- Email: [email protected]
Does the use of multiple FTAs force firms to raise local input share?: Evidence of the spaghetti bowl phenomenon
Economic Modelling 33 (2013) 458–461
Contents lists available at SciVerse ScienceDirect
Economic Modelling journal homepage: www.elsevier.com/locate/ecmod
Does the use of multiple FTAs force firms to raise local input share?: Evidence of the spaghetti bowl phenomenon☆ Kazunobu Hayakawa Bangkok Research Center, Japan External Trade Organization, 16th Floor, Nantawan Building, 161 Rajadamri Road, Pathumwan, Bangkok 10330, Thailand
a r t i c l e
i n f o
Article history: Accepted 22 April 2013 JEL classification: F15 F53 O53
a b s t r a c t This paper empirically investigates the firm-level relationship between the local input share and the number of used FTAs by employing the data on FTA utilization in Japanese affiliates in ASEAN. As a result, we do not find a robust linear relationship. However, affiliates using a large number of FTAs (seven or eight) have an extremely higher share of local inputs. This result might be interpreted as the first evidence of the “spaghetti bowl phenomenon”. © 2013 Elsevier B.V. All rights reserved.
Keywords: FTA Spaghetti bowl phenomenon
1. Introduction The surge of free trade agreements (FTAs) in each country has publicly yielded concerns on firms' inactive use of FTAs. According to the World Trade Organization (WTO) website, as of January 2012, around 500 regional trade agreements have been notified to the WTO. In order to use FTA schemes, firms need to meet the rules of origin (ROOs). To do that, the users may need to change their procurement sources from the optimal pattern of procurement. In particular, for the use of multiple FTA schemes, firms may raise the share of local inputs in total inputs because local inputs can always meet ROOs in any kinds of FTAs. If so, though the net benefits from such multiple-FTA use must be positive, some amount of benefits will be offset by the change of procurement sources from the optimal pattern of procurement. We may say that this offset is one form of the “spaghetti bowl phenomenon”.1 This paper empirically investigates whether the use of multiple FTAs forces firms to raise their local input share in total inputs. To do that, we employ the unique dataset, which has been collected by the Japan External Trade Organization (JETRO) in 2011 and 2012. In that survey, the very detailed firm-level information on FTA use is available
☆ I would like to thank an anonymous referee of this journal. The opinions expressed in this paper are those of the author and do not represent the views of any of the institutions with which I am affiliated. E-mail address: [email protected]. 1 As for the more precise concept of the spaghetti bowl phenomenon, see Bhagwati et al. (1998). 0264-9993/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.econmod.2013.04.041
according to FTA partner countries, in addition to some basic information on firms' activities such as employment, the breakdown of their export destinations, and their procurement sources. In the academic literature on FTAs, there are few studies on firm-level FTA utilization. Demidova and Krishna (2008) is the theoretical study which demonstrates that only the productive firms can use FTA schemes in exporting. To our best knowledge, Takahashi and Urata (2010) is the only empirical study at the firm level. They examine FTA usage by Japanese firms at the firm level by employing a questionnaire survey, finding that larger firms are more likely to use FTA schemes. Against this literature on firm-level FTA utilization, our paper is the first one that examines the firm-level use of multiple-FTAs. The rest of this paper is organized as follows. The next section provides our conceptual framework on the relationship of firms' multiple-FTA use with their local input share. Section 3 explains our empirical framework and data structure and presents an overview on firms' use of multiple FTAs. After providing the results of econometric analysis in Section 4, Section 5 concludes on this paper. 2. Conceptual framework This section presents a conceptual framework describing the relationship between firms' local input shares and the FTAs they use. For this purpose, we first consider firms' decisions concerning use of an FTA for exporting. We then extend consideration to the context of use of multiple FTA schemes. A firm will use an FTA when exporting its output if the gain from using the FTA is greater than the cost of using it. The gain depends on how much the firm can save on its tariff payment. Roughly speaking, the reduction in the tariff payment depends on the following two
K. Hayakawa / Economic Modelling 33 (2013) 458–461
459
Table 1 Average employment according to number of used FTAs. Number of used FTAs
IDN
KHM
LAO
MMR
MYS
PHL
SGP
THA
VNM
All
0 1 2 3 4 5 6 7 8
966 1406 1268 688 360 364
128 350 601
60 440
439 875 600
393 457 429 610 405 719 4135 1274 89
899 1322 1019 584
265 187 214 71 259 141 264 113 95
599 413 759 1394 2738 1812 550 3102
752 471 691 674 1625 515
615 597 742 917 1553 994 1492 998 93
Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
3. Empirical framework Fig. 1. Number of sample affiliates according to the number of FTAs. Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
elements: One is tariff margin, which is the difference between the FTA rate and the most-favored nation rate; the larger the tariff margin, the larger the reduction in the tariff payment. The second element is the size of exports because the use of FTA schemes for the larger size of exports, which is likely in the case of larger-sized firms, leads to a bigger reduction of tariff payment. On the other hand, the cost for the use of an FTA is the “procurement adjustment cost,” in addition to any administrative cost. To use an FTA, a firm needs to comply with the rules of origin of its product. In order to comply with the ROOs requirements, a firm may need to change its procurement sources from optimal pattern of procurement. We call such extra procurement costs the “procurement adjustment cost”; the lower the procurement adjustment cost, the greater the incentive for a firm to use an FTA. Thus, ROOs are among the crucial elements affecting a firm's decision to use an FTA. There are several types of ROOs. “Change in tariff classification criterion” and “value added content criterion” are the major ones. The former criterion determines the country of origin of a good according to whether the tariff classification assigned to the final goods produced in a country differs from the tariff classification of the input goods. The latter criterion determines the country of origin according to whether output production in the country adds sufficient value to the inputs used, usually 40% of the price of final goods. In FTAs concluded by ASEAN, an optional criterion tends to be adopted, which determines the country of origin of a good according to whether it meets either a value-added content criterion or a change in tariff classification criterion. Some FTAs, such as the ASEAN–China FTA, adopt a value-added content criterion. When firms export their products to multiple countries and further use FTA schemes when exporting to those countries, they need to satisfy ROOs in all FTAs that they use. If firms try to satisfy a value-added content criterion in all FTAs, it is difficult for them to use many imported inputs. For example, suppose that firms in Country A export to Countries B and C, which are, respectively, bilateral FTA partners for Country A. In addition, ROOs in those FTAs are assumed to use a value-added content criterion. Then, inputs imported from Country B cannot be regarded as having originated in Country A under the bilateral FTA between Countries A and C. Therefore, the higher share of inputs from Country B makes it more difficult for firms in Country A to satisfy ROOs when exporting to Country C. The same is true for the case of using inputs from Country C when exporting from Country A to Country B. The use of local inputs (i.e., inputs from Country A) is more likely to enable firms to satisfy ROOs under the FTAs with both Countries B and C, since inputs from Country A are regarded as having originated there under both bilateral FTAs. As a result, when multiple, bilateral FTA schemes are in place, firms are likely to use a relatively high share of local inputs.
This section explains the empirical framework we use to investigate the relationship between the share of local inputs in total inputs and the FTAs used. Specifically, we estimate the following simple equation. Local Input Shareijct ¼ β1 Number of FTAsit þ β2 ln Employmentit þβ3 ln Ageit þ uj þ uct þ εijct : ð1Þ Our dependent variable is the share of local inputs in total inputs in affiliate i in year t. Its industry and host country are denoted by j and c, respectively. Our main interest is the coefficient of “Number of FTAsit,” which is the number of FTAs that affiliate i uses in exporting in year t. Employmentit and Ageit are affiliate i's employment and operation duration in year t, respectively. uj and uct are industry dummy and host country-year dummy variables, respectively. εijct is an error term. There are some papers examining the determinants of the share of inputs that overseas affiliates import. Hanson et al. (2005) investigate the roles of host countries' factor prices and trade costs while controlling for overseas affiliates' output volumes. While lower factor prices lead to lower prices for local inputs, raising local input shares, lower trade costs also encourage multinationals to use more imported inputs. In addition, Kiyota et al. (2008) examine the role of affiliate age, which is expected to be positively correlated with local input share because of accumulation of knowledge concerning host economies. In our specification, Eq. (1), we use employment as a proxy for output volumes. Following Kiyota et al. (2008), affiliate age is also included. All host country characteristics, such as factor prices or trade costs, are controlled for by introducing host country-year dummy variables. Industry dummy variables control for the remaining variation across industries. The industry dummy variables also help control for differences in difficulty meeting ROOs across industries, and the host country-year dummy variables help control for differences in the number of available FTAs among countries.
Table 2 Average share of local inputs according to number of used FTAs (%). Number of used FTAs
IDN
KHM
LAO
MMR
MYS
PHL
SGP
THA
VNM
All
0 1 2 3 4 5 6 7 8
0.41 0.38 0.44 0.64 0.35 0.90
0.12 0.30 0.01
0.00 0.00
0.06 0.01 1.00
0.42 0.43 0.43 0.48 0.40 0.31 0.68 0.51 0.37
0.29 0.22 0.21 0.22
0.23 0.31 0.26 0.53 0.42 0.76 0.63 0.75 0.82
0.54 0.56 0.52 0.58 0.51 0.47 0.57 0.50
0.27 0.28 0.20 0.23 0.07 0.08
0.44 0.45 0.44 0.49 0.42 0.44 0.62 0.63 0.64
Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
460
K. Hayakawa / Economic Modelling 33 (2013) 458–461
As mentioned in the introductory section, our main data source is JETRO's “Survey of Japanese-Affiliated Firms in ASEAN” for 2010 and 2011. For example, in the survey for 2010, the questionnaires were sent to 4164 Japanese affiliates operating in seven ASEAN countries (Thailand, Malaysia, Singapore, Indonesia, the Philippines, Vietnam, and Myanmar), receiving 1865 valid responses. Of those, 1111 were from Japanese affiliates in the manufacturing industry. Our use of foreign plants' data enables us to avoid including firms not using FTA schemes because of their lack of knowledge on what FTAs are. Firms that invest abroad will have enough knowledge on FTAs because such companies are familiar with international activities and are also sensitive to the available tools in those activities. In short, foreign plants are more likely to behave rationally in their decisions on FTA use. Our focus on affiliates in ASEAN countries enables us to control for heterogeneity in rules of FTAs (e.g. ROOs) to some extent because ASEAN countries try to harmonize their respective FTAs among ASEAN members. In addition to the ASEAN Free Trade Area (AFTA), as of our sample period, ASEAN countries have bilateral or multilateral FTAs with Australia, China, India, Japan, Korea, and New Zealand. Also, while Malaysia concludes its FTA with Pakistan, Singapore does so on its FTAs with Chile, Peru, and the United States. As a result, according to countries in which affiliates locate, the number of their potentially available FTAs is different. The sample affiliates for our analysis are restricted only to those in the manufacturing industry. 2 Next, we make an overview of affiliates' use of FTAs. Fig. 1 depicts the number of affiliates according to the number of FTAs that they use, indicating that nearly 70% of sample affiliates are non-users of FTAs. Table 1 reports the relationship between the average employment and the number of FTAs that an affiliate uses. In the “All” column, we cannot find its clear relationship. Such an unclear relationship can be seen even when taking a look at the relationship by country. Table 2 reports the relationship between the average share of local inputs and the number of FTAs that an affiliate uses. In the “All” column, affiliates using more than five FTA schemes seem to have a relatively high share of local inputs. We may find a relatively clear positive relationship in Singapore.
In the above analyses, we did not find a robust linear relationship between the local input share and the number of FTAs. Finally, we investigate its non-linear relationship. In column (I) in Table 4, we introduce two dummy variables. While one is a dummy variable taking the value of one for the use of one or two FTA schemes, the other takes the value of one for the use of three or more FTA schemes. As is consistent with our expectation, the coefficient for the latter dummy variable is significantly positive and larger than that for the former dummy variable. In column (II) in Table 4, we also introduce dummy variables according to the number of used FTAs. It indicates that affiliates using more than six FTA schemes have a 20%–30% higher share of local inputs than do non-users of FTA schemes and the other affiliates, except for affiliates using three FTA schemes. It is puzzling that in spite of the significant result in the case of three FTA schemes, the results in the cases of four, five, and six FTA schemes are insignificant. The results on the other variables (i.e. employment and age) are not changed qualitatively.
4. Econometric results
5. Discussion
This section reports our econometric results on the relationship between firms' local input share and the number of FTAs that they use. The results of ordinary least square (OLS) estimation are provided in Table 3. As is consistent with our expectation, we can see that the number of FTAs has a significant association with the local input share. Also, affiliate age has a significant coefficient: The older affiliates are more likely to have a higher share of local inputs perhaps because of their greater knowledge on local suppliers. The firm size in terms of employment does not significantly affect the local input share. Such an unclear result in firm size is also consistent with the result in Hanson et al. (2005). 3 Next, we apply the instrument variable method to our model in order to take care of the simultaneity problem between local input share and the number of used FTAs. We use the number of destinations to which affiliates export as an instrument, which seems to be related to the number of used FTAs but not at least directly to the local input share. The results are reported in Table 3. The coefficient for the number of used FTAs turns out to be insignificant. The other interesting finding is that both the affiliate size and the number of export destinations are positively associated with the number of used FTAs.
Our findings for the analysis of Japanese affiliates in ASEAN are that there is no robust linear relationship between the local input share and the number of FTAs, but that affiliates using an extremely large number of FTAs in our sample have a high share of local inputs. In sum, when firms use so many FTA schemes at the same time, they
2
The industry classification in this dataset is rough; food industry, textile industry, wearing apparel, wooden products, furniture, paper industry, chemical industry, plastic products, medicine, rubber products, pottery, iron and steel, non-metallic mineral products, metal products, general machinery industry, electric machinery industry, transport equipment, precision machinery industry, and other manufacturing sectors. 3 We also estimate this model by the fractional logit technique, but the results are qualitatively unchanged. For the easier interpretation on marginal effect, we report only OLS results.
Table 3 Baseline estimation. Estimation method
OLS
Instrument variable method
Dependent variable
Local input share
Number of FTAs
Number of FTAs
0.011* [0.006] 0.0003 [0.006] 0.032*** [0.012]
ln employment ln age Number of export destination Industry dummy Country ∗ year dummy Observations R-squared
YES YES 2090 0.1913
0.117*** [0.020] 0.049 [0.040] 0.183*** [0.021] YES YES 2090 0.1830
Local input share 0.031 [0.024] −0.003 [0.007] 0.032*** [0.012]
YES YES 2090 0.1868
Notes: The parentheses are robust standard errors. *** and ** show 1% and 5% significance, respectively.
Table 4 OLS estimation according to number of FTAs. (I)
Number of FTAs = 1 or 2 = 3 or above =1 =2 =3 =4 =5 =6 =7 =8 ln employment ln age Dummy variables Industry Country ∗ year Observations R-squared
(II)
Coef.
R.S.E.
−0.025 0.057**
[0.015] [0.024]
−0.0000 0.034*** YES YES 2090 0.1936
[0.006] [0.012]
Coef.
R.S.E.
−0.020 −0.038 0.059** −0.016 0.058 0.167 0.200* 0.314** −0.0002 0.036***
[0.017] [0.024] [0.030] [0.045] [0.075] [0.125] [0.110] [0.159] [0.006] [0.012]
YES YES 2090 0.1968
Notes: The dependent variable is a share of local inputs in total inputs. R.S.E. indicates robust standard errors. *** and ** show 1% and 5% significance, respectively.
K. Hayakawa / Economic Modelling 33 (2013) 458–461
may need to raise their local input share so as to meet ROOs in all FTA schemes. In this sense, some amount of benefits from the use of multiple FTAs is offset by the change of procurement sources from the optimal pattern of procurement. Against this, there are two noteworthy points. First, we need to investigate our hypothesis with a larger number of sample firms. As confirmed in Fig. 1, the number of affiliates using seven or eight FTA schemes is less than 1% of our sample affiliates. Second, this weak evidence on the rise of local input share might be specific to FTAs concluded by ASEAN countries. As pointed out in Cadot and de Melo (2007), ROOs in their FTAs are much less restrictive than those in other FTAs existing in the world. Thus, it is important to investigate the same kind of hypothesis for the other regions, including America. References Bhagwati, J., Greenaway, D., Panagariya, A., 1998. Trading preferentially: theory and policy. The Economic Journal 108, 1128–1148. Cadot, O., de Melo, J., 2007. Why OECD countries should reform rules of origin. World Bank Research Observer 23 (1), 77–105. Demidova, S., Krishna, K., 2008. Firm heterogeneity and firm behavior with conditional policies. Economics Letters 98 (2), 122–128.
461
Hanson, G., Mataloni, R., Slaughter, M., 2005. Vertical production networks in multinational firms. The Review of Economics and Statistics 87 (4), 664–678. Kiyota, K., Matsuura, T., Urata, S., Wei, Y., 2008. Reconsidering the backward vertical linkages of foreign affiliates: evidence from Japanese multinationals. World Development 36 (8), 1398–1414. Takahashi, K., Urata, S., 2010. On the use of FTAs by Japanese firms: further evidence. Business and Politics 12 (1), 2. He is a researcher of the Bangkok Research Center, Japan External Trade Organization. He received his doctorate from Keio University with a dissertation on intermediate goods trade in 2008. His main research areas are international economics and economic geography. He has published in journals such as the Journal of Economic Surveys, the Journal of the Japanese and International Economies, the North American Journal of Economics and Finance, and the Economics Letters.
Contents lists available at SciVerse ScienceDirect
Economic Modelling journal homepage: www.elsevier.com/locate/ecmod
Does the use of multiple FTAs force firms to raise local input share?: Evidence of the spaghetti bowl phenomenon☆ Kazunobu Hayakawa Bangkok Research Center, Japan External Trade Organization, 16th Floor, Nantawan Building, 161 Rajadamri Road, Pathumwan, Bangkok 10330, Thailand
a r t i c l e
i n f o
Article history: Accepted 22 April 2013 JEL classification: F15 F53 O53
a b s t r a c t This paper empirically investigates the firm-level relationship between the local input share and the number of used FTAs by employing the data on FTA utilization in Japanese affiliates in ASEAN. As a result, we do not find a robust linear relationship. However, affiliates using a large number of FTAs (seven or eight) have an extremely higher share of local inputs. This result might be interpreted as the first evidence of the “spaghetti bowl phenomenon”. © 2013 Elsevier B.V. All rights reserved.
Keywords: FTA Spaghetti bowl phenomenon
1. Introduction The surge of free trade agreements (FTAs) in each country has publicly yielded concerns on firms' inactive use of FTAs. According to the World Trade Organization (WTO) website, as of January 2012, around 500 regional trade agreements have been notified to the WTO. In order to use FTA schemes, firms need to meet the rules of origin (ROOs). To do that, the users may need to change their procurement sources from the optimal pattern of procurement. In particular, for the use of multiple FTA schemes, firms may raise the share of local inputs in total inputs because local inputs can always meet ROOs in any kinds of FTAs. If so, though the net benefits from such multiple-FTA use must be positive, some amount of benefits will be offset by the change of procurement sources from the optimal pattern of procurement. We may say that this offset is one form of the “spaghetti bowl phenomenon”.1 This paper empirically investigates whether the use of multiple FTAs forces firms to raise their local input share in total inputs. To do that, we employ the unique dataset, which has been collected by the Japan External Trade Organization (JETRO) in 2011 and 2012. In that survey, the very detailed firm-level information on FTA use is available
☆ I would like to thank an anonymous referee of this journal. The opinions expressed in this paper are those of the author and do not represent the views of any of the institutions with which I am affiliated. E-mail address: [email protected]. 1 As for the more precise concept of the spaghetti bowl phenomenon, see Bhagwati et al. (1998). 0264-9993/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.econmod.2013.04.041
according to FTA partner countries, in addition to some basic information on firms' activities such as employment, the breakdown of their export destinations, and their procurement sources. In the academic literature on FTAs, there are few studies on firm-level FTA utilization. Demidova and Krishna (2008) is the theoretical study which demonstrates that only the productive firms can use FTA schemes in exporting. To our best knowledge, Takahashi and Urata (2010) is the only empirical study at the firm level. They examine FTA usage by Japanese firms at the firm level by employing a questionnaire survey, finding that larger firms are more likely to use FTA schemes. Against this literature on firm-level FTA utilization, our paper is the first one that examines the firm-level use of multiple-FTAs. The rest of this paper is organized as follows. The next section provides our conceptual framework on the relationship of firms' multiple-FTA use with their local input share. Section 3 explains our empirical framework and data structure and presents an overview on firms' use of multiple FTAs. After providing the results of econometric analysis in Section 4, Section 5 concludes on this paper. 2. Conceptual framework This section presents a conceptual framework describing the relationship between firms' local input shares and the FTAs they use. For this purpose, we first consider firms' decisions concerning use of an FTA for exporting. We then extend consideration to the context of use of multiple FTA schemes. A firm will use an FTA when exporting its output if the gain from using the FTA is greater than the cost of using it. The gain depends on how much the firm can save on its tariff payment. Roughly speaking, the reduction in the tariff payment depends on the following two
K. Hayakawa / Economic Modelling 33 (2013) 458–461
459
Table 1 Average employment according to number of used FTAs. Number of used FTAs
IDN
KHM
LAO
MMR
MYS
PHL
SGP
THA
VNM
All
0 1 2 3 4 5 6 7 8
966 1406 1268 688 360 364
128 350 601
60 440
439 875 600
393 457 429 610 405 719 4135 1274 89
899 1322 1019 584
265 187 214 71 259 141 264 113 95
599 413 759 1394 2738 1812 550 3102
752 471 691 674 1625 515
615 597 742 917 1553 994 1492 998 93
Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
3. Empirical framework Fig. 1. Number of sample affiliates according to the number of FTAs. Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
elements: One is tariff margin, which is the difference between the FTA rate and the most-favored nation rate; the larger the tariff margin, the larger the reduction in the tariff payment. The second element is the size of exports because the use of FTA schemes for the larger size of exports, which is likely in the case of larger-sized firms, leads to a bigger reduction of tariff payment. On the other hand, the cost for the use of an FTA is the “procurement adjustment cost,” in addition to any administrative cost. To use an FTA, a firm needs to comply with the rules of origin of its product. In order to comply with the ROOs requirements, a firm may need to change its procurement sources from optimal pattern of procurement. We call such extra procurement costs the “procurement adjustment cost”; the lower the procurement adjustment cost, the greater the incentive for a firm to use an FTA. Thus, ROOs are among the crucial elements affecting a firm's decision to use an FTA. There are several types of ROOs. “Change in tariff classification criterion” and “value added content criterion” are the major ones. The former criterion determines the country of origin of a good according to whether the tariff classification assigned to the final goods produced in a country differs from the tariff classification of the input goods. The latter criterion determines the country of origin according to whether output production in the country adds sufficient value to the inputs used, usually 40% of the price of final goods. In FTAs concluded by ASEAN, an optional criterion tends to be adopted, which determines the country of origin of a good according to whether it meets either a value-added content criterion or a change in tariff classification criterion. Some FTAs, such as the ASEAN–China FTA, adopt a value-added content criterion. When firms export their products to multiple countries and further use FTA schemes when exporting to those countries, they need to satisfy ROOs in all FTAs that they use. If firms try to satisfy a value-added content criterion in all FTAs, it is difficult for them to use many imported inputs. For example, suppose that firms in Country A export to Countries B and C, which are, respectively, bilateral FTA partners for Country A. In addition, ROOs in those FTAs are assumed to use a value-added content criterion. Then, inputs imported from Country B cannot be regarded as having originated in Country A under the bilateral FTA between Countries A and C. Therefore, the higher share of inputs from Country B makes it more difficult for firms in Country A to satisfy ROOs when exporting to Country C. The same is true for the case of using inputs from Country C when exporting from Country A to Country B. The use of local inputs (i.e., inputs from Country A) is more likely to enable firms to satisfy ROOs under the FTAs with both Countries B and C, since inputs from Country A are regarded as having originated there under both bilateral FTAs. As a result, when multiple, bilateral FTA schemes are in place, firms are likely to use a relatively high share of local inputs.
This section explains the empirical framework we use to investigate the relationship between the share of local inputs in total inputs and the FTAs used. Specifically, we estimate the following simple equation. Local Input Shareijct ¼ β1 Number of FTAsit þ β2 ln Employmentit þβ3 ln Ageit þ uj þ uct þ εijct : ð1Þ Our dependent variable is the share of local inputs in total inputs in affiliate i in year t. Its industry and host country are denoted by j and c, respectively. Our main interest is the coefficient of “Number of FTAsit,” which is the number of FTAs that affiliate i uses in exporting in year t. Employmentit and Ageit are affiliate i's employment and operation duration in year t, respectively. uj and uct are industry dummy and host country-year dummy variables, respectively. εijct is an error term. There are some papers examining the determinants of the share of inputs that overseas affiliates import. Hanson et al. (2005) investigate the roles of host countries' factor prices and trade costs while controlling for overseas affiliates' output volumes. While lower factor prices lead to lower prices for local inputs, raising local input shares, lower trade costs also encourage multinationals to use more imported inputs. In addition, Kiyota et al. (2008) examine the role of affiliate age, which is expected to be positively correlated with local input share because of accumulation of knowledge concerning host economies. In our specification, Eq. (1), we use employment as a proxy for output volumes. Following Kiyota et al. (2008), affiliate age is also included. All host country characteristics, such as factor prices or trade costs, are controlled for by introducing host country-year dummy variables. Industry dummy variables control for the remaining variation across industries. The industry dummy variables also help control for differences in difficulty meeting ROOs across industries, and the host country-year dummy variables help control for differences in the number of available FTAs among countries.
Table 2 Average share of local inputs according to number of used FTAs (%). Number of used FTAs
IDN
KHM
LAO
MMR
MYS
PHL
SGP
THA
VNM
All
0 1 2 3 4 5 6 7 8
0.41 0.38 0.44 0.64 0.35 0.90
0.12 0.30 0.01
0.00 0.00
0.06 0.01 1.00
0.42 0.43 0.43 0.48 0.40 0.31 0.68 0.51 0.37
0.29 0.22 0.21 0.22
0.23 0.31 0.26 0.53 0.42 0.76 0.63 0.75 0.82
0.54 0.56 0.52 0.58 0.51 0.47 0.57 0.50
0.27 0.28 0.20 0.23 0.07 0.08
0.44 0.45 0.44 0.49 0.42 0.44 0.62 0.63 0.64
Source: Survey of Japanese-affiliated firms in ASEAN, India, and Oceania (JETRO).
460
K. Hayakawa / Economic Modelling 33 (2013) 458–461
As mentioned in the introductory section, our main data source is JETRO's “Survey of Japanese-Affiliated Firms in ASEAN” for 2010 and 2011. For example, in the survey for 2010, the questionnaires were sent to 4164 Japanese affiliates operating in seven ASEAN countries (Thailand, Malaysia, Singapore, Indonesia, the Philippines, Vietnam, and Myanmar), receiving 1865 valid responses. Of those, 1111 were from Japanese affiliates in the manufacturing industry. Our use of foreign plants' data enables us to avoid including firms not using FTA schemes because of their lack of knowledge on what FTAs are. Firms that invest abroad will have enough knowledge on FTAs because such companies are familiar with international activities and are also sensitive to the available tools in those activities. In short, foreign plants are more likely to behave rationally in their decisions on FTA use. Our focus on affiliates in ASEAN countries enables us to control for heterogeneity in rules of FTAs (e.g. ROOs) to some extent because ASEAN countries try to harmonize their respective FTAs among ASEAN members. In addition to the ASEAN Free Trade Area (AFTA), as of our sample period, ASEAN countries have bilateral or multilateral FTAs with Australia, China, India, Japan, Korea, and New Zealand. Also, while Malaysia concludes its FTA with Pakistan, Singapore does so on its FTAs with Chile, Peru, and the United States. As a result, according to countries in which affiliates locate, the number of their potentially available FTAs is different. The sample affiliates for our analysis are restricted only to those in the manufacturing industry. 2 Next, we make an overview of affiliates' use of FTAs. Fig. 1 depicts the number of affiliates according to the number of FTAs that they use, indicating that nearly 70% of sample affiliates are non-users of FTAs. Table 1 reports the relationship between the average employment and the number of FTAs that an affiliate uses. In the “All” column, we cannot find its clear relationship. Such an unclear relationship can be seen even when taking a look at the relationship by country. Table 2 reports the relationship between the average share of local inputs and the number of FTAs that an affiliate uses. In the “All” column, affiliates using more than five FTA schemes seem to have a relatively high share of local inputs. We may find a relatively clear positive relationship in Singapore.
In the above analyses, we did not find a robust linear relationship between the local input share and the number of FTAs. Finally, we investigate its non-linear relationship. In column (I) in Table 4, we introduce two dummy variables. While one is a dummy variable taking the value of one for the use of one or two FTA schemes, the other takes the value of one for the use of three or more FTA schemes. As is consistent with our expectation, the coefficient for the latter dummy variable is significantly positive and larger than that for the former dummy variable. In column (II) in Table 4, we also introduce dummy variables according to the number of used FTAs. It indicates that affiliates using more than six FTA schemes have a 20%–30% higher share of local inputs than do non-users of FTA schemes and the other affiliates, except for affiliates using three FTA schemes. It is puzzling that in spite of the significant result in the case of three FTA schemes, the results in the cases of four, five, and six FTA schemes are insignificant. The results on the other variables (i.e. employment and age) are not changed qualitatively.
4. Econometric results
5. Discussion
This section reports our econometric results on the relationship between firms' local input share and the number of FTAs that they use. The results of ordinary least square (OLS) estimation are provided in Table 3. As is consistent with our expectation, we can see that the number of FTAs has a significant association with the local input share. Also, affiliate age has a significant coefficient: The older affiliates are more likely to have a higher share of local inputs perhaps because of their greater knowledge on local suppliers. The firm size in terms of employment does not significantly affect the local input share. Such an unclear result in firm size is also consistent with the result in Hanson et al. (2005). 3 Next, we apply the instrument variable method to our model in order to take care of the simultaneity problem between local input share and the number of used FTAs. We use the number of destinations to which affiliates export as an instrument, which seems to be related to the number of used FTAs but not at least directly to the local input share. The results are reported in Table 3. The coefficient for the number of used FTAs turns out to be insignificant. The other interesting finding is that both the affiliate size and the number of export destinations are positively associated with the number of used FTAs.
Our findings for the analysis of Japanese affiliates in ASEAN are that there is no robust linear relationship between the local input share and the number of FTAs, but that affiliates using an extremely large number of FTAs in our sample have a high share of local inputs. In sum, when firms use so many FTA schemes at the same time, they
2
The industry classification in this dataset is rough; food industry, textile industry, wearing apparel, wooden products, furniture, paper industry, chemical industry, plastic products, medicine, rubber products, pottery, iron and steel, non-metallic mineral products, metal products, general machinery industry, electric machinery industry, transport equipment, precision machinery industry, and other manufacturing sectors. 3 We also estimate this model by the fractional logit technique, but the results are qualitatively unchanged. For the easier interpretation on marginal effect, we report only OLS results.
Table 3 Baseline estimation. Estimation method
OLS
Instrument variable method
Dependent variable
Local input share
Number of FTAs
Number of FTAs
0.011* [0.006] 0.0003 [0.006] 0.032*** [0.012]
ln employment ln age Number of export destination Industry dummy Country ∗ year dummy Observations R-squared
YES YES 2090 0.1913
0.117*** [0.020] 0.049 [0.040] 0.183*** [0.021] YES YES 2090 0.1830
Local input share 0.031 [0.024] −0.003 [0.007] 0.032*** [0.012]
YES YES 2090 0.1868
Notes: The parentheses are robust standard errors. *** and ** show 1% and 5% significance, respectively.
Table 4 OLS estimation according to number of FTAs. (I)
Number of FTAs = 1 or 2 = 3 or above =1 =2 =3 =4 =5 =6 =7 =8 ln employment ln age Dummy variables Industry Country ∗ year Observations R-squared
(II)
Coef.
R.S.E.
−0.025 0.057**
[0.015] [0.024]
−0.0000 0.034*** YES YES 2090 0.1936
[0.006] [0.012]
Coef.
R.S.E.
−0.020 −0.038 0.059** −0.016 0.058 0.167 0.200* 0.314** −0.0002 0.036***
[0.017] [0.024] [0.030] [0.045] [0.075] [0.125] [0.110] [0.159] [0.006] [0.012]
YES YES 2090 0.1968
Notes: The dependent variable is a share of local inputs in total inputs. R.S.E. indicates robust standard errors. *** and ** show 1% and 5% significance, respectively.
K. Hayakawa / Economic Modelling 33 (2013) 458–461
may need to raise their local input share so as to meet ROOs in all FTA schemes. In this sense, some amount of benefits from the use of multiple FTAs is offset by the change of procurement sources from the optimal pattern of procurement. Against this, there are two noteworthy points. First, we need to investigate our hypothesis with a larger number of sample firms. As confirmed in Fig. 1, the number of affiliates using seven or eight FTA schemes is less than 1% of our sample affiliates. Second, this weak evidence on the rise of local input share might be specific to FTAs concluded by ASEAN countries. As pointed out in Cadot and de Melo (2007), ROOs in their FTAs are much less restrictive than those in other FTAs existing in the world. Thus, it is important to investigate the same kind of hypothesis for the other regions, including America. References Bhagwati, J., Greenaway, D., Panagariya, A., 1998. Trading preferentially: theory and policy. The Economic Journal 108, 1128–1148. Cadot, O., de Melo, J., 2007. Why OECD countries should reform rules of origin. World Bank Research Observer 23 (1), 77–105. Demidova, S., Krishna, K., 2008. Firm heterogeneity and firm behavior with conditional policies. Economics Letters 98 (2), 122–128.
461
Hanson, G., Mataloni, R., Slaughter, M., 2005. Vertical production networks in multinational firms. The Review of Economics and Statistics 87 (4), 664–678. Kiyota, K., Matsuura, T., Urata, S., Wei, Y., 2008. Reconsidering the backward vertical linkages of foreign affiliates: evidence from Japanese multinationals. World Development 36 (8), 1398–1414. Takahashi, K., Urata, S., 2010. On the use of FTAs by Japanese firms: further evidence. Business and Politics 12 (1), 2. He is a researcher of the Bangkok Research Center, Japan External Trade Organization. He received his doctorate from Keio University with a dissertation on intermediate goods trade in 2008. His main research areas are international economics and economic geography. He has published in journals such as the Journal of Economic Surveys, the Journal of the Japanese and International Economies, the North American Journal of Economics and Finance, and the Economics Letters.