MASTER THESIS IN THE FRAMEWORK OF THE DD INTERNATIONAL DEVELOPMENT ECONOMIES: M.A. INTERNATIONAL ECONOMICS - GEORG-AUGUST UNIVERSITÄT GÖTTINGEN AND
M.SC. INTERNATIONAL ECONOMICS AND BUSINESS – RIJKSUNIVERSITEIT GRONINGEN
Net energy exports and national wealth
Supervisor: Prof.dr.mr. C. J. Jepma (Rijksuniversiteit Groningen) Co-assessor: Dr. N. Behncke (Georg-August Universität Göttingen)
Miralda van Schot, s1887912/11403454
15-6-2015
Abstract
This paper investigates the effect of energy abundance on national wealth and more specifically whether countries richly-endowed with energy resources tend to have higher indirect energy exports. A new measure of energy abundance is applied that accounts for the re-export of energy embodied in goods and services and is established by Input-Output compilations of the WIOT for the period 2002-2009. Empirical findings indicate that energy abundance represses the level of national wealth, but enhances economic growth. Moreover, gas endowment enhances indirect export by 1,5 times, whereas endowments in oil represses indirect exports by 2/3, signifying that aggregation of energy resources into a single measure will be misleading.
Keywords: Energy abundance, direct-to-indirect energy export, WIOD; Input-Output; developed nations, economic prosperity
Acknowledgements
Table of content List of abbreviations ... 1 List of tables ... 1 List of figures ... 2 1. Introduction ... 3 2.Theory ... 6
2.1 Energy consumption and economic prosperity ... 6
2.2 Direct and indirect energy exports. ... 9
2.3 Other plausible determinants of indirect energy export ... 12
3. Methodology ... 14
3.1 The effect of energy export on GDP per capita ... 15
3.2 Determinants of indirect energy export ... 15
3.3 Accounting for endogeneity ... 16
4. Data ... 17
4.1 Data analysis of direct and indirect energy export ... 19
5. Results ... 21
5.1 Energy export contribution to GDP ... 21
5.1.1 Digging deeper: Economic development effects ... 24
5.2 Gravity analysis of indirect energy exports in 2009 ... 26
6. Discussion ... 28
6.1 A cross-country comparison: Russia versus the Netherlands ... 28
6.2 Limitations/avenues for future research ... 30
7. Conclusion ... 31
1
List of abbreviations
CBS = Centraal Bureau voor de Statistiek CPB = Centraal Plan Bureau
EEBT = embodied energy in bilateral trade GDP = Gross Domestic Product
GLS = generalized least square GVC = Global Value Chain
IEA = International Energy Agency LSG = Low Sulfur Gasoline
MRIO = multi-regional input-output NIOT = National-Input Output Tables
OECD = Organization of Economic Cooperation and Development OLS= ordinary least square
OPEC = Organization of the Petroleum Exporting Countries RIIC = Russia, India, Indonesia and China
SDA = Structural Decomposition Analysis TJ = Terajoule
UNSDESD =UN Statistics Division Energy Statistics Database WIOD = World Input-Output Database
WIOT = World Input-Output Tables WTO = World Trade Organization
List of tables
Table 1: Overview of countries included in empirical analysis
Table 2: Estimation regression direct-indirect energy ratio on GDP per capita Table 3: Interaction direct and -indirect energy on GDP per capita
Table 4: Differences in income-levels and GDP per capita. Table 5: Determinants of indirect energy export
2
List of figures
Figure 1: Apparent relationship energy exports and GDP growth Figure 2: Expected effects of energy exports on economic prosperity Figure 3: Variety in national energy intensity
Figure 4: Dutch sectorial changes 1960-2000
Figure 5: Share of natural gas exports to GDP in 2009 in TJ/US$ Figure 6: Share of crude oil exports to GDP in 2009 in TJ/US$ Figure 7: Share of hard coal exports to GDP in 2009 in TJ/US$ Figure 8: Share of brown coal exports to GDP in 2009 in TJ/US$ Figure 9: Determinants indirect energy use
Figure 10: Main sectorial indirect gas exports
3
1. Introduction
In 1959, large natural gas fields were discovered in the North of the Netherlands, which resulted in a thriving exploitation with large benefits for the Dutch government. In the recent months however, small earthquakes have startled the Northern Province, causing damage to 40,000 properties (Varlin, 2015). These earthquakes have been related to the exploitation of this natural gas, and now the call to stop exploiting natural gas has arisen from many inhabitants. For years, the Dutch government has made fortunes with the export of natural gas, making the Dutch government reluctant to halter the extraction process. This year governmental reluctance came to an end, and measures were taken to reduce gas extraction in 2015 by 7% toward an amount of 39,4 billion m³. The consequence of this decision for economic growth has been estimated by Centraal Plan Bureau (CPB, 2015) as respectively a 0.1 to 0.2 percentage-points.
Many scholars have investigated the effects of energy endowment on economic performance. Mideska (2013) and Keay (2007) found that energy endowment has a positive wealth effect for countries as Norway and Canada. Lederman and Malony (2003) also found by applying a panel data analysis that resource exports raises economic growth. However, not all scholars tend to find a positive relation between natural resource endowment and economic development. Richard Auty was the first who introduced the term ‘resource curse’, as he found a negative relation between natural resource abundance and economic growth. The main explanations for these findings include issues as “Dutch disease”, rent-seeking and improper institutions. As the literature is not able to answer the question whether energy endowment is good for national wealth, I want to investigate this relation in my thesis. Although energy endowment is most commonly measured by using the proxy energy exports, there are several limitations to this measurement (Torres, Afonso, and Soares, 2013). First, direct energy exports do not truly proxy for energy endowment, since energy endowment is a stock measure rather than a flow measure. This limitation is justified by the idea that without the extraction of the resources no effects of energy endowment on national wealth are expected to take place. However, a second and more serious limitation is that the proxy does not account for the re-exports of energy resources.
4
any exported good or service. The inclusion of indirect energy export is interesting because virtually all traded goods and services use energy as an input factor in their production. For instance, Treloar, Love and Crawford (2004) show that the production of one single Australian truck in 1998 required a total input of energy of 1.088 Terajoule (TJ). Therefore, by taking indirect energy exports into account I try to control for the re-export of energy embodied in the production of goods and services. An example of the importance of indirect energy exports is given by McKinsey Germany (2009) who estimated that in 2008 around 40% of all global revenues were attributable to sectors for which energy use is of strategic importance. The vast majority of this 40% was attributable to the construction, transport and other energy intensive sectors, whereas not more than 10% of these global revenues were generated by energy industries. For these sectors, high revenues provide a good indication of their dependence on energy resources, since these revenues are for a large extent determined by a company’s access to cheap and high quality sources of energy. Nevertheless, access to energy is not self-evident. The worldwide spread of shale oil and gas is, in contrast to general-use factors like labour and capita, extremely unequally distributed (Gerlagh and Mathys, 2011). This uneven distribution implies that having fossil energies is unique. However, not all types of fossil energy are unequal distributed. Whereas 81% of the crude oil reserves are concentrated in the Organization of the Petroleum Exporting Countries (OPEC, 2014), coal is more equally spread across nations. Although the world is looking for alternative sources of energy, it is still largely dependent on fossil fuels as inputs in their production processes. To illustrate the importance of access to low energy costs in economic development a reference can be made to Dutch tomato growers. In 2011, 32% of the total production costs for Dutch tomato growers were attributable to energy inputs (Wageningen UR, 2011), which is large in comparison to their Mediterranean competitors. However, the discovery of natural gas in the northern province of the Netherlands provided tomato growers with favourable access to low priced gas, making the Netherlands one of the largest exporters of tomatoes. Therefore, Dutch competitiveness does not only illustrate the importance of access to cheap energy resources, it also gives notion to the idea that endowment in energy resources can be used as a proxy for lower energy prices.
As I stated before, in this thesis I will investigate the effect of both direct and indirect energy
exports on national wealth, and more specifically whether countries richly-endowed with
5
has not been investigated before. Second, the inclusion of indirect energy exports contributes to the discussion on the well-known resource curse. Whereas, the resources curse has traditionally been associated by malfunctioning institution, rent-seeking and the “Dutch disease”, I propose that the absence of industrial linkages of the primary extractive energy sectors with other sectors provides a good indication of the presence of a resource curse. Figure 1 depicts the unconditional correlation between direct energy export and GDP per capita growth (left side) and between direct and indirect energy export and GDP per capita growth (right side). Although direct energy exports seem to hinder GDP per capita growth in 2002-2009, the inclusion of indirect energy exports in the measure of energy abundance changed the relationship. This change gives rise to the notion that resource endowment does not predetermine the fate of nations. In contrast, it can be argued that it is within countries capacity to adjust policies and to create inter-industrial linkages so that energy endowments work in favour of overall economic development rather than against it. In this thesis I will use a sample of 37 relative developed countries from 2002-2009 which could provide comprehensible results on the effect of indirect energy exports since they together make up about 85% of world’s GDP (Timmer, Dietzenbacher, Los, Stehrer & de Vries, 2014).
Figure 1: Apparent relationship energy exports and GDP growth
This thesis is comprised of six more sections. The next theory section discusses the general effect of energy abundance on national wealth and more specifically the effect of energy endowment on indirect energy export. Following that section, the methodologies used in this study will be explained in Section 3, and the data used for analysis will be presented in Section 4. The outcomes of the regression are outlined in Section 5, whereas more insights into the results for Russia and the Netherlands as well as the limitations and possibilities for further
AUS_1 AUT_1BEL_1 BGR_1 BRA_1CAN_1 CHN_1 CHL_1 DEU_1 DNK_1ESP_1 EST_1 FIN_1 FRA_1 GBR_1 GRC_1 HUN_1 IDN_1 IND_1 IRL_1 ITA_1 JPN_1 KOR_1 LTU_1 LUX_1 LVA_1 MEX_1 NLD_1 POL_1 PRT_1 ROU_1 RUS_1 SVK_1 SVN_1 SWE_1 TUR_1 USA_1 AUS_2 AUT_2 BEL_2 BGR_2 BRA_2 CAN_2 CHN_2 CHL_2 DEU_2 DNK_2 ESP_2 EST_2FIN_2FRA_2 GBR_2 GRC_2 HUN_2 IDN_2 IND_2 IRL_2 ITA_2 JPN_2 KOR_2 LTU_2 LUX_2 LVA_2 MEX_2 NLD_2 POL_2 PRT_2 ROU_2 RUS_2 SVK_2 SVN_2 SWE_2 TUR_2 USA_2 -5 0 5 10 Ave ra g e G D P g ro w th 2 0 0 2 -2 0 0 9
0 2.00e+07 4.00e+07 6.00e+07 8.00e+07 Average direct and indirect energy exports 2002-2009
Fitted values GDP growth 2002-2009
Relationship energy export and GDP growth
AUS_1 AUT_1BEL_1 BGR_1 BRA_1CAN_1 CHN_1 CHL_1 DEU_1 DNK_1ESP_1 EST_1 FIN_1 FRA_1 GBR_1 GRC_1 HUN_1 IDN_1 IND_1 IRL_1 ITA_1 JPN_1 KOR_1 LTU_1 LUX_1 LVA_1 MEX_1 NLD_1 POL_1 PRT_1 ROU_1 RUS_1 SVK_1 SVN_1 SWE_1 TUR_1 USA_1 AUS_2 AUT_2 BEL_2 BGR_2 BRA_2 CAN_2 CHN_2 CHL_2 DEU_2 DNK_2 ESP_2 EST_2FIN_2FRA_2 GBR_2 GRC_2 HUN_2 IDN_2 IND_2 IRL_2 ITA_2JPN_2 KOR_2 LTU_2 LUX_2 LVA_2 MEX_2 NLD_2 POL_2 PRT_2 ROU_2 RUS_2 SVK_2 SVN_2 SWE_2 TUR_2 USA_2 -5 0 5 10 Ave ra g e G D P g ro w th 2 0 0 2 -2 0 0 9 0 2000000 4000000 6000000 8000000 Average direct energy exports 2002-2009
Fitted values GDP growth
2002-2009
6
research will be discussed in Section 6. Concluding this research, Section 7 comprises an overview of the main findings.
2.Theory
Although there has been a great deal of literature on the effect of direct energy exports on economic prosperity, no consensus has been reached. Some studies find a positive effect of energy endowment on economic prosperity (Mideska, 2013; Keay, 2007); others find a negative economic performance of resource rich countries such as Ecuador, Nigeria, and Sierra Leone (Auty, 2001; Boschini, Pettersson, and Roine, 2007; Gylfason, 2001). In A survey of
literature on resource curse: critically analysis of the main explanation, empirical tests and resource proxies, Torres et al., (2013) provide an overview on the current stand of debates on
the resource curse. Whereas, these researchers summarize the possible negative effect of direct energy endowments on economic prosperity, I prefer to focus on the effect of indirect energy exports on national wealth. Figure 2 provides a schematic overview of the expected relation between direct and indirect energy exports and national wealth. I will first elaborate on the effect of indirect energy exports on national wealth, and later I will discuss the expected relation between direct energy and indirect energy export.
Figure 2: expected effects of energy exports on economic prosperity 2.1 Energy consumption and economic prosperity
In An international literature survey on energy-economic growth nexus: evidence from country
specific studies, Omri (2014) provides an indication that no consensus has been reached in the
7
energy to economic growth, 27% find that the unidirectional causal relationship runs from economic growth to energy; 27% support bi-directional causality, and 21% find no causal relation between both variables. The lack of consistency in findings on the energy-economic growth nexus is caused by the sensitivity of these analyses to varying econometric approaches and datasets. Although the authors focus on country specific studies, many studies using panel data share the same inconsistency in outcomes. For example, Narayan and Smyth (2008) found that energy consumption and real GDP of the G7 countries are co-integrated, implying that GDP corresponds to increased energy consumption and vice versa1. Moreover, their panel unit root estimation, co-integration analysis, and granger causality showed that a 1% increase in energy consumption causes a real GDP increment of 0,12% to 0,39%. In contrast, Coers and Sanders’s (2013) investigated 30 OECD countries by the mean of a vector error correction model and found that at least in the long run, the causal direction runs from economic growth to energy consumption. Although no consensus has been reached on the direction of causality, I will first elaborate more on the effect of energy consumption on national wealth. In Section 2.3 I will discuss the mechanisms through which national wealth might affect energy consumption is some more detail.
Traditional growth models emphasize on the production factors capital and labour, since these input factor account for the largest fraction of the total production cost. Although energy is not included in this model as a factor of production, it seems rational that the production of goods and services cannot take place without the input of energy. Ayres, Bergh, Lindenberger and Warr (2013) conducted an investigation on the output elasticity of petroleum in the US and found that a 50% cut in crude oil supply would reduce GDP by a factor of ten, which is greater than the 2% predicted on the basis of small cost share. One explanation for this stronger effect is that the measure of output elasticity includes the effect of crude oil cuts on other sectors (Ayres et al., 2013). Therefore, the output elasticity captures the existence of forward linkages between the extractive primary sector and other sectors. As a result, Ayres et al. (2013) argue that energy might be a much more important factor for economic growth than its relatively small share in total cost indicates. The effects of energy consumption on economic growth seem to depend on the establishment of inter-industrial linkages. To illustrate this, a short side step is made to the export of cacao beans. There are two ways to export cacao beans; export of
1
8
cacao as a raw product or embodied in the exports of e.g. chocolate. The primer, is generally associated with the exploitation of cacao farmers, since these farmers generally receive a negligible part of the overall retail price. In contrast, the export of chocolate involves higher levels of value addition and therefore contributes more to national wealth. However, whereas value creation in the cacao value chain is located downstream, 75% of the total value added in the oil and gas chain is concentrated in upstream production and exploration activities (Petrostrategies Inc., 2015). The upstream location of value adding activities signifies little incentive for vertical upgrading. Nevertheless, energy resources are still important factors for production, since neither labour nor capital can function without inputs from energy (Ayres et al., 2013) Therefore, rich-endowed countries should develop a resource based industry, which is characterized by strong linkages between manufacturing sectors and the primary extractive sectors.
The rise of global value chains (GVC) provides less developed countries with more opportunities to engage in the production of advanced goods, since firms no longer have to build complex value chains on their own2. Although economic fragmentation created opportunities to enter the GVC, the establishment of industrial linkages has become even more challenging. Morris and Fessehaie (2014) mention two specific challenges in building a resource based industry.
First, the reduction in transport cost reduce the possibilities to create forward linkages. For instance, the Brazilian metallurgical industry faced a re-location of its steel production to countries that most efficiently produce the final products (De Ferranti, Perry, Lederman and Maloney, 2002). Therefore, the proximity to fossil energies does not, in itself, confer sufficient cost advantages to establish energy intensive manufacturing sectors. For instance, Liu, Xi, Guo and Li (2010) show that China’s lack in energy endowments did not hold the country back in becoming a net exporter of embodied energy. Other factors, such as access to cheap labour, might be more deterministic for cost competiveness. In regard to this cost competitiveness, the relative immobility of labour and transportability of fossil energy might be essential.
Second, the characterization of extractive industries as capital intensive implies indirectly that these industries offer limited opportunities for employment and/or for the development of higher skill levels (Morris and Fessehaie 2014, p. 29). However, countries like Canada and
2
9
Norway show that these limitations can be overcome. Moreover, virtually all natural resource industries are to a certain degree vertically integrated. This is beneficial, since Keay (2007) mentions that well-established linkages between the primary extractive sector and manufacturing industry promotes the diversification into more capital intensive industries by transmitting domestic resource-intensive intermediate input prices. For instance, a 10% price increase in petroleum leads to a 5% price increase for refined products (Hanson, Robinson and Schluter, 1993)
Although it might be harder to establish strong linkages with the primary extractive sector, it is certainly not impossible with the right policies and conditions in place. Therefore, richly endowed nations should direct their energy resources to higher value adding activities. I argue that the value of energy for national wealth will be maximized with the creation of industrial linkages with the primary extractive sector. In this regard, directing energy toward the most value adding manufacturing industries, like heavy and chemical industries, maximizes energy’s contribution to GDP (Auty 2001). This leads to the follow hypothesis:
H1: the greater indirect energy exports in the measure of total energy abundance, the higher national wealth.
2.2 Direct and indirect energy exports
10
Although the traditional Heckscher-Ohlin theorem focuses on specialization in final commodities, a great part of the overall trade takes place in intermediate goods nowadays. The increased trade in intermediate goods can be ascribed to the process of economic fragmentation. Economic fragmentation, attributable to a drop in information and communication technology costs, signifies a separation of production processes in different tasks (Baldwin, 2006). This separation implies that the average efficiency in performing all production tasks is no longer relevant. Instead, what matters is the coordination of tasks and their allocation to the most efficient firms across nations. Therefore, specialisation occurs at a less aggregated level, as countries focus their activities on those stages of the production process in which they have a comparative advantage. As a result, the traditional Heckscher-Ohlin theorem should take a broader scope by including trade in intermediate goods. The importance of accounting for energy embodied in intermediaries was highlighted by Tolmasquim and Machado (2003). They found that the use of the direct coefficient methodology led to severe underestimation of Brazil’s embodied energy in trade of approximately 600 Petajoule3. Therefore, data on energy embodied in intermediate exports/imports are needed to assess the quantity of embodied energy export correctly. For example, the total energy embodied in the production of a car includes the direct energy use in the manufacturing of a car and all energy used indirectly to produce the components of the car (Constanza, 1980). A measure to account for indirect energy inputs is the Leontief inverse which is generated by a transformation of Input-Output tables. Input-Output tables specifically help to keep track of inter-industrial linkages; how output from one industrial sector is related to other industrial sectors and ultimately to final consumption. Several researches applied Input-Output tables, among them are Machado, Schaeffer, and Worrell (2001); and Tang, Snowden, and Höök (2013). Machado et al. (2001) find that exports and imports of embodied energy in non-energy goods comprise about 12% and 10% of the total of 6781 Petajoule of energy consumed in Brazil in 19954. However, whereas Brazil became net indirect energy exporter, the United Kingdom turned into a net consumer of embodied energy. The application of Input-Output tables by Tang et al. (2013) show that since 1997 the United Kingdom increasingly depends on indirect imports of fossil energy.
3
Tolasmaquim et al. (2003) used the energy intensity coefficient of sectors, which can be defined as the ratio between the final energy used directly in processing these goods and its respective gross output value. The use of the direct coefficient methodology is problematic, since it does account for inter-industrial linkages.
4
11
Moreover, the role of specialization in embodied trade has been analysed by Gasim (2015). His study on embodied energy in trade in 2009 on 40 countries confirmed the relevance of specialization in accounting for 50% of the indirect energy exports. In addition, the author distinguishes two methods to quantify embodied energy exports: The embodied energy in bilateral trade (EEBT) and the multi-regional Input-Output (MRIO) model. The main difference between these two models is the moment where consumption of energy is recorded. The EEBT attributes embodied energies in traded goods to the first point of consumption, whereas MRIO ascribes energy consumption to the final point of consumption. To illustrate the difference between the two models, consider the export of German car component to Slovakia, which in turn are assembled into a car for further export to the Netherlands. In the EEBT the energy embodied in German car components would be attributed to Slovakia, whereas MRIO records them as Dutch energy consumption. However, as EEBT fails to account for the re-export of embodied energy, it is less appropriate for application in an increasingly fragmented world. Therefore, in order to account for the re-exports of goods, Gasim (2015) suggests to use the MRIO method that treats the world as a single integrated production system. Although Gasim (2015) emphasized the importance of choosing the right method, he did not specifically test for the existence of the Heckscher-Ohlin theorem.
12
Figure 3: Variety in national energy intensity
(based on WIOD of Timmer et al., 2015)
To sum up, I argue that direct exporting countries have higher quantities of indirect energy export, as the relative advantage in natural energy prompts lower domestic energy prices and thereby stimulates specialization into energy intensive sectors. This stimulation of specialization signals a positive effect of direct energy exports on indirect energy exports. Nevertheless, direct energy export also implies that these resources cannot be used to establish sectorial linkages with the primary extractive sector, since exported resources are no longer available for use in the domestic industry. In this thesis I will focus on the effect of direct energy exports on indirect energy exports which results in the following hypothesis:
H2: Countries richly endowed with natural energy resources have higher quantities of embodied energy in bilateral trade
2.3 Other plausible determinants of indirect energy export
Although indirect energy exports seem to be positively affected by country’s endowment in energy resources, a number of other variables might play a role. Three of these variables include: governmental policies, the level of national wealth and country size.
13
fuel, oil, or coal use was technically inferior5. For example, the in 1966 established aluminium factory Aldel was highly depended on the access to low-priced gas and its failure to negotiate favourable energy deals resulted in application for bankruptcy by 2013. Moreover, Mulder and de Groot (2013) find that governmental policies facilitated the development of a relatively energy intensive agricultural sector, which is mainly driven by the horticultural subsectors and greenhouses. Figure 4 provides an overview of the sectorial changes in the Dutch economy from 1960 and onwards.
Figure 4:Dutch sectorial changes 1960-2000
(based on Centraal Bureau voor de Statistiek (CBS) 2007)
The development of energy intensive sectors, such as the chemical industry, after the discovery of natural gas is notable. Moreover, the relatively large clothing, shoe, leather and non-energy sectors industries diminished over time, partly due to an appreciation of the guilder (Correljé et al. 2003). The loss in competitiveness of the tradable goods sector became later well-known as the “Dutch disease”.
Second, Judson, Schmalensee and Stoker (1999) link indirect energy exports to the level of national wealth. They found that energy consumption follows an inverted U-shape pattern, implying that additional income first increases industrial energy demand, but will decrease after a certain income threshold is reached. This phenomenon is caused by significant declines in the income elasticity of energy demand with increasing income (Judson, et al., 1999). As a result, a substantial part of future energy demand will originate in the new industrialized economies, like China. China’s outstanding economic growth has come along with a voracious appetite for energy consumption, from 465 kg of oil equivalent at the beginning of the 1970s to 767 kg in the early 1990s, and has recently overtaken the world’s average per capita consumption of over
5 This includes chemical, metallurgical, and ceramic industries
14
2000 kg of oil equivalent (World Bank, 2015a). In addition, economic development amplifies social concerns on environmental issues such as greenhouse gases and the depletion of fossil energies. In an attempt to quantify the effect of income on renewable energy consumption, Omri and Nguyen (2014) conducted a study of 64 countries over the period 1990-2011 and found that a 1% increase in per capita GDP of high-income countries raises renewable energy consumption by 20%. In the same study, they do not find a statistical significant effect for renewable energy consumption from low-income countries.
Third, the argument that country size affects indirect energy exports is the last channel I would like to discuss. On the one hand, one might expect that larger countries have higher trade in embodied energies since their absolute value of trade is higher. On the other hand, trade is of greater significance for smaller nations. Dietzenbacher and van der Linden (1997) conducted an inter-country Input-Output analysis for seven countries of the European Committee and found that smaller countries, like the Netherlands, are two to three times more dependent on inter-industrial trade with foreign nations than larger countries6. This means that large resource-abundant countries seem to be more biased toward autarky since a greater domestic market for sales makes openness to foreign market less of a necessity (Auty and Kiiski cited in Auty, 2001)
To conclude, indirect energy is expected to be affected by governmental policies, national wealth and population size. Although the effects of these variables are considered, I mainly focus on the effect of energy endowments on indirect energy exports.
3. Methodology
To investigate whether direct and indirect energy exports enhance economic development, and more specifically whether countries richly-endowed with energy resources tend to have higher indirect energy exports, I will use two different frameworks. First, a panel data model is applied, covering the period of 2002 to 2009 and 37 countries, to empirically analyse the relationship between direct-to-indirect energy export and GDP per capita. Second, more insight into the determinants of indirect energy export in bilateral trade is gained though Tinbergen’s (1962) gravity model. Both frameworks are executed multiple times for different types of energy. This means that I use disaggregated energy data rather than aggregated data on energy
6
15
consumption. The advantage of this is that it allows me to control for certain aspects of economic utility such as: the scarcity of distribution, demand for cleaner types of energy, and the function energy might fulfil within the society.
3.1 The effect of energy export on GDP per capita
The effect of energy abundance, proxied by direct and indirect net energy exports, on a country’s level of GDP per capita will be estimated by a panel data model. This model builds on the subsequent equation:
+ (1)
where describes GDP per capita. , , and signify the quality of institutions, population size, and gross capital formation. depicts the variable of interest: the ratio of direct-to-indirect energy exports. The rationale behind this equation is that the smaller the ratio of direct-to-indirect energy exports the higher the GDP per capita. This smaller ratio is acquired by relatively large indirect energy exports in comparison to direct energy exports. In addition, I will run a second regression to account for the interaction between direct and indirect energy export. The equation for this model is:
(2)
where is a dummy variable that takes a value of one for direct exporters of natural energy resources and zero otherwise. In addition, the dummy captures a country’s position on indirect energy exports. The rationale behind this interaction effect is that indirect energy exports will be higher given country’s abundance in energy. Hence, indirect energy export is modified by direct energy export. Consequently, country features like institutional quality, population size, and gross capital formation are included, to deal with effects on GDP per capita arising from country specific characteristics.
3.2 Determinants of indirect energy export
16
- (3)
where - represents the energy embodied in the net export of goods and services from country to country for energy type . describe GDP per capita of both countries, captures the distance, and signify population size, and represents the
national energy intensity of country for energy type . and depict the variables of interest: the position of both countries in the direct export of energy. The justification of this equation is that county’s well-endowed with fossil energy resources, proxied by their direct energy export, will have higher quantities of embodied energy in their exports. Moreover, the traditional variables stay relevant. Thus, higher GDP per capita and greater distances predict higher and lower embodied energy exports, respectively. Nevertheless, the gravity model has some important shortcomings. Gómez-Herrera (2013) argues that the validity of the log-linearization in the presence of heteroskedasticity leads to biased and inefficient ordinary least squares (OLS) estimation. Therefore, different test and measures need to be taken to address the presence of heteroskedasticity.
3.3 Accounting for endogeneity
The issue of endogeneity is caused by the existence of unobserverables that influences the independent variable and thereby the trade pattern of embodied energy. Endogeneity can manifest itself in several forms (Wooldridge, 2010).
17
A second issue of endogeneity is reverse causality, which implies the mutual dependence of embodied energy and GDP per capita. As discussed in the theory, no consensus has been reached in the energy-economic growth nexus so far. Therefore, I use a one period lagged variable for GDP per capita in equation 3 to reduce the likelihood of reversed causality, arguing that indirect energy export in 2009 does not affect GDP in 2008. A second problem of reverse causality problem is the interdependence between indirect energy export and sectorial structure, making the interpretation of causation ambiguous. This causality implies that energy abundant countries have more energy intensive sectors, which might be caused by fiscal policies, such as energy subsidies. I expect that the inclusion of a lagged variable for national energy intensity encompasses, among others, the effect of policies on the indirect net energy exports. This expectation is based on the fact that direct energy export is separately included in the model. Lastly, the issue of omitted variables has not been specifically addressed. Although presence of omitted variables is hard to check for, the fixed effect model reduces at least all problems with time-invariant omitted variables. Moreover, although I take certain steps to reduce and addressed the impact of endogeneity on the outcomes, it is hard to completely rule out the existence of it.
4. Data
18
constant over time (Gerlagh et al., 2011). For the second analysis, energy embodied in bilateral trade is given for 2009 so that effects from time-invariant characteristics are excluded.
Table 1:Overview of countries included in the empirical analysis
European Union Latin America Asia and
Pacific
Austria (AUS) Greece (GRC) Poland (POL) Brazil (BRA) China (CHN) Belgium (BEL) Hungary (HUN) Portugal (PRT) Mexico (MEX) India (IND)
Bulgaria (BGR) Ireland (IRL) Romania (ROU) Japan (JPN)
Czech Republic (CZE)
Italy (ITA Slovak Republic (SVK)
North America South Korea
(KOR)
Denmark (DNK) Latvia (LVA) Slovenia (SVL) Canada (CAN) Australia (AUS) Estonia (EST) Lithuania (LTU) Spain (ESP) United States
(USA)
Turkey (TUR)
Finland (FIN) Luxembourg (LUX) Sweden (SWE) Indonesia (IDN)
France (FRA) Netherlands (NLD) United Kingdom (GBR)
Russia (RUS) Germany (DEU)
WIOD is not the only organization that offers information on Input-Output structures. The most well-know are Trade in Value Added (a joint project of the OECD and WTO), the Institute of Development Economies of the Japan External Trade Organization, and Global Trade Analysis project. The first two are characterized by lacking consecutive time-series, which will be particular problematic for this thesis. In addition, the WIOD is preferred above the others as it comprises a more selected set of developed countries, which together account for 85% of the world’s GDP (Timmer et al., 2014). The focus on these relatively well-developed nations might bias the results since these countries already established meaningful manufacturing sectors in which energy resources are consumed. Therefore, external validity is low, suggesting that results do not hold for developing nations.
19
The World Bank’s World Development Indicators (2014) provides information on several national accounts for the period 2002 to 2009. Here I obtained data real GDP per capita, growth in GDP per capita, institutional quality, population size, and gross capital formation. Nevertheless, I decided to include level terms of GDP per capita, rather than growth of GDP per capita as depicted in the unconditional correlation of Figure 1. The main reason for this is the minimal predictive power of the models on GDP per capita growth. Although the models on growth in GDP per capita have little predictive power, they are included to check for the robustness of the outcomes. Further on, the World Development Indicators do not hold data on the level and growth of GDP per capita for Taiwan. Therefore, Taiwan is excluded from the data sample. Further on, data on national energy intensity is compiled from aggregating the sectorial energy intensity from WIOD compilations and distance measures are taken from Mayer and Zignago (2011). A detailed overview and description of all included variables is given in Appendix B.
4.1 Data analysis of direct and indirect energy export
Before moving ahead, several tests are carried out to examine the data and enable decision making for appropriate model testing. An overview of these tests is given in appendix C and includes issues as: correlation, heteroskedasticity, autocorrelation and outliers.
Appendix D describes the direct and indirect energy exports of 37 countries in 2009. From this data, confirmation of the scarce distribution of natural gas in Australia, Canada, Denmark, Indonesia, the Netherlands, and Russia and crude oil in Brazil Canada, Mexico, and Russia is found in Figures D.1 and D.2. Although these observations differ substantially from the other observations, they are not excluded. Instead, these variables carry valuable information that helps to identify the effect of energy exports on national wealth. Nevertheless, Figures D.1 to D.4 indicate large differences in the quantity of direct energy exported across energy types. Remarkably, directly traded natural gas and crude oil are relatively large in comparison to coal resources, which might be attributable to a more equal distribution of coal resources across countries. Further on, countries that are well-endowed with natural resources seem to be indirect exporters of gas or oil.
20
Figure 5: Share of natural gas exports contribution to GDP in 2009 in TJ/US$
(based on: WIOD of Timmer et al., 2015 and UNSDESD 2014)
Figure 6: Share of crude oil exports contribution to GDP in 2009 in TJ/US$
(based on: WIOD of Timmer et al., 2015 and UNSDESD 2014)
Remarkably, indirect energy export in relation to GDP per capita is clustered toward the RIIC countries (Russia, India, Indonesia, China), which might be ascribed to the relative importance of manufacturing sectors in their overall economy. In the same manner, more industrialized nations slowly shift their production structure away from energy intensive industries toward more value-added, but less energy intensive service sectors (Warr et al, 2010). Moreover, Figure 5 and 6 give rise to the notion that a lack of oil fields, as in China and India, does not immediately imply that countries become net indirect energy importers. In contrast, it might suggest that cross-country difference in labour cost outweigh the transportation cost of energy, resulting in a shift of energy and labour-intensive activities toward low-wage countries.
Figure 7: Share of hard coal exports contribution to GDP in 2009 in TJ/US$
(based on: WIOD of Timmer et al., 2015 and UNSDESD 2014)
-500 0 500 1000 A US A UT B E L B GR B R A C A N C HN CZE DE U DNK ESP E ST FIN FR A GB R GR C
HUN IDN IND IRL IT
A JP N KOR LT U L UX L VA MEX NE D P OL P R T R OU RUS SVK SVN SW E T UR USA
direct exports indirect export
-6000 -1000 4000 A US A UT B E L B GR B R A C A N C HN CZE DE U DNK E SP E ST FIN FRA GB R GR C HUN ID N IN D IR L IT A JP N KOR LT U L UX L VA MEX NE D P OL PRT R OU RUS SVK SVN SW E T UR USA
direct export indirect export
-2000 3000 8000 13000 A US A UT B E L B GR B R A C A N C HN CZE DE U DNK E SP E ST FIN FRA GB R GR C HUN ID N IN D IR L IT A JP N KOR LT U L UX L VA MEX NE D P OL P R T R OU RUS SVK SVN SW E T UR USA
21
Figure 8: Share of brown coal exports contribution to GDP in 2009 in TJ/US$
(based on: WIOD of Timmer et al., 2015 and UNSDESD 2014)
Nevertheless, Appendix D and Figures 5 through 8 are not controlled for demand effects. This control is necessary since large cross-country variety is expected to come from discrepancies in demand. A structural decomposition of Dutch and Russian indirect gas exports into intensity, technology, and demand effects will provide more insight into the size of demand effects. The outcomes are discussed in more detail in Section 6.
5. Results
5.1 Energy export contribution to GDP
In a first step of the empirical analysis, the effect of direct-to-indirect energy export on GDP per capita is examined. Preference is given to the fixed effect model with clustered standard errors, as suggested by Hoechle (2007)7. The use of clustered standard errors implies that resulting standard errors are completely robust to any kind of serial correlation and/or heteroskedasticity8.
Table 2 and Table 3 display the results of Equations 1 and 2, respectively, with independent energy variables (1) natural gas export, (2) crude oil export, (3) hard coal export, (4) brown coal exports, and (5) all four energy types.
First, the control variables institutional quality, population size and gross capital formation are significant at a 1% or 5% level, with the expected signs. The ratio direct-to-indirect energy exports are positive and insignificant for all types of energy. In the case that a country is well-endowed with fossil energies the positive effect runs against my first hypothesis, since an
7
The Hausman test indicates that the random effect model is affected by an incorrect specification. Fixed effects are intercepts that capture all behavioral differences between individuals, referred to as individual heterogeneity (Hill, Griffiths, and Lim 2012). Although the R² of 23% is low, the model is jointly significant and the rule of thumb is not violated.
8
Wald and Wooldridge test results confirm the presence of heterogeneity and autocorrelation. Results are given in appendix C. -10 90 190 A US A UT B E L B GR B R A CA N CHN C Z E DE U DNK E SP E ST FIN FR A GB R GR C
HUN IDN IND IR
L IT A JP N KOR LT U L UX L VA MEX NE D P OL P R T ROU RUS SVK SVN SW E T UR US A
22
increase in indirect energy exports represses the ratio of direct-to-indirect energy exports, and thus lowers the level of GDP per capita. Therefore, additional regressions are performed to find a possible explanation for this unexpected effect.
Table 2: Estimation regression direct-indirect energy ratio on GDP per capita
Log GDP per capita - Fixed effects – cluster
(1) (2) (3) (4) (5)
Natural gas Crude oil Hard coal Brown coal All
Institutional quality 1,291*** 1,291*** 1,279*** 1,291*** 1,273***
(0,198) (0,197) (0,198) (0,198) (0,200)
Log population size 4,166*** 4,181*** 4,164*** 4,159*** 4,154***
(0,793) (0,783) (0,782) (0,786) (0,798)
Gross Capital Formation 0,00656** 0,00662** 0,00642** 0,00664** 0,00645**
(0,00286) (0,00284) (0,00284) (0,00285) (0,00289)
Direct/indirect natural gas ratio
2,38e-05 2,57e-05
(8,32e-05) (8,38e-05)
Direct/indirect crude oil ratio
0,00424 0,00444
(0,00620) (0,00625)
Direct/indirect hard coal ratio
0,000117 0,000120
(0,000134) (0,000137)
Direct/indirect brown coal ratio 1,14e-05 -1,20e-05 (0,000247) (0,000251) Constant -62,26*** -62,55*** -62,26*** -62,20*** -62,06*** (13,46) (13,29) (13,28) (13,34) (13,55) Observations 295 296 296 295 294 overall R² 0,228 0,229 0,230 0,227 0,231
Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
23
The results of Table 3 suggest that the repressive effect of simultaneous direct and indirect energy export is a markedly more negatively correlated than indirect energy on its own. Intuitively this effect makes only sense if the proxy direct energy exports implies that these exported cannot be used in the domestic industry. If this is the case, than embodied gas exports add more value to the domestic product, whereas the economic added value in direct energy export is minimal. This is in confirmed by (Keay 2007), who states that deep-rooted linkages of the primary extractive sector with other sectors promote diversification into more capital intensive industries, increasing the demand for both service sector output and non-resource-intensive manufactured products, and thereby stimulating national wealth. To quantify this, being an indirect energy exporter without simultaneous being a direct energy exporter, ceteris paribus, decreases GDP per capita on average by 21%; 18%; and 23%, whereas being simultaneous an indirect and direct exporter decreases GDP per capita on average by 36%; 27%; and 49%9. Further on, the inconsistency in the sign of interaction effect of (1-0) across the model is remarkable.
Table 3: Interaction of direct and indirect energy on GDP per capita
log GDP per Capita - Fixed effects – cluster
(1) (2) (3) (4)
Natural gas Crude oil Hard coal Brown coal
Institutional quality 1,169*** 1,253*** 1,186*** 1,256*** (0,193) (0,194) (0,189) (0,191) Log population size 3,135*** 3,966*** 3,373*** 3,597*** (0,790) (0,772) (0,756) (0,760) Gross Capital Formation 0,00746*** 0,00506* 0,00474* 0,00638** (0,00276) (0,00280) (0,00272) (0,00272) D.direct X D.indirect energy 0-1 -0,237*** -0,196*** -0,260*** -0,265*** (0,0581) (0,0642) (0,0552) (0,0702) 1-0 -0,174 0,126 -0,382* 0,173 (0,262) (0,158) (0,231) (0,148) 1-1 -0,445*** -0,310*** -0,683*** -0,220** (0,143) (0,104) (0,183) (0,0968) Constant -44,50*** -58,66*** -48,44*** -52,38*** (13,42) (13,11) (12,85) (12,90) Observations 296 296 296 296 R-squared 0,286 0,270 0,314 0,299
Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
The interaction effect (1-0) is negative related to GDP per capita in the first and third columns, but positive in the second and fourth column. One explanation for the positive sign for crude oil
9 I use the general formula (( -1)*100%) to reform dummy variables into log forms. Thus 23% is calculated as
24
might be the discrepancy in domestic and world oil prices. However, Section 5.2 provides more insight in the interaction between direct and indirect energy exports, and especially on the negative sign of the interaction term (1-0).
Third, most of the coefficient of the interaction terms of direct- and indirect energy export are significant, while insignificant in Table 2. This inconsistency in significance might be explained by the use of binary variables rather than level variables in Equation 2, signifying that the intensity of direct and indirect energy exports is less important than the incidence of exports.
Finally, enhancing effects of institutional quality, population size, and gross capital formation on GDP per capita are confirmed by all tables, with significance levels of 1% through 10%. Although these variables have the expected sign, they are only included as a mechanism of control and are not further explained. Further on, the robustness checks for equation 1 and 2 are given in Table E.1 and E.2. Table E.1 indicate that the findings are robust to other measurement techniques. Table E.2 shows the regression results of the interaction effect on the dependent variable GDP growth per capita. For a discussion of these results I refer to the Appendix. 5.1.1 Digging deeper: Economic development effects
In order to find possible explanations for the drivers of the negative effect of direct- and indirect energy export on GDP, the sample is divided into average and highly developed countries. More specifically, the data is split into medium-income and high-income countries based on a $22.500 threshold10. The results, without interaction effects, are depicted in Table 4. First, the results provide an indication that the negative effects are driven by medium-income countries. For medium-income countries, it is suggested that indirect energy export lowers GDP per capita by 38%; 30%; 32%; and 37% on average. In contrast, this effect is much smaller for high-income countries and even insignificant in the event of natural gas. Intuitively, the negative effect of indirect energy export is somewhat strange. For instance, China’s remarkable economic growth has often been attributed to its transition into the light-manufacturing sector, which went compared with the increase in energy consumption. Therefore, an additional regression is run on GDP per capita growth as dependent variable. The results of this regression for natural gas and oil are depicted in Table E.3 of Appendix E. The outcomes of column 1, 2 and 4 of this table show that indeed indirect energy exports
25
significantly enhances GDP per capita growth, whereas direct energy exports represses economic growth. For more details on this regression I refer to Appendix E.
Table 4: Differences in income-levels and GDP per capita11
Log GDP per capita - fixed effects - cluster
(1) Natural gas (2) Crude oil (3) Hard coal (4) Brown coal
Medium income High income Medium income High income Medium income High income Medium income High-income Institutional quality 1,663*** 0,104 1,932*** 0,0240 1,738*** 0,0522 1,742*** 0,0668 (0,294) (0,193) (0,295) (0,194) (0,285) (0,182) (0,286) (0,185) Log size population 2,622* 5,744*** 2,093 6,132*** 1,615 5,643*** 1,985 5,777*** (1,429) (0,660) (1,394) (0,575) (1,350) (0,563) (1,357) (0,571) Gross Capital Formation 0,00658 0,00420** 0,00563 0,00352* 0,00593 0,00329* 0,00882* 0,00358* (0,00514) (0,00194) (0,00528) (0,00186) (0,00495) (0,00177) (0,00488) (0,00182) D.direct energy -0,247 -0,148 -0,134 0,0487 -0,368* -0,140*** 0,109 -0,00851 (0,199) (0,111) (0,111) (0,0823) (0,209) (0,0328) (0,116) (0,0501) D.indirect energy -0,479*** -0,0500 -0,363*** -0,0984** -0,389*** Omitted -0,470*** -0,139*** (0,160) (0,0374) (0,125) (0,0417) (0,0987) (0,121) (0,0374) Constant -36,40 -85,68*** -27,47 -92,02*** -19,17 -83,86*** 1,742*** -86,12*** (24,53) (10,98) (23,95) (9,544) (23,21) (9,348) (0,286) (9,494) Observations 152 144 152 144 152 144 152 144 R-squared 0,316 0,545 0,311 0,553 0,354 0,593 0,344 0,580
Standard errors in parentheses;*** p<0.01, ** p<0.05, * p<0.1
One explanation for finding a higher repressive effect for developing nations in Table 5 might be related to the sample group of this thesis. This sample group is focused toward relatively well-developed nations, which concentrate their activities in high value added and energy-extensive sectors, such as the service sector. This is confirmed by Warr et al., (2010). In contrast, relatively energy intensive activities take place in medium-income countries, which are characterized with lower levels of GDP per capita. Other explanations might be that energy resources are directed to sectors that add relatively little to GDP per capita or that high-income nations have better technologies in place that reduces the quantity of energy embodied in trade. Secondly, although the repressive effect of direct energy export is insignificant in most columns, direct hard coal exports are highly significant for high-income countries. Whereas, this finding is somewhat remarkable, I do not have a clear explanation for it.
11 Note: Medium-income countries include Australia (2002), Bulgaria, Brazil, China, Czech Republic (except
2008), Spain (2002 and 2003), Estonia, Greece (until 2005), Hungary, Indonesia, India, Italy (2002), Korea (except 2007), Lithuania, Latvia, Mexico, Poland, Portugal (until 2006), Romania, Russia, Slovak Republic,
26
Finally, institutional quality, which is proxied by rule of law, is markedly more positively correlated to GDP per capita for medium-income countries. Therefore, institutional quality seems to become less relevant after a certain quality threshold has been reached. In the same manner, the enhancing effect of population size exhibits a larger impact on high-income nations, suggesting that in the sample of high-income nations larger countries have higher wealth levels. Although enhancing effects of gross capital formation are found for both countries, the effects are slightly greater and less significant for medium-income countries. 5.2 Gravity analysis of indirect energy exports in 2009
In this section, the determinants of embodied energy in exports will be examined. The Breusch-Pagan and Goldfeld-Quandt tests rejected the existence of homoskedasticity. Therefore, the robust variance-covariance estimator is used to provide an unbiased and consistent estimation of the Equation 3. Table 5 shows the outcomes of the robust-OLS regression with indirect energy export variables (1) natural gas export, (2) crude oil export, (3) hard coal export, (4) and brown coal export.
Table 5: Determinants of indirect energy export
Log indirect energy export - robust-OLS
(1) (2) (3) (4)
Natural gas Crude oil Hard coal Brown coal
log GDP per capita i 0,821*** 0,505*** 0,535*** 0,301*** (0,140) (0,0907) (0,0712) (0,105) log GDP per capita j 0,674*** 1,062*** 0,531*** 0,796*** (0,0846) (0,0922) (0,0853) (0,0720) log Distance ij -0,979*** -0,922*** -1,006*** -1,010*** (0,0638) (0,0787) (0,0707) (0,0667) log Population i 0,907*** 0,901*** 1,118*** 0,666*** (0,0487) (0,0651) (0,0456) (0,0529) log Population j 0,785*** 0,661*** 0,915*** 0,779*** (0,0384) (0,0543) (0,0468) (0,0446) log National energy intensity i -0,465 0,00562 0,129*** 0,0152 (0,581) (0,0114) (0,0129) (0,0147) D.direct energy i 0,943*** -0,972*** -0,0230 0,559*** (0,156) (0,179) (0,161) (0,160) D.direct energy j 0,468*** 0,236 0,351* 0,342** (0,174) (0,197) (0,202) (0,149) Constant -28,52*** -27,95*** -28,47*** -23,18*** (2,078) (2,064) (1,806) (2,207) Observations 615 639 658 659 R-squared 0,589 0,485 0,645 0,494
Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
27
thus for energy. Moreover, the enhancing effect of population size signifies that larger countries have higher absolute values of energy embodied in trade. Further on, with the exception of hard coal, national energy intensity relation to the quantity of indirect energy embodied is insignificant. This insignificant effect might be attributed to the inclusion of direct energy exports, which indicates that other factors affecting national energy intensity are less relevant in comparison to energy endowment.
Secondly, when considering direct energy endowment of the importing country , the term is positive and significant in all columns, except for crude oil. Moreover, indirect gas and brown coal exports are clearly significantly enhanced by direct energy abundance of the exporting country . This means that indirect gas and brown coal exports of well-endowed countries are respectively 1,5 and 0,7 times higher than those lacking resources. In contrast, the Heckscher-Ohlin prediction that countries specialize in their abundant factor endowment does not hold for crude oil and hard coal. The differences in the effect of direct energy export status on indirect energy exports prompts a discussion on the interpretation of direct energy export. This thesis used direct exporting status as a proxy of energy endowment and thereby followed the predictions of the classic Heckscher-Ohlin theorem that relative advantage in energy endowment prompts lower domestic energy prices and stimulates specialization into energy intensive sectors. Nevertheless, export of direct energy also implies that these resources cannot be used to establish sectorial linkages with the primary extractive sector, signalling a negative effect on indirect energy export. Hence, the latter seems the case for crude oil and hard coal.
28
To check for the robustness of the outcomes, the robust-OLS are compared with the weighted - General Least Squares (GLS) estimation. For a comparison of these results I refer to Table E.4 of Appendix E.
6. Discussion
6.1 A cross-country comparison: Russia versus the Netherlands
This discussion draws attention to the empirical effects of Section 5.2 on natural gas consumption in Russia and the Netherlands. Although both countries are well-endowed with gas resources, they differ significantly in size and prosperity. For this reason, they form an interesting basis for comparison. Figure 9 depicts the importance of population size, GDP per capita and direct energy exports in the discrepancy in Russian and Dutch indirect gas exports12. The Dutch embodied energy in exports is negatively affected with regard to population and energy endowment, whereas the higher prosperity level enhances Dutch gas embodied in exports.
Figure 9: Determinants indirect energy use
The importance of direct energy exports is confirmed by the structural decomposition analysis (SDA), discussed in Appendix F13. The results from this analysis show that 67% of the difference in energy consumption is explained by intensity, 31% by technology, and just 2% by the effect of demand. This is a surprising finding since a large share of demand was expected due to significant differences in absolute GDP per capita in 2009: $20.867,55 versus $45.512,12 (World Bank, 2015b). However, more insight in the drivers of this demand effect
12 This figure is based on level variables of national gas intensity and direct natural gas exports.
13 These SDA outcomes are quite similar to those found by Gasim’s (2015) MRIO for 41 economies indicating a
29
(foreign or domestic) indicates indeed a positive effect of domestic demand in favour of the Netherlands. In contrast, foreign demand for indirect energy exports is greater in Russia, which confirms that larger countries have higher absolute values of energy embodied in exports. While gas endowment is an important determinant for embodied gas export, it should be attributed to most value adding activities in order to maximize the economic rents. Taking relative contributions into account, Figure 10 features great variance in energy embodied in sectorial exports14. An overview of the covered sectors is depicted in Appendix G. Nonetheless, great variety in embodied energy exist in (1) Agriculture, Hunting, Forestry and Fishing; (2) Mining and Quarrying; (4) Textiles and Textile Products, (8) Coke, Refined Petroleum and Nuclear Fuel (9) Chemicals and Chemical Products, and (15) Transport Equipment. This variety in allocation of gas resources suggest that both countries specialize in different industries. Insight into the competitive advantage generated by the export of indirect gas is depicted in Figure 11.
Figure 10: Main sectorial indirect gas exports
(based on WIOD of Timmer et al., 2015)
Figure 11: Main sectorial indirect gas per value added exports
(based on WIOD of Timmer et al., 2015)
With regard to the sectorial value added, especially Russia seems to direct its energy resources to the wrong sectors. For instance, Russia’s relatively large shares of gas exports in (8) Coke,
14
Relative contributions are determined as the share of a sector’s indirect energy exports in the total quantity of indirect energy exported
30
Refined Petroleum and Nuclear Fuel provides minimal economic rents in terms of value added. In order to maximize its economic rents, gas should be directed through inter-industrial linkages toward (1) Agriculture, Hunting, Forestry and Fishing; (3) Food, Beverages and Tobacco; (9) Chemicals and Chemical Products and (13) Machinery, Nec.
6.2 Limitations/avenues for future research
Certain limitations are concerned with the estimation of this thesis’ outcomes. One of the limitations is related to the measure of indirect energy exports, which is established on the basis of strong simplifying assumption. For instance, production technologies are assumed to be homogeneous within industries signifying constancy of production processes across products within a firm or across firms (Timmer et al., 2014). This implies that exporting and non-exporting firms are assumed to have the same constant to scale technology. Nevertheless, the Melitz model (2008) suggests that firms participating in international trade differ significantly in terms of productivity levels from non-exporting firms.
Even more severe are potential problems regarding endogeneity. The thesis used lagged variables of GDP per capita and national energy intensity in Equation 3 to assess arising problems from simultaneity. The presence of lagged income and national energy intensity indicate a dynamic nature, suggesting interdependence of these variables across periods. However, this relative constancy of energy intensity overtime (Gerlagh et al., 2011) weakens the strength of lagged variables as instruments (Blundell and Bond, 1998). Moreover, although I took certain steps to reduce and addressed the impact of endogeneity on the outcomes, it is hard to completely rule out the existence of it. Therefore, an avenue for future research would be the use of General Methods of Moments (GMM) estimations, as this model deals with the existence of endogeneity.
31
Further on, although this thesis used national energy intensity in combination with energy endowment to capture the possible effect of governmental policies, concerns about the omission of policies cannot be completely ruled out. A better indication of the effect of policies would make the estimates of energy abundance more reliable.
7. Conclusion
In this thesis I sought to address the impact of energy abundance on national wealth as well as the influence of energy endowment on embodied energy in bilateral trade. For this purpose, energy abundance is redefined as a combination of direct and indirect energy exports. Thereby, relatively new comprehensive data on indirect energy exports is generated by the alignment of WIOT with several types of fossil resources of the energy account. In this regard, I controlled for the re-exports of goods and services by using WIOT, which treats the world as a single integrated production system. Using a fixed effect model with clustered standard errors estimation, no significant effects are found of the direct-to-indirect energy export ratio on national wealth and neither on economic growth. Therefore, I reject the H1 that greater indirect energy exports in the measure of total energy abundance leads to a higher level of national wealth. Nevertheless, simultaneous direct and indirect energy export represses national wealth further. This might be explained by the fact that direct exported resources cannot be used for specialization into energy intensive and value adding activities.
Drawing attention to the possible driver of this effect, the split in the data at a threshold of $22.500 to generate two income groups is empirically investigated. This investigation is conducted regarding the existence of a reverse U-shaped pattern of energy consumption. The outcomes showed that the negative effect of indirect gas export is driven by medium-income countries, which might be explained by a low levels of national wealth and relative energy intensive export sectors. Moreover, looking at the effect of direct- and indirect energy export on economic growth, it turns out that indirect gas exports enhances economic growth but that direct gas export represses growth.
32
