No additional evidence that proximity to the July 4th holiday affects affective polarization

One promising approach for reducing affective polarization is priming a shared American identity and one promising event to prime that identity is the 4 th of July. Prior work showed that proximity to the 4 th of July reduced affective polarization. We conceptually replicated this study using a 9-wave longitudinal design in 2019. We found no short-term or long-term effects of the 4 th of July on social distance from partisan and ideological ingroups or outgroups. Notably, our within-subjects design was able to identify the existence of individual differences in social distance trajectories across time, but there were not individual differences in short-terms changes in social distance in close proximity to the 4 th of July. Additional analyses, did not find consistent predictors of these individual differences, suggesting a clear gap for future studies. Although priming a shared American identity may be effective, these findings suggest that the salutary effects of the 4 th of July holiday do not emerge in 2019.

Over the last decades, affective polarization has increased in the United States (Iyengar & Krupenkin, 2018). People are likely to express dislike towards and a preference for social distance from both partisan and ideological outgroups (Brandt & Crawford, 2020;Iyengar et al. 2019;Mason 2018), resulting in both partisan and ideological manifestations of affective polarization. Affective polarization is the result of people's identification with their political party or particular ideological labels (e.g., liberal or conservative; Iyengar et al., 2019;Mason, 2018). Like any identification process, people who identify with their partisan or ideological ingroups are likely to sort their world into ingroups and outgroups based on these identities (Tajfel, 1981;Tajfel & Turner 1979). Driven partly by the need for positive identity, people then tend to favor their ingroup over outgroups. This is especially the case for competitive outgroups like political rivals where ingroup preference can also entail an outgroup dislike (Brewer, 1991;Mullen, Brown, &, Smith 1992;Voci, 2006).
One promising approach for reducing affective polarization is to remind people of their shared American identity with partisan and ideological opponents (Levendusky, 2018;Riek, Mania, Gaertner, McDonald, & Lamoreaux, 2010). People have multiple identities that can be hierarchically organized (e.g., as an American, as an American woman, as an American woman political psychologist). If it is possible to make a superordinate identity salient (within which both the ingroup and outgroup belong) that allows people to recategorize their political outgroups as part of an ingroup (Gaertner and Dovidio 2000). This emphasizes what is positive and shared between "us", and could consequently reduce affective polarization. One such identity in the United States is people's American identity (Huddy & Khatib 2007) which can encompass both American political ingroups and outgroups. When participants are primed with their American identity they report less negative views of the partisan outgroup (Levendusky 2018). Such primes do not make people more positive about the partisan ingroup, suggesting that the experimental treatment does not just make people more positive about all groups in general.
Overall, it appears that priming American's American identity reduces affective polarization by changing people's perceptions of the partisan outgroup.

July 4 th Holiday, Original
To complement experimental studies, Levendusky (2018) examined if proximity to the July 4 th holiday reduced affective partisan polarization. The idea was that the celebration of the July 4 th holiday involves reminders of people's American identities. This should increase the salience of the American identity and therefore reduce affective polarization between party outgroups. This was tested using data from the 2008 National Annenberg Election Study's online panel (N ~ 6,000). Participants in this panel were interviewed about their perception of the opposing party's nominee (Obama or McCain) on a feeling thermometer. Because interview day was randomly determined between participants, this allowed Levendusky to test if proximity to the holiday reduced affective polarization. He found, using a variety of operationalizations of "close," that closer proximity to the holiday reduced dislike of the outparty candidate (effect size ranged between 1.72 and 3.63 points on a 100-point scale). Mirroring the experimental effects, proximity to the July 4 th holiday did not affect people's evaluations of the inparty candidate. This result suggests that the nearer the July 4 th holiday was, the more strongly the American identity can be primed, and this reduces affective polarization. This highlights a naturally and annually occurring setting in which the superordinate American identity can be made salient. Moreover, it suggests that the July 4 th may be a unique setting of relative partisan understanding.

The Current Study
We aimed to conceptually replicate Levendusky (2018) for the July 4 th holiday in 2019. We use longitudinal data to gain within subjects comparisons and allow for more detailed assessment of the impact of this holiday on affective polarization. This allows us to identify different time trends and individual differences surrounding the holiday and its relationships with affective polarization. We adopt methods from personality psychology that have aimed to understand how life events affect the development of personality, life satisfaction, and other constructs. We follow the example of Denissen, Luhmann, Chung, and Bleidorn (2019) who studied the effect of life events on personality (e.g., the effect of marriage on life satisfaction). Their models help distinguish between linear changes as one approaches the July 4th, a linear effect as the July 4th recedes into the past, a stable change following the July 4th, short term changes around the July 4 th , and testing effects (i.e. repeated exposure to the survey; see Table 1). This will help us assess if the influence of the July 4 th holiday is because of people's experience on the holiday (e.g., effects of variables indicating short term changes, stable changes, or linear effects post July 4 th ), their anticipation of the holiday (e.g., effects of linear changes as one approaches July 4 th ), because of repeated exposure to the survey (e.g., testing effects), or some combination.
Based on Levendusky's (2018) theorizing, we would expect that negative evaluations of the out-party and out-ideology would decrease on or near July 4 th (i.e. significant effects for the short term changes in Table 1). Levendusky's theorizing could also be interpreted as predicting less negative evaluations of the out-party and out-ideology as July 4 th approaches (i.e. linear anticipation from Table 1) and that negative evaluations of the out-party and out-ideology increases as the July 4 th recedes (linear change post July 4 th from Table 1). We test these possibilities. Levendusky (2018) reports no clear differences between Republicans and Democrats in their analyses; however, there may be individual differences that do not map onto this particular moderator variable. An additional benefit of the longitudinal design, that is not available in typical between-subject designs, is that we will be able to estimate the existence of individual differences in response to the July 4 th holiday by testing if the effects of the time varying variables are significantly different across • firstDay: Changes on July 4 th only. Coded 1 on July 4 th or 0 otherwise • firstWeek: Changes starting on July 4 th and lasting for 1 week. Coded 1 for July 4 th -July 10 th and 0 otherwise. • withinDay: Changes that occur within one day of July 4 th . Coded 1 for the 3 rd , 4 th , and 5 th and 0 otherwise. • withinWeek: Changes that occurred within one week of July 4 th (July 3 rd to July 10 th in the data available). Coded 1 for July 3 rd -July 10 th and 0 otherwise.

Testing Effects
Changes in affective polarization associated with the number of waves completed.
The number of waves participants have completed since the start of the study, ranging from 0 to 8 people (i.e. the random slopes are significant; Bolger et al. 2019; Denissen et al. 2019;Whitsett and Shoda 2014). When there are individual differences, our study included a number of additional variables that might serve as moderators, including strength of identification, party and ideological identification, perceived economic and value threat, political interest, the importance of the July 4 th holiday, and demographic variables. In this way, we are able to test for the mere presence of individual differences and test a suite of possible moderators. In addition, our study included measures of affective polarization for both party identification and ideological identification. The original work on the July 4 th holiday only examined party identification; however, affective polarization also occurs for ideological identification and is thought to occur for similar theoretical reasons (e.g., Mason 2018;Brandt & Crawford, 2020). Therefore, we believe that the theoretical rationale would predict similar results for both types of political identification, although the analysis of party identification is the most similar to the original work. People who think that ideological affective polarization occurs for different reasons than partisan affective polarization, would not predict similar results for both types of political identification. We also included evaluations of moderates as an exploratory measure; we had no specific hypotheses.
The data we analyze for the project are from a yearlong longitudinal study. This study started on 8 May 2019 and ended on 29 April 2020, surveying the same participants every 2 weeks for a year. Crucially, the study overlaps with the July 4 th holiday and it includes measures of party identification, ideological identification, and affective polarization. By adopting this study design, we are able to retest Levendusky's (2018) hypothesis about the proximity to the July 4 th holiday, extend his analysis to ideological identification, identify the mere existence of individual differences, and assess several moderators that may account for any individual differences.

Method Preregistration
We used data from our year-long longitudinal study. This study was conducted between 8 May 2019 and 29 April 2020, surveying the same participants every 2 weeks for a year. There are three pre-registrations associated with this paper. First, the method for the entire year-long study is preregistered at this link: https://osf.io/p2eju/. This preregistration was fixed as read-only before any data were collected. Second, the measures and analysis for this specific study are preregistered at this link: https:// osf.io/t3gf5. This preregistration was fixed as read-only after the start of data collection, but before Wave 4 of data collection and before any other analyses were conducted with the data. Third, due to a minor error in adding a measure to Wave 4 (see below), the preregistration for this specific study was amended and is at this link: https://osf. io/xfgrq. This preregistration was fixed as read-only after the start of Wave 4 data collection and before any analyses were conducted on the data.

The Sample
Participants were recruited via Prolific (Palan and Schitter, 2018). The first wave recruited 552 participants who voted for Clinton, Trump, or someone else/did not vote in proportion to the results of the 2016 presidential election. It was open until 550 participants were collected (with 2 extra unintentional participants). Subsequent waves were open for 1 week and participants could complete the wave any time during that week. We analyzed the data collected between 8 May and 4 September. This included the first four waves before the holiday, Wave 5 which overlaps with the holiday, and the four waves following the holiday. The response rates for these waves ranged from 79% (Wave 9) to 92% (Wave 2). After removing participants without clear partisan or ideological identifications (see below), our sample consisted of 363 participants (M age = 36.4, SD age = 12.8, 184 women, 178 men, 1 reporting another option). Participants were primarily white (n = 290) and a plurality had bachelor degrees (n = 145). The most commonly selected income range was $50,000 to $74,999 (n = 76).

Preregistered measures
There are many measures in the study (see materials here: https://osf.io/x94rc/). We focus only on those relevant to this research question. All of the measures we used are in Table S1 in the supplemental materials.

Social distance
The primary outcome variables used to assess affective polarization are people's self-reported social distance from the political outgroup. Levendusky (2018) originally operationalized affective polarization as a feeling thermometer towards the in/outgroup presidential candidate. Although not identical to Levendusky's measure, we apply social distance as a common measure of affective polarization in the literature (e.g., Iyengar et al., 2012). It is a common measure of prejudice in the social psychology literature more broadly (Correll, Park, Judd, & Wittenbrink, 2010), correlates well with other potential measures of affective polarization (e.g., feeling thermometers; Crawford et al., 2017), and often shares predictors with other measures of affective polarization (Druckman & Levendusky 2019;Garrett & Bankert, 2020;Mason 2018). 1 We looked at social distance as people's self-reported willingness to be friends with people from ideological outgroups (i.e. liberals or conservatives) and partisan outgroups (i.e. Democrats or Republicans). We first reverse scored the social distance items from Table S1, so that higher scores indicated more social distance. We coded observations of "Don't know" or "I haven't thought much about it" as missing data. All social distance measures were standardized by their standard deviation. This rescaling was not preregistered.
Ideological outgroups were determined by their placement on the 1-7 ideological identification item. Ideological identification was asked at all waves. We used data from the first four waves (i.e. the 4 waves before the July 4 th holiday) to categorize participants. People who state a 1-3 (i.e. are on the liberal side of the measure) over all of Waves 1-4 have conservatives as their ideological outgroup. People who state 5-7 (i.e. are on the conservative side of the measure) over all of Waves 1-4 have liberals as their ideological outgroup. People who switch allegiance, state 4 (i.e. moderate), "Don't know," or "I haven't thought much about it" were excluded from this analysis (n included = 330).
Partisan outgroups were determined by their placement on the 1-7 partisan identification item. Partisan identification is asked at all waves. We used data from the first four waves (i.e. the 4 waves before the July 4 th holiday) to categorize participants. People who state a 1-3 (i.e. are on the Democrat side of the measure) over all of Waves 1-4 have Republicans as their partisan outgroup. People who state 5-7 (i.e. are on the Republican side of the measure) over all of Waves 1-4 have Democrats as their ideological outgroup. People who switch allegiance, state 4 (i.e. independent), "Don't know," or "I haven't thought much about it" were excluded from this analysis (n included = 314).
We test if proximity to the July 4 th holiday improves attitudes towards the ingroup as well (Levendusky's [2018] theoretical analysis would not expect such an effect). We test this using measures of social distance towards ideological and partisan ingroups. Ingroups were determined using the same strategy as described above. As a further exploratory test, we examined if proximity to the holiday influences attitudes towards moderates. For this analysis, people who state 4, "Don't know," or "I haven't thought much about it" on the ideological identification measure were excluded from this analysis.

Preregistered analytic strategy
We must depart from Levendusky (2018) analytic strategy because of differences in design. His design is essentially a between-subjects design where proximity to the holiday is randomly assigned. Our design is a within-subjects design and is conceptually similar to studies that test how life events (e.g., marriage, diverse, unemployment) affect personality or life satisfaction. In our study, the life event is the July 4 th holiday and the outcome is social distance. Because participants did not all complete the Wave 2 though Wave 9 surveys on the same day, we have a range of distances from the July 4 th holiday. See Figure S1 in the supplemental materials for sample sizes for partisan and ideological affective polarization analyses across the study period.
To test the hypothesis that proximity to the July 4 th holiday would be associated with less social distance from political outgroups, we used the analytic strategy of Denissen, Luhmann, Chung, and Bleidorn (2019). We modified their models to help distinguish between several different types of effects (see Table 1). A description of the factors we consider are in Table 1. The exact coding for each factor is in Table S2 in the supplemental materials.
Four time-varying factors were included in every model.
• Linear Anticipation: This indexes the linear change as one approaches July 4 th . • Post July 4 th Stable Change: this indexes stable change (for the duration of the study) following July 4 th . • Linear Change Post July 4 th : This indexes the linear change as one moves away from July 4 th . • Testing effects: This indexes the number of waves that have occurred since the start of the study.
We also aimed to assess short-term changes in social distance, consistent with Levedusky's (2018) observation that the analyses for data closest to the July 4 th holiday had the strongest effects. Because it is not obvious how close people need to be to the holiday for American identity to be primed, we follow Levendusky's lead and estimate additional, separate models with indicators representing different windows around July 4 th . Each of these time-varying factors estimating short-term changes will be included in separate models.
• July 4 th : This indexes changes on July 4 th only.
• 1 Week Post July 4 th : This indexes changes starting on July 4 th and lasting for 1 week. • Within 1 Day of July 4 th : This indexes changes that occur within one day of July 4 th (i.e. the 3 rd , 4 th , and 5 th ). • Within 1 Week of July 4 th : This indexes changes that occurred within one week of July 4 th (3 July to 10 July in the data available to us). 2 Observations were nested within participants, which were treated as a random effect. 3 All of the time-varying predictors were treated as fixed effects. Across all analyses, we use p < .005 as our preregistered alpha level (Benjamin et al. 2018). 4

Does Proximity to the July 4th Affect Affective Polarization
The mean social distance for each date with data are in Figure 1. To estimate the effects of the July 4 th holiday, we estimated 20 models (5 outcomes: [social distance out-ideology, social distance out-party, social distance in-ideology, social distance in-party, social distance moderates] × 4 individually-entered time-varying factors [July 4 th , 1 Week Post July 4 th , Within 1 Day of July 4 th , Within 1 Week of July 4 th ]). Across models, the number of participants range from 314 to 330 (M = 323.6) and the number of observations ranged from 2598 to 2747 (M = 2683.8). The results of these models are in Figure 2 and in Tables S3 to S22. Based on Levendusky (2018) findings and theorizing, we would expect that social distance from the outparty and out-ideology would decrease on or near July 4 th . However, we did not find significant effects for the any of the short-term effects (i.e., July 4 th , 1 Week post July 4 th , Within 1 Day of July 4 th , or Within 1 Week of July 4 th factors). 5 Translating the largest effect size from Levendusky (2018, Table 4, 3 Day Window) into our scaling would be a short-term effect of approximately b = -.11. The confidence intervals for the short-term effects when predicting both out-party and out-ideology social distance did not overlap with b = -.11 for Within 1 Day of July 4 th or Within 1 Week of July 4 th factors. For 1 Week post July 4 th , the confidence interval did not overlap when predicting out-ideology. This suggests that many of the estimates of the current study are largely inconsistent with the largest estimates of Levendusky's findings. 6 The remaining confidence intervals are very wide with a lot of uncertainty. Levendusky's findings could also be interpreted as predicting less social distance from the out-party and outideology as July 4 th approaches and that social distance from the out-party and out-ideology increases as the July 4 th recedes (i.e., on the longer-term time varying effects). However, we did not find significant effects of linear anticipation or linear post. These results are all inconsistent with our expectations. Notably, there were also no significant effects of any time-varying predictor on any of the potential outcome variables, including in-party, in-ideology, and moderate targeted social distance. Proximity to the July 4 th holiday does not have any clear effects on social distance.

Are there Individual Differences in the Effect of Proximity to the July 4th on Affective Polarization?
We tested for the mere existence of individual differences in the experience of the holiday by testing if there are significant random effects for the time-varying predictors. That is, are the fixed effects reported in Figure 2 significantly different between participants? Fixed effects are essentially averages of the effect across participants; meaning that it is possible that the average effect is not different from zero, but that the effect varies and is different for some subsets of participants. We test this possibility individually for each of the 20 models in Figure 2. For each model, we compare the model with a random effect for one of the fixed effects (e.g., Linear Post) to the model with only fixed effects (and no random effects, i.e. the model in Figure 1) using the anova function in R. Across all analyses, we use p < .005 as our alpha level. Of the 100 potentially significant random effects (5 time-varying fixed effects × 20 models), we found that 80 of the time-varying predictors were significantly different between participants. 7 These 80 significant random slopes suggest that there are individual differences in people's social distance trajectories over the course of the study.
Notably, the 20 non-significant random effects were for the short-term effects of July 4 th , 1 Week post July 4 th , Within 1 Day of July 4 th , and Within 1 Week of July 4 th . These time varying predictors are those that most clearly operationalize proximity to the July 4 th holiday, suggesting that the non-significant effect of the July 4 th is consistent across the participants in our sample. However, the remaining parameters of the model are more likely to differ between participants. Because these other parameters differed across participants, we conducted moderator analyses in order to try to explain these differences. However, because these were not the key parameters related to short-term effects of the 4 th of July, we report these analyses in the supplemental materials.

Selection Effect?
The analytic strategy focuses on within-subject comparisons, which make selection effects less of a concern. However, one place they may emerge is that people who find the July 4 th holiday particularly important might be less likely to participate in our study on July 4 th because they are busy celebrating. To test if this is a concern, we tested if the measure of the importance of the 4 th of July was associated with the participants' completion date of the Wave 5 survey using a one-way ANOVA with date as the factor and importance of the 4 th of July as the DV. Across all analyses we use p < .005 as our alpha level. This pre-registered ANOVA was not significant, F(6, 308) = 1.54, p = .16. A non-preregistered ANOVA recoding date to be dichotomous (on July 4 th vs. other date) was also not significant, F(1, 313) = 2.93, p = .09. Similarly, distance from the July 4 th holiday was not significantly correlated with the importance of the holiday, r(313) = -.08, p = .15, 99.5%CI [-.24, .08].

General Discussion
We found no clear effects of proximity to the 4 th of July in 2019 on social distance from partisan and ideological outgroups, ingroups, or ideological moderates using a preregistered 9-wave panel study. Although individual differences exist on a number of the relevant longitudinal trajectories, we did not find individual differences on any of the factors representing short terms changes in social distance near the 4 th of July.
The statistical power of this study was high enough to detect effects that are at least as large as the largest in Levendusky's (2018) original work when using the more lenient p < .05. However, if the true effect is smaller than this, our study is not well positioned to identify such an effect. Notably, the original effect size was somewhere between 1.72 and 3.63 points on a 100 point scale, suggesting that the original effect was quite small. Our study lends further credence to the likely small effect size of the July 4 th holiday. The lack of any hints of the July 4 th 's effectiveness in the moderator analyses should further curtail any enthusiasm that the holiday has salutary effects of an appreciable size in our data. In short, these results should cast doubt on the effectiveness of the 4 th of July to meaningfully reduce affective polarization in 2019.
There are important differences from Levendusky's (2018) original finding. Levendusky used a betweensubjects design in the election year of 2008 and asked participants to evaluate candidates using a feeling thermometer. We used a within-subjects design in the offelection year of 2019 and asked participants to evaluate partisan and ideological ingroups and outgroups using a measure of social distance. All of these methodological differences should not theoretically cause a problem.
For example, the original paper was about affective polarization broadly (i.e. not just about candidates) and the theorizing should apply to our measures of social distance, a commonly used measure when studying affective polarization. Indeed, the theory of common ingroup identity (Gaertner and Dovidio 2000) that motivated the original work is a theory about reducing intergroup animosity in general and not about reducing only a specific type of intergroup animosity. Nonetheless, recent work has found that social distance may represent a different facet of animosity and affective polarization than do feeling thermometers (Druckman & Levendusky 2019). People also tend to report more animosity about political elites and parties than they do about ordinary citizens who are part of the partisan outgroup (Druckman & Levendusky 2019;Kingzette, in press). Druckman and Levendusky (2019) show that when people evaluate the other party (as in Levendusky, 2018) they think of elites more than ordinary voters. However, our measure more explicitly focuses participants on ordinary voters. As such, this might mean that the measurement differences between our study and the original cause the differences in conclusions, similar to how feeling thermometers and social distance measures respond differently to some manipulations (Ahler & Sood, 2018). Readers who are persuaded that social distance measures are substantially different from other measures of affective polarization can interpret our findings as indicating that proximity to the 4 th of July does not appear to affect one specific sub-type of affective polarization directed at behavior outcomes.
Importantly, other work finds that measures of social distance correlate well with feeling thermometers (Crawford et al., 2017), that both types of measures have similar predictors (Druckman & Levendusky 2019;Garrett & Bankert, 2020;Mason 2018), and react similarly to some experimental manipulations (Levendusky & Malhotra, 2016). This set of findings suggests that our measure of social distance should not necessarily find different results than feeling thermometers. Readers who are persuaded that social distance and feeling thermometers are more similar than different can interpret our findings as indicating that proximity to the 4 th of July does not appear to meaningfully affect affective polarization more generally.
Similarly, conducting our study in a non-election year is different from the original. It seems that, if anything, a non-election year might be less polarizing because the political context is less competitive. Nonetheless, the political system in the United States is in a different place in 2019 compared to 2008. In the summer of 2008, both presidential candidates expressed support for working with members of the other party and bridging American divides. In the summer of 2019, Donald Trump held a polarizing military-style parade to mark the 4 th of July. These different political contexts may be enough to politicize the meaning of the 4 th of July and reduce its potential for the 4 th of July to serve as a remind of American's superordinate identity.
Levedusky's (2018) original theoretical insight was that a common ingroup identity might reduce affective polarization. Although we did not find support for the idea that this might occur via proximity to the 4 th of July, common ingroup identity could still be an effective depolarization strategy (Riek et al., 2010). This suggests that what serves as an effective prime of common ingroup identity is subject to change. According to Hornsey and Hogg (2000), making the superordinate identity salient while ignoring subgroup identities might induce identity threat and therefore perpetuate intergroup bias. As such, future application of Levendusky's (2018) July the 4 th paradigm may find it useful to explicitly acknowledge the American, Democrat and Republican identities simultaneously. However, it is possible that growing differences between Democrats and Republicans limit the effectiveness of the American identity to function as a common ingroup. Democrats and Republicans have different ideas about what American identity means (Hanson & O'Dwyer, 2019). Rutchick and Eccleston (2010) argue that because Democrats and Republicans are perceived to have rather different ideas about what the American identity means, it may be less able to harmoniously unite these subgroups. If this is the case, then carefully constructing primes to work their current context is important for replicating and extending the work on American identity primes, as well as using this work in practical settings. Future studies around the 4 th of July holiday in different political circumstances and amidst different types of national conversations may help uncover if and how the meaning of the holiday impacts people's American identities and levels of polarization.
The longitudinal design allowed us to identify the existence of individual differences in response to the proximity of the 4 th of July. However, this came at the cost of non-representativeness. Although our analyses suggest little heterogeneity in the effects of proximity to the 4 th of July, a larger and more heterogenous sample may identify the predicted effects. We were also only able to include a single-item measure of affective polarization, although we were able to use this measure for both ideological and partisan groups. Our results suggest that proximity to the 4 th of July does not impact social distance from ideological and partisan outgroups, ingroups, or ideological moderates in 2019. Other primes of American identity may be more effective.

Data Accessibility Statement
Data and replication code are available here: https://osf. io/26bua/.

Notes
1 After our study was launched, a paper (Druckman & Levendusky 2019) was published suggesting that social distance measures capture a different factor of affective polarization compared to feeling thermometers, trust ratings, and trait ratings. At the same time, these authors note that (p. 119), "To be clear, this does not mean that one measure is "better" than another; rather, they gauge different manifestations of affective polarization." They also note (p. 119) that the correlates of social distance measures and the other measures of affective polarization "all meaningfully capture variation in partisan animosity." Because nothing in the theory of Levendusky (2018) suggests that proximity to the 4 th of July should be limited to a particular type of affective polarization, we believe this is a reasonable test. We will return to this point in the discussion. 2 Due to an error in the preregistration we listed the date ranges for within 1 week as between both 26 June and 10 July and between 3 July and 10 July. Because the latter range is the only version to adhere to the within 1 week description, we chose this version for the analyses. 3 The models were estimated in R (R Core Team, 2019) with lme4 (Bates et al., 2015). P-values were calculated using lmerTest (Kuznetsova et al., 2017). Additional