0
Asterina Zarnia – s3138305
Thesis Draft
i ABSTRACT
Anti-Dumping (AD), which is imposed to a specific product (named-product) from a specific country (named-country), is expected to reduce import of the named product from the named country (direct effect). However, theory indicates that there are unintended or indirect effects of AD beyond the named-product. To analyze the true effect of AD, this thesis examines the impact of AD imposition both at aggregate and product level. Product level analysis will cover the direct effect, while aggregate level analysis will try to explore the indirect effect of AD. Besides trade, AD is also expected to affect FDI, hence, this thesis will examine the impact of AD toward trade and FDI in Indonesia. Using gravity model, the product-level results show that AD is not effective in reducing named-product import from named-country exporter. Meanwhile at aggregate level, we find that AD generates a higher import from the named-country. For FDI effects, this study find that AD leads to a tariff-jumping FDI both at the aggregate and product level.
ii
TABLE OF CONTENTS
1 Introduction ... 1
2 Literature Review ... 4
2.1 Understanding Anti-Dumping... 4
2.2 The Effect Of Anti-Dumping ... 5
2.2.1 Direct Effect ... 6
2.2.1.1 Trade Destruction Effect ... 6
2.2.2 Indirect Effect ... 7
2.2.2.1 Trade Diversion Effect ... 7
2.2.2.2 Trade Deterrent Effect ... 7
2.2.2.3 Downstream Effect ... 7
2.2.2.4 Fdi Effect ... 8
2.3 Previous Study Of Anti-Dumping... 11
2.4 Hypotheses ... 13
3 Aggregate Level Analysis ... 13
3.1 Model Specification ... 14
3.1.1 Trade ... 14
3.1.2 Fdi ... 15
3.2 Data And Description Of Variables ... 15
3.2.1 Dependent Variable... 15
3.2.2 Independent Variables... 15
3.3 Methodology ... 17
3.4 Estimation Result And Analysis ... 17
3.4.1 Trade ... 17
3.4.2 Fdi ... 19
4 Product Level Analisis ... 21
4.1 Model Specification ... 21 4.1.1 Trade ... 21 4.1.2 Fdi ... 21 4.2 Data Sources ... 22 4.2.1 Dependent Variable... 22 4.2.2 Independent Variables... 22 4.3 Methodology ... 22
4.4 Estimation Result And Analysis ... 23
4.4.1 Trade ... 23 Descriptive Analysis ... 23 Econometric Analysis ... 24 4.4.2 Fdi ... 26 4.5 Discussion ... 28 5 Conclusion ... 30 References ... 32 Appendix ... 36
Appendix A - Trade In Indonesia 2011-2018 ... 36
Appendix B - Case Of Anti-Dumping On Bi-Axially Oriented Polyethylene Terephthalate (Bopet) ... 36
Appendix C - Previous Literatures ... 38
Appendix D - Trade Aggregate Level ... 41
Appendix E - Fdi Aggregate Level ... 42
Appendix F - Detail Of Product Subjected To Ad In Indonesia ... 43
Appendix G - Concordance Of Product To Industrial Sector In Nswi Data ... 44
iii List Of Tables
Table 1 Estimation Result Of Ad Impacts On Aggregate Bilateral Import ... 18
Table 2 Pairwise Correlation ... 19
Table 3 Estimation Result Of Anti-Dumping Impacts On Aggregate Fdi Inflows ... 20
Table 4 Estimation Result Of Anti-Dumping Impacts On Product Level Import... 25
Table 5 Estimation Result Of Anti-Dumping Impacts On Product Level Fdi ... 27
List Of Figures Figure 1 Ad Imposition 1979-2007 ... 2
Figure 2 Direct And Indirect Effect Of Anti-Dumping ... 10
Figure 3 Import Quantity Of Named-Products From Named Country ... 23
1 1 INTRODUCTION
Major countries adopted various approaches to liberalize their trade during 1985-2009 (Bown, 2011). After General Agreement on Tariffs and Trade (GATT) transforms toward World Trade Organization (WTO) (based on Uruguay Round 1986-1994), WTO countries undertook further liberalization through negation with other members and generates diverse preferential trade agreement initiatives1. The major approaches are likely to result in low enforced import tariffs
(Bown, 2011). While there has been an effort to lower applied tariffs from WTO countries, in recent years, there is an increasing import protection actions such as Anti-dumping (AD), Countervailing Duties (CVD)2, and Safeguards3. Bown (2011) refers these actions as temporary trade barrier (TTB). Bown also mentions that the phenomena of liberalization and trade protection ensue an average lower import tariff, but concurrently, facing a variation of threats through imposition or TTB removal.
Among the TTB, WTO (2009, 2018) reports that Anti-Dumping (AD) is the most frequent trade policy for the majority of countries in the world. According to WTO (2009), until 2007, countries investigate 205 anti-dumping, 26 countervailing duties and 12 safeguard cases on average per year, and annually impose 113 AD, 11 CVD, and 7 safeguards. Bloningen and Prusa (2015) state that government agencies usually arrange AD implementation without any consideration from executives or legislatives. Without executives and legislatives intervention, AD becomes procedurally and economically easier to impose. Moreover, once a country imposes AD, it may stay in force for a long period of time, as it will be reviewed every 5 years, while safeguard may remain in force for 4 years and can be extended to another 4 years only. Therefore, these conditions make AD the most preferable action rather than other TTB actions.
Dumping refers to an activity in which a firm sells products at lower price in foreign markets than that in domestic market (Bloningen & Prusa, 2015). Thus, according to AD requirements (as documented and codified in GATT), an importer country can initiate AD if it can convince that there are signs of dumping action from its trading partners and generate material loss to the domestic industry. AD is formed as tariffs implementation toward particular import product that is suspected of being a dumping product (named product) by exporter country (named country). Figure 1 below illustrates that Anti-dumping imposition increases tremendously and reaches its peak in 2000-2002. Even though there is a lowering number of AD investigation from 2002 onwards, new AD investigations have been increasing again since 2012. According to the data, prior to the 1990s, developed countries such as Canada, Japan, Australia, United States (USA) and European Union (EU) usually utilize this AD action. However, developing countries start to dominate 64 percent of all AD usage since the 1990s (WTO, 2009, 2018). Bown (2013) notes that
1 Several trade agreement such as NAFTA, CUSFTA, the expansion of EU community, etc.
2
trade liberalization is one of the motives to why developing countries need more protection through AD.
AD in developing countries attract more world attention because the extent of AD proliferation is also followed by an increasing share of imports of the product subjected to AD (Bown, 2011). Bown (2011) shows that AD in India, as the most frequent user of AD, dealt with 6.09 percent of its imports based on HS4 count and 2.94 percent of its import value in 2009. In China, product
subjected to AD covers 2 percent of imports value in 2003. In contrast, coverage of product subjected to AD in developed countries shows a declining trend. For instance, Bown (2011) also observes that according to the import value, USA has only 2.33 percent of its imports covered in 2007, which shows a decline from 6.14 percent in 2001.
Figure 1 AD imposition 1979-2007
Source: WTO, 2009.
Figure 1 shows the AD pattern of developed and developing countries. Figure 1 indicates clearly that developing countries have a rapid increase of AD impositions. In line with the trend, Indonesia, as one of the developing countries, also becomes a frequent user of AD. Based on data from the Indonesia Anti-Dumping Committee (KADI)5, there are at least 125 cases of AD
investigations and 92 initiations since 1996.
The trade deficit that arises in recent year partly motivates the positive trend of AD imposition in Indonesia. Based on data from Central Bureau of Statistics (BPS) of Republic Indonesia, import value for 2001-2018 increases tremendously to approximately 13.5 percent per year, outweighing the trend of export which increases by only 8.0 percent per year. Moreover, Indonesia hits the highest trade deficit value at 8.5 billion US$ in 2018 (see Appendix, Figure 5). To overcome this deficit as well as to protect domestic industry and domestic market, Indonesia put attention on
4 As an nomenclature of international product standard which is utilized by WTO countries
3
trade protectionism policies, particularly to its TTB. Among several kinds of TTB (AD, CVD, and safeguard), Indonesia also favors the utilization of AD as it is easier to impose than others. According to the Ministry of Trade Republic Indonesia, during 1995-2012, Indonesia imposed 43 AD, while there were only 13 safeguards. This imposition is expected to reduce and control imported product in Indonesia market.
While Indonesia and other developing countries initiate numerous acts of AD, the effectiveness of AD is still uncertain (Prusa, 2005). Tharakan (1991) argues that AD is beneficial to the economy than other protection such as import quotas since its product-country-specific’s characteristic does not generate substantial extra cost borne by consumers. Contrary to this argument, Klitgaard and Schiele (1998) claim that AD may jeopardize the economy due to the direct price escalation at the expense of the importer. Moreover, there are empirical evidences of unintended or indirect effects6 of AD beyond the product that is subjected to AD. However, available literature mostly focused on developed countries such as the USA and EU, and there is still limited literature that investigates the impact of AD in a developing country (Ganguli, 2008). Some scholars (Prusa,1996; Lee et al, 2013; Brenton, 2001) identify that unintended effect (i.e. trade diversion7) weakens the
effectiveness of AD in the USA and EU. In contrast with the result in developed countries, Ganguli, (2008) and Aggarwal (2011) detect that there is an existence of trade diversion effect, to note however, this effect does not outweigh the effectiveness of AD. Besides the effect on import, AD is also expected to affect FDI in a country (Bloningen and Feenstra, 1996). Several literature (Girma et al, 1996; Barren and Pain, 1999) finds that AD triggers more FDI (tariff-jumping FDI) in developed countries, however, there is still limited literature to investigate this effect in a developing country. Thus, more studies about AD in developing countries are needed.
This limited study about AD is also the case in Indonesia. The analysis of AD’s impact and effectiveness at the product and aggregate level are needed as it can provide an empirical study about the effectiveness of AD implementation in reducing the import value of protected goods in Indonesia. Furthermore, the case of whether AD is able to reduce imports at a more aggregate level or not. The possibility of the downstream and deterrent effect8 of AD may generate different outcome between product and aggregate level analysis. If an AD is effective in reducing imports of named products only, then at a more aggregate level the imposition of the AD generates higher imports for other product (downstream effect), which thus may jeopardize Indonesia's economy9.
Accordingly, this study explores the literature gap on AD effectiveness at product and a more aggregate level in the case of Indonesia. As the biggest economy in ASEAN10, Indonesia can maintain a stable and tremendous growth. Indonesia also actively engages in trade partnership through bilateral or regional agreement. Today, Indonesia is the top 10th biggest economy in terms
6 Trade diversion, deterrent effect, downstream effect, and FDI effect (see section 2.1) 7 Trade diversion refers to an increasing of import of named product from non-named country.
8 Downstream effect refers to an increasing of non-named product import from named country which have higher value added. Deterrent effect refers to a decreasing of non-named product import from named country.
4
of GDP based on purchasing power parity in the world, and being part of the G-20, which pronounce the prominent role of Indonesia as a developing country. Using gravity model, we examine the effect of AD on import and FDI in Indonesia at product and aggregate level. In term of FDI, we identify that tariff-jumping FDI occurs at both aggregate and product level. Our finding also shows that AD will increase bilateral import at aggregate level, but has no effect at the product level. This result suggests that AD is ineffective to reduce named product import and harmful for Indonesia’s trade balance due to its effect at aggregate level. As a suggestion, Government of Indonesia (GOI) should evaluate and review the use of AD as trade protection.
2 LITERATURE REVIEW
AD is initially negotiated and adopted through article VI in GATT 1947 which follows US Antidumping Act 1921. The Uruguay Round 1994, then, generates substantial revision to the article VI by providing a very detailed description of how implement AD by WTO countries. While WTO provides the guidance, countries have wide range on the way to define the rules. Bloningen and Prusa (2015) mention that this discretion develops variation of effects, thus, motives them to investigate a lot of AD actions. Thus, in this section, we review some existing literature on what the impact of the AD imposition from different countries is, and how this study fills the gap on the literature. Finally, we will also provide the hypotheses of this study.
2.1 Understanding Anti-Dumping
The word “Dumping” refers to a condition where a company sets lower prices for the similar product in foreign market than prices in the domestic market (Bloningen and Prusa, 2015). According to Article VI of GATT, dumping is quoted literally as:
“…by which products of one country are introduced into the commerce of another country
at less than the normal value of the products, is to be condemned if it causes or threatens material ‘injury’ to an established industry in the territory of a contracting party or materially retards the establishment of a domestic industry”. (GATT, 1994)
Based on that description, dumping may threat domestic industry or generates industries’ injury in destination markets. Based on the Anti-Dumping Agreement, there are several conditions that can be classified as injury. Injury may occur in the form of declining:
a. Domestic sales,
b. Profit of domestic firms, c. Output,
d. Domestic market share, e. Productivity,
f. Capacity utilization, and g. Return on Investment.
5
prevent domestic injury that is caused by dumping action from other countries through several actions. For instance, the importing country can conduct an investigation for maximum 2 years and initiate AD duty to restrain the import of the relevant product. There are several forms of AD measures, which are ad valorem duty, specific duty, price or quantity undertakings11, or a combination of all AD’s forms. When a country initiates an AD, the importer will pay for the duties levied. However not all dumping action will end with protection, the act of dumping action should follow some criteria, in this case, WTO (The GATT Agreement, 1994) mentions several criteria:
1. A Dumping that categorized as Less Than Fair Value (LSTV). This condition is a measure that will be used to define the margin of Anti-Dumping.
2. There should be a material loss in the host country.
3. The material loss must have causal link with dumping action.
4. If it is proven that the exporters set product price less than fair value but this does not generate any loss for the importer country, then dumping is not prohibited
One of the countries that is known to use dumping is Canada. Canada first established an anti-dumping (AD) law in 1904 to protect its domestic industry from injury due to anti-dumping practices. In that time, Canada protected its steel industry from the US steel industry.
Based on Global Antidumping Database, in few decades, the use of antidumping as a form of trade protection has become increasingly common. It is known that the act of AD was mostly used by developed countries to protect its industries from growing competition (Bown, 2008). However, Jafta (2006) explains that since 1995, developing countries started to use this protection. Based on WTO Committee on Anti-dumping practices semi-annual reports, in 1995-2005, India becomes the first initiator and imposer of AD, meanwhile US leads in the second position. Bown (2008) mentions that the main reason for the increase in AD usage in developing countries is the demand for protection in the era of trade liberalization, thus motivates the increasing number of AD imposition by developing countries.
To gain a better understanding of dumping and the usage of Anti-Dumping, see Appendix B which provides a brief elaboration of an AD case in Indonesia.
2.2 The Effect of Anti-Dumping
Recent literature is triggered by the proliferation of AD in the world. AD becomes the most frequent TTB for most of the countries in the world (Bown, 2010, 2011a, b). However, there is a changing pattern from who use this AD policy. There has been a decline in AD imposition by developed countries (i.e USA and EU), and a significant increase in use by developing country. The rapid adoption of AD in developing countries obviously motivate further analysis as well.
11Undertaking is a concession by the exporter to increase its market price and/or decrease its export quantity to an importer
6
What previous literature finds empirically about AD comes mostly from the analysis of the USA and EU data. But developing countries have different economic condition with different structure and political force, and different position and role in the global economy system than US and EU. Thus, the motivation and outcome from the use of AD policy are likely to be different from developed countries’ study.
The increasing phenomenon of the AD triggers the question of whether AD is able to protect a country from ‘unfair’ trade practices or not. The imposition of AD, indeed, is very products and country-specific. We expect that an effective AD would restrain imports and treat the injuries of the domestic industry that requests protection. Many studies have documented substantial identification of the direct impact of AD. Prusa (1996), Cuyver and Dumont (2005) try to identify the effect of AD in the product level in US and EU respectively. Their study find that AD can reduce named product import from the named country. However, they also find that there is trade diversion effect which weaken the direct effect of AD. Recently, previous literature have documented indirect and unintended implications of this AD duty which are also important to consider for the policy maker as well.
Figure 2 provides the framework about the effect of Anti-Dumping and gives insights to understand the effect that have been documented so far. In the figure, the effect of AD can be distinguished into named and non-named country, and named and non-named product. In general, the imposition of Anti-Dumping is expected to directly affect import of named product from named country (Bloningen & Prusa, 2015). However, the unintended or indirect effect may affect non-named product both from named and non-named country. Besides the effect toward import, AD may also affect FDI in a destination country (Bloningen and Prusa, 2015; Vandenbussche and Zanardi, 2010). Theoretically, there are relations between trade and FDI as they may affect one another (Helpman et al, 2004). Thus, the disaggregation of the effect between import and FDI is connected with red dashed line (as seen in figure 2).
In the next section, we will review and elaborate the effects of AD which have found by previous literatures.
2.2.1 Direct Effect
2.2.1.1 Trade Destruction Effect
7
Several literatures have found empirical proof regarding this trade destruction effect. Prusa (1996), Baylis and Perloff (2010), and Lee et al (2013) discover this effect on AD in the case from US. This effect also happens in EU which is proven by empirical studies from Cuyver and Dumont (2005). Beside the developed country like US, trade destruction effect also takes place in some developing countries like India (Ganguli, 2008), Mexico (Niels, 2003), and China (Park, 2009). Thus, as the direct effect, AD can protect domestic industry by reduce named product import from named country.
2.2.2 Indirect Effect
2.2.2.1 Trade Diversion Effect
AD policy is a specific instrument of trade protection. AD is subject to specific products from specific countries. Besides the trade destruction effect as explained previously, the imposition of AD also has an unintended effect which is commonly called as trade diversion effect. This effect is a situation where there is a shifting of imports to countries that are not subject to AD. The effect will take place especially when domestic industry cannot fulfil the demand or there is lack of supply. The low production capacity in domestic industries or the production inefficiency are most likely to trigger the emergence of this effect. Some previous studies (e.g, Prusa, 1996; Brenton, 2004; Cuyvers and Dumont, 2005) have empirically found the presence of trade diversion from non-named countries because of the imposition of AD in the product level. This studies indicate that the decreasing import from named-countries will offset the increasing of import from non-named countries. Thus, the emergence of trade diversion effect will make AD less beneficial to domestic producer as it can increase import of named product from non-named country.
2.2.2.2 Trade Deterrent Effect
Another indirect effect of AD is called trade deterrent effect (Vandenbussche and Zanardi, 2010). This effect takes place when trade partners tend to become more careful in exporting goods to countries that are frequent users of AD and learns to avoid dumping filings by setting higher price or lower export volumes. Reitzes (1993) documents that AD duties generate several strategic actions from exporter firm, although it still depends on the competitiveness of the exporter firm. Higher prices and lower trade volumes are more likely to occur as the trade partner learn how to evade from dumping accusation. In relation to this, a study by Bloningen (2006) shows that a higher number of previous AD filing affects higher probability of AD filing in the present year. Consequently, we predict that this learning and signaling will negatively affect trade flows for non-named product.
2.2.2.3 Downstream Effect
8
In similar fashion, the impact of AD toward downstream products may also increase imports. Vandenbussche and Zanardi (2010) explain that AD increases imports of downstream goods through two channels. First, when AD is used for instance, to protect steel industry that is used as input for products such as cars, the increasing input price reduces the competitiveness of domestic car industry. As a consequence, the import of cars (downstream industry) increases. This effect suggests that AD may actually generate higher imports not only for the named product, but also for more aggregate level. Ferreira and Rossi (2003) examine manufacturing sector in Brazil from 1985 to 1997 and find that trade protection is negatively significant to the growth rates of downstream industry productivity and industry labor productivity. This implies higher imports for downstream products in domestic market. Lee et al (1995) also conduct study and find that trade protection also gives negatively significant effect to the growth rate of labor productivity and TFP in sector level’s Korea manufacturing. This result also suggest that AD can increase import at more aggregate level due to lower competitiveness of downstream industry.
Second, if the protected product is a raw product, this then provides an incentive for the exporter country to further process the product and export goods that have more value added. In the case for importer country, the import of named product decreases, but the import of non-named products, which have higher value added, increases. Thus, again, imposing AD increases import value in a more aggregate level.
2.2.2.4 FDI Effect
Economic theory indicates that trade barrier may have an effect on investment. Theoretically, Helpman et al (2004) explain that a firm can fulfil foreign demand through production in home then export the products to foreign or build a new factory in the destination market. In short, a firm has two options between export and ‘horizontal’ foreign direct investment (FDI). The decision of a firm will much depend on the profit comparison where export generates trade cost, while FDI needs fixed cost to build new plan abroad. When profit of doing export is higher than FDI, or in other words, cost of trade is less than cost of FDI, a firm should keep the production in home and trade. However, when trade cost becomes higher and doing FDI is likely more profitable, instead of export, a firm can choose to expand its production in the destination market through FDI. Helpman et al (2004) mention the arise of situation when a profit between export and FDI is equal (proximity-concentration trade-off) then at this point, a firm can choose between the two available options mentioned before.
Trade cost consists of each and every sum of money spent to ship the products until it arrive at the destination. This trade cost may also include import tariff set by the destination country. The imposition of AD from destination market usually increases the trade cost, thus AD is predicted to be a tool to attract foreign direct investment (FDI) (Vandenbussche and Zanardi, 2010). If named country chooses FDI as its response to AD imposition, import of the named product from the named country decrease. This effect is commonly known as ‘tariff-jumping FDI’.
9
a higher probability to choose FDI instead of export when tariff increases. In term of AD, Barrel and Pain (1999), Belderbos (1997), and Fenstra (1997) have found that AD has positive impact to FDI. This suggest the emergence of tariff-jumping FDI.
10
11 2.3 Previous Study of Anti-Dumping
Several studies have been conducted to empirically investigate the impact of AD (for further detail see Appendix C). In general, many studies document the existence of trade destruction effect and trade diversion effect. The findings are mostly gathered from the analysis towards subjected or named product to AD. Trade destruction and trade diversion effect occur both for the case of AD in developed country (Prusa, 1996, Lee et al, 2013, Brenton, 2001, Cuyvers and Dumont, 2005), Koning et al, 2002) and developing countries (Bown, 2008, Ganguli, 2008, Aggarwal, 2011, Chandra, 2017). Trade diversion weakens the effectiveness of AD. In a developed country, trade diversion outweighs trade destruction effect (Prusa, 1996, Lee et al, 2013). Meanwhile in developing country like India, trade destruction effect still dominates the impact, thus it seems that AD works better rather than in a developed country (Ganguli, 2008).
In term of AD in developed countries, Prusa (1996) and Lee et al (2013) provides empirical evidence about the effect of AD in United States (US) into trade. Using OLS estimation on panel data of 1978-1993, Prusa (1996) finds that AD reduce import from named country (trade
destruction effect). Beside trade destruction, he identifies the emergence of trade diversion effect
which outweighs trade destruction effect. Lee et al (2013) also find that the existence of trade
destruction effect as the impact of AD. Lee et al also identify that trade destruction effect occurs
in short period only.
Beside US, study about AD in developed countries are also conduct for EU countries. Brenton (2001), Cuyvers and Dumont (2005), and Konings et al (2001) using similar model applied by Prusa (1996), examine AD that initiated in EU. They also find that AD cause trade destruction
effect and trade diversion effect at the same time. However, Konings et al (2001) finds that trade
diversion effect in EU is substantially smaller compared to US. This differences are triggered by different legal rules and political economy issue between US and EU.
In recent years, a rapid increase of AD in developing countries motivates several studies about the impact of AD in developing countries. Ganguli (2008), using Arrellano-Bond procedures, investigates the impact of 285 AD petitions filed by India in 1992-2002 toward trade flows. As the most frequent user of AD recently, Ganguli finds that AD in India has significant trade destruction effect. Moreover, trade diversion effect occurs too but fails to wipe destruction impact out altogether. So, AD turns to be effective in India.
12
The studies described before mainly mitigate the impact of AD towards trade. However, as seen in figure 2, AD may also have an effect on FDI. Although there are still limited studies concerning this topic, Bloningen and Feenstra (1996) and Girma et al (2002) have written down about the FDI effect on AD imposition. Using data of FDI from Japanese manufacturers, they find that AD imposition triggers more FDI (tariff-jumping FDI) in the US and UK. While they find that AD generates tariff jumping FDI which will cause a domestic profit reduction, Bloningen (2002), using all FDI inflows, find that the effect in US is modest (not as high as using FDI from Japanese only). Bloningen argues that only big companies from developed countries could take advantage of this strategy. Bloningen (2002) also explains that tariff-jumping FDI mostly occur if foreign companies which facing AD threats have a competitive advantage that can be transferred abroad, and if the production of named-products subjected to AD can be moved abroad at relatively low costs But again, most of the studies are conducted for developed countries.
Past studies generally aim to see the impact of AD imposition at product level which capture trade destruction and trade diversion effect, whereas AD imposition has unintended and indirect impacts beyond the named product. However, previous literature comes mostly from the analysis of the USA and EU data. There is still little known about the impact of AD in developing countries, especially about its impact on FDI. Developing countries have different economic condition with different structure and political force, thus, the outcome from the use of AD are likely to be different from developed countries’ study.
Building upon the gap of the previous studies, this study aim to investigate the impact of AD in Indonesia as one of prominent developing countries. As far as author’s knowledge, the studies related to Indonesia are limited and mostly conducted specific for a subjected product, for example, the impact of Cold Rolled Coil/Sheet AD (Tjahjasari, 2015). Therefore, this study analyses the effectiveness of AD toward trade based on several levels of analysis, which include:
1. Aggregate level: analysis at this level is expected to be able to see how the impact of AD imposition both on named and non-named products from named country.
2. Product level: as in previous research, analysis at this level is expected to be able to investigate the effects of AD specifically for named product from named country only. Furthermore, given the possibility of the FDI effect from the imposition of AD, this study will also look at how the impact of the imposition of AD on imports and FDI both at the product and a more aggregate level.
These kinds of study are important due to the increasing frequency of the AD initiation as an instrument of trade protection in the middle of trade deficit experienced by Indonesia. This study aims to provide a more comprehensive empirical study about the effectiveness of AD in reducing the import value of protected goods in Indonesia, while there is possibility of downstream and
deterrent effect of AD may generate different outcome between product and aggregate level
13
condition may harmful for Indonesia's economy. The results of this study will certainly be useful for the Indonesian government to further evaluate the use of AD as a trade protection instrument in Indonesia.
2.4 Hypotheses
According to figure 2, this study will focus on AD’s effect on import and FDI inflows from named country exporter. We can see in the figure that there is direct effect of AD (trade destruction effect) which is expected to reduce import of named product from the named country exporter. However, beyond the named product, AD is also expected to affect import (positively or negatively) of non-named product from the non-named-country exporter through downstream effect and deterrent effect. As a result, the outcome at product and aggregate level might be different. Thus, since this study conducts at two levels of analysis, the hypotheses are:
Analysis at Product Level
At the product level, the imposition of AD is predicted to have a significant negative effect on the imports from named-country exporter which suggests the emergence of trade destruction effect. On the other hand, AD is predicted to have a significant positive effect on FDI inflows
(tariff-jumping FDI). This result may occur since FDI becomes more attractive compared to export due
to higher trade cost as AD is imposed.
Analysis at Aggregate Level
At the aggregate level, the impact of AD will capture both named and non-named product. Based on figure 2, there are at least 3 effects that can appear when analyzing named and non-named product at the same time which are trade destruction effects, trade deterrent effects, and
downstream effects. If AD is negative significant then it suggests that the trade destruction and
trade deterrent effect are dominating the effect. On the other hand, if AD is significant positive then it can be a sign that the downstream effect dominates the effect of the AD.
Building upon theory, trade effects of AD at aggregate level could be negative or positive. While in the case of FDI at aggregate level, AD is also expected to have a significant positive effect (tariff-jumping FDI) due to increasing trade barriers.
3 AGGREGATE LEVEL ANALYSIS
This study analyses the impact of Anti-dumping in Indonesia toward its bilateral trade and FDI inflows both at the aggregate and product level. Based on Global Anti-Dumping Database (GAD), there are at least 50 countries which imposes AD for the last decades. Using gravity model, we choose to investigate AD in Indonesia to explore the gap of limited study about impact of AD in developing countries.
14 3.1 Model Specification
The Gravity model is used recently to explain bilateral trade which other economic theory cannot explain. Tinbergen (1962) uses this model when he argues that bilateral trade can be explained by “gravity equation” law which motivated by the Newton theory of gravitation. In the model, Tinbergen uses bilateral trade flow as the dependent variable, while GDP and distance as the independent variable. The result shows that GDP has a positive effect on trade flows, while distance has a negative effect. This result suggests that larger and closer countries tend to trade more to one other.
3.1.1 Trade
Traditionally, the gravity model uses GDP, distance, and several variables which mostly used to address trade cost. Gravity model is simply an estimation model which includes all variables in natural logarithms and achieve a log-linear equation that can be predicted by using OLS regression. The usual problem with the ‘old’ gravity model estimation is that the so-called multilateral resistance terms (MRTs) are neglected. Feenstra et.al (2001) and Frankel and Rose (2002) argue that gravity model should consider MRT to address all trade barriers that are faced by every country with all their trading partners. To deal with this problem, importer and exporter dummies fixed effects can be used to replace this MRT. Moreover, gravity model estimation mostly includes year fixed effects to adjust for common trend and shocks (Kohl, 2014). MRT also may change over time and thus, we may need to include exporter and importer time-varying effects in the estimation.
The proposed model of gravity model in this study is shown below (equation 1). In the model, import value in country i (Indonesia) from all Indonesian trade partner country j is the dependent variable. To find the effect of Anti-dumping toward trade flows, following Vandenbussche and Zanardi (2010), we have variable AD dummy as the main interest. This variable equal to 1 if Indonesia imposes AD to the exporter in that year, and 0 if there is no AD imposition toward the exporter in that year. To address MRT, the model includes importer and exporter fixed effect (ai and aj). However, ai is constant because Indonesia is the only importer in our dataset. The inclusion of exporter fixed effects will address for unobserved variables that may correlate with bilateral import (left-hand side) and right-hand side variables. Not surprisingly, GDP importer (𝐺𝐷𝑃𝑖𝑡), distance, border, language, and colony are not estimated due to perfectly collinear association the exporter dummies. Moreover, we also include year dummies in the estimation. This is not a perfect model since MRT may vary over time. However, MRT may not change significantly over a rationally short sample period (Baldwin and Taglioni, 2006).
ln 𝑋𝑖𝑗𝑡 = 𝛼𝑗+ 𝛼𝑡+ 𝛽1𝐴𝐷𝑑𝑢𝑚𝑚𝑢𝑦𝑗𝑡 + 𝛽2ln(𝐺𝐷𝑃𝑖𝑡) + 𝛽3ln(𝐺𝐷𝑃𝑗𝑡) + 𝛽4ln(𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗)
15 3.1.2 FDI
Recently, the gravity model has also been intensively used for the analysis of FDI flows` determinants. Using gravity model approach, it suggests that FDI is positively correlated to GDP both in origin and destination countries and negatively correlated to the distance between them. A lot of theoretical framework supports the use of gravity model in describing FDI flows. Dunning` s (1958) OLI (Ownership, Location, Internalization) paradigm is one of the most acknowledged theoretical framework. According to the framework, market size is essential factors for FDI. Several studies have concerned on OLI and suggest that critical consideration of location is the market size of the destination market (Ethier and Markusen, 1991,1996; Barrel and Pain, 1999; Yeaple, 2001).
Based on the justification above, this study will also use gravity model which aims to find the impact of AD toward bilateral FDI inflows. To do so, this study will have FDI inflows from all country j which invest to country i (Indonesia) as the dependent variable and include similar independent variables as included in equation 1.
ln 𝐹𝐷𝐼𝑖𝑗𝑡 = 𝛼𝑗+ 𝛼𝑡+ 𝛽1𝐴𝐷𝑑𝑢𝑚𝑚𝑢𝑦𝑗𝑡 + 𝛽2ln(𝐺𝐷𝑃𝑖𝑡) + 𝛽3ln(𝐺𝐷𝑃𝑗𝑡) + 𝛽4ln(𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗)
+ 𝛽5𝑏𝑜𝑟𝑑𝑒𝑟𝑖𝑗+ 𝛽6𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒𝑖𝑗+ 𝛽7𝑐𝑜𝑙𝑜𝑛𝑦𝑖𝑗+ 𝛽8ln(𝐸𝑅𝑖𝑗) + 𝛽9𝑅𝑇𝐴𝑖𝑗 +𝜀𝑖𝑗𝑡 (2)
3.2 Data and Description of Variables 3.2.1 Dependent Variable
Our dependent variable is annual bilateral import or FDI inflows from country partner j to Indonesia i. The import data for this variable are obtained from Trademap and from National Single Window of Indonesia (NSWI) for FDI data with the period from 2001 to 2012. NSWI provides bilateral FDI data for 23 industry sectors in Indonesia. The database is maintained by BKPM or Indonesian Investment Coordinating Board.
3.2.2 Independent Variables
ADdummyjt - There are several independent variables used in this study. The main interest variable
in this study is ADdummyjt which equal to 1 if the importer imposes AD to the exporter in that year, and 0 if there is no AD imposition toward the exporter in that year. The value of this variable will suggest the effectiveness of AD both at the product and at aggregate level. The information about AD imposition is collected from GAD by World Bank. In the database, they provide detail information of AD case which are date of investigation and initiation, detail of HS of named product, margin of dumping, and tariff duty of AD.
Gross Domestic Product (GDP) - As we use gravity model in this study, this study will correlate
16
while GDP of exporter indicates the potential supply from the particular country. GDP is expected to come with positive coefficient on import and FDI since large countries mostly will trade and invest more. We obtain the GDP data from World Development Indicator provided by World Bank.
Distance - The other variable is distance between two countries (distanceij) which reflects transportation or trade cost. The data of this variable is provided by CEPII. The relationship between distance, trade, and FDI depends on whether the products in question is vertically or horizontally linked to foreign firm. If the products are related horizontally, Markusen (2002) argues that distance will positively affect FDI and negatively on trade because FDI and trade are considered as substitutes; unless associated with higher information cost. Further distance generates significant shipping cost which is expected to be replaced by FDI. On the other hand, if the products are vertically related, Bergstrand and Egger (2007) find that FDI and distance considered to be negative. Thus, we expect that distance variable will negatively affect import and the effect on FDI could be negative or positive.
Exchange Rate - Bergstand (1985) demonstrate that the inclusion of the exchange rate in gravity
model is important to define many variation among countries. Therefore, this study will also include exchange rate and computed by the formula:
𝐸𝑅𝑖𝑗𝑡=
𝐴𝑛𝑛𝑢𝑎𝑙 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝑟𝑎𝑡𝑒 𝑜𝑓 𝐼𝑛𝑑𝑜𝑛𝑒𝑠𝑖𝑎 𝑝𝑒𝑟 𝑈𝑆 𝑑𝑜𝑙𝑙𝑎𝑟
𝐴𝑛𝑛𝑢𝑎𝑙 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑗 𝑝𝑒𝑟 𝑈𝑆 𝑑𝑜𝑙𝑙𝑎𝑟(𝑖𝑛 𝑦𝑒𝑎𝑟 𝑡)
Using this formula and data from IFS by IMF, we define annual exchange rate by Indonesia’s currency units divided by one unit of exporter currency. A higher value of this variable means that there is devaluation of Indonesian currency, accordingly, imported products become more expensive rather than domestic product. Meanwhile for FDI, Froot and Stein (1991) argue that a depreciation of host country’s currency will increase the relative wealth of foreign investor, generating more investment inflows to host country. Therefore, we expect that this variable will have a negative impact both on import and FDI.
Regional Trade Agreement (RTA) – McCulloch (1985, 1993) mentions that introducing bilateral
RTA refers to a reduction in a bilateral trade cost and it is associated with a higher bilateral trade. However, he also demonstrates that a declining in trade costs tends to reduce bilateral FDI. Bergstrand and Egger (2007) also confirms that bilateral RTA should be associated with lower investment. Thus, we expect that import (FDI) is positively (negatively) related to RTA. The RTA data are obtained from Trade Agreement Heterogeneity Database12.
Border, Language, Colony - The model in this study also controls other variables which reflect
individual characteristics (time-invariant characteristic) of countries including border, language,
17
and colony. Respectively, the value of this variable will be 1 if Indonesia is contiguous, share a similar common language, and has ever had a colonial link with country partner. Information about this variable is collected from CEPII database. This time-invariant variables are expected to show the positive sign both on import and FDI.
3.3 Methodology
Our panel dataset is arranged by country-pair and year. There are 3 estimation ways that can be adopted to estimate panel data. They are pooled model least square (POLS), random effect model (REM), and fixed effect model (FEM). To decide the best model to estimate our gravity model, thus we need to pay attention to the characteristic of the data as well as our research interest. To find the best model, thus we conduct the Breusch-Pagan test and Hausman test. The Breush-Pagan test and Hausman test (see Appendix) show that there is systematic difference in the coefficient, thus we use FEM rather than REM. In the analysis section, we will provide both FEM result. Silva and Tenreyro (2006) argue that the standard empirical method in estimate gravity model is flawed due to several problems. First, log-linearization of the empirical model with the occurrence of heteroscedasticity generates to inconsistent estimation. Second, in the presence of zeros both in trade and FDI data which will be dropped when we use standard method. Thus, they conclude that using standard empirical method, the result will be biased and inconsistent.
Silva and Tenreyro (2006) propose the Poisson pseudo-maximum-likelihood (PPML) method to address the problem. They find that PPML is robust in the presence of heteroscedasticity and additionally provide a natural way to deal with zero in the data. Therefore, because our data set contains a lot of zero data and suffers from heteroscedasticity problem, we also show the result of PPML estimation in the analysis. In the PPLM estimation, we use import and FDI at level value instead of logarithm value.
3.4 Estimation Result and Analysis 3.4.1 Trade
The estimation results from the various specification of the gravity equation as in (1) are shown in Table 1. Column (2) reports FEM which is not accounting the presence of zero trade data and heteroscedasticity. FEM specification shows that the coefficient of ADdummy is positive but not significant. In column (4), (5), (6), (7), (8) and (9), we show PPML13 estimations to deal with heteroskedasticity and zero trade flows. In the PPML estimations, we include all variables as shown in equation (1) to our estimation, and as we have expected before GDP importer, distance, border, and language are omitted due to collinearity.
18
Table 1 Estimation Result of Anti-dumping Impacts on Aggregate Bilateral Import
Variable FEM PPML (1) (2) (4) (5) (6) (7) (8) (9) ADdummy 0.0291 (0.0964) 0.0932** (0.0341) -0.0073 (0.521) 0.0756* (0.0343) ADdummyt-1 0.112* (0.0467) -0.0208 (0.037) 0.115* (0.0450) ADdummyt-2 0.144* (0.0591) 0.104* (0.040) ADdummyt-3 0.0786 (0.063) LnFDI 0.0146 (0.0132) -0.00087 (0.0102) Ln(GDPit) 0.669 (0.666) (0) (0) (0) (0) (0) (0) Ln(GDPjt) 1.167 (0.782) 1.531*** (0.299) 1.403*** (0.304) 1.343** (0.310) 1.462*** (0.349) 1.655*** (0.291) 1.380*** (0.275) LnERij -0.0399 (0.0795) 0.0775 (0.144) 0.169 (0.160) -0.123 (0.163) -0.206 (0.179) -0.0315 (0.257) -0.0937 (0.255) RTAij -0.161 (0.165) -0.0446 (0.0935) -0.0521 (0.101) -0.0149 (0.113) -0.0106 (0.111) -0.0539 (0.0924) -0.0543 (0.0830) Number of Obs 1926 2062 1886 1709 1531 451 417 R-sq 0.069 0.967 0.973 0.975 0.975 0.978 0.981
Exporter Dummy Yes Yes Yes Yes Yes Yes Yes
Year Dummy Yes Yes Yes Yes Yes Yes Yes
Note: i) Standard errors in parentheses (clustered by country-pair) * p<0.05, ** p<0.01, *** p<0.0010; ii) other estimates (language, distance, border, colony) omitted to save space. iii) Panel data arranged by pair of exporter-importer.
PPML estimations show that ADdummy and ln(GDPjt) are the significant variables in our model.
According to column (4), (5), (6), as we include ADdummy and lag 1, 2, and 3 years of ADdummy (ADdummyt-1, ADdummyt-2, ADdummyt-3), the coefficient of these variables are positively significant on bilateral import. Positive coefficient of ADdummy and its lag variable suggest that AD will increase bilateral import from named country at aggregate level. However, when we try to include lag 4 years of ADdummy, the variable becomes insignificant. The insignificant of
ADdummyt-4 is probably because some of AD in Indonesia are imposed for only 3 years. In
addition, column (7) shows the result when we put ADdummy and all its significant lag together to find the effect of AD when phase-in effects have been accounted for. The results show that AD still have a positive effect over a prolonged period of time.
19
industry. The increasing input prices lead to a lower competitiveness of downstream industry (Vandenbussche and Zanardi, 2010). Consequently, this downstream effect generates higher import for non-named product. As a result of no-trade destruction effect and domination of the downstream effect, AD imposition leads to a higher import at the aggregate level.
Table 2 Pairwise Correlation
import ADdummy FDI Lngdp_exp lner RTA
import 1.0000 ADdummy 0.3581 1.0000 FDI 0.4521 0.2373 1.0000 lngdp_exp 0.3295 0.2287 0.1162 1.0000 lner -0.0677 -0.0526 0.0904 0.2854 1.0000 RTA 0.4437 0.3608 0.1655 0.1106 -0.1849 1.0000
Besides our main interest variable, another significant variable in our estimation result is GDP exporter. As we have expected, the coefficient of GDP exporter is positively significant. This result indicates that, holding other things constant, an increase in GDP of exporter’s country will increase import from the exporter’s country.
Vandenbussche and Zanardi (2010) argue that AD also generates FDI effects. To investigate the relevance between Import and FDI, we add FDI variable in our model. As seen in Table 9 column (8) and (9), FDI has no effect on import. However, in Table 2, we deliver the result of the pairwise correlation matrix and find that there is a positive correlation between FDI and import. This positive correlation suggests that for Indonesian case, FDI inflow is not a substitution for import. An increasing of FDI may also increase import. The reasonable explanation of this correlation are countries which tend to increase trade openness attract a higher level of FDI and most FDI inflows in Indonesia is part of a global production network or global value chain. If this is the case, a new plan in Indonesia due to FDI may need input materials from foreign which lead to higher import demand.
3.4.2 FDI
Besides trade, AD is also expected to affect FDI inflows of the importing country (Belderbos, 1997). Table 3 provides the estimation result of AD impact toward FDI inflows in Indonesia based on equation (2). The results show that using FEM and PPML our main interest variable, noted as
ADdummy, seems insignificant on FDI inflows. The insignificant of ADdummy indicates that we
20
results indicate that, in term of FDI, AD positively affects FDI inflows from-named country with some delays. The delay suggests that named countries need time to respond to the AD imposition. According to the results in table 3, we can find that the imposition of AD increases FDI at the aggregate level through the effect of tariff-jumping FDI. This finding is in line with the study from Barrel and Pain (1999) and Bloningen and Feenstra (1997) who find that AD correlates with an FDI increase in the USA and EU. Bloningen (2002) explains that tariff-jumping FDI is most likely to occur if foreign companies, which facing AD threats, have a competitive advantage that can be transferred abroad and if the named product can be moved overseas at relatively low costs. According to this opinion, tariff-jumping FDI that occurs in Indonesia may indicate that: first, products originating from named-country for both named and non-named products can be transferred to overseas production processes through FDI. Second, countries that are subject to dumping in Indonesia are countries that have high competitive advantages. In this regard, the majority named-countries in Indonesia, most likely are indeed industrialized countries such as China, South Korea, and India.
Table 3 Estimation Result of Anti-dumping Impacts on Aggregate FDI Inflows
Variable FEM PPML (1) (2) (4) (5) (6) (7) (8) (9) ADdummy 0.491 (0.251) 0.0398 (0.163) 0.229 (0.267) ADdummyt-1 0.0960 (0.233) 0.176 (0.229) ADdummyt-2 0.453 (0.239) 0.156 (0.131) ADdummyt-3 0.655** (0.218) 0.244 (0.158) ADdummyt-4 0.586* (0.237) 0.459* (0.020) Ln(GDPit) -1.246 (1.465) 0.935 (0.960) 0.860 (1.131) -0.656 (1.565) -1.394 (1.180) -2.384 (1.734) -2.340 (1.346) Ln(GDPjt) 3.512** (1.202) (0) (0) (0) (0) (0) (0) LnERij 0.337 (0.192) 0.861 (0.722) 0.856 (0.734) 1.299 (1.044) 0.571 (0.405) 0.431 (0.481) 0.435 (0.063) RTAij 0.394 (0.356) 0.127 (0.333) 0.124 (0.343) 0.155 (0.277) -0.0119 (0.144) 0.129 (0.229) -0.008 (0.340) Number of Obs 482 971 884 796 696 618 607 R-sq 0.118 0.861 0.861 0.877 0.898 0.899 0.905
Exporter Dummy Yes Yes Yes Yes Yes Yes Yes
Year Dummy Yes Yes Yes Yes Yes Yes Yes
21
Also in column (2) table 3, it can be seen that the Ln(GDPjt) as Indonesian GDP has a significantly positive coefficient. This coefficient means that an increasing of Indonesian GDP will increase bilateral FDI inflows in Indonesia. Indonesia has a big market size through its high GDP which makes AD in this country will also incentivize foreign countries to invest (FDI) in Indonesia. Thus, again, AD will increase FDI.
4 PRODUCT LEVEL ANALISIS
This section explains the model specification, data, methodology, estimation result and analysis for product level analysis. We have found previously that AD generates higher import at aggregate level and tariff-jumping FDI. To enrich the relevance of Anti-dumping impact in Indonesia, this study also aim to distinguishthe impact at the product level. In this product level analysis, we can elaborate further about the effect of AD on specific named-product. Similar to the analysis on the aggregate level which has elaborated before, we also determine the impact of antidumping on trade and FDI inflows in this section.
4.1 Model Specification 4.1.1 Trade
In analyzing the impact of AD toward trade at the product level, this study also uses gravity model as in equation 1 and has ADdummy as the main interest variable. The differences are, the dependent variable used in this section is import value and quantity in country i (Indonesia) from all Indonesia trade partner’s country j for the specific product that imposed AD (See appendix F for the product’s detail). In the model, we also consider the multilateral resistance terms (MRTs) by including importer and exporter dummies fixed effects (αi and αj). However, as Indonesia is the only
importer, αi is constant. This is not a perfect model since MRT might vary over time. However,
MRT may not change significantly over a rationally short sample period (Baldwin and Taglioni, 2006). This product gravity model estimation also includes year fixed effects to adjust for common trend and shocks (Kohl, 2014). Moreover, because our dataset consists of a lot of different products which subject to AD, we include product fixed effect (αp) to control with unobserved
HS-products-specific, time invariant characteristics.
ln 𝑋𝑖𝑗𝑡 = 𝛼𝑗+ 𝛼𝑡+ 𝛼𝑝+ 𝛽1𝐴𝐷𝑑𝑢𝑚𝑚𝑢𝑦𝑗𝑡 + 𝛽2ln(𝐺𝐷𝑃𝑖𝑡) + 𝛽3ln(𝐺𝐷𝑃𝑗𝑡) + 𝛽4ln(𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗)
+ 𝛽5𝑏𝑜𝑟𝑑𝑒𝑟𝑖𝑗+ 𝛽6𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒𝑖𝑗+ 𝛽7𝑐𝑜𝑙𝑜𝑛𝑦𝑖𝑗+ 𝛽8ln(𝐸𝑅𝑖𝑗) + 𝛽9𝑅𝑇𝐴𝑖𝑗 +𝜀𝑖𝑗𝑡 (3)
4.1.2 FDI
22
ln 𝑋𝑖𝑗𝑡 = 𝛼𝑗+ 𝛼𝑡+ 𝛼𝑠+ 𝛽1𝐴𝐷𝑑𝑢𝑚𝑚𝑢𝑦𝑗𝑡 + 𝛽2ln(𝐺𝐷𝑃𝑖𝑡) + 𝛽3ln(𝐺𝐷𝑃𝑗𝑡) + 𝛽4ln(𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒𝑖𝑗)
+ 𝛽5𝑏𝑜𝑟𝑑𝑒𝑟𝑖𝑗+ 𝛽6𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒𝑖𝑗+ 𝛽7𝑐𝑜𝑙𝑜𝑛𝑦𝑖𝑗+ 𝛽8ln(𝐸𝑅𝑖𝑗) + 𝛽9𝑅𝑇𝐴𝑖𝑗 +𝜀𝑖𝑗𝑡 (4)
4.2 Data Sources
4.2.1 Dependent Variable
For trade analysis, we use import value and quantity from country j in Indonesia. To construct the data, we acquired the Harmonized System (HS) product code for each AD filled in 2001-2012. Because duration of AD imposition is 3 or 5 years, we obtain trade and FDI data until 2017 to see the whole impact of AD for the named-product.
There are 10 products subjected to AD in that period with vary of HS. The primary data on AD cases are sourced from the Global Antidumping Database managed by Bown (2014) as it provides detail information about the type of product and the HS code imposed, the exporting country involved, the date of initiation and the date of imposition, and the amount of AD until 2015. To facilitate the collection of import data related to named-products, the classification of 6-digit HS codes is obtained from trademap.
In regards with FDI data, Indonesia through National Single Windows of Indonesia (NSWI) provides data of FDI inflows across sector and countries which can be utilized for our FDI analysis at product level. We therefore can concord the HS product that subjected to AD with sectors in NSWI data, World Integrated Trade Solution (WITS) provides information about various product nomenclatures and the concordances between various product nomenclatures (See Appendix G for further detail).
4.2.2 Independent Variables
ADdummyjt - The main interest variable in this study is ADdummyjt which equal to 1 if the importer
imposes AD to the named-products from exporter in that year, and 0 if there is no AD imposition toward the exporter in that year. This variable will suggest the effectiveness of AD at the product level. According to figure 2, we predict that this variable will have negative value for trade analysis as the existence of trade destruction effect. Meanwhile for FDI analysis, we expect that this variable will be positive or there is tariff-jumping FDI.
For the other independent the data source and explanation are similar with our aggregate analysis (see section 3).
4.3 Methodology
23
average, this analysis will also eliminate the outliers whose fluctuation range exceeds 350% compared to the base year amount in the descriptive statistical analysis.
To find the impact of AD, the percentage of change in t1, t2, and t3 will be compared with the value
in the t0. t0 is the year of investigation which is assumed as one year before AD imposition.
Moreover, to find the impact of investigation we will compare t0 with t-1.
Furthermore, this study also conducts econometric analysis for AD case in Indonesia as shown in Table 7. We use Ordinary Least Square (OLS) for our estimation following Prusa (1995) and Lee et al (2013) for both import and FDI analysis.
4.4 Estimation Result and Analysis 4.4.1 Trade
Descriptive Analysis
To investigate the effect of AD, we plot percentage change of import data from 1 year prior and 3 years subsequent to AD decision. to refers to the year of investigation which is assumed as one
year before the imposition of AD, while t1, t2, and t3 represent the year of AD imposition. By doing
this comparison, we can analyze the impact of AD investigation and imposition toward import from the named country. If AD investigation and imposition are able to reduce import from named country (trade destruction effect), then AD becomes efficient.
Figure 3 presents the changes of import quantity from named country during the investigation year and 3 years after AD decision. As seen in the figure, import increases in the investigation year and starts to decrease in t2 after AD imposed. This trend suggests that trade destruction effect occurs.
However, while the import quantity from named country also decrease further in t2, import starts
to increase again in t3. The increasing import in t3 suggests that AD is not effective in this year.
Figure 3 Import Quantity of Named-Products from Named Country
Source: Author’s own calculation
Figure 4 shows the changes of import value from named country during the investigation year and 3 years after AD decision. The pattern is quite similar with the pattern of import quantity. We can see in the figure, relative to t-1, there is an increase of import value in the investigation year. The
import volume then decrease in t1 which suggests that there is trade destruction effect of AD. In
24
the t1, the decreasing of import value is less than the decreasing in import quantity which might
suggest that AD cause a higher unit value of the named products. In t1 onwards, import value
decreases gradually, which also suggest that trade destruction effect exists until t3.
Figure 4 Import Value of Named-Products from Named Country
Source: Author’s own calculation
Based on the descriptive analysis, we may expect that AD positively affect import quantity and value in the investigation year. This initial finding is one of the reasons why beside ADdummy this study also considers investigation period (dummy variable) in the regression model14. In general, figure 3 and 4 also indicates that AD is not effective in Indonesia. This initial indication is based on two reasons, first, trade destruction effect of AD imposition is limited in Indonesia case, and second, the quantity of import also increase under AD measures (t3). To find out whether the investigation and trade destruction effect occurs significantly, this study conducts a regression that the results are presented in the next section.
Econometric Analysis
Table 4 shows the estimation result about the impact of AD at product level. The dependent variables are value and quantity of import for named product from country i to country j. The second and third column shows that AD is not significant on both the value and quantity of the imports. To investigate the possibility of delay in the impact of AD, we also use the lags of
ADdummy in our estimation15. The lags of ADdummy also result insignificant on both value and
quantity of import. Thus, we do not have strong evidence that AD generates trade destruction
effect toward bilateral named-product import from the named-country exporter.
14 The specifications of the regression models by Prusa (1996, 2001), Brenton (2001), Lee, Park, and Cui (2013) do not consider the period of investigation, in contrast to Niels (2003) and Ganguli (2008) which includes these variables in the model.
15 We have done a regression for ADdummy
, ADdummyt-1, ADdummyt-2, ADdummyt-3 separately and they are still insignificant. We also compound all those variable in one regression as shown in Table 4, column (4) and (5) to look the effect of AD in a prolonged period of time.
25
Table 4 Estimation Result of Anti-dumping Impacts on Product Level Import
Variable Ln Value Ln Quantity Ln Value Ln Quantity Ln Value Ln Quantity
(1) (2) (3) (4) (5) (6) (7) ADdummy 0.273 (0.381) 0.233 (0.410) 0.143 (0.280) 0.060 (0.283) ADdummyt-1 0.148 (0.138) 0.204 (0.176) ADdummyt-2 0.0924 (0.153) 0.0524 (0.235) ADdummyt-3 -0.102 (0.210) -0.0403 (0.272) T0 0.664* (0.309) 0.730* (0.349) T1 0.326 (0.307) 0.202 (0.322) T2 0.213 (0.328) 0.163 (0.385) T3 0.276 (0.315) 0.161 (0.428) Ln(GDPit) 0.462 (0.331) -0.213 (0.302) 0.312 (0.477) -0.0841 (0.515) 0.494 (0.345) -0.180 (0.317) Ln(GDPjt) 0.572* (0.276) 0.451 (0.242) -0.117 (0.447) -0.179 (0.485) 0.579* (0.269) 0.457* (0.218) LnERij -1.011* (0.423) -1.275*** (0.367) -0.893 (0.506) -1.170* (0.494) -0.985* (0.424) -1.249*** (0.362) RTAij 0.524** (0.192) 0.419 0.231) 0.446 (0.249) 0.291 (0.290) 0.513** (0.186) 0.408 (0.223) Border 0.180 (0.202) 0.572* (0.229) 0.468 (0.242) 0.961* (0.292) 0.150 (0.215) 0.536 (0.242) Language -9.187*** (1.666) -11.19*** (1.552) -9.507*** (2.118) -12.41*** (2.074) -8.841*** (1.666) -10.79*** (1.569) Colony 2.786* (1.166) 3.417*** (0.943) 3.167* (1.333) 4.015** (1.232) 2.690* (1.174) 3.321*** (0.939) LnDistance -5.545*** (0.938) -6.532*** (0.818) -5.642*** (1.207) -7.080*** (1.138) -5.387*** (0.930) -6.351*** (0.810) Number of Obs 6376 5766 5273 5050 6376 5766 R-sq 0.494 0.506 0.501 0.510 0.495 0.507
Exporter Dummy Yes Yes Yes Yes Yes Yes
Year Dummy Yes Yes Yes Yes Yes Yes
HS Dummy Yes Yes Yes Yes Yes Yes
Note: i) Standard errors in parentheses* p<0.05, ** p<0.01, *** p<0.0010. ii) Standard error clustered by exporter. iii) to: year of investigation, t1-t2-t3: 1st, 2nd, 3rd year of imposition
Besides using ADdummy and its lag, as explained in the descriptive analysis before, we expect that there is an investigation effect and limited trade destruction effect. To test the significance, we generate dummy variables for t0, t1, t2, and t3 and test it in the model. The result, as presented in
26
product from the named-countries increases. However, we still do not have strong evidence that
trade destruction effect happens due to the insignificant of t0, t1, t2, and t3.
The insignificant of t1, t2, t3 are in line with the insignificant of AD imposition as reflected in
ADdummy variable (column 2 until 5). The result indicates that AD imposition in Indonesia has
no significant effect on import at the product level (no trade destruction effect) even when we accounted for phase-in effect. Empirically, the imposition of AD in Indonesia will only increase the import in the investigation year.
Using our gravity model, we have no evidence of trade destruction effect as the impact of AD imposition at product level. This result is different with our hypotheses. The result suggest that, as a net importer country, Indonesia still has high dependency on named-product’s import. The possible explanation for this result is Indonesian industry still cannot provide the named-product domestically (see Appendix H). In several cases, the domestic production for named-product is still under supply or under qualified. So, import product is still needed to meet national demand with or without AD.
Besides the main variable, we also have the result of other variables in our gravity model. According to the result, GDP exporter, exchange rate, colony, and distance significantly affect import at the product level as we have predicted. GDP exporter is positive significant as shown in column (2), (6), and (7). This suggests that if the economy size of the exporter country becomes bigger, it will increase the amount of import from the exporter. Similar to GDP exporter, colony also has a positive significant coefficient. In contrast, also as theory and previous literature expected, distance negatively affects import – confirming the well-known theory that the further away the exporting country is from the importing country (Indonesia), the lower the trade flows between them. The exchange rate shows also the expected negative sign as a depreciation of Indonesian currency against that of exporter country leads to smaller exports, and highly significant in all specification. The language variable turns out to be negative which is different what we have expected. The possible argument is a country which shares common language with Indonesia (i.e. Malaysia, Brunei Darussalam) has similar comparative advantage or product specialization. Thus, they trade less each other. The other possible explanation is ‘measurement artifact’, as we cannot guarantee that the data reflects a real condition in society.
4.4.2 FDI
Another unintended effect of AD is tariff-jumping FDI which is expected to weaken the direct effect of AD. We predict that AD will have a tariff-jumping FDI effect at product level. Table 5 shows the estimation result based on equation (4). According to column (2) and (6), the coefficient of ADdummy is positive significant on FDI inflows which suggests that an imposition of AD increases FDI inflows from the named-country exporter in the protected sector (tariff-jumping FDI
exists). Based on previous literature, this positive significance of AD suggests: first, products that
