Contracting with Enemies? Vertical FDI with Outsourcing Contracts

Abstract

An exploration of Korean MNCs' foreign affiliate-level data reveals that a significant portion of manufacturing foreign affiliates sell both to related and unrelated firms at the same time. We refer to this as hybrid vertical FDI. We rationalize the presence of hybrid vertical FDI by modifying the otherwise standard property-rights model of global sourcing with the subsidiary-level option of supplying inputs to unrelated customers in addition to related firms. Given the positive production externality from serving additional customers (that is proportional to the MNC's productivity) and the costs of getting such benefit (that are increasing in relationship-specificity of the outsourced inputs), the model generates following testable hypotheses: Both MNCs' likelihood of choosing hybrid over pure vertical FDI and their foreign affiliate firms' related-firm sales ratio over unrelated-firm go up when the productivity of foreign affiliates increases (but such tendencies weaken when contractual complexity goes up), which our subsequent empirical analysis robustly confirms.

Keywords:

• FDI
• vertical FDI
• hybrid vertical FDI

Acknowledgments

We are grateful to Pol Antràs, Jaerim Choi, Alan Spearot, and seminar participants at Korea Development Institute, Korea Institute for International Economic Policy, 2020 KER International Conference, and 2023 Korea International Economic Association Summer Conference for their helpful comments. We are also grateful for the editor and a referee for their constructive comments.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

Please note that the data used in this paper is a proprietary subsidiary-level balance sheet data from the Export-Import Bank of Korea. Although the confidential data from the Export-Import Bank of Korea is typically not available for outside researchers, we should be able to facilitate the access to the dataset for interested researchers and we will make replication codes available.

Notes

1 Conceptual frameworks to understand the motive for other types of FDI activity such as export platform FDI (Ekholm et al., 2007; Tintelnot, 2017) and export-supporting FDI (Krautheim, 2013) have been also developed. Spearot (2012) and Díez and Spearot (2014) consider a separate aspect of greenfield vs. mergers and acquisition in a heterogeneous firms framework. Bilir et al. (2019) consider the role of host-country financial conditions on MNCs' location decisions for horizontal, vertical and platform FDI. See Antràs (2015) for a comprehensive review of the recent literature on global production.

2 To our knowledge, there are three countries that record such detailed subsidiary-level information: Japan, Korea, and the US. German Micro database Direct Investment (MiDi) employed in Krautheim (2013) and Tintelnot (2017) only records total sales of foreign affiliates.

3 Few exceptions include Corcos et al. (2013) and Kohler and Smolka (2014) that test the main predictions of the vertical FDI model by exploring the French and Spanish parent firm-level sourcing decision data, respectively. See Antràs and Yeaple (2014) and references reviewed therein for more discussion.

4 This is consistent with the main finding in Alfaro and Charlton (2009) that suggests high frequency of vertical FDI at global level based on the information on industry classification of worldwide parent and subsidiary firms.

5 It is in line with Hyun and Hur (2013) and Baldwin and Okubo (2014) that note a large share of Korean and Japanese foreign affiliates cannot be simply categorized as pure horizontal or vertical FDI. Feinberg and Keane (2006) and Hanson et al. (2001) also note that US MNC parents and their Canadian and Mexican affiliates show a similar pattern.

6 As a concrete example, one may consider the global sourcing structure of OLED (Organic Light-Emitting Diode) panels. LG Display, a Korean multinational company, has a production site in Guangzhou, China, that supplies OLED TV panels to many multinational corporations, such as Bang & Olufsen, Sony, Vizio, etc., in addition to its mother company. It is also well-known that Samsung provides specialized inputs to Apple despite their competition in the smartphone market.

7 As shown in Section 3.2, the sales ratio of unrelated firms to related firms decreases in the headquarter firm's productivity in our model. The corresponding sales ratio of the hybrid FDI of Grossman et al. (2006) would stay constant in response to an increase in the headquarter firm's productivity. We provide a brief discussion of their analysis of hybrid FDI in Section 3.3, explaining both the difference as well as a similarity in their prediction and ours.

8 If an MNC allows its horizontal FDI type subsidiary to sign outsourcing contracts with other MNC firms, such hybrid horizontal FDI will not generate sales to related parties on top of its existing sales to unrelated parties.

9 Since the dataset allows a short panel structure for only a limited set of firms, we focus on cross-sectional analysis in this study. While we obtained the permission of KEXIM to use their confidential foreign affiliate-level FDI data only over this period for our research, we had a chance to explore more recent (2007–2018) FDI data of KEXIM in a government-funded research project, of which the public revelation is strictly prohibited, confirming that the empirical findings of this paper largely hold for this more recent period.

10 The sample coverage has increased over time starting with about 100 parents and their 200 foreign affiliates in 2000 (Debaere et al., 2013).

11 A few studies that used the Korean MNCs' foreign affiliate-level data include Cho (2018), Chung (2014), Debaere et al. (2010), Debaere et al. (2013), Hyun and Hur (2013), etc.

12 The number of observations in Table 2 is smaller than that in Table 1 due to a greater number of missing information on input purchases.

13 Comparing with manufacturing subsidiaries created by vertical, horizontal, or export platform FDI, subsidiaries in other sectors such as the wholesale and retail are likely to have a higher sales ratio to unrelated parties reflecting the export-supporting FDI motive–keeping production at home while shifting distribution tasks to foreign affiliates in the local market, thereby saving the variable distribution cost in return for an additional fixed cost to set up subsidiaries abroad (Krautheim, 2013). In fact, the median firm's sales share to unrelated parties in the whole sale and retail sector in this data is 100%.

14 Although we may further distinguish horizontal FDI firms from export-platform FDI firms using the detailed geographical information of unrelated arm's length buyers, we simply grouped them together here because there are not many export-platform FDI firms as indicated in Table 1 by median firms' low share of sales to unaffiliated parties in third countries.

15 Our dataset does not allow us to identify whether foreign subsidiaries are multi-product firms, thereby selling different products to related firms from those sold to unrelated customers. That said, our model introduced in the subsequent section can be viewed as a general case in that a product supplied to related firms is not necessarily the same as the one supplied to unrelated firms.

16 As discussed in the introduction and in Section 3.2, the hybrid FDI case of Grossman et al. (2006) does not yield the prediction that the sales ratio of unrelated firms to related firms decreases in the headquarter firm's productivity, for which our empirical analysis find a supportive evidence.

17 The sequence of events in this global sourcing problem is the typical one: (1) F determines the ownership structure (v or o), signing an initial contract (that specifies a sidepayment, s, from F to I) and incurring fixed costs ($f_v$ or $f_o$); (2) F and I produce their inputs of the amounts of h and m, respectively; (3) F and I renegotiate over the share of revenue from the final product; (4) F produces and sells the final product by combining the inputs produced in the second stage.

18 As we do not analyze the general equilibrium consequence of adding hybrid vertical FDI to the list of possible organizational forms from which headquarter firms may choose, B, $w_N$, and $w_S$ are treated as fixed variables in the following analysis.

19 Because $I_v$ obtains no profit from its sales to $F_o$, it seems easier to assume that outsourcing sales are sold at the marginal cost and generate no profit, a typical assumption for the production of standardized inputs. We may justify our explicit modeling of production relationship between $I_v$ and $F_o$ as follows. The degree of relation-specificity in manufacturing outsourced inputs would affect the degree of hardship in transforming the production knowledge of outsourced inputs into the one applicable for input production for the headquarter. We capture such a hardship by a parameter κ, by making the cost of benefiting from producing outsourced inputs increase in the number of unrelated firms that $I_v$ serves, powered by κ, as shown in (3(3) $p\left(q_v({\phi }_v;\cdot )\right)q_v({\phi }_v;\cdot )-w_N{h}_v({\phi }_v;\cdot )-\tau w_S{m}_v({\phi }_v;\cdot )-w_N{f}_v-w_N({\left(x_o\right)}^{\kappa }-1)f_{vo}$(3) ) below. This parameter plays a crucial role in our model's theoretical predictions, for which our empirical analysis provides supporting evidences.

20 After $F_v$ allows $I_v$ to sign an outsourcing contract with $F_o$ at period 0, we assume the following sequence of events for an outsourcing contract: (1) $F_o$ signs an outsourcing contract (that determines the side payment from $F_o$ to $I_v$) with $I_v$, incurring fixed costs, $f_o$; (2) $F_o$ and $I_v$ produce their inputs of the amounts of $h_o$ and $m_v$, respectively; (3) $F_o$ and $I_v$ renegotiate over the share of revenue from the final product; (4) $F_o$ produces and sells the final product by combining the inputs produced in the second stage. For the in-sourcing contract between $F_v$ and $I_v$, with the number of outsourcing contracts being $x_o-1$, we assume that the following sequence of events occur concurrently: (1) $F_v$ signs an in-sourcing contract (that determines the sidepayment from $F_v$ to $I_v$) with $I_v$, incurring the fixed costs for tranferring the production knowledge from producing $m_o$ into producing $m_v$, that increases in $x_o-1$ as specified in (3(3) $p\left(q_v({\phi }_v;\cdot )\right)q_v({\phi }_v;\cdot )-w_N{h}_v({\phi }_v;\cdot )-\tau w_S{m}_v({\phi }_v;\cdot )-w_N{f}_v-w_N({\left(x_o\right)}^{\kappa }-1)f_{vo}$(3) ), as well as the typical fixed costs, $f_v$; (2) $F_v$ and $I_v$ produce their inputs of the amounts of $h_v$ and $m_v$, respectively; (3) $F_v$ and $I_v$ renegotiate over the share of revenue from the final product; (4) $F_v$ produces and sells the final product by combining the inputs produced in the second stage.

21 While we do not specify how $I_v$ produces $m_v$ and $m_o$ (i.e. does not define the production function of $m_v$ and $m_o$), following the standard property-rights model, $I_v$ may produce $m_v$ and $m_o$ using any combination of in-sourced or outsourced production inputs. We do not analyze the sourcing strategy of a foreign subsidiary firm for producing $m_v$ and $m_o$, thus we implicitly assume that it does not affect any of our results.

22 $\kappa >ϵ(1-\eta )(\sigma -1)$ is a stability condition that insures a bounded solution for the extensive-margin-decision problem in (6(6) $\begin{array}{rl}& \underset{x_o\ge 1}{\mathrm{Max}}[{\mathrm{\Lambda }}_v{\phi }_v^{(\sigma -1)}{x}_o^{ϵ(\sigma -1)(1-\eta )}-w_N({\left(x_o\right)}^{\kappa }-1)f_{vo}-w_N{f}_v],\mathrm{with}\\ & {\mathrm{\Lambda }}_v=\frac{B{\beta }_v^{\eta (\sigma -1)}{(1-{\beta }_v)}^{(1-\eta )(\sigma -1)}[\sigma -(\sigma -1){\beta }_v\eta -(\sigma -1)(1-{\beta }_v)(1-\eta )]}{{\left(w_N\right)}^{\eta (\sigma -1)}{\left(\tau w_S\right)}^{(1-\eta )(\sigma -1)}}\end{array}$(6) ). While $ϵ(1-\eta )(\sigma -1){\mathrm{\Lambda }}_v{\phi }_v^{(\sigma -1)}<\kappa w_N{f}_{vo}$ will hold for a headquarter firm with a sufficiently small productivity level, there is no loss of generality in assuming the other inequality (i.e. $ϵ(1-\eta )(\sigma -1){\mathrm{\Lambda }}_v{\phi }_v^{(\sigma -1)}\ge \kappa w_N{f}_{vo}$) as it would still allow the headquarter firm to choose pure vertical FDI over hybrid vertical FDI.

23 Figure 6 of Grossman et al. (2006) indicates this hybrid FDI case by an area denoted by ‘S, HRS,’ implying that the intermediate input production takes place in S (southern country) and the assembling operation takes place in all three consumption locations, H (home), R (another northern country), and S.

24 As discussed above, for this hybrid FDI to arise, the cross-country transportation cost of final goods needs to be high with the shipping cost of intermediate inputs and the fixed cost of having an intermediate-input-producing foreign subsidiary being sufficiently low.

25 This change in the MNC's extensive margin decision on hybrid FDI corresponds to moving from the area of ‘S, HR’ into the area of ‘S, HRS’ in Figure 6 of Grossman et al. (2006), which occurs with the MNC's higher productivity.

26 We acknowledge that subsidiary-level total factor productivity estimation was not feasible due to a limited panel structure and insufficient information from the dataset.

Additional information

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2019S1A 5A2A030 53880). Park also acknowledges the financial support by the Research Grant of the Center for Distributive Justice of Institute of Economic Research at Seoul National University.