2014 Midterm Election Language Sentiment Analytics

midterm elections

Are elections won by the most “negative” candidate? A 2014 study from Cambridge University found a connection between a person’s political views and the strength of their negativity bias – that is, their tendency to respond more strongly to negative than positive information. With the 2014 midterm elections happening today, we decided to analyze the language of political candidates to learn whether there is a significant difference in how the parties talk about major issues.

According to the Cambridge study, the negativity bias is a spectrum – some people respond more strongly to negative events than others – and conservatives are more likely to have a strong negativity bias. For example, conservatives are more likely to support strict punishments for criminals and defense spending in order to avoid threats to safety.

Negative language is influential not just in politics but in other forms of communication. Studies of negativity bias and relationships have found that healthy couples must be able to list five positive things about each other for each negative observation. Media studies have found a negative to positive news report ratio of seventeen to one. Academic research found the highest-performing business teams offered each other, on average, 5.6 positive comments as feedback for every criticism. As much as we pretend otherwise, negative language can have a lot of impact.

We were curious whether a strong negativity bias is reflected in communication styles, and whether negativity bias could be an accurate predictor for election results.

What we did:

We analyzed the language of 40 candidates in the 2014 senate midterm elections on a positive vs. negative spectrum. We focused on candidates from five of the closest races for this analysis. In order to keep the language comparable, we took public statements from each candidate regarding the same major issues. We also measured the language of several Republicans and Democrats already in office to use as a baseline.

What we found:

  1. The language of the politicians that we analyzed showed strong emotional content. Of the political candidates analyzed, 70% used above average levels of positive language and 97% used above average levels of negative language, when compared to the average person.
  2. The language of the Republican candidates analyzed was more negative than the language of the Democratic candidates analyzed. On average, the Republican candidates used 11.6% more negative language than the Democratic candidates.

The following graph shows the sentiment scores from the candidates of five of the closest 2014 senate races. The bars represent the range of sentiment displayed and the dots represent each candidate’s overall sentiment score (positive language minus negative language).

How will these results impact the 2014 midterm elections?

Communication analytics provide more evidence of a strong negativity bias in conservatives.

  • In three out of five of the close races (AK, IA, KS), the Republican candidate used more negative language than their opponents.
  • In the races where the Democratic candidates used more negative language (NC, GA), the negative outweighed the positive by only a small margin.

We look forward to evaluating whether the levels of positive or negative sentiment expressed by these candidates will affect the voters’ decisions today.