CESIS Electronic Working Paper Series

Paper No. 40

Knowledge Accessibility and Regional Economic Growth 1

Martin Andersson and Charlie Karlsson

(JIBS)

Sept 2005

1 Status of the paper: The Paper is submitted to a Journal.

- 1 -

Knowledge Accessibility and Regional Economic Growth

- a study on Swedish municipalities

by

Martin Andersson & Charlie Karlsson Department of Economics & CESIS

Jönköping International Business School P.O. Box 1026

S-551 11 Jönköping Sweden

Phone: +46 36 15 77 00 Fax: +46 36 12 18 32

E-mail: Martin.Andersson@jibs.hj.se Charlie.Karlsson@jibs.hj.se

Abstract

In this paper, we analyze the relationship between knowledge accessibility and regional economic growth in Sweden. The research question we ask in this paper is the following: can the variation between regions in knowledge accessibility in a given period explain the variation in the growth performance of Swedish regions in the following period? A main assumption in the paper is that the potential for human interaction a various spatial scales transforms into potential knowledge flows. Based upon an endogenous growth framework we show how the potential for knowledge flows at different spatial scales can be modelled using an accessibility approach. An advantage of modelling the interaction potentials at different spatial scales is that we in principle takes care of the spatial auto-correlation in the modelling, and that we then can estimate our models using ordinary OLS. Our results indicate that the intra-municipal and intra-regional knowledge accessibilities are both significant and capable of explaining a significant share of the variation in growth of value added per employee between Swedish municipalities. However, inter-regional knowledge accessibility turned out to be insignificant, indicating that knowledge flows are bounded in space.

Key words: knowledge, endogenous growth, region, accessibility, Sweden JEL codes: O10, O52, R11, O30

- 2 -

1. Introduction A rich literature has in recent decades grown up around the concept of the rise of the importance of regions in the global economic system (see, for example, Johansson, Karlsson & Stough, 2001, eds. & 2002, eds.). According to this view, global trade has the form of interaction between functional urban regions, rather than between countries. As this vision has evolved, empirical observations suggest that global and national economic change should be understood as a process which is dependent on dynamics operating at the local level. Perceptions and models of such change can be underpinned with various theoretical perspectives generally referred to as the new endogenous growth theory. The overall objective of the theoretical research on endogenous growth theory has been to develop dynamic general equilibrium models with precisely formulated microeconomic foundations. This enables a clearer understanding of how processes such as physical and human capital accumulation, innovation, knowledge and product differentiation, impact long-run economic growth.

The term endogenous implies that economic growth is influenced by the use of “investment resources” generated by the economy itself – in contradiction to the reference made to exogenous factors in the traditional Solow type growth models (Solow, 1956 & 1957). In aggregate macro models this form of endogeneity implies that investments in production capital, infrastructure capital, knowledge capital (R&D) and human capital (embodied knowledge capital) affect the growth rate of the economy. In particular, concepts related to knowledge generation, knowledge accumulation, knowledge appropriation and knowledge flows are prominent features of these models. Hence, education, learning, R&D and technical innovation play a fundamental role in the economic growth process. Today a large number of scholars agree that models of endogenous economic growth and dynamic increasing returns have a prominent role to play in the analyses of contemporary economies at both the national and the regional level. Moreover, according to endogenous growth models, policy matters – both regarding the supply of public services and regarding investments in tangible and intangible infrastructure.

Even though there are macro mechanisms that affect the growth of productivity and income per capita at the level of functional urban regions, there is a need to detect the micro mechanisms that are working at the regional level. It is often maintained that innovation and thus economic growth is a localized process. Innovation in each locality is among other things a function of the local technological level (Rodríguez-Pose, 2001), local knowledge generation in firms and in universities, local knowledge flows and knowledge flows from other localities in the region and from other regions. Much knowledge is embedded in human beings and this indicates that knowledge flows are a function of the mobility and interaction of people with the relevant knowledge, skills and experiences. Due to the “tyranny of distance”, most of the human interaction is bounded to the functional region and in particular to the locality where people live and work.

Presenting the growth problem in this way it is obvious that local decision makers face a substantial uncertainty which policies, if any, and which policy mixes to pursue if they want to stimulate local economic growth. Should they invest to increase local university R&D? Should they invest to increase local higher education? Should they invest to increase both local university R&D and local higher education to take advantage of possible synergy effects between university R&D and higher education? Should they invest in local amenities to increase the in-migration of

- 3 -

highly educated labor? Should they invest to reduce local travel times and travel cost? Should they invest in local interaction arenas? Or should they invest to reduce intra- and interregional travel times and travel costs to facilitate knowledge flows from other localities in the region and from other regions?2 Obviously, local policy makers face very interesting trade-offs, if they want to stimulate local economic growth via a higher rate of innovation (and imitation). Furthermore, interesting conflicts between local and national interests might exist, if there are economies of scale in R&D and higher education. Of course, it is a major research task to analyze how the various local knowledge processes is transformed into economic growth, and which local policies and policy mixes that might stimulate economic growth in various types of localities.

In this paper we analyze the relationship between knowledge accessibility and regional economic growth in Sweden. It is clearly the case that the accessibility to various types of knowledge resources differs between regions in Sweden. The question we ask in the present paper is the following: can the variation in this accessibility in a given period explain the variation in the growth performance of Swedish regions in subsequent periods? A main assumption in the paper is that the potential for interaction at various spatial scales transforms into potential knowledge flows. Hence, the local industry in regions with high accessibility to knowledge resources is expected to have a higher potential of exploiting such resources. To make the concept of potential interaction operational, accessibility measures are employed. Three types of knowledge resources are considered in the paper: (i) private R&D, (ii) university R&D and (iii) patent applications. Regional growth is measured by the growth in value-added per employee.

The outline of the paper is as follows: In Section 2 we introduce theoretically how knowledge accessibility can be integrated into various models of endogenous economic growth. The data used in the empirical analyses and our empirical analyses can be found in Section 3 and in Section 4 we summarize our conclusions and give suggestions for future research.

2. Knowledge Accessibility and Endogenous Economic Growth Solow (1956) and Swan (1956) developed what was to become the standard model of neo-classical growth theory. The key aspect of this model was the form of the production function, which assumed constant returns to scale, diminishing returns to each input, and some positive smooth elasticity of substitution between inputs (Barro & Sala-i-Martin, 1995). One important prediction from this model is that of conditional convergence deriving from the assumption of diminishing returns to capital. The lower the starting level of real GDP/capita, relative to the long run or steady-state position, the faster is the rate of growth. The issue of income convergence has during the years attracted a lot of empirical research.

However, the traditional neo-classical model of economic growth has not much to say about the major sources of long-run growth. Long-run economic growth is explained as a function of technological progress, which is assumed to be an exogenous factor. Dissatisfaction with the 2 One could also perhaps think of a case in which policy makers take a “free-rider “ position and do nothing, hoping that knowledge investments in other localities and regions will be transferred and spillover to an extent that is large enough to achieve the local growth targets. As a free rider, resources could instead be used for consumption or to other growth enhancing investments.

- 4 -

major determinant of long-run economic growth being treated as an exogenous factor and with the assumption of decreasing returns led to the development of endogenous growth models, where the rate of technical progress is determined within the model and where increasing returns exist.

2.1 Endogenous Growth Theory A natural starting point for gaining an increased theoretical understanding of the emerging knowledge economy is the new endogenous growth theory, which emphasizes the role of the stock of accumulated knowledge and the growth of this stock. It suggests that continuous, endogenously generated increases in the stock of technological knowledge (Romer, 1990) or in human capital accumulation, i.e. the embodiment of knowledge in human beings (Lucas, 1988) are the driving force behind economic growth. However, it takes for granted that economic growth is not emerging automatically as ‘manna from heaven’, but is the result of deliberate actions and choices of various stake-holders, including the government, which means that it can be promoted by public policy (Nijkamp, 2003). Lucas’ conceptualization of the process by means of which human capital is built up is simple. Economic agents decide according to their preferences for higher wages in the future about the allocation of their non-leisure time between current production and the accumulation of human capital. Thereby they also determine the growth rate of output. If the time spent on current production is reduced current output is reduced but at the same time it speeds up the formation of human capital and thus increases future growth. The Lucas approach includes an externality due to increasing returns in the sense that the more human capital society as a whole has accumulated, the more productive each individual member of society will be.

A distinguishing feature of the Romer approach is the modeling of technological progress as a result of profit-motivated investments in production of technological knowledge by private economic agents. These economic agents are acting in an economic environment characterized by monopolistic competition along the lines suggested by Dixit & Stiglitz (1977), which explicitly considers the trade-off between the output of goods and their variety. Markets are assumed to be imperfectly rather than perfectly competitive, because investments in R&D are feasible only if price exceeds production cost by some margin. Another distinguishing feature is the existence of increasing returns, since technological knowledge is treated as a non-rival, partially excludable good. Firms are able to maintain ownership of at least a portion of the value of the increased productivity or better product performance won through their R&D (Nelson, 1997). In such a framework, each firm developing new technological knowledge gain some market power and thus can earn some monopoly profits on its investments. According to the Romer approach, knowledge spillover compensates for private diminishing returns in investment in knowledge creation by increasing the stock of knowledge in the economy as a whole.

The spillover mechanism can easily be illustrated in a “learning-by-doing” variant of the AK-model in line with Romer (1986). The output per worker of firm j can be written as:

( ) αα −= 1kkAy jj (2.1)

- 5 -

In Equation (2.1), kj is firm j’s level of capital per worker and k is the average level of capital per worker in the economy as a whole. The set-up implies that the an individual firm faces diminishing returns to its own investment but the function exhibits constant returns to scale in kj and k taken together. The production function in Equation (2.1) implies that the marginal product of capital is increasing in the average capital intensity of the economy. The raison d’être for this is that higher capital intensity increases the extent of knowledge (or technological) spillovers (Obstfeld & Rogoff, 2002). In general, spillover mechanisms can be applied to new technological knowledge, since it is used in two ways in the economy:

• It is used in the production of a specific unique product by the firm who devel-

oped it. The use of this specific knowledge by another firm for producing the same product can be protected by means of, for example, patenting.

• It increases the total stock of technological knowledge and may spillover to other

firms investing in the production of technological knowledge by means of, for example, examinations of patent documentation (Romer. 1990). In this way it increases the productivity of knowledge production in the economy and it may very well be so that new knowledge benefits others as much or even more than they benefit the creator of the new knowledge (Quigley, 1998).

Various attempts have been made in the literature to endogenize technological progress in models of economic growth. Primary examples are Arrow (1962), Romer (1986) and Lucas (1988). 2.2 The Role of Knowledge Accessibility A limiting factor with these original approaches is the assumption of general accessibility of the stock of knowledge across space. If knowledge is not easily accessible at every point in space, the location of knowledge production and the characteristics of knowledge flows become critical issues in understanding economic growth. However, there are strong reasons to believe that the stock of knowledge is not evenly accessible across countries or even across functional regions within countries. New knowledge is often extremely complicated and contains complex (and sometimes tacit) elements which imply that it often only is accessible via interactions within either inter-firm innovation networks or general innovation systems that tend to be bounded by geographical proximity (Karlsson, 1997; Karlsson and Manduchi, 2001; Andersson and Karlsson, 2004 a & b). Strong evidence is also provided for both the US and Europe that knowledge flows measured by patent citations are bounded within a relatively narrow geographical range (Jaffe, Trajtenberg & Henderson, 1993; Almeida & Kogut, 1999; Maurseth & Verspagen, 1998; Verspagen & Schoenmakers, 2000). Of particular concern is also the volume of human capital engaged in the generation of new ideas, innovations and technologies.

The implications of these factors are far-reaching. Functional regions will differ not only in terms of their production of and access to knowledge but the mix of knowledge will also be different between functional regions. Thus, important elements of the production of knowledge will tend to be regional rather than national. This will probably have its strongest effects on sci-ence-based and high-technology industries but will in principle influence all industries. Empirical analyses also show that the production of new scientific and technological knowledge has a predominant tendency to cluster spatially (Varga, 1999; Caniëls, 2000). Sensitivity of the

- 6 -

transmission of new knowledge to distance seems to provide a principal reason for the devel-opment of regional innovation clusters (Acs, Anselin & Varga, 2002). Hence, it is natural that in the regional development literature, the geographical distribution of knowledge workers is hypothesized to be a key driver of existing and future patterns of regional growth (Nijkamp & Poot, 1998; Bal & Nijkamp, 1998; Mathur, 1999; Florida, 2000 & 2002). This implies that the kinds of work the regional economy does deserve at least as much attention as the kinds of products it makes (Thompson & Thompson, 1985 & 1987; Feser, 2003).

Recently, some economists have suggested an important link between national economic growth and the concentration of people and firms in large urban regions. The high concentration of people and firms in large urban regions creates an environment in which knowledge moves quickly from person to person and from firm to firm. This implies that large, dense locations encourage knowledge flows and knowledge exchange, thus facilitating the spread of new knowledge that underlies the creation of new goods and new ways of producing existing goods (Carlino, 2001). Glaeser, et al. (1992) show that localized inter-industry knowledge flows can explain the economic growth of US cities. Cheshire & Carbonaro (1995) present an analysis in a regional context which embodies increasing returns to human capital as a result of spillovers which occur due to the non-rival and partially non-excludable component of knowledge generation. They model the rate of growth of non-rival knowledge as a function of the total human capital that is employed in research multiplied with the stock of knowledge, allowing for the differential concentration of human capital among regions. Moreover, Fujita & Thisse (2003) show theoretically that the growth of the global economy depends on the spatial organization of the innovation sector across regions.

Given these considerations, it is apparent that there is a need for a modelling approach that can mirror existing variations within and between functional urban regions in terms of knowledge generation and conditions for intra- and inter-regional knowledge flows. Knowledge flows are related to the mobility and interaction of people. Thus, spatial proximity is generally assumed to be instrumental in facilitating knowledge flows among actors3. Given that mobility and interaction are time-consuming it is natural that mobility and interaction vary between different geographical scales, such as the local scale, the intra-regional scale and the inter-regional scale.

Against this background, Karlsson & Manduchi (2001) have suggested the use of accessibility measures to make the role of mobility and interaction patterns in knowledge production functions operational. What are then the benefits of accessibility measures? Weibull (1980) maintains that accessibility measures can be seen as measures of:

• Nearness

• Proximity

• Ease of spatial interaction

• Potential of opportunities of interaction

• Potentiality of contacts with activities or suppliers.

3 In this context, a common apprehension is that the most recent, and as such the most valuable type of knowledge, tends to have such a complex, uncertain and non-codified form that it can not be fully articulated and may only be transferred through personal interactions (Polanyi, 1996; Dosi, 1988; Feldman, 1994)

- 7 -

Physical intra- and interregional infrastructure

Location pattern of knowledge resources

Knowledge accessibility Potential for knowledge flows

This paper concentrates on the fourth interpretation, i.e. accessibility as a measure of potential of opportunities of interaction. As outlined in Figure 2.1, the potential for interaction at various spa-tial scales - measured by accessibility - is assumed to transform into potential knowledge flows.

Figure 2.1. Knowledge accessibility and potential for knowledge flows.

Knowledge accessibility is formed by (i) the location pattern of knowledge resources and (ii) the physical intra- and interregional infrastructure. A region with good physical infrastructure to regions with many knowledge resources will have high knowledge accessibility. Therefore, the extent of knowledge flows is expected to be large.

What is then meant by knowledge flows? Here, the term is used as a comprehensive term for different types of flows of knowledge. Figure 2.2, adapted from Johansson (2004), provides a general classification scheme of such flows. Firstly, knowledge flows can be purely transaction-based. In this case, there is an explicit agreement of transaction of knowledge between the parties involved. Such transactions can either be subject to monetary payments of knowledge or be constituted by R&D cooperation in which case the parties share losses and profits in some pre-specified fashion (cf. Johansson, 2004). Secondly, knowledge may flow in the form of knowledge spillovers, i.e. unintended side effects of ordinary activities. Such spillovers can in turn be divided into (i) spillovers mediated by market mechanisms and (ii) spillovers as pure externalities. Hence, in terms of the characteristics (i) is equivalent to pecuniary externalities and (ii) to technological externalities.

- 8 -

Figure 2.2. Classification of knowledge flows, (based on Johansson, 2004).

Market-mediated knowledge spillovers occur for example via the labor market and as a by-product of the purchasing and the selling of goods. For instance, a seller gains knowledge from a standard transaction with a customer. Knowledge spillovers as pure externalities occur, for example, when firms observe, e.g., certain routines and techniques and copy or imitate each other.

The next section presents the formal definition of accessibility used in the paper and makes a distinction between different types of accessibility.

2.3 A Distinction between Different Types of Accessibility

The starting point for a distinction between different types of accessibility is that a national economy can be divided into functional regions that consist of one or several localities. In this paper, such localities are labelled municipalities. Functional regions are connected to other functional regions by means of economic and infrastructure networks. The same prevails for the different localities (or municipalities) within a functional region. Moreover, each municipality can also be looked upon as a number of nodes connected by the same type of networks. The borders between functional regions are characterized by a decline in the intensity of economic interaction including commuting compared to the intraregional interaction. Thus, functional re-gions can be approximated with labor market regions.

With reference to such a structure, it is possible to define three different spatial levels with different characteristics in terms of mobility and interaction opportunities. Because of this, it is also possible to construct three different categories of accessibility. Johansson, Klaesson & Olsson (2002) separate between: (i) intra-municipal accessibility, (ii) intra-regional accessibility and (iii) extra-regional accessibility. Based on commuting data, they also show that the time sensitivity parameter λ is different for intra-municipal, intra-regional and extra-regional interaction. Inside a municipality parameter 1λ applies, inside the pertinent region parameter 2λ applies and for contacts outside the region parameter 3λ applies. These differ in size in the following way: 132 λλλ >> .

Knowledge spillovers

Mediated by market mechanisms (pecuniary)

Pure externalities (technological)

Knowledge flows

Transaction-based flows

- 9 -

In order to explain the three different accessibility measures in more detail, one has to start at the municipality level. The focus is on municipality s in a functional region R , so that Rs ∈ . The average time distance between zones in municipality s is denoted by sst and the size of the opportunity D in the same municipality is given by sD . From this, the intra-municipal accessibility to the opportunity sD is calculated as follows:

{ } sssDsM DtA 1exp λ−= (2.5)

However, the economic actors in municipality s have also accessibility to the opportunity D in all other municipalities r that belong to region R . By letting srt denote the time distance between municipality s and r the intra-regional accessibility of municipality s can be expressed as:

{ } rsrsrRrDsR DtA 2

, exp λ−∑= ≠∈ (2.6)

Finally, economic actors such as firms and households in municipality s also have accessibility to the opportunity kD in the k municipalities outside region R . This extra- or inter-regional accessibility is specified in formula (2.7):

{ } kskRkDsE DtA 3exp λ−∑= ∉ (2.7)

Di is here a measure of opportunities in each municipality and can relate to opportunities such as suppliers, customers, supply of producer services, supply of educated labor, universities and R&D institutes, R&D activities, higher education, patents, etc. (see, inter alia, Klaesson, 2001). The accessibility measure of the type that discussed here satisfies certain criteria of consistency and meaningfulness (see e.g. Weibull, 1976).

2.4 Accessibility to Knowledge and Economic Growth – an empirical application

In this section we present a simple model of growth in output per worker, which incorporates knowledge accessibility in a simple fashion. We assume that each municipality s has an aggregate production function in which the technological progress is labour-augmenting or Harrod-neutral:

( ) αα −=

1sss LAKY s (2.8)

- 10 -

In Equation (2.8), AsLs is effective labor. Writing Equation (2.8) in terms of labour gives us:

αα −= 1sss Aky

(2.9)

By taking logs and differentiating, the change in output per labour in region s, sy∆ , can be expressed as:

ss

ss

s

ss y

AA

ykk

y∆

−+∆

=∆ )1( αα (2.10)

Over a period of 6-7 years, it can be assumed that the change in capital intensity (i.e. capital per worker) is close to zero, i.e., 0≈∆ sk . This simplification implies that the change in output per worker is solely a function of the technological progress, as shown in Equation (2.11):

ss

ss y

AAy ∆

−=∆ )1( α (2.11)

The technological knowledge in a municipality is assumed to evolve according to:

⎟⎟⎠

⎞⎜⎜⎝

⎛=∆

s

sssss y

AAhA φσ )( (2.12)

where A denotes the stock of knowledge in municipality s and )( ss hσ is a productivity parameter. In Equation (2.12), this productivity parameter is expressed as a function of the human capital – in terms of education – in municipality s, sh . This formulation rests on the assumption that the absorptive capacity (c.f., Cohen & Levinthal, 1990) of the economic actors in a municipality increases with the workforce’s level of education. Thus, educated workers are expected to be better at exploiting the knowledge stock than non-educated workers. Moreover, Equation (2.12) implies that not only the stock of knowledge matters but also the knowledge intensity. Naturally, the knowledge stock has a positive effect on the change in knowledge, but the effect is larger the higher the knowledge intensity.

How is the knowledge stock As specified? We assume that the knowledge stock of a municipality in a given time period can be approximated by the municipality’s knowledge accessibility:

- 11 -

[ ]φβββ1

321WsE

WsR

WsMs AAAA ++≡ (2.13)

In Equation (2.13), the β’s represent the relative importance of each type of knowledge accessibility WA . This formulation implies that the knowledge resources in surrounding municipalities add to the knowledge stock of a given municipality. The strength of these effects is determined by the β’s. Substitution of (2.12) into (2.11) and making use of (2.13) leaves us with:

WsEss

WsRss

WsMsss AhAhAhy )()()( 321 σθσθσθ ++=∆

(2.14)

)1( αβθ −≡ ii

In (2.14), sy∆ is expressed as a function of the municipality’s accessibility to knowledge resources. The distinction between the three types of accessibilities makes it possible to estimate the influence from each type. This gives important information about the nature of knowledge flows, e.g. do they cross municipal borders? As stated previously, a main assumption in the paper is that the potential for interaction at various spatial scales transforms into potential knowledge flows.

3. The Relationship between Knowledge Accessibility and Regional Growth in Swedish Municipalities 3.1 Presentation of Variables and Data The equation to be estimated is presented in Equation (2.15). This empirical equation is based on Equation (2.14) but is expanded to include a number of control variables.

( ) ( ) ( )WsEs

WsRs

WsMss AAAy ×+×+×+=∆ ωθωθωθα 321 +…

(2.15)

sr srrsmigs DED εθθθ ++++ ∑ =

+ 80

1 ,5)(

4...

- 12 -

Table 3.1 explains and provides a description of the variables in Equation (2.15). As is evident from the table, most of the variables are constructed based on secondary material from Statistics Sweden (SCB).

Three different variables are used to measure knowledge accessibility: (i) university R&D, (ii) private R&D, (iii) the stock of patent applications. The data on patent applications comes from the European Patent Office (EPO)4. With exception of patent applications, the knowledge variables are all measured for the year 1993, which is the base period in the analysis. However, patent applications are available from EPO from 1977 and onwards. Therefore, we use the stock of applications. The R&D data originates from SCB. These data are collected by SCB via questionnaires that are sent out to firms and universities. The R&D data is measured in man-years. One man-year is the amount of work a full-time employee performs during a year. This means that a full-time employee who only spends 50 % of her work on R&D counts as 0.5 man-years.

The accessibility calculations are based on a Swedish travel time-distance matrix, which gives the minimum travel time by car: (i) between zones within municipalities and (ii) between municipalities. This matrix is provided by the Swedish Road Administration (SRA). As described in Section 2.3, three different values of the time-distance sensitivity parameter, λ, are used: (i) 0.02 for intra-municipal accessibility, (ii) 0.1 for intra-regional accessibility and (iii) 0.05 for extra-regional accessibility. These are the values found by Johansson, Klaesson & Olsson (2003), who estimated the value of the respective λ by using Swedish commuting data. These values represent the best information available. Moreover, following the discussion in Section 2.4, all the accessibility variables are weighted by the knowledge intensity of the municipality’s workforce. Thus, accessibility to knowledge is assumed to have a bigger effect if the absorptive capacity of the municipality’s workforce is large.

4 A patent application is classified as Swedish if at least one inventor is Swedish. The address(es) of Swedish inventor(s) has been used was used to allocate the applications to a specific municipality.

- 13 -

Table 3.1. Description of variables in Equation (2.15).

Variable Description Source

sy∆ Change in value-added per employee in municipality s 1993-2001. Statistics Sweden (SCB)

FRDs The number of R&D man years in 1993 of the private sector in municipality s. Statistics Sweden (SCB)

URDs The number of R&D man years in 1993 at the university(ies) in municipality s. Statistics Sweden (SCB)

Ps

The number of patent applications 1977-1993 filed to the EPO originating in municipality s.

European Patent Office (EPO)

AW Accessibility to variable W based on travel time distances by car.

The Swedish Road Administration (SRA) and SCB

sω Knowledge intensity of the workforce in municipality s5. Statistics Sweden (SCB)

)(+migsD

Dummy which takes the value 1 if municipality s experienced positive net migration 1993-2001, 0 otherwise.

Statistics Sweden (SCB)

sE Number of 1-person firms in municipality s normalized by employment in municipality s.

Statistics Sweden (SCB)

srD , Dummy which takes the value 1 if municipality s belongs to region r, 0 otherwise.

Statistics Sweden (SCB)

The specification in (2.15) also includes a dummy for whether the municipality has had a positive net migration under the period of investigation is included and an entrepreneurship variable, denoted by E. A large literature suggests that entrepreneurship spurs growth. As seen in Table 3.1, entrepreneurship is proxied by the number of 1-person firms. Although this is a crude proxy, it is frequently used in the literature (see e.g., Braunerhjelm & Borgman, 2004). In addition, Equation (2.15) also includes regional dummies to control for region-specific effects. The model includes 80 regional dummies6.

Table 3.2 presents some descriptive statistics of the variables in the empirical analysis, excluding

the dummy variables. The notation for the weighted accessibility variables follows

that in Section 2.4 and Equation (2.15). Inspection of the figures reveals that the

minimum value for all intra-regional accessibilities is zero. This is due to that some

5 The knowledge intensity is calculated as )1( sss δδω −= , where

sδ denotes the share of the employees in municipality s with a university education of at least three years. 6 There are 81 regions in Sweden.

- 14 -

regions only consist of one municipality, in which case the intra-regional

accessibility by definition is zero. Moreover, several municipalities have no R&D

(neither private nor university) and no registered patent applications filed to the

EPO.

Table 3.2. Descriptive statistics of the variables in the analysis.

Variable Min. Max. Std.dev Mean

FRDsMs A×ω

Intra-municipal accessibility to private R&D

0 1 864.94 117.48 14.53

×sω URDsMA

Intra-municipal accessibility to university R&D

0 694.70 63.54 8.86

×sω PsMA

Intra-municipal accessibility to

patent stock 0 401.15 30.25 6.77

×sω FRDsRA

Intra-regional accessibility to private R&D

0 822.57 91.16 24.33

×sω URDsRA

Intra-regional accessibility to university R&D

0 350.84 39.28 11.04

×sω PsRA

Intra-regional accessibility to

patent stock 0 390.49 39.86 11.51

×sω FRDsEA

Extra-regional accessibility to private R&D

0.001 308.95 20.42 8.18

×sω URDsEA

Extra-regional accessibility to university R&D

0.0004 108.74 15.59 5.90

×sω PsEA

Extra-regional accessibility to

patent stock 0.0007 174.72 12.01 5.05

Es Entrepreneurship 0.01 0.09 0.03 0.01 *) A signals accessibility and the superscript signals the variable to which the accessibility is calculated. In the sequel, we will analyze to what extent the municipalities’ accessibility to the variables above in 1993 can explain the variation in the growth performance of Swedish municipalities in subsequent periods, i.e. 1993-2001.

- 15 -

3.2 Empirical Analysis Table 3.3 presents OLS coefficient estimates7 of the variables in Equation (2.15) based on each of

the three knowledge variables, i.e. (i) private R&D, (ii) university R&D and (iii) the stock of

patents. The dependent variable is the change in value-added per employee during the period

1993-2001.

The results presented in Table 3.3 shows that the parameter estimates for the weighted intra-regional accessibility variables are statistically significant and positive for all the measures of knowledge. Except when knowledge is proxied by the stock of patent applications, the parameter estimates for the weighted intra-municipal accessibility variables are also statistically significant and positive. This suggests that knowledge flows transcend municipal borders. However, such knowledge flows seem to be bound within functional regions, since the parameter estimates for the weighted extra-regional accessibility are all insignificant. Thus, based on the results in Table 3.3, the hypothesis that there is a positive relationship between intra-municipal and intra-regional knowledge accessibility and municipal growth cannot be rejected. Table 3.3. OLS estimates of Equation (2.15) on the accessibility to private R&D (FRD), university R&D

(URD) and the stock of patent applications (P) in 1993.

Variables FRD URD P

Constant 198729.9* (21.22)

199497.7* (19.95)

199912.9* (20.72)

WsMs A×ω 98.43*

(3.03) 292.1* (3.02) n.s

WsRs A×ω 326.33*

(3.35) 683.21* (4.28)

709.13* (2.64)

WsEs A×ω n.s n.s n.s

)(+migsD n.s n.s n.s

sE -2384590.3* (-4.59)

-2429308.1* (-4.39)

-2451387.1* (4.59)

R2 0.39 0.39 0.40

Deg.fr 200 200 200

N 286 286 286

*) * denotes significance at the 0.05 level. **) ** denotes significance at the 0.1 level. ***) t-values are presented within brackets. ****) Standard errors are calculated using White’s (1980) heteroscedasticity-consistent covariance matrix.

7 However, the parameter estimates for the regional dummies are not included in the table to save space.

- 16 -

Moreover, the parameter estimate for the entrepreneurship variable comes out as significant and negative in all the estimations. This means that value-added per employee has shown a limited growth in municipalities with many 1-person firms per employee. The dummy for positive net migration is statistically insignificant in all estimations.

Optimally, one would like to include the various measures of knowledge accessibility in a single regression. However, it turns out that the different measures of knowledge accessibility are highly correlated. In order to account for all the different measures at the same time, we applied principal component analysis. Three principal components were constructed based on the three weighted intra-municipal variables, three based on the weighted intra-regional variables and three based on the weighted extra regional variables. Then, the first principal component of each was extracted, which means that we consider three principal components. The advantage of this method is that the new set of variables bares a natural relation to the original variables. Moreover, the correlation between the original variables and the first principal component is very high. All correlation coefficients are significant at the 0.01 level. This means that the contribution of the three measures of knowledge accessibility to the principal component is similar, which ease the interpretation. Table 3.4 shows the percentage of the variance of the original data explained by the first principal component.

Table 3.4. Percentage of variance explained by the 1st principal component.

Type of variables Variance explained (%)

Intra-municipal accessibility variables 87.89

Intra-regional accessibility variables 96.38

Extra-regional accessibility variables 86.30

Evidently, more than 86 % of the variance in the original data is explained by the first principal component of each type of accessibility. This means that the first principal components can be used without loosing too much information.

Table (3.5) presents the coefficient estimates of an OLS regression of Equation (2.15) by including the three principal components. Here,

• sMZ = 1st principal component of the three intra-municipal variables,

• sRZ = 1st principal component of the three intra-regional variables.

• sEZ = 1st principal component of the three extra-regional variables.

These principal components can be seen as measuring the overall accessibility (i) to knowledge inside the municipality, (ii) to knowledge in the region the municipality belongs to and (iii) to all knowledge outside the region the municipality belongs to, respectively.

- 17 -

As can be seen in Table 3.5, the principal components yield a similar result as those reported in Table 3.3. The coefficient estimates for K

sMZ and KsRZ are statistically significant and positive.

This confirms that knowledge flows may transcend municipal borders. However, the coefficient estimate for K

sEZ turns out to be statistically insignificant. Thus, knowledge flows seem to be bounded by regional borders.

Table 3.5. OLS estimates of Equation (2.15) by including the the principal components.

Variables Parameter estimates

Constant 210037.9* (18.09)

sMZ 14746.5** (1.91)

sRZ 28555.5* (3.39)

sEZ n.s

)(+migsD n.s

sE -2412534.3* (-4.49)

R2 0.40

Deg.fr 200

N 286 *) * denotes significance at the 0.05 level. **) ** denotes significance at the 0.1 level. ***) t-values are presented within brackets. ****) Standard errors are calculated using White’s (1980) heteroscedasticity-consistent covariance matrix. 4. Conclusions and Suggestions for Future Research This paper has presented some preliminary and tentative tests of the hypothesis that there is a relationship between knowledge accessibility and regional growth. Using data at the municipality level in Sweden, the hypothesis cannot be rejected. The knowledge accessibility in a given period has a statistically significant effect on the growth in subsequent periods.

By decomposing the total accessibility of a region into three types, (i) municipal accessibility, (ii) regional accessibility and (iii) extra-regional accessibility, the results of the analysis also indicate that knowledge flows transcend municipal borders, but that they tend to be bounded within functional regions.

Concerning the possible policy conclusions of our results, they indicate that knowledge accessibility contributes to growth. As knowledge flows seems to be bounded by regional boundaries regions that want to stimulate growth must try to improve intra-municipal and intra-regional accessibility. If this should be done via more investments in R&D or more investments

- 18 -

in transport infrastructure is a question that needs further research to be answered. One might expect that the answer differs between different regions.

To validate the research in this paper it would be interesting to re-estimate the models using a panel data approach. Another interesting avenue for future research is to analyse whether R&D or transport infrastructure investments have the strongest growth effects at the regional level and also to see if there are patterns in which type of regions that gain on the types of investments.

- 19 -

References Acs, Z.J., L. Anselin & A. Varga (2002), “Patents and Innovation Counts as Measures of

Regional Production of Knowledge”, Research Policy 31, 1069-1085

Almeida, P. & B. Kogut (1999), “The Localization of Knowledge and the Mobility of Engineers”, Management Science 45, 905-917

Andersson, M. and C. Karlsson (2004 a), “Regional Innovation Systems in Small and Medium-Sized Regions. A Critical Review and Assessment” in Johansson, Karlsson and Stough (2004) (Eds.)

Andersson, M. and C. Karlsson (2004 b), “The Role of Accessibility for Regional Innovation Systems”, in Karlsson, Flensburg and Hörte (2004) (Eds.)

Armstrong, H.W. & R.W. Vickerman (1995) (eds.), Convergence and Divergence among European Regions, London; Pion

Arrow, K. (1962), “Economic Welfare and the Allocation of Resources for Invention”, in Nelson (1962) (ed), 609-626

Bal, F. & P. Nijkamp (1998), “Exogenous and Endogenous Spatial Growth Models”, The Annals of Regional Science 32, 63-89

Barro, R.J. & X. Sala-i-Martin (1995), Economic Growth, Cambridge, MA; The MIT Press Braunerhjelm, P. & B.Borgman (2004), “Geographical Concentration, Entrepreneurship and

Regional Growth – evidence from regional data in Sweden 1975-1999”, Regional Studies, 38, 929-947

Caniëls, M.C.J. (2000), Knowledge Spillovers and Economic Growth. Regional Growth Differentials across Europe, Cheltenham; Edward Elgar

Carlino, G.A.(2001), “Knowledge Spillovers: Cities’ Role in the New Economy”, Federal Reserve Bank of Philadelphia Business Review, Q4,17-23

Cheshire, P. & G. Carbonaro (1995), “Convergence-Divergence in Regional Growth Rates: An Empty Black Box?”, in Armstrong & Vickerman (1995) (eds.), 89-111

Cohen, W.M. & D.A. Levinthal (1990), “Absorptive Capacity – a new perspective on learning and innovation”, Administrative Science Quarterly, 35, 128-152

Dixit, A. and J. Stiglitz (1977), “Monopolistic Competition and Optimum Product Diversity”, American Economic Review 67, 297-308

Dosi, G. (1988), “Sources, Procedures and Microeconomic Effects of Innovation”, Journal of Economic Literature 26, 1120-1171

Feldman, M. (1994), The Geography of Innovation, Boston; Kluwer Academic Publishers Feser, E.J. (2003), “What Regions Do Rather than Make: A Proposed Set of Knowledge-based

Occupation Clusters”, Urban Studies 40, 1973-1958 Fischer, M.M. & J. Fröhlich (2001) (Eds.), Knowledge, Complexity and Innovation Systems,

Berlin; Springer-Verlag Fischer, M.M., L. Suarez-Villa & M. Steiner (1999) (eds.), Innovation, Networks and Localities,

Berlin; Springer-Verlag Florida, R. (2000), Competing in the Age of Talent: Quality of Place and the New Economy,

Pittsburgh, PA; R.K. Mellon Foundation, Heinz Endowments, Sustainable Pittsburgh Florida, R. (2002), The Rise of the Creative Class, New York; Basic Books Fujita, M. & J.-F., Thisse (2003), “Does Geographical Agglomeration Foster Economic Growth?

And who Gains from It?”, The Japanese Economic Review 54, 121-145 Glaeser, E., et al. (1992), “Growth in Cities”, Journal of Political Economy 100, 1127-1152

- 20 -

Jaffe, A., M. Trajtenberg & R. Henderson (1993), “Geographic Localization of Knowledge Spillovers as Evidenced by Patent Citations”, Quarterly Journal of Economics 108, 577-598

Johansson, B., C. Karlsson & R.R. Stough (2001) (eds.), Theories of Endogenous Regional Growth. Lessons for Regional Policies, Berlin; Springer-Verlag

Johansson, B., C. Karlsson & R.R. Stough (2002) (eds.), Regional Policies and Comparative Advantages, Cheltenham, UK; Edward Elgar

Johansson, B., C. Karlsson and R. Stough (2004) (Eds.), The Emerging Digital Economy: Entrepreneurship, Clusters and Policy, Berlin; Springer-Verlag (Forthcoming)

Johansson, B., J. Klaesson & M. Olsson (2002), “Time Distances and Labor Market Integration”, Papers in Regional Science 81, 305-327

Johansson, B., J. Klaesson & M. Olsson (2003),”Commuters’ Non-Linear Response to Time Distances”, Journal of Geographical Systems, 5, 315-329

Karlsson, C. (1997), “Product Development, Innovation Networks and Agglomeration Economies”, The Annals of Regional Science 31,

Karlsson, C., P. Flensburg and S.-Å. Hörte (2004) (Eds.), Knowledge Spillovers and Knowledge Management, Cheltenham; Edward Elgar (Forthcoming)

Karlsson, C. and A. Manduchi (2001), “Knowledge Spillovers in a Spatial Context – A Critical Review and Assessment”, in Fischer and Fröhlich (2001) (Eds.), 101-123

Klaesson, J. (2001), “Spatial Interaction Models and Concepts: a Review of the Gravity Theory”, Jönköping International Business School, Jönköping (mimeo)

Lucas, R.E. (1988), “On the Mechanics of Economic Development”, Journal of Monetary Economics 22, 3-42

Mathur, V.K. (1999), “Human Capital-Based Strategy for Regional Economic Development”, Economic Development Quarterly 13, 203-216

Maurseth, P. & B. Verspagen (1998), “Knowledge Spillovers in Europe and its Consequences for Systems of Innovation”, Paper prepared for the TSER project ‘Technology, Economic Integration and Social Cohesion’ workshop in Gothenburg

Nelson, R.R. (1997), “How New is New Growth Theory?”, Challenge 40, 29-58 Nijkamp, P. (2003), “Entrepreneurship in a Modern Network Economy”, Regional Studies 37,

395-405 Nijkamp, P. & J. Poot (1998), “Spatial Perspectives on New Theories of Economic Growth”, The

Annals of Regional Science 32, 7-37 Obstfeld, M & Rogoff, K. (2002), Foundations of International Macroeconomics (5th printing),

The MIT Press, Massachusetts Polanyi, M. (1996), The Tacit Dimension, London; Routledge & Kegan Paul Quigley, J.M. (1998), “Urban Diversity and Economic Growth”, Journal of Economic

Perspectives 12, 127-138 Rodríguez-Pose, A. (2001), “Is R&D Investment in Lagging Areas of Europe Worthwhile?

Theory and Empirical Evidence”, Papers in Regional Science 80, 275-295 Romer, P.M. (1986), “Increasing Returns and Long Run Growth”, Journal of Political Economy

94, 1002-1037 Romer, P.M. (1990), “Endogenous Technological Change”, Journal of Political Economy 98, 71-

102 Solow, R. (1956), “A Contribution to the Theory of Economic Growth”, Quarterly Journal of

Economics 70, 65-94

- 21 -

Solow, R. (1957), “Technical Change in an Aggregate Model of Economic Growth”, International Economic Review 6, 18-31

Swan, T. (1956), “Economic Growth and Capital Accumulation”, Economic Record 32, 334-361

Thompson, W.R. & P.R. Thompson (1985), “From Industries to occupations: Rethinking Local Economic Development”, Economic Development Commentary 9, 12-18

Thompson, W.R. & P.R. Thompson (1987), “National Industries and Local Occupational Strengths: The Cross-hairs of Targeting”, Urban Studies 24, 547-560

Varga, A. (1999), “Time-Space Patterns of US Innovation: Stability or Change? A Detailed Analysis Based on Patent Data”, in Fischer, Suarez-Villa & Steiner (1999) (eds), 215-234

Verspagen, B. & W. Schoenmakers (2000), The Spatial Dimension of Knowledge Spillovers in Europe: Evidence from Firm Patenting Data, Maastricht; Merit

Weibull, J. (1976), “An Axiomatic Approach to the Measurement of Accessibility”, Regional Science and Urban Economics 6, 357-379

Weibull, J. (1980), “On the Numerical Measurement of Accessibility”, Environment and Plan-ning A 12, 53-67

White, H. (1980), “A Heteroscedasticity-Consistent Covariance Matrix Estimator and Direct Test for Heteroscedasticity”, Econometrica, 48, 817-838