Tests for the error component model in the presence of local misspecification

Abstract

It is well known that most of the standard speci¯cation tests are not valid when the alternative hypothesis is misspeci¯ed. This is particularly true in the error component model, when one tests for either random e®ects or serial correlation without taking account of the presence of the other e®ect. In this paper we study the size and power of the standard Rao's score tests analytically and by simulation when the data is contaminated by local misspeci¯cation. These tests are adversely a®ected under misspeci¯cation. We suggest simple procedures to test for random e®ects (or serial correlation) in the presence of local serial correlation (or random e®ects), and these tests require ordinary least squares residuals only. Our Monte Carlo results demonstrate that the suggested tests have good¯nite sample properties and are capable of detecting the right direction of the departure from the null hypothesis. We also provide some empirical illustrations to highlight the usefulness of our tests.

Can Gravity Model Explain BIMSTEC's Trade?

“…Also, the estimated Bera et al (2001) statistic for serial correlation indicates that both import and export panels suffer from serial correlation at 1% level of significance, which is again confirmed by Bhargava et al (1982) modified Durbin-Watson (DW) statistic.…”

A Study of the Determinants of Cost and Revenue Efficiency in the Context of Islamic and Conventional Banks: The Case of Malaysia from 2006 to 2012

“…However, Bera and Yoon (1993) show that in the presence of first-order autocorrelation, the Lagrange multiplier test developed by Baltagi and Li (1990) tends to reject the null hypothesis of the absence of random effects even if it is correct. For this reason, a modified Lagrange multiplier test was developed by Bera et al (2001) for balanced and unbalanced panel data. The results, reported in Table 6, show that the test of random effects (LM (Var (u) = 0)) and the joint test of Baltagi and Li (1991) for serial correlation and random effects reject the null hypothesis which indicates the absence of random effects.…”

“…The results, reported in Table 6, show that the test of random effects (LM (Var (u) = 0)) and the joint test of Baltagi and Li (1991) for serial correlation and random effects reject the null hypothesis which indicates the absence of random effects. However, the results of the modified version of the Lagrange multiplier test (ALM (Var (u) = 0)) developed by Bera et al (2001) fail to reject the null hypothesis and prove that rejecting the null hypothesis of the joint test is mainly the result of the presence of a serial correlation problem. This finding is confirmed by Wooldridge (2002) whose null hypothesis, which was rejected, assumes the absence of first-order autocorrelation.…”

“…Various tests will then be performed. The test of Bera et al (2001) and that of Wooldridge (2002) to study errors autocorrelation, and the test of Breusch and Pagan (1979) and the likelihood ratio to study the heteroskedasticity problem. Some previous studies have applied the random effects model taking into account errors heteroskedasticity problem (Hoff, 2007), McDonald (2009) and Johnes et al (2013), whereas others used a Tobit regression model (Casu & Molyneux, 2003) and Sufian (2009).…”

Aggressive and Defensive High-Frequency Trading and its Impact on Liquidity of German Stock Market

“…Next Hausman test for model specification rejected random effects model (Hausman, 1978). Serial correlation has been examined with likelihood-based conditional LM test (Baltagi and Li, 1995), locally robust LM test (Bera, Sosa-Escudero and Yoon, 2001), Breusch-Godfrey test for panel models and Wooldridge's first-difference test (Wooldridge, 2010). First three tests did not reject the presence of serial correlation.…”