【正文】 r’spoint of view. By assuming a particular distribution on eij, it is possible to make probabilitystatements with regard to the likelihood of an establishment choosing a particular location j. Theconditional logit model is obtained by assuming that eijfollows an . Weibull distribution. Inthis particular setting, the probability that establishment i, that belongs to industry s, chooseslocation j, psij, is of the conditional logit type given by exp exp′() ′()=∑yysjssjsjJββ1.The validity of the conditional logit model hinges on whether or not the independence ofirrelevant alternatives (IIA) assumption holds. This property follows from the independenceassumption made on the eijs. In the context of the firm decision problem, the eijs are unlikelyto be independent for two reasons. First, entrepreneurs may be attached to a particular area andhence the eijs will be correlated across nearby municipalities. Second, unobserved location13In cases in which there was no employment for the relevant variables, following Rosenthal and Strange (2020), thepetition variables are assumed to be zero.581The scope of agglomeration economiesPapers in Regional Science, Volume 88 Number 3 August 2020.Table 2. Summary statistics, year 2020Birth of new establishmentsMean Std. Dev. Min MaxTextiles 0 22Wood and furniture 0 10Chemical products 0 2Metal products except for machinery 0 8Motor vehicles 0 2Radio, television and munication equipments 0 2Medical, precision and optical instruments 0 4Location attributesMean Std. Dev. Min MaxUrban land area (Km2) 0 Textiles employmentmunicipal 0 4,9499 km 1, 0 11,4589–15 km 1, 2, 0 18,779Firmss/workerss 1Wood and furniture employmentmunicipal 0 5,8769 km 0 10,0529–15 km 1, 2, 0 16,229Firmss/workerss 1Chemical products employmentmunicipal 0 19,4889 km 2, 0 28,1029–15 km 1, 5, 0 35,239Firmss/workerss 1Metal products except for machinery employmentmunicipal 0 8,6709 km 2, 0 23,7239–15 km 1, 4, 0 33,763Firmss/workerss 1Motor Vehicles employmentmunicipal 0 24,6159 km 2, 0 29,5709–15 km 1, 5, 0 34,609Firmss/workerss 1Radio, television and munication equipment employment.municipal 0 2,3739 km 0 4,6499–15 km 0 5,213Firmss/workerss 1Medical, precision and optical instruments employmentmunicipal 0 2,0909 km 0 3,9649–15 km 0 5,121Firmss/workerss 1Overall manufacturing employmentmunicipal 4, 0 118,7119 km 6, 1, 0 178,6689–15 km 15, 37, 0 269,983Firms/workers 1Diversity indexmunicipal 1 9 km 2 9–15 km 1 Note: All petition variables statistics are puted excluding those municipalities with no positive appropriateemployment levels.582 J. JofreMonsenyPapers in Regional Science, Volume 88 Number 3 August 2020.determinants are also very likely to be spatially correlated, inducing correlation in the eijs acrossnearby municipalities. Our strategy to remove the spatial correlation in the eijs is to conditionthe choice set to be the local labour market (LLM) where the establishment locates.14This is toassume that the IIA assumption only holds across municipalities within a local labour market.15At an intuitive level, we are assuming that even though an entrepreneur may be attached toa particular area, she will still choose the best location for her establishment within the set ofmunicipalities found in her area (., her local labour market). Note that by restricting the choiceset to the local labour market level we are controlling for wages since these do not change acrossmunicipalities within a local labour market (and hence they drop out of the probabilities). Thisstrategy also provides us with a natural control for municipal variables that typically sh。 the less than 9 km concentric ringlevel (9 km)。 hence, theexternalities stemming from innovation are to a larger extent internalized. In contrast, Jacobs(1969) and Porter (1990) claim that local petition fosters the rapid adoption of technologiesand, therefore, favours growth. Glaeser et al. (1992) finds evidence to support Porter’s andJacobs’ theories. A related question is whether entering firms show a preference for small orlarge firms. Saxenian (1994) presents evidence that small firms are more open and innovative.7Also for Spain, Ala241。 2020 the author(s). Journal pilation 169。The scope of agglomeration economies: Evidence fromCatalonia*Jordi JofreMonseny11Universitat de Barcelona amp。blica, Avda.Diagonal 690, Torre 4, Planta 2a, 08034 Barcelona, Spain ( )Received: 14 January 2020 / Accepted: 28 October 2020Abstract. This paper is an empirical study of the geographic and industrial scope of agglomeration economies. We also explore if small establishments make better neighbours than theirlarger counterparts. We address these issues by studying the effects of local industrial characteristics on the location decisions of new establishments using the random profit maximizationframework. We carry out separate econometric estimations for seven industries in Catalonia, aSpanish region, using data from 1995–2020. Agglomeration economies seem to work at a verylocal level. Evidence of localization, urbanization and diversity effects is found. There isno strong evidence that establishments prefer to be located near to small rather than largeestablishments.JEL classification: L25, R30Key words: Agglomeration economies, firm location, conditional logit, Poisson regression1 IntroductionExternal effects exist when the economic scale of a firm’s geographical location enhances itsproductivity (Rosenthal and Strange 2020). There is a substantial body of literature on thequestion of why firms colocate in space and how this coloca