Location determinants of high-tech firms: an intra-urban approach

ABSTRACT

This paper analyses location determinants of high-tech firms at the intra-urban level, concretely for neighbourhoods of Barcelona. Mercantile Register data is used to analyse the location of 515 firms between 2011 and 2013 through count data estimations. The identification of the location patterns, followed by a typology of the firms, and the role played by neighbourhood characteristics in attracting them, constitutes a contribution to the empirical literature. Our results help in understanding the entry processes within cities and show that i) there are certain specificities at industry level, ii) that both amenities and economic-oriented neighbourhood characteristics matter, and iii) that spatial spillovers are relevant for some high-tech industries.

KEYWORDS:

• Location decisions
• count data
• cities
• intra-urban level
• spatial econometrics
• high-tech

JEL:

• O30
• R10
• R39

Acknowledgments

This research was partially funded by FEDER/Ministerio de Ciencia, Innovación y Universidades – Agencia Estatal de Investigación ECO2017-88888-P, the “Xarxa de Referència d’R+D+I en Economia i Polítiques Públiques” and the SGR Program (2017-SGR-159) of the Generalitat de Catalunya. I would like to acknowledge research assistance by M. Lleixà, supportive comments by A. Rodríguez-Posse, M. Manjón-Antolín, seminar and conference participants, and two anonymous referees, for their valuable comments and suggestions.

Disclosure of potential conflicts of interest

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

¹ Concretely, Video & TV Production, Wired Telecommunications and Other Telecommunications, Computer Programming & Consultancy, and Data processing & Hosting, Other Information Services and R&D Natural Sciences & Engineering, and R&D Social Sciences & Humanities.

² The Catalan case has been widely analysed using different approaches. Liviano and Arauzo-Carod (2014, 2013) focused on the ‘zero problem’ (i.e., the existence of a threshold in terms of whether a site can be chosen by a firm) and the role of spatial effects on extreme overdispersion. Arauzo-Carod (2008) and Arauzo-Carod and Manjón-Antolín (2012) considered the spatial units to be used for location analyses.

³ Coll-Martínez, Moreno-Monroy, and Arauzo-Carod (2019) analyse the spatial extent of agglomeration of creative industries in Barcelona (as a proxy of agglomeration economies) and conclude that it ranges between 0 and 1 km.

⁴ Nevertheless, there are also researchers who play down the role of agglomeration economies (see Rousseau 1995) and consider that skilled labour availability and sectorial specialisation are more important determinants of efficiency differentials at big urban areas.

⁵ Nevertheless, urban size is not the only source of efficiency gains, as shown by the concept of ‘borrowed size’ (Alonso 1973). According to this, proximity to big cities may allow smaller cities to borrow some of the advantages of their bigger neighbours.

⁶ Lack of empirical research for European cities is especially relevant in view of urban specificities across cities in different areas (e.g., Asian cities, U.S. cities, continental European cities, etc.) that make comparisons difficult.

⁷ There is plenty of evidence of suburbanisation of high-tech activities. See, for instance, Nunn and Warren (2000) for computer services in the metropolitan statistical areas of the U.S. or Guillain, Le Gallo, and Boiteux-Orain (2006) for a wide range of industries at the Paris region.

⁸ The attractiveness of large urban areas should be considered in net terms, as there are, as well, additional disagglomeration economies operating for these cities (Camagni, Capello, and Caragliu 2016).

⁹ See, among others, Jofre-Monseny and Solé-Ollé (2009) and Jofre-Monseny, Marín-López, and Viladecans-Marsal (2011).

¹⁰ Selection is made according to both standard classifications of high-tech activities and the typology of industries predominant in Barcelona.

¹¹ Following Babyak (2004) and in order to assess the (potential) impact of overfitting in our estimates, we have estimated different combinations of the estimations using a shorter number of covariates, and the results are essentially the same. Additionally, we checked AIC when adding/dropping parameters and although, in general terms, lower AIC were obtained when dropping out covariates (this effect was largely dependent on the specific covariates involved). Therefore, we consider that although overfitting is a relevant issue that has to be always monitored and controlled, in this case it is not driving results in a relevant way.

¹² See Alañón-Pardo and Arauzo-Carod (2013) and Melo, Graham, and Noland (2010) for an extended discussion.

¹³ For the sake of making the econometric estimation as simple as possible (and in order to avoid potential problems of overfitting -see also footnote 11) we decided not to include spatial lags of all independent covariates (additionally, it was unnecessary to lag all of them because some variables did not have a clear spatial pattern as values at a given neighbourhood had no or minor relationship with values in near areas, according to Moran’s I results). We selected stock of firms because of the expected spatial scope of agglomeration economies caused by the concentration of firms was expected to go beyond neighbourhood borders; 22@ district dummy, because of similar reasons, as although the official policies of 22@ target only this single district it is also true that many firms outside the 22@ may also benefit if they locate close to facilities existing there; population with college degree, because of well-known positive effects of human capital that are very difficult to constrain inside neighbourhood borders; and percentage of foreigners, because of benefits of a ‘melting pot’ imply a lot of amenities and economic activities occurring a limited distance from the neighbourhoods where foreigners live.

¹⁴ Results of these tests (overdispersion, dispersion statistic, and alpha = 0) are available upon request. We have also carried out correlation analysis and there are no relevant issues with the selected variables.

¹⁵ This clustering pattern at core areas is also observed for other ‘similar’ industries, as advanced services in Brussels (Waiengnier et al. 2020).

¹⁶ It is important to take into account that high levels of Gini index are partially explained by the few numbers of entries for these industries.

¹⁷ As for the Gini index (see previous footnote), the lower entropy levels seem to be explained by a few entries at peripheral neighbourhoods.

¹⁸ In order to calculate local and global measures of spatial autocorrelation, as well as spatially lagged variables, and in view of the size, shape and proximity of neighbourhoods of Barcelona, we have decided to use a contiguity matrix as a spatial weight matrix. Although there are alternative criteria (e.g., distance-based), these could have some limitations such as an inappropriate number of neighbours (i.e., very similar to that of total neighbourhoods).

¹⁹ Despite coefficients were non-significant, they had a negative sign in most specifications. In this sense, a similar analysis for Beijing postal areas (Zhang et al. 2013) shows that land prices deter entry of high-tech firms.

²⁰ Although there is theoretical and empirical evidence suggesting the positive relationship between amenities and wages (Roback 1982), the identification of such amenities (and its inclusion in an econometric estimation) is not obvious. Notwithstanding these difficulties, our variable measuring availability of the public bike renting system captures some amenities as public transport accessibility that are expected to be positively appreciated by workforce of high-tech firms. At this point, our approach is similar to the one known as ‘voting-with-your-feet’, initially proposed by Tiebout (1956) and later on developed by, among others, Wall (2001). Concretely, this approach implies that location decisions are taken depending on the trade-off between advantages and costs of different sites (e.g., amenities vs. local taxes), being that agents (e.g., firms) will move away if they feel that, for instance, quality of local amenities does not deserve the costs of locating there.

²¹ Fort the sake of simplicity, we did not include the spatial lagged counterparts of all covariates, as the spatial scope of the omitted ones is not expected to go beyond a neighbourhood’s borders. See also footnotes 11 and 14.

²² Each of the 73 neighbourhoods of Barcelona belongs to one of the 10 city districts. Results clustered at district level are available on request.

²³ These public efforts include, among others, the creation of research and innovation centres and the relocation of several public university facilities.