The Icesave Referendums

The contemporary global economic crisis broke calamitously over Iceland in September 2008 when 85 percent of its banking sector, the balance sheets of which showed liabilities in excess of 800 percent of assets, effectively went bankrupt.Footnote ¹ The Icelandic government swiftly nationalized the country's largest banks, including the third largest, Landsbanki, which had been marketing high-yield, Internet-based savings accounts under the “Icesave” brand to European consumers.Footnote ² Landsbanki could not repay the billions it owed to foreign depositors, including many British and Dutch nationals. These governments reimbursed their citizens' lost deposits and demanded reimbursement from the Icelandic treasury, which accepted the obligation, in principle. What remained to be established were the precise repayment terms. By the end of 2009, the Icelandic Parliament (the Alþingi) approved a $5.7 billion payout over fifteen years at 5.5 percent interest, beginning in 2016—which the UK and Netherlands promptly rejected.Footnote ³ At this point Iceland's president, Ólafur Ragnar Grímsson, interceded at the behest of a petition signed by more than 60,000 citizens—about a quarter of the electorate.Footnote ⁴ Acknowledging that the present terms amounted to approximately $17,000 for every man, woman, and child in the country of only 320,000, Grímsson stepped outside his traditionally ceremonial role and declared that the question should be brought to a public referendum under Article 26 of the Icelandic Constitution.Footnote ⁵

News of the referendum and doubts about its outcome led international credit agencies to downgrade Iceland's credit rating to “junk” status.Footnote ⁶ In hopes of averting the referendum, Iceland continued to negotiate for better terms and gained several term and interest-rate concessions. Nonetheless, a vote occurred on Saturday, 6 March 2010, on the basis of the inferior December 2009 terms. In Iceland's first-ever referendum, 62.7 percent of eligible voters turned out. Unsurprisingly, given that a superior offer was already on the table, Icelanders overwhelmingly rejected the original terms of repayment: 98.1 percent rejected the deal; only 1.8 percent approved.Footnote ⁷ With the repayment plan duly overruled by this direct democratic vote, interstate negotiations resumed, and in late 2010 a new agreement was reached calling for repayment over thirty years, beginning in 2016, at 3 percent interest to the Netherlands and 3.3 percent interest to the UK—terms much more favorable to Iceland. Though the Alþingi approved the settlement, in February 2011 President Grímsson “vetoed” this second deal as well, reiterating that the people should co-legislate the matter. This second Icesave referendum took place on 9 April 2011. 75.3 percent of eligible voters turned out, with 59.8 percent rejecting the proposal with “no” while 40.2 percent voted “yes.”Footnote ⁸

To our knowledge, these Icesave referendums represent the only occasions in history on which average citizens were asked to vote directly on sovereign debt resettlement. They thus provide an unparalleled opportunity to examine how citizens make such choices. Although referendums on debt will likely remain rare, analyzing voting behavior in these episodes sheds new light on public attitudes and behavior not only toward sovereign debt, which is presently a defining political issue across the advanced democratic world, but also toward questions of international economic policy more generally.Footnote ⁹

Mass International Political Economy

While historically unique, the Icesave referendums speak to a much broader set of issues. An interdependent era characterized by globalization and democratization demands an understanding of how everyday citizens think, choose, and behave politically regarding international economic questions. The past decade has witnessed considerable interest in mass IPE, as scholars seek to explain individuals' attitudes and, less commonly, behavior as they relate to the complex matters posed by the global economy. Mass IPE studies typically follow the basic Open Economy Politics (OEP) démarche of identifying the domestic distributional implications of international economic policy changes, imputing preferences to actors based especially on their income/employment profiles and how these would be affected by the policy changes in question, and then testing politics for empirical consistency with these theoretically derived economic preferences.Footnote ¹⁰ Mass IPE takes OEP models from the group to the individual level.

Self-interest initially played a central role in mass IPE literature.Footnote ¹¹ Scheve and Slaughter found that Americans' factor type (mobile versus immobile, proxied theoretically by skill level and empirically by education and occupational earnings) shapes attitudes in ways consistent with Heckscher-Olin (“factor endowments”) models of the income effects of trade.Footnote ¹² They found similar results for immigration attitudes: low-skilled US workers oppose open immigration policies that would suppress their incomes.Footnote ¹³ A wave of subsequent studies supported the role of material self-interest in shaping public attitudes toward policy changes in, inter alia, the areas of trade,Footnote ¹⁴ immigration,Footnote ¹⁵ and exchange rate regimes.Footnote ¹⁶

Despite these results, more recent research casts doubt on self-interest's role.Footnote ¹⁷ Hiscox warns against priming effects lurking in survey question wordings.Footnote ¹⁸ Hainmueller and Hiscox suggest that education's observed positive effects on attitudes toward trade liberalization cannot reflect factor income concerns, inasmuch as it operates similarly among respondents within and without the active labor market.Footnote ¹⁹ Since educational attainment was the standard measure for factor mobility, skill, human capital endowments, and other “self-interest” constructs, this finding called previous inferences into question. Similarly, factor returns' effect on immigration attitudes has been undermined by improved measures of exposure to wage competition.Footnote ²⁰ More recently, Mansfield and Mutz find no support for self-interest on trade attitudes in either Heckscher-Olin or Ricardo-Viner specifications.Footnote ²¹ They instead find strong effects for sociotropic economic evaluations and symbolic attachments.

In sum, work at the research frontier questions the role of self-interest in shaping mass engagements with IPE matters. Yet there is reason to think that conventional observational samples, drawn especially from attitudinal surveys, are ill-suited to making inferences about self-interest. Citrin and Green argue that self-interest explains mass attitudes only insofar as the stakes are visible, tangible, large, and certain.Footnote ²² IPE issues, by contrast, exhibit high complexity, lengthy causal chains, low salience, and small stakes.Footnote ²³ “Nonattitudes” prevail: the modal citizen appears to lack meaningful opinions from which to draw inferences.Footnote ²⁴

Sampling frames with three features can help overcome these problems. First, scholars can use data from crisis environments with heightened public attention, dense information, and clear stakes.Footnote ²⁵ The Icesave referendums clearly qualify: the very solvency and survival of the country were widely considered at risk. Related to this, inference benefits from circumstances in which individual stakes are tangible enough to pass the noise threshold and serve as actionable guides to behavior. While the costs of a “no” Icesave vote were, as far as we can find, never distillable into a summary statistic, the costs of a “yes” were quantified as the amount to be repaid divided by the total population: $17,000 per person. Finally, truly behavioral rather than merely attitudinal data can be especially informative. The cognitive contours of direct action (casting a vote, marching in the street, buying a basket of consumer goods) may differ from those that characterize a mere opinion, exhibiting clearer stakes, cause-effect relations, and other inferentially fruitful features.Footnote ²⁶ Here again, the act of voting in the referendum, as opposed to simply voicing attitudes about the underlying problem(s), provides a clearer signal than other empirical frames can generate.

Self Interest, Sovereign Debt, and Voting Behavior

We apply the standard OEP logic to mass material self-interest in relation to sovereign debt resettlement. For a most stringent test, we define self-interest narrowly as “the tangible, relatively immediate, personal or family benefits of a policy.”Footnote ²⁷ We identify three economic impacts, and associated domestic distributional consequences, linked with the decision to accept or reject the proposed repayment terms put to citizens in the 2011 Icesave referendum. We emphasize objective self-interest to avoid the risks of treating as causally distinct what are actually co-constituted dispositions and attitudes. Our dependent variable—vote choice—is behavioral, helping to avert the “attitudes on attitudes” problem for causal inference.Footnote ²⁸

First, nonrepayment could trigger interest rate increases through a default-downgrade process and an attendant rise in borrowing costs.Footnote ²⁹ Individuals are variably exposed to these costs. Those holding short-term and/or variable rate personal debt should be most sensitive to these potential pocketbook impacts, and most likely to seek to avoid them through repayment. This leads to our first hypothesis:

Second, repayment of $5.7 billion in debt, representing 5 to 8 percent of the gross domestic product (GDP), would divert roughly that same amount of treasury resources from other valued uses in the form of fiscal retrenchment.Footnote ³⁰ Citizens financially dependent on public expenditures, for example, those unemployed and/or employed in the public sector, should be most sensitive to such diversions and hence be least likely to support repayment. From this we derive a second hypothesis:

Third, default—or nonrepayment—would likely exacerbate capital outflows as investors would become increasingly concerned about the anticipated course of economic policy. Given Iceland's floating exchange rate, a capital outflow would result in a rapid and significant devaluation of the krona, diminishing the relative value of assets held in Icelandic portfolios. Individuals holding investments should be most likely to seek their stabilization by voting in favor of the referendum. Thus we have our third hypothesis:

Support for self-interest explanations need not preclude the importance of symbolic attachments or domestic political dynamics.Footnote ³¹ Many existing findings speak to the strong impact of nationalism;Footnote ³² cosmopolitanism, ethnocentrism, and sociotropic orientations (that is, attunement to broader societal benefits) also feature prominently in explaining attitudes toward trade and immigration.Footnote ³³ Our analyses thus account for these factors. At the same time, when it comes to voting on IPE questions, citizens should be attentive to the political messages they receive from partisan elites. Mass IPE studies in the American context sometimes consider partisan identification—though only as a control—given well-known differences between Republicans and Democrats toward trade liberalization.Footnote ³⁴ We embrace a more generic cueing model, such that individuals will follow the positions articulated by their preferred party.Footnote ³⁵

Education is ambiguous in mass IPE models: it is almost always positively related to attitudes and behavior that advance liberalization/openness. Initially a proxy for skills in factor endowment models of trade attitudes,Footnote ³⁶ it then came to be understood as capturing issue-comprehension (for example, exposure to the economics of trade).Footnote ³⁷ Most recently, this, too, has been called into question, with education now being postulated to foster cosmopolitan attitudes and thus induce more friendly attitudes toward “others” when engaging IPE questions.Footnote ³⁸ Based on existing work, we expect education to be positively associated with supporting repayment, but we remain agnostic regarding the specific causal mechanism.

Research Design and Data

We examine these expectations in the empirical context of the second Icesave referendum, held on 9 April 2011, after which we immediately fielded an original survey.Footnote ³⁹ Implemented by the Social Science Research Institute (SSRI) of the University of Iceland in Reykjavik, the survey was conducted via telephone interviews to a nationally representative sample drawn from the national register. In the survey, 785 respondents participated for a response rate of 65.4 percent. The referendum asked Icelanders whether they endorsed the terms by which their government was to settle its €3.8 billion ($5.7 billion) debt to the British and Dutch governments. In principle, citizens were not voting on whether to repay Iceland's sovereign debt, but when and how.

Variables

We asked those who confirmed they had voted the following question: “How did you vote in the April 9th ‘Icesave’ referendum? Did you vote in favor of having the repayment plan remain in force, did you vote to have it repealed, or did you submit a blank ballot?” Our dependent variable, vote choice, is dichotomous, equal to 1 if respondents voted in favor of the proposed repayment plan and 0 if they voted against it. Proportionate to the actual referendum outcome, 42.4 percent of our sample supported the terms and voted in favor while 55.8 percent voiced their opposition by voting against.Footnote ⁴⁰

Our key independent variables tap the various components of our material self-interest hypotheses. For our borrowing costs hypothesis (H1), we include a measure of respondents' reported personal credit card debt (None, Not very much, Some, A lot). We focus on credit card debt because variable rates and monthly repayments mean that individuals should be especially sensitive to the effects of interest rate changes; rising interest rates make individuals' existing debt more expensive to repay and make access to future funds uncertain. We expect that the more credit card debt individuals hold, the more likely they will vote in favor of repayment; thus, this indicator should be positively signed.Footnote ⁴¹

For our fiscal dependence hypothesis (H2), we follow Tomz in coding both unemployed and public-sector employee with dummy variables.Footnote ⁴² Status in either category implies reliance on state outlays for personal income. Insofar as debt repayment might detract from domestic spending, we expect the fiscally dependent to oppose repayment; hence, these coefficients should be negatively signed.

To assess our investment assets hypothesis (H3), we code has investments as 1 for individuals holding investments in Iceland or abroad, and 0 otherwise; this should be positively signed. In our sample, approximately 41 percent of respondents report having investments—a number consistent with reports from the Icelandic Chamber of Commerce.Footnote ⁴³ Thus, to test all three self-interest hypotheses, we have selected indicators of material well-being as they relate to the (non-)resettlement of Iceland's sovereign debt.

We account for alternative explanations by first including measures that tap symbolic attachments on two dimensions, one ranging from inclusive to exclusive identity and another ranging from cold/negative to warm/positive outgroup affect. exclusive icelandic identity is a four-point scale based on whether respondents proclaim themselves European only, European and Icelandic, Icelandic and European, or Icelandic only. Because claimed identity becomes more narrow and exclusive the higher the value on this indicator, we expect it to have a negative sign. For outgroup affect, we created a mean feeling thermometer index of respondents' ratings toward three countries (the UK, Netherlands, and India) on a scale of 0 to 100, where higher values indicate more positive, friendly opinions of the country in question. The UK and the Netherlands are the countries to which repayment would be made; India captures general disposition toward foreign countries. Because we expect cosmopolitanism to favor repayment, this variable should have a positive sign.Footnote ⁴⁴

Our conceptualization of partisan cues emphasizes the importance of elite messaging. Knowing the positions taken by the Icelandic political parties, our main specification (Table 1) includes a series of party identification dummies to capture respondents' professed support for the progressive, independence, social democratic alliance, or left-green movement parties. We expect partisans of the Social Democratic Alliance to vote in favor of repayment, since the deal was negotiated under and supported by the Social Democrat-led government and since the party endorsed the “yes” vote. While the Left-Greens were in government with the Social Democrats, the Icesave issue was controversial and there was no official “party line,” leaving us with mixed priors for these partisans. For the opposition parties, we expect partisans of the center-right Progressive Party to oppose repayment, following the leadership's position, while supporters of the conservative Independence Party received no clear cues from their party elites about how to vote.Footnote ⁴⁵ In Table 2, we consider an alternative specification centered on government-opposition dynamics by excluding the party identification measures and including a measure of pm approval (a five-point Likert scale of agreement with whether “Jóhanna Sigurðardóttir is doing a good job as Prime Minister”).

Table 1. Vote choice in Icesave 2 referendum

Notes: Logistic regression. Robust standard errors clustered by respondent postal code in parentheses. * p < .10; ** p < .05.

Table 2. Alternative measure of domestic political cues

Notes: Logistic regression. Robust standard errors clustered by respondent postal code in parentheses. * p < .10; ** p < .05.

Given its centrality to mass IPE debates, we include a measure of educational attainment, a seven-point scale from “No schooling” to “Postgraduate.” In most analyses, education correlates positively with more globally open attitudes, which in the Icesave case should induce greater likelihood of a “yes” vote.Footnote ⁴⁶ Finally, we control for gender (male).Footnote ⁴⁷

Primary Results: The Role of Self-interest in Mass IPE Behavior

Given our dichotomous dependent variable, we estimate logit models and report robust standard errors clustered by the respondent's postal code. Table 1 presents the main results, with the first three models corresponding to different specifications of material self-interest: borrowing costs (Model 1), fiscal dependence (Model 2), and investment assets (Model 3). Model 4 includes all self-interest measures together, from which we plot, in Figure 1, the first differences and associated confidence intervals of our key variables for a more substantive interpretation.

Figure 1. Marginal effects (from Table 1, Model 4)

With the exception of public-sector employee (included to evaluate the fiscal dependence hypothesis), the material self-interest variables are always signed as expected, though the effect is not robust across specifications. Those holding more credit card debt and whom we expect to be most sensitive to rising interest rates are approximately 4 percent more likely to vote in support of repayment. This coefficient is statistically significant, at p < .10, in the limited Model 1 but falls out in the fuller Model 4.Footnote ⁴⁸ Similarly, those with investment assets (H3) are almost 10 percent more likely to vote in favor of repayment.Footnote ⁴⁹ While, as expected, the unemployed voted against repayment, public-sector employees do not indicate the same opposition. (public-sector employee's insignificance is not sensitive to inclusion of the unemployment measure.) The fiscal dependence hypothesis (H2) thus finds mixed support, though we suspect that unemployment is a cleaner test of fiscal dependency than public-sector worker status, which might include many who could substitute private-sector employment in a fiscal pinch. Overall, we find some support for our overarching proposition that material self-interest shapes voting behavior in terms of sovereign debt resettlement, a single but perhaps increasingly important aspect of international economic policy.

Several variables were designed to account for existing findings on symbolic attachments and cosmopolitan dispositions in shaping attitudes and referendum voting behavior, both in the European Union (EU) and for mass IPE. exclusive icelandic identity reduces the likelihood of a vote for repayment and is statistically robust (p < .01) across our models. It is natural to suspect that the fact that repayment would be made to foreign creditors plays a part here, though we cannot, of course, verify that claim with the present empirical sample. The feeling thermometer index also operates as expected, this time positive, though our confidence decreases across specifications. Given our measure, the “warmer” individuals feel toward non-Icelanders, the greater likelihood they will support repayment.

Domestic political cues operate quite systematically, with expected effects across the models. Partisans of the governing social democratic alliance (the prime minister's party) and the left-green movement party strongly support the repayment plan—they are 42 percent and 37 percent, respectively, more likely to vote for it—while those identifying with the opposition independence and progressive parties oppose it. We note that the observed effects of party identification are fully predicted by a government-opposition distinction, rather than a cueing logic per se. Table 2 substitutes party identification for pm approval and finds, unsurprisingly, that it is always positive and significant. In this specification, credit card debt and feeling thermometer index both lose significance.

In all models, higher educational attainment is associated with a greater likelihood of voting “yes.” This replicates standard findings whereby more educated citizens express more support for open international economic policies. While sovereign debt resettlement does not fit neatly onto a continuum running, for example, from international economic closure to openness, repayment fits intuitively with support for cooperative international engagement. Mansfield and Mutz have suggested that this sense of “cosmopolitanism” is precisely what education seems to supply.Footnote ⁵⁰ In their analysis, education loses significance when modeled together with indicators of isolationism and outgroup affect, but is insensitive to the inclusion of nationalism. In our analyses, there is a consistently significant relationship between educational attainment and outgroup affect (as captured by the feeling thermometer index), though we find some sensitivities between education and exclusive icelandic identity. But, per Table 1, educational attainment is always positive and statistically significant, even with all variables included. Education's steady and strong effects in the presence of our symbolic attachment indicators suggests there is still more going on than we are tapping with various measures of human capital endowments or cosmopolitan attitudes.Footnote ⁵¹

Extension: The (Non-)Conditioning Impact of Political Sophistication

The foregoing results invite reconsideration of the recently common rejection of material self-interest as an explanatory factor in mass IPE models. We have suggested that conventional sampling frames, which survey attitudes about far-distant, often materially inconsequential, and poorly understood questions, may be inappropriate for testing pocketbook effects. The correct conclusion may be not that material self-interest is inconsequential, but that the inferential environment in which it tends to be studied is often too noisy to detect its true effects. We thus constructed a sampling frame characterized by a crisis environment, clear and high stakes, and true political behavior rather than mere attitudes. We find, contrary to the thrust of the literature, that material self-interest can shape mass political behavior toward IPE matters.

Thus far, we have treated these factors as additive. Yet sovereign debt resettlement involves complicated tradeoffs across countless dimensions, and rational choice is cognitively demanding.Footnote ⁵² Studies of economic voting suggest that the operation of material self-interest should be conditional on recognizing those interests (and understanding linkages between interests, behavior, and outcomes), and hence on political sophistication.Footnote ⁵³ These insights have been implemented in some work in mass comparative political behaviorFootnote ⁵⁴ but have not tended to characterize mass IPE work.Footnote ⁵⁵ Thus, we investigate these considerations in the second Icesave referendum context and put forward our final hypothesis:

H4

(Sophisticated interest): The more sophisticated an individual, the greater the impact of material self-interests on voting behavior.

Sophisticates, in short, should be better able to recognize and act on their interests than nonsophisticates.

In keeping with standard practice, we measure an individual's political sophistication using a mean objective knowledge index based on the number of correct answers respondents provided to objective questions. In our 2011 survey, respondents were asked to identify the issue at stake in the Icesave referendums, then how many EU member states there were at the time.Footnote ⁵⁶ We then interact this measure with each material self-interest measure (credit card debt, unemployment , and investment assets).

Table 3 examines the potential conditioning effect of political sophistication. As a baseline we add our knowledge measure to Model 1 of Table 1. By itself, the variable is not significant (Model 9). We then interact knowledge with the measures of economic interests (Model 10).Footnote ⁵⁷ There are revealing patterns here: credit card debt is not statistically significant by itself yet becomes significant once interacted with knowledge, suggesting that knowledge increases the willingness of those with large debt to vote “yes.” But, importantly, the aggregate effect of knowledge and credit card debt is not statistically significant. The same occurs with unemployment.

Table 3. The role of knowledge

Notes: Robust standard errors clustered by respondent postal code in parentheses. * p < .10; ** p < .05.

Why might this be the case? For our purposes, the premise of the sophistication hypothesis revolves around the availability of information and an individual's willingness/cognitive ability to access (or proxy for) it. We suspect—but cannot establish conclusively—that the null result around the conditioning role of political sophistication on interests may have to do with the informational environment surrounding the vote. Iceland was considered to face a near-existential threat from October 2008 forward and had seen the collapse of its government, currency, and banking system. Given this, it is unsurprising that the Icesave issue was regularly front-page news from summer 2009 forward, especially around both referendums. To complement our survey, we also commissioned a media study of the 2010 referendum campaign from SSRI, covering the two-month period from 6 January to 6 March 2010. The analysis yielded 941 different news items across the largest and most popular Icelandic newspapers, radio and television stations, and Internet news sites in Iceland. We do not have equivalent measures for 2011, but a search on the webpage of Iceland's main newspaper, Morgunblaðið, for references to “Icesave” returns more than 700 items in the eleven months preceding the referendum (May 2010 to March 2011), and more than 100 in the eight days preceding the actual vote. We conjecture that in such a saturated informational environment, most citizens were able to make well-informed choices, leaving relatively little variation in effective sophistication. In other words, the highly salient and publicized context surrounding this particular event may explain why, according to our data, almost all Icelanders appear highly “sophisticated.”Footnote ⁵⁸ While our results suggest an unconditional impact of material self-interest on voting behavior, this conditional conjecture certainly warrants further examination before final conclusions are drawn.

Discussion, Implications, and Conclusion

We have leveraged data from the world's only direct democratic vote on sovereign debt resettlement to test propositions about the role of material self-interest on mass political behavior relating to international economic policies. The Icesave referendums took place in a crisis context, involved high and clearly discernible individual stakes, and demanded the active political participation of each voter. Based on various specifications, we find that Icelanders did, to some extent, vote their pocketbook interests in the April 2011 “Icesave 2” referendum. Those expected to be more sensitive to prospective borrowing costs and who would potentially have much to lose from a rejection-default-downgrade scenario tended to vote in support of repayment, as did, in some specifications, those with substantial credit card debt. So also did those with investment assets, worried, we presume, about the inflationary effects of a currency devaluation on their portfolio.Footnote ⁵⁹ The unemployed, by contrast, who we postulated should worry about the fiscal retrenchment that would follow repayment of such a large obligation, tended to vote against repayment, though public-sector employees with putatively similar public finance preferences did not. Taken together, we find substantial support for the view that individual self interests shaped citizens' voting behavior regarding sovereign debt (non-)resettlement.

While not consistently strong, this self-interest result contradicts the recent thrust of mass IPE findings. In reconciling our results with these previous works, we suggest that standard observational approaches and samples may be inappropriate for accurately assessing the effect of pocketbook interests. For example, self-interest may be undetectable in surveys of trade attitudes because the signal on material interests sent by a given question is so weak that our measurement technology cannot capture it. Insignificant results may ultimately reflect the inferential terrain more than the real-world data-generating process.

At the same time, our results confirm and advance other aspects of the mass IPE literature. We find substantial effects for symbolic attachments on voting behavior, consistent with EU-based results on the role of “community” and IPE-based results on the importance of “cosmopolitanism.” What EU scholars would refer to as “cues” (or “second-order” effects), where domestic partisan-political orientations shape behavior toward IPE questions, also operate strongly. Party identification has the largest impact on vote choice among all of the factors that we consider.Footnote ⁶⁰ Education, which has been the object of vigorous examination elsewhere, is, as in most studies, positively correlated with support for international cooperation (here, repayment of Icelandic debt to sovereign creditors). Education's effects are strong and consistent across our models. While we confirm the relationship established by Mansfield and Mutz between educational attainment and our measures of symbolic attachment, in our analysis education continues to exert substantial independent effects on vote choice.Footnote ⁶¹ This thus reopens the puzzle of “educated preferences” over international economic policies.Footnote ⁶²

Finally, in keeping with the increasingly elaborated behavioral models in mass IPE studies, we considered a possible conditioning effect for political sophistication. It would make sense if, confronted with dauntingly complex political-economic questions, sophisticates could better recognize and act on their interests than nonsophisticates. Interestingly, we find no such separation in our data, which we expect stems from the information-rich environment surrounding the particular case of the second Icesave referendum. While we can provide no definitive answers here, we suggest that greater attention to informational and media environments would further enrich future mass IPE work and bring it into greater dialogue with more general political behavior models, among which such factors are a staple.

To what extent do our results from a unique referendum vote in a unique little place inform broader questions of interest to mass IPE scholars? We believe considerably. First, as scholars of rare events in international relations have argued, sometimes we need to study the “ones” more closely than the “zeroes”; random sampling is inefficient in a rare-events world.Footnote ⁶³ This is the place to look to understand mass political behavior around sovereign debt. Second, the contours of the Icesave referendums—crisis context, clear and high stakes, and truly behavioral rather than mere attitudinal outcomes—are exceptionally conducive to inference. Like other IPE issues, Icesave posed a complicated political economy problem also involving symbolic attachments and elite messages. But unlike many existing areas of study, it afforded us a cleaner signal on self-interest than do standard, even heavily primed, survey questions. Third, we have no reason to expect that Icelanders would choose to act any differently than voters in any of the other democratic open economies felled by the Eurozone crisis. Greeks, Portuguese, Cypriots, Italians, Hungarians, and many others continue to work through the mass politics and economics of sovereign debt resettlement, as do citizens beyond Europe. Like Icelanders, many have taken to the streets to protest their economic difficulties. While it is unclear whether publics elsewhere will get to vote directly on debt repayment, we still strive to understand how they think and act around the issue. There is no clear reason to expect that mass attitudes and behavior in other crisis-battered, highly indebted, advanced industrial democracies should manifest themselves differently than they did in Iceland.

The recent crisis has stretched political orders to the breaking point. Old understandings of the social contract binding states to citizens are being squeezed between the demands of everyday people, domestic banks, international officials, and global markets. Regardless of whether other referendums on sovereign debt resettlement arise, policy choices require political solutions, which in democracies will always involve citizens to some degree. Understanding the material self-interest, as well as symbolic and political underpinnings of mass behavior toward sovereign debt resettlement can inform and guide not only these policy discussions but also broader understandings of democratic politics in a global economy.