Saddlepoint Approximations for Spatial Panel Data Models

We develop new higher-order asymptotic techniques for the Gaussian maximum likelihood estimator of the parameters in a spatial panel data model, with fixed effects, time-varying covariates, and spatially correlated errors. We introduce a new saddlepoint density and tail area approximation to improve on the accuracy of the extant asymptotics. It features relative error of order O(m to the power of -1) for m = n(T -1) with n being the cross-sectional dimension and T the time-series dimension. The main theoretical tool is the tilted-Edgeworth technique. It yields a density approximation that is always non-negative, does not need resampling, and is accurate in the tails. We provide an algorithm to implement our saddlepoint approximation and we illustrate the good performance of our method via numerical examples. Monte Carlo experiments show that, for the spatial panel data model with fixed effects and T = 2, the saddlepoint approximation yields accuracy improvements over the routinely applied first-order asymptotics and Edgeworth expansions, in small to moderate sample sizes, while preserving analytical tractability. An empirical application on the investment-saving relationship in OECD countries shows disagreement between testing results based on first-order asymptotics and saddlepoint techniques, which questions some implications based on the former.