Economic integration in Europe has de facto become a multi-speed process. For the Central and Eastern European Countries (CEECs), this has come as a response to the challenge of integrating possibly twenty-five countries with very different income levels without halting altogether the deepening of integration. This is also the case for the process of monetary unification for which the conditions for being in the fast track group have been defined in the Maastricht Treaty. We do not know much of the effect of this new dimension of time and sequentiality in regional integration. One of the most interesting questions is whether the existence of different speeds of integration can have an impact on the long term characteristics of this integration.

Our paper looks at this question from the point of view of location of economic activities. Our objective is to use the tools of the “new geography”, in order to describe the possible impact of a multi-speed approach to integration on the location choice of industries and therefore on the long term geography of economic activities in Europe. We use a three-country model where two identical rich countries decide to integrate their economies and leave a third, poorer, country temporarily outside. The questions we ask are the following: if the concentration of economic activities in the core of Europe is a concern to policy makers, will a multi-speed approach help alleviate this problem or will it exacerbate it? Is the transition period, during which the poor country is excluded from the integrated area, necessary to avert massive relocation to the rich core until the income gap has sufficiently decreased? Or, on the contrary, will this transition period increase the risk of agglomeration in the rich countries of the core? The answers to these questions are important because they partially determine whether the integration of the outside countries (in particular the CEECs) should be conditioned on a decrease of the income gap or whether the imposition of such condition is perverse because the imposed transition period will generate agglomeration in the core, income divergence and therefore the indefinite postponement of complete integration.

To answer these questions, we use a stylised model in which history is given a strong role as a determinant of long term economic geography. We model the income gap between the two rich countries and the poor country as a difference in the endowment of human capital or " entrepreneurs " who also are the mobile factor. Integration is characterised by the elimination of transaction costs. We distinguish between two scenarios: one where agglomeration economies can not set in during the transition period and one where they can. In our model, the source of agglomeration forces is migration of human capital. However, we could as well have modelled agglomeration as arising from vertically linked industries.

When agglomeration economies can not set in during the transition because migration is not allowed, then "entrepreneurs" (the owners of human capital) can choose where to locate their plants but cannot migrate themselves. We show that when two out of the three countries lower their transaction costs, the excluded country will always face relocation of its firms towards the integrated area. This is true whatever the income differential between the countries: as the transaction costs are lowered in the integrated area, both integrated countries become a better export base to each other than the excluded country. Hence, from the point of view of location of economic activities, the excluded country would prefer that regional integration in the core had never happened. This does not readily imply that the excluded country should insist to enter as soon as possible. This is because, as transaction costs are lowered, the income gap effect will be reinforced as a determinant of industry location. If it joins the integrated area, firms will relocate in the richest markets which are in the core. If the transition period is associated with income convergence, then a multi-speed approach may be a good idea: it enables the " periphery " country to join the integrated area at a time where the income differential is not too large and does not generate massive relocation to the core countries. The transition period is then characterised by relocation of firms from the core to the periphery.

This conclusion is reversed when agglomeration economies can set in. This is because, when we let entrepreneurs free to migrate, incomes are not fixed geographically so that the temporary exclusion of the poor country may trigger agglomeration in the rich integrated core during the transition. We show that the transition period, during which the poor country is excluded, can induce the agglomeration of the increasing returns sector in the rich integrated countries, causing divergence in terms of incomes between the core and the periphery. This may occur even though convergence between the rich and poor countries was taking place before integration among the first track countries. We call this the " agglomeration effect " of multi-speed integration. If the length of the transition period itself is conditioned on income convergence, the country " temporarily " excluded may never be able to be integrated. We analyse in detail how this perverse scenario depends on the parameters of the model.

The policy implications follow naturally. If policy makers do not believe that migration of human capital and more generally agglomeration economies can set in, then a multi-speed approach to integration makes sense because income differentials between countries are the main determinant of industry location. However, if agglomeration can set in, this approach is dangerous because the attraction of the core during the transition period when the poor country is excluded is cumulative and may never be reversed. If we think of agglomeration as a long-term phenomenon, our model then suggests some theoretical ground for the common sense view according to which the transition period before integration of the periphery countries should not be too short but it should not be too long either.