The effects of protectionist trade measures: an Empirical Study on the Steel Sector

(1)

1

The effects of protectionist trade measures: An

Empirical Study on the Steel Sector

A.R. De Vos

UVA

Master’s Thesis

MSc Economics

Email: abel.de.vos@student.uva.nl

Student number: 11121998

Supervisor: dr. D.J.M. Veestraeten

Second supervisor:

drs.

N.J. Leefmans

(2)

2

Statement of Originality

This document is written by student Abel de Vos who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no other 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)

3

Abstract

This study provides an analysis of the effectiveness of protectionist measures in the US steel sector. We provide a history of US protectionist measures in the steel sector after which we perform an analysis on the effects of protectionist measures on raw steel production by US producers from 1961 until 2007. Moreover, we provide further analysis on the non-equal effects of quotas and tariffs on raw steel production. We find evidence of negative effects of tariffs on raw steel production as well as positive effects of quotas on raw steel production. We continue this analysis by providing evidence for non-equality between the effects of tariffs and the 2002-2003 safeguard on raw steel production. Finally, we find that it appears that Democratic administrations provide a more beneficial environment for the production of raw steel.

(4)

4

5.3 Economic growth ... 24 5.4 Democratic administration ... 24 5.5 Apparent consumption ... 24 5.6 Import penetration ... 24 5.7 Exports ... 25 5.8 Tariffs ... 25 5.9 Quotas ... 25 5.10 Safeguards ... 25 6. Empirical Analysis ... 26 6.1 Regressions ... 29 6.1.1 Safeguards ... 33 7. Conclusions ... 35 Bibliography ... 36

(5)

5

1. Introduction

Protectionism within the commodity sector is not a rare occurrence. In the United States steel has been a heavily politized product on which tariffs and quotas have been placed on many occasions. There is theoretical evidence for non-equivalent effects of quotas and tariffs, however empirical evidence remains scarce. The goal of this study is use an empirical model to research whether the theory matches real world data. This will be done from a US-market perspective in the steel sector.

Since World War II countries around the world have been working together on ways to promote a more interconnected world. Through trade liberalization the world saw a possibility to create interdependencies between countries as well as an opportunity to create and distribute wealth. By signing the General Agreement on Tariffs and Trade (GATT) a group of countries around the world created a platform to promote more liberal global trade policies. This proved to be the beginning for an environment in which global trade flourished. Economic unions and customs unions were formed, as well as the World Trade Organization (WTO). The number of bilateral, multilateral, free and preferential trade agreements has been growing steadily over the last decades. However, with Brexit approaching and lobbyist within the EU and US opposing certain trade agreements it seems as if protectionist policies are gaining ground again. The Trump-administration is already stating it will implement several measures in the coal and steel sector to ensure that the domestic production is protected against pressures from foreign markets.

Even though it seems as if protectionist measures have been gaining momentum recently, protectionist measures are not new topics and have been implemented by governments on different sectors for a long period of time. Some sectors experience more measures than others. Interest groups in grain sectors, mining sectors, and construction materials sectors often exerted their influence on administrations to put protectionist measures into place to protect domestic production from foreign competitors.

A commodity sector which has experienced a large number of such measures is the US steel sector. Historically the steel sector has experienced a fair amount of protectionist policies. For the larger part of the 20th_{century the US steel sector consisted of large plants supplying a large amount of}

jobs. Moreover, steel was essential during WWII for war efforts, as well as for post-WWII for rebuilding efforts. Even though the steel sector has changed since then, as the sector moved on to mostly smaller plants, the importance of the steel sector in the eyes of policymakers has not

necessarily declined. This is partially due to the efforts of interest groups which are trying to ensure the market power of US firms within the domestic sector remains the same. These efforts by interest groups stem from the fact that the US steel sector historically consisted of less competitive and less

(6)

6

innovative firms (Lenway, Morck, and Yeung, 1996). These firms were less competitive than foreign firms causing the domestic firms to lobby for protectionist measures to raise their competitiveness.

As a historically large producer of steel, US based production has by and large been stagnant over the past 4 decades (except a major fall during the financial crisis). In this period China has taken over as the worldwide leader in steel production. The US has had a sizeable amount of trade

regulations in this sector to make sure the market power of the domestic sector remains stable and US producers are not outcompeted by foreign producers. Due to the amount of protectionist measures in the US the focus of this study will lie on the US steel sector.

Interest groups within the steel sector have been pushing for measures to protect US steel production for decades. The US steel sector experienced many different trade protectionist measures since the late 1960s. As seen in table 1 which is gathered from Blonigen, Liebman, Pierce, and Wilson (2013), the US has implemented measures such as voluntary restraint agreements (VRAs),

antidumping investigations (AD), and countervailing duties (CVDs), and tariffs over several time periods since 1969. Due to the amount of protectionist measures over time in this sector it is clear that many different administrations have put in effort to protect the US steel producers from foreign steel producers.

The number of different measures implemented in the steel sector makes way for the question of effectiveness of the different measures. As multiple theoretical models (Harris, 1985; Krishna, 1989) suggest, there is theoretical evidence for the non-equivalence of quotas and tariffs on the degree of market power of domestic firms. Blonigen et al. (2013) test this theory in the steel sector with an empirical model which focuses on data gathered from individual steel producers and processors. Interestingly, the existing literature has not focused on the effects that quotas and tariffs have on steel production in the entire US steel sector, and more specifically the different ways these measures may affect domestic steel production. The existing literature has failed to provide a sector wide comparison between the effects of quotas on steel production in the US to the effects of tariffs on steel production. Therefore, the goal of this paper is to provide a clearer picture of the different effects that quotas and tariffs have on steel production in the US. The data that is used spans from 1961, around the time the first trade-policy measures were implemented, to 2007. 2007 as the final data point is chosen for several reasons. Firstly, trade-policy measures in the steel sector after this period have only been implemented around 2014, and since the available data only spans to 2014 the full effects of these measures are ambiguous. Secondly, 2007 is chosen to exclude the severe effects the financial crisis has had on the steel sector.

The structure of this paper is as follows. Chapter 2 describes the different protectionist measures have been used in the steel sector. Chapter 3 gives an overview of the nature of the US steel sector and provides a short history thereof. Chapter 4 reviews the existing literature on related topics

(7)

7

to the research question. Chapter 5 describes the empirical model and the choice of variables. Chapter 6 provides the analysis and the results of the study. Finally, chapter 7 presents a conclusion to the paper.

(8)

8

2. Protectionist measures

As can be seen in table 1, the application of protectionist measures in the US steel sector over time is quite broad. This chapter intends to cover the essential characteristics of these protectionist measures such as safeguards, quotas, tariffs, AD/CVD measures, and VRAs. It will also provide some theory regarding these measures and the differences in their use.

Table 1: US steel trade protection events (Blonigen et al., 2013). Date of implementation Type of US steel trade protection

1969-1974 Voluntary restraint agreements (VRAs) with Japan and the EC. 1978-1981 Trigger price mechanism that applied to all imports.

1982 Antidumping (AD) and countervailing duty (CVD) cases filed against EC countries. Subsequently terminated in favor of for VRAs on EC imports.

1984 AD and CVD cases filed against non-EC countries. Subsequently terminated for comprehensive VRAs.

1984-1989 Comprehensive VRAs with all significant sources of imports.

1989-1992 Extension of VRAs.

1992-1993 AD and CVD cases filed against significant sources of imports after VRAs expire. AD and CVD remedies applied to only subset of products.

1998-2000 Multiple AD and CVD cases against Japan and other Asian countries. 2002-2003 Safeguard remedies in form of tariffs placed on steel imports,

excluding FTA partners and developing countries.

2.1 Quotas and Tariffs

Quotas and tariffs are protectionist measures which can be placed on many individual or groups of products. Quotas are a quantitative restriction to the amount of imports of a certain product or product group in order to limit the imports of that product or product group. This is turn causes (foreign) supply in the country that imposes the quota to go down, giving more opportunities for the domestic sector to fill in the gap in the supply. This gap causes an increase in prices making way for larger profits in the domestic sector, as well larger domestic production. Similarly, tariffs cause a decrease in imports by way of an import tax on certain products. By implementing such an import tax, the price of foreign products increases, thus diminishing the demand for foreign products. Similarly to quotas, this is beneficial to the domestic sector as prices rise and thus the margins and production for domestic producers become larger.

(9)

9

2.1.1 Differential effects of Quotas and Tariffs

As quotas and tariffs are different measures, they can have different effects on the market. A simple model shows us that this is the case. Figure 1 shows two cases as described by Krugman and Obstfeld (2003). An assumption of a small country is made for simplicity, so that US producers have no influence on world prices. The left graph shows an importing economy where a tariff is placed on a product such that prices in the US are no longer at world prices (PW_{) but at price P}

1US. Similarly, in the

right graph a quota is put into place, such that the price of the product is at price P1US. Hence, in this

case of equivalent tariffs and quotas, US producers sell at the same price in both the case in the next period where a tariff is implemented, and the case where a quota is implemented. However, if a shock occurs, such as a demand shock where the demand curveshifts to the right and changes from D1 to D2,

there are different effects between tariffs and quotas. In the case of tariffs nothing changes for the US producers. Due to the assumption of a small country, all extra possible production due to the shift in demand is absorbed by foreign producers. This is because foreign producers are still able to supply products in the market as there is no restriction to the market, only higher prices. However, in case of a quota this is not the case. Here, foreign producers can only supply a limited amount to the US market. All extra demand cannot be supplied by foreign producers. Therefore, the demand shock is

transformed into a change of prices from P1US to prices at which the volume of imports equals the

quota (P2US). Thus, prices and production increase for US producers in case of a demand shock when

quotas are set in place, but not when tariffs are set in place. Figure 1: Effects of demand shock under tariffs and quotas

(10)

10

2.2 Voluntary Restraint Agreements (VRAs)

VRAs (also known as VERs, or Voluntary Export Restraints) are agreements between countries to place limits on the quantity of a product exported and imported between countries in a certain time period. VRAs are less intrusive and less flexible protectionist trade measures compared to many other traditional measures such as for instance embargoes, subsidies, and trade bans since they are voluntarily placed by the exporting country. VRAs are a method of protection for the domestic sector of the importing country, and are established in cooperation of the domestic country with the exporting countries’ producers. VRAs have been used often after the first 1947 GATT agreements. Since they were placed in cooperation with the exporting countries’ producers VRAs were seen as a grey-area measure, as opposed to some stricter trade measures of which the use was increasingly curtailed under this GATT agreement. VRAs are essentially self-imposed export quotas. In

cooperation with the importing country, the exporting countries’ producers agree to limit their exports to a certain amount. In turn the importing country refrains from imposing heavier trade measures on the country and/or producers. These agreements can thus be seen as voluntary, but are mostly put into place due to pressure from the importing country.

VRAs were seen as a way to get around the GATT agreements since they were placed by the exporting country. During the Uruguay round, spanning from 1986 to 1994, it was decided that

existing VRAs were to be phased out, and new VRAs were no longer to be put in place. This was done as the goal of the Uruguay rounds was to counter renewed protectionist policies and to promote further trade liberalization. It was thought that VRAs were compromising this goal. Due to the major

similarities with quotas, during the remainder of this study VRAs will be characterized as quotas.

2.3 Antidumping measures (AD) and countervailing duties (CVD)

Antidumping is a measure which can be taken to oppose situations in which a foreign producer exports a product below the ‘normal value’ that the foreign producer supplies that certain good for in its own domestic market. In this situation domestic producers can file for antidumping agreements in which the WTO determines what actions a country can take towards the foreign producer. Under the WTO rules antidumping procedures are initiated with a filing from the domestic sector towards its own government, which can be implemented if there is sufficient cause for the government to assume that a foreign exporter is dumping. However, the extent to which countries did and could apply antidumping measures did vary from country to country. Under the 1947 GATT agreement certain agreements were made such that antidumping measures could be challenged by the exporting countries, in order to prevent misuse. However, the rules agreed upon in the 1947 GATT agreements were ambiguous and often lacked in precision. During consequent rounds the antidumping agreement under which a country could pose antidumping measures were updated with respect to the procedures

(11)

11

to be followed, the implementation of countermeasures, determination of the cause of dumping, and furthermore a clarification of the role of the dispute settlement procedure.

Under the WTO’s agreement on Subsidies and Countervailing Measures the implementation of subsidies in the exporting country can be questioned, such that the WTO can rule if a subsidy should be withdrawn. Another option for a domestic sector is to file for a countervailing duty measure. Countervailing duties are duties that are to be imposed whenever a firm or an industry suspects that a foreign sector is unfairly subsidized by its government such that this foreign sector has the ability to undercut prices of the domestic producers, and thus outcompete the domestic producers with lower-cost imports. These duties are filed for at the domestic government, similarly to antidumping filings. Countervailing duties are duties which are imposed on a country-to-country basis, such that all producers in said country are equally hit whenever such a duty is imposed. As with antidumping measures, the protectionist use of subsidies is not easy to prove, as countries differ in the

determination whether a subsidy is actually imposed and whether it truly causes injury to the domestic sector. The fact that the determination is ambiguous can cause differences between countries regarding the rightfulness of the implementation of these measures. Countervailing duties are often spoken of in the same way as antidumping measures are. Both antidumping and countervailing duty measures are often filed at the same time for the same product in the same country.

So although there are many rules under which the WTO proposes that antidumping and countervailing duty measures should be determined, there is still enough room for countries to vary in their determination of rightfulness of a measure. These differences in determination often cause conflicts between countries, and when measures are imposed on certain countries, these countries may retaliate by imposing antidumping and countervailing duties on other products from the imposing country.

Both countervailing duties and antidumping measures are very similar to tariffs. Thus for the remainder of this study they will be characterized as tariffs.

2.4 Safeguards

Safeguards are a measure used to protect a domestic industry as an emergency action when an increase of imports of one or more products endangers the corresponding domestic production sector. The WTO allows safeguard measures under the threat of injury. This requires proving two

occurrences. First, injury: A significant impairment of the position of the domestic industry due to an increase in imports. Second, threat of serious injury: With a threat of serious injury there must be a significant threat to the position of the domestic industry, and this must be shown by facts and not merely based on conjecture or allegation.

(12)

12

A safeguard measure can be put into place by imposing either tariffs or quotas on products across that sector. This measure is allowed under WTO rules under the basis that the specific measures it imposes are placed on equal products from all countries on a most-favored-nation basis. This implies that all countries should be equal when imposing the terms. Safeguard (SG) measures were a measure proposed under the 1947 GATT. However, only under the Uruguay rounds in 1994 when the

Safeguard Agreement was created did it gain in popularity. Before the Uruguay rounds countries often preferred grey area measure such as VRAs over safeguard measures. However, in the WTO Safeguard Agreement grey area measures were prohibited under new WTO rules. Safeguard measures need to have sunset clauses, meaning that all safeguard measures must have a limited timeframe after which the safeguard measures are to be removed. This timeframe is 4 years, and can be extended if the situation still exists and the rules for applying a safeguard still hold. Furthermore, the Uruguay round imposed the necessity of an investigation by the WTO before a safeguard measure is approved, whereas this was not necessary under the previous GATT 1947 agreement.

Quotas under a safeguard agreement can be put into place, although it must be ensured that the level of the quotas do not go under the three latest representative (so pre- import shock) years, unless there is a real justification for it to be otherwise. Furthermore, there are rules allocating the

percentages of the quotas across the exporting countries.A tariff under a safeguard agreement can only be imposed to cover the extent of the damage under serious injury or expected damage under threat of serious injury.

(13)

13

3. Overview of the US steel sector

This chapter will provide a short overview of the history of the US steel sector. It will describe the main characteristics of the US steel sector, and furthermore it will provide an overview of historic trade measures in this sector.

3.1 A brief history of US steel production

In the early 20th_{century the US steel sector was booming. Technological advancements, a war,}

and rapid growth of domestic infrastructure caused substantial growth in the steel sector. These factors, in combination with a surge in demand of steel for buildings as well as large growth in the automobile sector caused the US steel sector to become world leader in the production of steel. During the ‘roaring’ 1920s US steel producers were responsible for about 40 percent of world steel

production. This rapid growth in the US steel sector was assisted by the discovery of iron ore deposits in ranges mainly in the mid-west. After a dip in production after the stock market crash of 1929 production did not rebound to pre-crash levels until 1939. At this point raw steel production in the US was about 61 million metric tons per year. After the US joined World War II production started to rise to accommodate the demand for steel for the US war efforts. Meanwhile, world production of raw steel declined from 172 million tons in 1943 to 101 million tons 1945 causing the share of US raw steel production to rise towards 72% of total world production. By this time the domestic production exceeded domestic demand in the US by large amounts, making the US a large exporter of steel.

In the beginning of the 1950s world production started to return back to pre-WW II levels, and steel production in the US continued to grow, partially due to efforts abroad to rebuild war-torn countries. However, as these countries were rebuilt and exports fell, imports grew. Transoceanic shipping costs fell, as did costs of exporting to the US. As import costs fell foreign-produced steel was a more viable option within the US, and the market share of imports in the US started to grow.

Whereas in 1960 the import share was only 4.7%, by 1968 this share had grown to 16.7% (Moore, 1996). Imports as well as domestic production continued to gradually grow. During the 1960s and 1970s raw steel production was booming, and in 1973 production peaked at 137 million tons. However, from around 1974 to 1986 the American steel sector experienced a massive depression. Caused in part by the OPEC oil embargo and the Iranian revolution prices fell rapidly and profits declined. The economic downturn caused a large part of steel workers to lose their jobs, as

employment in the US steel sector fell from 552 thousand in 1968 to 391 thousand in 1981, to 169 thousand in 1989 (Moore, 1996). This downturn also inspired a shift in production factors as well many trade measures to protect the domestic sector, such as a minimum import price put in place by President Carter from 1978 to 1981 as a means to increase prices and margins.

The steel sector never fully recovered from this decade-long downturn, and it started to lose out as the world production leader. Whereas the post-WW II period already caused shifts in the

(14)

14

percentage of world steel production by the US steel sector, this downturn helped bring the share down even further, to about 10 percent in 1982. US steel production went down from about 50% in 1946 to 20% of world production in 1973. However, due to shift in production mills the domestic production sector did manage to refocus from mainly major integrated mills towards a sector with more niche filling production, through usage of minimills. Through a more diverse sector, the steel sector managed to be relatively stable after this crisis. The domestic production of raw steel remained stable around 80-90 million tons until 2008.

3.2 Steel mills in the US

After World War II the bulk of US steel production originated from major integrated mills. These mills were for the large part responsible for domestic production. Whilst foreign producers were gaining ground, domestic producers maintained their control over the domestic market. In 1979 only 8 major integrated mills had a 64% share of total production in the US market (table 2). Major integrated mills were mainly used since they have large scale economies as all processes from the creation of steel from melting and blasting of iron ore to the rolling of steel and creation of ingots take place in the mill. However, around that time major integrated mills required a minimum production level of 7 million tons of steel per year in order to be efficient (Barnett and Crandall, 1993), and the large investments required to build such a mill proved to be a barrier to massive investments. This was also one of the reasons why during this time producers started to shift away from major integrated mills and started to invest in other types of mills. Minimills and reconstituted mills started to gain in market share, and in 1991 the market share of major integrated mills fell to 34%, while minimills’ and reconstituted mills saw their share grow to 25% and 24% respectively (see table 2). Reconstituted mills’ share grew after major integrated mills where sold off in parts to diminish costs. By splitting up the parts producers aimed to lower costs. Reconstituted mills were thus simply integrated mills split up in part, and operated by different smaller firms.

The shift was also due to a rise in investments in minimills, in which raw steel is not created from raw materials such as iron ore, but rather is created through melting scrap steel and then immediately recasting it into steel products. This was a more efficient way or forging, since it was a continuous process instead of first creating ingots and then processing. Continuous casting made it such that there were much less storage costs. Therefore, minimills had significant cost efficiencies over integrated mills. Nowadays technological progress has made it such that integrated mills can also continuously cast raw steel, and 99% of all steel in both minimills as well as other mills is cast

continuously. Also, these minimills not only do not need expensive coke ovens and blast furnaces, but they also have no significant need to locate next or near to raw material supplies such as coal and iron. Moreover, they have a smaller workforce than integrated mills. Combine this with the fact that these mills are more geographically dispersed throughout the country, communities do not solely rely on the producers to be the main supplier of employment. The rise of the minimills has created a much more

(15)

15

competitive sector with many more competitors in the US and a sector in which economies of scale play a smaller role than before.

As of 2015, according to USGS about 63% of steel is produced in electric arc furnaces, the type of furnace mostly used in minimills, as opposed to 39% in basic oxygen furnaces, which are used mostly in integrated mills. In 2016 approximately 108 minimills were up and running in the US, as well as 11 integrated steel mills. So minimills have gradually acquired a larger market share since 1991 compared to integrated mills. However, integrated steel mills still play a role in the creation of raw steel, albeit smaller than before.

Table 2 : Estimated Market Share of US participants (Moore, 1996)

1979 1991

Type Number

of firms

Shipments1_{Share (%) Number} of firms

Shipments Share (%)

Major integrated mills 8 73.4 64 5 60.3 34

Reconstituted mills 0 0 0 15 22.4 25

Other traditional mill 20 17.7 15 6 3.5 4

Minimills 48 8.2 7 52 21.3 24

Specialty steel mill 10 1.0 1 9 1.5 2

Domestic total 100.3 87 79 89

Imports 17.5 15 15.7 19

Exports 2.8 2 6.5 7

Total Market 115 100 88.2 100

1_{In metric millions of tons.}

3.3 Trade-policy measures used in the US Steel sector

As transoceanic transport started to become less costly for producers of steel, international trade in steel was starting to get easier and more viable. However, with international trade opening up towards the US, the large US steel producers had now started to face competition from producers from mostly the EC and Japan. Collectively, the large US producers argued through the American Iron and Steel Institute (AISI) that relief was required. The initial goal of these producers was to obtain a temporary tariff increase to counter the effects of relatively lower wages abroad, government subsidization for foreign steelmakers, and national security (US government, 1968).

Around September 1967 the idea of temporary tariffs was abandoned and in cooperation with British Steel Corporation officials the idea of voluntary export restraints was explored. Moreover, at that time a bill was proposed to induce a limit to steel imports to 9.6% of the US market, after which

(16)

16

Japanese and German producers agreed to VRAs if the US would shelve the bill as well as persuade other countries to also agree to VRAs.

In the end a VRA was agreed upon by the European Coal and Steel Community (ECSC) and Japan. This VRA stated that a total of 14 million tons of imports for 1969-1972 was agreed upon, and both Japan and the ECSC were given 41% of the total, about 5.75 million net tons in the first year, with an 5% increase each year. The initial VRA would last 3 years. The remainder of the suppliers, mainly the UK and Canada were allocated the remaining percentage of 18%. In 1972 the VRAs were extended for another 2 years, and unlike before, the UK was included (the UK joined the ECSC). A quota of about 8 million tons for the ECSC was agreed upon, and the quota for Japanese producers was about 6.5 million tons per year. The growth percentage for the coming years was however reduced to 1-2.5% a year. Interestingly enough, the quotas for both Japan and the ESCS were not fully fulfilled in almost all of the 1972-1974 period, whereas in 1971 the quota were exceeded by both.

After the 1969 VRAs ended a notable crisis occurred in the US steel sector, although not necessarily related to the ending of the VRAs. After pressure from the US steel sector President Carter implemented a trigger price mechanism (TPM) in 1978 to counter unfair practices from steel

producers abroad. Under the TPM the domestic steel sector agreed to refrain from issuing AD and CVD petitions. The idea was that the TPM could detect imports of steel at unfairly low prices, and implement relief for the US sector (Anderson, 1982). The establishment of the TPM was to create reference prices for steel imports. Any sale below a certain price would trigger an investigation, for instance it could trigger an anti-dumping investigation. Three main ideas were behind the TPM. First, it was to facilitate anti-dumping investigations. Second it was to create higher import prices as well as lower import volumes. Third, it was to avoid larger protectionist measures from being lobbied for by interest groups. In the end, the TPM failed because foreign producers found multiple ways to avoid the TPM. The TPM was flawed, as foreign producers mislabeled products and increased quality to subvert the TPM. Furthermore, it was difficult to assess the true quality of the goods and the true transaction payments. The failure of the TPM did not go unnoticed, and the US steel producers again initiated the filing of CVD and AD protection against ECSC countries in 1982. This caused the TPM to be

disbanded. In turn, VRAs for ECSC countries were negotiated by President Reagan in order to avoid trade frictions from the CVD and AD protection filings, which were disbanded due to the

implementation of the VRAs. The VRA that was negotiated contained a 5.5% market share limit of ECSC exports to the US market and it was to expire in the beginning of 1986. The fact that these VRAs only targeted ECSC countries caused trade diversion, since non-ECSC countries could simply fill in the gap left behind by the VRAs. Two years later similar events occurred, but this time for non-ECSC countries. Yet again AD and CVD procedures where filed, this time for non-non-ECSC countries. Subsequently, comprehensive VRAs with all significant import sources were put in place. This

(17)

17

happened under heavy pressure from the steel sector on President Reagan. He complied and agreed to put the VRA in place as his reelection was coming up.

The VRAs between 1984-1989 were put into place for all major steel exporters, and intended to limit imported steel to 18.4% of the domestic market. Furthermore, a quota of 1.7 million tons for semi-finished steel was put into place. One of the changes compared to the 1969 VRAs was that quotas were placed on both a product- and country-specific basis. Both supply diversion and product upgrading were countered this way, since these methods to circumvent the import restrictions had been a nuisance in the 1969 and the 1982 VRAs. Furthermore, net cash flow was to be invested in the US domestic steel plants to upgrade outdated techniques and to let them become more competitive compared to mainly the ECSC and Japanese producers. The implementation of the VRAs combined with several concessions of the workers unions and increased investments in mills caused the

competitiveness of the steel sector to rise, such that the imports’ percentage share of imports decreased from 26.4% in 1984 to 20.4% in 1988. Since the implementation of the 1984 VRAs were seen as a success story in the US steel sector, at the end of the period the sector desired to let the VRAs be extended.

As the elections were yet again approaching, and Presidential Nominee George H.W. Bush was behind in the polls, in an effort to gain votes Presidential Nominee Bush stated his support for the renewal of the VRAs. However, unlike before the earlier VRA negotiations this time the effects of a potential VRA on steel consumers was analyzed as well. An estimate of a weighted average price increase varying between 0.6% to 1.6% per year during the 1984 VRAs was made, clearly damaging the steel consumer sectors. Furthermore, a lobby organization named CASUM, headed by Caterpillar, Inc., argued that their exports were harmed due to the VRA. They argued that US producers of final products which used steel in their product had to purchase semi-finished or raw steel products at higher prices due to the VRA. Due to these higher prices US producers, which were mainly small businesses, were less competitive than foreign producers. As President Nominee George H.W. Bush won the elections, the VRAs were extended as promised. However, instead of a 5 year extension, a 2.5 year extension was put in place. Also, the Bush administration allowed for 1% increases per year for countries willing to end trade-distorting steel policies. Also, with this extension the final date no longer allowed for another extension. The final date was to be the end date of the VRA.

After the expiration of the extension of the VRAs, AD and CVD cases were yet again filed again important import sources. This time, President George H.W. Bush refused to extend, as promised during the previous extension. President Clinton was elected in 1992, and AD and CVD filings continued. Most of these AD and CVD filings were indeed affirmed, and preliminary duties were imposed. However, a large part of the claims was rejected.

(18)

18

During the larger part of the 1990s AD and CVD cases were filed on limited amounts of products. This changed in 1998, when currency crises in Asia led to cheaper East Asian steel, and this led to renewed AD and CVD filings. Due to the diminishing influence of the steel sector on policy makers, a large number of bankruptcies occurred in the sector. The smaller influence was mostly due to the rise in minimills, which were smaller than the larger integrated mills. These minimills were more abundant, and coordination of producers thus was harder to realize. Due to the number of bankruptcies President George W. Bush implemented a safeguard consisting of tariffs throughout the sector in 2002. President Bush opted for tariffs varying between 8% to 30% on various products to protect the domestic sector from more bankruptcies due to foreign competition. As stated by the WTO rules, a safeguard can only be implemented as an emergency action when unforeseen increases in imports damage or threaten to damage the domestic production sector, and can only be implemented on a most-favored-nation basis. President George W. Bush implemented these measures not only when imports had been relatively stable for a long period, but also distinguished between countries and made exceptions from the safeguards on countries such as Mexico and Canada. The WTO

therefore ruled that the safeguard should not be allowed and ruled against it, after which the safeguards were dropped.

(19)

19

4. Review of the existing literature

The amount of different measures implemented in the steel sector makes way for the question of effectiveness of the different measures. As multiple theoretical models (Harris, 1985; Krishna, 1989) suggest, there is theoretical evidence for the non-equivalence of the effects of quotas and tariffs on the degree of market power of domestic firms. Harris (1985) sets up an analysis on the effects of a VRA in an oligopolistic setting. In this setting there is one domestic firm and one foreign firm. Harris (1985) finds that by setting a form of price leadership on the domestic firm, VRAs (quotas) can be set at trade level of exports and improve both foreign as well as domestic firms’ profits. The free-trade level of exports would have been the level of exports for the foreign firm without any restrictions to trade. The improvement for both the profits of the domestic firm and the foreign firm is established in this model by including a price leadership for the domestic firm which causes a Stackelberg-competition, and in turn this can cause profits to go up for both firms. This means that in an oligopolistic setting, with one domestic firm and one foreign firm, a VRA does not have to be set below free-trade levels of exports to be beneficial, be it under certain plausible elasticity conditions. Larger profits can be made by setting a price-leadership on the domestic firm, under the constraint that the VRA imposes a restriction to the foreign firm such that the foreign firm cannot be a strategic price setter. The VRA causes a form of collusion beneficial to both foreign and domestic firms, and as Harris (1985) states, this trade policy essentially counters anti-trust policy. This theoretical study therefore suggests that imposing a VRA (which is essentially a quota, as mentioned in the second chapter) can cause the domestic firm to become a price-setter, essentially creating a Stackelberg leadership for the domestic firm. However, the model further finds that under certain price-elasticity conditions a VRA on the level of free-trade of exports for the foreign firm, a fall in market share for the domestic firm occurs. This would not occur in a setting in which a tariff is imposed. This presents a reason why foreign firms should accept a VRA in the model of Harris (1985).

Krishna (1989) also looks at an oligopolistic market, and at setting a VRA at free-trade levels. She finds that when the foreign products are substitutes for the domestic products, a VRA acts as method of collusion as foreign firms are competing less with the domestic firms within the domestic market, similar to the set-up in Harris (1985). Krishna (1989) also finds that due to the effects of quotas on strategic interaction, quotas are different from tariffs. Tariffs do not affect strategic

interaction between firms, whereas quotas cause a shift in market power. The model used by Krishna suggests 3 separate differences between quotas and tariffs. First, prices are lower under a tariff than under a VRA set at free-trade levels. Secondly, profits of the domestic firm are higher under a VRA compared to tariffs. Thirdly, similarly to Harris (1985), Krishna (1989) finds that the foreign firm should prefer a VRA set at free-trade levels over no VRA at all, as profits will be higher for the foreign firm as well. These foreign firms’ profits are higher even compared to the case with a tariff

(20)

20

implemented in which the tariff revenues are lump-sum transferred to the foreign firm. So Krishna (1989) essentially obtains similar results to Harris (1985) for the case in which the products produced by the domestic firm and foreign firm are substitutes. The model of Krishna (1989) further suggests that an import equivalent export tax imposed by the exporting nation would be preferred over a VRA for the importing country, as anti-competitive effects would be mitigated, as well as the fact that prices would be lower and foreign profits would be lower. This import equivalent export tax would be an export tax set by the foreign country in order to reduce imports into the domestic country’s market to where it is equal to the VRA-level of imports. Thus, in the case of an oligopolistic market with products which are substitutes tariffs would be preferred over a VRA.

Blonigen et al. (2013) test the theory of non-equivalence between quotas and tariffs in the steel sector with an empirical model that employs data gathered from individual steel producers and

processors. One of the views this empirical study lacks is a view on the steel sector as a whole. The study uses panel data from US steel plants as well as a control group of non-ferrous metal plants from 1967 until 2002. The study uses a price-cost margin (PCM) proxy model as a target variable, which is a variable that essentially describes the profit of a single firm at a single point of time. It is defined as follows:

𝑃𝐶𝑀𝑖𝑗𝑡=

𝑆𝑎𝑙𝑒𝑠𝑖𝑗𝑡− 𝑀𝑎𝑡𝑖𝑗𝑡− 𝐿𝑎𝑏𝑖𝑗𝑡

𝑆𝑎𝑙𝑒𝑠𝑖𝑗𝑡

This is done for plant i for product j in year t, where Mat is material and energy costs, Labis labor costs, and Sales are sales. It is called the price-cost margin proxy model as it observes variable costs, but not marginal costs directly. Blonigen et al. (2013) use this target variable and regress a set of trade measures on it. Furthermore, a set of control variables is included. Using this regression set-up Blonigen et al. (2013) find that quantitative restrictions such as quotas and VRAs significantly increase market power. The study does not find significant increases in market power for tariff-based policies. The results hold for cases in which the trade measures are installed on products which are substitutes for domestic products, whereas it does not hold for the control group, in which the foreign products are complements to domestic goods.

Blonigen et al. (2013) thus perform a study by using individual firm data in the steel sector and focus on the differences between the effects of tariffs and quotas on market power in the domestic sector by regressing trade measures on the PCM variable. Thus the study examines the effects of trade measures on specific firms through the use of the PCM model. However, as this study looks at individual data there is room for a broader type of study, i.e. for the entire steel sector. One could argue that these effects can be seen as sector-wide effects, however this neglects the effects of quotas and tariffs can have on the interaction between domestic firms. Some of the benefits domestic firms

(21)

21

may have had could have been obtained from other domestic firms. For instance, some of the benefits could be due to steel mills in the US stopping production during the time frame of this study and thus creating a possibility for the existing (and in this study the examined) firms to benefit in terms of demand, prices and production possibilities. Furthermore, no existing studies have focused on a broader view of the difference in effects of trade measures in the US. Therefore, the empirical study in this thesis will focus on statistics from the steel sector in its entirety. Furthermore, a restriction of the study of Blonigen et al. (2013) is that the study uses panel data provided by the US census bureau in 5-year intervals. Since the protectionist measures they use do not necessarily cover every 5-5-year interval there is an issue, as it does not fully capture the year-to-year effects of a measure. Therefore, for more comprehensive results this study will use year to year data.

Looking at table 1, it is clear that the steel sector’s protectionist measures consists mainly out of AD, CVD, SG measures and VRAs. There are some empirical studies that examine AD, CVD, and safeguard procedures (sometimes also known as temporary trade barriers, or TTB). Leidy (1997) looks at the correlation between AD and CVD procedures and macroeconomic activity and finds that there is evidence for an increase in the presence of protectionist measures during macroeconomic downturns. However, Leidy (1997) does note that there is a possibility of changes to the study’s findings due to reforms during the Uruguay round. Taking this study into account, the empirical model in this study should examine whether the macroeconomic situation affects the amount of measures in the steel sector.

When focusing on the effects of VRAs in the steel sector there are several interesting studies. Canto (1984) focuses specifically on VRAs in the steel sector in the US from the 1950s to the

beginning of the 1980s. The focus lies primarily on the VRAs in the period between 1969 to 1974, as that was the only time during the research period in which VRAs existed (see table 1). It found that even though there was a direct decline in the volume of imported steel, foreign producers of steel responded by altering the mix of steel products to a higher standard. By altering the mix of steel products Japanese and European steel producers shifted focus towards higher-value products. This consequentially caused a decline in the rate of returns on these higher-value steel products for US producers. Only lower-value steel producers profited from this VRA. Furthermore the value of imports remained the same, even though the volume of imports declined. Canto (1984) also finds that the decline in the number of US steelworkers in the US was not halted by this measure either. So, the effects of this specific VRA seem to be negligible and potentially even harmful for the domestic sector. In relation to the goals of this study this could be an interesting perspective, as the regression may confirm (or state the opposite) these findings for VRAs in later years within the steel sector.

Bown (2013) gives another interesting perspective on the difference between AD/CVD and SG procedures in the US steel sector. Whilst comparing different trade impacts between AD/CVD

(22)

22

measures between 1989-2003 and the 2002 SG procedures he found that there were significant differences of the effects on imports between AD/CVD and SG. Furthermore, the AD/CVG measures had a much larger negative impact on imports into the US than the SG. So, according to this study the impact on imports is dependent on the measure chosen. However, this study still lacks any impact analysis on domestic production, and only analyses differences in imports. This can only give an incomplete indication for the effects of these measures on domestic production.

The existing literature provides us with interesting research into the theoretical and empirical effects of different trade barriers. The existing literature which focuses on the steel sector fails to discuss effects in terms of domestic production. Whether these different trade barriers affect US steel production remains to be examined. Therefore there is a clear opportunity for this thesis to research whether or not different trade measures have a significant and/or different effect on domestic production.

(23)

23

5. The Empirical Model

In this section the empirical model including its variables will be discussed. The following model will be used for the study.

𝑅𝑎𝑤 𝑆𝑡𝑒𝑒𝑙 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛𝑡

= 𝛽0+ 𝛽1∗ 𝐸𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝐺𝑟𝑜𝑤𝑡ℎ 𝑡+ 𝛽2∗ 𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝑔𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝑡+ 𝛽3

∗ 𝐴𝑝𝑝𝑎𝑟𝑒𝑛𝑡 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑡+ 𝛽4∗ 𝐼𝑚𝑝𝑜𝑟𝑡 𝑝𝑒𝑛𝑒𝑡𝑟𝑎𝑡𝑖𝑜𝑛 𝑡+ 𝛽5∗ ∆ 𝐸𝑥𝑝𝑜𝑟𝑡𝑠 𝑡

+ 𝛾1∗ 𝑇𝑎𝑟𝑖𝑓𝑓𝑠 𝑡+ 𝜖1∗ 𝑄𝑢𝑜𝑡𝑎𝑠 𝑡+ 𝜀𝑡

In the following section the choice for these variables will be explained, and possible issues within the empirical results are noted.

5.1 Data

The data for this paper is gathered from several sources. The main dataset which contains the data on raw steel production, apparent consumption, import penetration and US exports is gathered from Kelly and Matos (2014). This rapport is published by the United States Geological Survey (USGS). Raw steel production, apparent consumption, and US exports are provided in millions of metric tons. Further data for economic growth is gathered from the IMF, and data on the use of quotas and tariffs is gathered from Blonigen et al. (2013). For determining US administrations over time the internet is consulted.

5.2 Raw Steel Production

The choice for raw steel production over raw steel prices as the dependent variable is made for several reasons. First, data for raw steel production in the US can more easily be obtained for longer periods. Moreover, all production and consumption data are obtained from one dataset which causes the data to be less ambiguous and more consistent compared with data that are gathered from multiple sources. Second, raw steel production can be more easily compared to both world production as well as foreign production. This is the case since the specialized form of the US steel final products are less comparable to foreign products than raw steel. Furthermore, raw steel production data can also be compared to actual consumption data. Market shares of US raw steel production in both the world market and the domestic market can be gathered, and any conclusions on whether import, tariffs and or quotas impact these market shares can be more easily made. Another advantage is that other data which are to be used for other variables, such as apparent consumption data, import penetration, and exports can be gathered from the same source, namely from the USGS. As raw steel is now often continuously cast, therefore it is defined by the American Iron and Steel Institute (AISI) as steel in the first solid state after melting, when it is suitable for rolling. In this study raw steel production is

(24)

24

determined as all steel which is produced from basic materials. This includes scrap steel, as well as raw materials. This term excludes all semi-finished and finished steel products.

Looking at the raw data, there seems to be a possibility for stationarity, and therefore there must be a test to dismiss the potential presence of a unit root. This will be done by using the Augmented Dickey-Fuller test in chapter 6.

5.3 Economic growth

This variable is to be included in the equation as economic growth is expected to influence steel production positively as economic growth can create growth in steel production over time. The argument can be made that a stronger economy creates a better environment to produce steel in the US. The data on economic growth in the US are gathered from the IMF database.

5.4 Democratic administration

This variable is to be included as a dummy variable, that is 1 in periods in which the US was led by a Democratic government, and 0 in the periods in which it was led by a Republican

administration. Looking at the past, the mining and the steel sector traditionally have received more support by Republican administrations. This therefore can be an interesting variable to include in this regression. This data is easily accessible.

It has to be noted that there can also be multicollinearity between this variable and tariffs and quotas, as it is probable that Republican administrations tend to implement more protectionist trade measures. This should be tested. This data is gathered from the internet.

5.5 Apparent consumption

Apparent consumption as a variable is to be included, as one would assume larger demand in the US would cause larger steel production in said country. Apparent consumption is defined by the USGS as raw steel production minus exports plus imports plus or minus stock changes. This includes the imports of semi-finished steel products. Thus apparent consumption is not directly observed, and calculated through other observed data. From this point on, this paper considers consumption a synonym for apparent consumption. Possible multicollinearity with economic growth could occur. This data is gathered directly from the USGS data source.

5.6 Import penetration

Import penetration is defined as imports as a percentage of total domestic steel consumption, in this thesis total apparent steel consumption. One would assume in this case import penetration affects local production as a high level of penetration reduces local production, as the market power of domestic producers is lower with high penetration, and thus any changes in demand can be picked up more easily by foreign producers. This data is constructed from data from the USGS data source.

(25)

25

5.7 Exports

Exports can have a direct effect on the production of raw steel in the US. Any shocks or changes in foreign demand of US steel imply an increase or decrease in US steel production.

Excluding this variable could lead to the omitted variable bias. This data is gathered directly from the USGS data source.

5.8 Tariffs

This variable is to be included as a dummy variable. In years in which tariffs were in place this variable is to be 1, and years in which they have not been implemented this variable is to be 0. This together with the quotas variable will be the most interesting variable for the main research question. These two variables will be used to determine whether implementing protectionist measures truly affect the production of steel. Looking at the sign and t-statistic of these two variables can determine whether there are truly differences between protectionist measures and whether tariffs and quotas have any effect on the production in the steel sector. This data is constructed from the collection of tariffs in the steel sector as seen in table 1.

5.9 Quotas

This variable is to be included as a dummy variable. In years in which quotas have occurred this variable is to be 1, and years in which they have not been implemented this variable is to be 0. This data is constructed on the basis of table 1.

5.10 Safeguards

Furthermore, an interesting inclusion could be to separate AD and CDV measures from SG measures as Bown (2013) showed there were differences in the effectiveness between these measures. This could lead to the introduction of an extra dummy variable for safeguards, separately from tariffs, in which AD and CDV are included.

(26)

26

6. Empirical Analysis

Before performing any regression, taking a look at the raw data can be a useful way to detect issues that could be present in the dataset. The raw steel production data from 1961 to 2007 does present some potential issues with the data. The first issue seems to be a large negative shock in the data around 1982. Interestingly, this large negative shock seems to coincide with the ending of the TPM, and the renewed AD and CVD filings by US producers in 1982. In figure 2 the drop in raw steel production in 1982 is clearly present in the data. The average production during the first period, starting in 1961 and ending in 1981, is denoted as Average Pre-1982. For the second period, starting in 1982 and ending in 2007 the average production is denoted as Average Post-1982. Comparing these two averages, raw steel production pre-1982 appears to jump down from about 115 million tons of raw steel each year to about 90 million tons every year post-1982. This seems to be suggest a ‘structural break’ in the data.

Figure 2: Raw steel production data in metric tons (Source: USGS)

The data first has to be tested for a unit root. Therefore the first test performed on the raw steel production data is the Augmented Dickey-Fuller test (ADF test). This test is done under the null hypothesis of a unit root in the data. For the raw steel production data there seems to be no apparent trend upwards or downwards (see figure 2). The data seems to be centered around the mean (albeit two means due to an apparent structural break). Thus an ADF test without any trends should be performed to test for a unit root. The results of two ADF tests are shown below. One without a lag, and one with a one-period lag. The test statistics from the ADF tests indicate the presence of a unit root in the data, since in both tests the null hypothesis of unit root is not rejected. Furthermore, a Kwiatkowski– Phillips–Schmidt–Shin (KPSS) test is rejected under H0. The H0 under this test is that the raw steel

0 20.000.000 40.000.000 60.000.000 80.000.000 100.000.000 120.000.000 140.000.000 160.000.000

(27)

27

production data is stationary. Since this H0 is rejected at the 5% level, this provides further evidence

for a unit root in the raw steel production data. Table 3: Unit root test results

A major issue with both the ADF and KPSS tests is that possible structural breaks are not accounted for within the test. As a result, these tests tend to overstate the occurrence of unit root when one or more structural breaks are present in the dataset. As mentioned before, in the case or raw steel production data eyeballing presents evidence of such a break in the data. Furthermore, when

performing ADF tests on the data points before 1982 and the data points after 1982 separately the null Augmented Dickey-Fuller test for unit root results

(No lags)

Number of observations = 47 Z(t) test statistic = -2.322 Z (t) 5% critical value = -2.938 Z (t) 10% critical value = -2.604 MacKinnon approximate p-value for test statistic p-value = 0.1649

Augmented Dickey-Fuller test for unit root results (One period lag)

Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test results

Maximum lag 5% test statistic = 0.146 Test statistic for maximum lag = 0.14 Maximum lag chosen by Schwert criterion Maximum lag = 3

Augmented Dickey-Fuller test for unit root results (No lags, pre-1982)

Augmented Dickey-Fuller test for unit root results (No lags, post-1982)

(28)

28

hypothesis is rejected at the 10% level (see table 3). This is another indication that the structural break in the data may have caused the ADF and KPSS tests to return test statistics that are not accurately representing the data.

A method to test whether this apparent structural break is actually causing the ADF and KPSS tests to be inaccurate is the Zivot-Andrews unit root test. This test not only tests for a unit root in the data, but it also simultaneously tests for the presence of a structural break. Under the Zivot-Andrews unit root test’s null hypothesis there is a unit root present while a structural break is accounted for. Thus, in the alternative hypothesis there is a stationary process with a structural break in either the trend, intercept, or both. Another advantage of the Zivot-Andrews unit root test is that it provides the minimal ADF test statistic (thus the most negative test statistic). In the Zivot-Andrews unit root test, ADF test statistics for all data points are calculated and the minimal ADF test statistic provides the point of one (of maybe multiple) structural break(s) (Zivot and Andrews, 1992).

Table 4: Zivot-Andrews unit root test results

Since there is no visual evidence for any trend in the data, an alternative hypothesis of a structural break in the trend is not examined. Therefore, a Zivot-Andrews unit root test is performed with an alternative hypothesis of a stationary process with a structural break in the intercept. As shown in table 4 the test provides evidence for a stationary process with a structural break in the intercept. The results show that the rejection of the null hypothesis is at the 1% level. Furthermore, the test finds the minimal ADF test-statistic at the 1982 data point, indicating that indeed the structural break is at the ‘eyeballed’ 1982 data point. The Zivot-Andrews unit root test thus provides enough evidence that a structural break took place in 1982 which distorted the results of other unit root tests. It furthermore provides evidence that if a structural break in the data is accounted for that there is no unit root present. Since there has been no indication of another structural break, and the unit root is already rejected when the structural break in 1982 is accounted for, no further tests to check for other structural breaks are necessary.

Zivot-Andrews unit root test results (Allowing for break in the intercept)

Lags included = 0 (Through BIC) 1% Critical value = -5.340 Minimum t-statistic = -5.808 Established year for minimum t-statistic Year = 1982

(29)

29

6.1 Regressions

Table 5: Regressions on raw steel production

Regression

(1)

(2)

(3)

(4)

R

2 _0.7324 _0.8652 _0.8644 _0.8670

F-test value

25.98 41.09 47.28 72.39

Growth

0.4860 0.7216 1 0.6732 0.6270 0.6973 0.6487 0.6425 0.6586

DemAdm

0.8011 3.6865 3.371 2.678 3.678 2.331 4.411 2.452*

Consumption

0.8987 0.1753*** 0.5160 0.1863*** 0.4814 0.1422*** 0.4950 0.1438***

ImPen

-149.8 16.03*** -12.42 28.15

Exports

0.3406 0.7568 1.294 0.5693** 1.345 0.5303** 1.267 0.5350**

Tariffs

-2.930 3.603 -0.9225 1.541 -1.083 1.563 -2.610 1.395*

Quotas

2.350 3.674 7.425 2.424*** 7.784 2.308*** 8.406 2.414***

Breakdummy

-26.80 4.904*** -28.58 2.292*** -28.41 2.333***

Safeguards

2.456 2.191

Constant

41.82 13.63*** 57.83 12.75*** 59.08 11.08*** 57.74 11.26*** 1

Standard errors are noted below the coefficients. Significance levels are noted as * _{at the 10% level,}**

at the 5% level, and *** _{at the 1% level.}

In regression (1) in table 5, raw steel production is regressed on the explanatory variables growth (Growth), Democratic administrations (DemAdm), apparent consumption (Consumption), import penetration (ImPen), exports (Exports), tariffs (Tariffs), and quotas (Quotas). All regressions in this table are performed using robust standard errors. This is done to counter any heteroskedasticity within the regressions. Regression (1) seems to give some interesting results. Neither the coefficients

(30)

30

for tariffs or quotas seem to be significant. Furthermore, there is no evidence that economic growth and the presence of Democratic administrations in the US have a significant effect on raw steel production. Consumption and import penetration have a significant effect on raw steel production at the 1% level. Consumption has a positive correlation with raw steel production with a sign of 0.8987. Import penetration has a negative correlation with raw steel production with a sign of -149.8. It seems that most of the shocks in raw steel production are attributed to these two variables. Changes in import penetration seems to be the major attributor to changes in raw steel production. It is logical that higher import penetration negatively affects domestic raw steel production, since imports are a substitute for domestic production. Although consumption has a significant effect on raw steel production, this effect is very small with a coefficient of 0.8988. As apparent consumption is correlated positively with import penetration (see table 6), it is likely that a large part of the shocks is absorbed by import as opposed to domestic raw steel production. Apparently domestic raw steel production only partially benefits (or loses, depending of the direction of the shock) from shocks in consumption. Most of the shock is absorbed by imports. This indicates that any changes in consumption have only a small effect on the domestic production of raw steel in the US.

However, as we have seen in the Zivot-Andrews test in the previous section there has been a structural break in the raw steel production data. Regression (1) above does not account for any structural break and as such may lead to misleading inference (cf. the omitted variable bias). A solution to this issue is to include a step dummy variable in the regression. A step dummy variable is set up in which all values for Breakdummy before 1982 are 0, and all values including 1982 and afterwards are 1. The estimation results for this model are shown as regression (2).

Taking the previous regression and adding the step dummy in regression (2) returns the coefficients that appear to be more in line with what we theorized they could be. Several variables now are significantly affecting raw steel production. In particular, with the addition of the breakdummy, quotas now have a significantly positive effect on the amount of raw steel production in the US. Tariffs have a negative sign, and thus a negative effect on raw steel production. However, tariffs do have an insignificant negative effect on raw steel production. A more detailed discussion of the effects of both quotas and tariffs is done in the discussion of regression (3). The breakdummy is significant at the 1% level, and the coefficient is negative and quite large. Apparently a rather large shock in 1982 has caused raw steel production to drop down by a large amount, as could be expected from figure 2 and the results of the Zivot-Andrews unit root test. However, with the addition of the breakdummy there seems to be a large change in some t-statistics. The large change in the t-statistics of some variables seems to indicate multicollinearity with the breakdummy. Also, multicollinearity was already assumed to be a possible issue for several predictor variables, and thus in table 6 a correlation matrix is constructed to check for multicollinearity.

(31)

31 Table 6: Correlation matrix

Growth DemAdm Cons ImPen Exports Tariffs Quotas Break dummy Growth 1 DemAdm 0.3276 1 Cons 1 ImPen 0.5489* 1 Exports -0.3964* 0.5922* 0.4441* 1 Tariffs 0.4408* 1 Quotas -0.5976* -0.3191 1 Break dummy 0.8280* 0.2975 0.3644 1

Only correlations at the 5% significance level are shown in this pairwise correlation matrix, and a * indicates they are significant at the 1% significance level.

This matrix provides some evidence for multicollinearity. Mostly between the breakdummy and import penetration, where the correlation coefficient is high with 0.8280. This is a clear indicator for multicollinearity between the variable import penetration and the breakdummy. Therefore, regression (3) shows the results of a regression in which the variable import penetration is omitted from the regression.

Regression (3) reveals some interesting results. As could be expected, coefficients are very similar to those in regression (2). First, apparent consumption still has a significant and positive effect on raw steel production. This indicates that any shocks in apparent consumption in the US does in turn affect production in the country. However, this effect is smaller than one would reckon. With a

coefficient of only 0.48 raw steel production is not affected by apparent consumption in the US in a one-to-one manner. This indicates that shocks that hit consumption are mainly absorbed by imports. As seen in the correlation matrix there is a significant positive correlation of 0.5489 between consumption and import penetration. As import penetration is an indicator of imports, this relation provides evidence of the possibility that a large portion of consumption shocks could indeed be transferred to an increase or decrease of imports.

Interestingly enough, in regression (3) still shows no indication that economic growth has a significant positive effect on raw steel production in the US. As discussed earlier in this chapter, there have been indications that there is no trend in raw steel production over time. This can perhaps be explained by the fact that the services sector accounts for a large part of the US economy. Economic growth is thus also driven to a large extent by the services sector, and as this sector barely uses any steel, economic growth may then have a fairly small (or no) effect on raw steel production in the US

The effects of protectionist trade measures: an Empirical Study on the Steel Sector