STUDIES
Modelling Network Interdependencies of Regional Economies using Spatial Econometric Techniques
Péter Járosi
Regional Research Institute West Virginia University
E-mail: peter.jarosi@mail.wvu.edu
This paper was presented at the 56 th Meeting of the Southern Regional Science Association, Memphis, Ten-nessee, April 2017.
Regional Statistics, Vol. 7. No.1. 2017: 003 – 016; DOI: 10.15196/RS07101
Multiregional dynamic models of economic growth rarely capture the interdependencies among regions that are geographically distant and/or often underestimate the importance of these linkages. This bias has become more and more serious because travel and transportation costs continue to decrease, while new telecommu-nication and information technologies enable business activities to readily take place between geographically remote locations. The conceptual framework in this study – modelling the network o regions – is based on well-known spatial economet-ric methods and provides alternative ways to inte-grate network interdependencies of economic ac-tivities into many fields, as well as modelling tech-niques such as spatial computable general equilib-rium, input-output, and dynamic econometric models.
Keywords: spatial econometrics, input-output analysis, network of regions, interregional trade
Introduction
Both spatial econometric models and network-based regressions can be considered as special cases of cross-sectional dependence regressions, where the relationships between regions are defined by spatial proximity and the edges of the network structure, respectively. LeSage and Debarsy (2016) suggest replacing the word ’ spa-tial’ with the term ’ cross-sectional dependence ’ when weight matrices are not creat-ed through spatial proximity, but via any other non-spatial proximity definition. In this article, I take the example of interindustry relationships to show how further development of these techniques is possible based on weight matrices that are cre-ated not only by contiguity and spatial distance algorithms, but also by defining network distance measurements. Empirically, long-distance cross-sectional depend-encies can be more important than spatial ones depending on the industry; never-theless, there are still many traditional industries where spatial autocorrelation can be dominant while other cross-sectional dependencies remain insignificant.
For regional policy, it is essential to: identify the key industries within a given re-gion and analyse their spatial and network dependencies, including geographically distant linkages; measure their dependencies from weaker to stronger; and know how to influence them to improve their economic performance. Before making policy decisions, it is crucial to run simulations using dynamic models, implement regional impact assessment, and take into consideration interregional and interindustry effects beyond macroeconomic aspects and regional impacts. In this respect, the main goal of this study is to provide theoretical and methodological insights about the importance of regional network interlinkages of key industries, as well as a proof of concept for future network econometrics and economic modelling research
Regional policy implications
The impact assessment can be implemented by changing the initial values of the exogenous variables. In this way, the shocks added to the system are defined as the differences between new and original values of the variables. This technique repre-sents a powerful tool to investigate the dynamics of the multiregional model re-sponding to an economic shock.
The dynamic simulations highlight the consequences of changes in the policy variables, which can be implemented by using macroeconomic and multiregional models separately, or by developing a mixed macroeconomic-multiregional model. The multisectoral regional blocks of the suggested model can support the identifica-tion of key industries within regions, while distinguishing spatially and network de-pendent sectors. Different types of industries justify different policies, depending on the observed interregional, backward, and forward linkages.
Conclusions and further research
The methodology to create network-based regions, weight matrices has been out-lined, and the existence of multiregional network autocorrelation has been proved through a simplified econometric example.
The construction of interregional linkages in the multiregional modelling frame- work via network and spatial econometric equations is more convenient than the burdensome creation of the identities to connect regions social accounting matri-ces. Collecting reliable data for these problematic identities also seems to be labori-ous or sometimes even impossible. Moreover, the assumptions of the spatial com-putable general equilibrium modelling framework (e.g. short-term spatial equilibri-um of the labour market) are not verifiable. In this new approach, the long-term dynamic simulation using the interregional econometric equations can provide re-sults as reasonable as the ones from spatial computable general equilibrium models. As for the macroeconomic and the intraregional blocks of the integrated model, the computable general equilibrium approach and the identities creation through the social accounting matrix are still recommended. The most appropriate solution might be a mixed type of econometric, input-output, and computable general equi-librium model, where the interregional linkages are simulated by network, gravity, and spatial econometric equations replacing the conventional identities.
In the future, an alternative way to build a lightweight version of dynamic mul-tiregional models will be developed by using the network econometrics equations.
حمله من هنا
للقراءة والتحميل اضغط هنا
ليست هناك تعليقات:
إرسال تعليق