The welfare effect in the European natural gas market as a result of the introduction of the TTIP agreement

1

Student: T. Burggraaf

Student number: 6035094

Supervisor: Mr. D. Smerdon

University: UvA: University of Amsterdam

Date: 29-06-2016 Bsc Economics

9582 words

The welfare effect in the European natural gas

market as a result of the introduction of the TTIP

agreement

Abstract

An important part of the current negotiations between the EU and the US on the TTIP treaty is the abandonment of the total export restraint on natural gas from the US towards the EU. This thesis provides a method of estimating demand and supply of natural gas in both markets and accordingly a method of calculating the welfare effects of the abandonment of the export restraint trade. This is a contribution to international trade theory, especially to the theory of trade liberalization. The outcome of the research is that welfare generated in the European Union rises with 78.3% when the trade restriction is abandoned. This result is highly debatable though, and should not be taken as the truth. The main contribution of this thesis lies in the method.

Keywords: TTIP, three stage least squares regression, Krugman model, international trade, trade liberalization, free trade, trade policy, natural gas

2

Statement of Originality

This document is written by Tijmen Burggraaf who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is 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

C

ONTENTS

1 Introduction ... 4

2 Literature review ... 6

2.1 History of international trade and free trade ... 6

2.2 International trade restrictions ... 7

2.3 Background on the natural gas market in the EU ... 7

2.4 Background on the natural gas market in the US ... 8

2.5 Theoretical model by Krugman ... 8

2.6 Hypothesis ... 13 3 Research method ... 14 3.1 Regression model ... 14 3.2 Welfare effect ... 18 4 Data ... 20 5 Results ... 24

5.1 Results of the demand and supply estimation ... 24

5.2 Current amount of welfare generated in the EU natural gas market ... 32

5.3 Determination of the world price ... 32

5.4 Determination of the welfare change due to international trade ... 33

5.5 Conclusion ... 34

6 Discussion ... 35

4

1 I

NTRODUCTION

As of July 2013, the European Union (EU) and the United States (US) have been negotiating on an international trade treaty called the Transatlantic Trade and Investment Partnership (TTIP. The goal is to reach agreement on a treaty that should eliminate all trade restrictions (tariffs) between the EU and the USA. It is expected that removing trade restrictions positively affects the welfare of both countries, as literature shows that free trade has positive consequences for all trade participants. (Loewy & David, 1998, pp. 146-148) (Achele, 2014) have shown that the real economies of the EU and the US could grow respectively 2.68% and 2.12% and the EU itself even claims that the total economic growth could be € 275 billion immediately, equalizing the GDP of countries like Austria or Denmark (European Commission, 2016). The preliminary TTIP treaty has several pillars and the most relevant are (Europa Nu, 2016):

 The abolishment or diminishing of several existing tariffs between the EU and the USA.  The abolishment of several non-tariff obstructions to trade, mostly trade law.

 The adjustment of rules and standards concerning trade and investment. The European Commission has proposed to establish a ‘regulatory cooperation body’, which has the task to create a system in which standards and rules from both economies are accepted by the other market, called the Investor-State Dispute Settlement (ISDS).

 The removal of one of the largest trade barriers: the voluntary export restraint that does not allow American natural gas to be exported to the EU.

This thesis focuses on the last pillar. The welfare effect of the removal of the export restraint for the European natural gas market is evaluated using the economic concepts of consumer and producer surplus. The corresponding research question is:

What effect does the introduction of the Transatlantic Trade and Investment Partnership have on welfare generated in the European natural gas market?

This study contributes to the existing international trade policy literature, as the results show how the demand and supply function in the European natural gas market are established and how the removal of the trade barrier with the US affects the welfare generated in the market. The demand and supply functions for the EU and US natural gas market are estimated using the three stage least squares (3sls)-regression method. The welfare effects of the removal of the trade barrier are interpreted accordingly in the Krugman model, making use of the estimated results of the 3sls-regressions. This thesis contributes

5 to the literature, as this model has not been used for this purpose (influence of TTIP on the welfare) before.

The results of this thesis are in line with the literature. As already shortly mentioned above the literature shows that removing trade barriers could increase market welfare, by lowering the trade price and increasing consumer surplus to such a level that it exceeds the decline in producer surplus. The findings of this thesis indicate that the introduction of TTIP increases the welfare in the market with 78.3%, thereby confirming the theoretical expectations.

The structure of the remainder of this thesis is as follows. Chapter 2 provides an overview of the relevant literature, including a detailed explanation of the Krugman model. The research method is explained in chapter 3, including a detailed explanation of the 3sls-regression method. Chapter 4 consists of the descriptive statistics of the dataset used. Chapter 5 discusses the demand and supply function (outcomes of the 3sls-regressions) and the impact of TTIP on the welfare (according to the Krugman model). Finally, the results will be discussed in Chapter 6.

6

2 L

ITERATURE REVIEW

This chapter provides an overview of the relevant theory. The chapter begins with a history of international trade and free trade in paragraph 2.1, followed by a background on international trade restrictions in paragraph 2.2. Paragraph 2.3 and 2.4 are background information on the EU and US natural gas market. The first four paragraphs are meant to give more background to the Krugman model, which is the model at the core of this thesis and which is explained in detail in paragraph 2.5.

2.1 HISTORY OF INTERNATIONAL TRADE AND FREE TRADE

In contemporary history the trend in trade policy is a trend of liberalization. Since the start of the so called ‘Washington consensus’ (the economic system based on cooperation, globalization and liberalization, guarded by the IMF, the World Bank and the World Trade Organization) (International Monetary Fund, 2016) after World War II, several negotiation rounds have taken place with the aim of liberalizing international trade. In 1947, 23 countries opened up multilateral trade negotiations which lead to the General Agreement on Tariffs and Trade (GATT). The main objective of the negotiation rounds and the GATT was trade liberalization by signing several multilateral agreements. Global trade liberalization was greatly brought forward in the Uruguay Round (1986-1984), where agreement was reached on the removal of a large number of trade barriers and tariffs between the 123 participating countries, widely ranging from barriers in the market for toothbrushes to the market of AIDS treatments. As a result of the Uruguay Round, the average import tariffs in the advanced countries fell almost 40%, from 6.3% to 3.9%. (World Trade Organization, 2016).

As a result of negotiations in the Uruguay Round, the World Trade Organization (WTO) emerged as the institutional successor of the GATT. The WTO is in fact the institutionalization of the liberal idea that free trade is beneficial for the economy. The WTO’s main function is to ensure that trade flows as smoothly, predictably and freely as possible. (World Trade Organization, 2016). Following the successful Uruguay Round, the WTO had the ambition to achieve complete free trade all over the world. In order to reach that goal, a new negotiation round was set up: the Doha Round. One of the major disputes in the Doha Round was the liberalization of the agricultural markets, which is the market with the highest degree of protection worldwide. (Krugman, Obstfeld, & Melitz, 2012)(Anderson & Martin, 2005)

7 Concluding, since World War II the trend regarding international trade has been towards more and more free trade. It is confirmed by empirical research and backed by theory that free trade leads to real economic growth, increasing welfare, as an effect of lower prices and an on average positive, diminishing marginal utility for commodities traded. (Loewy & David, 1998, pp. 146-148) The TTIP negotiations can be seen in this light as well. Despite the assumed welfare increase that occurs from liberalizing trade, there may be reasons to restrict international trade, for example because it is politically `desirable or because it is deemed necessary by government to protect the domestic economy. Next paragraph continues on possible reasons for international trade restrictions.

2.2 INTERNATIONAL TRADE RESTRICTIONS

Countries have used several forms of international trade restrictions over the years. International trade restrictions are mostly implemented by government in order to protect the domestic market. Examples of trade policy measures that countries can use are voluntary export restraints, import tariffs and import quotas. The American export restriction on natural gas (which is in effect at the start of the research in this thesis) to the European Union is an example of a voluntary export restraint by the USA. For the EU, this means that imports are restricted by a pre specified number and thus for the EU the restriction can be seen as an import quota. (Krugman, Obstfeld, & Melitz, 2012)

International trade restrictions can be imposed by countries for various reasons. The most common reason for countries to impose trade restrictions is to protect the national economy for factors that might harm the domestic industry: international trade may lower domestic prices and profits as a result of too much cheap alternatives abroad. Another reason a country might protect its own economy is an anti-inflationary policy: too much exports may cause inflation, which in turn harms a nation’s exports and economic growth. (Magee, 1972)

2.3 BACKGROUND ON THE NATURAL GAS MARKET IN THE EU

The European gas market is characterized by two factors: firstly, European production of natural gas in some parts of the EU cannot freely be transported to all parts of the EU. The reason for this problem lies in the fact that most of the natural gas is still transported by pipeline, limiting the possibilities of trade. A solution to this problem may be the production of liquefied natural gas, which can be transported in barrels. The second factor is political dependency: the European natural gas market is highly dependent on trends in the country the EU imports from. Market uncertainties are being created by the relationship

8 between the EU and Russia, which is unstable as an effect of the dispute over Ukraine. Other trends that may influence the European natural gas market are that the Algerian production is expected to have reached its limit and Russia is expected to export more and more towards eastern Asia, instead of the EU. (Westphal, 2014)

2.4 BACKGROUND ON THE NATURAL GAS MARKET IN THE US

Like the EU, the United States are a net importer of natural gas. Most of the imports and exports go through pipelines. The main trade partners of the USA are Canada and Mexico. Data coming from the American Energy Information Administration on the export volumes of natural gas coming out of the USA shows that export flows have grown significantly since the beginning of the millennium and import flows have declined, leading to declining net imports of natural gas.

The reason for the declining net imports of natural gas in the United States lies in increasing domestic dry natural gas production. Gas exports from the USA are highly regulated. Only countries with which the USA holds free trade agreements like the NAFTA are allowed to import gas from the United States. Exports to the NAFTA countries Mexico and Canada by pipeline accounts for 98.9% of total US exports in 2014. By this protection, and the rising production of gas, domestic prices can stay very low. The domestic prices are reported to be four times as low as in Europe, and six times as low as in Asia. Experts claim that the low prices still cause producers to produce less than they actually could. Most of the rise in natural gas production comes from rising production of shale gas. The production of shale gas is not very sustainable and thus prone to debate between environmental lobby groups and production actors. Because exports are mainly in the form of dry gas nowadays, which can be transported through pipelines, the gas has to be liquefied in order to be able to export them in barrels, which is necessary to export gas overseas. (American Energy Information Adminsitration, 2016)

2.5 THEORETICAL MODEL BY KRUGMAN

The background on international trade history outlined in paragraph 2.1 and 2.2 gives some background behind the driving forces and counterforces determining international trade. The model at the core of this thesis is the Krugman model. In this paragraph the Krugman model and its implications to welfare will be explained in detail.

9 The Krugman model is a simplified classical model laid out by Krugman et al. (2012), with an epistemological position that is purely positivist. The aim of the Krugman model is to calculate welfare effects when changes occur in trade policy. The welfare effects can be examined by standard instruments of international trade policy analysis. The model examines international trade between two actors: the domestic actor and the foreign actor. Sometimes, the foreign actor is regarded as the rest of the world. In our case, the domestic actor is the EU and the foreign actor is the United States. The model concerns a simplified world using several assumptions. All assumptions are debatable, and are indeed debated thoroughly in trade policy economics, but these discussions go beyond the scope of this thesis. (Krugman, Obstfeld, & Melitz, 2012) The assumptions are:

 The domestic and the foreign market are markets in perfect competition, where supply and demand are a function of price.

 There are no transportation costs of the commodity traded.

 There are no influences of exchange rates. Prices are given in the same (domestic) currency.  Welfare is measured in terms of utility. The model assumes that the direct gains and losses

measured in terms of utility capture all changes in social benefits and that these utility measure is valued identically by all actors.

Excess supply & demand

The Krugman model starts with exogenously given demand and supply curves for both the home and the foreign actor. How supply and demand are determined is not a feature of the model. We start with determining the equilibrium prices of the commodity in both the home and the foreign market, without allowing for international trade. In both markets the equilibrium price is where demand equals supply. If the price in a country is above the equilibrium price, supply is higher than demand at that price level and there is excess supply in the market at that price level. If the price level is below the equilibrium price, there is excess demand in the market at that price level. Figure 3.1 shows how excess demand (IM) and excess supply (ES) follows from the domestic demand (D) and supply (S) curves. The level of excess supply (which is only relevant for positive quantities) a country offers to the world market at a certain price level equals the difference between the quantity supplied and the quantity demanded at that price level in the domestic market. The same logic holds for the determination of excess demand.

10 Figure 3.1: the depiction of the determination of excess supply and excess demand

World equilibrium

An assumption of the Krugman model is that the home market is cleared at a higher market price than the foreign market when there is no international trade. This means that domestic consumers are willing to pay more for the commodity than foreign consumers. Allowing for international trade would than mean that home consumers are willing to demand more of the commodity at a price that is lower than the clearing price in the home market, but higher than the clearing price in the foreign market. Vice versa, foreign suppliers are willing to supply more at a price that is higher than the clearing price in the foreign market, but lower than the clearing price in the home market. This leads to an excess demand of the commodity in the home market and an excess supply in the foreign market. The model assumes that under perfect market conditions, the excess supply coming from foreign flows to home, where excess demand exists. (Krugman, Obstfeld, & Melitz, 2012, pp. 222-225.) In figure 3.2, the determination of the world market equilibrium is shown graphically. The equilibrium price clearing the world market of excess demand (Import demand: ID) and excess supply (Export supply: ES) is called the world price and will lay between the domestic and the foreign market equilibrium price.

11 Figure 3.2: the depiction of the determination on the world market equilibrium

Welfare changes as a result of a change in trade policy

Figure 3.2 shows that for the home country, where the domestic equilibrium price is higher than the foreign equilibrium price, the price of the commodity will be lowered toward the world price after allowing for international trade. In order to calculate the welfare effect of this decline in price, welfare generated in the market before and after the removal of the trade barrier has to be calculated. This can be done by calculating what happens in the market in terms of consumer surplus, producer surplus. (Krugman, Obstfeld, & Melitz, 2012, pp. 228-232) Consumer and producer surplus are measures of the amount consumers and producers gain from trade. Consumer surplus is the difference between the price consumers are willing to pay for a good and the amount they actually pay. Producer surplus is the difference between the price under which the producer is willing to produce and the actual price the commodity is traded for. The sum of consumer surplus and producer surplus is the total amount of welfare generated in the market (Krugman, Obstfeld, & Melitz, 2012, pp. 228-232). The consumer and producer surplus are shown graphically in figure 3.3.

q q

q

p

p p

12 This way of calculating the amount of welfare in a market is valid at all price levels, so at both the initial equilibrium price and the world price after the removal of the trade barrier. The difference between the amounts of welfare in both situations is the change in the total amount of welfare in the market. If total welfare in the market declines, the country loses and if total welfare increases, the country benefits from trade policy. (Krugman, Obstfeld, & Melitz, 2012, pp. 228-232)

Figure 3.3: the depiction of welfare calculation

Key assumptions, equations and parameters

As said, in order to answer the research question, the welfare change in the home market has to be calculated. In order to do so, first of all, the equations for demand and supply have to be written as a function of price. In the model, in both markets demand and supply are a linear function of price. The unit of price is $/MMBTU, a measure for energetic value. The unit of q is million cubic feet of natural gas. (2.1) 𝑞𝑑(𝑝) = 𝑎 + 𝑏𝑝

(2.2) 𝑞𝑠(𝑝) = 𝑐 + 𝑑𝑝

Formula 3.1 shows the linear demand function, where demand is negatively dependent on price. a is positive constant, showing at which price the demand is 0 and b is a negative coefficient showing how much more natural gas consumers are willing to buy when the price falls with 1 unit. So, the demand function is downward sloping. Formula 2.2 shows the linear supply function, where supply is positively dependent on price. c is a constant showing at which price suppliers are willing to produce when the price is zero and the coefficient d is a positive coefficient showing how much more producers are willing to supply when the price increases with 1 unit. So, the supply function is upward sloping. The market is in equilibrium where 𝑞𝑑(𝑝) = 𝑞𝑠(𝑝).

13 The demand and supply functions are necessary to calculate the consumer and producer surplus. The formulas for consumer surplus and producer surplus are shown in formula 2.3 and 2.4. The sum of consumer and producer surplus equals the total amount of welfare generated in the market. Since prices are measured in $/MMBTU and quantities are measured in trillion cubic feet per year, the unit for welfare (which holds for consumer surplus, producer surplus and total welfare) is yearly welfare of x trillion $/MMBTU. This value becomes more useful when a change in welfare occurs and the percentage change in the total welfare generated can be calculated. (Krugman, Obstfeld, & Melitz, 2012, pp. 228-232)

(2.3) 𝐶𝑆 =1

2(𝑎 − 𝑝) ∗ 𝑞𝑑 (2.4) 𝑃𝑆 =1

2(𝑝 − 𝑐) ∗ 𝑞𝑠

2.6 HYPOTHESIS

My hypothesis is that the removal of the trade barrier in the natural gas market between the EU and the US as an effect of TTIP leads to a larger increase in consumer surplus than the decrease in producer surplus. So, I expect the total welfare change in the European natural gas market to be positive.

14

3 R

ESEARCH METHOD

In order to estimate the net welfare effect in the EU following the removal of the voluntary export restraint by the USA on the natural gas market, several steps have to be taken. First of all, the demand and supply curves of the European and the US natural gas market have to be estimated. The method to estimate these is explained in section 3.1. Consequently, the EU net welfare that emerges from trade in the natural gas market before implementation of the TTIP has to be calculated by means of consumer and producer surplus. Next, holding the demand and supply curves constant, the excess demand and supply in the world market (consisting of only the EU and the US market) have to be calculated, followed by the world price under which the world market is cleared. The net welfare change in the EU as a result of a the removal of the trade barrier can be calculated by means of calculating the consumer and producer surplus generated under the world price. The method regarding these steps is outlined in section 3.2. All estimations of the demand and supply curves in this thesis are performed using the software package Stata 13.

It has to be pointed out that I will make 2 extra simplifying assumptions in this thesis.

 I assume that my estimates of the demand and supply functions of European natural gas market are representative for the demand and supply functions as if there is no trade at all with foreign countries. This is not realistic, (the main countries the EU imports from are Russia (2013: 39%), Norway (2013: 29.5%) and Algeria (2013: 12.8%)) ( (Eurostat, 2016), which does influence the supply function) but I make this assumption for simplifying reasons.

 Furthermore, I assume that the European natural gas market and the US natural gas market make up for the whole world market of natural gas. Put differently, the ‘world market’ I describe in this thesis is a closed economy which consists of only the EU and the US.

3.1 REGRESSION MODEL

The method used to estimate the demand and supply function of the natural gas markets follows the approach used by Lin in her 2011 article ‘Estimating Supply and Demand in the World Oil Market’, published in the Journal of Energy and Development. In this thesis, the demand and supply functions of the natural gas market are estimated using three regression methods: an ordinary least squares regression (OLS), a 2 stage least squares regression (2SLS) and a 3 stage least squares regression (3SLS). In this section it is explained how the three regression methods work theoretically and which variables are included in

15 each model. In chapter 4 the data used on the variables of the models are laid out. The results of the regression in chapter 5.

The first step in building the model is assuming that the market of natural gas is a perfect competition market in which both consumers and suppliers are price-takers. It is assumed that the market clears, meaning that the demanded quantity 𝑞𝑑 equals supplied quantity 𝑞𝑠, as is stated in formula 3.1.

(3.1) 𝑞_𝑑 (𝑝; 𝑥) = 𝑞𝑠 (𝑝; 𝑥) = 𝑞

The demand and supply functions are assumed to be linear with fixed coefficients and additive residuals. The structural form of the model is as follows.

Demand function: 𝑞𝑑 (𝑝; 𝑥) = 𝛽𝑝𝑑𝑝 + 𝛽𝑥𝑑𝑥′+ 𝜀𝑑 Supply function: 𝑞𝑠 (𝑝; 𝑥) = 𝛽𝑝𝑠𝑝 + 𝛽𝑥𝑠𝑥′+ 𝜀𝑠

Assumed that the market clears (function 1) the structural form can be reduced to: (3.2) Demand function: 𝑞 (𝑝; 𝑥) = 𝛽𝑝𝑑𝑝 + 𝛽𝑥𝑑𝑥′ + 𝜀𝑑

(3.3) Supply function: 𝑞 (𝑝; 𝑥) = 𝛽𝑝𝑠𝑝 + 𝛽𝑥𝑠𝑥′ + 𝜀𝑠

In the structural form of our model, in both the demand and the supply function, the endogenous variable 𝑞 is dependent of price 𝑝, some covariates 𝑥′ and an error term 𝜀. The β’s are the fixed coefficients that determine the slope of the curve. Covariates are included in the functions as exogenous control variables that shift the curve.

The model assumes that the vector of covariates 𝑥′ can be decomposed in three components: exogenous covariates that shift the demand curve, but not the supply curve, exogenous covariates that shift the supply curve, but not the demand curve and exogenous market control covariates that affect both supply and demand. Calling these categories of covariates respectively 𝑥𝑑, 𝑥𝑠, and 𝑥𝑐 and substituting them for 𝑥′ _{in functions 3.2 and 3.3 leads to the following structural demand and supply functions. In the demand} function, the covariate 𝑥𝑠 is left out and in the supply function the covariate 𝑥𝑑 is left out, since the correlation between these covariates and the dependent variable is assumed to be 0.

(3.4) Demand function: 𝑞 (𝑝; 𝑥) = 𝛽𝑝𝑑𝑝 + 𝛽𝑥,𝑑𝑑 𝑥𝑑′ + 𝛽𝑥,𝑐𝑑 𝑥𝑐′ + 𝜀𝑑 (3.5) Supply function: 𝑞 (𝑝; 𝑥) = 𝛽𝑝𝑠𝑝 + 𝛽𝑥,𝑠𝑠 𝑥𝑠′ + 𝛽𝑥,𝑐𝑠 𝑥𝑐′ + 𝜀𝑠

16 Since demand functions are predicted to be downward-sloping in economic theory and supply functions are predicted to be upward-sloping, 𝛽𝑝𝑑 is expected to be ≤ 0 and 𝛽𝑝𝑠 to be ≥ 0.

In the abovementioned model of natural gas, the following covariates are considered to influence demand exogenously, but not supply:

 GDP: a higher GDP generated in the economy is assumed to be a pushing factor to natural gas demand.

 Population: more people living in the geographic area means more potential consumers of electricity and gas, which is assumed to be a pushing factor to natural gas demand.

 Energy use: the more energy is used in, the more gas demands are expected to rise.

 Electricity production: the more electricity produced, the more gas demands are expected to rise. As exogenous variables that shift supply, but not demand, the following covariates are considered to be relevant:

 The world natural gas reserves

 The amount of rigs used globally in the production of gas

An exogenous variable that shifts both the supply and the demand curve is also introduced: a dummy variable indicating whether it is summer (1) or winter (0).

Substituting all these covariates into function (4) and (5) leads to the following demand and supply functions.

(3.6) Demand function: 𝑞𝑑 (𝑝; 𝑥) = 𝛽𝑝𝑝 + 𝛽𝐺𝐷𝑃𝐺𝐷𝑃 + 𝛽𝑃𝑜𝑝𝑃𝑜𝑝 + 𝛽𝐸𝑛𝑒𝑟𝐸𝑛𝑒𝑟 + 𝛽𝐸𝑙𝑒𝑐𝐸𝑙𝑒𝑐 +

𝛽𝑆𝑢𝑚𝑚𝑒𝑟𝑆𝑢𝑚𝑚𝑒𝑟 + 𝛽𝑊𝑖𝑛𝑡𝑒𝑟𝑊𝑖𝑛𝑡𝑒𝑟 + 𝜀𝑑

(3.7) Supply function: 𝑞𝑠 (𝑝; 𝑥) = 𝛽𝑝𝑝 + 𝛽𝑅𝑒𝑠𝑅𝑒𝑠 + 𝛽𝑤_𝑟𝑖𝑔𝑤_𝑟𝑖𝑔 + 𝛽𝑆𝑢𝑚𝑚𝑒𝑟𝑆𝑢𝑚𝑚𝑒𝑟 +

𝛽𝑊𝑖𝑛𝑡𝑒𝑟𝑊𝑖𝑛𝑡𝑒𝑟 + 𝜀𝑠

𝑞 = production of natural gas 𝑝 = price of natural gas

𝑊𝑖𝑛𝑡𝑒𝑟 = a dummy variable indicating whether it is winter (December, February and March)

𝐺𝐷𝑃 = real GDP in Europe

𝑃𝑜𝑝 = the amount of European inhabitants

𝑤_𝑟𝑖𝑔 = the amount of rigs worldwide used to extract natural gas 𝑔𝑑𝑝𝑟 = Russian real GDP

𝐸𝑛𝑒𝑟 = the amount of energy used 𝜀𝑑 = the error term of the demand estimation

𝐸𝑙𝑒𝑐 = the amount of electricity produced 𝜀𝑠 = the error term of the supply estimation

𝑆𝑢𝑚𝑚𝑒𝑟 = a dummy variable indicating whether it is summer (June, July and August)

17 The models specified in formulas 3.6 and 3.7 can be estimated by the three regression methods mentioned before. All methods hold several assumptions. The simplifying assumptions of each method and the problems and opportunities each method poses are outlined on the next page. (Lin, 2011)

General simplifying assumptions

Both the OLS, the 2SLS and the 3SLS method assume linearity of the regression model. As already mentioned, it is assumed that the markets clear. The natural gas market from which the demand and supply functions are estimated is assumed to be static and perfectly competitive. Lastly, the variable price is assumed to be endogenous in the model. All other explanatory variables are assumed to be exogenous: they are given and assumed not to influence the other explanatory variables in the model (except price). (Lin, 2011)

OLS-regression

The structural β’s of both the demand and the supply function (6) and (7) can be estimated using the OLS-method by regressing the explanatory variables on the dependent variable quantity. Thus, the variable price is considered as if it were exogenous. This creates several problems. Making use of OLS, the estimates will be neither efficient nor consistent. The problem of non-consistent estimates is an effect of prices being endogenously determined in the supply-and-demand system. This leads to a so called identification problem, meaning the true coefficient of prices can’t be identified due to other factors troubling the real value of the coefficient. (Lin, 2011)

2SLS-regression

The identification problem can be solved through the creation of instrument variables for price. Including instruments for price will lead to consistent estimates. The assumption made in conducting the 2SLS-regression is the exclusion restriction, which is that the exogenous variables affecting demand do not affect supply and vice versa. In building up the model, this is already done in forming functions (4) and (5). So, the exogenous variables that affect demand are assumed not to influence supply directly, but do affect price, which in turn does affect supply. So, the exogenous variables affecting demand affect supply indirectly through price. Inversely, the same is true for the exogenous variables affecting supply. Because this is the case, the exogenous variables that are not included in the functions, can function as instruments of price. So, for the natural gas model, in the demand function, the world gas reserves and the world rig count are used as instruments for price in the demand function and GDP, population, energy use and

18 electricity production are used as instruments of price in the supply function. The exogenous variables affecting both supply and demand can be used as instruments in both equations. After regressing the instruments on price, the instruments can be included in the regular formulas 3.6 and 3.7 and the estimators can be estimated by OLS-regression. This method is called a regression. The 2SLS-regression gives better estimates that an OLS-2SLS-regression, since the identification problem is solved and thus the 2SLS-regression estimates are expected to be consistent. Even though the estimates in using a 2SLS-regression are expected to be consistent, they are not efficient. (Lin, 2011)

3SLS-regression

The efficiency problem means that, given that we use the exclusion assumption in the 2SLS-regression, the instruments that we use affect price in both equations. So, by estimating the equations separately, the 2SLS-regression does not make use of all information available. The efficiency problem that arises in both the OLS-regression and the 2SLS-regression can be solved by estimating the demand and the supply function simultaneously. So, in order to reach both consistent and efficient estimates, the 2SLS-model has to be extended by estimating the functions simultaneously. The regression method that estimates the coefficients through these three stages is called the 3SLS-method. The 3SLS-method will yield both efficient and consistent estimates because both the identification problem and the efficiency problem are solved in this method. (Lin, 2011)

3.2 WELFARE EFFECT

The 3SLS-regression method estimates for the demand and supply of natural gas in both the EU and the US market are estimates for respectively function 2.1 and function 2.2. First of all, these estimates allow us to determine the equilibrium price in the EU natural gas market at the point where 𝑞𝑑(𝑝) = 𝑞𝑠(𝑝). Secondly, these estimates can be used to calculate the total amount of welfare generated in the EU natural gas market before the removal of the trade barrier by extrapolating the coefficients of the estimates into functions 3.3 and 3.4.

The third step is to determine the excess demand and excess supply that occurs when the trade barrier is removed and the EU and the US form one market. The excess demand function for a country can be calculated by taking the difference between demand and supply for each price level. The difference between demand and supply at price = 0, equals the constant a (or c) in formulas 3.8 and 3.9. The

19 difference between the value of excess demand (or supply) between p = 0 and p = 1 is equal to the coefficient of price (b and d).

(3.8) 𝐸𝑥𝑐𝑒𝑠𝑠 𝑑𝑒𝑚𝑎𝑛𝑑 (𝑝) = 𝑎 + 𝑏𝑝 (3.9) 𝐸𝑥𝑐𝑒𝑠𝑠 𝑠𝑢𝑝𝑝𝑙𝑦 (𝑝) = 𝑐 + 𝑑𝑝

Where formula 3.8 equals formula 3.9, the world market is in equilibrium. The price level where this occurs is the world price.

The fourth step is to calculate the consumer and producer surplus generated in the EU natural gas market again, using the method explained at the end of paragraph 2.5, but this time by replacing p with the new world price. The change in total welfare generated can be calculated by subtracting the old amount of welfare generated at the domestic equilibrium price and the amount of welfare generated at the world price.

20

4 D

ATA

In the empirical estimation of the demand and supply functions of the European natural gas market, a monthly data set, comprising of data from January 1990 until December 2013 is used. The summary of the descriptive statistics used in the estimation of the EU natural gas market is shown in table 4.1. Some variables include annual monthly data. For others, only annual data was available. For these variables, the annual data are converted into monthly data, by ‘smoothing’ them in a linear way. The data are obtained from several sources. For all variables it is aimed to use datasets including all 28 countries that are currently member of the European Union (EU-28). These countries are Austria, Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and the United Kingdom. Unfortunately, for some variables data were not available. In some cases data on the 27 European Union countries (EU-27) are used, which is the 28 European Union countries except the most recent member of the EU: Croatia. It is assumed throughout this thesis that the in- or exclusion of the 28th _{country Croatia does not make a significant difference.}

For the dependent variable of both the demand and the supply function q annual data on the gross natural gas production in the EU-27 are used. The data are obtained from the US Energy Information Administration (EIA). The data are measured in trillion cubic feet per year. Since the data are annual, they are transformed into monthly data by ‘smoothing’ them in a linear way. For example: given annual data points are a production of 7.72 trillion cubic feet per year for 1990 and 8.43 trillion cubic feet per year for 1991. In this case, 7.72 is taken as data point for January 1990. The data point for February 1990 is calculated by taking the difference between the years, divided by twelve, and adding the difference to the foregoing month: (8.43-7.72)/12+7.72 = 7.78. March 1990 is measured as (8.43-7.72)/12+7.78 = 7.94 and so on.

For the endogenous variable p monthly data on the price of natural gas in the EU coming from the World Bank is used. As a measure for this price, the average import border price, including the UK is taken. The data are said to include ‘Europe’. Which countries exactly are included and which are excluded is unknown. For the data between June 2000 and March 2010 the UK is excluded. The data are corrected for inflation to the price level of 1990 using Consumer Price Index data for the EU obtained from the Organization for Economic Co-operation and Development (OECD). The data on price are measured in $ / MMBTU, which is a measure for energetic value.

21 For the exogenous variables affecting only demand the following data set is used. For GDP data on annual real GDP of all European countries that are a member of the OECD are used. This measure includes a slightly higher GDP-level than for the EU-28 countries, but it is assumed not to make a significant difference. The data on GDP are measured in millions of dollars and are obtained from the OECD. Like the data on quantity, the data on GDP are ‘smoothed’ in a linear way. For population, annual data on the EU-28 population coming from Eurostat are used. The data are smoothed in a linear way in a similar way as is done with quantity and are measured in millions. For the variable energy use, annual data on the gross inland consumption of energy in the EU-28 countries are used. The data are measured in thousand tonnes of oil equivalent coming from Eurostat and are smoothed in a linear way, as is the case with quantity. For the electricity production variable elec data on the EU-28 countries production of electrical energy in thousand tonnes of oil equivalent coming from Eurostat is used. These data are smoothed in a linear way as well.

For the exogenous variables affecting only supply the following data set is used. For the world gas reserves annual data coming from the EIA are used. These are measured in million cubic feet and smoothed in a linear way, as is the case with quantity. For the world rig count a monthly dataset obtained from Baker Hughes Inc.is used. The data show the amount of rigs that were exploited in that month to extract natural gas from its resources.

The exogenous variables that affect both supply and demand are the dummy variables summer and winter. These variables are created in Stata and hold a 1 for the relevant months (June, July and August for summer and December, January and February for winter) and a 0 for the remaining months.

22

Table 4.1: a summary of the variables used in the model for estimation of the EU natural gas market

𝑞 = production of natural gas in the EU 𝑝 = price of natural gas in the EU

𝑊𝑖𝑛𝑡𝑒𝑟 = a dummy variable indicating whether it is winter (December, February and March)

𝐺𝐷𝑃 = real GDP in Europe

𝑃𝑜𝑝 = the amount of European inhabitants

𝑤_𝑟𝑖𝑔 = the amount of rigs worldwide used to extract natural gas 𝑔𝑑𝑝𝑟 = Russian real GDP

𝐸𝑛𝑒𝑟 = the amount of energy used 𝜀𝑑 = the error term of the demand estimation

𝐸𝑙𝑒𝑐 = the amount of electricity produced 𝜀𝑠 = the error term of the supply estimation

𝑆𝑢𝑚𝑚𝑒𝑟 = a dummy variable indicating whether it is summer (June, July and August)

The data for the estimation of the US market are established in the same way as the data on the EU market. A difference is that, instead of taking the GDP of Russia, for the US market the GDP of Canada is included because Canada is the prime gas supplier to the US. Another difference is that for the US estimation the variable on electricity production is excluded, since there were nog sufficient data available. The descriptive statistics for the data used in the US market are shown in table 4.2.

gdpr 288 8.07e+11 6.03e+11 1.96e+11 2.08e+12 w_rig 288 2356.368 669.673 1156 3900 res 288 5594.152 808.1947 3987.533 6961.906 summer 288 .25 .4337664 0 1 winter 288 .25 .4337664 0 1 elec 288 18333.26 1776.806 15513 20531 ener 288 144067.9 5318.551 134291 153330 pop 288 490.9049 9.733765 475.161 506.921 gdp 288 1.23e+07 3627342 6972163 1.85e+07 p 288 3.023045 1.498407 1.101633 7.801009 q 288 8.327283 1.108773 6.070154 9.890699 Variable Obs Mean Std. Dev. Min Max

23

Table 4.2: a summary of the variables used in the model for estimation of the US natural gas market

𝑞 = production of natural gas in the US 𝑝 = price of natural gas in the US

𝑊𝑖𝑛𝑡𝑒𝑟 = a dummy variable indicating whether it is winter (December, February and March)

𝐺𝐷𝑃 = real GDP in the US

𝑃𝑜𝑝 = the amount of American inhabitants

𝑤_𝑟𝑖𝑔 = the amount of rigs worldwide used to extract natural gas 𝑔𝑑𝑝𝑐 = Canadian real GDP

𝐸𝑛𝑒𝑟 = the amount of energy used 𝜀𝑑 = the error term of the demand estimation

𝑆𝑢𝑚𝑚𝑒𝑟 = a dummy variable indicating whether it is summer (June, July and August)

𝜀𝑠 = the error term of the supply estimation

gdpc 301 1036901 334593.3 558170.4 1600354 w_rig 312 2402.641 689.7721 1156 3900 res 300 5649.287 836.758 3987.533 6972.518 summer 312 .25 .4337083 0 1 winter 312 .25 .4337083 0 1 ener 301 7422.553 1014.68 0 8056.864 pop 300 287.6523 20.61621 249.6228 318.857 gdp 301 1.16e+07 3595240 5979589 1.79e+07 p 301 4.202658 2.819028 1.26 12.88 q 312 20.98397 2.503322 17 28 Variable Obs Mean Std. Dev. Min Max

24

5 R

ESULTS

This chapter presents the results of the research, as it is explained in chapter 3. The chapter starts with the outcomes of the estimates of the demand and supply functions of both the EU and the US natural gas market in paragraph 5.1. In paragraph 5.2, the total amount of welfare generated in the EU market, followed by the determination of the world price when allowing for international trade in paragraph 5.3. Paragraph 5.4 in turn calculates the welfare change in the EU natural gas market due to the removal of the trade restriction. The results chapter is concluded in paragraph 5.5 with a summary of the results.

5.1 RESULTS OF THE DEMAND AND SUPPLY ESTIMATION

The estimation of demand and supply in both the US and the EU market is conducted by an OLS-regression, a 2SLS-regression and a 3SLS-regression. Since the outcomes of the 3SLS-regression are most relevant, both scientifically and for the rest of this thesis, these outcomes are explained in more detail than the outcomes of the OLS-regression and the 2SLS-regression.

EU market

Estimating the demand function of the EU natural gas market using the OLS-regression gives the outcomes shown in table 5.1. The estimated coefficient of p is negative (α < 0,005), as one would expect, with a value of 0.201. The estimator is significant on the 0,005 level. Furthermore it shows that the only control variable that has a significant effect on the quantity demanded is the energy use in the EU. All other variables turn out to be insignificant at the 5%-level. The estimated constant is negative and highly insignificant.

25

Table 5.1.: outcomes of the OLS-estimation of the demand function of the EU natural gas market

The outcomes for the supply estimation using the OLS-regression are shown in table 5.2. In this regression the estimate on price turns out to be positive, with a value of 0.044, but the estimate is insignificant at the 5%-level. All exogenous variables are significant in the model, except for the control variables winter and summer. This may be due to too many annual data, even though they are smoothed. As argued in paragraph 6.2, the OLS-estimators are prone to bias in this model. So, the estimators are probably both inconsistent and inefficient.

Table 5.2: outcomes of the OLS-estimation of the supply function

_cons -66.41193 15.75856 -4.21 0.000 -97.43223 -35.39163 summer -.0096021 .054206 -0.18 0.860 -.1163052 .097101 winter .0005349 .0542434 0.01 0.992 -.1062416 .1073115 elec .0004258 .0001796 2.37 0.018 .0000722 .0007793 ener .0000537 .0000227 2.36 0.019 8.90e-06 .0000984 pop .1406665 .0336689 4.18 0.000 .0743902 .2069428 gdp -7.53e-07 1.25e-07 -6.02 0.000 -9.99e-07 -5.07e-07 p -.2009071 .0347667 -5.78 0.000 -.2693444 -.1324698 q Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 352.831548 287 1.22937822 Root MSE = .37537 Adj R-squared = 0.8854 Residual 39.4529092 280 .140903247 R-squared = 0.8882 Model 313.378639 7 44.768377 Prob > F = 0.0000 F( 7, 280) = 317.72 Source SS df MS Number of obs = 288

_cons 7.682763 .2167432 35.45 0.000 7.256117 8.10941 summer -.0030365 .0476114 -0.06 0.949 -.0967567 .0906837 winter -.0129577 .0487906 -0.27 0.791 -.108999 .0830837 gdpr -2.43e-12 1.02e-13 -23.96 0.000 -2.63e-12 -2.23e-12 w_rig .0002142 .0000814 2.63 0.009 .0000539 .0003744 res .000353 .0000487 7.25 0.000 .0002572 .0004489 p .0443107 .0373873 1.19 0.237 -.0292841 .1179056 q Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 352.831548 287 1.22937822 Root MSE = .3291 Adj R-squared = 0.9119 Residual 30.4342109 281 .1083068 R-squared = 0.9137 Model 322.397337 6 53.7328895 Prob > F = 0.0000 F( 6, 281) = 496.12 Source SS df MS Number of obs = 288

26 Estimating the demand function of the EU natural gas market using the 2SLS-regression gives the outcomes shown in table 5.3. The estimator of price in the demand function is -0.475 (α < 0,005). Just like the OLS-regression outcomes, the only significant exogenous variable that is significant on the 5%-level is the EU energy use. All others are insignificant at the 5%-5%-level. As is the case in the OLS-regression, the estimate of the constant is both highly negative and insignificant.

Table 5.3: outcomes of the 2SLS-estiamation of the EU natural gas demand function

Estimating the supply function of the EU natural gas market using the 2SLS-regression gives the outcomes shown in table 5.4. These estimates turn out to have the expected signs and turn out to be significant at the 5%-level, except for the variable w_rig. This is remarkable, since the coefficient for the variable w_rig is significant in the OLS-regression.

Table 5.4: outcomes of the 2SLS-estimation of the EU natural gas supply function

_cons -22.0988 18.53631 -1.19 0.233 -58.4293 14.23169 summer -.0228147 .0591203 -0.39 0.700 -.1386882 .0930589 winter .0206261 .0592088 0.35 0.728 -.095421 .1366733 elec -.0002889 .0002257 -1.28 0.201 -.0007313 .0001535 ener .0001194 .0000269 4.45 0.000 .0000668 .0001721 pop .0434255 .039757 1.09 0.275 -.0344968 .1213478 gdp -1.11e-07 1.70e-07 -0.66 0.512 -4.44e-07 2.21e-07 p -.4750208 .0573946 -8.28 0.000 -.5875122 -.3625294 q Coef. Std. Err. z P>|z| [95% Conf. Interval] Root MSE = .40915 R-squared = 0.8634 Prob > chi2 = 0.0000 Wald chi2(7) = 1912.40 Instrumental variables (2SLS) regression Number of obs = 288

_cons 7.368856 .2680283 27.49 0.000 6.84353 7.894182 gdpr -2.75e-12 1.87e-13 -14.70 0.000 -3.12e-12 -2.39e-12 w_rig .0000252 .0001211 0.21 0.835 -.0002121 .0002625 res .0004354 .0000634 6.87 0.000 .0003111 .0005597 p .2267516 .0962418 2.36 0.018 .0381212 .415382 q Coef. Std. Err. z P>|z| [95% Conf. Interval] Root MSE = .3386 R-squared = 0.9064 Prob > chi2 = 0.0000 Wald chi2(4) = 97.83 Instrumental variables (2SLS) regression Number of obs = 288

27

Table 5.5: outcomes of the 3SLS-estimation of both the demand and supply function of the EU natural gas market

Estimating the demand function of the EU natural gas market using the 3SLS-regression gives the outcomes shown in table 5.5. The estimates of these regressions are used as estimates for a and b in formula 2.1 and c and d in formula 2.2. Accordingly, these estimates are used in the following welfare analysis.

The estimated coefficient of p is negative (α < 0,005), with a value of -.463. The estimate of the constant has a negative value of -11.296 (α > 0,005). As is the case in the OLS-regression and the 2SLS-regression, the only significant exogenous explanatory variable is the energy use. All other estimates are insignificant at the 5%-level. Taking the estimated coefficient of price for b and the estimated constant for a in formula 2.1 leads to the following estimated demand function of the EU natural gas market in formula 5.1.

_cons 7.669325 .268545 28.56 0.000 7.142986 8.195663 summer .0037585 .0494378 0.08 0.939 -.0931378 .1006548 winter .0079187 .0509585 0.16 0.877 -.0919581 .1077956 gdpr -2.43e-12 1.80e-13 -13.54 0.000 -2.79e-12 -2.08e-12 w_rig .0000514 .0001158 0.44 0.657 -.0001756 .0002784 res .0003905 .0000628 6.22 0.000 .0002675 .0005135 p .1041675 .0913917 1.14 0.254 -.074957 .283292 qSupply _cons -11.29669 16.86279 -0.67 0.503 -44.34716 21.75377 summer -.0222028 .0591134 -0.38 0.707 -.1380629 .0936573 winter .0203508 .0591934 0.34 0.731 -.0956661 .1363677 elec .0001422 .0002053 0.69 0.489 -.0002602 .0005445 ener .0000568 .0000243 2.33 0.020 9.09e-06 .0001045 pop .0267184 .0361582 0.74 0.460 -.0441504 .0975872 gdp -2.35e-07 1.54e-07 -1.52 0.128 -5.37e-07 6.76e-08 p -.4630046 .0542763 -8.53 0.000 -.5693842 -.3566251 qDemand

Coef. Std. Err. z P>|z| [95% Conf. Interval]

qSupply 288 5 .3278587 0.9123 81.82 0.0000 qDemand 288 7 .4111729 0.8620 1876.13 0.0000 Equation Obs Parms RMSE "R-sq" chi2 P Three-stage least-squares regression

28 (5.1) 𝑞_𝑑(𝑝) = −11.296 − .463𝑝

The estimate of the price coefficient of the supply function d is positive, with a value of .104 (α > 0,005). The estimate of the constant holds the value of 7.669 (α < 0,005). As is the case in the 2SLS-regression model, the estimated coefficient for w_rig is insignificant. The gas reserves and Russian GDP are significant. It is remarkable that the estimated coefficient of price is significant in the 2SLS-regression, but is not in the 3SLS-regression. This implies that the exclusion restriction in the 3SLS-regression does not hold. Taking the estimated constant for c in formula 2.2 leads to the following supply function of the EU natural gas market:

(5.2) 𝑞𝑠(𝑝) = 7.669 + .104𝑝

Equating formula 5.1 and 5.2 leads to an equilibrium price in the EU natural gas market of -33.45 $/MMBTU. At this equilibrium price the quantity demanded (and thus supplied) 𝑞𝑑(−33.45) = 4.19 trillion cubic feet of natural gas per year.

US Market

Estimating the demand function of the US natural gas market using the OLS-regression gives the outcomes shown in table 5.6. The estimated coefficient of p is 0.08 and is significant at the 5%-level. Furthermore it shows that the only control variable that has a significant effect on the quantity demanded is the energy use in the EU. The estimated constant is highly negative, but significant. The only specific exogenous variable to the demand function that is insignificant is GDP.

Table 5.6: outcomes of the OLS-estimation of the demand function of the US natural gas market

_cons -66.75301 9.211402 -7.25 0.000 -84.88513 -48.6209 summer -.2478908 .1271893 -1.95 0.052 -.4982555 .0024739 winter .0789141 .1269719 0.62 0.535 -.1710228 .3288509 elec -1.47e-08 1.92e-09 -7.67 0.000 -1.85e-08 -1.09e-08 ener .0032474 .000838 3.88 0.000 .0015978 .004897 pop .422894 .036908 11.46 0.000 .3502426 .4955453 gdp -2.77e-07 2.08e-07 -1.33 0.183 -6.85e-07 1.32e-07 p -.0872836 .0362307 -2.41 0.017 -.1586016 -.0159656 q Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 1078.10381 288 3.74341599 Root MSE = .88117 Adj R-squared = 0.7926 Residual 218.186531 281 .776464522 R-squared = 0.7976 Model 859.917276 7 122.845325 Prob > F = 0.0000 F(7, 281) = 158.21 Source SS df MS Number of obs = 289 . regress q p gdp pop ener elec winter summer

29 Estimating the supply function of the US natural gas market using the OLS-regression gives the outcomes shown in table 5.7. The estimated coefficient of p is -0.08 and is significant at the 5%-level. All exogenous variables turn out to be insignificant in this regression.

Table 5.7: outcomes of the OLS-estimation of the supply function of the US natural gas market

The results of the estimated demand function in the US using the 2SLS-regression are shown in table 5.8. As is the case in Europe, the coefficient of the variable energy use is insignificant in the 2SLS-regression, where it is significant in the OLS-regression. Different from the OLS-regression is that the coefficient for GDP is significant in the 2SLS-regression. The result is that rising GDP has a decreasing effect to natural gas demand. The estimated constant is highly negative, but significant at the 5%-level.

_cons 11.88756 1.579598 7.53 0.000 8.778767 14.99636 summer -.257469 .1819949 -1.41 0.158 -.6156519 .1007139 winter -.0407013 .1852918 -0.22 0.826 -.4053728 .3239702 gdpc 2.47e-06 1.46e-06 1.70 0.091 -3.96e-07 5.34e-06 w_rig .0004635 .0002365 1.96 0.051 -2.02e-06 .0009291 res .0009938 .0004997 1.99 0.048 .0000103 .0019773 p -.089864 .0415057 -2.17 0.031 -.1715511 -.0081768 q Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 1410.03667 299 4.71584169 Root MSE = 1.2863 Adj R-squared = 0.6491 Residual 484.788846 293 1.65456944 R-squared = 0.6562 Model 925.247821 6 154.20797 Prob > F = 0.0000 F(6, 293) = 93.20 Source SS df MS Number of obs = 300 . regress q p res w_rig gdpc winter summer

30

Table 5.8: outcomes of the 2SLS-estimation of the demand function of the US natural gas market

The results of the estimated supply function in the US using the 2SLS-regression are shown in table 5.9. The most remarkable result of this regression is that the estimated coefficient for price is negative.

Table 5.9: outcomes of the 2SLS-estimation of the supply function of the US natural gas market

Instruments: gdp pop ener elec winter summer res w_rig gdpc Instrumented: p

_cons -52.7677 11.34517 -4.65 0.000 -75.00383 -30.53157 summer -.2556022 .1416985 -1.80 0.071 -.5333263 .0221218 winter -.0049306 .1443787 -0.03 0.973 -.2879076 .2780464 elec -7.04e-09 3.41e-09 -2.06 0.039 -1.37e-08 -3.49e-10 ener -.0005696 .0016178 -0.35 0.725 -.0037403 .0026012 pop .4169745 .0411621 10.13 0.000 .3362983 .4976508 gdp -1.45e-06 4.67e-07 -3.10 0.002 -2.36e-06 -5.34e-07 p .231813 .1176014 1.97 0.049 .0013184 .4623075 q Coef. Std. Err. z P>|z| [95% Conf. Interval] Root MSE = .98152 R-squared = 0.7418 Prob > chi2 = 0.0000 Wald chi2(7) = 440.76 Instrumental variables (2SLS) regression Number of obs = 289

Instruments: res w_rig gdpc gdp pop ener elec winter summer Instrumented: p

_cons 12.13429 1.432382 8.47 0.000 9.326874 14.94171 gdpc 2.62e-06 1.37e-06 1.91 0.056 -7.17e-08 5.32e-06 w_rig .0004341 .0002126 2.04 0.041 .0000174 .0008507 res .0009776 .0004587 2.13 0.033 .0000785 .0018766 p -.1934728 .0590906 -3.27 0.001 -.3092882 -.0776573 q Coef. Std. Err. z P>|z| [95% Conf. Interval] Root MSE = 1.158 R-squared = 0.6405 Prob > chi2 = 0.0000 Wald chi2(4) = 509.93 Instrumental variables (2SLS) regression Number of obs = 289

31 Estimating the demand function of the US natural gas market using the 3SLS-regression gives the outcomes shown in table 5.10. As done with the EU market, the estimates of the 3SLS-regression are used for the coeficcients of the suply and demand function in the US. The estimated coefficient op price in the demand function is negative and significant. Energy use and GDP turn out to have a positive effect on gas demand and are significant at the 5%-level. It is estimated that population is negatively correlated with gas demand in the US, but this estimator is insignificant.

The estimated coefficient of price in the US supply function is negative, which is remarkable. Furthermore, the world rig amount and the GDP level of Canada have a positive effect on gas supply and are significant at the 5%-level It is highly remarkable that world gas reserves are negatively correlated with gas supply, but this esimtator is insignifcant.

Table 5.10: outcomes of the 3SLS-estimation of both the demand and supply function of the US natural gas market

Exogenous variables: gdp pop ener winter summer res w_rig gdpc

Endogenous variables: q p

_cons 13.90896 1.291224 10.77 0.000 11.37821 16.43971 summer -.2567626 .2379858 -1.08 0.281 -.7232062 .2096809 winter .0880395 .2410838 0.37 0.715 -.3844761 .560555 gdpc 8.24e-06 1.46e-06 5.66 0.000 5.39e-06 .0000111 w_rig .0004505 .0001992 2.26 0.024 .0000601 .000841 res -.0000494 .0004018 -0.12 0.902 -.000837 .0007382 p -.5923853 .129406 -4.58 0.000 -.8460163 -.3387543 qSupply _cons 54.65942 30.13895 1.81 0.070 -4.411831 113.7307 summer -.2459334 .4153052 -0.59 0.554 -1.059917 .5680499 winter .2225257 .430598 0.52 0.605 -.6214308 1.066482 ener .0005405 .0001395 3.88 0.000 .0002671 .0008139 pop -.2109816 .1434497 -1.47 0.141 -.4921378 .0701747 gdp 2.30e-06 1.07e-06 2.15 0.031 2.06e-07 4.40e-06 p -.9164911 .4434918 -2.07 0.039 -1.785719 -.0472632 qDemand

Coef. Std. Err. z P>|z| [95% Conf. Interval] qSupply 300 6 1.562757 0.4804 343.34 0.0000 qDemand 300 6 2.119493 0.0442 124.30 0.0000 Equation Obs Parms RMSE "R-sq" chi2 P Three-stage least-squares regression

32 Taking the estimated coefficient of price for b and the estimated constant for a in formula 2.1 leads to the following estimated demand function of the US natural gas market in formula 5.4.

(5.3) 𝑞_𝑑(𝑝) = 54.659 − .916𝑝

Taking the estimated constant for c in formula 2.2 leads to the following supply function of the EU natural gas market:

(5.4) 𝑞𝑠(𝑝) = 13.909 − .592𝑝

Equating formula 5.4 and 5.5 leads to an equilibrium price in the US natural gas market of 125.77 $/MMBTU. At this equilibrium price the quantity demanded (and thus supplied) 𝑞𝑑(125.77) = −60.71 trillion cubic feet of natural gas per year.

5.2 CURRENT AMOUNT OF WELFARE GENERATED IN THE EU NATURAL GAS MARKET

Plotting the estimated coefficients of the estimated supply and demand functions of the EU natural gas market (5.1 and 5.1) and the equilibrium price in the EU into formula 3.3 and 3.4 allows us to calculate the total amount of welfare generated in the EU before the removal of the trade barrier. So, a = -11.296, b = 7.669, p = -33.45 and 𝑞𝑑 = 𝑞𝑠 = 4.19. Plotting these values in formula 3.3 and 3.4 for producer and consumer surplus yields the following results:

(5.3) 𝐶𝑆 =1

2(−11.296 − −33.45) ∗ 4.19 = yearly welfare of 46.41 trillion $/MMBTU (5.4) 𝑃𝑆 =1

2 (−33.45 − 7.669) ∗ 4.19 = yearly welfare of − 86.14 trillion $/MMBTU

The sum of the outcomes of 5.3 and 5.4 is the total amount of welfare generated in the EU natural gas market and is shown in formula 5.5.

(5.5) Total Welfare generated = 46.41 − 86.14 = yearly welfare of − 39.73 trillion $/MMBTU

5.3 DETERMINATION OF THE WORLD PRICE

Since the domestic equilibrium price is higher in the US market than in the EU market, there will be excess demand in the US and excess supply in the EU when allowing for international trade. Following the procedure explained at the end of paragraph 3.2, the right values can be plotted into formula 3.8 and 3.9. This leads to the excess supply and excess demand functions shown in formula 5.5 and 5.6. The formulas

33 are depicted graphically in figure 5.1. The equilibrium price that occurs in this market is the world equilibrium price.

(5.5) 𝐸𝑥𝑐𝑒𝑠𝑠 𝑆𝑢𝑝𝑝𝑙𝑦 𝐸𝑈 = 18.965 + 0.567𝑝 (5.6) 𝐸𝑥𝑐𝑒𝑠𝑠 𝐷𝑒𝑚𝑎𝑛𝑑 𝑈𝑆 = 40.75 − 0.324𝑝

Equalizing excess supply in the EU to excess demand in the US leads to a world price of 24.45 $/MMBTU.

Figure 5.1: The graphical depiction of excess supply in the EU and excess demand in the US forming the world market of natural gas after allowing for international trade

5.4 DETERMINATION OF THE WELFARE CHANGE DUE TO INTERNATIONAL TRADE

The determinations of the world price after allowing for international trade allows us to calculate the welfare effect in the EU. As a result of the removal of the trade barrier, the EU market price becomes 24.45 $/MMBTU. At this price level, in the domestic EU market, there would be a demand of -22.62 trillion cubic feet of natural gas. This leads an amount of consumer surplus, calculated in formula 5.7:

(5.7) 𝐶𝑆 =1

2(−11.296 − 24.45) ∗ −22.62 = yearly welfare of 808.574 trillion $/MMBTU

At the level of the world price, the quantity supplied is 10.21 trillion cubic feet of natural gas. The amount of producer surplus generated in the EU market after the removal of the trade barrier is shown in formula 5.8:

0

5

10

15

20

25

30

35

40

45

0

2

4

6

8

₁₀

₁₂

₁₄

₁₆

₁₈

₂₀

₂₂

₂₄

₂₆

₂₈

₃₀

Excess Supply

EU

Excecss

Demand US

34 (5.8) 𝑃𝑆 =1

2(24.45 − 7.669) ∗ 10.21 = yearly welfare of 85.67 trillion $/MMBTU

Accordingly, the total amount of welfare generated in the EU natural gas market after the removal of the trade barrier is calculated in formula 5.9:

(5.9) Total Welfare generated = 808.574 + 85.67 = yearly welfare of 894.244 trillion $/MMBTU

5.5 CONCLUSION

Before allowing for international trade the total amount of welfare generated in the EU natural gas market is a yearly welfare of 501.469 trillion $/MMBTU. After the removal of the trade barrier the amount of welfare generated in the EU natural gas market is a yearly welfare of 894.244 trillion $ / MMBTU. This rise corresponds to an increase in welfare of 78.3% as a result of removing the trade barrier.

So, following this result it can be concluded that the removal of the voluntary export restraint as an effect of signing the TTIP treaty will improve welfare generated in the EU natural gas market with 78.3%. The hypothesis can therefore be approved.

The approval of the hypothesis should not be taken as the truth though, because there are several reasons why the estimated results are not reliable. These reasons will be addressed in the discussion in chapter 6.

35

6 D

ISCUSSION

The estimates of the EU natural gas demand and supply functions show some unlikely results. First of all, the estimator of the constant of the demand function a in the EU natural gas market is negative. This would mean that consumers are only willing to buy zero amounts of natural gas when they receive 11.30$ per unit. Besides that, the estimator is insignificant at the 50%-level. Furthermore it shows that the only control variable that has a significant effect on the quantity demanded is the energy use in the EU. All other variables turn out to be insignificant at the 5%-level.

The insignificant outcomes and the large negative constant show that the estimation results of the regression are not valid. Despite the features of the 3sls-regression method to establish consistent and efficient estimates, the results are not. This flaw can have various reasons.

 The modelling is incorrect. It may be the case that there are more exogenous variables that influence the quantity of natural gas demanded. It may be the case as well that a linear estimation method is not the best way to approach the demand and supply functions, but that another estimation method is better suiting.

 The exclusion restriction may not hold. It may be the case that the exogenous variables of the demand function do affect the quantity supplied and that the exogenous variables of the supply function do affect the quantity demanded. For example, it may well be that gdp influences the quantity supplied. The fact that the estimator of price in the supply function estimated with the 2SLS-regression is significant fosters the idea that there are problems concerning the exclusion restriction.

 It may be that the dataset is not large enough. Another reason influencing the results is that most of the data are annual data. It may well be that gas demand and supply are dependent on the season of the year and that the large amount of annual data fails to address this feature.

A couple of other issues regarding this research have to be borne in mind as well, though. First of all, the research model holds several simplifying assumptions that do not hold in reality. The assumption that international trade between the US and the EU is not affected by exogenous factors, and therefore can be seen as a world market, is very unrealistic. Besides that, the assumption that the empirical estimation of the domestic markets can be regarded as demand and supply functions in a closed economy is very unrealistic too. Both countries trade with neighbouring countries in the natural gas market.