Durbin-Watson Statistic for Autocorrelation

College Depth 76 in the knowledge graph I know this Set as goal
Unlocks 3 downstream topics
autocorrelation diagnostics testing

Core Idea

The Durbin-Watson statistic DW = Σ(ûₜ - ûₜ₋₁)² / Σûₜ² approximates 2(1 - ρ̂) where ρ̂ is the first-order autocorrelation. Values near 2 suggest no autocorr, < 2 suggests positive autocorr, and > 2 suggests negative autocorr, providing a quick diagnostic.

Explainer

From your study of autocorrelation and lag structures, you know that when regression residuals are correlated over time, OLS standard errors are biased and t-statistics are unreliable. The Durbin-Watson statistic is the standard first-pass diagnostic for detecting this problem. Its formula — the sum of squared *differences* between consecutive residuals, divided by the sum of squared residuals — is designed to measure exactly how much each residual resembles the one that came before it.

The key insight is the relationship DW ≈ 2(1 − ρ̂). If there is no autocorrelation, ρ̂ ≈ 0, so DW ≈ 2. If residuals are strongly positively autocorrelated (ρ̂ close to +1, meaning each residual is similar to the previous one), then consecutive differences are small, making the numerator small, and DW approaches 0. If residuals are strongly negatively autocorrelated (ρ̂ close to −1, meaning residuals alternate in sign), then each difference is large, and DW approaches 4. So the full scale runs 0 to 4, with 2 as the "clean" value.

In practice, you compare the computed DW statistic to critical bounds dL and dU from the Durbin-Watson tables (which depend on sample size n and the number of regressors k). If DW < dL, reject the null of no positive autocorrelation. If DW > 4 − dL, reject the null of no negative autocorrelation. Between dL and dU is an inconclusive zone — not evidence of no autocorrelation, but not decisive evidence against it either. This inconclusive region is one of the test's known limitations.

The Durbin-Watson statistic has two important restrictions to internalize. First, it only tests for first-order autocorrelation — whether ûₜ is correlated with ûₜ₋₁. It will miss higher-order patterns (e.g., quarterly seasonality where ûₜ correlates with ûₜ₋₄). Second, it is invalid when a lagged dependent variable appears as a regressor, because in that case the residuals are mechanically correlated with the regressor, violating the test's assumptions. For those more complex situations, you will next encounter the Breusch-Godfrey test, which handles both higher-order autocorrelation and lagged-dependent-variable models.

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 DistributionClassical OLS Assumptions (Gauss-Markov)Autocorrelation: Structure and SourcesDurbin-Watson Statistic for Autocorrelation

Longest path: 77 steps · 440 total prerequisite topics

Prerequisites (1)

Leads To (2)