Bad polling is a behavioral problem

One source of annoyance over the past year has been the media’s blind portrayal of national and local polls. They are presented without context, framing and, most importantly, an acknowledgment of past history – just the blind repetition of useless “data”.

By failing to mention the abysmal pollster results, the media presents a highly distorted view of the upcoming election results.

Indeed, polls a year before an election frequently focus on candidates who don’t make it to the polls. Remember, in 2007, the polls had Rudy Giuliani and Hillary Clinton tied (neither became their party’s nominee in 2008). November 2023 polls showed Biden vs Trump. We know how it turned out.

But there’s an even bigger problem with the poll, one that’s unlikely to be resolved anytime soon: Our own inability to predict our future behaviors.

A quick caveat: I’m not a survey expert, but I’ve spent decades studying market sentiment data. Long-time readers know that, except in extreme cases, they find very little usable information in Sentiment data. The reason is that sentiment measures suffer from similar problems to political polls. (Behavioral economics provides insight into both surveys and modern survey errors).

Sentiment has five key issues that make it problematic. It is:

1. Looking back
2. Emotionally charged
3. It works with a substantial lag
4. Requires accurate self-reporting
5. Extremely dependent on precise wording of questions

It’s just about basic market, economic and asset allocation questions. My experiences interacting with many investors over time suggest that people tend to have a fluid sense of their sentiment, too dependent on what just happened in the markets. Our ability to self-report our optimism or bearishness is flawed. It typically reflects recent changes in your portfolio, not our true future expectations. Sentiment fails to measure these issues accurately.

To these five basic sentiment issues, political polls have additional issues:

1. Landlines
2. The intentionality of the voter
3. Mobile phones and caller ID
4. Voting attendance
5. Voter participation

Let’s briefly consider each.

1. Landlines: In 2000, 95% of American households still had a landline. Today, it is only 27%.

Losing three-quarters of households is a huge drop, and it radically affects who pollsters can reach. (I ignore text and online surveys because they are even worse than phone surveys). It is fair to conclude that this makes creating a representative group of American voters very difficult.¹

2. Intentionality: I think most (many?) people answering surveys are answering honestly. The problem is that people often don’t know what they really believe. (Behavioral finance helps explain why this is so).

Everyone is focused on the undecided. Yes, these”Persuadable” matter. But I think they make up less than 7% of voters – maybe even less than 3%. What really matters to the results is who and how many people actually voted. Whether you’re a die-hard political partisan or an independent, you can say you’ll vote – but data shows a third of you fail to do so. This behavior is what sways presidential elections.

3. Mobile Phones (Caller ID): Who answers unknown cell phone calls? Unless you’re waiting for a call from someone whose number you don’t have – delivery, contractor, doctor, etc. – your phone (like mine) is probably set to “Mute unknown callers.” They go straight to voicemail – and if they don’t leave a message, it’s probably spam.

Who answers calls from strangers and spends 20 minutes answering questions? I guess they are not a representative group of American voters.

4. Voter turnout: The United States has one of the lowest percentages of eligible voters who actually participate in presidential elections (it’s even worse for state and local elections, as well as non-POTUS election years).

PBS, citing data from the United States Election Project, reported that “only 36 percent of registered voters cast ballots during the 2014 election cycle, the lowest turnout in a general election since 1942.”

In 2020, after a massive voter registration drive, the Census estimated that 168.3 million people were registered to vote. This represented two thirds (66.7%) of the total voting age population. Most modern developed democracies have much higher voter registration rates. The United Kingdom has 91.8% (2019 general election); Germany, Australia and Canada also have over 90% of eligible registered voters. Sweden and Japan automatically register citizens once they become eligible – they have a voter registration rate of nearly 100%.

A surprising number of Americans assume they’re registered — and many aren’t. The 80 million unregistered eligibles are a huge variable when it comes to surveys. No wonder that the margin of error is actually double which is usually estimated.

5. Voting attendance: The key challenge for the survey is that people have no idea what their behavior will be in the future. This is why polls are only “correct” a month and “about right” a week or so, but a year, six months, or even two are completely useless months before most elections.

Since 1980, turnout in presidential elections has ranged from 50% to 67% of the voting-age population. The 2020 presidential election saw the highest turnout in decades at 66.8%, but this still pales in comparison to most other Western democracies.

Who gets off the couch, goes to the local school or library and casts their vote? The answer is an unknown giant. What is known is that one-third to one-half of eligible voters do not. This is also why a 2-3% margin of error is ridiculously wrong – it’s much closer to 6-8% margin of error.

For any early survey to be correct, it must accomplish 5 difficult tasks:

1. Reach a representative audience
2. Have people self-identify accurately
3. Use unbiased survey questions
4. Get honest answers
5. Get accurate predictions about people’s future behaviors.

The first four all create errors – surveyors can take steps to (partially) compensate for those problems, but it’s still full of mistakes.

The latter is devastating to polling accuracy.

Behavioral finance has taught us that human beings have no idea what they will do in the future. Whether it’s a year or over 30 days, we don’t know with any degree of reliable accuracy. We sometimes Think we know how we’re going to feel that day, we want to believe we’ll do what we say we’ll do, but at least in the history of finance, we know that people are just terrible at predicting their future behaviors.

How will you feel one month from now, Tuesday, November 5, 2024? What is your physical state of being? Your emotional perspective? Your mental health? Are you excited, depressed or apathetic? Have you just started or ended a relationship? What will the weather be like that day (a surprisingly important aspect)?

***

Over the past year, I’ve had this conversation with various television and radio personalities, analysts and experts. They mostly admit that they recognize this to be true. It didn’t stop them from ignoring the political polling history of the last 10 election cycles. The emphasis on the horse race, the artificial creation of a competition, is what the media does best. It’s not so much that they have a partisan bias – all human beings do – but rather their commercial interest in anything that makes the contest more interesting, artificial or otherwise. LOG IN NOW TO GET THE LATEST INDER! It is sensationalism in writing. The claim that this is a close race seems designed to manipulate viewers into watching more polls, panels, speculation and opinions. Most of it is useless filler, the rest is just nonsense.

It is disappointing to see essential aspects of democracy being replaced by what appear to be lazy monetization schemes.

The polls misled people in 2016 (Trump won), misunderstood 2020 (Biden won by a much larger margin than expected), and messed up the 2022 midterms (Red Wave lol). Why people assume it will be any different this time is simply an ongoing default. Perhaps the American media focuses more on elections as sports-like competitions than on real issues, because sports is what the American media does best.

Attention-grabbing clickbait rather than policy analysis is not a great way for the media to cover “Democracy.” The repercussions have been having a negative impact for decades now…

__________

1. I’ve been wanting to cancel my landline for years, but I live in an area with poor cell reception — I get calls home from my cell via Wi-Fi. If the power goes out and the backup generator doesn’t start, we can’t even call our local utility to alert them that we’ve lost power.