3

Tracing Value Added in the Presence of Multinational Firms with an Application to High-Tech Sectors

Paper Authors: Zhi Wang (George Mason University and University of International Business and Economics), Shang-Jin Wei (Columbia Business School, National Bureau of Economic Research [NBER], and Centre for Economic Policy Research [CEPR]), Xinding Yu (University of International Business and Economics), and Kunfu Zhu (Renmin University)

Presenter: Shang-Jin Wei (Columbia Business School, NBER, and CEPR)
Moderator: Sally Thompson, Bureau of Economic Analysis (retired)

The workshop paper by Zhi Wang, Shang-Jin Wei, Xinding Yu, and Kunfu Zhu explores the accurate measurement of global value chains (GVC) activity, which is vital to the validity and success of future international trade research.

Shang-Jin Wei, the N. T. Wang professor of Chinese business and economy and professor of finance and economics at Columbia University’s Graduate School of Business and School of International and Public Affairs, introduced the presentation by describing the goals and features of the paper. The paper develops a new accounting framework to better account for the role of foreign direct investment (FDI) in GVC accounting systems, consistent with the national accounts conceptual framework. The paper offers two accounting decompositions, which collapse to a single formula in matrix notation: a forward and backward linkage.

The paper’s major contribution is the addition of FDI-related activity to a GVC accounting framework. The authors found that previous estimates of GVC activities that do not account for the role of foreign-invested firms miss about half of GVC activities. Specifically, the authors found that GVC activity involving foreign-invested enterprises (FIEs) accounts for about 10 percent of global gross domestic product (GDP), similar in magnitude to the entirety of GVC activities identified in current approaches used to measure GVC activities using conventional input-output tables (that do not separately identify the contribution made by MNEs and foreign affiliates). These additional GVC activities are larger in high-tech sectors and higher-income countries.

The measurement framework in this paper decomposes a country’s GDP and final production into different value-added terms related to pure domestic activities, trade-related GVC activities that involve trade in intermediate inputs but not through FDI, and FDI-related GVC activities. The value of their approach is illustrating the importance of FDI in determining participation in GVCs and, in turn, informing bilateral investment treaties.

Wei indicated the importance of accounting for FDI by showing two data patterns:

• First are the trends in FDI stock and trade volumes. Both are generally increasing with time, but the global FDI stock has increased at a quicker and growing rate. Both are related to GVC activity, so why are they increasing at different rates?
• Second, about 60 percent of U.S. multinational corporation–affiliate sales in 2016 were in the host country of the affiliate. This pattern displays the importance of host-country sales in the GVC.

PAST LITERATURE AND CONTRIBUTION

Wei began the review of the literature on GVC accounting with Hummels and colleagues (2001), who introduced two measures of trade in intermediate goods. Koopman and colleagues (2012) updated this framework to include domestic value added. Noguera (2012) proposed a net measure of value added in trade, but did not capture the entirety of domestic value added in exports. Koopman and colleagues (2014) next proposed a framework containing four mutually exclusive terms that account for all gross exports. The framework accounts for domestic value added that is exported and consumed abroad, domestic value added that is exported and then imported, foreign value added used in the production of exports, and double-counted items.

Following the work of Koopman and colleagues (2014), there are two strands of literature. The first strand studies the decomposition of gross trade. Wei cited the following papers without additional comment: Nagengast and Stehrer (2016), Johnson (2018), Borin and Mancini (2019), Arto and colleagues (2019), and Miroudot and Ye (2020). The second strand of literature studies the decomposition of value added beyond trade. Wei cited Los and colleagues (2015), who decomposed final good production but made no distinction between GVC and non-GVC production activity. Wei also cited Wang and colleagues (2017), who decomposed GDP and final good production in a fashion that measures GVC activity.

According to Wei, this workshop paper contributed to the second strand of literature with three major contributions: (1) adding to the literature on measuring GVC activity using intercountry input-output (ICIO) tables by studying the contribution of FIEs, (2) contributing to literature on FDI in global production fragmentation and GVCs, and (3) quantifying FDI-related GVC activity by country income and industry research and development (R&D) intensity.

ACCOUNTING FRAMEWORK

Wei briefly discussed the underlying methodology. The authors started from ICIO tables with information on firm ownership. There are G economies, N industries, and two ownership types. This creates ICIO tables in which firm behavior is split by ownership, separating the total output of foreign- and domestic-held firms. The decomposition framework follows Wang and colleagues (2017), the major contribution of which is a single-equation decomposition “V̂LŶ.” This decomposition is extended in the workshop paper to include FDI.

For simplicity, Wei focused on forward-linkage decomposition only. The decomposition containing FDI includes 16 terms that can be characterized into five categories: pure domestic production activity, traditional trade production activity, trade-related GVC activity, FDI-related GVC activity, and trade- and FDI-related GVC activity. Wei claimed that the FDI-related GVC activity is missed in current estimates of GVC activity that do not fully account for the contribution of value added generated through FDI channels to GVCs.

Wei explained that the 16 terms can be compressed into nine components. These nine components are split into non-GVC activity and GVC activity. Three of the components are classified as missing GVC activity. The missing activity centers on FIE production and value added for final goods consumed in the host country.

DATA AND NUMERICAL APPLICATION

The workshop paper uses the Analytical Activities of Multinational Enterprises (AMNE) database from the Organisation for Economic Co-operation and Development (OECD). This database has the form of an ICIO table broken down by domestic- and foreign-owned firms. The dataset includes 59 countries and a rest-of-the-world composite and covers 34 industries from 2005 to 2016.

The paper shows that the largest contribution to global GDP is made by pure domestic activity, which appears to be countercyclical. The remainder of the global GDP can be expressed in the four other categories. FDI-related GVC activity is the largest nondomestic component of global GDP. Trade-related GVC activity is the next largest, contributing about 2 percent less to global GDP per year than the FDI GVC contribution. Non-GVC trade accounts for about 5.5 percent of global GDP across all years; this is about 4 percent less than the FDI-related GVC activity in a given year. The trade measures of GVC activity—pure and hybrid trade and FDI—shrank during the global financial crash in 2008. The GVC activity related to FDI did not have a strong response to the financial crash, and Wei suggested this may be an interesting channel for future macroeconomic research.

The results Wei presented are heterogeneous across countries in the sample. FDI-related GVC activity plays a larger role in smaller and more open economies, such as those of Hong Kong, the Czech Republic, Singapore,

Romania, and Hungary. In such countries as South Korea, Russia, Japan, Saudi Arabia, and Israel, the FDI-related component of GVC activity contributes less to GDP.

Wei explained that variation due to industry is also present in the sample. For this analysis, countries are split into low, middle, and high income, and industries are split into medium-low, medium, and high tech. Two patterns arise in the data:

• First, for high-income countries, the GVC-activity share of value added is highest in high-tech industries and lowest in low-tech industries. This result is robust across the three income groups. This trend is also true when considering only the FDI-related share of GVC activity.
• Second, in high–R&D-intensity industries, FDI GVC activity is more important than trade-related GVC, whereas in medium- to low-tech industries, the GVC components are equal.

SUMMARY

Wei concluded by summarizing the paper’s main results. According to Wei, the paper develops a framework to trace value added that accounts for FIEs. He indicated that the framework allows for both forward and backward linkage; however, he focused his discussion on the results from the forward-linked decomposition. He and his coauthors use the Analytical AMNE database from the OECD to study how the introduction of FDI-related GVC activity alters the existing estimates of the share of GVC activity in global GDP, finding that prior methods miss 9–10 percent of GVC activity contributing to global GDP. After correcting for this missing activity, the total contribution of GVC activity to global GDP is about 20 percent—double the prior estimates. There are some country and industry patterns to the missing data. The volume of missing activity is largest in high-tech industries and high-income countries, relative to medium-tech industries and middle-income countries. There is no formal hypothesis testing in this paper but the decomposition results provide a more accurate measurement of GVC activity that can be used in future work. Finally, Wei concluded that their future work may include incorporating bilateral FDI stock into the accounting framework, as well as decomposing gross exports.

DISCUSSION

Discussant: Thomas F. Howells III (Bureau of Economic Analysis, U.S. Department of Commerce)

Thomas Howells, chief of the Industry Analysis Division of the Bureau of Economic Analysis (BEA) in the U.S. Department of Commerce, began his discussion by acknowledging the contributions of the paper and providing some

background information to contextualize it in the current literature on measuring GVC activity. Howells stated that the paper is well placed in the literature and builds on the use of trade in value added (TiVA) statistics. The backward-linkage approach would take a final product and look backwards at its supply chain, whereas the forward-linkage approach would look at a given industry and analyze GVC activity in a forward direction. The literature typically offers a decomposition of either exports or final demand. Howells explained that this paper contributes by developing a framework containing both forward and backward linkages, as well as a decomposition of final demand. A decomposition based on exports is listed as future work by the authors.

Howells observed the importance of the ICIO tables to the literature, noting that this paper uses an expanded ICIO, including multiple other dimensions for variation—namely domestic- and foreign-owned firms. To help understand the main contribution, which is the framework itself, Howells presented a simple example of product flows that the framework would capture. In this example, a U.S. firm purchases intermediates from a domestic and a foreign firm. This U.S. firm may then trade intermediate goods back to these firms and/or may sell final products directly to consumers, both domestically and internationally. In this setting, U.S consumers may buy final products from the foreign firm as well.

Howells pointed out that standard TiVA analysis defines GVC as cross-border trade in intermediate goods and services, breaking trade into three categories: domestic production and consumption, pure trade of final goods and services, and GVCs. While the implied assumption in this framework indicates that a firm’s location is what matters, the paper posits that firm ownership is also an important dimension of analysis.

Howells described the empirical results presented in the paper as interesting and important to future work. The main takeaway from the empirical results is that current TiVA analysis misses about 10 percent of the FDI-related GVC activity in terms of global GDP.

Howells offered several comments and critiques. The first comment concerned the Analytical AMNE database from the OECD used in the empirical analysis. The framework itself is robust to choice of dataset, but the empirical results are not. The standard OECD ICIO tables contain around 6 million cells, while the Analytical AMNE database contains approximately 24 million cells. While data exist for many of these cells, the OECD did have to construct additional data, and doing so introduced assumptions into the data. The OECD team used tools like gravity modeling, ordinary least squares regression analysis, and linear interpolation. Howells noted that the task the OECD attempted was very difficult and the team did a fantastic job.

Howells moved to a discussion of the BEA research agenda. The BEA is developing extended supply-use tables, which will contain information on three types of firms: U.S. parents, U.S. affiliates, and an “other” category. The BEA is pursuing a top-down approach, using aggregated and readily available datasets, as well as a bottom-up approach, using extensive microdata. A National Bureau

of Economic Research (NBER) working paper characterizes the work on the top-down approach.¹

Howells then offered some final comments:

• First, the standard TiVA framework provides a breakdown of value added by labor and capital location. In the new framework, control of assets is an important dimension. Howells differentiated between control and ownership, and noted foreign minority stakes, which are larger than 10 percent in BEA data and may have sizable control of a firm, especially where the minority stake is a plurality of shares. In order to be considered an FIE, a firm must have 50 percent foreign ownership in the OECD database.
• Second, the new framework can indicate that a firm in country C manufactured an intermediate part assembled in country B and purchased in country A. The framework cannot indicate the location of the parent of the firm in country C. This means that the framework is still somewhat location locked, like the standard TiVA framework.
• Third, clarification is needed in how we think about the GVC decomposition terms outlined in the chapter. Rather than being mutually exclusive, we can consider the total value created in any GVC as coming, in different amounts, from the different components.
• Lastly, the standard TiVA framework accounts for trade in intermediate goods and services, but MNEs will also trade capital goods. This is an interesting future path for work.

Wei agreed that Howells made a useful distinction between FIEs by ownership and FIEs by control. If this is a concern, then the estimates are undercounting the role of FIEs in their empirical analysis. Wei also indicated that the conceptual framework is independent of the dataset used and encouraged its use with whatever data are available.

Finally, Wei indicated that current data do not allow for removing the location-based aspect of GVC, as discussed in the above example with three countries. Wei expressed hope for this data to become available, but he said that he and his colleagues are working on an alternative approach to answer this question.

___________________

¹ According to Howells, the NBER paper breaks down value-added content in exports for a set of industries in 2012 and shows that value-added content in exports varies widely across industries. U.S. MNEs appear to have large market share in the high-tech R&D-intensive industries, while U.S. non-MNEs have larger market share in manufacturing industries. Imported content is concentrated in the high-tech R&D-intensive industries, as is foreign MNE domestic value-added content. See Fetzer et al. (2021).

Nadim Ahmad, deputy director at the OECD Centre for Entrepreneurship, SMEs, Regions and Cities, added that the connection between the TiVA framework and the FDI flows was the most intriguing part of the presentation. This connection is vital to truly understand how GVCs work and was a large motivator for the OECD in the creation of its expert group on extended supply use tables, which it formalized in 2014. Ahmad cautioned that the presented framework is important but may go too far by quantifying any FDI activity by an MNE as GVC activity.

Wei responded that there is a distinction between simple and complex GVC activity and not all FDI-related GVC activity is treated as the same in the current framework.

This page intentionally left blank.