Local context of local government participation in the innovation networks: Evidence from Poland

Innovations¹ are of crucial importance because they are the driving force of economic development [Schumpeter, 1943] and are particularly important in the present turbulent times of the COVID-19 pandemic. However, over time, there has been a change in attitude toward innovation, and nowadays many scholars acknowledge the importance of an established innovation ecosystem as a crucial factor for successful value creation through innovation [Su et al., 2018]. An innovation ecosystem² may transform knowledge into additional value, and create synergy effects or power collaboration and co-creation [Durst and Poutanen, 2013]. An innovation ecosystem may consist of a network of central, regional, or local government authorities, financial institutions, customers, intermediaries, universities, suppliers, and/or competitors [Su et al., 2018]. The creation of innovation ecosystems can be facilitated by innovation networks established between at least three independent organizations connected by a set of long-term cooperation ties seeking innovation [Batterink et al., 2010, de Vasconcelos Gomes et al., 2018]. Innovation networks are locally rooted thanks to knowledge spillover and social embeddedness [Jaffe et al., 1993]; thus, the local context is growing significantly in importance [Martin and Rypestøl, 2018], in particular in times of the COVID-19 pandemic, which has affected global value chains and reinforced local cooperation. Local government units (LGUs) enter innovation networks to exchange ideas on local innovation and/or to foster local innovation initiatives.

Innovation networks may help to expand innovation capacity [Rampersad et al., 2010] or accelerate the creation of value through innovation [Barsh et al., 2008] at the local level. Moreover, Kim and Yoon [2015], based on a survey of 1,576 employees in the Seoul Metropolitan Government, argue that LGU participation in innovation networks may enhance local creativity and foster a local culture of innovation. However, LGUs differ not only due to factors such as geographical location, territorial size, population density, history, culture, local economy, natural resources, etc. [Koellinger and Thurik, 2012] but also due to a self-enforcement mechanism [Godlewska and Morawska, 2020] or good governance [Rodríguez-Pose and Garcilazo, 2015]. Having in mind all the advantages linked to participation of LGUs in innovation networks, we examined why only a few LGUs participate in such networks, and whether local context matters for the participation of LGUs in innovation networks.

The local contexts of innovative behaviors of entrepreneurs are well-recognized, but the determinants of LGU participation in innovation networks have not been thoroughly investigated by scholars. Although studies of innovative activities of LGUs do exist, there have been many calls for further studies to explore why LGUs engage in local value creation or participation in local innovation networks [Damanpour and Schneider, 2009; Gonzalez et al., 2013; Skog, 2016].

Previous research focused on: (i) regional innovation networks [Cooke, 1996; Pekkarinen and Harmaakorpi, 2006; Schoonmaker and Carayannis, 2010]; (ii) learning region in the context of institutions, innovation, and regional renewal [Morgan, 1997]; (iii) LGU manager characteristics in the context of the adoption of innovation [Damanpour and Schneider, 2009]; or (iv) typology or strategic profile of innovative activities of LGUs [Gonzalez et al., 2013].

This study responds to these calls by addressing a problematic local context understood by authors as the following set of factors measured at LGUs’ level: (i) institutional; (ii) economic; (iii) geographical; (iv) cultural; (v) demographic; (vi) administrative; (vii) political; and (viii) managerial. The selection of factors was formulated based on a literature review [North, 1990; Hodgson, 2006; Damanpour and Schneider, 2009; Reginato et al., 2011; Gonzalez et al., 2013; Rodríguez-Pose and Garcilazo, 2015; Ketterer and Rodríguez-Pose, 2016; Capello and Perucca, 2018]. Therefore, the purpose of the present research was to ascertain whether local context matters for LGUs’ participation in innovation networks.

The empirical analysis is based on data collected through an electronic survey sent in April 2020 to all the Polish LGUs (all the municipalities and cities on district rights). The local context analysis was based on data published by Statistics Poland in the Local Data Bank for 2019. The logistic regression analysis was performed to predict a binary outcome such as LGUs’ participation in innovation networks. A logistic regression model predicts the LGUs’ participation in innovation networks (dependent variable) by analyzing the relationship with the set of defined local factors (independent variables) such as: previous cooperation with special local institutions (SLI), the number of business environment institutions (BEI) located in a given LGU, the level of economic development, debt per capita, population, type of LGU, and gender, age, and education level of the LGU manager as well as his or her political independence.

This paper is organized as follows. The first section presents the findings of a literature review of local context factors, the second section describes the research approach, and the third section discusses the results. The summary highlights the most important conclusions of the study.

According to the new growth theory, new economic geography, the new economics of innovation, the concept of an innovation ecosystem, or new institutional economics, innovations are key factors in each national economy [Acs and Varga, 2002]. Institutional theory may give an insight into the innovative activities of LGUs. Rodrik et al. [2004] highlight that the quality of institutions such as LGUs matters more than development factors, such as geography or trade, for determining national or regional growth. This may be the case also for successful value creation through innovation or innovation networks. Freeman [1994] has noted that the innovativeness of regions is determined by several institutions and organizations, which, together with enterprises, form a relatively separate territorial system, the essence of which is interactive network connections.

An entrepreneurial or innovation ecosystem may be established at national [Wessner, 2005, Godlewska, 2021], regional [Cooke, 1996; Pekkarinen and Harmaakorpi, 2006; Schoonmaker and Carayannis, 2010], or local [Derlukiewicz et al., 2021, Pilewicz et al., 2021] levels. The significance of the local context factor for successful value creation through innovation or innovation networks has been repeatedly highlighted in literature. What is missing, however, is a discussion of which local context factors influence the LGUs’ participation in innovation networks. As there is neither a list of innovation networks nor data on participation in innovation networks, authors made an attempt to investigate the scale of such initiatives in Poland. Therefore, a bottom-up approach was needed, which indicated that a local-level analysis would be appropriate. Brenner and Broekel [2011] argue that regional or local level of empirical investigation is justified if the innovation network has a strong regional or local dimension or there are no firm-level data.

Entrepreneurship or innovation literature highlights that context³is usually taken for granted by scholars [Johns, 2006]. However, there is no consensus on what constitutes context in terms of entrepreneurship and innovation [Zahra et al., 2014]. Moreover, scholars argue that there is a need for additional studies of the context of entrepreneurship or innovation ecosystems [Zahra and Wright, 2012, Derlukiewicz et al., 2021]. This context is a crucial topic for entrepreneurship and innovation researchers, as it may help them to understand how different contexts may enhance or hamper entrepreneurial and innovation processes or affect outcomes at national, regional, or local levels [Zahra, 2007; Hindle, 2010; Welter, 2011; Zahra and Wright, 2012; Müller and Korsgaard 2018]. Context may be an asset or a liability [Welter, 2011]. Context matters [Welter, 2011] because it regulates the outcomes of the entrepreneurial and innovation ecosystems [Acs et al., 2014]. Contexts such as temporal, industry, spatial, institutional, economic, social, and organizational or ownership and governance contexts [Hindle, 2010; Welter, 2011; Zahra et al., 2014; Müller and Korsgaard, 2018] may frame and shape the entrepreneurial or innovative activities of local government. The contextual approach may investigate why not all local government authorities participate in innovation networks [Godlewska et al., 2022].

The theory of innovation networks or concept of innovation ecosystem highlights the key importance of local context [Damanpour and Schneider, 2009; Gonzalez et al., 2013; Skog, 2016] and LGUs to the study of the innovation ecosystem at a local level [Gorzelany-Dziadkowiec et al., 2019]. For this reason, in response to calls for research [Mazzucato, 2016] that aims to enhance theories of local context and innovation networks by focusing on the local determinants of LGUs’ participation in the innovation networks and to better understand the state of play or investigate research gaps, the authors chose a six-stage approach (Figure 1).

Figure 1.

Despite the widely held belief that local context matters when it comes to innovation networks, no studies have provided evidence of whether local context matters for LGUs’ participation in the innovation networks. Addressing this perceived gap in the literature was, accordingly, the main objective of this study.

Therefore, in this paper, based on the most recent studies by Ketterer and Rodríguez-Pose [2016], Drucker [2015], Phelps [2013], Gonzalez et al. [2013], Zeng et al. [2010], Damanpour and Schneirder [2009], Laukkanen and Niittykangas [2003], Oughton et al. [2002], Sabel [1995], and Lundvall [1995], the following hypothesis was introduced:

H1: The following set of factors, namely: (i) institutional (cooperation between LGUs and special local institutions or tiers of business environment institutions active on LGU territory); (ii) geographical (the seat of an LGU in a particular region, i.e., one of 16 voivodeships); (iii) economic (the tier of entrepreneurs active in the LGU territory or tier of total income, expenditure, or debt per capita); (iv) demographic (the number of inhabitants or local population density); (v) cultural (the amount of cultural infrastructure); (vi) administrative (the type of LGU); (vii) managerial (age, gender, or education of LGU mayors); and (viii) political (political support for the local mayor/city mayor), influence the LGUs’ participation in the innovation networks.

Aars and Fimreite [2005] or Gebauer et al. [2005] highlight that LGUs are members of innovation networks in Norway or Germany. In Poland, there is no available data on LGU participation in the innovation networks. This knowledge is important, as LGUs are a key element of an innovation or entrepreneurial ecosystem. That is why the authors have chosen Polish LGUs for empirical investigation.

Empirical research into Polish LGUs based on the Brenner and Broekel [2011] methodological approach for spatial units was carried out from April 6, 2020, to April 30, 2020. Data were acquired using an electronic survey in Polish (an English translation of the survey is provided in Appendix 1), which was sent to n = 2,477 LGUs, i.e., all rural, rural-urban, and urban municipalities and cities on district rights, Local Administrative Unit (LAU) level 2. The electronic survey questionnaire (Appendix 1) was based on a literature review and tested during the pilot study. Finally, the response rate from Polish LGUs was n = 184 complete records, i.e., 7.2% of the population (requiring n = min. 182). The sample had a 7% maximum error (2SE) and a 5% materiality level. The research sample was not fully representative of all LGUs, because rural municipalities and urban-rural municipalities were under-represented by 3%, in contrast to urban municipalities and cities on district rights, which were over-represented by 3% in the sample compared to the population.

The dependent variable was LGUs’ participation in the innovation networks of all kinds. The variable (MIN) has a value of “0” if LGUs do not participate in the innovation networks, and “1” if they do.

The list of independent variables would be as follows (Table 1): (i) cooperation with SLI; (ii) BEI per 10,000 entities of the national economy (ENE); (iii) newly registered entities (NRE) per 10,000 working-age population; (iv) ENE in total; (v) debt per capita (DEB); (vi) total income per capita (TIP); (vii) total expenditure per capita (TEP); (viii) number of cultural infrastructure (NCI); (ix) population density per 1 km² (PDE); (x) number of inhabitants (NIN); (xi) age of LGU manager (ALM); (xii) gender of LGU manager (GLM); (xiii) education of LGU manager (ELM); (xiv) political ILM; and (xv) type of local government (TLG), or voivodeship (VOI). The independent variables were chosen from among the following: (i) the Local Data Bank of Statistics Poland for 2019 [https://bdl.stat.gov.pl/bdl/metadane] (for the variables ENE, NRE, BEI, DEB, TIP, TEP, NCI, PDE, and NIN); (ii) a survey (for the variables SLI, ALM, GLM, TLG, and VOI); and (iii) the National Electoral Commission for local elections in 2018 [https://pkw.gov.pl/wybory-i-referenda/wybory-samorzadowe-i-referenda-lokalne/wybory-samorzadowe-w-2018ampnbspr] (for the variable ILM).

Statistical analyses were conducted using the (1 New Orchard Road Armonk, New York 10504-1722 United States) IBM SPSS Statistic Program Version 27. The analytical framework of the paper is based on the Kleinbaum and Klein [2010] methodology of logistic regression analysis. The logistic model is based on the following logistic function: f(z)=11+e−z$$f(z) = {1 \over {1 + {{\rm{e}}^{ - z}}}}$$

To obtain the logistic model from the logistic function, we express z as the linear sum α plus β₁ times X₁ plus β₂ times X₂, and so on to β_k times X_k, where the Xs are independent variables of interest like local context factors such as institutional (X₁), economic (X₂), geographical (X₃), demographic (X₄), cultural (X₅), political (X₆), managerial (X₇), and administrative (X₈), and α and the β_i are constant terms representing unknown parameters. z=α+β1X1+β2X2+…+βkXk$$z = \alpha + {\beta _1}{X_1} + {\beta _2}{X_2} + \ldots + {\beta _k}{X_k}$$

We substitute the linear sum expression for z in the right-hand side of the formula for f(z). The probability being modeled can be denoted in terms of the following: P(D=1|X1,X2,…,Xk).$${\rm{P}}\left( {D = 1|{X_1},{X_2}, \ldots ,{X_k}} \right).$$

Further, the logistic model for the probability of the LGUs’ participation in the innovation networks can be expressed as the following: P(D=1|X1,X2,…,Xk)=11+e−(α+ΣβiXi)$${\rm{P}}\left( {D = 1|{X_1},{X_2}, \ldots ,{X_k}} \right) = {1 \over {1 + {{\rm{e}}^{ - \left( {\alpha + {\rm{\Sigma }}{\beta _i}{X_i}} \right)}}}}$$

The authors feel the need to mention that the present study was characterized by certain limitations. First, the small sample size (184 LGUs participating in the study, and 49 LGUs who were members of innovation networks) is not fully representative and thus does not allow for full generalization of the research results for the LGU population. Second, there were no representatives of the possible variety of innovation network members. The authors studied only LGUs and assumed that other network members would answer in a similar way to LGUs. Third, the authors did not investigate all local context factors that may influence LGU participation in the innovation networks, but focused only on institutional, economic, geographical, demographic, cultural, political, and administrative factors. There are other local contexts such as historical, social, technological, or infrastructural factors that may matter. Fourth, the study is partly based on the truthfulness of the answers given by the LGUs. Finally, almost 50% of all LGUs that participated in the study were from 4 out of the 16 voivodeships that Poland totally contains.

The strengths of this methodology are: (i) a cohesive research sample of LGUs that participate and do not participate in innovation networks was considered, which allows for a comparison of local context between these two groups of LGUs; (ii) an electronic survey was used that was comprised of indirect questions, filtering questions, or questions with detailed definitions; (iii) local context factors were identified based on literature review; and (iv) the logistic regression model was employed, which predicts the dependent data variable by analyzing the relationship with the independent variables.

The statistical results (Table 2) of quantitative variables showed that LGUs who participated in the study had by average on their territory 411 (from 44 to 950) BEI (per 10,000 ENE) compared to the country average of 864; 135 (from 42 to 598) NRE (per 10,000 working-age population) compared to the country average of 95; or 1,370 (from 104 to 8,870) ENE in total compared to the country average of 1,175. Moreover, by average, the TIP of these LGUs were 5,431 PLN (from 3,989 to 9,798 PLN) compared to the country average of 5,969 PLN; and the TEP was 5,416 PLN (from 3,871 to 10,351 PLN) compared to the country average of 6,051 PLN. LGUs who participated in the study by average (from -1,200 to 2,000 PLN) had no indebtedness compared to the country average of -82 PLN. The amount of cultural infrastructure such as libraries, cinemas, theaters, and cultural centers characterizing LGUs was, by average, 3 (from 0 to 59), in comparison with the country average of 1. Further, LGUs had, by average, a population density per 1 km2 of 311 (from 9 to 3,243) compared to the country average of 123, as well as 11,567 (from 200 to 99,350) inhabitants compared to the country average of 15,496.

The categorical variables like SLI, MIN, VOI, ILM, ELM, GLM, or TGL showed that only 26.6% of LGUs who participated in the study were members of innovation networks, and only 21.20% of them cooperate with SLI. In the study, 57% of LGUs were rural municipalities (compared to 61% in population), 22% were rural-urban municipalities (compared to 26% in population), 15% were urban municipalities (compared to 10% in population), and 6% were cities on district rights (compared to 3% in population). Moreover, almost 50% of all LGUs that participated in this study were from the Mazowieckie, Malopolskie, Kujawsko-pomorskie, and Lubelskie voivodeships. Surprisingly, less than 2% of LGUs’ mayors were up to 30 years old and almost 90% of LGUs’ mayors had higher education and 83% of them were male. In addition, 65% of LGUs’ mayors were politically independent.

Independent variables (quantitative ones), such as (i) BEI per 10,000 ENE (BEI); (ii) NRE per 10,000 working-age population (NRE); (iii) ENE in total; (iv) debt per capita (DEB); and (v) TEP, are correlated with a dependent variable, i.e., the number of LGUs participating in the innovation networks (see Table 3). These correlations are significant at the 0.01 level (two-tailed).

For independent variables (categorical ones), the so-called contingency table was prepared, and Cramer’s V coefficient was chosen to evaluate the relationship between the number of LGUs participating in the innovation networks and independent variable such as: (i) cooperation with SLI; (ii) ALM; (iii) GLM; (iv) ELM; (v) political ILM; (vi) TLG; or (vii) voivodeship/province (VOI). Unfortunately, only TLG and cooperation with SLI were related to the number of LGUs participating in the innovation networks (Table 4).

A logistic regression model predicts the LGUs’ participation in the innovation networks (dependent data variable) by analyzing the relationship with the set of defined local factors (independent variables). Logistic regression also estimates the probability of an event occurring, such as LGUs’ participation in the innovation networks or lack thereof, based on a given local factor dataset of independent variables. Independent variables, which were correlated with dependent variables such as (i) BEI per 10,000 ENE; (ii) NRE per 10,000 working-age population; (iii) ENE in total; (iv) debt per capita (DEB); (v) TEP; (vi) cooperation with SLI; and (vii) TLG, were introduced into the model.

The research hypothesis has not been confirmed because only one institutional factor, namely cooperation with SLI, influenced the number of LGUs participating in the innovation networks. The model was based on only one independent variable, an institutional factor, namely cooperation with SLI; despite this being so, it was able to make a correct prediction in 91.1% of the cases of LGUs’ non-participation in the innovation networks as well as in 51% of the cases of their participation. The overall percentage of correct prediction was 80.4% (Table 5).

This study examined whether local context defined by the following set of indicators, that is to say – (i) institutional; (ii) economic; (iii) geographical; (iv) cultural; (v) demographic; (vi) administrative, (vii) managerial; or (viii) political, influence the LGUs’ participation in innovation networks. Our results are in line with those of Ketterer and Rodríguez-Pose [2016], who emphasize that for economic growth in the European regions in the period 1995–2009 institutional factors were more relevant than geographical ones. Similarly, Skog [2016] argues that local institutions are very important for each local network evolution.

Fitjar et al. [2016] highlighted that in the case of Norwegian innovation networks the “right” distance, i.e., “not too close and not too far” from another member of the network, matters, and that geographical factors are important. Our study has shown that almost 50% of all LGUs (who comprised the respondents for this study) were from 4 out of the 16 voivodeships totally contained in Poland. This may indicate that LGUs cooperating in the network tend to be concentrated, although ultimately the analysis showed that their location is not statistically significant.

The present study’s results for Poland are in contrast with those of Gonzalez et al. [2013], who provided evidence that the innovative activities of LGUs in Spain depend on their typologies or strategic profile. Polish LGU administrative factors such as types of LGU did not affect LGUs’ participation in the innovation networks. Surprisingly, LGUs from rural municipalities were the most active in participation in the innovation networks. This interesting observation may be indicative of a strategic approach to the development of municipalities; it may be explained that LGUs that are aware of their limitations are characterized by a greater tendency to seek support from the network. This is an approach that allows them to strengthen their development potential, as many of the innovative solutions can be relatively easily replicated and scaled-up at low cost.

Based on survey results of LGUs in the United States, Damanpour and Schneider [2009] highlighted that both innovation characteristics and LGU manager characteristics influenced the adoption of innovation. This is in contrast to our results, because in the case of Polish LGUs, managerial factors such as the LGU manager’s age, gender, or education and political factors such as the LGU manager’s political independence did not influence the LGUs’ participation in the innovation networks.

Based on the survey carried out with Italian LGUs, Reginato et al. [2011] stressed that the social, cultural, and structural (demographic and cultural) characteristics affected the innovation ecosystem at a local level. According to our study, a similar relationship did not exist in Polish conditions. Cultural factors measured by cultural infrastructure, demographic factors reflected by the number of inhabitants, and local population density did not affect LGUs’ participation in the innovation networks. This is consistent with the conclusion made earlier that networking may be driven by weaknesses. The need for networking arises rather in rural municipalities, which are smaller and economically weaker than urban municipalities.

The research hypothesis has not been confirmed. The cultural, demographic, managerial, political, and geographical factors were not correlated with the number of LGUs participating in the innovation networks. The economic and administrative factors were not statistically significant for the number of LGUs participating in the innovation networks. The study confirmed the importance of the institutional factor.

In 80.4% of the cases, the analyzed institutional factor correctly predicted LGUs’ participation in the innovation networks.

Polish LGUs highlighted how important local factors are for the innovation ecosystem at a local level and that cooperation with SLIs such as technology incubators or technology transfer centers may change LGUs’ activities to make the ecosystem more innovative and supportive for the evolution of local innovation network ecosystems. Operative local innovation networks may enhance local economic development by increasing the potential to undertake various initiatives that a single municipality would not be able to undertake alone. This may explain why rural municipalities were the most active of all LGUs in participation in the innovation networks (intermunicipal ones).

Taking into account that: (a) innovation networks are not popular among Polish LGUs; (b) participation in innovation networks has many advantages (e.g., fostering local innovation culture or local development process); and (c) factors that proved to be significant can be influenced by the public policy, a new policy instrument should be applied to enhance LGUs’ participation in the innovation networks. These can be incentives for supralocal development strategies and supralocal cooperation. This instrument may help to link complementary projects and facilitate the implementation of innovative projects that need the cooperation of several municipalities (in the fields of transport, a smart city, social innovation, etc.).

Further investigation is required to determine whether other local context factors of LGUs, i.e., historical, social, technological, geographical location (division of regions), or infrastructural indicators, may influence the engagement of LGUs in innovation networks.