Click the different terms below to find out what Twitter in the USA thinks about them!
For more information on how these results were obtained, please visit the methodology section at the bottom of this page.
On the videos, the red areas indicate tweets with negative sentiment and green indicates positive sentiment towards the search terms concerned. The more yellow the area becomes, the more the opinion towards those terms is mixed.
Data up to 01/10/2020
Data based on 48,116 tweets
‘Mail in Ballot’
The map below shows the sentiment of tweets containing the term ‘mail in ballot’ and a ‘sentiment’ analysis of those tweets. Overall we can see that although there are areas of dominant red or green, the majority of the country is divided on the subject.
The map below shows the sentiment of tweets containing the term ‘absentee voting’ and a ‘sentiment’ analysis of those tweets. Overall we can see that although there are areas of dominant red or green, the majority of the country is divided on the subject.
The map below shows the sentiment of tweets containing the term ‘early voting’ and a ‘sentiment’ analysis of those tweets. Overall we can see that although there are areas of dominant red or green, the majority of the country is divided on the subject.
Since their inception, Facebook and Twitter have garnered an increasing number of users who use it both for social purposes and as a source of ‘information’. Whilst wide-spread use of social and internet media played a central role in the 2008 Presidential Election, previous campaigns had endeavoured to use this ‘real-time’ data to better their chances. As one of the most notable users of Twitter, President Donald Trump has encouraged an increasing number of his supporters and opponents to join and/or actively use Twitter as a platform for political discourse. The purpose of our data monitoring is to determine how Twitter users are responding to key electoral issues in the run up to the 2020 US Election. This monitoring allows Democracy Volunteers to identify key problems voters may be experiencing around ‘mail in ballots’, ‘early voting’, ‘absentee voting’ and the like.
What is Twitter Monitoring?
Twitter Monitoring involves utilising all publicly accessible tweets within a defined timeframe and analysing them based on equally definable variables of user, content and location. This process utilises Twitter’s application programming interface (API) to read tweets either liver or base on a retrospective search. This process requires the application and approval of a user for a developer account, after which the use and authorisation of API Keys allows a researcher to access all publically accessible data of a user on a wide scale basis. Figure 1 shows a simplified scheme of how an application searches and analyses Twitter information.
How is sentiment determined?
Sentiment of a given Tweet is divided in three categories based on the contents of said tweet: Positive, Negative and Neutral. Table 1 shows examples of which words are categorised as either positive or negative.
Participation; Love; Like; Preferred; Timely; Right; Easy; Accepted; Successfully. Cheating; Fraud; Issues; Problem; Late; Steal; Cheat; Plot; Outrageous; Bad; Failing.
A sentiment score is applied to each tweet depending on the frequency of positive and negative key terms and is scored on a scale of very negative (-1) to very positive (+1).
For additional information on how Twitter scrapping and analysis may conducted see below:
Bovet, A., Morone, F. and Makse, H.A. (2018) ‘Validation of Twitter opinion trends with national polling aggregates: Hillary Clinton vs Donald Trump’. Scientific Reports, 8(1), pp.1-16.
Grover, P., Kar, A.K., Dwivedi, Y.K. and Janssen, M. (2019) ‘Polarization and acculturation in US Election 2016 outcomes–Can twitter analytics predict changes in voting preferences’. Technological Forecasting and Social Change, 145, pp.438-460.
Yaqub, U., Chun, S.A., Atluri, V. and Vaidya, J. (2017) ‘Analysis of political discourse on twitter in the context of the 2016 US presidential elections’. Government Information Quarterly, 34(4), pp.613-626.