In U.S. presidential elections since 1964, voter turnout among male and female voters has changed gradually but significantly, with women consistently voting at a higher rate than men since the 1980 election. 67 percent of eligible female voters took part in the 1964 election, compared to 72 percent of male voters. This difference has been reversed in recent elections, where the share of women who voted has been larger than the share of men by around four percent since 2004.

According to exit polling in the 2020 Presidential Election in the United States, ** percent of surveyed females reported voting for former Vice President Joe Biden. In the race to become the next President of the United States, ** percent of men reported voting for incumbent President Donald Trump.

According to exit polling data for presidential elections since 1980, women voters tend to gravitate towards the Democratic candidate for president. Since the 1992 presidential election when George Bush ran against Bill Clinton, a higher share of women have supported the Democratic candidate. In the 2020 election, ** percent of women are thought to have voted for Joe Biden, a Democrat, while ** percent voted for Donald Trump, a Republican..

PROBLEM AND OPPORTUNITY In the United States, voting is largely a private matter. A registered voter is given a randomized ballot form or machine to prevent linkage between their voting choices and their identity. This disconnect supports confidence in the election process, but it provides obstacles to an election's analysis. A common solution is to field exit polls, interviewing voters immediately after leaving their polling location. This method is rife with bias, however, and functionally limited in direct demographics data collected. For the 2020 general election, though, most states published their election results for each voting location. These publications were additionally supported by the geographical areas assigned to each location, the voting precincts. As a result, geographic processing can now be applied to project precinct election results onto Census block groups. While precinct have few demographic traits directly, their geographies have characteristics that make them projectable onto U.S. Census geographies. Both state voting precincts and U.S. Census block groups: are exclusive, and do not overlap are adjacent, fully covering their corresponding state and potentially county have roughly the same size in area, population and voter presence Analytically, a projection of local demographics does not allow conclusions about voters themselves. However, the dataset does allow statements related to the geographies that yield voting behavior. One could say, for example, that an area dominated by a particular voting pattern would have mean traits of age, race, income or household structure. The dataset that results from this programming provides voting results allocated by Census block groups. The block group identifier can be joined to Census Decennial and American Community Survey demographic estimates. DATA SOURCES The state election results and geographies have been compiled by Voting and Election Science team on Harvard's dataverse. State voting precincts lie within state and county boundaries. The Census Bureau, on the other hand, publishes its estimates across a variety of geographic definitions including a hierarchy of states, counties, census tracts and block groups. Their definitions can be found here. The geometric shapefiles for each block group are available here. The lowest level of this geography changes often and can obsolesce before the next census survey (Decennial or American Community Survey programs). The second to lowest census level, block groups, have the benefit of both granularity and stability however. The 2020 Decennial survey details US demographics into 217,740 block groups with between a few hundred and a few thousand people. Dataset Structure The dataset's columns include: Column Definition BLOCKGROUP_GEOID 12 digit primary key. Census GEOID of the block group row. This code concatenates: 2 digit state 3 digit county within state 6 digit Census Tract identifier 1 digit Census Block Group identifier within tract STATE State abbreviation, redundent with 2 digit state FIPS code above REP Votes for Republican party candidate for president DEM Votes for Democratic party candidate for president LIB Votes for Libertarian party candidate for president OTH Votes for presidential candidates other than Republican, Democratic or Libertarian AREA square kilometers of area associated with this block group GAP total area of the block group, net of area attributed to voting precincts PRECINCTS Number of voting precincts that intersect this block group ASSUMPTIONS, NOTES AND CONCERNS: Votes are attributed based upon the proportion of the precinct's area that intersects the corresponding block group. Alternative methods are left to the analyst's initiative. 50 states and the District of Columbia are in scope as those U.S. possessions voting in the general election for the U.S. Presidency. Three states did not report their results at the precinct level: South Dakota, Kentucky and West Virginia. A dummy block group is added for each of these states to maintain national totals. These states represent 2.1% of all votes cast. Counties are commonly coded using FIPS codes. However, each election result file may have the county field named differently. Also, three states do not share county definitions - Delaware, Massachusetts, Alaska and the District of Columbia. Block groups may be used to capture geographies that do not have population like bodies of water. As a result, block groups without intersection voting precincts are not uncommon. In the U.S., elections are administered at a state level with the Federal Elections Commission compiling state totals against the Electoral College weights. The states have liberty, though, to define and change their own voting precincts https://en.wikipedia.org/wiki/Electoral_precinct. The Census Bureau... Visit https://dataone.org/datasets/sha256%3A05707c1dc04a814129f751937a6ea56b08413546b18b351a85bc96da16a7f8b5 for complete metadata about this dataset.

AP VoteCast is a survey of the American electorate conducted by NORC at the University of Chicago for Fox News, NPR, PBS NewsHour, Univision News, USA Today Network, The Wall Street Journal and The Associated Press.

AP VoteCast combines interviews with a random sample of registered voters drawn from state voter files with self-identified registered voters selected using nonprobability approaches. In general elections, it also includes interviews with self-identified registered voters conducted using NORC’s probability-based AmeriSpeak® panel, which is designed to be representative of the U.S. population.

Interviews are conducted in English and Spanish. Respondents may receive a small monetary incentive for completing the survey. Participants selected as part of the random sample can be contacted by phone and mail and can take the survey by phone or online. Participants selected as part of the nonprobability sample complete the survey online.

In the 2020 general election, the survey of 133,103 interviews with registered voters was conducted between Oct. 26 and Nov. 3, concluding as polls closed on Election Day. AP VoteCast delivered data about the presidential election in all 50 states as well as all Senate and governors’ races in 2020.

Using this Data - IMPORTANT

This is survey data and must be properly weighted during analysis: DO NOT REPORT THIS DATA AS RAW OR AGGREGATE NUMBERS!!

Instead, use statistical software such as R or SPSS to weight the data.

National Survey

The national AP VoteCast survey of voters and nonvoters in 2020 is based on the results of the 50 state-based surveys and a nationally representative survey of 4,141 registered voters conducted between Nov. 1 and Nov. 3 on the probability-based AmeriSpeak panel. It included 41,776 probability interviews completed online and via telephone, and 87,186 nonprobability interviews completed online. The margin of sampling error is plus or minus 0.4 percentage points for voters and 0.9 percentage points for nonvoters.

State Surveys

In 20 states in 2020, AP VoteCast is based on roughly 1,000 probability-based interviews conducted online and by phone, and roughly 3,000 nonprobability interviews conducted online. In these states, the margin of sampling error is about plus or minus 2.3 percentage points for voters and 5.5 percentage points for nonvoters.

In an additional 20 states, AP VoteCast is based on roughly 500 probability-based interviews conducted online and by phone, and roughly 2,000 nonprobability interviews conducted online. In these states, the margin of sampling error is about plus or minus 2.9 percentage points for voters and 6.9 percentage points for nonvoters.

In the remaining 10 states, AP VoteCast is based on about 1,000 nonprobability interviews conducted online. In these states, the margin of sampling error is about plus or minus 4.5 percentage points for voters and 11.0 percentage points for nonvoters.

Although there is no statistically agreed upon approach for calculating margins of error for nonprobability samples, these margins of error were estimated using a measure of uncertainty that incorporates the variability associated with the poll estimates, as well as the variability associated with the survey weights as a result of calibration. After calibration, the nonprobability sample yields approximately unbiased estimates.

As with all surveys, AP VoteCast is subject to multiple sources of error, including from sampling, question wording and order, and nonresponse.

Sampling Details

Probability-based Registered Voter Sample

In each of the 40 states in which AP VoteCast included a probability-based sample, NORC obtained a sample of registered voters from Catalist LLC’s registered voter database. This database includes demographic information, as well as addresses and phone numbers for registered voters, allowing potential respondents to be contacted via mail and telephone. The sample is stratified by state, partisanship, and a modeled likelihood to respond to the postcard based on factors such as age, race, gender, voting history, and census block group education. In addition, NORC attempted to match sampled records to a registered voter database maintained by L2, which provided additional phone numbers and demographic information.

Prior to dialing, all probability sample records were mailed a postcard inviting them to complete the survey either online using a unique PIN or via telephone by calling a toll-free number. Postcards were addressed by name to the sampled registered voter if that individual was under age 35; postcards were addressed to “registered voter” in all other cases. Telephone interviews were conducted with the adult that answered the phone following confirmation of registered voter status in the state.

Nonprobability Sample

Nonprobability participants include panelists from Dynata or Lucid, including members of its third-party panels. In addition, some registered voters were selected from the voter file, matched to email addresses by V12, and recruited via an email invitation to the survey. Digital fingerprint software and panel-level ID validation is used to prevent respondents from completing the AP VoteCast survey multiple times.

AmeriSpeak Sample

During the initial recruitment phase of the AmeriSpeak panel, randomly selected U.S. households were sampled with a known, non-zero probability of selection from the NORC National Sample Frame and then contacted by mail, email, telephone and field interviewers (face-to-face). The panel provides sample coverage of approximately 97% of the U.S. household population. Those excluded from the sample include people with P.O. Box-only addresses, some addresses not listed in the U.S. Postal Service Delivery Sequence File and some newly constructed dwellings. Registered voter status was confirmed in field for all sampled panelists.

Weighting Details

AP VoteCast employs a four-step weighting approach that combines the probability sample with the nonprobability sample and refines estimates at a subregional level within each state. In a general election, the 50 state surveys and the AmeriSpeak survey are weighted separately and then combined into a survey representative of voters in all 50 states.

State Surveys

First, weights are constructed separately for the probability sample (when available) and the nonprobability sample for each state survey. These weights are adjusted to population totals to correct for demographic imbalances in age, gender, education and race/ethnicity of the responding sample compared to the population of registered voters in each state. In 2020, the adjustment targets are derived from a combination of data from the U.S. Census Bureau’s November 2018 Current Population Survey Voting and Registration Supplement, Catalist’s voter file and the Census Bureau’s 2018 American Community Survey. Prior to adjusting to population totals, the probability-based registered voter list sample weights are adjusted for differential non-response related to factors such as availability of phone numbers, age, race and partisanship.

Second, all respondents receive a calibration weight. The calibration weight is designed to ensure the nonprobability sample is similar to the probability sample in regard to variables that are predictive of vote choice, such as partisanship or direction of the country, which cannot be fully captured through the prior demographic adjustments. The calibration benchmarks are based on regional level estimates from regression models that incorporate all probability and nonprobability cases nationwide.

Third, all respondents in each state are weighted to improve estimates for substate geographic regions. This weight combines the weighted probability (if available) and nonprobability samples, and then uses a small area model to improve the estimate within subregions of a state.

Fourth, the survey results are weighted to the actual vote count following the completion of the election. This weighting is done in 10–30 subregions within each state.

National Survey

In a general election, the national survey is weighted to combine the 50 state surveys with the nationwide AmeriSpeak survey. Each of the state surveys is weighted as described. The AmeriSpeak survey receives a nonresponse-adjusted weight that is then adjusted to national totals for registered voters that in 2020 were derived from the U.S. Census Bureau’s November 2018 Current Population Survey Voting and Registration Supplement, the Catalist voter file and the Census Bureau’s 2018 American Community Survey. The state surveys are further adjusted to represent their appropriate proportion of the registered voter population for the country and combined with the AmeriSpeak survey. After all votes are counted, the national data file is adjusted to match the national popular vote for president.

The IPUMS Contextual Determinants of Health (CDOH) data series provides access to measures of disparities, policies, and counts, by state or county, for historically marginalized populations in the United States including Black, Asian, Hispanic/Latina/o/e/x, and LGBTQ+ persons, and women. The IPUMS CDOH data are made available through ICPSR/DSDR for merging with the National Couples' Health and Time Study (NCHAT), United States, 2020-2021 (ICPSR 38417) by approved restricted data researchers. All other researchers can access the IPUMS CDOH data via the IPUMS CDOH website. Unlike other IPUMS products, the CDOH data are organized into multiple categories related to Race and Ethnicity, Sexual and Gender Minority, Gender, and Politics. The measures were created from a wide variety of data sources (e.g., IPUMS NHGIS, the Census Bureau, the Bureau of Labor Statistics, the Movement Advancement Project, and Myers Abortion Facility Database). Measures are currently available for states or counties from approximately 2015 to 2020. The Politics measures in this release include county-level presidential election results from 2000-2020, indicating the proportion of votes cast for the Democratic candidate or the Republican candidate in each presidential election. To work with the IPUMS CDOH data, researchers will need to use the variable MATCH_ID to merge the data in DS1 with NCHAT surveys within the virtual data enclave (VDE).

The 2020 Taiwan Presidential election and Legislative Yuan election datasets. Raw data files downloaded from the election database of the Central Election Commission.

2020 臺灣總統和立法委員選舉資料集。原始資料下載自中選會選舉資料庫。

TODO: Transform the raw xls files into a more machine-friendly format.

Cover photo credit: The Reporter.

Abstract: We examine how the office a candidate seeks influences their use of feminine stereotypes in campaign messaging, focusing on the interplay between government branches (legislative and executive) and jurisdictions (local, state, and federal). We argue that when a particular office is seen as more feminine, such as school boards or city councils, women candidates will perceive a strategic opportunity to emphasize feminine traits to showcase their qualifications for that office. Women candidates will be more likely to perceive and leverage these strategic opportunities relative to men because of the congruence between being a woman and feminine stereotypes. Drawing on an exhaustive and novel dataset of about 49,000 televised campaign ads from the 2012 to 2020 election cycles, we analyze candidates' strategic use of masculine and feminine traits across local, state, and federal legislative and executive offices. Contrary to our initial expectations, our findings suggest that women candidates do not significantly tailor their use of feminine stereotypes to match the perceived femininity of the office they seek. We find that women and men are more likely to use masculine traits over feminine traits across all offices. We also find that women appear to employ a dual stereotype strategy across offices with a higher likelihood than men of emphasizing both feminine and masculine traits in strategic messages. The results from the analyses of our novel campaign data contribute to an understanding of how candidates strategically emphasize feminine characteristics across contexts, and how scholars consider the gendered dimensions of political offices. Replication code will be available on Dataverse here: https://dataverse.harvard.edu/dataverse/polbehavior. Note that the data we use from the Wesleyan Media Project prohibits us from posting the raw data online. Others can obtain the data we use here: https://mediaproject.wesleyan.edu/dataaccess/. Our raw data file is therefore not publicly posted. Any scholar who purchases the WMP data and signs their contract can email Nichole Bauer, nbauer4@lsu.edu, for our replication data.

According to exit polling in ten key states of the 2024 presidential election in the United States, Donald Trump received the most support from men between the ages of ** and **. In comparison, ** percent of women between the ages of ** and ** reported voting for Kamala Harris.

Included here is the replication data for "What Girls Do: The Effects of Exposure to Women Candidates on Adolescents' Attitudes Toward Women Leaders," as published in Public Opinion Quarterly. There are four files: (1) the Supplementary Material for the published article, which contains the full models and coding details; (2) the STATA do file to run the models; (3) the replication dataset of the Family Matters 2 study in STATA format; (4) the replication dataset of the Notre Dame module of the 2020 Cooperative Election Study, also in STATA format.

According to exit polls for the 2020 South Carolina Democratic primary, former Vice President Joe Biden led the way among voters, receiving ** percent of the vote from men and ** percent of the vote from women. Vermont Senator Bernie Sanders came in second place among both male and female voters.

Compiled French National Assembly election data for 2002, 2007, and 2012 from the Centre de Données Socio-Politiques (CDSP, 2020) and for 2017 from the Plateforme ouverte des données publiques françaises (2020). Includes immigration data for 2012 and derived variables for research on party differences in the Glass Cliff effect for women and ethnic, racial, or immigrant minority candidates.

Percentage of people who voted in the last federal, provincial and municipal elections, by groups designated as visible minorities and selected sociodemographic characteristics (age group, gender, immigrant status, generation status, first official language spoken and highest certificate, diploma or degree).

According to exit polling in the 2020 Presidential Election in the United States, ** percent of surveyed 18 to 29 year old voters reported voting for former Vice President Joe Biden. In the race to become the next president of the United States, ** percent of voters aged 65 and older reported voting for incumbent President Donald Trump.

This is information on the candidates for the General Election 2020, including name, gender, political party, votes received and their constituency.

Context

The 2020 US election is happening on the 3rd November 2020 and the resulting impact to the world will no doubt be large, irrespective of which candidate is elected! After reading the two papers, here and here, I was inspired to attempt a similar sentiment analysis myself!

Content

Tweets collected, using the Twitter API statuses_lookup and snsscrape for keywords, with the original intention to try to update this dataset daily so that the timeframe will eventually cover 15.10.2020 and 04.11.2020. Added 06.11.2020 With the events of the election still ongoing as of the date that this comment was added, I've decided to keep updating the dataset with tweets until at least the end of the 6th Nov. Added 08.11.2020, just one more version pending to include tweets until at the end of the 8th Nov.

Columns are as follows: - created_at: Date and time of tweet creation - tweet_id: Unique ID of the tweet - tweet: Full tweet text - likes: Number of likes - retweet_count: Number of retweets - source: Utility used to post tweet - user_id: User ID of tweet creator - user_name: Username of tweet creator - user_screen_name: Screen name of tweet creator - user_description: Description of self by tweet creator - user_join_date: Join date of tweet creator - user_followers_count: Followers count on tweet creator - user_location: Location given on tweet creator's profile - lat: Latitude parsed from user_location - long: Longitude parsed from user_location - city: City parsed from user_location - country: Country parsed from user_location - state: State parsed from user_location - state_code: State code parsed from user_location - collected_at: Date and time tweet data was mined from twitter*

Acknowledgements

• @taniaj and her great Australian Election 2019 Tweets dataset, I used as a template for this dataset.

• Thanks to Twitter for providing the free API and snsscrape to allow collection of the tweet_ids.

Cover photo by Jorge Alcala on Unsplash Unsplash Images are distributed under a unique Unsplash License.

Inspiration

My primary interest for creating this dataset is to ascertain if there is a correlation between the sentiment of users on Twitter and the eventual election results. Other ideas that might be interesting to investigate include:

• Can we detect if there are or were any attempts at manipulating the election.
• Can we predict the candidate from tweet text only.
• Can we predict the election outcome of each state.

I also included still valid and interesting ideas from the Australian Election 2019 Tweets dataset below:

• Take into account retweets and favourites to weight overall sentiment analysis.
• Which parts of the world are interested (ie: tweet about) in the US elections, apart from the US?
• How do the users who tweet about this sort of thing tend to describe themselves?
• Is there a correlation between when the user joined Twitter and their political views (this assumes the sentiment analysis is already working well)?
• Predict gender from username/screen name and segment tweet count and sentiment by gender

Version

• Version 3 - 355,000 tweets collected, using the Twitter API statuses_lookup and snsscrape for keywords between 15.10.2020 and 22.10.2020.

• Version 5 - New tweets collected for the date of 23.10.2020, with a new total number of tweets at around 387,000 tweets.

• Version 6 - New tweets collected for the date of 24.10.2020, with a new total number of tweets at around 418,000 tweets. Additionally the "coordinates" column was removed with "lat" and "long" columns added for geolocation data (where possible).

• Version 7 - New tweets collected for the date of 25.10.2020, with a new total number of tweets at around 456,000 tweets. Added column "collected_at" to indicate when the data was mined from twitter. *Note this data is only accurate from 21.10.2020 onwards, data in the subject column before this date is an estimation.

• Version 8 - New tweets collected for the date of 26.10.2020, with a new total number of tweets at around 49...

Online poll on the day of the 8 February 2020 General Election in the Republic of Ireland. The poll is executed by Ireland Thinks (https://www.irelandthinks.ie) based on their online panel, which was alerted by SMS on the day of the election. The survey data is collected using SurveyGizmo (https://www.surveygizmo.com). Among the first questions is a fake ballot with all candidates in the respective constituency, which has been used to populate the vote choice variables. The age and gender variables were based on earlier data collection on the same panel.

A list that includes everyone on the electoral database for the election of the President of 2020. The list includes information on gender, age, nationality, whether legal domicile is in Iceland, constituency and municipality.

Since 1964, voter turnout rates in U.S. presidential elections have generally fluctuated across all age groups, falling to a national low in 1996, before rising again in the past two decades. Since 1988, there has been a direct correlation with voter participation and age, as people become more likely to vote as they get older. Participation among eligible voters under the age of 25 is the lowest of all age groups, and in the 1996 and 2000 elections, fewer than one third of eligible voters under the age of 25 participated, compared with more than two thirds of voters over 65 years.

Election return data from all Democratic and Republican primaries for the US House of Representatives for 1956-2010. Includes district-specific variables, as well as data about candidate backgrounds and gender in recent years. For an appendable version of the 2012-2018 data, see Miller/Camberg 2020: https://doi.org/10.7910/DVN/CXVMSY