Abstract
Based on the common correlated effects (CCE) method and the Lagrange multiplier (LM) principle, this paper proposes a slope homogeneity test in a panel data model with a multifactor error structure that allows unobserved factors to be correlated with explanatory variables. The CCE method is first used to transform the regression equation to control for the unobserved common factors. Then, we adopt the idea of an LM-type test to conduct a homogeneity test. Our asymptotic analysis indicates that the test statistic is asymptotically normally distributed under the null hypothesis of homogeneity, regardless of the errors’ normality or homoskedasticity, as both N and T go to infinity, with \(T^{2/3}N^{-1}\rightarrow 0\) and \(T^{2}N^{-1}\rightarrow \infty \). It is also proved that the test is asymptotically powerful under a sequence of Pitman local alternatives. Monte Carlo simulations indicate that the test has good finite sample properties for all combinations of N and T, with the exception of a large N / T. The simulation results also suggest that the proposed test is robust to the errors’ non-normality and conditional heteroskedasticity.
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs00362-017-0929-1/MediaObjects/362_2017_929_Fig1_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs00362-017-0929-1/MediaObjects/362_2017_929_Fig2_HTML.gif)
Similar content being viewed by others
References
Ando T, Bai J (2015) A simple new test for slope homogeneity in panel data models with interactive effects. Econ Lett 136:112–117
Bai J (2009) Panel data models with interactive fixed effects. Econometrica 77:1229–1279
Bai J, Ng S (2004) A PANIC attack on unit roots and cointegration. Econometrica 72:1127–1177
Baltagi BH, Griffin JM (1997) Pooled estimators vs. their heterogeneous counterparts in the context of dynamic demand for gasoline. J Econom 77:303–327
Baltagi BH, Griffin JM, **ong W (2000) To pool or not to pool: homogeneous versus heterogeneous estimators applied to cigarette demand. Rev Econ Stat 82:117–126
Bao Y, Ullah A (2010) Expectation of quadratic forms in normal and nonnormal variables with applications. J Stat Plan Inference 140:1193–1205
Blomquist J, Westerlund J (2013) Testing slope homogeneity in large panels with serial correlation. Econ Lett 121:374–378
Breitung J, Roling C, Salish N (2016) LM-type tests for slope homogeneity in panel data models. Econom J 19:166–202
Hobza T, Morales D, Pardo L (2014) Divergence-based tests of homogeneity for spatial data. Stat Papers 55:1059–1077
Juhl T, Lugovskyy O (2014) A test for slope heterogeneity in fixed effects model. Econom Rev 33:906–935
Lieberman O (1994) A laplace approximation to the moments of a ratio of quadratic forms. Biometrika 81:681–690
Matyas L, Blanchard P (1998) Misspecified heterogeneity in panel data models. Stat Papers 39:1–27
Moon HR, Perron B (2004) Testing for a unit root in panels with dynamic factors. J Econom 122:81–126
Pesaran MH (2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74:967–1012
Pesaran MH (2007) A simple panel unit root test in the presence of cross-section dependence. J Appl Econom 22:265–312
Pesaran MH, Yamagata T (2008) Testing slope homogeneity in large panels. J Econom 142:50–93
Phillips PCB, Sul D (2003) Dynamic panel estimation and homogeneity testing under cross section dependence. Econom J 6:217–259
Su L, Chen Q (2013) Testing homogeneity in panel data models with interactive fixed effects. Econom Theory 29:1079–1135
Ullah A (2004) Finite sample econometrics. Oxford University Press, New York
Ullah A, Srivastava VK, Chandra R (1983) Properties of shrinkage estimators when the disturbances are not normal. J Econom 21:389–402
Westerlund J, Urbain J (2015) Cross-sectional averages versus principal components. J Econom 185:372–377
Acknowledgements
We are grateful to editor-in-chief professor Werner G. Müller and two anonymous referees for many constructive comments on the previous version of the paper. We thank Professor Zhonglin Bai and Professor Shi Li for helpful comments at the 2015 annual conference of Chinese association of quantitative economics. Financial supports for this paper are provided by the National Natural Science Foundation of China (NSFC) (Project No. 71071130) “Stationary and non-stationary linear panel data modelling research under cross-sectional dependence and non-spherical disturbances conditions,” the Fundamental research funds for the central universities (Project Nos. JBK171115, JBK170141, JBK160125, and JBK160138), and fundamental research funds for excellent overseas talents.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Auxiliary lemmas
Lemma 1
Suppose that \(\left\| {\varvec{\beta } _i } \right\| <K\) for each i, and Assumptions A1–A4 hold. Then
- (i)
\({\bar{{\varvec{H}}}}'\bar{{\varvec{H}}}/T=\bar{{{\varvec{P}}'}}{\varvec{G}}'G\bar{{P}}/T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\)
- (ii)
\({\bar{{\varvec{H}}}}'\varvec{F}/T=\bar{{{\varvec{P}}'}}({\varvec{G}}'\varvec{F}/T)+\mathrm{O}_p ((NT)^{-1/2})\);
- (iii)
\({\varvec{X}}'_i \bar{{\varvec{M}}}\varvec{F}/T={\varvec{X}}'_i \bar{{\varvec{M}}}_q F/T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\)
- (iv)
\(\varvec{Q}_{\textit{iT}} ={X}'_i \bar{{\varvec{M}}}_q X_i /T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\),
- (v)
\(\varvec{\xi } _{\textit{iT}} =T^{-1/2}{\varvec{X}}'_i \bar{{\varvec{M}}}_{\varvec{q}} \varvec{v}_{\varvec{i}} +\mathrm{O}_p (T^{1/2}N^{-1})\)
- (vi)
\(\bar{{\varvec{\varepsilon }}}^{*\prime }v_i /T =\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\).
Proof
These results are directly from Eqs. (36)–(38), (43), (44) and (A.13) in Pesaran (2006). \(\square \)
Lemma 2
Suppose that Assumptions A1–A5 hold, then
- (i)
\({\varvec{\lambda }}'_i {\varvec{F}}'\bar{{\varvec{M}}}\varvec{F}\varvec{\lambda } _i /T=\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\)
- (ii)
\({\varvec{\lambda }}'_i {\varvec{F}}'\bar{{\varvec{M}}}\varvec{v}_i /T=\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\);
- (iii)
\({\varvec{v}}'_i \bar{{\varvec{M}}}\varvec{v}_i /T={\varvec{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i /T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\);
- (iv)
\(T^{-1/2}{\varvec{X}}'_{\varvec{i}} \bar{{\varvec{M}}}_{\varvec{q}} \varvec{v}_{\varvec{i}} =\mathrm{O}_p (1)\) and \((NT)^{-1/2} \sum \nolimits _{i=1}^{N} {{\varvec{X}}'_{\varvec{i}} \bar{{\varvec{M}}}_{\varvec{q}} \varvec{v}_{\varvec{i}}} =\mathrm{O}_p (1)\).
Proof
-
(i)
First noting that \(\bar{{\varvec{H}}}=G\bar{{\varvec{P}}}+\bar{{\varvec{\varepsilon } }}^{*}\) and \(\bar{{\varvec{M}}}_{\varvec{q}} =\varvec{I}_T -G\bar{{\varvec{P}}}({\bar{{\varvec{P}}}}'{\varvec{G}}'\varvec{G}\bar{{\varvec{P}}})^{-}{\bar{{\varvec{P}}}}'{\varvec{G}}'\), we have \({\varvec{\lambda }}'_i {\varvec{F}}'\bar{{\varvec{M}}}\varvec{F}\varvec{\lambda } _i /T={\varvec{\lambda }}'_i {\varvec{F}}'\varvec{F}\varvec{\lambda }_i /T-{\varvec{\lambda }}'_i {\varvec{F}}'\bar{{\varvec{H}}}({\bar{{\varvec{H}}}}'\bar{{H}})^{-}{\bar{{H}}}'F\lambda _i /T= {\lambda }'_i {F}'\bar{{\varvec{M}}}_q \varvec{F}\varvec{\lambda } _i /T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT))^{-1/2}, \) where the last equality uses Lemma 1 (i) and (ii). When Assumption A5 is satisfied, using the results on generalized inverse, \(\bar{{\varvec{M}}}_q =\varvec{M}_g\). Since \(\varvec{F}\subset \varvec{G},\) then \(\bar{{\varvec{M}}}_q \varvec{F}=\varvec{M}_g \varvec{F}=0\). Therefore \(\varvec{{\lambda }}'_i {\varvec{F}}'\bar{{\varvec{M}}}\varvec{F}\varvec{\lambda } _i /T=\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\).
-
(ii)
Since \({\varvec{\bar{{H}}}}'\varvec{v}_i /T={\bar{{\varvec{P}'}}{G}'v}_i /T+{\bar{\varvec{\varepsilon }}^{*\prime }v}_i /T\), then \({\varvec{\bar{{H}}}'v_i} /T={\bar{{\varvec{P}'}}{G}'v}_i /T+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})\) follows from Lemma 1 (vi). Using the same line of reasoning as in Lemma 2 (i), the result is proved to be true.
-
(iii)
The result can be shown in the same way as (i) and (ii).
-
(iv)
Since \(v_i\) is independent of \(\bar{{\varvec{M}}}_q \) and \(\varvec{X}_i \), then \(E\left( {T^{-1/2}\varvec{{X}}'_i \bar{{\varvec{M}}}_q v_i } \right) =0\) and \(E\left( {T^{-1/2}\varvec{{X}}'_i \bar{{\varvec{M}}}_q v_i } \right) ^{2}=T^{-1}\sum \nolimits _{t=1}^T {a_{\textit{it}} E(v_{\textit{it}}^2 } )=\mathrm{O}_p (1)\) where \(a_{\textit{it}} \)’s are coefficients. Therefore, \(T^{-1/2}\varvec{{X}}'_i \bar{{\varvec{M}}}_q v_i =\mathrm{O}_p (1)\). Then it is easily to prove that \((NT)^{-1/2}\sum \nolimits _{i=1}^N {\varvec{{X}}'_i \bar{{\varvec{M}}}_q v_i } =\mathrm{O}_p (1)\).
\(\square \)
Lemma 3
Suppose that Assumptions A1–A5 hold. Then under \(\varvec{H}_0 \), we have
- (i)
\(\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}} _i^2 ={v}'_i \bar{{\varvec{M}}}_q v_i /(T-k-2)+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2})=\sigma _i^2 +\mathrm{O}_p (T^{-1/2})+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((TN)^{-1/2})\);
- (ii)
\(N^{-1/2}\sum \nolimits _{i=1}^N {\varvec{{\xi }}'_{\textit{iT}} \varvec{Q}_{\textit{iT}}^{-1} \varvec{\xi } _{\textit{iT}} /\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}}} _i^2 =N^{-1/2}\sum \nolimits _{i=1}^N {\tilde{z}_{\textit{iT}} } +\mathrm{O}_p (N^{1/2}T^{-1})+\mathrm{O}_p (TN^{-3/2})+\mathrm{O}_p (N^{-1/2^{}})+\mathrm{O}_p (T^{1/2}N^{-1})+\mathrm{O}_p (T^{-1/2})\),
where \(\tilde{z}_{\textit{iT}} =(T-r-1)\varvec{{v}}'_i \varvec{D}_i \varvec{v}_i /\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i \) with \(\varvec{D}_i =\bar{{\varvec{M}}}_q \varvec{X}_i (\varvec{{X}}'_i \bar{{\varvec{M}}}_q \varvec{X}_i )^{-1}\varvec{{X}}'_i \bar{{\varvec{M}}}_q\).
Proof
-
(i)
According to the definition of \(\overset{\scriptscriptstyle \smile }{{\varvec{U}}} _i \) in Sect. 2.3, it is easily seen that
$$\begin{aligned} \overset{\scriptscriptstyle \smile }{{\varvec{U}}} {\prime }_i \overset{\scriptscriptstyle \smile }{{\varvec{U}}} _i= & {} \left( {\bar{{\varvec{M}}}X_i (\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} )+\bar{{\varvec{M}}}(F\lambda _i +v_i )} \right) ^{\prime }\\&\quad \times \,\left( {\bar{{\varvec{M}}}X_i (\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} )+\bar{{\varvec{M}}}(F\lambda _i +v_i )} \right) \\= & {} \left( {T(\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} {)}'Q_{\textit{iT}} (\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} )} \right. +2(\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} {)}'T^{1/2}\xi _{\textit{iT}} \\&\quad +\,2(\beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} {)}' {X}'_i \bar{{\varvec{M}}}F\lambda +{v}'_i \bar{{\varvec{M}}}v_i +{\lambda }'_i F^{\prime }\bar{{\varvec{M}}}F\lambda _i +\left. {2{\lambda }'_i F^{\prime }\bar{{\varvec{M}}}v_i } \right) . \end{aligned}$$Under \(\varvec{H}_0 \),
$$\begin{aligned} \beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} =(NT)^{-1/2}Q_{NT}^{-1} \xi _{NT} +(NT)^{-1}Q_{NT}^{-1} \sum \nolimits _{i=1}^N {{X}'_i \bar{{\varvec{M}}}F\lambda _i } . \end{aligned}$$$$\begin{aligned} \varvec{\xi } _{NT} =\mathrm{O}_p (1)+\mathrm{O}_p (T^{1/2}N^{-1/2}). \end{aligned}$$Combining Assumptions A4–A5 and Lemma 1 (iii), we have
$$\begin{aligned} \beta _i -\overset{\scriptscriptstyle \frown }{{\varvec{\beta }}} _{CCEP} =\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2}). \end{aligned}$$Using Lemmas 1 and 2, we establish that
$$\begin{aligned} \overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}} _i^2 =\overset{\scriptscriptstyle \smile }{{\varvec{U}}} {\prime }_i \overset{\scriptscriptstyle \smile }{{\varvec{U}}} _i /(T-k-2)={v}'_i \bar{{\varvec{M}}}_q v_i /(T-k-2)+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((NT)^{-1/2}). \end{aligned}$$In addition, we can further develop the relationship between \(\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}} _i^2 \) and \(\varvec{\sigma }_i^2 \) . Since \(\varvec{v}_i \) is independent of \(\bar{{\varvec{M}}}_q \), combining the law of iterated expectation and Theorem 1 in Bao and Ullah (2010), we have
$$\begin{aligned} E(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )= & {} \sigma _i^2 Tr(\bar{{\varvec{M}}}_q )=\sigma _i^2 (T-r-1),\\ E(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )^{2}= & {} E(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )E(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )+2Tr(\bar{{\varvec{M}}}_q \bar{{\varvec{M}}}_q )\\= & {} \sigma _i^4 (T-r-1)^{2}+2(T-r-1). \end{aligned}$$Hence, for fixed k,
$$\begin{aligned} E[\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i /(T-k-2)-\sigma _i^2 ]^{2} =\mathrm{O}_p (T^{-1}), \end{aligned}$$which implies
$$\begin{aligned} \varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i /(T-k-2)=\sigma _i^2 +\mathrm{O}_p (T^{-1/2}). \end{aligned}$$Therefore
$$\begin{aligned} \overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}} _i^2 =\sigma _i^2 +\mathrm{O}_p (T^{-1/2})+\mathrm{O}_p (N^{-1})+\mathrm{O}_p ((TN)^{-1/2}). \end{aligned}$$ -
(ii)
Combining Lemmas 1 and 3, we get
$$\begin{aligned} N^{-1/2}\sum \nolimits _{i=1}^N{\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}}} _i^{-2} \varvec{{\xi }}'_{\textit{iT}} \varvec{Q}_{\textit{iT}}^{-1} \varvec{\xi } _{\textit{iT}}= & {} N^{-1/2}(T-k-2)/(T-r-1)\\&\quad \times \,\sum \nolimits _{i=1}^N (T-r-1)\varvec{{\xi }}'_{\textit{iT}} \varvec{Q}_{\textit{iT}}^{-1} \varvec{\xi } _{\textit{iT}} /(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )\\&\quad +\,\mathrm{O}_p (N^{-1/2})+\mathrm{O}_p (T^{-1/2})\\= & {} N^{-1/2}\sum \nolimits _{i=1}^N {(T-r-1)\varvec{{\xi }}'_{\textit{iT}} \varvec{Q}_{\textit{iT}}^{-1} \varvec{\xi } _{\textit{iT}} /(\varvec{{v}}'_i \bar{{\varvec{M}}}_q \varvec{v}_i )}\\&\quad +\,\mathrm{O}_p (N^{1/2}T^{-1})+\mathrm{O}_p (N^{-1/2})+\mathrm{O}_p (T^{-1/2})\\= & {} N^{-1/2}\sum \nolimits _{i=1}^N {\tilde{z}_{\textit{iT}} } +\mathrm{O}_p (TN^{-3/2})+\mathrm{O}_p (T^{1/2}N^{-1})\\&\quad +\,\mathrm{O}_p (N^{1/2}T^{-1})+\mathrm{O}_p (N^{-1/2})+\mathrm{O}_p (T^{-1/2}). \end{aligned}$$The first equality follows from Lemma 3 (i), and the last equality follows from Lemma 1 (iv)–(v).\(\square \)
Lemma 4
Suppose that Assumptions A1–A5 hold, then under \(\varvec{H}_0 \)
Proof
According to the definition of \(S_{lm}^f \) in Sect. 2.3, we have
For the last five terms in the above equality, first note \(\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}} _i^2 =\mathrm{O}_p (1)\), \(\varvec{Q}_{\textit{iT}} =\mathrm{O}_p (1)\), and \(\varvec{\xi } _{NT} =\mathrm{O}_p (1)+\mathrm{O}_p (T^{1/2}N^{-1/2})\). Using Lemmas 1 and 3, it follows that
Using Lemma 3 (ii) and the above five equations, we have
Lemma 5
Let \(\varvec{\Phi } \) be a \(T\times T\) symmetric matrix and \(\varvec{\Gamma } \) a positive definite \(T\times T\) matrix, and suppose that \(\varvec{\upsilon }\sim {\textit{IID}}(0,\varvec{I}_T )\), \(\varvec{\upsilon }=(\varvec{\upsilon }_1 ,\varvec{\upsilon }_2 ,\ldots ,\varvec{\upsilon }_T )'\). Denote the pth cumulant of \(\varvec{{\upsilon }'\varvec{\Gamma } \varvec{\upsilon }}\) by \(\kappa _p \), and the \(\gamma =1+m\) order, \(\delta =r+m\) degree generalized cumulant of \((\varvec{{\upsilon }'\Phi \upsilon })^{r}\) and \(\varvec{{\upsilon }'\Gamma \upsilon }\) by \(\kappa _{rm} \) and assume that the following conditions hold:
Condition A: for \(p=1,2,\ldots ,\kappa _p =\mathrm{O}(T)\); Condition B: \(r=1,2,\ldots ,\kappa _{r0} =E(\varvec{{\upsilon }'\Phi \upsilon })^{r}=\mathrm{O}(T^{r})\);
Condition C: \(r,m=1,2,\ldots ,\kappa _{rm} =\mathrm{O}(T^{l})\) with \(l\le r\).
Then the Laplace approximate expansion for the rth moment of \(\varvec{{\upsilon }'\Phi \upsilon }/\varvec{{\upsilon }'\Gamma \upsilon }\) is given by
where \(\phi _{1T} =\frac{r(r+1)}{2}\left[ {\frac{E[(\varvec{{\upsilon }'\Phi \upsilon })^{r}]\kappa _2 }{[E(\varvec{{\upsilon }'\Gamma \upsilon })]^{r+2}}} \right] -r\frac{\kappa _{r1} }{[E(\varvec{{\upsilon }'\Gamma \upsilon })]^{r+1}}\),
And \(\kappa _{r1} =E[(\varvec{{\upsilon }'\Phi \upsilon })^{r}\varvec{{\upsilon }'\Gamma \upsilon }]-E[(\varvec{{\upsilon }'\Phi \upsilon })^{r}]E(\varvec{{\upsilon }'\Gamma \upsilon })\),
Proof
See Lieberman (1994). \(\square \)
Remark
Pesaran and Yamagata (2008) extends the results that allow \(\varvec{\Gamma } \) being a semi-positive definite matrix.
Appendix 2: Proof of Theorems 1–2 and Corollary 1
Proof of Theorem 1
We first consider \(E(\tilde{z}_{\textit{iT}} )\) and \(Var(\tilde{z}_{\textit{iT}}\) ). By definition, \(\bar{{\varvec{M}}}_q \) and \(\varvec{D}_i \) are all idempotent matrices and independent of \(v_i \). Using results of Bao and Ullah (2010), we have \(E[v_i ^{\prime }\varvec{D}_i \varvec{v}_i /\sigma _i^2 ]=Tr(\varvec{D}_i )=k\), \(E[v_i ^{\prime }\bar{{\varvec{M}}}_q \varvec{v}_i /\sigma _i^2 ]=Tr(\bar{{\varvec{M}}}_q )=T-r-1\). Conditions A, B and C of Lemma 5 are satisfied. After some tedious algebra, we have \(\phi _{i1T} =\mathrm{O}_p (T^{-1})\), and \(\phi _{i2T} =\mathrm{O}_p (T^{-2})\). Hence, \(E(\tilde{z}_{\textit{iT}} )=k+\mathrm{O}_p (T^{-1})\). Similarly, we can obtain \(Var(\tilde{z}_{\textit{iT}} )=2k+\mathrm{O}_p (T^{-1})\). Then by Lemma 4,
Applying central limit theorem (CLT) to the first term of the last equality above, we finally establish that
\(LM_s^c \Rightarrow N(0,1),\) as \((N,T)\rightarrow \infty \) such that \(T^{2/3}N^{-1}\rightarrow 0\) and \(T^{2}N^{-1}\rightarrow \infty \). \(\square \)
Proof of Theorem 2
For the modified PY test, under \(\varvec{H}_0 \), we have
Use the definition of \(\tilde{S}_{cce} \), it is easily seen that
Combining Lemmas 1–3, it follows that
Using Eqs. (A.1)–(A.4) and the above three equations, we obtain
\(\square \)
Proof of Corollary 1
Under \(H_{1,NT} \),
where \(\gamma _{iNT} =\bar{{\varvec{M}}}\varvec{v}_i +\bar{{\varvec{M}}}\varvec{F}\varvec{\lambda } _i -(NT)^{-1}\bar{{\varvec{M}}}\varvec{X}_i \varvec{Q}_{NT}^{-1} \sum \nolimits _{i=1}^N {({\varvec{X}}'_i \bar{{\varvec{M}}}\varvec{v}_i +{\varvec{X}}'_i \bar{{\varvec{M}}}\varvec{F}\varvec{\lambda } _i )} \), and \(\varvec{\chi } _{iNT} =N^{-1/4}T^{-1/2}\bar{{\varvec{M}}}\varvec{X}_i \big [\varvec{\delta } _i -\varvec{Q}_{NT}^{-1} \tilde{\varvec{\delta } }_x \big ]\) with \(\tilde{\varvec{\delta } }_x =N^{-1}\sum \nolimits _{i=1}^N {\varvec{Q}_{\textit{iT}} \varvec{\delta } _i } =O_p (1)\). Then
For the first term in the above equality, we use the same line of reasoning in the proof of Lemma 4 and establish that
For the second term,
And for the last term, using Lemmas 1 (v) and 2 (iv), we get
Therefore
and
Finally, apply the CLT to the first term of the last equality above and conclude that \(LM_s^c \Rightarrow N(\psi _{NT} /\varvec{\sqrt{2k}},1),\) as \((N,T)\rightarrow \infty \) such that \(T^{2/3}N^{-1}\rightarrow 0\) and \(T^{2}N^{-1}\rightarrow \infty \), where \(\psi _{NT} =\mathop {\lim }\limits _{N\rightarrow \infty } N^{-1}\sum \nolimits _{i=1}^N {\overset{\scriptscriptstyle \smile }{{\varvec{\sigma }}}} _i^{-2} \big [\varvec{\delta } _i -\varvec{Q}_{NT}^{-1} \tilde{\varvec{\delta } }_x \big ]'\varvec{Q}_{\textit{iT}} \big [\varvec{\delta } _i -\varvec{Q}_{NT}^{-1} \tilde{\varvec{\delta } }_x \big ] \ge 0\). \(\square \)
Rights and permissions
About this article
Cite this article
Xu, F., He, Z. Testing slope homogeneity in panel data models with a multifactor error structure. Stat Papers 61, 201–224 (2020). https://doi.org/10.1007/s00362-017-0929-1
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-017-0929-1