Skip to content

Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root

Why this mattered

Dickey and Fuller’s 1981 paper mattered because it gave econometricians a practical likelihood-ratio framework for testing whether an autoregressive time series contains a unit root. Before this work, many empirical time-series models implicitly treated macroeconomic and financial series as stationary after fitting deterministic trends or autoregressions. The paper showed that the unit-root case is statistically nonstandard: conventional asymptotic distributions for likelihood-ratio and related tests do not apply. By deriving the relevant distributions and tabulating critical values, Dickey and Fuller made it possible to distinguish persistent but stationary dynamics from stochastic trends in a disciplined way.

The paradigm shift was methodological and conceptual. A unit root meant that shocks could have permanent effects, so the choice between trend-stationary and difference-stationary modeling was not a technical detail but a claim about the structure of economic fluctuations. After this paper, researchers had a standard testable language for asking whether output, prices, exchange rates, interest rates, and other series should be modeled in levels, detrended levels, or differences. That changed empirical macroeconomics, forecasting, and finance by making nonstationarity a central diagnostic rather than an afterthought.

Its influence also set the stage for later breakthroughs in time-series econometrics. The Dickey-Fuller testing framework became the basis for augmented unit-root tests, for the Phillips-Perron line of robust unit-root testing, and for the cointegration revolution of Engle and Granger, where nonstationary series could still have meaningful long-run equilibrium relations. In that sense, the paper did not merely add a test; it helped reorganize empirical time-series analysis around integration, persistence, and long-run structure.

Abstract

(no abstract available)

Sources