How to Promote Clusters: Policy Experiences from

Latin America

TILMAN ALTENBURGGerman Development Institute, Berlin, Germany

and

J ORG MEYER-STAMER *University of Duisburg, Duisburg, Germany

Summary. In this paper we propose to dierentiate between three types of clusters whenit comes to formulating cluster-oriented policies in Latin America. Survival clusters ofmicro- and small-scale enterprises owe their existence more to unfavorable macroeco-nomic conditions and less to entrepreneurial competence and dynamism. Theircompetitive potential is limited. Support measures should mainly aim at improving theconditions for survival since these clusters are important in creating employmentopportunities. The impetus should be to break through the low skills/low investmentvicious circle. More advanced and dierentiated mass producers have been ourishing inthe import-substitution era but are coming under enormous pressure with the transitionto open economies. In these clusters the main challenge is to create an environment thatstimulates and supports learning, innovation, and constant upgrading. Clusters oftransnational corporations are typically dominated by foreign rms not only at the nalassembly stage but also in parts production. These clusters often are showcases of best-practice manufacturing; this can be used to stimulate the upgrading of domestic rms,notably by involving them in the supply-chain of transnationals. 1999 ElsevierScience Ltd. All rights reserved.

1. INTRODUCTION

Clustering seems to enable rms, especiallysmall and medium-sized enterprises (SMEs), togrow and upgrade more easily. SMEs may evenbecome players in world markets if a highdegree of interrm specialization and theirproximity to other rms performing comple-mentary functions oset the disadvantages ofbeing small. Clusters often create positiveexternalities which help managerial and tech-nical learning. Empirical evidence shows thatclustering is especially common among tradi-tional small-scale and labor-intensive activities.Upgrading these activities contributes to amore balanced rm size structure and a morelabor-intensive growth pattern. These featuresof clustering have attracted the interest of

policymakers in developing countries anddevelopment assistance agencies. Recently,numerous bilateral and multilateral agencies(including the World Bank, UNIDO,UNCTAD, and ILO)1 have begun to recognizethe benets of clustering and to reframe theirSME and private sector developmentprograms.

This article focuses on Latin America anddevelops two arguments to advance theongoing debate on clusters and cluster-relatedpolicies. First, industrial clusters in LatinAmerica are very heterogeneous, and they are

World Development Vol. 27, No. 9, pp. 16931713, 1999 1999 Elsevier Science Ltd. All rights reserved

Printed in Great Britain0305-750X/99/$ - see front matter

PII: S0305-750X(99)00081-9www.elsevier.com/locate/worlddev

* The authors would like to thank John Humphrey,Hubert Schmitz, and Regine Qualmann for their

comments on earlier drafts of this paper.

1693

quite dierent from those stylized role modelswhich have inuenced the academic and policydiscussion in Europe and North America.2 Weidentify, for Latin America, three ideal typeclusters, each having a dierent genesis, rmstructure, organization principles, developmenttrajectory, opportunities, and bottlenecks.Second, as development opportunities andbottlenecks dier, each type of cluster requiresa tailor-made policy approach. This mayappear obvious. In practice, nevertheless, clus-ter support in Latin America often proceeds onunrealistic assumptions about developmentpotentials of (groups of) rms and clusters,ignores the specic logic guiding entrepreneur-ial decision-making, and understates the inertiaof the local sociocultural milieu, thus failing todesign adequate policies.

The paper is structured as follows: InSection 2 we will specify the term ``industrialcluster.'' Section 3 addresses the justicationfor cluster-related policy interventions. InSection 4 we present a typology of LatinAmerican clusters. In Sections 57 we discussthe main characteristics and possible policyinterventions in three dierent types of clusters.In Section 8 we draw some conclusions forpolicymakers as well as for future research. Agood part of the paper is based on our own eldresearch and advisory work but we also drawextensively on the work of other researchers.

2. CONCEPTUAL ISSUES

There is no generally accepted denition ofindustrial clusters. The term is used quiteindiscriminately for a broad range of businessarrangements. In its broadest sense, the term``cluster'' only depicts local concentrations ofcertain economic activities. What makes clus-ters so attractive for policymakers however arethe opportunities for collective eciency(Schmitz, 1995a) emanating from positiveexternal economies, low transaction costs, andjoint action. Pure agglomerations of unrelatedrms do not give rise to collective eciency.Most denitions of clusters thus add someingredients to the basic notion of spatiallyconcentrated rms and focus on external eectsand interaction:

positive external eects emanating from theexistence of a local pool of skilled labor andthe attraction of buyers;

forward and backward linkages betweenrms inside the clusters;intensive information exchange betweenrms, institutions, and individuals in the cluster,which gives rise to a creative milieu;joint action geared to creating locationaladvantages;the existence of a diversied institutionalinfrastructure supporting the specic activitiesof the cluster;a sociocultural identity made up of commonvalues and the embeddedness of local actors ina local milieu which facilitates trust.

Given the complexity of patterns of inter-action in clusters and the emphasis of clusterliterature on ``soft,'' nonmeasurable vari-ablessuch as trust, social embeddedness, orcreative milieuit is impossible to formulate aprecise denition of clusters or to draw aclear-cut borderline between pure agglomera-tions and complex clusters with strong exter-nalities. In the real world, even the simplestindustrial agglomerations, such as ExportProcessing Zones (EPZs) in developing coun-tries, where no forward and backward link-ages exist at all, will generate some very basicexternalities, e.g., contribute to the formationof a specialized local labor pool or augmentthe demand for certain complementaryservices. The notion of clustering thereforerefers to a variety of industrial agglomera-tions.

We propose an operational denition ofclustering based on measurable variables: Acluster is a sizable agglomeration of rms in aspatially delimited area3 which has a distinc-tive specialization prole and in which inter-rm specialization and trade is substantial.4

This excludes agglomerations of the EPZ-type, as these do not build upon intensivelinkages. In our understanding, those localbusiness networks in which a dense socialfabric based on shared cultural norms andvalues and an elaborate network of institu-tions facilitate the dissemination of knowledgeand innovation, constitute a specic type ofcluster and may be termed ``industrialdistricts''5. The above-mentioned qualitativefeatures considerably augment the perfor-mance of clusters. Scarcity of entrepreneurialspirit, barriers to information-sharing, lack oftrust, and similar ``soft'' constraints oftenconstitute the main bottlenecks for clusterdevelopment.

1694 WORLD DEVELOPMENT

3. A TYPOLOGY OF LATIN AMERICANCLUSTERS

Clustering implies a continuum. Moreover,clusters as well as their environment areconstantly changing, so that each cluster has aspecic, open-ended trajectory. Clusters shouldhence be interpreted as dynamic systems. Thisis especially relevant in the case of LatinAmerican clusters, where we can observe far-reaching structural changes in response to theeconomic liberalization of the 1980s and 1990s.Any typology therefore implies the risk ofoversimplifying matters. Nevertheless, andbeing well aware that not every existing clusterwill t neatly into one of our categories, weconsider it useful to distinguish the followingthree ideal type clusters. This distinction helpsto identify some specic policy packages as wellas some widespread policy errors resulting froman uncritical, one-size-ts-all approach.

(a) Most frequent are survival clusters of microand small-scale enterprises which produce low-quality consumer goods for local markets,mainly in activities where barriers to entry arelow. Firms in these clusters display many char-acteristics of the informal sector, with produc-tivity and wages being much lower than thoseof medium- and large-scale enterprises. The de-gree of interrm specialization and cooperationis low, reecting the lack of specialists in the lo-cal labor force as well as a fragile social fabric.(b) Some clusters are made up of more advancedand dierentiated mass producers which for themost part prospered in the import substitutionperiod and mainly produce for the domesticmarket. They typically comprise a heteroge-neous mix of enterprises ranging from pettyproducers to large Fordist industries. Tradeliberalization forced these clusters to face inter-national competition and induced far-reachingstructural change.(c) Clusters of transnational corporations(TNCs) exist in technologically more complexactivities, such as the electronics and autoindustries. These clusters are dominated bylarge branch plants of world-class manufactur-ers and usually serve both national and interna-tional markets. They typically have few linkageswith domestic SMEs and institutions. Theycomply with the criteria of a cluster becausein contrast to simple assembly plants (e.g., inthe clothing industry)they include severalstages of the value chain and derive competitiveadvantages from local external economies.