Democracy and foreign direct investment at the industry level: evidence for US multinationals

Abstract

Theories of multinational enterprises emphasize that foreign direct investment (FDI) is undertaken in different industries for different reasons, yet studies of the effects of democracy on FDI most commonly use aggregate-level FDI data. This paper evaluates US FDI outflows to 15 industries (eight manufacturing, seven nonmanufacturing) in 54 countries in a linear dynamic panel-data gravity FDI model using a “system” generalized method of moments estimator and three widely used democracy indicators. At the aggregate-level, we estimate a positive effect of democracy on FDI, consistent with most prior studies. At the industry level, we estimate larger positive effects of democracy on FDI for service than manufacturing industries, particularly for finance and insurance and information, and negative effects for mining and oil and gas extraction.

Appendix

1.1 Data sources

Freedom House civil liberties and political rights indexes are available online from Freedom House at: http://www.freedomhouse.org/.

The Polity IV Polity2 index is available online from the Polity IV Project at: http://www.systemicpeace.org/polity/polity4.htm.

US FDI data are available online from the US Bureau of Economic Analysis at: http://www.bea.gov/international/di1usdbal.htm.

Crosswalk tables between SICS and NAICS are available online from the NAICS Association at: http://www.naics.com/.

Data on English as an official language and distances between capital cities are available online from CEPII at: http://www.cepii.fr/anglaisgraph/bdd/distances.htm.

World Development Indicators are available online from the World Bank at: http://data.worldbank.org/data-catalog/world-development-indicators.

However, we downloaded the World Development Indicators in Stata format from the Graduate Institute of International and Development Studies website at: http://graduateinstitute.ch/md4stata/datasets/wdi.html.

All websites for data were accessed in May 2012.

1.2 Technical notes

For US FDI data, the SIC (1987) industry classification was used from 1994 to 1998 and the NAICS (1997/2002) industry classification from 1999 to 2010. For most of our 15 industries, we use data from 1999 to 2010, as we were unable to adequately match SIC with NAICS. But for six industries, we matched SIC and NAICS as follows, using crosswalk tables provided by the NAICS Association. For chemicals and allied products, SIC 28 was matched with NAICS 325, with no difference between the two. For food and kindred products, SIC 20 is matched with NAICS 311, with beverages included in the SIC period but combined with other manufacturing thereafter. For primary and fabricated metals, SIC 33 and 34 are matched with NAICS 331 and 332, for which 93 of 106 NAICS 331 and 332 products correspond. For industrial machinery and equipment, SIC 35 is matched with NAICS 333, for which 58 of 82 NAICS 333 products correspond. For transportation equipment, SIC 37 is matched with NAICS 336, for which 31 of 41 NAICS 336 products correspond. For wholesale trade, SIC 50 and 51 are matched with NAICS 421 and 422, for which 73 of 95 NAICS 421 and 422 products correspond.

Regarding the specification of the FDI model, several additional host-country variables commonly used in FDI models were tested and not found to be statistically significant. These include the number of phone lines per 100 inhabitants (as natural logarithms, where 1 is added to values to retain observations with zero values—a measure of infrastructure), the change in the consumer price index (as percentage change—a measure of inflation), GDP growth (as percentage change based on valuation in constant 2,000 US dollars—a measure of market dynamics) and total FDI stock from all countries (as natural logarithms based on valuation in constant 2,000 US dollars—a measure of agglomeration effects and treated endogenously in the system GMM context). While data for the first three variables come from the World Bank’s World Development Indicators, the last comes from the United Nations Conference on Trade and Development (UNCTAD), and is available online at: http://unctadstat.unctad.org/ReportFolders/reportFolders.aspx?sCS_referer=&sCS_ChosenLang=en.

Following Busse et al. (2011), we set negative FDI values to zero and then added 1 to all FDI values, which the authors refer to as their “preferred FDI variable” (Busse et al. 2011, p. 150). But for regressions using this variable construction, we reject the null hypothesis for the Hansen J test that the instruments are not correlated with the residuals. Following Asiedu and Lien (2011), we used measures of total trade relative to GDP and gross fixed capital formation relative to GDP constructed as percentages, but we also used these constructed as natural logarithms, and this did not substantively change results.

1.3 Countries in the sample (grouped according to World Bank geographical classification)

East Asia and Pacific (7): China, Indonesia, the Republic of Korea, Malaysia, the Philippines, Singapore, Thailand.

Europe and Central Asia (5): the Czech Republic, Hungary, Poland, the Russian Federation, Turkey (data for the first four countries are only available from 1999 to 2010).

Developed (21): Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Luxembourg, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom.

Latin American and Caribbean (13): Argentina, Barbados, Brazil, Chile, Colombia, Costa Rica, the Dominican Republic, Ecuador, Honduras, Mexico, Panama, Peru, Venezuela.

Offshore (1): Bermuda.

Middle East and North Africa (4): Egypt, Israel, Saudi Arabia, the United Arab Emirates.

South Asia (1): India.

Sub-Saharan Africa (2): Nigeria, South Africa.