A finite sample correction for the panel Durbin–Watson test

ABSTRACT

This study shows that the panel Durbin–Watson $d$ test is incapable of distinguishing temporal heteroskedasticity and serial correlation with short time series. The study proposes a new finite sample correction for solving this weakness of the panel Durbin–Watson test. Monte Carlo simulations demonstrate that the corrected statistic performs better in terms of size and power than the extant panel $d$ tests. It leads to more accurate inferences in empirical studies.

KEYWORDS:

• Autocorrelation
• Baltagi–Wu test
• Durbin–Watson test
• serial correlation
• temporal heteroskedasticity

Acknowledgement

The author thanks Kaarkuzhali Gunasekaran and Taylor & Francis's RAEC production team for highly professional proofreading and typesetting.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ For example, in Applied Economics or Applied Economics Letters, Farag and Cressy (2010), Otsuka (2012), Fuerst and Matysiak (2013), Jawadi, Idi Cheffou, and Jawadi (2016), Chen and Lee (2017), Mensi et al. (2017), Racicot and Rentz (2017), and Chen, Chen, and Lin (2020).

² It is implemented in Stata command xtregar.

³ The econometrics textbooks (e.g. Baltagi 2013) and empirical results also indicate the same thing.