Vertical Specialization in Multinational Firms

1
Vertical Specialization in Multinational Firms

• Gordon H. Hanson
• University of California, San Diego and NBER
• Raymond J. Mataloni, Jr.
• U.S. Bureau of Economic Analysis
• Matthew J. Slaughter
• Tuck School of Business at Dartmouth and NBER

2
Introduction

• The expansion of multinational firms has
  contributed to the dramatic recent growth in
  world trade.
• Multinationals create trade in part by setting up
  global vertical production networks.
• In semiconductors, firms perform RD in US,
  fabricate silicon wafers in Taiwan, and assemble
  chips in China.
• In autos, firms produce high-end components in US
  and locate downstream tasks near markets in S.
  America, Asia, and Europe.
• This vertical specialization accounts for
  one-third of recent growth in world trade
  (Hummels et al., 2001). Parts and components are
  responsible for 30 of world trade in
  manufactures (Yeats, 2001).

3
Introduction

• Why do multinationals vertically fragment
  production?
• Low import tariffs, transport costs.
• In theory, low trade costs make back-and-forth
  trade feasible between parents and their foreign
  affiliates. Implies that small reductions in
  tariffs can boost trade by a large amount (Yi,
  2002).
• But previous empirical research suggests that
  tariffs raise FDI by inducing firms to replace
  exports with production abroad by foreign
  affiliates (Brainard, 1997 Carr et al., 2001).
• International differences in wages.
• In theory, firms prefer to locate labor-intensive
  tasks in low-wage countries, skill-intensive
  tasks in high-wage ones (Helpman, 1984).
• But previous empirical work (see above) suggests
  that large local markets, and not low wages, are
  what attract multinational firms.

4
Regional Shares of Employment by Foreign
Affiliates of US Multinational Firms ()
5
Table 2 Share of Imported Inputs in Total
Sales of US Foreign Affiliates ()
6
Our Paper

• We use BEA firm-level data on US multinationals
  to analyze a direct measure of vertical
  specialization
• The demand for imported intermediate inputs (for
  further processing) by foreign affiliates in
  manufacturing in 1994.
• We are able to measure the extent to which
  foreign affiliates specialize in processing
  inputs supplied by their US parents.
• We estimate the sensitivity of demand for
  imported inputs to host-country tariffs, wages,
  and taxes.
• Previous work uses aggregate data on
  multinationals and so may miss how production is
  organized in foreign affiliates.

7
Preview of Empirical Findings

• A prominent role for tariffs in vertical
  specialization.
• Trade in imported intermediate inputs is greater
  where host-country tariffs are lower.
• Sensitivity of vertical specialization to host
  wages.
• Input trade is greater where wages for
  low-skilled workers are lower and wages for
  high-skilled workers are higher.
• Impact of other host policies, characteristics.
• Input trade is greater where corporate tax rates
  are lower, free trade zones exist, and host
  countries are smaller.

8
Empirical Framework

• Consider a US multinational firm at a point in
  time that is deciding how to organize production
  in its foreign manufacturing affiliates.
• For simplicity, assume production involves two
  stages input manufacturing and input
  processing.
• The US parent chooses between fragmenting these
  two stages across borders (vertical
  specialization) or co-locating them at home and
  abroad (vertical integration).
• Our framework captures the role of trade costs,
  labor costs, and other factors in this decision.

9
Empirical Framework

• Using a cost-minimization framework, we derive a
  foreign affiliates demand for imported
  intermediate inputs (from US parent and other
  entities).
• The greater are imported intermediate inputs as a
  share of total costs the more vertically
  specialized is the affiliate.
• According to theory, input demand will be a
  function of
• Host-country import tariffs and transport costs.
• Host-country wages for low-skilled and
  high-skilled labor.
• Host-country corporate taxes.
• Other host-country policies and characteristics.

10
Estimation Issues

• We do not observe transaction prices that
  affiliates pay for inputs that they import from
  their parents.
• We control for input prices by decomposing them
  into a parent-industry fixed effect (to capture
  price in US of inputs parents supply to their
  affiliates) and host-country trade costs.
• Parents may manipulate input prices they charge
  affiliates in order to shift income between tax
  jurisdictions.
• The incentive to transfer price is influenced by
  the foreign tax credit status of the US parent
  (captured by parent-industry fixed effects), and
  the corporate tax rate in the host country.
• Other estimation issues
• Zero values for dependent variable, how to
  measure wages, whether other host-country
  characteristics matter.

11
Summary of Estimation Strategy

• To explain the variation in imported intermediate
  inputs across foreign affiliates (of same
  parent), we estimate the responsiveness of
  processing imports to wages, tariffs, tax rates,
  and other variables.
• From coefficient estimates, calculate
  elasticities of input demand w.r.t. trade costs
  and factor prices.

12
Data Description

• We use data on majority-owned foreign affiliates
  of US multinationals in manufacturing for 1994.
• Sample 4,285 affiliates that belong to 640 US
  parents.
• Variables Imported intermediate inputs for
  further processing as a share of total sales,
  capital stock, total output, average earnings of
  high-skilled labor and low-skilled labor.
• Other data
• Trade costs tariffs, NTBs, freight costs,
  distance, language.
• Corporate income taxes statutory rate,
  effective average rate.
• Wages for skilled, unskilled workers market
  size EPZs exchange rate policies host-country
  institutions.

13
Table 1 Summary Statistics
14
Main Empirical Results

• The share of imported intermediate inputs in
  total costs is higher for affiliates in countries
  with
• Lower tariffs and lower freight rates.
• Membership in NAFTA.
• Lower wages for low-skilled labor.
• Higher wages for high-skilled labor.
• Lower corporate income tax rates.
• Export processing zones and smaller domestic
  markets.

15
Interpreting the Results

• A reduction in imports tariffs from 15 to 5
  (from two ss above the mean to the mean)
• Raises demand for imported intermediate inputs by
  18.
• A reduction in average wages for low-skilled
  labor from US levels (15,540) to Mexico levels
  (4,622)
• Raises demand for imported intermediate inputs by
  41.
• A reduction in the statutory corporate tax rate
  from 55 to 26 (from two ss above mean to
  mean)
• Raises demand for imported intermediate inputs by
  10.

16
Robustness Checks

• Drop observations with high tariff values.
• Tobit estimation with industry dummy variables.
• Control for country-specific fixed effects.
• Change variable definitions.

17
Conclusion

• In this paper, we examine trade in imported
  inputs for further processing between US parent
  firms and their affiliates in foreign countries.
• These shipments of intermediate inputs provide a
  direct measure of vertical specialization within
  the global production networks of multinational
  firms.
• We find input processing by foreign affiliates to
  be driven by trade costs, factor prices, and
  other host-country policies and characteristics.

18
Conclusion

• Where to locate processing of intermediate
  inputs
• Multinationals appear to prefer countries with
  abundant low-skilled labor, low tariffs, low
  corporate taxes, good access to the US, and
  small domestic markets.
• This is not the message of previous research,
  which emphasizes desire of multinationals to
  locate behind tariff walls in order to serve
  large, local markets.
• As low-wage countries continue to lower trade
  barriers, we expect them to play a larger role in
  the vertical production networks of multinational
  firms.