Reminders About Polling
A new Washington Post-ABC Poll so alarmed its authors that they discredited their own work and buried the lede - who voters look to blame if there's a government shutdown.
After Friday’s earthquake federal indictment (second in eight years) of New Jersey’s Senior US Senator and Chair of the Foreign Relations Committee, the political world woke up to another one on Sunday morning.
One of the legacy media’s go-to national surveys, sponsored by the Washington Post and ABC News, shows Donald Trump leading incumbent Joe Biden by 10 points.
Legacy media and fellow travelers in the political polling world immediately shouted, “Outlier!” Even the poll’s sponsors called it an outlier: “The difference between this poll and others, as well as the unusual makeup of Trump’s and Biden’s coalitions in this survey, suggest it is probably an outlier,” wrote the Post’s Dan Balz and two others and even put it on the subhead at their website, to drive home the point.
Not so fast. It is consistent with previous Post-ABC polling, said Pollster Gary Langer:
“Head-to-head in a hypothetical November 2024 matchup, Trump has 51% support while Biden has 42% -- numerically up 3 points for Trump and down 2 points for Biden from an ABC/Post poll in February, shifts that are not statistically significant.
“There's even less change from the most recent ABC/Post poll in May,” he added, “which had the race at 49-42% (again with a different, but comparable, question wording).”
If it’s consistent with your past two polls, where’s the “outlier?” Sounds and reads like so much spin to me. That dog doesn’t hunt.
Langer:
In terms of sampling, this survey was conducted using the ABC/Post poll's longstanding methodology. Demographic results are typical. So are partisan preferences; 25% of respondents identify themselves as Democrats, 25% as Republicans and 42% as independents. Forty-one percent are Democrats or independents who lean toward the Democratic Party; 45% are Republicans or lean toward the GOP, consistent this year.
Additionally, survey respondents who say they voted in 2020 report having supported Biden over Trump by 50-46%, very close to the actual outcome, 51-47%.
Some partisans quickly noted an NBC poll released Sunday, showing a race between Trump and Biden tied. However, there are notable questionnaire and party identification differences. NBC poses the ballot test late in the survey; WaPo-ABC put it up front. The latter poll included more self-identified independents; NBC features more partisans, with 40 percent Democrats and 39 percent Republicans (both numbers seem high). Both polled registered voters: 848 by NBC and over 1,000 by the WaPo-ABC folks. It’s way too early to screen for “likely voters” at this point.
These seemingly little and rather technical things underscore that polling is both an art (asking and ordering questions, guestimating the composition of the electorate, etc.) and a science, not a very exact one.
Disclaimer: I’m not a pollster nor a polling expert, nor did I stay at a Holiday Inn last night. But over 40 years of consuming and analyzing polls for political campaigns and working and learning directly from pros have helped me separate the wheat from the chaff.
This is your reminder that polls are (cliche alert, but still valid) snapshots in time and do not predict outcomes (although prognosticators use polling to make their own predictions). Things like the number and sample of voters culled (registered voters, random digit dialing, etc.), how they’re contacted (landline and cell phones, online, texts, etc.), when they’re contacted (the “field dates”), and the questions and their order all matter.
The New York Times in 2016 ran an experiment. They gave four separate pollsters the raw data from a survey of Florida voters. That poll by Siena College gave Hillary Clinton a one-point lead over Donald Trump. The lone Democratic poller, Penn Schoen, gave Clinton a three-point lead; the Republican firm, Echalon Insights, matched the Siena results. Others gave Trump a one-point lead and Clinton a four-point lead.
Trump won Florida by +2, over 100,000 votes.
The WaPo-ABC survey has a history of outliers and may be worse. In October 2020, days before the Presidential election in Wisconsin, their survey gave Biden a 17-point lead. But on election day, Biden won by only 20,000 votes, a .6 percent margin of victory. Was it an outlier or a suppression poll?
The real value of any well-crafted poll, 16 months out from the general election, is how it measures voter attitudes on various factors that shape the political environment. The Washington Post-ABC delivers on that in spades. We’ll let Langer summarize the key findings (emphasis added).
President Joe Biden's job approval rating is 19 points underwater, his ratings for handling the economy and immigration are at career lows. A record number of Americans say they've become worse off under his presidency, three-quarters say he's too old for another term and Donald Trump is looking better in retrospect -- all severe challenges for Biden in his reelection campaign ahead.
Forty-four percent of Americans in the latest ABC News/Washington Post poll say they've gotten worse off financially under Biden's presidency, the most for any president in ABC/Post polls since 1986. Just 37% approve of his job performance, while 56% disapprove. Still fewer approve of Biden's performance on the economy, 30%.
On handling immigration at the U.S.-Mexico border, Biden's rating is even lower, with 23% approval. In terms of intensity of sentiment, 20% strongly approve of his work overall, while 45% strongly disapprove. And the 74% who say he's too old for a second term is up 6 percentage points since May. Views that Trump is too old also are up, but to 50% in this poll. . .
And just like that, Democrats aren’t quoting Bill Clinton’s 1992 campaign manager, James Carville, anymore: “It’s the economy, stupid.” So much for the “Bidenomics” spin job. More voters are concerned by Biden’s age (74%) than Trump’s 91 criminal indictments (62%).
To me, the real shocker of the poll was who would get the blame for a government shutdown. Langer again: “Such is down-on-Biden sentiment that if a government shutdown occurs at month's end, 40% say they'd chiefly blame him and the Democrats in Congress, versus 33% who'd pin it on the Republicans in Congress -- even given the GOP infighting behind the budget impasse.”
That doesn’t seem right, but it should be a wake-up call to House Democratic Leader Hakeem Jeffries and Senate Majority Leader Chuck Schumer, who have sat on their hands while House Republicans fight among themselves. Shutdowns are never popular. Conversely, it may only inspire the small band of House GOP recalcitrants who seem to be cheering for a shutdown, but that would be misreading the numbers.
Yes, Bidenomics contrasts very nicely with Donald Trump’s record, so much so that Trump’s job approval is up by ten points over when he left office.
A third poll was released this weekend and deserves much more attention. “On the issue of free expression, at least, Republicans are not the authoritarian party,” wrote RealClearPolitics.com editor Carl Cannon. “That distinction belongs to the Democrats, the party launched by Thomas Jefferson — the Founding Father who famously said that if he were forced to choose between “a government without newspapers or newspapers without a government, I should not hesitate a moment to prefer the latter.”
I used to run that quote in the masthead of the Henryetta Daily-Freelance, back in the day.
“It was Democrats who more often employed situational ethics giving a pass to misinformation that helped their side,” Cannon, no partisan, added. “Most Republicans didn’t differentiate based on which way the false headline cut.”
Ballot tests at this stage of an election aren’t very valuable; job approval numbers are, along with perceptions of candidates and the issues swirling around them. National polls at this stage are also much less valuable than early state caucus and primary polls, especially in Iowa (for the GOP), New Hampshire, South Carolina, and Nevada. Caucus states like Iowas are especially hard to poll, given the relatively smaller universe of caucus participants who venture out on a cold, dark, and icy Iowa winter night for a long and tedious caucus).
Bottom lines: Don’t place too much value in public polling data now, and don’t assume or predict outcomes based on them. The best ones are those commissioned by the candidates that you never see. Public polling is far less valuable. Circumstances can and will change, perhaps dramatically. And I’m persuaded that some are used for nefarious purposes, with samples manipulated to inflate or deflate numbers. It helps to learn how to spot them.
I look at things like field dates (I discount polls conducted on Friday and Saturday nights or longer than four nights); the composition of the sample, especially partisan identification and other demographics, such as age cohorts and education levels; the questionnaire design; and how voters were contacted (I prefer a traditional mix of landline and cell phones, but a broader mix that includes online polling can arguably reach some Republican-leaving voters who won’t talk to pollsters over the phone). Refusal rates are high, often reaching 90 percent, so pollsters are getting creative on enticing participation (such as texting a link to a cell phone).
But they increasingly remind us that most voters aren’t buying the sympathetic and wildly misleading or false spin the White House and legacy media are feeding us. That makes me hopeful.
What was funny was watching a news organization work so desperately to trash their own survey.