Journal of Economic Geography4(2004)pp.565–582doi:10.1093/jnlecg/lbh040 Agglomeration economies and industrial
location:city-level evidence
Elisabet Viladecans-Marsal*
Abstract
There is clear evidence that economic activity,in particular industrial
activity,is unequally located in Spain.Further,the results from the
analysis of single manufacturing ctors show an even higher spatial
读书分享会主持词
concentration.The aim of this paper is to demonstrate the extent to
which agglomeration economies account for this high industrial
concentration.To this end,I analy the influence of various types of
agglomeration on the location of manufacturing employment in Spanish
cities.I consider two types of agglomeration economies:urbanization
economies(associated with a city’s population and employment levels
and the diversity of its productive structure)and localization economies
(associated with a city’s specialization in one specific ctor).Special
attention is given to the geographical unit of analysis by employing spatial
econometric techniques that allow the influence of agglomeration effects
extending beyond a city’s limits to be considered.The results demonstrate
that agglomeration economies influence the location of manufacturing
activity,with most ctors being influenced by urbanization economies
and a few by localization economies.In some ctors,population or
employment levels in neighboring cities were found to enhance a city’s
agglomeration economies.
Keywords:Agglomeration economies,cities,industrial location
JEL classifications:R30,C5,L60
Date submitted:26January2003Date accepted:26March2004
1.Introduction
There is a marked spatial concentration of production and employment patterns in Spain, which is even more evident in analys conducted for individual manufacturing ctors. Within thefield of economics,a range of approaches have been developed to identify the factors influencing this geographic concentration,and the objective of this paper is to analy the impact of one of the—agglomeration economies—on the geographical distribution of various manufacturing ctors.From the outt,it should be stresd that this concept,agglomeration economies,encompass many types of element, ranging from the productivity enhancing effects of geographic specialization in specific activities(localization economies)to the growth promotion effects of agglomeration economies that ari from the size of an area and its diver economic environment(urbanization economies).
*Departament d’Econometria,Estadı´stica i Economia Espanyola,Facultat de Ci e ncies Economiques i Empresarials,Universitat de Barcelona,Avda.Diagonal,690,08034Barcelona,Spain.
email<eviladecans@ub.edu>
Journal of Economic Geography,Vol.4,No.5,#Oxford University Press2004;all rights rerved.
566 Viladecans-Marsal
Recent analys of cities and their evolution(Glaer,1998;Quigley,1998)assume that a city prents veral advantages of location for thefirm(including lower transport costs to thefinal market,availability of intermediate products,and the advantages of labor division)that increa with a growing population.Further,the analys point to the disadvantages experienced once cities surpass a specific size and urbanization economies turn to diconomies(characterized by pollution,congestion,and high land prices).It should be remembered that cities have invariably been associated with the diversification of economic activities and that this element has always been considered an indicator of growth capacity(Audretsch,1998),although few authors have attempted to parate in empirical terms the effects of productive diversification from tho due to the size of the city(Combes,2000,and Beeson et al.,2001,are two recent exceptions).This paper pays particular atte
ntion to this point by eking to dintangle localization from urbanization economies and,in the ca of the latter,by eking to dintangle the effects of diversification from the other factors related to city size.
The current debate focus not only on the dichotomy between specialized and diversified environments as elements which influence afirm’s location decisions,but also on the coexistence of diversified and specialized areas within the same urban systems (Feldman and Audretsch,1999;Duranton and Puga,2000,2001).The papers have as their objective not only to analy the effects of just one type of agglomeration economy, but to establish the circumstances in which industrial activities come to value specialized or diversified environments,or both characteristics at once.The results obtained from this study should contribute to asssing the validity of such a hypothesis.
Empirical rearch examining the nature and extent of agglomeration economies is rich and includes Carlino(1982),Nakamura(1985),Henderson(1986),Glaer et al.(1992), Henderson et al.(1995),Ciccone and Hall(1996),Moomaw(1998),Costa and Viladecans (1999),Henderson et al.(2001),and Ronthal and Strange(2003).This paper eks to add furtherfindings to the on-going discussion by analysing the incidence of different types of agglomeration economy on the location of
industrial activity in Spanish cities. More specifically,I analy the effects of urbanization and localization economies on city employment in certain industrial ctors.Urbanization economies are measured using a diversification index of the city’s productive structure,city population and industrial employment statistics.The equation relating the variables and city-level employment by ctor is a labor demand function obtained from a CES production function(e Calem and Carlino,1991,and Moomaw,1998).By adopting this approach,the parametric relationship between the output and wage variables(included in the employment equation)can be ud to test for localization economies.Given that my databa only provides complete information for a cross-ction of cities,this is the only feasible way to test for localization economies.
Special attention is paid here to ensuring the study is conducted within the appropriate geographic area of analysis.Most authors agree that this area should be local,since various applied studies have argued that the state,in the ca of America,and the region,in the ca of Europe,are too large to be considered suitable units of analysis (Audretsch and Feldman,1996;Ciccone and Hall,1996).In the ca of Spain,a geographical classification of metropolitan areas has yet to be undertaken and, therefore,it would em more viable to u the city and its surrounding area as a unit of analysis,and then subquently to consider any neighboring effects.This is,to my knowledge,thefirst paper using d
ata at this level of geographic detail in Spain.The main reason for this is the abnce of Spanish databas holding local information.The
梁家驹Agglomeration economies and industrial location 567 databa ud in this paper provides city-level data on industrial activities that are drawn from thefiscal databas of VAT,Wage Taxes,and Customs Revenues provided by the Institute for Fiscal Studies(Spain)for1994.This databa has not been ud before, neither for studies at this level of geographical detail nor in eking to fulfil the aims as t out in this paper.The introduction of the neighbor effect in this analysis of agglomeration economies is also a novelty,with the exception of Ronthal and Strange(2003).
The article is organized as follows.In the cond ction,I compute veral indices of geographical concentration in order to illustrate the geographic distribution of certain manufacturing ctors in Spain.In the third ction,I derive the labor demand equation ud to analy the effects of agglomeration economies on the location of industrial employment.In the fourth ction,I discuss the econometric procedure ud in the estimation and prent the results obtained.Finally,the last ction prents the main conclusions of the paper.
2.The geographical concentration of manufacturing
电脑分盘怎么分
activities in Spain
The objective of this ction is to illustrate the geographical distribution of certain manufacturing ctors in Spain.To do so,Ifirst calculate veral indices of concentration.The information ud in this empirical analysis is drawn from the officialfiscal databas of VAT,Wage Taxes and Customs Revenues for Spanish firms in1994and is provided by the Institute for Fiscal Studies(Ministry of Finance, Spain).It contains data that refer to manufacturingfirms in tho Spanish cities with a population over15,000inhabitants1In Spain,a city is understood as a ,a local political and administrative unit.The variables available are employment,output, wage per worker and number offirms and are recorded for each manufacturing ctor in each of the cities.They reprent70%of the population and80%of the industrial employment of the regions included in the databa.This percentage differs,however, from that recorded for the six industrial ctors analyd here.For example,in the ca of Chemical Products the cities in the databa reprent87%of the total employment and, similarly,for Food,thefigure is around63%.Finally,it is worth reiterating that the u of information at this level of geographical detail is new in this kind of analysis as applied to Spain and reprents a marked advance on analys conducted at a regional level, especially given the purpos of the current analysis.酒店运营
The analysis carried out in this paper us information for six manufacturing activities: Office&Computing Machinery,Chemical Products,Motor Vehicles,Food,Textiles,and Leather&Footwear.The activities account for nearly50%of the whole industrial work force in the cities under analysis,but the weighting varies form one ctor to another. Specifically,Office&Computing Machinery reprents0.56%of the overall industrial employment,Chemical Products11.28%,Motor Vehicles7.01%,Food15.7%,Textiles 11.04%,and Leather&Footwear3.55%.The activities prent differing productive ,level of productivity,degree of innovation,labor skills,and prence of foreign investment)while their patterns of geographical distribution in the territory are also different.Thefirst of the manufacturing ctors can be considered as a high-tech 1In Spain,the regions of the Basque Country,Navarre,the Canary Islands,and Ceuta and Melilla have differentfiscal systems and,for this reason,fiscal information is not available for them.
activity,the cond and the third as reprenting a medium level of technology,while the remaining three are traditional manufacturing activities.It is interesting to determine whether the influence of agglomeration economies on the respective location patterns and concentration levels of the activities differs in line with their technological levels. Previous studies suggest that different types of
agglomeration economies have dif-ferent effects.For example,most of the evidence indicates that localization eco-nomies are more important for traditional activities whereas urbanization economies affect high-tech activities.
Three concentration indices werefirst calculated.Thefirst was the Gini index.This is an inequality index ud normally in analysing the distribution of personal income.This index goes from0(when,in our ca,the industrial activity under analysis is equally distributed among all the cities)to1(when one city concentrates all the employment of the industrial activity under analysis).The cond was an Index of Relative Concentration (ICR),computed with the following expression:
ICR k¼1
2
X
j
L kj
L k
À
L j
L
where k is the ctor,j the city,and L kj is the number of employees in ctor k and city j,L k is the national employment in the k ctor,L j is total industrial employment in the city j, and L is national industrial employment.If employment in one ctor is distributed across the areas in such a way that it matches the distribution of employment in all ctors,the index is given the value zero,since L kj/L k¼L j/L for all areas.If employment in one specific ctor is concentrated in just one area,the index approaches zero as the participation of this area in the total employment approaches zero,since L kj/L k¼1 and L j/L approaches0for this area,and L kj/L k¼0and L j/L approaches1for the other areas.For this reason,we can say that a ctor is relatively concentrated when the ICR approaches one2and relatively disperd when the ICR approaches zero. The third index ud was Moran’s I statistic of spatial association(Moran,1948). Moran’s I statistic,in common with other similar indicators,allows us to determine whether an economic variable clusters together(e Anlin,1988,for alternative indices of spatial association).This index was ud in an attempt to solve
the limitations resulting from the u of administratively-defined city borders when computing other measures of concentration.For example,the concentration levels obtained when using traditional measures depend on the level of geographical ,the Gini index tends to be higher when computed at the city-level than when computed at the regional level,becau in the latter ca intra-regional inequalities are ignored.A similar situation occurs when using the ICR).Moreover, when computed using city-level data,the measures provide the same values regardless of whether the cities in each region are adjacent or distant in space. Clearly,when one wishes to identify agglomeration effects(as is the ca here),I am concerned specifically with being able to determine thefirst of the two situations. Moran’s I test,however,can inform us at to which of the two situations(clustered vs sparly localized employment)provides the bestfit with my data.Moran’s I is computed 2This occurs at the limit as the number of cities increas.
双软认证568 Viladecans-Marsal
using the following expression:
IM k¼P
i
P
j
w ijðL kiÀ L kÞðL kjÀ L kÞ
P
i
ðL kiÀL kÞ
where L is employment,k is the ctor,i and j are the cities,and L is the average.A positive, statistically significant index indicates that the geographical distribution of this ctor is not random.In other words,the cities with a high level of employment in this particular ctor are located near other cities with a similarly high level of employment in the same ctor.In fact,such an index shows that employment in the ctor clusters in an area larger than that of the city itlf.收纳框
怎么种韭菜To compute this index,a JÂJ neighborhood matrix must be ud,with J being the number of cities in the sample.The elements in this matrix(w ij)define the geographical relationship between any two pai
rs of cities.Although other alternatives are available, here,a binary matrix is ud,with w ij¼1when two cities are considered neighbors and w ij¼0when there is no relationship between them.Two cities are considered neighbors when the linear distance between both city centers is under20km.This distance was chon after running trials with10,20,30,40,and50km binary matrices.
Thefirst two columns in Table1correspond to the Gini index,computed for both city-level and regional(NUTS-3)data.The third and fourth columns correspond to the ICR index,similarly computed at the two levels of geographical aggregation.Columnsfive and six contain the values for Moran’s I statistic(20km)and its z value,respectively.Both the Gini and the ICR indices show that the most highly concentrated ctors are Leather& Footwear,Office&Computing Machinery,and Motor Vehicles.Interestingly,it is not only the most advanced ctors that are the most highly concentrated,since here a traditional activity(Leather&Footwear)showed a high level of concentration.The exact ranking of geographical concentration provided by the two indices(Gini index and ICR index)did,however,differ.Moreover,while the Food and Textile ctors were found to be fairly evenly distributed throughout the territory,Office&Computing Machinery and Motor Vehicles were found not to cluster in the same areas.
大国重器观后感Table1.The concentration of manufacturing employment in the Spanish cities.Selected ctors
Gini index IRC index Morgan’s I index
Cities Regions Cities Regions Index Z
Office&computing machinery0.9710.8940.5210.3520.001(0.125) Chemical products0.8760.8360.3000.2590.080(2.923)** Motor vehicles0.9650.8410.4550.302À0.008(À0.188) Food0.6880.5210.2910.233À0.005(À0.086) Textile0.8230.7140.3740.2210.061(2.262)** Leather and footwear0.9510.8250.7840.5770.176(6.278)** Notes:(1)Employment data come from the Institute of Fiscal Studies’(IEF,Spain)fiscal databa as described in the text;the data correspond to the year1994and the cities included are tho with more than15,000inhabitants, (2)IRC¼index of relative concentration(3)Moran-I is a measure of spatial association developed by Moran (1948)computed with a binary contact matrix with a20-km threshold;**indicates statistical significance at the 95%significance level.
Agglomeration economies and industrial location 569
570 Viladecans-Marsal
It should also be noted that the levels of concentration computed with regional data are much lower t
han tho computed with city-level data.This occurred when applying both the Gini and the ICR indices and for all activities.However,this fall in the level of concentration is negligible in certain cas(Chemical Products)but much greater in ,Textiles and Leather&Footwear).This suggests that intra-regional inequality is low in thefirst ca but considerably higher in the cond.
Table1shows Moran’s I index to be statistically significant in three cas:Chemical Products,Textiles,and Leather&Footwear.This suggests that in the ca of cities with employment in the ctors,a significant amount of employment in the same ctor is located within a20-km radius.It should be noted that only two of the spatially clustered ctors are included among tho that traditionally have the highest levels of concentration(Chemical Products and Leather&Footwear);that one of them is spatially clustered despite the rather small intra-regional inequality(Textiles);and that some activities with high concentration levels(Motor vehicles)are not spatially clustered.The results em to indicate that Moran’s I statistic(in common with other measures of spatial association)can provide information not captured by trad-itional measures of concentration measures and that the analysis of agglomeration economies should consider a geographical area that is larger than the city but smaller than the region.This latter aspect is taken into account in the econometric analysis conducted here as employment in a city is allowed to be influenced by both city and neighborhood agglomeration economies.
Thus,it is clear that patterns of geographic concentration in Spain vary greatly from one manufacturing ctor to another.Yet,the patterns of uneven distribution are not solely attributable to the effects of agglomeration economies.Other factors may well contribute to explaining the location patterns of manufacturing activities:the cost of productive factors,the availability of raw materials,the existence of networks of infrastructure,the local tax level,the incentives of industrial and regional policies, and,for certain activities,even the climatic conditions.Therefore,the econometric analysis undertaken in this paper is needed in order to determine the effect of the agglomeration economies on the location of manufacturing activities.
3.The effect of agglomeration economies on industrial location 3.1.The employment equation
In this ction,an equation is developed that relates the level of employment for a given ctor in a city with a number of variables that measure the agglomeration economies of that city.The approach described by Calem and Carlino(1991)and Moomaw(1998)is adopted,whereby a labor demand function is derived from a CES production function. This approach has veral advantages.First,by estimating a labor demand function (instead of the production function)the need to obtain information concerning the capital stock(figures which are unavailable in my databa)is avoided.Second,the specific parameterization of the model allows localization economies(and not only urbanization econ
omies)to be tested for even with a cross-ction of data.
Thefirm’s production function in an industrial ctor is:
q¼gð Þr lÀsþð1ÀrÞkÀs
½ Àð1=sÞð1Þ
