Abstract
Bruna F., Lopez-Rodriguez J. and Faíña A. Market potential, spatial dependences and spillovers in European regions, Regional Studies. This paper reinterprets the new economic geography (NEG) ‘wage’ equation by distinguishing two different types of spatial dependences: a global spatial trend and local spatial autocorrelation. A measure of the variable Market Potential in this equation can capture both a global core–periphery pattern and spillovers, while the standard weights matrices of spatial econometrics tend to be designed to capture short-distance interactions among neighbours. Using cross-sectional European regional data, the paper compares different weighting schemes to build spatial lags. The estimation of spatial models of an NEG equation for gross value added per capita (GVApc) reveals new research challenges.
Bruna F., Lopez-Rodriguez J. and Faíña A. 欧盟的市场潜能、空间依赖与外溢,区域研究。本文透过区别两类不同的空间依赖—全球空间趋势与地方空间自相关,重新诠释新经济地理(NEG)的“薪资”方程式。在此一方程式中,市场潜能变量的测量,能够同时捕捉全球的核心—边陲模式与外溢,而空间经济计量学的标准权重矩阵,则倾向设计用来捕捉邻里间的短距离互动。本文运用欧洲区域的横截面数据,比较建立空间落迟的不同权重计算方法。人均毛附加价值(GVApc)的NEG方程式之空间模型评估,揭露了崭新的研究挑战。
Bruna F., Lopez-Rodriguez J. et Faíña A. Le potentiel du marché, les dépendances spatiales et les retombées dans les régions européennes, Regional Studies. Cet article cherche à réinterpréter l’équation de salaire de la Nouvelle géographie économique en distinguant deux types différentes de dépendances spatiales: une tendance spatiale globale et l'autocorrélation spatiale locale. Dans cette équation, une mesure de la variable Potentiel du marché peut capter à la fois une structure centre-périphérie globale et des retombées, tandis que les matrices de poids standard de l’économétrie spatiale ont tendance à être concues afin de capter les interactions à courte distance entre voisins. Employant des données régionales européennes transversales, l'article compare différentes formules de pondération pour créer des retards spatiaux. L'estimation des modèles spatiaux à base d'une équation de la Nouvelle géographie économique relative à la valeur ajoutée brute par tête indique de nouveaux défis.
Bruna F., Lopez-Rodriguez J. und Faíña A. Marktpotenzial, räumliche Dependenzen und Übertragungseffekte in europäischen Regionen, Regional Studies. In diesem Beitrag wird die Lohngleichung der neuen Wirtschaftsgeografie durch die Unterscheidung zwischen zwei verschiedenen Arten von räumlichen Dependenzen neu interpretiert: einem globalen räumlichen Trend und einer lokalen räumlichen Autokorrelation. Eine Messung der Variablen Marktpotenzial in dieser Gleichung kann sowohl ein globales Kern-Peripherie-Muster als auch Übertragungseffekte erfassen, während die herkömmlichen Gewichtsmatrizen der räumlichen Ökonometrie tendenziell zur Erfassung von Wechselwirkungen zwischen Nachbarn über kurze Entfernungen hinweg ausgelegt sind. Anhand von regionalen europäischen Querschnittsdaten werden verschiedene Gewichtungsschemata zum Aufbau räumlicher Verzögerungen miteinander verglichen. Die Schätzung der räumlichen Modelle einer Gleichung der neuen Wirtschaftsgeografie für die Bruttowertschöpfung pro Kopf verdeutlicht neue Herausforderungen für die Forschung.
Bruna F., Lopez-Rodriguez J. y Faíña A. Potencial del mercado, dependencias espaciales y derrames en las regiones europeas, Regional Studies. En este artículo reinterpretamos la ecuación de salarios de la Nueva Geografía Económica al distinguir entre dos tipos diferentes de dependencias espaciales: una tendencia espacial global y la autocorrelación espacial local. Una medida de la variable Potencial de Mercado en esta ecuación puede captar tanto una pauta centro-periferia global como derrames, mientras que las matrices de ponderaciones estándar de la econometría espacial suelen estar diseñadas para captar interacciones de corta distancia entre vecinos. Mediante datos transversales de regiones europeas, comparamos diferentes esquemas de ponderaciones para construir retardos espaciales. La estimación de modelos espaciales de una ecuación de la Nueva Geografía Económica para el valor añadido bruto per cápita revela nuevos desafíos de investigación.
Acknowledgments
The authors wish to thank Tomasz Mickiewicz and two anonymous referees for very useful contributions. They are grateful to the participants who attended presentations of previous drafts of this paper given at the XVth Applied Economic Meeting (A Coruña, Spain, 2012), the XIIIth Conference on International Economics (Granada, Spain, 2012), and VIth World Conference of the Spatial Econometrics Association (Salvador, Brazil, 2012). The views expressed in this paper are purely those of the authors and cannot under any circumstances be regarded as stating an official position of the European Commission. The authors also thank Eduardo Giménez, Coro Chasco, Julie Le Gallo, Robert Bivand, Frank van Oort, David G. Rossiter, Jose-María Montero, Laura Varela, Nicolas Devarsy and Xosé Manuel Martínez Filguera for helpful insights made on particular aspects of previous drafts. Jon Stenning made helpful clarifications about the Cambridge Econometrics' data. The first author is also grateful to Giuseppe Arbia for conversations had whilst on a research visit at the Università Cattolica del Sacro Cuore, Rome. The valuable research assistance of Adrián Gutiérrez, Vanessa Mato and Paulino Montes is highly recognized. The usual disclaimers apply.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Fernando Bruna http://orcid.org/0000-0002-6433-0987
Jesus Lopez-Rodriguez http://orcid.org/0000-0002-2142-3913
Andrés Faíña http://orcid.org/0000-0003-2323-7521
Notes
1 Breinlich (Citation2006), as well as other authors, finds that using travel times does not alter the results significantly.
2 GVApc is preferred to wages for theoretical and empirical reasons. Its correlation with nominal remuneration per worker is 0.81 in the sample under study, but the data of the latter variable have lower quality.
3 The words ‘global’ and ‘local’ are not always used with this meaning in spatial econometrics. A ‘global’ measure of spatial autocorrelation is that which applies a common weights matrix to the space of observations, even if that matrix is designed to capture (average) local dependence.
4 Row standardization guarantees that the maximum eigenvalue of W is 1 and the invertibility of a linear combination of W. It prevents the estimated spatial parameter from implying explosive models with unknown properties. Alternatively, Kelejian and Prucha (Citation2010) propose dividing each element of W by the spectral radius of W, the maximum absolute value of its eigenvalues. This method is not discussed here.
5 The heterogeneous size of the observational units (and, therefore, sample selection) is always an issue when modelling space (). That is related to the large literature on the modifiable areal unit problem and is beyond the scope of this paper.
6 See below, note 10.
7 See above, note 5.
8 The two outliers at the top of the first two plots of correspond to Inner and Outer London, which have a somewhat arbitrary small distance between centroids.
9 The authors appreciate the suggestions of an anonymous referee to study correlations and improve the explanation of .
10 On an experimental basis, the SAR model in column (4) of was estimated using the spatial lags of the explanatory variables (X, WX, WWX) as instruments for the spatially lagged dependent variable. This last variable becomes non-significant and the estimates of External Market Potential and Human Capital get close to the OLS ones in column (4) of . The additional instrumentation of External Market Potential by mean distances does not change the qualitative results.
11 The authors thank Tomasz Mickiewicz for raising this point.