Mapping Local Production Systems in the UK

203

LOS DISTRITOS INDUSTRIALES EN EL MUNDO ANGLOSAJÓN.EL CASO BRITÁNICO

Lisa de Propis *

* Universidad de Birmingham.

Este ensayo presenta una reflexión sobre la presencia y elpapel de los distritos industriales en Gran Bretaña. Cien añosdespués de los escritos de Marshall, los distritos industriales hansido redescubiertos por Becattini (1987), y se han convertido en elobjeto de una dramática revuelta en el estudio sobre las dinámicasde los lugares, de las empresas y de las comunidades. Los ríosde contribuciones que se han llevado a cabo en el debateanglosajón han participado aún más de la literatura americana yfilo-porteriana, que ha introducido el concepto de cluster y lo hapropuesto como el motor vital de la competitividad local. Hasido el hermanamiento cluster y competitiveness lo que hallamado la atención de los policy-makers británicos en los años90, luchando contra una amenazante Globalización, un sectormanufacturero en total exfoliación y un sector servicios-sobre todo aquéllos de mayor valor añadido- únicamenteconcentrado en Londres. A diferencia de las líneas que ha seguidoel debate británico, el presente análisis quiere ser un intento deverificar el fenómeno de los «distritos industriales marshallianos»en Gran Bretaña, aplicando una ya bien testada metodologíaque hace referencia a los trabajos de Sforzi en Italia y Boix enEspaña. Este análisis para Gran Bretaña permite unacomparación internacional más amplia sobre los distritos, y estotiene valor sobre todo en contraposición a la más recientetendencia a explorar el tejido económico inglés, y verificar coninstrumentos más cualitativos si hay todavía realidadesterritoriales que tienen un espesor socio-económico.

RESUMEN ABSTRACTThe paper presents some considerations on the role and extentof industrial districts in Great Britain. One hundred years afterthe seminal work by Marshall on «industrial districts» andlocalised industries, these were reawakened and brought torenewed attention by Becattini (1987) in his work on industrialdistricts in Tuscany. Such contribution paved the way for a newapproach to analyse the dynamics, functioning and trends ofplaces, firms and socio-economic communities. The breath ofthe academic literature that followed has had resounding effectsalso in the Anglo-Saxon debate, which, however, has tended tobe pegged to Porter's concept of clusters (1990) and to considerclusters as factors of local competitiveness. The twinningbetween clusters and competitiveness has crucially caught theattention of British policy-makers, especially since the 1990s.Increasingly aware of the challenges and threats of productionglobalisation, manufacturing decline especially across theEnglish regions and Wales, and the rise of the high value addedservice sectors mostly in London, British policy-makers haveturned to clusters as possible objects and vehicles policy actions.Unlike the current line of investigation in the UK, this paperpresents the findings of an analysis of the industrial districtphenomenon across England, Scotland and Wales, drawing onsimilar studies in Italy by Sforzi and in Spain by Boix. Thispaper contributes to a possible international comparison of thephenomenon; whilst at the same time providing a picture oflocal production system across regions.

1. Introducción

Este ensayo presenta una reflexión sobre la presencia y el papel de los distritos industria-les en Gran Bretaña. Como sabemos, el concepto de distrito industrial fue introducido por Marshallpara describir las «industrias localizadas» de los Midlands, la cuna de la Revolución Industrial.Cien años después de los escritos de Marshall, los distritos industriales fueron redescubiertospor Becattini (1987), y se han convertido en el objeto de una dramática revuelta en el estudiosobre las dinámicas de los lugares, de las empresas y de las comunidades. Los ríos de contribu-ciones que les han seguido han contribuido conceptualmente a expandir y detallar tal modelo,mientras que los estudios casuísticos han establecido que tal forma de sistema local está pre-sente en todo el mundo, incluso aunque no sean siempre llamados como tales.

Artíc

ulo

publ

icad

o en

el n

úm. 1

3 de

la C

olec

ción

Med

iterrá

neo

Econ

ómic

o: "L

os d

istri

tos

indu

stria

les"

Coo

rdin

ador

: Vic

ent S

oler

- IS

BN: 9

78-8

4-95

531-

40-7

- D

epós

ito L

egal

: AL-

728-

2008

Edita

: CAJ

AMAR

Caj

a R

ural

, Soc

ieda

d C

oope

rativ

a de

Cré

dito

- P

rodu

cido

por

: Fun

daci

ón C

ajam

ar

LOS DISTRITOS INDUSTRIALES

204

De hecho, y sobre todo en los países anglosajones, a menudo se encuentran realidadessimilares o diversas a los distritos industriales, pero en todos los casos muy groseramentereagrupadas bajo el término porteriano de clusters.

Frente al debate académico entre naciones que ha desmenuzado las varias formas desistema local y ha analizado cada matiz: -véanse los distritos industriales (Pyke et alii, 1990;Becattini et alii, 2003); los medios innovadores (Camagni, 1995) y los sistemas regionales deinnovación (Cooke, 2001; Braczyk et alii 1998), corriéndose también el riesgo de crear caóticasrepeticiones y confusiones (Martín y Sunley, 2003)-; los análisis que han resultado en GranBretaña han participado más de la literatura americana y filo-porteriana, que ha introducido elconcepto de cluster y lo ha propuesto como el motor vital de la competitividad local. Fue elhermanamiento clusters y competitiveness lo que atrajo la atención de los policy-makers británi-cos en los años 90, luchando contra una amenazante Globalización, un sector manufacturero entotal exfoliación y un sector servicios -sobre todo aquéllos de mayor valor añadido- únicamenteconcentrado en Londres.

El debate sobre los clusters en el Reino Unido se ha desarrollado con estudios casuísticosy poco con contribuciones sustancialmente conceptuales. Los policy-makers ingleses han ab-sorbido la idea de que los sectores están geográficamente concentrados y que esto tiene razo-nes, potencialidades y a veces límites, sólo desde 2001. Siguiendo un documento publicado porel Department of Trade and Industry, donde se hace una cartografía de los clusters en GranBretaña (Inglaterra, Gales y Escocia), las Agencias de Desarrollo Regional (Regional DevelopmentAgencies) empezaron a ver sus economías regionales con distintos ojos; esto es, como coágu-los de sectores en lugares particulares, y a focalizar las acciones de política sobre éstos. No hayduda de que para los observadores expertos el descubrimiento en Gran Bretaña del papel de lossistemas locales y de su uso como objeto de intervención para acciones para el desarrolloregional, se ha llevado a cabo sin una verdadera y profunda comprensión del fenómeno, y con laidea de que fuese «la última tendencia» en los términos de policy-making, similar a un tren enmovimiento al que debemos saltar o dejar atrás.

A diferencia de las líneas que ha seguido el debate británico, el presente análisis quiere serun intento de verificar el fenómeno de los «distritos industriales marshallianos» en Gran Bretaña, yde encuadrar el fenómeno en una discusión crítica de su papel en las políticas regionales.

2. El debate actual

Las razones de este relativismo son dos: por una parte, en los países anglosajones laliteratura sobre el desarrollo local y sobre los sistemas locales ha sido, por así decirlo, «porterizada»,por lo que, a partir de una moda, el uso del término cluster se ha multiplicado, dejando ainterlocutores o lectores a menudo no muy seguros de lo que verdaderamente se está hablando.

205

LOS DISTRITOS INDUSTRIALES EN EL MUNDO ANGLOSAJÓN. EL CASO BRITÁNICO / LISA DE PROPRIS

La «porterización» del debate sobre sistemas locales ha sido también avalada por la desco-nexión entre sociedad y territorio que caracteriza la realidad económica británica. El concepto decluster propuesto por Porter no parte del territorio, de las localizaciones de tradiciones producti-vas, sino de la estructura y organización de la producción; esto es cierto, en línea con la literaturaamericana de la flexible specialisation, que siempre ha considerado los sistemas locales comoel resultado de la desmembración de la gran empresa en el ocaso del sistema fordista. En otraspalabras, los sistemas locales siempre han sido considerados sobre la base de su funcionalidadproductiva. En Gran Bretaña, la erradicación de las actividades productivas del territorio puedeexplicarse por varios factores: en primer lugar, el hecho de que Inglaterra se haya industrializadomucho antes que el resto de Europa hizo que empezase a terciarizarse también antes; ensegundo lugar, el capitalismo inglés es quizá más parecido al americano que al europeo, empre-sas de dimensiones más grandes, un desarrollado sistema de venture capital y una economíamuy abierta a las inversiones extranjeras.

Precisamente este último fenómeno, con muchas de las mayores empresas inglesasahora de propiedad extranjera, ha acelerado el susodicho proceso de desconexión, por lo quedecisiones que impactan sobre una localidad vienen tomadas en head offices a miles de kilóme-tros de distancia (Bailey y Driffield, 2007). Este fenómeno se ha llamado wimbledonización, enreferencia al hecho de que Wimbledon, uno de los eventos tenísticos más importantes de latemporada, lleva desde hace años sin campeones ingleses.

Traducido en términos económicos, esto quiere decir que la presencia de propiedad ex-tranjera en empresas y sectores punteros en Gran Bretaña, ha producido una separación entreterritorio, sociedad y economía que se ha reflejado en el debate sobre el desarrollo local y lossistemas locales.

La segunda razón es que el debate británico sobre los sistemas locales está fragmentadoy desunido, en cuanto que las varias disciplinas que podrían concurrir a desarrollar el conceptode manera multiforme no se hablan, siendo éstas, por ejemplo, la Economía, la Geografía y laSociología. La fragmentación disciplinar y la desconexión territorio/economía lleva a considerarel distrito industrial marshalliano (Becattini, 1987, 1994, 2000 y 2001) como una realidad muycompleja y casi un ideal-tipo al cual aspirar, sin ser alcanzado jamás.

3. Los distritos industriales en la época de Marshall

No creo que sea posible escribir un ensayo sobre los distritos industriales en Gran Breta-ña sin comenzar por Marshall, y en cierto modo por la Revolución Industrial en los Midlands.

El propio Marshall, en Industry and Trade y en los Principles, hace referencia explícita-mente a algunas realidades concretas, como las industrias metalúrgicas en Staffordshire,


206

Shropshire y, más vagamente, en Gales y Escocia; el potteries en Staffordshire; chair-making enBuckinghamshire; el straw plaiting en Bedfordshire; la cuchillería en Sheffield; los cotton mills deLancashire; y el textil de los Midlands.

No era el interés de Marshall hacer un inventario de todas las industrias localizadas. Dehecho, otros interesantes distritos emergen deshojando los trabajos de historia económica. Comose ha dicho anteriormente, la investigación de los historiadores económicos ingleses no hasabido apreciar el modelo distritual y crear un filón de estudios sobre esto; por tanto, cuando heintentado desentrañar el fenómeno a caballo del novecientos, me he encontrado juntando nume-rosos estudios que describen varias industrias y localidades sin un acople conceptual a la mane-ra de las «industrias localizadas» de Marshall. Esta estimulante caza del tesoro ha sacado a laluz una multitud de distritos en el ochocientos y a caballo del novecientos, exactamente en lostiempos de Marshall.

Éstos incluyen, por ejemplo:

• La industria de los guantes en Worchester y Taunton (Coopey, 2003).

• La industria del ribbon-making machinery en Coventry –Popp (2003) sostiene que des-pués se ha convertido en la industria de las bicicletas, primero, y de la mecánica deautomóviles después–; y el textil en Manchester (Lancashire) (Wilson y Singleton, 2003).

• La elaboración de la lana en Est Anglia, del lino en Norkfolk y de la seda en Essex; lacamisería en Leicester; la elaboración del metal (metal bashing) en los Midlands y enYorkshire; la cuchillería en Sheffield y en Yorkshire (Hudson, 2004).

• La elaboración de la lana en Yorkshire; la producción de toys and button en Birmingham(Berg, 1994).

• El iron district en Dudley; las destilerías de ginebra en Londres y de spirits en Escocia;llaves y candados en Wolverhampton; la guarnicionería en Walsall1; la producción deencajes (bone-pillow lace industry) en Buckinghamshire; la producción de botones paracamisas en Shaftesbury y Blandford (Clapham, 1930).

• La producción de zapatos y botas para hombre en Northampton, y de zapatos de mujeren Norwich y Leicester; el mecano-textil en Lancashire; la producción de bicicletas enCoventry y Birmingham (Aldcroft, 1968).

• El jewellery quarter en Birmingham (Wise, 1950; De Propris y Lazzeretti, 2007).

• El gun quarter en Birmingham (Wise, 1950).

1 El distrito de la guarnicionería de Walsall está en rápido crecimiento por el boom de la equitación «a la inglesa», sobre todo en losEstados Unidos.

207


Mapa 1.Algunos distritos industriales en Inglaterra y Gales (1892)


208

4. Sistemas productivos locales y distritos en Gran Bretaña

En Gran Bretaña ha habido pocos estudios que hayan dado lugar a un mapa de losclusters: un informe del Ministerio de Comercio e Industria (Department of Trade and Industry,2001), titulado Business clusters in the UK, presentaba una cartografía general de los clusters enGran Bretaña para el sector manufacturero y terciario; un ensayo de Crouch y Farell (2001),donde se usan los cocientes de localización para diferenciar coágulos ocupacionales; y, final-mente, mi contribución (De Propris, 2005), donde aplico los cuatro criterios de Sforzi (1990) paralocalizar varias formas de sistema local y los distritos industriales en particular.

El informe del DTI (2001) ha provisto de algún modo una primera descripción, si bien muyaproximativa, del fenómeno de los clusters en las distintas regiones británicas. La contribuciónmás importante de este informe ha sido la de dar a conocer el fenómeno de los sistemas localesa los policy-makers, a los que las resonancias del debate académico no les habían llegado. Paracada región el informe señala una serie de sectores que se configuran como aglomerados, usan-do un cociene de localización que de hecho asevera la presencia de concentraciones sectorialessobre la base de densidad de ocupaciones a nivel regional (con un LQ inferior a 1). En particular,los high points de esta polvareda de clusters son aquéllos cuyo LQ es mayor que 1,25, y laocupación del sector corresponde al 2% de las ocupaciones regionales.

A esta debilidad metodológica, se une el hecho de haber usado informaciones cualitativasy ad hoc para clasificar tales clusters según el nivel de desarrollo (embrionarios, en crecimientoo maduros); la profundidad, entendida como extensión y complejidad de las relaciones entreempresas (profundos, superficiales o desconocida); el crecimiento (en crecimiento, estables oen declive), y la relevancia a nivel regional, nacional o internacional.

En De Propris (2005) se hace un esfuerzo por proceder a un análisis-diagnóstico máscompleto de los sistemas locales, proponiendo una metodología que combina un análisis espa-cial a nivel regional o nacional (sobre la base de Sforzi, 1990 y Brusco y Paba, 1997), conestudios casuísticos cualitativos sobre la organización de la producción y sobre el aumento deimportancia de las instituciones.

4.1. Metodología para el análisis espacial

Para diseñar un mapa de sistemas locales hay que considerar cuatro criterios: a) intensi-dad manufacturera; b) dimensión de las empresas; c) especialización industrial; y d) conjunta-mente la especialización industrial y la dimensión de las empresas (Sforzi, 1990). La aplicaciónde estos criterios requiere una apropiada definición ya sea de la clasificación sectorial, ya de launidad geográfica de referencia. Sforzi (1990) considera sectores de dos cifras y sistemas loca-les del trabajo. Los sistemas locales de producción que surgen de la aplicación de estos criteriospueden ser múltiples, incluidos los distritos industriales.

209


Mapa 2. Mapa de los clusters en el Reino Unido elaborado por el DTI (2001)


210

En Gran Bretaña, la Oficina Nacional de Estadística (Office of National Statistics) facilitadatos sobre la ocupación clasificados por sectores (dos, tres o cuatro cifras) y por sistemaslocales de trabajo (297 travel-to-work-area2). Esto ha hecho posible aplicar los cuatro criterios ydiseñar un mapa de los sistemas locales en general, y de los distritos industriales en particularen Gran Bretaña.

A. Intensidad manufacturera

La intensidad manufacturera mide las economías de aglomeración, en tanto que asumeque la proximidad de múltiples sectores manufactureros genera externalidades positivas para lasempresas en términos de intercambio intersectorial, transferencia de conocimiento y tecnología.Desgraciadamente, el sector manufacturero en Gran Bretaña se está retirando de manera visible,aunque no siempre para dar paso a un emergente terciario; no obstante, los sistemas locales detrabajo de alta intensidad manufacturera son ahora más numerosos que los no manufactureros:165 sobre un total de 297.

( )( )sllEmanufsllE .,

≷( )

( )...,

natEmanufnatE

(1)

Notas del traductor: Las abreviaturas utilizadas son las originales en italiano.

B. Dimensión de las empresas

El segundo criterio sirve para aseverar la composición de la población de las empresas enun cierto sistema local; en otras palabras, está caracterizado por pequeñas, medianas o grandesempresas. Las clases de dimensión pueden variar según la disponibilidad de los datos estadís-ticos. En Gran Bretaña, la ONS clasifica las empresas con menos de 99 trabajadores (quepodremos llamar pequeñas empresas) en 152; empresas con 100-299 trabajadores (medianasempresas), 186; y empresas con más de 300 trabajadores (grandes empresas), 113.

( )( ).,

.,,manufsllEoccupatimanufsllE

≷( )

( )..,.,.,manufnatEoccupatimanufnatE

(2)

2 En De Propris (2005) se usa la definición de sistema local del trabajo de 1998, sobre la base del censo de población de 1991.

211


C. Especialización industrial

El tercer criterio sirve para identificar la especialización sectorial del sistema local. Estomide las economías de localización del sistema, esto es, las externalidades que surgen de laacumulación de conocimientos y competencias específicas por un cierto sector en un ciertolugar. Las economías de localización hacen a un lugar distinto de otro en cuanto contenedor deconocimientos no transferibles sino radicados. El análisis del índice de especializaciones indus-triales puede suministrar importantes informaciones no sólo sobre el sector dominante de unsistema local, sino también indicar sectores secundarios que, sin embargo, forman parte de lamisma filière de producción.

En el caso británico hemos aplicado el índice de especialización industrial al sector dedos cifras por un total de 23 sectores.

( )( )

( )( )

1

..,.,

,,

⟩

manufnatEsettorenatE

settoresllEsettoresllE

(3)

D. Especialización industrial y dimensión de las empresas

En definitiva, si combinamos el índice de especialización industrial con la dimensión delas empresas, estamos en condiciones de verificar si un cierto sistema local del trabajo especia-lizado en un sector particular tiene un sistema local de pequeñas, medianas o grandes empre-sas. Este índice es muy importante porque, a partir de los datos sobre la ocupación, nos permiteindividualizar aglomeraciones de empresas y no de trabajadores.

Además, estudios sobre la governance (De Propris, 2001; Markusen 1997) han sugeridoque la dimensión de las empresas influye sobre la capacidad de decisión y de negociación de lasmismas, lo cual a su vez define la governance del sistema local.

( )( )settoresllE

occupatisettoresllE,

,,≷ ( )

( )settorenatEoccupatisettorenatE

.,,.,

(4)


212

3 Los 23 sectores manufactureros son: alimentos y bebidas; metales básicos; productos del tabaco; productos fabricados de metal;textiles; maquinaria y equipo; vestir ropa/pieles; equipos de oficina y ordenadores; cuero; maquinaria y aparatos eléctricos;madera y corcho; radio, TV, equipo de comunicaciones; pasta papelera, papel y productos de papel; instrumental médico deprecisión; publicaciones, impresión, soportes gráficos; vehículos de motor, remolques; coque, productos de petróleo refinado;otros equipos de transporte; productos químicos; mobiliario; de caucho y plástico; mercancías; reciclaje; otros productos nometálicos.

4 Brusco y Paba (1997).

4.2. Resultados

En resumen, la aplicación de la metodología para el análisis espacial de los sistemasproductivos locales en Gran Bretaña se ha llevado a cabo con datos sobre la ocupación de 1997,estratificados por 23 sectores manufactureros3 y por 297 sistemas locales del trabajo, desdeAndover (SO) a Wick (Scotland). La dimensión de las empresas habían permitido distinguir entrepequeñas (menos de 99 trabajadores), medianas (entre 100 y 299 trabajadores) y grandes (másde 300 trabajadores).

A diferencia de Sforzi (1990), primero se ha distinguido y luego clasificado los sistemaslocales sobre la base de tres coordenadas: intensidad manufacturera (manufacturero o no manu-facturero); especialización sectorial (sector primario o sector secundario); y, por último, la dimen-sión de la empresa (pyme o gran empresa). Dada la avanzada terciarización del sistema econó-mico inglés, tras la progresiva contracción del sector manufacturero que de 1984 al 2004 haperdido más del 30% de la ocupación, al inicio de 2000 éste supone sólo el 15% de la misma. Sedecidió entonces considerar también sistemas en contextos de baja intensidad manufacturera.En segundo lugar, se decidió tener en cuenta también sistemas locales cuya especializaciónsectorial en pequeñas y medianas empresas, o bien grandes empresas, correspondiese a sec-tores secundarios y no primarios, es decir, con un LQ segundo en el ranking. Finalmente, se handistinguido sectores caracterizados por pequeñas y medianas empresas de aquéllos dominadospor grandes empresas.

Haciendo esto se han obtenido ocho categorías de sistemas locales: (1) proto-distrito4; (2)proto-distrito no especializado (porque el sector en el que se encuentra una aglomeración depequeñas y medianas empresas no es aquél con el LQ más alto); (3) sistema local no manufac-turero / especializado de pequeñas y medianas empresas; (4) sistema local no manufacturero /no especializado de pequeñas y medianas empresas; (5) sistema local manufacturero / especia-lizado de grandes empresas; (6) sistema local manufacturero / no especializado de grandesempresas; (7) sistema local no manufacturero / especializado de grandes empresas; y (8) siste-ma local no manufacturero / no especializado de grandes empresas (ver Tablas 1 y 2).

Respecto al informe del DTI (2002), este ejercicio ha permitido detectar de manera exactay rigurosa diversos tipos de sistemas locales (vale la pena considerar que esta clasificacióninicial se basa puramente en datos estadísticos agregados; si se efectuase también un extensoanálisis cualitativo surgiría una más detallada heterogeneidad); y entre éstos, formas distrituales.

213


De hecho, la aplicación de los cuatro criterios ha señalado 47 proto-distritos, contradicien-do la famosa afirmación de Zeitlin (1995) de que no habría distritos industriales en el Reino Unido.Más genéricamente, se han encontrado 165 sistemas locales en áreas de alta intensidad manu-facturera y 132 en áreas de baja intensidad manufacturera; 105 sistemas locales de grandesempresas y 192 de pequeña y mediana empresa.

Los sistemas locales de grandes empresas se concentran mayoritariamente en los Midlands,el Noroeste, el Sur de Gales y las Tierras Altas del Sur de Ecocia. Mientras los sistemas localesde pequeñas y medianas empresas están espolvoreados por todas partes. En particular, se detec-tan más distritos industriales en los Midlands, el Este de Gales y el Norte (ver Mapas 3 y 4).

En particular, a los sistemas locales especializados correspondería el 21% de la ocu-pación manufacturera; pero si se consideran también las especializaciones secundarias (quea menudo forman parte de una filière productiva o bien son el resultado de economías deurbanización), entonces la relevancia de los sistemas locales alcanza cuotas del 50% (véase elcaso de Coventry).

Tabla 1. Tipos de sistemas locales (1997)

Fuente: ONS. Elaboración propia.

Tabla 2. Tipos de sistemas locales (1997)


LS manufactureros SL no manufactureros

SL especializados PROTO–DISTRITOS SL de PMISL no especializados PROTO-DISTRITOS NO-ESPECIALIZADOS SL de PMI

Proto-distrito Proto-distrito no especializado

SL de PMI no manuf. y

especializado

SL de PMI no manuf. y no

especializado

SL de LI manuf. y especializado

SL de LI manuf. y no especializado

SL de LI no manuf. y

especializado

SL de LI no manuf. y no

especializado Otros

1) Intensidad manufacturera √ √ √ √

2) empleados <100

empleados <300 √ √ √ √

empleados >300 √ √ √ √

√ √ √ √ √ √ √ √

Sector primario Sector secundario Sector primario Sector secundario Sector secundario Sector secundario Sector primario Sector secundario

4) Espec.+ PI

√ √ √ √

Sector primario Sector secundario Sector primario Sector secundario

√ √ √ √

Sector primario Sector secundario Sector primario Sector secundario

No SLT 47 31 80 22 73 5 28 5 6

3) Especialización sectorial

Espec.+LI

Espec.+ PMI


214

Mapa 3.Mapa de los sistemas locales de gran empresa en Gran Bretaña (1997)

Fuente: ONS (2003). Elaboración propia.

215


Mapa 4.Mapa de los sistemas locales de pequeña y mediana empresa en Gran Bretaña (1997)

Fuente: ONS (2003). Elaboración propia.


216

5. Los distritos industriales

En el análisis presentado en De Propris (2005), se habían localizado 47 proto-distritosindustriales. Éstos estaban distribuidos por el territorio británico, concentrándose sobre todo enel más manufacturero Centro y Norte, si bien dejando aún fuera centros importantes comoBirmingham, donde se ha registrado un alto impacto de grandes empresas; Manchester, dondese ha destacado un sistema local pero de un sector secundario; o finalmente Liverpool, dondeel sector manufacturero ha alcanzado dimensiones modestas. Los sectores en los que se hanencontrado formas distrituales van desde aquéllos más tradicionales, como alimentos y bebi-das en Escocia (Whiskey Valley), o el textil en Leicester, hasta aquéllos con un contenido másalto de tecnología, como el sector mecánico, la composición de automóviles, el plástico y lagoma (ver Tabla 3).

En 1998 hubo una reclasificación de los confines de los sistemas locales de trabajo, y senecesitó entonces verificar el impacto del fenómeno con las nuevas fronteras. Siempre consi-derando 297 sistemas locales de trabajo y 23 sectores manufactureros, recientemente se haactualizado este trabajo con datos de 2002. Se ha reconsiderado el fenómeno de los distritosindustriales en Gran Bretaña, siempre aplicando la metodología de Sforzi (1990), pero distinguien-do los distritos de pequeñas empresas de los distritos de mediana empresa (ver Tabla 4).

Uno de los primeros resultados de relieve y que, reteniendo la información sobre los siste-mas de pequeña empresa (sin amalgamar juntos sistemas de pequeña y mediana empresa,como en De Propris, 2005), surgen realidades interesantes donde existen proto-distritos indus-triales sólo de pequeña empresa. En algunos casos, hay proto-distritos de pequeña y medianaempresa y en otros casos sólo proto-distritos de mediana empresa. Conjuntamente se hanlocalizado 77 distritos industriales que cubren un poco todas las regiones, desde el sector textilal médico. En particular, el estudio ha individualizado 18 distritos de pequeña empresa, 9 depequeña y mediana empresa y 54 distritos de mediana empresa.

Algunos de estos distritos son históricos distritos marshallianos, como el metalúrgico enDudley/Sandwell y Worchester en los Black Countries; el de Sheffield (cuchillería) y el textil(ropa, lencería, peletería) de Leicester. El famoso distrito cerámico de las potteries de Stoke-on-Trent ha sido detectado en nuestros estudios como un sistema local, pero de grandes empresasy, por lo tanto, no ya como un distrito industrial. En general, por sectores como el metalúrgico,mecánico, eléctrico y de los transportes, que están presentes en los Midlands, es fácil ver larelación indirecta con las industrias localizadas a las que Marshall hacía referencia en sus obras.

Otros sectores son típicamente tradicionales, como el alimentario, de la elaboración de lamadera y el textil. Finalmente, otros distritos están asociados a sectores «modernos», como elde la goma y el plástico, el médico, el de los ordenadores/office machinery, el de las comunica-ciones y el de las ciencias ambientales (reciclaje).

217


SPL REGIÓN SECTOR

Campbeltown Escocia RopaFishguard y St David’s Gales RopaKnighton y Radnor Gales Metales básicosHaltwhistle Noreste QuímicaKelso y Jedburgh Escocia MetalPoole Suroeste MetalWelshpool Gales MetalDudley and Sandwell Medio Oeste MetalGainsborough Medio Este Alimentos y bebidasMatlock Medio Este Alimentos y bebidasDiss Este Alimentos y bebidasFakenham Este Alimentos y bebidasKing’s Lynn Este Alimentos y bebidasWorkington Noreste Alimentos y bebidasBanff Escocia Alimentos y bebidasGirvan Escocia Alimentos y bebidasKeith y Buckie Escocia Alimentos y bebidasPeterhead Escocia Alimentos y bebidasCamelford Suroeste Alimentos y bebidasDevizes Suroeste Alimentos y bebidasLaunceston Suroeste Alimentos y bebidasShaftesbury Suroeste Alimentos y bebidasWadebridge y Bodmin Suroeste Alimentos y bebidasLlangefni y Amlwch Gales Alimentos y bebidasGoole y Selby Yorkshire y Humber Alimentos y bebidasRetford Medio Este Maquinaria y equipoAndover Sureste Maquinaria y equipoBedford Sureste Maquinaria y equipoWorcester Medio Oeste Maquinaria y equipoCalderdale Yorkshire y Humber Maquinaria y equipoRhymney y Abergavenny Gales Vehículos a motorFalmouth Suroeste Otros transportesHarlow Sureste Publicidad y ediciónHorncastle Medio Este Caucho y plásticoMalvern Suroeste Caucho y plásticoLeominster Medio Oeste Caucho y plásticoWellingsborough Sureste Curtidos / cueroLeicester Medio Este TextilesNottingham Medio Este TextilesBlackburn Noroeste TextilesRochdale Noroeste TextilesEast Ayrshire Escocia TextilesGalashiels y Peebles Escocia TextilesLeek Medio Este TextilesHuddersfield Yorkshire y Humber TextilesKeighley y Skipton Yorkshire y Humber TextilesHuntly Escocia Madera

Tabla 3. Los distritos industriales en Gran Bretaña (1997)


218

Tabla 4. Distritos industriales en Gran Bretaña (2002)

Sistemas productivos locales Distritos industriales de Distritos industriales de(clasificación de 1998) pequeña empresa (0-99) mediana empresa (100-299)

ESTE King’s Lynn Alimentos y bebidasESTE Peterborough Maquinaria y equipoESTE Wisbech PapelESTE Huntingdon Caucho y plásticoESTE Mildenhall MaderaMEDIO ESTE Gainsborough MaderaMEDIO ESTE Leicester Prendas de vestir y pielesMEDIO ESTE Retford ReciclajeMEDIO ESTE Chesterfield Metales básicosMEDIO ESTE Matlock Metales básicosMEDIO ESTE Stamford Productos no metálicosMEDIO ESTE Worksop Productos no metálicosMEDIO ESTE Louth PapelMEDIO ESTE Horncastle Prendas de vestir y pielesMEDIO ESTE Boston MaderaMEDIO ESTE Skegness y Mablethorpe Madera MEDIO ESTE Kettering y Corby Curtido de pieles y vestidoNORESTE Bishop Auckland Maquinaria y equipoNORESTE Sunderland y Durham Vehículos a motorNORESTE Haltwhistle Caucho y plásticoNORESTE Berwick-upon-Tweed MaderaNORESTE Hartlepool MaderaNOROESTE Wigan y St Helens Productos no metálicosNOROESTE Blackburn TextilesNOROESTE Nelson y Colne TextilesNOROESTE Rochdale TextilesESCOCIA Keith y Buckie Alimentos y bebidasESCOCIA Newton Stewart Alimentos y bebidasESCOCIA North Ayrshire Maquinaria de oficina y ordenadoresESCOCIA Peterhead Alimentos y bebidasESCOCIA Dingwall MetalESCOCIA Huntly MetalESCOCIA Fraserburgh Alimentos y bebidasESCOCIA Girvan PapelESCOCIA Kelso y Jedburgh Radio, TV, comunicacionesESCOCIA East Ayrshire TextilesESCOCIA Forfar TextilesESCOCIA Hawick TextilesSURESTE Bedford Maquinaria y equipamientoSURESTE Southend Coque y petróleoSURESTE Wellingborough Curtido de pieles y vestidoSUROESTE Evesham ReciclajeSUROESTE Holsworthy Instrumental médico de precisiónSUROESTE Okehampton Alimentos y bebidasSUROESTE Camelford Metales básicosSUROESTE Launceston Alimentos y bebidasSUROESTE Stroud Instrumental médico de precisiónSUROESTE Falmouth Otros equipos de transporteSUROESTE Gloucester Otros equipos de transporteSUROESTE Poole Otros equipos de transporte Otros equipos de transporte

219


Sistemas locales de trabajo Distritos industriales de Distritos industriales de(clasificación de 1998) pequeña empresa (0-99) mediana empresa (100-299)

SUROESTE Bridgwater Caucho y plásticoSUROESTE Malvern Caucho y plástico Caucho y plásticoSUROESTE Chard Curtido de pieles y vestidoSUROESTE Wells Curtido de pieles y vestido Curtido de pieles y vestidoGALES Welshpool MaderaGALES Knighton y Radnor Metales básicosGALES Rhymney y Abergavenny Maquinaria eléctricaGALES Pontypridd y Aberdare MobiliarioGALES Merthyr Maquinaria y equipoGALES Newtown TextilesMEDIO OESTE Leominster MaderaMEDIO OESTE Ludlow Caucho y plásticoMEDIO OESTE Wolverhampton y Walsall Curtido de pieles y vestidoMEDIO OESTE Dudley y Sandwell Metales básicos Metales básicosMEDIO OESTE Worcester Metales básicosMEDIO OESTE Stafford Maquinaria eléctricaMEDIO OESTE Oswestry MaderaMEDIO OESTE Kidderminster TextilesMEDIO OESTE Leek TextilesY&H Calderdale TextilesY&H Keighley y Skipton TextilesY&H Sheffield y Rotherham Metales básicosY&H Scarborough Maquinaria eléctricaY&H Bridlington y Driffield Alimentos y bebidas Alimentos y bebidasY&H Barnsley Productos no metálicosY&H Pickering Otros equipos de transporte Otros equipos de transporteY&H Wakefield Prendas de vestir y pielesY&H Huddersfield Textiles


Continuación Tabla 4. Distritos industriales en Gran Bretaña (2002)


220

6. Conclusiones

Las reflexiones presentadas en este ensayo constituyen un primer intento de describir yanalizar el fenómeno de los sistemas locales, y en particular de los distritos industriales en GranBretaña. Los resultados de la cartografía aportan importantes indicaciones a tres niveles: (1)muestran la relevancia espacial de los sistemas locales; (2) ayudan a individualizar tipologías desistemas locales; y finalmente, (3) suministran una exacta descripción espacial de los distritosindustriales (proto-distritos).

No hay duda de que esto es sólo el primer paso hacia un análisis más detallado de lasespecíficas realidades territoriales de los distritos industriales, para el que es necesario un diver-so y más complejo set de datos e indicadores.

Tal análisis tiene un gran valor, dada la relevancia que los sistemas productivos localeshan asumido en el debate político en el Reino Unido; de hecho, el Gobierno central ha delegadoa las agencias locales de desarrollo (regional developemt agencies) la tarea de identificar priori-dades económicas sobre las que focalizar fondos y energía. En este sentido, al viejo enfoque depolicy que veía los sectores como objetivo de acciones y decisiones, a partir de 2001 talesagencias han revisado sus estrategias de política, señalando priority clusters como pilares deldesarrollo económico regional. La individualización de tales sistemas locales es, por lo tanto, nosólo relevante, sino sobre todo necesaria.

7. Bibliografia

• ALDCROFT, D. H. (1968): The Development of British Industry and Foreign Competition.Londres, George Allen and Unwin.

• BAILEY, D. y DRIFFIELD, N. (2007): «Industrial Policy, FDI and Employment: Still a MissingLink»; en Journal of Industry Competition and Trade (7); pp. 189-211.

• BECATTINI, G. (1987) : Mercato e Forze Locali: Il Distretto Industriale. Bolonia, Il Mulino.

• BECATTINI, G. (1994): «The Development of Light industry in Tuscany: An Interpretation»; enLEONARDI, R. y NANETTI, R. Y., eds.: Regional Development in a Modern European Economy.Londres, Pinter; pp. 69-85.

• BECATTINI, G. (2000): Il Bruco E La Farfalla: Prato Nel Mondo Che Cambia. Florencia, LeMonnier.

• BECATTINI, G. (2001): Il Caledoscopio dello sviluppo locale: trasformazioni economiche nell’Italiache Cambia. Turín, Rosemberg & Sellier.

221


• BECATTINI, G.; BELLANDI, M.; DEI OTTATI, G. y SFORZI, F. (2003): From Industrial Districtsto Local Development. Cheltenham, Edward Elgar.

• BERG, M. (1994): «Factories, workshops and industrial organisation»; en FLOUD, R. yMcCloskey, D., eds.: The Economic History of Britain since 1700. Cambridge, CambridgeUniversity Press; vol. 1.

• BRACZYK, H.; COOKE, P. y HEIDENREICH, M., eds. (1998): Regional Innovation Systems.Londres, UCL Press.

• BRUSCO, S. y PABA, S. (1997): «Per una storia dei distretti italiani dal secondo dopoguerraagli anni novanta»; en BARCA, F., ed.: Storia del capitalismo italiano. Roma, Donzelli; pp.265-333.

• CAMAGNI, R. P. (1995): «The Concept of Innovative Milieu and its Relevance for Public Policiesin European Lagging Regions»; en Papers in Regional Science (74, 4).

• CLAPHAM, J. H. (1930): An Economic History of Modern Britain. Cambridge, CambridgeUniversity Press.

• COOKE, P. (2001): «Regional Innovation Systems, Clusters and the Knowledge Economy»;en Industrial and Corporate Change (10, 4); pp. 945-974.

• COOPEY, R. (2003): «The British Glove Industry, 1750-1979: The Advanategs and Vulnerabilityof a Regional Industry»; en WILSON, J. F. y POPP, A., eds.: Industrial Clusters and RegionalBusiness Networks in England, 1750-1970. Aldershot, Ashgate.

• CROUCH, C. y FARREL, H. (2001): «Great Britain: Falling through the Holes in the NetworkConcept»; en CROUCH, C.; LE GALÈS, P.; TRIGILIA, C. y VOELZKOW, H., eds.: LocalProduction Systems in Europe. Rise or Demise? Oxford, Oxford University Press.

• DEPARTAMENT OF INDUSTRY AND TRADE (2001): Business Clusters in the UK. A FirstAssessment. Londres.

• DE PROPIS, L. (2001): «Systemic Flexibility, Production Fragmentation and Local IndustrialSystem Governance»; en European Planning Studies (9, 6); pp. 739-753.

• DE PROPIS, L. (2005): «Mapping Local Production Systems in the UK: Methodology andApplication»; (39, 2).

• DE PROPIS, L. y LAZZERETTI, L. (2007): «The Birmingham Jewellery Quarter: A MarshallianIndustrial District»; en European Planning Studies (15, 10); pp. 1.295-1.325.

• HUDSON, P. (2004): «Industrial Organisation and Structure»; en FLOUD, R. y JOHNSON, P.,eds.: The Cambridge Economic History of Modern Britain 1700-1860. Cambridge, CambridgeUniversity Press; vol. 1.

• MARKUSEN, A. (1996): «Sticky Places in Slippery Space: A Typology of Industrial Districts»;en Econ. Geogr. (72, 3); pp. 293-313.

Jesus A. Trevino

Highlight

Jesus A. Trevino

Typewritten Text

Jesus A. Trevino

Typewritten Text

Enseguida


222

• MARSHALL, A. (1920): Priciples of Economics. Londres, MacMillan (8ª edición).

• MARTIN, R. y SUNLEY, P. (2003): «Deconstructing clusters: chaotic concept or policy pana-cea?»; en Journal of Economic Geography (3); pp. 5-35.

• PORTER, M. E. (1998): «Clusters and the new economics of competition»; en Harvard Busi-ness Review; pp. 77-90.

• PYKE, F.; BECATTINI, G. y SENGENBERGER, W. (1990): Industrial districts and inter-firmcooperation in Italy. Génova, ILLS.

• SFORZI, F. (1989): «The Geography of Industrial Districts in Italy»; en GOODMAN, E. yBAMFORD, J., eds.: Small Firms and Industrial Districts in Italy. Londres, Routledge; pp.153-173.

• SFORZI, F. (1990): «The Quantitative Importance of Marshallian Industrial Districts in theItalian Economy»; en PYKE, F.; BECATTINI, G. y SENGENBERGER, W., eds.: IndustrialDistricts and Inter-firm Cooperation in Italy. Génova, IILO; pp. 75-107.

• SFORZI, F. (1996): Italy: Local Production Systems of Small and Medium-sized Firms andIndustrial Changes. OCDE, Networks of Enterprises and Local Development.

• SFROZI, F. (1999): «Economic Change»; en BONAVERO, P.; DEMATTEIS, G. y SFORZI, F.,eds.: The Italian Urban System. Aldershot, Ashgate.

• WILSON, J. F. y SINGLETON, J. (2003): «The Manchester Industrial District, 1750-1939:Clustering, Networking and Performance»; en WILSON, J. F. y POPP, A., eds.: IndustrialClusters and Regional Business Networks in England, 1750-1970. Aldershot, Ashgate.

• WISE, M. J. (1950): «On the Evolution of the Jewellery and Gun Quarters in Birmingham»; enInstitute of British Geographers (15).

• ZEITLIN, J. (1995): «Why are there no Industrial Districts in the United Kingdom?»; enBAGNASCO, A. y SABEL, C. F., eds.: Small and Medium Enterprises. Derby, Pinter.

udies. Vol. 39.2. pp. l')7-211. Apnl 2t)()5 O Routledge

Mapping Local Production Systems in the UK:Methodology and Application

LISA DE PROPRISInstitute for Industrial Deuelopmetit Policy, Birmingham Business School, University of Bimiitigham, Edgbaston, Birmingham

B15 2'rr, UK. Email: [email protected]

(Received June 20{)2: in revised form December 2003)

Dr PnoPHis, L, (2005) Mapping local production systems in the UK: mcthodolot!;y and application. Regional Studies 39. 197-211. The paper outhnes a possible methodology to map and study local production systems. Tht- three-level diagnosticmethodology enables researchers to map, classify and analyse in depth firms' agglomerations in regions or countries where thereis little intorniation about the presence and location oflocal production systems. The spatial diagnostic procedure is applied toilie UK to ni.ip local production systems.

Location Industrial districts Agglomerations

is, L. (2(105) L'elaboration des systemes de production iocaux au Royaume-Uni; une methodologie et son application.Regional Studies 39, 197-211. Cet article cherche a esquisser une methodologie eventuelle qui sert a elaborer et a etudier lessystemes de production Iocaux. Cette methodologie diagnostique a trois niveaux permet aux chercheurs d'elabiirer, de classeret d'approfondir les agglomerations des entreprises dans les regions ou les pays ou il y a peu de renseignements a propos ou dela presence, ou de la localisation, des systemes de production Iocaux. La demarche diagnostique geographique se voit appliquerau Royaume-Uni afin d'elaborer les systemes de production Iocaux.

Localisation Districts industriels Agglomerations

D E PROPRIS, L. (2005) Kartierurig iirtlichcr Produktionssyteme im Vereinigten Konigreich: Methodik und Anwcndnng,Regiotiiil Studies 39, 197-21 1. Dieser Autsatz unireilJt eine in Betracht zu ziehende Methodik der Kartierung und Untersuchungortlicher Produktionssysteme. Die diagnostische Methodik auf drei Ebenen gestattet Forschern, Firnienballungen in Regionenoder Landern. wo es an Information liber Vorhandensein und Standort ortlicher Produktioiissysteme nuiigclt. zu kartieren, zuklassifizieren und eingehend zu analysicren. Das raumlich-diagnostische Vertahren wird zur Kartierung ortlicher Produktionssys-teme angewandt.

Standort Industriegebiet Dallungen

Dr. FROPKIS, L. (2005) El mapeo de los sistemas de produccion locales en el Reino Unido: planteaniiento metodologico yaplicacion, Regional Studies 39, 197—2] ], El articuio resume un posible planteaniiento metodologico para niapear y estudiar lossistemas de produccion locales. La metodologia de diagnostico de tres niveles propuesta permite a los mvestigadores que lautilicen mapear, clasificar y analizar en profundidad aglonieraciones de cmpresas en aquellas regiones o paises donde existe escasaintormacion sobre la presencia y la localizacion de sistemas locales do produccion. El procedimiento de diagnostico espacial seaplica al Rcino Unido para mapear sistemas locales de produccion.

Localizaci6n Distritos industriales Aglomeraciones

JEL classification: R12

INTRODUCTION drawing from them generalizable observations. Oneconimon element across all these studies was their

The flourishing debate on local production systems confirmation that LPSs have been crucially important(LPSs). e.g. industrial districts, clusters and networks, for industrial development, regional development andhas mostly involved in-depth analyses of existing phe- employment. Over the last decade, studies on industrialnoniena in Europe and the USA in the attempt to districts in Italy (LEONABDI and NANETTI . 1994;understand their nature and dynamics. These studies COSSENTINO et ai, 1996; BECATTINI, 2001), onaimed at identifying case-specific elements, as well as clusters in Portugal (PORTER, 1998), the USA

01)34-3404 pnnt/13WI-()5<91 online/05/020147-15 ©2005 Regional Studies Association DOI: t0.1U80/U03434O()5200059983

li ttp://www.rfgioiial-studies-assoc.ac.uk

198 Lisa De Propris

(PORTER, 2000) and Norway (BJOKG and ISAKSEN,

1997), and innovative milieu in France (LONGHI,

1999) and developing countries (RABELLOTTI, 1997;GuERRiERi ct al. 2001) have all strengthened theargument that agglomerations of small- and medium-sized enterprises (i.e. firms) (SMEs) can catalyseregional industrial competitiveness.

Most LPSs, however, tend to be studied once theybecome so successful as to catch the attention ofgeographers or economists. At that point, they areanalysed to trace their evolution, to understand theirfunctioning and the reasons for their success and, finally,to puU out elements ot replicability. In other words,researchers' analysis of production systems tends to startonly after the phenomenon has become visible to theeye. One shortcoming of this 'approach" is that researchtends to omit two important categories of LPSs: emerg-ing production systems in industrialized countries andthose in embryonic form in developing countries. Infact, in industrialized countries, there are sectors, sub-sectors or even product niches that tend to grow quicklyand show a localized concentration. If, as happens inmost cases, localization contributes to agglomerationand extemal economies, which in turn boost competi-tiveness, then such phenomena could be singled outbefore they become common knowledge.

The thrust of this paper is, therefore, to present a setof tools to map and diagnose LPSs in a region orcountry where there is httle or no knowledge aboutthe existence and location of such systems. LPSs arereferred to generally as the geographical agglomerationof firms specialized in one or a few complementarysectors. Such producdon systems are characterized byan external division of labour, a more or less developedsocial capital and a more or less engaged institutionalframework. This definition was kept as general aspossible since any further categorization would implythe need for more detailed and accurate informationabout individual clusters. In the current literature manytypologies of LPSs have been proposed to the pointwhere, as argued by MARTIN and SUNLEY (2003), theconcept of a cluster has been given so many labels andhas been defined in so many ways that it has become achaotic and ambiguous one. After the first conceptual-ization ofindustrial districts (BECATTINI, 1987, 2000,2001; PYKE et ai, 1990), further definitions includehub-and-spoke districts (MARKUSEN, 1996), clusters(PORTER, 1998, 2000; COOKE, 2002) and, if theemphasis is on inter-firm innovation and technologyprocesses, innovative milieu (CAMAGNI, 1995). Sim-ilarly, GORDON and M C C A N N (2000) distinguishbetween three forms of clustering; the classic model ofpure agglomeration, the industrial complex model andthe social network model. Meanwhile, GARCJFOLI'S

(1991) typology of LPSs includes areas of specializedproduction, local productive systems and systemareas. SIMMIE and SENNETT (1999) add cities toMARKUSEN'S (1996) typology; while BELUSSI and

ARCANGELI (1998) distinguish between evolutionary,reversible and steady-state networks.

The set of tools outlined in this paper is a method-ology that allows one to carry out the followingdiagnostic study of production systems: (1) to localizeagglomerations of small, medium and large firms in anarea producing in the same sector; (2) to map suchproduction systems across a territory; (3) to provide ataxonomy' of LPSs according to basic categories (e.g.sector, firm size) in order to identity patterns of regionalspecialization; (4) to understand the structure of inter-firm relationships; and (5) to assess the role of theinstitutional framework.

The diagnostic methodology has three levels ofanalysis: spatial diagnostic, diagnostic analysis of inter-firm relationships and analysis of the institutional frame-work. It relies both on statistical and primary datacollected via questionnaires. Qualitative and quantita-tive tools tend to be complementary, allowing for arounded research approach that moves from a generalmapping of LPSs to the in-depth study of each indi-vidual case or a selection of them. The mix of data andcase studies suggested in this methodology is coherentwith the types of information it brings together. Oneis not distinguishing between measurable and non-measurable variables, since case studies can also producequantifiable information sometimes as dummy variables(e.g. belonging to a civic club or association) or discretevariables (e.g. the output share of a firm's largest buyer).Rather, what matters is the depth of analysis. Thegeneral description of the phenomenon (levels 1 and 2)must rely on genenilized indexes and formulas thatrequire data. However, LPSs are characterized by morethan employment concentration and inter-sectoralinput-output flows: there are intangible factors andsystem-specific elements that can be only analysedproperly with a case study. To some extent, the data-based approach enables general conclusions to be drawnon the phenomenon, whereas the case study approachprovides precise information that can be used to showthe variants and shades in which the phenomenonpresents itself in reality.

The possibility of identifying LPSs since their originand studying them throughout their developmentwould enable research to move away from a typical ex-post static approach (e.g. PORTER'S (1998) work onthe wine valley in California) that explores successfulsystems and tries to explain them by looking back atthe causes of their success. Rather, researchers couldadopt a more dynamic approach where the patterns ofproduction systems' evolution can be evaluated as theyunfold. Note too that the analysis of 'success stories'only provides a partial description of the more generalphenomenon of local industrial systems because itneglects to study 'failure stories', i.e. production systemsthat, although displaying industrial specialization andgeographical proximity, have not emerged or continuedsuccessfully but rather have lingered before dechning.

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Mappinji Local Production Systems in the UK 199

Yet, LPSs have also proved to be engines ofindustrialdevelopment (BELLANDI, 2001) alternative to largefirms, and for [his reason they are deserving specialattention in developing countries. More recently,BELLANDI and SFORZI (2001) recognize that a multi-plicity of hybrid forms of local systems can be foundby combining four elements: small firms, large firms,and rural and urban environments. In particular, theyidentify five possible categories of LFSs: industrialpoles, rural local systems, big dynamic cities, buyer-dependent local systems and industrial districts. Eachform has the potential to trigger and catalyse regionaldevelopment.

The usefulness of this methodology with respect tothe clustering phenomenon is related not only to itsexploratory value, but also to the fact that it can provideinformation for the design and definition of objectivesfor regional policy. Once LPSs have been mapped andanalysed, regions have a better understanding ot theirstrengths and weaknesses. This awareness can assistpolicy-makers in formulating a set of policy measures tocreate the conditions for LPSs to grow and consolidate.These measures could include training, business-to-business marketing, research and development,exporting, finance and specialized service provision. Inthe UK, the publication of the Department of Tradeand Industry's (UTI) report on clusters in the UK(DEPARTMENT OF TRADE AND INDUSTRY, 2001) has

been used by regional development agencies across thecountry to incorporate the targeting and support ofclusters in their economic strategy. The author feelsthis methodology would be even more valuable toinform policy-makers in transition and developingcountries where the need to trigger the combinedregional and industrial development often lies initiallyin catalysing the further development of existingcompetencies and specializations.

Little work has been done so far to diagnose LPSs.MARTIN and SUNLEY (2003) divide the empiricalwork done to identify and map clusters in two broadapproaches: top-down mapping exercises and bottom-up qualitative studies. Both approaches seem to havefundamental shortcomings. Pt)RTER's (199H) mappingof the wine cluster in California and the footwearcluster in Portugal possibly falls under the latter categorysince the study of these clusters was based on knowledgeand information gathered on the ground. In the UK,there have been two main quantitative mapping studies;the DTI report Busimss Clusters in the UK (2001) andwork of CROUCH and FABREH- (2001). The DTIreport, in particular, has opened the debate on whatare the possible and appropriate methodologies todiagnose LPSs, which can then be applied to mapknown and unknown realities.

The main differences between the DTI methodologyand that presented herein are manifold. First, differentgeographical units of analysis are referred to for statisticaldata. The DTI uses both regional and local authority

data from various databases." In contrast, in the spatialdiagnostic procedure, a case is made for using travel-to-work areas (TTWAs) data supplied by the UKOffice for National Statistics. The reason for using thesedata is that as they correspond to self-contained workingand living areas, they are more appropriate to identifyLPSs. As PANICCIA (2002, p. 48) notes, TTWAs'encapsulate the features of a locally bounded communi-ty, given that it identifies a restricted area in whichinteractions between firms and populations are verydense'.

Second, the DTI methodology relies on a data-basedanalysis and qualitative information gathered from localinstitutions and organizations about the identifiedclusters. The objective of the data-based analysis is tomeasure employment density via a location quotientand data about industries' regional employment. Thereport highlights "high points' as those industries torwhich regional employment accounts f~or at least 0.2%of the total regional workforce and for which thelocation quotient shows that the regional industry is25% more concentrated than at the national level. Forsector /, region r and country n, the local quotient (LQ)is defined as follows:

>1.25

ln other words, the data-based analysis of the DTIonly measures industrial concentration as mirrored byemployment density. Qualitative analysis is aimed atranking and classifying the identified clusters underfour headings: stage of development (embryonic, estab-lished, mature), deptb (deep, shallow, unknown),employment dynamics between 1991 and 1998 (grow-ing, declining, stable) and significance (regionally,nationally, internationally).

The spatial diagnostic procedure presented in thispaper is more complete in that in addition to measuringindustrial concentration (step 3), it considers firms' sizeand the use of data on units of production by sectorand by TTWAs. Both elements provide a much moreinformative picture of firm clustering as opposed simplyto mapping employment density. Furthermore, theanalysis of case studies by means of the quantitativeinput-output analysis and the qualitative analysis of theinstitutional framework would provide a richer pictureof the phenomenon. For instance, in the input-outputanalysis, one could provide imponant information onthe flow of inter-industry exchanges and on inter- andintra-LPS sectoral networks.

The paper is structured as follows. The second tofourth sections will discuss the three levels of thediagnostic methodology. The fifth section wall presentthe results of the application of the spatial methodologyto map LPSs in the UK. Finiilly, the sixth section willpresent some concluding remarks.

200 Lisa De Propris

SPATIAL DIAGNOSTICS

The obvious first level of the diagnostic study of LPSsin a country or region is to pinpoint on a map wherethey are and to attempt an initial understanding ofwhattliey are. For this purpose, it is important to have aconsistent statistical data set that contains informationabout firms' production activities according to threecoordinates: sector, size and location. A nuip of firms"agglomerations provides the basic information aboutproduction systems across a country or region: in somecases, it might confirm the existence of already knownphenomena, while in some other cases, it can flag upnew infant systems.

A methodology already tested and proved successfulis that used by SFORZI (1990) to map Italian industrialdistricts and other types of LPSs. Until then, industrialdistricts were only perceived as regional experiencesand almost unique cases attached to local traditionalexpertise. The application of this methodology high-lighted the presence of LPSs across Italy in the 19H()sand gave an initial picture of the magnitude of thephenomenon." Furthermore, it showed that in additionto industrial districts, there were other types of produc-tion systems that played a diflerent, and crucial, role forregional industrial development such as 'non-manutacturing local production systems' or 'local pro-duction systems of large firms' (BHUSCO and PABA,

1997).

Sforzi's methodology 'does not identify districts, butlocal systems whose production structure is compatiblewith that of districts" (13RUSCO and PABA, 1997, p. 278;author's translation) and this represents its strength,especially when applied with an open perspective withrespect to what types of LPSs could be found. In fact,even before the case-based analysis, it enables an initialclassification of LPSs according to sector, firm size andurban/non-urban area.

The methodology relies on data on employmentshares in firms (considered as units of production)according to size, sector and location. Small firms areseen as having fewer than 100 employees, SMEs havefewer than 250 employees and large firms have morethan 2S() employees."' Manufacturing sectors are definedaccording to the national classifications, which can bemore or less aggregated. Moving from a two- to athree-digit sector classification might alter the mappingot local systems, but it tends to be a very interestingexercise if one wants to single out extremely dynamicproduct niches.^ Finally, location is defined by thestatistical geographical units that partition the nationalor regional territory. These can correspond to regions,counties, provinces, cities or any meaningful combina-tion of them. SFORZI (1990) argued that the mostappropriate geographical units to identify LPSs tend tooverlap with local labour markets. As self-containedTTWAs, they reflect the overlapping of the communityof people living in a certain area and the population

of economic activities. Local labour markets rarelycorrespond to standard geographical classifications inthat they often cut across regions or counties and grouptogether towns and villages that would i>therwise nothave anything in common. C R O U C H and FARHELL

(2001) use T T W A data for employment and units ofproduction figures to identify clusters in the UK. Inparticular, they identified four types of firm agglomera-tion: industrial districts, concentrated clusters, weakclusters and simple clusters.

Before going any further, note that there is no reasonfor restricting analysis only to the manufîcturing sector.However, the author feels that the mapping of servicelocal systems might require a slightly difference set ofcriteria. For instance, firm size might not be so relevantfor the service sector. The location quotient a.s definedin the third criterion could, nevertheless, provide usefulinformation about the geographical concentration ofservice industries (for tlie mapping of Italian urbansystems of services, see SFORZI, 1999).''

For the spatial diagnostic analysis described in thispaper, four criteria need to be considered: the share ofemployment in the manufacturing sector; the size offirms; industrial specialization; and industrial specializa-tion and the size of firms.

Share of employment in the nuiiiufacturing sector

This criterion assesses the proportion of employmentin the manufacturing sector (manuf.) in each T T W Aout of the total non-agriculture employment (nat.),which is benchmarked against the national employmentshare of the manufacturing sector out of the nationalnon-agricultural employment:

E(TTWA, manuf.) H(nat., manuf~.

E(TTWA) E(nat.)(1)

If the ratio of manufacturing to total employment inthe TTWA is greater than the national average, itmeans that manufacturing employment dominates thelocal labour market in question. In contrast, if the ratiois below the national average, it indicates that servicesectors are dominant. SFORZI (1990) and 13RUSCO andPABA (1997) stress that one of the key features ofindustrial districts is that of being embedded in amanutacturing environment. In a further application ofSforzi's methodology, SOLINAS and BARONI (1998)distinguish between manufacturing local systems andnon-manufacturing local systems to highlight theincreasing importance of local systems (not necessarilyindustrial districts) in de-industrialized or still agricul-tural areas where manufacturing sectors account for alimited share of the local economy.

Size of firms

The second criterion concerns firms' size and assessesthe composition of firms' population in the local labour

Mapping Local Production Systems in the UK 201

market according to whether firms are mostly small,medium or large sized. In particular, it considers theproportion of employment in firms with fewer than100, fewer than 250 and more than 250 employeesout of the total TTWA's manufacturing employment.Again, the benchmark is the share of the nationalmanufacturing employment by firms' size out of totalmanufacturing employment:

H(TTWA, manuf., size) .E(nat., manuf, size)

E(TTWA. manuf.) H(nat., manuf)(2)

This criterion allows researchers to ascertain the domin-ance of small, medium or large firms in the local labourmarket. In the case ofindustrial districts, for instance,SFORZI (1990) argued that firms had to have fewerthan 250 employees. In contrast, if medium or largefirms dominate, other t>'pes of local systems can befound. Finns" size is an important element when startinganalysis of LPSs. It can be argued that inter-firmdynamics and the governance of LPSs cliange greatlyaccording to whether the system is dominated by smallor large firms (DE PROPRIS, 2001). Besides, firms' sizeis important in order to uncover geographical dividesin the organization of production activities.

Industrial specialization

The third criterion to assess LPSs is to look at theirindustrial specialization. In fact, whichever types ofagglomeration of firms one aims to find, they tend tobe characterized by production specialization in onesector or, in the case of a filicre, in a few related andcomplementary sectors.

Manufacturing sectors can be classified by relyingon different levels of aggregation: a two-digit sectorclassification usually allows researchers to identify themain production sectors (e.g. textiles, chemicals), whilea three-digit sector classification allows a breakdownof each sector into sub-sectors. Within the presentmethodology, the use of either classification has advan-tages and disadvantages. On the one hand, the use of asub-sector classification means the possibility of identi-fying production systems on the basis of a preciseproducdon specialization such as ceramic tiles (a sub-sector of the non-metallic mineral product sector) orhosiery (a sub-sector of the textiles sector). However,it means having to handle a large data set given thatsub-sector data are in turn broken down by TTWAs.On the other hand, using the two-digit classificationmakes the handling of the methodology more straight-fbi-Vk'ard, although some information might be lost.

SFORZI (1990) relied on the two-digit sector classifi-cation. Industrial specialization is given by the localquotient:

In other words, the proportion of employment ina sector out oi the total TTWA's manufacturingemployment has to be greater than the share of thesector at the national level. Industrial concentrationmirrors, therefore, the accumulation of sector employ-ment across all firms' size. In so doing, it can be saidthat an area is specialized in one sector. As alreadymentioned, the same area can be specialized in a fewsectors; sometimes they tend to be related in a type ofsector chain (Jiiicre), or can be completely unrelated. Inthe latter case, one possible explanation is that the areahas been or is still manufacturing-intensive so that avariety of sectors have developed and are still present.Another explanation could be the presence of urbanareas that tend to be centres of accumulation of manyindustrial activities. Sforzi argued that industrial districtstend to be specialized in only one sector.

Industrial specialization and firms' size

Finally, the last criterion brings together informationabout firms' size and industrial specialization so as toclarify whether the sector(s) in which a certain area isspecialized is (are) characterized by small, mediumor large firms. This means looking at whether theproportion of employment in sector / by firms' size outof the total TTWA's sector employment is smaller orgreater than the share of the national employment bysector and firms' size out of each sector's nationalemployment:

£(TTWA, sector,size)

E(TTWA, sector)

'(nat., sector,size)

E{nat,, sector)(4)

H(TTWA, sector)/E(TTWA, manuf)

H(nat., sector)/E(nat., manuf)(3)

This last criterion enables one to understand whether acertain agglomeration of firms is formed by small orlarge firms, or whether some sectors are more likely tobe characterized by small or large firms. In industrialdistricts, industrial specialization must be characterizedby SMEs, whilst other types of production systems canbe completely different in their make-up. For instance,hub-and-spoke districts are formed by one or a fewlarge firms and many small firms respectively producingin related industries.

Pros aud cons of the spatial diaj^noslic analysis

The spatial diagnostic analysis just described producedthe first systematic assessment of Italian industrial dis-tricts and identified a wide range of LPSs different fromindustrial districts. In particular, to identity industnaldistricts, SFORZI (1990) suggested that local labourmarkets: (1) had to be dominated by manufacturingsectors; (2) had to be characterized by firms with fewerthan 25(1 employees; (3) had to be specialized inone sector; and (4) the specialized sector had to becharacterized by firms with fewer than 250 employees.The result was the well-known maps of Italian districts

202 Lisa De Propris

Table 1. Local production systems (LPS) iisin;^ Sforzi's methodology

Pre-district

LPS of sectors outsidespecialized small LPS ot medium the industrial LI'S of large

firms and large firms district firms

Non-iractLPS

{]) Manufacturing dominance

{2a) Dominance of small firms{fewer than 25(1 employees)

(2b) Dominance of large firms{more than 250 employees)

(3) liidiiscrial specialization

(4a) Size and specialization(fewer than I(H) employees)

(4b) Size and specialization{fewer than 251) employees)

{4c) Size and specializarion(more than 250 empioyecs!

Soimes: BBUSCO and PAHA (1 'W7), SOLINAS and BARONI (1998), SFOBZI (199U, 1996).

in Sforzi. More precisely, what was found were 'pre-districts' (j)roto-disrretti). namely, LPSs with the spatialcharacteristics of districts (small firms, sector specializa-tion, agglomeration), but that could be labelled as suchonly after a deeper analysis of the structure of inter-firm relationships and of the institutional framework.

Overall, its main advantage is to provide an automaticprocedure to carry out a spatial diagnostic analysis ofLPSs. It relies on statistical data available at the regionalor national levels. It must be stressed that the fourcriteria do not Just enable the identification of onlyindustrial districts, but all sorts of production systems.SFORZI (199(3) also mapped 'northern and centralurban systems', 'southern urban systems' and 'northernmanufacturing local systems'. In the same way, BRUSCO

and PABA (1997) mentioned 'specialized sectors ofsmall firms', 'non-specialized sectors of small firms','small- and medium-sized sectors' and 'large firm sec-tors' (Table 1). The breadth of the classification thatemerged from the use of Sforzi's methodology showedthat it can be a powerful tool to carry out spatialdiagnostics when one keeps an open perspective onwhat could be found. This is especially true if onebegins to pinpoint LPSs on a blank geographical map.

The main drawback of the procedure is that it onlycaptures localized industrial specialization according tothree dimensions: firms' size, sector and location. It issilent, however, on firm interactions. Only a deeperanalysis can help in terms of quahfying the nature andgovernance of the LPS. This explains why Sforzi'smethodology constitutes only a first step. The secondstep is to analyse inter-firm relationships to understandthe nature of the production system. The fifth sectionpresents the findings of an application of the spatialdiagnostic procedure to map LPSs in the UK (DEPROPKIS, 2003).

DIAGNOSTICS OF INTER-FIRMRELATIONSHIPS

The second level in the assessment of LPSs is to analysethe structure and nature of inter-firm relationships.Before attempting this, two questions need to be raised:Do firms in LPSs necessarily interact?; and How dothey interact? It cannot be taken for granted that inlocal systems firms engage in some form of relationship,but there is much evidence showing that the localizedagglomeration of production activities is tightly relatedto the presence of firms' interdependencies (traded anduntraded). Firms locate close to other firms to interactwith them as much as they are encouraged to interactwith other firms because of proximity. COOKE andMORGAN (1998) emphasize the associational nature oflocalities; PioRE and SABEL (1984) refer to regionalconglomerations of firms; MARKUSEN (1996) accentu-ates the stickiness of some places; whilst STORPER

(1996) looks at the locality as a nexus of untradedinterdependencies. It is also well established that prox-imity and firm cooperation contribute to agglomerationand external economies." There are many ways inwhich firms can interact. The most common is throughthe exchange of outputs and inputs along the produc-tion chain. LPSs are almost by definition complexnetworks of buyers and suppliers, together contributingto the production of the final goods. Subcontractingrelationships can often be accompanied by other formsof interaction, such as cooperation over innovation,'^joint ventures, joint purchase of inputs and jointtraining.'"

However, one needs to distinguish between quantita-tive and qualitative aspects of inter-firm relationships.The existence of production relationships betweenbuyers and suppliers can be detected and measured with


an input-output analysis, whereas the nature of suchrelationships and the dimensions of inter-firmcooperation can be only detected with case-by-casequalitative studies that rely on questionnaires or inter-views. Therefore, the need to handle a multifacetedmethodology that involves both quantitative andqualitative approaches is again stressed.

Input-output analysis has been traditionally used tomeasure inter-industry trade at the macro level, inter-industry multipliers and inter-industry linkages{RASMUSSEN, 1956; CHENERY and WATANABE,

1958; IsARD el ai, 1998). In reality, it can also be anextremely useful instrument for the study of LPSs.Input-output tables enable researchers to quantify inter-industry exchanges and localize the origin and destina-tion of such exchanges (i.e. interregional flows), input-output tables quantify the monetary value of sales (orpurchases) from (to) sector i to (from) sector j . It ispossible to construct a map of the network of inter-industry exchanges that mirrors the network of theexternal division of labour (inter-firm production link-ages). This can be a national or regional map dependingon the data available froni countries' statistical offices.In the USA, such data are supplied at the state level;in the UK, there are data only at the national level andfor Scotland. EUROSTAT provides input and outputtables for the 15 European Union Member States.

An example of input-output analysis is in FESER andIiER(;MAN (2000), who map the 'vehicles manu-facturing cluster' in the USA, namely the complexnetwork of inter-industrial linkages. Such linkages areassumed to be underpinned by inter-firm linkages, sothat the result is a map of the buyer-supplier networkaround the automotive sector. FESER and BERGMAN

(2000) map the automotive cluster across the USAproviding no information about the geographicaldimension of the cluster. However, when regional dataare available, it is possible to map an inter-industrynetwork at the local level.

Input-output analysis can therefore be a useful tool:(1) to identify the set of industries involved in an LPS;(2) to see the direction of the exchange flows {i.e. whobuys from whom); and (3) to evaluate the relativestrength of the direct and indirect inter-industry link-ages. Fig. 1 shows in stylized form the map that canbe derived from an input-output analysis, where thedirection of the exchange between buying and supply-ing industries is described by the arrows betweenindustries (e.g. F buys from Y and sells to C). Someindustries, like C, are the hubs of the systems becausemany inputs are channelled in that direction wherethey are assembled into a final or semi-final outputthat in turn is sold to a downstream industry beforeapproaching the fmal market. The thickness of thearrows corresponds to the intensity of inter-industryflows.

The main limitations of the input-output analysis

Fig. 1. huer-indusfry linkages

are twofold. One is that it needs a predefined geograph-ical unit of analysis; the other is that there is noinformation about firm size. In fact, if the input-outputanalysis is carried out at the national level, it will mapthe network of inter-industry linkages across the entirecountry with no information about localization oragglomeration. Because of these limitations, input-output analysis is best seen as complementary to Sforzi'sspatial diagnostic analysis in that it adds key informationabout the dynamics of inter-firm linkages between sub-sectors.

A further step to understand the nature of suchrelationships would involve a qualitative assessment viaquestionnaires and case studies. Given that a qualitativemethodology is very labour intensive and time consum-ing, it should be adopted only to carry out an in-depthanalysis ot one particular LPS, and in any case, it cannotbe the dominant methodology for the study of LPSs.

A qualitative assessment can clarify in how manytypes of linkages firms tend to engage in additionto production input—output exchanges; u^hether firmsshare or exchange information and knowledge,cooperate over innovation, training, marketing andexports, or embark on different sorts of joint ventures.Moreover, a qualitative assessment could also capture

204 Lisa De Propris

elements that cannot be measured such as trust,cooperation, the degree of firms' einbeddedness andthe role of social capital. These intangible elements arecrucial for the functioning of LPSs. However, they canonly be identified by means of primary informationgathered directly from agents (e.g. firms and institutions)that operate within the local system. Because of theirdaily involvement in the lite ot the local system, localagents can be unique sources of information about thedegree of trust between firms, the degree of cooperationbetween firms and insritutions, the role of the civicsociety behind the functioning of the production systemand, finally, the trade-ort~ between einbeddedness andopenness for the sustainability of the LPS.

To summarize, once a map of LPSs has been drawnby means of Sforzi's methodology, the analysis of inter-firm relationships enables a better understanding of thenature and dynamics of the linkages between firms.Here an in-depth analysis should rely on both input-output analysis and a qualitative assessment of therelationships.

INSTITUTIONAL FRAMEWORK

Tlie development, and often survival, of LPSs can belinked to the presence and role of the local and regionalinstitutional framework. This comprises a network ofpublic and private institutions related in many differentways and to different degrees to hrms in an LPS.Analysing this institutional framework is the final stageill the diagnostic analysis of the LPSs.

The geographical dimension of the institutionalframework is crucial to define its interaction with firms.LPSs are often bounded within a territory that isnarrower than the administrative regional borders and.therefore, need local business support organizations andbodies. These are meant to reflect the specific interestsand needs of firms in the LPS, for instance in relationto their sector specialization. Local and regional bodiescan include universities, laboratories, science parks,trade unions, trade associations, entrepreneur associa-tions, training and other service centres. Besides these,there could be public institutions such as city, provincialand regional councils, and government offices.Examples from Italian industrial districts and Germanclusters have illustrated how crucial the role of businesssupport organizations can be in addressing situations ofmarket failure that damage firms." In particular, thereare three main functions that business support organiza-tions undertake:'"^

• To provide advice and support to individual firms(e.g. training, fmance, innovation, export).

• To provide collective goods (e.g. training, computer-aided design, testing centres, joint marketing orbranding, joint export).

• To act as brokers to facilitate and promote business-to-business and business-to-institution networking.

As far as the first two points are concerned, it is wellknown that the growth of SMEs is often constrainedby their limited access to finance and managementskills, and that for this reason they might struggle tocarry out lu-house innovation projects, trainingschemes or marketing/export strategies. Externalsupport is needed to bypass these problems. Businesssupport organizations need to tailor their support tosatisfy precisely those needs offering either individual or,wherever possible, collective help. In LPSs, collectiveinitiatives can work particularly well because they canbe activated in an environment where there alreadyexists a high degree of inter-firm networking due tosubcontracting linkages.

The brokerage role of business support organizationsis aimed at encouraging or strengthening linkagesbetween firms and between firms and organizations.'"'This is particularly important where the competitivenessof the LPS is hindered by "rusty' or malfunctioninginter-firm relationships caused by changes in the com-petitive scenario, sector decline, internal restructuring,the disrupting entry of new competitors, etc. All thesefactors can temporarily alter the equilibrium of thenetwork and stall the fluidity of the dynamics of inter-firm linkages. Tbe system can take time and eflort toadjust to the new context. In the best-case scenario, itwill redefine its functioning possibly with a differentset of firms. This adjustment process can be eased byan external intervention that can recreate or facilitatethe re-organization of the production network. In theworst-case scenario, where the system is incapable ofaltering its structure to accommodate the new contextand where there is no external support to help, thesystem might never recover from the exogenous shockand decline until it disappears.

Needless to say, some LPSs do not have a set ofbusiness support organizations on which to rely. Thelack of an institutional framework leaves firms on theirown to solve both individual and collective problems,exposing the entire system to decline. Both inter-firmnetworking and business support organizations allowfirms to overcome their internal constraints so that theycan improve their innovative capability, grow and becompetitive.

The assessment of the institutional framework neces-sarily involves a qualitative analysis based on primaryinformation gathered through questionnaires from bothfirms and business support organizations, as well as onsecondary information. The objective of this assessmentis twotold: (1) to list and map all organizations andinstitutions of the region or locality that have contactswith firms of the LPS; and (2) to appraise the role andinvolvement of such organizations and institutions inthe production system's life. For the first objective,secondary information can be used together with directquestions to firms. The outcome should be a map ofthe various organizations and institutions that specifiestheir objectives, activities and types of firms they aim


CO target. For the second objective, questionnaires haveto be designed to address specific issues directed at firmsand organizations. This enables an evaluation of businesssupport organizations' initiatives for firms and localities.

To conclude, analysis of the institutional frameworkis the hnal stage of the diagnostic analysis of an LPS. Itshould not be surprising that the rise and growth of anLPS often tends to be related to the effectiveness of theinstitutional setting, which can intervene in situations ofmarket failure (i.e. fmance and innovation) and triggervirtuous circles of cooperation, competition, innovationand competitiveness. The study of effective institutionalframeworks can also be used as a reference point toenhance and strengthen the institutional capability oflocalities and/or systems experiencing challenges andchanges.

LOCAL PRODUCTION SYSTEMS INTHE UK

To illustrate how such an approach can be used, theapplication of the spatial diagnostic analysis to the UKis now presented. The data set constructed for mappingUK LPSs consists of 1997 industry employment datasupplied by the Office for National Statistics, originallybroken down into five size classes: 1-24, 25-49,5()-99, 100-299 and more than or equal to 300employees. Such data are aggregated in three main

bands: small firms with fewer than 100 employees,SMEs with fewer than 3(K) employees and large firms(LFs) with more than or equal to 300 employees. Thereare 23 manufacturing sectors (two-digit level)'"' and 297TTWAs from Andover to Wick.

The application of the four criteria of the spatial diag-nostic procedure identified eight categories of LPSsaccording to their manufacturing intensity, degree ofspecialization and firm size. First, LPSs were distin-guished according to whether the economy of theTTWAwas dominated by manufacturing sectors or not{in line with criterion 1). Second, LPSs' sector special-izations were identified and such specializations wereranked according to the location quotient of criterion3. In so doing, we distinguished between primary andsecondary specializations. The former corresponds tothe sector specialization with the highest location quo-tient, while the latter corresponds to the second or thirdranked specializations. Finally, LPS were distinguishedbetween those dominated by SMEs and those domi-nated by large firms (according to criteria 2 and 4).Based on these divides, a matrix of types of LPS in theUK was constructed (Table 2 and Fig. 2).

The eight categories of LPSs were identified asfollows: (1) pre-district; (2) non-specialized pre-district;(3) non-manufacturing and specialized LPS of SMEs;(4) non-manufacturing and non-specialized LPS ofSMEs; (5) manufacturing and specialized LPS of LFs;

Jiihlc 2. Types of ]oai\ production systems (LPSs)

Pre-district

Non-specializedprc-district

Non-tnanufaf-Curiiig andspecialized

LPS ofSMEs

Non-maiiufac-CuHng and

non-specialized

LPS ofSMEs

M;inufac-turing andspecialized

LPS oflarge firms

Manufac-turing and

iion-specialized

LPS oflarue firms

Non-manufac-turing andspecialized

LPS oflarije firms

Non-man utac-turiiig and

non-specialized

LPS oflaree firms Other

(1)Manufacturing specialization

(2)Size <!00Size <3(X)Size >300

(3)Specialization

(4)Specialization plus size <100Specialization plus size <300

Specialization plus size > 300

primary secondary primary secondary primary secondary primary- secondarysector sector sector sector sector sector sector sector

primary secondary primary secondarysector sector sector sector

y / / /primary secondary primary secondarysector sector sector sector

Number of travel-rn-work areas 47 31 73

Source: Liata from the Office of National Statistics (ONS), author's elaboration.Note: SMEs, small- to medium-sized firms.

206 Lisa De Propris

Manu^cturing LPSs Non-maniifecturing LPSs

Specialized LPSs

Non-speciailzed LPSs

NS Pre-district of SMEs

Fig. 2. Types of heal production systems (LPSs) in the UKS-NM, specialized non-inanufacturing; S-M, specialized manufacturing; NS, non-speciaiized: NS-NM, non-specialized non-manufacturing;NS-M, non-spccializcd manufacturing; LF. large firm; SMEs. small- and medium-sized enterprises {i.e. firms). Source: Data are from che Officeof National Statistics, author's elaboration

(6) manufacniring and non-specialized LPS of LFs;(7) non-manufacturing and specialized LPS of LFs; and(8) non-manufacturing and non-specialized LPS ofLFs.'^

A few points are worth mentioning. First, with thismethodology, we were able to identify diverse types ofLPSs in the UK directly from the statistical data beforeany qualitative assessment was made. This is a veryuseful step to provide a broad {e.g. country-wide)picture of the LPSs' phenomenon and, at the sametime, single out diversity across such systems by meansof a quantitative analysis. This is in contrast witb theDTI (2001) report where only clusters were identifiedand differences across them were purely based onqualitative information. Second, 47 industrial districtswere identified that strictly satisfy criteria 1-4. Thisfinding contrasts with the statement by ZEITLIN (1995)that there are no industrial districts in the UK. Third,more than half the LPSs were characterized by SMEs:170 LPSs were dominated by SMEs against 111 beingdominated by large firms. The geographical distributionof LPSs in the UK is shown in Figs 3 and 4. LPSs ofLFs are distributed across England, with a concentration

in the Midlands and the North East, and in SouthWales and the Southern Upperlands of Scotland. Onthe other hand, LPSs of SMEs are scattered acrossEngland and cover most of Wales and Scotland. Inparticular, industrial districts are scattered across theentire country with a significant presence in the Mid-lands and the North of England. Manufacturing LPSsare concentrated in Scodand, Wales and the Southof England. Fourth, there seems to be more non-manufacturing LPSs (185 were counted) than manu-facturing LPSs (156). This finding is consistent withthe rapid and extensive expansion of the service sectoracross the UK. Finally, those LPSs that were strictlyidentified according to criteria 1—4 accounted for 21%of the total UK liianufactiiring employment. This resultis not insignificant given that the rigid application ofthe criteria identifies only the main localized industryfor each TTWA. For instance, for the CoventryTTWA, only its main LPS in the motor vehiclessector were considered as satisfying criteria 1—4 and itaccounted for 22% of the sector national employment.However, in Coventry there are also two other sectorsfor which the location quotient was greater than tbe


Fig. 3. Local producfioii systems of large firms in the UK hy trauel-to-u>ork area. For ahbreuiations, see Fig. 2Soiirtr. Figs 3 and 4 are the author's elaboration with data from the Office of National Statistics (ONS). © ONS, 1998; map produced fay ONS,2003

208 Lisa De Propris

NM-NS-LPS

NM-5-LPS

N5-ID

ID

Fig. 4. Local production systems of small and medium-sizefrms in the UK by travcl-to-work area. For abbreviations, sec Fig. 2


national average: the fabricated metal product sector,and machinery and equipment. These account for aconsiderable share of national employnienc: the metalprodLict sector accounted for 13% of the sector nationalemployment, and the machinery and equipment for12%. Therefore, should one consider such secondaryLPSs for each TTWA, it would certainly be found thattheir total employment share of the U K manufacturingemployment is ahnost 50%.

An evaluation of these initial fmdings already revealshow a better Linderstanding of the location, type and(if one were also to apply the second and third steps ofthe methodology) kinctioningot UK LPSs is extremelyimportant for regional industrial development catalysingaround localized industries and, associated with this,the formulation of appropriate industrial developmentpolicies.

CONCLUSIONS

The paper has presented a methodology to map andanalyse LPSs in regions and countries where there islittle or no knowledge about whether LPSs exist, wherethese are and what form they take. The diagnosticniethodology presents three stages: the spatial diagnos-tic, the diagnostic of inter-firm relationships and theassessment of the institutional framework, hi particular,the spatial diagnostic procedure is aimed at equippingus with an autoniatic procedure that relies on statisticaldata and that can provide basic but fundamentalinformation about LPSs. The three-tiered method-ology involves both quantitative and qualitative analyses,and it proceeds from the big to the small, from thegeneral to tbe specific. In fact, spatial diagnostic analysisreveals a map and a classification of LPSs; after this firststage, tbe further two stages apply to in-depth casestudies on specific production systems.

The present paper fills a gap in the debate on localizedindustries and firm agglomeration in [hat it suggests amethodology to identify, map, classify and analyse in-depth LPSs. The methodology' has the potential to flagup any agglomeration of firms: inflint systems, successfulsystems, systems in decline. This means that it enablesresearch to explore the factors driving the rise, develop-ment and success of production systems, as well as thecauses of their decline and possibly disappearance.

The second half of the paper presented an applicationof the spatial diagnostic analysis to the UK (excludingNorthern Ireland). It identified eight types of LPS,inchiding industrial districts, according to sector special-ization, firm size and manufacturing intensity. Thethree-step methodology has also been applied to mapLPSs in Central and Eastern Europe. It has provided avery useful insight into the presence of LPSs in Lithu-ania, Bulgaria, Romania and Bosnia, where they canbe sparks for regional industrial development {D EPROPRIS and PITELIS, 2000).

To conclude, this methodology discusses the desir-ability of criteria and measures to study LPSs; theauthor feels that the debate on the methodology tomap LPSs shoLild drive rather than be constrained bythe availability of statistical data. In other words, cheacademic debate should challenge the 'goodness' of thepresent data provision, should this be insufficient toprovide a clear picture of a particular phenomenon. Forinstance, the absence of regional or sub-regional input-output tables is a serious limit to the study of LPSs inthe UK.

Acknowledgements - The author thanks four anonymousreferees aTid Marco BcUandi, Christos Pitelis and DavidBailey for useflil comments and suggestions. The usualdisclaimers apply.

NOTES

1. For a very interesting discussion on the conceptualdifference between typology and taxonomy, seeGRANDORI (1990). A typology is defined as an ex-antcclassification that identifies analytical categories to predictvariables not included in it. By contrast, a taxonomy isa classification generated ex-post from empiricalobservation.

2. The main data sources for the DTI cluster analysis are;the Inter-departmental Business Register, the Dun andBnidstreet database, the National Online ManpowerInformation System, ;ind local aucbority- data.

3. The niain contributions in Italian industrial districtsinclude: LEONARDI and NANETTI (1994) on Tuscany;

BRUSCO (1982) and LAZERSON (1990) on Emilia

Romagna, and RABI;1.I.OTTI (1997) on the footweardistrict in che Marche and Veneto.

4. Thf Eumpean Union considers small firms as havingfewer than 50 employees and SMEs as having fewerthan 250.

5. Note that SIC-s are a device designed to a.ssist the otficeof Customs and Excise and the collection of statisticaldata. However, they do not always provide a realisticgrouping of production activities. For this reason, oneneeds to consider sector classifications with cautionespecially if used to map industrial agglomerations. Herethe degree of specialization of production activities makesit very difficult to match them with standard SIC codes.

6. Most studies have looked at LPSs in the manufacturingsectors. However, there has been some work on clusteKin the service sector, such as the entertainment industryin Hollywood, CA (PORTER, 1998), and the financialservice cluster in Hong Kong (ENBIGHT, 2000).

7. On the case of tbe 'three Italies', see KlNc; (1985) andSFOBZI (1999).

8. According to MARSHALL (1920), agglomeration econo-mies are related to firms' external economies of scaleand a rise from information spillovers, local non-tradedinputs and a pool of local skilled labour ( M C C A N N .2001). In the case of LPSs, external economies areassociated with the organization of production and firms'specialization. Production fragmentation and specializa-tion enable firnis to reap economies that are not internal

210 Lisa De Propris

to the firm (although producing increasing return toscale) but external to the firm and internal to theproduction system of which the firm is a part. Seminalcontributions on agglomeration economies includeMARSHALL (1920), PER RO U X (1950) and, morerecently. PORTER (1998) and STCJRPER (1996). Forempirical work, see BRESNAHAN ct ai (2001), R O S E N -THAL and STRANGE (2001) and PARR (2002).

9. For a case study on the link between productionnetworking and cooperation over innovation, see D EPROPRIS (2000).

10. There is a fundamental difference between inter-firmrelationships in the industrial district model and the post-Fordist model, i.e. the structure of governance mirroredby input output exchanges. In post-Fordist systems,where the main buyer outsources its non-core activities,governance tends to be characterized by external verticalcontrol. In contrast, in industrial districts, subcontractingtends to be associated with a more dispersed and hori-zontal system of governance. For fijrther information,seeBECATTiNi (1994).

11. For a description of the institutional frameworks in Baden

Wuttemberg and Emilia Romagna, see COOKE andMORGAN (1994).

12. This analysis draws on D E PROPRIS and PITELIS (2000).13. Oil Real Service Centres in Emilia Romagna, see L E O N -

ARDi and NANETTI (1990).14. The 23 manufacturing sectors are as follows; Food

products and beverages; Basic metals; Tobacco products;Fabricated metal products; Textiles; Machinery andequipment; Apparel dressing/dying fur; Officemachinery and computers; Tanning/dressing of leather;Electrical machinery/apparatus; Wood/products/cork;Radio, television, communications equipment; Pulp,paper and paper products; Medical precision instruments;Publishing, printing, recorded media; Motor vehicles,trailers; Coke, refined petroleum products; Othertransport equipment; Chemicals and chemical products;Furniture: Rubber and plastic goods; Recycling; andOther non-metallic products.

15. Six T T W A cannot be classified because criterion 4is not respected: Barnsley, Bolton, Burnley, Dundee,Havi'ick, and Sheffield and Rotherhani.

REFERENCES

BECATTINI G. (1987) Mcrcam e Forze Locali: II DimeUo liiditslriale. II Mulino, Bologna.BECATTINI G. (1994) The development oflight industry in Tuscany: an interpretation, in LtoNAKDl R. and NANKTTI R . Y.

(Eds) Regional Devehypment in a Modern European Economy, pp. 69-85. Pinter, London.BECATTINI G. (2000) H Bmco e la Ftirfalla: Prato ml Mondo cite Canihia. Le Monnier, Florence.BECATTINI G. (2001) [I Calcdoscopio dello Sviluppo Locale: Trastoriua.zioi)i Eamomichc m-IVltalia che Gambia. Rosenberg &

Sellier, Turin.BELLANDI M . (2001) Local development and embedded large firms. Entreprencurship and Regional Development 13, 189-211.BELLANDI M . and SFORZI E (2001) La molteplicita' dei sentieri di sviluppo locale, in BECATTINI G. et al. (Eds) // calcdoscopio

dello sviluppo locale, pp. 4t—f>6. Rosenberg &: Sellier, Turin.BELUSSI F and ARCANGELI R (1998) A typology of networks: flexible and evolutionary tirnis. Research Policy 27. 415^28.BjORc; A. and ISAKSHN A. (1997) Location, a^lomeration and innovation; towards regional innovation systems in Norway,

European Planning Studies 5, 299-331.URESNAHAN T , GAMBARDELLA A. and SAXENIAN A. (2001) Old economy inputs for new economy outcomes: cluster

formation in the New Silicon Valley, Industrial and Corporate C/JIJ/JSJC 10, 853—860.BRUSCO S. (1982) The Emihan model: productive decentralisation and social interaction, Cambridge Journal of' Ecouomia 6,

167-184.BRUSCO S. and PABA S. (1997) Per una stoHa dei distretti italiani dal secondo dopoguerra agli anni novanta. in BARCA F (Ed.)

Storia del capitalismo ilaliaiio, pp. 265-333. Donzelli, Rome.CAMAGNI R . P. (1995) The concept of innovative milieu. Its relevance for public policies in European lagging regions. Papers

in Regional Scien^ê 74, 317-340.CHENERY H . B. and WATANABE T (1958) International comparison of the structure of production, Econometrica 26, 487-521.COOKE P. (2002) Knowledge Economy: Clusters, Learning and Cooperative Advantage. Routledge. London.COOKE P. and MORGAN K. (1994) Growth regions under duress: renewal strategies in Baden Wiitteniberg and Emilia Romagna,

in AMTN A. and THRIFT N . (Eds) Globalisation, Institutions and Regional Development in Europe, pp. 91-117. Oxford UniversityPress, Oxford.

COOKE P. and MORGAN K. (1998) Tlte Associational Economy. Oxford University Press, Oxford.CosspNTiNo F, PYKK F and SEGENBERGER W (1996) Local and Regional Responses to Glohal Pressure: The Case of Italy and its

Industrial Districts. International Institute for Labour Studies, Geneva.C R O U C H C . and FARRELL H . (2001) Great Britain; falling through the holes in the network concept, in C R O U C H C , LE

GALfes P., TRIGILIA C . and VOELZKOW H . (Eds) Local Production Systems in Europe. Rise or Demise? Oxford UniversityPress, Oxford.

DEPARTMENT OF TRADE AND INDUSTRY (2001) Business Clusters in the UK- A First Assessment. DTI, London.D E PROPRIS L. (2000) Innovation and inter-tirin cooperation: the case of the West Midlands, Economic Innovation and New

Technology 9, 421^46.D E PROPRIS L. (2001) Systemic flexibility, production fragincntation and cluster governance, European Planning Studie.^ 9,

739-753.D E PROPRIS L. (2003) Mapping Local Production Systems iti Britaiu. Working Paper No. 14. Birmingham Business School.D E PROPRIS L. and PITELIS C . (2000) Glusier Diagnostics. Phare AC£ Project Report.


ENRICH r M. J. (2000) Regional clusters and multinational enterprises, Intemaliotial Studies in Mariaêmetit and Organization 30,114-138.

FESEB E. J. and BERGMAN E. M . (2000) National industry cluster templates: a framework for applied regional cluster analysis.Regional Studies 34, 1-19.

GABOFOLI G. (1991) Local networks, innovation and policy in Italian industrial districts, in Br;Rc;MAN E., MAIER G. andToDTLiNG E (Eds) Regions Reconsidered, pp. 119-140. Mansell, London,

GORDON I. R. and M C C A N N P. (2000) Industrial clusters: complexes, agglomeration and/or social networks, Urban Studies37, 513-532.

GRANDORI A. (1990) Epistemologia ed economia aziendale: note per un dibattito, L'indmtria 11, 76-103.Gui'.RRiERi P., IAMMARINO S. and PL[^TROBBI.I.I L. (2001) The Global Change and Industrial Districts. Edward Elgar, Cheltenham.IsARP W.. Azis I. j . , DRBNNAN M . P.. Miii-ER R. E., SALT/^MAN S, and THORBEKE E. (1998) Methods of Interregional and

Regional Analysis. Ashgate, Aldershot.KING R . (1985) Tlie Industrial Geography of Italy Groom Helm, London.LAZERSON M . H . (1990) Subcontracting in the Modena knitwear industry, in PYKE E, BECATTTNI G. and SENGENBERGER

W. (Eds) (1990) Industrial Districts and Inter-jirni Cooperation in Italy. International Institute for Labour Studies, Geneva.LEONARDI R . and NANETTI R . (1990) The Region and European Integration: The Case of Emilia Romagna. Pinter, London.LEONARDI R . and NANETTI R . (1994) Regional Development in a Modern European Economy: Tlie Case of Tuscany. Pinter,

London.LoNGHi C. (1999) Networks, collective learning and technology development in innovative high technology regions: the case

of Sophia Antipolis, Regional Studies 33, 333-342.MARKUSEN A. (1996) Sticky places in slippery space: a typology of industrial districts. Economic Geography 72, 293—313.MARSHALL A. (1920) Principles of Economics, 8th Edn. Macmillan, London.MARTIN R . and SUNLEY P. (2003) Deconstructing clusters: chaotic concept or policy panacea?, Jowmd/ of Economic Geography

3, 5-35.M C C A N N G. (2001) Urhan and Regional Economics. Oxford University Press, Oxford.PANICCIA I. (2002) Industrial Districts. Edward Elgar, Cheltenham.PARR.J . B. (2002) Missing elements in the analysis of agglomeration economies. International Regional Science Reuiew 25, 151-170.PERROUX H (1950) Economic space: theory and application. Quarterly Journal of Economics 64, 89—104.PioKE M. and SABEL C . {19K4) Tlie Second Industrial Divide. Basic, New York.PORTER M . P. (1998) Clusters and the new economics competition, Harrard Busitiess Rei'iews November-December, 77-90.PORTER M . P. (2000) Location, competition and economic development: local clusters in a global economy. Economic Deuelopment

Quarterly 14, 15-24.PYKE H, BECATiiNt G. and SENGENBERGER W. (Eds) (1990) Industrial Districts and Inter-firm Cooperation in Italy. International

Institute for Labour Studies, Geneva.RABELLOTTI R . (1997) External Economies and Cooperation in Industrial Di.trnWi. Macmillan, Basingstoke.RASMUSSEN P. N. (1956) Studies in Inter-sectoral Relations. North Holland, Amsterdam.RoSENTHAi., S. S. and STRANGE W. C . (2001) The determinants of agglomeration, yoMr«ii/ ofUthan Economics 50, 191-220.SFORZI E (1989) The geography of industrial districts in Italy, in GOODMAN E. and BAMFORD J, (Eds) Smalt Eirnis and

Industrial Districts in Italy, pp. 153-173. Routledge, London.SFORZI E (1990) The quantitative importance of Marshallian industrial districts in the Italian economy, in PYKE F . , BECATTINI

G. and SENGENBERGER W. (Eds) Industrial Districts and Inter-firm Cooperation in Italy, pp. 75-107. International Institute forLabour Studies, Geneva.

Sr-ORZi E (1996) Italy: Local Production Systems of Small and Medium-sized Eirms and Industrial Changes. O E C D Networks ofEnterprises and Local Development, Paris.

SFORZI E (1999) Economic change, in BONAVERO P., DEMATTEIS G. and SFORZI E (Eds) Tlte Italian Urban System. A.shgate,Aldershot.

SiMMiE J. and SENNETT J. (1999) Innovative clusters: global and local linkages?. National Institute of Economic Reviews 170, 87-98.SOLINAS G. and BARONI D. (1998) / sislemi locali manufatturieri in Italia: 1991-1996. Mimeo, Dipartimento di economia

politica, Universita' di Modena e Reggio Emilia.STORPER M . (1996) The resurgence of regional economies, ten years later: the region as a nexus of untraded interdependencies,

European Urban and Regional Studies 2. 191-221.ZKITLIN J. (1995) Why are there no industrial districts in the United Kingdom?, in BAGNASCO A. and SABEL C . E (Eds)

Stnall and Medium Enterprises. Pinter, Derby.

http://usj.sagepub.com/Urban Studies

http://usj.sagepub.com/content/early/2012/12/20/0042098012466601The online version of this article can be found at:

DOI: 10.1177/0042098012466601

published online 20 December 2012Urban StudAndrew Crawley, Malcolm Beynon and Max Munday

AnalysisMaking Location Quotients More Relevant as a Policy Aid in Regional Spatial

Published by:

http://www.sagepublications.com

On behalf of:

Urban Studies Journal Foundation

can be found at:Urban StudiesAdditional services and information for

http://usj.sagepub.com/cgi/alertsEmail Alerts:

http://usj.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

What is This?

- Dec 20, 2012OnlineFirst Version of Record >>

at UNIV OF TEXAS ARLINGTON on April 7, 2013usj.sagepub.comDownloaded from

http://usj.sagepub.com/

http://usj.sagepub.com/content/early/2012/12/20/0042098012466601

http://www.sagepublications.com

http://www.urbanstudiesfoundation.org/

http://usj.sagepub.com/cgi/alerts

http://usj.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://usj.sagepub.com/content/early/2012/12/20/0042098012466601.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml


Making Location Quotients More Relevantas a Policy Aid in Regional Spatial Analysis

Andrew Crawley, Malcolm Beynon and Max Munday

[Paper first received, October 2011; in final form, June 2012]

Abstract

Location Quotients (LQs) remain an important tool for geographical analysis, par-ticularly in terms of assessing industrial specialisation and clustering. LQs as decisionaids are typically understood through the use of arbitrarily set cut-off values.However, LQs are rarely accompanied by an associated level of variance that can beconnected with the estimated data used to calculate them. This paper reveals theimportance of understanding this variance and shows how confidence intervals canbe estimated for employment-based LQs. A systematic process is introduced,through which the arbitrariness of cut-off-value choice can be mitigated and border-line industry cases in terms of their LQ values and the considered cut-off value,identified. A case from a UK region is used to illustrate the issues covered in thepaper.

1. Introduction

Since its conception with the advent of theeconomic base model (Haig, 1926), the loca-tion quotient (LQ) has provided one meansof assessing the relative specialisation of aparticular characteristic within a popula-tion.1 Its popularity has not diminished andits use is still prevalent (see recent work byBishop et al., 2003; Tonts and Taylor, 2010).

The LQ has been used in different typesof analysis to accomplish numerous analy-tical tasks. One factor encouraging the ana-lytical use of LQs, according to Isserman

(1977), is that relatively few data arerequired for their computation and thismay also have encouraged its use as a deci-sion aid. Indeed, the notion of the LQ hasbeen consistently used in geographical anal-ysis since the 1940s (Gibson et al., 1991).Its application has increased recently withits adoption in a number of studies investi-gating industry specialisation and cluster-ing. For example, Porter (2000) and theUK cluster mapping study (DTI, 2001),used LQs to identify clusters of industries

Andrew Crawley, Malcolm Beynon and Max Munday are in the Business School, CardiffUniversity, Colum Drive, Cardiff, CF10 3EU, UK. E-mail: [email protected],[email protected] and [email protected]

1–16, 2012

0042-0980 Print/1360-063X Online� 2012 Urban Studies Journal Limited

DOI: 10.1177/0042098012466601 at UNIV OF TEXAS ARLINGTON on April 7, 2013usj.sagepub.comDownloaded from


at sub-national scale (see also for recentstudies, Boix and Galletto, 2009; Feser andIsserman, 2009; Chiang, 2009; Carroll et al.,2008). The LQ is also used by the EuropeanCluster Observatory (ECO) programme, apolicy-making group of the EuropeanUnion (Crawley and Pickernell, 2012).

Many researchers are interested in thepractical application of LQs. For example,epidemiologists examine the spatial distri-bution of diseases (see Clayton and Hills,1993), whilst criminologists often want tounderstand differences in reported crimesin different neighbourhoods (see de Francesand Titus, 1993). Climatologists have usedthe technique to measure certain meteoro-logical effects (see Wong and Chen, 2009).The LQ has also been widely used in psy-chology as a way of establishing patterns ofpsychological disorders in populations (seeMetraux et al., 2007).

Returning to the traditional use of LQs inindustry analysis, a common problem is theuse of arbitrary cut-off values to designatesets of industries based on calculated LQvalues. Moreover, while LQ-based analysisleaves the analyst with a set of ‘included’ and‘excluded’ industries there are inevitably‘borderline’ industries with LQs close to theconsidered cut-off value. The systematicprocess developed in this paper elucidatesthe ‘interest’ potential of identified border-line industries, in terms of the uncertainty oftheir membership to either of the two desig-nation sets (inclusion or exclusion). In par-ticular, this paper demonstrates that suchanalysis should take careful account of theunderlying variance in calculated LQs. Toillustrate the points made in our paper, weexamine LQs calculated for Welsh manufac-turing industries.

The structure of the rest of the paper isas follows. The second section reviews howLQs are typically used to aid decision-making within industrial specialisationanalysis, and reviews the usefulness of the

measure. This section focuses on the prob-lems associated with practically using LQswithout taking into account associated var-iances. The third section discusses onemethod that can be used to analyse varianceassociated with LQs. This demonstrateshow employment-based LQs,2 calculatedfor the purpose of industrial specialisationanalysis, might be understood when theconfidence intervals around them are con-sidered. The fourth section introduces asystematic process by which the arbitrari-ness of cut-off value choice can be miti-gated and how the ‘interest’ potential forborderline industries, with LQs near a cut-off value, can be identified. The fifth sectionoffers conclusions and directions for futureresearch.

2. The Usefulness of LocationQuotients: A Review

LQs were originally used as a means of deter-mining the levels of manufacturing versusservice industries within a given spatial area(see Haig, 1926). They have since been usedto quantify and identify numerous spatialphenomena. The method quickly becameone means of determining the presence ofindustrial complexes (Czamanski, 1974). Itsuse to describe industrial specialisationwithin national and regional economies hasincreased in the past 20 years following theacademic work by, among others, Porter(1990, 1999, 2000), Bergman and Feser(2000) and Carroll et al. (2008).

Gibson et al. (1991) describe the LQ asdemonstrating how strongly an industry isrepresented in a region. Studies have usedthe technique to identify industrial clusterswhich might become the focus of govern-ment support (see DTI, 2001; see alsoWorld Economic Forum, 2007). LQs havefurther applications. For example, the esti-mation of LQs forms one basis for

2 ANDREW CRAWLEY ET AL.



Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

‘regionalising’ national input–output tablesin the absence of extensive regional surveydata (see Isard et al., 1968; Miernyk, 1976;Brand et al., 1998). In non-survey input–output table development, LQs are used toadjust national industry coefficients to takeinto account regional differences in indus-try structure (for a full discussion on this,see Miller and Blair, 2009).

The use of LQs in these applications is notwithout problems. For example, in the case ofdeveloping non-survey tables, there are aseries of assumptions typically made (i.e. withrespect to spatial differences in labour pro-ductivity, cross-hauling, etc.), constrainingthe use of LQs (see Schaffer, 1999). Moreover,in terms of the methodology itself, it has beenargued both for cluster analysis (Malmbergand Maskell, 2002) and input–output tableconstruction (West, 1986), that LQs lackreliability and consistency in estimation.

Linked to these issues, practical analysisusing LQs often means the employment ofarbitrary cut-off values to determine desig-nation sets of industries (see Bishop et al.,2003). The cut-off value is simply a valueagainst which the industry LQs are com-pared, whether an industry LQ value isabove or below it, so forming two differentdesignation sets. For example, in one of thelargest UK studies utilising LQs, undertakenby the DTI (2001), a cut-off value of 1.25was employed. This allowed the inclusionand exclusion of industries for further study.Table 1 summarises some of the othercommon cut-off values used in prior research.

An issue arising from the variation in thepreviously employed cut-off values shownin Table 1 is that the key to understandingwhich value is the most appropriate (as acut-off value) might not be the most impor-tant issue given the applications for whichthe LQ is typically used. The exercise shouldbe to establish whether or not a particularLQ value is noticeably greater than somearbitrarily developed cut-off value. It is also

important to recognise when calculatingLQs with government employment data inindustries, that the data themselves may beestimated.

The accuracy of small-area employmentdata has been called into question beforeand this has been shown to have impact onLQ estimates (see Silcocks, 1994; Thrallet al., 1995; Beyene and Moineddin, 2005).For example, in treating with these problemsO’Donoghue and Gleave (2004) suggest astandardised LQ approach which revealsagglomerations as being made up of locationswith statistically significant (rather than arbi-trarily defined) LQ values for industries beingconsidered. The problem of the LQ methodgenerating single point estimates is alsoexamined by Beyene and Moineddin (2005).Basically, there is a need to consider the mea-surement of statistical error, or the degree ofaccuracy of a particular LQ value, basedupon the overall dataset being employed.

A simple technique that could be appliedis the calculation of a t-test comparing theLQ for a particular industry in a region,with that of other LQs calculated from otherregions. Thrall et al. (1995) adopt a similarapproach when looking at the calculation ofLQs for the number of mortgages given bydifferent financial institutions across theUS. They used a t-test to determine howsimilar an individual LQ for one institutionwas compared with an aggregated LQ deter-mined by looking at all the institutions.This, however, still had the drawback ofassuming that the calculated point estimateswere absolute rather than estimates.

Beyene and Moineddin (2005) choseto utilise a more sophisticated approach,based upon the linear approximation of thevariance, allowing the establishment ofconfidence intervals around LQs. This cal-culation is of benefit as it is constructed totake into account the inherent uncertaintycontained within the estimation procedureinvolved in the initial calculation of the LQs.

MAKING LOCATION QUOTIENTS MORE RELEVANT 3



Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

This paper in the next section applies theBeyene and Moineddin (2005) method toindustrial employment data.3 However, wethen use the resulting LQ confidence inter-vals in a novel way to offer an informedexpectation to the use of arbitrary cut-offvalues, in terms of designation of industries.By analysing the relative numbers of indus-tries designated as included or excluded,based on different cut-off values, this paperprovides a unique graphical-based systematicprocess to allow decision-makers a consistentstatistically based alternative to arbitrary cut-off-value choice.

3. Estimating Confidence Intervalsfor LQs

In this section, and the following, we use anexample to illustrate the intended systematicprocess. The case used is the estimation ofLQs for manufacturing industries in Walesin the UK. To test the effectiveness of themethod, a region needs to be selected thatallows a manageable and yet representativesample of a nation as a whole. Wales hasaround 5 per cent of the UK population andhas a representative number of industriesthat also feature at the national UK level. Toexplore industry specialisation, we use themost disaggregated data possible. The UKOffice of National Statistics holds estimateddata on firms and industry size in terms ofemployment. These data are classified by theStandard Industrial Classification (SIC),4

within which the level of detail is denomi-nated by the number of digits used toexpress the data. The other potential level ofdisaggregation is spatial, with SIC employ-ment data being available at national,regional, travel-to-work area and localauthority levels. When calculating LQs, thespatial domain is important (see McCannand Dewhurst, 1998). Mulligan and Schmidt(2005) also found that LQs are sensitive tochanges in the reference geographical scale.The issue known as the moveable area unitproblem (Openshaw and Taylor, 1979)means that if one uses different spatialdomains it changes the findings dramati-cally. To this effect, for the purposes of thispaper, a large region case is used to give afair representation of what the methodcan allow. In summary, to illustrate theconfidence interval approach, we use LQscalculated for SIC 5-digit manufacturingindustries within Wales for 2010. For illus-trative purposes, we also initially adopt anarbitrary cut-off value for the LQ of 1.25 tohighlight the issues we are seeking to addresslater. For Wales, there were 218 differentmanufacturing industries (5-digit SIC)where data were available.5

In order to explore the issue of cut-offvalues, this work uses the LQ formulationgiven by de Propris (2005), for measuringmanufacturing specialisations (for industry i)6

LQi =xi

ni

�x

n

Table 1. Arbitrary cut-off values used previously in economic studies

LQ cut-off value References

3 Malmberg and Maskell (2002)1 Held (1996); Bishop et al. (2003); Tonts and Taylor (2010)1.25 Bergman and Feser (2000)2 Solvell et al. (2003)1.25 to 5 Kumral and Deger (2006)




where, xi and ni represent the levels ofmanufacturing employment in industry i atthe regional and national levels (in our caseWales and the UK); and x and n are thetotal levels of manufacturing employmentat the regional and national levels. In whatfollows, the word regional is taken to repre-sent some geographical scale below thenational. It follows, for industry i, that ifthe LQi is greater than one, then the regionhas a greater share in that industry than thenational average and thus it could beinferred that the regional area is more spe-cialised in the given industry.

To estimate the variance associated withan LQi value, the Delta method is adopted(see Oehlert, 1992), which utilises the Taylorseries expansion. The Taylor series is therepresentation of any given function by aninfinite sum of terms. These are calculatedfrom the values of its derivatives at a singlepoint (see Thomas and Finney, 1996). Forfull details on the utilisation of this methodwith respect to LQs and their associated var-iances, see Moineddin et al. (2003) andBeyene and Moineddin (2005). In summary,given the two expressions ri = xi

ni, and r = x

n(so LQi = ri

r), denoting respectively the

regional share of national employment inmanufacturing industry i, and then theregional share of total national manufactur-ing employment, it can be shown the confi-dence interval limits at the 100 1� að Þ percent confidence level, with associated za=2

z-score, are given by

ri

r6za=2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ri(1� ri)

nir2+

r2i (1� ri)

nr3� 2r2

i (1� ri)

nr3

s

It follows that the relative size of industryemployment and the regional area in whichit operates, both have a role in determiningthe relative variance associated with the

estimated LQ. For example, if the industryemployment is relatively large and theregional area is small, the LQ will appearlarge and the relative variance will appearsmall. Conversely, if the industry employ-ment is relatively small and the regionalarea employment is larger, then the LQ willbe small and the associated variance will bemuch higher.

With the use of regions it is possible tomake the assumption that the referencearea is partitioned into k non-overlappingregions. It is assumed that the distributionof xi is binomial with the parameters oftotal population ni, and the true incidencerate ri. Following Moineddin et al. (2003)and Beyene and Moineddin (2005), we canmake the assumption that x/n is fixed notrandom (the incidence rate is the same inall areas for an industry, ignoring any spa-tial variation). It is acknowledged that thisis a restrictive assumption and means that agreater component of the measure is saidto be non-random. This approximation isused as a substitute for individual incidencerates as, with large sample sizes, the relativechange in the true incidence rate is negligi-ble (Moineddin et al., 2003).

Assumptions can then be made regardingthe covariance if spatial autocorrelation isconsidered negligible—that is, if the pres-ence of an industry in one regional area hasno impact on whether that industry isrepresented in areas close by. With thesedata, the lack of continuity—that is tosay, not all regions have a presence fromevery industry—means that autocorrelationbecomes less of an issue. Due to the largenumber of industries and the resulting obser-vations for Wales used in this study, it can beassumed that the statistic has a Gaussian dis-tribution, subject to a z-score (here consid-ered over a range of the 100 1� að Þ per centconfidence interval). The size of the confi-dence interval then depends upon just thethree variables, ri, r and n. Hence the earlier




assertion that the magnitude of the variancevaries with the levels of spatial and industrialdisaggregation.

In Table 2, we present the constituentLQ values and 95 per cent (a = 0.05 in100 1� að Þ per cent) confidence intervalsfor a sample of the manufacturing indus-tries. These were not randomly chosen fromthe 218 manufacturing Welsh industriesconsidered. Instead, they were identifiedsubject to a chosen LQ cut-off value (andconfidence level). As noted earlier for illus-trative purposes, an LQ cut-off value of 1.25is considered, this being a common value inprevious studies (see Table 1).7 The choiceof 1.25 is without loss of generality to theconsideration of any LQ cut-off value.These industries are then of interest sinceeach of their respective 95 per cent confi-dence intervals surrounds the considered1.25 LQ cut-off value (whereas for the other203 industries at the 95 per cent confidenceinterval, their confidence intervals are eitherfully below or above the 1.25 cut-off value).In this situation, due to the consideration ofthe variances of industry LQs, there is thepossibility that, if the point LQ is above orbelow the considered LQ cut-off value, theactual LQ could be below or above thisrespectively. Even within the reportedsample of industries shown in Table 2, thevariation in the LQ point values and associ-ated variances is demonstrated. The subse-quent impact on the constructed confidenceintervals is also displayed in Table 2.

Rather than only considering the usualsingle confidence interval for a specific100 1� að Þ per cent confidence level(see Moineddin et al., 2003; Beyene andMoineddin, 2005) as shown in Table 2, theanalysis here is extended in Figure 1 toshow the relative change in the confidenceintervals as the respective 100 1� að Þ percent confidence level changes from 0 percent (point values) to 95 per cent (often

employed), for the 15 industries high-lighted in Table 2.

In Figure 1, in the main part of the graph,each defined industry, from Table 2, isrespresented by a two-dimensional ‘trum-pet’ shape. At the back of the graph, on theconfidence level axis labelled ‘Point(0%)and 95 % Confidence Interval’ is the pointLQ value and 95 per cent confidence intervalrepresentation (see Moineddin et al., 2003;Beyene and Moineddin, 2005), for all 218industries including the 15 industries whosevalues are reported in Table 2. The trumpetshape shows the change in the associatedconfidence interval of the industry LQ valueas the 100 1� að Þ per cent confidence levelchanges. Clearly, as the 100 1� að Þ per centconfidence level increases, so the confidenceinterval range increases with the limits rep-resented by the respective top and bottompoints along the trumpet (at the relativeconfidence level). The important point tonote here is the variability of the confidenceinterval ranges across the different industriesconsidered, a consequence of the differentvariances associated with each LQ value.Figure A1 in the Appendix summarily showsthe trumpet shapes associated with all 218manufacturing industries.

The results in Figure 1 show the indus-tries, in terms of their LQ ‘trumpet’ repre-sentations, which, subject to a 95 per centconfidence interval, could be considered inone of two following incidences (the dashedline shows the 1.25 cut-off value across thewhole figure)

(1) Their LQ value is above the 1.25 cut-offvalue, but their 95 per cent confidenceinterval includes values below 1.25 (socould actually be below it).

(2) Their LQ value is below the 1.25 cut-offvalue, but their 95 per cent confidenceinterval includes values above 1.25 (socould actually be above it).




Tab

le2.

LQco

nfi

den

cein

terv

als

for

15

Wel

shm

anufa

cturi

ng

indust

ries

(ori

ginal

poin

tLQ

,va

rian

cean

d95

per

centco

nfi

den

cein

terv

al)

SIC

no.

Man

ufa

ctu

rin

gL

QV

ar(s

2)

95pe

rce

nt

con

fid

ence

inte

rval

(LQ

61.

960s

)

1091

0P

rep

ared

feed

sfo

rfa

rman

imal

s1.

352

0.00

3116

(1.2

431,

1.46

19)

1610

0Sa

wm

illi

ng

and

pla

nin

go

fw

oo

d1.

305

0.00

2964

(1.1

981,

1.41

15)

1629

0O

ther

pro

du

cts

of

wo

od

1.19

70.

0032

34(1

.085

9,1.

3088

)17

211

Co

rru

gate

dp

aper

and

pap

erb

oar

d;

man

ufa

ctu

reo

fsa

cks

and

bag

so

fp

aper

1.15

50.

0065

18(0

.996

7,1.

3131

)18

202

Rep

rod

uct

ion

of

vid

eore

cord

ing

1.05

50.

0174

45(0

.796

4,1.

3141

)22

230

Bu

ild

ers-

war

eo

fp

last

ic1.

232

0.00

0494

(1.1

886,

1.27

58)

2314

0G

lass

fib

res

1.31

70.

0075

99(1

.146

4,1.

4881

)23

490

Oth

erce

ram

icp

rod

uct

s0.

918

0.05

7111

(0.4

500,

1.38

67)

2452

0C

asti

ng

of

stee

l1.

206

0.00

6838

(1.0

442,

1.36

84)

2521

0C

entr

alh

eati

ng

rad

iato

rsan

db

oil

ers

1.15

50.

0033

95(1

.040

4,1.

2688

)25

300

Stea

mge

ner

ato

rs,

exce

pt

cen

tral

hea

tin

gh

ot

wat

erb

oil

ers

1.27

10.

0127

00(1

.050

3,1.

4921

)27

120

Ele

ctri

city

dis

trib

uti

on

and

con

tro

lap

par

atu

s1.

195

0.00

0992

(1.1

331,

1.25

66)

2823

0O

ffic

em

ach

iner

yan

deq

uip

men

t(e

xcep

tco

mp

ute

rsan

dp

erip

her

aleq

uip

men

t)1.

234

0.00

3462

(1.1

188,

1.34

95)

2894

0M

ach

iner

yfo

rte

xtil

e,ap

par

elan

dle

ath

erp

rod

uct

ion

1.22

70.

0137

32(0

.997

6,1.

4569

)30

920

Bic

ycle

san

din

vali

dca

rria

ges

1.28

10.

0107

78(1

.077

1,1.

4841

)




A count of these industries shows 15 (6.9per cent) out of the 218 considered indus-tries fall into one of these two incidences.Two example industries are highlighted inFigure 1 with their trumpet and SIC codesin bolder lines, for illustration purposes,Other Ceramic Products is SIC 23490 (andso labelled in Figure 1) and ElectricityDistribution and Control Apparatus (here-after Electricity Distribution etc.) is SIC27120. These two industries illustrate thevariation in the levels of variance associatedwith their respective LQ values. OtherCeramic Products has a large variance (widetrumpet) as the confidence level 100 1� að Þper cent increases, whereas for ElectricityDistribution etc. there is a smaller variance(thin trumpet). These results illustrate theearlier assumption that employment sizeand variance are interconnected. OtherCeramic Products employ just 14 employeeswhereas Electricity Distribution etc. has rel-atively large employment (1328) comparedwith the UK average (470). Inspection ofthese two industries in Figure 1 shows thatthey have LQ values below the cut-off value

1.25 and, as the confidence level increases,they each subsequently have confidenceintervals inclusive of this cut-off value(both fall into incidence (2) describedpreviously).

The implication is that industries whosederived LQ confidence intervals surroundthe chosen cut-off value might be flaggedup for further consideration, such thatthere may be contradictions in their initialdesignations.

4. Implications

The method described in the previous sec-tion shows a way of determining the var-iances associated with the calculation ofLQs. A consequence of the consideration ofvariances is the potential uncertainty in theindustries’ designations based on a chosenLQ cut-off value. As demonstrated in Table1, numerous different reports and academicpapers have chosen to adopt arbitrary cut-off values, with little statistical rationale.Understanding this uncertainty issue nextallows us to consider whether this method,

Figure 1. LQ confidence intervals for manufacturing industries in Wales (with varying confi-dence levels of 100(1� a) per cent), for industries whose LQ confidence interval at the 95 percent confidence level is inclusive of 1.25 cut-off value.




using the associated LQ variance values,can play a role in improving one of thelongstanding issues of the utilisation ofLQs—namely, the arbitrariness of thechoice of cut-off values used (O’Donoghueand Gleave, 2004).

This paper proposes a systematic processwhereby utilising the confidence intervalmethod one is able to develop a statisticallybased interpretive rationale to the inclusionor exclusion of industries for further analy-sis. The process is generated on the principlethat statistical variance should be consideredwhen choosing industries for inclusion, notsimply how close they are to an arbitrarycut-off value based on their LQ values.Essentially, this process allows the assess-ment of the potential of borderline indus-tries to be included or excluded from furtheranalysis. Figure 1 illustrated that the rangesof the confidence intervals associated withthe industry LQs varied markedly (over thesame 100 1� að Þ per cent confidence level).It follows that, for those industries whoseLQ values are close to the considered cut-offvalue, the associated confidence intervals,beyond some 100 1� að Þ per cent confi-dence level value, may include the cut-offvalue. This means that parts of the associ-ated confidence intervals are on the oppositeside of the cut-off value from their respec-tive point LQs—the implication being thatan industry’s LQ confidence interval sug-gests a possible contradiction in what theLQ value implies, relative to the consideredcut-off value.

Figure 2 further illustrates the issue,examining the industries which have apotential for contradiction in their initialdesignation based on their LQ confidenceinterval (now considered up to the 99 percent confidence interval).8 In the industrypotential graph reported in Figure 2, eachpoint represents an industry, which has apoint LQ value below or above the cut-offvalue 1.25, but which has a concomitant

confidence interval, beyond a value of100 1� að Þ per cent, which goes above orbelow the considered cut-off value respec-tively. The implication, following on fromthis, is that the industry could have beeninaccurately included or excluded from fur-ther analysis (since there is a part of theconfidence interval suggesting the oppositeof what the industry’s LQ value was imply-ing). The vertical axis in the industrypotential graph is the LQ scale and, foreach industry point shown, their valueagainst the vertical axis is their point LQvalue. Along the horizontal axis, the level ofpotential is shown (least 100 1� að Þ percent confidence level), here from 0 per centto 99 per cent, signifying the confidencelevel at which the considered cut-off valueis included in an industry’s LQ confidenceinterval. Here, the closer to 99 per cent theless potential (left to right in decreasingpotential). For example, Other CeramicProducts (SIC code 23490) has a LQ valueof 0.917 (on the vertical axis) and a confi-dence level potential value of 91.8 per cent(on the horizontal axis) indicating it wouldrequire the consideration of the LQ’s confi-dence interval at the 91.8 per cent confi-dence level or above for it to suggestpotential for contradiction of its initialdesignation.

There are two points to note from Figure 2.First, industries occurring above or below theline need to be considered separately: ‘aboveline’—these industries have the potential tobe excluded even though they were originallyincluded based on the considered LQ cut-offvalue; ‘below line’—because these are indus-tries which have the potential to be includedeven though they were originally excludedbased on the considered LQ cut-off value.Secondly, the scattered nature of the points,either side of the cut-off line, is due to thevariations in confidence interval ranges asshown in Figure 1 (different-sized trumpets).This is a consequence of the different levels




of LQ variance found amongst the consid-ered industries (see, for example, Table 2).

In this analysis, considering the 218Welsh manufacturing industries and up to a99 per cent confidence interval, out of the 50of these whose LQ values are above the cut-off value 1.25, six have the potential to fallbelow based on their associated confidenceintervals (for different levels of 100 1� að Þper cent), and out of the 168 whose LQvalues are below 1.25, 12 have the potentialto go above the cut-off value (examples ofthese industries are shown in Figure 1labelled with their SIC codes). Clearly,these 18 industries might be further consid-ered. The industry potential graph providesdecision-makers with a tool to assess thepotential of industries to change their ini-tial designation. To utilise the graph inFigure 2, decision-makers read, from left toright, the order of the points they encoun-ter, which denotes the decreasing potentialfor the associated industries to be consid-ered, in terms of the possible contradictionof their initial designation.

There is an important point here: theindustry order of potential shown in Figure2 is not simply based on the order of theindustries’ LQ values moving away, eitherdown or up, from the considered cut-offvalue. As well as the nearness of an indus-try’s LQ value to the considered cut-off

value (on the vertical axis), their potentialalso needs to be considered (on the hori-zontal axis). To illustrate this point, takethe two industries highlighted earlier.Other Ceramic Products, labelled 23490,has an LQ value further away from the cut-off value 1.25 than the other, ElectricityDistribution etc., labelled 27120. It maynot be the case that an industry with anLQ value further away from the cut-offvalue should be thought of any less thanan industry with an LQ value closer; it isthe assessment of its associated potentialthat is also a factor (based on 1100 1� að Þper cent confidence level). For these twoindustries, the larger variance associatedwith Other Ceramic Products means thatits 100 1� að Þ per cent confidence-level-based confidence interval goes over thecut-off value 1.25 for a smaller 100 1� að Þper cent confidence level than that of theindustry Electricity Distribution etc. withan LQ value nearer the cut-off value.

The example results and inference onindustries’ potentials for inclusion or exclu-sion from further analysis so far describedhere are referenced on the specific cut-offvalue of 1.25. Moreover, the industry poten-tial graph shown in Figure 2 is specific tothe cut-off value 1.25; it follows that, if a dif-ferent cut-off value were to be considered,then a new associated industry potential

Figure 2. Industry potential graph, with example LQ cut-off value of 1.25.




graph, like that shown in Figure 2, wouldhave to be considered. Figure 3 displays theindustry potential graphs for two further‘theoretical’ cut-off values—namely, 1.75(Held, 1996) and 0.75 (Bergman and Feser,1999).

In Figure 3, the two industry potentialgraphs shown reveal the potential forindustries whose LQ associated variancesimply that they are possibly worthy of fur-ther consideration, based on the cut-offvalues 1.75 and 0.75. The results show dif-ferent numbers of industries, above andbelow the respective cut-off values, fromthose reported in Figure 2.

Instead of choosing an arbitrary orextant-literature-based cut-off value, thisanalysis suggests that consideration of arange of cut-off values, and also the conco-mitant variances of the established industry

LQ values, be taken into account. This is animportant point—namely, that the pre-scribed systematic process will naturallyconsider a continuous range of cut-offvalues rather than the typical single cut-off-value approach. Figure 4 graphically displaysthe relationship over a range of cut-offvalues and the subsequent number of indus-tries included or excluded in designationterms based on different cut-off values inthis range. Using this, decision-makers canmake an initial, more subjective, choice ofcut-off value. This means choosing a cut-offvalue which offsets the overinclusion orunderinclusion or exclusion of industries.

In Figure 4, the two curves represent, fora specific cut-off value shown on the hori-zontal axis, the number of industries whichwould be included or excluded, based onwhether their LQ values are above or below

Figure 3. Industry potential graphs, with example cut-off values of 1.75 (above) and 0.75(below).




the respective cut-off value. With 218industries considered here, the ‘inclusion’and ‘exclusion’ lines monotonically decreaseand increase from 218 and 0 respectively,as the considered cut-off value increases. Itis of note that the range of considered cut-off values is from zero to around 16; theupper bound of near 16 was necessarydue to one industry having an LQ value ofnear 16.

This graphical information can be takenfurther over the same range of cut-off valuesas considered in Figure 4. Figure 5 shows therelationship between a considered cut-offvalue and the number of industries whichcould be incorrectly designated for inclusion

or exclusion, termed here (see Figure 2) ashaving the potential to change (contradict)their initial designation.

In Figure 5, the two graphs report thenumbers of industries whose LQ values areabove (top graph) and below (bottomgraph) the considered cut-off value indi-cated on the horizontal axis, but have thepotential to be on the other side of the cut-off value. When combined with the infor-mation in graphs such as Figures 2 and 3,and the choice of a cut-off value, this gra-phical information can give valuable insightinto industry designation and potential. Itfollows, using the information contained,such as that shown in Figures 2, 3, 4 and 5,

Figure 4. Industry ‘inclusion–exclusion’ graph.

Figure 5. Numbers of potential industries identified in inclusion or exclusion groups,depending on considered cut-off value.




that a systematic process can be constructedelucidating the impact of cut-off-valuechoice and industry designation.

To summarise, the goal of this overallprocess is to provide an understanding ofcut-off-value choice and the resultingimpact on the numbers of industries con-sidered for inclusion and exclusion. Theprocess also highlights which industries areborderline cases in which case their poten-tial may need to be reconsidered, as follows.

First, Figure 4 offers the decision-makertwo possible options to begin the process.They can ascertain the chosen cut-off valueby selecting the number of industries theywish to include or exclude from furtheranalysis. Alternatively, they can initiallyselect a cut-off value and establish howmany industries will be included or excludedfrom further investigation.

Secondly, using the selected cut-off valuefrom Figure 4, Figure 5 offers the ability toestablish how many industries within thetwo designations (inclusion or exclusion)have the potential to have their designationchanged.

Finally, after the consultation of Figures4 and 5, and after the selection of a cut-offvalue, an industry potential graph can beconstructed, similar to those shown inFigures 2 and 3, reporting the possibleorder in which the industries have thepotential to change their designation (basedon inclusion of cut-off value in their associ-ated confidence interval around their LQvalue). As described previously, this order-ing is based on moving left to right acrossthe graph.

In summary, the systematic process out-lined allows a consistent robust techniquefor the consideration and interpretation ofLQ values. By incorporating a diagnosticelement into the interpretation of LQs,some of the problems of arbitrariness asso-ciated with more conventional analyses areavoided.

5. Conclusions

The context of this paper was the continueduse of location quotients (LQs) for decision-making purposes, particularly with respectto the establishment of designation sets inindustry cluster analysis. Of practical con-cern was the use of arbitrary cut-off valuesto establish industries into two designationsets—namely, inclusion and exclusion fromfurther analysis. This issue has been a longstanding concern in the literature (seeO’Donaghue and Gleave, 2004). While thereis a literature that questions the wisdom inselecting key clusters of industries, there areparallel concerns on the methods to accom-plish this selection of industries on which totarget policy resources. This paper suggeststhat where such policies are being adopted,real care needs to be taken in selecting desig-nation sets of industries based on arbitrarycut-off values.

However, the paper aimed to go one stepfurther and establish a test of significance foremployment-based LQs. This test obviatesagainst the use of arbitrary cut-off values forLQ-based industry designation. Moreover,we argue that undertaking this more rigor-ous analysis permits greater precision in thedefinition and identification of spatial spe-cialisations. Simply using the traditionalpoint estimates for LQs provides little indi-cation of the significance attatched to suchestimates. Furthermore, this contributionprovides a systematic process through whichthe analyst can develop a statistically basedrationale for a designation set of industriesadopted within a cluster or key-sectorpolicy. Importantly, the process describedpermits the analyst to consider the potentialfor ‘borderline’ industries to be included orexcluded from designation sets and providesa ready visualisation of the technique forclearer interpretation purposes.

The method derives answers that aresimple in calculation and ease of use, and




yet which demonstrate the variance that isoverlooked when using the standard for-mulation of the LQ. In short, we believethat future studies using LQs need to be alittle more transparent in the limits of theapproach and more ready to acknowledgethe ways in which uncertainty around esti-mates and resulting decision choices can bebetter accounted for.

Future research might include an explicitanalysis of the determinants of the range ofthe confidence interval associated with anLQ. Furthermore, there is a need to considerhow far the better understanding of uncer-tainty around LQ estimates can be incorpo-rated into techniques that are commonlyused to regionalise input–output tables foreconomic and geographical analysis.

Notes

1. Using Google scholar, it is possible to iden-tify over 700 articles using the LQ methodol-ogy published in 2008 and 2009 alone.

2. The paper here utilises the employment-based LQ—i.e. for industry i

LQi =xi

ni

�x

n

where, the numerator term is employment inthe industry divided by the national employ-ment in the industry; and the denominator isthe reference area employment over nationalemployment.

3. Although employment data are being usingin this illustration, this method could also beapplied to any estimated economic data usedto calculate location quotients such as, unitor GVA data.

4. Traditional industrial specialisation analysishas utilised sectors. Some have arguedthat commodities are a more accurate wayof identifying these agglomerations(Voigtlander, 2011). UK official surveys donot collect employment data by commoditygroups, but rather industry groups.

5. This work utilised 5-digit industry data fromthe 2010 Annual Business Inquiry, Office ofNational Statistics, UK, that the authors feelprovides a good coverage of the UK andWales manufacturing base. This list is notexhaustive, but was used to illustrate the use-fulness of the technique.

6. For the purposes of this illustration, the dePropris (2005) form of the LQ has beenused, but other forms would yield similarresults.

7. The 1.25 cut-off value is used as an examplebecause this was adopted in DTI (2001), butlater we show pictorially the consequences ofchanging cut-off values.

8. The reason for 18 industries shown in Figure2 and only 15 shown in Figure 1 is that inFigure 2 the upper limit on the consideredconfidence level is 99 per cent in contrast to95 per cent in Figure 1. The differences inconfidence levels employed are only to allowthe clarity of the associated graphs to bemaintained.

Funding Statement

This research received no specific grant from anyfunding agency in the public, commercial or not-for-profit sectors.

References

Bergman, E. M. and Feser, E. (1999) Industrialand Regional Clusters: Concept and Compara-tive Applications. Web book, RegionalResearch Institute, West Virginia University(http://www.rri.wvu.edu/WebBook/BergmanFeser/contents.htm).

Bergman, E. M. and Feser, E. J. (2000) Nationalindustry cluster templates: a framework forapplied regional cluster analysis, RegionalStudies, 34(1), pp. 1–19.

Beyene, J. and Moineddin, R. (2005) Methodsfor confidence interval estimation of a ratioparameter with application to location quo-tients, BMC Medical Research Methodology,5(32), pp. 1–7.

Bishop, P., Gripaios, P. and Bristow, G. (2003)Determinants of call centre location: someevidence for UK urban areas, Urban Studies,40(13), pp. 2751–2768.




Jesus A. Trevino

Highlight

Boix, R. and Galletto, V. (2009) Innovationand industrial districts: a first approach tothe measurement and determinants of theI-district effect, Regional Studies, 43(9), pp.1117–1133.

Brand, S., Hill, S. and Roberts, A. (1998) WelshInput–Output Tables for 1995. Cardiff: Uni-versity of Wales Press.

Carroll, M., Reid, N. and Smith, B. (2008) Loca-tion quotients versus spatial autocorrelationin identifying potential cluster regions, Annalsof Regional Science, 42(3), pp. 449–463.

Chiang, S. (2009) Location quotient and trade,Annals of Regional Science, 43(2), pp.399–414.

Clayton, D. and Hills, M. (1993) StatisticalModels in Epidemiology. Oxford: OxfordUniversity Press.

Crawley, A. and Pickernell, D. (2012) An apprai-sal of the European cluster observatory, Eur-opean Urban and Regional Studies, 19(2), pp.207–211.

Czamanski, S. (1974) Study of clustering of indus-tries: Halifax, Nova Scotia. Institute of PublicAffairs, Dalhousie University, Canada.

DTI (Department of Trade and Industry) (2001)Business clusters in the UK: a first assessment.DTI, London.

Feser, E. and Isserman, M. (2009) The rural rolein national value chains, Regional Studies,43(1), pp. 89–109.

Frances, C. J. de and Titus, R. M. (1993) Urbanplanning and residential burglary outcomes,Landscape and Urban Planning, 26, pp.179–191.

Gibson, J., Miller, M. and Wright, G. (1991)Location quotient: a basic tool for economicdevelopment analysis, Economic DevelopmentReview, 9(2), pp. 65–68.

Haig, R. M. (1926) Toward an understanding ofthe metropolis, Quarterly Journal of Econom-ics, 40, pp. 421–433.

Held, J. R. (1996) Clusters as an economic devel-opment tool: beyond the pitfalls, EconomicDevelopment Quarterly, 10(3), pp. 249–261.

Isard, W. H., Langford, T. W. and Romanoff, E.(1968) Philadelphia region input–output study.Working Papers, Regional Science ResearchInstitute, Philadelphia, PA.

Isserman, A. (1977) The location quotientapproach to estimating regional economic

impacts, Journal of the American Institute ofPlanners, 43(1), pp. 33–41.

Kumral, N. and Deger, C. (2006) An industrialcluster study: as a basis for the Aegean region’sdevelopment policy. Working Paper No. 06/01,Department of Economics, Ege University.

Malmberg, A. and Maskell, P. (2002) The elusiveconcept of localization economies: towards aknowledge-based theory of spatial clustering,Environment and Planning A, 34(3), pp.429–449.

McCann, P. and Dewhurst, J. (1998) Regionalsize, industrial location and input–outputexpenditure coefficients, Regional Studies,32(5), pp. 435–444.

Metraux, S., Caplan, J., Klugman, D. andHadley, T. (2007) Assessing residential segre-gation among Medicaid recipients with psy-chiatric disability in Philadelphia, Journal ofCommunity Psychology, 35(2), pp. 239–255.

Miernyk, W. H. (1976) Comments on recentdevelopments in regional input–output anal-ysis, International Regional Science Review, 1,pp. 47–55.

Miller, R. E. and Blair, P. (2009) Input–OutputAnalysis: Foundations and Extensions, 2ndedn. New York: Cambridge University Press.

Moineddin, R., Beyene, J. and Boyle, E. (2003)On the location quotient confidence interval,Geographical Analysis, 35(3), pp. 249–256.

Mulligan, G. F. and Schmidt, C. (2005) A note onlocalization and specialization, Growth andChange, 36(4), pp. 565–575.

O’Donoghue, D. and Gleave, B. (2004) A note onmethods for measuring industrial agglomera-tion, Regional Studies, 38(4), pp. 419–427.

Oehlert, G. (1992) Note on the Delta method,American Statistician, 46, pp. 27–29.

Openshaw, S. and Taylor, P. J. (1979) A millionor so correlation coefficients: three experimentson the modifiable areal unit problem, in:N. Wrigley (Ed.) Statistical Applications inthe Spatial Sciences, pp. 127–144. London:Pion.

Porter, M. E. (1990) The Competitive Advantageof Nations. Basingstoke: Macmillan.

Porter, M. E. (1999) Clusters and the new eco-nomics of competition, Harvard BusinessReview, November/December, pp. 77–90.

Porter, M. E. (2000) Location, competition andeconomic development: local clusters in a




global economy, Economic DevelopmentQuarterly, 14, pp. 15–24.

Propris, L. de (2005) Mapping local productionsystems in the UK, Regional Studies, 39(2), pp.197–211.

Schaffer, W. (1999) Regional Impact Models. Webbook, Regional Research Institute, West Vir-ginia University (http://www.rri.wvu.edu/regscweb.htm).

Silcocks, P. (1994) Estimating confidence limitson a standardised mortality ratio when theexpected number is not error free, Journal ofEpidemiology and Community Health, 48, pp.313–317.

Solvell, O, Lindqvist, G. and Ketels, C. (2003)The Cluster Initiative Greenbook. Stockholm:Ivory Tower AB.

Thomas, G. B. Jr and Finney, R. L. (1996) Calcu-lus and Analytic Geometry, 9th edn. Reading,MA: Addison Wesley.

Thrall, G., Fandrich, J. and Elshaw-Thrall, S.(1995) Location quotient: descriptive geogra-phy for the Community Reinvestment Act,Geo Info Systems, 5(6), pp. 18–22.

Tonts, M. and Taylor, M. (2010) Corporatelocation, concentration and performance:large company headquarters in the Austra-lian urban system, Urban Studies, 47(12), pp.2641–2664.

Voigtlander, N. (2011) Skill bias magnified: inter-sectoral linkages and white-collar labor demandin U.S. manufacturing. Economic WorkingPaper Series, University of California, LosAngeles.

West, G. R. (1986) A stochastic analysis of aninput output model, Econometrica, 54(2), pp.363–374.

Wong, N. H. and Chen, Y. (2009) Tropical UrbanHeat Islands: Climate Building and Greenery.New York: Taylor and Francis.

World Economic Forum (2007) Global competi-tiveness report 2007–2008 (http://www.gcr.weforum.org/).

AppendixThis appendix reports the trumpet shapes for all218 defined industries considered (see Figure A1).

Figure A1. LQ confidence intervals for SIC manufacturing industries in Wales, with varyingconfidence levels of 100(1� a) per cent, between 0 per cent and 95 per cent.




Jesus A. Trevino

Highlight

Jesus A. Trevino

Highlight

Documents

Mapping Local Production Systems in the UK