Arbing Intraday Cross-Asset Correlations

Tyler Durden's picture

By now, after Zero Hedge has been demonstrating for about a year, even the kitchen sink is aware that cross-asset correlations between stocks, bonds, FX, and commodities is at or near all time highs, which in itself is a very deplorable situation simply because it eliminates virtually all long/short hedging opportunities, courtesy of the Synthetic CDO redux boom whereby most of the trading in stock is conducted via ETFs, as both high beta and low beta, or quality and crap assets all trade as one. But few if anyone was aware of peculiar intraday correlation patterns which may be an eye opener to some readers who believe that stocks are uniformly broken during the day. That is not true: in fact, stocks are only untradeable for the rational investor during the times when the market is most active, around open and close. In fact, in a paper by Michael Bommarito II, "Intraday Correlation Patterns Between the S&P 500 and Sector Indices", we discover that average return correlations have a very distinct U-shape, whereby correlations are near their highs (0.75) just after the open, and before close, while dropping to a statistically significant 0.6 at 1 pm, when volume is the lowest. This merely confirms that increasingly more market participants, read - electronic traders and algos, trade exactly the same strategies at the time when volume is at its peak, indicating that most strategies have nothing to do with actual fundamental investing and all to do with gaming market structure, and hoping to capture some idiot who thinks they can beat the machine. And as we demonstrated recently, many traders no longer trade during the hours between 10am and 3pm. Which means that this is actually a very interesting arb opportunity, for those who wish to take advantage of the machines' downtime, but shorting correlation at open and close, and bidding it up during the day. In fact the trade can be structured as a pair trade with almost no capital downside opportunity.

In his paper Bommarito summarizes his findings as follows:

Notably, there appears to be evidence of two previously unreported patterns in intraday correlation. First, there is a "U-shaped" trend in return correlation, characterized by higher correlation at open and close and lower correlation during mid-day hours. Second, volume  correlation is marked by lower values in the morning and increasing values in the afternoon. In some cases, this trend even takes the infamous "hockey-stick" shape, exhibiting stable values in the morning but sharply increasing values in the late afternoon.

The U-shape of correlation can be seen in the chart below:

In addition to return correlation, another observation which however should be less surprising, is a comparable finding in volume correlations between the SPY and various sector indices. However, unlike the symmetric U shape above, here we notice that the volume pick up is materially more pronounced toward the close of the market:

Bommarito's summary of these surprising observations is spot on:

The results contained above provide evidence for the existence of two striking and previously unreported patterns. First, the correlation between the S&P 500 and sector indices is not time-stationary within a day. This is the result of an extreme U-shape in covariance and is a phenomena separate from the previously observed U-shape in volatility. Furthermore, this U-shape is possibly evidence of market inefficiency. Whereas market liquidity or volatility may vary with natural intraday cycles of market interest, the structural relationship between the S&P 500 and market sectors should not exhibit intraday patterns. The second phenomenon observed above is that the correlation of volume between the S&P 500 and sectors indices exhibits lower correlation in the morning and rapidly increasing correlation in the afternoon. This trend even takes the infamous "hockey stick" shape for some sector indices and values of tau. Both of these results have implications for strategies that are based on the assumption that correlation between sector indices and the market are time-stationary within a day.

All this merely confirms what we have been witnesses to since the Flash Crash: the market has now crossed into the plane where it is essentially has a "mind" of its own- positive feedback loops are the only major factor in the market, the bulk of whose volume is now contained within such synthetic constructs as ETFs, which in essence has completely broken the cause-effect relationship between input signals and output prices. Stunningly, we have reached a point where ETFs are a primary gating factor to determining what the underlying price of constituents stocks is. When one throws in underlying derivatives as well, such as plain vanilla stock options, and increasingly more complex option-derived strategies, and consider that the bulk of such trades continue to be illegally manipulated on a daily basis courtesy of the HFT marauder brigade, and one must be amazed we don't have daily flash crashes for the entire market.

This article will likely prompt a variety of adjustments to trading signals and quant factors, once the Ph.D's in the market realize that the abovementioned patterns allow a "reversion to the mean" strategy to be implemented, and which can be accelerated by yet more feedback loops: both things that the HFTs substructure of the market does better than anything, even if neither has any reflection in reality.

Ultimately, this is nothing but another signal of just how broken and fragmented the market has become, and just how little real capital formation is left in the stock market, which is nothing more than a vehicle for those with the fastest computers to take advantage of all everyone else, courtesy of precisely this two-tiered market structure. For everyone else, who still has the urge to gamble in stocks at odds far worse than Vegas, we believe the only way to make money is to beat the robots at their own game, which means to exploit the weaknesses in the market which are created precisely due to the monopolistic dominance of computerized trading. Everything else is a sure way to lose your capital.