Introduction
Over the past few election cycles, there has been growing interest in the concept of political anxiety. The idea that politics is a significant, and perhaps increasing, source of anxiety has captured the attention of psychologists, political scientists, media outlets, universities, nonprofit organizations, medical providers, and professional associations.Footnote 1 For example, in October 2020 the American Psychiatric Association released survey data indicating that 72% of Americans reported being extremely or somewhat anxious about the upcoming presidential election, 61% were extremely or somewhat anxious about the impact of politics on daily life, and 51% were extremely or somewhat anxious about discussions about the 2020 election at work or in their personal life.Footnote 2 Around the same time, the Mayo Clinic released an article entitled “Is election stress disorder real?” In it, they noted that “Election stress disorder isn’t a scientific diagnosis, but the concept is real….It’s an experience of overwhelming anxiety that can manifest in a number of ways” (Howland, Reference Howland2020). According to Robert Bright, a Mayo Clinic psychiatrist featured in the article, “We notice it [election stress] in our bodies, the tension in our shoulders. Sometimes people get GI (gastrointestinal) upset or headaches. People have trouble sleeping. There’s a lot of sleep disturbance going on right now—tossing, turning, and worrying, and not being able to get to sleep—or having bad dreams about the election” (qtd. in Howland, Reference Howland2020).
Although measures of generalized anxiety have existed for quite some time (e.g., Crocq, Reference Crocq2017; Kroenke et al., Reference Kroenke, Spitzer, Williams, Monahan and Löwe2007; Löwe et al., Reference Löwe, Decker, Müller, Brähler, Schellberg, Herzog and Herzberg2008; Spitzer et al., Reference Spitzer, Kroenke, Williams and Löwe2006), researchers have only just begun to explicitly measure political anxiety.Footnote 3 In a recent article, Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) developed and validated a general measure designed to capture anxiety specifically attributed to politics. Their measure is based on eight items that index how a range of features and situations commonly associated with the contemporary political environment (e.g., polarization and conflict, caring too much about politics, etc.) make people feel anxious. This measure performed well psychometrically; it was found to be reliable (α = .89 and α = .91 in two different survey waves) and valid (i.e., individual-level changes in political anxiety were related to changes in the political environment). Though this suggests political anxiety is quantifiable, it does not necessarily imply that it is a concept independent of general anxiety. Political anxiety may be just one symptom of general anxiety; in other words, those reporting anxiety tied to politics may be the same individuals who would already score higher on measures of general anxiety. Generalized anxiety is typically characterized as “chronic free-floating anxiety accompanied with anxious apprehension or worry about many circumstances of daily life” (Crocq, Reference Crocq2017, p. 107, emphasis added). Politics may simply be one of those many circumstances.
An alternate possibility is that political anxiety is a unique concept, i.e., that politics invokes anxiety even in those who are not experiencing anxiety more generally. Politics is conceptualized here as an environmental source of state-anxiety, something that repeatedly triggers transitory emotional responses characterized by feelings of worry or even fear, even in those who are low in trait anxiety. Indeed, this concept of anxiety appears core to the concept of election stress disorder discussed above. That definition explicitly suggests that politics is a uniquely salient source of worry and uncertainty, one that induces anxiety even in those who would not score highly on standard batteries of general anxiety, particularly around highly salient events such as elections. So, is political anxiety simply a symptom of general anxiety, or is it conceptually distinct?
In this article, we address this question by using data from a university-sponsored survey conducted in October 2023. On the survey, we were able to include the Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) political anxiety battery along with four questions from Spitzer et al. (Reference Spitzer, Kroenke, Williams and Löwe2006) that measure generalized anxiety (these constitute a shortened version of the widely employed General Anxiety Disorder 7 instrument, which is used to capture the DSM-V criteria for General Anxiety Disorder). The survey also included various demographic measures that enable us to examine similarities and differences in the correlates of political and generalized anxiety.
Methods
Participants
We make use of data collected as part of a university-sponsored survey [University of Wisconsin-Green Bay public opinion poll] that was fielded between October 16 and October 21, 2023.Footnote 4 In total, our sample contains 451 Wisconsin residents who completed the survey online. To recruit respondents, we used Dynata, a well-known and widely used research firm that specializes in online panels. Although online panels are different than random samples, we examined the representativeness of our sample by comparing various demographic measures in our data set to US Census estimates for Wisconsin.Footnote 5 In general, the demographic composition of our sample appears to mirror the state’s demographic profile fairly well, though we note that our sample is slightly more educated and older than the state as a whole. After informed consent, we required respondents to complete a CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart), which is designed to prevent bots from taking the survey. In addition, early on in the survey, we included an attention check where respondents were asked to select a particular response to make sure they were paying attention.Footnote 6 If a respondent did not answer the question correctly, the survey was terminated, and their responses were not used. We also included a second attention check toward the end of the survey.Footnote 7 Overall, 97% of respondents passed this attention check. In the analyses that follow, we exclude the 14 respondents who did not answer this attention check question correctly. The survey was designed to be fairly short and took respondents an average of 9.16 minutes to complete.
Measures
Political anxiety
As noted above, we included Smith et al.’ (Reference Smith, Weinschenk and Panagopoulos2023) political anxiety battery. Respondents were asked: “How much anxiety does each of the following give you: (1) The election of a disliked candidate or political party, (2) The level of polarization and conflict in the current political climate, (3) That the American public is insufficiently informed about politics, (4) That you care too much about politics, (5) That you are insufficiently informed about politics, (6) The poor quality of political leaders/candidates, (7) The uncivil nature of modern politics, (8) The extent to which ordinary people are disinterested in politics.” Responses to each item were recorded on a 1–10 scale where 1 corresponded to “no anxiety at all” and 10 corresponded to “a great deal of anxiety.” In order to avoid any ordering effects, the list of eight items was presented in random order for each respondent. As noted by Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023), the items were selected to try to tap into anxiety that individuals associate specifically with the contemporaneous political environment. The goal of the measurement battery is to be general enough to be relevant to any political environment (i.e., it avoids items that are tied to a particular set of environmental circumstances, such as a specific politician, piece of legislation, or election). To create an overall measure of political anxiety, we combined the eight items to form an index that ranges from 1 to 10 (M = 6.15, SD = 2.01).
Table 1 reports the descriptive statistics for the anxiety measure items from our Wisconsin sample and from the data used by Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023), which was a two-wave (pre- and post-election) panel that was designed to be nationally representative. The Wisconsin sample is designed to be representative of the state rather than the nation, but the means are in the same ballpark as the Smith et al. numbers. This suggests results from the Wisconsin sample, at least in the context of political anxiety, may be generalized.
Descriptive statistics for political anxiety items
Note: Mean (standard deviation) reported, 2020 data are from Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) and are provided for comparative purposes.
Generalized anxiety
We were also able to include on the survey four questions from Spitzer et al. (Reference Spitzer, Kroenke, Williams and Löwe2006) designed to measure generalized anxiety.Footnote 8 More specifically, we asked respondents “Over the last 2 weeks, how often have you been bothered by the following problems? (1) Feeling nervous, anxious or on edge, (2) Not being able to stop or control worrying, (3) Worrying too much about different things, (4) Trouble relaxing.” For each item, the response categories were as follows: Not at all (coded 0), Several days (coded 1), More than half the days (coded 2), and Nearly every day (coded 3). We note that follow-up research has shown that the Spitzer et al. (Reference Spitzer, Kroenke, Williams and Löwe2006) measure has “strong psychometric properties in terms of reliability and validity” (Rutter & Brown, Reference Rutter and Brown2017, p. 145).Footnote 9 Again, to avoid any ordering effects, the measures were presented in random order for each respondent. To create an overall measure of generalized anxiety, we combined the four items to form an index that ranges from 0 to 3 (M = 1.00, SD = .87).Footnote 10
Results
Factor analysis
We start by confirming that our two anxiety batteries (political and general) have good psychometric properties. The political anxiety items had good internal validity (α = .89) as did the items for general anxiety (α = .92).Footnote 11 Factor analyzing the political anxiety items yielded a single factor with an eigenvalue > 1.0 that explained ~54% of the variance.Footnote 12 A similar analysis of the general anxiety items yielded a single factor with an eigenvalue > 1.0 that explained ~73% of the variance. These results were robust to extraction method (principal components, principal axis, and maximum likelihood approaches all yielded similar results). In short, both the political and general anxiety batteries seem to reflect a single underlying latent construct that captures the concepts these instruments are designed to measure. The key question, however, is whether they are tapping independent concepts or are indexing a single underlying anxiety dimension.
Basic evidence that the concepts are independent comes from the correlation of the additive indexes constructed from the respective survey items. The correlation between our political anxiety and general anxiety indexes is r(437) = .24 (p < .01). That suggests a relationship that is nontrivial but is also far from suggesting that these are two measures tapping into an identical underlying concept. Clearly, the concepts are related, but their ability to predict one another is fairly modest—a correlation of .24 means one of these measures is capable of accounting for ~5% of the variance in the other. Again, not trivial, but limited enough to suggest that these are a long way from two measures of the same thing.
We conducted a more stringent test of the independence of these measures using an exploratory factor analysis (the EFA) of all the items used in both measures. If these measures are tapping into distinguishable concepts the underlying structure of the items should fall into two distinct dimensions, one associated with political anxiety and one with general anxiety. A principal axis factor analysis using an oblique (oblimin) rotation (see Table 2) confirms this expectation, returning two factors with eigenvalues >1.0. The political anxiety items consistently load highly on one of these factors, the general anxiety items on the other. In short, the results presented in Table 2 clearly suggest these two measures are not expressing different measures of a single underlying latent structure. Political and general anxiety, at least as measured by these instruments, clearly seem to represent two distinct empirical concepts.
Factor analysis of political and general anxiety items
Notes: N = 437. The extraction method was principal axis factoring with an oblique (Oblimin with Kaiser Normalization) rotation. Factor loadings <.30 are suppressed. Eigenvalue for factor 1 = 5.1 (SS/ percent of variance = 39.8), for factor 2 = 2.8 (SS/percent of variance = 21).
Correlates of political and general anxiety
If political and general anxiety are distinct empirically it implies, prima facie, that they are not necessarily products of the same social forces, or at least that these forces do not affect these concepts equally. Existing research does suggest there are likely areas of overlap, i.e., that certain social traits will be associated with both political and general anxiety. For example, previous studies suggest females and younger people are more likely to report higher levels of general (Michael et al., Reference Michael, Zetsche and Margraf2007) and political anxiety (Smith, Reference Smith2022). Yet there is at least as much evidence suggesting political anxiety and general anxiety are driven by different types of social stressors. For example, studies of general anxiety suggest racial/ethnic minorities and those socio-economically stressed (e.g., unemployment, low income) are more likely to score higher in general anxiety, while those with higher levels of education report lower levels of anxiety (see Nunes et al., Reference Nunes, Carroll, Mahaffey, Califf, Doraiswamy, Short, Shah, Swope, Williams, Hernandez and Hong2022). This contrasts with studies of political stress and anxiety, which report little relationship with socioeconomic status and, if anything, show racial minorities have lower levels of political anxiety (Smith, Reference Smith2022). The latter studies, not surprisingly, have focused more on political traits, and consistently found political orientation, interest, and engagement to be positive correlates with political anxiety (e.g., Smith, Reference Smith2022; Smith et al., Reference Smith, Weinschenk and Panagopoulos2023). The impact of such political variables on general anxiety is largely unknown, though rates of depression have been reported to be higher among liberals than conservatives (e.g., Gimbrone et al., Reference Gimbrone, Bates, Prins and Keyes2022), and a study by Helminen et al. (Reference Helminen, Elovainio and Jokela2022) found that in Great Britain those with left-leaning attitudes on several issues (e.g., inequality, concern for the environment) had higher anxiety scores than those with right-leaning attitudes.
To investigate the impact of differing social traits on political and general anxiety, we conducted two regression analyses, identical except for the dependent variables. The dependent variables were the additive political and general anxiety indexes described above, and the predictors included in both models were annual household income (a 9-item scale where 1 = <$10,000, 9 = $150,000 or more), race (dummy variable where 1 = White), sex (dummy variable where 1 = male), age (in years), education (a 9-item scale where 1 = no high school, 9 = advanced graduate work), ideology (self-identification on 5-point scale where 1 = very liberal and 5 = very conservative), political interest (5-point scale where respondents indicate how often they pay attention to government and politics, where 1 = never, 5 = all the time), and political knowledge (number correct of three questions total, two of which focused on state politics and one of which focused on national politics). These variables have all been identified by previous research as robust predictors of general and/or political anxiety. Before proceeding, we note that while these variables were included by Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) as predictors of political anxiety, we are not able to exactly replicate their models. Indeed, some of the variables that Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) included in their model (e.g., negative emotionality, vote choice, political participation, etc.) were not measured in the current survey (as noted above, space was limited on the survey, and we therefore had to make decisions about what to include). Even so, the inclusion of many of the measures used by Smith et al. (Reference Smith, Weinschenk and Panagopoulos2023) in our model is valuable as it allows us to compare our findings to previous results on the correlates of political anxiety. We strongly encourage future scholars to replicate the models presented in this article and to build on them by incorporating other theoretically relevant predictors of political anxiety.
In Table 3, we report the standardized betas from these regressions—this is done so effect sizes can be directly compared across models (standardized betas can be roughly thought of as partial correlations that control for all other variables; full regression results are reported in the Supplementary Appendix). In terms of effects, these results largely confirm findings from previous studies. The novel contribution here is the examination of comparative variation in the predictors by type of anxiety measured. That comparison is illuminating in that it suggests while there are clearly common sources of both political and general anxiety, there are also notable differences up to and including predictors that switch signs between models.
Effect sizes of political anxiety and general anxiety predictors
Notes: Standardized betas from OLS regression models are reported. Full models are reported in Supplementary Appendix along with descriptive statistics for predictor variables. *p <. 05, #p <. 10 (two-tailed tests). VIF statistics are as follows: 1.21 (Model 1), 1.24 (Model 2), 1.21 (Model 3), and 1.24 (Model 4).
Table 3 also reports the results of follow-up regressions where we included the anxiety measures as predictors of each other. This was done to assess the effect on inference—if our argument that these can be treated as independent concepts is correct, then the results of the expanded model should not radically differ from the constrained model. Given the established correlation between the anxiety measures it is not surprising they are significant predictors of each other, but not to the extent that they capture the variance accounted for in the unconstrained models by the predictors existing research associated with, respectively, political and general anxiety.
Consistent with that previous research, established predictors of general anxiety (income, race, gender, age, and education) are all in the expected direction in the unconstrained model (Model 3), though race and education have small effect sizes and are statistically insignificant. None of these predictors cross the standard threshold of statistical significance (p < .05) in the unconstrained political anxiety model (Model 1), though three (race, gender, and education) do meet a more liberal standard of p < .10. Of those three, two flip signs compared to the general anxiety model: Whites tend to report higher levels of political anxiety relative to underrepresented minorities/ethnicities, and more highly educated individuals also report higher levels of political anxiety than those with lower levels of education. Similarly, the political measures predict political anxiety consistent with results reported in earlier studies: conservatives report less political anxiety, while those who are more interested in politics and are more knowledgeable about politics report higher levels of political anxiety. Interestingly, two of these variables also display significant and nontrivial relationships with general anxiety. Again, conservatives report less general anxiety and those who are more interested in politics report more. The effect sizes are of some interest: At least in our data, ideology has a moderately larger impact on general anxiety compared to political anxiety. It is also worth noting that political knowledge positively predicts political anxiety and negatively predicts general anxiety. The positive relationship with political anxiety is consistent with previous research. The negative relationship with general anxiety perhaps reflects a similar sort of prophylactic effect as education, i.e., that people who are better informed tend to score lower in general anxiety.
In the unconstrained models (Models 2 and 4) general anxiety is a significant predictor of political anxiety and vice versa. With few exceptions, the effect sizes and inferences taken from the other predictors remain unchanged. Perhaps the most notable difference across the constrained and unconstrained models is political interest as a predictor of general anxiety. When political anxiety is included as a predictor of general anxiety, political interest is indistinguishable from zero—the estimated effect size is only a fifth of that reported for the constrained model and is statistically insignificant. We highlight this finding as it may have relevance for a broader literature on the relationship between emotion and political engagement, especially the framework of affective intelligence theory proposed by Marcus et al. (Reference Marcus, Neuman and MacKuen2000). That framework argues political interest is emotionally motivated and generates an a priori hypothesis that political interest and political anxiety are strongly correlated. This is indeed the case—the bivariate correlation between the two is r = .47 in our data, thus strongly suggesting the same variance in general anxiety political attention accounts for in Model 3 is accounted for by political anxiety in Model 4. We believe that future researchers would be well served by devoting attention to the interplay between general anxiety, political anxiety, and political interest in models of political decision-making and political engagement.
Discussion
The primary objective of this article was to empirically investigate the conceptual similarity of political and general anxiety. General anxiety is well-established as a concept, and validated instruments to capture that concept have been in use for some time. Political anxiety is increasingly capturing the attention of both academics and clinicians, but both the concept and its measurement are comparatively much more recent. The existing literature provides little insight into whether political anxiety is just an example of one of the many circumstances of daily life that produce worry and apprehension or whether it is something that is conceptually distinct, a source of worry and apprehension that affects not only those who suffer from more general forms of anxiety. Our analysis suggests political anxiety can be considered an empirically distinct concept. This has implications not just for how anxiety is measured but for broader research literatures examining the relationship between emotion and politics (e.g., research with the affective intelligence framework).
We do find a relationship between political and general anxiety, but this is fairly modest with the relationship between instruments amounting to r = .24. The items that make up those instruments do not appear to be tapping into a single underlying construct, and the systematic correlates of these measures identified by previous literature are not equivalent predictors of the different types of anxiety. Politics seems to be a source of apprehensiveness and worry that affects individuals who are not necessarily suffering from general anxiety.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/pls.2024.6.
Open access
