The campaign to deny the 2020 election results, which culminated in the January 6, 2021, attack on the US Capitol, rightly led many to fear for the stability of American democratic institutions. But although the peaceful transfer of power is a necessary component of democracy, it is also crucial to understand the collective choice of whether to transfer that power or not. This is all the more important when the country is racked by multiple crises, which may upset the usual rules of mass voting behavior.

Judging by this masterful account by the Annenberg Institutions of Democracy (IOD) Collaborative, however, the 2020 election operated in much the same way as prior campaigns. US voters, they write in chapter 1, “saw the crises of the pandemic, the economy, and the protests over racial justice … through the lens of their polarized partisan predispositions filtered by their preferred media sources” (14). Opinions of the candidates changed little over the course of the campaign, and what little change there was came from the same factors that have long mattered in political campaigns.

To understand how such stability could still produce political change, the authors argue we should consider more than the standard factors of retrospection, partisanship, and the media, all of which are reviewed and helpfully placed in the context of 2020 in chapter 2. Instead, understanding 2020 requires more nuanced consideration of each factor, starting with partisanship in chapter 3. Here they introduce the survey data that power the bulk of the analysis: a 14-wave panel of voters from four battleground states—Florida, Pennsylvania, Michigan, and Wisconsin—collected over the course of the 2020 campaign. These data are invaluable, allowing us to zoom in on key voters in key states. But even in these closely divided states, “vote preferences changed little over [2020]. Even before Joe Biden had won a single Democratic primary, many voters knew how they would cast their ballots” (49). Specifically, 9 of 10 voters in these surveys voted the way they initially predicted they would.

I emphasize this marked partisan stability in the face of social and economic instability, in part because the authors undersell their contribution by writing, “Few would be surprised by a finding that stability more than change characterizes voters’ preferences” (p. 49). At least to the uninitiated—or to those “who might dust off a copy” of the book “many years from now” (p. 18), an audience to whom the book’s many supplemental historical narratives will be very useful—such stability would be remarkable even in a normal campaign, let alone one conducted amid such upheaval.

With so little opinion change, how can we study the influence of campaigns? Borrowing from epidemiology, the authors begin by dividing the sample in all waves into “unwavering” supporters of a candidate, or those who said they would turn out to vote and would vote for a particular candidate; “wavering” supporters, or those who did not report a consistent preference for a particular candidate; and “everyone else,” or those not voting or voting but supporting a minor-party candidate. This strategy enables the authors to drill down even further, identifying the “at once elusive and seminal” subgroups (p. 49) that, in today’s polarized environment, play an outsized role in determining presidential elections.

In chapter 4, the authors introduce another measurement innovation, which they aim at observing media use. Recognizing that we now live in a high-choice media environment, they ask respondents not only what types of media they consume but also which sources they actively seek out or actively avoid. The result is a fivefold classification of voters where, for media of different ideological slants, we can place voters as “successful seeking,” “unsuccessful avoidance,” “casual consumption,” “successful avoidance,” or “disengaged.” Comparing this classification to that of the previous chapter, we learn that although most unwavering supporters were predictably able to avoid dissonant news, majorities of wavering supporters were not (95).

The subsequent chapters apply these novel classifications to the key issues in the 2020 campaign: the pandemic (chap. 5), the economic downturn and recovery (chap. 6), racial justice (chap. 7), and status threat (chap. 8). Not surprisingly, the authors find all these factors are important for explaining the decisions of wavering voters in these four states. Just as interesting as the findings are the novel ways in which the authors measure the influence of these events. To highlight just a few: using changes in foot traffic to measure the local economic effects of the pandemic; using a survey measure of stimulus check receipt to measure the effect of personal economic circumstances; and using geocoded distance to the August 2020 Kenosha, Wisconsin, protests to measure the effect of racial unrest.

The final two chapters move to the “fourth crisis”: the crisis of democratic legitimacy. Although their surveys do not include anyone who participated in the January 6 attack, the authors do observe changes in beliefs about election fraud and reactions to the attack in more or less real time. On the former, what I found most interesting were the postelection trends. When asked whether the election was “free and fair,” responses predictably diverge after Election Day, but then change little thereafter (303). And although the postelection drop in confidence is larger for those who consume more conservative media, there is little over-time change there as well (315). In contrast, Trump voters do seem to have become less confident that their own ballot was counted as intended as the propaganda campaign ground on (305). These trends raise interesting questions about the role of elite rhetoric and media in shaping both broad and specific confidence in elections.

One of the authors’ conclusions in these final chapters is that beliefs in voter fraud led to the attack and, more broadly, to support for future political violence. To measure support for violence, they ask respondents about support for the insurrection, beliefs that the rioters were patriotic, beliefs that any violence was a “false flag,” beliefs that force may be necessary to save the American way of life, and beliefs that it may be necessary to take the law into one’s own hands. The cross-tabulations of these responses are chilling and certainly tell us something important about how the January 6 attack was framed and perceived by partisans. But is it really the case that a substantial share of voters will embrace violence as a substitute for political action when elections do not go their way?

There is actually little direct evidence that either beliefs in fraud, support for violence, or participation in the January 6 attack is associated with lower political participation; thus, there is no evidence that fraud beliefs lead voters to choose “force, rather than the ballot box” (p. 340) as the authors predict. There is also, as far as I can tell, no direct statistical test of the relationship between fraud beliefs and support for violence in chapter 9. Recent survey experimental work also suggests that, although elite rhetoric about fraud undermines confidence in elections, it does not also undermine support for democracy or increase support for violence; see Nicolas Berlinski et al., “The Effects of Unsubstantiated Claims of Voter Fraud on Confidence in Elections,” Journal of Experimental Political Science (2021); Katherine Clayton et al., “Elite Rhetoric Can Undermine Democratic Norms,” Proceedings of the National Academy of Sciences (2021). On the other hand, and as we can clearly see in these pages, a substantial share of Republican voters will act as apologists once political violence is committed, which is perhaps no less unsettling.

Overall, there is probably no better accounting of the behavioral causes and effects of January 6, and there is almost certainly no better accounting of voter behavior in the 2020 election campaign. But as if making sense of a single chaotic election and its aftermath were not enough, the members of the IOD Collaborative have achieved even more. Through the use of numerous innovations in measurement and data collection, they have set us all up to better understand the elections to come.