What Is Serial Correlation?
Let me explain serial correlation to you directly: it happens in a time series when a variable and a lagged version of itself—for example, at times T and T-1—show correlation over periods of time. You'll see repeating patterns in this, where the level of a variable influences its future level. In finance, I use this correlation as a technical analyst to figure out how well a security's past price predicts its future price.
You should know that serial correlation is much like the statistical ideas of autocorrelation or lagged correlation.
Key Takeaways
Here's what you need to grasp: serial correlation is the relationship between a given variable and a lagged version of itself over various time intervals. It measures how a variable's current value connects to its past values. If a variable is serially correlated, it suggests it's not random. As a technical analyst, I validate profitable patterns in a security or group of securities this way and determine the risks in investment opportunities.
Serial Correlation Explained
In statistics, serial correlation describes the relationship between observations of the same variable over specific periods. If I measure a variable's serial correlation as zero, there's no correlation, and each observation is independent. But if it skews toward one, the observations are serially correlated, meaning future observations are affected by past values. Basically, a serially correlated variable has a pattern and isn't random.
Error terms come up when a model isn't fully accurate, leading to different results in real-world use. When error terms from different periods—usually adjacent ones—are correlated, that's serial correlation in the error term. This occurs in time-series studies when errors from one period carry over to future periods. For instance, if I overestimate stock dividend growth in one year, it'll lead to overestimates in the years that follow.
Important Note on Trading Models
Serial correlation can make simulated trading models more accurate, which directly helps you as an investor develop a less risky investment strategy.
Serial Correlation in Technical Analysis
Technical analysis relies on measures of serial correlation to analyze a security's pattern. This is based entirely on the stock's price movement and volume, not the company's fundamentals. If you practice technical analysis and use serial correlation correctly, you can identify and validate profitable patterns in a security or group of securities and spot investment opportunities.
The Concept of Serial Correlation
Serial correlation started in engineering to determine how a signal, like a computer signal or radio wave, varies compared to itself over time. It gained traction in economics as economists and econometricians applied it to analyze economic data over time.
Today, almost all large financial institutions employ quantitative analysts, or quants. These analysts use technical analysis and statistical methods to analyze and predict the stock market. They identify correlation structures to improve forecasts and potential profitability of strategies. Plus, spotting the correlation structure makes simulated time series more realistic, reducing the risk in investment strategies.
Quants are crucial to these institutions' success because they build market models that form the basis for investment strategies.
Fast Fact
Serial correlation originated in signal processing and systems engineering to see how a signal varies with itself over time. In the 1980s, economists and mathematicians flocked to Wall Street to apply it for predicting stock prices.
Determining Serial Correlation
Among quants, serial correlation is determined using the Durbin-Watson (DW) test. The correlation can be positive or negative. A stock price with positive serial correlation shows a positive pattern. One with negative serial correlation has a negative influence on itself over time.
