PRELUDE OF THE BREXIT: DETERMINING THE EX-ANTE TRADE EFFECTS ON UK AND EU EXPORT AND IMPORT VOLUMES
Master Thesis University of Groningen Faculty of Economics and Business MSc International Economics and Business
December 2017
Author: Jeroen Hakkaart Student number: 2319683
E-mail address: J.h.a.hakkaart@student.rug.nl Supervisor: Dr. Tristan Kohl
Abstract
This thesis investigates the ex-ante effect of the Brexit on import and export flows between the United Kingdom and European Union using a gravity model of international trade. The four datasets used consists of panel data existing of 245 countries over the months January 2015 and July 2017. The results suggest that the Brexit does not affect total import flows, meaning that countries do not adapt their import pattern. Total export flows do show presence of ex-ante effects around the announcement that a Brexit referendum will be held, but not around other relevant dates of the Brexit. The effects are strong for four of the five largest industries of import and export flows between the European Union and United Kingdom. The strong ex-ante effects hint at a strong revealed comparative disadvantage in industries within the EU28, however such strong comparative disadvantage is not confirmed by the results of a model of revealed comparative advantage leaving the ex-ante effects of the Brexit in the middle.
Table of Contents
List of Figures………...… 4
List of Tables………..….….4
1. Introduction………5
2. Literature………..………7
2.1 Brexit – What Do We Already Know?...7
2.1.1 Timeline……….………….7
2.1.2 Causes of the Brexit……….……….8
2.1.3 Brexit Models... ………....……9
2.1.4 Economic Consequences of the Brexit………...10
2.2 International Trade – The Clear Benefits of the EU Single Market………..11
2.3 Do Countries Anticipate on Brexit?... ………...12
2.3.1 Economic Effect of Trade Agreements………..……12
2.3.2 Anticipation Effects in Trade Agreements………..13
2.4 Factors Negatively Affecting International Trade………..…15
2.4.1 Political Instability……….15
2.4.2 Economic Uncertainty……….15
2.5 Revealed Comparative Advantage……….…………..16
3. Theoretical framework – Gravity Model of International Trade………...……..17
3.1 Gravity Model of International Trade………..…..17
3.2 The Problem of Unobserved Price Indices………..18
3.3 The Problem of Self-Selection……….…………18
3.4 Gravity Model and Anticipation Effects………...………….19
4. Methodology………20
4.1 Data Sources………..…21
4.2 Variable Description and Pre-Estimates……….….21
4.3 Gravity Models……….………23
5. Results & Discussion………...………25
5.1 Effects on the Total Flow of Import and Export……….………….25
5.2 Lagged Effects………...………26
5.3 Effects on Import and Export Flows per Product Group……….…...26
5.4 Revealed Comparative Advantage Model……….……...28
6. Conclusion………..………...30
6.1 Limitations……….…….31
6.2 Recommendations……….………31
7. Bibliography……….…32
List of Figures
Figure 1 - Timeline of Critical Dates in the Prelude of Brexit………8
Figure 2 - UK Net Benefits from EU Membership……….…12
List of Tables Table 1 – Alternative Brexit Models………..……9
Table 2 – Stages in the Lifetime of a EU EIA and their Definitions. ………20
Table 3 – Total Volume of Import………....25
Table 4 – Total Volume of Export……….26
Table 5 – Volume of Export at Product Level By Date and Product Group……….27
Table 6 – Volume of Import at Product Level By Date and Product Group………28
Table 7 – RCA Per Product Group………29
Table 8 – Results Hypotheses……… 29
Table 9 – Summary Statistics ExportsAggregated……….…39
Table 10 – Summary Statistics ImportsAggregated……….…39
Table 11 – Summary Statistics Export per Product Group………...……39
Table 12 – Summary Statistics Import per Product Group………..…40
Table 13 – Correlation Matrix Exports/(Imports) Aggregate...………41
Table 14 – Correlation Matrix Exports/(Imports) per Product Group……….41
Table 15 – Hausman Test: Exports Aggregated………...………..…42
Table 16 – Hausman Test: Imports Aggregated………...……….……42
Table 17 – Main Import Flows Per Product Group, EU-28, 2011 and 2016…………..……43
1. Introduction
For decennia we have been in a period of globalization, a process that integrates social, cultural and economic processes across countries. This integration has stimulated the economic development in many regions of the world. However, recent developments such as the rise of nationalistic thoughts in Europe reflected by the emergence of populist parties, and the election of Donald Trump as the president of the United States with his ‘’America first’’ policy seem to hamper the process of global integration. One of the most prominent and influential disintegrative activities at the moment is the exit of the United Kingdom (UK) out of the European Union (EU): the Brexit.
The decision to leave the EU breaks with the long period of (economic) integration and marks the beginning of an uncertain period for the UK and the EU. This uncertainty is reflected in politics, economics, and society. One of many examples is the uncertain status of EU citizens living in the UK causing a sharp decline of European nurses willing to work at the National Health Service of the UK resulting in a personnel shortage (Tebbens, 2017).
The uncertainty about the consequences of the Brexit forms an interesting topic for researchers and policy makers. Over the past decades, much literature has been written on the economic effects of trade agreements (Aitken, 1973; Winters, 1993; Baier and Bergstrand, 2001; Glick and Rose, 2002; Soete, 2017). At the same time, current Brexit research focuses mostly on the possible economic, social, and political effects of the Brexit on the UK and EU. What literature is clearly lacking is research on the effects of trade agreements before they enter into force, that is: research on ex-ante effects of trade agreements that are ended. Soete (2017) is, to the best of my knowledge, the only authors who found ex-ante effects of trade agreements before those were officially enforced. Without correcting for these ex-ante effects the impact of trade agreements might be underestimated.
Up till now, there has been no published article that investigates the anticipation effects of the Brexit. The focus of research has mainly been on the possible economic effects after the Brexit is official, thereby investigating consequences of different Brexit scenarios.
This thesis investigates imports and export flows among European countries on aggregated level, based on all products as classified by the EU1. The period ranges from January 2015 and July 2017, thereby covering the key dates towards the Brexit. The important variables in this data are import and export flows, as well as the key dates during the prelude of the Brexit. Next to the impact on the total export and imports, the focus is on five product groups that contribute more than 80% to the total export and imports flows between the UK and EU.
This thesis makes two contributions to the current literature. First of all, it extends the current literature about anticipation effects by measuring possible anticipation effects occurring in a situation that a trade agreement is going to end. Second, it makes a contribution by investigating the possible anticipation effects of the Brexit, consequently adding to current literature about the possible economic effects of the Brexit. These economic effects could be higher since Baier and Bergstrand (2007) and Kohl (2014) found the (strong) presence of phase-ineffects. Phase-in effects are the effects that take into account the years after a trade agreement has officially been enforced, this in order to capture the full impact of the specific trade agreements. Soete (2017) elaborates on these phase-in effects in order to determine the presence of anticipation effects. Soete finds strong anticipation effects in EU trade agreements.
In order to determine the presence of anticipation effects of the Brexit, the high dimensional fixed effects regression model is used. The results on total import and export flows level do not show strong presence of anticipation effects. However, focusing on different industries results in strong negative anticipation effects occurring in the machines and transportation vehicles, chemicals, and Food Tobacco and Beverages industry.
This thesis is structured as follows: The second chapter focuses on all relevant information available concerning the Brexit, the benefits of the UK’s membership of the EU and the benefits of trade agreements in general, the concept of anticipation effects, factors negatively affecting international trade, and also looks at the concept of revealed comparative advantage. The third chapter focuses on the gravity model of international trade; Chapter four explains the methodology, which is followed by a section with results and discussion. Finally, a conclusion combined with limitations and recommendations is provided.
2. Literature
In this chapter the information and literature available about the Brexit is discussed
first. The focus is on the timeline of the Brexit, the factors causing the Brexit to occur, the multiple forms that a new trade agreement between the EU and UK can take, and lastly the possible ex-post effects of the Brexit. The next sections discus the benefits that the UK reaped from its position within the EU as well as the benefits of trade agreements in general. Then, ex-ante effects are discussed and how failing to account for ex-ante effects leads to underestimated effects of trade agreements. Next, discusses how political and economic uncertainty might influence export and import trade flows. This chapter ends with the concept of revealed comparative advantage, a concept which is used to determine which industries are more to be affected by trade agreements than others.
2.1 Brexit – What do we already know?
2.1.1 Timeline
On February 22, 2016, former Prime Minister David Cameron announced a referendum that would decide over the fate of the relationship between United Kingdom (UK) and the European Union (EU). By announcing the referendum Cameron fulfilled his election promise to let the people decide over the relation between the UK and EU. On June 23, 2016, the majority (52%) of the people of the UK decided, to end this relationship. The ending of this relationship is now referred to as the Brexit.
The Brexit has some critical dates that need to be taken into account in order to assess whether and in which stage of the prelude of the Brexit anticipation effects occur which is when companies anticipate on the consequences of the Brexit. Figure 1 presents five critical dates, which might play a role in this anticipation effect. The five main dates are:
1. May 7 2015: the announcement of the referendum by Prime Minister David Cameron in the House of Commons that a referendum about the future of the UK as a member of the EU will be held.
2. February 22 2016: the official announcement of the date of the referendum by David Cameron.
3. June 23 2016: the outcome of the referendum. UK citizens could vote in favour or against remaining member of the EU.
4. March 29 2017: trigger of Article 50 which gives every EU member the right to leave the EU. Prime Minister Theresa May triggered Article 50. Consequently, the UK should leave the EU no later than April 2019.
Figure 1: Timeline of Critical Dates in The Prelude of the Brexit
2.1.2 Causes of the Brexit
In light of the outcome of the referendum Paul Krugman (2016) a Distinguished Professor of Economics at the Graduate Center of the City University of New York and a columnist for the New York Times made the following statement:
‘’A number of people deserve vast condemnation here, from David Cameron, who may go down in history as the man who risked wrecking Europe and his own nation for the sake of a momentary political advantage, to the seriously evil editors of Britain’s tabloids, who fed the public a steady diet of lies’’
The first part of his statement refers to the statement of Cameron that he would hold a referendum when he would be re-elected. More importantly, the second part refers to the fact that the voting behavior of people was influenced by tabloids that did not focus on the consequences for the UK were only to be positive.
The most popular arguments posed by the “leave’’ campaign were based on sovereignty and immigration, where benefits of international trade were neglected. Factors that influenced the voting behavior were education, employment rate, and sex (Zhang, 2016). The people that did not profit much from the growing globalization and international economic integration felt left behind and were more likely to be in favor of the Brexit. Multiple papers are written about the causes and consequences of the Brexit and the key findings are presented below.
The main focus of the debate around the Brexit was on national identity and sovereignty (Scruton, 2016), issues concerning immigration (Goodhart, 2016), globalization (Field, 2016), and the presence of a European elite (Tombs, 2016) or elite in London (Street-Porter, 2016). The authors found that UK citizens with a strong sense of national identity, preference for sovereignty, in general against immigration were more likely to vote in favor of the Brexit.
the outcome of the vote, resulting in the fact that people negatively experiencing this unequal distribution voted in favor of the Brexit. Therefore, David Miles (2016) argues that to some extent the Brexit is also because of the failure of economists, besides the politicians and government, to educate and convince the people of the UK that the Brexit would come along with economic costs. Miles argument is strengthened by the research of Los et al. (2017) who find that regions voting in favor of the Brexit are also the regions with the largest dependency on the European single market. This results in the fact that regions that have voted in favor of the Brexit will face most negative consequences of the Brexit.
2.1.3 Brexit models
In order to understand the possible consequences it is important to understand the distinction between two broadly used terms for the outcomes of the Brexit: Hard and Soft Brexit. Some possible outcomes of the Brexit are depicted in Table 1.2 It is unclear whether the outcomes of the negotiations between the EU and the UK will turn into a Soft or Hard Brexit. Soft Brexit in general refers to a model comparable to those of Norway or Switzerland, and a Hard Brexit refers to the scenarios of Canada, Turkey, or a general WTO agreement. An important distinction is whether the UK will keep its access to the European single market, because not having this access is crucial to the size of the economic loss.
Table 1. Alternative Brexit models
Source: BBC, 12 June 20173
As becomes apparent from Table 1, an important consequence of the Brexit is the fact that negotiations are about the new relationship with the EU, of which the outcome can
2 For a detailed description of Alternative Brexit models see Report Global Counsel (2015), and Dhingra,
Ottaviano,and Sampson (2016)
take multiple forms. The EU is the UK’s largest trading partner with 53% of it’s total import and export flows (Armstrong, 2016). By leaving the EU the UK most probably also leaves the EU single market, which puts the UK in a new international trading position. It has to take a new position in the WTO agreements and other trade agreements created during the UKs membership of the EU. For example, the EU has 53 preferential trade agreements from which the UK is now excluded (Armstrong, 2016). The UK has to reestablish these trade agreements on its own in order to avoid increased trading barriers with other countries.
2.1.4 Economic consequences of the Brexit
Baldwin (2016) argues that the Brexit might lead to a GDP reduction ranging from 1.5% to more than 7% for the UK. Ottoviano et al. (2014) found similar results, and Campos (2014) states that economists’ reports support the conclusion that the Brexit comes with costs. The short run estimates confirm the fact that the Brexit will cost the UK money, and consequently have a negative impact on the wealth of the UK. The Centraal Plan Bureau (2016) finds that in the case of the WTO scenario, falling under Hard Brexit as can be observed Table 1, trade will be reduced by 23% and GDP by 4.1%. Brakman et al. (2017) find that in the case of a hard Brexit the value added export will be reduced by 18%, because costs to trade with the EU increase. In the case of a soft Brexit this will be reduced to 14%. The authors also find that the Brexit has negative consequences for the EU countries, though being significantly less than for the UK. The conclusion of the paper of Brakman et al. is that there is no alternative to the EU membership to compensate for the large negative impact of the Brexit on trade flows.
Yet, the economic consequences are not only limited to GDP as a recent study by Los (2017) shows that 2.5 million jobs are at risk. The effect in the long run on GDP, ex-post, can only be measured in 2030. This possible reduction in GDP is not expected to be equally distributed across the UK. Ass et al. (2017) argue that London is far less dependent on the European market due to the fact that it is a global city. It focuses mostly on the international (financial) service sector, and the economy is much more diverse and much larger than other regions in the UK. Furthermore, Los (2017) has found that manufacturing and primary industries are extensively exposed to the Brexit. Also the service industry is affected, but indirectly since many of their UK clients are directly engaged with the EU.
but also heavily affects the service sector. Los, Chen, McCann and Ortega-Argilés (2017) find that chemicals, machinery and equipment, and food beverages and tobacco risk a loss of 33%, 24%, and 14% respectively.
So, from the above it becomes apparent that Brexit can come along with a reduction in GDP, loss of jobs, and a loss of share in the total value added. If leaving the EU truly has such negative consequences, the membership of the EU should have had positive consequences for GDP growth. The next section focuses on these benefits that the membership of the EU has brought the UK.
2.2 International trade – the clear benefits of the EU single market
Campos (2014) argues that there are three main mechanisms via which the UK was able to benefit substantially from the EU. These three mechanisms are trade, Foreign Direct Investment (FDI), and finance. This thesis focuses on the first mechanism: international trade volumes consisting of export and import.
The goal of the European single market was to boost economic growth and to eliminate conflicts (Martin et al. 2012). In general there is a strong consensus between economists that opening up of markets stimulates competition and innovation, consequently leading to increased wealth. Feyer (2009), Melitz and Trefler (2012), Her Majesty’s Treasury (2016), and CPB (2016) built a relation in their economic models between trade volumes and productivity, resulting in the fact that a reduction in import and export flows decreases productivity. Consequently, the impact of decreasing trade volumes can have serious negative implications for productivity growth. This is confirmed by Awokuse (2006) who finds a positive relationship between trade and economic growth by performing a Granger causality test with panel data of the countries: Bulgaria, Czech Republic, and Poland.
Campos and Coricelli (2015) emphasize that trade with the EU is so beneficial because it concerns intra-industry trade. This means that competition and technological innovation are very important, two factors forming the basis for GDP growth and Total Factor Productivity (TFP) growth. Figure 2 below shows UK’s increase in GDP per capita from its EU membership and consequently the benefits of international trade. The synthetic line reflects the real GDP per capita of the UK if it had would never have joined the EU. The black line reflects the actual development of real GDP in the UK. The benefits originating from the trade agreement between the EU and the UK become very clear here.
Figure 2. UK net benefits from EU membership
Source: Campos et al 2014.
But even with these results present, the people of the UK made the decision to terminate the relationship with the EU. This means that, depending on the outcomes of the negotiations, it might have a large impact on the volume of international trade and consequently on the real GDP per capita. One of the important consequences might be that the UK has to leave the European single market, which brought much wealth as it can be observed in Figure 2.
The trade agreement between the UK and the EU has brought significantly more wealth to the UK, as a logical consequence it is very much possible that after ending the trade agreement a strong reduction in real GDP per capita in the UK will be observed. As previously stated the conclusion of the paper of Brakman et al. (2017) is that there is no alternative to the EU membership to compensate for the large negative impact of the Brexit on trade flows. Thus it can be expected that the volume of imports and exports will be reduced.
It seems that from an international trade perspective the decision of the UK to leave the EU is quite remarkable. The trade agreement between the UK and EU brought real wealth, and ending this agreement might break this accumulated wealth down.
2.3. Do Countries Anticipate On The Brexit?
As established in the previous section, the economic integration between the EU and UK has brought a significant increase in real GDP per capita. This section focuses on the economic effects of trade agreements in general, and introduces the concepts of phase-in and anticipation effects.
2.3.1 Economic Effect Of Trade Agreements
In general, literature focuses on the economic effects of trade agreements after they are implemented and enforced. Thus, the focus was on the ex-post effects of trade agreements. There has been substantial research on the economic impact of trade agreements, focusing on FDI, export and import flows and GDP. For policy makers, it is most interesting to understand how large the impact of trade agreements can be. Kohl (2014) has conducted an extensive survey about the impact of trade agreements, focusing on different sorts of trade agreements and focusing on multiple regions. In order to obtain an extensive overview of the diverse findings of studies on the impact of trade agreements, examining this survey is recommended. Diverse findings in research are reflected by the findings of Brada and mendex (1983) and and Ghosh and Yamarik (2004b) who did not find significant results on an increase or decrease in trade, Frankel and Wei (1997) who found an increase in trade and Soloaga and Winters (2001) who found a decrease in trade.
But, do these results reflect reality? Baier and Bergstrand, henceforward B&B, conclude that because of differing findings in research there was no reliable ex-post treatment effect measurable after trade agreements were installed. Among studies of international trade the gravity model is broadly used. However, according to Egger (2002), large opposing results between predicted and actual trade flows hint at the deficiencies and problems of the model. These methodological problems regarding the model could potentially be the source of these unreliable ex-post effects of trade agreements. Chapter 4 will go into more detailed discussion.
2.3.2 Anticipation Effects In Trade Agreements
In 2007, B&B introduced the concept of phase-in effects, effects that account for the fact that it might take multiple years before the potential of a trade agreement is completely captured. This was new in literature on trade agreements since previously the focus was on the static effects of trade agreements, thereby possibly not fully capturing the economic effects of trade agreements. B&B (2007) make use of lagged variables of Economic Integration Agreements (EIA) to test for these possible phase-in effects. EIAs can refer to Free Trade Agreements (FTA), Economic Unions (such as the EU), and Custom Unions (CU) (Baier, Bergstrand, Feng, 2014). The authors find that trade flows are doubled after ten years between countries with trade agreements and that previous research has been underestimating the effect of trade agreements. B&B (2007) suggest the possibility that trade increases before governments officially enforce a trade agreement. Thereby, the authors follow Mclaren (1997), who argued that sunk costs, in for example infrastructure, can lead to an increase in trade before the trade agreement has been officially enforced. As an extension of their research B&B also test for possible anticipation effects of trade agreements, yet they do not find these anticipation effects to exist.
the effect of a trade agreement can be explained by factors such as year of enforcement and specific characteristics of the trade agreement. He recommends that further research is needed on possible anticipation effects that occurs before the economic agreement is signed by governments of nations. In anticipation of a trade agreement it can be the case that multiple factors of the trade agreement are already in place, meaning that trade increases before the agreement is officially enforced.
B&B (2007) and Kohl (2014) find phase-in effects of trade agreements. Their findings are confirmed by Zylkin (2016) who finds that trade almost doubles taking into account phase-in effects. Grassnick (2017) finds, by using the Poisson pseudo maximum likelihood method, that phase-out periods have a significant impact on the Mexican agricultural imports sector. Also Grassnick found that the longer the period was the larger the impact on trade was under NAFTA.
Constantini and Melitz (2008) find in their research that the announcement of the introduction of a more liberal trade agreement causes firms to anticipate on this liberalization of the market and that it stimulates firms to innovate in order to grasp first mover advantages in export markets. Burnstein and Melitz (2011) predict in their model, in which they created stages that firms anticipate their level of exports already on the moment of the announcement and consequently before the agreement enters into force. It is not clear about whether these anticipation effects exists. Also, if they do exists it is not clear if complete countries are affected or that only specific industries are tackled. Because of these uncertainties and the fact that negative anticipation effects caused by the Brexit are taken into account this thesis is written, so it can contribute to the theory of anticipation effects.
The papers of B&B and Kohl who do not find such anticipation effects do not confirm the prediction of anticipation effects by Burnstein and Melitz. However, Soete (2017) confirms the predictions of Burtstein and Melitz and finds that anticipation effects do exist for EU agreements by making use of the gravity model of international trade.Soete focused on different stages before the official enforcement of the trade agreement. The contribution of Soete (2017) was that the existence of anticipation effects in different stages in the development towards a trade agreement was confirmed.
Above we have observed literature predicts negative consequences of the Brexit. Furthermore, this section discusses literature on phase-in effects, and the anticipation effect as found by Soete. Following from these sections the first hypothesis are derived: H1a: The total volume of Import between the UK and EU decrease during the prelude of the Brexit.
That the Brexit has serious negative consequences is now broadly discussed, however this thesis did not yet focus on other factors that negatively affect international trade. Therefore, the next section focuses on those factors.
2.4 Factors Negatively Affecting International Trade
There are multiple factors that effect international trade. In case of the Brexit the focus is on possible negative side effects in politics and economics, since the Brexit does not affect factors such as geographical location and natural resources.
2.4.1 Political Instability
The Brexit has introduced a period of political, economic, and social instability. There is an extensive body of literature that argues that political instability hampers economic development. It limits the focus of policy makers in the long term, leading to short-term macroeconomic policies (El-Masry and Badr, 2017). Furthermore, Alesina et al. (1996) find that in times of political instability GDP-growth is significantly lower. The trigger of Article50 changes the political relationship between the UK and EU as well, which might be reflected in a lower GDP growth. Aiezenman and Marion (1993) find a negative relationship between policy uncertainty and real GDP growth per capita. Furthermore, it leads to a negative impact on the macroeconomic performance in the long run (Aisen and Veiga, 2011). Jong-a-Pin (2009) also finds, looking at 25 political instability indicators, that instability of the political regime has a significant negative impact on economic growth including factors such as GDP growth, private investment, and inflation. Campos and Nugent (2002) find, by conducting a sensitivity analysis on the relationship between economic growth and political instability that there is only a contemporaneous negative relationship on the short-term, but not in the long run. Alesina and Perotti (1996) find that political and social instability reduced investment due to the increased risks in the new economic environment. In the case of the Brexit it might mean that investors in European countries are less likely to make new investments in the UK, and the other way around. Aisen van Veiga (2011), supported by the findings of Özler and Rodrik (1992), conclude that political instability reduces investments and that productivity growth is negatively affected. This reduction in investments also leads to a decrease in competition and innovation, which can make products make less attractive on the global market, and consequently possibly decreasing the export and import flows (Campos, 2014).
2.4.2. Economic Uncertainty
The previous section about political instability signaled towards the negative implications for the economy. In this section the focus is more specifically on the consequences of economic uncertainty on international trade.
authors also find that the attitude towards the risk of exchange rate volatility does not have a direct impact on the volume of trade, but it does have an impact on the distribution of the production. The authors found that risk-averse firms reduce their optimal production point in order to avoid risk. This means that some firms are willing to reduce their production and make it consequently more expensive per product if that is what it takes to avoid risk. A reduction in production is also most likely reflected in the export streams of countries, since increasing prices make products less attractive for foreign markets. Vice-versa it could also be expected that when the expected foreign exchange rate is about to increase, the risk-averse firms will increase their export. Burnstein and Melitz (2011) find that the level of uncertainty does not affect the level of exports. According to the authors the reason for this is that firms make the decision of setting a price and export after trade costs have already been realized. An interesting finding is that uncertainty does have an impact on firms that enter the export market. In this situation firms make the decision about their export before the initial trade costs are made, thus no sunk costs are present (Burnstein and Melitz, 2011). In the case of the Brexit and the uncertain outcome of it might postpone or change the new export entry decision of firms, meaning that the Brexit can have a serious impact on new export entry decision of EU and UK firms. But, the effects will not be visible in the prelude of the Brexit, but might be visible after the outcomes of the negotiations are presented. Following the arguments of Burnstein and Melitz the second hypothesis is derived, which contradict hypothesis 1b:
H2: The total volume of export between the EU and UK are not affected by the Brexit. Different theories on the effects on the volume of trade exists, therefore the contradicting hypotheses H1b and H2 are included.
2.5 Revealed Comparative Advantage
Revealed comparative advantage (RCA) is a concept, based on the Ricardian concept of comparative advantage, that is introduced by Liesner (1958) and extended by Balassa (1965) RCA-index determines whether a certain industry of a certain region or country has an advantage or disadvantage observing certain trade flows. Consequently, RCA is a useful method to reveal a comparative advantage within a certain commodity or product group.
Utkulu and Seymen (2004) analyzed the competitiveness of certain industries observing trade flows between Turkey and the EU. The authors found that the Custom Union agreements that were enforced in 1996 between the EU and Turkey increased the pressure for competitiveness for both sides. In their paper they refer to the following model of RCA:
𝑅𝐶𝐴 = (𝑋(𝑖𝑗) − 𝑀(𝑖𝑗)) ÷ (𝑋(𝑖𝑗) + 𝑀(𝑖𝑗))
This model can be used to assess the trade performance of a country or an economic union in commodity or product groups.4 The ratio varies from -1 to 1, where -1 means a revealed comparative disadvantages and where 1 means a revealed comparative advantage. The advantage of this model of RCA is that it takes simultaneous import and export flows within industries into account within a country or economic region. Via this method it is possible to calculate the RCA of the five largest product groups responsible for more than 80% of trade between the EU and UK as can be observed in Table 17 and 18 in Appendix H.
This means that the Brexit might reveal in which industries the UK has a comparative (dis)advantage. Following the latest’s findings of Los et al. (2017) in combination with the results on the presence of anticipation effects of Soete (2017) a decrease in import and export flows in all five industries that are observed is expected. Therefore, the third hypotheses are:
H3a: The total volume of export of each product group does decrease during the prelude of the Brexit.
H3b: The total volume of import of each product group does decrease during the prelude of the Brexit.
3. Theoretical Framework – Gravity Model of International Trade
In this section the theoretical development of the gravity model of international trade is discussed in detail. This includes key contributions to the gravity model such as tackling problems of unobserved price indices and endogeneity.
3.1 Gravity Model of International Trade
The gravity model of international trade is generally used to understand existing bilateral trade flows between countries or regions and takes the form of a log linear function. The gravity equation presents the volume of the trade flows, export and import, by also correcting for the multiple factors that possibly influence trade. A basic gravity model is composed as follows:
𝑙𝑛𝑋𝑖𝑗𝑡 = 𝛽𝑜 + 𝛽1 𝑙𝑛𝐺𝐷𝑃(𝑖𝑡) + 𝛽2 𝑙𝑛𝐺𝐷𝑃(𝑗𝑡) − 𝛽3 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑖𝑗) + 𝛽4 𝐹𝑇𝐴(𝑖𝑗) + 𝛽5 𝐶𝑜𝑙𝑜𝑛𝑖𝑎𝑙𝐻𝑒𝑟𝑖𝑎𝑡𝑔𝑒(𝑖𝑗) + 𝛽6 𝐶𝑜𝑚𝑚𝑜𝑛𝑙𝑎𝑔𝑢𝑎𝑔𝑒(𝑖𝑗) + 𝜀𝑟𝑟𝑜𝑟(𝑖𝑗𝑡)
The model has been subject to some important econometrical adaptations. For a detailed description on all developments of the gravity model, the paper of Head and Mayer (2014) is recommended since it provides a broad overview of the developments focusing on for example Multilateral Resistance Term (MRT), first differencing and the inclusion of fixed effect. In this section, the focus will be on the important developments most relevant for the gravity model used in this thesis.
3.2 The problem of unobserved price indices
Multiple authors such as Feenstra et al. (2001) and Frankel and Rose (2002) showed that trade agreements stimulate the net trade creation. But, Anderson and van Wincoop (2003) argued that research on the economic effect of trade agreements were biased since unobserved price indices were not taken into account. Anderson and van Wincoop (2003) have contributed to the development of the gravity mode by implementing the MRT, instead of using a regular Ordinary Least Square model. According to Anderson and van Wincoop (2003) and Feenstra (2004, Ch.5) country specific fixed effects correct for the omitted variable bias. The omitted variable bias occurs when factors that are important for the model are left out, consequently the model can under- or overestimate the values of the other factors included in the model. MRT also accounts for the fact that substitutability among partners of a country is a possibility, which was lacking until then (Adam & Cobham, 2007). For example, trade between the Netherlands and Belgium is determined by the relative costs of trading with each other compared to other countries. If a trade barrier between the Netherlands and the UK diminishes, a third country such as Belgium might be affected since the Netherlands and UK switch their export and import production towards each other. Trade flows between Belgium and The Netherlands will decrease even though the bilateral trade barrier between the two countries remained the same. The contribution of Anderson and van Wincoop, now broadly used in literature, leading to the inclusion of country fixed effects that tackle the problem of endogeneity caused by the unobserved price term. Endogeneity occurs when an independent variable correlates with the error term (B&B, 2007).
3.3 The Problem of Self-Selection
B&B (2007) addressed this problem of endogeneity and argue that it could arise from three different sources. Namely, from omitted variables, simultaneity and a measurement error. According to B&B, the omitted variable and the selection bias is the most important source of the problem of endogeneity. Panel regression techniques, and consequently using panel data, are used to tackle the problem of endogeneity and to create stable results on the country fixed effects (Cheng and Wall, 2005). B&B solve the problem of endogeneity by using panel data in combination with country-pair fixed effects as well as country-time fixed effects: Fit, Fjt, Fij. These fixed effect take into account the variation of export and import flows between countries, while also taking into account trade between these countries and the rest of world. This introduction of usage of fixed effects is important in literature and now frequently used by other authors (Kohl, 2014; Soete 2017). Their research shows that free trade agreements increase trade flows up to five times when correcting for endogeneity and phase-in effects.
3.4 The Gravity model and anticipation effects
Soete (2017) is first to successfully find anticipation effects in EAIs by using the fixed effects in the gravity model as implied by B&B (2007). Soete focused on EIA’s in the European Union from the period 1988-2013 and their impact on bilateral import and export flows before they came into force. Soete determined seven critical stages towards enforcement of an EAI, and created a dummy variable for every phase into the model. By using this refined technique of creating different stages, Soete was able to determine that anticipation effects occur. Note that the contributions of Andersson and van Wincoop (2003) and B&B (2007) are included in order to tackle the problem of the unobserved price terms and the problem of endogeneity. Table 2 provides an overview of the different stages that are included in the gravity model by Soete. Soete has delicately defined every step that brings an EIA closer to being installed, and divided them in seven different stages. Soete created dummy variables for each stage.
Table 2: Stages in the lifetime of a EU EIA and their definitions.
Source: Soete 2017, Table 3.1, page 61.
Soete finds that the level of trade indeed shows an increase from the movement that involved governmental bodies start negotiations. Trade increased even further after the trade agreement had come into force. Thus, Soete provides evidence for the existence of the anticipation effect concerning trade agreements. The findings show the importance of the anticipation effect since static effects underestimate the impact of trade agreements. Soete found that trade flows increased already with 21% as a result of the anticipation effect. Five years after the enforcement of the trade agreement the trade flows had increased with 43%, this means that half of the increased trade flows comes from anticipation effects. In conclusion the refined gravity model of Soete has found strong presence of anticipation effects.
4. Methodology
Furthermore on product level the focus will be on the following five industries: 1) Machinery and transport equipment,
2) Miscellaneous Manufacturing Goods, 3) Chemicals,
4) Food, Drinks & Tobacco, 5) Minerals and Fuels.
The reason that these five industries are selected for the import and export flows is simply that those industries take more than 80% of import and export flows between the UK and EU into account (Eurostat, 2017). Table 14 and 15 in Appendix H show these main product groups responsible for a large share of the import- and export flows within the EU28. If the Brexit affects these industries, it will have a substantial impact on the export and import flows within EU28.
4.1 Data Sources
In order to conduct a research on the hypotheses and the conceptual model multiple data sources have been used. Eurostat provides statistical information about the European Union including the international trade per EU member. Specifically the Eurostat provides the trading volume per product type on a monthly basis per EU member. The dataset consists of the EU member states and the trading partner outside the EU as described in Appendix B and C. The data is available and can be downloaded by everyone from the website of Eurostat. The dataset provides export and also import data on a product level for the dependent variable. Both dataset have been converted to importer and exporter aggregated level5. The data retrieved is panel data.
4.2 Variable Description and Pre-Estimates
This thesis makes use of four different dataset. The first two datasets contain export and import aggregated data. The other two datasets contain export and import data on product level. Appendix A provides a detailed description on the variables that are included in these datasets. Furthermore, Appendix D contains summary statistics for each dataset of these variables. The aggregated datasets consist of 132.068 observations for import and 160.565 observations for export. The datasets on product level contain 55.121.470 observation for import and 78.130.704 observations for export data. Both export datasets contain 28 declarants and 246 partners. Both import datasets contain 28 declarants and 245 partners. In this thesis the dependent variable is export volume or import volume at aggregated level or product level. A variable of international trade (Xijt) is constructed in terms of these import and export volumes.
From Appendix E, containing tests for skewness and kurtosis, follows that the datasets are not normally distributed. Furthermore the summary statistics show high values of skewness from which can be concluded that the residuals of the import and export
variables are not perfectly symmetric and consequently are skewed positively to the right. Also the value of kurtosis differs strongly from 3. Therefore, in order to create a normal distribution, the logarithms of export (lnexportsaggregated) and import (lnImportaggregated) are created. Appendix F displays the correlation matrix of the relevant variables of the datasets; Tables 13 and 14 do not show extreme correlation between variables.
Dummy variables are created for the critical dates containing the value of one from the month of the critical date onwards, zero otherwise. These dummy variables were created for all four datasets.
In general there are two estimation techniques that are used in the case of these gravity models, the fixed effects method and the random effect method. Recent literature strongly suggests that the fixed effect method is preferred over the random effect method by estimating trade flows. In order to provide additional proof the Hausman test is performed as depicted in Appendix G. As can be observed the values are positive and also significant, confirming the existing literature, the differences in coefficient are important and consequently the fixed effects model needs to be used.
The fixed effects do not correlate with other characteristics and are therefore used. By the usage of this set of fixed effects, Fij, Fit, Fjt, it is possible to control for the time dimension as well as the country pair dimension, so investigating the expected variance within an observation over time.
The models contain the assumption that that the error term correlates with the explanatory variables. The mentioned fixed effects are implemented in the model in order to correct for this correlation, so that the coefficients of the independent variables truly say something about the dependent variable.
The two datasets containing product groups make use of fixed effects. The Hausman test was not performed for these datasets since literature strongly suggest the usage of fixed effects over random effects. Furthermore, the usage of fixed effect is also confirmed by the datasets of total import and export flows. In the case of the two datasets of the product level the fixed effect also include a product dimensions resulting in time fixed effects (period declarant product & period partner product) and country pair (declarant partner product) fixed effects. Next, testing for normality and observing the values of skewness and kurtosis lead to the same conclusion as for the datasets on total exports and import flows to use log functions for import and export.
on aggregated level in order to wider the perspective on possible anticipation effects during the prelude of the Brexit. Longer lag periods are not included since it would result in a substantial loss of data. Every extra lag leads to an extra month loss of data. In general a huge loss of data decreases the reliability of results.
4.3 Gravity models
The gravity models used in this thesis are based on the following model, as also mentioned in section 3.1:
𝑙𝑛𝑋𝑖𝑗𝑡 = 𝛽𝑜 + 𝛽1 𝑙𝑛𝐺𝐷𝑃(𝑖𝑡) + 𝛽2 𝑙𝑛𝐺𝐷𝑃(𝑗𝑡) − 𝛽3 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑖𝑗) + 𝛽4 𝐹𝑇𝐴(𝑖𝑗) + 𝛽5 𝐶𝑜𝑙𝑜𝑛𝑖𝑎𝑙𝐻𝑒𝑟𝑖𝑎𝑡𝑔𝑒(𝑖𝑗) + 𝛽6 𝐶𝑜𝑚𝑚𝑜𝑛𝑙𝑎𝑔𝑢𝑎𝑔𝑒(𝑖𝑗) + 𝜀𝑟𝑟𝑜𝑟(𝑖𝑗𝑡)
We learned that accounting for the MRT and endogeneity the usage of panel data and the inclusion of importer-time and exporter-time fixed effects the model could be adapted in order to retrieve more detailed results. Also we learned from Soete that by creating dummy variables for each critical date anticipation effects per date could be grasped specifically. When including these specifications the gravity model on aggregated level (1) in created:
(1)
𝑙𝑛𝑋𝑖𝑗𝑡 = 𝛽𝑜 + 𝛽1 𝑙𝑛𝐺𝐷𝑃(𝑖𝑡) + 𝛽2 𝑙𝑛𝐺𝐷𝑃(𝑗𝑡) − 𝛽3 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑖𝑗) + 𝛽4 𝐹𝑇𝐴(𝑖𝑗𝑡)
+ 𝛽5 𝐶𝑜𝑙𝑜𝑛𝑖𝑎𝑙𝐻𝑒𝑟𝑖𝑡𝑎𝑔𝑒(𝑖𝑗) + 𝛽6 𝐶𝑜𝑚𝑚𝑜𝑛𝑙𝑎𝑔𝑢𝑎𝑔𝑒(𝑖𝑗) + 𝛽7 𝑎𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡(𝑖𝑗𝑡) + 𝛽8𝑑𝑎𝑡𝑒𝑠𝑒𝑡(𝑖𝑗𝑡) + 𝛽9𝑜𝑢𝑡𝑐𝑜𝑚𝑒(𝑖𝑗𝑡) + 𝛽10 𝐴𝑟𝑡50(𝑖𝑗𝑡) + 𝛽11 𝑛𝑒𝑔𝑜𝑡𝑖𝑎𝑡𝑖𝑜𝑛𝑠(𝑖𝑗𝑡) + 𝛿(𝑗𝑖𝑡) + 𝜁(𝑗𝑡) + 𝜂(𝑖𝑗) + 𝜀(𝑖𝑗𝑡)
By including the aforementioned variables in the gravity model and to let it interact with the different variables representing the different dates, it is possible to test the hypothesis 1a, 1b, and 2. Following from Soete the dummy variable of the dates have the value of one when the time passes their specific month. First, the effects of the dummy variables of the critical dates are estimated separately. Next, the combined effects of those dummy variables are determined.
In order to test the effect of lagged effects the following models is created. The model displays the lagged effect correcting for one month.
(2) 𝑙𝑛𝑋𝑖𝑗𝑡 = 𝛽𝑜 + 𝛽1 𝑙𝑛𝐺𝐷𝑃(𝑖𝑡 − (𝑡 − 1)) + 𝛽2 𝑙𝑛𝐺𝐷𝑃(𝑗𝑡 − (𝑡 − 1)) − 𝛽3 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑖𝑗) + 𝛽4 𝐹𝑇𝐴(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛽5 𝐶𝑜𝑙𝑜𝑛𝑖𝑎𝑙𝐻𝑒𝑟𝑖𝑡𝑎𝑔𝑒(𝑖𝑗) + 𝛽6 𝐶𝑜𝑚𝑚𝑜𝑛𝑙𝑎𝑔𝑢𝑎𝑔𝑒(𝑖𝑗) + 𝛽7 𝑎𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛽8𝑑𝑎𝑡𝑒𝑠𝑒𝑡(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛽9𝑜𝑢𝑡𝑐𝑜𝑚𝑒(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛽10 𝐴𝑟𝑡50(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛽11 𝑛𝑒𝑔𝑜𝑡𝑖𝑎𝑡𝑖𝑜𝑛𝑠(𝑖𝑗𝑡 − (𝑡 − 1)) + 𝛿(𝑗𝑖𝑡) + 𝜁(𝑗𝑡) + 𝜂(𝑖𝑗) + 𝜀(𝑖𝑗𝑡 − (𝑡 − 1))
model on aggregated level, only the fixed effects have to account for changes in products over time. Furthermore, variables for the product groups are included in order to determine possible anticipation effects per industry.
(3) 𝑙𝑛𝑋𝑖𝑗𝑡 = 𝛽𝑜 + 𝛽1 𝑙𝑛𝐺𝐷𝑃(𝑖𝑡) + 𝛽2 𝑙𝑛𝐺𝐷𝑃(𝑗𝑡) − 𝛽3 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒(𝑖𝑗) + 𝛽4 𝐹𝑇𝐴(𝑖𝑗𝑡) + 𝛽5 𝐶𝑜𝑙𝑜𝑛𝑖𝑎𝑙𝐻𝑒𝑟𝑖𝑡𝑎𝑔𝑒(𝑖𝑗) + 𝛽6 𝐶𝑜𝑚𝑚𝑜𝑛𝑙𝑎𝑔𝑢𝑎𝑔𝑒(𝑖𝑗) + 𝛽7 𝑎𝑛𝑛𝑜𝑢𝑛𝑐𝑒𝑚𝑒𝑛𝑡/𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦(𝑖𝑗𝑡𝑝) + 𝛽8𝑑𝑎𝑡𝑒𝑠𝑒𝑡/𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦(𝑖𝑗𝑡𝑝) + 𝛽9𝑜𝑢𝑡𝑐𝑜𝑚𝑒/𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦(𝑖𝑗𝑡𝑝) + 𝛽10 𝐴𝑟𝑡50/𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦(𝑖𝑗𝑡𝑝) + 𝛽11 𝑛𝑒𝑔𝑜𝑡𝑖𝑎𝑡𝑖𝑜𝑛𝑠/𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑦(𝑖𝑗𝑡𝑝) + 𝛿(𝑖𝑡𝑝) + 𝜁(𝑗𝑡𝑝) + 𝜂(𝑖𝑗) + 𝜀(𝑖𝑗𝑡𝑝) By including the aforementioned variables in the gravity model and to let it interact with the different variables representing the different dates, it is possible to test the hypothesis 3a and 3b. Following from Soete the dummy variable of the dates have the value of 1 when the time passes their specific month. First, the effects of the dummy variables of the critical dates are estimated separately. Next, the combined effects of those dummy variables are determined.
5. Results and Discussion
In this section the results of the different regressions are discussed in four different subsections: Import and export aggregated, lagged effects on aggregated level, import and export at product level, and the model of RCA.
5.1 Effects on the Total Flows of Import and Export Aggregated
This subsection observes whether anticipation effects occur at an aggregated import and export level between the EU and the UK. As can be observed in Table 3, import aggregate does not find any significant anticipation effects. This would mean that import flows between the EU and UK are not affected during the prelude of the Brexit and leads not to the rejection of Hypothesis 1a.
Table 3: Total volume of Import
(1) (2) (3) (4) (5) (6) Announcement (0.807) .0113 (0.985) .0010 Dateset (0.636) .0161 (0.774) .0166 Outcome (0.721) .0121 (0.879) .0086 Art50 (0.739) -.0167 (0.573) -.0312 Negotiations (0.721) .0121 Omitted N 131,247 131,247 131,247 131,247 131,247 131,247 R-Squared 0.9127 0.9127 0.9127 0.9127 0.9127 0.9127
Note: Standard errors are reported in parentheses. ***, **, * denote significance at 1%, 5%
and 10% respectively. The variables in this Table are explained in Appendix A.
Table 4: Total volume of Exports (1) (2) (3) (4) (5) (6) Announcement -0.0635** -05826 (0.049) (0.106) Dateset -0.0288 .0029 (0.214) (0.936) Outcome -0. 0.028 -.0114 (0.231) (0.760) Art50 -.02628 -.0218 (0.448) (0.611) Negotiation 0.0149 .0497 (0.824) (0.511) N 160120 160120 160120 160120 160120 160120 R-Squared 0.9183 0.9183 0.9183 0.9183 0.9183 0.9183
Note: Standard errors are reported in parentheses. ***, **, * denote significance at 1%, 5%
and 10% respectively. The variables in this Table are explained in Appendix A. 5.2 Lagged effects
Both on export as well as import aggregated level presence of possible lagged effects are investigated. The findings are presented in Appendix I and J. It is important to understand that the coefficient of lagged variables do not mean precisely the same as the coefficient of the normal dates. This is because of the fact that the lagged variables include the value of the extra lagged period, so in this thesis a maximum of four years. On import aggregate no significant coefficients of the lagged variables are found, in the case of exports only the one- and two lagged variables of announcement show significance, meaning that the periods before the announcement have had an impact on export flows. This would indicate that the anticipation effects measured during announcement are not only explained by announcement but also the months before, which would indicate that the anticipation effects of announcement are not as strong as previously assumed.
When combining all the lagged affect in a regression the lagged-one variable shows an increase of 5% and lagged-two variable shows a -5% decrease variables representing the announcement of Cameron prove to remain significant as can be observed in X. For the laggedone variables the explanatory power becomes even stronger and grows to -17%.
5.3 Effects on Import and Export Flows Per Product Group
Table 5: Volume of Exports at Product Level By Period and Product Group
Panel A: Individual regressions for every unique combination of period and product group.
Missmanu Machinetransp Chemicals FDT Minfuel
Announcement (0.859) .0040 -.0648*** (0.000) -.0559*** (0.000) -.0637*** (0.008) (0.300) .0655
Dateset (0.955) .0009 -.0684*** (0.000) -.0889*** (0.000 ) -.0677*** (0.000) (0.811) .0113
Outcome (0.733) -.0057 -.0706*** (0.000) -.0908*** (0.000) -.0915*** (0.000) (0.092) -.0812*
Art50 -.0933*** (0.000) -.1330*** (0.000) -.1397*** (0.000) -.1790*** (0.000) (0.062) -.1301*
Negotiations -.2256*** (0.000) -.1204*** (0.000) -.0908*** (0.000) -.0915*** (0.000) (0.092) -.0812*
Panel B: Regressions for every product group with all periods included.
Missmanu Machinetransp Chemicals FDT Minfuel
Announcement (0.864) .0044 -.0256*** (0.005) (0.777) .0042 (0.401) -.0152 (0.291) .0766
Dateset (0.671) .0116 -.0259*** (0.008) -.0484*** (0.002) -.0596*** (0.002) .1527** (0.046)
Outcome (0.432) .0214 Omitted (0.184) -.0213 -.0531*** (0.006) -.1829** (0.018)
Art50 -.1139*** (0.000) -.1091*** (0.000) -.1016*** (0.000) -.0758*** (0.000 ) (0.204) -.1004
Negotiations Omitted (0.356) -.0090 Omitted Omitted Omitted
N 3,209,849 16,308,335 6,364,628 4,717,429 708,831
R-Squared 0.6438 0.4872 0.5039 0.4705 0.7121
Note: Standard errors are reported in parentheses. ***, **, * denote significance at 1%, 5%
and 10% respectively. The variables in this Table are explained in Appendix A.
The results at exports on product level show significant results for all critical dates of the product groups: Machinetransp, Chemicals, and FDT. Furthermore, anticipation effects around Art50 and negotiations show significant anticipation effects, though the strength of the coefficients is different. Trade in the FDT industry decreases with 18% around the critical date Art50 implying a strong loss caused by the Brexit. Also the product groups Chemicals and Machinetransp show strong decreases of 13% and 14% in export during Art50.
Table 6: Volume of Import at Product Level By Date and Product Group
Panel A: Individual regressions for every unique combination of period and product group.
Missmanu Machinetransp Chemicals FDT Minfuel
Announcement .0704** -.0356** -.0747*** -.0321 -.0457 (0.027) (0.015) (0.005) (0.344) (0.608) Dateset .0061 -.0561*** -.1224*** -.1247*** -.0150 (0.789) (0.000) (0.000) (0.000) (0.817) Outcome -.0513** -.0702*** -.1309*** -.1144*** .0188 (0.025) (0.000) (0.000) (0.000) (0.773) Art50 -.0062 -.0941*** -.1270*** -.1661*** -.1632* (0.851) (0.000) (0.000) (0.000) (0.091) Negotiations -.018* -.069** -.1304*** -.1970*** .0188 (0.081) (0.021) (0.017) (0.007) (0.294)
Panel B: Regressions for every product group with all periods included.
Missmanu Machinetransp Chemicals FDT Minfuel
Announcement .0863** .0015 .0075 .0655* -.0467 (0.017) (0.930) (0.805) (0.090) (0.646) Dateset .0800** -.0068 -.0578* -.1125*** -.0547 (0.035) (0.704) (0.074) (0.006) (0.605) Outcome -.1469*** -.0469*** -.073** -.0104 .1446 (0.000 ) (0.009) (0.025) (0.799) (0.167) Art50 -.003 -.0644*** -.045 -.099** -.207** (0.952) (0.001) (0.212) (0.034) (0.082) Negotiations .147** .0195 -.021 -.062 -.060 (0.038) (0.570) (0.734) (0.453) (0.773) N 2,315,046 11,144,179 4,141,118 2,936,987 531,975 R-Squared 0.6326 0.6111 0.6584 0.6594 0.7406
Note: Standard errors are reported in parentheses. ***, **, * denote significance at 1%, 5%
and 10% respectively. The variables in this Table are explained in Appendix A.
Again the data shows strong anticipation effects occurring in the product groups Machinetransp, chemicals, and FDT. Also all dates show high significant results. The anticipation effects are not strongly present in the product group Minerals and Fuels, meaning that this industry is not affected as strong by the Brexit as the other industries are. These findings are supporting the hypothesis 3a and 3b that state that import and export flows decrease during the prelude of the Brexit.
5.4 Revealed Comparative Advantage Model
Table 7: RCA Per Product Group. 201501 201506 201511 201604 201609 201702 201707 Miss manu 0.07 0.06 0.06 0,07 0,08 0,08 0.07 Manutransp 0.04 0.04 0.02 0.04 0.04 0.03 0.03 Chemicals 0.01 -0,01 -0,02 -0,01 -0,01 -0,02 -0,02 Fdt 0.03 0.02 0.01 0.02 0.02 0.02 0.02 Minfuels -0,05 -0,03 -0,05 -0,03 -0,04 -0,07 -0,03 Other 0.03 0.01 0.001 0.01 0.01 0.01 0.004
Note: the variables in this Table are explained in Appendix A.
The import and export flows on product level suggest that revealed comparative disadvantages are present in the industries that show strong decreases in their import and export flows such as FDT, Manutransp, and Chemicals. However, the results of the RCA model do not confirm these findings. As can be observed in Table 7 the products show stable revealed comparative (dis)advantages, leading to the conclusion that the prelude of the Brexit does not have an impact on the export and import flows of the Brexit through their revealed comparative (dis)advantages.
An overview of the hypothesis and their results can be observed in Table 8 below.
Table 8: Results Hypotheses
Expectation Results
H1a The total volume of Import between the UK and EU decrease during the
prelude of the Brexit.
The total volume of Import between the UK and EU does not decrease during the prelude of the Brexit.
H1b The total volume of Export between the UK and EU decrease during the
prelude of the Brexit.
The total volume of Import between the UK and EU does only decrease around the announcement.
H2
The Brexit does not affect the total volume of export between the EU and UK.
The Brexit does not affect the total volume of export between the EU and UK.
H3a
The total volume of export of each product group does decrease during the prelude of the Brexit.
The total volume of export of each product group does decrease during the prelude of the Brexit.
H3b
The total volume of import of each product group does decrease during the prelude of the Brexit.
6. Conclusion
This thesis investigated whether the Brexit affected EU and UK import and export flows. More specifically, this thesis sought to investigate whether anticipation effects occurred around relevant dates of the Brexit, which are expected to impact the probability of changing the trade agreement between the EU and the UK. The thesis made use of the gravity model of international trade, a regression model that includes multiple fixed effects, and four different datasets. With the model specification it is possible to correct for three sets of fixed effects and has, by increasing the internal validity, created a realistic image of the impact of the Brexit. The datasets on total import and exports flows only confirmed the existence of negative anticipation effects around the announcement by David Cameron of a Brexit referendum, in the exports dataset. The datasets focusing on the five industries separately show the existence of anticipation effects. Especially, strong negative anticipation effects occurred in the industries Machines and transports vehicles, Chemicals and related products, and Food, Drinks, and Tobacco. Also miscellaneous manufacturing and minerals and fuels showed anticipation effects, only less strong than the other three industries. These findings are contesting the results from the RCA-model, leaving the effects of the Brexit in the middle.
On product group level, the findings of this thesis are in line with other papers which suggest a negative ex-ante effect of the Brexit. The findings of this thesis show the presence of anticipation effects. Without taking these anticipation effects into account the negative impact of Brexit can be underestimated.
The contribution of this paper is relevant for policy makers since anticipation effects are suggested to increase the total impact of trade agreements. For researchers, this thesis is one of the first papers on anticipation effects of ending trade agreements, by using the latest econometrical adaptions in the Gravity Equation theory. So, on the one hand it contributes to the literature by introducing new literature for the event of ending a trade agreement for which the Brexit is a unique case. On the other hand, it contributes by extending the relatively scarce literature that makes use of time and country-pair fixed effects, and that makes use of stage variables as introduced by the most recent literature.
6.1 Limitations
A drawback of this thesis is the usage of aggregated data since aggregated data could lead to loss of information. Generally, aggregated data lowers the quality of estimation techniques too. Another limitation of this dataset is that it focuses on trade flows between the UK and EU. The research focuses on the revealed comparative advantage of the EU28 as a whole, meaning that the distinction between the EU and UK misses, which means that it is not clear in which countries the revealed comparative disadvantage on product level will exactly occur. Therefore, estimating the impact per specific country or per specific region is not possible.
Lastly, this study could not make a distinction between primary, manufacturing and service industries. The data was limited to the NC8 product groups, but a more detailed distinction within these product groups is most interesting in order to be more precise on the actual differences of impact on the different industries.
7.2 Recommendations
The first recommendation is to conduct a study related to the link between uncertainty and the behavior of firms, on the export and import flows. It would be interesting to know when people perceive the risk of the Brexit high enough to switch flows of production or investments, a degree of uncertainty that is not measurable in this thesis. Another recommendation is to solely focus on FDI when investigating anticipation effects of the Brexit. FDI is suggested to be influenced by the Brexit, and is also expected to have anticipation effects, when uncertainty regarding the Brexit outcome increases. It is possible that investors postpone their investments due to the increasing uncertainty. This makes that FDI is also an interesting variable to examine.
Furthermore, Soete and van Hove (2017) have shown that trade agreements have a different impact on EU member states. This means that heterogeneous effects are present. Consequently the anticipation effect could differ between EU member states. Soete (2017) has found that there are some member states of the EU that experience a substantially larger impact on their trade flows, caused by the anticipation effect of trade agreements. The different level of anticipation effects on EU members is out of the scope of this research and forms an interesting topic that might need further inquiry.
7. Bibliography
Abrams, R.K., 1980. International trade flows under flexible exchange rates. Federal Reserve Bank of Kansas City. Economic Review 65 (3), 3–10.
Adam, C. Cobham, D, (2007) modeling multilateral trade resistance in a gravity model with exchange rate regimes. Centre for Dynamic macroeconomic analysis conference papers 2007
Aisen, A. Veiga, F.J. (2011) How Does Political instability Affect Economic Growth? IMF Working paper, WP/11/12
Aitken, Norman D., 1973. The effect of the EEC and EFTA on European trade: a temporal cross-section analysis. American Economic Review 5, 881–892 (December).
Aizenman, J., and N. Marion. (1993). “Policy Uncertainty, Persistence and Growth.” Review of International Economics 1, 145–163.
Alesina, A., Özler, S., Roubini, N. et al. J Econ Growth (1996) 1: 189. https://doi.org/10.1007/BF00138862 Alesina,A.andPerotti,R.(1996).“Income distribution,political instability, and investment.” European Economic Review 40, 1203- 1228.
Armstrong, A. (2016), in Brexit Beckons: Thinking ahead by leading economists. Assembled by Baldwin, R.E. Chapter 5 ‘’ The UK’s new trade priorities,’’ p.53-p.63.
Awokuse, T. (2006). Causality between exports, imports, and economic growth: Evidence from transition economies. Economics Letters 94 (2007) 389–395, Elsevier.
Baier, S. L. and Bergstrand J. H. (2007). “Do Free Trade Agreements actually Increase Members’ international Trade?” Journal of InternationalEconomics 71.1, pp. 72–95.
Baier, S. L. and Bergstrand J. H, Feng, M. (2014). ‘’Economic integration agreements and the margins of international trade’’ Journal of International Economics 93 (2014) 339–350.
Baldwin, R.E. (2016), Brexit Beckons: Thinking ahead by Leading Economists, VoxEU.org eBook, CEPR press, London.
Balassa, B. (1965), Trade Liberalisation and “Revealed” Comparative Advantage1. The Manchester School, 33: 99–123. doi:10.1111/j.1467-9957.1965.tb00050.x
Bergstrand, Jeffrey H., 1985. The gravity equation in international trade: some microeconomic foundations and empirical evidence. Review of Economics and Statistics 67 (3), 474–481 (August). Brakman, S., Garretsen, J., & Kohl, T. (2017). Consequences of Brexit and Options for a "Global Britain".(CESifo Working Paper; No. 6648). Munich: CESifo.
Burnstein and Melitz (2011) Trade Liberalization and Firm Dynamics, NBER Working Paper No. 16960. Issued in April 2011, Revised in October 2011, Document Object Identifier (DOI): 10.3386/w16960 Campos, N. and Nugent, J. (2002). “Who is afraid of political instability?” Journal of
Development Economics 67, 157–172
Campos, N., F. Coricelli and L. Moretti (2014) “Economic growth and political integration: Estimating the benefits from membership in the EU using the synthetic counterfactuals method”, CEPR Discussion Paper No. 9968.
