Is the city centre synonymous with (non-)cooperative innovation? Innovation in Knowledge -Intensive Services through intra-metropolitan space of Mexico City

Abstract

Chinese

Innovation in Knowledge-Intensive Services (KIS) holds a central position in the global economy. Innovation developed by KIS can occur in either a cooperative or non-cooperative mode with other economic agents. The cooperative mode has traditionally been associated with the central area of the city, while the non-cooperative mode has been linked to the urban periphery, where KIS presumably finds favourable conditions for their innovative activities. By distinguishing between cooperative and non-cooperative innovation in KIS, this article aims to analyse whether this spatial distribution is followed in the intra-metropolitan area of Mexico City. Microdata on industrial innovation activity and statistical tools are employed to evaluate the global and local effects of agglomeration economies and urban innovation dynamics. The results indicate that, in addition to the spatial concentrations suggested in the literature, there are also clusters of cooperative innovation in the periphery and of non-cooperative innovation in the ‘central milieu’, both of which are associated with location economies and innovative dynamics. These findings challenge the notion that a specific mode of innovation in KIS is inherently linked to a particular urban space, and instead highlight the spatially differentiated utilisation of intra-urban conditions.

Keywords

cooperative innovation innovation in services innovation mode intra-urban geography knowledge intensive services

Introduction

Innovation developed by knowledge intensive services (KIS) is a central element for success in the global economy. An economy based on the continual development of innovations now places KIS at its core (Gallouj and Windrum, 2009). Services have evolved from being ‘takers’ and ‘co-producers’ to also acting as ‘generators’ of innovations, a concept defined as ‘innovation in KIS’ (Barras, 1990; Den Hertog, 2000; Miles, 2005). KIS fulfil this role due to the characteristics of their workforce, which includes high levels of qualifications, skills and the capacity for technological adoption (Daniels and Bryson, 2002). It is recognised that innovation processes in KIS have distinct qualities, compared to other industries (Corrocher et al., 2009; Tether, 2005).

The intra-urban geography of KIS has garnered significant interest, with various studies analysing its characteristics in cities around the world (Arauzo-Carod et al., 2017; Méndez-Ortega et al., 2022; Romero de Ávila Serrano, 2019; Santiago and Sánchez-Zárate, 2023; Shearmur and Alvergne, 2002; Solis et al., 2021; Ženka et al., 2020). However, research on their innovative performance in intra-urban areas remains limited. Some studies have explored the microgeography of innovation in KIS within specific urban spaces (Rammer et al., 2020; Roche, 2020). To our knowledge, Shearmur (2012) conducted the only systematic research on the intensity of innovation in KIS across intra-urban spaces. His findings indicate that innovation in KIS is not confined to the central business district (CBD) in Montreal (Canada) but is, in fact, more intense outside the CBD.

The literature on innovation in intra-urban space assumes that KIS located in central areas benefit from urbanisation economies and the intense knowledge exchange that characterises these areas. These firms tend to develop innovation through a cooperative mode, relying on external relations and porous boundaries in their interactions with other agents and their environment. In contrast, other KIS opt to situate in peripheral zones, driven by factors such as secrecy and a relatively autonomous capacity for innovation (Shearmur, 2012). This location is associated with a non-cooperative mode of innovation, whereby firms exert greater control over their interactions with their surroundings, and avoid agglomerations (Katz and Wagner, 2014; Shearmur, 2012).

The main objective of this research is to analyse and verify whether the central location patterns of innovation in cooperative KIS, and the peripheral location pattern of non-cooperative KIS, are present in the Mexico City Metropolitan Area (MCMA). The MCMA is chosen as the case study because it concentrates a large share of the national services sector (Alvarez-Lobato et al., 2023; Santiago, 2020) and presents strong internal contrasts (Escamilla et al., 2016). These features make it a suitable setting for assessing whether the intra-urban pattern of KIS innovation proposed in the literature is also present in a major Global South metropolis. Data on innovation in KIS are derived from the microdata of the Economic Censuses (EC) conducted by National Institute of Statistics and Geography (INEGI). A Geographically Weighted Logistic Regression (GWLR) was employed to evaluate the spatial relationship between innovation in KIS, agglomeration economies, and innovation dynamics across the city.

The article contributes to the research on the intra-urban geography of innovation in KIS, with particular attention to large cities of the Global South. The results obtained for MCMA provide partial support for the study’s hypothesis, identifying concentrations of cooperative innovations in KIS within the central area and non-cooperative innovation outside it, as suggested by the literature (Katz and Wagner, 2014; Shearmur, 2012). However, the findings also reveal concentrations of cooperative innovation in the urban periphery and non-cooperative innovation in the central area of the national metropolis. The evaluation of the global and local effects of agglomeration economies and intra-urban innovation dynamics suggests that these mechanisms operate across innovation modes. KIS appear to make extensive use of these factors both through proximity and over distance, enabled by the accessibility and technological development within urban space (Parr, 2002).

These results challenge the assumption that innovation modes are tied to specific urban spaces. Instead, they point to the central role of agglomeration economies and intra-urban dynamics in shaping innovation in KIS across the metropolitan area. By highlighting this diversity of spatial configurations, the case of the MCMA complicates centre-periphery approaches in the intra-urban geography of innovation, which have largely been derived from cities in the Global North. From this perspective, the study cautions against reinforcing spatial frameworks that may accentuate intra-urban inequalities and emphasises the need to recognise diverse spaces of innovation within development strategies for large cities of the Global South.

The rest of this article is structured as follows. Firstly, the conceptual framework on innovation in KIS, and the approaches that seek to explain its intra-urban geography, is presented. Secondly, the analytical strategy on the sources of information and tools of global and local analysis is detailed. Thirdly, the descriptive results and results obtained in the statistical models are described. Fourthly, as a conclusion, the results are contrasted and discussed with the literature on the subject.

Conceptual framework: KIS, innovation modes and intra-urban geography

Innovation in KIS: Background and (non-) cooperative modes of innovation

Changes in global production and consumption systems have resulted in: (i) the hegemony of the service sector as the main employer and generator of wealth and (ii) a growing demand for continuous innovation as key economic driver (Daniels and Bryson, 2002). Both results are intertwined with each other (Wood, 2005: 434). The role of the service sector in innovation processes has been recognised since the last century. KIS have gone from being considered ‘takers’ or ‘co-producers’ (Barras, 1990; Den Hertog, 2000) to being recognised as ‘generators’ of innovation, a role that is conceptualised as ‘innovation in KIS’ (Gallouj and Windrum, 2009; Miles, 2005).

Innovation in KIS occurs in the production field (Corrocher and Cusmano, 2014; Lee and Miozzo, 2019; Silva et al., 2011), and the consumption sphere (Hall, 2009; Santoro et al., 2020). The innovation developed by KIS has its own particularities, that distinguish it from other innovative segments, such as manufacturing (Evangelista and Sirilli, 1995; Gallouj and Windrum, 2009; Miles, 2005). It has been indicated that the vertical disintegration and modularisation of economic activity opened the possibility for the innovative process to be carried out in either a ‘collaborative’ or a ‘non-collaborative’ mode (Chesbrough, 2003; Dankbaar, 2007; Mina et al., 2014).

The collaborative mode is based on the exchange of ideas and knowledge that originate outside the company (Aslesen and Freel, 2012). This mode of innovation seeks to obtain value from assets that are in the ‘environment’ (Santoro et al., 2020), and accelerate innovation cycles (Enkel et al., 2009). Interactions are carried out with customers, competitors, universities and/or public sectors (Freel, 2006; Lee and Miozzo, 2019). It is a network of agents that can extend both locally and globally (Hervas-Oliver et al., 2021). Its development is mediated by human capital (Corrocher et al., 2009), technologies (Barrett et al., 2015), and firms’ capabilities to absorb and manage the heterogeneity of knowledge sources to which they are exposed (Mubarak and Petraite, 2020; Santoro et al., 2020).

The non-collaborative mode involves the development of research and development (R&D) within the firm itself, which is mainly undertaken by large companies with the capacity to sustain the necessary investments (Chesbrough, 2003; Silva et al., 2011). Among the reasons for the adoption of this mode of innovation are: sustaining competitive positions (Barrett et al., 2015), the technological complexity that limits the possibilities of interaction with their clients (Freel, 2006), and/or the technological intensity that allows them to cover large markets (Corrocher et al., 2009).

Intra-urban geography of innovation in KIS: Local and non-local effects

The innovation activity of KIS in the intra-urban area has been understood from two main perspectives. One line of research emphasises the local function, which assumes that co-location between industries is a necessary condition for the development of innovation processes (Kekezi and Klaesson, 2020). Innovation, in this line, relies mainly on the exchange of tacit knowledge, which requires spatial proximity to activate learning, matching, and sharing processes that occur formally and informally among economic agents (Malecki, 2021).

Such co-location does not occur randomly. The ‘innovative milieu’ of the central area has been identified as the space where innovation and creative processes are encouraged (Camagni, 2003; Hutton, 2004). This characteristic is associated with urbanisation economies (Alvarez-Lobato et al., 2023) and the availability of urban amenities (Duvivier and Polèse, 2018). These conditions foster learning, diversification, and dense inter-firm interaction, particularly among highly skilled workers and knowledge-intensive firms. At the same time, other studies have recognised the role of districts or clusters located outside central areas. In these contexts, innovation dynamics emerge through co-location with other innovative agents (Shearmur, 2012) and through localisation economies linked to related industries (Rammer et al., 2020).

An alternative perspective emphasises the non-local function, derived from empirical studies that have identified innovative industries in relatively isolated locations on the periphery of, or outside, the urban area (Suarez-Villa and Walrod, 1997). This spatial pattern suggests that not all innovative industries need to be co-located in order to launch their innovation processes (Bathelt and Turi, 2011; Fitjar and Rodríguez-Pose, 2017). Three mechanisms, which are not mutually exclusive, have been proposed to explain this.

One mechanism involves accessing the advantages of the central milieu from a distance, either frequently or sporadically, through communications and transport infrastructure, thereby avoiding the costs that are associated with the central location (Parr, 2002). A second explanation highlights how the peripheral location enables firms to maintain secrecy in their innovative activities, providing them with competitive advantages in introducing new products (Suarez-Villa and Walrod, 1997). A third mechanism concern firms operating in advanced stages of innovation, which do not require them to establish frequent face-to-face contact for the exchange of knowledge (Torre, 2008).

Based on the latter two perspectives, the literature suggests that the intra-urban geography of KIS innovation modes follows a core-periphery pattern. Central areas favour collaborative innovation, thanks to the urbanisation economies, urban amenities and intensive knowledge exchanges that occur there. Non-collaborative innovation, in contrast, tends to locate in the periphery. This spatial pattern is associated with mechanisms such as the remote exploitation of core advantages, secrecy, and a reduced need for face-to-face interaction in advanced stages of the innovation cycle (Hutton, 2004; Katz and Wagner, 2014; Suarez-Villa and Walrod, 1997).

Although this approach has been explored in cities of the Global North (Shearmur, 2012), its analysis in structurally distinct contexts, such as the MCMA is still limited. The MCMA, the main centre of activity for KIS in Mexico, but with a secondary role in the global economy (Graizbord et al., 2003) presents marked asymmetries between its core area and the periphery in urban-functional, economic, institutional, and social terms (Escamilla et al., 2016). These conditions are expected to accentuate the centre-periphery pattern described in the literature. Thus, the study of the intra-urban geography of KIS innovation modes in the MCMA allows us to examine the extent to which the centre-periphery pattern suggested by the literature is reproduced or intensified in one of the large metropolises of the Global South. The following section outlines the strategy used to address the central objective of the study.

Methodology

This study follows a five-phase empirical workflow, that is designed to analyse the spatial patterns of innovation in KIS. The phases are presented sequentially in Figure 1 and described below. The first step consists of defining the scope of KIS. While several proposals for operationalising KIS concentrate on production-oriented services (i.e. Knowledge-Intensive Business Services, KIBS; Miles et al., 2018), the recent literature highlights that consumer-oriented services also generate significant innovation (Hall, 2009). Excluding the latter would underestimate the overall scope of innovation, as they represent a key segment of the global economy (Castaldi, 2023). In the present study, the definition proposed by Santiago (2020) is adopted, which jointly incorporates production- and consumer-oriented services.

Figure 1.

Workflow diagram.

In the second phase, definition of variables, innovation is classified according to whether firms collaborate with external actors (cooperative), or innovate exclusively in-house (non-cooperative), following Chesbrough (2003). The operationalisation uses microdata from the Science, Technology and Innovation module of the 2019 Economic Censuses (INEGI, 2020, 2024), based on four dichotomous questions (yes = 1, no = 0):

Q1: Collaboration with universities or research centres.

Q2: Collaboration with firms with no productive relationship.

Q3: Collaboration with clients, suppliers, or subsidiaries.

Q4: Innovation carried out exclusively by the firm, without collaboration.

Firms reporting any of the first three items are classified as cooperative innovators, while those responding only to item four are classified as non-cooperative innovators.

In the third phase, we built a spatial matrix containing information from establishments with cooperative and non-cooperative innovation. The units of observation are the Basic Geostatistical Areas (BGAs) of the MCMA, defined as the built-up area within the metropolitan boundaries of the city in 2015 (SEDATU et al., 2018; Figure 2). The choice of BGAs responds to two criteria: (i) they enable the aggregation of firm-level innovation data without compromising confidentiality and (ii) their official delineation allows integration with other government datasets, comparability across sources and temporal monitoring of urban phenomena, as well as their potential applicability to urban planning. This selection is analogous to the use of Census Tracts in US research and offers a suitable level of detail for intra-urban analysis. Using these units, a spatial matrix of innovation was created, recording for each BGA the number of firms by innovation type (cooperative and non-cooperative).

Figure 2.

Mexico City Metropolitan Area (MCMA) Basic Geostatistical Areas (BGAs) and political delimitation.

In the fourth stage, based on the spatial matrix, a dichotomous identifier of innovation specialisation was generated for each BGA using the Location Quotient (LQ). In its classical formulation, the LQ measures the relative sectoral specialisation (sector i) of a spatial unit (j), with values above one indicating a relative concentration. BGAs exceeding this threshold were categorised as specialised in cooperative or non-cooperative innovation.

Finally, inferential techniques were applied, in order to examine spatial concentration and the functional dynamics of innovation specialisation. Firstly, the Nearest Neighbour Index (NNI) was used to provide an initial assessment of spatial clustering or dispersion among BGAs that were categorised by innovation type. Subsequently, the dichotomous variables of specialisation were employed as inputs in two Geographically Weighted Logistic Regression (GWLR) models. These models allow the estimation of local variations in the relationship between innovation specialisation and spatial context.

Table 1 describes the independent variables incorporated into the GWLR, reflecting the dimensions outlined in the conceptual framework. Employment specialisation in scientific KIS (analytical), technical-professional KIS (synthetic), and artistic-cultural KIS (symbolic) seeks to capture the effects of location economies (X1–X3). Co-location with innovative manufacturing industries accounts for their relationship with innovation dynamics (X4–X5). Urbanisation economies and proximity to knowledge networks are proxied by distance to the political centre and green space density as proxies for urban amenities (X6, X7), respectively.

Table 1.

Description of the independent variables of the logistic model.

Name	Code	Label	Coding
A. Innovation dynamics	X1	Medium-tech manufacturing	Yes = 1
			No = 0
	X2	High-tech manufacturing	Yes = 1
			No = 0
B. Localisation economies	X3	Analytical KIS	Yes = 1
			No = 0
	X4	Synthetic KIS	Yes = 1
			No = 0
	X5	Symbolic KIS	Yes = 1
			No = 0
C. Urbanisation economies	X6	Distance to the city centre	Km from the political centre
D. Urban amenities	X7	Green space density	Hectares

Source: Own elaboration.

The methodological strategy adopted to estimate the GWLR is based on a prior selection of variables from non-spatial univariate and multivariate logistics models. The statistical significance of the independent variables in these models is assessed, retaining only those significant (p < 0.05) in both exercises. This approach follows the methodological sequence suggested by Fotheringham et al. (2002), who recommend first estimating a global model as a baseline, to evaluate spatial heterogeneity and justify the use of the local model. This strategy, also applied in studies such as Zhou et al. (2016), may limit the detection of local effects in globally non-significant variables but prioritises statistical parsimony and robustness. Nevertheless, it is recognised that an exploratory extension of the analysis could reveal complementary spatial dynamics. The logistic models and the GWLR are expressed as follows:

\begin{matrix} l o g (\frac{P (y_{i j} = 1)}{1 - P (y_{i j} = 1)}) = β_{0} + \sum_{k = 1}^{K} β_{k} x_{j k}; \\ i = I (i) \end{matrix}

(1)

\begin{matrix} l o g (\frac{P (y_{i j} = 1)}{1 - P (y_{i j} = 1)}) = β_{0} (u_{j}, v_{j}) \\ + \sum_{k = 1}^{K} β_{k} (u_{j}, v_{j}) x_{j k}; i = I (i) \end{matrix}

(2)

Where, equation (1) shows the global logistic model (Agresti, 2002), and allows to obtain the probability that BGA j is specialised in innovation i, as a function of the independent variables defined in Table 1. Equation (2) shows the construction of the GWLR model, where β_k (u_j, v_j) are the local coefficients of the GWLR for each BGA j, and where u_j and v_j are the geographical position of the centroid of BGA j.

As with all spatial models, the GWLR is sensitive to the determination of spatial weights, that are used for parameter estimation. In this regard, two analytical meta-parameters are particularly important, as they influence both estimation and interpretation of results. The first meta-parameter is the selection of the bandwidth. To determine the optimal bandwidth size, an automatic parameter detection procedure was implemented, which optimises model performance through cross-validation (Fotheringham et al., 2002; Lu et al., 2025).

The second meta-parameter is the choice of kernel function. For this study, a bi-square kernel density was selected for two reasons. Firstly, it allows continuous values to be assigned to the geographical weights of each spatial unit, thereby capturing local effects across segments of the MCMA. Secondly, unlike the Gaussian kernel, which assigns values greater than zero to all observational units, the bi-square kernel clearly delimits the area of influence, thereby enabling a sharper focus on local effects.

The logistic model parameters are presented as odds ratios (ORs), allowing a straightforward interpretation of the probability that Y_ij = 1. The optimal bandwidth was estimated, and the GWLR model was fitted using GWR 4.0 software, and the ORs and p-values are visualised using choropleth maps. The empirical results of this work are presented in the following section.

Results: Intra-urban geography of innovation in KIS in Mexico City

Descriptions and spatial distribution

In the MCMA, 15.60% (548) of the surveyed KIS firms reported having developed innovation. Of this group, 68.24% indicated having done so cooperatively, and 31.75% indicated having done so non-cooperatively. These firms are mainly concentrated in 3.45% of the BGAs of the MCMA, where they achieve a LQ above 1.00, indicating territorial specialisation. BGAs specialising in cooperative and non-cooperative innovation represent 2.49% and 0.96% of the metropolitan total, respectively.

The analysis of the spatial distribution of these BGAs, using the NNI, shows that both innovation modalities do not have a random pattern within the metropolitan area. BGAs specialised in cooperative innovation have a more intense spatial concentration (NNI = 0.80). This concentration occurs mainly outside the central polygon formed by the four main boroughs of the MCMA (or Central Business District (CBD)). Clusters are predominantly located around Reforma–Insurgentes Avenues and also appear in other BGAs outside the metropolitan centre (Figure 3(a)). Non-cooperative innovation shows a less accentuated clustering pattern (NNI = 0.94). Non-cooperative innovation is also concentrated in Reforma–Insurgentes Avenues, as well as in areas linked to higher education institutions and industrial corridors located in relatively central and peripheral areas of the metropolis (Figure 3(b)).

Figure 3.

Spatial distribution of BGAs specialised in ‘innovation in KIS’. (a) Cooperative innovation and (b) Non-cooperative innovation.

Global logistic models indicate that nearly all variables are statistically significant in explaining specialisation in each mode of innovation, except for green area density, which reflects local amenities (Table 2). The multivariate models show a different pattern. For cooperative innovation, not all variables remain significant. The significance of X1 suggests that the presence of medium-technology industries with innovation increases the likelihood of BGA specialisation in cooperative innovation by 2.56 times compared to their absence. No interactions were found with high-tech industries. Regarding localisation economies, analytical-KIS raise the probability of cooperative innovation by 6.16 times, and symbolic-KIS by 3.78 times.

Table 2.

MCMA: Global logistics models of innovation in KIS.

		Cooperative				Non-cooperative
		Univariate		Multivariate^a		Univariate		Multivariate^a
		Coef.	OR	Coef.	OR	Coef.	OR	Coef.	OR
	Cons	−3.66	0.03***	−5.59	0.00***	−4.64	0.01***	−4.89	0.01***
A. Innovation dynamics	X1	1.72	5.59***	0.94	2.56**	1.22	3.38*	0.43	1.54
	X2	1.78	5.92***	0.65	1.91	1.75	5.75*	0.79	2.2
B. Localisation economies	X3	2.62	13.75***	1.82	6.16*	2.35	10.48***	1.78	5.93*
	X4	3.78	43.73***	1.94	6.98	1.65	5.22**	0.29	1.34
	X5	2.48	11.95***	1.33	3.78*	1.64	5.13***	0.77	2.17
C. Urbanisation economies	X6	0.00	1.00***	0.00	1.00***	0.00	1.00***	0.00	1.00*
D. Urban amenities	X7	5.5	240.6*	−2.31	0.1	2.14	8.52	−6.48	0.00
N	5728					5728
Y_i = 1	143					55
AIC	1000.39					547.76
BIC	1053.62					600.99
Log likelihood	−492.2					−265.88
Deviance	984.4					531.77
Pseudo R²	0.264					0.143

Source: Own elaboration with information from INEGI (2024).

Note: The test for the absence of multicollinearity is performed for all variables in the multivariate models with the VIF measure. No variable with VIF greater than 5.00.

OR: odds ratio; BIC: Bayesian inference criterion; AIC, Akaike information criterion; VIF: Variance inflation factor.

These are results of nested models. The other results can be requested from the authors.

***

p < 0.001. **p < 0.01. *p < 0.05.

Only analytical-KIS are significant in the non-cooperative mode, increasing the probability of innovation by 5.93 times. The statistical non-significance of innovation in the two types of manufacturing industry is striking, which may suggest the innovation processes are isolated from these activities. The specialisation of the other KIS (synthetic (X4) and symbolic (X5)) also shows no significance with non-cooperation in innovation, suggesting an absence of intra-urban co-location for knowledge exchange in the MCMA (Table 2). Urbanisation economies (X6), measured by the distance to the central area of the MCMA, are significant in both innovation modes, though with marginal coefficient, reflecting a certain level of metropolitan accessibility to them.

The urban amenities variable (X7), measured by the density of green areas, loses statistical significance for the cooperative mode once included in the multivariate model. This suggests that the chosen indicator may not fully capture the types or diversity of amenities relevant to this innovation mode. Symbolic-KIS and urbanisation economies, both significant, likely encompass a broader range of urban attributes beyond green spaces. This points to the limitations of the current operationalisation and the need to move toward more refined ways of operationalising urban amenities, that would allow capturing spatial qualities and broader urban interactions (Duvivier and Polèse, 2018). The next section examines the significant variables in both models, incorporating their spatial dimension through GWLR estimation for each innovation mode.

GWLR: Multivariate models and local coefficients

The GWLR shows an improvement in its explanatory power compared to global regressions, confirming the relevance of incorporating the spatial dimension in this analysis. This improvement is supported by the reduction in Akaike information criterion (AIC) values between the global and geographically weighted models, indicating a better model fit (Fotheringham et al., 2002). Specifically, for cooperative and non-cooperative innovation, AIC values decrease from 1000.40 to 928.40 and from 547.77 to 541.89, respectively (Tables 2 and 3).

Table 3.

MCMA: GWLR for innovation in KIS in the intra-metropolitan area.

		Cooperative						Non-cooperative
		Min		Average		Max		Min		Average		Max
		Coef.	OR	Coef.	OR	Coef.	OR	Coef.	OR	Coef.	OR	Coef.	OR
	Cons	−9.42	0.00	−4.96	0.01	−2.52	0.08	−6.05	0.00	−4.36	0.01	−3.36	0.04
A. Innovation dynamics	X1	−0.83	0.44	1.29	3.61	4.79	120.72	—	—	—	—	—	—
	X2	—	—	—	—	—	—	—	—	—	—	—	—
B. Localisation economies	X3	1.3	3.66	2.07	7.96	5.86	350.84	1.39	4.03	1.99	7.37	3.13	22.92
	X4	—	—	—	—	—	—	—	—	—	—	—	—
	X5	−4.79	0.01	1.54	4.67	5.76	317.25	—	—	—	—	—	—
C. Urbanisation economies	X6	0.00	1.00	0.00	1.00	0.00	1.00	0.00	1.00	0.00	1.00	0.00	1.00
D. Urban Amenities	X7	—	—	—	—	—	—	—	—	—	—	—	—
Best bandwidth size		2532						3751
AIC		928.55						541.92
BIC/MDL		1061.5						601.6
PDE		0.34						0.16

Source: Own elaboration with information from INEGI (2024).

Note:‘—’ variables not incorporated into the model, as they were non-significant in their overall evaluations.

PDE: percent deviance explained; AIC: Akaike information criterion; ML: minimum description length; BIC: Bayesian inference criterion.

The GWLR model identified optimal bandwidths of 2535 BGAs for the cooperative innovation mode, and 3571 BGAs for the non-cooperative mode. This is equivalent to comparing each unit with approximately 44% and 62% of the BGAs that make up the MCMA, respectively (Table 3). This bandwidth size, together with the application of the adaptive bi-square kernel, models the probability of innovation by favouring the effects of relative proximity over those of more distant units. This modelling choice is consistent with the heterogeneous spatial structure of the MCMA, discussed in the methodology section.

The coefficients of the significant local variables in the cooperative and non-cooperative innovation models increase, compared to those obtained with the global model. In cooperative innovation, the OR of innovation in the medium-tech industry is 3.61 and reaches a maximum value of 120.7 times more likely to explain specialisation in that type of innovation. The analytical and symbolic-KIS also increase, on average, the OR reported in the global model. In the non-cooperative innovation model, the average of the OR of the KIS-analytical does the same (Table 3).

Spatial patterns of explanatory variables for innovation modes

The independent variables related to cooperative innovation in KIS, ordered according to their coefficient values: analytical-KIS, symbolic-KIS, medium-tech innovative manufacturing, and distance to the city centre. Figure 4 displays the direction and weight of each variable across the metropolitan area.

Figure 4.

GWLR of ‘innovation in KIS’ mode in the MCMA. (a–d) Cooperative innovation; (e and f) Non-cooperative innovation. CBD (Central Business District) = Four main boroughs of the MCMA. Complete GWLR results can be provided by the authors upon request.

The analytical-KIS have a positive effect on the probability of cooperative innovation. Their highest OR is in the unconsolidated periphery in the north of the city. Within the consolidated urban structure, this variable has a smaller and practically homogeneous effect. This last characteristic suggests being the result of a dispersed spatial distribution of higher education and scientific activities within the urban area (Figure 4(a)).

The symbolic-KIS have a positive and negative impact on the dependent variable. Almost all (99.3%) of the BGAs that are specialised in cooperative innovation are in the area with a positive effect. The ORs with the highest and most significant values coincide with the cluster formed on Reforma–Insurgentes Avenues (Figure 4(b)). The negative values in this variable invite us to reflect on the supposed positive effect that is generally attributed to artistic-cultural activities on innovation (Polèse, 2011: 1820).

The results for medium-tech manufacturing suggest that they reflect the location of these industries in the MCMA, which is mainly centred on the highways that connect the city with the north of the country. This is the case of the industrial zones of Tlanepantla and Cuamatla (Figure 4(c)). This characteristic would explain why the areas with the highest levels of significance, which cover 55.15% of the BGA observations, and those where the ORs are accentuated, are in the vicinity of these industrial areas (Figure 4(d)). A feature of this variable is that its effect is practically limited to these industrial zones, which suggests innovation synergies that require spatial proximity.

Compared to cooperative innovation, the local effects of the explanatory variables for non-cooperative innovation in KIS are more spatially concentrated in northern peripheral areas and show less spatial variability. Only two variables are significant; they operate in opposite directions and with different weights through the MCMA. Analytical-KIS have the highest coefficients, which are positive and statistically significant across the full set of BGAs specialised in non-cooperative innovation. Their ORs are accentuated in the northern peripheral area of the city (Figure 4(e)). This suggests that, in this area, employment concentration in higher education and R&D increases the likelihood of non-cooperative innovation.

Distance to the city centre consistently shows a negative effect across the entire sample, with only marginal variations in the OR, meaning that the greater the distance the lower the probability of non-cooperative innovation development. However, the weight of this variable is not spatially uniform: ORs are more accentuated in the north of the consolidated urban area and decrease progressively towards the southeast of the MCMA, where distance becomes statistically significant (Figure 4(f)). The next section presents the analysis and discussion of these results.

Discussion and final reflections

The objective of this article is to distinguish innovation in KIS in terms of its cooperative and non-cooperative mode, and to question whether the former tends to concentrate in the central area and the latter in the periphery of the MCMA. The empirical results indicate that innovation activity in KIS is not random in intra-urban space. Cooperative innovation in KIS exhibits greater spatial concentration than non-cooperative innovation. Four key spatial features of the intra-urban distribution of innovation in KIS in the MCMA are identified, two aligning with prior studies, and two that are novel to this research.

Firstly, the concentration of BGAs specialising in cooperative innovation along the Reforma–Insurgentes Avenues supports findings from previous studies (Alvarez-Lobato et al., 2023). However, the present study reveals that this area is not merely a hub of employment or firms, but rather it functions as a ‘milieu of innovation’, where the highest proportion of KIS innovation occurs collaboratively. This suggests that the corridor facilitates intense formal and informal exchanges of tacit knowledge (Hutton, 2004; Katz and Wagner, 2014), underpinned by agglomeration economies extending from the city’s central area.

Secondly, non-collaborative innovation in the periphery aligns with cases identified by Suarez-Villa and Walrod (1997) and Shearmur (2012). As those authors suggest, the peripheral location of such KIS can be attributed to competitive strategies or internal R&D capabilities. The positive correlation between this mode of innovation and employment in analytical KIS indicates a need for spatial proximity to higher education centres and R&D institutions, leveraging localisation economies.

Thirdly, a novel finding of this research is the presence of non-cooperative innovation in the city’s central milieu (Reforma–Insurgentes Avenues). This suggests that innovation is not solely reliant on internal resources, but also on external factors that are mediated by absorption and management capacities (Bathelt et al., 2004; Santoro et al., 2020). While internally driven, these firms remain embedded in central learning dynamics, maintaining strict control over their inputs and outputs.

Fourthly, another new contribution of this study is the identification of a dispersed pattern of cooperative innovation beyond the central area. This distribution correlates with employment clusters in analytical KIS (science), symbolic KIS (art and culture) and innovative medium-tech manufacturing in peripheral areas. Such co-location fosters synergies between industries, driving cooperative innovation.

Statistical analysis of the global and local effects of independent variables shows that cooperative innovation is associated with innovative medium-technology industries and employment specialisation in analytical and symbolic KIS. In contrast, non-cooperative innovation is explained only by analytical KIS specialisation. Co-location with high-tech firms or urban amenities is not significant for either mode. The former may reflect more discreet innovation practices, while the latter likely results from a narrow operationalisation of amenities limited to green space. A broader definition, indirectly captured by symbolic KIS and urbanisation economies, may better reflect these effects. Although beyond the scope of this study, this issue remains open for future research.

A further limitation concerns the data, which capture innovation only among firms physically located within the BGAs. The analysis therefore relies on declared intra-firm innovation and inter-firm cooperation, excluding informal innovation dynamics, which should be examined in future studies.

These findings partially confirm the hypothesis guiding this study. They extend prior evidence by revealing additional concentrations of cooperative innovation in the urban periphery, and non-cooperative innovation in the central area. The results underscore the importance of agglomeration economies and innovation dynamics, which vary spatially in shaping the location of cooperative and non-cooperative innovation in intra-urban areas. KIS utilise these economies and dynamics both locally (through spatial proximity) and non-locally (over distance), regardless of the innovation mode. This aligns with Parr’s (2002) assertion that intra-urban features, such as accessibility and transport, enable extensive use of agglomeration economies and innovation dynamics, albeit with varying purposes and costs, depending on location.

The findings of this article add to the perspective of authors who question the idea of pigeonholing innovation into a single space. While such criticisms have been formulated mainly at the interurban level (Castaldi, 2023), this study provides evidence that KIS generate innovation through extensive use of space under differentiated logics within the MCMA. These results caution against reinforcing the notion of privileged spaces for certain types of innovation (Hutton, 2004; Katz and Wagner, 2014), which may accentuate socio-spatial imbalances (Breznitz, 2021). Instead, they point to the need for strategies that recognise the diversity of spaces where innovation takes place (Duvivier and Polèse, 2018). In this way, central and peripheral areas can be configured as ‘milieus of innovation’ that drive economic development and contribute to reducing the spatial inequalities that characterise large cities in the Global South (Graizbord et al., 2003; Ruiz-Porras and Zagaceta-García, 2016).

In summary, the findings of this study contribute to understanding the intra-urban geography of innovation in KIS. However, the study presents some limitations that also open up research opportunities. One of these concerns the resolution of the data used. The available data do not allow for distinguishing certain characteristics of innovation, such as its orientation (technological or non-technological) or its degree of novelty (radical or incremental), nor for capturing innovation processes that occur beyond formally registered agents. Addressing these limitations could help expand the analysis of intra-urban location patterns and explore their potential links with informal sectors, a structural component of Latin American economies. Advancing this research agenda is relevant not only for public policy but also considering the rapid development of Artificial Intelligence (AI), which poses new challenges and opportunities for urban economies at multiple scales.

Footnotes

Acknowledgements

The authors thank the anonymous reviewers for their valuable comments and suggestions, which helped improve the final version of the article. We also acknowledge the support provided by the INEGI Microdata Laboratory (Aguascalientes), particularly for assistance during data processing. Assistance in preparing the cartography for this article was provided by urbanism students Miguel Ángel Macias, Ximena Flores, Sahery Reyes and Raúl Nieves.

ORCID iDs

Luis Enrique Santiago

Alejandro Sánchez Zárate

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The preparation of this article was supported financially by the Autonomous University of Aguascalientes (internal project: PIU-22-3).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Agresti

(2002) Categorical Data Analysis, 2nd edn. Wiley.

Alvarez-Lobato

Cabrera-Pereyra

Garrocho

(2023) The clustering of knowledge-intensive business services in Greater Mexico City, 2010 through 2020. Papers in Applied Geography 9(2): 193–213.

Arauzo-Carod

J-M

Coll-Martínez

Méndez-Ortega

(2017) Aglomeración de sectores intensivos en conocimiento: Una aproximación intraurbana. Papeles de Economía Española 153: 110–126.

Aslesen

Freel

(2012) Industrial knowledge bases as drivers of open innovation? Industry and Innovation 19(7): 563–584.

Barras

(1990) Interactive innovation in financial and business services: The vanguard of the service revolution. Research Policy 19: 215–237.

Barrett

Davidson

Prabhu

, et al. (2015) Service innovation in the digital age: Key contributions and future directions. MIS Quarterly 39(1): 135–154.

Bathelt

Turi

(2011) Local, global and virtual buzz: The importance of face-to-face contact in economic interaction and possibilities to go beyond. Geoforum 42(5): 520–529.

Bathelt

Malmberg

Maskell

(2004) Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography 28(1): 31–56.

Breznitz

(2021) Innovation in Real Places: Strategies for Prosperity in an Unforgiving World. Oxford University Press.

10.

Camagni

(2003) Incertidumbre, capital social y desarrollo local: Enseñanzas para una gobernabilidad sostenible del territorio. Investigaciones Regionales 2: 31–57.

11.

Castaldi

(2023) The geography of urban innovation beyond patents only: New evidence on large and secondary cities in the United States. Urban Studies 61(7): 1248–1272.

12.

Chesbrough

(2003) The era of open innovation. MIT Sloan Management Review 44(3): 35–41.

13.

Corrocher

Cusmano

(2014) The ‘KIBS Engine’ of regional innovation systems: Empirical evidence from European regions. Regional Studies 48(7): 1212–1226.

14.

Corrocher

Cusmano

Morrison

(2009) Modes of innovation in knowledge-intensive business services evidence from Lombardy. Journal of Evolutionary Economics 19(2): 173–196.

15.

Daniels

Bryson

(2002) Manufacturing services and servicing manufacturing: Knowledge-based cities and changing forms of production. Urban Studies 39(5–6): 977–991.

16.

Dankbaar

(2007) Global sourcing and innovation: The consequences of losing both organizational and geographical proximity. European Planning Studies 15(2): 271–288.

17.

Den Hertog

(2000) Knowledge-intensive business services as co-producers of innovation. International Journal of Innovation Management 4(4): 491–528.

18.

Duvivier

Polèse

(2018) The great urban techno shift: Are central neighbourhoods the next silicon valleys? Evidence from three Canadian metropolitan areas. Papers in Regional Science 97(4): 1083–1111.

19.

Enkel

Gassmann

Chesbrough

(2009) Open R&D and open innovation: Exploring the phenomenon. R&D Management 39(4): 311–316.

20.

Escamilla

Cos

Cárdenas

(2016) Contesting Mexico City’s alleged polycentric condition through a centrality-mixed land-use composite index. Urban Studies 53(11): 2380–2396.

21.

Evangelista

Sirilli

(1995) Measuring innovation in services. Research Evaluation 5(3): 207–215.

22.

Fitjar

Rodríguez-Pose

(2017) Nothing is in the air. Growth and Change 48(1): 22–39.

23.

Fotheringham

Brunsdon

Charlton

(2002) Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. Wiley.

24.

Freel

(2006) Patterns of technological innovation in knowledge-intensive business services. Industry and Innovation 13(3): 335–358.

25.

Gallouj

Windrum

(2009) Services and services innovation. Journal of Evolutionary Economics 19(2): 141–148.

26.

Graizbord

Rowland

Aguilar

(2003) Mexico City as a peripheral global player: The two sides of the coin. The Annals of Regional Science 37: 501–518.

27.

Hall

(2009) Looking backward, looking forward: The city region of the mid-21st century. Regional Studies 43(6): 803–817.

28.

Hervas-Oliver

Sempere-Ripoll

Boronat-Moll

(2021) Technological innovation typologies and open innovation in SMEs: Beyond internal and external sources of knowledge. Technological Forecasting and Social Change 162: 120338.

29.

Hutton

(2004) The new economy of the inner city. Cities 21(2): 89–108.

30.

INEGI (2020) Censos Económicos 2019: Metodología. Instituto Nacional de Estadística y Geografía. Available at: https://www.inegi.org.mx/contenidos/productos/prod_serv/contenidos/espanol/bvinegi/productos/nueva_estruc/702825196530.pdf (accessed 19 February 2026).

31.

INEGI (2024) Economic Census 2019 microdata. Microdata Laboratory, National Institute of Statistics and Geography (INEGI). Access granted through research project 727.

32.

Katz

Wagner

(2014) The Rise of Innovation Districts: A New Geography of Innovation in America. Brookings Institution. Available at: https://www.brookings.edu/wp-content/uploads/2016/07/InnovationDistricts1.pdf (accessed 18 October 2025).

33.

Kekezi

Klaesson

(2020) Agglomeration and innovation of knowledge intensive business services. Industry and Innovation 27(5): 538–561.

34.

Lee

Miozzo

(2019) Which types of knowledge-intensive business services firms collaborate with universities for innovation? Research Policy 48(7): 1633–1646.

35.

Yang

, et al. (2025) GWmodelS: A standalone software to train geographically weighted models. Geo-Spatial Information Science 28(2): 648–670.

36.

Malecki

(2021) The geography of innovation. In: Fischer

Nijkamp

(eds) Handbook of Regional Science. Springer, pp. 819–834.

37.

Méndez-Ortega

Micek

Malochleb

(2022) How do coworking spaces coagglomerate with service industries? The tale of three European cities. Cities 130: 103875.

38.

Miles

(2005) Innovation in services. In: Fagerberg

Mowery

(eds) The Oxford Handbook of Innovation. Oxford University Press, pp. 433–458.

39.

Miles

Belousova

Chichkanov

(2018) Knowledge intensive business services: Ambiguities and continuities. Foresight 20(1): 1–26.

40.

Mina

Bascavusoglu-Moreau

Hughes

(2014) Open service innovation and the firm’s search for external knowledge. Research Policy 43(5): 853–866.

41.

Mubarak

Petraite

(2020) Industry 4.0 technologies, digital trust and technological orientation: What matters in open innovation? Technological Forecasting and Social Change 161: 120332.

42.

Parr

(2002) Agglomeration economies: Ambiguities and confusions. Environment and Planning A 34(4): 717–731.

43.

Polèse

(2011) The arts and local economic development: Can a strong arts presence uplift local economies? A study of 135 Canadian cities. Urban Studies 49(8): 1811–1835.

44.

Rammer

Kinne

Blind

(2020) Knowledge proximity and firm innovation: A microgeographic analysis for Berlin. Urban Studies 57(5): 996–1014.

45.

Roche

(2020) Taking innovation to the streets: Microgeography, physical structure, and innovation. Review of Economics and Statistics 102(5): 912–928.

46.

Romero de Ávila Serrano

(2019) The intrametropolitan geography of knowledge-intensive business services (KIBS): A comparative analysis of six European and U.S. city-regions. Economic Development Quarterly 33(4): 279–295.

47.

Ruiz-Porras

Zagaceta-García

(2016) La innovación en las empresas mexicanas de servicios: Un análisis a nivel de sectores, subsectores y ramas económicas. Análisis Económicos 31(76): 29–45.

48.

Santiago

(2020) The industries of the future in Mexico: Local and non-local effects in the localization of “knowledge-intensive services.” Growth and Change 51(2): 584–606.

49.

Santiago

Sánchez-Zárate

(2023) Geografía intraurbana del empleo en ‘servicios intensivos en conocimiento’ (SIC) con alta calificación en el Área Metropolitana de la Ciudad de México. Investigaciones Geográficas 112: e60777.

50.

Santoro

Bresciani

Papa

(2020) Collaborative modes with cultural and creative industries and innovation performance: The moderating role of heterogeneous sources of knowledge and absorptive capacity. Technovation 92–93: 102040.

51.

SEDATU, CONAPO and INEGI (2018) Sistema urbano nacional 2018. Secretaría de Desarrollo Agrario, Territorial y Urbano. Available at: https://www.gob.mx/conapo/documentos/sistema-urbano-nacional-2018 (accessed 28 April 2022).

52.

Shearmur

(2012) The geography of intrametropolitan KIBS innovation: Distinguishing agglomeration economies from innovation dynamics. Urban Studies 49(11): 2331–2356.

53.

Shearmur

Alvergne

(2002) Intrametropolitan patterns of high-order business service location: A comparative study of seventeen sectors in Ile-de-France. Urban Studies 39(7): 1143–1163.

54.

Silva

Sousa

Moreira

, et al. (2011) Innovation activities in the service sector: Empirical evidence from Portuguese firms. Journal of Knowledge Management, Economics and Information Technology 1(6): 1–17.

55.

Solis

Karimi

Garcia

, et al. (2021) Knowledge economy clustering at the intrametropolitan level: Evidence from Madrid. Journal of the Knowledge Economy 13: 1268–1299.

56.

Suarez-Villa

Walrod

(1997) Operational strategy, R&D and intra-metropolitan clustering in a polycentric structure: The advanced electronics industries of the Los Angeles Basin. Urban Studies 34(9): 1343–1380.

57.

Tether

(2005) Do services innovate (differently)? Insights from the European Innobarometer survey. Industry and Innovation 12(2): 153–184.

58.

Torre

(2008) On the role played by temporary geographical proximity in knowledge transmission. Regional Studies 42(6): 869–889.

59.

Wood

(2005) A service-informed approach to regional innovation – or adaptation? The Service Industries Journal 25(4): 429–445.

60.

Ženka

Slach

Ivan

(2020) Spatial patterns of knowledge-intensive business services in cities of various sizes, morphologies and economies. Sustainability 12(5): 1–19.

61.

Zhou

Wang

Yang

, et al. (2016) Geographical variations of risk factors associated with HCV infection in drug users in southwestern China. Epidemiology and Infection 144(6): 1291–1300.