Master Thesis
Cultural Capital and Regional Labour Productivity Growth in
the Netherlands Between 1995 and 2005
Jaap Beemster∗
University of Groningen Faculty of Economics and Business
Supervisor: prof. dr. J. Oosterhaven∗∗
August 2008
∗ Email: j.n.beemster@student.rug.nl
∗∗ Professor of Spatial Economics, Faculty of Economics and Business, University of Groningen, Postbus
Abstract
Table of Contents Abstract……….2 List of Tables………4 List of Figures………...………4 1. Introduction………5 2. Literature Review………...7
2.1 History of Cultural Economic Thought………...7
2.2 The Value of Culture………...8
2.3 Assessing Culture….………...9
2.4 Culture and Productivity Growth………..10
2.5 Creative Class………14 3. Data………..17 3.1 Variables………...……….17 3.2 Descriptive Statistics……….18 4. Model………20 4.1 Model Specification…..………...……….20 4.2 Methodology……….22
5. Results & Discussion………....23
5.1 Model Output……...……….23
5.2 Discussion……….25
6. Conclusion………....27
References………...28
Appendix I: Corop Specification………31
Appendix II: Variables: Model Name, Sources and Explanation………...…………36
Appendix III: Descriptive Statistics………..………..37
Appendix IV: ADF Test Results……….42
Appendix V: Correlation Matrix…..………...………..………..43
List of Tables
Table 1: Regression results of (11) for 40 corop regions from 1995 to 2005……….23 Table 2: Period fixed effects: test output………24
List of Figures
1. Introduction
The role of culture in economics becomes more and more important. Economic growth is long explained by Solow’s argument of capital accumulation, i.e. physical and human capital. More recently, ‘social capital’ and institutions act as additional explanatory factors (Hall and Jones, 1999). However, economists agree that there is still a large part to be further explained. That is, the so-called residual in economic growth models is still a challenge for economic researchers over the world. A recent wave of literature argues that cultural capital may be an important explanatory factor (Chartrand, (1990), Klamer (2002, 2003)). The common thought is that cultural heritage, a nice ‘green’ environment to live in, great literature, museums, theatre etc. in a region enhance the knowledge and innovative power of the working population, which lead to productivity improvements. As a result, economic growth takes place. Florida (2002) goes even further and argues that a large ‘creative class’ is the main source of regional growth.
However, the above arguments are mainly based on theory. The objective of this study is to test the influence of cultural capital empirically. To be more specific, we want to examine the relationship between the growth of cultural capital and the regional labour productivity growth in the Netherlands. Another question we want to answer is whether the impact of cultural capital is spatially correlated. That is, the growth rate of a particular region can be affected by the cultural expenditures of adjacent regions. A basic regional growth model is used to answer the questions. Cultural capital is measured by local government cultural expenditures and the potential of culture measures the spatial effect. The control variables included are physical capital, human capital, R&D expenditures, and job density. The analysis is estimated using data of 40 corop-regions for the period 1995-2005.
The research performed fills a gap in cultural economics. An exception is Roo (2005). She studied the role between cultural capital and regional productivity growth in the Netherlands from 1995 to 2001. The effect she found is positive. However, she uses a different measure for culture, i.e. national cultural subsidies.
In addition, this thesis takes the usual empirical pitfalls into account, i.e. problems of multicollinearity, heteroskedasticity, and stationarity are unraveled. Unfortunately, the results do not indicate that cultural capital growth is an important additional variable. All explanatory variables, except R&D, are insignificant and have an unexpected sign. A positive and significant effect of culture on productivity growth is only found when all other variables are excluded.
2. Literature Review
Cultural economics is a relatively young field of research. As a result, cultural economists do not agree on fundamental questions like ‘what is culture’ and ‘what is the exact effect of culture on growth’. In addition, cultural economists find it hard to measure culture and to derive an empirical model that explains the nexus between culture and productivity growth.
However, there are some authors who have done a great effort in bundling the different views of cultural economics through time (Chartrand (1990), Dieckmann (1996), and Fukuyama (2001)). I present their analyses first and give comments where needed. Klamer’s work is discussed secondly. He focuses explicitly on the theoretical value of culture and the role of culture in economics (2002, 2003). Explanations of his ideas help to understand the topic. The third paragraph explains how we may assess culture, using the framework of Hofstede (1978). The fourth paragraph explains how culture can influence growth and discusses briefly some empirical findings. The fifth and final paragraph introduces Florida’s (2002) theory. He argues that economic growth in (US) regions can be explained by a large ‘creative class’. However his work doesn’t receive a lot of support by economists. This is due to weak empirical support and because his ideas are very subjective and distinctive from the current stand of growth literature. Nevertheless, there are interesting findings of studies that test Florida’s ideas on Dutch regions (Marlet and Woerkens (2004a, 2004b) and Boschma (2005)).
2.1 History of Cultural Economic Thought
Many cultural economists agree that culture is a neglected field in the history of economic thought (Boulding (1972), Meisel (1974), and Chartrand (1990)). According to Chartrand (1990) this is mainly due to Bentam (1748-1832), since he introduced the standard that culture, custom, and tradition are irrational factors that should be neglected in economic analysis. He believed in ‘radical egalitarianism’; a world where the happiness of all individuals is the same and is based on the holdings of money, i.e. the more money an individual has, the happier he is. From the mid- to late-19th century, the irrelevance of culture in economic analysis was reinforced by mainstream economics, who adapted the same ideas regarding the role of culture.
inductive method and general systems analysis. On the other hand, mainstream economics is characterized by positivism, deductive method and mechanistic systems analysis”. The opposing views with respect to role of culture in economic thinking remained a problem in the 20th century.
Today, neoclassical economists still neglect the importance of culture, i.e. they make the simplifying assumption that human beings are rational utility-maximizing individuals. Growth is based on capital and labour accumulation. Their arguments against the use of culture in economic analysis are intuitive. That is, they argue that cultural factors are, methodologically, very difficult to measure and that empirical evidence is weak. However, a counterargument against their theory is that they fail to explain the so-called ‘Solow residual’, i.e. it has been shown that it is impossible to explain economic growth by economic factors only (Dieckmann, 1996).
In the mid-eighties of the 20th century, endogenous growth theory has been introduced (Barro and Sala-i-Martin, 1995). As Dieckmann (1996) argues, endogenous growth models are based on imperfect competition and they explicitly include human capital accumulation, R&D expenditures and other determinants that are related to culture. Hence, in this new way of economic thinking, culture is more and more treated as an additional explanatory factor.
Moreover, new institutional economists in the late 20th century showed their renewed interest in culture. According to Fukuyama (2001) institutional economists nowadays seek to give rational, maximizing accounts of the origins of institutions, but as a group they are much more aware of the importance of history, culture, tradition, and other so-called ‘path dependent’ factors in shaping economic behaviour.
In short, being invisible for centuries, cultural economics finally receives some attention in modern economic thinking. However, a lot remains to be explained; what is culture exactly, how can we assess culture, what is the evidence regarding the relation between culture and productivity growth, and how can we explain the relation? The next sub-sections provide answers to these questions.
2.2 The Value of Culture
in a social and cultural sense. A monument or museum often generates national pride and cultural identity. Social values play a role in arguments that favour the subsidization of cultural goods, since they improve the integration of minorities, have educational values, and are good for personal development (Klamer, 2003). A third interpretation is that culture has anthropological meaning. That is, culture leads to the shared values, stories and aspirations that distinguish one group of people from another (a community, an organization, a region). The economic value of culture then implies the economic contribution of these shared values. This interpretation is analogous to Weber’s work (1930). He argued that a particular culture may improve economic performance or hinder it. The final interpretation combines all previous interpretations. That is, culture may stand for both the arts and culture in the anthropological sense, and value for economic value as well as social and cultural values.
Klamer (2003) explains the link between the four interpretations as follows: “Whereas the common justification of cultural policy evokes the economic value of culture (think of the income generated, the increase in tourism and the attraction for new businesses in town) or social values (educations, inclusion of minorities, low thresholds for people with a low income), culture can also be said to have value in and of itself. It could even be argued that all economic activity serves the enhancement of cultural ‘capital’ of a community, such as a town, or a country” (Klamer (2003), Ch. 59).
We now have an understanding of the value of culture in economics. Still, the difference between cultural economics and the economics of culture remains to be explained. Chartrand (1990) provides a definition of both. According to him, cultural economics can be defined as the study of the evolutionary influence of cultural differences on economic thought and behaviour. That is, economic behaviour varies because of a changing cultural context. The economics of culture, on the other hand, can be defined as the study of the allocation of scarce resources within the cultural sector. Objective laws apply to economic behaviour regardless of cultural differences, i.e. a positivist approach. My study is a cultural economics study since I focus on the influence of cultural differences on economic performance, i.e. labour productivity growth.
2.3 Assessing Culture
first is power distance. It relates to the degree of inequality in power between people in a particular society. Power is the potential to determine the behaviour of other persons. The higher the power distance, the more inequality in power between people, i.e. the more hierarchy in decision making. The second is individualism. This dimension focuses on the degree to which a society reinforces individual or collective achievement and interpersonal relationships. Individuality and individual rights are important for cultures based on individualism and the ties between individuals, i.e. family relationships are important for cultures based on collectivism. Masculinity is the third dimension. It refers to the inequality between genders in a society. It relates to the degree of a society to reinforce or not the traditional masculine culture of assertiveness, ambition, and the accumulation of wealth and material possessions. Relationships and the quality of social life correspond to a low degree of masculinity. The fourth is uncertainty avoidance. This dimension concerns the level of acceptance for uncertainty within a society. Cultures that score high in uncertainty avoidance prefer a rule-orientated society that follows well defined and established laws, regulations and controls.
2.4 Culture and Productivity Growth
The next issue we focus on is the relation between culture and productivity growth. Again, cultural economists do not agree on the exact relation between the two. Due to measurement problems and data limitations, empirical evidence scarce. As a result, cultural economists mainly focus on a theoretical foundation.
One of these authors is Chartrand (1990). He argues that culture influences productivity growth via technological change. Technological change is widely accepted as being an important factor influencing growth. In general, technological change is achieved by creating new ideas, innovations, inventive organization structures, and developments in advertising techniques. Chartrand (ibid) argues that cultural capital leads to technological change, since cultural experience of the labour force leads to the creation of new ideas and innovations. This theory seems rather unconvincing at first sight, but it comes close to reality. Look for example at the multinationals nowadays; they are looking for new, creative employees constantly. That is, workers who have a personality, who are diverse, and who have refreshing and new ideas. Hence, Chartrand’s work deserves attention.
consumption goods on the other. Nonetheless, he stresses the importance of data requirements; he shows that there are a lot of opportunities when more cultural data becomes available.
Fukuyama (2001) also provides a theoretical explanation for the nexus between culture and productivity growth. He notes that cultural factors affect economic behaviour generally in four ways. The first is the impact of culture on organization and production. Firms around the globe have different hierarchical structures, norms, production processes etcetera (Hofstede, 1978). Some ‘organization cultures’ do better than others and make more profit. The second is the role of culture in consumption behaviour and work ethic. Different classes of societies, like rich and poor, have different cultural habits and different consumption patterns. Income and cultural expenditures are thought to correlate positively. The third is the ability of culture to create and manage institutions. Many economists agree that institutions are important determinants of the economic performance of countries. Culture in its broad sense affects the ability of societies to create and manage institutions. The fourth is the impact of culture on social networks. The impact cultural values have on networks of social relations is the basis of social capital. Social capital consists of norms and values shared among a group of people that promote cooperation and confidence among them. Hence, it refers to the flow of information in an economy. Information indeed becomes more and more important in today’s world, notice the current credit crisis. A lot of miscommunication and imperfect information lies at the basis of the crisis. Nevertheless, the concept of social capital has been criticized for being a rather vague concept and the lack of a clear measurement method.
The analyses of Chartrand and Fukuyama have one thing in common: culture affects productivity growth indirectly. It implicates that the effect of cultural capital on growth is difficult to disentangle from other factors of influence, like institutions and social networks.
where y is output, A is the level of technology, kis private capital, his public expenditures for cultural goods that are tax financed, and
β
is the production elasticity.β
=1 implies no influence of culture on output, i.e. the standard Rebelo model. Constant returns to scale and positive but diminishing returns to single inputs are assumed. Following Dieckmann (ibid), I use a Ramsey-type iso-elastic utility function in which each individual maximize overall utility, U:dt e c u U =
∫
∞ − t 0 ( ). ρ andσ
σ − − = − 1 1 ) ( 1 c c u (σ
>0,σ
≠1) (2)where u(c) denotes period utility, e−ρt is the discount factor,
ρ
is the discount rate, t is time, andσ
refers to the intertemporal substitution elasticity. Each individual is assumed to work for a given working time, i.e. there is no further decision between leisure and work. The budget constraint for the representative agent then becomes:h c y
k= − −
∆ (3)
Hence, gross savings (∆k; capital accumulation) refer to income after consumption expenditures and taxes paid for public cultural expenditures. Dieckmann assumes that public expenditures for cultural goods are taxed financed via constant legal charges on capital. That is, h=
α
k, whereα
is a constant and refers to the (cultural) tax rate. Hence, public expenditures for cultural goods change over time as private capital changes. In that sense it is an endogenous variable. This assumption is questionable since it leads to the problem of causality. However, the assumption is needed to complete the model.where k is the state variable, c is the control variable, and
λ
is the co-state variable; it represents the marginal utility of wealth. Maximization gives the following first order conditions1:λ
σ ρ = ⇔ = ∂ ∂ − − c e c H t 0 (5a) • − • − = − ⇔ − = ∂ ∂λ
α
α
λ
λ
β ) (A 1 k H (5b)Differentiation of (5a) with respect to time and substitution for
λ
from this equation and for •λ
from (5b) gives the consumption Euler equation:c A c c
γ
ρ
α
α
σ
β − − ≡ = − • ) ( 1 1 (6)The growth rate of consumption (
γ
c) is proportional to the difference between the marginal product of capitalα
1−βA and the sum of the cultural tax rate
α
and the discounting rateρ
. In this way the growth rate of countries or regions can easily be compared since every country has a unique cultural tax rate in order to spend on cultural goods. A constant tax rateα
and capital accumulation k lead to an increase in cultural expenditures h. As a result, the higher h solves the limits of decreasing returns to private capital in (1). The corresponding cultural tax rate that maximizes growth is:0 1 ) 1 ( 0 max = − − ⇔ = ∂ ∂ −β
βα
β
α
γ
A c ⇒α
max = A[ (1−β
)β
] (8)A country’s growth maximizing cultural expenditures thus depends on the level of technology and on the elasticity of production. In short, Dieckmann’s adjusted theoretical analysis shows how an endogenous growth model can be used to explain the impact of culture on economic performance. Nonetheless, a drawback of his theory is that he doesn’t take the negative effect of
1
Note that the transversality condition holds; lim =0
∞ → k
t λ
the cultural tax into account. That is, when k rises more tax revenue is received by the government which can be used for cultural expenditures. However, employees dislike taxes since it reduces their income and savings. As a result, capital accumulations diminish.
Testing Dieckmann’s model empirically is very complicated. Indeed, Dieckmann follows a different approach in testing the relation between culture and economic growth. He runs cross-country growth regressions in the period 1960-1990 using data from Penn World Table 5.6a, World Bank database, and cross-cultural studies (Hofstede, 1978).
Hofstede’s uncertainty avoidance index, explained in section 2.3, is used as a proxy for culture. Dieckmann’s results indicate that economic growth requires a cultural component but that the effect of culture on growth is weak, i.e. not robust to sensitivity analyses. The reason for this weak finding is probably the choice of Hofstede’s uncertainty avoidance variable. Hofstede admits that the theoretical net effect of culture on economic performance is not clear, using the uncertainty avoidance index as a measure for culture. That is, different values of the uncertainty avoidance index may lead to contradicting results. In addition, information on the uncertainty avoidance index is scarce for most societies. As a result, the dimension is difficult to construct and empirical testing is hard.
The above theories and empirical research are mainly based on cross-country (national level) studies. Regional growth studies that include cultural capital as a determinant are scarce. In most regional growth studies other factors are important, e.g. agglomeration effects (Broersma and Oosterhaven, 2006). Broersma and Oosterhaven note that the positive effect of the growth of job density refers to agglomeration. In general, agglomeration leads to productivity growth since the resulting knowledge spillovers lead to innovations and new ideas, i.e. productivity improvements (Fujita, Krugman, and Venables (1999).
A regional growth study that includes cultural capital is Roo (2005). She found a positive link between cultural capital and productivity growth in the Netherlands, between 1995 and 2001.
2.5 Creative Class
Figure 1. Creative class
1. Super Creative Core
- Computer and mathematical occupations - Architecture and engineering occupations - Life, physical, and social occupations - Education, training, and library occupations
- Arts, design, entertainment, sports, and media occupations (occupations in art, design, and for a part in entertainment are the so-called Bohemians, which are defined in figure 2 below)
2. Creative professionals - Management occupations
- Business and financial operations occupations - Legal occupations
- Healthcare practitioners and technical occupations - High-end sales and sales management
Source: Florida (2002), p. 328
Figure 2. Bohemians
- Musicians, singers, actors, dancers, choreographers, and directors - Decorators and commercial designers
- Sculptors, painters, and other figurative artists - Conductors and composers
- Authors and other writers - Photographers
Source: Florida (2002), p. 328
In addition, there are six indices that show a relationship with the creative class, according to Florida (2002). They are: the bohemian index (see figure 2), which measures the level of bohemian occupations in a region; the openness index, which measures the relative percentage of citizens in a region that are from abroad; the employment growth index; the cultural opportunity index, which measures the cultural supply in a region, i.e. the number of museums, theatres, art galleries, restaurants etcetera; the social cohesion index, which measures the social climate in a region, and finally; the public provision index, which measures the public nursing climate in a region. All indexes are supposed to correlate positively with the creative class. In summary: a large creative class, which finds and solves problems and lives in regions that are have high scores on the above indexes, lead to economic growth.
The main problem with Florida’s ideas is that they are based on very weak empirical evidence2 (Glaeser, 2004). For that reason many scholars do not pay much attention to his work. In fact, his theory is neglected in most cultural economic papers. However, Marlet and Woerkens (2004a, 2004b) and Boschma et al. (2005) show that the creative class is an important explanatory factor of Dutch regional growth. Marlet and Woerkens (2004a) reconstruct the creative index for the Netherlands and test whether a relative large creative class explains employment growth in the 50 largest cities in the period 1996-2002. They show that there is a significant link. Moreover, they show that cultural amenities in cities lead to a larger creative class, i.e. they find a positive link between the cultural potential index and the creative class. On the other hand, Boschma et al. (2005) do not find strong empirical evidence in favour of Florida’s theory. Nonetheless, they emphasize the role of the creative class in regional economic theory and the need for a worldwide identical definition of the creative class.
To conclude, the above theoretical analysis shows that cultural capital has potential to explain productivity growth, theoretically. However, the measurement problem of culture is severe. Therefore, more and better cultural data should become available in the following years and cultural economist must find universal ways to measure it. For the time being, it is sufficient that we accept the role of cultural capital in the analysis of productivity growth.
3. Data
This section presents and explains the data used. All data is taken from Statline (the electronic data-bank of the Dutch national statistical institute CBS). The time span runs from 1995 to 2005 and the level of investigation is corop-regions3. In total the Netherlands can be subdivided in 40 corop-regions. The 40 corop-regions with corresponding municipalities are presented in Appendix I. Details on the data are presented in Appendix II. Since I study relative productivity growth, all variables, except human capital and job density, are divided by the volume of labour (the total number of labour years used in the production process, see next section). In addition, all of the currency denominated variables are in 2000 euro’s; they are converted using 2000 purchasing-power parities (PPP).
3.1 Variables
The dependent variable–labour productivity growth–often is defined as the growth of GDP per capita. In this study however, it is measured by the relative growth of value added. The reason is similar to the argument of Broersma and Oosterhaven (2006). That is, in contrast to regional GDP data, regional value added data is available for all sectors. This classification is important since excluding the business sector ‘mineral winning’ in all regions provides more realistic results4. Hence, the value added data used refers to total relative value added excluding the mineral winning sector. The important explanatory variables are cultural capital and its potential. The control variables are physical capital, human capital, research and development (R&D) and job density.
As explained in the former section, measuring cultural capital is difficult since there is no single universally accepted definition. As a result, adequate statistics are scarce. Fortunately, Statline provides cultural expenditure data of local governments so the relation can be tested for these expenditures. The cultural expenditures are subdivided in 6 categories, which are: public
3
A corop-region is a regional area within the Netherlands that was classified in 1970 by the Regional Coordination Commission Investigation. The commission uses a nodal classification principle. That is, every corop-region has a central core (e.g. a city) with surrounding catchment area. A corop-region always exists of a few municipalities that belong to the same province and is always an interconnected area. Corop-regions are often characterized as labour market regions.
4 The value added of the mineral winning sector is over presented in the corop-region ‘Overig Groningen’.
library; informal education & development projects; art; archeology & museums; social quality of life & open-air recreation, and; remaining expenditures on culture and recreation. Obviously, the cultural proxy I use doesn’t comprise all expenditures on culture. For example, private consumption of cultural goods, national cultural subsidies, and private donations are not included. Furthermore, data on private consumption is scarce and information regarding the receiving cultural organizations of private donations is generally not available.
The potential of culture measures the spatial effect of cultural expenditures in a region on the productivity growth of its surrounding regions, and vice versa. The longer the distance between regions, the weaker the influence of cultural expenditures on each other’s productivity growth. Hence, a region that is surrounded by regions with high cultural capital at the workplace have a high potential.
Measuring physical capital is difficult since data is not readily available on the corop level. A good alternative is gross investments by firms. Investment in housing is excluded because it generally doesn’t contribute to productivity growth, i.e. investment in housing is not really a capital good that contributes to value added growth. Human
capital is measured by the percentage of workers at the workplace with a high level of education, i.e. with a bachelor or master degree. In addition, R&D expenditures by businesses and research institutions proxy for the level of research and development at the workplace. Finally, the number of jobs per km² measure job density. According to Fujita, Krugman, and Venables (1999), a dense working population in a region can have two effects on productivity growth; a positive (agglomeration) and a negative effect (congestion).
3.2 Descriptive Statistics
The regional specification (part b) shows that the highest values of almost all variables occur in the Rim-city5, e.g. corop-regions ‘Utrecht’ (17), ‘Groot-Amsterdam’ (23), ‘Agglomeratie ‘s-Gravenhage’ (The Hague, 26), ‘Delft en Westland’ (27), and ‘Groot-Rijnmond’ (Rotterdam, 29). Regions outside the Rim-area show lower values. ‘Overig Groningen’ (Groningen, 03) is an exception. This is logical since the city of Groningen is the economic and cultural metropole of the North of the Netherlands. Interestingly, the above regions with high values are regions with a very large creative class (Marlet and Woerkens, 2004a, 2004b).
The expenditures on culture show an interesting pattern. Again, corop-regions located in the Rim-area spend relatively a lot on culture. However, regions in the province of Groningen and Zeeland also show high levels of cultural expenditure. Local governments that spend relatively little on culture are found in the corop-regions in the province of Gelderland, Overijssel, Limburg and Flevoland.
5 The Rim-city (Randstad in Dutch) is a conurbation which consists of the four largest cities (Amsterdam,
4. Model
As explained before, empirical research on the nexus between cultural capital and productivity growth is scarce. However, Dieckmann’s (1996) work shows that is possible to embed culture in a neoclassical Ak model. He focuses on a micro economic analysis, using cultural expenditures by households. We instead focus on a regional analysis, using local government expenditures on culture.
4.1 Model Specification
Fortunately, theoretical work on the relation between culture and growth is more widely available. As described in the theoretical part, many scholars argue that culture should be integrated in growth analysis. The goal of this paper is to test this argument. I start with a Cobb-Douglas production function:
η η − = 1 L AK Y (0<
η
<1) (9)In addition to capital K and labour L, the level of technology A is the main driver of output. The relevant factors that influence the level of technology (and therefore fall under A) are discussed in section 2 and 3, i.e. human capital, R&D expenditures, job density, and cultural capital. To continue, the Cobb-Douglas production function (9) can be written as a log-linear growth function: ) log ) 1 (( ) log ( log logY = ∆ A+∆
η
K +∆ −η
L ∆ (10)In order to estimate the growth of labour productivity we divide both sides of equation (10) by L. The operational form is then as follows6:
6
rt rt rt rt rt rt rt rt rt rt rt rt rt rt L D R JD L CULT P L CULT HC L K L VA
ε
β
β
β
β
β
β
β
+ ∆ + ∆ + ∆ + ∆ + ∆ + ∆ + = ∆ ) & ( log log ) ( log log log log log 6 5 4 3 2 1 0 (11) ) exp( ) ( , ,s r rs s s r CULT t CULT P =∑
− ⋅ ≠β
(12) ∆ is the growth operator, VA is total value added (excluding mineral winning),β
0 is a constant, K is physical capital input, and L is the volume of labour used in the production process. The other explanatory variables represent technological progress. That is, HC represents human capital, CULT stands for cultural capital, P(CULT) is the potential of cultural capital, JD is job density, and R&D represents research and development. In addition, r refers to a corop region and t to a time period. Finally,ε
is the normal distributed error term.HC and JD are not divided by L since that is not reasonable: the HC variable is a proxy for the relative input of education in the production process and JD already includes the density effect of labour in a region.
Furthermore, all variables are presented in growth terms, since that ensures stationarity. That is, the augmented Dickey-Fuller tests (ADF) for every time series show that the growth term of every variable should be used: the null hypothesis of a unit root can be rejected for all series in first differences at a significance level of 1% (see Appendix IV). Intuitively, it is preferred to measure K, R&D, and JD in levels. That is, K and R&D are gross investment, i.e. growth of the capital stock and growth of innovations. Moreover, testing the effect of the growth of the density of jobs is rather ineffective. However, the time series of these variables are not stationary when measured in levels.
The error term can be specified such that the basic regression of (11) includes fixed or random effects. Fixed effects can be subdivided in period- (13) and cross-section fixed effects (14) (Baltagi, 1995):
rt t
rt
µ
ν
rt r
rt
µ
ν
ε
= + (14)t
µ
andµ
r denote, respectively, the unobservable period fixed effect and the cross-sectional fixed effect.ν
rt denotes the remaining white noise. When (13) holds it implies thatµ
t is cross-sectional-invariant, i.e. every time period has a specific fixed effect in the regression. Alternatively, when (14) holds it implies that cross-sections (corop-regions) have a time-invariant specific effect in the regression. In addition,ν
rt varies with cross-sections and periods and shouldthus be interpreted as normal distributed white noise.
A disadvantage of the fixed effects model is that it includes too many parameters. The loss of degrees of freedom is another problem (Baltagi, 1995). These problems can be solved under a random effects model, i.e.
µ
t (µ
r) are assumed to behave randomly. In that case botht
µ
(µ
r) andν
rt are normally distributed andµ
t(µ
r) are independent of theν
rt. In the nextsection I will test whether these specifications are appropriate. For the moment, I focus on the basic equation without fixed and random effects.
4.2 Methodology
Equation (11) is estimated for the complete panel using OLS in Eviews. Following the theoretical overview, a positive relation between the explanatory variables and the dependent variable is expected. In order to ensure that the results are the best linear unbiased estimators, some analytical tests are conducted. First, the residuals should follow a normal distribution. A histogram is constructed and the Jarque-Bera (JB) test is used to examine this assumption. The JB test output presented in the next section fails to reject the null hypothesis of a normal distribution. Second, the problem of heteroskedasticity is eliminated since the estimated output in the next section is based on White corrected standard errors and covariance. Finally, the problem of multicollinearity is not present since all explanatory variables have mutual correlations below 0.6 (see Appendix V for an overview).
5. Results 5.1 Model Output
The basic test result of equation (11) is presented in the first column of table 1. Since R&D data is only available from 1996 to 2003 the total number of observations is 320 (8 years times 40 cross-sections). Unfortunately, the results do not favour the inclusion of cultural capital growth. Its influence on labour productivity is insignificantly positive. This finding is in line with the weak effect of Dieckmann’s (1996) but contradicts the findings of Roo (2005). Furthermore, the influence of physical capital, job density, human capital, and the potential of culture are also insignificant. Moreover, they all have a surprising sign. Hence, they are not in line with theory and do not support former research. The only effect that is significant and in line with theory is research and development. Its effect on labour productivity growth is strong: a 10 percent increase in R&D expenditures leads to a 1.3 percent increase of labour productivity growth.
Table 1. Regression results of (11) for 40 corop regions from 1995 to 2005
Method: OLS: basic equation OLS: period fixeda OLS: period fixedb
Variables: constant 0.005 (0.009) -0.026 (-0.418) ∆log(K/L) -0.005 (-0.013) 0.003 (0.462) ∆log(HC) -0.009 (0.033) -0.013 (-0.378) ∆log(CULT/L) 0.014 (0.514) 0.043 (0.114) 0.065* (1.923) ∆log(PCULT/L) -0.058 (-1.639) 0.056 (0.307) ∆log(JD) 0.001 (0.450) -0.002 (-0.873) ∆log(R&D/L) 0.013* (-1741) 0.003* (-1872) Adjusted R² 0.126 0.125 0.175 JB test statistic 342.0 (0.000) 279.4 (0.000) 601.9 (0.000) Observations 320 320 320
Table 2. Period fixed effects: test output
test statistic d.f. p-value
F-test 7.162 (-7,306) 0.000
Chi-square test 48.554 7 0.000
The results in the first column are not the end of the story, however. That is, the model may contain fixed or random effects and may therefore be misspecified. I therefore test whether the basic model should include fixed or random effects, and whether the effects are period or cross-section related. The used test are the sums-of-squares (F-test) and the likelihood function (Chi-square test). The test results strongly favour period-fixed effects, i.e. the two statistical values and the associated p-values strongly reject the null hypothesis that the fixed-effects are redundant (see table 2). Test results for the inclusion of cross-section fixed effects and random effects are statistically insignificant. Hence, from now on we run equation (11) with period fixed effects (13).
The output of the model with period fixed effects is presented in the second column of table 1. The period-fixed dummies are presented in the note below the table. Again, the results are not satisfactory. The effect of all variables, except R&D, are insignificant. The dummy variables imply that for each year the intercept changes. However, the coefficients are unaffected, i.e. the influence of all explanatory variables remains the same.
Not only fixed and random effect are important to detect, some variables may be misspecified and should be excluded from the regression. To test whether the variables in the model are redundant or not I perform a redundant F-test. The F-test shows that all variables except culture are redundant in the regression. That is, for all variables except culture I fail to reject the null hypotheses that they are redundant. Hence, only culture should be included. I therefore run the OLS regression with period-fixed effects again but with culture only. The results are presented in the third column of table I. The coefficient now has a positive and significant effect.
5.2 Discussion
In the empirical process we performed several trial and error versions of the model. That is, test for the effect of cultural capital and the potential of cultural capital when divided by the working labour force instead of the volume of labour. The working labour force refers to the place where workers live. At their living environment, workers enjoy culture. According to theory, the experience of culture improves labour productivity. Therefore, this specification is sensible. Workers can thus experience culture at home and at the place where they work, both places are expected to have a positive effect on labour productivity growth. In the above model however, I have used the volume of labour as the denominator, since that follows naturally from theory. Unfortunately, dividing cultural capital and its potential by the working labour force doesn’t give significant results and is therefore not reported.
Surprisingly, when all variables, except human capital and job density, are divided by the working labour force the model performs satisfactory, i.e. all variables have the expected sign and have a significant influence on productivity growth. However, such a specification of the model is not in line with regional growth theory, e.g. physical capital accumulation and R&D expenditures only have an effect at the working place, not at the place where workers live. The corresponding output is therefore not reported.
The specification of physical capital is another discussion point. In section 3 I argued that gross investments by firms is a good proxy. This interpretation is very general and should be treated more specifically. However, the construction of high-quality data on the capital stock is very complicated and beyond the scope of this paper. Nonetheless, an alternative specification for the growth of the capital stock is provided by Dorwick and Nguyen (1989)7. I have constructed physical capital following their specification. Unfortunately, its effect on labour productivity growth is insignificant in the model.
The final discussion point is the existence of Granger causality between the variables. Granger causality implies that variables share information. It does not indicate causality in the more common use of the term. Especially causality between an explanatory variable and the dependent variable is important to detect. That is, when the dependent variable ‘Granger causes’ an explanatory variable, one should be cautious in drawing conclusions. In cultural economic studies Granger causality is often present since cultural- and social capital influence growth via 7 1 1 / / log ≈ − ≈ ⋅ −
6. Conclusion
This paper investigates the relationship between cultural capital growth and the growth of labour productivity for 40 corop-regions within the Netherlands between 1995 and 2005. Theory predicts that cultural capital should lead to productivity improvements by the labour force, i.e. technological improvements. Ultimately, economic growth should follow. A regional growth model is constructed in which regional productivity growth is explained by cultural capital, the potential of cultural capital, physical capital, R&D expenditures, and job density.
The estimation of the basic regression shows a positive but insignificant link between the growth of cultural capital and productivity growth. Moreover, the potential of cultural capital is insignificantly negative. The other explanatory variables are insignificant as well. R&D is the only exception. Its effect is positive and significant. The results do not change when significant period fixed effects are included. However, the direct effect of cultural capital on productivity growth is significantly positive. That is, when all other variables are excluded.
The weak performance of the model is probably due to some severe limitations. One limitation is the measurement problem for some of the variables. For example, cultural capital is measured by the cultural expenditures of local government. Expenditures by consumers, national cultural subsidies, and private gifts are neglected due to data limitations. In addition, data on regional physical capital inputs and regional human capital are difficult to attain. Moreover, the time series are relatively short, which reduces the long term power of the model.
References
Baltagi, B.H. (1995): Econometric Analysis of Panel Data, John Wiley & Sons Ltd, Chichester (UK).
Barro, R.J. (1991): Economic Growth in a Cross Section of Countries, Quarterly Journal of Economics 106 (2), p. 407–443.
Barro, R.J. and Sala-i-Martin, X. (1995): Economic Growth, McGraw-Hill, New York (US).
Boschma, R.A. et al. (2005): Creatieve Klasse en Regionaal-Economische Groei, Den Haag: OCW, Available at: http://www.cultuureneconomie.nl/_pdf/onderzoek_Universiteit_Utrecht.pdf.
Boulding, K.E. (1972): Toward the Development of a Cultural Economics, Social Science Quarterly 53 (2), p. 267–284.
Broersma, L. and Oosterhaven, J. (2006): Regionale Arbeidsproductiviteit in Nederland: Agglomeratie- en Congestie-Effecten, Kwartaalschrift Economie 4, p. 397–419.
Chartrand, H.H. (1990): The Hard Facts: Perspectives of Cultural Economics, in: Transactions of the Royal Society of Canada 1989/Fifth Series/Vol. IV, University of Toronto Press, Toronto (CA), p. 1–4.
Dieckmann, O. (1996): Cultural Determinants of Economic Growth: Theory and Evidence, Journal of Cultural Economics 20 (4), p.297–320.
Florida, R. (2002): The rise of the creative class, Basic Books, New York (US).
Fujita, M., Krugman, P.R., and Venables, A.J. (1999): The Spatial Economy, Cities, Regions and International Trade, MIT Press, Cambridge MA.
Glaeser, E.L. (2004): Book Review of Richard Florida’s “The Rise of the Creative Class”, Available at: http://post.economics.harvard.edu/faculty/glaeser/papers/Review_Florida.pdf.
Hall, R.E. and Jones, C.I. (1999): Why Do Some Countries Produce So Much More Output per Worker than Others?, The Quarterly Journal of Economics 114 (1), p. 83–116.
Heijdra, B.J. and van der Ploeg, F. (2002): Foundations of Modern Macroeconomics, Oxford University Press, New York (US).
Hofstede, G. (1978): Organization-Related Value Systems in Forty Countries, European Institute for Advanced Studies in Management, Working Paper Series.
Klamer, A. (2002): Accounting for Social and Cultural Values, De Economist 150 (4), p.453– 473.
Klamer, A. (2003): Value of Culture, in: Towse, R. (ed.), A Handbook of Cultural Economics, Edward Elgar Publishing, Cheltenham (UK), p. 465–469.
Marlet, G.A. and van Woerkens, C.M.C.M. (2004): Het Economisch Belang van de Creatieve Klasse, Economisch Statistische Berichten 89 (4435), p. 280–286.
Marlet, G.A. and van Woerkens, C.M.C.M. (2004): Skills and Creativity in a Cross-Section of Dutch Cities, Utrecht School of Economics, Discussion Paper Series.
Meisel, J. (1974): Political Culture and the Politics of Culture, Canadian Journal of Political Science 7 (4), p. 601–615.
Rebelo, S.T. (1991): Long-Run Policy Analysis and Long-Run Growth, Journal of Political Economy 99 (3), p. 500–521.
Roo, M. (2005): Cultuur en Economische Groei?!, Wetenschapswinkel Economie
Appendix I: Corop Specification
a. Corop code (CR) and region
CR Region CR Region
1 Oost-Groningen 21 Agglomeratie Haarlem
2 Delfzijl en omgeving 22 Zaanstreek
3 Overig Groningen 23 Groot-Amsterdam
4 Noord-Friesland 24 Het Gooi en Vechtstreek
5 Zuidwest-Friesland 25 Agglomeratie Leiden en Bollenstreek 6 Zuidoost-Friesland 26 Agglomeratie 's-Gravenhage
7 Noord-Drenthe 27 Delft en Westland
8 Zuidoost-Drenthe 28 Oost-Zuid-Holland
9 Zuidwest-Drenthe 29 Groot-Rijnmond
10 Noord-Overijssel 30 Zuidoost-Zuid-Holland 11 Zuidwest-Overijssel 31 Zeeuwsch-Vlaanderen
12 Twente 32 Overig Zeeland
13 Veluwe 33 West-Noord-Brabant
14 Achterhoek 34 Midden-Noord-Brabant
15 Arnhem/Nijmegen 35 Noordoost-Noord-Brabant 16 Zuidwest-Gelderland 36 Zuidoost-Noord-Brabant
17 Utrecht 37 Noord-Limburg
18 Kop van Noord-Holland 38 Midden-Limburg 19 Alkmaar en omgeving 39 Zuid-Limburg
20 IJmond 40 Flevoland
CR Municipality CR Municipality CR Municipality CR Municipality 1 Bellingwolde 4 Tytjerksteradiel 12 Ensched 14 Lochem 1 Menterwolde 4 Vlieland 12 Haaksbergen 14 Neede 1 Pekela 5 Bolsward 12 Hellendoorn 14 Ruurlo 1 Reiderland 5 Gaasterlan-Sleat 12 Hengelo 14 Steenderen 1 Scheemda 5 Lemsterland 12 Hof van 14 Vorden 1 Stadskanaal 5 Nijefurd 12 Twente 14 Warnsveld
1 Veendam 5 Sneek 12 Losser 14 Wehl
1 Vlagtwedde 5 Wunseradiel 12 Oldenzaal 14 Winterswijk 1 Winschoten 5 Wymbritseradiel 12 Rijssen-Holten 14 Wisch 2 Appingedam 6 Heerenveen 12 Tubbergen 14 Zelhem 2 Delfzijl 6 Ooststellingwerf 12 Twenterand 14 Zutphen 2 Loppersum 6 Opsterland 12 Wierden 15 Angerlo 3 Bedum 6 Skarsterlan 13 Apeldoorn 15 Arnhem 3 De Marne 6 Smallingerland 13 Barneveld 15 Beuningen 3 Eemsmond 6 Weststellingwerf 13 Ede 15 Didam 3 Gronigen 7 Aa en Hunze 13 Elburg 15 Doesburg
3 Grootegast 7 Assen 13 Epe 15 Druten
3 Haren 7 Midden-Drenthe 13 Ermelo 15 Duiven 3 Hoogezand- 7 Noordenveld 13 Harderwijk 15 Groesbeek
3 Sappermeer 7 Tynaarlo 13 Hattem 15 Heumen
3 Leek 8 Borger-Odoorn 13 Heerde 15 Lingewaard
3 Marum 8 Coevorden 13 Nijkerk 15 Millingen aan
3 Slochteren 8 Emmen 13 Nunspeet 15 de Rijn
3 Ten Boer 9 De Wolden 13 Oldebroek 15 Nijmegen
3 Winsum 9 Hoogeveen 13 Putten 15 Overbetuwe
3 Zuidhorn 9 Meppel 13 Scherpenzeel 15 Renkum 4 Achtkarspelen 9 Westerveld 13 Voorst 15 Rheden 4 Ameland 10 Dalfsen 13 Wageningen 15 Rijnwaarden 4 Boarnsterhirn 10 Hardenberg 14 Aalten 15 Rozendaal
4 Dantumadeel 10 Kampen 14 Bergh 15 Ubbergen
4 Dongeradeel 10 Ommen 14 Borculo 15 Westervoort 4 Ferwerderadiel 10 Staphorst 14 Brummen 15 Wijchen 4 Franekeradeel 10 Steenwijkerland 14 Dinxperlo 15 Zevenaar 4 Harlingen 10 Zwartewaterland 14 Doetinchem 16 Buren 4 Het Bildt 10 Zwolle 14 Eibergen 16 Culemborg 4 Kollumerland c.a. 11 Bathmen 14 Gendringen 16 Geldermalsem 4 Leeuwarden 11 Deventer 14 Gorssel 16 Neder-Betuwe 4 Leeuwarderadeel 11 Olst-Wijhe 14 Groenlo 16 Lingewaal 4 Littenseradiel 11 Raalte 14 Hengelo 16 Maasdriel 4 Menaldumadeel 12 Almelo 14 Hummelo en 16 Neerijnen 4 Schiermonnik-oog 12 Borne 14 Keppel 16 Tiel
CR Municipality CR Municipality CR Municipality CR Municipality 16 Waal 18 Harenkarspel 23 Diemen 26 Wassenaar 16 Zaltbommel 18 Hoorn 23 Edam-Volendam 26 Zoetermeer 17 Abcoude 18 Medemblik 23 Graft-de Rijp 27 Delft 17 Amerongen 18 Niedorp 23 Haarlemmermer 27 Westland 17 Amersfoort 18 Noorder- 23 Landsmeer 27 Midden- 17 Baarn 18 Koggenland 23 Oostzaan 27 Delftland 17 Breukelen 18 Obdam 23 Ouder-Amstel 28 Alphen aan de
17 Bunnik 18 Opmeer 23 Purmerend 28 Rijn
17 Bunschoten 18 Schagen 23 Uithoorn 28 Bergambacht 17 De Bilt 18 Stede Broec 23 Waterland 28 Bodegraven 17 De Ronde Venen 18 Texel 23 Zeevang 28 Boskoop
17 Doorn 18 Venhuizen 24 Blaricum 28 Gouda
17 Driebergen- 18 Wester- 24 Bussum 28 Jacobswoude 17 Rijsenburg 18 Koggenland 24 Hilversum 28 Liemeer
17 Eemnes 18 Wieringen 24 Huizen 28 Moordrecht
17 Houten 18 Wieringermeer 24 Laren 28 Nieuwkoop 17 Ijsselstein 18 Wognum 24 Muiden 28 Reeuwijk
17 Leersum 18 Zijpe 24 Naarden 28 Rijnwoude
17 Leusden 19 Alkmaar 24 Weesp 28 Schoonhoven
17 Loenen 19 Bergen 24 Wijdemeren 28 Ter Aar
17 Lopik 19 Heerhugowaard 25 Alkemade 28 Waddinxveen
17 Maarn 19 Heiloo 25 Hillegom 28 Zevenhuizen-
17 Maarssen 19 Langedijk 25 Katwijk 28 Moerkapelle 17 Montfoort 19 Schermer 25 Leiden 29 Albrandswaard 17 Nieuwegein 20 Beverwijk 25 Leiderdorp 29 Barendrecht 17 Oudewater 20 Castricum 25 Lisse 29 Bergschenhoek 17 Renswoude 20 Heemskerk 25 Noordwijk 29 Berkel en 17 Rhenen 20 Uitgeest 25 Noordwijkerhout 29 Rodenrijs
17 Soest 20 Velsen 25 Oegstgeest 29 Bernisse
CR Municipality CR Municipality CR Municipality CR Municipality 29 Middelharnis 32 Schouwen- 35 Maasdonk 37 Maasbree 29 Nederlek 32 Duiveland 35 Mill en Sint 37 Meerlo- 29 Nieuwerkerk aan 32 Tholen 35 Hubert 37 Wanssum
29 den Ijssel 32 Veere 35 Oss 37 Meijel
29 Oostflakkee 32 Vlissingen 35 Schijndel 37 Mook en 29 Oud-Beijerland 33 Bergen op Zoom 35 Sint Antonis 37 Middelaar
29 Ouderkerk 33 Breda 35 Sint- 37 Sevenum
29 Ridderkerk 33 Drimmelen 35 Michielsgestel 37 Venlo 29 Rotterdam 33 Etten-Leur 35 Sint-Oedenrode 37 Venray 29 Rozenburg 33 Geertruidenberg 35 Uden 38 Ambt 29 Schiedam 33 Halderberge 35 Veghel 38 Montfort 29 Spijkenisse 33 Moerdijk 35 Vught 38 Echt- 29 Strijen 33 Oosterhout 36 Asten 38 Sustern 29 Vlaardingen 33 Roosendaal 36 Bergeijk 38 Haelen
29 Westvoorne 33 Rucphen 36 Best 38 Heel
30 Alblasserdam 33 Steenbergen 36 Bladel 38 Heythuysen 30 Dordrecht 33 Woensdrecht 36 Cranendonck 38 Hunsel 30 Giessenlanden 33 Zundert 36 Deurne 38 Maasbracht 30 Gorinchem 34 Aalburg 36 Eersel 38 Nederweert 30 Graafstroom 34 Alphen-Chaam 36 Eindhoven 38 Roerdalen 30 ‘s-Gravendeel 34 Baarle-Nassau 36 Geldrop-Mierlo 38 Roermond 30 Hardinxveld- 34 Dongen 36 Gemert-Bakel 38 Roggel en 30 Giessendam 34 Gilze en Rijen 36 Heeze-Leende 38 Neer 30 Hendrik-Ido- 34 Goirle 36 Helmond 38 Swalmen 30 Ambacht 34 Hilvarenbeek 36 Laarbeek 38 Thorn 30 Leerdam 34 Loon op Zand 36 Nuenen c.a. 38 Weert 30 Liesveld 34 Oisterwijk 36 Oirschot 39 Beek
30 Nieuw- 34 Tilburg 36 Reusel- 39 Brunssum
30 Lekkerland 34 Waalwijk 36 De Mierden 39 Eijsden 30 Papendrecht 34 Werkendam 36 Someren 39 Gulpen- 30 Sliedrecht 34 Woudrichem 36 Son en Breugel 39 Wittem 30 Zederik 35 Bernheze 36 Valkenswaard 39 Heerlen 30 Zwijndrecht 35 Boekel 36 Veldhoven 39 Kerkrade
31 Hulst 35 Boxmeer 36 Waalre 39 Landgraaf
31 Sluis 35 Boxtel 37 Arcen en Velden 39 Maastricht
31 Terneuzen 35 Cuijk 37 Beesel 39 Margraten
32 Borsele 35 Grave 37 Bergen 39 Meerssen
32 Goes 35 Haaren 37 Gennep 39 Nuth
Appendix II: Variables: Model Name, Source, and Explanation
Variable Source Explanation
Value added (1995-2005)
CBS, REY
- Economic sectors included: agriculture, forestry, and fishery; industry; energy and water supply; construction; trade,
catering and repair; transport, storing, and communication; financial and business services; environmental services. - Value added of the mineral winning sector is excluded in all regions. Cultural capital (1995-2007) CBS, FM
- Cultural expenditures of local governments, aggregated to corop level.
- Types of cultural expenditure: public library; informal education & development projects; art; archeology & museums; social quality of life & open-air recreation, and; remaining expenditures on culture and recreation.
Physical capital (1996-2005)
CBS, REY
- Gross investment of companies to type of (fixed) asset: business buildings; land, road, and water operations; vehicles; machinery; maintenance; remaining fixed assets.
Volume of labour (1995-2005)
CBS, REY
The amount of labour years used in the production process. -- The mineral winning sector is excluded.
Human capital (1995-2005)
CBS, LFS
- Percentage of the working labour force with a bachelor or master degree.
Job density (1995-2005)
CBS, LFS
- The number of jobs per km².
- Km² implies the surface of land. Water surface is excluded. R&D
expenditures (1996-2003)
CBS, IRD
- Research and development expenditures (R&D) of
businesses and research institutions. They are only available at the province level.
- The value added share of each corop-region in its province is multiplied by the R&D expenditures of businesses and
research institutions in that province in order to get R&D expenditures at the corop level.
Potential of cultural capital (1995-2005) CBS/ B&O
- The potential of culture is computed as follows ) exp( ) ( , ,s r rs s s r CULT t CULT P =
∑
− ⋅≠
β
, with tr,s being thetraveling time between corop r and corop s and β = -0.0392 (-LOG(15)/30), i.e. the half-value of cultural capital after 30 minutes of traveling; the starting point being the border of region r.
Note: CBS: Dutch statistical institute; REY: Regional Economic Year figures; FM: Finances Municipalities; LFS: Labour Force Survey; IRD: Innovation and Research&Development; B&O: Broersma and
Appendix III: Descriptive Statistics
a. Total
Variable Notation Specification Mean Max Min StDev a. VA/L euro p/labyr 62740 93967 50249 6754 b. K/L euro p/labyr 10965 30418 6432 2957
d. HC % 0.26 0.45 0.10 0.07
e. CULT/L euro p/labyr 419 701 293 79 f. P(CULT)/L euro p/labyr 1030 3668 114 813 g. JD nr of jobs p/km² 270 1688 44 288 h. R&D/L euro p/labyr 1182 4475 192 818 Note: p/labyr = per employed labour year
Variable a. VA/L b. K/L
Variable e. CULT/L f. P(CULT)/L
CR Mean Max Min StDev Mean Max Min StDev
1 480 518 450 25 583 640 544 35 2 537 616 481 45 1859 2125 1702 168 3 467 643 415 67 185 194 181 4 4 419 432 405 11 134 142 131 4 5 498 549 471 22 1581 1644 1517 44 6 400 446 368 28 734 766 690 29 7 401 478 347 42 744 770 707 19 8 454 505 385 38 609 657 583 30 9 437 492 396 33 1295 1418 1151 102 10 381 419 348 24 460 477 441 13 11 426 540 354 64 1726 1769 1675 34 12 342 440 312 37 192 199 187 4 13 325 382 300 24 457 473 447 8 14 354 401 322 27 480 507 466 14 15 389 422 356 23 348 355 337 7 16 332 372 305 24 2340 2506 2169 115 17 333 382 293 28 262 275 251 9 18 471 521 445 24 707 730 681 16 19 451 503 419 30 1535 1601 1455 52 20 456 523 425 35 2448 2542 2352 57 21 608 692 533 60 2267 2385 2178 65 22 392 482 343 45 2951 3052 2873 77 23 416 492 365 42 182 192 170 8 24 334 369 305 24 1832 1983 1768 71 25 404 443 384 21 1577 1624 1530 32 26 458 544 399 48 575 593 555 13 27 373 409 342 23 2410 2541 2275 90 28 349 381 318 23 1776 1839 1709 47 29 624 701 582 39 225 234 217 5 30 382 412 358 20 1226 1298 1169 46 31 428 507 398 30 450 461 442 7 32 559 604 511 25 572 590 553 11 33 400 413 379 11 528 545 509 12 34 415 446 384 21 841 876 816 21 35 369 402 346 20 472 508 446 21 36 344 384 319 22 290 301 278 8 37 322 358 303 19 655 676 639 11 38 370 394 356 14 892 916 873 17 39 414 486 390 27 116 123 114 3 40 446 505 393 40 955 1061 856 77
Variable g. JD h. R&D/L
CR Mean Max Min StDev Mean Max Min StDev
1 52 55 48 3 432 586 284 111 2 61 64 59 2 685 979 481 183 3 125 134 110 9 2032 2240 1851 180 4 72 76 62 5 427 560 295 85 5 52 58 44 5 438 571 302 88 6 61 67 51 6 421 539 290 81 7 63 67 59 3 425 792 329 153 8 61 64 58 2 433 789 333 149 9 62 71 44 10 432 783 336 147 10 96 110 78 13 705 848 599 97 11 137 147 124 8 669 808 562 91 12 156 166 137 10 1194 1316 1098 83 13 137 148 114 11 1624 1706 1378 111 14 91 97 80 6 914 1006 710 103 15 301 327 258 25 1041 1699 742 284 16 106 117 87 12 1032 1145 813 116 17 401 433 339 33 1458 1798 1182 229 18 103 114 84 10 807 915 756 50 19 293 329 244 33 796 897 746 49 20 427 459 404 19 922 1085 843 73 21 639 678 591 28 815 932 762 56 22 507 533 464 22 837 939 787 47 23 946 1021 789 87 1546 1669 1483 62 24 529 575 483 31 856 995 787 70 25 610 653 537 45 1948 2191 1699 204 26 1513 1688 1362 104 858 1064 626 174 27 581 670 483 76 4099 4475 3772 299 28 231 250 199 19 868 1069 645 172 29 462 493 407 33 1220 1443 962 190 30 309 334 263 25 859 1062 631 172 31 54 56 49 3 593 845 253 215 32 86 93 75 6 469 650 192 169 33 197 212 161 19 1884 2196 1595 212 34 190 207 153 20 1962 2258 1679 195 35 191 205 164 15 1671 1949 1397 206 36 215 231 188 16 2007 2294 1696 200 37 133 143 115 10 3698 4381 3108 465 38 128 139 111 9 1062 1222 779 149 39 358 380 317 22 1616 1818 1400 146 40 68 82 49 12 1536 1935 1166 289
Variable d. HC
Appendix IV: ADF Test Results
variable ADF t-statistic p-value obs
∆log(VA/L) 176.009 0.000 320 ∆log(K/L) 298.726 0.000 320 ∆log(CULT/L) 204.318 0.000 320 ∆log(PCULT/L) 171.845 0.000 320 ∆log(R&D/L) 190.222 0.000 200 ∆log(JD) 187.104 0.000 320 ∆log(HC) 317.691 0.000 320 log(K/L) 46.3753 0.9991 360 log(R&D/L) 50.7802 0.9956 240 log(JD) 29.5556 1.0000 360
Appendix V: Correlation Matrix
∆log: (VA/L) (K/L) (PCULT/L) (CULT/L) (R&D/L) (JD) (HC)
Appendix VI: Granger Causality Output
∆log (variable)
Granger causes: (VA/L) (K/L) (PCULT/L) (CULT/L) (R&D/L) (JD) (HC) (variable): (VA/L) - 0.015 0.001 0.040 - - - (K/L) 0.005 - 0.000 - 0.073 0.028 - (PCULT/L) 0.000 - - 0.000 - 0.000 0.000 (CULT/L) 0.000 - - - - 0.016 0.014 (R&D/L) - - 0.001 0.000 - 0.004 - (JD) 0.083 - - 0.012 - - 0.078 (HC) - 0.011 - - - - -
