Unobserved heterogeneity in panel time series models (original) (raw)
Related papers
Journal of Applied Econometrics, 2007
The papers included in this special issue are primarily concerned with the problem of cross section dependence and heterogeneity in the analysis of panel data models and their relevance in applied econometric research. Cross section dependence can arise due to spatial or spill over effects, or could be due to unobserved (or unobservable) common factors. Much of the recent research on non-stationary panel data have focussed on this problem. It was clear that the first generation panel unit root and cointegration tests developed in the 1990's, which assumed cross-sectional independence, are inadequate and could lead to significant size distortions in the presence of neglected cross-section dependence. Second generation panel unit root and cointegration tests that take account of possible cross-section dependence in the data have been developed, see the recent surveys by Choi (2006) and Breitung and Pesaran (2007). The papers by Baltagi, Bresson and Pirotte, Choi and Chue, Kapetanios, and Pesaran in this special issue are further contributions to this literature. The papers by Fachin, and Moon and Perron are empirical studies in this area. Controlling for heterogeneity has also been an important concern for empirical researchers with panel data methods promising better handle on heterogeneity than cross-section data methods. The papers by Hsiao, Shen, Wang and Weeks, Pedroni and Serlenga and Shin are empirical contributions to this area.
The correlation between lagged endogenous regressor and the error term, in case of dynamic panel data models, causes the least squares dummy variable (LSDV) estimator to be biased and inconsistent. This problem persists even in case of heteroscedastic errors. In 2006, Bun and Carree, firstly addressed this situation and proposed a bias-corrected LSDV estimator. Earlier other authors have considered the case of simple (static) panel data models allowing heteroscedasticity, and have proposed several versions of estimated generalized least squares (EGLS) estimators using different ways, through which the variance components are estimated. In this study, we have customized them for the dynamic panel data models with cross-sectional heteroscedastic remainder errors and have analyzed their performance as compared to the bias-corrected LSDV estimator using Monte Carlo experiments. The experimental evidences showed that the proposed estimators, particularly extended Heteroscedastic Consistent Covariance Matrix (HCCM)-based EGLS estimators HGLS1s, are attractive choices in the sense of bias and mean squared error (MSE).
Panel Macroeconometric Modeling
SSRN Electronic Journal, 2014
This paper provides a selective survey of the panel macroeconometric techniques that focus on controlling the impact of "unobserved heterogeneity" across individuals and over time to obtain valid inference for "structures" that are common across individuals and over time. We consider issues of (i) estimating vector autoregressive models; (ii) testing of unit root or cointegration; (iii) statistical inference for dynamic simultaneous equations models; (iv) policy evaluation; and (v) aggregation and prediction.
Biases in GLS Estimators for Dynamic Panel Data Models allowing Cross- Sectional Heteroscedasticity
The inclusion of lagged dependent variable in the list of explanatory variables introduces the specific estimation problems even the generalized least squares estimator for the dynamic panel data models allowing cross sectional heteroscedasticity becomes biased and inconsistent. In this study, the analytical expressions for the inconsistency have been derived in the first order autoregressive case. A comparison between asymptotic bias and small sample simulated bias has also been carried out. The analytical biases emerged coincident with or a little above the small sample simulated biases. The closeness of the two types of biases mainly depends on coefficient of lagged dependent variable (Gamma) and the number of cross sectional units N.
The Robustness of Estimators for Dynamic Panel Data Models to Misspecification
The Singapore Economic Review, 2009
Transition from economic theory to a testable form of model invariably involves the use of certain "simplifying assumptions." If, however, these are not valid, misspecified models result. This article considers estimation of the dynamic linear panel data model, which often forms the basis of testable economic hypotheses. The estimators of such a model are frequently similarly based on certain assumptions which appear to be often untenable in practice. Here, the performance of these estimators is analyzed in scenarios where the theoretically required conditions are not met. Specifically, we consider three such instances of serial correlation of the idiosyncratic disturbance terms; correlation of the idiosyncratic disturbance terms and explanatory variables; and, finally, cross-sectional dependence (as a robustness check to these findings, we also consider correlations between observed and unobserved heterogeneity terms). The major findings are that the limited tests readily...
Panel Time Series. Review of the Methodological Evolution
RePEc: Research Papers in Economics, 2015
In this article, we discuss the econometric treatment of macropanels, also known as panel time series. This new approach rejects the assumption of slope homogeneity and handles nonstationarity. It also recognizes that cross-section dependence (that is, some correlation structure in the error term between units due to unobservable common factors) squanders efficiency gains by operating with a panel. This approach uses a new set of estimators known in the literature as the common correlated effect, which essentially consists of increasing the model to be fit by adding the averages of the individuals in each time t, of both the dependent variable and the specific regressors of each individual. We present two commands developed for the evaluation and treatment of cross-section dependence.
Some cautions on the use of panel methods for integrated series of macroeconomic data
The Econometrics Journal, 2004
Existing panel cointegration tests rule out cross-unit cointegrating relationships, while economic theory and empirical observation argue strongly in favour of their presence. Using an extensive set of simulation experiments, we show that both univariate and multivariate panel cointegration tests can be substantially oversized in the presence of crossunit cointegration. We also propose a test for cross-unit cointegration that performs well in practice and can be used to decide upon the usefulness of panel methods.
Autocorrelation and masked heterogeneity in panel data models estimated by maximum likelihood
Empirical Economics, 2012
In a panel data model with random effects, when autocorrelation in the error is considered, (Gaussian) maximum likelihood estimation produces a dramatically large number of corner solutions: the variance of the random effect appears (incorrectly) to be zero, and a larger autocorrelation is (incorrectly) assigned to the idiosyncratic component. Thus heterogeneity could (incorrectly) be lost in applications to panel data with customarily available time dimension, even in a correctly specified model. The problem occurs in linear as well as nonlinear models. This paper aims at pointing out how serious this problem can be (largely neglected by the panel data literature). A set of Monte Carlo experiments is conducted to highlight its relevance, and we explain this unpleasant effect showing that, along a direction, the expected log-likelihood is nearly flat. We also provide two examples of applications with corner solutions.
Mathematical Theory and Modeling, 2013
This paper investigates the efficiency of four methods of estimating panel data models (Pooling (OLS), First-Differenced (FD), Between (BTW) and Feasible Generalized Least Squares (FGLS)) when the assumptions of homoscedasticity, no autocorrelation and no collinearity are jointly violated. Monte-Carlo studies were carried out at different sample sizes, at varying degrees of heteroscedasticity, different levels of collinearity and autocorrelation all at different time periods. The results from this work showed that in small sample situation, irrespective of number of time length, FGLS estimator is efficient when heteroscedasticity is severe regardless of levels of autocorrelation and multicollinearity. However, when heteroscedasticity is low or mild with moderate autocorrelation level, both FD and FGLS are efficient, while BTW performs better only when there is no autocorrelation and low degree of heteroscedasticity. However, in large sample with short time periods, both FD and BTW could be used when there is no autocorrelation and low degree of heteroscedasticity, while FGLS is preferred elsewise. Meanwhile, Pooling estimator performs better when the assumptions of homoscedasticity, independent of error terms and orthogonality among the explanatory variables are justifiably valid.
LSE Research Online Documents on Economics, 2019
This paper proposes efficient estimation methods for panel data limited dependent variables (LDV) models possessing a variety of complications: non-ignorable persistent heterogeneity; contemporaneous and intertemporal endogeneity; and observable and unobservable dynamics. An important problem handled by the novel framework of this paper involves contemporaneous and intertemporal simultaneity caused by social strategic interactive effects or contagion across economic agents over time. The paper first shows how a simple modification of estimators based on the Random Effects principle can preserve the consistency and asymptotic efficiency of the method in panel data despite non-ignorable persistent heterogeneity driven by correlations between the individual-specific component of the error term and the regressors. The approach is extremely easy to implement and allows straightforward classical and omnibus tests of the significance of such correlations that lie behind the non-ignorable p...