The debt ceiling is a peculiar and increasingly fraught feature of American politics. From its inception in 1917 to the mid-1990s, Congress routinely raised the debt ceiling without controversy. Over the past 30 years, however, Republican Members of Congress used the debt ceiling as a brinkmanship ploy five times.Footnote 1 Failure to raise the debt ceiling risks a sovereign default and unprecedented negative consequences for the domestic and international economy.Footnote 2 In May 2023, Moody’s Analytics predicted that a protracted default from a failed debt ceiling negotiation would eliminate millions of American jobs, upend financial markets, and have lasting effects for generations of Americans.Footnote 3 Given the costs of a debt ceiling breach and default, it is puzzling how politicians engage in such costly games of chicken without public backlash.
In this research article, we examine whether voter ignorance and uncertainty around the consequence of a debt ceiling breach affect political support during negotiations. Public support is assumed to enhance bargaining credibility by raising the political costs of capitulation (Fearon Reference Fearon1994; Groseclose and McCarty Reference Groseclose and McCarty2001). Less acknowledged is that public support or acquiescence might stem from public ignorance or misinformation, a phenomenon particularly likely in debt ceiling debates. Fiscal politics, the debt ceiling, and the consequences of a ceiling breach are difficult for most citizens to understand given the absence of a precedent. Furthermore, conflicting statements from politicians can contribute to public ignorance. For example, Donald Trump argued that a default would be preferable to an agreement over raising the debt limit, while other Republicans questioned whether a default would occur if the debt limit were not raised.Footnote 4 Even non-partisan experts are uncertain about the consequences of a debt ceiling breach given the legal ambiguity over government responses.Footnote 5 As such, the problem has structural characteristics that make public assessment more challenging. Both ignorance and outcome uncertainty, we argue, limit the ability of the public to reign in politicians during this risky political game.
To test our expectations, we fielded a survey experiment on a quota sample (reflective of the US Population along with age, gender, and race/ethnicity) two weeks before the expected default on US debt in 2023. We find that information about the costs of a debt ceiling breach increases support for resolving the issue, even if that requires conceding to the other party’s demands. The effect is consistent across the political spectrum. Further, we find that more certain information has a stronger and significantly different effect than less certain information, suggesting that the unpredictability of the event has independent effects on public attitudes.
Scholars have only just begun to examine legislative dynamics in debt ceiling negotiations (Herrera, Macé, and Núnez Reference Herrera, Macé and Núnez2023) and other domestic brinkmanship scenarios (Pitsoulis and Schwuchow Reference Pitsoulis and Schwuchow2017). No research, to our knowledge, has investigated the mass politics aspect of this high-stakes game despite strong theoretical priors that the public plays an important role. Beyond the consequential but rare debt ceiling negotiations, our study also contributes to comprehending the role of information and uncertainty in political processes (Christensen Reference Christensen2022; Jacobs and Matthews Reference Jacobs and Matthews2017). Furthermore, it holds significance for models involving third-party constraints on politicians’ bargaining space (Fearon Reference Fearon1994; Martin Reference Martin1993).
Public Ignorance and Uncertainty in Bargaining
Politicians can use brinkmanship for political gain. Brinkmanship can alleviate government gridlock (Herrera, Macé, and Núnez Reference Herrera, Macé and Núnez2023), send signals to constituents (Patty Reference Patty2016), and can also be used to undermine political rivals if rivals take a greater share of the blame (Gieczewski and Li Reference Gieczewski and Li2022; Hirsch and Kastellec Reference Hirsch and Kastellec2022).
In the case of debt ceiling brinkmanship, it is hard to make a case that the public sees any benefit given the high costs of bargaining failure. Still, the public is not a passive audience to political theatrics. Public support likely motivates or sanctions the use of the debt ceiling as a bargaining ploy. Public appraisals constrain the bargaining space of negotiators by imposing costs on negotiators that make unfavourable concessions (Putnam Reference Putnam1988). The public’s role has been analyzed in crisis bargaining in international relations (Debs and Weiss Reference Debs and Weiss2016; Fearon Reference Fearon1994; Kertzer and Brutger Reference Kertzer and Brutger2016; Martin Reference Martin1993; Schultz Reference Schultz1998; Tomz Reference Tomz2007) and extends to legislative bargaining (Anderson, Butler, and Harbridge-Yong Reference Anderson, Butler and Harbridge-Yong2020; Bauer, Yong, and Krupnikov Reference Bauer, Yong and Krupnikov2017; Groseclose and McCarty Reference Groseclose and McCarty2001).Footnote 6 A key implication of these models is that when the public punishes concessions and backing down, politicians are more likely to issue meaningful threats, but the risk of bargaining failure increases. Most of these models assume, explicitly or implicitly, that the public’s preferences are fixed and grounded in preferences for competent leaders or in their bargaining outcome preferences. What has not been fully appreciated is that the public may be ignorant or misinformed about the costs of bargaining failure.
Voters who are ignorant or uncertain about the consequences of a debt ceiling breach affect brinkmanship bargaining in several ways. First, voters who underestimate the costs of a default may signal that they will impose greater political costs on politicians who accept concessions and back down. With more accurate information, citizens may be more likely to accept concessions in negotiations. Theoretically, this should mean a wider bargaining range, a lower chance of bargaining failure, and the avoidance of crisis scenarios.
If voters care about competence beyond outcome, they can still influence bargaining outcomes. Levendusky and Horowitz (Reference Levendusky and Horowitz2012) find that backing down when it is the ‘right thing to do’ increases competence assessments rather than harming leaders in foreign policy crises. Kertzer and Brutger (Reference Kertzer and Brutger2016) find that leaders are not only punished for backing down in international crises but also for belligerency. In the context of the debt ceiling, if voters perceive concessions as incompetence, politicians may be more inclined to pursue brinkmanship to please voters. Conversely, better-informed voters might provide different incentives. Risking an economic catastrophe for relatively minor political gains could be seen as irresponsible while showing flexibility or making concessions could be seen favourably.
While the public is likely misinformed about many crisis bargaining outcomes, the role of information is particularly important in the context of the debt ceiling. Under normal conditions, fiscal politics and sovereign debt are difficult concepts for the public to understand (Grigoli and Sandri Reference Grigoli and Sandri2023; Roth, Settele, and Wohlfart Reference Roth, Settele and Wohlfart2022). As a result, the public often relies upon co-partisan cues or trusted media sources to form their opinions (Alt, Marshall, and Lassen Reference Alt, Marshall and Lassen2016; Bansak, Bechtel, and Margalit Reference Bansak, Bechtel and Margalit2021; Barnes and Hicks Reference Barnes and Hicks2018). While government budgeting happens consistently and annually, brinkmanship over raising the debt ceiling happens more infrequently. Further, US voters have no direct experience with the possible outcome of an extended debt ceiling breach – a sovereign default. This lack of experience and the complexities of fiscal policy and default mean the voters are likely to be misinformed about the consequences of a debt ceiling breach and are subject to misinformation by politicians trying to strengthen their own bargaining position.
Observational survey data reflects the expectation of ignorance. An early May 2023 Associate Press poll found that only 21 per cent of Americans closely followed the debt ceiling negotiations. A YouGov/Economist poll in April 2023 suggests that this lack of attention has consequences for understanding the stakes involved. Only 27 per cent of Americans believe that a default on the US would result in a crisis.Footnote 7 While Republican voices have been loudest in signalling that a default would not be a crisis, the poll demonstrates that the assessment of the consequences of a default is fairly consistent among Democrats, Independents, and Republicans. Lastly, a non-experimental CBS News/YouGov Poll, in early April 2023, finds that respondents update their support for raising the debt ceiling when learning that not doing so would cause a default.Footnote 8
Given the expectation that the public’s ignorance about the consequences of a debt ceiling breach influences their opposition to concessions in debt ceiling negotiation, we state our first pre-registered hypothesis:Footnote 9
Hypothesis 1 Information about the negative consequences of failing to increase the debt ceiling increases support for concessions.
Beyond ignorance or naivety that can be remedied with facts or, at least, strong and well-informed predictions, outcome uncertainty may have an independent effect of support for brinkmanship.Footnote 10 While experts are in agreement that a debt ceiling breach would have crisis-like consequences, they disagree on the severity of the consequences (Vaitilingam Reference Vaitilingam2023). Even putting a bound on the estimates is difficult to do given the uniqueness of a debt ceiling breach and the American centrality to the world economic system. This uncertainty should play in the public’s ability to assess the costs and benefits of the bargaining process. As Christensen (Reference Christensen2022) shows, voters discount outcomes when presented with uncertain predictions of outcomes of general policy reforms. The same logic should extend to preferences in bargaining scenarios as well. Thus, we expect that respondents are more supportive of concessions when provided with certain estimates of the costs of a crisis than when presented with uncertain estimates. Uncertain estimates should, in effect, lead citizens to revise downward their assessment of consequences and reduce support for concessions.
Hypothesis 2 More certain information about the negative consequences of failing to increase the debt ceiling increases support for concessions more than less certain information.
Finding that uncertain information has a smaller effect than certain information allows us to assess how the structural nature of outcome uncertainty, rather than information that can theoretically be addressed with accurate information, contributes to the underlying political support for concessions in debt-ceiling-like scenarios.
Experimental Design
To test our hypotheses, we recruited a quota sample on the Prolific survey platform approximately two weeks before the expected early June deadline to pass a debt ceiling agreement (16 May 2023). We chose this platform because it allowed us to recruit a sample in a single day that reflects the US population along age, sex, and ethnicity. Further, recent studies demonstrate that Prolific respondents are more attentive than those recruited via Mechanical Turk, Dynata, or Qualtrics (Douglas, Ewell, and Brauer Reference Douglas, Ewell and Brauer2023; Peer et al. Reference Peer2022).
The timing of the survey was planned to capture a moment when the issue was salient but still allowed us to leverage general unfamiliarity with the topic. As we show in Fig. 1, our survey was fielded just as news mentions of the ‘debt ceiling’ increased in the national media.
Debt Ceiling News and Survey Date: Google Trends news media mentions of ‘debt ceiling’ in the months before and after the debt ceiling deadline of 1 June 2023.
The survey begins with a consent form, demographic questions, and a module to collect general attitudes toward public debt. The latter serve as quasi-pretreatment questions to increase the precision of the estimates (Clifford, Sheagley, and Piston Reference Clifford, Sheagley and Piston2021). Following this introduction, we provided a brief background on the debt ceiling debates that described the debt ceiling rule and the consequences of breach, and that Republicans sought budget cuts and limits on future spending in exchange for raising the debt ceiling. This was delivered in a matter-of-fact tone.
We then evenly assigned respondents by block randomization to one of three conditions: more certain information about the consequences of a debt ceiling breach, less certain information, or no information about the consequences.Footnote 11 To assess the effect of certain and uncertain information treatments, we present text and accompanying figures that illustrate the estimated job losses of a clean debt ceiling increase versus a default and prolonged breach of the debt ceiling. Table 1 presents the text and figures for each of our treatment conditions: certain (1/3) and uncertain (1/3). Our control condition does not receive any additional information.
Treatment images and text
The ‘certain’ treatment condition provides actual estimates from Moody’s Analytics on the consequences of a prolonged debt ceiling breach.Footnote 12 We use these estimates to create a figure and mention the source in the text treatment. The figure depicts job growth under the two stark conditions of a clean debt ceiling increase and a breach of the debt ceiling. Moody’s Analytics is unlikely to be associated with any partisan bias among the few readers who are familiar with the service. Through its role as an information provider to firms, Moody’s Analytics should signal neutrality to those unfamiliar with the service.
The uncertain treatment again uses a single estimate of a clean debt ceiling increase, but provides estimates of the potential costs. We use the debt ceiling breach estimates from Moody’s as a maximum bound and generate additional lines between the two scenarios. This figure demonstrates that between 3 and 7 million jobs will be lost while the text highlights these are just a few possible estimates of job losses. The text emphasizes uncertainty by highlighting the unprecedented nature of such an event and the inherent unpredictability of its consequences.Footnote 13
Following the treatment conditions, respondents answer outcome questions. We are interested primarily in citizens’ willingness to support backing down and accepting the terms of the other party in a brinkmanship crisis. We see this as the most relevant attitude, which provides support for politicians’ ability to carry out such bargaining and is indicative of an intention to punish making concessions. After asking the respondents about the general importance of reducing the debt ceiling (see Supplementary Appendix), we asked those Democrats who lean closer to the Democratic Party and true independents, ‘How important is it for the government to increase the debt limit even if Democrats must make concessions?’ We asked the respondents who identified as Republicans or closer to the Republican party, ‘How important do you think it is for the government to increase the debt limit even if Republicans CAN NOT get Democrats to agree to spending cuts?’ Respondents replied on a 5-point scale of ‘Not Important at all’ to ‘Extremely Important.’ We chose to ask different outcome questions based on partisanship to reflect the actual stakes as closely as possible. Asking an identical question of both respondents risks obfuscating what a ‘concession’ would entail and thus takes us further from the outcome of interest.
Results
The top panels of Fig. 2 present the mean outcome by treatment group when all respondents are pooled. In addition to the means, we also indicate the statistical significance of the difference. These p-values are from ordinary least squares models in which we regressed the treatment(s) and a set of covariates on the outcome variable. The bottom panel (c) presents the standardized coefficients to demonstrate the substantive effect as a percentage of a standard deviation. Following Bloniarz et al. (Reference Bloniarz2016), we selected the covariates agnostically by first estimating a Lasso model regressing the outcome on all collected pre-treatment variables. We then included those variables that had a non-zero coefficient in our final estimation. The bottom panel indicates the coefficients from these models. The dependent variable is standardized (mean = 0, sd = 1). Consequently, all treatment effects can be interpreted as a percentage of a standard deviation.
Importance of Conceding by Treatment Condition: Panel (a) presents the raw means of reported importance of conceding when pooling the uncertain and certain information treatments. Panel (b) presents the raw means of all three treatment conditions. Above each comparison is the p-value resulting from an Ordinary Least Squares model in which the covariates were selected via a preceding Lasso model. Panel (c) presents the standardized coefficients and 95 per cent confidence intervals from these two regressions. N = 1,486. All models were estimated with robust standard errors. Missing observations of covariates are replaced with the mean or median value.
Panel (a) of Fig. 2 illustrates a statistically significant difference when we compare the control condition with the pooled information treatments. Providing either certain or uncertain information about the consequences of breaching the debt limit increases support for concessions by 27 per cent of a standard deviation. Panel (b) illustrates the effect of two treatment conditions and the difference between them. We find that both conditions (certain and uncertain) increase support for concessions. Consistent with expectations, the ‘certain’ condition has a larger effect (35 per cent vs. 21 per cent of an S.D.). Importantly, the bottom-most coefficient in panel (c) indicates the difference between the two is significant at the 0.05 level. This provides evidence in support of Hypothesis 2.
Given that we ask different outcome questions conditional on partisanship, we compare treatment groups among those that identify as strong or not strong Democrats (N = 685) and Republicans (N = 332) in Fig. 3. Across all conditions, we see that Democrats are more likely to support spending cuts than Republicans are willing to support a clean debt ceiling increase. Importantly, we find that each information treatment has a significant effect compared to the control condition in each subgroup. The certain and uncertain condition increases support for concessions by 51 per cent and 22 per cent of S.D. among Democrats and 37 per cent and 21 per cent of an S.D. among Republicans. The difference between the certain and uncertain treatments is significant among Democrats in the expected direction. However, among Republicans, the uncertain condition is surprisingly higher but not statistically different from the certain condition. We refrain from speculating about the possible reasons for this anomalous insignificant difference given the smaller sample of Republicans.Footnote 14
Importance of Conceding Among Republicans and Democrats: Each of the top panels (a and b) presents the unadjusted means of the reported importance of conceding across the three treatment conditions for Democrats or Republicans. Above each comparison is the p-value resulting from an Ordinary Least Squares model in which the covariates were selected via a preceding Lasso model. Respondents are considered Democrats (N = 685) and Republicans (N = 332) if they responded that they are strong or not strong Democrats or Republicans on a traditional party identification two-question survey module. The bottom panels (c and d) present the coefficients and 95 per cent confidence intervals from the two regressions for each sub-sample estimating either the effect of any information or the separate effects of certain and uncertain information. All models were estimated with robust standard errors. Missing observations of covariates are replaced with the mean or median value.
The results are consistent with our hypotheses. First, information about the consequences of a debt ceiling breach increases the perceived importance of increasing the debt limit with concessions. Further, the effect holds across both parties, albeit with varying strengths. This suggests that political support for brinkmanship in this context is partially a product of voter ignorance. The significant difference between treatments suggests that voter attitudes are a product of both a lack of awareness of the likely dire consequences of sovereign default and a product of the uncertain context. It is important to point out that these are not negligible effects. As Coppock (Reference Coppock2023) notes, informational treatments generally have small effects of around 10 per cent of a standard deviation. Among Democrats, at least, we find results 5x greater than ‘traditional’ informational treatments. Lastly, we report in the Supplementary Appendix that our information treatment influenced reported willingness to sign a petition for a ‘clean’ debt ceiling increase without conditions. Treated respondents were 7 per cent more likely to express interest in signing a petition.
Conclusion
The use of the debt ceiling to renegotiate fiscal policy under the threat of financial ruin often dumbfounds observers and has questioned the underlying faith of credit markets in the US. In this article, we have attempted to understand the dynamics behind this fraught policy and understand why agreement is so hard to find. To do so, we examine the mass politics that serves as the basis of political support and find that ignorance and uncertainty about the consequences of a debt ceiling breach help explain the lack of public support for compromise. This is an important piece of the puzzle but is not sufficient to understand this perplexing feature of US politics. More research is needed to examine if politicians suffer the same informational deficits and which dynamics prevent them from clearly conveying the consequences to constituents and finding agreement amongst themselves.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123424000462.
Data availability statement
Replication Data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/K6MKOS.
Acknowledgements
The authors thank Kathleen J. Brown, Alessia Aspide and the anonymous reviewers for their constructive feedback.
Financial support
This work is funded by ERC Horizon2020 Starting Grant #852334.
Competing interests
The author(s) declare none.
Ethical standards
This study is part of a larger project approved by the Leiden Faculty of Social Sciences Ethics Review Committee Social Sciences. Participants were compensated above the minimum wage equivalent in the United States.
Open access
