Skip to main content

"Markets Were Up Today Because They Were Down Yesterday"

  • Author:
  • Updated:

If you read only one thing about ancient fire management methods and modern volatility, it should be this report by Chris Cole of Artemis Capital Management.

There is much goodness here that others have discussed. But perhaps the most interesting for the current world is the move from positive to negative serial correlation in stock prices over time. (Positive serial correlation means that stocks are more likely to go up the day after an up day, negative that they're more likely to go down on the day after an up day, and zero that yesterday's performance is not predictive of today's.) Serial correlation peaked in 1971, floated around zero-ish from 1980 to 2000, and is now strikingly negative:

As Cole says:

The degree of up and down days in the DJIA is at the most extreme level in recorded history representing a pinnacle in an era of daily mean reversion. It only takes a casual observer of markets to see that the propensity for large-up days followed by large-down days seems particularly vicious in today's cycle. The excessive intra-day and day-to-day volatility is nauseating to professional and retail investors alike and multi-100 point swings in the DJIA are all too common. ...

We are going nowhere at the fastest pace in market history.

If you like the efficient markets hypothesis, serial correlation is a bummer: if it is consistently nonzero, then you can make money consistently just based on pretty simply observable past results (if it's positive, just buy on the day after stocks go up; if it's negative, buy on the day after stocks go down). So this data set is fun insofar as the hypothesis was popularized in the late 1960s and early 1970s - by this measure, exactly when it was most wrong.

Because the Dodd-Frank Act requires that anyone who writes interesting markets research has to be a gold bug, Cole points out that the peak in serial correlation occurred three months before the US went off the gold standard, and therefore blames monetary expansion for the decline in serial correlation: "is it possible that monetary expansion has artificially rewarded stock market mean reversion strategies (such as value investing and buying on dips) for the past 40 years?"

Sure, why the hell not? In particular, if your model for the Fed is "any time stock prices drop there will be easing," that does explain the trend - all you need is for the stock market to puke hard enough, and the monetary authorities will come in and bring prices back up.

You could, though, pick other events that occurred around the same time as the shift in serial correlation. For instance, the early 1970s marked the invention of microprocessors and computer networking and generally saw the beginning of computer use in finance. They also saw the development of options theory and arbitrage - which, of course, benefited from the use of computers to crunch data.

Much discussion has focused on computer trading fueling cross-asset correlations. My sense is that computerized and computer-assisted strategies tend to drive serial correlations down as well: if your statistical-arbitrage model is "we'll pick things that are statistically out of whack and buy them," and everyone does that, then that will drive them back into whack. Meaning that as prices deviate from trend, lots of robots are buying them to bring them back to the trend.

Whatever your explanation, though, someone loses in this environment: people who are set up to be "long serial correlation." My suspicion is that in practice that means a lot of retail investors who think "ooh, the market was up today, maybe I'll buy stocks" - or, more relevant these days, "holy shit was the market down today, I'd better sell all my stocks and stash my money in a PVC pipe buried four feet underground."

If you take Cole's thesis seriously, ordinary investors are ill served by overly strenuous efforts to find meaning in market movements: those invented meanings suggest that stocks went down today for some real reason that might drive them still lower tomorrow. The message of serial correlation is to let go of those explanations and find inner peace. Markets went down, whatever - tomorrow, the Fed and/or the robots will drive them back up again.

Volatility is change, and the world is changing [FTAV]

"Your Alpha Is Burning" - Fighting Greek Fire with Fire: Volatility, Correlation, and Truth [ZH]


Banks Were Asked If They Would Prefer To Make More Money Or Less Money, Chose More Money

One kind of obvious thing about financial markets is that you can't just call everyone into a room and tell them, "look, guys, just be honest about the price that you would pay / receive for Thing X." This is because financial industry traders are degenerate lying scumbags. No, wait, that's not right. This is because if everyone just told each other their reserve prices then it would be really hard for them to make any money trading and so we, like, wouldn't have a financial system. So you have things like anonymous execution on stock exchanges and dark pools and, um, lying scumbag traders. And that allows you to have profitable trading. Of course you have to put some limits on the lying scumbaggery: you can't tell people you're investing their money while really blowing it on hookers, and I guess now you can't sell someone synthetic CDOs without telling them who was on the other side. But a little fudging around the edges about the price you're willing to pay or receive - or the price you could pay or receive elsewhere - is kind of at the heart of what trading is. So in a sense the amazing thing about the Libor scandal is that people are amazed by it. A quick recap: