![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
A growing body of literature has emerged on the links between free trade agreement (FTA) and agricultural export margins. As a key component of intensive margins, export survival and duration play essential roles in long-term and sustainable export growth. We investigated the impact of FTAs on agricultural export duration from the perspective of firm-destination pair. Based on highly disaggregated firm-level panel data of Chinese agricultural exporters covering the period of 2000–2016, we applied the proper survival method to empirically analyze the effect of FTA on export duration and survival. The results identify dynamic export stylized facts among firm-destination country pairs: although most of the agricultural firm-destination export relations demonstrated longer duration than those of other industries in China, they were still short-lived. Most importantly, we confirmed that FTA produced two opposite effects: Compared to non-FTA partners, the FTA membership role helps extend the length of export duration for agricultural export relation in general. However, for newly created relations, the implementation of FTA increased the hazard rate of export failure and shortened the length of duration. Our results cast a new light on firm-level export survival issue and FTA policy, and provide implications for both agricultural exporters and policymakers.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 RTA mean any preferential agreement for members including free trade agreements (FTAs) as well as preferential trade agreements (PTAs), partial scope agreements (PSAs), customs unions (CUs) and economic integration agreements (EIAs). These different typologies of RTA are categorized by the depth of economic integration and differ on the issues of trade liberalization degree. FTAs varies among the agreements themselves, to help eliminate restrictions between bilateral trade but also to maintain respective external trade restrictions.
2 There is a simple rule regarding the objectives of any international trade negotiations or agreements: Exports are good and export growth contributes to greater economic growth (Krugman Citation1991; Greenaway, Morgan, and Wright Citation1998). Therefore, it is reasonable for governors and scholars to investigate FTA strategy through export growth.
3 China has grown into a major competitor in the world agricultural market. According to the WTO Trade Database, China has been the fourth largest agricultural exporter since 2010, after the EU, the USA and Brazil.
4 As indicated by Besedeš and Prusa (Citation2011), export duration is an intensive margin for export growth in time dimension, even a small difference in survival rates and duration could lead to significant differences in the long-run export growth.
5 According to heterogeneous firm trade theory, the involvement in the destination country market is predicted to continuing exporting with high persistence due to the existence of sunk costs for entry and exit (Melitz Citation2003; Medin Citation2003). Nevertheless, for instance Volpe Martincus and Carballo (Citation2008), Görg, Kneller, and Muraközy (Citation2012) and Esteve-PÉrez, Requena-Silvente, and PallardÓ-Lopez (Citation2013), found that most firm-level export relationships exited the destination market quickly and few survived for a long duration.
6 Recent studies explored the impact of particular FTA (e.g., Besedeš, for North American Free Trade Agreement [NAFTA]; Bojnec and Imre, for EU, Yang and Inmaculada, for Asian–China) on export performance.
7 Currently, it is easy to notice that China has signed FTAs with diverse partners, including both developing counties and developed countries, the neighboring countries and geographically distant countries, as well as the agricultural trade surplus partners and agricultural trade deficit partners (according to the net of bilateral agricultural trade flows). With the discussion focus on a typical developing country, China for instance, there are sets of FTAs with agricultural terms that vary across the FTAs of South–South cooperation (e.g., the China–Pakistan FTA) and North–South cooperation (e.g., the China–New Zealand FTA) which have distinct features in certain different contents.
8 This fact does not only apply to China’s exporters. Takahashi and Urata (Citation2010) concluded that the utilization rates of FTAs are also not optimistic for Japan’s firms. According to a recent survey conducted by Klynveld Peat Marwick Goerdeler (KPMG), only 22% of Asian firms have taken advantage of the preferential policies of FTAs.
9 e.g., market access to particular agricultural product, special trade subsidies and remedies, requirements of sanitary and phytosanitary (SPS) measures and technical barriers to trade (TBT) in agreements.
10 By this way, any firm that exports agricultural products would be chosen to our dataset. The range of agricultural product ruled in external trade statistic system is published by Ministry of Agricultural and Rural Affairs of the People’s Republic of China (See agricultural products and corresponding HS codes are shown specifically in appendix B). Note that the HS code system for product categorization has been adjusted in 1996, 2002, 2007 and 2012 in origin dataset. To keep the consistency of the agricultural product categorization, we use the conversion tables to convert HS 1996/2007/2012 codes into the HS 2002 codes. These conversion tables can be extracted from the UN Comtrade Website: https://comtrade.un.org/data/doc/api/#classifications.
11 China’s FTA information is available at http://fta.mofcom.gov.cn/ (accessed at 1 January 2020).
12 Although time is continuous, so is export duration, information records of export relation current state can only be observed on a yearly basis in our data. This is also the main reason why we use discrete-time hazards models later.
13 During such a long period, the exporter name and ID code would have changed if restructured. Based on other firm information provided by raw data, we observe a certain agricultural firm not only by its name, also through its special code, telephones and emails. Hence, we would identify the same exporter to the greatest extent and measure the export duration more precisely.
14 Note that multiple spells for one same trade relationship might violate the assumption that all export spells are conditional independence on both the observed and unobserved effects.
15 This finding is consistent with Chen, Li, and Zhou’s research (2012), an empirical investigation based on entire China’s exporters across all industries.
16 To address the multiple spell issues, we do not treat all export spells as independent spells in our analysis. In many empirical studies, export experience has a significant effect on trade survival. There are some alternative spells used in robustness check and the results keep consistent (see more in Section 5.2).
17 Many econometricians have proven that adequate prerequisites needed to solve the left-censoring problem efficiently, including making sufficient assumptions on the baseline hazards or using supplementary data (Rosholm, Citation2002). As these prerequisites are not available based on our dataset, we decided to drop particular spells to minimize the possible bias.
18 Note that durations of export are grouped into yearly intervals but FTA implemented at a specified date, here we identify the subsamples B1 and B2 by regarding the year of FTA implementation as a benchmark. That is, if the export relation started in the year that FTA become effective, it would be allocated to post-FTA implementation, the group B2. This criterion seems quite reasonable. As the existence of warm-up time before any FTAs coming into force officially, export spells could be influenced to some extent no matter whether it started after the actual implementation day or not within a certain period of time.
19 See similar findings documented in Pakes and Ericson (Citation1998), Esteve–Pérez et al. (Citation2013), and Brenton et al. (Citation2010).
20 For a more detailed introduction of this test, see in Cleves, Gould, and Gutierrez (Citation2004).
21 As shown in the Section 3.3 stylized facts, nearly half of the exports ceased after the first year of export.
22 The clog–log model specification makes the proportionality assumption of extreme-value minimum distribution, which is an analog of the discrete-time Cox proportional hazards model. Additionally, probit and logit specification can also be applied if the distribution form of the hazard rate encounters the normal and logistic distribution, respectively. We also use these alternative distribution assumptions to check the robustness in Section 5.3.
23 By 2016, the FTA regions of China included 11 countries (the FTAs with Hong Kong, China; and Macao, China are excluded).
24 Since all export firms of each trade relationship belong to China, the characteristic variables of the exporting countries are consistent. Thus, the model does not include the any control variables for China.
25 According to Baldwin and Krugman (Citation1989), a firm's export decisions are history-dependent to maintain export status in the world market.
26 Some empirical studies, however, have come to the opposite conclusion for Linder’s proposition or failed to support it (Kennedy and McHugh Citation1983; Linnemann and van Beers Citation1988).
27 In addition to China’s customs dataset and the FTAs Websites mentioned in Section 3 as our data sources, the Food and Agriculture Organization (FAO), World Data Bank (WDB), Penn World Table and CEPII also provided free access to other important statistics for our explanatory variables. See more in Table E1.
28 Taking multiple spells into consideration might violate the assumption that all export spells in the sample are independent. Researchers have argued that the former export experience plays an essential role in later export decisions through a learning effect (Araujo, Mion, and Ornelas Citation2016). Although Besedeš and Prusa (Citation2006a) argued that this had little effect on the final result in their analysis, here, we attempt to explore whether our results were robust if alternative samples were applied to the original empirical models.
29 This data was originally collected and issued monthly by Ministry of Commerce of People's Republic of China. See more information in http://wms.mofcom.gov.cn/article/ztxx/ncpmy/.