Regression modelEconometricsModel

Augmented Dickey-Fuller (ADF) Unit-Root Test

Also known as: ADF test, Dickey-Fuller test, unit root test, Genişletilmiş Dickey-Fuller testi

OriginatorDavid A. Dickey & Wayne A. FullerYear1979Sources2Related methods16

The Augmented Dickey-Fuller (ADF) test is the most widely used test for a unit root — that is, for whether a time series is non-stationary and must be differenced before modelling. Introduced by David Dickey and Wayne Fuller in 1979 and extended by Said and Dickey in 1984 to series with higher-order autocorrelation, it regresses the change in the series on its lagged level plus lagged differences and asks whether the lagged-level coefficient is zero.

Key highlights

  • The de facto standard unit-root test, available in every econometrics package and widely understood by reviewers.
  • The augmentation by lagged differences accommodates higher-order autocorrelation without modelling it explicitly.
  • Flexible deterministic specification (constant, trend) lets it match a wide range of series behaviour.
  • Provides a clear order-of-integration decision that feeds directly into ARIMA and cointegration analysis.

Intuition

This section is available to Pro members. Upgrade to Pro

How it works

This section is available to Pro members. Upgrade to Pro

When to use it

Apply the ADF test to any single time series before fitting models that assume stationarity, such as ARMA/ARIMA specification, regression among time series, or vector autoregressions — non-stationary inputs can produce spurious regressions. It is also the workhorse for determining the order of integration and as a building block of the Engle-Granger cointegration test. Choose the deterministic terms (none, constant, or constant-plus-trend) to match the visual behaviour of the series, and select the lag length to remove residual autocorrelation. Because the test has low power against near-unit-root (highly persistent but stationary) processes, pair it with a confirmatory KPSS test, whose null is the opposite, and with the Phillips-Perron test, which handles autocorrelation non-parametrically.

Strengths & limitations

Strengths
  • The de facto standard unit-root test, available in every econometrics package and widely understood by reviewers.
  • The augmentation by lagged differences accommodates higher-order autocorrelation without modelling it explicitly.
  • Flexible deterministic specification (constant, trend) lets it match a wide range of series behaviour.
  • Provides a clear order-of-integration decision that feeds directly into ARIMA and cointegration analysis.
Limitations
  • Low power against persistent stationary alternatives: it often fails to reject a unit root that is not truly present.
  • Results are sensitive to the chosen lag length and to whether a constant and trend are included.
  • The non-standard Dickey-Fuller distribution must be used; ordinary t critical values are invalid.
  • Structural breaks in the series can masquerade as a unit root, biasing the test toward non-rejection.

Common pitfalls

This section is available to Pro members. Upgrade to Pro

Applications

This section is available to Pro members. Upgrade to Pro

Frequently asked

What does failing to reject the ADF null mean?

It means there is insufficient evidence against a unit root, so the series is treated as non-stationary and usually differenced before modelling. Because the test has low power, failing to reject is not strong proof of a unit root — confirm with a KPSS test, which reverses the hypotheses.

How is the ADF test different from KPSS?

Their hypotheses are opposite. ADF takes non-stationarity (a unit root) as the null and looks for evidence of stationarity; KPSS takes stationarity as the null and looks for evidence of a unit root. Running both gives a more reliable, confirmatory conclusion about the order of integration.

Why can't I use ordinary t critical values?

Under the unit-root null the lagged level is a non-stationary regressor, so the t-ratio does not converge to a normal distribution. It follows the non-standard Dickey-Fuller distribution, whose critical values are more negative than the usual ones; using ordinary t values would reject far too often.

How many lagged differences should I include?

Enough to remove serial correlation from the residuals but no more, since extra lags cost power. Common practice is to select p by an information criterion (AIC/BIC) or a data-dependent rule, then verify the residuals are approximately white noise.

Sources

  1. 1.
    Dickey, D. A., & Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 74(366a), 427–431.
  2. 2.
    Said, S. E., & Dickey, D. A. (1984). Testing for unit roots in autoregressive-moving average models of unknown order. Biometrika, 71(3), 599–607.

You have read it. What now?

Cite this page

ScholarGate. (2026, June 2). Augmented Dickey-Fuller Test. ScholarGate. https://scholargate.app/econometrics/adf-test