Integration, agglomeration and the political economics of factor mobility

Journal of Public Economics 83 (2002) 429–456www.elsevier.com/ locate /econbase

Integration, agglomeration and the political economics offactor mobility

a,b,c b,d ,*Gianmarco I.P. Ottaviano , Jacques-Francois Thissea Ùniversita ‘L. Bocconi’, Milan, Italy

b ĆORE–Universite catholique de Louvain, 34 voie du Roman Pays, B-1348 Louvain-la-Neuve,Belgium

cCEPR, 90 –98 Goswell Road, London EC1V7RR, UKd ´ `CERAS-Ecole Nationale des ponts et chaussees, 28 Rue des Saints-Pares, 75343 Paris Cedex 07,

France

Received 8 August 2000; received in revised form 15 November 2000; accepted 21 November 2000

Abstract

This paper tackles the issue of the optimality of agglomeration in a two-region economywith skilled /mobile and unskilled / immobile workers. The market leads to the optimaloutcome when transport costs are high or low. However, for intermediate values, it yieldsagglomeration whereas dispersion is socially desirable. The reasons for this inefficiency arethen analyzed. Finally, we show that competitive lobbying on factor mobility by the twogroups of workers sustains the second best optimum. 2002 Elsevier Science B.V. Allrights reserved.

Keywords: Agglomeration; Trade; Monopolistic competition; Political economics; Factor mobility

JEL classification: F12; F22; H23; R13; R23

1. Introduction

One of the most severe challenges posed by the ongoing integration of nationaleconomies within the European Union is the strengthening of the core regions

*Corresponding author: Tel.: 132-10-47-4312; fax: 132-10-47-4301.E-mail address: [email protected] (J.-F. Thisse).

0047-2727/02/$ – see front matter 2002 Elsevier Science B.V. All rights reserved.PI I : S0047-2727( 00 )00166-3

430 G.I.P. Ottaviano, J.-F. Thisse / Journal of Public Economics 83 (2002) 429 –456

accommodating modern production sectors at the expense of the peripheral regionsretaining only traditional and local activities (Ottaviano and Puga, 1998). In theFirst Report on Economic and Social Cohesion, the European Commission (1996)observes that ‘‘[t]he average disparity in income per head in the EU is twice that incomparable regions in the US’’, while ‘‘[e]conomic activity is strongly concen-trated in the most urbanised areas of the Community. Regions with more than 500inhabitants per square kilometers account for only 4% of the land area of theUnion but for more than half the population. This implies that between two thirdsand three quarters of the EU’s total wealth creation occurs in urban areas’’ (p. 24).

Although Sala-i-Martin (1996) has found slow interregional convergence from1950 to 1990, it is more and more widely accepted that regional disparities havebeen growing within the EU since the deepening of the integration process startedin the 1980s (Armstrong and Vickerman, 1995; Neven and Goyette, 1995;Magrini, 1999). Furthermore, even slow convergence between countries may wellhide a process of regional divergence inside each country as argued by de laFuente and Vives (1995) who show that about half of the income inequality acrossEU regions corresponds to differences within each member state. Taking adifferent angle and comparing the spatial distributions of activities in varioussectors between 1968 and 1990, Amiti (1998) notes both an increase in thegeographical concentration of economic activities within most EU state membersand, for a vast majority of sectors, a tendency toward more agglomeration withinthe EU as a whole.

This lack of cohesion in social development and economic growth has led theEuropean Commission to pay more attention to this issue at the time of the SingleEuropean Act (SEA). The reform of the Structural Funds and the creation of theCohesion Funds, both aiming partially at the reduction of regional disparities, havebeen the concretization of this awareness. For the period 1994–1999, these fundscorrespond to approximately one third of the EU budget, thus showing thegrowing importance that the ‘regional question’ plays for the European

1construction. For the European Commission (1996, p. 13):

‘‘Imbalances do not just imply a poorer quality of life for the mostdisadvantaged regions and the lack of life-chances open to their citizens, butindicate an under-utilisation of human potential and the failure to takeadvantage of economic opportunities which could benefit the Union as awhole.’’

The statement is clear-cut and reveals that the concern of the Commission istwofold. On the one side it is about equity (‘‘ . . . a poorer quality of life for the

1For example, this is emphasized in Article 130a of the Amsterdam Treaty, which states that the‘‘Community shall aim at reducing disparities between the level of development of the various regionsand the backwardness of the least favoured regions or islands, including rural areas’’.

G.I.P. Ottaviano, J.-F. Thisse / Journal of Public Economics 83 (2002) 429 –456 431

most disadvantaged regions and the lack of life-chances open to their~citizens . . . ’’). On the other it is also about efficiency (‘‘ . . . an under-utilization

of human potential and the failure to take advantage of economic opportunitieswhich could benefit the Union as a whole . . . ’’). However, while the equityimplications of regional imbalances are self-evident, the self-confidence of thestatement about efficiency is puzzling. The reason why is the shortage oftheoretical arguments on the emergence, not to say on the welfare impact, ofregional imbalances that might support the Commission’s view.

Indeed, if anything, conventional wisdom supports the opposing view, accordingto which the concentration of means within the most productive regions is oftenthe optimal strategy to maximize global income (Rahman, 1963; Takayama,1967). Accounting for regional disparities mitigates such recommendations butdoes not upset them: important discrepancies are usually required between thebeginning and the end of the planning period because it is desirable in themeantime to invest into the most efficient regions (Michel et al., 1983). Thus,conventional wisdom seems to accept the existence of an inherent trade-off

2between regional equity and efficiency (Martin, 1999).This paper has two objectives. The first is to investigate the conditions under

which the equity–efficiency trade-off does not exist, i.e. agglomeration is indeedsocially undesirable. In doing so, we are forced to depart from the neoclassicalworld of constant returns to scale and perfect competition in which economicintegration has no dramatic effects on the spatial distribution of economicactivities: geographical discrepancies are not amplified and eventually disappear.Indeed, as shown by Starrett (1978), the competitive model cannot generate

3economic agglomerations unless technological externalities are assumed to exist.Without denying the relevance of such externalities at the local level, they areunlikely to play a major role in explaining interregional imbalances. Suchphenomena are better understood by appealing to pecuniary externalities, whicharise from imperfect competition and increasing returns in the presence ofmarket-mediated linkages between firms and consumers–workers. In addition,while both types of externalities are natural components of any completeexplanation of economic clustering, pecuniary externalities have one majorintellectual advantage. Technological externalities are often ‘black boxes’. On thecontrary, being focused on economic interactions mediated by the market, theorigin of pecuniary externalities is clearer. In particular, as it will be shown in this

2Note that there exists a huge literature in urban economics that focusses on the impact of variousexternalities on the city structure. This literature casts doubt on the efficiency of the market outcomewithin cities as well as across cities. A major difference between these contributions and ours is that weassume imperfect competition and pecuniary externalities, whereas urban economists, e.g. Henderson(1974) and Kanemoto (1980), typically consider perfect competition and technological externalities. Werefer the reader to Anas et al. (1998).

3See, e.g. Ottaviano and Thisse (2001) for more details.


paper, their intensity can be traced back to the values of fundamental micro-economic parameters such as the intensity of returns to scale, the strength of firms’market power, the level of barriers to goods and factor mobility. This is crucial

4from the normative viewpoint. Whatever externalities are at work, prices do notreflect the social values of goods and services so that market outcomes aretypically inefficient. However, we will show that, since the nature of pecuniaryexternalities is easier to identify, the corresponding inefficiencies can be betterunderstood.

In this paper, we build on some recent developments of economic geographythat study the impact of frictions to goods and factors mobility on the location ofimperfectly competitive industries in the presence of increasing returns to scale(Fujita and Thisse, 1996; Ottaviano and Puga, 1998; Fujita et al., 1999).Specifically, we rely on Krugman’s (1991) classical ‘core-periphery model’ asmodified by Ottaviano et al. (2001) who propose the adoption of an alternativedemand system that, contrary to the original one, is amenable to analytical

5solutions and proper welfare comparisons.In Krugman’s set-up there are two regions, two sectors and two specific factors.

The traditional sector employs unskilled workers to produce a homogeneous goodwhich is perfectly competitive and freely traded. The modern sector employsskilled workers to supply a differentiated good which is monopolistically competi-tive and costly traded. Finally, unskilled workers are assumed to be much lessmobile than skilled workers.

In the light of past migration flows (see, e.g., Sowell, 1996), this last assumptionneeds some additional justification. Empirical evidence regarding migration flowsacross the EU suggests that skilled workers are more mobile than the unskilled(e.g. Shields and Shields (1989, pp. 279, 285 and 287)). This may be explained asfollows. It is a documented fact that welfare differentials based on marketconsumption goods and wages decrease across industrial regions as economicdevelopment proceeds (Alonso, 1980). In turn, this implies that nonmarket factors,such as the attachment of people to their region of origin, are able to show agrowing importance in individual decisions to migrate. More precisely, once theyhave reached some living thresholds, individuals are less willing to trade theirfamily and social environment against more personal consumption. Within the EU,this is well illustrated by the fact that traditional migration flows, such as thosefrom Southern to Northern Italy have stopped during the last decades. By contrast,there has been an increase in the mobility of skilled workers within the EU. Thisphenomenon may be explained by the following factors. First, education generates

4See Abdel-Rahman and Fujita (1990).5Most of the models surveyed in Fujita et al. (1999) and Ottaviano and Puga (1998) are based on a

nested-CES demand system which is not really suited to address the question of the efficiency oféconomic agglomeration because the marginal utility of the numeraire is not constant but differs

between agents with different endowments.


human capital which is easily transferable to another region and eases the searchfor job, residency and a social environment to live in. Hence, the marginal rate ofsubstitution between the social environment and individual consumption would belarger for the skilled than for the unskilled. Second, regional wage differentials arelikely to be larger for the skilled than for the unskilled (Black, 1999). Put together,these two factors make migration more attractive for the former than for the latter.As a result, unskilled may be assumed to be much less mobile than skilled withinan area such as the EU.

Of course, one may wonder if the assumption of mobility of the skilled isreasonable in an area where linguistic and cultural barriers have often beenconsidered as major impediments to mobility. Some transformations are takingplace within the EU that make this assumption more plausible than it might seemat first sight. First, English is becoming the lingua franca of the business andscientific communities as well as that of most international institutions (‘‘They allspeak English, don’t they?’’). Second, the construction of a High Speed Railconnecting the main conurbations of the Union (such as the TGV) seems toenhance the mobility of the skilled. Although it is premature to conclude, it isworth mentioning that, since the Channel Tunnel has been built, the City ofLondon has already attracted a fairly large number of French financial operatorswho used to work in Paris. To be sure, services account for a substantial share ofthe modern sector and many services are produced where they are consumed.However, it is reasonable to expect high-level services workers to be mobile, thusmaking their agglomeration in a few large metropolitan regions possible. In thisperspective, our setting is reminiscent of the brain drain problem within large

6industrialized areas, as briefly discussed in the concluding section.Having said that, the core–periphery model reveals that for low transport costs

(broadly defined to encompass any impediment to trade) skilled labor mobilitycauses the modern sector to cluster in one region because the positive aggregatedemand effect of skilled workers’ immigration on profits more than offsets thenegative competition effect of modern firms’ inflow. Though path-breaking,Krugman’s seminal contribution falls short of providing any welfare evaluation ofthe implied pattern of regional specialization. This is our first aim: we want todetermine whether or not the spatial concentration of the modern sector is sociallydesirable and to assess its pros and cons by disentangling the various externaleffects which cause its emergence. When skilled workers move from one region toanother, they impose negative pecuniary externalities on the whole economy, thusmaking our analysis different from the competitive approach to migration (see, e.g.Borjas, 1995). More precisely, we will see that the skilled workers neglect theimpact of their move on product and labor markets in both destination andoriginating regions. Consequently, agglomeration may lower aggregate surplus.

6See Fujita and Tabuchi (1997) for a discussion of the Japanese case, which seems to bear someresemblance with the EU case.


´This inefficiency result has much in common with that obtained by Benabou(1993, 1996) who obtains suboptimal urban segregation between unskilled andskilled workers. However, the underlying mechanisms are quite different because,while he considers externalities of the ‘social capital’ variety, we focus onpecuniary externalities associated with imperfect competition.

Previewing our results, we begin by pointing out that the market outcome issocially desirable when transport costs are either high or low. While in the formercase activities are dispersed, in the latter they are agglomerated. By contrast, forintermediate values of these costs, the market leads to the agglomeration of themodern sector whereas it is efficient to keep it geographically dispersed. In thiscase, the equity–efficiency trade-off disappears. This is an interesting finding inthat low transport costs may be viewed as corresponding to shipping costs betweensmall-sized regions, whereas large costs would instead be the counterpart of tradecosts between large trading blocks. Intermediate values would therefore corre-spond to shipping costs between regions belonging to medium-sized areas such asthe EU. This interpretation would suggest that there could well be too muchgeographical concentration within the EU after the effects of the Single Marketwere completed (Martin, 1998). If so, our finding would provide an efficiency-grounded case for an active regional policy at the level of the European Union.

Once we have established the suboptimality of the market-delivered economiclandscape, we can move on towards our second objective, that is to understandwhether the actual mechanics of European policy making has any chance ofdelivering the economically efficient outcome. In principle, the effectiveness of theCommission is hampered by at least two crucial problems. First, the costs of aninefficient economic landscape are likely to be unevenly distributed amongdifferent interest groups so that any active regional intervention is bound togenerate a conflict between opposite interests (Mazey and Mitchell, 1993).Second, effective intervention requires enormous information-gathering capa-bilities that, due its relatively small bureaucracy, the Commission is unlikely topossess (Gorges, 1996). This second aspect appears even more severe if oneconsiders that, even though sometimes (under the ‘assent’ and ‘cooperation’procedures) it has to ask for the opinion of the European Parliament, theCommission is an unelected body and thus it receives no direct feed-back fromany electorate (Mazey and Richardson, 1993). Both problems have led theCommission to adopt the role of an intermediator between conflicting interests thatare thus urged to get organized as pressure groups in order to improve thetransmission of information about their collective needs (Gorges, 1996). In 1993as many as 525 interests groups were officially recognized and regularly consultedby the Commission (Mazey and Richardson, 1993) while the current estimate ofpeople involved in interest representation in Brussels is close to 13,000 (TheEconomist, 1998).

Our model lends itself quite naturally to studying the political economics of thedistributive effects generated by the formation and the development of a coreregion. Specifically, it can be used to show that the population of unskilled


workers as a whole is always negatively affected by the emergence of a core–periphery structure because their access to the output of the modern sectorbecomes too costly. In other words, members of the traditional sector oppose thedeepening of the European unification process. By contrast, the skilled workersmay be either better off or worse off when they are agglomerated. In the formercase, the two groups of workers have conflicting interests, which may give rise tothe emergence of competing interest groups.

By building on the literature on informational lobbying (Potters and vanWinden, 1992; Austen-Smith, 1993) and on the theory of menu auctions(Bernheim and Whinston, 1986), we are able to show that, first, lobbying byconflicting pressure groups for or against the mobility of the skilled workersallows one to implement the second best optimum economic landscape and,second, the actual coalitions of interests that are formed are immaterial for theattainment of efficiency since their composition has only redistributive implica-tions. These results hold if lobbying is ‘competitive’ in the sense that all interestshave free access to the Commission, all face the same costs of collective actionand the Commission simply acts as a mediating organization (‘honest broker’).This provides a useful benchmark for political conduct. It implies that, forimplementing efficient regional policies, the Commission has to lend a candid earto all interests as well as to promote their participation to the lobbying processwith particular emphasis on those interests that, because they are diffuse, facelarger costs of collective action (Olson, 1980). Under this respect, our resultsprovide some rationale to the extension of qualified majority voting by the SEA,which in 1992 substituted the national veto power with qualified majority votingon measures necessary to complete the internal market and to new policy sectors(R&D, health and safety, environment). This has increased ‘‘the appeal of, and theneed for, rapid transnational interest definition, aggregation, and coordination’’(Gorges, 1996), thus favoring the coalition of diffuse Euro-wide interests.

The remainder of the paper is organized as follows. The next section presents asimple model yielding a core–periphery structure when transport costs are low.Section 3 identifies the conditions under which the geographical agglomeration ofskilled workers is a market equilibrium, the first best optimum or a second bestoptimum in which firms must break even. Section 4 deals with the welfare ofskilled and unskilled workers. Section 5 discusses the desirability of agglomerationfrom the social point of view as well as from the standpoint of each group ofworkers. Section 6 analyzes the lobbying game involving the groups of skilled andunskilled workers. Section 7 concludes.

2. A simple model of agglomeration

Consider a spatial economy formed by two sectors, called modern andtraditional, two regions, H and F, and two production factors, L and A, the firstbeing mobile and the second immobile. Although other interpretations are


possible, it is convenient to think of the mobile factor as skilled workers and theimmobile one as unskilled workers. The modern sector exhibits scale economiesand operates under monopolistic competition. The corresponding output ishorizontally differentiated. There is a continuum of firms whose number N isdetermined under free entry. Each firm produces a single variety by means of afixed amount f of skilled labor L. The traditional sector produces a homogeneousgood under constant returns to scale, using unskilled labor A as the only input: oneunit of A is required to produce one unit of output. This good is freely traded and

ís chosen as the numeraire. The varieties of the modern sector are traded at a costóf t units of the numeraire per unit shipped between the two regions. The

economy is endowed with A units of unskilled who are equally split between7regions, and with L units of skilled whose fraction 0 # l # 1 is located in region

H.Preferences are identical across individuals and described by the following

quasi-linear indirect utility function which is symmetric in all varieties:

N N N 2

b 1 cN c2]] ]V( y; p(i), i [ [0,N]) 5 2 aEp(i) di 1 E[ p(i)] di 2 Ep(i) di2 23 40 0 0

]1 y 1q (1)0

]where p(i) is the price of variety i [ [0,N], y the consumer labor income, and q08´her initial endowment in the numeraire. In (1), a . 0 expresses the intensity of

preferences for the differentiated product; b . 0 means that the representativeconsumer is biased toward a dispersed consumption of varieties, thus reflecting alove for variety; c . 0 expresses the substitutability between varieties so that thehigher c, the closer substitutes the varieties. Finally, we assume that the initial

] éndowment q in the numeraire is large enough for the consumption of the0

ńumeraire to be strictly positive at the market equilibrium and optimal solutions.Throughout the paper, we will encounter several conditions involving the variousparameters of the model. In particular, we will assume that the least demandingcondition regarding trade costs and preference parameters, that is,

t # a /b (2)

is always satisfied. This condition is necessary for any single firm to find itprofitable to sell in the foreign market, regardless of whom pays for the tradecosts.

7This assumption is made to avoid exogenous differences between regions. However, the analysiscould be extended to the case of an uneven distribution of the traditional sector without affecting itsmain insights.

8The direct utility behind (1) is the standard quadratic utility yielding linear demand functions (seeOttaviano et al., 2001, for more details).


Skilled labor market clearing implies that the number n of firms belonging toH

the modern sector and located in region H is equal to:

n 5 lL /f (3)H

so that the number of firms in F is

n 5 (1 2 l)L /f (4)F

Consequently, the total number of firms (varieties) in the economy is fixed andequal to N 5 L /f.

Entry and exit are free so that profits are zero in equilibrium. Hence, (3) and (4)imply that any change in the population of skilled located in one region must beaccompanied by a corresponding change in the mass of firms. By (3) and (4), thedemand and supply of skilled in each region are equal. As a result, thecorresponding equilibrium wages are determined by a bidding process amongfirms which ends when no firm can earn a strictly positive profit at the equilibriummarket prices.

Firms are assumed to take advantage of positive trade costs to segment markets,that is, each firm sets a price specific to the market in which its product is sold.Even within a unified economic area, empirical work shows that firms succeed toprice discriminate between spatially separated markets (McCallum, 1995; Headand Mayer, 2000). As will be shown below, in equilibrium arbitrage is neverprofitable by third parties.

In the sequel, we focus on region H. Things pertaining to region F can bederived by symmetry. Using the assumption of symmetry between varieties andRoy’s identity, individual demands for a representative firm in H are given by:

q 5 a 2 (b 1 cN) p 1 cP (5)HH HH H

and

q 5 a 2 (b 1 cN) p 1 cP (6)HF HF F

where p ( p ) is the price set in H (F ) by a firm located in H andHH HF

P ; n p 1 n pH H HH F FH

P ; n p 1 n pF H HF F FF

Clearly, P /N and P /N can be interpreted as the price indexes prevailing inH F

regions H and F.A representative firm in H maximizes its profits defined by:

P 5 p [a 2 (b 1 cN) p 1 cP ] [A /2 1 lL] 1H HH HH H

( p 2 t) [a 2 (b 1 cN) p 1 cP ] [A /2 1 (1 2 l)L] 2 fwHF HF F H

where A /2 stands for the unskilled population in each region while w is theH

skilled workers’ wage prevailing in H.


Market prices are obtained by maximizing profits, while the wages of the skilledare determined as described above by equating the resulting profits to zero. Sincewe have a continuum of firms, each one is negligible in the sense that its actionhas no impact on the market. Hence, when choosing its prices, a firm in Haccurately neglects the impact of its decision over the two price indices P and P .H F

In addition, because firms sell differentiated varieties, each one has somemonopoly power in that it faces a demand function with finite elasticity. On theother hand, since the price index enters the demand function as an additive term(see (5) and (6)), a firm must account for the distribution of the firms’ pricesthrough some aggregate statistics, given here by the average market price, in orderto find its equilibrium price. As a consequence, our market solution is given by aNash equilibrium with a continuum of players in which prices are interdependent:each firm neglects its impact on the market but is aware that the market as awhole has a non-negligible impact on its behavior.

Since profit functions are concave in own price, solving the first orderconditions for profit maximization with respect to prices yields the equilibriumprices:

1 2a 1 tcN(1 2 l)]]]]]]*p 5 (7)HH 2 2b 1 cN

1 2a 1 tcNl]]]]*p 5 (8)FF 2 2b 1 cN

t]* *p 5 p 1 (9)HF FF 2

t]* *p 5 p 1 (10)FH HH 2

Thus, the equilibrium prices under monopolistic competition depend on the totalnumber of active firms as well as on their distribution between the two regions. Inparticular, using (2) we observe that more firms in the economy leads to lowermarket prices for the same spatial distribution (l, 1 2 l) because there is morecompetition in each regional market. Similarly, both the prices charged by localand foreign firms fall when the mass of local firms increases because competitionis fiercer. Equilibrium prices also rise when the degree of product differentiation,inversely measured by c, increases provided that (2) holds. All these results agreewith what is known in industrial organization and spatial pricing theory (Tirole,

* *1988, Chapter 7). In addition, it is readily verified that p 2 p , t so that theHF HH

prohibition of arbitrage associated with the assumption of segmented markets isnot binding.

Subtracting t from (9) and (10), we see that firms’ prices net of trade costs arepositive regardless of the distribution of the skilled if and only if


2af]]]t , t ; (11)trade 2bf 1 cL

The same condition must hold for consumers in F (H ) to buy from firms in H (F ),i.e. for the demand (6) evaluated at the prices (7) and (8) to be positive for all l.Clearly (11) is more restrictive than (2). From now on, condition (11) is assumedto hold. Consequently, we consider a setting in which there is a priori intra-industry trade.

Finally, local sales rise with t because of the higher protection enjoyed by thelocal firms but exports fall for the same reason. It is easy to check that theequilibrium operating profits earned by a firm established in H on each separatedmarket are as follows:

2* *P 5 (b 1 cN) ( p ) (A /2 1 lNf) (12)HH HH

*where P denotes the profits earned in H while the profits made from selling inHH

F are

2* *P 5 (b 1 cN) ( p 2 t) [A /2 1 (1 2 l)Nf] (13)HF HF

*Increasing l has two opposite effects on P . First, the equilibrium price (7)HH

falls as well as the quantity of each variety bought by each consumer living inregion H. However, the total population of consumers residing in this region isnow larger so that the profits made by a firm located in H on local sales mayincrease. What is at work here is a positive demand effect due to the increase inthe local population that may compensate firms for both the adverse price effectand the reduction in individual demand due to the presence of a larger number oflocal varieties.

The individual consumer surplus S in region H is then as follows (a symmetricH

expression holds in region F ):

b 1 cN 2 2]]S (l) 5 2 a[l p 1 (1 2 l) p ]N 1 [l p 1 (1 2 l) p ] NH HH FH HH FH2c 2 2]2 [l p 1 (1 2 l) p ] NHH FH2

When evaluated at the equilibrium prices (7) and (10), the resulting expression*S (l) can be shown to be quadratic in l. Differentiating twice this expressionH

*with respect to l shows that S (l) is concave. Furthermore, (11) implies thatH

*S (l) is always increasing in l over the interval [0,1]. Hence, as more firms enterH

in H, the surplus of residents rises because local competition becomes fiercer;however, this effect gets weaker and weaker as the number of local firmsincreases.

The equilibrium wage of the skilled prevailing in region H may be obtained by* * *evaluating w (l) 5 (P 1 P ) /f, thus yielding the following expression:H HH HF


bf 1 cL A2]]]]] ]* H S Dw (l) 5 [2af 1 tcL(1 2 l)] 1 lLH 2 2 24(2bf 1 cL) f

A2 ]F GJ1 [2af 2 2tbf 2 tcL(1 2 l)] 1 (1 2 l)L (14)2

which, after simplifying, turns out to be quadratic in l. Standard, but cumbersome,*investigations reveal that w (l) is concave and increasing (convex and decreasing)H

in l when f is large (small) as well as when t, c, A, and L are small (large). Inother words, the equilibrium wage rises with the local number of firms whenincreasing returns are strong, trade costs are low, and varieties are poor substitutes.

* *In this case, both S (l) and w (l) increase with l.H H

3. Equilibrium and optimum spatial configurations

3.1. The market outcome

We now ask whether for a given spatial distribution of skilled workers, (l,1 2 l), there is an incentive for them to migrate and, if so, what direction the flowof migrants will take. Following the tradition of economic geography, we assumethat skilled workers care only about their current utility levels. Accordingly, if theyobserve that one location offers a higher indirect utility than the other, they want tomove to that location.

Formally, the distribution l [ [0,1] is a spatial equilibrium when no skilledworker may get a higher utility level by changing location. Given that the indirectutility in region H is as follows:

]* *V (l) 5 S (l) 1 w (l) 1qH H H 0

a spatial equilibrium arises at l [ (0,1) when

* * * *DV(l) ;V (l) 2V (l) 5 S (l) 2 S (l) 1 w (l) 2 w (l) 5 0H F H F H F

in which case we have a dispersed configuration. However, an equilibrium mayalso arise at l 5 0 when DV(0) # 0 or at l 5 1 when DV(1)$0, in which case wehave an agglomerated configuration.

In order to study the stability of a spatial equilibrium, we assume that local labormarkets adjust instantaneously when some skilled move from one region to theother. More precisely, the number of firms in each region must be such that thelabor market clearing conditions (3) and (4) remain valid for the new distributionof skilled. Wages are then adjusted in each region for each firm to earn zero profitseverywhere. The assumed adjustment process is as follows:

~l ; dl /dt 5 l(1 2 l)DV(l) (15)


in which t is time while the nonnegative factor l ? (1 2 l) has been introduced for9~l to be zero when l reaches the boundary of the unit interval. Clearly, a spatial

~equilibrium implies l 5 0. A spatial equilibrium is stable for (15) if, for anymarginal deviation from the equilibrium, this equation of motion brings thedistribution of skilled back to the original one. Therefore, the agglomeratedconfiguration is always stable when it is an equilibrium while the dispersedconfiguration is stable if and only if the slope of DV(l) is nonpositive in aneighborhood of this point.

The following comments are in order about the dynamics of migration of bothconsumers and firms. The immobility of the unskilled is a centrifugal force, atleast as long as there is trade in the differentiated product between the two regions.The centripetal force is more elaborated. If an increasing number of firms arelocated in region H, there are two effects at work. First, more varieties are locallyavailable (trade cost effect). Second, (7) and (10) imply that the equilibrium pricesof all varieties sold in H are lower (competition effect). Both effects generate ahigher indirect utility. This in turn induces some consumers to migrate toward thisregion. The resulting increase in the number of consumers creates a larger demandfor the output of the modern sector in region H (demand effect), which thereforeleads more firms to locate there.

The indirect utility differential DV(l) is obtained by plugging the equilibriumprices (7)–(10) and the equilibrium wages (14) into (1):

MDV(l) 5 Ct(t 2 t) ? (l 2 1/2) (16)

where

L(bf 1 cL) 2]]]]]C ; [2bf(3bf 1 3cL 1 cA) 1 c L(A 1 L)] . 02 22f (2bf 1 cL)

and

4af(3bf 1 2cL)M ]]]]]]]]]]t ; . 022bf(3bf 1 3cL 1 cA) 1 c L(A 1 L)

It follows immediately from (16) that l 5 1/2 is always an equilibrium. SinceC . 0, for l ± 1/2 the indirect utility differential has always the same sign as

Ml 2 1/2 if and only if t , t , otherwise it has the opposite sign. In particular,when there are no increasing returns in the manufacturing sector (f 5 0) or whenvarieties are not differentiated (c → `), the coefficient of (l 2 1/2) is always

Mnegative since t 5 0 so that dispersion is the only (stable) equilibrium. This

9This ad hoc process leads to a fairly good approximation of forward-looking behavior in thepresence of quadratic adjustment costs whenever skilled workers discount the future heavily and/ortheir migration process is very slow (Ottaviano, 1999; Ottaviano et al., 2001).


shows the importance of increasing returns and product differentiation for thepossible emergence of an agglomeration.

MIt remains to determine when t is lower than t . This is so if and only iftrade

2 2 2 26b f 1 8bcfL 1 3c L]]]]]]]A /L . . 3 (17)

cL(2bf 1 cL)

This inequality means that the population of unskilled is large relative to thepopulation of skilled. When (17) does not hold, the coefficient of (l 2 1/2) in (16)is always positive for all t , t so that agglomeration is the only equilibriumtrade

configuration with trade.MWhen t , t , the symmetric equilibrium is unstable and skilled agglomerate in

region H (F ) provided that the initial fraction of skilled residing in this regionexceeds 1/2. In other words, agglomeration arises when the trade cost is lowenough, as in Krugman (1991) and for similar reasons. In contrast, for large trade

Mcosts, that is, when t . t , it is straightforward to see that the symmetricMconfiguration is the only stable equilibrium. Hence, the threshold t corresponds

to both the critical value of t at which symmetry ceases to be stable (the ‘breakpoint’) and the value below which agglomeration is stable (the ‘sustain point’);this follows from the fact that (16) is linear in l.

MProposition 1. Assume that t , t and that (17) holds. Then, if t . t thetradeMsymmetric configuration is the only stable equilibrium with trade; if t , t there

are two stable equilibria corresponding to the agglomerated configurations withMtrade; if t 5 t there is a continuum of equilibria with trade.

The nature of trade varies with the type of configuration emerging inequilibrium. In the dispersed configuration, there is only intra-industry trade in themodern sector; in the agglomerated equilibrium, trade is only inter-industry: thecore region only imports the output of the traditional sector from the peripheralregion which only imports the output of the modern sector from the core region.

3.2. Efficiency: the first best outcome

In our economy, there are four homogeneous groups of individuals: theunskilled and the skilled residing in region H or F. We assume that the planner isable (i) to assign any number of skilled workers (or, equivalently, of firmsbelonging to the modern sector) to a specific region, (ii) to use lump sum transfersfrom all workers to pay for the loss firms incur while pricing at marginal cost, and(iii) to use interpersonal transfers across the above four groups of individuals.Since we assume transferable utility, the social welfare may be expressed as thesum of all workers’ indirect utilities. Since workers may be aggregated within fourhomogeneous groups of workers, namely the unskilled and skilled residing in


region H or F, the planner chooses l in order to maximize the following socialwelfare function:

A A] ]W(l) ; [S (l) 1 1] 1 lL[S (l) 1 w (l)] 1 [S (l) 1 1]H H H F2 2

1 (1 2 l)L[S (l) 1 w (l)] (18)F F

which is the sum of each type workers’ indirect utilities and where all prices havebeen set equal to marginal cost:

o o o op 5 p 5 0 and p 5 p 5 tHH FF HF FH

thus implying by (12) and (13) that operating profits are zero. Consequently,o ow (l) 5 w (l) 5 0 for every l so that firms do not incur any loss and lump sumH F

transfers to them are unnecessary. Hence (18) becomes:

O OW(l) 5 (C /2)Lt(t 2 t)l(l 2 1) 1 constant (19)

where

LO ]C ; [2bf 1 c(A 1 L)] . 02f

and

4afO ]]]]]t ; . 0 (20)2bf 1 c(A 1 L)

O OThe function (19) is strictly concave in l if t . t and strictly convex if t , t .2Furthermore, since the coefficients of l and of l are the same (up to their sign),

this expression has always an interior extremum at l 5 1/2. As a result, the2optimal choice of the planner is determined by the sign of the coefficient of l ,

Othat is, by the value of t with respect to of t .Thus we have:

OProposition 2. If t . t , then the symmetric configuration is the first bestO Ooptimum; if t , t any agglomerated configurations is an optimum; if t 5 t

there is a continuum of optima.

As expected, it is socially desirable to agglomerate the modern sector into asingle region once trade costs are low, increasing returns in the modern sector arestrong enough and/or the output of the modern sector is differentiated enough. Inparticular, in the absence of increasing returns (f 5 0) and of product differentia-tion (c → `), it is never efficient to have agglomeration.


3.3. Efficiency: the second best outcome

We now assume that lump sum transfers are not available to the planner who isonly able to assign locations to the skilled workers. In such a context, the socialwelfare function is still given by (18) but it is evaluated at the equilibrium prices(7)–(10):

S SW (l) 5 (C /2)Lt(t 2 t)l(l 2 1) 1 constantS

where

L(bf 1 cL)S 2]]]]]C ; [8bf(3bf 1 2cL 1 cA) 1 3c L(A 1 L)] . 02 24f (2bf 1 cL)

and

16af(3bf 1 cL)S ]]]]]]]]]]]t ; . 0 (21)28bf(3bf 1 2cL 1 cA) 1 3c L(A 1 L)

The choice of the planner is similar to that described in the first best case exceptfor the critical value of t. In particular, when f 5 0, the second best outcomenever involves agglomeration. Hence we have:

SProposition 3. Assume that (17) holds. If t . t , then the symmetric configurationSis the second best optimum; if t , t any agglomerated configuration is a second

Sbest optimum; if t 5 t there is a continuum of second best optima.

4. Welfare for the skilled and unskilled workers

Our purpose is now to study the welfare of the four groups of workers describedabove (see (18)). To this end, we proceed in two steps. In the first one, we studythe aggregate welfare of the skilled and of the unskilled, each as a whole. We thenstudy how differences in location affect the well-being of individuals within eachof these two groups. This division turns out to be convenient to disentangle thevarious effects at work.

Consider the welfare of the skilled, which is defined up to an additive constantby:

W (l) 5 lL[S (l) 1 w (l)] 1 (1 2 l)L[S (l) 1 w (l)]L H H F F

which, evaluated at at the equilibrium prices (7)–(10), becomes:

L LW (l) 5 (C /2)Lt(t 2 t)l(l 2 1) 1 constantL

where


L(bf 1 cL)L 2]]]]]C ; [8bf(3bf 1 2cL 1 cA) 1 c L(2A 1 3L)]2 24f (2bf 1 cL)

and

16af(3bf 1 cL)L ]]]]]]]]]]]t ; (22)28bf(3bf 1 2cL 1 cA) 1 c L(2A 1 3L)

which is the critical value of t for which the whole group of skilled preferagglomeration to any other distribution. Hence, l 5 1/2 is always an extremum

Lfor the skilled. This extremum is a maximum if and only if t . t . Otherwise, it isa minimum and the welfare of the skilled is maximized at l 5 0, 1. So, it is clearthat the skilled are not always better off when they are agglomerated. When skilledworkers agglomerate in one region they do not take into account the negativeeffect on local firms’ profits stemming from increased local competition. Also,they do not take into account the positive effect on local firms’ profits due toincreased local market size. While the sign of the net pecuniary externality theygenerate is a priori ambiguous, the foregoing analysis reveals that skilled workers’neglect of the former effect dominates their neglect of the latter.

As to the welfare of the unskilled, up to a constant, it is equal to:

A]W (l) 5 [S (l) 1 S (l)]A H F2

which, evaluated at the equilibrium prices (7)–(10), becomes:

A 2W (l) 5 2 C t l(l 2 1) 1 constantA

where

3 2L Ac (bf 1 cL)A ]]]]]C ; 2 28f (2bf 1 cL)

Since W is concave, l 5 1/2 is the unique maximizer, implying that the unskilledA

workers as a whole always prefer the dispersed configuration to the agglomeratedone.

However, agglomeration gives rise to redistributive effects inside the population*of unskilled. Since, as seen in Section 2, the surplus S is increasing in l, anyH

increase from l 5 1/2 makes the unskilled in H better off and, by symmetry, the


10 *unskilled in F worse off. Furthermore, since we know that S is concave in l,H

the marginal gain of the former is lower than the marginal loss of the latter:agglomeration hurts the unskilled in the periphery more than it helps those livingin the core region.

To sum-up, we have:

LProposition 4. If t . t , then the skilled workers as a group prefer the symmetricLconfiguration to any other; if t , t they prefer the agglomerated configuration to

Lany other; if t 5 t they are indifferent. Regardless of the value of t, the unskilledworkers as a group prefer the dispersed configuration to any other.

5. Is agglomeration desirable? And for whom?

As seen in Section 3.1, deepening integration is likely to lead to the emergenceof a core region accommodating the entire modern sector. We now wish todetermine whether or not such an agglomeration is efficient.

Since there are several distortions and external effects at work, the analysis isnot obvious. Besides the standard distortion due to the fact that firms do not priceat marginal cost, there are pecuniary externalities that matter for the level ofwelfare. In particular, when moving skilled workers impose a pecuniary externalityon the unskilled in both regions. Their move affects the intensity of localcompetition as well as the amount spent by the unskilled on trade costs. Inaddition, when skilled workers move from one region to the other, they do notaccount either for the impact of their migration on the other skilled. Such move bythe skilled affects not only the intensity of competition but also the level ofdemand inside each region, and, therefore, their wages. Note, however, that thereis no over- or under-entry effect: the number N of firms is the same in equilibriumand at the optimum since it is determined by the technology and equal to L /f.

In order to study the impact of these various effects, we need to rank thedifferent threshold values for t obtained in the previous sections.

5.1. Equilibrium vs. optimum

O S MSome simple calculations show that t , t , t when (17) holds. TheseO Sinequalities can be interpreted as follows. First, t , t , namely agglomeration is

desirable for higher values of the transport cost in the second best. This is becausethe individual demand elasticity is lower in the first best (marginal cost pricing)than in the second best (Nash equilibrium pricing), so that regional price indices

10Note that this comparison is made for a given level of transport costs. Starting from any t forwhich equilibrium involves dispersion and lowering transport costs so that agglomeration arises in H, itcan be shown that a sufficiently large decrease in t also makes the unskilled in F better off.


are less sensitive to a decrease in t. The fall in transport costs must be sufficiently11large to make the agglomeration of the mobile workers socially desirable.

S MSecond, we also have t , t . This is because skilled workers do notinternalize the negative external effects they impose on the unskilled who alwaysprefer dispersion as shown by Proposition 4. Although the unskilled who reside inthe region of destination are better off since the price index of the modern sectorgoes down, they are not able to compensate the unskilled who stay behind. Hence,even though the skilled have incentives to move, these incentives do not reflect thesocial value of their move.

This discrepancy is even stronger when we compare the first best outcome andthe market equilibrium because we have just seen that marginal cost pricing ismore favorable to dispersion. As a result, the market yields an agglomerated

O Mconfiguration for a whole range (t , t , t ) of trade cost values for which it issocially desirable to have a dispersed pattern of activities.

O MAccordingly, when transport costs are low (t , t ) or high (t . t ) the marketyields the optimal configuration so that no regional policy is required from theefficiency point of view, although equity considerations might still justify such apolicy when agglomeration arises. On the contrary, for intermediate values of

O Mtransport costs (t , t , t ), the market provides excessive agglomeration, thusjustifying the need for an active regional policy in order to foster the dispersion ofthe modern sector on both the efficiency and equity grounds.

O MWhen t , t , t , the first best optimum requires the implementation oflump-sum transfers. If such transfers are not feasible while the planner is still ableto control the locations of the skilled workers, it remains true that, for some values

S Mof the trade costs (t , t , t ), the market outcome is agglomeration whereas thesecond best optimum involves dispersion. Nevertheless, the discrepancy is less

M Ssevere than with the first best outcome since the gap t 2 t is narrower thanM O

t 2 t . Still, in both cases, we have an illustration of what was claimed by theEuropean Commission in the quotation made in the introduction.

5.2. Agglomeration and the welfare of skilled and unskilled

We now want to determine to what extent the cooperative behavior of skilledworkers in the migration process affects the well-being of the two groups ofworkers evaluated at the equilibrium prices. In this case, the skilled are assumed toact together in order to internalize all the effects that an individual mover has onthe other skilled. However, they still disregard the impact that their decision has on

11Although the argument above seems to rest very much on the linearity of demands, it is likely tohold true for any functional form such that the elasticity of demand of a variety increases with its price.This happens for any demand function that is concave or convex but less convex than a negativeexponential, a condition often met in spatial pricing theory (Greenhut et al., 1987, p. 25). A notableviolation of this property is given by isoelastic demands.


L S L Othe unskilled. This explains why t . t (as well as t . t ), as shown in Section4.

L MTurning to the well-being of the skilled, we must compare t and t . Theranking is not a priori obvious because skilled are free to move but movers do notinternalize the competition, trade cost and demand effects (described in Section3.1) that they impose upon the other skilled. Nevertheless, a direct comparison

M Lshows that t . t . Hence, since the unskilled as a whole always preferL Mdispersion, there exists a domain of trade cost values (t , t , t ) for which each

group of workers as a whole is worse off at the market agglomeration than at thedispersed configuration.

To summarize, thresholds can be unambiguously ranked as follows:

O S L M0 , t , t , t , t

Observe that all these inequalities tend to vanish when varieties are very goodsubstitutes and /or when increasing returns are weak. More precisely, all thesethresholds go to zero as c goes to infinity (perfect substitutability) or f goes tozero (constant returns to scale). As discussed above, under perfect competition andconstant returns to scale, there is no inefficiency due to pecuniary externalities. Inother words, our inefficiency results rest crucially on the pecuniary externalitiesgenerated by skilled workers when moving, an issue that we now address.

5.3. Dissecting the inefficiency

To summarize our welfare results, it is useful to consider once more the problemof the second-best in which the planner maximizes (18) with respect to l subjectto market prices. Dropping the label * for notational convenience, the first ordercondition requires to equate to zero the sum of three terms:

A]9 9 9 9 9 9 9W 5 DV ? L 1 [(S 1 w ) ? l 1 (S 1 w ) ? (1 2 l)] ? L 1 (S 1 S ) ?S H H F F H F 2

(23)

On the right hand side, the first term is the indirect utility differential that anindependent skilled mover takes into account when migrating. The second andthird terms stand for the external effects that she imposes on, respectively, theother skilled and the unskilled. From previous results, we know that these threeterms equate zero at l 5 1/2 so that the dispersed configuration always satisfies

9the first order conditions of the planner (W 5 0), of the skilled as a wholeS

9 9 9 9(DV 1 [(S 1 w ) ? l 1 (S 1 w ) ? (1 2 l)] 5 0) and of the unskilled as a wholeH H F F

9 9(S 1 S 5 0), as well as the indifference condition of an individual skilled workerH F

(DV5 0). Consequently, any discrepancy between the market outcome and theoptimum must arise from the second order condition:


99 9 9 9 9W 5 2 ? [(S 1 w ) 2 (S 1 w )] ? LS H H F F

A]99 99 99 99 99 991 [(S 1 w ) ? l 1 (S 1 w ) ? (1 2 l)] ? L 1 (S 1 S ) ? (24)H H F F H F 2

where we have used the fact that the derivative of DV with respect to l is equal to9 9 9 9(S 1 w ) 2 (S 1 w ). Condition (24) can be rearranged so as to give:H H F F

A 1 L L]] ]99 9 9 9 9 99 99 99 99W 5 2 ? [(S 1 w ) 2 (S 1 w )] ? L 1 (S 1 S ) ? 1 (w 1 w ) ?S H H F F H F H F2 2

(25)

where the first term on the right hand side is positive whenever agglomeration isthe stable outcome, the second term on the right hand side is always negative,while the third is positive (negative) whenever trade costs are above (below) thevalue:

8afP ]]]]]t ;4bf 1 c(A 1 L)

Mwhich is larger than t .Therefore, when, for l ± 1/2, a skilled worker finds it individually rational to

9 9 9 9move to the larger region (i.e., when (S 1 w ) 2 (S 1 w ) . 0 as implied byH H F FM

t , t ), she neglects the collective welfare losses caused to the unskilled as99 99expressed by (S 1 S )A /2. She also neglects the negative effect on the welfareH F

of the skilled, expressed by

99 99 99 99(S 1 S )L /2 1 (w 1 w )L /2 (26)H F H F

M Pin which both terms are negative since t , t . Hence, within the skilledLpopulation the gainers cannot compensate the losers, thus explaining why t ,

Mt . This is because of the following general equilibrium effect: at the margin,fiercer price competition in the larger region depresses operating profits, thuswages, more than it increases them in the smaller region.

By giving weight only to the first positive term of (25), the skilled moveralways imposes a negative external effect on the economy as a whole. In addition,since the absolute value of (26) decreases with t, as trade costs fall the negativeexternal effects become weak enough to align the agglomerated market equilib-rium with the second best outcome. This explains why a planner would chooseagglomeration over a smaller interval of trade costs than the market does.


5.4. How to reach the optimum?

12How can the planner implement the second best? We know from (16) thatDV(l) is linear in l. Furthermore, (19) implies that the first order condition foroptimality is also linear and given by

S S9W (l) 5 C t(t 2 t)(l 2 1/2)S

9Since both DV(l) and W (l) are linear and pass through the point (1 /2, 0), theS

9equilibrium and the optimal configurations coincide when DV(l) ? W (l) $ 0. ByS

9contrast, the need for a corrective policy arises when DV(l) ? W (l) , 0. This isSS Mthe case for the range t , t , t . Then, using (23), it is readily verified that the

second best can be reached by imposing a lump sum tax on skilled workers (firms)in the larger region. More precisely, such a tax Y(l) is equal to the per capitavalue of the pecuniary externalities generated by migration, that is

S T9Y(l) ; DV(l) 2 W (l) /L 5 (C 2 C)t(t 2 t )(l 2 1/2)S

where

S S MC t 2 CtT ]]]]t ; . 0SC 2 C

T S S Mand, since t , t and t , t , t , Y(l) is always positive (negative) whenl . 1/2 (l , 1/2). When the tax is imposed, individual incentives to move aregiven by

9DV(l) 2Y(l) 5 W (l) /LS

implying that the equilibrium is aligned with the second best optimum. However,it must be kept in mind that implementing the tax schedule Y(l) requires a lot ofinformation on the part of the planner. This is why we turn to an alternativeinstitutional setting in the subsequent section.

6. The lobbying game

We now assume that a central authority (the European Commission?) has thepower to restrict the mobility of the skilled workers but does not have enoughinformation to enforce the second best optimum as described above. This can beviewed as an extreme form of admission fees, a tool often used in the literature onlocal public finance. Which interest groups are going to be formed and how can we

12Note that a similar analysis could be performed for the first best.


expect them to influence the political choice between free mobility (denoted by f )and mobility barriers (denoted by b) for the skilled workers?

We start with the analytically convenient case in which decisions are madebehind the veil of ignorance about which region is going to accommodate themodern sector so that it is reasonable to think of the following two interest groups:(i) the skilled who are potentially mobile and (ii) the unskilled who are immobile.This turns out to be useful to explain the more relevant case in which the would-becore region is known ex ante.

Following Bernheim and Whinston (1986), we model the political process as anauction over alternative policies whose bidders are the interest groups. Theauctioneer is the central authority that has no information about the welfareimplications of policies and uses the bid offers to extract information about thelobbies’ valuations of the alternative policies. Hence, this is a signalling game inwhich the two interest groups make offers, anticipating that the central authoritywill implement the policy that gets the higher bid. An interesting aspect of thisauction is the presence of externalities: the adoption of the policy favored by onegroup affects the welfare of the other. For simplicity, it is assumed that the firststage game is a game of complete information whose outcome is given by a (purestrategy) Nash equilibrium. Finally, we assume that bidding is ‘competitive’: bothinterest groups have free access to the bidding process, both face zero cost ofcollective action and the central authority simply acts as a neutral intermediator(‘honest broker’).

LWe already know that for t . t both groups of workers are better off whenthere is dispersion. In this case, there is multiplicity of equilibria in which one

Lgroup makes an arbitrarily small but strictly positive offer (when t 5 t only theunskilled will make an offer). Regardless of the equilibrium outcome that emerges,the chosen policy, that is free mobility or mobility barriers, turns out to be socially

Loptimal. When t , t , there is always a conflict between the two groups becausethe unskilled do not want the skilled to agglomerate. The winning group will bethe group which enjoys the higher surplus relative to the alternative policy,respectively given by W ( f ) 2 W (b) for the skilled and W (b) 2 W ( f ) for theL L A A

unskilled, where W (s) is the welfare evaluated at the equilibrium prices (7)–(10)K

of the group K [ hL, Aj when the solution s [ h f, bj is chosen. Clearly, in theequilibrium of the lobbying game, only the winning group bids and its offer isequal to the surplus of the other group. Hence, we have to determine under whichconditions we have

W ( f ) 2 W (b) . ( , )W (b) 2 W ( f )L L A A

These inequalities are equivalent to

W ( f ) 1 W ( f ) . ( , )W (b) 1 W (b) (27)L A L A

which corresponds to the second best condition. As a consequence, it follows fromSProposition 3 that the group of skilled will win the auction when t , t . On the


S Lother hand, when t , t , t , the auction will be won by the group of unskilled.In other words, in our model competitive lobbying leads to the second bestoptimum, a result that has a familiar flavor in political economics (see, e.g. Becker,

131985; Wittman, 1989). This implies that a deepening of the integration processwill favor the freeness of the skilled once trade costs have reached a value which

Sis below t . Stated differently, moving sufficiently far in the direction of economicintegration should be accompanied by a political process that spurs the mobility ofthe skilled workers and the formation of a core region.

SFinally, in the conflicting case in which t , t , since the winning skilled groupoffers to pay the relative surplus of the losing unskilled group, the central authorityis able to raise the amount:

2t A]T 5 C4

Clearly, this amount is increasing in the trade cost t. This property suggests thatincreased economic integration should reduce the amount of funds that can beraised by the central authority from interest groups on a voluntary base.

Thus far, we have assumed that special interests organize on the basis of skillbecause of uncertainty about the final spatial equilibrium. By contrast, in the morerealistic case where it is clear which region will become the core region, oneexpects the skilled to cooperate with the unskilled living in that region. Such achange in the composition of interest groups does not affect our results. Indeed, if

H HH is the elected region, we have to compare W ( f ) 2 W (b) 1 W ( f ) 2 W (b), theL L A A

differential surplus earned by the skilled and the subgroup formed by the unskilledF Fliving in region H, to W (b) 2 W ( f ) which represents the differential surplus ofA A

the unskilled in region F. Even if the values of the bids are not the same as before,the policy choice is still the same since the inequalities

H H F FW ( f ) 2 W (b) 1 W ( f ) 2 W (b) . ( , )W (b) 2 W ( f )L L A A A A

are again equivalent to (27). Of course, the whole burden of agglomeration is nowon the shoulders of the traditional workers living in the peripheral region.

H H F FConsequently, since W (b) 2 W ( f ) , 0, we have W (b) 2 W ( f ) . W (b) 2A A A A A

W ( f ), so that the new winning coalition has to pay more than the previous oneA

because of the absence of internal compensation among the unskilled in the tworegions. Therefore, cooperation between the skilled and some unskilled has strongredistributive implications within the population. More generally, this implies that,

13A caveat is in order. Clearly, any violation of the ‘competitive’ assumption would lead tosuboptimal policy choices. This would happen, for example, if (i) not all skilled or unskilled workerswere members of their corresponding interest group, since different weights would be introduced in theabove inequalities, (ii) if different interest groups faced different costs of collective action, (iii) if thecentral authority were not a neutral intermediator.


while the actual coalitions that are formed are crucial in terms of redistribution,they might well be immaterial in terms of efficiency.

7. Policy implications and concluding remarks

Our model of agglomeration has allowed us to investigate the much debatedquestion of whether or not the deepening of the integration process within the EUis going to lead to more concentration of economic activities, and when such anagglomeration is desirable from the collective standpoint. Our results suggest thatthere is no clear-cut answer. Indeed, too much agglomeration arises for inter-mediate values of the trade costs, while the market outcome is optimal for low orhigh values of these costs. However, there seems to be a need for an activeregional policy within the EU if, as discussed in the introduction, the actual valuesof trade costs are not likely to become very low. Such a need might become evenmore stringent after the enlargement to Central and Eastern European countriesthat could well trigger some brain drain toward the richest regions of the EU. Thisraises the question of the implementation of the desirable policy.

We concentrated here on imposing restrictions on factor mobility as a policyinstrument to prevent trade integration from causing undesired agglomeration.Both the instrument and its aim seem to be consistent with EU attitudes towardsthe accession of new ‘peripheral’ members. For example, during the negotiationson Eastern enlargement, the extension of the free mobility of people has beenrecognized as a serious issue both for current and prospective members that maylead to a mutual agreement on a slow and gradual liberalization. In particular, ithas been argued that ‘‘it could be in the advantage of the new members to restrictthis right for a set period of time. The economies of the applicant countries aremore likely to suffer a brain drain and loss of a valuable part of [their] youngmobile workforce if early, unrestricted liberalisation of labour markets is granted.’’(European Commission DGX, 1999). For similar reasons, in the case of theaccession of Spain, Portugal and Greece, ‘‘migration flows were forestalledthrough inclusion in the accession treaties of restrictions on the free movement ofworkers for a set period of time’’ (ibidem). Finally, the Schengen agreements of1985 and 1990, that represent a first step towards the full implementation of thefree mobility of people inside the EU, involve member states with fairly similarlevels of institutional and economic development.

Assuming that the central authority has little or poor information about thewelfare implications of alternative decisions, we have shown that, if interestgroups are allowed to reveal their preferences through competitive lobbying, thepolitical outcome may well foster a solution which is efficient. For such amechanism to work, all interests should have free access to the bidding process, allshould face the same cost of collective action and the central authority shouldsimply act as a neutral intermediator. In perspective, our analysis also yields


potentially interesting insights on the issue of interregional transfers. First, themodel suggests the existence of an upper bound on the amount of redistributionthat could be performed without distorting efficiency. This level is equal to half thedifference between the relative surpluses attained by the winning and losinginterest groups. Second, while, for simplicity, we have restricted our attention tofactor mobility controls so that the potential scope of transfers is merelyredistributive, inspired by the European experience one might wonder whethersimilar results could also be attained by using transfers for shaping the efficient

14economic landscape. This should be possible in principle whenever equity andefficiency go hand in hand (no equity–efficiency trade-off). However, under bothcounts, the lack of information by the central authority would require some sort ofmore sophisticated mechanism design than our simple menu auction.

Acknowledgements

We thank R. Gordon and two referees for very useful comments and sug-gestions. We have also benefited from helpful discussions with P. Battigalli, G.Bertola, M. Lebreton, Ph. Martin, M. Motta,V. Norman and seminar participants atCERAS Paris, EUI Florence, NHH Bergen, and Sorbonne Paris, as well as atCEPR ERWIT99 in Bergen, at the 1999 Annual Conference of the Society forEconomic Dynamics in Alghero, and at PETS 2000 in Warwick. We are indebtedto FNRS (Belgium) for funding. The first author gratefully acknowledges financialsupport from a Jean Monnet Fellowship at EUI Florence.

References

Abdel-Rahman, H., Fujita, M., 1990. Product variety, Marshallian externalities, and city sizes. Journalof Regional Science 30, 165–183.

Alonso, W., 1980. Five bell shapes in development. Papers of the Regional Science Association 45,5–16.

Amiti, M., 1998. New trade theory and industrial location in the EU: a survey of evidence. OxfordReview of Economic Policy 14, 45–53.

Anas, A., Arnott, R., Small, K.A., 1998. Urban spatial structure. Journal of Economic Literature 36,1426–1464.

Armstrong, H.W., Vickerman, R.W., 1995. Convergence and Divergence Among European Regions.Pion, London.

Austen-Smith, D., 1993. Information and influence: lobbying for agendas and votes. American Journalof Political Science 37, 799–833.

Becker, G.S., 1985. Public policies, pressure groups, and deadweight costs. Journal of PublicEconomics 28, 329–347.

14For a discussion of the effectiveness of EU regional policies, see European Investment Bank(2000).


´Benabou, R., 1993. Working of a city: location, education, and production. Quarterly Journal ofEconomics 108, 619–652.

´Benabou, R., 1996. Efficiency and equity in human capital investment: the local connection. Review ofEconomic Studies 63, 237–264.

Bernheim, B.D., Whinston, M.D., 1986. Menu auctions, resource allocation, and economic influence.Quarterly Journal of Economics 101, 1–31.

Black, D., 1999. Local knowledge spillovers and inequality. University of California at Irvine,Department of Economics, mimeo.

Borjas, G.J., 1995. The economic benefits from immigration. Journal of Economic Perspectives 9,3–22.

de la Fuente, A., Vives, X., 1995. Infrastructure and education as instruments of regional policy:evidence from Spain. Economic Policy 20, 13–51.

European Commission, 1996. First Report On Economic and Social Cohesion. Office for OfficialPublications of the European Communities, Luxembourg.

European Commission DGX, 1999. Free movement of people. European Dialogue Jan–Feb.European Investment Bank, 2000. Regional development in Europe. An assessment of policy strategies.

EIB Papers 5 (1).The Economist, 1998. The Brussels lobbyist and the struggle for ear-time, August 15th, 25.Fujita, M., Krugman, P., Venables, A.J., 1999. The Spatial Economy. Cities, Regions and International

Trade. MIT Press, Cambridge (Mass).Fujita, M., Tabuchi, T., 1997. Regional growth in postwar Japan. Regional Science and Urban

Economics 27, 643–670.Fujita, M., Thisse, J.-F., 1996. Economics of agglomeration. Journal of the Japanese and International

Economies 10, 339–378.Gorges, M.J., 1996. Eurocorporatism? Interest Intermediation in the EC. University Press of America,

London.Greenhut, M.L., Norman, G., Hung, C.-S., 1987. The Economics of Imperfect Competition. A Spatial

Approach. Cambridge University Press, Cambridge.Head, K., Mayer, T., 2000. Non-Europe. The magnitude and causes of market fragmentation in the EU.

Weltwirtschaftliches Archiv 136, 284–314.Henderson, J.V., 1974. The sizes and types of cities. American Economic Review 64, 640–656.Kanemoto, Y., 1980. Theory of Urban Externalities. North-Holland, Amsterdam.Krugman, P., 1991. Increasing returns and economic geography. Journal of Political Economy 99,

483–499.Magrini, S., 1999. The evolution of income disparities among the regions of the European Union.

Regional Science and Urban Economics 29, 257–281.Martin, Ph., 1998. Regional policies, growth and geography in Europe. World Economy 21, 757–774.Martin, Ph., 1999. Public policies, regional inequalities and growth. Journal of Public Economics 73,

85–105.Mazey, S., Mitchell, J., 1993. Europe of the regions? Territorial interests and European integration: The

Scottish experience. In: Mazey, S., Richardson, J. (Eds.), Lobbying in the European Community.Oxford University Press, Oxford, pp. 95–121.

Mazey, S., Richardson, J., 1993. Introduction: transference of power, decision rules, and rules of thegame. In: Mazey, S., Richardson, J. (Eds.), Lobbying in the European Community. OxfordUniversity Press, Oxford, pp. 3–26.

McCallum, J., 1995. National borders matter: Canada–US regional trade patterns. American EconomicReview 85, 615–623.

Michel, Ph., Pestieau, P., Thisse, J.-F., 1983. Regional allocation of investment with distributiveobjectives. Journal of Regional Science 23, 199–209.

Neven, D., Goyette, C., 1995. Regional convergence in the European Community. Journal of CommonMarket Studies 33, 47–65.

Olson, M., 1980. The Logic of Collective Action. Harvard University Press, Cambridge (MA).


Ottaviano, G.I.P., 1999. Integration, geography and the burden of history. Regional Science and UrbanEconomics 29, 245–256.

Ottaviano, G.I.P., Puga, D., 1998. Agglomeration in the global economy: a survey of the ‘neweconomic geography’. World Economy 21, 707–731.

Ottaviano, G.I.P., Tabuchi, T., Thisse, J.-F., 2001. Agglomeration and trade revisited. InternationalEconomic Review, forthcoming.

Ottaviano, G.I.P., Thisse, J.-F., 2001. On economic geography in economic theory: increasing returnsand pecuniary externalities. Journal of Economic Geography, forthcoming.

Potters, J., van Winden, F., 1992. Lobbying and asymmetric information. Public Choice 74, 269–292.Rahman, M.A., 1963. Regional allocation of investment. Quarterly Journal of Economics 77, 26–39.Sala-i-Martin, X., 1996. Regional cohesion: evidences and theories of regional growth and conver-

gence. European Economic Review 40, 1325–1352.Shields, G.M., Shields, M.P., 1989. The emergence of migration theory and a suggested new direction.

Journal of Economic Surveys 3, 277–304.Sowell, T., 1996. Migrations and Cultures: A World View. Basic Books, New York.Starrett, D., 1978. Market allocations of location choice in a model with free mobility. Journal of

Economic Theory 17, 21–37.Takayama, A., 1967. Regional allocation of investment: a further analysis. Quarterly Journal of

Economics 81, 330–337.Tirole, J., 1988. The Theory of Industrial Organization. MIT Press, Cambridge (MA).Wittman, D., 1989. Why democracies produce efficient results. Journal of Political Economy 97,

1395–1424.

Documents

Integration, agglomeration and the political economics of factor mobility