1,721,000 research outputs found
Cross-sectional and spatial dependence in panels
No full textEconometricians have recently turned towards the problems posed by cross-sectional dependence across individuals, which may range from inefficiency of
the standard estimators and invalid inference to inconsistency. Panel data are especially useful in this respect, as their double dimensionality allows robust approaches to general cross-sectional dependence.
A general object oriented approach to robust inference is available in the R system (Zeileis, 2004), for which all that’s needed are coefficients β and robust covariance estimators. An useful implementation is, e.g., in linear hypotheses
estimators for vcov(β).
testing (see Fox, package car). The plm package for pael data econometrics
already has features for heteroskedasticity– and serial correlation–robust inference (Croissant and Millo, forthcoming). If cross-sectional dependence is detected, using a robust covariance estimator
allows valid inference. I describe the implementation in the plm package for panel
data econometrics of:
- tests for detecting cross-sectional dependence in the errors of a panel model
(Friedman 1928, Frees 1995, Pesaran 2004)
- robust estimators of covariance matrices for doing valid inference in the presence of cross-sectional dependence (White 1980, Beck and Katz 1995, Driscoll and Kraay 1998). If a particular spatial structure is assumed, this allows a parsimonious characterization of spatial dependence but, on the converse, the resulting models are computationally expensive to estimate, all the more so in the panel case. Efficient ML estimators for spatial models on a cross-section (Anselin 1988) are implemented in the spdep package (Bivand et al.). I describe implementation in a forthcoming package of
- marginal and conditional LM tests for spatial correlation, serial correlation and random effects (Baltagi, Song, Jung and Koh 2007)
- ML estimators for panel models including spatial lags, spatial errors and possibly serial correlation (Anselin 1988, Elhorst 2003, Baltagi, Song, Jung and Koh 2007) I illustrate the functionalities by application to Munnell’s (1990) data on 48 USA states observed over 17 years. On an ordinary desktop machine, the estimators and tests all take under one minute (few seconds for the basic ones). The ML
approach is nevertheless structurally limited to a few hundred cross-sectional ob-
servations, so further work is warranted to implement Kapoor, Kelejian and Prucha
(2007)’s GM approach, which promises to handle problems with n in the thousands
Robust standard errors for panel data: a general, software-oriented framework
No full textA comprehensive, modular and flexible framework is described for estimation of robust standard errors in panel data. Heteroskedasticity and autocorrelation robust estimators are brought together with the SCC mixing-fields based estimator, the unconditional PCSE estimator and the recent double-clustering approach, trying to bring together the applied literatures in macroeconometrics, finance, political science and accounting by demonstrating the common features of these apparently different approaches. The covariance estimators are integrated in the R package 'plm' and allow robust specification and restriction testing over a number of different panel models
Detecting Spatial Dependence in Panels with Common Factors through Permutations of Pesaran's CD(p) Test
No full textIn the spatial econometrics literature, spatial error dependence is characterized by spatial autoregressive processes, which relate every observation in the cross-section to any other with distance-decaying intensity: i.e., dependence obeys Tobler's First Law of Geography (''everything is related to everything else, but near things are more related than distant things''). In the literature on factor models, on the converse, the degree of correlation between cross-sectional units depends only on factor loadings.
Standard spatial correlation tests have power against both types of dependence, while the economic meaning of the two can be much different; so it may be useful to devise a test for detecting ''distance-related'' dependence in the presence of a ''factor-type'' one.
Pesaran's CD is a test for global cross-sectional dependence with good properties. The CD(p) variant only takes into account p-th order neighbouring units to test for local cross-sectional dependence. The pattern of CD(p) as p increases can be informative about the type of dependence in the errors, but the test power changes as new pairs of observations are taken into account.
I propose a permutation test based on the values taken by the CD(p) test under permutations of the neighbourhood matrix, i.e. when ''reshuffling the neighbours''. I provide Montecarlo evidence of it being able to tell the presence of spatial-type dependence in the errors of a typical spatial panel irrespective of the presence of an unobserved factor structure
The income elasticity of non-life insurance: a reassessment
No full textIn aggregate insurance regressions at the country level, the question whether
insurance is a normal or superior good translates into whether income elasticity is significantly greater than one or not. Twenty-five years after a seminal article, I reassess the income elasticity of nonlife insurance by means of homogeneous and heterogeneous versions of the common correlated effects estimator, controlling for common factors and individual trends and characterizing the average behavior of insurance markets while allowing for individual heterogeneity. The evidence supports the existence of a cointegrating behavior between insurance consumption and GDP and the view of nonlife insurance as a normal good
The S-curve and Reality
No full textWe challenge the common wisdom that the income elasticity of insurance is higher, ceteris paribus, in developing countries (the so-called S-curve hypothesis). Focusing on non-life insurance, we show that the available evidence is contradictory and heavily dependent on methodology. Based on a recent approach to consistent inference on the income elasticity of insurance, we show counterexamples to the theory. Although not supporting it in general, we argue that it could still be relevant for explaining the behaviour of particular lines of business
ML estimation of spatially and serially correlated panels with random effects: an estimation framework and a software implementation
No full textIt is described a procedure for maximum likelihood estimation of panel models incorporating: random effects and spatial dependence in the error terms; and/or a spatially lagged dependent variable; and possibly also a serial dependence structure in the remainder of the error term. We start by sketching a taxonomy of spatial panel models, beginning with the two basic random effects (RE) specifications used in the literature: the spatial
autoregressive (SAR) RE model containing a spatially lagged dependent variable and a group-specific, time-invariant component in the error term, and the spatial error (SEM) RE model, with both a group-specific component and a spatial dependence structure in the error term. Extending the
SEM specification, an encompassing model allowing for serial correlation in the residuals is considered. Restrictions of the full model give rise to 18 different specifications. It is discussed an efficient implementation of the estimation procedure in R, to be added to the splm package for estimation and testing of spatial panel models, illustrating it through some well-known examples from the literature
Spatial Panel Data Models
No full textA synthetic treatment of spatial panel data models with either fixed or random effects, estimated by either maximum likelihood or generalized moments, with examples in
D in the presence of spillovers, revisited
Full text linkThis is both a replication of Eberhardt et al. (Review of Economics and Statistics, 2013, 95(2), 436–448) using different software, and a critical extension and diagnostic reassessment of the original results. The main findings of the paper are confirmed and sometimes reinforced. We point out some inconsistencies, in particular in the calculation of standard errors for the common correlated effects pooled model; we extend the diagnostic checks; lastly, in the spirit of the original contribution, we show how local cross-sectional dependence diagnostics can be used to provide a first assessment of the direction of spillovers. We provide complete replication code in open source R
A simple randomization test for spatial correlation in the presence of common factors and serial correlation
No full textA randomization test is proposed for detecting spatial dependence in panel models with cross-sectional dependence induced by an unobserved common factor structure. Spatial dependence is related to the position of observations in space while cross-sectional dependence is generally not; yet spatial correlation tests have power against both. Permuting the pairs of neighbouring observations in the proximity matrix yields a simple spatial dependence test which is robust to the presence of non-spatial cross-sectional correlation, serial correlation and can accommodate short and unbalanced panels. The proposed procedure is evaluated and compared to alternatives through Monte Carlo simulation; it is then illustrated by an application to recent research on technology spillovers. A user-friendly R implementation is provided
splm: Econometric analysis of panel data
No full textWe illustrate the new splm package, aimed at providing a comprehensive resource for spatial panel econometrics. The package fills a gap in applied practice, as the relevant estimators and tests are well established in the literature but to date they lack user-friendly and widely available software implementations.
Building on the infrastructure for spatially referenced data in package spdep, we provide estimators for the standard panel models in the spatial econometrics literature: fixed and random effects with either a
spatial lag or spatial correlation in the error term, based on both the concurrent approaches prevailing in the literature, i.e. the Maximum Likelihood framework pioneered by Anselin (1988) and the Generalized Moments framework of Kapoor, Kelejian and Prucha (2007).
Some of the model estimation procedures are generalized to the case of spatially and serially correlated
error terms. GM estimators for systems of equations are also available.
We also provide the Lagrange Multiplier joint, marginal and conditional specification tests from the work
of Baltagi et al. (2003, 2007).
The user interface aims at consistency w.r.t. the spatial (non-panel) estimators in package spdep and the panel (non-spatial) estimators in package plm.
We briefly discuss code optimization aspects of the computationally heavy Maximum Likelihood routines
that have up to now hindered the practical implementation of these estimators. The GM approach, on its part, yields very fast estimators that can be applied to comparatively big datasets.
We conclude with an empirical illustration on a well-known data set from the panel data literature
- …
