April 4, 2023

The power of word choice

Does slanted language affect public opinion?

Source: hadrian

US political parties use noticeably different language when talking about the exact same issues. Republicans discuss “gun control” and the “death tax,” whereas Democrats  refer to “gun safety,” and the “estate tax.” When talking about immigration, Republicans prefer “illegal immigrants,” while Democrats tend to use the phrase “undocumented immigrants.”

The hope is that the right word choice will give their preferred policies an edge in the public’s mind. But do these variations in vocabulary really make a difference? 

In a paper in the American Economic Review, author Milena Djourelova argues that it does. By examining a ban on the politically charged term “illegal immigrant” by the Associated Press (AP) newswire, she found that slanted language significantly lowered support for restrictive immigration policies.

The AP newswire is a significant source of news content for many US newspapers and broadcasters across the ideological spectrum. Moreover, member outlets often use AP-produced content with little to no changes.  

In April 2013, the AP abruptly changed its guidelines on the acceptable terminology to be used to refer to people living in the United States without legal authorization, switching from recommending “illegal immigrant” to banning its use in AP copy.

 Using the text of all immigration-related AP dispatches released between 2009 and 2017, Djourelova first documented how the ban affected the AP’s coverage of immigration.

She found that the label “illegal immigrant” virtually disappeared from the AP’s feed after the ban—largely through the omission of direct references to legal status. But importantly, other dimensions of the AP’s content, such as volume of immigration coverage, slant, and sentiment, did not change with the ban.


Slant and sentiment over time
The chart below shows the slant and sentiment of AP dispatches on immigration over time. Panel A plots a slant index, which measures the similarity of AP language to the language used by Republicans (positive values on y-axis) versus Democrats (negative values on y-axis) in congressional speech. Panel B plots a sentiment polarity index, which measures positive versus negative sentiment. The vertical red lines indicate when the ban went into effect.
Source: Djourelova (2023) 


The ban in turn changed the way that many outlets covered immigration. Outlets more reliant on the AP for content used the term “illegal immigrant” much less frequently, but the decline was more notable for left-leaning outlets.

As the shift in language spread through the media ecosystem, it also moved opinions on immigration. Using a large electoral survey, Djourelova compared public opinion on immigration before and after the ban. In particular, she compared counties with newspaper circulations that relied more or less heavily on AP content (“AP intensity.”)

She found that a one standard deviation increase in AP intensity was associated with 0.8 percentage point lower support for increasing border security after the ban—roughly 2 percent of the gap between Republicans and Democrats. These results disappeared when the circulation intensity of another major newswire, Reuters, was used for comparison. Reuters didn’t adjust its language related to immigration until 2014, and, even then not as abruptly as the AP.

There was a similar shift in support for other immigration-related issues, such as allowing police to question suspected unauthorized immigrants, fining firms that employ them, and prohibiting their access to public services. But there was no effect on other policy issues, such as healthcare and gay marriage.

One lesson is that partisan language can be used in a way that may influence what some people think about policy-relevant issues. Perhaps we should be careful about how we use language and think about what consequences it may have.

Milena Djourelova

Moreover, these changes in opinion were concentrated among self-described moderates with less exposure to immigrants, which is consistent with the theory that the change in language was responsible for persuading respondents.

The findings may be important for media outlets to consider. 

“One lesson is that partisan language can be used in a way that may influence what some people think about policy-relevant issues,” Djourelova told the AEA. “Perhaps we should be careful about how we use language and think about what consequences it may have.”

Persuasion through Slanted Language: Evidence from the Media Coverage of Immigration appears in the March 2023 issue of the American Economic Review.