Types of Agglomeration Economies: Effects on Business Innovation

A review of the literature does not provide conclusive results about the effects caused by firm agglomeration on innovation. In order to shed light on this issue, this paper draws a distinction among three kinds of agglomeration economies and empirically tests their respective impact on business innovation. The advantage that external knowledge generated through concentration can bring to each company depends on its absorptive capacity. Hence, it is posited that this dynamic capability acts as a mediator in the relationship between agglomeration and innovation. Using data from a survey conducted in 2013 by the Technological Innovation Panel (PITEC), an analysis of these ideas was performed using a sample of 2,906 high and medium-high technology companies. The results obtained indicate that several types of agglomeration economies exist and that the net effect each one of them has on innovation is different. More specifically, only urbanization economies favor innovation. Additionally, all of our findings reveal that firms increase their greater absorptive capacity in the context of agglomeration.


Introduction
Given that innovation −which generates growth, efficiency and profit in today's world− constitutes a key element in competitiveness, firms must innovate if they

Effects of agglomeration economies on business innovation
According to studies on agglomeration, the concentration of economic activity generates different types of externalities (Anselin, Varga, & Acs, 1997;Audretsch, 2003). These external economies, also known as economies of agglomeration, assume that the profits of a firm located near other firms increase as the number of firms in the same location increases (Appold, 1995). However, recent studies find that agglomeration can also have negative effects on business profits because greater competition exists among companies to obtain necessary inputs, such as land, employees, etc. (Arikan & Schilling, 2010;Flyer & Shaver, 2003;Folta, Cooper, & Baik, 2006;Glaesmeier, 1991;Pouder & St. John, 1996;Prevezer, 1997).
The purpose of this paper is to clarify the ambiguity surrounding the relationship between agglomeration and innovation. We begin by distinguishing among the three types of agglomeration economies that may prove beneficial for innovation depending on the type of co-located firms, namely, urbanization economies, localization economies, and knowledge-intensive economies.
Urbanization economies (Jacobs, 1969) are those derived from the concentration of companies that develop various economic activities in a particular area or region. This plurality of technological and commercial realities carries multiple and varied types of knowledge that firms can share and combine, thus enhancing innovation (Frenken, van Oort, & Verburg, 2007). It is in this context that inter-firm cooperation becomes feasible and allows for the generation of new knowledge, insofar as these firms are not rivals because they come from different industrial sectors. Moreover, creativity and innovation are likely to be favored through the combination of heterogeneous knowledge stemming from various industrial and commercial environments. As such a spatial concentration of activities without sectorial or industrial specialization is also characterized by a wider range of infrastructures, specialized services, and agents that act as middlemen and are responsible, to some extent, for the Vizja Press&IT www.ce.vizja.pl

219
Types of agglomeration economies: effects on business innovation investment attraction effect. A first hypothesis is formulated based on these thoughts.

Hypothesis 1: Business innovation increases with urbanization economies.
Localization economies (Glaeser, Kallal, Scheinkman, & Shleifer, 1992;Marshall, 1920) are those derived from the concentration of companies that develop the same economic activity in a specific area or region.
This geographical concentration produces externalities that allow firms to learn from each other. In this case, apart from the transmission of knowledge and ideas across companies, the use of the same language, together with the existence of a common knowledge base, permits greater interaction among firms and generates greater possibilities for new knowledge creation.
Thus, we present the following hypothesis:

Hypothesis 2a: Business innovation increases with localization economies.
Conversely, a higher concentration of potential competitors in the same place implies a greater relative shortage of resources, particularly that of valuable knowledge. In other words, although it is true that knowledge is not necessarily exhausted, it stops being valuable once it becomes indiscriminately accessible to any rival. That is why, within a context of physical proximity characterized by greater exposure to possible imitators, those firms that generate and take advantage of external knowledge must invest in the protection of that knowledge. This reallocation of resources meant for isolation and protection will most probably prove detrimental to investments in innovation.
Furthermore, in this regard, aside from the risk of being plundered by imitators, those firms that are best equipped in terms of knowledge will choose not to be located in places characterized by the concentration of competitors.
Taking into account both lines of reasoning allows us to conclude that being located in an environment with a higher concentration of firms belonging to the same industrial sector favors innovation up to a certain level of agglomeration after which saturation becomes excessive and the net effect on innovation then becomes negative (Marco-Lajara, Claver-Cortés, Úbeda-García, & Zaragoza-Sáez, 2016;Melo, Graham, & Noland, 2009;Sorensen & Sorenson, 2003). Consequently, we advance the following hypothesis: Hypothesis 2b: An inverted U-shaped relationship exists between business innovation and localization economies.
Knowledge-intensive economies (Knoben, Raspe, Arikan, & Oort, 2016) arise in locations next to firms and/or organizations that produce knowledge, in an environment where knowledge is valued, transferred, and generated. This knowledge-intensity is an essential feature of regions without industry specialization and where innovative and knowledge-oriented agents are located. Thus, we formulate Hypothesis 3.

The mediation effect of absorptive capacity
Even though a large number of firms may actually be exposed to identical environmental conditions, not all of them are able to convert outside knowledge into results with the same levels of success because they differ in their abilities to use these sources of knowledge (Caloghirou, Kastelli, & Tsakanikas, 2004;Rothaermel & Hess, 2007).
In fact, the existence of more external sources of potentially useful knowledge increases the possible combinations of knowledge and, therefore, the complexity of its management. As a result, the inability of firms to manage and exploit that knowledge can limit their possibilities for innovation (Henderson & Clark, 1990;Laursen & Salter, 2006), which is why it is necessary to highlight the role of AC (Cohen & Levinthal, 1990 Together, Hypotheses 4 and 5 lead to another hypothesis that predicts a mediating effect of AC on the link between agglomeration and innovation: Hypothesis 6: AC mediates the relationship between agglomeration and business innovation.

Methodology
Although the present study tests the hypotheses using a multiple linear regression, a variety of models are estimated because not all hypotheses can be tested in the same way. H1, H2a, and H3 predict a direct effect of the independent variable (agglomeration) on the dependent variable (innovation); H4, H5, and H6 forecast a mediating effect; and H2b predicts a nonlinear effect.
Furthermore, with respect to H1, H2a, and H3, the model exhibits a general formulation such that Y = β 10 + β 11 * X + β 12 * C, where Y is the dependent variable, X is the independent variable, and C is the control variable.
According to Judd and Kenny (1981) and Baron and Kenny (1986), the analysis of the mediating effect requires the formulation of three equations.
In the first equation, the dependent variable is estimated using independent and control variables, and the equation is the same as that of the direct effect. In the second equation, the mediator variable is estimated using independent and control variables. Regarding the third, the dependent variable
The IBM Statistical Package for the Social Sciences (SPSS) version 23 was used to conduct the calculations.

Measurement of variables
The work performed to measure all the variables specified in the equations is supported by the PITEC (Panel de Innovación Tecnológica) database, elaborated using a questionnaire about innovation in business.

Dependent variable
Two approaches were considered when estimating IN-NOVATION. One of the commonly used measures refers to the number of patents for which a company has filed an application (Dutta & Weiss, 1997;Henderson & Cockburn, 1994;Squicciarini, 2008;2009 A principal components factor analysis was applied with both approaches, which explains 65.55% of variance. nizations. Cooperation between businesses and organizations is fostered in such locations, and a physical as well as social infrastructure exists that stimulates the creation, access, and acquisition of external knowledge. Accordingly, innovation is encouraged in such locations (Squicciarini, 2008;2009;Siegel, Westhead, & Wright, 2003).

Independent variables
LOC.AGGL. Localization economies result from the geographic concentration of similar firms in a given area, an autonomous region in our case. The fact that many firms are located in several regions caused us to adopt a criterion when determining the autonomous region associated with each company, more precisely, the place where each firm develops its internal R&D. In this sense, it was decided that this place corresponds to the physical location of R&D workers.
The data reveal that 59.73% (1,736)  With respect to PAC, the variable EXTSOURC scores the number of external sources of knowledge (machine suppliers, clients/customers, competitors, consultants, private laboratories, universities, public research bodies, technology centers, conferences, fairs or exhibitions, scientific journals, industrial and professional associations) that can be assessed as being of high importance by firms. The value of EXTSOURC ranges from 0 to 9. The opportunity for each firm to access external knowledge through successful alliances, for which the variable SUCCALL was estimated, was also considered. SUCCALL receives a score of 1 if the firm has engaged in developing or innovating products, technological processes, organizational practices, or commercial strategies with other firms or institutions. A principal components factor analysis performed with these two variables enabled us to extract one factor that explains 60.46% of the PAC variance.
As for RAC, the literature often uses measures related to R&D expenditure. Therefore, our study focused on the percentage of internal expenditure R&D over the total expenditure in R&D (INTR&D). RAC not only depends on the implementation of such investments, but its use as an indicator could actually penalize the importance of smaller-sized organizations that are unable to carry out R&D activities on a regular basis. Perhaps for this reason, several studies ultimately stressed the importance of human resources when identifying and assessing this capability (Mangematin & Nesta, 1999). In keeping with the previous line of reasoning, the following indicators were considered: the relative importance of research staff with respect to the entire staff −denoted by the variable RESTAFF− and whether the RAC has a twofold dimension depending on the source of information, i.e., scientific or market-related, on which it is supported (Caloghirou et al., 2004;Murovec & Prodan, 2009). Hence our decision to take two variables into account, namely, R&DSTAFF (percentage of R&D staff employed in internal R&D) and HIGHEDU (percentage of employees who have completed higher education). A principal components factor analysis performed with the aforementioned four variables allowed us to extract a Types of agglomeration economies: effects on business innovation factor that explains 77.81% of the RAC variance. Finally, the measure that represents the AC was obtained from a factorial analysis of these two factors (PAC and RAC), which explains 73.35% of the variance.

Control variables
AGE. The number of years during which a firm has been operating since its foundation can influence innovation both positively and negatively. Indeed, greater experience is likely to permit a higher accumulation of knowledge, but it may also become an inertiagenerating source that hinders both adaptation and the introduction of novelties in products and processes. SIZE. Size is significantly correlated with innovation, even though no consensus exists as to whether this relationship is positive or negative. On the one hand, larger firms can be more innovative due to their greater financial holdings, but on the other hand, the higher flexibility and the better communication level actually allows smaller-sized firms to be the most innovative (Damanpour & Gopalakrishnan, 1998). A dichotomous variable, SIZE200, which shows whether the firm is defined as large because it has over 200 employees or is not large, was added to our model to assess this effect.
GEOGRAPHIC SCOPE. The greater or lesser predisposition to innovate may additionally be determined by sales expectations, which in turn are going to depend on the breadth of the geographical markets that constitute the product or service target (Löfsten & Lindelöf, 2003). Thus, sales dispersion will also most likely encourage innovation because of the need to adapt products to the local demand and to the regulations of foreign markets (Vernon, 1966

Population and sample
The study object of this paper is formed by those firms that have innovated, that is, firms that have engaged in any activity oriented to achieve new or significantly improved products or processes. Previous studies suggest that the dynamism or the technological tur-

Results
The following tables summarize the results of all the models estimated. 1 As presented in Table 2, H1, H4 (this effect is included in every table), and H5 were verified when considering the urbanization economies generated in a STP. Moreover, confirmation was equally obtained for H6, according to which AC mediates the influence of agglomeration on business innovation.

Instead, Tables 3 and 4 test the effect of localization economies on business innovation (H2a and
H2b). The absence of a direct and significant effect becomes evident, which means that neither H2a nor H2b are verified. It can, however, be proved that these economies directly and significantly impact the AC of firms (H5). An additional regression was performed in which the relationship between AC and the localization economies was non-linear. The results of that regression indicate that AC increases up to a certain level of agglomeration at which point it then begins to decrease. This is the effect as predicted by H2b for business innovation.
Finally, knowledge-intensive economies exhibit the same patterns as do the other types of economies (Table 5). Moreover, they do not have a direct and significant effect on innovation, which prevents us from validating H3 and H6, but they do cause such an effect on AC (H5).
With regard to control variables, despite the fact that none of the control variables has a significant effect on absolute value as none of the estimated coefficients exceeds one, all but AGE cause a positive effect and are statistically significant in every estimated equation.     and use seem to be restricted to certain specific communities or networks inside the district. In this way, knowledge remains external to each one of the firms;

Discussion, conclusions, and future lines of research
it cannot be stated that it is freely disseminated across the district, becoming a club asset (Morrison, 2008).
This makes us reflect not only on the fact that being located near sources of external knowledge does not guarantee access to it but also on the importance of developing the ability to establish and manage alliances as well as on the extent to which this influences the generation of and access to external valuable knowledge resulting from cooperation.
It could also be assumed that firms characterized for being more autonomous and self-sufficient from the knowledge point of view depend to a lesser extent on the advantages that location in agglomerations can bring and that they have a stronger need for protection against exposure and potential imitation. This probably justifies why they seem less willing to be located close to other rival firms or to those for which external knowledge really is a must. Being able to prove this argument would allow us to state that the appeal of agglomerations is much greater for firms that are more strongly dependent on external knowledge.