What Is Autocorrelation?
Let me explain autocorrelation directly to you: it's the degree of similarity between a variable in two successive time intervals. More precisely, autocorrelation is a mathematical way to represent how similar a time series is to a lagged version of itself over successive periods. Think of it as correlation, but instead of two different series, you're using the same one twice—once original and once shifted by one or more lags.
For instance, if it's raining today, data might show it's more likely to rain tomorrow than if it were clear. In investing, a stock with strong positive autocorrelation in returns means if it's up today, it's probably up tomorrow too. As a trader, you can use this tool, especially if you're into technical analysis.
Key Takeaways
- Autocorrelation shows the similarity between a time series and its lagged version over intervals.
- It measures how a variable's current value relates to its past values.
- A value of +1 means perfect positive correlation, while -1 means perfect negative.
- Technical analysts use it to see how past prices influence future ones for a security.
Understanding Autocorrelation
You might also hear autocorrelation called lagged correlation or serial correlation, since it looks at a variable's current value against its past ones. Take this simple example: imagine five daily percentage changes in a stock. Compare them to the same values shifted up by one day. The calculation gives a result between -1 and +1.
A +1 means perfect positive correlation—an increase in one leads to a proportionate increase in the lagged series. A -1 means perfect negative—an increase in one causes a decrease in the other. Remember, this measures linear relationships; even if autocorrelation is low, there could be nonlinear patterns at play.
Autocorrelation Tests
The Durbin-Watson test is the go-to method for checking autocorrelation. It comes from regression analysis and gives a statistic between 0 and 4. Closer to 0 means strong positive autocorrelation, closer to 4 means strong negative, and around 2 suggests little to none.
Correlation vs. Autocorrelation
Don't confuse the two: correlation is between two different variables, while autocorrelation is a variable correlated with its own lagged values. In finance, autocorrelation helps you analyze historical prices to predict future moves, like whether a momentum strategy fits.
Autocorrelation in Technical Analysis
Technical analysis focuses on price trends and relationships through charts, unlike fundamental analysis which digs into company health. You can use autocorrelation to gauge how past prices affect future ones. For a stock with high positive autocorrelation, two days of gains might signal more ahead.
Example of Autocorrelation
Suppose you're checking if a stock's returns show autocorrelation. Run a regression with prior session returns as independent and current as dependent. If you get 0.8, that's close to +1, meaning past returns strongly predict future ones positively. You could hold or buy more shares to capitalize on this momentum.
Frequently Asked Questions
What's the difference between autocorrelation and multicollinearity? Autocorrelation is a variable's correlation over time with itself; multicollinearity is when independent variables correlate, like predicting one from another.
Why is autocorrelation problematic? Most stats tests assume independent observations; autocorrelation violates that by showing dependence between values.
What is autocorrelation used for? It's common in technical analysis to spot trends and forecast security performance based on them.
The Bottom Line
Autocorrelation correlates a time series with its lagged self. It's like regular correlation but uses the same series twice. You, as a financial analyst or trader, can apply it to historical prices for future predictions. Technical folks use it to measure past price impacts on futures, often alongside other stats tools.
Other articles for you
Optimized Portfolio as Listed Securities (OPALS) is a single-country equity index product created by Morgan Stanley in 1994 as a precursor to ETFs, designed to track and outperform benchmarks with fewer holdings.
A naked put is an options strategy where you sell put options without a short position in the underlying security to capture premiums if the price rises.
Long-term care insurance provides coverage for nursing home, home health, and personal care for those over 65 or with chronic conditions, offering more options than public programs like Medicaid.
A virtual data room is a secure online platform for storing and sharing sensitive documents during business deals like mergers and acquisitions.
A high-deductible health plan (HDHP) is a type of health insurance with high deductibles and low premiums, often paired with HSAs for tax benefits.
Profitability ratios evaluate a company's ability to generate profits relative to its assets, equity, and sales.
A price taker is an entity that must accept market prices without the power to influence them due to limited market share.
Value of Risk (VOR) evaluates the financial benefits of risk-taking activities to help organizations achieve objectives while treating risks as investments.
Candlestick charts are visual tools used in technical analysis to track and interpret price movements of securities like stocks over time.