Unit Roots and Testing for Stationarity

College Depth 82 in the knowledge graph I know this Set as goal
Unlocks 5 downstream topics
time-series unit-roots stationarity

Core Idea

A time series is stationary if its mean, variance, and autocovariances are time-invariant. A unit root (coefficient on lagged dependent variable equal to 1) induces nonstationarity and persistence. Tests like the Augmented Dickey-Fuller (ADF) and KPSS test detect unit roots; differencing restores stationarity for I(1) series.

How It's Best Learned

Simulate AR(1) processes with different coefficients (e.g., 0.9 vs 1.0) and observe how unit roots produce very different time path behavior.

Common Misconceptions

A nearly unit root process (e.g., φ = 0.99) is not the same as a unit root process; small differences have large implications for statistical properties.

Explainer

Think back to what makes a time series useful for standard inference: you need the statistical properties of the series to be stable over time. If the mean, variance, and autocorrelations all stay the same regardless of when you sample, the series is stationary, and the familiar tools of regression and hypothesis testing apply. If those properties drift — the mean trends upward, the variance grows, or the autocovariances depend on where you are in time — the series is nonstationary, and standard inference breaks down in subtle and serious ways.

The most important source of nonstationarity in economic data is the unit root. Consider the simplest case: an AR(1) process yₜ = φyₜ₋₁ + εₜ. If |φ| < 1, the effect of a shock to y dies away over time — the series reverts toward its mean and is stationary. If φ = 1 exactly, shocks never die away; every εₜ is permanently incorporated into the level of the series. This is a random walk: yₜ = yₜ₋₁ + εₜ, the most common unit root process. GDP, stock prices, and interest rates often behave this way — a shock today shifts the entire future path of the series, not just the next few periods. The crucial distinction between φ = 0.99 and φ = 1.0 seems numerically small but is statistically enormous: the first eventually reverts, the second never does.

Running a regression between two unrelated random walk series produces the spurious regression problem: you'll find high R² and significant t-statistics even though there is no true relationship. This is why testing for unit roots before modeling is essential. The Augmented Dickey-Fuller (ADF) test is the standard tool. It tests the null hypothesis that a unit root is present (φ = 1) against the alternative of stationarity (|φ| < 1). A key quirk: the test statistic does not follow a standard t-distribution under the null, so you must use Dickey-Fuller critical values, which are more negative than standard thresholds. Failing to reject the null means the series likely has a unit root.

The KPSS test takes the opposite approach: it tests the null of stationarity against the alternative of a unit root. Using ADF and KPSS together is good practice — if ADF fails to reject and KPSS rejects, both tests point to a unit root. When a series has a unit root, the standard remedy is differencing: taking first differences Δyₜ = yₜ - yₜ₋₁ removes one unit root. If a series requires one difference to become stationary, it is called integrated of order 1, or I(1); requiring two differences gives I(2), and so on. Once differenced to stationarity, you can apply AR models and standard regression — which is exactly where the next topic, AR models, begins.

Practice Questions 5 questions

Prerequisite Chain

Counting to 10Counting to 20Understanding ZeroThe Number ZeroCounting to FiveOne-to-One CorrespondenceCombining Small Groups Within 5Addition Within 10Addition Within 20Two-Digit Addition Without RegroupingTwo-Digit Addition with RegroupingAddition Within 100Repeated Addition as MultiplicationMultiplication Facts Within 100Division as Equal SharingDivision as Grouping (Measurement Division)Division: Grouping (Repeated Subtraction) ModelDivision: Fair Sharing ModelDivision as Equal SharingDivision as GroupingBasic Division FactsDivision Facts Within 100Two-Digit by One-Digit DivisionDivision with RemaindersRemainders and Quotients in DivisionDivision Word ProblemsIntroduction to Long DivisionFactors and MultiplesPrime and Composite NumbersEquivalent FractionsRelating Fractions and DecimalsDecimal Place ValueReading and Writing DecimalsComparing and Ordering DecimalsAdding and Subtracting DecimalsMultiplying DecimalsDividing DecimalsDividing FractionsMixed Number ArithmeticOrder of OperationsInteger Order of OperationsVariable ExpressionsCombining Like TermsOne-Step EquationsTwo-Step EquationsSolving Multi-Step EquationsEquations with Variables on Both SidesAngle Pairs: Complementary, Supplementary, and VerticalParallel Lines and TransversalsCorresponding AnglesAlternate Interior AnglesTriangle Angle Sum TheoremExterior Angle TheoremTriangle Inequality TheoremSimilar Triangles: AA SimilaritySimilar Triangles: SSS and SAS SimilarityProportions in Similar TrianglesRight Triangle Trigonometry IntroductionTrigonometric Ratios ReviewRadian MeasureConverting Between Degrees and RadiansThe Unit CircleGraphing Sine and CosineGraphing Tangent and Reciprocal Trigonometric FunctionsDerivatives of Trigonometric FunctionsAntiderivativesIndefinite IntegralsBasic Integration RulesRiemann SumsDefinite Integral DefinitionProbability Density Functions and Continuous DistributionsCumulative Distribution FunctionsContinuous Random VariablesNormal DistributionCentral Limit TheoremConfidence Intervals for MeansZ-Tests and T-Tests for MeansOne-Sample Z-Test for MeansOne-Sample and Two-Sample T-TestsOne-Way ANOVAF-Test and Joint SignificanceChow Test and Detection of Structural BreaksUnit Roots and Testing for Stationarity

Longest path: 83 steps · 496 total prerequisite topics

Prerequisites (4)

Leads To (1)