7

Economies of Scope and Relational Contracts: Exploring Global Value Chains in the Automotive Industry

Paper Authors: Susan Helper (Case Western Reserve University) and Abdul Munasib (Bureau of Economic Analysis)

Presenter: Susan Helper (Case Western Reserve University)
Moderator: Wolfgang Keller (University of Colorado Boulder and National Bureau of Economic Research)

The workshop paper by Susan Helper and Abdul Munasib explores the development of supply networks by considering multiple transactions at a time. The authors report that transaction-based measures are significant and organization-based measures generally have much greater explanatory power.

Susan Helper, the Frank Tracy Carlton professor of economics at the Weatherhead School of Management at Case Western Reserve University, introduced her presentation with a story about the Kojima Press. In the late 1930s, Hamakichi Kojima asked Toyota for work, eventually obtaining a contract for procurement of sand buckets for use in fire protection. Toyota trained Kojima and, after some time, ordered additional automotive parts for a truck radiator. Now, Kojima supplies a wide variety of metal and plastic automotive parts to Toyota through a global network of affiliates. This story offers several puzzles, particularly from the point of view of transaction cost analysis.

Helper explained that Toyota often offers suppliers a “permanent deal,” giving long-term contracts and technical assistance to suppliers of both parts and commodities. Suppliers engaged in these agreements are rewarded with greater volume and contracts for new products. Toyota’s manufacturing strategy is counterintuitive to the work of Oliver Williamson, a Nobel laureate, who theorized that decisions regarding whether transactions should take place within a firm or across firms are determined by the need to minimize the costs of transactions. Williamson’s transactions cost theory lends itself to asking the following questions:

• First, why sign contracts with firms that lack the required technical capability and help them to develop that capability?
• Second, markets for commodities have low transaction cost hazards and there is little product differentiation, so why offer long-term contracts when one could buy from the cheapest producer?

Helper said that the workshop paper proposes transaction spillovers as an answer. Many economic theories analyze a single transaction at a time, but products generally require multiple intermediate inputs. Spillovers exist across these relationships, increasing the clarity and credibility of contracts with long-term suppliers. A network of tenacious problem-solving suppliers can quickly identify root causes of quality issues and implement solutions. To create tenacious firms, Toyota makes an investment in each firm and benefits by learning from each supplying firm. The supplying firms also learn from each other, as Toyota offers long-term contracts to just a few firms. Due to these network spillovers, Toyota can use a firm-wide strategy for governance instead of a product-level strategy. Clarity in one relationship aids a second relationship even in the case of different parts, which creates an economy of scope for the customer.

Helper stated that in transaction cost analysis, the focus is on individual inputs rather than firm-level relationships. This type of analysis predicts that firms will use spot markets for commodities and enter long-term relationships for high research and development (R&D) aspects of the value chain (Antràs and Chor, 2013; Hart and Moore, 1990; Williamson, 1979). U.S. automotive manufactures appear to follow these predictions; for example, Chrysler sources stampings from Die-Matic, but this relationship doesn’t have a long-term guarantee, since stampings are a commodity product with many suppliers. Chrysler sources an innovative crankshaft from Mahle as a sole source provider. Mahle enjoys a long-term relationship and collaboration with Chrysler as its sole producer of an R&D-intensive product. Spillovers do exist in this example, namely as an incentive to reduce switching costs.

CONTRIBUTION

As contributions of the workshop paper, Helper pointed to new insights into the development of supply networks by considering multiple transactions at a time. In this framework, lead firms and suppliers experience economies of scope by developing a clear and credible network. Because of network spillovers, firms and nations have unique and stable strategies for global value chain (GVC) governance, including the number of suppliers per part, degree of vertical integration, and length of relationships with suppliers. The authors applied their framework to U.S. Customs microdata, organized as a panel, on imports from automotive manufactures from 1997 to 2015. They found that transaction-based measures are significant, but organization-based measures generally have much greater explanatory power. This is important as it suggests that response to shock differs by organizational strategy, not just by product; it also suggests that it is

necessary to understand not only the patterns of products in trade, but also how those are products are integrating into value chains by lead firms.

CONCEPTUAL FRAMEWORK

Helper pointed out several reasons a firm might only enter purchase agreements in a nonspot/collaborative fashion:

• First, this organizational structure facilitates relational contracts that strengthen the supply network. By creating tight-knit supply networks, firms build clarity and credibility. This allows the lead firm to dictate the specifications of parts more easily, leading to a higher-quality final good. In addition, the supply firm can better withstand exogenous market shocks.
• Second, spillovers exist across functions in the buyer’s organization.
• Third, complementarities in adoption of practices and the development of capabilities exist within the network of supplying firms (e.g., just-in-time inventory; Dyer, 1996) and learning to write good contracts (Argyres, 2013).
• Fourth, some buyers may see two products as similar when others in the industry do not; for example, Toyota wants all suppliers to both explore and exploit (Aoki and Wilhelm, 2017).

Helper argued that well-connected supply network spillovers allow for purchasing strategies to interact with other functions in a cohesive manner. Collaborative supply networks incentivize firms to learn from one another, increasing the efficiency of all firms in the network. Supply networks organized at arm’s length allow firms to interact only through purchasing, which may reduce average quality across many competing suppliers. Another important dimension is what entity designs the product. Lead firms will have to design predominantly in-house with arm’s-length transactions.

Lastly, as Helper described, the career path for purchasing agents differs in the two frameworks. In a collaborative network with strong firm-to-firm connections, purchasing agents can move between a variety of functions, while in an arm’s-length agreement, agents should be rotated between commodities to avoid capture. A benefit of the arm’s-length network is that agents don’t need a deep technical understanding of the product, as they are more reliant on market forces.

To test the propositions presented in the conceptual framework, Helper and Munasib first ran regressions to understand the relationship between product attributes and the style of supply chain governance. Measures for understanding this relationship include product fixed effects, a measure of R&D intensity for a product category, a measure of product differentiation, and a measure of the extent to which an industry is upstream.

Next, the authors tested the theory for firms or nations with organization-wide relational contracts (a collaborative framework). They found that U.S. automotive manufacturers are more likely to engage in spot purchasing relative to Japanese firms. This result is robust to removing variation due to product attributes. The authors also found that Japanese automotive manufacturers are more loyal to their supply network than U.S. firms, as discussed below.

DATA

Helper explained that she and Munasib used microdata from the Longitudinal Firm Trade Transactions Database (LFTTD), which contains transaction-level data by firm, collected by U.S. Customs and Border Protection. All automotive manufacturers assembling cars and trucks in the United States are included in the dataset, using 10-digit harmonized tariff schedule (HTS) categories and providing detailed information about each part (e.g., vulcanized gaskets; washers and other seals; mountings, fittings, and similar articles of base metal). Lastly, the dataset identifies related-party transactions, defined as transactions between a lead firm and supplier in which the lead firm owns at least a 5 percent stake in the supplying firm. This measure differs from the prior literature that typically only accounted for firms that were 100-percent owned by the lead firm.

The sample contains the annual supplier transactions for all automotive manufacturers headquartered in the United States or Japan from 1997 to 2015, and includes 44 countries and 600 products that have been purchased for at least 5 years by a firm headquartered in either the United States or Japan, and 2 years by firms in both groups. If a supplier consists of multiple affiliate firms in one country, then the authors aggregated these affiliates into a single firm. Lastly, the authors excluded purchases from the automotive manufacturer’s home country because they do not have data on purchases from U.S. suppliers. For consistency, the authors dropped the same data from Japan.

EMPIRICAL STRATEGY AND MEASURES

Helper described analysis conducted on two broad categories of firms: representative U.S. firms and Japanese-owned firms. Aggregation used original equipment manufacturer–level imports as weights, and the sample was representative of larger market dynamics. There are four outcome variables of interest: number of suppliers per component per year, Herfindahl–Hirschman Index (HHI) of suppliers for a component per year (a horizontal concentration measure), the share of related-party imports of a component per year (a vertical integration measure), and the average number of years supplied by a supplier of a component. At least 25 percent of sales by U.S.-headquartered automotive manufacturers are sourced internationally, while for Japanese firms this number is over 40 percent. In their paper, the authors argue that use of nonspot, or

collaborative, purchase agreements from suppliers in the United States is similar to that of foreign suppliers (see Helper et al., 2012).

RESULTS

Helper then introduced the paper’s results by noting that there is some support for some predictions of transaction cost theory. As R&D intensity increases to three standard deviations from the mean, the number of U.S. suppliers falls from 8.96 to 7.55, while the number of Japanese suppliers decreases from 3.32 to 3.09. These results are significant, but organization-based theories have more explanatory power. For all levels of R&D intensity, U.S. automotive manufactures have at least seven times more suppliers than Japanese manufacturers. Japanese firms are less sensitive to R&D intensity when choosing the number of suppliers, and even at the highest levels of R&D intensity, U.S. firms use more suppliers than Japanese firms use for commodity parts. Japanese firms are also more loyal, as they are less likely to drop a supplier if an exchange-rate shock occurs.

As Helper described, any of the results discussed so far looked at the country of ownership and therefore a national culture, but the evidence also points to variation on the firm level, indicating that corporate culture is a determining factor. This variation is identified in the number of suppliers used and suggests that U.S. firms have a high variance relative to the mean.

An alternative story to explain the facts and patterns outlined in the work is one of product complexity. Novak and Stern (2008) show that luxury automotive manufacturers with complex parts are more likely to integrate vertically. Novak and Stern find that Japanese firms that produce luxury vehicles engage less in spot purchasing, relative to other luxury automobile manufacturers. This would imply that final product attributes, and not organizational strategy, determine whether a firm chooses to outsource the supply of components or vertically integrate its supply chain. Helper argued that this is unlikely to explain differences between U.S. and Japanese firms, because U.S. firms have more luxury varieties than their Japanese counterparts in the automotive sector.

Lastly, Helper returned to her opening story, the Kojima–Toyota versus Chrysler–Die-Matic comparison. The authors propose that network spillovers due to clear and credible relationships are the reason lead firms give business to a supplier that lacks technical capability. In the United States, a higher discount rate, more concerns about hold-up, and more faith in markets leads to a lower estimated value of a strong network of suppliers. U.S. firms only establish the relational contracts with high-R&D suppliers and maintain an option to exit from these relationships.

SUMMARY

Helper concluded by summarizing the contributions of her work with Munasib, stating that it has contributed to the literature on relational contracts and

the importance of network spillovers across nonaffiliated firms in the network. Using U.S. Customs microdata from 1997 to 2015, their results suggest that transaction theory–based measures are significant and have some explanatory power, but organization-based measures have much greater explanatory power. Lead firms and suppliers experience economies of scope when they develop clear and credible relational contracts. This implies that firms and countries have clear and stable strategies for GVC governance.

DISCUSSION

Discussant: Davin Chor (Dartmouth College)

Davin Chor, associate professor of business administration and chair of globalization in the Tuck School of Business at Dartmouth College, stated that the 2020 World Development report highlighted two channels through which trade in GVCs differs from the traditional view of trade: hyperspecialization and durable firm-to-firm relationships. While there is an extensive prior literature on the hyperspecialization of GVCs, less empirical work has been done on firm-to-firm relationships. There is clearly a relational dimension to sourcing decisions for some inputs in the GVC. The data requirements to study the firm-to-firm relationships are very high and can make studying this problem difficult.

The authors overcame this data challenge by using the U.S. Customs LFTTD data, which offer detailed information on the product level. They exploited a feature of the dataset, manufacturer identification (MID), to pin down the identity of each foreign supplier. Caveats to the approach include the focus on U.S.- and Japanese-owned firms only and the exclusion of domestic suppliers.

Chor described the key finding as the large contrast in the input sourcing patterns in U.S.- and Japanese-owned firms. Controlling for product-level variation, Japanese firms have a smaller number of suppliers, higher supplier HHI, higher related-party trade share, and longer supplier longevity. Japanese ownership (organizational strategy) explains more variation in sourcing patterns than alternative product-level characteristics associated with the transaction cost theory.

Chor stated that the work of Helper and Munasib fits in the broader literature, as defined in Bernard and Moxnes (2018) and Antràs and Chor (2021), in four areas:

• First, the literature on firm-to-firm links, which itself has two approaches: the cross-border trade transactions approach (Benguria, 2021; Bernard et al., 2018a; Blum et al., 2010; Eaton et al., 2016b; and Sugita et al., 2014), and the domestic value-added tax data approach (Adão et al., 2020; Alfaro-Urena et al., 2021; Bernard et al., 2019; Dhyne et al., 2021; Demir et al., 2021; Huneeus, 2018).
• Second, the literature on the duration of firm-to-firm links. Work exploiting the MID in U.S. data comes from Kamal and Monarch

• (2018), Monarch and Schmidt-Eisenlohr (2017), and Monarch (2021). Using French data, Martin and colleagues (2020) studied this relationship as well. The broad takeaway from past literature is that newly formed relationships are broken quickly; but, conditional on survival, a durable network emerges.
• Third, the literature that discusses relational contracting in specific markets, often in the setting of developing countries. This literature includes the works of Macchiavello and Miquel-Florensa (2018), Macchiavello and Morjaria (2015, 2021), Cajal-Grossi and colleagues (2020), and Brugues (2020).
• Fourth, the literature that studies teardown analysis; for example, Dedrick et al. (2010) and the iPod. Relating to the conference, the work of Surana and colleagues (see Chapter 6) falls in this fourth category.

Chor described this work as an exciting empirical setting, laying in the middle ground between broad cross-sectional analysis and specific industry-level analysis. The automotive industry is a natural industry to study, as it has many components and is deeply immersed in GVCs. The results on the number of suppliers per HTS-10 products is an issue of growing importance given the renewed emphasis on supply chain resilience in the age of COVID-19 and trade frictions. Chor offered insights along three dimensions: sharpening of the empirical work, economies of scope, and alternative interpretations of the Japanese-ownership effect.

The main finding—the number of suppliers—is striking, but Chor pointed out that further clarification on how some results may be driven by the MID construction would enhance the work. As an example, Toyota may have multiple affiliate firms in Mexico listed under the same MID. The authors aggregated all of these firms into a single supplier, which may have created an endogenous result. This would be a larger concern for Japanese-owned automotive manufacturers that source from related parties more often. If the results are robust to this comment, they would be very strong. The unconditioned results on supplier longevity suggest that U.S. firms have longer relationships than their Japanese counterparts; these results flip only when the mean is conditioned on a set of controls.

Chor suggested that the addition of summary statistics on the number of products exported per foreign supplier to U.S. or Japanese firms would be a helpful addition for understanding the role of economies of scope. This is especially true if this measure turns out to be larger for Japanese firms, consistent with anecdotes from Kojima Industries. Further, looking within sourcing networks, are there temporal dimensions to the number of products per supplier, the total value purchased from each supplier, and the quality of each product? A structural model of transactions cost and a fixed cost per supplier, rather than product, may offer some insight into the mechanisms at work. Are there

diseconomies of scope if each supplier is responsible for more inputs (Fally and Hillberry, 2018; Kikuchi et al., 2018)?

Chor concluded that the work is a great opportunity to understand network sourcing patterns with a lot of potential for future work, especially in studying U.S. and Japanese GVC resiliency in the wake of COVID-19 supply shocks. Does greater loyalty or a greater number of suppliers help weather these supply disruptions?

During the presentation, one audience member asked about the difference between a U.S. firm, which sees the United States as the home market, and a Japanese firm, which sees the United States as a foreign market. Abdul Munasib, economist at the Bureau of Economic Analysis in the U.S. Department of Commerce and a coauthor of the presented work, responded that there are data privacy concerns when looking at this detailed data. This places artificial constraints on adding outside firm-level data.