An ad-free and cookie-free website.
Autocorrelation, or serial correlation, is the correlation of one variable with itself over different time periods. This is important for testing whether stock returns behave like independent variables, which is an assumption of statistical tests of several important theories, including the Efficient Market Hypothesis (EMH).
To calculate autocorrelation for monthly returns with a 1-month lag, use nearly the same return stream except stagger the y-variable dependent variable by 1 month. In a basic 36-month example, the x-variable in the regression could go from Jan 2014-Dec 2016 and the y-variable would then go from Feb 2014-Jan 2017.
Synonym: serial correlation, cross correlation
For context, imagine running an autocorrelation of risk measures. Here, historically, statistically significant autocorrelation has been observed. A likely reason is that volatile periods in financial markets are often followed by other volatile periods. Some refer to this as a 'volatility cluster'.
With respect to correlations of stock returns, some use autocorrelation studies to determine the level of decay in alpha signals and for research on stock momentum.
Bud: Hey Guy, nice chart, eh? Solid uptrend,
nice doji, MACD, and strong momentum.
Guy: You are the Technician, but technically, it looks like autocorrelation to me.
Many terms have 4-5 minute videos showing a derivation and explanation. If this term had one, it would appear here.
Videos can also be accessed from our YouTube Channel.
If this term had a video, the script would be here.
Click box for answer.
False, performance chasing could result in positive autocorrelation.
Still unclear on Autocorrelation? Check out our 27-video deep-dive in Excel in the Quant 101 Series.
Our trained humans found other terms in the category statistics for finance you may find helpful.
Subscribe on YouTube so you catch the latest video. Email list too.
A newly-updated free resource. Connect and refer a friend today.