Testing panel cointegration with unobservable dynamic common factors that are correlated with the regressors

Abstract

The paper proposes statistics to test the null hypothesis of no cointegration in panel data when common factors drive the cross-sectional dependence. We focus on the case in which regressors and the common factors are correlated, although the uncorrelated case is also discussed. Both endogenous and strictly exogenous regressors are considered. The test statistics are shown to have limiting distributions independent of the common factors, making it possible to pool the individual statistics. Simulations indicate that the proposed procedures have good finite sample performance. JournalThe Econometrics Panel cointegration with common factors correlated with regressors

Fiscal policy and TFP in the OECD: measuring direct and indirect effects

“…In line with the results in Kao (1999) and Phillips and Moon (1999), for a model with no cross-sectional dependence, and in Banerjee and Carrion-i-Silvestre (2011), for a model with cross-sectional dependence as in equation (6), we conjecture that the CCEPnl estimator yields consistent estimates for the homogenous coefficients β and λ and therefore, using equation (15), also for Ft 8 . This implies that we can obtain a consistent estimate for the composite 7 Using the PANIC approach to testing for panel cointegration in the presence of common factors has also been suggested by Gengenbach et al (2006), Banerjee and Carrion-i-Silvestre (2006) and Bai and Carrion-i-Silvestre (2013). The main difference between these approaches and ours lies in the estimation of the unknown coefficients in the cointegrating relation, for which we use the CCEP estimator while the above references estimate a model in first-differences with the common factors and factor loadings estimated using principal components.…”

Novel panel cointegration tests emending for cross-section dependence withNfixed

“…Then, we have for ≥ 0, Using Lemma A.3, it is not difficult to see that R T = o p (1). As a result, we obtain the bias.…”

“…For i = 2, by Cauchy-Schwarz inequality, we have (1) in exactly the same manner as the proof for i = 1.…”

“…The main techniques employed to account for cross-sectional dependence are: (1) bootstrapping which generally accommodates cross-sectional dependence of general form, (2) common factors based approaches which very often requires the estimation of the number of factors and the factor loadings (3) non-linear instrumental variables, (4) rank tests which is the most recent method, and finally (5) long-run variance which could be considered as a nonparametric approach. The use of bootstrapping methods was pioneered by Maddala and Wu (1999) followed by many who applied the same technique but all of them did not establish the validity of their bootstrap.…”

Stochastic Frontier Models for Long Panel Data Sets: Measurement of the Underlying Energy Efficiency for the OECD Countries

“…Recently, a growing literature on panel time series econometrics has introduced new methods for estimating long, possibly non-stationary panels, also allowing for contemporaneous correlation in the errors (Bai and Ng (2004), Pesaran (2006), Kapetanios and Pesaran (2007), Bai (2009), and Bai and Carrion-i-Silvestre (2013)). In this paper, we draw from this literature to propose a new general approach for estimating stochastic frontier models, suitable for long panel data sets.…”