Multinational Firms and Global Supply Chains

ECON 171 · Spring 2026 · Week 8

Sasha Petrov

Today’s Agenda

  1. Types of FDI: horizontal vs. vertical
  2. The concentration-proximity trade-off
  3. Vertical FDI and the cost-coordination trade-off
  4. Outsourcing vs. in-house production, and transaction costs

What We Aim to Learn

How much trade is multinational?

  • Multinationals move production across borders, not just goods. The right unit of analysis is the firm, not the country.
  • In 2024, 44% of US goods trade ($2.3 T of $5.3 T) — and 49.5% of US imports — was related-party.
  • Global FDI flows reached $1.6 T in 2025, but have lost ground to GDP and trade since 2010.

Half of what the US buys from abroad is shipped between a firm and its own affiliate.

FDI inflows as % of host-country GDP · OWID (WB WDI / IMF BoP). Headline figures: US Census, Related-Party Trade 2024; UNCTAD, WIR 2025.

Horizontal FDI: Toyota in Kentucky

Replicate the plant abroad

  • 1981: US–Japan VRA caps Japanese auto exports to the US at 1.6 M cars/year.
  • 1988: Toyota opens TMMK in Georgetown KY — its first wholly-owned US assembly plant.
  • Today: ~ 9,400 employees, 14 M+ vehicles built. Builds Camry, RAV4 hybrid, Lexus ES for the US market.
  • Same product, replicated inside the tariff wall.

Horizontal FDI — replicate production where demand is, to bypass trade costs.

Toyota Motor Manufacturing Kentucky visitor center and main entrance, Georgetown, KY (Wikimedia Commons, CC BY-SA 3.0).

Plant Georgetown, KY — Toyota’s first wholly-owned US assembly facility (opened May 1988).
Image: Wikimedia Commons (Censusdata, CC BY-SA 3.0).

Source: APEC, Toyota Motor Manufacturing Kentucky case study (2025); Toyota Global, 75 Years of Toyota: Voluntary Export Restraints (2012).

Vertical FDI: GM in Mexico

Fragment the chain across borders

  • GM owns three Mexican assembly plants — Ramos Arizpe (Equinox / Blazer EV, Cadillac Optiq), San Luis Potosí (Equinox, GMC Terrain), Silao (light-duty Silverado / Sierra).
  • In 2024, GM exported ~ 85% of Mexican output — about 712 K vehicles — back to the US market.
  • Design, R&D, and powertrain engineering remain in Detroit; stamping, paint, and final assembly fragmented across NAFTA / USMCA.
  • Same firm, different stages, different countries.

Vertical FDI — chain across borders inside the firm, to chase factor-cost gradients.

GM's North American production network — Detroit HQ to Ramos Arizpe, San Luis Potosí, and Silao plants in Mexico, with finished-vehicle exports back to the US market.

GM’s Mexican production footprint — three wholly-owned plants, 85% of output exported back to the US in 2024.

Source: Council on Foreign Relations, Tariff Turbulence (2025); AméricaEconomía, Mexico’s Automotive Surplus with the US (2024).

Outsourcing: Apple’s iPhone

… or buy it on the market

  • iPhone is designed in Cupertino; iOS, the App Store, the brand are owned by Apple.
  • Components are bought from independents: TSMC chips (Taiwan), Samsung OLED (Korea), Sony sensors (Japan).
  • Assembly by Foxconn / Pegatron / Luxshare — ~ 9 of 10 iPhones still from China; Vietnam and India ramping fast.
  • Apple owns none of these plants — every stage is an arm’s-length contract.

Outsourcing — fragment the chain without owning any of it.

iPhone supply chain — Apple Cupertino orchestrates component suppliers (TSMC, Samsung, Sony) and assembly partners (Foxconn/Pegatron in China, Vietnam, India) to deliver phones to the world market.

The iPhone supply chain in 2025 — Apple designs in Cupertino, buys components and assembly from independent firms.

Source: Business Standard, India builds China-light Apple supply chain (2025); AEI, Apple’s Supply Chain: Economic and Geopolitical Implications (2024).

Multinational Firms

Types of FDI: horizontal vs. vertical

Horizontal FDI — replicate the plant

  • Replicate production abroad to serve a foreign market locally (e.g., Toyota Kentucky).
  • Substitute for exporting: skip the trade cost \(\tau\), duplicate the plant fixed cost \(F_{\text{plant}}\). Common between similar economies (US ↔︎ EU autos, pharma).

Vertical FDI — fragment the chain

  • Fragment the value chain across countries (e.g., GM Mexico); offshore the labour-intensive stages, keep capital-intensive at home.
  • Driven by factor-cost differences; common between dissimilar economies (North → South).

Two-panel schematic. Top: horizontal FDI — Home HQ → Foreign plant → Foreign market. Bottom: vertical FDI — Home HQ + final assembly sends IP to a foreign parts plant; parts return to home for assembly; finished goods ship to home or world market.

Each arrow shows where the finished good ends up — that’s the distinguishing feature across the three FDI types.

Export-platform FDI & taking stock

Export-platform FDI — third-country sales

  • Plant abroad serves a third-country market, not the host. Examples: BMW Spartanburg → world; Toyota Mexico → US; Intel Ireland → EU.
  • A hybrid — replication (be inside the destination region) and factor-cost arbitrage motives.

Taking stock

  • The three patterns aren’t exclusive — most large MNEs run all three. Horizontal dominates between developed countries; vertical and export-platform dominate North-South flows.

Export-platform FDI schematic: Home HQ sends IP to a plant in Foreign A; the plant exports finished goods to a third-country market (Foreign B / world).

Further reading: Markusen, Multinational Firms and the Theory of International Trade (MIT Press, 2002); Antràs, Global Production (Princeton, 2016); Helpman, Melitz & Yeaple, “Export Versus FDI with Heterogeneous Firms” (AER, 2004).

Pricing under export vs. FDI

  • Foreign demand \(p = a - b\,Q\) — same in both panels.
  • Firm picks \(Q\) to max \((p - \mathrm{MC})\,Q\) \(\Rightarrow\) \(\mathrm{MR} = \mathrm{MC}\).
  • Export. \(\mathrm{MC} = c + \tau\). FDI. \(\mathrm{MC} = c\).
  • Lower \(\mathrm{MC}\) slides the optimum down \(D\): higher \(Q^{*}\), lower \(p^{*}\), larger variable profit \(\pi^{\text{var}} = (p^{*}- \mathrm{MC})\,Q^{*}\).

The gap \(\pi^{\text{var}}_{\text{FDI}} - \pi^{\text{var}}_{\text{export}}\) is what FDI buys you. It must clear the extra fixed cost \(F_{\text{plant}}\).

Concentration-Proximity Trade-off

  • Concentration motive. One plant at home, export — pay trade cost \(\tau\) on every unit.
  • Proximity motive. Plant in foreign market — skip \(\tau\) but pay extra fixed cost \(F_{\text{plant}}\).
  • Both strategies face the same demand, but at different \(\mathrm{MC}\). The trade cost \(\tau\) is the wedge.
  • FDI wins when the variable-profit gap \(\pi^{\text{var}}_{\text{FDI}} - \pi^{\text{var}}_{\text{export}}\) exceeds \(F_{\text{plant}}\).

Higher \(\tau\) widens the gap; higher \(F_{\text{plant}}\) raises the bar. Try both sliders.

Trade policy and horizontal FDI

  • Tariff \(\Delta\tau\) shifts \(\mathrm{MC}_{\text{export}}\) up; \(\mathrm{MC}_{\text{FDI}}\) is unaffected.
  • Export optimum slides up demand: smaller \(Q^{*}\), higher \(p^{*}\), smaller \(\pi^{\text{var}}_{\text{export}}\).
  • Variable-profit gap widens. If it now exceeds \(F_{\text{plant}}\), the exporter jumps the tariff and opens an MNE plant.
  • Classic case: 1981 US–Japan VRA \(\Rightarrow\) Honda, Nissan, Toyota open US transplants in the 1980s.

Vertical FDI

Intel: one chip, four countries

Intel splits production of one chip into three stages — every box is wholly owned by Intel:

  • R&D · design (skill-intensive) — Santa Clara, Hillsboro, Haifa, Bangalore.
  • Wafer fabrication (capital-intensive) — Oregon, Arizona, Ireland, Israel.
  • Assembly · test · packaging (labor-intensive) — Vietnam, Malaysia, Costa Rica.

The same firm runs the chain in four countries, picking each stage’s location separately.

Vertical FDI at full extent — chain across borders, all in-house.

Intel global production network — R&D in California/Oregon/Israel/India, wafer fabrication in Oregon/Arizona/Ireland/Israel, assembly-test-packaging in Vietnam/Malaysia/Costa Rica, finished chips to the world market. Every box is wholly owned by Intel.

Source: Intel, Manufacturing Locations (2024); Intel CSR Report 2023–24; Intel newsroom press releases (2021–2024).

The cost-coordination trade-off

  • Stage \(j\) uses inputs at intensities \(a^{L}_{j}\) (labor) and \(a^{K}_{j}\) (capital).
  • Unit cost at country \(k\): \(\;C_{k}(j) = w_{k}\, a^{L}_{j} + r_{k}\, a^{K}_{j}\).
  • Saving from offshoring: \(\;\Delta(j) = (w_{H} - w_{F})\, a^{L}_{j} + (r_{H} - r_{F})\, a^{K}_{j}\).
  • Rule: offshore stage \(j\) when \(\Delta(j) > \tau_{\text{coord}}\) (coordination friction).
  • Takeaways: with \(w_{H} > w_{F}\), \(\Delta(j)\) rises in \(a^{L}_{j}\) — labor-intensive stages go first; falling \(\tau_{\text{coord}}\) \(\Rightarrow\) more stages cross over.

For now: every offshored stage stays in-house. We relax this assumption in the next section.

Schematic of the cost-coordination trade-off. The factor-cost saving from offshoring rises with the labor-intensity of the stage. Two horizontal threshold lines mark coordination friction in 1980 (high) and 2020 (low). As the threshold drops, more stages cross above it — the shaded band between the two intersections is the newly offshored region.

Frameworks: Baldwin, The Great Convergence (Harvard, 2016); Grossman & Rossi-Hansberg, “Trading Tasks: A Simple Theory of Offshoring” (AER, 2008).

The great unbundling: evidence

  • Country level, 1970s → 2008: foreign value-added share of OECD exports roughly doubled; intermediate goods now ~56% of OECD goods trade.
  • Firm level, 2023: ~32% of US multinationals’ worldwide workforce sits in foreign affiliates; ~24% of their value added is produced abroad (BEA).
  • Driver: real shipping costs \(\downarrow\) ~80% vs. 1930; international call cost ≈ zero (chart →).
  • Since 2008 (“slowbalisation”): global foreign value-added share flat; US fell 12.1% (2008) → 7.5% (2020).

30 years of falling \(\tau_{\text{coord}}\) pushed firms toward fragmentation — but the trend has stalled.

OWID chart: the decline of transport and communication costs relative to 1930. Sea freight, passenger air travel, and international call costs all fall sharply from 1930 to ~2000, then plateau.

Sources: OECD-WTO TiVA (1995–2020); BEA, Activities of US Multinational Enterprises 2023; OWID Trade & Globalisation (OECD Economic Outlook 2007).

Outsourcing and Supply Chains

Outsourcing vs. In-house Production

Relax the in-house assumption

  • Until now each offshored stage was in-house. Now allow independent suppliers.
  • In-house → own the foreign affiliate (FDI). Arm’s-length → buy on the market (outsourcing).
  • Each stage is now here vs. abroad and inside vs. outside the firm — a 2 × 2.

Whether to integrate depends on transaction costs — see the next slide.

2x2 matrix of organisation of production. Rows: home vs. foreign. Columns: in-house vs. arm's-length (outsourced). Cells: Intel Oregon (home, in-house); Magna/Lear/BorgWarner (home, outsourced); GM Mexico (foreign, in-house, vertical FDI); Apple-Foxconn (foreign, outsourced, offshoring).

Transaction costs

Coase (Economica, 1937): firms exist because using the market is not free. Beyond the price \(p\) a buyer also pays:

  • Search & info — find suppliers, screen for quality.
  • Bargaining — negotiate price, terms, contingencies.
  • Enforcement — monitor delivery, defects, IP.
  • Adaptation — renegotiate the unforeseen.

Williamson’s three drivers (1975, 1985):

  • Asset specificity — supplier’s investment has no value outside this deal. Decisive → hold-up risk (right).
  • Uncertainty — more contingencies ⇒ more renegotiation.
  • Frequency — repeated deals need governance.

The holdup problem in four steps. Step 1: supplier sinks a relationship-specific investment worth V in the deal, zero outside. Step 2: contract signed at price p. Step 3: buyer threatens to walk unless supplier accepts p minus Delta; supplier's outside option is zero because the equipment is locked in. Step 4: supplier accepts; buyer captures the rent.

Rule. High asset specificity, uncertainty, or frequency ⇒ integrate. Standardised, thick-market inputs ⇒ outsource.

References: Coase (Economica, 1937); Klein–Crawford–Alchian (JLE, 1978); Williamson, Markets and Hierarchies (1975) and The Economic Institutions of Capitalism (1985); Grossman & Hart (JPE, 1986); Hart & Moore (JPE, 1990).

What gets outsourced?

Antràs (2003) headline result

  • Capital-intensive inputs ⇒ integrate (intra-firm).
  • Labour-intensive inputs ⇒ outsource (arm’s-length).
  • Logic: when HQ’s investment is decisive (R&D, brand), the firm wants residual control. When the supplier’s effort is decisive, it cedes ownership.
  • Country corollary (Nunn, 2007): weaker contract enforcement ⇒ more integration.

The same firm chooses different governance for different inputs — pharma intra-firm, packaging arm’s-length.

Horizontal bar chart of intra-firm share of US imports by industry. Capital- and R&D-intensive industries (pharmaceuticals 75%, transportation equipment 74%, chemicals 72%, computer & electronic 62%, industrial machinery 49%) shown in maroon, dominating the top of the chart. Labor-intensive industries (plastics & rubber 38%, textile mills 15%, apparel 7%, footwear/leather 5%) shown in gold at the bottom. Source: US Census Related-Party Trade 2009 exhibits; pattern from Antràs QJE 2003.

Further reading: Antràs, “Firms, Contracts, and Trade Structure” (QJE, 2003); Nunn, “Relationship-Specificity, Incomplete Contracts, and the Pattern of Trade” (QJE, 2007); Antràs, Global Production (Princeton, 2016).

Main take-aways

  • Multinational firms cross borders as producers, not just exporters. Distinguish horizontal FDI (replicate the plant in a foreign market) from vertical FDI (fragment the chain across borders).
  • Concentration-proximity trade-off. Exporting pays a per-unit trade cost \(\tau\); FDI skips \(\tau\) but pays a fixed plant cost \(F_{\text{plant}}\). FDI wins when the variable-profit gap \(\pi^{\text{var}}_{\text{FDI}} - \pi^{\text{var}}_{\text{export}}\) exceeds \(F_{\text{plant}}\).
  • Tariffs are an asymmetric shock. They raise \(\mathrm{MC}_{\text{export}}\) only — \(\mathrm{MC}_{\text{FDI}}\) is unchanged. A large enough \(\Delta\tau\) flips the regime to tariff-jumping FDI (1981 VRA, Honda/Nissan/Toyota in the US).
  • Vertical FDI follows cost-coordination logic. Offshore stage \(j\) when factor-cost saving \(\Delta(j)\) exceeds coordination friction \(\tau_{\text{coord}}\). Labor-intensive stages go first; falling \(\tau_{\text{coord}}\) explains the great unbundling since the 1970s.
  • Outsourcing vs. in-house is about transaction costs. Hold-up risk, asset specificity, and incomplete contracts push relationship-specific stages inside the firm.

Multinationals make a location choice (where to put each stage) and an organisational choice (in-house vs. arm’s length). Both choices respond to the primitives — factor prices, \(\tau\), \(F_{\text{plant}}\), contracts.

Skills you can now use

Solve the model

  • From \((a, b, c, \tau, F_{\text{plant}})\), compute the monopolist’s \((Q^{*}, p^{*}, \pi^{\text{var}})\) under export vs. FDI and pick the winning mode.
  • Find the threshold \(\Delta\tau\) at which an exporter becomes a tariff-jumping MNE for given \(F_{\text{plant}}\).
  • Rank stages by \(a^{L}_{j}/a^{K}_{j}\), compute \(\Delta(j)\), and predict which stages cross \(\tau_{\text{coord}}\).
  • Trace comparative statics in \(\tau\), \(F_{\text{plant}}\), \(\tau_{\text{coord}}\), \(w_{H} - w_{F}\).

Interpret and apply

  • Classify a real firm’s setup (Toyota Kentucky, GM Mexico, Intel, Apple–Foxconn) as horizontal/vertical and in-house/outsourced.
  • Predict the direction of FDI from a tariff change or transport-cost shock.
  • Diagnose when transaction costs (hold-up, specificity) justify vertical integration of a stage.
  • Suggest data measurements that can indicate trends and patterns in vertical FDI and outsourcing.

You can now move between primitives \((\tau, F_{\text{plant}}, \tau_{\text{coord}}, \text{contracts})\) and observables \((Q^{*}, p^{*}, \text{mode}, \text{location})\) — the firm-level layer that complements the country-level theories from earlier in the course.