A Multi-Regional Computable General Equilibrium Model for New Zealand by Nathaniel Robson A thesis submitted to the Victoria University of Wellington in fulfilment of the requirements for the degree of Doctor of Philosophy Victoria University of Wellington 2012 Abstract Although New Zealand has had an active CGE modelling community since the 1980's, a multi-regional CGE model for the country has not been devel- oped until now. This thesis presents a prototype multi-regional CGE model to demonstrate the feasibility of developing a comprehensive model that cap- tures the benefits of modelling agent behaviour with a bottom-up approach. The prototype model is built upon bottom-up regional micro-foundations and New Zealand data is used to operationalise a particular implementation of the model. The thesis fills an important gap in the New Zealand CGE mod- elling literature as none of the models in current use have a structure involv- ing bottom-up regional modelling. The method of implementation is also a key contribution, utilising a maximum-entropy approach to overcome data shortages. An illustrative simulation of a natural disaster that strikes the Wellington central business district demonstrates the strengths of the bottom- up multi-regional approach | that the model can capture differential effects across regions of shocks that occur at the regional level, and incorporate flow- on and feedback effects between regions. Sensitivity testing of the substitution elasticity between domestic sources of products reinforces the importance of empirically-estimated parameters in CGE models. The basic model is extended in two ways. The first is to introduce mod- elling of distribution services as has been done in the ORANI and subsequently FEDERAL models. The key structural difference here is that products identi- fied as distribution services are required to facilitate movement of other prod- ucts from seller to buyer. Thus there are no opportunities to substitute away from these services if they become relatively more expensive. To implement the additional structure, sets of coefficients are specified to control technical i ii possibilities in the usage of the distribution services. These include switches that can dictate, for example, that wholesale trade is only involved in the delivery of tangible products, that retail trade is only used by in-region pur- chasers, and that transport is required for moving physical products across regional borders or to exporters. That these assumptions can be integrated seamlessly into the database highlights the strength of the maximum-entropy approach used to generate the multi-regional input-output database. Simula- tions of an oil price shock show that the regional assumptions surrounding the distribution networks are material to the results. The second extension to the model is the addition of a module to control the degree of inter-regional labour mobility. Essentially the user is given the ability to specify the extent to which households respond to regional real wage differences by moving to regions with relatively higher rates. Therefore, in short-run simulations labour can be made more mobile than capital, while in the long-run it can be less mobile than capital. The module also introduces additional structure to link populations, households, and labour market com- ponents. One important element of this new structure is a link back to the endogenous labour supply theory of the basic model. Publicly available demo- graphic and labour market data are used to implement the mobility module. The importance of a mobility response to relative real wage changes is explored in an illustrative application looking at the impact of regionally-concentrated immigration flows. The simulations suggest that population movements can work to dissipate the welfare effects of such migration inflows. Acknowledgements In achieving the completion of this thesis, I owe a debt of gratitude to my supervisors, Professor Viv Hall and Dr. Stephen Burnell, for their continuous support. Their excellent academic advice and direction in the research process was essential, along with their kind assistance in securing financial support. I gratefully acknowledge the following sources of financial assistance provided during various stages of this research: • School of Economics and Finance PhD Scholarship • Philpott-BERL Scholarship • G. G. G. Watson Award (on two separate occasions) • Department of Labour Graduate Research Award In addition, the School of Economics and Finance (SEF) and Business and Economics Ltd. (BERL) have provided fixed-term employment opportunities which have extended my range of professional experience and complemented my research. I wish to express my thanks to Dr. Ganesh Nana of BERL for very helpful discussions and the provision of data which aided this research. The data provision was funded through the Department of Labour Graduate Research Award. The New Zealand Institute of Economic Research (NZIER) invited me to join a CGE modelling workshop they had organised, and later to make a presentation to the Institute. Both of these provided valuable opportunities for discussion and very useful feedback. Professor John Madden of the Centre of Policy Studies, Monash University generously supplied me with a copy of his doctoral thesis and Dr. Adolf Stroombergen of Infometrics kindly granted me permission to use his input-output tables. I wish to also thank all iii iv participants of the presentations I have made to SEF, the Post-Graduate Students As- sociation (PGSA), and the New Zealand Association of Economists (NZAE). Professor Jacques Poot of the National Institute of Demographic and Economic Analysis, University of Waikato gave particularly useful advice following my NZAE conference presentation. I acknowledge the contribution of Dr. Peter Chang, formerly of SEF, who provided the initial impetus and supervision during the early stages of this research. Members of the SEF staff have provided valuable assistance on various matters throughout. A special thanks to Dr. Paul Calcott for the LATEX help. Complementing the assistance elucidated above has been the fantastic support I have received from family and friends. First and foremost has of course been that of my wife Mayumi and daughter Jennifer. Thank you for your endless support and patience, standing by while I pursued my dream. Thanks are also due to the wider family, whose encouragement and assistance have been vital to our survival in the interim. And last but certainly not least, my appreciation to all my friends who endured the awkwardly constructed descriptions of my research when they asked \what's your topic?" and incomprehensible rambling when I encountered difficulties. Thanks for providing the opportunities I needed to restore my sanity. My thanks go to one friend especially, for being there as a source of strength when I needed it most. ∗ ∗ ∗ ∗ In all affairs it's a healthy thing now and then to hang a question mark on the things you have long taken for granted. Bertrand Russell 花は盛りに、月は隈なきをのみ、見るものかは。 吉田 兼好 [徒然草第 137 段] (Translation) Why only when flowers are in full bloom and the moon is shining in spotless perfection should they be looked upon? Yoshida Kenk¯o[Idle time essay #137] Contents Glossary xv 1 Introduction1 1.1 Bottom-up Micro-foundations.........................4 1.2 Multi-regional IO Data and Information Theory...............6 1.3 Implementation in GAMS...........................8 1.4 Regional Focus of Applications........................ 10 2 The Basic Model 13 2.1 Introduction................................... 13 2.2 Basic Description................................ 13 2.3 Theoretical Structure.............................. 18 2.4 Model Implementation............................. 63 2.5 Simulations Using the JENNIFER Model................... 102 3 Distribution Services 109 3.1 Introduction................................... 109 3.2 Modelling Distribution Services........................ 110 3.3 Additional Model Structure.......................... 114 3.4 Implementation of the Margins Modelling.................. 119 3.5 An Illustrative Application........................... 139 4 Labour Mobility 147 4.1 Introduction................................... 147 4.2 Modelling Labour Mobility........................... 148 4.3 Additional Model Structure.......................... 150 v vi CONTENTS 4.4 Implementation of the Labour Mobility Modelling.............. 158 4.5 An Illustrative Application........................... 165 5 Conclusion 181 5.1 The JENNIFER Model............................. 181 5.2 The Geographic Nature of Distribution.................... 182 5.3 Partial Labour Mobility............................ 182 5.4 Future Development.............................. 183 Appendices 185 A CGE Models in the Literature......................... 187 B Selected RPEP Papers............................. 191 C List of Model Variables............................. 195 D List of Model Equations............................ 199 E Solutions to Constrained Optimisation Problems............... 211 F Linearisation of Demand Functions...................... 215 G Labour Supply Functions........................... 219 H Endogenous Investment Allocation...................... 221 I Regional Consumption and Propensities to Save............... 225 J Conversion Factors............................... 227 K Calibration of Demand Functions....................... 231 L Walras' Law................................... 233 M Margins Demands: A Worked Example...................