Spatial and Spatiotemporal Econometrics: Volume 18

For this discussion, assume there are n sample observations of the dependent variable y at unique locations. In spatial samples, often each observation is uniquely associated with a particular location or region, so that observations and regions are equivalent. Spatial dependence arises when an observation at one location, say y i is dependent on “neighboring” observations y j, y j∈ϒi. We use ϒi to denote the set of observations that are “neighboring” to observation i, where some metric is used to define the set of observations that are spatially connected to observation i. For general definitions of the sets ϒi,i=1,…,n, typically at least one observation exhibits simultaneous dependence, so that an observation y j, also depends on y i. That is, the set ϒj contains the observation y i, creating simultaneous dependence among observations. This situation constitutes a difference between time series analysis and spatial analysis. In time series, temporal dependence relations could be such that a “one-period-behind relation” exists, ruling out simultaneous dependence among observations. The time series one-observation-behind relation could arise if spatial observations were located along a line and the dependence of each observation were strictly on the observation located to the left. However, this is not in general true of spatial samples, requiring construction of estimation and inference methods that accommodate the more plausible case of simultaneous dependence among observations.

Baltagi and Li (2001) derived Lagrangian multiplier tests to jointly test for functional form and spatial error correlation. This companion paper derives Lagrangian multiplier tests to jointly test for functional form and spatial lag dependence. In particular, this paper tests for linear or log-linear models with no spatial lag dependence against a more general Box-Cox model with spatial lag dependence. Conditional LM tests are also derived which test for (i) zero spatial lag dependence conditional on an unknown Box-Cox functional form, as well as, (ii) linear or log-linear functional form given spatial lag dependence. In addition, modified Rao-Score tests are also derived that guard against local misspecification. The performance of these tests are investigated using Monte Carlo experiments.

From a theoretical point of view, a spatial econometric model can contain both a spatially lagged dependent variable (spatial lag) and a spatially autocorrelated error term (spatial error). However, such models are rarely used in practice. This is because (assuming a lattice model approach is used for both the spatial lag and spatial error) the model is difficult to estimate¹ unless the weight matrices are different for the spatial lag and the spatial error.

Within spatial econometrics a whole family of different spatial specifications has been developed, with associated estimators and tests. This lead to issues of model comparison and model choice, measuring the relative merits of alternative specifications and then using appropriate criteria to choose the “best” model or relative model probabilities. Bayesian theory provides a comprehensive and coherent framework for such model choice, including both nested and non-nested models within the choice set. The paper reviews the potential application of this Bayesian theory to spatial econometric models, examining the conditions and assumptions under which application is possible. Problems of prior distributions are outlined, and Bayes factors and marginal likelihoods are derived for a particular subset of spatial econometric specifications. These are then applied to two well-known spatial data-sets to illustrate the methods. Future possibilities, and comparisons with other approaches to both Bayesian and non-Bayesian model choice are discussed.

A Bayesian probit model with individual effects that exhibit spatial dependencies is set forth. Since probit models are often used to explain variation in individual choices, these models may well exhibit spatial interaction effects due to the varying spatial location of the decision makers. That is, individuals located at similar points in space may tend to exhibit similar choice behavior. The model proposed here allows for a parameter vector of spatial interaction effects that takes the form of a spatial autoregression. This model extends the class of Bayesian spatial logit/probit models presented in LeSage (2000) and relies on a hierachical construct that we estimate via Markov Chain Monte Carlo methods. We illustrate the model by applying it to the 1996 presidential election results for 3,110 U.S. counties.

The purpose of this paper is two-fold. First, on a theoretical level we introduce a series-type instrumental variable (IV) estimator of the parameters of a spatial first order autoregressive model with first order autoregressive disturbances. We demonstrate that our estimator is asymptotically efficient within the class of IV estimators, and has a lower computational count than an efficient IV estimator that was introduced by Lee (2003). Second, via Monte Carlo techniques we give small sample results relating to our suggested estimator, the maximum likelihood (ML) estimator, and other IV estimators suggested in the literature. Among other things we find that the ML estimator, both of the asymptotically efficient IV estimators, as well as an IV estimator introduced in Kelejian and Prucha (1998), have quite similar small sample properties. Our results also suggest the use of iterated versions of the IV estimators.

We formulate generalized maximum entropy estimators for the general linear model and the censored regression model when there is first order spatial autoregression in the dependent variable. Monte Carlo experiments are provided to compare the performance of spatial entropy estimators relative to classical estimators. Finally, the estimators are applied to an illustrative model allocating agricultural disaster payments.

The size and number of employment subcenters have increased in large metropolitan areas as the spatial distribution of jobs has become increasingly decentralized. Although employment decentralization is not a new phenomenon, only recently have concentrations of employment outside the central city begun to rival the traditional central business district (CBD) in size and scope. Because of this change, neither theoretical nor empirical models in urban economics now rely solely on the traditional monocentric city model of Muth (1969) and Mills (1972). Instead, recent research incorporates some version of a polycentric model, a trend that Anas et al. (1998) document in their excellent review article.

This paper presents a mixture of linear models (or hedonic regressions) for defining housing submarkets. Two different mixture models are considered: the first model allows all the regression coefficients to vary among the clusters (random coefficients); and the second model allows only the intercept term to change (random intercept). The model with a random intercept can be seen as a linear mixed model where the random effects distribution is estimated via non-parametric maximum likelihood (NPML). The models are illustrated using a real data set of 293 properties in Pamplona, Spain. These mixture models provide a classification of the dwellings into homogeneous groups that determine the structure of the submarkets.

We analyze spatio-temporal data on U.S. unemployment rates. For this purpose, we present a family of models designed for the analysis and time-forward prediction of spatio-temporal econometric data. Our model is aimed at applications with spatially sparse but temporally rich data, i.e. for observations collected at few spatial regions, but at many regular time intervals. The family of models utilized does not make spatial stationarity assumptions and consists in a vector autoregressive (VAR) specification, where there are as many time series as spatial regions. A model building strategy is used that takes into account the spatial dependence structure of the data. Model building may be performed either by displaying sample partial correlation functions, or automatically with an information criterion. Monthly data on unemployment rates in the nine census divisions of the U.S. are analyzed. We show with a residual analysis that our autoregressive model captures the dependence structure of the data better than with univariate time series modeling.

A common feature of certain kinds of data is a high level of statistical dependence across space and time. This spatial and temporal dependence contains useful information that can be exploited to significantly reduce the uncertainty surrounding local distributions. This chapter develops a methodology for inferring local distributions that incorporates these dependencies. The approach accommodates active learning over space and time, and from aggregate data and distributions to disaggregate individual data and distributions. We combine data sets on Kansas winter wheat yields – annual county-level yields over the period from 1947 through 2000 for all 105 counties in the state of Kansas, and 20,720 individual farm-level sample moments, based on ten years of the reported actual production histories for the winter wheat yields of farmers participating in the United States Department of Agriculture Federal Crop Insurance Corporation Multiple Peril Crop Insurance Program in each of the years 1991–2000. We derive a learning rule that combines statewide, county, and local farm-level data using Bayes’ rule to estimate the moments of individual farm-level crop yield distributions. Information theory and the maximum entropy criterion are used to estimate farm-level crop yield densities from these moments. These posterior densities are found to substantially reduce the bias and volatility of crop insurance premium rates.