Introduction
Amid concerns over political violence and other risky behaviors, an emerging body of evidence suggests some individuals express a desire to ‘watch the world burn’ (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021, p. 2). According to this literature, those who perceive themselves as losing out from society and the political system have a stronger ‘mindset to gain status’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489) and are therefore more willing to engage in these behaviors, independent of political identity and beliefs (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023; Landry et al., Reference Landry, Druckman and Willer2024). The ‘Need for Chaos’ seemingly captures much of the discontent apparent in modern politics (see, e.g., Thompson, Reference Thompson2024), underscoring the need to further understand those motivated to ‘burn it all down’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489).
Yet, we do not know whether the expressed willingness to engage in these acts, as measured in previous studies on Need for Chaos, translates into actual behavior (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023; see also Westwood et al., Reference Westwood, Grimmer, Tyler and Nall2022). Few studies have examined whether Need for Chaos shapes real-world contexts of decision making, such that we can better understand the potential motivations for, and limits of, the desire to ‘burn it all down’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489). Previous evidence further suggests Need for Chaos is shared by ‘a divergent set of malcontents’, not all of whom endorse violence or ‘want destruction for the sake of destruction’ (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021, p. 1; 14). Thus, in this paper, we ask whether this ‘mindset to gain’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489) moderates decision making about salient political events, and whether the relationship between Need for Chaos and preference for risky outcomes depends on how choices are framed.
Perhaps the most widely accepted and validated explanation for decision making under risk, prospect theory emphasizes that most people are generally risk-averse, but losses ‘loom larger’ than gains (Tversky and Kahneman, Reference Tversky and Kahneman1981, p. 456). In turn, people often reverse preferences for objectively equivalent options, such that they are risk-averse toward choices framed as prospective gains but risk-seeking toward choices framed as prospective losses (Kahneman and Tversky, Reference Kahneman and Tversky1979; Tversky and Kahneman, Reference Tversky and Kahneman1981). However, personality and individual differences can moderate and even reverse framing effects (Kahneman and Tversky, Reference Kahneman and Tversky1979; Tversky and Kahneman, Reference Tversky and Kahneman1981; Slovic et al., Reference Slovic, Finucane, Peters and MacGregor2004; Chong and Druckman, Reference Chong and Druckman2007; Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024), particularly among individuals who situate prospective outcomes in the context of what they already know. In turn, we argue that whether individuals high in Need for Chaos perceive themselves as having ‘more to gain’ or ‘less to lose’ depends on the perceived consequences of a decision, i.e., how choices are framed.
Using a simple framing experiment based on prospect theory (Tversky and Kahneman, Reference Tversky and Kahneman1981), we test whether Need for Chaos – a previously validated measure representing one’s context-specific discontent with society and the political system (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023) – moderates framing effects. Results from two studies of US adults (n = 2,704; n = 3,002), fielded on the heels of salient funding negotiations in the US Congress in 2023, suggest that while most respondents are risk-averse toward potential gains but risk-seeking toward potential losses, those high in Need for Chaos exhibit the opposite pattern, seeking risk when anticipating gains and avoiding risk when anticipating losses. These results suggest that individuals high in Need for Chaos may differently interpret gains and losses and adjust their risk preferences based on the perceived consequences of their decision.
These patterns are consistent with prospect theory’s fourfold pattern of decision making (Tversky and Kahneman, Reference Tversky and Kahneman1992) and evidence indicating Need for Chaos is a context-specific adaptation, rather than merely a combination of dark or otherwise risk-oriented personality traits (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023). That those high in Need for Chaos systematically reversed preferences also adds to evidence indicating that support for political violence is often overstated when measured in abstract or ambiguous terms (Westwood et al., Reference Westwood, Grimmer, Tyler and Nall2022). The implication of our findings is that while we should not discount the significance of broader discontent with existing social and political systems, we cannot predict how certain individuals will evaluate and respond to information based on expressed sentiment alone.
Need for Chaos and decision making under risk
Prospect theory contrasts with expected utility theory to emphasize the importance of gain and loss frames, positing that decision making reflects a two-stage process in which choices are framed, then individuals evaluate choices based on that frame Kahneman and Tversky, Reference Kahneman and Tversky1979; Tversky and Kahneman, Reference Tversky and Kahneman1981). Because most people evaluate choices based on the direct consequences of an outcome and losses tend to ‘loom larger’ than gains, people are typically risk-averse toward prospective gains but risk-seeking toward prospective losses (Tversky and Kahneman, Reference Tversky and Kahneman1981, p. 456).
However, one’s response to a given frame can depend on context (e.g., incentives and the perceived probability of an event, see Tversky and Kahneman, Reference Tversky and Kahneman1992; Laury and Holt, Reference Laury and Holt2005) and ‘the norms, habits, and personal characteristics of the decision-maker’ (Tversky and Kahneman, Reference Tversky and Kahneman1992, p. 453). For example, how one perceives and responds to prospective gains or losses can depend on one’s familiarity with certain types of information (Ansolabehere et al., Reference Ansolabehere, Meredith and Snowberg2013), their ability to discriminate between choice options (Fetherstonhaugh et al., Reference Fetherstonhaugh, Slovic, Johnson and Friedrich1997), motivation to consider outcomes (Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024), and whether choice frames evoke emotions consistent with one’s existing attitudes, identity, and beliefs (see, e.g., Gross and D’Ambrosio, Reference Gross and D’Ambrosio2004; Kahan et al., Reference Kahan, Peters, Dawson and Slovic2017).
We propose that another potential moderator is Need for Chaos: a combination of ‘status-oriented personality traits’ and ‘contexts of marginalization’ among those who are ‘disaffected with current society and their (perceived) status in it’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 2; 6). Evidence thus far suggests Need for Chaos is positively associated with psychological traits like the Dark Triad (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021) and willingness to engage in risky behaviors, e.g., conflict-oriented decision making and illegal protests (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021), acts of partisan violence (Landry et al., Reference Landry, Druckman and Willer2024), and sharing conspiracy theories online (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023). Moreover, those high in Need for Chaos are more willing to engage in these acts independent of political identity (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023) and beliefs (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023) suggesting they may approach decision making differently than those low in Need for Chaos. As such, we anticipate that Need for Chaos will moderate framing effects.
More specifically, we would expect these individuals to be less risk averse (or more risk-seeking) than those low in Need for Chaos, when presented with choices framed as prospective gains. Prior work finds that the mere ‘image[] of winning’ (Slovic et al., Reference Slovic, Finucane, Peters and MacGregor2004, p. 316) can evoke a more affective mode of information processing, motivating riskier choices even when one knows the probability of a better outcome is low (see also Denes-Raj and Epstein, Reference Denes-Raj and Epstein1994). Tversky and Kahneman (Reference Tversky and Kahneman1992) also proposed that those who consider previous losses may respond differently to choice frames, and that they may interpret gain-framed options as the choice between outcomes that are as good as, or better than, their current state. If it is the case that those higher in Need for Chaos perceive themselves as already ‘losing out’ from society and the political system, and these individuals consider these losses when making certain decisions, then we would expect them to have a more affective response toward the prospect of winning. In other words, we would expect their ‘mindset to gain’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489) to moderate framing effects, such that they prefer risky, rather than certain, options when choices are positively framed.
But what might we expect when those with greater Need for Chaos are faced with options framed as prospective losses? Importantly, the assumption that these individuals would prefer riskier outcomes when presented with the possibility of winning does not account for why they might seek chaos in the first place – a question that, we argue, is essential to understanding how they respond to loss frames. Research in social psychology consistently finds that how people perceive and respond to information depends on their motivations (Chen et al., Reference Chen, Duckworth and Chaiken1999), e.g., whether they seek to reaffirm existing attitudes or achieve a specific goal (Kunda, Reference Kunda1990). Thus, if the motivation among those high in Need for Chaos is to ‘burn it all down’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489), then we might expect them to prefer the outcome most consistent with those destructive goals – in this case, the risky option, regardless of frame.
However, this assumption loses sight of evidence indicating that Need for Chaos is a personality adaptation (i.e., status-oriented traits that materialize relative to perceived marginalization, see Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), which would suggest that there are specific contexts in which the desire to ‘burn it all down’ would emerge. In turn, it may be that whether these individuals make risky decisions depends on not only whether they consider previous losses, but also whether the framing of choice outcomes influences their expected success. Extant literature on prospect theory further suggests that salient losses can increase one’s sensitivity to possibility, reversing the expected response to framing effects (Tversky and Kahneman, Reference Tversky and Kahneman1992). In this case, those who consider outcomes in the context of salient losses may prefer risky, rather than certain, outcomes when deciding between choices framed as prospective gains; conversely, they may prefer certain, rather than risky, outcomes when deciding between choices framed as prospective losses (Tversky and Kahneman, Reference Tversky and Kahneman1992). We show this fourfold pattern of decision making in Table 1.
Prospect theory’s fourfold pattern of decision making
Note: We adapted Table 1 from Oliver (Reference Oliver2018).
Furthermore, just as the mere ‘image[] of winning’ (Slovic et al., Reference Slovic, Finucane, Peters and MacGregor2004, p. 316) can evoke a more affective mode of information processing and decrease risk aversion (Denes-Raj and Epstein, Reference Denes-Raj and Epstein1994), considering previous losses can increase the perceived stakes of a decision and increase risk aversion (Tversky and Kahneman, Reference Tversky and Kahneman1981). Previous research on Need for Chaos asked respondents about their willingness to share conspiracy theories ‘[a]ssuming you could choose the people who saw it’ (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), their likelihood of taking part in illegal protests (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021), and their likelihood of engaging in politically violent behaviors, should the ‘other party’s nominee win[] a contested 2024 presidential election’ (Landry et al., Reference Landry, Druckman and Willer2024, p. 9). However, none of these scenarios provided respondents with information about the potential consequences of these acts, nor did they consider how individuals might evaluate outcomes in the context of previous loss. Thus, just as we might expect those high in Need for Chaos to perceive gain-oriented options as the choice between outcomes as good as, or better than, their current state, they may perceive loss-oriented options as the choice between outcomes as good as, or worse than, their current state (Tversky and Kahneman, Reference Tversky and Kahneman1992). As such, it may be that those high in Need for Chaos are risk-averse, rather than risk-seeking, when the choices are negatively framed.
To reiterate, we argue that if the contexts and characteristics captured by Need for Chaos influence one’s approach to decision making, then Need for Chaos should moderate framing effects. Given their ‘mindset to gain status’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023, p. 1489), we expect those high in Need for Chaos to be less risk-averse (or more risk-seeking) when deciding between options framed as prospective gains (H1). Although we might expect these individuals to prefer the risky option regardless of frame, previous evidence suggests Need for Chaos emerges only in specific contexts (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023) and that considering salient losses can reverse the expected framing effects (Tversky and Kahneman, Reference Tversky and Kahneman1992). Therefore, we also expect individuals high in Need for Chaos to be more risk-averse (or less risk-seeking) when deciding between options framed as prospective losses (H2).
The 2023 debt ceiling and government shutdown debates in US Congress
We test our hypotheses with data from two studies, both of which were fielded on the heels of two salient and contentious funding negotiations in the US Congress. Economic outcomes consistently rank high on Americans’ list of policy priorities (see, e.g., Linde and Vis, Reference Linde and Vis2016) and perceptions of the economy are endogenous to politics (Enns et al., Reference Enns, Kellstedt and McAvoy2012). Therefore, while not every citizen is equally focused on the economy (Singer, Reference Singer2011), nor politics in general (Miller et al., Reference Miller, Peterson, Saunders and McClurg2023), using frames focused on real-world, economic contexts of decision making should help us to examine how people respond to frames when material issues are at stake. We expect economic decisions to be particularly relevant for individuals high in Need for Chaos, whose status-seeking orientations and perceptions of marginalization may influence how they respond to these gain and loss frames (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023).
For context, we plot the salience of the 2023 US debt limit and government shutdown debates (over a one-year period, limited to searches in the US) using Google Trends data in Figure 1. Albeit an indirect measure, it helps to demonstrate what we might expect regarding issue salience over time: The first spike in popularity for the US debt ceiling issue corresponds with when the US government hit its debt limit in January 2023; the peak spike in popularity for this issue corresponds with congressional negotiations and the subsequent funding bill in May 2023. Similarly, the US government shutdown reached its peak popularity during congressional negotiations and the subsequent funding bill in late September 2023. Both studies were fielded around these peaks (indicated by the vertical, dashed lines in late May for Study 1 and late September for Study 2), suggesting these data have the potential to capture periods when these issues were of greater interest among the American public.
2023 US Google search trends for the US debt ceiling and US government shutdown.
Study 1: May 2023 debt ceiling debate in US Congress
Data for Study 1 include a sample of US adults (n = 3,463) obtained via Lucid, an online survey provider used in similar research (Guay and Johnston, Reference Guay and Johnston2022; Kreps and Kriner, Reference Kreps and Kriner2022; Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023). Lucid provides survey respondents with proprietary compensation and provides clients with samples aligned with US Census demographic benchmarks (Coppock and McClellan, Reference Coppock and McClellan2019). We launched Study 1 on 27 May 2023, shortly after the announcement that the White House and US Congress reached a deal to raise the federal debt ceiling. The study concluded on 28 May 2023, prior to the bill’s release (see also Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024).
We designed Study 1 such that respondents first received a series of demographic questions (i.e., age, income, race, ethnicity, gender, and religiosity), followed by a series of randomized questions assessing Need for Chaos, objective numeracy, partisanship, political knowledge, need for closure, need for cognition, trust in government, interest, authoritarianism, and Big Five personality traits. Respondents who failed a randomized attention check (n = 650) were automatically eliminated from the analysis. After accounting for other missed values, Study 1 yielded a total sample of n = 2,704. Study 1 respondents then received the following prompt:
As you know, the U.S. government has reached its debt limit and is at risk of defaulting on its debt as of 5 June 2023. The White House and GOP negotiators have reached a compromise to raise the debt ceiling and avoid default, but Congress has yet to vote on this deal and the economic impact remains unknown.
Imagine that raising the debt ceiling is expected to affect 6 million jobs and you are tasked with deciding between two alternative deals. Which would you choose? (Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024, p. 3).
One half of Study 1 respondents were randomly assigned to receive choices framed as potential gains; the other half of respondents received choices framed as potential losses:
[Jobs Gained Frame] If Deal A is chosen, 2 million jobs will be preserved; If Deal B is chosen, there is a ⅓ probability that 6 million jobs will be preserved and ⅔ probability that no jobs will be preserved.
[Jobs Lost Frame] If Deal A is chosen, 4 million jobs will be lost; If Deal B is chosen, there is a ⅓ probability that no jobs will be lost and ⅔ probability that
6 million jobs will be lost. (Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024, p. 3).
Study 2: September 2023 government shutdown debate in US Congress
We designed Study 2 like Study 1.Footnote 1 We launched Study 2 on 30 September 2023, shortly after Congress passed a funding bill to avert the government shutdown and concluded Study 2 on 6 October 2023. We used Qualtrics to collect a sample of US adults (n = 3,100) via Prodege, another online survey provider used in similar research (Benoit et al., Reference Benoit, Conway, Lauderdale, Laver and Mikhaylov2016; Chen et al., Reference Chen, Kteily and Ho2018) that provides respondents with proprietary compensation in exchange for their survey participation and clients with survey samples that are balanced with the US Census (ESOMAR, 2023).
As in Study 1, we first asked respondents a series of demographic questions, followed by attention check question. Those who answered the attention check question incorrectly were automatically eliminated from the survey.Footnote 2 After accounting for other missed values, Study 1 yielded a total sample of n = 3,002. Respondents then received a series of randomized questions on Need for Chaos, as well as the social, political, and psychological measures included in Study 1.Footnote 3 At the end of the survey, all Study 2 respondents received the following prompt and were randomly assigned to receive choices framed as either ‘jobs gained’ or ‘jobs lost’. While we adapted the prompt such that it referred to the September 2023 government shutdown debate in US Congress, respondents (who were randomly assigned to receive one of two choice frames) received ‘jobs gained’ or ‘jobs lost’ choices identical to those presented in Study 1:
As you know, the U.S. government’s funding is set to expire on 1 October 2023, upon which the federal government is at risk of a shutdown. Congress has yet to reach a funding agreement that would avoid a federal government shutdown and the economic impact of a government shutdown remains unknown.
Imagine that a federal government shutdown is expected to affect 6 million jobs and you are tasked with deciding between two alternative deals. Which would you choose?
[Jobs Gained Frame] If Deal A is chosen, 2 million jobs will be preserved; If Deal B is chosen, there is a ⅓ probability that 6 million jobs will be preserved and ⅔ probability that no jobs will be preserved.
[Jobs Lost Frame] If Deal A is chosen, 4 million jobs will be lost; If Deal B is chosen, there is a ⅓ probability that no jobs will be lost and ⅔ probability that 6 million jobs will be lost.Footnote 4, Footnote 5
What Is Need for Chaos?
Given the recency of existing research on Need for Chaos, it is important to more closely examine this measure and its relationship with other covariates before proceeding to our main results. Consistent with extant literature (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021; Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), respondents in both samples expressed the most agreement with statements on discontent with social and political institutions (Q4 and Q5) and the least agreement with the statement on general destruction (Q7; see Table 2), but item and factor analyses indicate all seven items fit together into a unidimensional Need for Chaos scale.Footnote 6 We illustrate the distribution of the seven-item Need for Chaos scale (recoded to range from 0 to 1) for Study 1 (M = 0.17, Mdn = 0.10, SD = 0.20) and Study 2 (M = 0.18, Mdn = 0.10, SD = 0.21) in Figure 2. Similar to previous findings (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021; Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), approximately 26 per cent and 27 per cent of respondents (in Study 1 and Study 2, respectively) had a combined Need for Chaos score of zero, indicating those who strongly disagreed with all statements in the Need for Chaos battery. Considering the majority of each sample had a non-zero Need for Chaos score, we take this as more evidence to suggest that many individuals do not agree with chaotic sentiments for chaos’s sake; rather, they may harbor a more diffuse dissatisfaction with the current state of society and the political system.
Distribution of the seven-item Need for Chaos scale in Study 1 and Study 2.
Items in Need for Chaos scale
Note: Table values include the mean (M), median (Mdn), minimum (Min), and maximum (Max) values, as well as the factor loadings (FL) for each item. For each statement, respondents were asked to indicate, ‘How much do you disagree or agree with the following statements?’ with response options on a seven-point scale ranging from Strongly disagree to Strongly agree.
We also regressed Need for Chaos on covariates included in previous studies (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), as well as measures previously shown to moderate how people process information and respond to decision frames. Similar to previous literature (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023; Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), data from both studies suggest Need for Chaos is negatively associated with Age, Female gender, Conscientiousness, and Agreeableness. Need for Chaos is also positively associated with Openness, Need for Cognition, and Need for Closure. In line with prior evidence indicating Need for Chaos is higher among marginalized individuals (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023), we find White race is negatively associated with Need for Chaos in Study 1. Like Farhart et al. (Reference Farhart, Fitz, Miller and Saunders2023), we also find Religiosity is positively associated with Need for Chaos in Study 2. Both studies also demonstrate that Need for Chaos is negatively associated with objective Numeracy and Political Knowledge, both of which are positively associated with greater susceptibility to gain and loss frames (Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024).
Though not shown in Figure 3, we also tested whether other select covariates are associated with Need for Chaos.Footnote 7 Study 1 included a measure of general Risk Orientation (i.e., one’s general tolerance for risk, see Kam and Simas, Reference Kam and Simas2010); Study 2 included measures of perceived Social Status (Adler et al., Reference Adler, Epel, Castallazzo and Ickovics2000), Cognitive Reflection (Frederick, Reference Frederick2005), and White ID (i.e., one’s attitudes regarding the importance of their white racial identity, see Jardina, Reference Jardina2019). Adding these covariates to models for Study 1 and Study 2 indicate Need for Chaos is positively associated with Risk Orientation and Cognitive Reflection. We find no significant relationship between Need for Chaos and Social Status or White ID.Footnote 8
Correlates of Need for Chaos. Horizontal bands represent 95 per cent confidence intervals, two-tailed tests.
As for how we interpret these initial results, the negative relationships between Need for Chaos and Age, Female gender, Agreeableness, and Conscientiousness are consistent with previous scholarship associating status seeking with ‘young male syndrome’ (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021, p. 7; see also Wilson and Daly, Reference Wilson and Daly1985). They are also consistent with evidence linking ‘undercontrolled’ personality types to externalizing tendencies, e.g., greater interpersonal conflict (Asendorpf et al., Reference Asendorpf, Borkenau, Ostendorf and Van Aken2001; see also Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021). The positive relationship between Need for Chaos and Need for Cognition also supports our assumption that those who take a more inclusive account of previous losses when tasked with decision making under risk and may respond differently to decision frames (Fitz et al., Reference Fitz, Stecuła, Hitt and Saunders2024). Results from both samples further suggest that neither Party ID nor Income are associated with Need for Chaos, echoing previous research demonstrating ‘chaotic motivations can supersede those rooted in partisanship’ (Farhart et al., Reference Farhart, Fitz, Miller and Saunders2023, p. 1) and that those high in Need for Chaos ‘do not need to be deprived in an absolute sense’ (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021, p. 2). Altogether, these findings help to further validate Need for Chaos and its ability to capture whether those disaffected with society and the political system respond differently to gain and loss frames.
Results
Turning to our main analysis, we first examine whether our samples respondents exhibit conventional framing effects, that is, whether they are risk-averse toward potential gains but risk-seeking toward potential losses. As expected, most respondents (67 per cent) randomly assigned to the ‘jobs gained’ frame (M = 0.67, SD = 0.47) chose the ‘certain’ Deal A; most (59 per cent) who received the ‘jobs lost’ frame (M = 0.59, SD = 0.49) selected the ‘risky’ Deal B. The majority of Study 2 respondents (67 per cent) randomly assigned to the ‘jobs gained’ frame (M = 0.67, SD = 0.47) also chose the ‘certain’ Deal A, while the majority (61 per cent) of those who received the ‘jobs lost’ frame (M = 0.61, SD = 0.49) chose the ‘risky’ Deal B.
We provide results from logistic regression models in Table 3. We use Choice of Risky Deal B as the dependent variable (coded 0 for respondents who chose the ‘certain’ Deal A and 1 for respondents who chose the ‘risky’ Deal B, regardless of frame). All models include a binary independent variable indicating which Frame they received (coded 0 for respondents who received the ‘jobs gained’ frame and 1 for respondents who received the ‘jobs lost’ frame) to test whether exposure to loss frames is positively associated with Choice of Risky Deal B. To test whether Need for Chaos moderates the effects we might typically expect because of gain and loss frames, models 2 and 5 also include an interaction term for Frame x Need for Chaos. Here again, Frame is positively associated with Choice of Risky Deal B; we also find that the interaction term for Frame x Need for Chaos is negatively associated with Choice of Risky Deal B. These results hold when controlling for the full range of previously mentioned covariates (shown in models 3 and 6), speaking to the robustness of our results.Footnote 9
Framing effects and Need for Chaos
* Note: B denotes logistic regression coefficients; SE denotes standard errors.p < 0.05, **p < 0.01, ***p < 0.001.
To interpret results for the interaction term, we calculated the predicted probabilities for Frame x Need for Chaos for Study 1 and Study 2 (based on Table 3, Models 3 and 6) in Figure 4. As expected, results from both studies indicate that Need for Chaos is positively associated with Choice of Risky Deal B among those presented with the ‘jobs gained’ frame, but negatively associated with Choice of Risky Deal B among those presented with the ‘jobs lost’ frame. Thus, in line with our hypotheses and prospect theory’s fourfold pattern of decision making, this suggests those with high Need for Chaos are more risk-seeking, rather than risk-averse, toward potential gains (H1), but more risk-averse, rather than risk-seeking, toward potential losses (H2).
Predicted probabilities for Frame x Need for Chaos in Study 1 and Study 2. Corresponds with results from Table 3, Models 3 and 6. Shaded areas are 95 per cent confidence intervals, two-tailed tests.
Given the overlap in confidence intervals at higher levels of Need for Chaos, one might question whether these individuals are systematically responding to each frame or disregarding the frames altogether. As previously discussed, prospect theory contrasts with rational choice models to posit that individuals often express inconsistent preferences for objectively equivalent options, depending on how choices are framed; however, personality and other individual differences can moderate, and in some cases reverse, framing effects (Tversky and Kahneman, Reference Tversky and Kahneman1981, Reference Tversky and Kahneman1992). Whereas a lack of response to framing would suggest these individuals have diminished sensitivity to the manipulation (Chong and Druckman, Reference Chong and Druckman2007), a stronger or reversed response would suggest they remain sensitive to, but interpret differently, gain and loss frames (Tversky and Kahneman, Reference Tversky and Kahneman1981, Reference Tversky and Kahneman1992). Importantly, the latter would align with our theory, whereas the former would be consistent with the idea that Need for Chaos represents a broader desire to disrupt the status quo.
To test this empirically, we conducted a series of post-estimation tests that compare the predicted probabilities of choosing the risky option across frames and levels of Need for Chaos. Results from both studies indicate that those with low-to-moderate Need for Chaos scores (e.g., Need for Chaos = 0.00–0.30) exhibited the classic framing effect posited by prospect theory: More specifically, they had a significantly higher probability of choosing the risky option when presented with choice options in the loss frame (p = 0.45 and p = 0.004 when Need for Chaos = 0.50 in Study 1 and Study 2, respectively). In line with our main results, the pattern reverses at higher values of Need for Chaos, indicating these individuals had a lower probability of choosing the risky option when presented with options in the loss frame. Although differences among those with moderately high Need for Chaos (e.g., Need for Chaos = 0.60–0.80) were non-significant, we found significant differences among those with high Need for Chaos scores (e.g., Need for Chaos = 0.90–1.00). For example, the predicted probabilities of choosing the risky outcome when presented with choices in the loss frame were 20 (p = 0.011) and 19 (p = 0.008) percentage points lower for those highest in Need for Chaos (i.e., when Need for Chaos = 1.00) in Study 1 and Study 2, respectively. Together, these patterns help to reinforce our hypotheses that individuals high in Need for Chaos are not merely indifferent to framing; rather, they evaluate and systematically respond differently to gain and loss frames (H1 and H2).
Discussion and conclusion
Using original data collected during the May 2023 US debt ceiling and September 2023 US government shutdown debates in the US Congress, we test whether Need for Chaos moderates decision making in response to gain and loss frames. Consistent with fourfold pattern of decision making elucidated by prospect theory (Kahneman and Tversky, Reference Kahneman and Tversky1979; Tversky and Kahneman, Reference Tversky and Kahneman1981, Reference Tversky and Kahneman1992), we find most respondents are risk-averse when presented with choices framed as prospective gains, but risk-seeking when presented with choices framed as prospective losses. Conversely, Need for Chaos moderates framing effects, such that those high in Need for Chaos are risk-seeking toward prospective gains but risk-averse toward prospective losses. Our results are consistent across both studies, providing strong evidence to suggest those high in Need for Chaos respond differently to strategic frames.
Still, a few caveats are in order. First, although we interpret our findings as evidence of context-dependent sensitivity to frames, we do not directly measure the mechanism(s) underlying this effect. Second, the non-significant and narrower framing effects observed for those with moderate to high Need for Chaos could reflect not only transitional reasoning or measurement error (see, e.g., Ansolabehere and Hersh, Reference Ansolabehere and Hersh2012), but also meaningful variation among this group (see, e.g., Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021). Lastly, only a small proportion of respondents in our samples expressed sufficiently high Need for Chaos scores to exhibit reversed framing effects, raising important questions about the prevalence and behaviors of these individuals in the broader population. However, we argue that any number of respondents who express these sentiments – and who systematically respond to frames in ways that are opposite from most others – should be cause for concern. In turn, we look forward to future research that can better assess whether certain types of chaos-seeking individuals are more sensitive to choice frames, and for whom these expressed sentiments map onto actual behavior.
In any case, our findings are consistent with previous evidence on the reversal of framing effects (Tversky and Kahneman, Reference Tversky and Kahneman1992). Our findings also help to advance our theoretical understanding of Need for Chaos by demonstrating that not all ‘chaotic’ individuals are merely disaffected risk-takers who want to ‘burn it all down’ (Petersen et al., Reference Petersen, Osmundsen and Arceneaux2023; p. 4). Despite its provocative label, they are ‘a divergent set of malcontents’ (Arceneaux et al., Reference Arceneaux, Gravelle, Osmundsen, Petersen, Reifler and Scotto2021, p. 1) who, like others influenced by strategic frames, make decisions that reflect both the information provided and the ‘circumstances and the environment that people find themselves in’ (Oliver, Reference Oliver2024; p. 183). Thus, while we in no way aim to justify the sentiments or actions of those with nefarious intent, we do agree with others (e.g., Oliver, Reference Oliver2018, Reference Oliver2024) who emphasized that assumptions based on a singular conceptualization of rationality ignores heterogeneous preferences. Given more than two-thirds of respondents in both samples had a non-zero Need for Chaos score, the implication is that researchers should direct greater attention to what is a gain and loss, in which contexts, and for whom.
These findings also have broader implications for elite decision making. Evidence to suggest Need for Chaos is both trait- and state-specific, and that those high in Need for Chaos respond differently to strategic frames, suggests discontent with existing systems may be particularly influential on one’s preferences in contentious political environments. This presents a unique set of challenges for elites looking to gain or maintain political support, as they need to not only develop strategic messaging, but also anticipate how various segments of their target audience will interpret that messaging and respond. Attempts to appeal to various groups with different frames may prove ineffective, and even costly, as targeted messaging can be seen as less credible among those who do not identify with the targeted group (see, e.g., Hersh and Schaffner, Reference Hersh and Schaffner2013). Thus, we look forward to scholarship that continues to examine the causes and consequences of Need for Chaos, and that which investigates whether (and which) elites shift their approach to strategic messaging over time. For example, following Kamala Harris’s loss to Donald Trump in November 2024, Harris–Walz campaign officials expressed frustration with the strategy of testing and using only the highest-scoring political ads, regardless of whether ad messages aligned with each other or the broader campaign. As one official concluded, ‘You can’t get to a storyline if you’re optimizing for each individual product’ (Stuart, Reference Stuart2024).
To that end, we also look forward to research that continues to investigate the stability and downstream effects of Need for Chaos, and whether Need for Chaos moderates framing effects in other social and political contexts. Assessing whether Need for Chaos moderates policy frames on abortion, for example, may be worthwhile considering who tends to score higher in this measure (i.e., those who are younger, white, and male), as well as links between attitudes on abortion and those toward historically marginalized groups (Deckman et al., Reference Deckman, Elder, Greene and Lizotte2023). Regardless, we suspect that while Need for Chaos may help to signal a particular sentiment of disaffection, many of these individuals – when tasked with making judgments and decisions on matters of real-world importance – will focus more on what they might have to gain or lose, as opposed to lofty goals aimed at dismantling society and the political system. As we continue to broaden our understanding of Need for Chaos amidst the current social and political climate, our hope is that these objectives are not one and the same.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/bpp.2025.10022.
Acknowledgements
The authors thank the anonymous reviewers, editors of Behavioural Public Policy, Nicholas Davis, Christopher Federico, Kathryn Haglin, Kristina LaPlant, Joanne Miller, and participants of the Psychological Needs, Cynicism, and Politics panel at the 2024 Midwest Political Science Association Annual Conference for their helpful feedback in improving this project.
Funding statement
The authors received no support for the design, execution, analysis and interpretation of data, or writing of this study.
Competing interests
The authors declare none.
Data statement
All data and code required to replicate these analyses are located at https://doi.org/10.7910/DVN/4WXM3B.
Open access
