reading discussion
xiang
annurev-economics-Yeaple.pdf
The Multinational Firm Stephen Ross Yeaple Department of Economics, Pennsylvania State University, University Park, Pennsylvania 16802, and the National Bureau of Economic Research, Cambridge, Massachusetts 02138; email: [email protected]
Annu. Rev. Econ. 2013. 5:193–217
First published online as a Review in Advance on March 20, 2013
The Annual Review of Economics is online at economics.annualreviews.org
This article’s doi: 10.1146/annurev-economics-081612-071350
Copyright © 2013 by Annual Reviews. All rights reserved
JEL codes: F23, F12, L22, L24, L25
Keywords
foreign direct investment, horizontal integration, vertical integration, internalization, offshoring
Abstract
This article documents the recent advances in the international trade literature toward understanding the role of multinational firms in the conduct of international commerce. Over the past 10 years, we have developed a better understanding of the incentives firms face in their choice of production location, and we know more about the incen- tives that induce firms to vertically integrate. Furthermore, the theory literature has progressed from two-country models that cannot cap- ture the richness of multinational firms’ activities to multicountry models that do. The empirics have advanced as well but at a slower pace. Progress has been slowed by the lack of comprehensive data and the difficulties of distinguishing between the various mechanisms proposed by theory.
193
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
1. INTRODUCTION
International trade theory has increasingly focused on the firm as an important unit of analysis. This trend is a response to the empirical observation that international activity is concentrated in a small number of very large firms and to the new theory that has emphasized increasing returns to scale, imperfect competition, and contracting frictions between agents.
This article reviews the state of the economics literature on the multinational firm. Multi- national firms own production facilities in multiple countries. They obtain these facilities by en- gaging in foreign direct investment (FDI), in which the investments involve either acquiring a substantial controlling interest in an existing foreign firm (cross-border acquisitions) or estab- lishing an entirely new facility in a foreign country (greenfield investment). Although the corporate structure of a multinational firm can be complicated, it is useful to define two types of entities within a multinational firm that are associated with ownership structure: the parent and the affiliate. Parents are entities that are located in one country that own affiliates. Affiliates are establishments that are located in other countries.
The study of the multinational firm is centered on several broad and interrelated questions. First, what induces firms to open production facilities in some countries but not in others? This is a question about the economic activities the firm undertakes and the attractiveness of certain locations for performing these activities. Second, why is it that so few firms become multina- tional, and how do firms that become multinational differ from those that do not? Third, when firms choose to operate a foreign affiliate, how do they obtain this facility? Do they open a new establishment, buy an existing facility, or open a joint venture with a local firm? Finally, why do firms own foreign facilities rather than simply contract with local producers or distributors? This is a question about the boundaries of the firm in an international context.
The ownership, location, and internalization framework of Dunning (1981) is often used to organize ideas regarding the answers to these questions. There must be some firm-specific advantage (ownership), such as technology, that explains why a firm would be able to compete in an unfamiliar environment, a location advantage associated with various countries to mo- tivate a desire for international production, and an internalization advantage that explains why markets are not a good substitute for hierarchical control within the firm. Although this classification scheme is useful, a moment’s reflection suggests it might be excessively tidy. A firm characteristic might provide an advantage in one location and a disadvantage in another. In some locations, a firm may be able to use arm’s length markets effectively, whereas in others, it is better to vertically integrate. Nevertheless, this framework will prove useful in our discussion of the literature.
As we are focused on the firm and how it organizes its global production, it is natural to consider first what types of ownership advantages a firm might have. Much of the literature focuses on the role of intangible assets that have been either created internally or acquired ex- ternally. These assets, which include proprietary technology and reputation, have the character- istics of a public good in the sense that they exhibit a degree of nonrivalry within the firm. The development of these assets often has a fixed or sunk cost nature, and the use of these factors in many different locations simultaneously allows economies of scale to be exploited. Replicating and horizontal investment are names given to the phenomenon when the same intangible asset is used to support the same production activity in multiple locations.
Another strand of this literature focuses on situations in which the production process for a particular good is amenable to fragmentation into activities that can be geographically sepa- rated. When different countries might then have a comparative advantage in different activities, the geographic dispersion of production is efficient. We refer to the relocation of production
194 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
activities overseas as offshoring. When various activities are done within the firm, we refer to this as vertical integration. When activities are done for the firm by arm’s length suppliers, we refer to this as outsourcing.
Both strands of this literature emphasize that multinational firms arise when contracting frictions make the integration of activities within the firm superior to arm’s length trans- actions on markets. Coase (1937) made the observation that firms exist when the hierar- chical control of factors of production is more efficient than the use of a market mechanism. Extending that logic to an international context, multinational firms exist when the common ownership of productive facilities across international borders is more efficient than market transactions are. For instance, a firm might prefer to manage multiple plants rather than license the use of its intangible assets to unrelated parties when it is difficult to prevent an unrelated party from abusing the asset. Outsourcing may be difficult for many activities because of the nature of their production processes, such as variation in the quality of output or the degree of relationship-specific investments that must be made by individual agents for production to take place.
The review is limited in focus. There is no effort to be encyclopedic given space constraints and the numerous useful surveys of the literature. More comprehensive treatments include the monographs of Caves (2007), Markusen (2002), and Navaratti & Venables (2004) (see also Markusen 1995). Although I cannot avoid covering some of the same ground as these earlier surveys, I focus primarily on the theoretical and empirical work of the past 10 years.
The remainder of this review is organized into five sections. Section 2 briefly describes the data available to analyze the global operations ofmultinational firms and provides a list of stylized facts about the multinational firm. Section 3 focuses on the integration of multinational firms into standard trade models and the relevant empirical literature that both informs the design of these models and tests these models’ predictions. The papers discussed in Section 3 have the common feature that they take the need for internalization as given. Section 4 discusses the literature that explores why some multinationals enter foreign markets via cross-border acquisitions, whereas others enter through greenfield investment. Section 5 addresses the literature that is primarily focused on the issue of internalization. Section 6 concludes.
2. FACTS ABOUT MULTINATIONAL FIRMS
This section serves two purposes. First, it describes some of the available data that can be used to infer features of multinational firms’ behavior. Second, it discusses several stylized facts con- cerning the structure of multinational firms’ operations.
2.1. Data Sources
Measuring the global operations of multinational firms is difficult. Most data are collected by government agencies with narrow mandates and limited resources. Publicly available data are typically collected at the country level, whereas the multinational firm spans multiple countries. Here I describe three classes of data that can be accessed with various degrees of difficulty: balance-of-payments data, government-collected operations data, and customs data on intrafirm trade.1
1There are several private data sets on the global operations of multinationals. Examples are the Amadeus database, compiled by Bureau van Dijk, and Worldbase, compiled by Dun and Bradstreet. Thompson Financial also compiles data on cross-border mergers and acquisition activities.
195www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
All countries collect data on their balance of payments, and one component of these data is FDI, which occurs when a firm from one country obtains a controlling ownership stake (usually 10%) in an enterprise in another country or when a financial flow occurs between parties that reside in different countries but are related by ownership.2 These financial flows include equity capital, reinvested earnings, and capital associated with intercompany debt transactions. As the balance-of-payments data measure financial flows between countries, they may say little about the extent of real economic activity in either the sending or receiving country. Data on FDI flows are reported by the International Monetary Fund, the United Nations Center for Trade and Development (UNCTAD), and the Organization for Economic Cooperation and Development (OECD).
Most governments conduct a census of the firms that operate in their country. Many surveys require a firm to identify whether it is foreign owned, allowing researchers to assess the differences between domestic firms and local affiliates of foreign firms. Furthermore, some countries require their resident firms to report whether they own foreign affiliates so that it is possible to compare parent firms’ characteristics to the population of firms. These data cannot be used to assess the structure of multinational firms’ global operations.
Publicly collected data on the global operations of firms are rare. One of the few sources is the Bureau of Economic Analysis (BEA) of the United States. The BEA conducts extensive and mandatory surveys of US multinational firms, collecting information on the activities of the US resident parents and their foreign affiliates.3 A few other countries, such as Germany and Japan, also conduct surveys of multinational firms’ global operations. These data tend to be proprietary and require special permission to obtain access.
Coverage of the global structure of multinational firms’ operations, although uneven, is far better than the alternatives to multinational production, such as licensing contracts and contracts with arm’s length suppliers. Some information can be gleaned by customs data that distinguish between related and unrelated party trade, for instance, US customs records for each transaction, whether the transaction is between related parties (defined by ownership between exporter and importer) or unrelated parties. Transaction-level data are confidential, but industry-level data are available online.4 In the remainder of this section, I sketch a portrait of the operations of mul- tinational firms, bearing in mind that this portrait is unavoidably incomplete.
2.2. Stylized Facts
Here our discussion of robust patterns in the data on multinational firms is organized into four areas. First, we discuss the geographic structure of global production within multinational firms, focusing on the level of development of the countries in which these firms are most active and the physical distance between their facilities. Second, we discuss the industrial characteristics asso- ciated with high levels of multinational activity across industries and the characteristics of mul- tinational firms relative to nonmultinational firms within industries. Third, we discuss aspects of vertical specialization across countries within multinational firms. Finally, we discuss how parent firms become associated with their foreign affiliates, concentrating on the trade-off between buying existing concerns and opening entirely new establishments.
2Where possible, I limit attention to majority-owned affiliates (for which foreign ownership exceeds 50%) because, for these firms, actual foreign control is more likely. 3The BEA also collects data on the US affiliates of foreign companies, but these data do not provide information about the characteristics of the non-US activities of these firms. 4US related party trade data can be downloaded from http://sasweb.ssd.census.gov/relatedparty/.
196 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Multinational firms account for a large share of global production. According to the esti- mates of UNCTAD in itsWorld Investment Report 2011, multinational firms account for 25% of global GDP and one-third of international trade. The importance of multinational firms in economic activity varies considerably across country pairs within industries and within country pairs across industries. For developed countries, sales by foreign affiliates to foreign customers tend to be a far more important means of serving foreign markets than are exports from the home country. For instance, according to BEA data, in 2009 the sales of the foreign affiliates of US firms were nearly $5 trillion, while US exports were only slightly greater than $1 trillion. Our discussion of the variation in multinational activity across countries begins with stark patterns associated with the level of economic development.
The most comprehensive data on the relative importance of multinational firms in economic activity across countries can be gleaned from FDI data. It is well documented that most multi- national activity is confined to developed countries. According to UNCTAD’s World Investment Report 2011, developed countries accounted for 82% of the outward FDI stock and 66% of the inward stock. The tendency of multinational firm activity to be concentrated in developed countries is confirmed by the BEA data. In 2009, the share of the value added by the affiliates of US multinationals in developed countries was 73% (Barefoot & Mataloni 2011).5 In 2009, the value added of the affiliates of foreign firms operating in the United States accounted for $550 billion, with seven countries—Canada, France, Germany, the Netherlands, Switzerland, the United Kingdom, and Japan—accounting for 75% of foreign affiliate value added in the United States (Anderson 2011).6
We summarize this information as the following fact.
Fact 1: The parents and affiliates of multinational firms are primarily located in developed countries. Affiliates are better represented in developing countries than are parents.
We now turn to the geographic patterns concerning the activity of multinational firms. According to data from the 2009 Benchmark Survey of the BEA, 68% of US-based multina- tionals’ employment is at their parent firm, while only 32% is accounted for by their majority- owned foreign affiliates. In this sense, the national identity of a firm appears to be strong. Over time, however, the share of parent firm employment in a US multinational’s global employment has fallen from nearly 79% in 1989.
Now let us consider the location of affiliates relative to their parents. Estimates from gravity equations show that in the aggregate and at the level of the firm, multinational firms’ production activities drop rapidly in distance between the parent and the foreign location, and the pace of this decline is only modestly lower than the decline in export sales. These gravity results have been shown both for multinationals that originate in developed countries (Buch et al. 2005, Yeaple 2009, Chen&Moore 2010) and for multinationals that originate in developing countries (Fajgelbaum et al. 2011). Brainard (1997) and Helpman et al. (2004), among others, show that the foreign sales of US multinationals’ affiliates are suppressed less by measures of trade costs than are exports from the United States.
5The concentration of activity in developed countries has dropped in recent years. In 1999, the value added of US-owned affiliates in developed countries accounted for 89% of the global value added of US-owned affiliates. 6Multinational firms that originate from developed countries may have a different investment behavior than those from developed countries. Lipsey & Sjöholm (2011) discuss some evidence that developing country multinationals may be more likely to invest in other developing countries than are firms from developed countries (see also Fajgelbaum et al. 2011).
197www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
This information is summarized as Fact 2.
Fact 2: Most economic activity within multinational firms is concentrated in the parent country, and the economic activity of affiliates is concentrated in countries that are close to their parents. Affiliates’ sales are less geographically concentrated than are exports, however.
We now turn our attention to the cross-industry variation in the importance of multinational firms in economic activity. According to BEA andOECDdata, multinational activity as measured by the share of an industry’s sales or employment can be predicted by features of an industry’s technology. Affiliates located in developed countries account for a larger share of host country value added and employment in high research and development (R&D) and capital-intensive industries. US customs data also show that industry R&D and capital intensity play an important role in predicting the share of international trade that is conducted within multinational firms rather than at arm’s length (e.g., Bernard et al. 2012).
It is important to note that much of the existing evidence is gleaned from developed country data. There is some evidence that multinationals that originate from developing countries may be predominantly in different industries than are those from developed countries. Lipsey & Sjöholm (2011) provide ameta-analysis of a number of studies and conclude that the affiliates of developing country multinationals are more concentrated in unskilled labor–intensive industries, such as food, textiles, apparel, and wood products than are the affiliates of developed country mul- tinationals. This suggests that the nature-of-ownership advantages of multinationals depend in part on their source countries.
We summarize this information in the following fact.
Fact 3: Multinational firms account for a larger share of economic activity in capital- and R&D-intensive industries, although that pattern is less pronounced among developing country multinationals.
We now turn to howmultinational firms differ fromnonmultinationalswithin industries. There is substantial evidence that both the parents and affiliates of multinational firms are much larger, more productive, andmore export oriented than other firms within the same industries. For instance, BEA and census data for 2009 reveal that the 1,079USparent firms inmanufacturing industries accounted for less than 1% of USmanufacturing firms but 59% of employment, 61% of value added, and 44% of exports. Similar patterns emerge for the affiliates of multinational firms.7 In France, multinational affiliates accounted for 2% of manufacturing enterprises but 26% of employment, 32% of sales, and 40% of exports (OECD 2007; see also Mayer & Ottaviano 2007 for total factor productivity comparisons of multinational and nonmultinational firms for a number of European countries). The export participation by affiliates of multinational firms relative to domestic firms is particularly high for a number of export-oriented economies. According to Manova et al. (2011), firms with foreign ownership participation account for 77% of Chinese manufacturing exports. OECD (2007) data indicate that 92% of Irish manufacturing exports were accounted for by the affiliates of foreign firms.
Although much production activity is concentrated within multinational firms, a relatively small number of multinationals account for a large share of multinational activity. Data from the 1999 BEA Benchmark Survey indicate that the top 1% of USmultinational firms, ranked in terms of the size of their global operations, accounted for 38% of all affiliate sales, while the bottom
7Using customs data on related party trade, Bernard et al. (2009) show that multinational entities (US parents and the US affiliates of foreign firms) account for up to 90% of US trade.
198 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
50% of firms accounted for less than 2% of affiliate sales. This size disparity is even more striking with respect to intrafirm trade, with the top 1% of firms, ranked in terms of their global operations, accounting for 45% of parent exports to affiliates and 53% of parent imports from their affiliates. The bottom 50% again accounted for roughly 3% of sales, respectively (see Ramondo et al. 2012 for further details on heterogeneity in the distribution of intrafirm trade).8
We summarize this information in the following fact.
Fact 4: The operations ofmultinational firms are concentrated in a very small number of firms, and both the affiliates and parents of these firms tend to be larger, more productive, and more export oriented than other firms.
Within multinational firms, parent and affiliates concentrate to some degree in different activities. The share of parents in the R&D expenditures of multinational firms in 2009 was nearly 85%, while their share of employment was only 68%. Affiliates tend to be more focused on serving foreign markets than on providing inputs to their parent firms. In 2009, on average 61% of the sales of the foreign affiliates of US multinationals were made in their host country markets, and only 9% of these sales were back to the United States. Interestingly, the affiliates of multinational firms tend to be more export oriented than their parents. In 2009, parents exported only approximately 11% of their output, while 39% of sales of affiliates were outside their host country markets.
We summarize this information in the following fact.
Fact 5: Within the multinational firm, parents are relatively specialized in R&D activities, whereas affiliates are specialized in serving foreign markets rather than exporting to their parent country.
We conclude this section with empirical regularities on the mode of entry of multinationals into foreign markets. According to UNCTAD (2011), the value of cross-border mergers and acquisitions relative to FDI flows exceeded 50% in 2007. This number masks empirical regularities across countries, however. Among developed countries, cross-border acquisitions accounted for nearly 70% of FDI inflows, while for developing countries the share was less than 20%.
At the microlevel, many empirical regularities have been documented. Numerous papers have shown that multinational firms are choosy when selecting a local firm to acquire. For instance, Arnold & Javorcik (2009) and Guadalupe et al. (2012) show that firms acquired by foreign multinationals tend to have above-average productivity. Nocke & Yeaple (2008) show that among the US multinationals entering a foreign market, the more productive the investing firm is, the more likely it is to enter through greenfield investment rather than cross-border acquisition.
We summarize this information in the following fact.
Fact 6: Cross-border acquisitions account for a large portion of global FDI, par- ticularly into developed countries. More productive firms are more likely to enter through greenfield investment than to acquire an existing local firm, but affiliates acquired through a cross-border acquisition tend to be more productive than average in the target country.
8Among all firms engaged in international trade, the concentration in a handful of firms is even higher. Bernard et al. (2009) find that the top 1% of US exporters accounted for 90% of US trade.
199www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
3. WHO PRODUCES WHERE?
Geographic frictions define the field of international trade. Political jurisdictions are defined by borders, and policies tend to discriminate against agents from outside the border. Tariffs dis- criminate against agents located in other countries, and contracts may be harder to enforce across borders. Legal restrictions and cultural differences limit factor mobility across space. Goods are mobile geographically, but their movement is subject to physical shipping costs, tariffs, and legal barriers to trade. Finally, information available at one location may be costly to obtain elsewhere.
This section addresses the reasons why some firms facing the various types of geographic frictions choose to own plants in multiple countries. It begins with traditional approaches and the relevant empirical evidence. More recent theory is then described, along with the extent to which it has improved the fit between theory and data.
3.1. Traditional Approaches and Recent Extensions
Historically, the literature on international trade ignored the role played by multinational firms, treating FDI as faceless capital flows between countries. In 1984, two important papers by Helpman and Markusen, respectively, were published that integrated multinational firms into traditional trade models. Subsequently, much of literature on the multinational firm builds on the modeling techniques of these two papers.
Helpman (1984) embeds multinational firms into a two-country, two-good Heckscher-Ohlin model that is augmented to include product differentiation, scale economies, and monopolistic competition in one sector. A firm that enters in the differentiated goods sector produces a dis- tinct variety over which it is the sole producer (its ownership advantage). The countries have endowments of capital and labor that are immobile internationally, and the differentiated good industry is relatively capital intensive, giving capital-abundant countries a comparative advantage in its production. Trade in goods is frictionless. If the two countries are sufficiently similar in terms of their relative factor abundances, then factor prices are equalized by trade in final goods, and the capital-abundant country exports the capital-intensive good. If the relative factor abundances of the two countries become sufficiently different, then capital prices rise in the capital-scarce country, and wages rise in the labor-scarce country.
Now suppose that theproductionof the capital-intensive good canbe split into a capital-intensive headquarter services (e.g., coordination of production, development of intangible assets) and a labor-intensive production activity. Firms from the capital-abundant country now have an incentive to fragment the production of their differentiated good vertically, with the capital-intensive headquarter service in the capital-abundant country and the labor-intensive production activ- ity located in the labor-abundant country. No factors have moved, but the services of those factors are embodied in the communication from the headquarters to the plant. This factor service flow arises only because there is a foreign affiliate. This type of multinational activity is often called vertical investment.9
Markusen (1984) follows a different approach, focusing on the public good nature of knowledge (an ownership advantage) within the firm. Once an intangible asset, such as the
9Helpman (1985) extends the model to include moderately capital-intensive intermediate inputs that the parent (headquarter) provides to its affiliates. The amended model generates a rich array of multinational firm behavior. Parent firms provide (invisible) headquarter services to their affiliates and export intermediate inputs to their affiliates (intrafirm trade). Affiliates sell their product in the host country market (local affiliate sales) and export their product back to the parent firm (intrafirm trade). He also allows for firms that producemultiple products so that it is possible for a firm to export final goods to the same country in which it owns an affiliate.
200 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
blueprint to produce a particular good, has been developed, it is then nonrival within the firm. Multinational production exploits a particular form of economies of scale whose origin lies in replicating the same production activities across production locations. That is, once the blueprint has been developed, it can be combined with immobile factors in multiple locations by a single firm. In the absence of such a firm, redundant development costs would have to be incurred to make the technology available to geographically dispersed immobile factors. Because this type of multinational activity involves replicating the same activity in multiple countries, it is referred to as horizontal investment.
Neither Helpman (1984) nor Markusen (1984) considers geographic frictions (other than international factor immobility). The role of trade frictions in generatingmultinational production is prominently featured by Horstmann & Markusen (1992), who focus on a partial equilibrium model in which up to two firms (one in each of two symmetric countries) may enter by paying a fixed product development cost. After entering, the firms can choose which of two countries (or both) in which to produce their good. Each production location requires the firm to pay a plant-level fixed cost. If the firm exports its good, then it pays a variable trade cost.10
Several implications arise in the analysis of the model. First, firms face a proximity-concentration trade-off in their production decisions that arises because consumers are immobile and goods are costly to ship.11 If a firm opens a plant in each country, it foregoes economies of scale by incurring the plant-level fixed cost twice but avoids the shipping cost. Second, the number of entrants (duopoly versus monopoly market structures) also depends on these technological var- iables. Not surprisingly, multiplant production is more likely when plant-level fixed costs are low and trade costs are high. Less obvious is the role of the corporate fixed cost. When corporate fixed costs are high, only one firm will be active, and its sales in each market will be large, making multiplant operation more profitable.12
Markusen (2002, chapter 7) presents a general equilibrium model that encompasses both vertical and horizontal motives in a two-country setting. This model focuses on the interaction among comparative advantage, trade costs, and economies of scale. The assumptions are similar to those of Helpman (1984) but allow for the existence of trade costs. Trade costs substantially complicate the model analysis, and the predictions of the model, which also hinge on various factor intensity assumptions, have to be teased out of simulation exercises (for an effort to test particular features of these models, see Carr et al. 2001). An important prediction is that horizontal or rep- licating multinationals are pervasive between similarly sized and endowed countries (for certain factor intensity assumptions, they are consistent with Fact 1). When countries are very asymmetric in terms of size and endowments, the incentive to export goods becomes strong so that either national exporters or vertically organized multinationals prevail.
Brainard (1997) tests the implications of the proximity-concentration framework. (Examples of earlier studies are Horst 1972 and Swedenborg 1979.) Using industry-country-level data from the 1989 BEA survey, Brainard estimates a simple econometric model in which the dependent variable is the logarithm of the ratio of affiliate sales by country and by industry made in the host country market to the sum of affiliate sales and arm’s length exports from the source country.
10A related paper is by Horstmann & Markusen (1987), who focus on the possibility that a firm might invest in a given market to preempt entry by a local firm. 11The term proximity-concentration trade-off appears to be due to Brainard (1993), whose model differs fromHorstmann& Markusen’s by starting with the Krugman (1980) monopolistic competition framework rather than the reciprocal dumping framework. The idea that affiliate production is motivated by costs of trade goes back many decades. 12Markusen & Venables (1998, 2000) embed horizontal multinationals in a general equilibrium setting to analyze the role of country factor abundance and size in a proximity-concentration setting.
201www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
This independent variable has the benefit of describing the degree of substitution between local affiliate sales and exports while controlling for industry and country characteristics that jointly determine total sales. The explanatory variables are proxies for key proximity-concentration variables, such as freight and insurance costs, tariffs, plant-level fixed costs, and corporate-level fixed costs. Also incorporated are controls, including a crude measure of factor endowment differences. Brainard finds that firms substitute affiliate sales for exports when trade costs and corporate fixed costs are high and plant-level fixed costs are low, as predicted by the proximity- concentration framework.Moreover, she shows that differences between host country and source country GDP per capita are associated with less foreign affiliate sales, which is inconsistent with simple models of vertical investment.13
Brainard (1997) has been influential in supporting the horizontal or proximity-concentration approach tomultinational firms, but there are limitations. For instance, themeasures of fixed costs are problematic, particularly in light of a growing literature on firm heterogeneity (discussed below). Plant-level fixed costs are measured as the number of nonproduction workers at the median-sized plant ranked by value addedwithin an industry, but given the extent of heterogeneity across firms within industries in terms of their size, it is not clear how to interpret this measure. More worryingly, in specifications that include both country and industry fixed effects, the coefficients on tariffs and freight costs go to 0.
Themodels of horizontal investment described above consider environments in which all firms are identical so that if some firms export and others engage in multinational production, all firms are indifferent between the twomodes. This is at oddswith Fact 4, that the parents and affiliates of multinational firms are very different from other firms operating in a given economy. Helpman et al. (2004) combine the proximity-concentration model of Brainard (1993) with the firm heterogeneitymodel ofMelitz (2003). Specifically, firms pay a fixed cost to develop a firm-specific variety and in the process learn their firm-specific productivity (intangible asset conferring an ownership advantage). Once this productivity is drawn, a firm may serve its domestic market by paying a fixed cost to open a plant. In addition, a firm may export to a given foreign market, but doing so requires a firm to incur two types of trade costs, a fixed cost ofmarketing its product in the foreign country and a variable iceberg-type trade cost. Finally, a firm may open a plant in the foreignmarket and transfer its productivity to that plant by incurring the fixedmarketing cost plus the fixed cost of managing an additional plant. In so doing, the firm avoids trade costs.
Unlike the case of Brainard (1993), both exporters and multinationals appear in equilibrium, and all (but the firms at the cutoff) strictly prefer their organization of international production to all alternatives. Firms with greater productivity sell more in any given country and so can spread the plant-level fixed costs over a larger number of units sold. By avoiding trade costs, a firm serves an effectively larger market. In equilibrium, firms sort into modes of serving world markets, with the least productive firms exiting, the least productive active firms serving only their domestic market, moderately productive firms exporting to foreign markets, and the most productive firms opening a foreign affiliate. Using Compustat data, Helpman et al. (2004) show that multinational firms are more productive than exporters who are more productive than domestically oriented firms.14
13Using similar data for 1994, Yeaple (2003b) argues that this result obscures variation across industries. He shows that there is a tendency for US multinationals in skill-intensive industries to locate in skill-abundant countries and for US multinationals in less skill-intensive industries to locate in skill-scarce countries. 14This hypothesis finds further support in other papers, such as Girma et al. (2004). Fillat & Garetto (2012) show that the shares of multinational firms trade at a discount relative to nonmultinationals. This would seem to contradict their supposed productivity advantage. They show that amending Helpman et al. (2004) to allow for risk aversion in a dynamic setting eliminates this tension.
202 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Another implication of Helpman et al. (2004) is that the extent of firm productivity heterogeneity within an industry is a determinant of aggregate volumes of trade and multinational production. Using data similar to Brainard (1997) and controlling for standard proximity-concentration variables, Helpman et al. (2004) show that US firms substitute foreign production for exports from the United States in industries characterized by highly dispersed firm sales, and the effect, as measured by beta coefficients, is large relative to other industry characteristics.
Another prediction of Helpman et al.’s (2004) model is that more productive firms are not only more likely to become multinationals, but are also more likely to own affiliates in a larger number of countries. This prediction is explored in depth by Chen&Moore (2010). Using French firm-level data from the Amadeus data set, they calculate firm-level productivity and investigate the mapping from parent productivity to the production locations chosen by the firm. They find that more productive firms are more likely to be found in any given country, including those with high labor costs and high barriers to entry.15
Finally, Baldwin & Ottaviano (2001) extend the proximity-concentration framework to the case of multiproduct firms. This extension is desirable because large multinational firms produce a large portfolio of products, because firms are frequently observed to serve customers in a given country by both exports and affiliate sales, and because these firms may internalize effects of expanding the production of one product on the profitability of another. Moving production of a good offshore avoids marginal trade costs, which raises profits earned for that good, but the resulting cannibalization effect lowers the profits earned by the firm’s other products. When fixed costs are high so that one plant is optimal for each good, a profit-maximizing firm will produce some products in one country and some in the other to minimize cannibalization effects, therefore generating two-way trade between countries within the same firm.
3.2. Hybrid Models with Simple Geographies
Although the proximity-concentration trade-off is consistent with the share of sales that is through local affiliates rather than by exports from the source country, standard proximity- concentration models do not predict correctly the levels that are observed in the data. According to Fact 2, a source country’s exports and the local sales of its affiliates are decreasing in measures of trade cost. Yeaple (2009) shows that the decline of aggregate affiliate sales in distance occurs both because fewer firms locate affiliates in more distant locations and because the affiliates of those that do sell less the further they are from the United States. (The same pattern is found for German multinationals in Buch et al. 2005.) This latter result suggests that distance is associated with either risingmarginal costs at the affiliate level or lower demand.16 In this section, we discuss some models that blend vertical and horizontal elements in essentially a two-country setting to account for this fact.
There are several mechanisms that can cause distance to discourage multinational operations. Although simple models of vertical multinational production may have trouble explaining why most FDI is two way between similarly endowed countries, integrating features of vertical FDI models with features of horizontal FDI models to produce a hybrid model can help explain why distance deters foreign production. We start off with simple adjustments to the simple geography of a two-country framework.
15Yeaple (2009) also explores similar predictions using data forUSmultinationals but focuses onmodel-consistent aggregates. The results are consistent with those of Chen & Moore (2010). 16Also, multinational production has grown faster than exports during a time period in which trade costs were falling. This is the time dimension to the cross-section puzzle.
203www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Suppose that a production technology features a continuum of intermediates that are costlessly assembled into a final good. Each intermediate differs in its knowledge intensity, and there is an efficiency loss associated with transmitting knowledge across locations. The efficiency loss is in- creasing in the knowledge intensity of the intermediate. Now suppose the alternative to trans- mitting the technology is to produce the intermediate near the parent company and then to ship that intermediate to the foreign location. Goods that require knowledge-intensive intermediates would then involve high levels of intrafirm trade between the parent country and affiliate (vertical FDI), but as trade costs rise, more value added is done at the affiliate (horizontal FDI).17
This would mean that imported intermediates in total affiliate costs would fall with distance, and affiliate sales would also drop as the firm becomes increasingly exposed to knowledge transmission costs. Furthermore, the intrafirm trade share would be less sensitive to trade costs in high-knowledge-intensive industries, while affiliate sales would be more sensitive to trade costs in high-knowledge-intensive industries.
Keller & Yeaple (2013) embed this mechanism in Helpman et al.’s (2004) model, test these predictions using a panel of US multinational firms from the BEA data set, and provide a cal- culation for how much of the gravity result can be explained by this mechanism. (Earlier theory in this literature includes Zhang & Markusen 1999, and earlier empirical work on the intrafirm trade of US multinationals is Hanson et al. 2005.) Trade and technology transfer costs can account for approximately 30% of the effect of gravity, with the remaining effect of gravity resulting from fixed costs that rise in distance (15%) and distance-dampening effects on demand for US goods (55%).
Irarrazabal et al. (2010) structurally estimate a version of Helpman et al.’s (2004) model that has been extended to include parent-affiliate trade in intermediates. As in Keller&Yeaple (2013), this trade exposes affiliates to trade costs. The key object to estimate is the share of intermediate inputs in affiliate costs that are necessary to rationalize gravity in the firm-level data. (The authors use an interesting data set for Norwegian firms that contains both export and affiliate sales in- formation but does not break out intrafirm trade.) Their answer is 90%. Thiswhopping number is obtained after having accounted for distance-related variation in fixed costs of entry. One way to interpret this result is that the proximity-concentration framework is only marginally relevant because making it fit requires the marginal costs of affiliates to rise in distance from the parent at nearly the same rate as trade costs.
The interaction between vertical and horizontal motives for multinational production can also be analyzed by observing the structure of multinational firms’ operations before and after a trade liberalization for a single country pair. This is the strategy of Feinberg & Keane (2006), who structurally estimate a dynamic hybrid model using firm-level BEA data that span the Canada-US Free Trade Agreement. Such a hybrid model is important in the Canadian-US context as the value of intermediate inputs imported by affiliates is more than one-third of the aggregate affiliate value added, and two-thirds of Canadian affiliates import intermediates, export intermediates, and sell their product locally. Over a 10-year period that spans the free trade agreement, Canadian affiliates of US multinationals became much more integrated with their parents, selling a substantially smaller portion of their output in the Canadian market. The authors find that their model attributes a one-third increase in the volume of arm’s length multinational firm trade and a 5–10% increase in intrafirm trade to the tariff reduction. These numbers, although large, are much less than the actual changes, leading the model to attribute
17Blonigen (2001) documents such a mixture of activities at work in data on the operations of Japanese automobiles in the United States.
204 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
much of the change in multinationals’ organization to technological change, such as the rise of just-in-time delivery.18
3.3. Hybrid Models with Complex Geographies
Until recently, the literature has been confined to two-country models that cannot capture the full complexity of actual multinational operations. As noted above, about one-third of sales of the affiliates of US multinational firms are to countries other than the affiliate’s host country or the United States. Some portion of these sales undoubtedly is consistent with a proximity- concentration framework in which an affiliate in a continent sells to multiple countries, but the primary motive is to avoid shipping costs from the source (or parent) country. However, for at least an important subset of firms, the structure of their global operations involves fragmenting the production of some inputs to low-cost locations where factor prices encourage concen- tration, while other activities, such as assembly, might be replicated in many locations. We refer to firms that replicate some activities in many countries while concentrating other activities in a few countries as firms that follow complex strategies.
Complex strategies aremodeled explicitly byYeaple (2003a), who considers an environment in which there are three production activities that differ in their relative factor intensities and three countries, of which two are identical and skill abundant (the north) and one is abundant in unskilled labor (the south). The differences in factor abundances mean that entry will be done in the north, while one of the intermediate inputs is least costly to produce in the south. Trade costs between countries motivate the replication of production activities while discouraging fragmentation. Any kind of multinational production incurs a fixed cost. In the model, both types of multinational production lower marginal cost in a complementary manner. Replicating pro- duction abroad lowers the marginal cost by avoiding transport costs, whereas fragmenting pro- duction processes lowers the marginal cost by accessing low-cost southern labor. A firm that avoids trade costs by replicating one stage of production increases its sales and so gains even more from fragmenting production and vice versa.
This mechanism has many implications. For instance, although high trade costs encourage replication and discourage fragmentation, complementarity means that for intermediate levels of trade costs, both activities must be undertaken for multinational production to be viable. As a result, a reduction in trade costs can induce replication by increasing the return to comple- mentary fragmentation, and an increase in trade costs can induce fragmentation by increasing the return to complementary replication. Finally, policy changes in one country (e.g., the Czech Republic) could make multinational production in another location (e.g., Germany) more desir- able. Additional issues concerning various complementarities and the role of firm heterogeneity are taken up in Grossman et al. (2006) and Yeaple (2008).
Although Yeaple (2003a) provides some guidance on how to think about the geographic complexities facing multinationals in their location decisions, the geographic and production structure considered is special.19 For instance, there is no sense of asymmetric access to the
18Feinberg & Keane (2001) consider a reduced-form approach to the same experience. One notable result from that paper is the extent of parameter heterogeneity across firms within the same industry:Within-industry responses to trade liberalization show greater variance than across-industry responses. This result raises the question as to the usefulness of industry classifications. 19We assume throughout that intermediate inputs are gathered worldwide and then assembled into final goods at a single location. Baldwin&Venables (2010) refer to this as a “spider”: Intermediates are drawn to a central location. An alternative production structure is the “snake” in which a sequence of activities needs to be done on the good as it progresses toward its final stage. Baldwin & Venables show that the location activity can be very different depending on the production structure.
205www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
export platform of the south, and there is no particular sequence in which production must occur. Some of these issues are addressed by Ekholm et al. (2007), who also consider a three- country model in which intermediates must be provided by the multinational firm’s home country. They explicitly allow one of the two northern countries to have favored access to a southern export platform, which leads to very different cross-country dependencies (for further analysis of export platform models, see Neary & Mrázová 2011).
Models of export-platform FDI or complex strategies suggest that empirical work needs to account for the possibility that changes in the characteristics of a country’s neighbors can alter the attractiveness of that country as a production location. Both Head & Mayer (2004) and Yeaple (2008) show that affiliates cluster in centrally located production sites, as measured by the foreign market potential of the region.Alfaro&Charlton (2009) go further in integrating vertical linkages by introducing input-output relationships into the analysis of affiliate clustering. They show that there is a tendency of vertically linked affiliates to cluster geographically but cannot provide direct evidence of trade links between affiliates. This may be problematic. Ramondo et al. (2012) provide evidence that input trade between affiliates and that between affiliates and parents may not be important for a large set of firms so that input-output relationships may be poor proxies for trade linkages.
Empirical work accounting for these interdependencies is scarce as the spatial econometric techniques aredifficult touse. TheworkofBlonigen et al. (2007) is representative of such research. Using a spatial lag and measures of the market potential of a particular location, they show that a country’s neighborhood predicts the volume of US multinational affiliate activity in a particular location. More importantly, they show that the inclusion of these measures does not dramatically alter the coefficient estimates on other variables. Less encouraging is their result that the coefficient estimates are not robust across subsamples in their data. This latter result may reflect the limitations of existing spatial econometric techniques that require a priori knowledge of the exact nature of cross-country interactions.
Differences in relative demand for goods across space add another layer of complexity to geography. The role of nonhomothetic preferences in understanding the structure of global multinational operations is taken up by Fajgelbaum et al. (2011), who consider a four-country, north-south model. They show empirically that multinational firms that originate in developing countries are more likely to invest in other developing countries and that firms originating in developed countries invest primarily in other developed countries. They consider a non- homothetic preference system in which goods differ in their quality and are horizontally dif- ferentiated aswell. Fixed costs to entry and geographic frictions interact to create a homemarket effect: Firms producing high-quality goods are more likely to enter in developed countries and then invest (owing to proximity benefits) in other developed countries, whereas the opposite is true for firms producing lower-quality goods. In terms of the ownership, location, and in- ternalization framework, developed country firms have ownership advantages that are more valuable in developed countries, whereas developing country firms have ownership advantages that are more valuable in developing countries.
3.4. Quantifiable Models with Complex Geographies
We now turn our attention to a new branch of the literature that develops quantifiable models that incorporate many of the mechanisms described above and that allows for counterfactuals that fully incorporate general equilibrium feedback effects in a multicountry setting.
Arkolakis et al. (2012) introduce a model in which countries are distinguished by their size, by their comparative advantage in introducing new technologies, and by their location. Firms pay a fixed cost to invent a new product and receive a vector of productivities in which each element of
206 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
this vector corresponds to a particular country. In this way, comparative advantage is introduced into the model without the complex issues regarding the factor intensities of goods and the factor abundances of countries.Having paid a fixedmarketing cost in a given country, the firmminimizes its cost of serving that country by choosing the lowest-cost location, where local costs depend on the constellation of trade costs, technology transfer costs, the local wage, and comparative ad- vantage. In this way, the model captures a trade-off between proximity and comparative ad- vantage in an environment that provides firms the full set of locational opportunities, including export platform investment.
Arkolakis et al. (2012) focus on how geography, increasing returns, and comparative ad- vantage determine the location of innovation versus production.With both technology and output costly to move internationally, fixed costs of entry make the market size a key determinant of the location of innovation versus production. This home market effect coexists with the com- parative advantage of entry (the quality of productivity draws on average by country) in de- termining the structure of global innovation and production by multinational firms. The authors fit the model to aggregate bilateral trade and multinational production data. Although the home market and the comparative advantage effects are not separately identified by the available data, comparative static exercises are readily implemented because changes in trade and technology transfer cost frictions between countries can be considered independently of other country characteristics. The authors find that the gains from openness are large, particularly for small countries located in densely populated areas. The counterfactual exercises suggest, however, that around the calibrated values of trade and information costs, small reductions in these costs can lower the welfare of individual countries.
Arkolakis et al. (2012) achieve tractability in their framework by abstracting from fixed costs of production to focus on the location of global entry versus production. But, as noted above, fixed costs of production are thought to play an important role in the structure of multinational firms’ operations. In the presence of fixed costs at the plant level and the possibility of export platform FDI, a firm’s decision to establish foreign plants is interdependent across countries, leading to a potentially very difficult discrete choice problem at the firm level.
Tintelnot (2012) quantifies the size and importance of these fixed costs in a general equilibrium model that allows for export platform FDI. He achieves tractability in his model by making the problem of the firm smooth through several assumptions. First, in each country, there is a fixed set of potential parent firms, and each firm is endowed with the ability to produce a continuum of goods that it will sell in every country. Second, the firm is also endowed with a core productivity over its goods and a vector of idiosyncratic fixed costs of opening a plant in each foreign country. Third, if a firm incurs the fixed cost in a foreign country, it then obtains a country-specific draw for each of its goods (comparative advantage). Fourth, given the productivity draws made available from its plant location choices, the firm minimizes its cost of serving global markets, taking into account technology transfer costs, shipping costs, and wages. Each additional location in which a firmopens aplant raises a firm’s variable profits (adding an entirely new set of productivity draws for each product) but at a decreasing rate as new locations cannibalize sales from other locations. In a sense, the mechanism of Eaton&Kortum (2002) has been subsumed inside the firm to create smoothness in the payoffs of the firm’s options.
Tintelnot (2012) uses his model to two ends. First, he structurally estimates the model on firm- level data for German multinationals. He finds that the fixed costs of production, which lie between 4 and 8 million euros on average for those firms that established a plant in the respective country, account for a significant amount of the home bias in production that exists at the firm level. Tintelnot then calibrates his model to bilateral trade andmultinational production data. The model fits the datawell: Although he does not fit themodel to affiliate exports, themodel generates
207www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
artificial export platformdata that are remarkably consistentwith the data.Model counterfactuals demonstrate the importance of allowing for export platform investment.
3.5. Financial Markets and the Multinational Firm
This section concludes with a brief discussion of how financial frictions across countries may shape the operations of multinational firms. Multinational firms tend to be large and highly di- versified firms that can access credit markets in multiple jurisdictions. As such, they may be in a favoredpositionvis-à-vis local firms in countries with dysfunctional financial systems as they can shift financing needs from local credit markets to their internal credit markets. There is a sub- stantial body of empirical evidence that this is so. For instance, Desai et al. (2004a) show that in countries with poor creditor rights, US multinational firms shift their borrowing away from external markets and toward their parent firms. Another example is Manova et al. (2011), who show that the foreign affiliates of multinational firms in China are better represented in China’s exports in industries in which external financial needs are high and opportunities for collater- alization are low.
One way to establish the importance of credit to multinational expansion is to see what hap- pens during a financial crisis. Klein et al. (2002) demonstrate that foreign investments were highly curtailed for Japanese multinationals that had strong ties with the Japanese banks that were most affected by the collapse of Japanese assets prices in the early 1990s. Alternatively, during a financial crisis, credit-constrained local firms are often sold to multinationals with deep pockets (e.g., Acharya et al. 2011).
4. CROSS-BORDER ACQUISITIONS VERSUS GREENFIELD INVESTMENT
According to Fact 6, many firms obtain ownership of foreign production facilities by acquiring an existing facility rather than by opening a new affiliate, but there is little agreement as to how to interpret this fact. On the one hand, the models presented in the previous section may need to be amended only modestly if the acquired firms are simply a bundle of primary factors, such as real estate or capital. On the other hand, it may be that foreign firms acquire domestic firms because theywant to obtain access to intangible assets that they have difficulty creating themselves. If this is the case, then a new question arises: Why do foreign firms value these assets more than domestic firms? Another possibility is that cross-border acquisitions involve an attempt to reduce compe- tition in a given market. If this is the case, then we might think very differently about the facts. If the goal of a merger is to blunt product market competition, then it may make sense that firms acquire other firms in countries in which their competitors are located (a possible explanation for Fact 1).
We consider first the possibility that cross-border acquisitions are driven by the desire to ac- quire intangible assets. Suppose that firms are bundles of intangible assets and that these assets are complementary in generating profits with the firms. In the context of an international envi- ronment, we might think of some types of firm-specific assets as internationally mobile, such as high-quality management techniques or access to proprietary technology, and other types of firm-specific assets as location specific, such as reputation, knowledge of local conditions, or in- tegration into local production networks. This is the environment analyzed by Nocke & Yeaple (2007), who add synergy-driven cross-border acquisitions to the model of Helpman et al. (2004). Their focus is on the types of firms that engage in cross-border acquisitions relative to the types of firms that engage in other modes of foreign operations. The authors show that firm hetero- geneity matters. When firm heterogeneity is primarily in the quality of mobile assets, the firms
208 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
that engage in cross-border acquisitions tend to be the most productive firms. When firm het- erogeneity is primarily in the quality of immobile assets, the least productive firms engage in cross-border acquisitions.
Nocke & Yeaple (2008) consider a different twist on synergy-driven mergers in a world in which countries differ in their cost of production but trade between countries is frictionless. They show that greenfield FDI will tend to be one way from high-cost to low-cost locations as firms are willing to pay a lot to move highly productive assets to a low-cost production location. Two-way cross-border acquisitions arise between countries in which there are small differences in the cost of production in order to exploit small differences in the distribution of intangible assets across countries. Hence the model captures Fact 6: Most multinational entry between similar developed countries takes the form of cross-border acquisition, whereas most multinational entry into developing countries tends to be greenfield FDI.20 It also shows how large two-way volumes of FDI can be generated between similar developed countries (Fact 1) even in the absence of trade costs.
Guadalupe et al. (2012) consider a large panel of Spanish firms for which some firms came to be acquired by a foreign multinational. Using matching techniques to address selection issues, the authors show that foreign firms tend to acquire relatively more productive Spanish firms, which subsequently tend to become even more productive through process innovation. Acquired firms are also likely to become exporters after acquisition, often through exporting to the acquiring multinational. The results are consistent with the view that cross-border acquisitions involve synergies between foreign parents (access to foreign markets and ability in process innovation) and target firm characteristics (high productivity in the target country). Arnold& Javorcik (2009) use similar econometric techniques to those of Guadalupe et al. (2012) to study the impact of foreign acquisitions in the developing country context of Indonesia. Foreign firms acquire rela- tively well-performing Indonesian firms, and the subsequent total factor productivity and wage growth of foreign-acquired firms relative to matched domestic firms is faster. They also show that foreign acquisition is associated with increased international trade.21
Head & Ries (2008) address the multilateral structure of cross-border mergers and acquis- itions. In their framework, each country has a set of management teams and a set of firms that may be proportional to country size. Managers have different abilities in monitoring production in different plants. Cross-bordermergers occurwhen the optimalmanager for a given firm is located in a foreign country. Pushing against foreign control are monitoring costs that are a function of geography, and this gives rise to a gravity equation that rationalizes Fact 2.
Other research treats cross-border acquisition as a mechanism for reducing the degree of competition in an industry. For example, Neary (2007) models cross-border acquisitions in an oligopolistic environment in which firms from low-cost locations acquire firms from high-cost locations in order to reduce competition. Here the cost asymmetry across countries that results from comparative advantage is fundamental in driving international mergers. A central result in the paper is that a reduction in trade costs between countries can then spur consolidation through a cross-border merger wave. This is one possible explanation for how trade liberalization within Europe may have led to an increase in cross-border ownership. [Horn & Persson (2001) also
20Blonigen (1997) explains the acquisitionwave in the United States by Japanese firms as exchange-rate driven. Japanese firms have an ownership advantage serving their homemarket, and the appreciation of the yenmakesUS intangible assets that could be used to serve the Japanese market cheaper. This is a synergy story. 21These results contrast with a number of models, such as that of Gordon&Bovenberg (1996), who hypothesize that adverse selection would result in acquisitions being concentrated in poorly performing firms.
209www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
consider cross-border acquisitions in the context of a coalition-formation game.] It is not clear that this mechanism would be associated with the increased size of affiliate operations over the same period, however.
5. THE BOUNDARIES OF THE FIRM
Above we cover papers that assume that all production activities must be conducted within the firm and so ignore alternatives, such as outsourcing the production activities to an in- dependent firm. In practice, there is huge variation, even within the same industry, in the way that firms set their boundaries. For instance, whereas Apple contracts out much of the value added for its products, Lenovo keeps most of its operations in-house. We now briefly discuss some of the recent literature on the internalization problem. [Antràs & Rossi-Hansberg (2009) and Antràs (2013) also provide a discussion of much of the recent literature (see also Spencer 2005).]
In structuring their global operations, firms face difficulties associated with contracting with outside contractors on the one hand andwith firms’ own employees on the other hand. The nature and severity of imperfect contracting will have implications that vary across industries and across firms within an industry, depending on the nature of technology. For instance, in some industries, firms earn rents from intangible assets, and those rents could be diminished by contract failure (asset dissipation).Anarm’s length contractormight damagea firm’s reputation for quality or steal proprietary secrets. In other industries, relationship-specific investments might be necessary, giving rise to the potential for hold-up problems. Having produced an input for a firm, an outside contractor may find that it is denied full payment, whereas the purchaser may find courts unlikely to support a claim of shoddy manufacturing.
An important paper in the asset dissipation literature is by Ethier & Markusen (1996), who construct a model in which firms compete to invent a new product that can generate rents for a fixed amount of time before becoming obsolete.22 The firm then chooses among exporting the product, licensing the technology to a foreign producer, or opening a multinational affiliate to serve a foreign market. Exporting forces the firm to incur trade costs but prevents the technology from becoming available to local producers prematurely, whereas both licensing and multina- tional production require the firm to teach a foreignmanagement team the technology and to incur a fixed cost. Asset dissipation occurs when a foreign management team forms a separate firm to compete with the inventor, thereby adversely affecting the inventor’s rents.23
To prevent the dissipation of its technology, the firm can export early in the technology’s life cycle and then later license the technology to a local producer as the technology becomes generally available. If a local producer is involved early in the technology’s life cycle, the contract that is arranged, whether the local producer is an owned affiliate or is an arm’s length contractor, must be such that neither the inventor nor the local producer will defect from the agreement. The inventor may always export the product to the market to compete with a local firm that has defected, and this creates both a punishment to the defector and a temptation to defect for the inventor. The model gives rise to interesting interactions between locational characteristics and the internalization problem. For instance, if trade costs rise, this makes the desire to transfer
22Horstmann & Markusen (1987) is another important early work. There, a firm may internalize production when a local supplier may lower product quality and so damage the firm’s reputation. 23Examples of such behavior are common in the popular business press, particularly in developing countries with poor legal institutions. The possibility of such technology spillovers has led to a large literature that asks whether multinational production results in the increased productivity of local firms (see Keller 2010 for a review).
210 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
technology stronger but also limits the ability of the innovator to punish the defector. High trade costs may even induce a firm to export early in the technology’s life cycle.
Recent empirical work by Bilir (2011) supports the hypothesis that asset dissipation may be an important factor in the location decisions of US multinational firms. She presents a model in which products are developed in the north and may be imitated in the south. Industries differ in their technological life cycle, which is the length of time that a technology avoids becoming obsolete. Imitation is costly and limited to some extent by the patent protection provided by the south. Bilir shows that multinational firms will not open an affiliate in the south until the expected time to obsolescence hits a critical level, which will be early in short life-cycle industries and late in long life-cycle industries. Patent protection intuitively plays a more important role in attracting multinational firms in long life-cycle industries.
Bilir (2011) tests this prediction (and other related ones) of this model on panel data for US multinationals. By constructing a measure of the product life cycle from the average lag of patent citations and interacting this measure with (time-varying) measures of intellectual property rights protection, she makes a convincing case that multinational activity is more likely to be proscribed in long life-cycle industries in which patent protections are weak. Although the analysis has nothing to say about arm’s length transactions per se, it provides strong support to the idea that asset dissipation is a serious concern for multinational firms.24
Much recent work, both in terms of theory and empirics, has been motivated by the work of Antràs (2003). Antràs (2003) adapts Grossman &Hart’s (1986) property rights approach to the firm to the case of the international sourcing of intermediate inputs in a simple and intuitive way. To produce a final good, intermediates must first be produced. These intermediates require two agents, a final-good producing firm F and a supplier of the intermediate S, to make relationship- specific investments. These investments are aggregated via a Cobb-Douglas production function, where the cost share on F’s investment represents the relative importance of the firm’s contribution to joint output, while the remaining cost share is the contribution of the manager. The two agents cannot contract over the level of investment made. Instead, after the investments have been made by F and S, the two agents engage in Nash bargaining to split the surplus. Critically, the outside option of the two agents depends on the organizational form. If the two agents were not part of the same firm, then both have an outside option of 0, whereas if S were an employee of F, F may fire S, seize the inputs, and assemble the final good with some loss of productivity.
Given the ex post Nash bargaining, there will always be underinvestment by both agents, but the extent of the underinvestment of a given agent depends on the organizational form.25 When intermediate input production is outsourced, S getsmore of the surplus in theNash bargaining and so has an incentive to make a larger investment, whereas the opposite is true when intermediate input production is done within the firm. Hence we should be more likely to see vertical integration in industries in which the technology dictates that Fmakes the relatively more important investment.
An important feature of the Antràs framework is that it can make direct contact with the data. To the extent that intermediate input trade occurs across borders, firms’ integration decisions are reflected in the share of international trade that is conducted between parties related by ownership. These data are readily available, although firm-level data on contracting are not.
24This result is consistent with survey evidence from Mansfield & Romeo (1980), who find that firms were reluctant to transfer their newer technologies to affiliates in countries in which intellectual property rights were poor. 25Note that Antràs does not allow partial ownership of the intermediate producer plant such as might be the case in a joint venture. This can be justified by the fact that partial ownership is uncommon among developed country affiliates. Desai et al. (2004b) provide an analysis of why joint ventures are rare and are becoming rarer.
211www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
The Antràs mechanism has been applied to a number of interesting contexts. Antràs (2003) embeds this mechanism in a Heckscher-Ohlin setting. The model predicts that related party trade will predominate between similar countries that are abundant in the resource used most intensively in industries whose technologies favor vertical integration (and so can explain Fact 1). Antràs & Helpman (2004, 2008) derive predictions over the share of trade that is intrafirm in a north-south partial equilibrium setting that integrates firm heterogeneity and variation across countries in contract enforcement. The integration of firm heterogeneity allows the analysis of firm-level decisions and allows smooth aggregation to the industry level, as in Helpman et al. (2004).26
The intrafirm trade share implications of Antràs & Helpman’s model have motivated much empirical work. Given space constraints, we discuss only three of the most recent and most comprehensive papers. Nunn & Trefler (2012) work with a highly disaggregated sample (Harmonized System six digit) of US import data that distinguishes between related party trade and arm’s length trade. They construct measures of industry characteristics that are plausibly correlated with the relative importance of relationship-specific investments and then regress the intrafirm trade share on these variables, letting “the data speak.” The authors find that an industry’s skill, capital, and R&D intensity predict intrafirm trade shares, as one might expect. Going further, they show that the type of capital intensity matters: Industries that use a lot of capital that is not firm specific (i.e., autos or computers) do not tend to display high levels of intraindustry trade. The authors extend their analysis to address the specific predictions of Antràs & Helpman (2004, 2008) regarding firm heterogeneity and find that the data are consistent with these predictions.27
Bernard et al. (2012) use even more disaggregated US import data. Observing trade patterns at the level of the firm and product, the authors provide finer detail on the structure of intrafirm versus arm’s length trade. In particular, they create an index of a product’s contractability based on the importance of the wholesale activity of the firms that import the product. Goods asso- ciated with within-firm distribution presumably are those for which contracting problems are most severe. The authors show that an improvement in an exporting country’s governance raises the probability of related party trade but lowers the share of imports that is intrafirm, whereas goods with lower contractibility are associated with more intrafirm trade. Finally, lower contractability is associated with a greater reduction in the intrafirm trade share as an exporting country’s governance improves. [Other important studies of the property rights approach are by Carluccio & Fally (2012), who use French firm-level data, and by Feenstra &Hanson (2005), who investigate the organization of Chinese processing firms in the context of government-imposed sourcing rules.]
Costinot et al. (2011) consider a variant of intrafirm trade share regression. The authors hypothesize that production technologies across goods differ in the extent to which unexpected problems arise in production. When problems are unlikely to arise, outsourcing to an arm’s length producer is efficient, whereas vertical integration is preferredwhen agents are likely to need to adapt to unforeseen circumstances. The authors test this hypothesis by constructing from occupational data an index of how routine an industry’s production technology is. They find that routineness
26As in Helpman et al. (2004), Antràs & Helpman (2004) have strong sorting implications: The most productive firms in an industry in which vertical integration arises source their intermediates from an owned affiliate, whereas less productive firms buy inputs from arm’s length suppliers. Direct evidence supporting this implication is provided by Kohler & Smolka (2012). 27Yeaple (2006) uses BEA data to explore related questions. The advantage of BEA data relative to the customs data is that they allow the researcher the opportunity to treat US affiliates’ exports to their US parents separately from trade between foreign affiliates operating in the United States and their foreign parents.
212 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
strongly predicts lower intrafirm trade shares and that the inclusion of this variable causes other industry characteristics (such as skill intensity and intermediation indexes) to become statistically insignificant.
The models discussed above are highly stylized in that only two agents contribute value to a final product. Modern value chains may involve hundreds of inputs produced by many distinct firms or divisions of a firm. Antràs & Chor (2012) tackle the question of how vertical integration versus outsourcing decisions are made when stages of production occur sequentially. They show how the prevalence of vertical integration along the value chain depends critically on the degree of the substitutability of final goods relative to the substitutability of investments made by individual agents along the value chain. In deciding whether to vertically integrate or outsource a particular activity, firms must take into account the share of value added that they obtain, the incentives that this creates for the individual undertaking that activity, and the manner in which the decision affects the incentives of other agents producing inputs further down the value-added chain. The authors show that when demand for the final output is very inelastic, firms should integrate only the final stages of production and outsource those upstream, whereas the opposite occurs when demand for the final good is relatively elastic. The authors then test these implications using intrafirm trade data in which a good’s relative downstreamness (measured using input-output techniques) interacts with the elasticity of substitution of final-goods-producing industries.28
Finally, there are numerous papers that consider other features of moral hazard. We focus on thework ofAntràs et al. (2009) because it combines theory and empirics.29 These authors expand the scopeof the analysis to link these issues to the financial structure of themultinational firm.They consider an application of Holmstrom & Tirole (1997) in which there are three types of agents. There is an innovator who had developed a technology, a local entrepreneur who can implement the project, and a continuum of potential local investors with low opportunity cost of funds who can contribute to the funding of the project. The project requires funding that is conditional on the scale of the project, and this funding can be contributed in part or in full by any of the agents. Two related agency problems arise. First, the local entrepreneur can misbehave in ways that lower the expected earnings of the project. Second, the innovator, who can monitor the manager and reduce the benefits of shirking, can only do so at a cost and so needs incentives to monitor. In general, contracts cannot restrain the opportunistic behavior of either the local entrepreneur or the innovator, but local investor protections can partially restrain opportunistic local entrepreneurs. In countries with weak investor protections, arm’s length contracts will be discouraged, multi- national affiliates will rely more heavily on their parent firms for funding, parent firms will take larger equity stakes in their affiliates, and the size of affiliates will be stunted relative to that of countries with stronger investor protections. The authors find empirical support for these impli- cations using US data on licensing and royalty income and multinational production data.
6. CONCLUSION
There is now a large catalog of models of the multinational firm that have been cleverly designed to explain various stylized facts. The list of factors relevant to understanding multinational production is long: increasing returns, contracting frictions, comparative advantage, trade and
28Garetto (2013) develops a very different model in which the elasticity of substitution between intermediates predicts the extent of related party trade in total trade. In her model, input sourcing firms vertically integrate in response to market power on the part of unaffiliated suppliers. Her paper is unique in providing a calibration of the gains from foreign input sourcing. 29Grossman&Helpman (2004) present an example of a pure theory paper in this literature. They investigate agency problems within the firm and when it is better to avoid these problems by outsourcing production.
213www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
communication costs, et cetera. Distinguishing empirically between these factors is difficult be- cause models are rarely nested, data are largely inadequate, and identification even in an ideal setting is tricky.
Even interpreting stylized facts is still difficult. Why is it that most multinational production is north-north rather than north-south? Is this because multinational firms can find many existing facilities to acquire in developed countries through a cross-border acquisition, or is it because developing countries have a comparative advantage in production activities that do not require ownership? If there is little multinational production in a given location, is this because it is an unattractive production location or because it is easy to write contracts in that country so that ownership is unnecessary?
One area in which there has been notable progress in the literature in the past 10 years is the trend toward developing multicountry models in which geography can play a central role. For these models to be tractable, strong simplifying assumptions must be made, and it is not clear how much violence these assumptions do to reality. A particularly nagging question is the role of cross-country dependencies driven by the nature of vertical production chains. We have every reason to believe that the nature of production chains is important, but there is little guidance from the empirical literature as how to treat them in aggregative models.
A further complicating factor is the lack of data on arm’s length contracting. The scope of a firm’s global operations is limited by the foreign affiliates that it reports, but firms may be highly integrated with and dependent on unaffiliated contractors. To date, most of the empirical literature on the boundaries of the multinational firm has been limited to documenting correlations between industry characteristics and the scale of multinational production. This is fine as far as it goes, but it does little to help us to understand how substitutable arm’s length contracts and vertical in- tegration are (see Garetto 2013 for an example of early work in this dimension). Integrating these mechanisms into quantifiable multicountry general equilibrium models would be a valuable direction for further research.
DISCLOSURE STATEMENT
The author is not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.
ACKNOWLEDGMENTS
I thank Pol Antràs for much help in improving earlier versions of this article.
LITERATURE CITED
AcharyaV, ShinHS,YorulmazerT. 2011.Fire sale FDI. Unpublished document, Fed.Reserve BankNewYork Alfaro A, Charlton A. 2009. Intra-industry foreign direct investment. Am. Econ. Rev. 99:2096–119 Anderson T. 2011. U.S. affiliates of foreign companies: operations in 2009. Surv. Curr. Bus. 91(8):211–26 Antràs P. 2003. Firms, contracts, and trade structure. Q. J. Econ. 118:1375–418 Antràs P. 2013. Grossman-Hart (1986) goes global: incomplete contracts, property rights, and the in-
ternational organization of production. J. Law Econ. Organ. In press Antràs P, Chor D. 2012. Organizing the global value chain. Unpublished manuscript, Harvard Univ.,
Cambridge, MA Antràs P, Desai M, Foley C. 2009. Multinational firms, FDI flows, and imperfect capital markets.Q. J. Econ.
124:1171–219 Antràs P, Helpman E. 2004. Global sourcing. J. Polit. Econ. 112:552–80
214 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Antràs P, Helpman E. 2008. Contractual frictions and global sourcing. In The Organization of Firms in a Global Economy, ed. E Helpman, D Marin, T Verdier, pp. 9–54. Cambridge, MA: Harvard Univ. Press
Antràs P, Rossi-Hansberg E. 2009. Organizations and trade. Annu. Rev. Econ. 1:43–64 Arkolakis C, Ramondo N, Rodriguez-Clare A, Yeaple SR. 2012. Innovation and production in the global
economy. Unpublished manuscript, Pennsylvania State Univ., University Park Arnold F, Javorcik B. 2009. Gifted kids or pushy parents? Foreign direct investment and plant productivity
in Indonesia. J. Int. Econ. 79:42–53 Baldwin R, Ottaviano G. 2001. Multiproduct multinationals and reciprocal FDI dumping. J. Int. Econ. 54:429–48 BaldwinR,VenablesA. 2010. Spiders and snakes: offshoring and agglomeration in the global economy.Work.
Pap. 8163, Cent. Econ. Policy Res., London Barefoot K, Mataloni R. 2011. Operations of U.S. multinational companies in the United States and abroad:
preliminary results from the 2009 Benchmark Survey. Surv. Curr. Bus. 91(11):29–47 Bernard A, Jensen J, Redding S, Schott P. 2009. Importers, exporters and multinationals: a portrait of firms
in the U.S. that trade goods. In Producer Dynamics: New Evidence from Micro Data, ed. T Dunne, JB Jensen, MJ Roberts, pp. 513–52. Chicago: Univ. Chicago Press
Bernard A, Jensen J, Redding S, Schott P. 2010. Intra-firm trade and product contractibility. Am. Econ. Rev. 100:444–48
Bernard A, Jensen J, Redding S, Schott P. 2012. The empirics of firm heterogeneity and international trade. Annu. Rev. Econ. 4:283–313
Bilir K. 2011. Patent laws, product lifecycle lengths, and the global sourcing decisions of U.S. multinationals. Unpublished manuscript, Univ. Wisconsin–Madison
Blonigen B. 1997. Firm-specific assets and the link between exchange rates and foreign direct investment.Am. Econ. Rev. 87:447–65
Blonigen B. 2001. In search of substitution between foreign production and exports. J. Int. Econ. 53:81–104 Blonigen B, Davies R, Waddell G, Naughton H. 2007. FDI in space: spatial autoregressive relationships in
foreign direct investment. Eur. Econ. Rev. 51:1303–25 Brainard SL. 1993. A simple theory of multinational corporations and trade with a trade-off between
proximity and concentration. NBER Work. Pap. 4269 Brainard SL. 1997. An empirical assessment of the proximity-concentration tradeoff between multinational
sales and trade. Am. Econ. Rev. 86:520–44 Buch C, Kleinart J, Lipponer A, Toubal F. 2005. Determinants and effects of foreign direct investment:
evidence from German firm-level data. Econ. Policy 20:52–110 Carluccio J, Fally T. 2012. Global sourcing under imperfect capital markets. Rev. Econ. Stat. 94:740–63 Carr DL, Markusen JR, Maskus KE. 2001. Estimating the knowledge-capital model of the multinational
enterprise. Am. Econ. Rev. 91:693–708 Caves RE. 2007.Multinational Enterprise and Economic Analysis. Cambridge, MA: Cambridge Univ. Press.
3rd ed. ChenMX,MooreM. 2010. Location decision of heterogeneous multinational firms. J. Int. Econ. 80:188–99 Coase RH. 1937. The nature of the firm. Economica 4:386–405 Costinot A, Oldenski L, Rauch J. 2011. Adaptation and the boundary of multinational firms.Rev. Econ. Stat.
93:298–308 DesaiM, FoleyCF,Hines J. 2004a. Amultinational perspective on capital structure choice and internal capital
markets. J. Finance 59:2451–88 Desai M, Foley CF, Hines J. 2004b. The cost of shared ownership: evidence from international joint ventures.
J. Financ. Econ. 73:323–74 Dunning JH. 1981. International Production and the Multinational Enterprise. London: Allen & Unwin Eaton J, Kortum S. 2002. Technology, geography, and trade. J. Polit. Econ. 70:1741–79 Ekholm K, Forslid R, Markusen J. 2007. Export-platform foreign direct investment. J. Eur. Econ. Assoc.
5:776–95 Ethier W, Markusen J. 1996. Multinational firms, technology diffusion, and trade. J. Int. Econ. 41:1–28 Fajgelbaum P, Grossman G, Helpman E. 2011. A Linder hypothesis for foreign direct investment. NBER
Work. Pap. 17550
215www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Feenstra R, Hanson G. 2005. Ownership and control in outsourcing to China: estimating the property-rights theory of the firm. Q. J. Econ. 120:729–62
Feinberg S, Keane M. 2001. U.S.-Canadian trade liberalization and MNC production location. Rev. Econ. Stat. 83:118–32
Feinberg S, Keane M. 2006. Account for the growth of MNC-based trade using a structural model of U.S. MNC. Am. Econ. Rev. 96:1515–58
Fillat J, Garetto S. 2012. Risk, returns, and multinational production. Unpublished manuscript, Boston Univ. Garetto S. 2013. Input sourcing and multinational production. Am. Econ. J. Macroecon. 5:118–51 Girma S, Gorg H, Strobl E. 2004. Exports, international investment, and plant performance: evidence from
a non-parametric test. Econ. Lett. 83:317–24 Gordon RH, Bovenberg A. 1996. Why is capital so immobile internationally? Possible explanations and
implications for capital income taxation. Am. Econ. Rev. 86:1057–75 Grossman SJ, Hart O. 1986. The costs and benefits of ownership. J. Polit. Econ. 94:691–719 Grossman G, Helpman E. 2004. Managerial incentives and the international organization of production.
J. Int. Econ. 63:237–62 GrossmanG,HelpmanE, SzeidlA. 2006.Optimal integration strategies for themultinational firm. J. Int. Econ.
70:216–38 Guadalupe M, Kuzmina O, Thomas C. 2012. Innovation and foreign ownership. Am. Econ. Rev. 102:3594–627 Hanson G, Mataloni R, Slaughter M. 2005. Vertical production networks in multinational firms. Rev. Econ.
Stat. 87:667–78 HeadK,Mayer T. 2004.Market potential and the location of Japanese investment in Europe.Rev. Econ. Stat.
86:959–72 Head K, Ries J. 2008. FDI as an outcome of the market for corporate control. J. Int. Econ. 74:2–20 Helpman E. 1984. A simple theory of trade with multinational corporations. J. Polit. Econ. 92:451–71 Helpman E. 1985. Multinational corporations and trade structure. Rev. Econ. Stud. 52:443–57 Helpman E, Melitz M, Yeaple SR. 2004. Exports versus FDI with heterogeneous firms. Am. Econ. Rev.
94:300–16 Holmstrom B, Tirole J. 1997. Financial intermediation, loanable funds, and the real sector. Q. J. Econ.
112:663–91 Horn H, Persson L. 2001. The equilibrium ownership of an international oligopoly. J. Int. Econ. 53:307–33 Horst T. 1972. The industrial composition of U.S. exports and subsidiary sales to the Canadian market. Am.
Econ. Rev. 62:37–45 Horstmann I,Markusen J. 1987. Strategic investments and the development of multinationals. Int. Econ. Rev.
28:109–21 Horstmann I, Markusen J. 1992. Endogenous market structures in international trade (natura facit saltum).
J. Int. Econ. 32:109–29 Irarrazabal A, Moxnes A, Opramolla L. 2010. The margins of multinational production and the role of
intra-firm trade. Unpublished manuscript, Dartmouth Coll., Hanover, NH KellerW. 2010. International trade, foreign direct investment, and technology spillovers. InHandbook of the
Economics of Innovation, Vol. 2, ed. B Hall, N Rosenberg, pp. 793–829. Amsterdam: North-Holland Keller W, Yeaple SR. 2013. The gravity of knowledge. Am. Econ. Rev. In press Klein M, Peek J, Rosengren E. 2002. Troubled banks, impaired foreign direct investment: the role of relative
access to credit. Am. Econ. Rev. 92:664–82 Kohler WK, Smolka M. 2012. Global sourcing decisions and firm productivity: evidence from Spanish
firm-level data. In Quantitative Analysis of Newly Evolving Patterns of International Trade: Frag- mentation, Offshoring of Activities, and Vertical Intra-Industry Trade, ed. RM Stern, pp. 139–89. World Sci. Stud. Int. Econ. Singapore: World Sci.
Krugman PR. 1980. Scale economies, product differentiation, and the pattern of trade. Am. Econ. Rev. 70:950–59
Lipsey R, Sjöholm F. 2011. South-south FDI and development in East Asia. Asian Dev. Rev. 28:11–31 Manova K, Wei S, Zhang Z. 2011. Firm exports and multinational activity under credit constraints. NBER
Work. Pap. 16905
216 Yeaple
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Mansfield E, Romeo A. 1980. Technology transfer to overseas subsidiaries by U.S. based firms. Q. J. Econ. 95:737–50
Markusen J. 1984. Multinationals, multi-plant economies, and the gains from trade. J. Int. Econ. 16:205–26 Markusen J. 1995. The boundaries of multinational enterprises and the theory of international trade. J. Econ.
Perspect. 9(2):169–89 Markusen J. 2002. Multinational Firms and the Theory of International Trade. Cambridge, MA: MIT Press Markusen J, Venables A. 1998. Multinational firms and the new trade theory. J. Int. Econ. 46:183–203 Markusen J, Venables A. 2000. The theory of the endowment, intra-industry and multinational trade. J. Int.
Econ. 52:209–35 MayerT,OttavianoG. 2007.Thehappy few: the internationalisationof European firms.Rep., Bruegel Bluepr.
Ser. 12, Brussels Melitz M. 2003. The impact of trade on intra-industry reallocations and aggregate industry productivity.
Econometrica 71:1695–725 Navaratti G, Venables A. 2004. Multinational Firms in the World Economy. Princeton, NJ: Princeton Univ.
Press Neary P. 2007. Cross-border mergers as instruments of comparative advantage.Rev. Econ. Stud. 74:1229–57 Neary P, Mrázová M. 2011. Selection effects with heterogeneous firms. Discuss. Pap., Oxford Univ. Nocke V, Yeaple SR. 2007. Cross-border mergers and acquisitions versus greenfield foreign direct investment:
the role of firm heterogeneity. J. Int. Econ. 72:336–65 Nocke V, Yeaple SR. 2008. An assignment theory of foreign direct investment. Rev. Econ. Stud. 75:529–57 Nunn N, Trefler D. 2012. Incomplete contracts and the boundaries of the multinational firm. Unpublished
document, Univ. Toronto Org. Econ. Coop. Dev. (OECD). 2007. Measuring Globalisation: Activities of Multinationals, Vol.
1: Manufacturing, 2000–2004. Paris: OECD Ramondo N, Rappoport V, Ruhl K. 2012. Horizontal versus vertical FDI: revisiting evidence about U.S.
multinationals. Unpublished document, Arizona State Univ., Tempe Spencer BJ. 2005. International outsourcing and incomplete contracts. Can. J. Econ. 38:1107–35 Swedenborg B. 1979. The Multinational Operations of Swedish Firms: An Analysis of the Determinants and
Effects. Stockholm: Ind. Inst. Econ. Soc. Res. Tintelnot F. 2012. Global production with export platforms. Unpublished document, Pennsylvania State
Univ., University Park United Nations Conf. Trade Dev. (UNCTAD). 2011. World Investment Report 2011. New York: United
Nations Yeaple SR. 2003a. The complex integration strategies of multinationals and cross country dependencies in
the structure of foreign direct investment. J. Int. Econ. 60:293–314 Yeaple SR. 2003b. The role of skill endowments in the structure of U.S. outward foreign direct investment.
Rev. Econ. Stat. 85:293–314 Yeaple SR. 2006. Offshoring, foreign direct investment and the structure of U.S. trade. J. Eur. Econ. Assoc.
4:602–11 Yeaple SR. 2008. Firm heterogeneity, intra-firm trade, and the role of central locations. In The Organization
of Firms in the Global Economy, ed. E Helpman, D Marin, pp. 200–30. Cambridge, MA: Harvard Univ. Press
Yeaple SR. 2009. Firm heterogeneity and the structure of U.S. multinational activity: an empirical analysis. J. Int. Econ. 78:206–15
Zhang K, Markusen J. 1999. Vertical multinationals and host country characteristics. J. Dev. Econ. 59:233–52
217www.annualreviews.org � The Multinational Firm
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Annual Review of
Economics
Volume 5, 2013Contents
Early-Life Health and Adult Circumstance in Developing Countries Janet Currie and Tom Vogl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Fetal Origins and Parental Responses Douglas Almond and Bhashkar Mazumder . . . . . . . . . . . . . . . . . . . . . . . . 37
Quantile Models with Endogeneity V. Chernozhukov and C. Hansen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Deterrence: A Review of the Evidence by a Criminologist for Economists Daniel S. Nagin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Econometric Analysis of Games with Multiple Equilibria Áureo de Paula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Price Rigidity: Microeconomic Evidence and Macroeconomic Implications Emi Nakamura and Jón Steinsson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Immigration and Production Technology Ethan Lewis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
The Multinational Firm Stephen Ross Yeaple . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Heterogeneity in the Dynamics of Labor Earnings Martin Browning and Mette Ejrnæs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Empirical Research on Sovereign Debt and Default Michael Tomz and Mark L.J. Wright . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Measuring Inflation Expectations Olivier Armantier, Wändi Bruine de Bruin, Simon Potter, Giorgio Topa, Wilbert van der Klaauw, and Basit Zafar . . . . . . . . . . . . . . . . . . . . . . . . 273
Macroeconomic Analysis Without the Rational Expectations Hypothesis Michael Woodford . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
v
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
Financial Literacy, Financial Education, and Economic Outcomes Justine S. Hastings, Brigitte C. Madrian, and William L. Skimmyhorn . . . 347
The Great Trade Collapse Rudolfs Bems, Robert C. Johnson, and Kei-Mu Yi . . . . . . . . . . . . . . . . . . 375
Biological Measures of Economic History Richard H. Steckel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
Goals, Methods, and Progress in Neuroeconomics Colin F. Camerer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
Nonparametric Identification in Structural Economic Models Rosa L. Matzkin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
Microcredit Under the Microscope: What Have We Learned in the Past Two Decades, and What Do We Need to Know? Abhijit Vinayak Banerjee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
Trust and Growth Yann Algan and Pierre Cahuc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
Indexes
Cumulative Index of Contributing Authors, Volumes 1–5 . . . . . . . . . . . . . . 551 Cumulative Index of Article Titles, Volumes 1–5 . . . . . . . . . . . . . . . . . . . . . 554
Errata
An online log of corrections toAnnual Review of Economics articles may be found at http://econ.annualreviews.org
vi Contents
A nn
u. R
ev . E
co n.
2 01
3. 5:
19 3-
21 7.
D ow
nl oa
de d
fr om
w w
w .a
nn ua
lr ev
ie w
s. or
g A
cc es
s pr
ov id
ed b
y U
ni ve
rs ity
o f
Il lin
oi s
- U
rb an
a C
ha m
pa ig
n on
0 8/
15 /1
9. F
or p
er so
na l u
se o
nl y.
- Annual Reviews Online
- Search Annual Reviews
- Annual Review of Economics Online
- Most Downloaded Economics Reviews
- Most Cited Economics Reviews
- Annual Review of Economics Errata
- View Current Editorial Committee
- All Articles in the Annual Review of Economics, Vol. 5
- Early-Life Health and Adult Circumstance in Developing Countries
- Fetal Origins and Parental Responses
- Quantile Models with Endogeneity
- Deterrence: A Review of the Evidence by a Criminologist for Economists
- Econometric Analysis of Games with Multiple Equilibria
- Price Rigidity: Microeconomic Evidence and Macroeconomic Implications
- Immigration and Production Technology
- The Multinational Firm
- Heterogeneity in the Dynamics of Labor Earnings
- Empirical Research on Sovereign Debt and Default
- Measuring Inflation Expectations
- Macroeconomic Analysis Without the Rational Expectations Hypothesis
- Financial Literacy, Financial Education, and Economic Outcomes
- The Great Trade Collapse
- Biological Measures of Economic History
- Goals, Methods, and Progress in Neuroeconomics
- Nonparametric Identification in Structural Economic Models
- Microcredit Under the Microscope: What Have We Learned in the Past TwoDecades, and What Do We Need to Know?
- Trust and Growth
- ar:
- logo:
