you can predict when the spread between them will narrow.

    Another tricky type of model is a Garch model that shows that volatility is mean-reverting. The mean return is modeled with an AR(1), so can best be considered a random walk, very EMH-like. However, since the volatility tends to follow a more predictable process than the return, a mean-variance investor would substantially improve risk-adjusted returns by accounting for this fact. Not very EMH-like.

    Markov regime-switching models are also tricky to categorize as EMH or non-EMH. In these models, the investor is able to calculate an expected mean and variance in each regime and then calculates the probability they are in each regime to to adjust the expected means and variances depending on that. Since each regime could be thought of as different random walks, it is sort of EMH-like. However, there is evidence that portfolio allocation using regime-switching estimates can produce excess returns. This should not occur under many definitions of even weak form market efficiency (since only prices are used).

  15. Gravatar of StatsGuy StatsGuy
    9. April 2012 at 10:51

    @ John Hall

    “However, there is evidence that portfolio allocation using regime-switching estimates can produce excess returns.”

    I’m going to raise the notion that regime switching models are simply strategies that create additional serial correlation. Consider the following facts:

    1) Small schools are more likely to be among America’s best performing schools
    2) Small schools are more likely to be among America’s worst performing schools

    BOTH are true.

    Regime switching models effectively reduce the number of observations in the dataset. Their only advantage would be if they identified a weakness in existing human investors (like a cognitive bias) or quant investors (like an algorithm bias). As soon as this bias was identified, it would tend to be hedged.

    Also, any retrospective study of strategy performance is subject to overfitting – that is, for every 20 studies run, one will be significant. This affects all areas of statistical inquiry.

  16. Gravatar of Major_Freedom Major_Freedom
    9. April 2012 at 11:50

    ssumner:

    Ouch! Three years ago I argued that even if markets were perfectly efficient, they would look inefficient. That’s because for every 1,000,000 investors you’d expect one guy to outperform the market for 20 consecutive years. The masses would call that lucky guy a genius, even if he was just an average bloke from Nebraska.

    You’re just reinforcing your own worldview by using it as a premise to support the same conclusion. One could just as easily have written:

    “Ouch! Three years ago I argued that even if markets were efficient, they would look inefficient. That’s because for every 1,000,000 investors you’d expect one guy to outperform the market for 20 consecutive years. The masses would call that genius “lucky”, even if he was an above average bloke from Nebraska.”

    Every pro-EMH argument falls like a house of cards. None of them stand up to close scrutiny.

    Successful investors who take advantage of market inefficiencies, if only for a finite period of time, are the falsifications of EMH, and yet those who argue in favor of EMH wait, and wait, and wait, and wait, they will wait 20 years until some investor makes a loss, then they will say “AHA! Told you so!”, as if the market being inefficient somehow requires an individual to permanently exploit market inefficiencies without anyone else learning more than he does over time, such that they can exploit the inefficiencies better than him.

    That the market is not efficient as “efficient” is understood by those who argue in favor of EMH, does not mean that the same individual is immune from making losses. It means that profits can be made by anyone who acquires the knowledge and skill to exploit market inefficiencies.

    Those who call folks like Warren Buffet “lucky” are just losers trying to justify their own shortcomings by ignoring his knowledge and skill and pretending it’s all luck. It is so obvious in your tone. You say Buffet “shouldn’t get too cocky.” This is just your jealousy and resentment speaking. It has nothing to so with his knowledge and skill enabling him to exploit market inefficiencies that you and others are too dimwitted to exploit yourselves.

    Calling an investor who outperforms the market for 20 years “lucky” is like calling a guy who wins 20 lotteries in a row “skilled and knowledgeable.”

    And while we are at it, how about those bearish stock market comments by Nouriel Roubini in May 2009? And why wasn’t Robert Shiller screaming “buy” in March 2009?” One by one all the anti-EMH arguments that have been thrown at me seem to be melting away. The EMH is widely ridiculed, but will outlive all its critics.

    If EMH is (not precisely) true, then it doesn’t require that Roubini and Shiller in particular make good calls. It just means they were not knowledgeable or skilled enough to exploit market inefficiencies compared to their competition.

    Market inefficiencies does not require that a particular person is able to exploit a single constant market inefficiency until they die, after which they pass their knowledge down to the next person whereby he does the same thing.

    Market inefficiencies themselves change over time, by virtue of the inefficiencies being exploited, after which new information sets taking center stage and new market inefficiencies arise.

    You have it backwards. It’s not that perfect market efficiency will look inefficient. It’s that perfect market inefficiency will look efficient.

    What’s hilarious is contemplating Sumner’s EMH logic, and then asking how could an efficient market change prevailing stock and bond prices on the basis of the mere comments and investment choices of some lucky regular bloke from Nebraska! Hahaha. I mean, if the market is so efficient, how in the world could it move prices on the basis of just one lucky dude’s comments and investment choices?

    Sumner’s EMH thesis refutes itself, no help from anti-EMH folks is needed.

  17. Gravatar of Floccina Floccina
    9. April 2012 at 11:59

    How about Eric Falkenstein’s (http://www.efalken.com/video/index.html) assertion that people are willing to pay for a chance at out-performance (they buy lottery tickets) thus more volatile stock are overbought and as a group under perform? He says that in the past the safest stocks out performed the riskier stocks.

    If you consider lottery tickets as an investment (and why not) then it is easy to outperform the typical American, just by not buying lottery tickets.

  18. Gravatar of 123 123
    9. April 2012 at 12:12

    “From that quotation it looks like he’s wise to ignore macro.”

    He no longer ignores macro. Today he is in a median macroeconomist mode, making pronouncements last September like “I don’t think that fiscal stimulus or monetary policy from this point forth will do a lot, to be perfectly honest. I do think that the natural regenerative juices of capitalism will do, and are doing, plenty.”

    I was going to argue that Buffet might be right that Berkshire stock is undervalued these days, but then I realized that I need to look deeper into it, and if a more detailed analysis shows that Buffet is right, I need to adjust my portfolio first before commenting here 🙂

  19. Gravatar of 123 123
    9. April 2012 at 12:16

    @Cthorm

    Hedge funds outperforming the market – do you believe the illiquidity premium is larger than 2 and 20?

    I don’t agree that monetary policy presents an obstacle to a hedge fund performance. John Paulson was long monetary policy screw-ups in 2007.

  20. Gravatar of John hall John hall
    9. April 2012 at 12:37

    @Statsguy
    I’m not sure I follow all of your post, but I agree with parts…

    Your first point about schools. If I understand the implied model, there appears to be a problem of cross-sectional heterogeneity in school performance that would need to get resolved with GLS. I’m not sure a regime-switching model would be needed to explain this effect.

    Consider a bunch of equities, you don’t need a regime-switching model to account for the size effect; you need a factor model.

    A more suitable example is estimating a simple regime-switching model to explain the log changes of the S&P500 index. The regime-switching model would allow the mean and error to both switch between two regimes. The choice of regime will be almost entirely determined by the volatility. Like your school example, during the high-volatility regime you will see both the strongest positive and strongest negative returns. That need not result in serial correlation.

    Assume for the moment, you estimate there’s a 99% probability you’re in the high-volatility regime today and you predict you’ll be 90% in it the next period and 85% in it the period after. If the mean return for each regime are equal, then the only difference would be that you sample from different distributions in the next two periods (the earlier with higher volatility, the second with less). Even if the means are substantially different, it doesn’t necessarily mean that there will be autocorrelation. Just that the distribution you would sample from has a different mean in the different periods.

    It is possible to create regime-switching models that allow for regime-switching autoregression (so the amount of equity market momentum or mean-reversion changes over time). The problem with this tends to be that the number of parameters explode so your estimates are more uncertain. It is also unclear a priori whether the returns of an investment strategy that accounts for regime-switching will exhibit more or less serial correlation than the market. It would depend on portfolio construction.

    I’m with you on effectively reducing the number of data points. No disagreement there, it is a concern.

    Regarding the theoretical reason why they work, it has more to do with identifying periods where there is high vs. low volatility (like Garch models) than necessarily periods of high vs. low returns. Similar to your point about small schools being responsible for the best/worst performance, the most volatile periods tend to have both the best and worst returns. The hedge that would effectively resolve this, unless I’m mistaken, is not static but a dynamic one. If you consider your initial allocation as a 100% investment in the equity index, then the ideal hedge would effectively reduce that allocation during periods where you are in a high regime. My understanding is that it would make it difficult to hedge away.

    As for your final point about overfitting, perhaps then we should all just give up and never do statistics again. 🙂

  21. Gravatar of Cthorm Cthorm
    9. April 2012 at 13:41

    @123

    Certainly not on average! But that’s really just a guess on my part; 2/20 is such a rich compensation model I don’t see how it could, but there isn’t exactly data to back that idea up. But 2/20 isn’t necessarily the universal structure either. In real estate development waterfall profit agreements are more common (i.e. the developer received a higher share of profits if it hits certain IRR thresholds, typically progressing 2/5/10/20 like a marginal tax rate).

  22. Gravatar of Cthorm Cthorm
    9. April 2012 at 13:52

    @123

    Re: monetary policy and hedge funds, I did not mean to imply that monetary policy is a special obstacle for hedge funds. Paulson was short MBS, which was just the right call based on the macro environment. But like any other investment vehicle, hedge funds can have Long positions that would do well if the macro environment was stable, but get swamped by a NGDP shock. Cerberus would have made a killing on the Chrysler deal if NGDP had been stable, but they ended just a little positive (thanks to taxpayers). LBOs, restructurings, venture capital, sweat equity etc. are all difficult to compare returns. Each deal is so unique and private information dominates in such situations.

  23. Gravatar of Disappointing Employment Numbers Send Down Major Indices (MSFT, SNE) – InvestorGuide Disappointing Employment Numbers Send Down Major Indices (MSFT, SNE) - InvestorGuide
    9. April 2012 at 14:00

    […] A critical look at Warren Bufett’s performance over the last five years. […]

  24. Gravatar of StatsGuy StatsGuy
    9. April 2012 at 16:29

    @ John Hall

    At the simplest level, regime switching models identify when a trend begins and ends, or tries to predict when it will begin (or has begun), and when it will end (or has ended). Markov state models can do this, but so can a dozen other techniques. Let’s presume that things such as trends do exist – certainly if they didn’t in the past, the sheer number of algo-driven strategies is creating them (the low volatility climb from Dec to mid-march is a good example). The issue is, there aren’t that many transition points, and there are a lot of possible predictors.

    On top of this, there’s a likelihood that there are systemic “trends of trends”. Let’s imagine that for a period of several years, a certain transition strategy works, then there is a systemic change that causes that strategy to fail (possibly catastrophically). If “trends of trends” persist for periods of a few years, it’s not at all unlikely for someone to have a very successful run of 5 or 10 years following what amounts to a single class of strategies, only to have that strategy catastrophically fail. In fitting the new “regime of regimes”, one now needs to begin with one data point.

    All I’m saying is that any short term evidence has to be taken very skeptically – and even a twenty year run is a lot less statistically robust than a sequence of 20 independent data points. Separately, any successful strategy is doomed to become unsuccessful as soon as it’s discovered by too many people (and there is a lot of incentive to discover things).

  25. Gravatar of StatsGuy StatsGuy
    9. April 2012 at 16:33

    … with regard to overfitting, a good rule of thumb in epidemiology is to look for effects that are significant at a much higher level than minimum, and then try to replicate on a different dataset

    On overfitting in data mining, there’s been some interesting work on random matrix theory trying to identify what sort of “significant” effects one might expect to find in uncorrelated data, to help set a more reliable threshold for significance.

  26. Gravatar of AFG AFG
    9. April 2012 at 16:36

    Scott,

    1.Other people kind of said it, but not clear enough. Buffet is a buy-and-hold value investor, you would expect him to lose money when the market is down.

    –“That fact would only be important if he’d outperformed the market in all 56 years. But he hasn’t.”

    2. This is an absolutely insane statement. Consider:

    a. Anti-EMH does not mean a belief in perfect foresight. It means that you believe that there are some situations in which one can identify deviations from the equilibrium. Just because I can predict housing bubbles, doesn’t mean I can predict oil prices. And the fact that I can’t predict oil prices doesn’t mean I shouldn’t stop a housing bubble.

    b. If I offered you a business opportunity that caused you to lose 50% this, but make up twice your loss in the second year, would you take it? As long as there is a systemic way to predict that EMH is violated.

    c. Forget timing. Let’s bet on a coin flip. Heads you pay me $5, Tails a pay you $1. I will certainly lose in some periods, but it the long run…

    If I, for systemic and non-lucky reasons, devised an investment strategy that minimally underperforms the market in bad years and massively outperforms them in good years, then I have disproven EMH.

    Roubini might have timing, but if you had his personal portfolio you would have beat the market. Read his interviews, from late 2006 when he published the famous prediction article until 2010, he consistently said that his porfolio was 100% cash. He might have missed the rally, but he also missed the even bigger crash. He’s been bullish on the US (although his estimates were far too low) since sometime in 2010.

    d. Timing might be (semi) relevant to investing, but it is mostly irrelevant to the field you typically apply it to (Policy). I may need to know when the Housing bubble will pop to make money from it, but the mere knowledge that it will pop sometime soon is sufficient reason for the government to do X, Y, or Z.

    2. That article is incredibly dishonest about Buffet, they are choosing a random peak to make his losses appear bigger. Berkshire is down since the end of 2007, because he massively out performed the market in the first 2/3 of the year and then regressed to the mean. According to Google News, measuring from November 2, 2007 to today, Berkshire significantly outperformed the market (although was still negative). From November 23, 2007 to today, Berkshire underperformed. Since August 10, 2007 until today, Berkshire is up 9%, while the S&P is down 5%.

    3. Is my arbitrary date equally dishonest? Sort of. If you pick pretty much any date before November 2007 to today, Buffet outperforms. He also outperformed from the market bottom until late 2010, but has been basically flat since then while the market gained 20%. See my point above though, he doesn’t need to outperform in every year.

  27. Gravatar of John Hall John Hall
    9. April 2012 at 16:57

    Random matrix theory is an interesting bit of research, but more in its value as a general approach to dimension reduction and identifying what factors in PCA are significant or not and how to adjust the correlation matrix in response. I don’t use it to generate alpha.

  28. Gravatar of StatsGuy StatsGuy
    9. April 2012 at 17:35

    Actually, there was a string of papers a while back using random matrix theory for portfolio balancing, primarily to help identify relatively independent groupings of stocks for optimal volatility reduction and periodic re-balancing.

  29. Gravatar of It’s even worse than Buffett thinks. « Economics Info It’s even worse than Buffett thinks. « Economics Info
    9. April 2012 at 23:00

    […] Source […]

  30. Gravatar of John Hall John Hall
    10. April 2012 at 04:23

    I’d be interested in reading them. You have the references?

  31. Gravatar of StatsGuy StatsGuy
    10. April 2012 at 05:23

    http://arxiv.org/PS_cache/cond-mat/pdf/0111/0111537v1.pdf

    There’s been other stuff done since. Do a search on random matrix theory optimal portfolio construction.

    I would suspect that it’s precisely this type of work that’s causing some subsectors (like consumer package goods) to show very high PEs relative to the market. Earnings stability is one thing, but there’s additional demand for low capitalization sectors that offer relative independence (or inverse correlation) to larger sectors because of their value in reducing portfolio volatility.

  32. Gravatar of John hall John hall
    10. April 2012 at 07:51

    That paper fits in with my understanding of the application of RMT (focus on key factors, the rest are noise). I view RMT more like a methodology to reduce estimation risk, like other Bayesian methodologies. I tend to leave the portfolio optimization to handle the re-balancing and optimal allocations. An RMT-cleaned correlation matrix is merely an input to the process rather than using a sample correlation matrix.

    I’m not sure I would put RMT as being responsible for the P/Es of a particular sector or industry (since this industry would be correlated with other defensive industries). Models like Barra have been around for a long time and have tried to capture industry effects and correlations.

  33. Gravatar of ssumner ssumner
    10. April 2012 at 09:18

    Jason, But if managers have skill, why would the mutual fund industry be efficient?

    Statsguy. You have it backwards. If markets are efficient then excess returns are not serially correlated. If they are serially correlated, then the odds are less than 1 in a million of 20 good year sin a row.

    John Hall, By anti-EMH models I mean:

    1. Academics can predict market outcomes better than markets (aka old monetarism.)

    2. People should buy managed mutual funds that have been successful.

    3. The government should regulate banks on the basis that low paid government regulators can spot housing bubbles better than investment bankers.

    MF, You said;

    “Successful investors who take advantage of market inefficiencies, if only for a finite period of time, are the falsifications of EMH,”

    Not even close. And that’s precisely the problem, most people don’t have the faintest clue as to what the EMH really says.

    Floccina, That’s possible.

    123, If you stop commenting here I’ll know you struck it rich and are off on a yacht somewhere. Otherwise I’ll assume you can’t beat the market.

    Statsguy, You said;

    “… with regard to overfitting, a good rule of thumb in epidemiology is to look for effects that are significant at a much higher level than minimum, and then try to replicate on a different dataset”

    I completely agree. The 95% confidence criterion is an embarrassment to both economics and science.

    AFG, I completely disagree on Roubini, I have seen one prediction after another miss over the past three years, both for stocks and for economic growth.

    You misunderstood my 56 year comment. I had said there’s a one in 20 chance that Buffett could beat the market 20 straight years. The commenter replied that I should have used 56 years, not 20. I never claimed he should have been able to beat the market for 56 years, that would be crazy. I said his criticism of me was completely wrong, because it was based on the premise that he had beaten the market for 56 straight years, which isn’t true. What is the actual number? I don’t know, so I can’t compute the odds. Was it 46 out of 56 years? 42 out of 56 years? Without knowing we have nothing to work with.

    I agree that Buffett’s performance depends on the period examined, indeed that’s exactly my point. It’s clearly possible to find three year periods where he underperformed. I’m not claiming he’s done a poor job overall, but rather we have no evidence that his performance is likely to keep beating the markets.

  34. Gravatar of 123 123
    10. April 2012 at 09:40

    Scott,
    serial correlations are meaningless:
    http://soberlook.com/2012/04/performance-persistence-in-hedge-funds.html

    “If you stop commenting here I’ll know you struck it rich and are off on a yacht somewhere. Otherwise I’ll assume you can’t beat the market.”

    🙂

    The superb performance of my portfolio allows me to work less and comment more.

  35. Gravatar of Jim Glass Jim Glass
    10. April 2012 at 13:58

    .. for every 1,000,000 investors you’d expect one guy to outperform the market for 20 consecutive years. The masses would call that lucky guy a genius, even if he was just an average bloke from Nebraska.

    Buffett is the wrong guy to pick for this example. Much better is Bill Miller who beat the market picking stocks 15 straight years, becoming a super-hero fund manager — then lost it all in a crash and wound up exactly even with the S&P over all that time.

    Buffett isn’t a stock picker, he runs an operating business (highly tax-sheltered, insurance) and uses his leverage to make deals on terms not available to anyone else, see how he cleaned up on the deal he made with Goldman. That was not available to you and me and the mutual funds we might invest in.

    The most impressive stat to me about investing through stock-picking funds is that during the long-term secular bull market the average fund investor *lost* money (or barely broke even, depending on the time period and data used) while the funds went up, up, up.

    That’s because so many fund investors chase the market, market-timing with a lag. After the market has been going up they get in, then after it tops they ride down, then when it is down they sell and don’t get in again until after they miss a good part of the next rise. They notoriously do the same thing in sector funds, jumping into sectors after they get hot, then after they cool jumping to the next hot ones, etc., buying high and selling low. So as the market goes up investors manage to go down.

    Buffett has said people should invest as if they can make only 20 purchases/sales in their entire lives to avoid this. Commodore Vanderbilt, our first true self-made billionaire, and the great buy-and-hold value investor of the Wild Wild West era of US capitalism, said: Buy only what you actually want to own at the price. You’ll always be happy with what you own, and if it goes up in value and makes you rich so much the better.

  36. Gravatar of StatsGuy StatsGuy
    10. April 2012 at 17:19

    “If markets are efficient then excess returns are not serially correlated.”

    No, markets can be efficient and excess returns can still be highly serially correlated in the short term. Consider:

    If my investment strategy in Y2000 was “buy gold”, I would have outperformed substantially over a long time period. Was the market for gold inefficient? Not necessarily – merely because too few people recognized the likely increase in value early enough does not mean the markets didn’t process information properly. It just means that the markets failed to recognize a dynamic which was persistent, and a stategy that consistently bet on gold would have won… for 10 years.

  37. Gravatar of Mark A. Sadowski Mark A. Sadowski
    10. April 2012 at 17:43

    I don’t have investments. I have property. I enjoy my property. If you have investments you’re always dreaming about what you will buy with them when they pan out. I already have most of what I want.

    That’s why I call it property.

  38. Gravatar of Hal Hal
    10. April 2012 at 19:05

    If market is so efficient, why do active managers get paid? Isn’t that a conflict or a paradox? In other words, wouldn’t the “efficient” market mostly wipe out fees predicated on inefficiency?

  39. Gravatar of Major_Freedom Major_Freedom
    10. April 2012 at 20:15

    ssumner:

    “Successful investors who take advantage of market inefficiencies, if only for a finite period of time, are the falsifications of EMH,”

    Not even close. And that’s precisely the problem, most people don’t have the faintest clue as to what the EMH really says.

    Hahaha, the truest irony of all ironies. How typical is it that those who adhere to the EMH don’t have even the remotest understanding of it? This is a joke.

    The EMH, if you actually understood it, which you clearly don’t, is very clear. It holds that an individual cannot consistently achieve returns in excess of the market (risk-adjusted), given the available information at the time.

    Well, Warren Buffet achieved above average risk adjusted returns for 20 years, which is “consistently” earning above average returns, and, as I said, it is a “finite period of time.”

    Warren Buffet, along with many other investors, refute the EMH.

    The main problem with those who adhere to EMH is that they almost universally fail to grasp that the EMH is a way of viewing the world only; it is not an empirical based doctrine. It leads to viewing 20 year periods of achieving higher than average returns “luck”.

  40. Gravatar of Ritwik Ritwik
    10. April 2012 at 23:54

    Scott

    Buffett has handled the ‘one lucky guy out of 100,000’ argument amply in his biography (with Lowenstein) when he showed the sucesses of Munger and his other buddies from the Ben Graham school. Why don’t we compute the probability of that? Samuelson, when confronted with this argument, basically ended up accusing Buffett of insider trading.

    One key aspect that keeps getting ignored in all EMH related discussions is – ‘how replicable is your performance?’. Academic skill probably follows the bell curve. One in a 100,000 might score a 5/5 in MIT(or something comparable, I’m not quite up to date with all the grade inflation). Yet nobody would say – ‘random walk, lucky guy, doesn’t prove anything’ etc. Everyone will believe that with the right level of intelligence and effort, this success is replicable. For some reason that I’ve nerve understood people completely diss this structural explanation when analysing stock market returns of investors. Buffett read a 1000 balance sheets a year in his prime. How many have done that?

  41. Gravatar of Majorajam Majorajam
    11. April 2012 at 19:41

    Yipe. I presume you realize that Berkshire looks nothing at all like the stock market??… I presume, as it appears you are advising people on investing (gulp), you know there are factors that have to be controlled for when making a comparison like you’ve made here? I mean, at the very least you get that where ownership shares are traded matters less than what those shares actually own an interest in… right Scott?

    Wow this is some ignorant stuff. Buffet as proof of EMH. That’s probably not where I’d choose to run for the unenviable task of trying to support EMH, and not even because it’s one data point with utterly arbitrary and cherry picked start and end dates, but then I’m not quite as cavalier with the facts as you.

    No doubt it would interest you to know that GMO has had the best record in the business at timing markets, as per the extraordinary decades long performance of their asset allocation models. But then Grantham like Buffet and most professionals that have compiled similar records wouldn’t be caught dead making decisions based on comically obtuse Bethamite calculations formulated based on rational expectations.

    There’s a correlation between these capabilities and those beliefs, and it’s not incidental.

  42. Gravatar of ssumner ssumner
    12. April 2012 at 11:30

    123, I’d need to see more research on that. Might it reflect different risk strategies? If it’s legit, why wouldn’t everyone put all their money in the successful funds?

    Jim Glass, Good points.

    Statsguy, You are data mining–it’s always possible to find patterns looking at past data.

    Mark, Good point.

    Hal, Because labor markets are less efficient than financial markets.

    MF, To see what’s wrong with your claim, apply it to a roulette wheel in Vegas. Could someone get 20 black numbers in a row? And when they did, exactly what would that prove about the “efficiency” of the wheel?

    Ritwik, I’ve never denied that Buffett might be smarter than the market.

    Majorajam, You said:

    “Wow this is some ignorant stuff. Buffet as proof of EMH.”

    Yup, that would be ignorant. When did I say Buffett proved the EMH? I responded to others who said Buffett proves the EMH is wrong. But he doesn’t.

  43. Gravatar of Hal Hal
    12. April 2012 at 16:45

    Sophisticated capital sources are spending untold billions every year on active management. Nearly all these fees could be eliminated with indexing. Capital sources have a choice. Endowments, pension funds and other fiduciaries relying on their sophisticated consultants select active management. As someone who believes in markets, what does this tell you? It’s not about sticky labor costs. You gotta do better than that. If your argument is that they are all fools, then that would be mother of all inefficiencies, hardly an anomaly. To me it’s a paradox. If the market is efficient, then participants would not throw away money for active management fees. But they do. Which implies the market is not efficient. Such an argument does not suggest the nature of the inefficiency nor any method to exploit it.

  44. Gravatar of 123 123
    13. April 2012 at 07:47

    “I’d need to see more research on that. Might it reflect different risk strategies? If it’s legit, why wouldn’t everyone put all their money in the successful funds?”

    1. They are putting all their money.
    2. Yes, this is all about the risk strategies. See A.Lo and Capital Decimation Partners:

    http://krugman.blogs.nytimes.com/2008/03/16/capital-decimation-partners/

  45. Gravatar of Greg Ransom Greg Ransom
    13. April 2012 at 09:12

    Scott,

    Your “EMH” is whatever you want it to be — it’s a moving vapor.

    I.e. it isn’t grounded in any fixed form, it isn’t defined, and becomes anything you want it to be from one post to another.

    And you use this moving vapor to do all sorts of different work it just can’t do.

    Try making your arguments without appeal to the magic phrase “EMH” and lets see if there is anything left to your “arguments” or causal stories …

  46. Gravatar of Greg Ransom Greg Ransom
    13. April 2012 at 09:15

    Fama identifies at least THREE different things as the “EMF”, i.e. three cognitively very different things which have been labeled the “EMF”.

    Which one are you talking about?

  47. Gravatar of Major_Freedom Major_Freedom
    18. April 2012 at 06:02

    ssumner:

    MF, To see what’s wrong with your claim, apply it to a roulette wheel in Vegas. Could someone get 20 black numbers in a row? And when they did, exactly what would that prove about the “efficiency” of the wheel?

    To see what’s wrong with your claim, apply it to a market. Could someone earn profits 20 years in a row through luck? And when they did, exactly what would that prove about the “efficiency” of the market?

    Sumner, your view of the market is a priori. You insist that making returns in the market is a product of luck, whereas I insist that making returns in the market is a product of skill and foresight.

    You can use analogies of roulette wheels all you want, and I can use analogies of markets all I want.

    Let’s turn your scenario around. Suppose someone did win at roulette 20 times in a row. But now suppose this: Suppose they won 20 times in a row because they learned of an inefficiency in the roulette wheel. So they exploited that inefficiency to earn consistent gains concomitant with their wealth endowments. But then the casino learned of this inefficiency in the roulette wheel, so they repaired that inefficiency. Our man no longer makes gains playing roulette because his knowledge is now dated.

    But then someone else learns of a different inefficiency in the game. This new person then consistently exploits the inefficiency to make gains, until the casino again learns of this inefficiency and corrects it.

    And so on.

    Here’s the moral of the story: This is exactly what happens in casinos. Casino owners are constantly dealing with gamblers who exploit inefficiencies is casino games. From counting cards, to signals, to other inefficiencies that more closely resemble real life, than the sterile pure and perfect EMH model you have in your mind that a priori excludes even the possibility of existing inefficiencies.

    OF COURSE if your worldview STARTS with the assumption that inefficiencies don’t exist, as in using the analogy of a perfectly efficient casino game (which of course does not exist), then you will never understand pro-inefficient market arguments.

Leave a Reply