Next Article in Journal
Erratum: Lee, J. and et al. 2018. The Effects of Knowledge Sharing on Individual Creativity in Higher Education Institutions: Socio-Technical View. Administrative Sciences 8: 21
Previous Article in Journal
Requirements Engineering for an Industrial Symbiosis Tool for Industrial Parks Covering System Analysis, Transformation Simulation and Goal Setting
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

What Shapes Local Innovation Policies? Empirical Evidence from Japanese Cities

1
Graduate School of Economics, Hitotsubashi University, Naka 2-1, Kunitachi, Tokyo 186-8601, Japan
2
Faculty of Economics, Gakushuin University, Mejiro 1-5-1, Toshima, Tokyo 171-0031, Japan
*
Author to whom correspondence should be addressed.
Adm. Sci. 2020, 10(1), 11; https://doi.org/10.3390/admsci10010011
Submission received: 22 December 2019 / Revised: 31 January 2020 / Accepted: 10 February 2020 / Published: 13 February 2020

Abstract

:
Increasing attention has been paid to regional innovation systems. However, previous studies have so far only focused on (the regional impact of) national policies or specific regions. Despite increasing attention to regional and local innovation policies, no studies have been carried out to date on the factors of implementation and design of local research and development (R&D) subsidy programs at the city level. Our research fills this gap by using information on R&D subsidy programs from local authorities in Japan collected via websites and our original survey. Thus, our research aims at empirically investigating the determinants of both implementation and design of local R&D subsidy programs at the city level (length and upper limit of subsidies, and flexibility of subsidy conditions) considering both demand- and supply-side factors. We employ probit models for basic empirical estimations and provide some robustness checks. The empirical results suggest that, after controlling for city type and population size, supply-side factors including local government conditions significantly affect the implementation of public R&D subsidy programs. In contrast, we find that demand-side factors matter more for the design of subsidy programs than supply-side factors.

1. Introduction

Generating innovation and new business through R&D is important from the viewpoint of promotion of the regional economy. Accordingly, regional innovation systems have attracted much attention both from academia and practitioners (Cooke 2001). Whereas firms, universities, and regional governments ought to play important roles in the conceptualization of regional innovation systems (Leydesdorff and Etzkowitz 1996), there is not sufficient empirical evidence on the role of local governments in the discussion of regional innovation systems1. On the one hand, while the central government is expected to plan national policies considering the general welfare of the nation, local authorities play a complementary role, as they are often required to implement these national policies for local parties. On the other hand, they are also expected to develop original policies that may be better suited to local conditions and needs (Perry and May 2007). We may therefore expect to encounter a large variety of local policies according to different conditions of local authorities and different needs of local firms.
To date, there have been numerous studies on the central government’s innovation policies including the promotion of regional clusters (Martin et al. 2011; Nishimura and Okamuro 2011a; Töpfer et al. 2019). In addition, several studies targeted specific regional policies and provided detailed case studies on these policies (Bellucci et al. 2019; Sharif and Xing 2019; Morisson and Doussineau 2019). However, despite increasing attention to local policy initiatives, econometric studies on the determinants of implementing local government policies, especially city level innovation policies, are scarce. Regarding environmental sustainability policies, some empirical studies examine the determinants of such policies at city and county levels in the USA, focusing on political factors (Hawkins et al. 2016; Laurian et al. 2017). However, no studies analyze the determinants of regional innovation policies except for a study of state-level Small Business Innovation Research (SBIR) program (Lanahan and Feldman 2015). Moreover, no studies address the determinants of program designs. Hence, our research aims are to empirically investigate the determinants of both implementation and design of R&D subsidy programs by city governments using unique Japanese data.
In Japan, with a traditional centralization of policy-making, national government has always played a major role in promoting regional economies. Even though the promotion of regional innovation has been recognized as an essential policy issue since the beginning of this century under the “Science and Technology Basic Plan”, it was the Ministries of Economy, Trade, and Industry (METI) and of Education, Culture, Sport, Science and Technology (MEXT) that took the initiative in cluster policies (See (Kitagawa 2007; Okamuro and Nishimura 2018) for these innovation policies). However, in 2014, the Japanese government initiated the “Chiho Sosei” (regional revitalization) policy, encouraging local authorities to plan and design their own “General Regional Strategy” considering regional economic conditions and development trends. An important background of this new policy is the decline of the Japanese population especially in rural areas due to rapid aging and low fertility rate. This new policy trend gives local governments a new, more important position in regional revitalization, although policy initiatives may differ significantly according to local needs and conditions. This is the major reason for our research focus on local authorities, especially city governments.
Also in Europe, multilevel policy mix including both EU and regional innovation support policies has already been implemented. Fitjar et al. (2019) suggest that, in 2011, the EU Directorate-General for the Regional and Urban Policy (DG Regio) introduced the Research and Innovation Strategies for Smart Specialization (RIS3), “a place-based policy which foregrounds the role of regions and emphasizes research and innovation policy, building competitive advantage based on regional strength and potentials” (p. 1). In the period 2014–2020, EU regions were required to develop their own smart specialization strategies as a condition for access to European Structural and Investment Funds. In European studies, however, regional policies are not clearly distinguished from municipality policies, so that municipality policies may be still negligible. In this sense, recent innovation policy in Japan involving and encouraging city governments is distinctly unique worldwide and worth researching.
It is well known that the structure of public expenditures (the balance between different purposes) varies across local governments. Previous studies in public policy research find that the balance of expenditures depends on economic and institutional constraints, political imperatives, and actual needs (Hajnal and Trounstine 2010). In fact, for a direct and concrete policy targeting local firms, such as an R&D subsidy program, our original survey data found a large variety of program designs among cities regarding volume, conditions, and selection procedures. For example, the upper limit of subsidy per project varies from to less than one million to 50 million yen with a median of two million yen. Subsidy ratio to total project budget varies from 20% to 100% with a median of 50%. Regarding the flexibility of R&D expenses, only 34% of local governments allow public subsidies to be spent on personnel expenses. Thirty percent allow recipients to obtain other public subsidies for the same project. Before estimating the effects of public support with respect to program design, it is important to examine whether such variety is random, and if not, what determines such variations. However, to the best of our knowledge, no previous studies have empirically addressed the determinants of policy design.
Our research will fill this gap by targeting the variety of local innovation policies at the city level in Japan. Thus, our research aims are to empirically investigate the determinants of both implementation and design of R&D subsidy programs by city governments using unique Japanese data. Based on website information of local authorities in the entire country and using original and unique survey data for public officers responsible for local authority R&D subsidy programs, we explore the variety of regional innovation policies. Next, we investigate the determinants local R&D subsidy program implementation and their design at the city level using multiple regression analysis. As Tödling and Trippl (2005) argue, no innovation policy can fit all regions because of a wide variety of regional characteristics. Rather, it is important to consider regional variety in the implementation and design of innovation policies by local authorities.
The remainder of this paper is organized as follows: In Section 2, we provide a brief review of the related literature. In Section 3, we describe the process of the R&D project by local firms and its relationship with local R&D subsidy program. In Section 4, we provide theoretical background and hypotheses. In Section 5, we explain our sample and data from our original survey for empirical estimations. In Section 6, we elaborate estimation models and report the empirical results. In Section 7, we conclude this paper by presenting some future research agendas.

2. Literature Review

Regional innovation systems comprising local Small and Medium Enterprises (SMEs), research institutes, and local authorities have attracted increasing attention since the seminal works by Cooke et al. (1997) and Cooke (2001), whereas there have been few studies on the innovation policies of local authorities. A special issue in Regional Studies focused on the multilevel governance of science (and innovation) policy and provided a comparison of the changing balance between central government and regional governments towards greater decentralization by contrasting centralized countries (England, France, Finland, and Japan) with federal countries (Germany and Canada) (Perry and May 2007). The articles in this special issue provide descriptive studies of science policy in each country with program-level (Salazar and Holbrook (2007) for Canada) or regional case studies (Crespy et al. (2007) for France, Sotaraura and Kautonen (2007) for Finland, Koschatzky and Kroll (2007) for Germany) and historical review (Perry (2007) for England, Kitagawa (2007) for Japan). It is also noteworthy that they all target regions or states instead of cities or counties2.
An OECD report addresses the multilevel governance of innovation policies at different administrative levels (sub-national, national, and supra-national) and refers to some different patterns of governance among member and non-member countries (OECD 2011). There are some conceptual papers on the multilevel policy mix on innovation (Flanagan et al. 2011; Laranja et al. 2008), while Fernandez-Ribas (2009) compares the effects of innovation programs in EU regions between different levels of governments (regional, national, and EU levels), without considering municipality or city level.
More recent literature of science policy suggests that the decentralization of science and innovation policy has been promoted in the last decade. In the EU, the Directorate-General for Regional and Urban Policy (DG Regio) implemented in 2010 the Research and Innovation Strategies for Smart Specialization (RIS3). In the program period 2014–2020, regions are requested to develop their own strategies as a condition for access to European Structural and Investment Funds (Fitjar et al. 2019). In Norway, a non-EU country, nineteen country governments were given responsibility for one percent of the nation’s total publicly funded research by the central government decision in 2010. These country governments were then required to develop their own R&D strategies and to cooperate with each other through new institutional arrangements (Kolltveit and Askim 2017).
Several previous studies focus on the impact of national cluster policies. Specifically, some empirically investigate local effects of national cluster policies in France (Martin et al. 2011; Fontagné et al. 2013) and Germany (Engel et al. 2013; Töpfer et al. 2019) at firm and local levels. For Japan, Nishimura and Okamuro (2011a, 2011b, 2016) and Okubo et al. (2016) empirically examine the effects of METI’s cluster policy at the firm level. Using a unique micro dataset, Okamuro and Nishimura (2018) provide a comparative econometric analysis of the project-level effects of METI’s and MEXT’s cluster policies that have similar aims but contrasting schemes. However, they do not consider the role of regional or local governments3.
In addition, several studies targeted specific regional innovation policies and provided detailed case studies on these policies (Bellucci et al. 2019; Morisson and Doussineau 2019; Sharif and Xing 2019). However, despite increasing attention to local policy initiatives, econometric studies on the determinants of implementing local government policies, especially city level innovation policies, are scarce.
Since the seminal papers by Leydesdorff and Etzkowitz (1996) and Etzkowitz and Leydesdorff (2000), the concept of Triple Helix has paid attention to the coordinator roles of national and regional governments in university–industry–government collaboration. However, most studies in this line provide conceptual discussion and case studies of specific industries or nations. Empirical studies targeting city government’s roles are scarce.
A special issue in the journal Technological Forecasting and Social Change in 2019 (Appio et al. 2019) provides a systematic assessment and integrative view of the “Smart City” policy that recently attracted worldwide attention. The purpose of this policy is “to increase the competitiveness of local communities through innovation while increasing the quality of life for its citizens through better public services and a cleaner environment” (Appio et al. 2019, p. 1). In this special issue, Caragliu and Del Bo (2019) find positive impact of this policy on urban innovation (measured by patenting) using a panel data set of 309 European metropolitan areas and Propensity Score Matching. However, they neither consider the determinants of nor the variety in these policies.
Why is it so important to consider the city level policy initiatives? A recent empirical study from Japan confirms that multilevel R&D support is highly effective on the recipient’s productivity, especially when it includes city level support (Okamuro and Nishimura 2020), although city level programs usually aim at small and no high-tech innovation. Their finding is consistent with the argument for the decentralization of policy-making that “activating regional R&D potentials can be difficult if science policy governance is highly centralized” (Kolltveit and Askim 2017).
In other policy fields in which local initiatives and multi-level governance are considered to be essential, some empirical studies have been conducted at various levels of local government. For example, local governments are increasingly involved recently in the development of environmental sustainability policies. Accordingly, a couple of empirical studies investigate the determinants of implementation of such policies at city and county levels in the USA, focusing on political factors (Hawkins et al. 2016; Laurian and Crawford 2016; Laurian et al. 2017). More generally, Foucault et al. (2008), using a dynamic panel data model of French municipalities, examine interactions between neighboring municipalities regarding primary and investment expenditures. Hajnal and Trounstine (2010) analyze the effects of economic and institutional constraints, political imperatives, and actual needs on the balance between redistributive, allocational, and developmental spending in the USA at city and county levels.
However, apart from a few recent studies (Lanahan and Feldman 2015; Lanahan 2016) that target state-level small business innovation research (SBIR) programs in the USA, to the best of our knowledge, no empirical studies have so far been carried out on the determinants or effects of regional innovation policies by local authorities with respect to their variety. The 2015 study by Lanahan and Feldman explores the determinants of additional financial support by the states for the recipients of federal SBIR programs, while the latter by Lanahan (2016) examines the effects of additional state support for SBIR recipients. Lanahan and Feldman (2015) use a panel dataset of all states in the USA for the years 1983–2010. They argue that top–down (national policy variables), lateral (neighbor states’ policy), and bottom–up (political and scientific environment of the state) factors incur the implementation of additional state-level support for the federal policy and find that most of these variables significantly affect the propensity of state SBIR. However, they do not target any local authorities lower than state-level, nor do they address any variations across state SBIR programs.
This paper will fill this gap by empirically investigating the determinants of both implementation and design of city-level R&D support policies in Japan using original survey data. Focusing on city-level innovation policies, which are independent of national policies, is especially important, because they lack empirical investigation and evidence, despite recent transformation of local governments’ roles from passive (regions as stages and implementers) to active (regions as partners and independent policy-makers) (Perry and May 2007), which is striking in the case of Japan. This paper is original in that it targets not only the implementation, but also the variety in the design of R&D support policies by local authorities in Japan, especially in cities, where authorities are smaller and have less administrative power and responsibility than those in federal country states. Due to the cross-section nature of our dataset, we focus on the bottom–up factors used by Lanahan and Feldman (2015), which we reorganize into demand and supply factors.

3. Process of Local R&D Support Programs and R&D Projects

In this section, we describe the whole process of the local R&D subsidy program in relation to local R&D projects in order to illustrate how R&D support programs of local governments are managed, interacting with local R&D projects. The description in this section is based on the authors’ own survey and hearing information from the program officers of city governments. We found a large variety of support programs across cities and local governments, but show just an average pattern of program implementation. Based on the description, we can argue how program design, especially the term and ratio of subsidy to total project budget and the flexibility in project management (the dependent variables of our estimations), is important for the innovation by recipient firms and projects.
(1) Call for applications: R&D subsidy programs by city governments mostly depend on their own annual budget (sometimes supported by central government’s fund). For two-thirds (67%) of the respondents (city programs), the support is limited within one year (by the end of the same fiscal year). For the rest of the programs, public support is usually renewed every year after submitting project report. Since the fiscal year starts in April in Japan, call for applications of most programs are announced in April or later. The selection starts after the application deadline in May, June, or July. Since the recipients have to accomplish their projects and present project report by February or March in the following year, the real project term is limited to at most nine months. Call for applications is announced on the city government’s website and through the local chambers of industry and commerce and industry associations to potential applicants. Due to short terms, the applicants often organize and start R&D projects before they submit project proposals.
(2) Selection procedure: Due to a limited program budget (on average 18 million yen on average in 2015, but the median is six million yen), there is often a competitive selection of recipients. Around 71% of the programs organize an evaluation committee, which has 5.8 members on average. Most of the juries are experts from outside the city government, such as local university professors and board members of local chambers of industry and commerce. In 2015, on average, each program received 6.1 applications (median is four), among which 4.6 (median is three) projects were accepted for subsidy. Project evaluation is often (85%) made through application documents and interviews. About 10% of the programs got no applications in 2015. The ratio of acceptance was 85% on average.
(3) R&D project implementation: The upper limit of the R&D subsidy significantly differs across programs, but its mean is 3.7 million yen (median 2.0 million yen). Usually R&D project can only partially be supported, and the rest should be self-financed. For half of the programs (51%), the subsidy ratio is up to 50% of the total project budget. Each accepted project received 2.4 million yen on average in 2015 (median 1.3 million yen), but the recipients have to conduct the R&D projects with their own funding. In almost one-third of the cases (36%) the recipients are officially allowed to request partial advance payment of R&D expenditures, but usually they do not receive advance payment, but reimbursement of their R&D expenditures after they accomplished the R&D project and submitted the project report. Moreover, in two-thirds of the programs (66%), it is not allowed to expend the subsidy for personnel costs. For several city programs (46%), the recipients can obtain technological and managerial support for their projects through the city government. As the small amount and short term of the subsidy suggest, city R&D support programs usually promote small, no high-tech, and development-oriented innovation rather than high-tech and science-based R&D.
(4) Project report/evaluation and subsidy grant: In two-thirds (65%) of the programs, the accepted projects are subject to mid-term evaluation (even for a short-term project within one year). Formally, the city government can discontinue a project or reduce the subsidy amount depending on the mid-term evaluation in many cases (73%). After the subsidy period, all programs require final reports about project outcomes, such as new product prototypes, from the recipients. In some cases (39%), an additional interview (oral presentation) is requested for final evaluation. Formally, in several programs (66%), the city government can reduce or even disapprove the subsidy depending on the final evaluation.
(5) Follow-up: After the support program finishes, program officers often (in 55% of the programs) follow up with the recipient projects for three to five years to check the development of outcomes. Sometimes the recipients are invited by program officers to other related programs of the same city government or prefecture/national support programs.
It is noteworthy that, although a minority (24%) of the programs is designed considering the relationship with prefecture and national R&D support policies, two-thirds of the programs (67%) prohibit multiple subsidization of the same project by prefecture or national programs. According to our interviews with city program officers, the reasons for this restriction are to avoid the concentration of public subsidies on a few projects and to secure diversity in public support. As mentioned above, city government programs usually support small, non high-tech projects of small local firms that aim at relatively simple development of new products within one year.

4. Theoretical Backgrounds and Hypotheses

As previously mentioned, with a traditional centralization of policy-making, central government in Japan always played a major role in promoting regional economies. Even though the promotion of regional innovation has been recognized as an essential policy issue since the beginning of this century, METI and MEXT adopted the initiatives of cluster policies. Local authorities have been gradually acknowledged as partners in cluster policy, but it was not until the introduction of the Regional Revitalization Policy (Chiho Sosei) in 2014, that local authorities were encouraged and competitively supported by the central government to plan and design their own development policies with respect to local economic conditions. With this new national policy trend, local governments obtain a new and more important position in regional development.
Perry and May (2007) provide a conceptual framework for the “regional dimension” to science policy. Regions may play passive roles as stages and implementers of national policies and active roles as partners of central government and independent policy-makers. Recent developments in Japan demonstrate that Japanese local authorities (prefectures and cities) are changing rapidly and increasingly from passive to active players with the pressure and support of national policy. However, the extent of such a transformation may vary significantly across local governments. Perry and May (2007) explain the role of regions as independent policy-makers as follows. “Regional authorities and bodies are increasingly devoting their own finance and resources to funding regionally significant scientific investments or projects. The emergence of ‘regional science policies’ may be characterized by independent agenda-setting, institutional creation and new governance arrangements, new mechanisms and policy tools or strategic intelligence and capacity building” (p. 1042). Local authorities may differ regarding their motivation, resources, and capability.
Local government policy decisions are not random, but may be influenced by different factors. As we reported in the previous section, we found at least two empirical studies using US regional data that consider the factors for decision-making by local governments from multiple perspectives. In this section, we first explain the empirical models in these studies in more detail in order to derive our own theoretical basis for the hypotheses, considering the specific Japanese context.
Hajnal and Trounstine (2010) consider four factors for the balance of public expenditures: economic, political, and institutional factors, as well as actual needs or local context, and found that each of these factors is important. To elaborate, economic factors include the economic resources of and competition between local governments, for which they use government revenue per capita and the central city dummy as proxies. The political factor refers to local public preference which is represented by share of votes for the Democratic candidate in presidential elections. Institutional factors cover local, state, and federal institutions, among which differences in local administration patterns are regarded as especially important. Actual needs (local context) are measured by regional dummies, poverty rate, unemployment rate, and population size.
By focusing on local R&D subsidy programs, which are not essential for local governments, we can more clearly address local economic conditions and distinguish between supply- and demand-side factors than previous studies. Lanahan and Feldman (2015), who estimate fixed-effect models using US state data for 28 years, consider top–down, lateral, and bottom–up factors. Top–down factors (time-series variation in the national policy) consist of the variables of federal R&D funding and industrial specialization in public procurement. Lateral factors reflect policy imitation or bandwagon effects, which are measured as the proportion of neighboring states with policy implementation. Bottom–up factors (regional conditions) are the Democratic governor dummy, its interactive term with early scientific advisory system, fiscal health (state revenue), and the variables of high-tech employment and higher education capacity.
We are more interested in the effects of local (cross-section) variations than time-series variations on public subsidy. Local R&D subsidies at city level are relatively new policy programs in Japan, where most programs have traditionally been centralized and, until recently, there has been very little room for local policy initiatives. In this context, top–down or institutional factors have little importance compared to the US, where institutions and laws differ significantly across federal states. Moreover, in the context of Japanese policy, competition between two major political parties has not functioned well. Most mayors are either “conservative” or “independent,” but seldom belong to the opposition party in the central government, in which the conservative Liberal Democratic Party has almost always been the ruling party. Therefore, we concentrate on economic factors and actual needs (Hajnal and Trounstine 2010) and in particular, bottom–up factors (Lanahan and Feldman 2015), which we reorganize into demand- and supply-side factors.
Public expenditures by local governments depend on their revenues, which are mainly based on tax and public debt. Therefore, local governments can be viewed as being under fiscal constraints and they should consider fiscal health (McDonald 2018). In light of these constraints, different policy schemes compete for limited budgets, and since it is often difficult to avoid or decrease the expenses of existing policies, launching a new policy program, especially a subsidy program, is a challenge for local governments. Therefore, local government’s revenue is an important factor for R&D subsidy (Lanahan and Feldman 2015). Moreover, administrative capabilities regarding R&D support programs may differ significantly across local authorities due to the quantity and quality of available human resources (Charron et al. 2014; Morisson and Doussineau 2019). Indeed, using American municipality data, Hajnal and Trounstine (2010) argue and find that both economic and political constraints affect local government’s decisions on public expenditures. Therefore, we argue that the implementation of local R&D support programs depends on supply-side factors including fiscal constraints and the administrative capability of local authorities. Thus, we propose the following hypotheses on the effects of supply-side factors:
Hypothesis 1a:
Local authorities that are financially less constrained and that provide more administrative services, especially for local firms, are more likely to implement R&D subsidy program.
Hypothesis 1b:
Local authorities that are financially less constrained and that provide more administrative services, especially for local firms, are more likely to design R&D subsidy program in a more favorable way for potential recipients.
Regarding demand-side factors, actual policy needs of local firms are also important, as argued by Hajnal and Trounstine (2010). Here, we argue that not only actual needs, but also potential needs are important (Morisson and Doussineau 2019). Therefore, the larger the potential local needs for such programs, the more likely local authorities are to implement them. Although it can be also argued that local authorities of regions that lag behind in R&D activities are more eager to catch up with subsidies to local firms, as Lanahan and Feldman (2015) point out, we argue that efficient policy measures should consider local needs. Therefore, we propose the following hypotheses on the effect of demand-side factors:
Hypothesis 2a:
Local authorities of regions with higher research potential of local SMEs are more likely to provide R&D subsidy programs in response to higher demand by local firms.
Hypothesis 2b:
Local authorities of regions with higher research potential of local SMEs are more likely to design R&D subsidy programs in a more favorable way for potential recipients in response to higher demand by local firms.
Regarding R&D subsidies for local firms, local needs may be specific in the sense that not every local firm is eligible for the subsidy. It is well known that R&D is concentrated in the manufacturing sector, and more so in some industries within the manufacturing sector. Therefore, agglomeration of manufacturing firms, and specifically high-tech firms in R&D intensive industries, are important demand measures. When considering the competition between different policy purposes, we argue that the share of potential firms with R&D investment in each city is an appropriate variable for the demand for R&D subsidy, when controlled for economic size (population size) of the city. Moreover, we propose using average labor productivity of local manufacturing firms as a proxy for potential innovativeness of local firms4.
Moreover, the national government promotes collaborative R&D with other firms, universities, or public research institutes through the allocation of specific subsidies (Nishimura and Okamuro 2011b). Triple Helix theory argues that governments play coordinator roles in promoting university–industry–government collaboration (Etzkowitz and Leydesdorff 2000). Accordingly, most local R&D support programs target collaborative R&D projects, as we show later in the data section. Therefore, the availability of potential R&D partners in the region including universities and public research institutes may be an essential element of the implementation of local R&D subsidy programs. Hence, we propose the following hypotheses on further effects of demand-side factors:
Hypothesis 3a:
Local authorities of regions with more potential R&D partners (including universities and public research institutes) are more likely to implement subsidy programs in response to higher demand.
Hypothesis 3b:
Local authorities of regions with more potential R&D partners (including universities and public research institutes) are more likely to design subsidy programs in a more favorable way for potential recipients in response to higher demand.
We propose the above hypotheses regarding the effects of supply- and demand-side factors both on the implementation and design of local R&D subsidy programs. Basically, we consider that the same factors of program implementation apply to program design, thus both supply- and demand-side factors affect program implementation as well as program design. Yet, as the capacity of supply is the first and major hurdle of public policy, supply-side factors like fiscal constraints and administrative capability may be considered to be more important than local needs for implementing such programs. However, once the local authorities decide to implement their own R&D subsidy programs, they will also consider local needs as far as possible when designing their programs, according to the program goals. Therefore, as a whole, supply-side factors of local authorities may be more important for implementing R&D subsidy programs than demand-side factors, and demand-side factors are more important in designing these programs. We will compare the effects of supply- and demand-side factors in our empirical estimations.

5. Sample and Data

This paper aims at empirical (thus econometric) estimations of the determinants of both implementation and design of local R&D subsidy programs at the city level. Specifically, we consider both demand- and supply-side factors and compare their effects (we explain the estimation models in detail in the following section). For the first purpose, to investigate the determinants of program implementation, our sample should cover all cities in Japan, while we need information about which cities implement their own R&D subsidy programs. For the second purpose, to investigate the determinants of program designs, we target the cities with R&D subsidy programs, for which we need detailed information about program designs.
The sample for the first estimation covers all 813 cities in Japan (including special wards in Tokyo)5. We checked the website of each city in Japan in 2015 to derive some basic information about public R&D subsidy programs for local SMEs and start-ups, which local authorities (city governments) implemented for the fiscal year 2015. The target projects of these programs include single firm projects, inter-firm collaboration, and university–industry collaboration. We exclude those support programs that ended before 2014 and those in which the maximum amount of subsidy per project is under one million yen in order to select active programs with a substantial amount of subsidy per project. Finally, we confirmed 179 local R&D subsidy programs in 154 cities among 813 cities (18.9%) as of 2015.
Then, we collected local and administrative data of each city from various public sources including the Economic Census (Statistics Bureau) and the Census of Manufacture (METI), and public information about city governments. Eventually, our sample was reduced to 772 cities due to missing data for administrative mergers. Among these 772 cities, 129 cities (16.7%) implemented R&D subsidy programs as of 2015.
Then, we conducted an original online survey of relevant city government officers on the content and strategy of these programs6. Questionnaire items include (1) program contents (start year, targets and conditions, limit and ratio of subsidies, program budget and expenditure, number of applications, and selected projects); (2) procedures of selection and evaluation (selection method, support for selected projects, progress check, project report, and final inspection); and (3) background of program designs (relatedness to national policy, past programs, self-evaluation, factors for subsidy criteria, and future agenda). After a pre-test with two programs, questionnaires were sent to the relevant government officers of 177 local programs in January 2016. We obtained responses on 151 programs in 131 cities (response rate: 85.3%)7. Among them, we did not use two cities for missing data, so the sample size was reduced to 129 cities (16.7% of 772 cities). The second estimation on program designs focuses on at most 120 cities among them, since some cities had to be omitted due to missing local data.
In the following part, we briefly describe our survey results focusing on 151 programs in 131 cities. The description of program design partially overlaps with that in Section 3. The earliest local policies date back to the 1980s, but the median year of the implementation of the current policy program is 2009, which suggests that most local administration began their own R&D support programs relatively recently. Although there are few R&D subsidy programs that support only single-firm projects (11%), most programs also target single-firm projects. Half of the programs include university–industry collaboration in their targets. Most programs (88%) support only local firms in their own region, but in most programs project partners can be located elsewhere.
Subsidy duration is generally short: Two-thirds (67%) are limited to within one year. The average subsidy ratio to total R&D expenses is 58% (median: 50%). Most programs request partial self-payment of R&D expenditures by subsidy recipients, which may restrain their moral hazard. The average ceiling on R&D subsidy per project is 3.66 million yen (median: two million yen). Personnel expenses are allowed in one-third of the programs (34%), as is the advance payment of the subsidy (36%). The program budget in the fiscal year 2015 is on average 18 million yen (median six million yen) while the actual expenditure is on average 9.6 million yen (median four million yen in median).
Finally, regarding the relatedness to national programs, 67% of city programs do not support projects subsidized by prefecture or national programs in the same year. Thus, city programs are often designed as substitution for prefecture or national ones. The amount and conditions of subsidy are determined considering those of prefecture or national programs in 24% of the cases, while 27% of the programs create rules based on the situation of local firms.

6. Empirical Estimations on Local R&D Subsidy Programs

6.1. Estimation Models

We explain the empirical models to estimate the local authority’s propensity to implement its own R&D subsidy program and the determinants of the designs of these programs. We employ a probit model for both estimations because all dependent variables are dummy variables.
In the first estimation, the dependent variable is the dummy variable that takes one if a city implements its own R&D support policy in 2015 and zero otherwise. Almost 17 percent of the 772 cities in the sample (129 cities) implemented their own R&D subsidy program in 2015. We use the following city-level independent variables in the estimation: (1) proxies for local R&D capability (the share of manufacturing plants, share of high-tech plants in the manufacturing sector, and average labor productivity in manufacturing); (2) availability of potential partners for R&D collaboration (the number of large manufacturing firms with more than 300 employees, the number of R&D institutes, and the number of universities in the prefecture8); and (3) the scale and structure of public expenditures (the number of administrative officers per capita in natural logarithm, the share of industry and commerce expenditures, and the current account9).
The variables in Groups (1) and (2) represent the demand factors (proxies for policy needs of local firms). We expect that the propensity of implementing R&D support policy may increase with local firms’ policy needs (Hajnal and Trounstine 2010). However, if it is a general policy aimed at catching up with other regions, then we may also expect that policy support may be provided in higher propensity in regions with relatively lower R&D potential (Lanahan and Feldman 2015). The variables in Group (3) indicate the supply factors as the proxies for local authorities’ level of administrative services and budget constraints. We hypothesize that these factors affect the probability of local authorities implementing public R&D support programs for local SMEs as economic and institutional factors (Hajnal and Trounstine 2010). We also control for city scale measured as the population size in natural logarithm and the status of the city (the dummy for prefecture capitals)10. We assume that cities with a large population and higher administrative centrality in the region may be more likely to implement their own R&D support policies as “leaders” and “role models.” In the estimation, we use the data before 2009, the median of the start years of the support policies, in order to consider the time-series consistency.
In the next estimations of the determinants of program design, we use the following four dummies as the dependent variables: dummies for (1) the length of subsidy, which takes one if public support exceeds one year and zero otherwise; (2) the high ratio of subsidy, which takes one if the upper limit of subsidy ratio is above 50% and zero otherwise; (3) the flexibility in expenses, which takes one if the subsidy is allowed to cover personnel expenses and zero otherwise; and (4) the generosity in receiving public support, which takes one if the same project is allowed to be simultaneously funded by other public subsidies and zero otherwise. We focus on these dependent variables, since they represent basic conditions of public R&D subsidy and show a large variety across city programs, as we show in Section 3.
We employ probit models again because the dependent variables are dummy variables. For these estimations, we use the same independent variables as the first estimation, except for business field dummies that are not used in the first estimation. We include them as additional control variables in the estimation models of program designs because they may differ across the target business fields of subsidy programs. In the following subsections, we conduct these estimations separately and independently, but we will also check possible sample selection bias and the robustness of our estimations.

6.2. Determinants of Program Implementation

First, we estimate the factors of implementing an R&D subsidy program using a probit model. Our research question is which city (or local authority) provides R&D subsidy programs for local firms. Table 1 and Table 2 show the basic statistics and correlation matrix of the variables. The final sample is limited to 772 cities because some data was missing due to administrative mergers. Table 3 presents the estimation results on the propensity for implementing local R&D support policies. In this table, we also show the marginal effects, and not the coefficients, of each independent variable.
These estimation results suggest that, even controlling for the size and administrative status of the cities, regional specialization in manufacturing (the share of manufacturing establishments) and active administrative services (number of administrative officers per capita and the share of expenditures for industry and commerce) significantly increase the propensity of implementing R&D support programs. For example, a 1% increase in the number of administrative officers per capita will lead to a 0.216% increase in the propensity of implementing R&D support programs, which suggests a large impact. We also find that the current account has a negative and significant effect on the probability of R&D subsidy programs. As a negative value of current account in this case means budget surplus, this result suggests that the cities with a sound public finance structure are likely to implement R&D support programs.
These findings fully support H1a that the implementation of local R&D support policies depends on supply-side factors of local government, but only partially support H2a on potential local needs (regarding the agglomeration of manufacturing firms) and do not support H3a on potential R&D partners. These results are consistent with those of Lanahan and Feldman (2015), in which state revenue has a positive and significant effect on federal states’ propensity to implement their own SBIR subsidy. Overall, these results suggest that supply-side factors (H1a) may be more important than demand-side factors (H2a and H3a) in implementing local R&D policies. In this regard, we may argue that negative effects of demand-side factors (local authorities in the region with weak R&D activity tend to implement these programs for catch-up) offset and thus weaken their positive effects. These results are consistent with those of Lanahan and Feldman (2015), in which state revenue and high-tech employment have positive and significant effects on federal states’ propensity to implement their own SBIR subsidy.
Contrary to our expectation, we find that the number of large manufacturing firms in the region negatively affects the probability of public R&D subsidy. This result may suggest that cities in which employment relies on large firms find it less necessary to implement support programs for local SMEs. Regarding control variables, the results suggest that larger cities (measured as population size) have significantly higher propensity to implement their own R&D subsidies. A 1% increase in the population leads to a 0.165% increase in the propensity of policy implementation. However, when controlling for population size, prefecture capitals are not significantly different from other cities with regard to the propensity of policy implementation. This may be because prefecture capitals may be expected to play a leadership role in local policies, but may also be more dependent on prefecture’s policies in the sense that they do not need their own policies when a similar policy is implemented by their prefecture. Thus, the positive and negative impact may offset each other.
The above estimation results do not significantly change when the upper bound of subsidy per project is changed from one million to two or three million yen, with an exception for the result on the number of large manufacturing firms. Thus, we may regard these estimation results as robust regarding the amount of subsidy per project. Moreover, the results do not considerably change when we use the number of students or only national universities instead of the total number universities in the prefecture. From these findings, we may cautiously conclude that the supply-side factors (the conditions of public service of each city administration) are relatively important for the decisions of R&D support programs in relation to local SMEs.

6.3. Determinants of Program Design

In the next step, we examine the determinants of the design (scale, contents, conditions, and procedures) of local R&D subsidy programs. More specifically, we investigate whether and how the local factors used in the estimation in the previous section may affect program features, such as the length of subsidy, the upper limit of subsidy ratio (up to 50%, two-thirds etc.), whether the subsidy may cover personnel expenses, and whether multiple subsidization by different levels of governments is permitted. We focus on these variables because we find that, according to our survey, the features of R&D subsidy programs by local authorities differ considerably. This may also influence the attractiveness and user-friendliness of these programs. For example, a subsidy program for less than one year would be too short to accomplish any fundamental innovation.
Thus, we use the following four dependent variables: (1) subsidy length (dummy variable of whether the program covers a period longer than one year); (2) subsidy ratio (dummy variable of whether the subsidy’s upper limit is above half of the total budget); (3) cost coverage (dummy variable of whether the subsidy can be used for personnel expenses); and (4) the relationship with national and prefectural programs (dummy variable of whether an overlapping subsidy with a national or prefectural program is possible). The value one for these variables mean that local R&D subsidy programs provide favorable conditions for the recipients in that (1) the term is longer than one year, (2) more than a half of project budget can be publicly funded, (3) research and support staff may also be employed, and (4) the same project can also be supported by prefecture or central government, respectively.
In addition to the independent variables previously used, estimation models in this stage include business field dummies. With these new variables, we control for any unobservable differences in program designs across business fields. We hypothesize that, similar to the determinants of program implementation, both demand and supply factors matter for the design of local public R&D subsidy programs (H1b, H2b, and H3b).
In this stage, we exclude all cities that did not implement an R&D support policy in 2015. Our sample is therefore limited to 129 cities (including special wards in Tokyo) that conducted at least one R&D support program in 2015. As some cities implemented two or more related policy programs in 2015, in these cases we selected the major program for each city (in terms of total budget size) and excluded others from the sample in order to obtain one-by-one matching between cities and subsidy programs. Table 4 shows basic statistics of this final sample of 120 cities. A correlation matrix of the variables for this sample is provided in Table 5.
In the following estimations, we again employ probit models because all dependent variables are dummy variables. After these basic estimations, we also provide the results of other estimation models (Heckman’s two-step estimations, structural estimation modeling, and bivariate probit models) as robustness checks in the following subsection.
Table 6 shows the estimation results of the designs of local R&D subsidy programs. First, regarding subsidy length (dummy for subsidy longer than one year), high-tech ratio, and number of large manufacturing firms have positive and significant effect. Second, the upper limit of subsidy ratio (dummy for upper limit above a half) is positively and significantly affected by manufacturing share, number of large manufacturing firms and universities (in the same prefecture), and the current account of local government. Third, public subsidy may cover personnel expenses in cities where many high-tech firms as well as large manufacturing firms are located. Fourth, multilevel subsidization with prefecture or national government’s subsidy is permitted by cities with a higher ratio of high-tech firms but with lower productivity of manufacturing firms.
These results support H2b and H3b partially, but do not support H1b. It is noteworthy that the factors which significantly affect program design differ considerably across design measures. Yet, almost no supply-side factors have significant effects with expected signs, while either high-tech ratio in manufacturing firms or number of large manufacturing firms (demand-side factors) positively and significantly affects all measures of subsidy program design in favor of recipient firms. In contrast to the results on the implementation of local R&D subsidy programs, we find that as a whole program designs depend on the demand- rather than supply-side factors, but some specific factors affect it, program design may also be biased. The results are shown in Table 7. We find that the inverse Mill’s ratio is statistically not significant and that the error terms of the first and second stage estimations are not significantly correlated. We can therefore reject the null hypothesis that the estimation results on program designs are subject to sample selection bias. These results suggest that the determinants of implementing R&D subsidy programs are not related to those of program features, which is consistent with the results shown in Table 3 and Table 6.
Second, we simultaneously estimated four models of program design in Table 6 using structural equation modeling (SEM). The results are presented in Table 8. We find statistically significant correlations between the error terms between Models 1 and 2, Models 1 and 4, and Models 3 and 4, but do not observe significant changes in the estimated results from those in Table 6. Finally, we conducted bivariate probit estimations with Models 1 and 2, Models 1 and 4, and Models 3 and 4, in order to take correlations between the error terms of estimation models into consideration (Table 9). The estimated results did not change significantly from those in Table 6. The results of these additional estimations suggest that our findings are robust to possible biases.
Regarding control variables, population size, which is highly significant in the implementation of local R&D support programs, has no significant effects on program designs. Larger cities are more likely to implement their own R&D subsidy programs, but program designs do not depend on city size.

6.4. Sample Selection Models and Other Robustness Checks

We conducted some robustness checks in this stage. First, we estimated Heckman’s sample selection models (two-stage probit models) in order to check sample selection bias in the second stage (determinants of program designs). If the program implementation is not randomly selected, but some specific factors affect it, program design may also be biased. The results are shown in Table 7. We find that the inverse Mill’s ratio is statistically not significant and that the error terms of the first and second stage estimations are not significantly correlated. We can therefore reject the null hypothesis that the estimation results on program designs are subject to sample selection bias. These results suggest that the determinants of implementing R&D subsidy programs are not related to those of program features, which is consistent with the results shown in Table 3 and Table 6.
Second, we simultaneously estimated four models of program design in Table 6 using structural equation modeling (SEM). The results are presented in Table 8. We find statistically significant correlations between the error terms between Models 1 and 2, Models 1 and 4, and Models 3 and 4, but do not observe significant changes in the estimated results from those in Table 6. Finally, we conducted bivariate probit estimations with Models 1 and 2, Models 1 and 4, and Models 3 and 4, in order to take correlations between the error terms of estimation models into consideration (Table 9). The estimated results did not change significantly from those in Table 6. The results of these additional estimations suggest that our findings are robust to possible biases.

7. Conclusions

For several years, regional innovation systems and multilevel governance of innovation policies have attracted much attention. Local authorities are now expected to play a more active and independent role in regional development, including innovation policies for local SMEs in Japan, where regional and innovation policies have traditionally been strongly centralized. However, despite these new policy trends of regional decentralization, few empirical studies have so far been conducted on innovation policies by local authorities. In this sense, as far as we know, this paper is a pioneering empirical study on this research topic.
This paper paid particular attention to local R&D subsidy policies in Japan and, based on our original survey data on local governments, investigated the factors of implementation and design of local innovation policies. The results of the empirical estimation on the determinants of local R&D support programs show that both demand-side and supply-side factors are important in the decisions of local authorities to implement local R&D support program, but that supply-side factors seem to be more important. Further empirical estimations on the determinants of program design suggest that demand factors, especially the presence of high-tech firms and large firms in the manufacturing sector, are correlated with subsidy designs that are favorable for local recipients. These results as a whole support our hypotheses.
A major limitation of this study is its static nature due to cross-section data. It is not possible to obtain information about the start year of each local R&D support program from the websites. Moreover, we cannot collect information about local R&D support programs that ended before 2015. In the questionnaire survey, we asked for the start year of the current program, but 30% reported that they implemented another R&D support program prior to the current one. In some cases, nobody was able to identify the start year partly because some programs remain live with modifications to titles, targets, and contents. Therefore, it is often difficult to identify when R&D subsidy programs really began. With a more dynamic setting with a panel dataset, in which start years of each program could be clearly identified, we would quantitatively examine the effects of national policy schemes and decisions of neighboring local governments as considered by Lanahan and Feldman (2015).
According to the results of our original survey, in only a few cases, local authorities started R&D subsidy programs because neighboring cities started similar programs. In contrast, however, 23% and 22% of the respondents report that they determined the design of subsidy programs by considering neighboring cities’ programs or considering the relationship with national and prefectural government, respectively. Hence, the bandwagon effect in local innovation policies may be more than negligible.
A major academic implication of this study is that we obtained clear empirical evidence for the endogeneity of local innovation policies and the main sources of this endogeneity. Based on our estimation results, it is advised that future empirical studies on the effects of innovation policies are undertaken to consider this endogeneity issue. Moreover, from a practical point of view, the findings of this study may be considered when designing local public policies and the relationship between public policies at different levels.
Major policy implications may also be derived from this study. National government should consider the existing conditions of local economies and authorities in encouraging and evaluating policy initiatives by local authorities. Local authorities should consider local conditions (demand-side factors) more seriously in order to provide original local policies. We hope our paper may encourage empirical studies on the roles and strategies of local authorities in regional innovation systems and the multilevel policy mix in innovation policy.

Author Contributions

The first author mainly designed research and wrote the paper, while the second author mainly conducted empirical analyses, but otherwise both authors equally contributed to this paper. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by JSPS KAKENHI Grant Number 15H03342.

Acknowledgments

Early versions of this paper were presented at the 36th Annual Meeting of the Japan Academy of Small Business Studies (JASBS) at Meiji University, Tokyo, Japan, on 11 September 2016, the 4th Asian SME Conference of the Asia Council for Small Business (ACSB) in Jakarta, Indonesia, on 14 September 2016, the 7th Asia-Pacific Innovation Conference (APIC) at Kyushu University in Fukuoka, Japan, on 19 November 2016, the 31st RENT (Research in Entrepreneurship and Small Business) Conference at Lund University in Lund, Sweden, on 15 November 2017, the 6th International Conference on Innovation Management, Entrepreneurship and Sustainability (IMES) at the University of Economics in Prague, Czech Republic, on 31 May 2018, and the Regional Studies Association Annual Conference in Santiago de Compostela, Spain, on 6 June 2019. The authors are grateful to the participants of these conferences for their helpful comments and suggestions. Moreover, the authors thank two anonymous reviewers of this journal for their valuable comments. Any remaining errors and omissions are our own responsibility.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Appio, F. P., M. Lima, and S. Paroutis. 2019. Understanding Smart Cities: Innovation ecosystems, technological advancements, and societal challenges. Technological Forecasting and Social Change 142: 1–14. [Google Scholar] [CrossRef]
  2. Bellucci, Andrea, Luca Pennacchio, and Alberto Zazzaro. 2019. Public R&D subsidies: Collaborative versus individual place-based programs for SMEs. Small Business Economics 52: 213–40. [Google Scholar]
  3. Caragliu, Andrea, and Chiara F. Del Bo. 2019. Smart innovative cities: The impact of Smart City policies on urban innovation. Technological Forecasting and Social Change 142: 373–83. [Google Scholar] [CrossRef]
  4. Charron, Nicholas, Lewis Dijkstra, and Victor Lapuente. 2014. Regional governance matters: Quality of government within European Union member states. Regional Studies 48: 68–90. [Google Scholar] [CrossRef]
  5. Cooke, Philip. 2001. Regional innovation systems, clusters, and the knowledge economy. Industrial and Corporate Change 10: 945–74. [Google Scholar] [CrossRef]
  6. Cooke, Philip, Mikel Gomez Uranga, and Goio Etxebarria. 1997. Regional innovation systems: Institutional and organizational dimensions. Research Policy 26: 475–91. [Google Scholar] [CrossRef]
  7. Crespy, Cecile, Jean-Alain Heraud, and Beth Perry. 2007. Multi-level governance, regions and science in France: Between competition and equality. Regional Studies 41: 1069–84. [Google Scholar] [CrossRef] [Green Version]
  8. Engel, Dirk, Timo Mitze, Roberto Patuelli, and Janina Reinkowski. 2013. Does cluster policy trigger R&D activity? Evidence from German biotech contests. European Planning Studies 21: 1735–59. [Google Scholar]
  9. Etzkowitz, Henry, and Loet Leydesdorff. 2000. The dynamics of innovation: from National Systems and “Mode 2” to a Triple Helix of university-industry-government relations. Research Policy 29: 109–23. [Google Scholar] [CrossRef]
  10. Falck, Oliver, Stephan Heblich, and Stefan Kipar. 2010. Industrial innovation: Direct evidence from a cluster-oriented policy. Regional Science and Urban Economics 40: 574–82. [Google Scholar] [CrossRef]
  11. Fernandez-Ribas, Andrea. 2009. Public support to private innovation in multi-level governance systems: An empirical investigation. Science and Public Policy 36: 457–67. [Google Scholar] [CrossRef]
  12. Fitjar, R., P. Benneworth, and B. T. Asheim. 2019. Towards regional responsible research and innovation? Integrating RRI and RIS3 in European innovation policy. Science and Public Policy. [Google Scholar] [CrossRef] [Green Version]
  13. Flanagan, Kieron, Elvira Uyarra, and Manuel Laranja. 2011. The ‘policy mix’ for innovation: Rethinking innovation policy in a multi-level, multi-actor context. Research Policy 40: 702–13. [Google Scholar] [CrossRef]
  14. Fontagné, Lionel, Pamina Koenig, Florian Mayneris, and Sandra Poncet. 2013. Cluster policies and firm selection: Evidence from France. Journal of Regional Science 53: 897–922. [Google Scholar] [CrossRef] [Green Version]
  15. Foucault, Martial, Thierry Madies, and Sonia Paty. 2008. Public spending interactions and local politics. Empirical evidence from French municipalities. Public Choice 137: 57–80. [Google Scholar] [CrossRef]
  16. Hajnal, Zoltan L., and Jessica Trounstine. 2010. Who or what governs?: The effects of economics, politics, institutions, and needs on local spending. American Politics Research 38: 1130–63. [Google Scholar] [CrossRef] [Green Version]
  17. Hawkins, Christopher V., Sung-Wook Kwon, and Jungah Bae. 2016. Balance between local economic development and environmental sustainability: A multi-level governance perspective. International Journal of Public Administration 39: 803–11. [Google Scholar] [CrossRef]
  18. Kitagawa, Fumi. 2007. The regionalization of science and innovation governance in Japan? Regional Studies 41: 1099–14. [Google Scholar] [CrossRef]
  19. Kolltveit, Kristoffer, and Jostein Askim. 2017. Decentralisation as substantial and institutional policy change: scrutinising the regionalisation of science policy in Norway. Science and Public Policy 44: 546–55. [Google Scholar] [CrossRef]
  20. Koschatzky, Knut, and Henning Kroll. 2007. Which side of the coin? The regional governance of science and innovation. Regional Studies 41: 1115–27. [Google Scholar] [CrossRef] [Green Version]
  21. Lanahan, Lauren. 2016. Multilevel public funding for small business innovation: A review of US state SBIR match program. Journal of Technology Transfer 41: 220–49. [Google Scholar] [CrossRef]
  22. Lanahan, Lauren, and Maryann P. Feldman. 2015. Multilevel innovation policy mix: A closer look at state policies that augment the federal SBIR program. Research Policy 44: 1387–402. [Google Scholar] [CrossRef]
  23. Laranja, M., E. Uyarra, and K. Flanagan. 2008. Policies for science, technology and innovation: Translating rationales into regional policies in a multi-level setting. Research Policy 37: 823–35. [Google Scholar] [CrossRef]
  24. Laurian, Lucie, and Jan Crawford. 2016. Organizational factors of environmental sustainability implementation: An empirical analysis of US cities and counties. Journal of Environmental Policy & Planning 18: 482–506. [Google Scholar]
  25. Laurian, Lucie, Mark Walker, and Jan Crawford. 2017. Implementing environmental sustainability in local government: The impacts of framing, agency culture, and structure in US cities and counties. International Journal of Public Administration 40: 270–83. [Google Scholar] [CrossRef]
  26. Leydesdorff, Loet, and Henry Etzkowitz. 1996. Emergence of a Triple Helix of university-industry-government relations. Science and Public Policy 23: 279–86. [Google Scholar]
  27. Martin, Philippe, Thierry Mayer, and Florian Mayneris. 2011. Public support to clusters: A firm-level study of French “Local Productive Systems”. Regional Science and Urban Economics 41: 108–23. [Google Scholar] [CrossRef]
  28. McDonald, Bruce D. 2018. Local governance and the issue of fiscal health. State and Local Government Review. [Google Scholar] [CrossRef]
  29. Morisson, Arnault, and Mathieu Doussineau. 2019. Regional innovation governance and place-based policies: design, implementation and implications. Regional Studies, Regional Science 6: 101–16. [Google Scholar] [CrossRef] [Green Version]
  30. Nishimura, Junichi, and Hiroyuki Okamuro. 2011a. R&D productivity and the organization of cluster policy: An empirical evaluation of the Industrial Cluster Project in Japan. Journal of Technology Transfer 38: 117–44. [Google Scholar]
  31. Nishimura, Junichi, and Hiroyuki Okamuro. 2011b. Subsidy and networking: The effects of direct and indirect support programs of the cluster policy. Research Policy 40: 714–27. [Google Scholar] [CrossRef] [Green Version]
  32. Nishimura, Junichi, and Hiroyuki Okamuro. 2016. Knowledge and rent spillovers through government-sponsored R&D consortia. Science and Public Policy 43: 207–25. [Google Scholar]
  33. OECD. 2011. Regions and Innovation Policy. Paris: OECD Publishing. [Google Scholar]
  34. Okamuro, Hiroyuki, and Junichi Nishimura. 2015. Local management of national cluster policies: Comparative case studies of Japanese, German, and French biotechnology clusters. Administrative Sciences 5: 213–39. [Google Scholar] [CrossRef]
  35. Okamuro, Hiroyuki, and Junichi Nishimura. 2018. Whose business is your project? A comparative study of different policy schemes for collaborative R&D. Technological Forecasting and Social Change 127: 85–96. [Google Scholar]
  36. Okamuro, Hiroyuki, and Junichi Nishimura. 2020. Effects of Multilevel Policy Mix of Public R&D Subsidies: Empirical Evidence from Japanese Local SMEs. CCES Discussion Paper Series No. 70. Tokyo: Graduate School of Economics, Hitotsubashi University, January. [Google Scholar]
  37. Okubo, Toshihiro, Tetsuji Okazaki, and Eiichi Tomiura. 2016. Industrial Cluster Policy and Transaction Networks: Evidence from Firm-Level Data in Japan. RIETI Discussion Paper Series 16-E-071. Tokyo: Research Institute of Economy, Trade and Industry (RIETI). [Google Scholar]
  38. Perry, Beth. 2007. The multi-level governance of science policy in England. Regional Studies 41: 1051–67. [Google Scholar] [CrossRef] [Green Version]
  39. Perry, Beth, and Tim May. 2007. Governance, science policy and regions: An introduction. Regional Studies 41: 1039–50. [Google Scholar] [CrossRef]
  40. Salazar, Monica, and Adam Holbrook. 2007. Canadian science, technology and innovation policy: The product of regional networking? Regional Studies 41: 1129–41. [Google Scholar] [CrossRef]
  41. Sharif, Naubahar, and Jack Linzhou Xing. 2019. Restricted generalizability of city innovation policies: The case of E-hailing in China. Science and Public Policy 46: 805–19. [Google Scholar] [CrossRef]
  42. Sotaraura, Markku, and Mika Kautonen. 2007. Co-evolution of the Finnish national and local innovation and science arenas: Towards a dynamic understanding of multi-level governance. Regional Studies 41: 1085–98. [Google Scholar] [CrossRef]
  43. Tödtling, Franz, and Michaela Trippl. 2005. One size fits all? Towards a differentiated regional innovation policy approach. Research Policy 34: 1203–19. [Google Scholar] [CrossRef]
  44. Töpfer, Stefan, Holger Graf, and Uwe Cantner. 2019. Structural dynamics of innovation networks in German Leading-Edge Clusters. Journal of Technology Transfer 44: 1816–39. [Google Scholar] [CrossRef] [Green Version]
1
This relationship has been discussed in the literature on regional studies as an important issue with the keyword of multi-level governance (MLG) since the special issue in Regional Studies in November 2007, which we will discuss later in more detail.
2
The German study (Koschatzky and Kroll 2007) provides a case study of Bremen, which they do not regard as a city, but as a federal state.
3
As an exception, Falck et al. (2010) empirically analyze the firm-level impact of the regional cluster policy of the Federal State of Bavaria in Germany, without considering the roles of cities. Moreover, Okamuro and Nishimura (2015) provide a descriptive comparison between Germany, France, and Japan on the role of local cluster management in national cluster policies with case studies of biotechnology cluster areas.
4
Lanahan and Feldman (2015) use only the measures of high-tech employment and higher education capacity as the variables for bottom–up factors on the demand side, and so ignore regional industry characteristics.
5
We do not target towns and villages because these municipalities rarely implement their own R&D subsidy programs.
6
In fact, we conducted this survey both for city and prefecture government officers. We confirmed 112 programs in 42 prefectures and obtained response for 90 programs (80.4%). In the following estimations, however, we use city data only and do not use prefecture data, as program implementation and design are markedly different between prefectures and cities, and because the variety of local programs appears much larger across cities than prefectures. Thus, we focus on city governments’ R&D subsidy programs that are unique for regional innovation systems in Japan.
7
Because we consider the determinants of R&D subsidy policies, we do not use programs, but cities as the units of our estimations. Thus, if a city implements two or more R&D subsidy programs, which is the case for some cities, we focus on the program with the largest budget.
8
We use the number of universities in the same prefecture (and not in the same city) because partner universities of private firms are often located outside of the city.
9
This measure denotes the degree of financial soundness of local authorities. A higher (lower) value of this variable means higher (lower) deficit.
10
Alternatively, we used the dummy variable for ordinance-designated cities that are empowered to issue their own ordinances. However, estimation results did not change significantly.
Table 1. Basic statistics of all cities (including Tokyo Wards) in Japan.
Table 1. Basic statistics of all cities (including Tokyo Wards) in Japan.
VariablesObsMeanStd. Dev.MinMax
d_subsidy7720.170.3701
share of manufacturing7729.405.102.0836.36
high-tech ratio77229.2412.87067.86
ln (manufacturing productivity)7726.950.475.038.85
# large manufacturing firms77215.2363.120945
# research institutes7728.1119.540211
# universities in prefecture77226.2431.332132
ln (# officers per capita)772−5.250.32−5.95−4.26
share of industry expend.7722.562.540.1625.58
current account77291.366.4058.63106.6
ln (population size)77211.330.898.5615.09
prefecture capital7720.050.2201
Table 2. Correlation matrix of variables used in the first empirical analysis.
Table 2. Correlation matrix of variables used in the first empirical analysis.
Obs = 772123456
1d_subsidy1
2share of manufacturing0.061
3high-tech ratio0.050.181
4ln (manufacturing productivity)0.060.070.381
5# large manufacturing firms0.08−0.07-0.010.041
6# research institutes0.17-0.110.060.100.811
7# universities in prefecture0.04−0.010.300.030.310.24
8ln (# officers per capita)−0.09−0.10−0.48−0.30−0.10−0.21
9share of industry expend.0.170.05−0.11−0.020.110.13
10current account−0.12−0.10−0.17−0.16−0.23−0.19
11ln (population size)0.27−0.070.310.280.400.56
12prefecture capital0.10−0.16−0.030.060.330.47
789101112
7# universities in prefecture1
8ln (# officers per capita)−0.341
9share of industry expend.−0.210.111
10current account−0.270.190.011
11ln (population size)0.33−0.670.01−0.211
12prefecture capital−0.09−0.220.140.020.491
Table 3. Estimation results on the propensity of the R&D subsidy program (probit model).
Table 3. Estimation results on the propensity of the R&D subsidy program (probit model).
VariablesR&D Subsidy Program
Marginal EffectRobust S.E.
share of manufacturing0.006 ***0.002
high-tech ratio−0.0000.001
ln (manufacturing productivity)−0.0090.029
# large manufacturing firms−0.001 **0.000
# research institutes0.0010.001
# universities in prefecture−0.0000.000
ln (# officers per capita)0.216 ***0.061
share of industry expend.0.019 ***0.005
current account−0.005 ***0.002
ln (population size)0.165 ***0.023
prefecture capital−0.0820.038
N772
Wald chi2102.15 ***
Log pseudolikelihood−298.456
Pseudo R20.143
(1) This table shows marginal effects instead of coefficients. (2) Level of significance: *** 1%, ** 5%.
Table 4. Basic statistics of cities that implemented R&D subsidy programs in 2015.
Table 4. Basic statistics of cities that implemented R&D subsidy programs in 2015.
VariablesObsMeanStd. Dev.MinMax
subsidy longer than 1 year1200.340.4801
upper limit of subsidy ratio above 1/21170.430.5001
personal expenses covered1180.380.4901
overlap with other subsidy programs possible1150.370.4801
share of manufacturing12010.095.263.7236.36
high-tech ratio12030.5112.433.8855.43
ln (manufacturing productivity)1207.020.435.928.22
# large manufacturing firms12027.5455.370353
# research institutes12015.8030.090211
# universities in prefecture12028.3134.672132
ln (# officers per capita)120−5.310.34−5.86−4.52
share of industry expend.1203.532.740.2413.56
current account12089.417.3661.07105.40
ln (population size)12011.871.099.5815.09
prefecture capital1200.110.3101
Table 5. Correlation matrix of variables used in the second empirical analysis.
Table 5. Correlation matrix of variables used in the second empirical analysis.
Obs = 11012345678
1subsidy longer than 1 year1
2upper limit of subsidy ratio above 1/20.151
3personal expenses covered0.09−0.051
4overlap with other subsidy programs possible0.23−0.100.151
5share of manufacturing0.000.080.130.081
6High-tech ratio0.16−0.120.210.070.151
7ln (manufacturing productivity)−0.08−0.060.04−0.12−0.060.521
8# large manufacturing firms0.090.250.19−0.14−0.070.120.121
9# research institutes0.030.110.27−0.15−0.160.200.170.73
10# universities in prefecture−0.080.250.09−0.050.200.10−0.020.48
11ln (# officers per capita)−0.07−0.10−0.170.09−0.07−0.51−0.54−0.35
12share of industry expend.0.080.110.05−0.120.19−0.11−0.140.00
13current account−0.020.00−0.040.04−0.28−0.30−0.11−0.28
14ln (population size)0.060.180.23−0.18−0.130.400.440.59
15prefecture capital0.130.080.00−0.15−0.23−0.020.070.29
9101112131415
9# research institutes1
10# universities in prefecture0.181
11ln (# officers per capita)−0.36−0.301
12share of industry expend.0.03−0.240.151
13current account−0.04−0.580.380.241
14ln (population size)0.610.36−0.76−0.11−0.251
15prefecture capital0.38−0.17−0.250.050.130.431
Table 6. Estimation results on the design of R&D subsidy programs.
Table 6. Estimation results on the design of R&D subsidy programs.
VariablesModel 1Model 2Model 3Model 4
Subsidy Longer than 1 yearUpper Limit of Subsidy Ratio above 1/2Personal Expenses CoveredOverlap with Other Subsidy Programs Possible
share of manufacturing−0.0090.024 **0.008−0.003
0.0090.0090.0100.011
high-tech ratio0.014 ***−0.0080.009 **0.008 *
0.0040.0050.0040.005
ln (manufacturing productivity)−0.418 ***−0.112−0.209−0.236 *
0.1430.1750.1730.138
# large manufacturing firms0.003 **0.004 **0.002 *0.000
0.0010.0020.0010.001
# research institutes−0.004 **−0.0050.004−0.003
0.0020.0040.0030.003
# universities in prefecture−0.0030.006 **−0.002−0.000
0.0020.0030.0020.002
ln (# officers per capita)−0.167−0.0310.032−0.096
0.2550.2970.2530.251
share of industry expend.0.0140.0180.012−0.026
0.0160.0210.0190.019
current account−0.0060.027 **−0.0040.001
0.0090.0110.0090.009
ln (population size)−0.0060.0850.143−0.045
0.0940.1130.1040.090
prefecture capital0.2310.239−0.262−0.192
0.2030.2020.1390.140
business field dummiesyesyesyesyes
N120117118115
Wald chi228.47 **31.95 **29.39 **25.78 *
Log pseudolikelihood−64.849−58.548−64.689−65.208
Pseudo R20.1600.2670.1750.146
(1) This table shows marginal effects instead of coefficients. (2) Level of significance: *** 1%, ** 5%, * 10%. (3) Robust standard errors of coefficients in italics under marginal effects.
Table 7. Heckman’s sample selection model.
Table 7. Heckman’s sample selection model.
VariablesModel 1Model 2Model 3Model 4
Subsidy Longer than 1 yearUpper Limit of Subsidy Ratio above 1/2Personal Expenses CoveredOverlap with Other Subsidy Programs Possible
share of manufacturing0.0160.0210.0730.013
0.0420.0260.1160.025
high-tech ratio0.013 *−0.0060.0040.005
0.0070.0040.0190.006
ln (manufacturing productivity)−0.379 *−0.098−0.251−0.196 *
0.2040.1270.5500.111
# large manufacturing firms0.0000.002−0.0060.002
0.0050.0030.0140.004
# research institutes−0.000−0.0020.0100.002
0.0070.0040.0190.005
# universities in prefecture−0.0040.005 **−0.005−0.001
0.0040.0020.0110.003
ln (# officers per capita)0.7060.0292.3920.561
1.5190.9844.1101.080
share of industry expend.0.0760.0140.1760.029
0.1120.0720.2950.089
current account−0.0280.016−0.062−0.018
0.0410.0280.1090.034
ln (population size)0.6070.0941.7830.435
1.0540.6792.8110.822
prefecture capital−0.0940.202−1.037−0.432
0.6200.3541.5580.512
business field dummiesyesyesyesyes
N120117118115
Wald chi228.45 **51.27 ***2.1725.13 *
Mills lambda1.0950.0272.8650.829
1.8041.2264.7961.361
(1) We show only the estimation results in the second stage. (2) This table shows marginal effects instead of coefficients. (3) Level of significance: *** 1%, ** 5%, * 10%. (4) Robust standard errors of coefficients in italic under marginal effects.
Table 8. Structural estimation modeling.
Table 8. Structural estimation modeling.
VariablesModel 1Model 2Model 3Model 4
Subsidy Longer than 1 yearUpper Limit of Subsidy Ratio above 1/2Personal Expenses CoveredOverlap with Other Subsidy Programs Possible
share of manufacturing−0.0030.020 ***0.015 *0.005
0.0090.0070.0090.009
high-tech ratio0.012 ***−0.0060.008 **0.009 **
0.0040.0040.0040.004
ln (manufacturing productivity)−0.362 ***−0.139−0.246 *−0.218 *
0.1240.1300.1450.123
# large manufacturing firms0.002 **0.002 **0.001−0.000
0.0010.0010.0010.001
# research institutes−0.004 **−0.0010.000−0.001
0.0020.0020.0020.002
# universities in prefecture−0.0020.004 ***−0.001−0.000
0.0020.0010.0020.002
ln (# officers per capita)−0.191−0.040−0.071−0.042
0.2350.2290.2270.206
share of industry expend.0.0080.0120.006−0.028
0.0140.0160.0180.016
current account−0.0020.016 **0.0030.006
0.0100.0070.0070.008
ln (population size)−0.0110.0830.064−0.064
0.0790.0800.0910.084
prefecture capital0.284 *0.237−0.024−0.096
0.1610.1580.1640.139
business field dummiesyesyesyesyes
N110
Log pseudolikelihood−3854.191
cov (e1, e2) =0.024 *cov (e2, e3) =−0.023
cov (e1, e3) =0.009cov (e2, e4) =−0.017
cov (e1, e4) =0.036 **cov (e3, e4) =0.034 **
(1) This table shows marginal effects instead of coefficients. (2) Level of significance: *** 1%, ** 5%, * 10%. (3) Robust standard errors of coefficients in italics under marginal effects.
Table 9. Bivariate probit estimations.
Table 9. Bivariate probit estimations.
VariablesModel 1Model 2Model 3
Subsidy Longer than 1 yearUpper Limit of Subsidy Ratio above 1/2Subsidy Longer than 1 yearOverlap with Other Subsidy Programs PossiblePersonal Expenses CoveredOverlap with Other Subsidy Programs Possible
share of manufacturing−0.0170.065 ***−0.021−0.0090.024−0.010
0.0260.0240.0260.0300.0280.030
high−tech ratio0.040 ***−0.0200.038 ***0.025 **0.025 **0.026 **
0.0130.0120.0130.0120.0120.013
ln (manufacturing productivity)−1.345 ***−0.332−1.128 ***−0.608 *−0.696 *−0.640 *
0.4080.4390.4020.3670.4040.384
# large manufacturing firms0.008 **0.008 *0.008 **0.0000.003−0.001
0.0040.0050.0040.0030.0030.003
# research institutes−0.014 **−0.009−0.013 **−0.009−0.001−0.009
0.0060.0090.0060.0070.0060.006
# universities in prefecture−0.0070.014 **−0.007−0.001−0.0030.000
0.0060.0060.0060.0050.0050.005
ln (# officers per capita)−0.655−0.144−0.529−0.332−0.471−0.322
0.6990.7600.7010.6450.6870.672
share of industry expend.0.0260.0390.033−0.0670.032−0.066
0.0460.0530.0470.0520.0530.050
current account−0.0080.064 **−0.0110.006−0.0000.012
0.0290.0260.0280.0230.0230.025
ln (population size)0.0070.221−0.069−0.1470.123−0.164
0.2530.2840.2540.2340.2590.241
prefecture capital0.7690.6690.836 *−0.539−0.157−0.566
0.5060.4960.4980.5210.4960.533
business field dummiesyesyesyesyesyesyes
N117115113
Wald chi281.82 ***62.51 ***59.24 ***
Log pseudolikelihood−128.208−133.215−135.106
Wald test of rho=01.8924.098 **5.739 **
(1) This table shows marginal effects instead of coefficients. (2) Level of significance: *** 1%, ** 5%, * 10%. (3) Robust standard errors of coefficients in italics under marginal effects.

Share and Cite

MDPI and ACS Style

Okamuro, H.; Nishimura, J. What Shapes Local Innovation Policies? Empirical Evidence from Japanese Cities. Adm. Sci. 2020, 10, 11. https://doi.org/10.3390/admsci10010011

AMA Style

Okamuro H, Nishimura J. What Shapes Local Innovation Policies? Empirical Evidence from Japanese Cities. Administrative Sciences. 2020; 10(1):11. https://doi.org/10.3390/admsci10010011

Chicago/Turabian Style

Okamuro, Hiroyuki, and Junichi Nishimura. 2020. "What Shapes Local Innovation Policies? Empirical Evidence from Japanese Cities" Administrative Sciences 10, no. 1: 11. https://doi.org/10.3390/admsci10010011

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop