Approximately Exact Inference in Dynamic Panel Models

Article · January 2006with10 Reads
Source: RePEc
This paper develops a general method for conducting exact small-sample inference in models which allow the estimator of the (scalar) parameter of interest to be expressed as the root of an estimating function, and which is particularly simple to implement for linear models with a covariance matrix depending on a single parameter. The method involves the computation of tail probabilities of the estimating function. In the context of dynamic panel models, both the least squares and maximum likelihood paradigms give rise to estimating functions involving sums of ratios in quadratic forms in normal variates, the distribution of which cannot be straightforwardly computed. We overcome this obstacle by deriving a saddlepoint approximation that is both readily evaluated and remarkably accurate. A simulation study demonstrates the validity of the procedure, and shows the resulting estimators to be vastly superior over existing ones
  • [Show abstract] [Hide abstract] ABSTRACT: This paper presents specification tests that are applicable after estimating a dynamic model from panel data by the generalized method of moments (GMM), and studies the practical performance of these procedures using both generated and real data. Our GMM estimator optimally exploits all the linear moment restrictions that follow from the assumption of no serial correlation in the errors, in an equation which contains individual effects, lagged dependent variables and no strictly exogenous variables. We propose a test of serial correlation based on the GMM residuals and compare this with Sargan tests of over-identifying restrictions and Hausman specification tests.
    Article · Apr 1991
  • [Show abstract] [Hide abstract] ABSTRACT: Observations on N cross-section units at T time points are used to estimate a simple statistical model involving an autoregressive process with an additive term specific to the unit. Different assumptions about the initial conditions are (a) initial state fixed, (b) initial state random, (c) the unobserved individual effect independent of the unobserved dynamic process with the initial value fixed, and (d) the unobserved individual effect independent of the unobserved dynamic process with initial value random. Asymptotic properties of the maximum likelihood and “covariance” estimators are obtained when T → ∞ and when N → ∞. The relationship between the pseudo and conditional maximum likelihood estimators is clarified. A simple consistent estimator that is independent of the initial conditions and the way in which T or N → ∞ is also suggested.
    Article · Sep 1980
  • [Show abstract] [Hide abstract] ABSTRACT: For parametric models it is shown in general that by adjusting the mean and variance of the signed log likelihood ratio for a single parameter of interest ψ one obtains a statistic which is asymptotically standard normally distributed to order O(n-3/2), under repeated sampling. This statistic may also be used as an ancillary in the associated problem of drawing inference on the complementary parameter χ for known value of ψ in which case it entails accuracy to the same order of a simple formula (Barndorff-Nielsen, 1983) for the conditional distribution of the maximum likelihood estimator of χ. By iterated application, these results are extended to the case of multivariate parameters of interest. The asymptotic normality result may be used to set confidence regions for the parameter of interest which are correct to order O(n-3/2), conditionally as well as unconditionally. Several examples are discussed. In the course of the argument the concept of the affine ancillary (Barndorff-Nielsen, 1980) is extended from curved exponential families to rather general models.
    Article · Aug 1986
Show more