Resource productivity and growth : what is the effect of rising resource productivity on GDP growth in the European Union?

(1)

Resource Productivity and Growth

What is the effect of rising resource productivity on

GDP growth in the European Union?

Author: Djoeke Jolien Kraaijeveld Student nr.: 10538913

University: University of Amsterdam

Bachelor course: Bèta-Gamma, major Economy Supervisor: Péter Foldvari

(2)

1

Abstract:

This thesis looks at the question: What is the effect of rising resource productivity on GDP growth in

the European Union? This is an interesting question because the European Commission claims that

improving resource productivity is not only good for the environment but will also lead to growth in the European economy.

To address this question, the theories on GDP growth of Solow (1956), Mankiw et al. (1992) and Barro (2003) (1991) are used. In this paper a dynamic panel data regression is done to see if resource productivity affects GDP growth positively. This is done using the fixed effects estimator and the system GMM estimator introduced by Roodman (2006).

What is found, is that when looking at GDP per capita growth, resource productivity has a positive effect on growth. However, the control variables used where not all significant when using system-GMM.

This document is written by student Djoeke Jolien Kraaijeveld who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

1. Introduction

Climate change is currently a prevalent threat to life on this planet. And as the number of people on earth keeps rising, the amount of resources is declining rapidly. Globally there is an increasing pressure on our resources and the environment. Sustainable development, including sustainable resource use, is seen as a solution to this problem (Krausmann et al., 2017).

The United Nations has devised a plan consisting of 17 Sustainable Development Goals (SDG). The European Union (EU) has committed to implement the SDG’s both in its internal and external policies. Through the ‘Roadmap to a resource efficient Europe’ (2011), and the more recent ‘The Action Plan towards the Circular Economy’ (2015) the European Union has committed to create an environment to improve the resource efficiency and aid the transition towards the circular economy. Hoping this will bring economic benefits and can contribute to innovation, growth and job creation (Domenecha & Bahn-Walkowiakb, 2017). The European Union is aiming towards higher resource productivity and a lower impact on the environment. But not only climate change is their motivator to work toward this goal, there is also the notion that this will lead to growth in the European economy.

Because resource productivity has such a prominent role in the EU plans to reach the Sustainable Development Goals (SGD), this research will focus on this specific factor of sustainable development. The research will investigate if this really leads to the promised growth in the European economy. From these research topics, the research question was formed: What is the effect of rising resource

productivity on GDP growth in the European Union?

This paper is divided in 5 sections, including section 1, the introduction. Section 2 gives an overview of literature on GDP growth, resource productivity and the relationship between these factors. This is followed by the methodology and data collection in section 3, where the dynamic model is described and how the problems with this model are solved is explained. Then, the results are analysed in section 4. And finally, the paper will be concluded and discussed in section 5.

(5)

4

2. Literature review

2.1 GDP growth

This part analyses aspects of economic growth that are of use to answer the research question in this paper. It discusses the neoclassical growth model of Solow (1956) and the endogenous growth models of Mankiw et al. (1992) and Barro (1991) (2003).

Gross Domestic Product (GDP) per capita is an essential indicator of economic growth and activity. It is the monetary measure of the market value of all final goods and services produced in a period. This is often used to express the standard of living. The higher the GDP per capita, the better the standard of life in an economy.

In 1965 Solow used the Harrod-Damar model to argue a model in which there is a growth equilibrium in the long run in the economic system. This neo-classical model became very influential to explain GDP growth. The model attempts answer why total output of a country grows and why different countries grow at different speed (Solow, 1956).

Solow (1956) models GDP as total output in the function Y = F(K, AL), in this function L is labour, C is capital stock and A is the level of technology. This means that growth in the Solow model can only be explained by these variables. Technology and labour grow at exogenous rates and the level capital growth depends on the savings rate. The model predicts that higher saving lead to higher investment and subsequently higher income (Solow, 1956).

This function expresses labour as effective labour and has constant return to scale, so it is homogeneous of first degree. This implies that the input factors have a one to one relationship with the output.

When 𝑘 = 𝐾/𝐿 is defined, this means 𝑘 is capital per worker and 𝐹(𝑘, 1) = 𝑓(𝑘) is the production function. The slope of this function is the marginal product of capital. This function is equal to the output per worker 𝑦 = 𝑌/𝐿. 𝑦 can be divided into two parts; consumption per worker, and investment per worker. The model assumes that every year a person saves a fraction of its income, the savings rate 𝑠, the amount of money saved will be the investment 𝑖 = 𝑠𝑦 = 𝑠𝑓(𝑘). This investment will cause the capital stock to rise. But this capital that was invested in will similarly depreciate when it gets older, the speed at which this depreciates is the depreciation rate 𝛿. This means that the deprecation of capital per worker can be expressed by the function 𝛿𝑘 (Solow, 1956). When the function for the depreciation intercepts the investment function there is a steady state level of capital stock, as can be seen in figure 1.

(6)

5

Figure 1:The Steady State

There is convergence to the steady state where the investment is equal to the depreciation rate. This is when there is a constant capital labour ratio. At that point an economy still grows at the exogenous rate. Whatever the initial value of the capital-labour ratio is, the system will develop towards this state of balanced growth at natural rate. When the initial capital stock is below the equilibrium ratio, capital and output will grow faster than labour force until the steady state is reached. But when the initial capital-labour ratio is above the equilibrium ratio, capital and output will grow more slowly than the labour force. The economy will always converge to this steady state. The steady state references the long-run equilibrium of the economy (Solow, 1956).

Solow (1956) states that initial growth is a major indicator for future growth. And the differences in the speed of technological change are said to explain much of the variation in growth rates seen between countries.

According to Mankiw, Romer and Weil (1992), the direction of the variables in this original Solow model is modelled correctly, but because of omitted variable bias the coefficients are too large. They suggest that this omitted variable is human capital. They argue that human capital, population growth, physical capita and technological improvements are important variables to predict GDP growth per capita. This more extended model proved to be a better fit to the real world than the traditional Solow model (Mankiw, Romer, & Weil, 1992).

The version of the Solow model predicts that the output levels of poor countries will converge towards that of rich countries only when the poor countries have similar savings rates and human capital. This is known as conditional convergence. Still, savings rates are very different between countries and so

(7)

6 is the amount of human capital. This can be explained by factors like culture, geography and religion (Mankiw, Romer, & Weil, 1992). This means that absolute convergence can only take place when the countries have similar characteristics in education policy, institutions, the same level of free market and openness to trade.

On the last careerists Barro elaborated further. He conducted a research in which he did cross-country panel regressions to show the differences in per capita growth rates in countries, relate systematically to a set of explanatory variables. He stresses the importance of human capital and finds that for given per capita GDP, high human capital predicts higher growth. He also looks to other explanatory variables like life expectancy, fertility, government consumption, openness to international trade, rule of law and democracy (Barro, 2003).

Barro uses literacy rate, school enrolment and the student-teacher ratio in secondary education as proxies for human capital. School enrolment rate is a proxy that is very often used in economic research (Barro, 1991).

In these theories lots of variables have been suggested to predict GDP growth. Initial level of GDP per capita is an important factor to determine growth (Mankiw, Romer, & Weil, 1992).

(8)

7

2.2 Resource productivity

In this part, the paper will dwell upon resource productivity. First a short explanation about why this is an important notion will be given. Then how it is defined and what definition will be use in the remaining paper. And lastly how it correlates to other sustainability concepts.

When the exploitation of natural resources in the western world started to lead to long-term growth, these resources appeared inexhaustible. Through that, it was viewed as an inexhaustible source of wealth. Since then the world population has grown rapidly and more will join the middle class in the coming decade like the BRIC countries. This all increase, the strain on the not so inexhaustible resources (Schulte, 2013). Driven by rapid growth of emerging economies, from 1970 to 2010 annual global extraction of raw materials has tripled (Domenecha & Bahn-Walkowiakb, 2017).

To downplay the problems with supply security and competitiveness by changes in the commodity market and price volatility of raw resources, the attention has significantly shifted towards resource efficiency/productivity (Domenecha & Bahn-Walkowiakb, 2017). By means of the article of Domenecha & Bahn-Walkowiakb resource productivity can also be called resource efficiency but for simplicity resource productivity will be used for the remainder of this paper.

Resource productivity refers to the economic value extracted from natural resources (Domenecha & Bahn-Walkowiakb, 2017). The definition of the European Commission for resource productivity is domestic material consumption (DMC) divided by gross domestic product (GDP), this is their headline indicator of its ‘resource efficiency roadmap’. Domestic Material Consumption (DMC) is the total amount of raw materials extracted from the domestic territory minus all physical export plus all imports in tons (Wiedmann, et al., 2015). There are other less mainstream ways to measure resource productivity, these measures take in account raw materials used in finished products for import and export (Domenecha & Bahn-Walkowiakb, 2017).

The often-used concept of circular economy is much wider than that of resource productivity, but never the less, reaching a higher level of resource productivity is part of the transition to a circular economy (Domenecha & Bahn-Walkowiakb, 2017). Most of the economy can be described as a linear sequence from extraction to production to consumption to waste to disposal. Schulte (2013) proposes a circular economy, where waste is used as a resource and new products are produced with the same materials. By maximizing the productivity of the natural resources, new business models will emerge that will ensure prosperity despite population growth, with its demand on exhaustible resources (Schulte, 2013). Schulte (2013) discusses the incentives to stimulate enterprises to enter more resource productive ways, into a circular economy.

(9)

8 A high level of raw resource use is not a new problem, Krausmann et al. (2009) gives us a more historical review of raw material use in the 20th century and the effects/correlations with gross domestic product (GDP). To do this, the material flow accounts are discussed. It is concluded that resource productivity has risen, but this does not slow down the growth of global material use (Krausmann, et al., 2009).

In 2006, Stern described the economic implications of climate change. He described this as one of the most difficult problems the world faces today. It involves externalities, that will impact those people that are the most poorly represented; the future generation, long-term horizons on a global scale. It is not only those that pollute and use the natural resources now that will be affected. He states that the government needs to understand that growth and climate-change are highly correlated. He concludes his article by emphasizing the role the European Union in the international action needed for an effective response to climate change (Stern, 2006).

(10)

9

2.3 Policy

This section will look at the policies concerning resource productivity.

Domenecha & Bahn-Walkowiakb (2017) give a comprehensive overview of current policy on resource productivity and circular economy. They discuss the most recent data and the notable differences between European countries. Also, it shows the efforts towards decoupling domestic material consumption (DMC) from GDP (Domenecha & Bahn-Walkowiakb, 2017).

The United Nations set Sustainable Development Goals (SDGs) this is a series of goals set to improve the sustainability of the planet. These goals are promoted as the worldwide goals in sustainable development. Goal number 12 is the goal to strive towards ensuring sustainable consumption and production patterns (United Nations, 2016). The main way this goal is interpreted is by policies to increase resource productivity.

In the European Union suitable development has been a goal long before the SDGs were introduced, and environmental protection is one of the key factors in EU law (Domenecha & Bahn-Walkowiakb, 2017). In 1997 sustainable development became one of the European Union’s objectives after it was included in the Treaty of Amsterdam as a main objective of the Union’s policies. In 2001 in Gothenburg the Sustainable Development Strategy (SDS) was set. This was a list of goals to make the European Union more sustainable. These goals are recently replaced with the Sustainable Development Goals of the United Nations.

To reach the goal of a more resource productive and efficient Europe, the TFEU treaty signed in Lisbon gives guidelines for EU environmental policy. This includes the principal to ensure “prudent and rational utilization of resources” which provides a legal foundation for policy on resource productivity and circular economy (Domenecha & Bahn-Walkowiakb, 2017).

The European Commission has initiated two main policies: ‘’ Roadmap to a Resource Efficient Europe’’ and ‘’ A resource-efficient Europe’’ which is a flagship initiative under the Europe 2020 Strategy. The ‘’Roadmap to a Resource Efficient Europe’’ outlines how resource efficient growth can be achieved and describes the importance of this for the future wellbeing in the European Union. It suggests actions for the sectors that consume the most resources. The roadmap is presented as an agenda for competitiveness and growth, based on creating business and job opportunities by improving resource productivity. It allows governments to shift taxation away from labour and activities such as recycling, better product design, materials substitution and eco-engineering and toward resource use and pollution. The roadmap implies that rising resource productivity will lead to

(11)

10 more growth in the European economy. This roadmap also discusses the role of the European Union in rising resource productivity in international context (European Commission, 2011) .

The policy “A resource-efficient Europe’’ also states that rising resource productivity will lead to growth. And pleads for taxation on resources and other measures to ensure a rising resource productivity in Europe. There many are other policies that promote resource efficacy; ‘’the Circular economy action plan’’, ‘’The flagship initiative on resource efficiency and the resource efficiency roadmap’’ and ‘’ The circular economy package’’ (European Commission, 2011) .

The importance of the European Union implementing policy to reduce climate change has been pressed by Stern in 2006. He describes how the European Union can play an effective role in fostering and promoting cooperation, effective action on climate change can only be taken in cooperation with the rest of the world (Stern, 2006).

(12)

11

2.4 Previous research on GDP and resource use

That resource productivity influences GDP and GDP growth is a concept most economics in this field agree on. But on how to measure this are different opinions. That is why this section looks at these different ideas.

In his article Krausmann et al. (2009) research the concept of resource decoupling. This concept often used to show that if resource use. When domestic material consumption (DMC) is growing at the same rate as GDP there is no decoupling. But the lower the growth rate of DMC in contrast with GDP growth rate the higher the resource decoupling. When there is GDP growth and resource use decreases there is absolute decoupling. The concept of resource decoupling is illustrated in figure 2.

Figure 2: Resource decoupling

Decoupling human well-being from resource consumption means using less resources per unit of economic output and reducing the environmental impact of these resources that are used. This is also the way Eurostat, the statistical analysing agency of the European Commission, looks at resource use in the European Union (Eurostat, 2017).

In the article of Wiedmann et al. (2015) different measures for resource productivity are used. In this article, multivariate regressions are done on both indicators to see what indicator can produce a more statistical accurate model to look at effects of GDP, domestic extraction (DE) and population density in different countries. He concludes that it would be better to also consider the raw materials used to make products that are imported by a country (Wiedmann, et al., 2015).

Tanning (2014) looks at raw resource use in upcoming Eastern European countries and finds that for these countries that higher resource use not always leads to GDP growth. Also, he finds that analysing

(13)

12 resource productivity would be better strive toward in order to achieve economic growth in these countries.

Cross-country regressions have been done for GDP and resource productivity both for European countries and worldwide for all countries with available data. All these researches showed a positive effect of resource productivity measured by GDP divided by domestic material consumption between the different countries.

In the article of Tampakoudis et al. (2014) a pooled ordinarily squares regression is done to look at the EU over time and see which of the highlighted sustainable development indicators of the European Union have a significantly positive or negative impact on GDP growth and what indicators have no significant impact. In this research, resource productivity is one of the indicators that has a positive impact with 5% significance. However, this was a static model.

Nordhaus introduced the concept of green GDP. This was a measure of the economy that adjusts for environmental damages. In 1972 Nordhaus together with Tobin enthused their Measure of Economic Welfare (MEW) as an alternative to GDP to measure economic wellbeing. This measure consists of value of GDP, value of leisure time, value of unpaid work and value of environmental damage. This value of environment damage by industrial production and consumption, reduces the value of GDP (Nordhaus & Tobin, 1972) (Nordhaus W. D., 1993).

It can be concluded that there are many articles on the effects of resource productivity and efficiency and GDP growth and GDP, but most are about decoupling or do a cross-country analysis at one moment in time. This paper will look at the effect of resource productivity in the European Union over time.

(14)

13

3. Data and Methodology

In this paper a panel data regression on 25 countries over 16 years is conducted.

3.1 Empirical design

The aim of this section is to explain the empirical design of this research. The model is used to explain the effect of resource productivity on GDP growth and the econometric tools to obtain robust and consistent indicators.

3.1.1 Model

A conditional convergence model is considered:

𝐿𝑛(𝑦𝑖,𝑡) − 𝐿𝑛(𝑦𝑖,𝑡−1) = 𝛽0+ 𝛽1𝐿𝑛(𝑦𝑖,𝑡−1) + 𝛽2𝑟𝑝 + 𝛽2𝑔𝑓𝑐𝑓 + 𝛽3𝑢𝑚𝑝 + 𝛽4𝑒𝑑𝑢 + 𝛽5𝑔𝑜𝑣 + 𝜇𝑡+ 𝜀𝑖𝑡

Where 𝑦𝑖,𝑡 is GDP per capita in country i at year t. The explanatory variables are initial GDP per capita

𝑦𝑖,𝑡−1 , resource productivity and the other control variables mentioned in section 2; these variables

vary across time and country. rp is resource productivity, gfcf is Gross Fixed Capital Formation in percentages of GDP of the country, ump is the unemployment rate, edu is the education rate described above that will be use as a proxy for human capital, gov is government consumption in percentages of GDP in a country. 𝜇𝑡 denotes the unobserved country specific effects that affect GDP per capita growth, like political and geographic factors. 𝜀𝑖,𝑡 is the stochastic error-term and are the unobserved time specific effects.

The former equation can be written in the following format, where 𝛽̂1= 1 + 𝛽1

𝐿𝑛(𝑦𝑖,𝑡) = 𝛽0+ 𝛽̂1𝐿𝑛(𝑦𝑖,𝑡−1) + 𝛽2𝑟𝑝 + 𝛽2𝑔𝑓𝑐𝑓 + 𝛽3𝑢𝑚𝑝 + 𝛽4𝑒𝑑𝑢 + 𝛽5𝑔𝑜𝑣 + 𝜇𝑖+ 𝜀𝑖,𝑡

This is a dynamic panel data model, for it contains a lagged dependent variable; GDP in the previous period is a predictor for the current value of the dependent variable. We want to include this because according to Solow (1956) described in section 2 this is an important predictor.

Now we look at the Least Squares assumptions;

• The error them has conditional mean zero. 𝐸(𝜀𝑖,𝑡|𝑋𝑖) = 0

• (𝑋_𝑖, 𝑌𝑖), 𝑖 = 1,2, … , 𝑛, are independent and identically distributed (i.i.d.) • Large outliers are unlikely

(15)

14 There is assumed that large outliers are unlikely. The table shows the descriptive statistics, there can be seen that large values are found but can be explained by economic events. As one can see in Table 1 this is unlikely.

In the model there are problems, some regressors can be correlated with the error term and there are time-nonspecific country effects, such as location, demographics and these may be correlated with the explanatory variables. These effects will be part of the error term. These are fixed effects and there should be control for these unobserved variables that differ from one country to the next (Stock & Watson, 2015). There are two common ways to solve this; by using fixed effect estimator or first difference estimator. The first difference method starts with taking the first difference of the equation above, this way we get rid of the fixed effects; ∆𝐿𝑛(𝑦𝑖,𝑡) = ∆𝛽̂1𝐿𝑛(𝑦𝑖,𝑡−1) + ∆𝛽2𝑟𝑝 + ∆𝛽2𝑔𝑓𝑐𝑓 + ∆𝛽3𝑢𝑚𝑝 + ∆𝛽4𝑒𝑑𝑢 + ∆𝛽5𝑔𝑜𝑣 + ∆𝜀𝑖,𝑡 because these are the same for every period ∆𝜇𝑖 = (𝜇𝑖− 𝜇𝑖) = 0. The fixed effects estimator takes the equation above and does this minus its average; 𝐿𝑛(𝑦𝑖,𝑡) − 𝐿𝑛(𝑦̅̅̅̅̅̅̅̅̅̅ = (𝛽̂𝑖,𝑡) 1𝐿𝑛(𝑦𝑖,𝑡−1) − 𝛽̂̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅) + (𝛽1𝐿𝑛(𝑦𝑖,𝑡−1) 2𝑟𝑝 − 𝛽̅̅̅̅̅̅) + (𝛽2𝑟𝑝 2𝑔𝑓𝑐𝑓 − 𝛽̅̅̅̅̅̅̅̅̅̅) +2𝑔𝑓𝑐𝑓 (𝛽3𝑢𝑚𝑝 − 𝛽̅̅̅̅̅̅̅̅̅) + (𝛽3𝑢𝑚𝑝 4𝑒𝑑𝑢 − 𝛽̅̅̅̅̅̅̅̅) + (𝛽4𝑒𝑑𝑢 5𝑔𝑜𝑣 − 𝛽̅̅̅̅̅̅̅̅) + (𝜀5𝑔𝑜𝑣 𝑖,𝑡− 𝜀̅̅̅̅). By using this method 𝑖,𝑡 we also get rid of the fixed effect because the average of the fixed effects is the fixed effects 𝜇𝑖− 𝜇̅ =𝑖 𝜇𝑖−

1

𝑇∑ 𝜇𝑖

𝑇

𝑡=1 = 0.

Another problem is that this lagged variable is also determined with its lagged past variable. So, the lagged dependent variable can be correlated to the error term and some explanatory variables might be endogenous all leading to bias. This is called Nickell-bias.

To compensate for this bias, Arellano and Bond (1991) developed the General Methods of Moments (GMM) estimators. This considers a system of equations. Every time-period is a question, and these are all equal except for the moment conditions.

The method starts with first taking the first difference like the first differed estimator, to get rid of the fixed effects; ∆𝐿𝑛(𝑦𝑖,𝑡) = 𝛽0+ ∆𝛽̂1𝐿𝑛(𝑦𝑖,𝑡−1) + ∆𝛽2𝑟𝑝 + ∆𝛽2𝑔𝑓𝑐𝑓 + ∆𝛽3𝑢𝑚𝑝 + ∆𝛽4𝑒𝑑𝑢 + ∆𝛽5𝑔𝑜𝑣 + ∆𝜀𝑖,𝑡. This does not solve the problem that the new error term ∆𝜀𝑖,𝑡 is correlated with the lagged dependent variable ∆𝛽̂1𝐿𝑛(𝑦𝑖,𝑡−1). Then it uses the Methods of Moments (GMM). This estimator developed by Arellano and Bond is often called the difference-GMM estimator (Roodman, 2006).

Arellano and Bover (1995) and Blundell and Bond (1998) later found weaknesses in this estimator when the dependent variable is a unit-root process. This means that they are close to a random walk, and then lagged variables can be very weak instruments for the first difference variables. Therefore,

(16)

15 they developed a adjected estimator. System-GMM was introduced by Blundell and Bond (1998), which combines the system of differences with the regression in lagged level. This adds the level lags to the set of moment conditions. The new assumption used is that in the first different equation, the variables are not correlated with the fixed effects. We use one-step GMM that means that the parameters are based on the initial weight matrix (Roodman, 2006).

This method provides consistent estimates for n sufficiently large and t relatively small (Arellano & Bond, 1991), in our research n, the amount of countries, is larger than t, the amount of years. The regression instruments are based on controlling for unobserved effects and using previous observations and lagged variables as instruments.

There are a few additional assumptions that must be checked before the GMM estimators are acceptable to use. The key assumption for the instrument set in GMM to work is that

𝐸(Δ𝑦𝑖,𝑡−𝑗 |Δε ) = 0 for ∀j ≥ 2, so we test for second-order autocorrelation. And for first-order this

should be there, otherwise we cannot use the results of the Arellano and Bond test (Arellano & Bond, 1991). The regression will show the test value AR(1) for first-order autocorrelation, we want this p-value to be low, as the null-hypothesis is that there is no autocorrelation and AR(2) for second-order autocorrelation for which we want the p-value to be high (Roodman, 2006).

To test the overall validity of the instruments the Hansen test is widely used. This reports a p-value, what we would like this to be high, the null-hypothesis is that the instruments as a group are exogenous.

Arellano and Bond developed this test and incorporated the command xtabond for it in Stata. After that Roodman (2006) introduced the command xtabond2 to make it possible to also do system-GMM and add additional correction.

In the research we assume that the data is heteroskedastic, and compensate by using the robust version of the regression, resulting in standard errors that are consistent with heteroscedasticity. Furthermore, we decided to limit the amount of lagged used in the system of equation in the GMM model to 4. And used the option collapse to create one instrument for every variable and lag distance instate of one for every time-period, variable and lag distance. In small samples this helps to avoid the bias that can occurs when the number of instruments comes close to the number of observations (Roodman, 2006).

(17)

16

3.2 Variables and data

There is data from 2000 to 2016 for the European Union as a whole and most European countries separately. A simple ordinary squares regression on the European countries will not give significant results, because it only consists of 16 data points. Therefore, will have to do a panel regression on multiple countries in the European Union (Stock & Watson, 2015).

Of the 28 countries of the European Union Romania, Croatia and Austria are left out. On Romania and Croatia not all data is available, which can be because the countries joined the Union more recent. Austria is left out because there is no data on schooling available for the first years of the time series that is compatible to the data of the other countries, because of the difference in school system. With excluding these countries this is a balanced panel. A list of the counties included can be found in the appendix.

The data needed for this regression is available at Eurostat, the European Union Open Data Portal that gives access to open data published by EU institutions and bodies.

3.2.1 GDP growth

GDP per capita and its lagged variable used, according to Solow and Barro this will have an impact on the growth rate. There was decided to use GDP per capita because in most previous research this is the form of GDP used.

3.2.2 Resource productivity

Resource productivity is by the European Union defined as Resource productivity by GDP divided by domestic material consumption (DMC). We use the same definition in this research, also because of the articles of the European Commission described in section 2.3. The hypothesis is that a rise in resource productivity will have a positive effect on GDP growth. Therefore, a positive sign in the regression analysis is expected.

3.2.3 Other variables

The choice of the control variables has its basis in the theory described in section 2 on GDP growth. From the literature it is expected that investment and human capital to have a positive effect on GDP growth and unemployment negative.

(18)

17 In table 1 the descriptive statistics are shown. It shows the mean, standard deviation, maximum and minimum of al variables over all, but also in one country and between the different countries.

Table 1: Descriptive statistics

Variable Mean Std. Dev. Min Max Observations

lngpc overall 9.840499 0.73139 7.495542 11.42409 N = 425 between 0.703983 8.285997 11.18835 n = 25 within 0.240918 9.027207 10.38437 T = 17 rp overall 1.449737 0.963629 0.1417 4.4771 N = 425 between 0.892109 0.244406 3.220065 n = 25 within 0.403428 0.297673 3.026002 T = 17 lngfcf overall 3.077375 0.184512 2.442347 3.602777 N = 425 between 0.122656 2.806453 3.333105 n = 25 within 0.139885 2.578274 3.468813 T = 17 lngov overall 3.79053 0.143706 3.299534 4.175925 N = 425 between 0.125245 3.602243 3.994319 n = 25 within 0.074546 3.482542 4.358933 T = 17 lnedu overall 4.198534 0.270507 3.020425 4.472781 N = 425 between 0.256578 3.433645 4.434651 n = 25 within 0.099126 3.711936 4.617162 T = 17 ump overall 9.091529 4.398217 1.9 27.5 N = 425 between 3.093366 4.594118 15.88235 n = 25 within 3.183786 1.409176 21.59741 T = 17

(19)

18

4. Results and analysis

In this section the results of the Fixed effects estimator and system GMM will be discussed.

Table 2:Regression

(1)

(2)

Fixed-effects

System-gmm

VARIABLES

lngpc

L.lngpc

0.817***

0.298***

(0.0745)

(0.0999)

L2.lngpc

-0.352***

-0.0943

(0.0915)

(0.0775)

L3.lngpc

0.342***

0.446***

(0.0463)

(0.0651)

rp

0.0359**

0.159***

(0.0144)

(0.0413)

lngfcf

0.0961**

0.154*

(0.0347)

(0.0859)

lnedu

0.168***

0.00755

(0.0241)

(0.0706)

ump

-0.00878***

-0.0141***

(0.00139)

(0.00395)

Constant

0.990***

2.906***

(0.332)

(0.848)

Observations

350

350 R-squared

0.943 Number of c

25

25 Hansen J-stat

3.450 d.f.

2 p-value

0.178 AR(1) p-value

0.109 AR(2) p-value

0.576 Robust standard errors in parentheses

* p<0.01, p<0.05, * p<0.1

In the fixed effects regression, all variables are significant and have the expected sign. Investment is positive, human capital is positive and unemployment negative all like expected form the theory in section 2. Resource productivity is positive with coefficient of 0.0359 and significant with a significance level of 5%.

(20)

19 Looking at the second regression it can be seen that not all the estimates form the first regression were correct. In this regression there is corrected for the Nickell-bias. All variables still have the same sign, but not all variables are significant.

Resource productivity has a significant positive effect, the null-hypothesis can be rejected with 0.01 significance, so this value is positively different from zero.

Human capital measured by Lnedu is not significant and second lag of GDP isn’t either, which means our regression is not completely significant. This could be explained by the correlation between resource productivity and education in counties being higher than the correlation between GDP per capita and education. For the correlation table see appendix part 3.

Subsequently the assumptions to use the GMM method are checked.

Observing the p-value for the AR(1) test for first order autocorrelation in the residuals, for a significance level of 10.9 % reject the null-hypothesis of no autocorrelation. This is a high value, usually it should be below 10% or 5% to have a significant result.

Meanwhile the test for second order serial correlation shows to have no problems, the test statistic is significantly high to not be able to reject the null-hypothesis.

The Hansen test has a low degree of freedom because of the collapse option used. The p-value is significantly high, the null-hypothesis cannot be rejected, which means the instruments are valid.

(21)

20

5. Conclusion and discussion

This thesis looks at the question: What is the effect of rising resource productivity on GDP growth in

the European Union? To address this question, the theory of GDP growth of Solow was used, that was

elaborated on by Mankiw et al. (1992) and later by Barro (2003). Barro (1991) also looked at panel data like in this research. In this theory there is a steady state of growth, and there can be diverted from this state but eventually growth will go asymptoticly back to a steady state. This paper used a dynamic model is used to see if resource productivity affects this steady state positively. This is done by using the fixed effects estimator and system GMM introduced by Roodman. Looking at the effect of rising resource productivity on GDP growth is an interesting question because the European Commission claims that improving resource productivity is not only good for the environment but will also lead to growth in the European economy.

What is found, is that when looking at GDP per capita growth, resource productivity has a positive effect on growth. The control variables used where not all significant, when estimated using system-GMM. That is why we cannot say how big this effect truly is, but we can conclude this positive effect. This means that the claim that the European Commission makes in its roadmaps to a resource efficient Europe and other policies is true, in the sense that when growth in the economy is defined solely by growth in GDP per capita. And, when not looking at the other effects the policies, that aim to make a rise in resource productivity possible, have on the economy.

For further research on this question it would be good to add dummy variables to the regression to compensate for time specific effects, like the subprime financial crisis. Also, more control variables could be added and more elaborate research could be done to see what is the best dataset to use for the control variables.

(22)

21

Bibliography

Arellano, M., & Bond, B. (1991). Equations, Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment. The Review of Economic Studies, 58(2), 277-297. doi:https://doi.org/10.2307/2297968

Barro, R. J. (1991). Economic Growth in a Cross Section of Countries. The Quarterly Journal of

Economics, 106(2), 407–443.

Barro, R. J. (2003). Determinants of Economic Growth in a Panel of Countries. Annals of economics and

finance, 4, 231–274.

Domenecha, T., & Bahn-Walkowiakb, B. (2017). Transition Towards a Resource Efficient Circular Economy in Europe: Policy. Ecological Economics, In Press. doi:https://doi.org/10.1016/j.ecolecon.2017.11.001

Krausmann, F., Gingrich, S., Eisenmenger, N., Erb, K. H., Haberl, H., & Fischer-Kowalski, M. (2009). Growth in global materials use, GDP and population during the 20th century. Ecological

Economics, 68(10), 2696-2705.

Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A Contribution to the Empirics of Economic Growth.

The Quarterly Journal of Economics, 107(2), 407-437. doi:https://doi.org/10.2307/2118477

Nordhaus, D., & Tobin, J. (1972). Is Growth Obsolete? Economic Research: Retrospect and Prospect,

5(Economic Growth), 1-80. Retrieved from http://www.nber.org/chapters/c7620

Nordhaus, W. D. (1993). Reflections on the Economics of Climate Change. The Journal of Economic

Perspectives, 7(4), 11-25. Retrieved from http://www.jstor.org/stable/2138498

Roodman, D. (2006). An Introduction to "Difference" and "system" GMM in stata.

Schulte, U. G. (2013). New business models for a radical change in resource efficiency. Environmental

Innovation and Societal Transitions, 9, 43-47. doi:http://dx.doi.org/10.1016/j.eist.2013.09.006

Solow, R. M. (1956). A Contribution to the Theory of Economic Growth. The Quarterly Journal of

Economics, 70(1), 65-94.

Stern, N. (2006). What is the economics of climate change? WORLD ECONOMICS-HENLEY ON THAMES,

(23)

22 Stock, J. H., & Watson, M. W. (2015). Introduction to econometrics. Boston: pearson.

Wiedmann, T. O., Schandl, H., Lenzen, M., Moran, D., Suh, S., West, J., & Kanemoto, K. (2015). The material footprint of nations. Proceedings of the National Academy of Sciences, 112(20), 6271-6276.

(24)

23

Appendix

1. Countries included in the data panel

 Belgium  Latvia

 Bulgaria  Lithuania

 Cyprus  Luxembourg

 Czech Republic  Malta

 Denmark  Netherlands  Estonia  Poland  Finland  Portugal  France  Slovakia  Germany  Slovenia  Greece  Spain  Hungary  Sweden

 Ireland  United Kingdom

 Italy

2. Data origin

What Code Explanation Origin

GDP gdp Current prices, million

euro

Gross domestic product at market prices

Eurostat

GDP per capita pc Eurostat

Recourse productivity rp Domestic Resource

Consumption divide by GDP. Where DMC total material consumption in thousand tonnes.

Eurostat

Gross fixed Capital formation (GFCF)

gfcfcl Current prices, million euro

(25)

24 Gross fixed capital

formation Ratio Gross fixed

Capital formation (GFCF) of GDP

gfcf In percentages of GDP Eurostat

Education ratio edu Percentage of 15 to 64

years olds with upper secondary, post-secondary non-tertiary and tertiary education (levels 3-8).

Eurostat

Unemployment rate ump Percentage of active

population

Eurostat

Government expenditure

gov In percentages of GDP Eurostat

3. Correlation table

(obs=425)

lngpc rp gfcf gov edu ump

lngpc 1.0000 rp 0.7671 1.0000 gfcf -0.2896 -0.4576 1.0000 gov 0.3809 0.3146 -0.3304 1.0000 edu -0.1104 -0.1143 0.0953 -0.0554 1.0000 ump -0.4229 -0.2365 -0.2378 0.0609 0.0337 1.0000

Resource productivity and growth : what is the effect of rising resource productivity on GDP growth in the European Union?