Inter-regional Spillovers of Policy Shocks in China

Abstract

Groenewold N., Chen A. and Lee G. Inter-regional spillovers of policy shocks in China, Regional Studies. Inter-regional output spillovers from policy shocks are analysed in a three-region vector-autoregressive (VAR)-based model of China using investment as the policy variable. The contemporaneous effect of policy is greater in the coastal region than in the other two regions, and the effect in the central region is larger than in the western region, suggesting that at least part of the expenditure boosts in the poorer inland regions find their way to the coastal provinces. These results are confirmed when the effects of the policy shocks are simulated over time and are found to be generally robust to alternative model specifications.

Groenewold N., Chen A. et Lee G. Les retombées régionales des chocs politiques en Chine, Regional Studies. A partir d'un modèle de la Chine à trois régions du type VAR, on analyse les retombées interrégionales de production qui résultent des chocs politiques, employant l'investissement comme variable de politique. L'effet contemporain de la politique s'avère plus important dans la zone côtière qu'il ne l'est dans les deux autres régions, et l'effet dans la région centrale s'avère plus important qu'il ne l'est dans la région occidentale, ce qui laisse supposer que l'impulsion donnée aux dépenses dans les zones intérieures plus défavorisées alimente dans une certaine mesure les provinces côtières. Les résultats se confirment suite à une simulation des effets temporels des chocs de politique et s'avèrent fiables en règle générale par rapport aux autres spécifications des modèles.

Chine Régional Retombées de politique

Groenewold N., Chen A. und Lee G. Interregionale Übertragungen aufgrund politischer Schocks in China, Regional Studies. Wir analysieren interregionale Leistungsübertragungen aufgrund politischer Schocks in einem VAR-basierten Dreiregionen-Modell von China, wobei wir die Investitionen als politische Variable heranziehen. Die zeitgleichen Effekte der Politik fallen in der Küstenregion stärker aus als in den anderen beiden Regionen, und der Effekt in der zentralen Region ist größer als in der westlichen Region. Dies lässt darauf schließen, dass zumindest ein Teil der Ausgabensteigerungen in den ärmeren Inlandregionen seinen Weg in die Küstenprovinzen findet. Die Ergebnisse bestätigen sich, wenn wir die Effekte über den Zeitraum der politischen Schocks hinweg simulieren, und erweisen sich generell als robust gegenüber alternativen Modellspezifikationen.

China Regional Politikübertragungen

Groenewold N., Chen A. y Lee G. Desbordamientos interregionales en los choques políticos en China, Regional Studies. Con ayuda de las inversiones como nuestra variable política, analizamos los desbordamientos de la producción interregional de los choques de la política a través de un modelo VAR basado en tres regiones de China. El efecto contemporáneo de la política es superior en la región costera que en las otras dos y el efecto en la región central es mayor que en la región occidental lo que sugiere que como mínimo parte de los impulsos inversores en las regiones interiores más pobres se orientan hacia las provincias costeras. Estos resultados se confirman cuando simulamos los efectos durante el transcurso del tiempo de los choques políticos y observamos que, en general, son contundentes comparándolos con especificaciones de modelos alternativos.

China Regional Desbordamientos políticos

Keywords:

• China
• Regional
• Policy spillovers

JEL classifications:

• R11
• R12

Acknowledgements

The authors have benefited from comments by their discussants at the ACE International Conference in Hong Kong, China, in December 2006, and at the UNU-Wider Project Meeting in Beijing, China, in January 2007. The authors are grateful to the Department of International Co-operation at Xi'an Jiaotong University, to the Department of International Co-operation at Jinan University, and to the School of Economics and Commerce at the University of Western Australia for grants which supported the visit of the first-named author to Guangzhou and Xi'an in November 2006. Two rounds of comments by two anonymous referees are gratefully acknowledged.

Notes

Data for per capita gross domestic product are from State Statistical Bureau (2006).

For a more extensive discussion of the regional implications of the various Five-Year Plans, see Groenewold et al. (2008).

This is not to say that regional economists have not been interested in inter-regional spillovers. Quite the contrary is the case. For recent examples, see Alvarez et al. (2006) and Ezcurra et al. (2005), but these publications deal with productivity spillovers from public capital and use a production–function framework. Other literature such as Kamps (2005), Marcellino (2006), and Moreno and Lopez-Bazo (2007) analyse the effects of fiscal policy on output in the spirit of earlier work by Blanchard and Perotti (2002). None of these, however, deals specifically with the issue addressed herein.

The literature goes back at least to the work of Kuznets (1955) and Williamson (1965); the concept was more recently developed in Barro and Sala -i-Martin (1992). There is a vast empirical literature. For some important papers with a bias to Chinese applications, see: Chen and Fleisher (1996), Fleisher and Chen (1997), Kanbur and Zhang (1999), Yao and Zhang (2001a, 2001b), Demurger (2001), Chang (2002), Lu (2002), Cai et al. (2002), Yang (2002), Demurger et al. (2002a, 2002b), and Bao et al. (2002).

For a set of papers using a modelling approach similar to that used by Groenewold et al. (2006, 2007, 2008) applied to the USA, see Sherwood-Call (1988), Cromwell (1992), Carlino and Defina (1995), Clark (1998), Rissman (1999), and Kouparitsas (2002).

Generally, independence of the e_i terms is not guaranteed by theoretical restrictions nor supported by the data and must be imposed by transforming them to ensure orthogonality. The standard method for achieving orthogonality is based on the use of the Choleski orthogonalization of the covariance matrix of the errors and this procedure makes the IRFs dependent on the order in which the variables are entered into the model, although the issue is not serious in applications where the correlation of the errors is weak. For an extensive discussion of this issue in a bivariate context, see Enders (2004, ch. 5).

These are not standard IRFs since, as explained in the text, they result from shocks not to the model errors, but to the exogenous variables. ‘IRFs’ is therefore used in the same loose way in which ‘VAR’ is used to describe the model. The alternative is to use terms like ‘quasi-VAR’ and ‘quasi-IRFs’, but this seems unnecessarily cumbersome for such small deviations from standard. An alternative to ‘IRF’ is to use ‘dynamic multiplier’ as suggested by a referee. This is more accurate, but also less specific, and the use of ‘IRFs’ is maintained – they are not too far from standard IRFs and, given that the model from which they are derived is clearly specified, it should not result in confusion.

Since beginning this project, the 2004 data have become available. The data set has not been updated in the interests of preserving comparability with the results reported by Groenewold et al. (2006, 2007a), whose sample period stops at 2003.

Note that Hainan, Chongqing, and Tibet are missing. Hainan is included in Guangdong and Chongqing in Sichuan. Tibet has been omitted altogether due to missing data.

The p-values were 0.83, 0.49, and 0.01 for coastal, central, and western region equations, respectively.

The p-values were 0.87, 0.04, and 0.01 for coastal, central, and western region equations, respectively.