Roll Up, Roll Up... Warren Buffett Never Had An Edge...

5th Oct '12 by Edward Croft

22 comments 22605 reads

At times the world of modern finance resembles a standoff at the OK Corral. On the one side stand aloof the esteemed fund managers of the era with their long records of outperformance. On the other the academics and quants who wish to shoot them down by 'proving' that their excess profits can be explained away as either lucky or systematically repeatable. In their sights has always been the biggest scalp of all, Warren Buffett, and where many have failed a recent research paper titled 'Buffett's Alpha' looks to have finally shot him down. Essentially the paper finds that Buffett may not have had an 'edge' after all.

The esteemed researchers looked at Buffett's investment vehicle, Berkshire Hathaway, and broke out the performance related solely to publicly traded companies. They then ran some mathematical regressions over the data to figure out what 'factors' could explain the portfolio's outperformance.

All geared up... Firstly, they found that Buffett's equity portfolio performance benefited from his ability to access cheap borrowed money. Basically he was investing $160 into stocks for every $100 he owned. How? Well Berkshire always had a great credit rating which helped, but it also owned insurance businesses which benefit by taking in premiums from customers often years before having to pay out claims. Effectively this 'float' acts as an ongoing interest free loan which Buffett can then invest back into stocks.

...and boring... But what of his stock picking prowess? Buffett has made 19% annually for a zillion years and the market plus the leverage could only explain 10% of that - where did the rest come from? Well it's very simple.

There are a few standard factors that academics use to explain the majority of stock market outperformance. The ones normally mentioned are the following: small caps beat big caps (size), cheap stocks beat expensive stocks (value) and recently rising stocks beat recently falling (momentum). But they found that while Buffett did have a preference for cheap stocks, he had little exposure to small caps or momentum. Previous studies had found the same, suggesting that Buffett's extra profits were due to some unexplainable 'alpha'.

But this set of quants decided to try out a couple of more modern factors to see if they could explain his performance: the propensity for lower volatility stocks to beat higher volatility stocks (i.e. safe beats exciting ) and the tendency for…

Unlock with Email

Unlock with Google

Already have an account?
Login here

Edward Croft

5th Oct '12

PS - re. Mechanical approaches not working in the long run... as far as I can tell index trackers have outperformed 90%+ funds over the long term. By definition they are mechanical approaches!

marben100

Fair point! But, by definition, they can't outperform the market. ;0) [and charges mean a small but persistent underperformance].

You haven't responded to or explained the excerpt I quoted from the paper, which makes it clear that their portfolios are based on hindsight data analysis - and hence all they do is show that Buffett's methodology, cheap financing and judicious gearing work... which is something we kinda knew!

I don't find the paper exactly clear - but then I'm not a quant.

macd2510

Having just finished reading 'the snowball' I can only quote the great man himself

Investors should be skeptical of history-based models. Constructed by a nerdy-sounding priesthood using esoteric terms such as beta, gamma, sigma and the like, these models tend to look impressive. Too often, though, investors forget to examine the assumptions behind the symbols. Our advice: Beware of geeks bearing formulas.

Since the simulated Buffett-style portfolios do not account for transaction costs and other costs and benefit from hindsight, their apparent outperformance should be discounted.

N.B. they state their outperformance should be discounted NOT their performance.

And of course these studies are based on hindsight - we can only build models using data that exists. Don't you a invest in 'undervalued' companies because in your experience they have tended to be revalued? Isn't that also hindsight bias? Where's the proof that that will occur again in future? It's really just a form of hope - even Ben Graham couldn't answer that question.

PS - and on another note... in my opinion Index trackers do outperform the market - depending on how you define the market. The 'market' is something that people can buy and sell... that's NOT the FTSE 100 or equivalent index... it's got to be the average 'marketed' investment fund.

Ha! Great quote there macd... but to be fair that quote was from the 2008 Berkshire Annual Letter and was in the context of the mortgage backed CDO disaster.

'Indeed, the stupefying losses in mortgage-related securities came in large part because of flawed, history-based models used by salesmen, rating agencies and investors.

I don't think that Warren Buffett would say that investing in safe, cheap, high quality stocks (on the basis that his own historical record suggests it works) would be a bad thing.

toddwenning

Hi Ed,

This was indeed a thought-provoking study. Thanks for sharing it, as well as your thoughts on it.

In a nutshell (and correct me if I'm wrong), the argument is that Buffett's performance can be explained by a formula and therefore can be replicated by purchasing cheap/quality/safe stocks and levering up 1.6x. This, then, opens the door to ETFs that can mirror the strategy, and so on.

Here's a problem with that -- Buffett still had to be correct most of the time or the leverage would have crippled his operations a long time ago. Even quality/cheap/safe shares have down years (the US small cap value index lost 32% in 2008), but looking at the 2011 BRK annual report, BRK book value wasn't negative year-over-year once between 1965-2001. Had WB even posted two bad years in a row, the leverage would have surely impaired long-term performance.

My point is this -- leverage cuts both ways and Buffett had to be consistently correct for the leverage to consistently work in his favour. That, I would argue, requires a certain measure of learned skill that probably can't be mirrored by a formula.

The other problem with creating an tracker for this strategy, assuming it was indeed successful, would be managing the increasing cash inflows to deliver sustained out-performance.

WB admitted as BRK got bigger, for example, that it would become more difficult to deliver the high rates of return seen in the early BRK years. The more cash you have to invest, the more difficult it is to move the needle. You can't keep buying the deep value small cap shares without moving the market for those shares; as such, you're forced to invest in larger companies whose performance tends to track the market.

Thanks again for starting this conversation!

Best,

Todd

jpblaw

The ongoing and longstanding debate about the secret of Buffett's outperformance is going to generate copious if not endless amounts of articles and discussions. It's definitely interesting to have focused on the publicly traded stocks part of Berkshire's portfolio but I suppose my first question is how you would identify those beyond the periodic regulatory filings. Got to agree as well that the float is a mixed blessing, as obviously leverage amplifies losses as well as gains, where its effects can be far more serious. The whole idea of trying to analyze the portfolio and create a replicable model is an interesting one but seems to be an exercise in futility doomed to failure due to the aforementioned lack of the necessary information. Certainly the effect of the float on amplifying his investment returns is an interesting question, as is his preferential access to exceptional investment opportunities as exemplified by his investments during the financial crisis (eg GE or GS). There can be no doubt Buffett is an exceptional investor but trying to gain some sort of advantage or insight through this sort of quantitative analysis strikes me as missing the point. For most investors it's what he says and the advice he gives, particularly in his annual letter, that is most useful. Also to be noteworthy is the lack of any kind of (out)performance of his vehicle (Berkshire) over recent years, especially as it's got so much larger and increasingly like a hedge fund.

Mark Simpson

That you can generate excess return by building expected-return-factor models is hardly new - Robert Haugen for one has been publishing books on the subject since the mid-90s. And as Haugen shows volatility, rather than being the only source of excess return as the EM theorists would believe, gives negative excess return in reality!

What Buffet does and in fact all of us are doing is the same but informally without the precise model - long term horizon investors like Buffet buy cheap, quality companies that have low volatility because these are the high coefficient factors for long term excess return.

So in my mind he's still a genius investor because he was one of the first to realise how owning insurance companies gives you cheap leverage via the float and picks outperforming companies without a formal expected-return-factor model. Which is cleverer doing it with the stats or in your brain? - neither in my mind they are different skills that generate the same outcome.

He has also remained an effective investor even as his portfolio size grows because the corporate culture he develops means that when he buys whole companies the owners and managers love to stay with the business and run it for him! I'm not sure many of us could do that.

Ed, any chance we could run our own expected-return-factors models on the stockopedia dataset? Now that would be a powerful PI feature to compete with the hedgies :-)

dangersimpson - now you read my mind! I've read several of Haugen's most famous research papers and admired the performance of his strategies. I'm still mulling over which of his books to read first, so if you recommend one please let me know - I think the ones on my list are Inefficient Markets and the New Finance.

I am absolutely determined to build exactly the features you describe, but certainly it's going to take us some time to get there. We are still working on the stability and scalability of the data set rather than a lot of the more 'meta' features layered on top, but given time and investment the product plan is fairly set in our minds. I won't give it all away here but I'm sure that you can put two and two together.

The critical thing for us is how to make complicated things simple, understandable, usable and their output investable for a wider audience. It's something we labour over here. It does sound like you have a lot of interest in the area, so if you want to share any ideas about what's most important please do let me know.

Hi Ed,

I think 'The Inefficient Stock Market' is probably the better read for how expected-risk-factor & expected-return-factor models are built and why certain factors might be persistent anomalies in the market e.g. how most managers are rewarded with a bonus means they may prefer a risky portfolio to a non-risky one at the same expected return.

The New Finance is more concerned with the evidence that various anomalies exist and therefore that the market is not efficient - I think quite a bit of the evidence you will have seen before in things like Dremen Contrarian Investment Strategies though it is always good to have a reminder. Both books are quite slim and are organised in short chapters with not much pre-amble. I think they are quite expensive new but I picked mine up pretty cheap on amazon market place when someone had one on cheap.

In terms of using the stockopedia dataset I think it would be relatively easy to run the factors that Haugen comes up with for the UK. If my reading of The Inefficient Market is correct you'd take the 12 highest impact factors given, like Earnings to Price, 1 Month relative return etc., work out the mean for all companies and then how many standard deviations each company is away from the mean (call it sigma) and then do a sum product of those sigmas with the factors given in the book to get a predicted excess return for that stock over the next (iirc) 3 months. Then rank all the stocks by expected return and give a decile ranking like you do for the magic formula.

The problems I see with doing that though are:

1. 3 months might be too short a horizon for most PI's - and if you did longer term studies I'd expect valuation factors like P/B to be more significant and momentum factors less.

2. The data he gives is from a study 1985-1993 and the factors driving the market may be different.

There may be other studies out there that are different time frames or more up to date data which could be used - I also think Haugen may sell the output of his models with up to date data but I would expect they wouldn't be cheap.

At least you could have the test portfolio of the long top decile and short bottom decile to see how well it actually works in practice before putting forward as a serious ranking to consider.

What would be really powerful is to do our own regression analysis on the stockopedia data to come up with our own model of the significant factors. However this would be a pretty advanced feature to be able to do this and would require a lot of historical data not just the ratios now. What essentially you'd have to do is record the stock prices and potential factors such as EV/EBITDA, P/S, 1 Month RS (basically all of the major screening ratios on stockopedia) periodically - say every month. Do that for a reasonable period of time and you have enough data to run a multiple regression to calculate the excess return factors over a given time period. Say we pick 3 months like Haugen. You'd look at the regression between each factor, say the EV/EBITDA, of all stocks at the end of all months you have data for and their relative performance over the following 3 months. [actually I think if I understand correctly you may do this regression to the sigmas (standard deviations from the mean) of the factors and the performance but can't remember exactly why.] This tells you if that factor is significant or not and what factor to use in your model.

Therefore even if you are nowhere near being able to implement something like this you may want to start taking a snap-shot of the data in the database each month. Then if you come to try and do this you'd have historical price and ratio data to work with.

I think if you are really clever when you have your expected return and expect risk value for each stock you can use Markovitz portfolio optimisation to build the lowest risk highest return portfolio form all available stocks. Interestingly Haugen seems a fan of Markowitz even if he disagrees with a lot of the EM theories. I guess he believes that there is an efficient frontier just that due to these anomalies the market portfolio is not on it!

Anyway, not sure any of that helps make things simple and understandable for the average PI [I think the only hope here would be to do what you did for greenblatt and have like a magic formula rank. A Haugen rank as it were.] but good to have someone else interested in this stuff! Let me know how you find The Inefficient Stock Market.

Cheers,

Danger

UK Value Investor

6th Oct '12

Hi Ed. I agree with everything in the article and the related research, because it's more or less what I do myself.

However, I would say that just because Buffett is replicable doesn't mean he isn't a genius.

Virtually everything a person can do is replicable by other people. I like F1 so I'll use Schumacher as an example. He won more races and more championships than anybody else over the longest career ever. He's a genius.

Why? Because he took fitness to new levels; he took new driving techniques like left-foot braking to new levels; he made sure his team mate couldn't compete with him (often contractually), he worked longer with the engineers than anybody else.

These things are all replicable, and they have been replicated which is why his 'edge' was gradually eroded (plus he got old).

Does that mean he wasn't a genius for doing it 10 years before anybody else and taking it further than anybody else? No. And the same applies to Buffett. He put the approach together and stuck to it for decades. Just because other people can reverse engineer what he's done doesn't mean he wasn't a genius for doing it!

However, research like this gives other investors the right mental framework for stock picking (safe, quality, cheap), so thanks for highlighting it.

John

Hi Ed, yes agree, I did apply the quote in a un-related context. What I felt however, is that's all to easy to create a 'look back' model that produces great resutls. Buffet's brilliance is to avoid trouble before it happens (Be fearfull when others are greedy) and to incorporate qualtiative factors (the competitive moat, the quality of management... 'the ham sandwich test' etc..)

BTW - I'm using a similar 'Total Returns' strategy myself - a long only, large cap, low beta (Read 'boring') stocks with a blend of value / income / quality strategies to pick individual stocks.

Love the Guru Strategies on Stockopedia... it''s nice to see so many of my picks on those lists..
Also loving the Dividend Investing Arcticles on Seeking Alpha (they've fundamentally changed my physcological approach to investing)

Thanks for the kind comments! One of the things I love about the screener is that you can also see the results from the bottom up - i.e. from an individual Stock Report - it's a big time saver.

On dividends... we are just putting the finishing touches to our e-book. It's definitely a bit different to the typical dividend book - it's built from a lot of our recent articles on dividends heavily edited into a single and I hope very readable document. We are hoping to publish it next week - if you want to get an email when it's finished the link is here (please excuse the flagrant marketing copy on the landing page!) http://www.stockopedia.com/courses/how-to-make-money-in-dividend-stocks/

thebuffoon

7th Oct '12

If anyone thinks they can automate what Buffett has been doing for decades, then they are seriously misguided.

Sure, he took over a textile company, not knowing the inherently bad fundamentals of the business. He did learn quite quickly though, and worked out that by using it as a holding company for other assets, he could benefit from the far superior fundamentals of an insurance company, where you take the money up front giving you an interest free float. For sure that supercharged his returns. No surprise there.

Prior to this he had a private partnership, producing fantastic returns. He still asserts that with smaller funds he could guarantee a CAGR of 50% for quite some time.

You can talk about data, algorithms and screens until the proverbial cows come home. The easy bit is identifying what to invest in. The hard bit is sitting on your hands.

Buffy

Isaac

He did learn quite quickly though, and worked out that by using it as a holding company for other assets, he could benefit from the far superior fundamentals of an insurance company, where you take the money up front giving you an interest free float

Buffy

Do you have most of your money in Insurance companies?

Murakami

Another interesting take on the paper from Monevator: http://monevator.com/weekend-reading-unpicking-on-warren-buffetts-alpha/ (see the comments below the blog too)

Buffett himself was a living, breathing businessman who several times changed course as an investor over his six-decade long career. He went from Ben Graham ‘cigar butts’ to private companies to blue chip brands to buying mega-railroads. Only this year he said he’d rather be buying residential real estate!

So I don’t think a mechanical ‘robo-Buffett’ would have a hope in hell of predicting what a real Buffett would do if he were to live to 140 and invest for another 60 years...

Unpicking Buffett’s edge as an academic activity is one thing – leaping to the conclusion that you can build and buy a Buffett ETF is quite another.

A few points on that quote. The study under discussion actually stripped out Buffett's private investments and he himself said he had to learn the hard way that cigar butt investing didn't work. But the blue chip brands and mega railroads are exactly the kind of safe, cheap, quality stocks that the above algorithm might highlight.

It's good to debate these issues, but I must say I did find Monevator's statement "if you believe in Buffett, the most logical thing is to ... buy shares in his company Berkshire Hathaway" quite illogical.

Aside from the statement making no consideration of value, Berkshire's stock has performed moderately of late and doesn't look like it can return to former glories. It has only returned an annualised 6.3% over the last decade. It has a market cap of $224bn and will struggle to ever compound returns at a high rate. Warren Buffett is also now 82 years old and recently underwent prostate cancer treatment bringing succession worries. There are questions of cultural issues at the top of the firm too - his previously favoured successor was given the boot for making trades based on inside information.

Given these constraints, if I had a choice as to whether to invest in Berkshire Hathaway today or a low cost strategy that focused systematically on cheap, high quality, low beta, dividend paying mid/big cap global stocks I know which I'd choose. Each to their own I guess.

Buffy
Do you have most of your money in Insurance companies?

Do you want to give me the punchline to that very strange question young 'un?

13th Dec '12

This piece on Buffett has been published in this quarter's edition of the London Stock Exchange's Private Investor Magazine.

See: http://www.londonstockexchange.com/traders-and-brokers/private-investors/magazine/magazine.htm

Fangorn

14th Dec '12

I presume Isaac asked you that question Buffy because Buffet has built Berkshire Hathaway on the back of cash flows from GenRE to purchase other cashflow positive businesses.

Roll up, Roll up... Warren Buffett never had an edge...

Unlock the rest of this article with a 14 day trial

About the Author

Edward Croft

22 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...