![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
The industry standard for studying multinational corporations (MNCs) has been to evaluate patterns in aggregate country-level measures of foreign direct investment (FDI). Though certainly related, these data are at best a proxy for the actual commercial and productive activities of multinationals that most political scientists purport to be interested in. Simply put, this is a very indirect way of testing theories about the sociopolitical and economic factors that motivate MNCs’ choice of host countries. This article introduces a new firm-level data set designed to get around this problem by permitting more direct analysis of multinationals’ foreign operations. It then revisits the relationship between regime type and direct investment, finding evidence that MNCs are more likely to establish new subsidiaries in democracies than in nondemocracies. However, further analysis reveals that the strength of this relationship varies by context. Specifically, MNCs rely on regime type as an indicator of political risk when they lack an existing relationship with the host state. In addition, those operating in extractive industries are generally less responsive to political institutions than those operating in manufacturing or services. These results suggest that firm- and sector-specific factors deserve greater consideration than they have been given in the existing literature.
KEYWORDS:
Acknowledgments
I would like to thank Ben Fordham, David Cingranelli, Katja Kleinberg, the editors of International Interactions, and the anonymous reviewers for their helpful comments and suggestions.
Supplementary material
A supplemental appendix to this article (containing Tables A1 through A15 and Figures A1 and A2) is available on the publisher’s Web site at http://dx.doi.org/10.1080/03050629.2015.1065698.
Notes
1 This critique is directed mostly at the mainstream political science literature. There is a large body of work in international business that has utilized firm-level data. Some studies have also surveyed corporate executives as an alternative to analyzing FDI data (Biglaiser and Staats Citation2010; Rolfe, Ricks, Pointer, and McCarthy Citation1993).
2 Blonigen and Wang (Citation2005) demonstrate the importance of distinguishing between developed and developing countries when analyzing patterns in FDI. There is a long tradition of FDI between advanced economies that has been studied extensively, and the volume of investment activity dwarfs that which we see directed toward emerging markets. Similarly, investors seem more sensitive to certain political stimuli when doing business with developing states (Barry, Clay, and Flynn Citation2013). This study thus focuses on developing countries. However, the reported findings are robust to the inclusion of the highly developed countries.
3 Kerner (Citation2014) recommends using data on fixed capital expenditures instead, which isolate the foreign commercial activity of MNCs. This is certainly a step in the right direction, but these data are limited to US outward investments and aggregated across firms and industries.
4 This does not mean that FDI data are ill-suited for answering all questions. Analyzing FDI flows is perfectly appropriate for research concerning cross-border capital flows in general, for example. Rather, the point is that these data are less appropriate for studies more narrowly concerned with multinationals’ commercial activities in foreign markets.
5 These generally include the World Bank, United Nations Conference on Trade Development (UNCTAD), International Money Fund (IMF), Organization for Economic Cooperation and Development (OECD), and the Bureau of Economic Analysis (BEA).
6 For example, it entails an implicit bias toward capital-intensive industries that involve larger financial transactions. It is not necessarily the case that these sorts of foreign investments are more important to the host market than those in labor-intensive industries that generate smaller financial transfers but that create more jobs locally.
7 See, for example, Henisz’s work on how political hazards affect firms’ mode of entry into foreign markets (2000).
8 See the Web appendix for all the companies included in the sample.
9 The sampling procedure was semi-random, as firms that made the list more frequently had a higher baseline likelihood of being selected. I had no knowledge of the locations in which the firms operated prior to selection, as sampling preceded data collection. Relevant information was available for all selected firms, meaning that data limitations did not determine which MNCs were sampled. However, some firms enter the sample later than others. For example, while I have data on most companies starting in 1993, data were only available for Statoil starting in 1994. These slight variations in the availability of historical data do not seem to be systematic, and there is little reason to suspect that they bias the analyses in any significant way.
10 Indeed, size and experience are firm-level factors that may condition the effects of political risk (or other causal variables) on a firm’s investment decisions (Henisz and Delios Citation2001). Although this particular feature is absent from the present study, I intend to gather data on a larger and more diverse sample of companies in the future.
11 LexisNexis (Citation2012) relies on its own extensive collection of financial records and reports and communicates regularly with both parent and subsidiary enterprises to ensure that the information provided is both accurate and up to date. Corporate Affiliations is a proprietary service designed for business-sector research and networking and is thus considered a reliable source. See http://www.corporateaffiliations.com/nonsub/aboutUs.asp for more details.
12 I was able to code data on 77 MNCs before my subscription to Corporate Affiliations terminated. The data set will be expanded and updated in the future.
13 A total of 90% are majority owned (that is, >50% ownership stake) by the parent MNC, with the modal category being wholly owned.
14 The “developing world” includes all countries except Andorra, Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Liechtenstein, Luxembourg, Malta, Monaco, Netherlands, New Zealand, Norway, Portugal, San Marino, South Korea, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Data were also collected on foreign subsidiaries located in each of these developed countries except the United States, but they are excluded from the following analyses.
15 The exit of a subsidiary ID code from the sample does not necessarily imply the exit of the subsidiary itself. Likewise, the entry of a new subsidiary ID code does not necessarily imply the entry of a new subsidiary. As such, measuring new investment or divestment in this way would be subject to error.
16 Because some companies have two registered home states, and because several firms’ primary industries change over this span, the sum of the home-country and main-sector figures exceeds 77.
17 These figures do not include new subsidiaries opened in semi- or nonautonomous territories, like Hong Kong, Puerto Rico, and the Cayman Islands.
18 Subsidiaries with SIC codes falling between 0111 and 1499 are identified as “extracting”: This includes agriculture, forestry, fishing, and mining. Those with SIC codes falling between 2000 and 3999 are identified as “manufacturing.” SIC codes falling between 1500 and 1799 (construction) and those falling between 4000 and 9999 (utilities, wholesale, retail, finance, services, and public administration) are coded as “services.” Many subsidiaries operate in multiple industries. For example, ExxonMobil’s subsidiary “Esso Malaysia Berhad” mined petroleum and gas, manufactured petrol fertilizers, and wholesaled petroleum and petrol-based products in Malaysia. It is coded as operating in all three sectors accordingly.
19 FDI stock and flows data are from UNCTAD.
20 Firm-level studies in the international business literature, for example, consider not just investment but mode and sequence of entry (Delios and Henisz Citation2003; Henisz Citation2000; Kesternich and Schnitzer Citation2010) and firm performance (Delios and Beamish Citation2005; Desai, Foley and Hines Citation2008).These represent potentially fruitful avenues for future research in political science.
21 Forced divestment, for example, is more frequent in countries that lack checks on executive power (Li Citation2009).
22 Other studies suggest that it may not be “democracy” per se, so much as particular institutions more often found in democracies—numerous veto players (Choi and Samy Citation2008), constraints on the executive (Büthe and Milner Citation2008; Jensen Citation2008), or an independent court system (Bigliaser and Staats Citation2010)—that are especially important to MNCs. More recent research considering the “demand” for FDI also suggests that, because labor benefits from MNC entry (Pandya Citation2010) and because democrats are more accountable to the broad working class, democracies are more likely to reduce barriers to foreign investment (Pandya Citation2014). This suggests an alternative mechanism by which regime type can influence patterns in FDI. Although it goes beyond the scope of this article, future research should build on this insight in order to clarify the mechanism that is driving this empirical regularity.
23 In the vast majority of cases, zero merely reflects a continuation of the status quo. In some instances, however, the firm is divesting from the host state. Though divestment is itself an interesting outcome worthy of study, it goes beyond the scope of this article. While cases of divestment are subsumed in the zero category in the models estimated here, the results are robust to dropping these cases from the analysis.
24 This statistic is based on the binary indicator of expansion by company i into country j during year t. In some cases, the firm in question may actually be opening multiple new subsidiaries in the host state during a given year, implying an even higher rate of investment activity than is reported.
25 The total across these three sectors is greater than the total number of new ventures in the aggregate because some subsidiaries operate in multiple sectors.
26 Indeed, Henisz and Delios (Citation2001) note that MNCs also look at other firms’ investment activities as a means of gauging risk in a host market, which is conceptually related to the object of study here.
27 As most MNCs establish new subsidiaries in only one or two developing countries per year, positive outcomes are recorded in only about 1% of cases. Though this is an artifact of the firm-level structure of the data, it is possible that some of the company-country dyads included in the sample are borderline irrelevant. To account for this, I also estimated bivariate probit models with partial observability (Poirier Citation1980). This approach permits simultaneous estimation of a “relevance” equation alongside the model of new investment (Xiang Citation2010). The results are robust to this specification and are included in the Web appendix.
28 Increasing these three variables—GDP/capita, population, and new ventures—to their specified levels simultaneously increases the predicted probability of investment to 0.032, which is 2,500% higher than the baseline probability.
29 These findings are robust to controlling directly for property rights (Gwartney, Hall, and Lawson Citation2010; see Li and Resnick Citation2003).
30 Regime duration demonstrates a very strong positive correlation with the total number of existing subsidiaries present in a country. What these estimates suggest is that new investments have been drawn more frequently into the states that emerged over this time span, such as those in Eastern Europe (Bevan and Estrin Citation2004).
31 While it is important to control for this spatial dependence, some caution is warranted because this variable is endogenously related to the outcome. The key findings are robust to excluding this variable from the equation.
32 Confidence intervals are not shown here, to preserve clarity. Graphs with the confidence intervals are included in the Web appendix.
33 The hypotheses concern the general relationship between regime type and direct investment in each industry. Models including interactions between firm presence and Polity are included in the Web appendix. The performance of the variables across sectors is similar to the noninteractive models shown here.
34 The magnitudes of these estimated effects are reported in the Web appendix.