The Determinants of Regional Freight Transport: A Spatial, Semiparametric Approach

Abstract

In the context of modeling regional freight the four‐stage model is a popular choice. The first stage of the model, freight generation and attraction, however, suffers from three shortcomings: first of all, it does not take spatial dependencies among regions into account, thus potentially yielding biased estimates. Second, there is no clear consensus in the literature as to the choice of explanatory variables. Second, sectoral employment and gross value added are used to explain freight generation, whereas some recent publications emphasize the importance of variables which measure the amount of logistical activity in a region. Third, there is a lack of consensus regarding the functional form of the explanatory variables. Multiple recent studies emphasize nonlinear influences of selected variables. This article addresses these shortcomings by using a spatial variant of the classic freight generation and attraction models combined with a penalized spline framework to model the explanatory variables in a semiparametric fashion. Moreover, a Bayesian estimation approach is used, coupled with a penalized Normal inverse‐Gamma prior structure, to introduce uncertainty regarding the choice and functional form of explanatory variables. The performance of the model is assessed on a real‐world example of freight generation and attraction of 258 European NUTS‐2 level regions, covering 25 European countries.

MCMCINLA estimation of varying coefficient spatial lag model—A study of China’s economic development in the context of population aging

“…Chen et al (2015, 2016) [ 16 , 17 ] successively proposed Two-Stage Least Squares estimation and General Moment estimation for semiparametric panel spatial lag models. Krisztin (2017) [ 18 ] combined penalized splines with Bayesian methods to propose a new Bayesian semiparametric estimation method for semiparametric spatial lag models. Li and Chen (2018) [ 19 ] constructed a nonparametric spatial lag model with random independent variables and gave a General Moment estimation for this model.…”

Bayesian Estimation of Partially Linear Additive Spatial Autoregressive Models with P-Splines

“…To capture the underlying relationships between the response variables and their associated covariates, semiparametric spatial autoregressive models have gained much attention in the literature of statistics and econometrics. For example, Su and Jin [22] discussed the quasi-likelihood estimator for the semiparametric partially linear spatial autoregressive model; Su [23] proposed semiparametric generalized method of moment (GMM) estimation of the semiparametric spatial autoregressive model; Sun et al [24] developed a profile likelihood estimator for the semiparametric spatial dynamic model; Chen et al [25] presented a two-step Bayesian approach for the semiparametric spatial autoregressive model; Wei and Sun [26] considered the semiparametric GMM estimation of a spatial model with space-varying coefficients; Hoshino [27] proposed a semiparametric series GMM estimator for the semiparametric spatial autoregressive model; Krisztin [28] studied a novel Bayesian semiparametric estimation for the penalized spline spatial autoregressive model; and Wei et al [29] considered profile quasi-maximum likelihood method to estimate the partially linear varying-coefficient spatial autoregressive model. Sun and Wu [30] discussed the GMM to estimate the partially linear single-index spatial autoregressive model.…”

Using Constraints in Freight Volume to Identify Regional Needs for Roadway Infrastructure

“…Alina [14] analysed the usage of the speed-flow curve in traffic modelling. Krisztin [15] used a four-stage model of regional transportation planning for evaluating the transport needs and for creating and harmonization of the transport policy. Economic models incorporate the indirect effects of infrastructure investments, such as employment or economic growth [16].…”