Does stakeholder involvement foster democratic legitimacy in international organizations? An empirical assessment of a normative theory

Data and supplemental information necessary to reproduce the numerical results are available at the website of this journal and Lisa Maria Dellmuth at Stockholm University (www.​lisadellmuth.​net). Earlier versions of the article benefited from reactions to presentations at ECPR (European Consortium for Political Research) in Bordeaux in September 2013, EISA (European International Studies Association) in Warsaw in September 2013, SWEPSA (Swedish Political Science Association) in Stockholm in October 2013, and the Transdemos Concluding Conference in Lund in June 2014. We want to thank, in particular, Thomas Gehring, Lisbeth Hooghe, Mathias Koenig-Archibugi, Jonathan Kuyper, Sofia Näsström, Thomas Risse, Jan Aart Scholte, Jens Steffek, the editor, and three anonymous reviewers of this journal for important comments without suggesting that they agree with our argument. This research was financially supported by Riksbankens Jubileumsfond and the European Research Council (200971DII).

E.g., Archibugi et al. (2012b); Bohman (2007); Brown (2011); Bäckstrand (2006); Dingwerth (2007); Dryzek and Niemeyer (2008); Frey and Stutzer (2006); Goodhart (2011); Gould (2004); Hardt and Negri (2005); Held (1995); Keohane et al. (2009); Macdonald (2008); Marchetti (2008); Miller (2010); Scholte (2014); Smith (2008); Scholte (2013); Steffek et al. (2008); Tännsjö (2008); Erman (2013).

E.g., Held (1995); Archibugi and Held (1995); Marchetti (2008); Tännsjö (2008).

E.g., Bäckstrand (2006); Frey and Stutzer (2006); Gould (2004); Hardt and Negri (2005); Dingwerth (2007); Dryzek (2006); Macdonald (2008); Macdonald and Macdonald (2006); Montanaro (2012); Saward (2010, 2011); Scholte (2013); Sechooler (2009); Smith (2008); Steffek et al. (2008); van Rooy (2004); Kuyper (2014).

Previous formulations of this principle vary slightly; see Bäckstrand (2006: 474); Goodin (2007: 51); Macdonald (2012: 47); Montanaro (2012: 1094).

It should be clear that both sociological and non-sociological variants of stakeholder democracy are normative theories, in the sense that they claim to specify how political institutions should be constructed and evaluated. Only the sociological variant of the theory, however, makes the causal assumption that the normative legitimacy of global institutions depends on their ability to foster experiences and perceptions of democracy through involvement of stakeholders or their organizations. Whether democratic legitimacy as understood in the sociological variant of the theory is an end in itself, or valuable in light of its consequences or necessity for still other matters, for example stability or effectiveness of global institutions, is a question not addressed in this article. The relevance of our research does not depend on this question since both sociological and non-sociological variants of stakeholder theory assume that democratic legitimacy is valuable (either intrinsically or extrinsically).

E.g., Archibugi et al. (2012a, b); Bäckstrand (2006); Dingwerth (2007); Macdonald (2008); Steffek et al. (2008).

See theoretical section below for elaboration on both points.

While Macdonald writes primarily about democratization of NGOs, she explicitly extends her argument to IOs: “[I]t is not difficult to imagine how the kinds of institutions I have discussed in this book could be employed to democratize their [IOs’] power” (Macdonald (2008: 224).

Stakeholder democracy may in theory be defended by arguing that political equality is normatively less important than political influence in proportion to varying stakes (Brighouse and Fleurbaey 2010). However, abandoning political equality as a fundamental principle of democracy is a theoretical weakness in itself (cf. Erman and Näsström 2013).

The sociological conception of stakeholder democracy is worth empirical examination because of the questions it addresses and the theoretical problems it may solve, not because of the number of researchers who have explicitly commit themselves to it so far. Nonetheless, sociological observations are commonplace in normative research on stakeholder democracy. Scholte (2013b) suggests that research on global democracy should as a first step inquire into what ordinary people mean by and desire in terms of democracy, and he also emphasises that all people should experience political “circumstances as being democratic in their own terms (Scholte 2013b: 15). Bäckstrand (2006) as well as Tallberg and Uhlin (2012), together with many others, motivate their interest in stakeholder democracy by pointing to the support for this ideal among real actors – as would be normatively irrelevant unless democratic legitimacy is defined partly in sociological terms. More generally, the recent interest in sociology in normative stakeholder theory parallels a shift towards non-ideal theory in debates on global justice (e.g., Sangiovanni 2008).

We thank an anonymous reviewer for suggesting this point.

The ability of stakeholder organizations to represent stakeholder communities democratically may be questioned (e.g., Pallas 2013). To be of interest in practice and normative theory, however, the stakeholder model of global democracy must assume that on balance, or in important cases, stakeholder organizations have this ability. For-profit actors are no exception. Firms and business associations may represent owners, consumers and experts in different sectors of the economy.

A completion rate of 30 % is similar to response rates of previous web surveys conducted among interest groups in the EU (see, e.g., Dür and Mateo 2013); Klüver 2013). In the absence of a 100 % response rate, the question arises whether non-respondents differ systematically from respondents, since this may give rise to a non-response error. Ideally, we would want to compare the characteristics of survey respondents with the characteristics of organizations in the total population that is surveyed (Rogelberg et al. 2003). Since we lack information about the total population of organizations involved and their characteristics, however, we cannot make such a comparison. Instead, we inquired for the reason of non-response when conducting the telephone survey. About half of the organizations that did not respond to the telephone survey indicated a lack of resources as a reason for not participating in the survey, which leads us to raise the cautionary note that organizations with relatively few resources could be underrepresented. Yet, although resourceful organizations may be overrepresented compared to the universe, the dataset involves organizations of all sizes, with a relatively similar distribution that is skewed towards smaller organizations across all IOs.

We trained four interviewers to conduct the interviews on the phone.

See Council of Europe (2011); Organization of American States (2011); UN DESA (2011). Data for non-state actors accredited to the AU was compiled using documents obtained from the AU Secretariat.

Before drawing the random samples, the sampling frames were checked for non-state actors appearing in both frames, but no such problem was detected.

The exact question wording of this and other variables in our analysis is summarized in Online Appendix A. See Tables B1 and B2 in Online Appendix B for descriptive statistics and correlations between all independent variables.

The exact question wording is as follows: Altogether, how would you rank the opportunities for your organisation to be involved in the following [IO] bodies? 1 “no opportunities,” 2 “few opportunities,” 3 “some opportunities,” 4 “many opportunities” (see Online Appendix C for an exhaustive list of the included IO bodies).

We tested whether several assumptions underlying OLS regression hold. Specifically, we examined the distribution of the residuals, which appear to be homoscedastically distributed. Furthermore, we tested for non-normal distribution of the residuals, but no such problem was detected. Last, a test of how much multi-collinearity may cause harm to the precision of the estimates reveals a Variance Inflation Factor (VIF) of less than 2 for all variables, indicating that multi-collinearity should not inflate the coefficient estimates (cf. Fox and Monette 1992).

To demonstrate how the predicted evaluation of representation, deliberation, and accountability changes when altering the explanatory variables from their minimum to their maximum value, while holding other variables at their means, we simulated first differences for models 2, 4, and 6. For each stakeholder organizations, we repeat the expected value algorithm M = 1000 times to approximate a 95 percent confidence interval around the expected value of influence, using the software package CLARIFY (King et al. 2000).

The base weight was calculated as the reciprocal of the probability of selection, BW=N/n, and reflect a non-state actors’ probability of being selected into the sample.