A national Washington Post-ABC News poll earlier this month asked Americans what issue is most important in their vote — with no options provided — and found 28 percent naming the economy and jobs. That was followed by terrorism and foreign policy (12 percent), immigration (11 percent), ethics (8 percent) and a slew of others.

The results make broad sense — the economy is important, and its health is a good predictor of which party wins elections — but as a Fix reader, you might notice these top-named issues are different from the high-decibel debates the candidates have waged over fighting terrorism, racial inequality, deportation of undocumented immigrants and economic inequality.

And it's not just the candidates. On social media, the politically inclined seem to be much more interested in talking about these other issues and are comparatively uninterested in the economy, according to a new analysis.

Tapping into a random sample of ordinary political conversations is practically impossible, but it is feasible to analyze what issues people are talking about on social media networks. While the public and self-selected nature of social media interactions gives greater voice to those who are most politically engaged and technologically capable, the sheer volume and interconnected nature of discussions make them worth a look.

To investigate, we asked the data scientists at MIT's Laboratory for Social Machines to analyze Twitter conversation data during the same period the poll was conducted — March 3-6. The MIT team has access to Twitter's complete fire hose and has honed a classification algorithm to identify tweets that mention any presidential candidate along with terms related to 20 election issues that crop up most. (More about the Twitter classification methodology here.)

The chart below summarizes the comparison — with several key caveats. This is by nature an apples-to-oranges comparison, not only in the differences in populations and activities, but the issue classification schemes for the poll and Twitter were developed separately, so categories are only roughly comparable.


Among the most striking differences, foreign policy and racial issues play a far bigger role in Twitter conversation. Fully 36 percent of all election-related tweets mentioned such issues, while only 12 percent of Americans said foreign policy was their top issue (most mentioning terrorism) in the Post-ABC poll.

The content of tweets was wide-ranging, with top terms including everything from #Benghazi to the "iraq war," as well as "torture laws" and #israel. Some of the most retweeted foreign policy tweets include Donald Trump's proposal to use torture against captured Islamic State fighters, the German vice chancellor's comments on Trump and a criticism of Bernie Sanders through an image of strife in Venezuela.

The heavy focus of Twitter election discussion on foreign policy is nothing new. It also was the largest component of Twitter discussion from November through mid-December — a period encompassing high-profile terrorist attacks in Paris and the United States.


Proportion discussing each election-related issue on Twitter. Data from MIT Laboratory for Social Machines. Graphic by The Washington Post.

The second big difference between voting priorities and Twitter discussion was on racial issues. Nearly 1 in 8 election-related tweets focused on racial issues, while no respondents in the Post-ABC poll volunteered this as their top voting issue. It's worth noting that race and race relations were mentioned by 5 percent of Americans in a February Gallup poll asking the nation's most important problem, so it's not as though it is never cited as a top issue.

Trump was the dominant focus of race-related tweets, featuring accusations of racism and mentions of "Saturday Night Live's" sketch mocking Trump's supporters as racists and defenses of Trump.

The Twitter analysis found economic tweets made up a significant portion of the discussion (8 percent), but this was far less than the share who said it was their top voting issue (28 percent). Campaign finance was also discussed more prominently on Twitter than named as the voters' top deciding issue (5 percent vs. zero percent).

One interesting similarity between the two data sets is that a similar share of Americans named immigration as their top voting issue (11 percent) as the issue made up within the Twitter discussion. Given the big focus on race and international issues and Trump's controversial proposals on the issue, immigration might be expected to be bigger on Twitter than with the public at large.

What do the big gaps between Twitter's and America's top voting issues nationally mean? The reasons for the divergence are likely numerous, most notably the differing basic populations being examined (tweets vs. general public); those who tweet are different from those who don't. An additional and not mutually exclusive possibility is that the economy is seen as very important in the abstract but more difficult to apply when discussing individual candidates, at least in 140-character messages. Another is that many tweets receiving widespread attention — as measured by the number of retweets — were passionate expressions of support and opposition to candidate's statements or proposals.

Special thanks to Bill Powers and data scientists Soroush Vosoughi and Prashanth Vijayaraghaven at MIT's Laboratory for Social Machines for conducting this custom analysis.