![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
Quantitative assessments of ‘country risk’ produced by private rating agencies have become influential among multinational firms considering investment in the developing world. This article considers whether developing countries exhibiting more labour rights violations are associated with higher risk ratings. It draws from existing literature on the determinants of foreign direct investment (FDI) and the relationship between labour rights and FDI to illuminate the ways in which labour repression may indicate increased risk for potential investors. It implements a set of time series of cross-sections models for 95 developing countries, from 1985 to 2003. These models present predictors of a long-established risk index, plus an additional latent risk measure created through factor analysis of three separate and distinct risk ratings. Results indicate a consistent association between increased labour rights violations and increased risk ratings. The article extends the analysis to highlight the heterogeneous nature of FDI and possible sector-specific relationships between labour rights and risk ratings. The paper argues that the economic profile of a country may condition the relationship between labour rights and risk ratings.
ACKNOWLEDGEMENTS
I wish to thank representatives from the Political Risk Services Group, Euromoney, and Institutional Investor for providing data and insight on their risk ratings. I also thank Layna Mosley for generously providing labour rights data, and the participants at the workshop on multinationals and labour rights at the University of North Carolina and Duke University in September 2006. I thank four anonymous reviewers for their comments; and Amanda Henley for research assistance. All errors and omissions in this work are my own.
Notes
1. For a good summary of previous studies on the economic determinants of FDI, see Chakrabarti (Citation2001).
3. For an explanation of the methodology used to construct the risk ratings, see http://www.prsgroup.com/ICRG_Methodology.aspx#PolRiskRating/.
4. There are a number of other risk measures that might have been utilized in this study. However, many of the alternate risk measures, such as Economist Intelligence Unit (EIU), began their publicly available risk measures only in the 1990s. Other organizations, such as Control Risks Group (CRG), do not make their risk rankings available to non-clients.
5. This follows the approach used by Oetzel et al. (2001) in their study of the reliability of various risk indicators.
6. For an elaboration on the multi-step methodology used to construct the Euromoney risk ratings, see http://www.euromoney.com/Poll/10683/PollsAndAwards/Country-Risk.html?Type=Polls&PageID=10683/. For Institutional Investor, for a description of the risk calculation methodology, see http://www.iimagazinerankings.com/countrycredit/methodology.asp/. There have been some semi-successful attempts to replicate the risk measures using commonly available socioeconomic data (Cossett and Roy, Citation1991).
7. For the coding template, along with the weighting structure used to measure the severity of violations, see Mosley and Uno (Citation2007).
8. Thanks to Andrew Schrank for the suggestion that I include this variable.
9. Unlike Vaaler et al. (Citation2005), I do not combine the centrist and rightist parties in the ideology measure, but maintain the three-part distinction in the original data. I also do not lag the political variables one year, as do Archer et al. (Citation2007). I contend that many of these political variables (e.g. the election year variable) will have more of an immediate impact than the economic predictors, which are often subject to reporting delays. This is consistent with the approach used by Biglaiser and Brown (Citation2005).
10. The presence of unit effects in the data was confirmed by a test of residuals after a simple OLS regression. An analysis of variance (ANOVA) test of the residuals revealed systematic differences of means across countries (significant at the 0.01 level).
11. See, in particular, the exchange between Green et al. (Citation2001) and Beck and Katz (Citation2001).
12. Only one instance of obvious measurement error was identified and recoded as a missing observation.
13. This test was implemented using the xtserial command, developed by Drukker (Citation2003). The test returned a statistic of 262.857, significant at the 0.01 level, rejecting the null hypothesis of no serial correlation.
14. If the Mosley and Uno indicator is moved from the sample minimum (zero) to the sample maximum (37), the associated risk rating increase would be 1.85. This is larger than the increase associated with the presence of armed conflict in the country (1.25). While countries in the sample do not move from the minimum to the maximum in practice, labour rights violations often appear in groups. Therefore, the accumulated influence of labour rights violations on the risk rating may be substantial.
15. The changed signs on some of the coefficients in this model most likely derive from the much reduced sample size of model (4). Some of the coefficients that were significant in the baseline model lose their significance in model (4). The coefficients associated with the inflation measure and the trade measure reverse their sign and gain significance. All of the textile data points occur in the 1990s and later, which may reduce the impact of inflationary environments of the 1980s. Also, many of the least developed countries in the main sample drop out of model (4), because they do not have textile data available. The limited time period and country coverage for the textile data introduce selection effects. However, it is noteworthy that the labour rights violations indicator remains significant.
16. My thanks to an anonymous reviewer for pointing this out.