Editorial

The autumn of 2007 saw one of the big events of the Regional Science year, namely the 54th Annual Meetings of the Regional Science Association International, held in Savannah, Georgia, USA. Savannah is a small city in the long shadow of big neighbour Atlanta, but it has its own unique charm, due largely to the distinctive quality of its urban landscape and architecture. Savannah is a planned city, laid out by 18th-century colonists, and was destined to become a place of privilege and style. French philosopher Bernard-Henri Lévy describes it thus … ‘Italianate and Hellenic architecture … [are] witnesses to bygone eras when rich shipowners from London, settled in Savannah, [and] vied with one another for style and splendor …’ It is one of the largest National Historic Districts in the USA. From the perspective of spatial economics, its location near river and sea imparts inherent characteristics that help support a metro area of 325,000. Savannah is not just a tourist centre, it is a logistics hub, container port and distribution centre of national importance.

It is impossible to describe in a few words the contents of a programme stretching over 4 days, comprising parallel sessions, optional social events (‘ice-breaker reception’, ‘midnight in the garden of good and evil’, etc.), editorial board meetings, formal lunches or exhibitions. So it will not be attempted. However, much was on show that is relevant to this journal, and this journal was also on show, featuring prominently and colourfully (alongside sister journal Regional Studies) in the conference brochure. This input is being rewarded by the quality of the submissions we are receiving, and by the very high-quality referees who have committed themselves in both time and effort to make this journal a success.

The next large event organized by Regional Science Association International will take place in March. This year the Regional Science Association International World Conference will be held in Sao Paulo, Brazil. Topics such as world integration, emerging countries, lagging regions, and sustainability will provide a platform for wide-ranging discussions of current research. Once again Spatial Economic Analysis will be present and represented by our co-editor Danilo Igliori, who is a member of the local organizing committee.

Closer to home, events are moving rapidly after a slow response to the bids from institutions and consortia to become the UK's new ESRC Centre for Spatial Economics. After a long wait, a press release finally announced the result of a closely fought contest, ultimately between consortia headed up by Cambridge University and the London School of Economics (LSE). The outcome was a decision in favour of the LSE. The other participating institutions in the LSE consortium are Nuffield College (Oxford University), the University of Wales (Swansea), Newcastle University and the Centre for Public Policy for Regions (CPPR), a joint institution of Strathclyde and Glasgow Universities in Scotland.

The aim is to identify fundamental determinants of spatial disparities in economic prosperity across the UK, providing analysis that will inform UK spatial policy. None the less, research outcomes will be of global relevance. Looking at the details of the proposal (three of our editors—Bernard Fingleton, Strathclyde University, Paul Cheshire, LSE, and Patricia Rice, Oxford—are co-investigators), we can see some clear indications of likely directions for spatial economics over the coming years. Notably, to a large extent the centre will:

• Use economic theory and concepts to develop empirical models with a basis in real data and real places.

• Focus on the micro scale.

As Cheshire & Malecki (2004) state, ‘the ultimate actors are not regions but households, establishments and firms and how these interact’.

This emphasis on the micro level has its basis in contemporary approaches to spatial economics, as evident in the explicit microeconomic underpinnings of aggregate models, notably in urban economics and the New Economic Geography (NEG). It is also an outcome of the availability of underexploited geographically referenced microeconomic data sets, and reflects recent advances in the techniques of spatial data analysis, spatial econometrics and spatial panel analysis. More precisely, ‘the Centre's strategy with respect to the choice of spatial scale is to access the finest-grain data available and then to analyse at the most appropriate scale for the process concerned’.

• Test whether spatial disparities are due mainly to place-based effects or to sorting.

Place-based effects attribute differences in welfare, income and productivity to the inherent geographical characteristics of places, for instance the (dis)advantages associated with a peripheral, coastal or coalfield locations. The assumption is that placing essentially identical actors in different locations will, per se, result in different outcomes for income and productivity. Rather than emphasizing the inherent advantages and disadvantages of places, sorting refers to the tendency for factors of production and consumers to gravitate towards (disperse away from) mutually beneficial (disadvantageous) locations, thus reinforcing externalities and amplifying initial differences to create larger, more permanent spatial disparities. However, there are costs as well as self-reinforcing benefits arising from agglomeration. As the proposal asserts:

As activity concentrates, the price of scarce resources increases; roads become more congested; pollution increases. In a modern economy, the trade off between these costs and benefits determines which areas are rich and which are poor.

• Explore appropriate policy responses, for instance to the costs and benefits of agglomeration.

There are many issues relating to spatial economic policy and governance that will be considered and to which is it difficult to do justice here. However, one of the more controversial topics is the appropriate scale of policy intervention if one allows amenity and cost-of-living differences to offset wage differentials across space. More generally, the intention is to consider how policy might best address spatial economic disparities, and the governance mechanisms most appropriate to policy deliverance and development.

• Demonstrate the value added from combining diverse data sets and different methodologies.

Different approaches to specific problems should, when considered together, help us better understand causes and effects in the spatial economy. For example, in order to better comprehend intra-urban inequalities and changing urban form, diverse data will be linked together, such as individual micro data, house price data, plant location and workplace employment. Likewise, complementary modelling strategies should jointly enhance understanding of dynamics and responses to shocks and policy instruments. One strategy will be to employ explicit theoretical NEG-based general equilibrium (GE) models. Formal GE models impose a lot of theoretical structure which has the advantage of setting out unequivocally the assumed linkages, but one has to take care to get these assumptions right, and outcomes may differ significantly according to the assumptions imposed. The complementary approach is to use VAR-type models. These have the advantage of being much more inductive in orientation, and therefore impose few untested restrictions, although the disadvantage is that they are not necessarily identified with specific economic processes. Taken together, however, both approaches should add to our understanding of the structure and evolution of the spatial economy.

A novel feature of the first paper in the current issue, by Sandy Dall'erba, Marco Percoco & Gianfranco Piras, is their application of an endogenous ‘convergence club’ (spatial heterogeneity) detection methodology that also takes account of spatial effects. The methodology allows more than two regimes, which is a necessity given their heterogeneous data, covering 244 regions across 25 EU countries. In contrast, the literature so far has focused on exploratory spatial data analysis methods and been limited to only two regimes. In addition, traditional endogenous regime detection methods do not consider spatial effects as in this paper. Moreover, this approach is applied not to the usual Solow–Swann-inspired neoclassical convergence model but to Verdoorn's law. This law implies increasing returns to scale, rather than the constant returns to scale (diminishing returns to factor inputs) embodied within the neoclassical approach. The implication is that productivity growth will not converge to the same rate, governed ultimately by a constant exogenously determined rate of technical progress. Rather, under the basic Verdoorn law growth can converge to its own, different steady state in each location. This means that productivity levels may forever diverge. Indeed, embodied within the Dixon–Thirlwall model of circular causation, which is a system of equations involving output, export and price growth, Verdoorn's law can even produce non-convergent differentiated growth dynamics, depending on the model parameterization. Under this scenario, productivity levels diverge at an even faster rate. However, when, standing alone as in this paper, the Verdoorn law specification includes a productivity-level gap or catch-up term, with feedback from productivity growth to productivity gap, then equilibrium is (normally) characterized by constant productivity growth rates across regions, which is observationally equivalent to the outcome of a neoclassical convergence process. However, Dall'erba, Percoco & Piras find that, on the whole, the level of technology gap is not significant across all regimes, and this, if true, has far-reaching implications for the long-run structure of the EU space-economy.

Spatial effects are commonly treated as autoregressive processes in spatial econometrics, and little consideration is given to moving average processes. However, the distinction between the two is fundamental, and moving averages remain a serious modelling option. With an autoregressive error process a shock in one region cascades outwards to all other regions ad infinitum. In contrast, the footprint of a moving average error process is much more restricted, the impact of the shock being confined to ‘neighbouring’ locations. This distinction is also important when we come to the burgeoning literature on modelling spatial panels with spatially dependent observations. The literature considers the spatial autoregressive error process exclusively. The paper by Bernard Fingleton seeks to remedy this by giving a generalized method of moments (GMM) estimator for spatial panels with moving average errors, extending to also incorporate an endogenous spatial lag. The paper illustrates the application of the estimator using a panel of UK real estate data, and uses Monte Carlo methods to illustrate its evident consistency.

The issues relating to spatial partitioning are important for the choice of computable general equilibrium (CGE) modelling approach advocated by James Giesecke, in his paper ‘A top-down framework for regional historical analysis’. Giesecke argues that while bottom-up does not have the same theoretical limitations as top-down, if you want to do serious applied work at the level of spatial and sectoral disaggregation that is increasingly demanded, then top-down has a lot going for it. Compared with bottom-up, it requires far less in terms of data. Nevertheless, the top-down approach can explain recent economic history for many regions, and also generate forecasts that are at a level of detail required by policy makers. It appears that ‘Given the experience of the past two decades, it is reasonable to expect that growth in demand for sectoral and regional detail will continue to outpace growth in computing power. In particular, top-down modelling will continue to play an important role in applications where the number of sectors and regions must necessarily be very large.’ By detail, we are talking here about 57 Australian regions and 106 commodities and industries. One limitation of the top-down methodology is that region-specific supply-side detail is absent. In the paper, Giesecke seeks to remedy this so as to accommodate some observable features of regional economies. The work is carried out using a top-down regional version of MONASH, which is a national CGE model.

The paper by Alastair Bailey, Alberto Zanni & Sophia Davidova is an excellent example of what is meant by micro-level analysis in the remit of the ESRC Centre for Spatial Economics. It shows how experimental techniques at the level of individuals can provide insights relevant to the operation and planning of the aggregate spatial economy. In their case the issues are the determinants of job and place of residence (in rural Slovenia), and their choice experiment methodology involves presenting people with hypothetical choice options and then analysing trade-offs. They show that wages are important in the choice of job, but are not the only factor. In addition, mobility and migration are not purely a response to wage differentials as there are other important factors. These finding correspond to what we already know, and are contrary to what we tend to assume in our formal models. For instance, the earliest versions of the NEG model go to equilibrium as a result of labour migrating purely in response to real wage differences. Reality is different, and the challenge is to take account of this evidence without destroying the benefits of formal modelling.

NEG analysis received a fresh and important policy-oriented impetus as a result of the publication of Baldwin et al.'s (2003) Economic Geography and Public Policy. In contrast, policy prescriptions, typically embodying trade policy, taxation and regional policy, are purposefully absent from Fujita et al. (1999). Pasquale Commendatore, Martin Currie & Ingrid Kubin take up the theme of the policy implications of NEG theory, and argue that truly genuinely periodic or chaotic complex dynamical behaviour is associated with a discrete rather than continuous time framework. They show that even minute changes in a tax rate can cause a dramatic, catastrophic and irreversible change in the location of manufacturing industry, rather than a marginal change in the tax rate producing a marginal relocation of a mobile factor. They argue that this is not a result of irrationality on the part of entrepreneurs; in other words, their conclusions hold under non-irrationality assumptions about entrepreneurial expectations. Consider rational expectations based on perfect information and unlimited computational ability. This requires a capability to predict regional post-tax incomes on the basis of perfect knowledge of migration flows and an ability to solve for the short-run equilibrium. Consider the case of heterogeneous agents switching between naïve expectations (simple rules) and rational expectations, depending on estimated performance. The outcome is complex dynamical systems behaviour that is not simply a consequence of entrepreneurial irrationality. These outcomes in a discrete time framework occur because entrepreneurial migration can overshoot—evidently an impossibility with continuous time. Overshoot may occur because there is a strong reaction to real entrepreneurial income differences between regions, although whether this will really occur depends on the weight given to industry location and the deterrent effect of migration costs, but this is an empirical issue not considered in this purely theoretical paper, so it is difficult to know the likelihood of periodic or chaotic behaviour without additional empirical analysis. They conclude that it is wrong to base policy conclusions on a paradigm (NEG) which produces substantially different dynamical outcomes simply by switching from a continuous to a discrete time framework.