To continue our political media monitoring and tracking of the US Democratic Party presidential primaries and our June update, here we track and analyse media coverage for the major democratic candidates in the month of July.
Using Artificial Intelligence (AI), we have tracked every time the candidates are referred to in the media throughout July. As the race is moving on, this month we are expanding our index. We’ve conducted a deep-dive into media coverage in Iowa, where the first of the Democratic caucuses and primaries will take place.
|8||Bill de Blasio||15,433||23,498||15,422||18,691||29,261||49,048||43,725||195,078|
July saw less coverage on the primaries than June (around 75% of June’s total).
|5 (+1)||Pete Buttigieg||49,080|
|6 (-1)||Cory Booker||44,848|
|7 (+1)||Bill de Blasio||43,725|
|8 (-1)||Beto O’Rourke||36,055|
We have tracked the coverage of the candidates in 45 Iowan news sources. We analysed the number of pieces each candidate is in, the number of unique references to each candidate, and the sentiment towards each candidate in every news piece they appear in.
For an explanation of our methodology in examining sentiment, and a list of the 45 outlets examined, see below.
Following the dramatic spike in coverage of Kamala Harris in June, we wanted to investigate her prominence in the media in Iowa. While she is a close fourth in the US as a whole, the data from Iowa suggests that until recently she has been lagging behind Biden, Sanders and Warren.
Additionally, Elizabeth Warren was in the lead alongside Sanders and Biden in Iowa. However, she hasn’t been able to sustain this. The articles that she has been appearing in mention her less often than other top candidates. (See below in the comparison of total mentions to total number of articles.)
The last two months have been seen Biden and Sanders jostling for the top spot. But in the wake of the July debates Sanders has over taken Biden.
This data is available upon request, simply email [email protected].
When analysing the sentiment of news coverage in Iowa, it suggests that Biden is the most divisive candidate. When compared to coverage of Sanders and Warren, a similar number of articles reflect on him positively. However, significantly more bear negative sentiment towards him.
In fact, almost 12% of articles towards Biden were negative. This is significantly more than the next candidate, Hickenlooper, with just over 7%.
Candidates with an exceptionally high portion of positive coverage include Bullock and Castro. Moulton and Tang were the only candidates with more negative coverage than positive.
Signal AI’s sentiment model assesses sentiment at the ‘entity’ level. This means it is able to detect sentiment for more than one organisation, person, or location in the same article. For this project, our AI has been trained to recognise the candidates as entities. For more information on what an entity is see here.
Our AI seeks to recognise words or phrases against a dictionary built by our researchers based on state-of-the-art linguistics research. Each word or phrase in the dictionary has a sentiment ‘intensity’ attached to it, on a scale of -1 to +1. The AI also negates double negative sentiment (for example, “not bad” is seen as a positive phrase) and overrides known colloquialisms (“that’s sick” becomes positive).
Sentiment from each individual word/phrase is combined to give the article an overall sentiment score for each entity based on its mentions, where:
Image Source: Cropped use of U.S. Senator Elizabeth Warren speaking with attendees at the 2019 National Forum on Wages and Working People hosted by the Center for the American Progress Action Fund and the SEIU at the Enclave in Las Vegas, Nevada (27 April 2019), by Gage Skidmore via flickr.com.