Augmented Dickey-Fuller Test
A statistical test used to determine whether a time series contains a unit root, indicating non-stationarity.
What is Augmented Dickey-Fuller Test?
The Augmented Dickey-Fuller (ADF) test evaluates the null hypothesis that a time series has a unit root (is non-stationary and has a stochastic trend). The 'augmented' version includes lagged differences of the series to account for higher-order serial correlation in the error terms. Rejecting the null hypothesis suggests the series is stationary.
Why Augmented Dickey-Fuller Test Matters
Regressing non-stationary time series on one another can lead to spurious regressions, where completely unrelated variables appear to be highly correlated simply because they both have a trend over time. Stationarity is a fundamental requirement for valid time-series forecasting and causal inference.
Example
A researcher examining daily stock market returns runs the ADF test on the raw stock prices and fails to reject the null hypothesis, confirming the prices are non-stationary. They then take the first differences of the prices (returns) and find them to be stationary.
Common Mistakes
- Misinterpreting a failure to reject the null hypothesis as absolute proof of a unit root, when it might just be due to low test power in short time series.
- Failing to correctly specify the test equation by omitting a necessary constant or trend term, which completely changes the critical values and validity of the test.