Tuesday October 29. New survey! exclaim American political commentators. New investigation! Observers from other latitudes exclaim throughout the day. On this occasion the study comes from the tandem formed by The economist and YouGov and claims that the Democratic Party presidential candidate, Kamala Harrisis two percentage points ahead of Donald Trumphis rival and Republican Party candidate to occupy the White House, one week before the elections.
Articles and opinion columns are built around this survey – carried out over four days, which involved a thousand and a half citizens and which claims that it has 49% support against 47% that Trump would have and many messages on him. social networks. In other words: it saves the day for many information professionals.
But two days earlier, another poll, this one concocted in the Ipsos offices, and Reuterssaid Harris had 47% support and that while Trump had even less, 46%, the distance between the two was minimal. In addition: the previous week, two surveys, one carried out by a team made up of analysts from the New York Times and Siena College and another from CNN, offered a technical tie at 48% and 47%, respectively. And the day before, the last poll of the The Wall Street Journal So far, he has Trump in the lead: 47% to 45%.
By seeing everything with a little perspective, we can see an evolution. Namely: On October 22, Trump was in the lead, on October 23, Harris was tied and from there he took the advantage.. The problem is that one day before the election of The Wall Street Journalthe one that said Trump was more popular than Harris, two different polls – from Times English, SAY and YouGov the first, and Monmouth University the second – gave the opposite result: she was ahead of him… already several percentage points ahead.
The results of five of the most reliable surveys in the country.
Art EE
Conclusion: Each survey says a different thing. Sometimes this difference is not big and the trend is shared. Others don’t have much to do with it. And sometimes, two surveys on the same thing can even say the opposite.
The question arises itself: which one to watch?
The experts’ warning
As explained Olivier Röderone of the best data analysts in the world Financial TimesPolling experts warn against turning a poll, no matter how good, into a prediction. This means, they add, suffering abuse. “That’s when surveys start to lose their usefulness,” he explains. Patrick Murraydirector of the Monmouth University Polling Institute. A usefulness which should be limited to the social sciences seasoned with infinite nuances depending on where the demographic workers have placed their magnifying glass.
And the fact is that there are too many variables, the margin of error is wider than you think and the samples are generally insufficient when it comes to reflecting anything beyond – in the best case – of a certain popular feeling. In 2024, in this case, this state of affairs would be none other than an electorate divided practically in two and, consequently, very close elections.
The problem is, no matter how much it’s repeated, too many people want to ignore these experts’ warning. In short, wanting to take too much advantage of these percentages. Because? According to Roeder, because of the “need” to have something new to get excited or alarmed about and something new to project philias and phobias onto. Practices that are both closely related and widespread in Western societies. “It’s like a black hole that traps every ray of political light”. A dynamic that prevails over other analyzes and the cross-sectional examination depending on the circumstances.
“People want this new little pill,” he says. Courtney Kennedyvice president for methods and innovation at the Pew Research Center. “Let’s see the new score, let’s see if there is a landing; It’s always the same phenomenon.
The loss of credibility
There are two reasons why many people view surveys as predictions. The first is that in the United States, historically, the trends they highlight have been proven to be more true than false. And the second is that many pollsters can’t help but be the first to do it.
That’s how things were when eight years ago, in the weeks immediately before the 2016 election, many of them began juggling the data they were collecting and giving predictions to anyone came to ask for them. That of Huffington Post He was one of the most famous; the information portal founded by Arianna Huffington said that Hillary Clinton he had a 98% chance of beating Trump. Also the New York Timeswhose model gave Clinton an 85% chance, was represented. Even the portal FiveThirtyEighta poll aggregator intended for political analysis after an exhaustive examination of the polls, gave Trump a 29% chance. As everyone knows, Trump won this election.
Four years later, in 2020, many firms specializing in demoscopy arrived in the elections of that same year after having reviewed their methodology and perfected their way of investigating. They reduced telephone consultations while strengthening those carried out via onlinefor example, to try to reach more sectors of the population. They also included new variables, such as education level, in their maps.
What they got, thanks to this, was a glimpse of the headlines that later appeared in the press: Joe Biden He actually won the election. He did so, however, with a much more discreet difference than many had previously inferred. The pollsters therefore continued to refine their filters and, at the same time, midterms of 2022 – when Congress and part of the Senate were renewed – qualified the Republican wave that many commentators expected. They understood well: These local elections were not the disaster for the Democratic Party that so many people had anticipated.. The compass was declared repaired.
Who to trust?
In any case, and despite everything No that we should expect from a survey, there are still courses among those dedicated to preparing this type of study in order to shed light on the currents of the moment. There are people who work better than others, wow.
“What I would tell people is to be careful,” he once commented. Andrew Smithprofessor of political science at the University of New Hampshire and director of its polling center. A caution that creeps into the advice he gives to all those who really want to look at this type of data: observe the different methodologies implemented, choose the one that gives the most confidence then, in this category, look for the pollsters more transparent.
There are several rankings about this. Among them, the one created, based precisely on transparency and also on historical accuracy, generally stands out, the portal FiveThirtyEight. First appears the tandem formed by the New York Times and Siena College followed by the tandem formed by ABC News and the Washington Post. In third, fourth and fifth place would be Marquette University Law School, YouGov (which typically partners with various media outlets when it comes to publishing its polls) and the Monmouth University Polling Institute.
The key is found in seven specific places
However, even if national polls show a consistent advantage for one of the two candidates, everything seems to indicate that next week’s elections will be played in the seven states where we cannot guess (and therefore cannot predict ‘discard) who will win the vote. most votes. They are: Nevada, North Carolina, Arizona, Georgia, Michigan, Pennsylvania and Wisconsin.
This is why, in recent weeks, media like Policy They have relegated national polls to the background, giving priority to any resulting polls. They continue to limit themselves to reflecting popular sentiment, it’s true, but In these seven territories popular sentiment will be far more decisive than in any other part of the country..
It turns out that, for the moment, they echo what we already know: a technical equality which advances or falls by a few tenths depending on the day. In other words: the night from Tuesday to Wednesday will certainly be long.
