【正文】 lization results in productivitydifferences across firms in different locations.1The existence of these external scale effects hasimportant policy implications. A good understanding of these phenomena can help to designpolicies aimed at fostering the development of particular industries at the local and regional leveland guiding more general policies for local and regional growth.The empirical literature on agglomeration economies is substantial.2A great deal of thisliterature has focused on whether specialized economic environments (localization/MarshallArrowRomer externalities) or large and diversified cities (urbanization/Jacobs diversity effects)* The author acknowledges financial support from SEJ202065086 and 2020SGR002851See Duranton and Puga (2020) for an extensive review.2See Rosenthal and Strange (2020) for an extensive review.doi:169。es et al. (2020) we achieve the following. First, we cangive our estimates a random profit maximization interpretation. Second, we take into accountthe discrete nature of the variable being modelled. Third, we can deal with the fact that manylocations experience zero births. Fourth, we avoid the incidental parameters problem.This paper is organized as follows. After this introduction, Section 2 reviews the strand of theliterature that has studied the nature and scope of agglomeration economies. Section 3 deals withthe empirical analysis. After describing the data and variables, the chosen econometric specification is explained and justified and the results are presented and discussed. Finally, a summaryand concluding remarks are provided in Section 4.2 Agglomeration economiesAs mentioned in the Introduction, external economies exist when the scale of the urbanenvironment adds to productivity (Rosenthal and Strange 2020).4This is to say that agglomeration economies emerge as a consequence of adding up the individual external effects of theinteraction of firms located in the same geographical environment. Many mechanisms have beenproposed in the literature to explain the rationale for firms to colocate. The work of Marshallinspired one very wellknown typology. Marshall (1890) points out three main advantagesstemming from the colocalization of agents: Labour market pooling, input sharing andknowledge spillovers. What is meant by the labour market pooling externality is that thecolocalization of industrial activity in the same geographical area enables both firms andworkers to share risks of demand fluctuations at the individual level.5‘Input sharing’ refers to thebenefits that the concentration of firms may have for specialized input industries. Finally,knowledge spillovers occur because geographic proximity fosters knowledge transmissionbetween firms. However, incentives for agents to disperse may appear as city size increases.Agglomeration of economic activity may increase petition for immobile factors of production, thus raising the price of production inputs. Other incentives for firms to disperse mayinclude nonpriced congestion costs such as traffic, congestion and pollution.Data at a very fine spatial level is required to empirically analyse the geographic scope ofagglomeration economies. The lack of such datasets has limited the number of studies focusingon this question.6Rosenthal and Strange (2020) use US zip code level data to shed light on thisissue. By georeferencing the dataset, these authors are able to study how the effects of externaleconomies vary when the interactions of agents located at different geographic distances areconsidered. They found that external effects fall sharply after the first mile and attenuatesmoothly afterwards for the next 10–15 miles. This issue was also addressed by van Soest et al.(2020), who look at zip code level data for SouthHolland, a Dutch province. Their results also4Note that this does not imply that all forces driving colocation of agents in space take place through productivityshifts.5An alternative, related interpretation is based on better matching quality. Helsley and Strange (1990) show that anincrease in the number of agents trying to match in each location improves the expected quality of the match.6For instance, seminal papers on agglomeration economies by Glaeser et al. (1992) and Henderson et al. (1995) bothuse data at the US Metropolitan Statistical Area (MSA) level.577The scope of agglomeration economiesPapers in Regional Science, Volume 88 Number 3 August 2020.suggest that external effects take place at a geographic scale which is smaller than the city. ForSpain, ViladecansMarsal (2020) also finds evidence pointing in this direction.7The benefits for two firms to locate close to each other are very likely to vary depending notonly on the geographic distance involved, but also on the nature of their activities. Beforeaddressing interfirm industrial proximity, we must first arrive at a definition of the term, whichwill always be to some extent arbitrary. Ellison and Glaeser (1997) defined the concept ofcoagglomeration and derived a measure that is easy to pute. Nevertheless, this is only onepossibility. Probably because of the conceptual difficulty, most studies treat industrial distancein a binary fashion, namely, firms that belong to the same industry and those that do not. Thisleads to the distinction between localization and urbanization economies first proposed byHoover (1936). Localization economies are externalities arising between firms with the sameindustrial activity, whereas urbanization economies are external effects taking place betweenfirms operating in different industries.8Romer (1986) places knowledge spillovers and learning by doing at the core of economicgrowth. Glaeser et al. (1992) test some growth implications at the local level, stressing the roleof knowledge spillovers as a mechanism