Report

A New Jackknife Variance Estimator for Time-Series and Panel Regressions


Abstract: We introduce a new jackknife variance estimator for time-series and panel-data regressions. The novelty in our approach is that we first rotate the data using a particular choice of trigonometric basis functions. This rotation removes serial correlation in a broad class of time-series processes, including random walks, and enables the use of the conventional leave-one-out jackknife on the transformed space of the regressors and residuals. The procedure is tuning-parameter free and naturally adapts to the degree of persistence of the data. We prove the asymptotic validity of our variance estimator under general conditions and demonstrate excellent finite-sample properties in extensive simulation experiments, spanning a wide range of time-series and panel-data designs.

JEL Classification: C12; C13; C22; C23;

https://doi.org/10.59576/sr.1133

Access Documents

File(s): File format is application/pdf https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr1133.pdf
Description: Full text

File(s): File format is text/html https://www.newyorkfed.org/research/staff_reports/sr1133.html
Description: Summary

Authors

Bibliographic Information

Provider: Federal Reserve Bank of New York

Part of Series: Staff Reports

Publication Date: 2024-10-01

Number: 1133

Note: Revised February 2026. Previous title: “A Jackknife Variance Estimator for Panel Regressions.”