Labor Rights and Foreign Direct Investment: Is There a Race to the Bottom?

Abstract

As global awareness of human rights continues to increase, the practices and investment decisions of multinational corporations have come under increased scrutiny. This is particularly the case with labor rights. Popular arguments concerning a “race to the bottom” abound, though a growing amount of scholarship portrays a more positive relationship between foreign capital and such rights. Yet empirical studies have yielded mixed results. To clarify the linkages between labor rights and foreign direct investment (FDI), we add to the extant literature in two ways. First, we take into account the reciprocal relationship between FDI and labor rights. Second, as FDI is a heterogeneous phenomenon, we assess these linkages across three different industrial sectors. Our findings show that there is a significant relationship between labor rights and FDI, though its exact nature varies across sectors.

Keywords:

• foreign direct investment
• globalization
• human rights
• labor rights

Acknowledgments

The authors would like to thank the editors and three anonymous reviewers, as well as Susan Aaronson and Indra de Soysa, for their comments on previous versions of this manuscript. We are also grateful for Roxanne Buckman, Jack Busbee, Clint Thompson, and Drew Wagstaff for their research assistance. Any errors are the responsibility of the authors. All data used in this analysis, as well as a Stata do-file for the models shown here, can be accessed at https://umdrive.memphis.edu/rblanton/public/blanton_II_2012_data_labor or http://dvn.iq.harvard.edu/dvn/dv/internationalinteractions.

Notes

¹Though such industries may use technologically-advanced equipment, and require some higher skilled labor, their overall skill requirements are comparatively low, as evidenced by their low rankings in empirically-based classifications of the “technological intensity” (Hatzichronoglou 1997; see also Peneder 2003) of industries.

²For example, the Thai government preceded the privatization of its petroleum services sector by decertifying the union, citing privatization as a change in the “legal status” of the industry which rendered the existing union irrelevant in that sector (ICFTU 2003).

³The biggest limitation on the sample size is data availability on the dependent variable. In all, the average time span in the study was approximately 8 years per country, with availability being greater from the mid-1990s to 2002. Though ideally we would like to have complete data on more countries, the sample of countries being examined here (see Appendix) is a diverse pool that includes developing countries at various economic levels with a range of political systems and labor practices—the full range of variation on key political variables, including labor rights, democracy, and human rights, is represented in this sample. Moreover, the sample means on these variables are not significantly different from the mean scores of developing countries as a whole (that is, though there is some variation between the means the difference does not rise to statistical significance).

⁴As noted by Li (2009), the logged measure is a “promising” way to account for outliers and is less problematic than using a proportional index to model FDI (that is, using FDI as a percentage of GDP as the dependent variable). Some manipulation of the data is necessary to make logging possible. Values of zero and below are reset to 1 (which produces a logged value of 0). This results in a slight truncation of the range of data—there were zero negative values on 2.6% of the observations of total flows, 8.6% in primary sectors, 3.3% in secondary sectors, and 4.9% in service sectors.

⁵There were some instances where only stock data were provided. In these instances, we followed the convention that UNCTAD employs in their World Investment Reports (UNCTAD 2004) and converted the stock measures to flows by taking the differences in stock across year.

⁶Though these rights are conceivably able to be broken down into their individual components, data on the constituent parts of this index is not publicly available. Moreover, Mosley (2011:124–125) reports that there is a substantial correlation between these types of rights, as factor analyses revealed strong factor loadings on a single pattern.

⁷Though the conceivable maximum score would be 76.5, in practice no country severely violates every category of labor rights. The scores of countries in this sample range from 0 to 34.5.

⁸The only other data set on labor rights that has comprehensive coverage across multiple countries for multiple years is the worker rights index contained in the CIRI (Cingranelli-Richards) dataset (Cingranelli and Richards 2006). They provide a nominal scale ranging from 0 to 2, based on the presence of key labor rights, including the right to collectively bargain and the right to strike. However, the Mosley dataset provides a more fully-developed indicator of labor rights in several respects. First, it yields insights into both the legal framework for labor rights as well as the extent to which these rights are respected, whereas the CIRI measure only includes the former. Moreover, the CIRI measure relies on a single source, namely the State Department reports, while the Mosley data uses multiple sources. The nominal scale used by the CIRI data is also problematic in that it is left unclear whether and in what way a country that scores a zero is worse off than a country with a score of one, as it cannot be inferred whether the latter has better labor rights or simply different types of problems that are concentrated in a single aspect of labor rights (Mosley 2011:112). Indeed, though the variables are intended to tap into a common type of rights, the actual correlation between the CIRI and the Mosley indices are very small among developing countries (.20). Another potential measure of labor rights could be the number of strikes which occur in a given country within a year (Robertson and Teitelbaum 2011). However, such a measure is problematic, as the implication of strikes is ambiguous—for example, it is difficult to unambiguously infer whether or not an absence of strikes is indicative of a very satisfied labor force or an effectively repressed one (Mosley 2011).

⁹Though Büthe and Milner (2009) find trade institutions—particular WTO membership and membership in preferential trading arrangements (PTAs)—to be related to FDI, we do not use these measures due to the years covered in our study. In particular, data on PTA membership is only available until 2000. As for WTO membership, given the shorter time span of our study (8 years) there is virtually no variation in WTO membership in the sample of countries assessed here—only three countries became WTO members during years assessed in this study. As such, the WTO variable would be very highly correlated with the fixed-effects measure used in our analysis.

¹⁰Specifically the “POLITY 2” index (Marshall, Jaggers, and Gurr 2007), which is the difference between the combined democracy and autocracy measures across a given polity. The index is rescaled to range from 0 (autocracy) to 21 (fully democratic).

¹¹Though human rights is a multifaceted phenomena, our narrow measurement is advantageous in that it allows us to better distinguish human rights from the economic rights implied by labor standards and the institutional safeguards of democratic regimes. Empirically, the small correlations between labor rights, human rights, and democracy (the highest correlation among these variables is .287 between human rights and democracy) enable us better model the independent impacts of each variable upon FDI.

¹²Data were obtained from their website: http://www.cid.harvard.edu/ciddata/barrolee/appendix_ data_tables.xls. The index is reported in five-year increments, so data in the missing years are filled in with the value from the most recent for which data is available. Though this measure has been refined in subsequent studies (Cohen and Soto 2007; see also Noorbakhsh, Paloni, and Youssef 2001), the newer measures do not cover a sufficient amount of years and/or countries for our purposes.

¹³As the intent of this measure is to represent the overall prevalence of FDI within a given country, FDI stock is measured as the cumulative amount of total FDI within a country as a percentage of GDP. The measure is logged to correct for skewness, and is lagged one year (unless indicated otherwise, all the other control variables are also lagged one year). Data are from UNCTAD: http://www.unctad.org/fdistatistics. Inclusion of this variable also helps to control for overall trends regarding inbound FDI.

¹⁴Market size is simply total GDP (adjusted for PPP). The index for trade openness is total trade as a proportion of GDP, while income is measured as GDP (adjusted for PPP) per capita. All three are logged to account for skewness. Economic growth is the yearly change in real GDP. Data for these economic variables are from the World Development Indicators (World Bank 2009).

¹⁵Data for the exchange rate variable are from Shambaugh (2004); the index is simply a dummy variable indicating whether a country has a fixed exchange rate regime during a given year. The fluctuation variable is the ratio of the value of the exchange rate in a given year (relative to the U.S. dollar) to its mean value over the two previous years (see Globerman and Shapiro 2003:25). Exchange rate data for this measure are from the Penn World Tables 6.2 (Heston, Summers, and Aten 2006).

¹⁶This measure is operationalized as the change in the total labor force over a year (LF_t – LF_t-1) divided by the total population of a country. Data for constructing this index are from the World Development Indicators (World Bank 2009).

¹⁷The ratio is simply the proportion of the total capital stock of the host country) to the size of the labor force. Capital stock data are obtained from the Penn World Tables version 7, while labor force data are from the World Development Indicators (World Bank 2009).

¹⁸The Privatization Database, compiled by the World Bank and the International Finance Corporation, includes all privatization transactions for over $1 million (U.S.). For 1988 to 1999, the data represent the government asking price for the transactions. Data from 2000 to 2002 represent the proceeds gained from privatization sales. Though these are not identical they do provide information into the prevalence of these efforts within a country. The index used here is the privatization transactions as a percentage of GDP, logged to correct for skewness.

¹⁹There are two specific differences. Due to the nature of our FDI indices, we do not incorporate the interaction term of human rights, NGO's and FDI inflows, as the NGO measure does not differentiate between different investment sectors (neither of these variables was statistically significant in the Mosley and Uno (2007) study). Also, our use of a fixed-effects model (which contains a dummy variable for each individual country) makes the use of regional dummies superfluous.

²⁰The trade measure is identical to the one in the FDI model; external debt is operationalized as total external debt as a percentage of GDP. Data for both were obtained from the World Development Indicators (World Bank 2009).

²¹The democracy, economic growth, and income variables are identical to the ones used in the FDI model. Total population data were obtained from the World Development Indicators (World Bank 2009), while the civil war variable was obtained from the PRIO dataset on armed conflict (http://www.prio.no/CSCW/Datasets/Armed-Conflict/).

²²Data for this variable were obtained from Mosley and Uno (2007) who gathered the data from the Human Rights Internet's Master List of Organizations. This variable is logged.

²³The need for the inclusion of the time dummies was also verified by using the “testparm” diagnostic in Stata. Hausman tests, which measure the extent that the error terms are correlated with the regressors, were also run to confirm that fixed effects are necessary. We use the “ivreg2” command in Stata, with the year and country dummies entered manually.

²⁴As there is some evidence that labor rights variables do trend over time (Mosley and Uno 2007; Neumayer and de Soysa 2006), we ran alternate models using yearly dummies. However, diagnostic tests (the “testparm” command in Stata) revealed that yearly dummies did not have a significant impact upon the dependent variable and were thus not necessary in our case. We did, however, run sensitivity analyses that included yearly dummies. The results were largely consistent with the findings reported below—the only difference was that the impact of labor rights on services FDI no longer rose to statistical significance (though the impact of services FDI on labor rights remained negative and significantly related to labor rights).

²⁵As mentioned earlier, though there is some degree of conceptual overlap between labor rights, human rights, democracy, and human capital, we use rather specific measures of each and the actual empirical correlation between the four—as measured in this study—is quite low. Bivariate correlations between the variables range from .08 (between democracy and labor rights) to .31 (democracy and human capital). Moreover, dropping individual variables from the model did not result in raising any of the insignificant variables to statistical significance; specifically, dropping human rights and human capital from the models did not change the lack of significance of the democracy variable in any of the models.

²⁶Though democracy is not our primary variable of interest, our findings do merit some elaboration as they vary from those of some of the literature in this area (Jakobsen and de Soysa 2006; Jensen 2006; Schulz 2007). First, we use a different dependent variable and a different sample of countries than some of these studies. Moreover, these studies do not incorporate measures for the other sociopolitical variables used in this study.

²⁷Fixed effects models control for all time-invariant country-specific effects, and consequently tend to obscure the significance of other variables that do not change over time (Beck and Katz 2001). As mentioned earlier, there is reason to suspect that country-specific effects are an important consideration when assessing labor rights in a cross-national manner (Neumayer and de Soysa 2006). Given the relatively short temporal range of this study, as well as the tendency of these economic variables to remain relatively stable over time, their lack of statistical significance is not surprising. However, we did run alternate models to see how these variables performed without the presence of the fixed-effects. Of the economic variables, market size was positively and significantly related to FDI in all four models, while income was negatively and significantly related to FDI in each model. The substantive interpretation of this would be that while within-country changes in overall market size and income are not related to FDI, size and income are significant determinants of variation in FDI inflows across different countries.