White Test and Detection of Heteroskedasticity

College Depth 81 in the knowledge graph I know this Set as goal
Unlocks 15 downstream topics
heteroskedasticity testing diagnostics

Core Idea

White's test detects heteroskedasticity by regressing squared residuals on all regressors and their interactions, testing whether these variables explain squared residuals. Unlike Breusch-Pagan, it is robust to specific forms of heteroskedasticity, making it practical for applied work when the source of heteroskedasticity is unknown.

Explainer

You know from studying heteroskedasticity that the problem is non-constant error variance — Var(u|X) depends on X. You also know from the F-test that you can test joint significance of a group of variables: does including these regressors significantly improve fit? White's test combines these two ideas into a single diagnostic: it asks whether the squared residuals — your proxy for the unobservable error variance — are systematically explained by the regressors in your model.

The mechanics of White's test follow a structured three-step procedure. First, run your original OLS regression and save the residuals ê. Second, construct a new regression where the dependent variable is ê² and the regressors are all the original X variables, all their squares (X₁², X₂², ...), and all their pairwise cross-products (X₁·X₂, X₁·X₃, ...). Third, test whether this auxiliary regression has any explanatory power, using the test statistic nR² from the auxiliary regression, which follows a chi-squared distribution under the null of homoskedasticity. The null hypothesis is that none of these terms explains variance in ê², meaning heteroskedasticity is absent.

The key advantage of White's test over simpler alternatives like Breusch-Pagan is its generality. Breusch-Pagan assumes that if heteroskedasticity exists, it is a linear function of X — a specific functional form. White's test is nonparametric in its approach: by including squares and interactions, it captures nonlinear and interactive patterns in error variance without assuming a particular structure. This is what "robust to specific forms of heteroskedasticity" means — you do not have to guess how variance depends on X; the test searches broadly.

The practical downside is a trade-off between power and parsimony. With many regressors, the auxiliary regression can include a very large number of terms (k regressors produce up to k + k + k(k−1)/2 auxiliary variables), consuming degrees of freedom and potentially generating spurious detections. An important caution: White's test can reject homoskedasticity even when the true problem is model misspecification rather than genuine heteroskedasticity — a misspecified functional form also produces systematic residual patterns. If White's test triggers, the appropriate diagnostic question is not just "should I use robust standard errors?" but also "is my model correctly specified in the first place?"

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-TestsHypothesis Testing in RegressionSpecification Error: RESET TestWhite Test and Detection of Heteroskedasticity

Longest path: 82 steps · 424 total prerequisite topics

Prerequisites (4)

Leads To (1)