MIT Digital Media Survey Methodology

What is slant?

When talking about politics, people usually have different views. Some might favor, for instance, an active role of the government in the economy thus supporting higher taxation and expenditure, while others might favor a more limited intervention. To simplify the multidimensional components of a policy or a group of policies, we tend to classify ideologies in a linear scale, usually using the terms "left", "center" and "right" or "liberal", "conservative" and "centrist". In fact, people frequently use terms as left-wing or liberal when the ideology of someone is characterized by an emphasis on ideas such as freedom, equality, fraternity, rights, progress, reform and internationalism while the right-wing or conservative ideology is characterized by someone with an emphasis on notions such as authority, hierarchy, order, duty, tradition, reaction and nationalism. The Democratic and Republican parties in the U.S. has been traditionally considered to have a leftist and a rightist ideology respectively.

Our measure of slant is a value from -1 to 1 that aims to capture how "leftist/liberal" or "rightist/conservative" is a Twitter account based on the publishers it follows (see more details below). The more negative the number is, the more likely an account is associated to publishers that have a "leftist/liberal" bias. Analogously, The more positive the number is, the more likely an account is associated to publishers that have a "rightist/conservative" bias. Twitter accounts with a Slant score close to zero, have a news diet relatively balanced.

How is slant measured?

In a nutshell, we compute the slant of each participant's account following a simple two-stage procedure:

1. Through Twitter's public API, we download the list of accounts that each participant follows and match this against our list of news publishers (mostly online newspapers). A complete list of publishers is available here.


2. For each news publisher the participant follows, we find the slant of that publisher following the methodology and data shared in the academic publication Robertson et al (2018)¹. This methodology calculates the slant of a publisher by comparing the relative number of shares by Democrats and Republicans on Twitter. The full paper is linked below if you would like additional details.

Example: Consider an example participant that, in their Twitter account, follows only two publishers let's say the New York Times (nytimes.com) and Fox News (foxnews.com). According to our dataset, the New York Times is a left-leaning publisher and Fox News is right-leaning, with scores of -0.15 and 0.40 respectively. Thus, the Slant score for the participant will be 0.125 (=(-0.15+0.40)/2). This positive number reflects the fact that on average, the participant is exposed to content with a conservative bias.