About this poll
POLITICO’s polling partner, Morning Consult, surveys more than 5,000 registered voters across the United States every day. This poll represents one week’s worth of responses from likely Democratic primary voters.
For each poll, Morning Consult asks those likely voters to tell us their first and second choice among the field of candidates vying to be the Democratic nominee for president. Morning Consult also asks how they feel about each candidate: whether they have a favorable or unfavorable opinion, or whether they’re unfamiliar with a candidate.
A balanced base
Ultimately, candidates are searching for broad appeal with as many voters as possible. But even though the number of supporters each candidate can count varies widely, the makeup of candidates’ support among different voter groups can show interesting strengths and weaknesses in their campaigns. These charts compare the demographic breakdown of voters supporting each candidate to all likely Democratic primary voters in our poll.
Think of each chart as a child’s paper chatterbox. If a specific demographic group is under-represented among a candidate’s supporters — a smaller share than they are of voters at-large — the chart will fold in. If a group is over-represented — a larger share — it will fold out. Both could suggest an imbalance in the campaign’s support. The question is whether candidates can make up ground with groups that are under-represented in their base or if they can use an over-represented group to make a wider case for their candidacy.
How to read these charts
Compare candidates’ supporters by:
Age & Gender
Interest in Politics
Race & Community
Race & Education
Race & Gender
Demographic charts methodology
The number of supporters each candidate has varies widely, but for these charts, we wanted to compare the makeup of their support. For each of the top ten candidates, we want to know whether a demographic group is over or under-represented among their supporters. That means measuring whether a candidate has more or fewer supporters of one group than we’d expect based on how many voters belong to that group overall. We used proportions to compare across candidates.
For each demographic group we calculated the difference between the proportion of a demographic group among all likely Democratic voters and that group’s share of a candidate’s supporters (those who said that candidate was their first choice to be the Democratic nominee). For example, if Hispanic voters made up 10% of all likely voters but 15% of a candidate’s supporters in the same poll, we’d show that candidate as +5 for Hispanic support.
On our charts, we show that data point on a range set by the candidate with the largest score. So, say another candidate had a difference of -10 (0% Hispanic supporters), we’d show our original candidate at +5 halfway up a scale that runs to +10, the absolute value of the largest score.
There are several more ways we can describe that +5 score. We chose to use the “absolute risk difference” in the cards that appear when you tap our charts. Absolute risk difference describes the difference in the “risk” or likelihood that something will be true for you depending on whether you’re a member of one group versus another. (It’s often used in medicine to measure clinical outcomes between different treatment groups.)
In our hypothetical case, we’re looking for the difference in the likelihood that people will be Hispanic if we pick them randomly from both the candidate’s supporters and from all likely Democratic voters. Since the likelihood that someone is Hispanic is 15% (think 15 out of 100) among the candidate’s supporters and 10% (10 / 100) among all voters, we say that a random supporter was 5% (15 – 10 = +5 / 100) more likely to be Hispanic than a random voter was in our poll.
That’s just one way to interpret the result of our analysis, and we always welcome arguments for better alternatives. Send them to email@example.com.