3.1 Measurement of the GVCs of Agricultural Products

A GVC position index refers to the relative position (upstream or downstream) of a country (or industry) in a GVC when participating in global economic activities. Previous studies have adopted a variety of methods to estimate a country's position in GVCs. These methods include: (1) the traditional export trade data statistics method (Qiu et al., Reference Qiu, Ye and Sun2012); (2) the vertical specialization index method (Hummels et al., Reference Hummels, Ishii and Yi2001; Yu and Deng, Reference Yu and Deng2014); (3) the value-added trade measurement method (Koopman et al., Reference Koopman, Powers, Wang and Wei2010, Koopman et al., Reference Koopman, Wang and Wei2014); (4) the net trade index (NET) measure method (Tang, Reference Tang2015); and (5) the import and export unit price comparison method (Mulder et al., Reference Mulder, Paillacar and Zignago2009; Lin and He, Reference Lin and He2015). Following Koopman et al. (Reference Koopman, Powers, Wang and Wei2010) and Wang (Reference Wang2014), the global value chain of agricultural products is interpreted as the global chain of added value generated from international trade in the agricultural products. Hence, the added value obtained by a country from international trade is used to calculate the GVC position of the country at a sectoral level such as agriculture. Assume the world includes G countries and each country produces tradable goods in N sectors. In each sector, goods can be used as intermediate inputs in production activities and to meet final demand from both domestic and foreign economic agents:

(1)$$x_{is} = \mathop \sum \limits_{\,j = 1}^N a_{ijss}x_{\,js} + \mathop \sum \limits_{r\ne s}^G \mathop \sum \limits_{\,j = 1}^N a_{ijsr}x_{\,jr} + y_{iss} + \mathop \sum \limits_{r\ne s}^G y_{isr}, \;\;\;i\in ( {1, \;N} ) , \;s\in ( {1, \;G} ) $$

where x_is is the total output of sector i in country s; a_ijsr is the direct input–output (IO) coefficient describing units of the product of sector i produced in country s and consumed in the production of one unit gross output of sector j in country r; and y_isr is the final demand in country r for goods of sector i produced in country s. Eq. (1) can be written in matrix form,

(2)$$\left[{\matrix{ {X_1} \cr {\matrix{ {X_2} \cr {X_3} \cr {X_4} \cr } } \cr } } \right] = \left[{\matrix{ {\matrix{ {I-A_{11}} & {-A_{12}} \cr } } & \cdots & {-A_{1G}} \cr {\matrix{ {-A_{21}} & {I-A_{22}} \cr } } & \cdots & {-A_{2G}} \cr {\matrix{ {\matrix{ \cdots \cr {-A_{G1}} \cr } } & {\matrix{ \cdots \cr {-A_{G2}} \cr } } \cr } } & {\matrix{ \cdots \cr \cdots \cr } } & {\matrix{ \cdots \cr {I-A_{GG}} \cr } } \cr } } \right]^{{-}1}\left[{\matrix{ {\mathop \sum \limits_r^G Y_{1r}} \cr {\mathop \sum \limits_r^G Y_{2r}} \cr {\matrix{ \cdots \cr {\mathop \sum \limits_r^G Y_{Gr}} \cr } } \cr } } \right]$$

where X_s is the sectoral gross output vector of country s; A_sr is the IO coefficient matrix indicating intermediate inputs from country s to country r Footnote ¹; and y_sr is the vector of final demand in country r produced in country s. The Leontief inverse matrix (or total requirement matrix) can be expressed by

$$B = \left[{\matrix{ {\matrix{ {B_{11}} & {B_{12}} \cr } } & \cdots & {B_{1G}} \cr {\matrix{ {B_{21}} & {B_{22}} \cr } } & \cdots & {B_{2G}} \cr {\matrix{ {\matrix{ \cdots \cr {B_{G1}} \cr } } & {\matrix{ \cdots \cr {B_{G2}} \cr } } \cr } } & {\matrix{ \cdots \cr \cdots \cr } } & {\matrix{ \cdots \cr {B_{GG}} \cr } } \cr } } \right] = \left[{\matrix{ {\matrix{ {I-A_{11}} & {-A_{12}} \cr } } & \cdots & {-A_{1G}} \cr {\matrix{ {-A_{21}} & {I-A_{22}} \cr } } & \cdots & {-A_{2G}} \cr {\matrix{ {\matrix{ \cdots \cr {-A_{G1}} \cr } } & {\matrix{ \cdots \cr {-A_{G2}} \cr } } \cr } } & {\matrix{ \cdots \cr \cdots \cr } } & {\matrix{ \cdots \cr {I-A_{GG}} \cr } } \cr } } \right]^{{-}1}$$

where B_sr denotes the N × N block Leontief inverse matrix. Furthermore, we define a direct value-added coefficient matrix by

(3)$$\hat{V} = \left[{\matrix{ {\matrix{ {\widehat{{V_1}}} & 0 \cr 0 & {\widehat{{V_2}}} \cr } } & {\matrix{ \cdots & 0 \cr \cdots & 0 \cr } } \cr {\matrix{ \cdots & \cdots \cr 0 & 0 \cr } } & {\matrix{ \cdots & \cdots \cr \cdots & {\widehat{{V_G}}} \cr } } \cr } } \right]_{( {GN \times GN} ) }$$

where each element ($\hat{V}_s$) shows the share of direct domestic value added in total output in country s. Similarly, the global export matrix can be defined by

(4)$$\hat{E} = \left[{\matrix{ {\matrix{ {\widehat{{E_1}}} & 0 \cr 0 & {\widehat{{E_2}}} \cr } } & {\matrix{ \cdots & 0 \cr \cdots & 0 \cr } } \cr {\matrix{ \cdots & \cdots \cr 0 & 0 \cr } } & {\matrix{ \cdots & \cdots \cr \cdots & {\widehat{{E_G}}} \cr } } \cr } } \right]_{( {GN \times GN} ) }$$

where each element ($\hat{E}_s$) shows sectoral gross exports in country s. Hence, the value-added of country-sectoral exports can be decomposed as follows:

(5)$$\hat{V}B\hat{E} = \left[{\matrix{ {\matrix{ {{\hat{V}}_1B_{11}{\hat{E}}_1} & {{\hat{V}}_1B_{12}{\hat{E}}_2} \cr {{\hat{V}}_2B_{21}{\hat{E}}_1} & {{\hat{V}}_2B_{22}{\hat{E}}_2} \cr } } & {\matrix{ \cdots & {{\hat{V}}_1B_{1G}{\hat{E}}_G} \cr \cdots & {{\hat{V}}_2B_{2G}{\hat{E}}_G} \cr } } \cr {\matrix{ \cdots & \cdots \cr {{\hat{V}}_GB_{G1}{\hat{E}}_1} & {{\hat{V}}_GB_{G2}{\hat{E}}_2} \cr } } & {\matrix{ \cdots & \cdots \cr \cdots & {{\hat{V}}_GB_{GG}{\hat{E}}_G} \cr } } \cr } } \right]_{( {GN \times GN} ) }$$

where each element ($\hat{V}_rB_{rs}\hat{E}_s$) is an N × N block matrix, representing the domestic sectoral value-added in country r embodied in the exports of country s. To sum up the off-diagonal elements along a row, we obtain a country's indirect value-added (IV) embodied in exports of other countries (Eq. (6)). Similarly, to sum up off-diagonal elements along a column, we obtain the foreign value-added (FV) embodied in a country's gross exports (Eq. (7)).

(6)$$IV_r = \hat{V}_r\mathop \sum \limits_{s\ne r}^G ( {B_{rs}{\hat{E}}_s} ) $$

(7)$$FV_r = \mathop \sum \limits_{s\ne r} ( {{\hat{V}}_sB_{sr}} ) \hat{E}_r$$

If the IV of a country is high relative to its gross exports, then a relatively large share of the country's exports comes from demand for domestic intermediates induced by the exports of other countries, which indicates the country's upstreamness in the GVC. On the contrary, if the FV of a country is high relative to its gross exports, then a relatively large share of the country's exports goes to the value-added of other countries that export to the country's intermediates, which indicates the country's downstreamness in the GVC. Hence, both indicators should be considered when determining a country's position in the GVC. Following Koopman et al. (Reference Koopman, Powers, Wang and Wei2010), the position in the GVC at a country-sector level is defined by the index

(8)$$GVC\_position_{ir} = Ln\left({1 + \displaystyle{{IV_{ir}} \over {E_{ir}}}} \right)-Ln\left({1 + \displaystyle{{FV_{ir}} \over {E_{ir}}}} \right)$$

where i represents sector i and r represents country r. The index is valid only if a country participates in international trade, i.e., has positive exports. If the index is large, the country lies upstream in the GVC in the sense that the country obtains more value-added from its exports of intermediates compared to the value-added that another country receives from its exports. If the index is small, the country lies downstream in the GVC. If a country moves from downstream to upstream in the GVC, the country becomes relatively more important as a destination to which other countries export their products, implying a stronger position in the GVC.

In this study, since the Global Trade Analysis Project (GTAP) model described in Section 3.2 does not directly report the value of the GVC position index, we use the relevant results from the model to calculate the GVC based on the above-mentioned method. We compare values of the GVC position index in alternative scenarios to identify changes of a country's position in the agricultural GVC.

3.2 GTAP Model

This study adopts the static version of the GTAP model, a global multi-region, multi-sector CGE model (Hertel, Reference Hertel1997). Key assumptions of this model include: (1) complete market competition; (2) constant return to scale in production; (3) cost minimization for producers; (4) utility maximization for consumers; and (5) market clearing for all products and production factors. The model allocates household consumption, savings, and government consumption to regions according to Cobb–Douglas functions, while distinguishing between domestic and imported products. In particular, the model describes the bilateral import and export between regions. More details about the GTAP model can be found at Hertel et al. (Reference Hertel, McDougall and Walmsley2007).

The GTAP static model in this study is built with a base year of 2011. The base year is only two years ahead of 2013, when the BRI was officially announced, providing a plausible benchmark economy without any BRI impact. To isolate the BRI impacts from any other factors affecting the GVC positions, a comparative static analysis is conducted to examine the BRI impacts by simulating reduction in trade barriers in the 2011 economy in alternative scenarios. We acknowledge that it would be more interesting to provide estimates for the coming years by extending the static model to a dynamic model, to better simulate the impact over time of investments in, for example, infrastructure. However, this study intends to explore the impact of removal of trade barriers rather than the impact of investments. In our case, a dynamic model needs a schedule of stepwise reductions of trade tariffs and non-tariff barriers, which is not available and difficult to assume.

3.3 Data

The GTAP model in this study is based on the GTAP 9 database, which contains the 2011 data for 140 countries (or regions), 57 industries (or products), and eight basic production factors (Aguiar et al., Reference Aguiar, Narayanan and McDougall2016). We aggregate the regions, industries, and production factors to simplify the calculation while still meeting the needs of this study.

In this study, the 65 BR countries are aggregated into seven regions: Central Asia Mongolia and the Commonwealth of Independent States (MCIS), ASEANFootnote ², West Asia, South Asia, Central and Eastern Europe, and China. All the other regions are treated as one region (rest of the world – ROW).

The economy of each country is divided into three major sectors: agriculture, manufacturing, and services. Agriculture is further divided into nine sub-sectors: cereal crops (CCPs); vegetables, fruit, and nuts (VFN); oil crops (OV); sugar (SU); other crops (OCPs); livestock meat (LM); fishery (FSH); forestry (FRS); and milk and dairy products (MD). The detailed classifications are shown in Table 1.

Table 1.

Aggregated regions and sectors based on the GTAP database classification

Note: *In this article, the TB reduction in Central and Eastern Europe is the same as in other regions. This article assumes that these states face lower TB and NTB than other EU member states, although some Central and Eastern Europe countries are part of the EU. These eastern EU member states are affected by BRI much more than the other EU member states in any case. Hence, this article assumes that Central and Eastern Europe has the same TB reduction as other BR regions.

Source: GTAP v9 database and own aggregation.

For 2011, the populations, savings, imports and exports, and tariffs of the BR regions are shown in Table 2. The population in South Asia was slightly more than that in China, while the population in ASEAN is slightly less than half of China's population. On the contrary, the amount of savings in China is nine times that in South Asia and 6.6 times that in ASEAN, indicating a dominant role of China in available funding for investments. The regions with the highest total imports among all the BR regions are China, ASEAN, and West Asia, which are also the only regions with total imports exceeding USD 1,000 billion. The regions with the lowest total imports are Central Asia, MCIS, and South Asia, of which Central Asia is the only region that does not exceed USD 100 billion. China's total imports are more than 20 times those of Central Asia, 3 times those of MCIS, and nearly 2.5 times those of South Asia; they account for approximately 8% of the world's total imports. The total imports of all the BR regions accounted for approximately 31% of the world's total imports. In addition, China, ASEAN, and West Asia were the regions with the highest total exports among all the BR regions. The regions with the lowest total exports were also Central Asia, MCIS, and South Asia. China's total exports accounted for about 10% of the world's total exports; these exports exceeded 16 times those of Central Asia. The total exports of all the BR regions accounted for approximately 34% of the world's total exports. The weighted average import tariff varied among the BR regions. The highest tariff level was the import tariff of China against South Asia, where the weighted average import tariff exceeds 200%. The second is ASEAN's import tariffs on South Asia, with a weighted average tariff of 194%. Only a few regions had an import tariff of less than 10%; as an example of these few regions, the import tariff of MCIS on Central Asia was only 1%.

Table 2.

Economic situation of the regions along the route in 2011

Source: GTAP v9 database and own aggregation.

3.4 Trade Barriers in the GTAP Model and Potential Impacts on GVC Positions

We assume that all products being traded among the BR countries enjoy the reduction of trade barriers assumed in the scenarios described in Section 4. In the GTAP model adopted in this study, trade barriers are represented by two parameters of bilateral tariff and non-tariff levels. A change in the tariff level leads to changes in the prices of imported products, which subsequently affect the quantity of the imported products. The NTBs are indicated in the GTAP model by the parameter representing the impact of foreign technology on the country, an increase of which reduces export volume and prices of the corresponding products. The effects of both trade barriers disturb the whole global economic system and finally the economy reaches an equilibrium different from the original one. By comparing the two equilibria, we identify the impact of the reduction in trade barriers for the BR countries on the GVCs of agricultural products.

Notice that the reduction in trade barriers reduces the purchasing prices of imported products and induces more demand for imports to a country. From the perspective of other countries, this means an increase in demand for their exported products, which pushes up the prices of the products being exported from these countries. Hence, the reduction in trade barriers generally reduces the purchasing prices in the importing countries while increasing the producer prices in the exporting countries. If a country has both imports and exports of specific products, then how prices of the product evolve in its domestic market depends on whether the effect of the imports is stronger than that of the exports.

According to the definition in Eq. (8), the position of a country in a GVC depends on exports of all regions and on prices in the international market, which modify both the IV embodied in exports of other countries and the FV embodied in a country's gross exports. A removal of trade barriers among the BR countries can encourage exports among these countries. If barriers are reduced, more intermediate products from one country will be demanded due to the additional exports of the other countries, which leads to more IV embodied in the exports of other countries. On the other hand, additional exports from one country require more intermediate products from other countries, which induces more FV embodied in a country's gross exports. In addition, the changes in prices can affect exports, IV and FV. Hence, it is unclear whether the GVC position of a BR country will move up or not following a removal of trade barriers among the BR countries. However, the GVC positions of the BR countries as a group and of other countries are respectively enhanced and weakened since a removal of trade barriers among the BR countries discourages their trade with the other countries.

4.1 Best Estimate Scenario

This study mainly focuses on the Best Estimate Scenario, where reductions in both TB and NTB are estimated from historical data and previous studies. In the long term, the BRI countries are expected to enter agreements similar to FTAs (State Council, 2015). Hence, in the Best Estimate Scenario, we assume that all BR countries enjoy the same reduction in trade barriers as those BR countries that currently have FTAs with China. Since the relevant adjustment parameters of TB and NTB are built into the model, this article needs to calculate the specific reduction values of TB and NTB, and then adjust the two parameters in the model. Changes in TB and NTB in the model can have an impact on import and export prices, income, welfare, etc. This affects total volumes of imports and exports until the model reaches a new general equilibrium. Although we consider the joint impacts of the reduced TB and NTB in the ‘Best Estimate’ scenario, the reduced TB and NTB may have different impacts on the economies of different BR countries, as indicated by the simulation results reported in the appendix.

The reduction in TB is estimated by a simplified method derived from Petri et al. (Reference Petri, Plummer and Zhai2012) based on the FTAs that have been signed with China. In our estimate, the tariff reduction is assumed symmetrical in bilateral relationships. Because of data availability, we follow previous studies (Huang and Yu, Reference Huang and Yu2018) to assume that the tariffs of agricultural products in all the BR countries have been reduced at the same rate, which is comprehensively measured from current FTAs between China and 13 FTA countries (or regions) as shown in Table 4,

(9)$$\Delta tariff = \displaystyle{{\sum tariff_i \times im_i^{2017} } \over {\sum im_i^{2017} }}$$

where Δtariff is the estimated overall tariff reduction in all 65 BR countries (or the 7 BR regions in our model); Δtariff_i is the average tariff reduction of China in the FTA with country i, expressed as the share of imports of all FTA duty-free goods in total imports; and $im_i^{2017} $ represents the imports from country i to China in 2017. Our calculation shows that the average tariff reduction among the BR countries is 92%, meaning 92% imports of all products in the bilateral trade among the BR countries are tax free. In our model, the tariff on all products is assumed to be reduced by 92%.

Table 4.

The FTAs between China and other BR countries (or regions) as of 2018

Note: Tariff reduction is based on data from China FTA Network (http://fta.mofcom.gov.cn/english/index.shtml), accessed on 26 January 2019. Imports in 2017 are from the UN Comtrade database (https://comtrade.un.org/data/).

We also consider reduction in NTB among the BR countries based on the iceberg trade cost approach. Following Petri et al. (Reference Petri, Plummer and Zhai2012), we divide the aspects covered by the FTAs into 24 issue areas and score the reduction in NTB of each area based on the content of the FTAs. The overall reduction of the NTB is then calculated as the weighted sum of the reduction in NTB across the 24 areas. Specifically, the NTB reduction (ΔNTB) is associated with three factors: average score S _j, reflecting subjective assessments of the issues in the FTAs; policy weight P _j, indicating relative importance of different FTAs for NTS; and coefficient λ,⋅

(10)$$\Delta NTB = \lambda \mathop \sum \limits_{\,j\in A} P_jS_j$$

The values of policy weights and average scores are directly from Petri et al. (Reference Petri, Plummer and Zhai2012). The coefficient λ is calculated using the iceberg trade cost based on the Overall Trade Restrictiveness Index (OTRI) (Huang and Yu, Reference Huang and Yu2018). The OTRI data are derived from Looi Kee, et al. (Reference Looi Kee, Nicita and Olarreaga2009) by estimating the ad-valorem equivalents (AVEs) of NTBs for each country at the tariff line level. The OTRI index summarizes the impact of each country's trade policies on its imports, depending on the overall level of protection and the elasticity of import demand and imports. In this paper, the OTRI of a country is weighted by the shares of imports from other countries relative to the gross imports of the country in the base year 2011, i.e.,

(11)$${\rm \lambda } = \mathop \sum \limits_{i\in N} OTRI_i \times w_i^{import} $$

where OTRI _i is the overall trade restrictiveness index of country i, and $w_i^{import} $ indicates the share of imports from country i relative to the gross imports of all countries in the collection N, where the countries in collection N are the BR countries that have OTRI.

Based on Eq. (10), we adopt the higher scores of S _j from Petri et al. (Reference Petri, Plummer and Zhai2012) and the data of imports and OTRIs of countries in collection N (Table 5) to calculate the overall reduction of NTB as 15.36%, which is assumed for all the BR countries in the Best Estimate Scenario.

Table 5.

Countries’ imports and OTRI

Source: The OTRI data are derived from the study by Looi Kee et al. (Reference Looi Kee, Nicita and Olarreaga2009). Imports in 2011 are from the UN Comtrade database (https://comtrade.un.org/data/).

4.2 Two Alternative Scenarios

In the best estimate scenario, we assume a certain level of reduction in trade barriers. However, the assumption might be too optimistic or too pessimistic from the perspective of the potential influence of the BRI on trade barrier reduction. Hence, we also consider two alternative scenarios: one is the pessimistic scenario and the other is the optimistic scenario.

4.21 Pessimistic Scenario

In this scenario, we assume that the effect of BRI on trade tariffs had been fully realized by 2019, meaning there will be no further tariff reductions due to BRI after 2019. We determine the reduction in trade barriers by comparing the trade barriers among the BR countries before and after the BRI was proposed in 2013 (Table 6). We obtain the average import tariff data from the BR countriesFootnote ³ to China in the WTO database and calculate the average tariff in 2011 weighted by the imports of each country in 2011, the same as the base year in our model. The average tariff is taken as the tariff level of China and these BR countries before the BRI. The tariff level after the BRI is then calculated as the average tariff data of the same BR countries still weighted by the imports in 2011 based on the latest tariff data from the WTO database as of February 2019. By comparing the tariff levels before and after the BRI, we obtain an average reduction in trade tariffs of only 1.01%, which is assumed for all the BR countries in the pessimistic scenario.

Table 6.

National tariff reduction and trade data along the BRI route

Source: Tariff level is based on data from WTO Network (http://tao.wto.org/report/TariffAverages.aspx), accessed on 22 February 2019.

Tariff reduction data are from author calculation and the negative signs signal tariff increases. China's imports to the country in 2011 are from the UN Comtrade database (https://comtrade.un.org/data/).

As for the calculation of NTB, the calculation method of formula (10) is still adopted. In order to compare with the Best Estimate Scenario, we adopt the lower scores of S _j from Petri et al. (Reference Petri, Plummer and Zhai2012) for the calculation. The reduction of NTB in this scenario appears to be only 10.56%.

4.2.2 Optimistic Scenario

This scenario follows the typical assumptions in previous studies to set the tariffs between the BR countries as zero. Table 7 shows the differences between the weighted average import tariffs of various industries and regions between the optimistic and pessimistic scenarios. For the non-tariff barriers, a reduction by 20% is assumed, which is almost double the reduction level in the pessimistic scenario.

Table 7.

The differences in tariff cuts between the pessimistic and the best estimate scenario. %

4.3 Key Differences between Scenarios

Table 8 summarizes the key assumptions about the reduced TB and NTB in the three scenarios mentioned above. Compared to the ‘best estimate’ scenario, the TB are reduced to a much larger extent than the NTB in the Pessimistic scenario, while the opposite is true in the optimistic scenario.

Table 8.

The assumed reductions of TB and NTB in alternative scenarios