Assorted links


It is always amusing that economists prefer things than are easy to model rather than things that can make money. Not sure what that says about them.

I will review all links, give substantive reviews, and STILL be (close to) first! How many can do that? Few.

1) - simply the observation that the flow of fluid around an obstacle--be it humans, ants, whatever--creates a "vacuum" behind the obstacle, and this creates less chance of clumping of humans. Rather neat and probably patentable, but unlikely to be enforceable sadly.

2) Another example of behavioral economics: a fixed change in income should not alter perceptions of inequality but it does

3) Momo investing "myths" but there's an error in the paper: see myth #9. This paper is simply backwards looking "data mining", where patterns can be found ("Myth #9: Different measures of momentum can give different results over a given period! OK, this isn’t a myth, it’s actually true, but it’s tritely obvious and yet still often hurled as a critique of momentum so we’ve chosen to include it. The myth, more properly stated, would say “different measures of momentum give different results over a given time period and that’s a terrible thing.” )

4) medical pot shops rely on cash, which increases crime, and is an argument of why banks are useful

5) Title says it all. Comedy in China is probably something lost in translation.

6) Id.

We're waiting for you to start your own blog so we can skip this one. You could be the next Ezra Klein.

We all promise to be faithful readers too. :)

I love you.

By the way, you now understand the articles better than their linker.

I feel cheated on 6. Have to do better for the big commenting $$$.

1. Ray is right. This is a mathematical, not a behavioral optimum.

Ants follow chemical trails of other ants. Humans engage in psychological herding - two entirely different mechanisms producing the same result.

If there seems to be no connection between X and Y that behave nearly identically, then the likely cause is Z that exogenously drives both X and Y. As Ray said, a dumb fluid will react the same way.

I have to give it more thought, but I think many venues are in fact designed for optimal throughput though perhaps unintentionally. Im thinking theaters that have doors in the corners and large barriers blocking the two main exits to the rear. Many modern sports stadiums restrict traffic to only left and right.

#3. It's right up there with all the other technical strategies for 'investing'?

Nope. Maybe you should read the paper.

Maybe. I skimmed it. Sounds like a bunch of crap. Suppose you know a stock price and have transparent financials. If I understand momentum theory, the stock price one year ago is another key piece of data in making an investment decision in such a company, but I fail to see the relevance of this data.

Since this seems to be the place for #3 comments, I will shed some light on the paper.

The paper found patterns in backtesting of data that suggests there is such a thing as "momentum trading" and you can make money off it. One problem (as I hinted upstream): you can only make money off it "in hindsight". Since random data will ALWAYS have patterns in it (long and short term). Hence the paper is indeed "a bunch of crap" but it's not for the usual reasons as the authors have bad data. Instead, they simply are seeing patterns of trade-worthy significance where in fact none exist. And their colloquial tone merely makes their smugness more insufferable. They are fooled by randomness.

A real world example: the "January Effect" was noticed by a bunch of finance professors, who tried to trade on it, and it disappeared about the time they acted. Case in point, I rest my case.

and yet backtesting is OK with climate models, which is dealing with a system that is perhaps 100x the complexity of the stock market.

Correct again, Ray.

"One problem (as I hinted upstream): you can only make money off it “in hindsight”.'

Says you. Three of the authors are at an investment firm that apparently employs this strategy, and seems to make money.

'A real world example: the “January Effect” was noticed by a bunch of finance professors, who tried to trade on it, and it disappeared about the time they acted. Case in point, I rest my case.'

Once some academics were wrong about a way to make money investing. Therefore, all academics are always wrong about ways to make money investing?!

Gene, 'seems to make money'. If these guys want to publish actual results (maybe they have), I'd be prepared to keep an open mind on this, but the paper is just theorizing.

The January Effect wasn't wrong- it's just that when the word got out, the effect disappeared. If momentum investing provides a free lunch edge, I would expect that edge to be eroded over time as well.


Momentum in returns was first discovered in a published paper in 1993 and has been stable OUT OF SAMPLE for 20 years. The Sharpe ratio on momentum strategies are strikingly high over this out of sample period. And momentum has been discovered in several markets beyond the equity market, which is additional out of sample validation.

So, no, it is not the paper that is full of crap, and it is not the authors who are insufferably smug. But somebody certainly is.

"If I understand momentum theory, the stock price one year ago is another key piece of data in making an investment decision in such a company, but I fail to see the relevance of this data."

Well, if momentum theory is right, it is relevant. Saying it can't be relevant is just another way of saying you don't believe the theory, and not an independent disproof of the theory.

I understand that, but if something is relevant, I'd like to hear an explanation of why. Imagine two companies with similar financials and stock prices of $50 a share. Company A was a darling that has 'lost favor'- a year ago, it's stock traded at $100 per share. Company B was unloved- a year ago it traded at $25 per share.

Momentum investing tells us that Company B is the stock you want, based solely on the 'lookback' price. I gotta tell you, this sounds pretty much exactly like the rear view mirror method of investing.

I know the authors distinguish what they're doing from a simple "the trend is your friend" approach, but without some theoretical basis for understanding why Company B is the better bet, it sounds spurious to me.

Examples abound. Microsoft was an awesome momentum investment, until it wasn't. When does the momentum end? How do you know?


The correct way to understand momentum is to think of security markets as large pools of participants, many with private information. If you could know everyone's private information, in addition to the standard public information, you would be able to outperform the market. Stock prices change in reaction to fluctuations in supply and demand. Some of this is due to participants with no private information, uninformed noise traders. But some of it is due to informed traders monetizing their edge. Uninformed traders are equally likely to trade in either direction, whereas informed traders are likely to be trading in the same direction as each other (if they didn't then they wouldn't have any edge). Since uninformed traders are trading randomly their trading has no persistence over time. However those acting on underlying information will likely keep trading in the same direction over time to build up a position.

At a large enough time frame, the noise of the uninformed traders tends to cancel out. But at a short enough time frame the informed traders are mostly going to keep trading in the same direction. Thus if you at the medium-run past returns of stocks, those that have risen probably on net balance have overall bullish private information. And vice versa for those that have fallen. Tilting your portfolio to say the 12 month minus 1 month past returns, is going to expose you to positive private information.

Momentum investing works because the market is not an autonomous computing machine that instantly reprices all securities immediately based upon new information. It is a market filled with now hundreds of millions of people who are prone to emotion and herd behavior.

Bear markets are incredibly painful and short-term. People dump their shares and in the really bad bear markets, they also forswear ever investing again.

Bull markets tend to be the opposite, they take time to unfold as investors need a reason to get back into the market. The biggest bull markets and biggest tops are made when the last buyers enter and there's no one left to buy. The bear market unfolds and the cycle repeats.

Momentum investing is just this in a microcosm at the sector or stock level. It works and the gains are very large over the course of a bull market. The key is getting out, since when the losses come, they are big because you are eventually herding with all the other momentum investors/traders in the market and also most of the suckers. The suckers one need not worry about, but they do buy late and dump late too, increasing volatility.

I would say it is riskier than something like value investing because it involves timing, but even with value investing there are "value traps."


I don't know how momentum works. But it was noticed and published over 20 years ago, and has continued to significantly outperform until now. This is in contrast to most other anomalies you can name. One of the authors of the linked paper, Cliff Asness, did his dissertation on this 20 years ago working under Eugene Fama, who is not known to be a fan of "technical analysis."

Maybe it will end tomorrow, I don't know. It certainly seems that the more people that believe in it the more likely it is to be arbitraged away eventually. But to lump it with the January effect or various technical signals is just ignorance on your part.

@Doug, not sure I grok completely, but this sounds more like a trading strategy than an investing strategy, based on the timeframes involved.

@8, you reinforce my impression of this as a trading strategy when you say stuff like: "The key is getting out,". Also, it sounds like you're saying that bear markets overshoot while bull markets are a bit late getting the news. If markets are late getting the news, I suppose that is some basis for momentum. My own, simplistic view is that investors are prone to overshoot in both directions, in which case a contrarian investment stance, which is basically the opposite of momentum, would seem to be the right call. I have also seen arguments (not sure of pedigree) of simple contrarian strategies that have empirically outperformed (e.g. buy the 10-worst performers in the S&P from last year.)

@ed, no need to be a dick, although you are certainly within your rights. Maybe semantics, but I thought ANY investing strategy predicated on stock price movements (including contrarianism) is by definition 'technical.'

6) Will be interesting to see more data on this over time. I long thought that legalized marijuana would save thousands of lives a year in reduction of alcohol abuse alone. (of course not counting the lives saved from prison and organized crime) This view was originally based on knowing plenty of alcoholics and marijuana users first hand, (growing up in an economically deprived rust belt town) and supported later by studies on the matter.

" the lives saved from prison"

Prisoners are the necessary raw material of the prison industry. If people who do things that other people don't like aren't incarcerated what will happen to the people that design and build prisons and their fixtures, the administration and guards, the parole officers, the attorneys that later become the judges, the cops that make the initial arrest, the legislators that enact the laws and so on? We need more prisoners, not fewer.

"knowing plenty of alcoholics and marijuana users first hand, (growing up in an economically deprived rust belt town)"

So there's a correlation between substance abuse and economic deprivation due to ferric oxygenation? That must mean that there's much less, if any, substance abuse in Manhattan or Silicon Valley or Beverly Hills, right?

Without reading the article, I'm not even sure its clear that the potheads are necessarily the ones binging on booze. A few possibilities:

1. Medical marijuana helps improve quality of life so that people can do more of what they want to do, including more drinking.
2. Medical marijuana may or many not improve quality of life, but it impairs judgment and encourages less healthy choices.
3. Medical marijuana increases socialization opportunities for the recipient, as well as friends and family, some of whom will drink too much.

This binge drinking connection makes no sense to me. Potheads I've known don't drink while smoking pot. They say it causes "the spins" which I think means either nausea or dizziness or both. They seem to use marijuana as an alternative to getting drunk.

I've be interested if anyone has any anecdotal evidence on 2) or 3), or any other theory.

My thinking is that medical marijuana marginally ups the hassle of getting weed. As opposed to just calling a dealer, more people are going through the process of getting it legally, and controls minimize the amount of weed diverted for recreational purposes. Some people just want to get blotto no matter their poison. If you make it harder to get weed, it will become easier to get drunk by default. Hence the slight uptick in binge drinking.

You may find this article interesting.

#3 the problem is momentum stocks fall really ,really fast when the tide turns

uhh mm if it were possible to construct a winning strategy with it they would be trading it instead of publishing it

I got the sensation the authors had realized there was more money to be made by persuading people you had a unique hedge fund strategy than in academia and were practicing their persuasive writing skills.

They are trading with it. Three of them are at a investment firm.

You are absolutely right that momentum crashes... hard. But that doesn't mean that it's a bad investment, only that one has to consider more than traditional mean-variance optimization. A portfolio properly tilted to momentum will over the long-run certainly earn more return than standard indexing. The issue is that some people may care about the short term, particularly their liquidity at times when momentum crashes (1929, 2008, etc.). But that depends on the individual characteristics of the investor.

AQR is publishing it, not because it's a losing strategy, but rather because it's a strategy that has huge capacity. In other words a risk factor. You could also say in that sense that the stock market itself is a winning strategy (you'd do much better to invest there than pretty much anywhere else), yet plenty of people popularize equity investing. In contrast you're not going to see people publish the details of their HFT or credit arbitrage strategies, because they have highly limited capacity. A marginal increase of investors in the strategy will squeeze returns, not really the case for MOM.

"A portfolio properly tilted to momentum will over the long-run certainly earn more return than standard indexing."

Why do you say this? The word 'certainly' is jarring, but even if its 'nearly certainly', why?

Without reading the paper (gated, $35.99 - cheap price to expose their winning strategy, perhaps), it seems clear to me that the only way one can win with momentum is by correctly identifying momentum and correctly identifying the end of momentum. If one has that information, then I have a simpler strategy: "Buy high, sell low". And, if you act now, I'll reduce the price from $35.99 to three easy payment of $9.99.

I used certainly in the probability-statistics standpoint. As in a biased coin will flip *certainly* heads more times than an unbiased coin given enough trials. This was a response to ummm about momentum being a losing strategy because of its drawdown. Consider the context of claims we both accept (and are well supported by historical data): momentum (MOM) has higher mean return than standard passive indexing (MKT) but that MOM is more prone to large and periodic losses. The long run return of a portfolio of some random variable X is exp(E(log(X))), so if E(log(X)) > E(log(Y)) than one can say X is a winning strategy relative to Y. ummm seems to contend that despite E(MOM) > E(MKT), that E(k * log(MOM) + (1-k) * log(MKT)) 0: E(log(k*MOM + (1-k)*MKT)) -> E(k*MOM + (1-k) * log(MKT)). To see this intuitively consider that long-run compounded returns underperform average arithmetic in the face of volatility (the drawdown effect or "volatility drag"), yet if an uncorrelated component of the portfolio is made small enough it will always be rebalanced to the same expected size regardless of its previous performance. Hence volatility drag disappears for that component of the portfolio.

Since E(MOM) > E(MKT) and E(MKT) > E(log(MKT)) (as it is for all cases where E(X) > 0), that implies E(MOM) > E(log(MKT). Thus for some sufficiently small k: E(log(k*MOM + (1-k)*MKT)) > E(log(MKT)). That is there is some non-zero allocation to MOM that will in expectation outperform a portfolio solely invested in MKT. From the standpoint of actually outperforming in the long run, that's just a simple application of the law of large numbers given the superiority of expectation. So we can say given the original conditions accepted by both me and ummm (and highly supported by historical evidence), one can say that with sufficiently small k and a sufficiently large n-year horizon a momentum tilted portfolio will certainly (i.e. with arbitrarily high probability) outperform standard indexing.

Every damn time I try to put equations in this stupid comment system it drops entire sentences of mine. I think it should still be relatively clear, but let me know if I should clarify on anything that seems to have been dropped.

While this is true, it is true for any uncorrelated asset class or trading strategy. It's not particular to momentum in any way.

Yes, that is absolutely correct, the same logic can be applied to any uncorrelated strategy. The problem is that there's really only a handful of strategies that are robustly positive expected value, uncorrelated and don't require massive infrastructure investment. I'd say off hand: momentum, value, low volatility, and maybe liquidity and size factors.

@Doug, thanks. This is helpful. If I understand what you're saying about volatility drag, somehow MOM is more likely to deliver, say, a 6%/6% return versus, say 8%/4% for MKT, which produces compounding benefits over time.

About those large losses: is this a candidate for a Talebian black swan? Do we really know how fat those tails are? How many elegant financial theories have foundered on these rocks?

Thanks again.

Well, that's a general criticism that can be lobbed at any strategy. The problem with compounded log returns, compared to arithmetic returns is that the potential left tail goes to negative infinity, instead of negative 100%. Historical estimates always face the possibility of censored data that would significantly alter the distribution. However with MOM you have 100+ years of data (people have reconstructed the factor going back to the Victorian era), which also includes major financial crises of 1929, 2008, 1987, 1907, etc. MOM has serious drawdowns, but not that much more serious than MKT. If the possibility of unobserved black swans scares you off from MOM, then its not that much of a leap to reject equities in general, or even possibly the notion of any positive real return investments.

It was the researchers who blocked the exits, not the ants.

#1: Please forward to Port Authority of New York and New Jersey so they can justify the complete mess of pillars, blocked stairways and obstructions on the Penn Station platforms!

"Port Authority: there's more to do here than catch a bus!"

No there isn't. isn't.

Well you could trip, slip, bump into stuff, get your wallet stolen, step into crap. Lots to do indeed.

Oops how forgetful of me!

To this day thinking about those signs still makes me laugh. I mean, just...the very notion that somebody would actively desire to go there to have fun or grab a drink or "meet me at McAnn's" stands in such sharp contrast to the fact that it's the worst place on earth. It just kills me every time.

#5 comedy in China

Any article that contains this assertion:

Sarcastic, Chinese, and female (a rare combination)

can not be taken seriously.

The only Chinese females the author of that article and his ilk are familiar with are rub n tug girls. I agree they are probably not sarcastic on the job.

5) What an asinine, ill-informed article. Of course expats in their little bubble wouldn't know anything about the Chinese comedy scene beyond what comes to them in expat hangouts IN ENGLISH. An article on this subject with no mention of a) Guo Degang b) Zhou Libo c) Zhao Benshan d) xiaopin is not even worth skimming - the distorted half-truths you come away with are inferior to total ignorance.
cliffs: Chinese people DO enjoy dirty jokes and political jokes; sorry your friends Rowswell and Wong are washed up; traditional practitioners who ARE actually funny will continue to outsell flaccid, unfunny "Western" comics

Thanks for posting article #3 on momentum. Good stuff that I wouldn't have seen otherwise.

These momentum strategies can be pretty sensitive to how momentum is measured (as mentioned in the article). Also, reporting results as Sharpe Ratios of trading strategies based on particular momentum horizons is very "practioner" oriented, and is sensitive to the trading strategy and portfolio construction to some extent. A better way to look at the reliability and persistence of this stuff is by looking at the autocorrelation structure with lags from 1-24 months and looking that cross-sectionally and over time. I'm not saying the effect doesn't exist, but reporting aggregate trading strategy results over long periods of time doesn't really capture the complexity and dynamic nature of the effect. I'm more familiar with the results in commodity markets where this is a good approach - I suppose it would be in equity markets as well. I also suspect the AQR guys have done this and understand it.

Comments for this post are closed