Polling Realism - Part I
Well, FoxNews started broadcasting its own poll about the 2024 presidential race and it was not good news for Donald Trump.
Joe Biden leads Donald Trump 50% to 48%.
Here's a brief primer on how to assess a poll.
If it’s registered voters, it's likely garbage and should be heavily discounted.
If the question posed is anything other than “if the election were held today…”, it's also garbage and should be disregarded.
Why?
Well, in the first instance, a lot of registered voters end up not voting on Election Day. Although early voting has reduced the problem of registered voters failing to vote somewhat, there is still a sizable percentage of registered votes that do not actually vote. In some states, Texas comes to mind, low turnouts are the norm.
And in the second instance, data on future actions are rarely reliable.
Think about it. “What are you and your girlfriend having for Valentine's Day dinner?” yields a different response than the answer to “If you were going out for a romantic dinner with your girlfriend tonight, what would you order?”
When pollsters asked about voting intentions 5 months from now, a lot of unknowns get fitted into the answer. Immediacy denies one the ability to recalculate one’s choices
The FoxNews poll appears to have avoided those two pitfalls.
Next, you need to plug in two sizable problems that feed the margin of error: over reliance on landlines and undercounting voters under 30.
Millennials and Post-Millennials constitute roughly 35% of the voting population. If you don't account for them, the results get skewed.
Ditto if your call list is short of cell numbers. I am a Boomer and have not had a landline in over 15 years.
Trust me…too many landlines and you are talking to too many people over 70 and overweighting rural areas.
The margin of error (MOE) in most polls is about 4-4.5 points. If it is higher, it’s again garbage. If it’s below 3.5 percentage points, it’s pure fiction.
And polls, good ones anyway, will give you the demographic breakdown of who was questioned by the polling firm.
In Pennsylvania for example, if your poll demographics have less than 10% African-Americans or more than 20% of the respondents over 65, the results will be heavily skewed towards GOP candidates. (The African-American population of PA is over 11% and the 65 plus population of PA is 17.6%.)
“Okay, Kevin, I think I have it. Is this poll any good?”
Well, it’s good for Joe Biden, but it’s particularly worrisome for Donald Trump.
There is an old rule of thumb I learned from a pollster in the 90s.
Take the poll results. Add half the MOE to the person ahead and subtract half of the MOE from the losing party. Then take away 3/4 of a point from the person in front. That’ll give you a likely true state of the race.
If applied to the FoxNews poll, it looks like a 51.25%-46% advantage for Biden.
“Wait, Kevin, isn’t that more or less the election result in 2020?”
Yes, but what has changed to produce a different result?
Donald Trump failed to break 47% of the popular vote in 2016 and 2020.
From April/May 2017 to the end of December 2020, President Donald Trump had a metadata approval/disapproval of 42%/52%. It really did not fluctuate much.
Inelastic metadata results like that means that 2024 GOP candidate Donald Trump is going to be hard pressed to suddenly be polling over 48% under any scenario.
From late 2023 until the State of the Union, the news has been “all Donald Trump, all the time” and President Joe Biden was subsumed into that media narrative.
Ask yourself this question, what has changed for people who voted 4 years ago — especially in reelection contests.
In 1980, a bad economy and the Iran Hostage crisis shifted rough 7 points from Jimmy Carter’s 1976 election results to former California Governor Ronald Reagan (R) and Congressman John Anderson (R-IL).1
In 1984, a better economy and no John Anderson in the mix led to a Reagan landslide.
In 1992, incumbent George H.W. Bush lost in a three way race owing to a bad economy and a populace extremely soft on both the Democrats and the GOP.2
In 1996, a much much better economy boosted President Bill Clinton to over 49% (a six plus point improvement over his 1992 results) and a huge rout of the GOP.
In 2004, President George W Bush got reelected on “9/11” alone . It was his political mantra, slogan, and battle cry. And politics in 2004 were as tight as they are now.
W won the presidency with 47.9% of the vote and got reelected with 50.7%. A less than 3% swing told its tale in the Electoral College results. Bush won majorities of the Electoral College twice; both times with fewer than 300 Electors.
The last time a president was elected and reelected with fewer than 300 Electoral College votes was with William McKinley in 1896 and 1900 and we had only 45 states. And McKinley won over 50% of the popular vote both times.
What kills incumbents is cataclysmic events (politically speaking). What brings victories almost always good economies.
Anderson made an independent run for the presidency after losing to Reagan in the GOP primary. He garnered over 6% of the vote nationally.
Reform Party candidate Ross Perot received over 18% of the vote.
Great explanations. While it is early, I hope that Biden’s edge improves in the next few weeks and holds. No. 45’s rhetoric is getting scarier by the day.
Anderson was a throwback to a kinder, gentler less self-aggrandizing Republican.