Skip to main content
ABC News
Silver Bulletpoints: The Union Vote Could Swing The Election

Today is Thursday, May 2, and this is Silver Bulletpoints, the column where we tackle three topics related to the 2020 primary in 300 words or less. Today’s edition is sponsored by EATT, the Endowment for the Advancement of Taco Trucks, an organization I just made up that advocates for the placement of taco trucks on every American street corner.

Bulletpoint No. 1: The union vote could be key in both the primary and the general election

In 2016, unionized workers were essential in helping Hillary Clinton maintain her lead over Bernie Sanders. Although a handful of major unions endorsed Sanders, most of the biggest ones — including the NEA, the SEIU, AFSCME, UFCW and the UAW — endorsed Clinton. And while some of those unions’ members defected to Sanders, Clinton mostly held the support of their rank and file, with Clinton winning union voters 62-36 over Sanders, per the Cooperative Congressional Election Study.

So it’s not surprising that Joe Biden kicked off his campaign on Monday by touting his endorsement from the International Association of Fire Fighters. It’s also not surprising that President Trump, perhaps sensing that Biden was encroaching on his turf, spent much of Wednesday morning retweeting accounts from firefighters who said they were planning to vote for Trump. Most union members voted Democratic in 2016 — but Trump did much better with them than Mitt Romney had four years earlier.

Nor was Trump’s union support merely a matter of white men shifting en masse to Trump. While white women and nonwhite men in unions mostly voted for Clinton, her margins with those groups were considerably narrower than Barack Obama’s in 2012.

Union voters shifted toward Trump in 2016

Presidential vote share among union members in 2012 and 2016, by race and gender

All labor union members
Year Margin
2012 Obama 64.8% Romney 30.4% +34.4
2016 Clinton 55.2 Trump 38.4 +16.7
White men in labor unions
Year Margin
2012 Obama 52.3% Romney 41.9% +10.5
2016 Clinton 40.7 Trump 52.5 -11.9
Nonwhite men in labor unions
Year Margin
2012 Obama 81.4% Romney 13.9% +67.5
2016 Clinton 73.2 Trump 18.6 +54.7
White women in labor unions
Year Margin
2012 Obama 64.5% Romney 31.0% +33.5
2016 Clinton 55.7 Trump 38.6 +17.2
Nonwhite women in labor unions
Year Margin
2012 Obama 88.5% Romney 8.9% +79.7
2016 Clinton 83.0 Trump 12.9 +70.2

2012 election results reflect voters who were union members as of 2016 and participated in the 2016 CCES


In fact, the shift among union voters was enough to swing the election to Trump. According to the CCES, Obama won union voters by 34.4 percentage points in 2012, but Clinton did so by only 16.7 points in 2016. That roughly 18-point swing was worth a net of 1.2 percentage points for Trump in Pennsylvania, 1.1 points in Wisconsin and 1.7 points in Michigan based on their rates of union membership1 — and those totals were larger than his margins of victory in those states.

Bulletpoint No. 2: Is Elizabeth Warren’s outspokenness on impeachment driving her mini polling surge?

If it wasn’t for Biden’s big polling surge, everyone might be talking about Elizabeth Warren’s improving numbers instead. In the four national polls out this week, Warren gained an average of 3 percentage points, even surpassing Sanders for second place in one of them — although that result remains an outlier for now.

The timing of Warren’s gains in the polls lines up reasonably well with the release of the Mueller report, after which she became one of the most outspoken Democrats in calling for Trump’s impeachment. Although impeachment proceedings are unpopular with the general electorate — 56 percent of Americans are opposed to impeachment hearings and 37 are in favor, according to the latest ABC News/Washington Post poll — a 62-29 majority of Democrats are in favor, including 53 percent strongly in favor.

Counterargument: It’s not clear that impeachment is the highest priority for Democrats. In CNN’s poll this week, only 43 percent of Democrats said it was “very important” for a presidential candidate to support impeaching Trump — much less than the number who said the same about climate change (82 percent), providing health insurance for all (75 percent) or passing stricter gun laws (65 percent).

Nonetheless, in a crowded field, any issue that allows candidates to differentiate themselves probably helps. And in what’s liable to be a relatively dull stretch of the campaign until the first debates on June 26-27, the ongoing arguments about Mueller and impeachment will provide candidates an opportunity to make news. That includes Warren and the candidates on the Senate Judiciary Committee, such as Kamala Harris, who was effective in her questioning of Attorney General William Barr on Wednesday.

Bulletpoint No. 3: Stop complaining about small sample sizes

Surveys showing bad numbers for candidates who are popular on the internet tend to produce lots of specious criticism about the polls. This week, for instance, some Sanders backers complained about the small sample size in CNN’s poll of Democrats, which surveyed 411 people, and they erroneously claimed that it hadn’t surveyed voters under the age of 50 when it had.

While 411 voters isn’t a huge sample, it’s fairly typical as primary polls go, especially for high-quality polls like CNN’s that make an effort to contact a representative sample of voters. The Democratic electorate contains lots of voters in hard-to-reach groups, such as college students and Latinos. Moreover, turnout in the primaries is relatively low. All of that plus ever-declining survey response rates make primary polling expensive; you have to place a lot of phone calls to find enough actual Democratic voters.

But complaining about sample size doesn’t make sense when there are several polls and they all show the same thing. CNN’s wasn’t the only poll to show Sanders declining; so did every other national survey since Biden’s announcement, including Morning Consult’s poll, which surveyed more than 15,000(!) voters in its online panel.

Nor are small-sample polls quite as error-prone as you might think. The margin of error in polls depends not only on the sample size but also on how many people choose a particular option. The further away a candidate is from 50 percent, the less the margin of error — so you can get away with a smaller sample when lots of candidates are polling in the single digits or low double digits. For instance, for a candidate at 15 percent (about where Sanders is), the MOE is a comparatively modest plus or minus 3.5 points even if the poll surveys only 400 people.

How big is the margin of error?

Margin of error with a 95 percent confidence interval by sample size and vote share of candidate

Sample size
Candidate vote share 200 400 600 1,000 2,000 10,000
90% ±4.2 ±2.9 ±2.4 ±1.9 ±1.3 ±0.6
75 6.0 4.2 3.5 2.7 1.9 0.8
50 6.9 4.9 4.0 3.1 2.2 1.0
40 6.8 4.8 3.9 3.0 2.1 1.0
30 6.4 4.5 3.7 2.8 2.0 0.9
20 5.5 3.9 3.2 2.5 1.8 0.8
15 4.9 3.5 2.9 2.2 1.6 0.7
10 4.2 2.9 2.4 1.9 1.3 0.6
5 3.0 2.1 1.7 1.4 1.0 0.4
2 1.9 1.4 1.1 0.9 0.6 0.3

From ABC News:

Trump unleashes tweetstorm amid Biden firefighter union endorsement

CORRECTION (May 2, 2019, 2:11 p.m.): A previous version of the second table in this article incorrectly labeled the top row. That row showed the margin of error for a candidate with 90 percent vote share, not 95 percent.

Check out our 2020 polls tracker.


  1. According to the CCES.

Nate Silver is the founder and editor in chief of FiveThirtyEight.