An Asymptotically F-Distributed Chow Test in the Presence of Heteroscedasticity and Autocorrelation

TL;DR

This paper introduces a robust F-distributed Chow test that accurately detects structural breaks in regression models even when heteroscedasticity and autocorrelation are present, using a novel variance estimator and fixed-smoothing asymptotics.

Contribution

It develops a new Chow test based on a heteroscedasticity and autocorrelation robust variance estimator, justified by fixed-smoothing asymptotics, improving reliability over traditional tests.

Findings

The proposed test maintains the correct size better than chi-square based tests.

Monte Carlo simulations confirm the test's accuracy in various heteroscedastic and autocorrelated settings.

The test employs the standard F distribution, simplifying implementation and interpretation.

Abstract

This study proposes a simple, trustworthy Chow test in the presence of heteroscedasticity and autocorrelation. The test is based on a series heteroscedasticity and autocorrelation robust variance estimator with judiciously crafted basis functions. Like the Chow test in a classical normal linear regression, the proposed test employs the standard F distribution as the reference distribution, which is justified under fixed-smoothing asymptotics. Monte Carlo simulations show that the null rejection probability of the asymptotic F test is closer to the nominal level than that of the chi-square test.