Search

What spurs voters to vote in person, despite an established universal vote-by-mail (VBM) system and a once-in-a-century pandemic? We explore this question with official voter data from Colorado, a vote-by-mail state since 2013, but where 6% of voters still vote in person. Using multiclass classification, we analyze (1) the choice between voting by mail (VBM), voting in person, and not voting in the 2020 general election, and (2) the choice to switch to in-person voting despite having used VBM in previous cycles. The results suggest that the choice of voting modes is mainly habitual, and local variations of COVID-19 and demographics hardly mattered. Notably, Republican partisanship plays an important role in predicting "switchers" to in-person voting; indeed, the probability of switching to in-person voting was 5.2% conditional on being a Republican as opposed to 1.9% conditional on being a Democrat.

Are political representatives responsive to their constituents in determining the issues they address? Our paper takes a novel approach to this critical question using innovative data and state-of-the-art methodology. While most previous research on issue responsiveness has looked at the United States and its unique political institutions, we study a multi-party, parliamentary setting: Canada. We focus our analysis on two prominent issues: the environment and climate change. Using transcripts from the House of Commons’ Question Period from 2006 to 2021, we measure the attention political parties attribute to the various issues. We use Google Trends data to measure policy issues’ public salience. We implement an instrumental variable estimation strategy to identify and estimate how the environment and climate change’s public salience drives elite attention. Our analysis demonstrates that public salience significantly affects political parties’ interest in the environment and climate change but unveils appreciable partisan heterogeneity in this regard.

Objective: Assuring voter confidence is important for the legitimacy of democratic elections. In this article, we test four hypotheses about the correlates of voter confidence at various levels of election administration. Methods: We take advantage of a large online survey of registered voters in a single election jurisdiction, Orange County (CA), which was implemented immediately after the November 2018 midterm elections. We measure voting experience and social media usage using item response theory as well as voters' perceptions of various election or voter fraud, and estimate their association with voter confidence via logistic regression models. Results: Our results show that voters who cast mail ballots are less confident about their own votes being counted correctly than in‐person voters. For both types of voters, those who have poor experiences with the voting process are much less likely to report confidence in the election. We also find that voters who have strong concerns about election fraud are less likely to report being confident in the election. Our last result indicates that information from news and social media is associated with a decline in voter confidence in election administration at the national level. Conclusion: Given the many conversations about election fraud that have occurred since the 2016 presidential election, this research provides important implications for election administration in future elections in the United States.

Objective: We analyze the significant increase in the residual vote rate in the 2016 presidential election. The residual vote rate, which is the percentage of ballots cast in a presidential election that contain no vote for president, rose nationwide from 0.99 to 1.41 percent between 2012 and 2016. Method: We use election return data and public opinion data to examine why the residual vote rate increased in 2016. Results: The primary explanation for this rise is an increase in abstentions, which we argue results primarily from disaffected Republican voters rather than alienated Democratic voters. In addition, other factors related to election administration and electoral competition explain variation in the residual vote rates across states, particularly the use of mail/absentee ballots and the lack of competition at the top of the ticket in nonbattleground states. However, we note that the rise in the residual vote rate was not due to changes in voting technologies. Conclusion: Our research has implications for the use of the residual vote as a metric for studying election administration and voting technologies.

Inferences about individual campaign contributors are limited by how the Federal Election Commission collects and reports data. Only transactions that exceed a cycle-to-date total of $200 are individually disclosed, so that contributions of many donors are unobserved. We contrast visible and "hidden" donors, i.e., small donors who are invisible due to censoring—and routinely ignored in existing research. We use the Sanders presidential campaign in 2016, whose unique campaign structure received money only through an intermediary/conduit committee. These are governed by stricter disclosure statutes, allowing us to study donors who are normally hidden. For Sanders, there were seven hidden donors for every visible donor, and altogether, hidden donors were responsible for 33.8% of Sanders' campaign funds. We show that hidden donors start giving relatively later, with contributions concentrated around early primaries. We suggest that as presidential campaign strategies change towards wooing smaller donors, more research on what motivates them is necessary.

In recent years, scholars, journals, and professional organizations in political science have been working to improve research transparency. Although better transparency is a laudable goal, the implementation of standards for reproducibility still leaves much to be desired. This article identifies two practices that political science should adopt to improve research transparency: (1) journals must provide detailed replication guidance and run provided material; and (2) authors must begin their work with replication in mind. We focus on problems that occur when scholars provide research materials to journals for replication, and we outline best practices regarding documentation and code structure for researchers to use.

Some American states have transitioned to universal voting-by-mail, where all registered voters receive a ballot in the mail. While this practice was growing in popularity prior to the 2020 general election, universal voting-by-mail was suddenly used in a larger number of states due to the COVID-19 pandemic. In this paper, we utilize a unique situation in which registered voters in some legislative districts in Los Angeles County were subjected to universal voting-by-mail in the March 2020 primary, while most of the rest of the Los Angeles County electorate was not. Using difference-in-difference and regression discontinuity designs, we estimate the causal effects of universal voting-by-mail on voter turnout and on who votes. Our results indicate that voter turnout increased by around 3\% for voters who do not automatically receive a ballot in the mail otherwise, and the increase is larger for registered partisan voters than those without a party affiliation.

The administration of elections depends crucially upon the quality and integrity of voter registration databases. In addition, political scientists are increasingly using these databases in their research. However, these databases are dynamic and may be subject to external manipulation and unintentional errors. In this article, using data from Orange County, California, we develop two methods for evaluating the quality of voter registration data as it changes over time: (a) generating audit data by repeated record linkage across periodic snapshots of a given database and monitoring it for sudden anomalous changes and (b) identifying duplicates via an efficient, automated duplicate detection, and tracking new duplicates and deduplication efforts over time. We show that the generated data can serve not only to evaluate voter file quality and election integrity but also as a novel source of data on election administration practices.

This paper is the result of a nationwide study of polling place dynamics in the 2016 presidential election. Research teams, recruited from local colleges and universities and located in twenty-eight election jurisdictions across the United States, observed and timed voters as they entered the queue at their respective polling places and then voted. We report results about four specific polling place operations and practices: the length of the check-in line, the number of voters leaving the check-in line once they have joined it, the time for a voter to check in to vote (i.e., verify voter’s identification and obtain a ballot), and the time to complete a ballot. Long lines, waiting times, and times to vote are closely related to time of day (mornings are busiest for polling places). We found the recent adoption of photographic voter identification (ID) requirements to have a disparate effect on the time to check in among white and nonwhite polling places. In majority-white polling places, scanning a voter’s driver’s license speeds up the check-in process. In majority nonwhite polling locations, the effect of strict voter ID requirements increases time to check in, albeit modestly.

Does attentiveness matter in survey responses? Do more attentive survey participants give higher quality responses? Using data from a recent online survey that identified inattentive respondents using instructed-response items, we demonstrate that ignoring attentiveness provides a biased portrait of the distribution of critical political attitudes and behavior. We show that this bias occurs in the context of both typical closed-ended questions and in list experiments. Inattentive respondents are common and are more prevalent among the young and less educated. Those who do not pass the trap questions interact with the survey instrument in distinctive ways: they take less time to respond; are more likely to report nonattitudes; and display lower consistency in their reported choices. Inattentiveness does not occur completely at random and failing to properly account for it may lead to inaccurate estimates of the prevalence of key political attitudes and behaviors, of both sensitive and more prosaic nature.

Who leads and who follows in Congress? By leveraging the Twitter accounts of U.S. House of Representatives members, we develop a new understanding of House leadership power using natural language processing methods in new ways. Formal theoretic work on congressional leadership suggests a tension in legislative party members' policy stances as they balance coordination and information problems. When their coordination problem is more pressing, the model predicts that legislative members will follow their party leaders' policy positions. When the information problem is more acute, party members coordinate and give their leaders direction for the party's agenda. Specifically, we exploit the network structure of retweets to derive measures of House leadership centrality within each party. We also use Joint Sentiment Topic modeling to quantify the discussion space for House members on Twitter. Our results provide support for the theoretical insights.

Tracking and collecting fast-evolving online discussions provides vast data for studying social media usage and its role in people's public lives. However, collecting social media data using a static set of keywords fails to satisfy the growing need to monitor dynamic conversations and to study fast-changing topics. We propose a dynamic keyword search method to maximize the coverage of relevant information in fast-evolving online discussions. The method uses word embedding models to represent the semantic relations between keywords and predictive models to forecast the future time series. We also implement a visual user interface to aid in the decision-making process in each round of keyword updates. This allows for both human-assisted tracking and fully-automated data collection. In simulations using historical #MeToo data in 2017, our human-assisted tracking method outperforms the traditional static baseline method significantly, with 37.1% higher F-1 score than traditional static monitors in tracking the top trending keywords. We conduct a contemporary case study to cover dynamic conversations about the recent Presidential Inauguration and to test the dynamic data collection system. Our case studies reflect the effectiveness of our process and also points to the potential challenges in future deployment., Preprint, Under Review

The integrity of democratic elections, both in the United States and abroad, is an important problem. In this Element, we present a data-driven approach that evaluates the performance of the administration of a democratic election, before, during, and after Election Day. We show that this data-driven method can help to improve confidence in the integrity of American elections.

Recent, non-peer-reviewed reports and meta-analyses suggest that smoking may reduce the risk of hospitalization with COVID-19, because the prevalence of smoking among hospitalized COVID-19 is less than that of the general population. However, there are two alternative classes of explanations for this phenomenon. Class (1) is the failure to report, or to accurately record, smoking history during emergency hospital admissions and other interviews. Face-to-face interviews can introduce bias into the responses to attitudinal and behavioral questions not present in the self-completion interviews typically used to measure smoking prevalence in the general population. Subjects in face-to-face interviews may be unwilling to admit socially undesirable behavior and attitudes under direct questioning. For example, urine testing of hospitalized patients for cotinine showed that smokers were under-counted by 37% because incoming patients failed to inform staff about their smoking behavior. Class (2), various types of “reverse” causation, including a pre-disposition to avoid smoking among COVID-19 patients with tobacco-related comorbidities, may also contribute to the difference between smoking prevalence in the COVID-19 and general population. Patient cohorts hospitalized with COVID-19 may be less prone to use tobacco than the general population. A potentially robust “reverse causation” hypothesis for reduced prevalence of smokers in the COVID-19 population is enrichment of patients in the population with serious comorbidities that have previously motivated them to quit smoking. We ask whether this “smoking cessation” mechanism accounts for a detectable fraction of the reduced prevalence of smokers in the COVID-19 population. Testing this hypothesis will require a focused research program.

Building on past research, we implement a hierarchical latent class model to analyze political participation from a comparative perspective. Our methodology allows simultaneously: (i) estimating citizens’ propensity to engage in conventional and unconventional modes of participation; (ii) classifying individuals into underlying “types” capturing within- and cross-country variations in participation; and (iii) assessing how this classification varies with micro- and macro-level factors. We apply our model to Latin American survey data. We show that our method outperforms alternative approaches used to study participation and derive typologies of political engagement. Substantively, we find that the distribution of participatory types is similar throughout the continent, and that it correlates strongly with respondents’ socio-demographic characteristics and crime victimization.

How do we ensure the accuracy and integrity of a statewide voter registration database, which often depends on aggregating decentralized, sub-state data with different list maintenance practices? We present Bayesian multivariate multilevel model to account for common patterns in local data while detecting anomalous patterns, using Florida as our example. We use monthly snapshots of state's voter database to estimate countywide change rates for multiple response variables (e.g., changes in voter's partisan affiliation), and then jointly model their changes. We show that there is much heterogeneity in how counties manage voter lists, resulting in very different patterns in additions, deletions, or changes of records. Our method identifies several Florida counties with anomalous rates of changes in the 2016 election.

The counting of votes in contemporary American elections is usually not completed on Election Night. There has been an increasing tendency for vote shares to shift toward Democratic candidates after Election Day in general elections, in particular, in recent U.S. elections. Leveraging important snapshots of precinct-level election returns and precinct-level demographic and political composition from Orange County, California, we conduct the first full-fledged analysis of the potential drivers of vote share shifts. Using an original large-scale post-election survey and unique snapshots of individual-level administrative records, we also provide the first analysis of the characteristics of voters whose ballots were tallied later versus earlier in the process. Far from being anomalous, our results indicate that the shifts are consistent with underlying precinct voter compositions and the order of precinct and mail ballot processing at the individual level in accordance with election administration practices. We find the same driving forces in North Carolina and Colorado, and discuss the consequences of the “Blue Shift” for public concerns about election integrity as states push policy changes regarding access to voting by mail.

Objective: What can machine learning tell us about who voted in 2016? There are numerous competing voter turnout theories, and a large number of covariates are required to assess which theory best explains turnout. This article is a proof of concept that machine learning can help overcome this curse of dimensionality and reveal important insights in studies of political phenomena. Methods: We use fuzzy forests, an extension of random forests, to screen variables for a parsimonious but accurate prediction. Fuzzy forests achieve accurate variable importance measures in the face of high‐dimensional and highly correlated data. The data that we use are from the 2016 Cooperative Congressional Election Study. Results: Fuzzy forests chose only a small number of covariates as major correlates of 2016 turnout and still boasted high predictive performance. Conclusion: Our analysis provides three important conclusions about turnout in 2016: registration and voting procedures were important, political issues were important (especially Obamacare, climate change, and fiscal policy), but few demographic variables other than age were strongly associated with turnout. We conclude that fuzzy forests is an important methodology for studying overdetermined questions in social sciences.

Social media data is now widely used by many academic researchers. However, long-term social media data collection projects, which most typically involve collecting data from public-use APIs, often encounter issues when relying on local area network servers (LANs) to collect high-volume streaming social media data over long periods of time. In this paper, we present a cloud-based data collection, pre-processing, and archiving infrastructure, and argue that this system mitigates or resolves the problems most typically encountered when running social media data collection projects on LANs at minimal cloud-computing costs. We show how this approach works in different cloud computing architectures, and how to adapt the method to collect streaming data from other social media platforms. The contribution of our research lies in the development of methodologies that researchers can use to monitor and analyze phenomena including how public opinion and public discourse change in response to events, monitoring the evolution and change of misinformation campaigns, and studying how organizations and entities change how they present and frame information online.

Observers of elections often report that voters have engaged in protest voting. We find that “protest voting” refers to a wide range of behaviors, and we create a taxonomy of these phenomena. Support for fringe or insurgent parties is often labeled as protest voting. Voting theorists have used the term in a completely different way, identifying an unusual type of tactical voting as protest voting. Protest voting also occurs when voters cast blank, null, or spoiled ballots. There are also instances when protest voting is organized and directed by political elites. Finally, several countries provide voters with the option of casting a vote for “None of the Above,” which some see as a form of protest voting. In addition to developing this taxonomy, we discuss the analytical and empirical challenges confronting research on each type of protest voting.

In this article, we study previous experiences with voting technologies, support for e-voting, and perceptions of voter fraud, using data from the 2015 Cooperative Congressional Election Study. We ...

With the discipline’s push toward data access and research transparency (DA-RT), journal replication archives are becoming increasingly common. As researchers work to ensure that replication materials are provided, they also should pay attention to the content—rather than simply the provision—of journal archives. Based on our experience in analyzing and handling journal replication materials, we present a series of recommendations that can make them easier to understand and use. The provision of clear, functional, and well-documented replication materials is key for achieving the goals of transparent and replicable research. Furthermore, good replication materials enhance the development of extensions and related research by making state-of-the-art methodologies and analyses more accessible.

There at least two important questions about voter behavior in California's blanket primary. The first is: How many voters took the opportunity to cast a ballot for a candidate of a party different than that of the voter's registration or identification? The second question is: Why did these crossover voters decide to defect from their own party's nomination campaign and support a candidate from another party? In this chapter, we focus on these two questions at the Assembly district level.

Survey responses in public health surveys are heterogeneous. The quality of a respondent’s answers depends on many factors, including cognitive abilities, interview context, and whether the interview is in person or self-administered. A largely unexplored issue is how the language used for public health survey interviews is associated with the survey response. We introduce a machine learning approach, Fuzzy Forests, which we use for model selection. We use the 2013 California Health Interview Survey (CHIS) as our training sample and the 2014 CHIS as the test sample. We found that non-English language survey responses differ substantially from English responses in reported health outcomes. We also found heterogeneity among the Asian languages suggesting that caution should be used when interpreting results that compare across these languages. The 2013 Fuzzy Forests model also correctly predicted 86% of good health outcomes using 2014 data as the test set. We show that the Fuzzy Forests methodology is potentially useful for screening for and understanding other types of survey response heterogeneity. This is especially true in high-dimensional and complex surveys.

Social scientists have long made use of advanced computational techniques in their research. There have been two very different applications of computational technologies in social science: one largely about simulation of complex political interactions, the other focused on the use of computational approaches with data to estimate models of political phenomenon. In this chapter, our emphasis is generally on the latter area of computational social science, and while we provide some grounding in the early and foundational work in this area, we mainly discuss recent research in computational social science applications involving data and model estimation.

Assuring election integrity is essential for the legitimacy of elected representative democratic government. Until recently, other than in-person election observation, there have been few quantitative methods for determining the integrity of a democratic election. Here we present a machine learning methodology for identifying polling places at risk of election fraud and estimating the extent of potential electoral manipulation, using synthetic training data. We apply this methodology to mesa-level data from Argentina's 2015 national elections.

Given the fundamental role that race and ethnicity play in U.S. society, sensitive survey items on this subject can often lead individuals to underreport their true attitudes. Previous studies have shown that the absence of an interviewer reduces the pressure to provide socially desirable responses. The 2012 and 2016 American National Election Studies (ANES), where both interviewer and self-administered surveys were used, allows us to test whether mode effects emerge in the way respondents answer survey items related to racial attitudes. We also expect mode effects to vary based on the extent to which individuals are politically socialized in the United States. We find that respondents tend to underreport their racial animosity in interview-administered versus online surveys. Moreover, underreporting is nonexistent in the responses provided by foreign-born Latinos, but emerges for U.S.-born Latinos, Blacks, and Whites. These findings pose a number of implications for our understanding of racial attitudes and survey mode.

Supplemental material, appendix_revised_redacted for Evaluating the Quality of Changes in Voter Registration Databases by Seo-young Silvia Kim, Spencer Schneider and R. Michael Alvarez in American Politics Research

Voting advice applications (VAAs) are voter information tools that millions of individuals have used in recent elections throughout the world. However, little is known about how they affect political behavior. Until now, observational studies of VAA have produced inconclusive results. Here we present the results from a randomized field experiment in Switzerland that estimates the causal effects of VAA use on voters’ vote intentions. Our results suggest that usage of the Swiss VAA smartvote strengthened the vote intention for the most preferred party and also increased the number of parties considered as potential vote options. These results imply that VAAs can influence voting behavior and that they can play an important role in electoral politics. published

How do voters make decisions in low-information contests? Although some research has looked at low-information voter decision making, scant research has focused on data from actual ballots cast in low-information elections. We focus on three 2008 Pierce County (Washington) Instant-Runoff Voting (IRV) elections. Using individual-level ballot image data, we evaluate the structure of individual rankings for specific contests to determine whether partisan cues underlying partisan rankings are correlated with choices made in nonpartisan races. This is the first time that individual-level data from real elections have been used to evaluate the role of partisan cues in nonpartisan races. We find that, in partisan contests, voters make avid use of partisan cues in constructing their preference rankings, rank-ordering candidates based on the correspondence between voters’ own partisan preferences and candidates’ reported partisan affiliation. However, in nonpartisan contests where candidates have no explicit partisan affiliation, voters rely on cues other than partisanship to develop complete candidate rankings.

While inferences of traits from unfamiliar faces prominently reveal stereotypes, some facial inferences also correlate with real-world outcomes. We investigated whether facial inferences are associated with an important real-world outcome closely linked to the face bearer’s behavior: political corruption. In four preregistered studies ( N = 325), participants made trait judgments of unfamiliar government officials on the basis of their photos. Relative to peers with clean records, federal and state officials convicted of political corruption (Study 1) and local officials who violated campaign finance laws (Study 2) were perceived as more corruptible, dishonest, selfish, and aggressive but similarly competent, ambitious, and masculine (Study 3). Mediation analyses and experiments in which the photos were digitally manipulated showed that participants’ judgments of how corruptible an official looked were causally influenced by the face width of the stimuli (Study 4). The findings shed new light on the complex causal mechanisms linking facial appearances with social behavior.

Polling and survey methods is an interdisciplinary activity and includes actors in all areas of society, including academia, government, and the private sector. Designing, implementing, and analyzing high-quality, accurate, and cost-effective polls and surveys requires a combination of skills and methodological perspectives. Despite the well-publicized issues that have cropped up in recent political polling, a great deal is known today about how to collect high-quality polling and survey data even in complex and difficult environments. Quality surveys and good survey data are important because social scientists are only as good as the data produced. Therefore, it is critical to follow best practices and guidelines and help researchers assess a variety of factors to make good choices when collecting and analyzing data. Equally important is transmitting those results to others in a clear and accessible manner.

Supplemental material, Appendix for Answering Questions About Race: How Racial and Ethnic Identities Influence Survey Response by Marisa Abrajano and R. Michael Alvarez in American Politics Research

Supplemental material, LinSupplementalMaterial for Inferring Whether Officials Are Corruptible From Looking At Their Faces by Chujun Lin, Ralph Adolphs and R. Michael Alvarez in Psychological Science

Using data from eight recent California initiatives and data provided by the California Secretary of State's Office, this research explores the geographic source of signatures and their distribution across counties, investigates both total signatures and valid signatures, and presents a regression analysis to study how characteristics of counties relate to the number of signatures gathered. The findings indicate a high rate of equity in the distribution of signature gathering and little demographic targeting across measures. The study also finds that a variety of social and political factors influence the number of total and valid signatures across counties, though the results are more consistent for the former.

In this paper we use latent class analysis to identify the four faces of political participation. Previous research has generally focused on conventional forms of political participation (for example, voting), with some research looking as well at unconventional forms of political participation, like protesting. Moreover, most research studies these forms of participation separately. However, citizens actually engage in both conventional and unconventional participation simultaneously, and here we present a methodology that can identify citizens who engage in both, neither, or only one form of participation. Using our approach, we examine a series of hypotheses about how social, political, and economic grievances lead citizens to engage in each face of political participation. We apply this methodology to recent survey data from Argentina, which we argue is an excellent case for studying both forms of participation simultaneously. This application demonstrates the utility of the latent class approach for studying the four faces of political participation.