The S&P 500 autocorrelation crash in the last six months is significant. Below are relevant charts and reference to possible causes.
The 1-lag autocorrelation of S&P 500 daily returns in periods of less than 120 days has crashed to levels not seen since 1960. Below is a chart that shows the 1-lag, 60-day autocorrelation of S&P 500 daily returns.
As it may be seen above, the autocorrelation has been low since last November. This also holds in the case of 1-lag, 120-day autocorrelation, which since the start of this year is near zero.
In the chart below we include a rolling count of the number of days in a 120-day period with 1-lag, 60-day autocorrelation in -0.04 to +0.04 range.
The last value of the count at 91 is the highest since 1960. Note how the count also rose before the 2008 market top.
The autocorrelation crash pattern is not necessarily bearish but could be related to potential drivers of a bear market. The pattern also points to random price action and the lack of both momentum and mean-reversion in daily returns. In our opinions these are the result of the following among other drivers:
- Dominant algo activity
- Crowded trades in momentum and statistical arbitrage spaces
- Higher market efficiency (EMH regime)
If this phenomenon persists it may mark a significant market regime change and could impose additional burden on struggling CTAs, as well as, on most quant strategies. However, this may be a temporary phenomenon but only time will tell.
If you found this article interesting, I invite you follow this blog via any of the methods below.
If you have any questions or comments, happy to connect on Twitter: @mikeharrisNY
Charting and backtesting program: Amibroker