Tuesday, November 12, 2024

Polls increased in accuracy yet still underestimated Trump


Now that practically every single presidential vote has been counted, Donald Trump has swept every single swing state, a feat that appeared outright insane just a week and a half ago. Ann Selzer, a pollster known for her reliability particularly in Iowa, even predicted a three-point lead for Harris in due to outrage over abortion rights, a prediction that fell flat as Donald Trump and Republican candidates mounted an unprecedented, swallowing red wave after Democratic gains in elections and midterm elections since 2018. Most other major polling platforms set Harris and Trump in a dead tie, but suggested that Harris had the potential to eke out victories where it mattered: namely Pennsylvania, Michigan, and the Blue Wall that soon turned into Blue Wall or Bust… and then bust.


Similarly, a Times/Siena poll found Kamala Harris had a lead over Donald Trump and saw her as the candidate of change, which also turned out to be sharply inaccurate as a micro survey conducted by Representative Alexandria Ocasio-Cortez asked her audience why they chose to engage in “split ticket” voting (both voting for her and Donald Trump), and while this is not close to a reliable survey that speaks for all Americans, the overwhelming message was change. Change from the establishment, and casting aside the policymakers that voters felt had ingrained themselves in elite institutions, no matter how inaccurate, even as many cite President Biden as the most pro-union president in the last century.


Hindsight is 20/20, and it is difficult to offer political conjecture without posting self-righteously about “hindsight-obvious” mistakes. However, there is clearly something amiss with the polling. It is not that polling is biased towards Democrats, that would be absurd, but Pew Research Center's vice president of methods and innovation Courtney Kennedy suggests that Democrats participate in surveys more than Republicans.


Our studies and materials in class also demonstrated the occasional unreliability of public opinion and public polling. The type of people participating in surveys, combined with the types of questions asked and even the phrasing of an individual question affects results. Using charged language inherently alters the results to a visible margin. Polling is also simply inaccurate because the real poll is Election Day, and there’s not much to be done to change that without devolving into an overly reliant manner of horse race journalism, an unhelpful method of journalism.



Given the current events, it would not be a ridiculous assertion to suggest that the modern polling strategy as we know it is deeply flawed, underestimating Trump in both 2016 and 2024. The poll was supposedly “fixed” to accommodate for Trump’s win in 2016, then re-optimized for Democrat success in midterms and so on, yet despite a ton of money and time being put into polling, the results of being neck-and-neck are incorrect.

However, numbers are always deceptive and manipulated. Conservatives took to social platforms claiming that Trump had beat Harris by 20 million votes (implying that 20 million votes Biden had in 2020 were in fact, rigged and fraud) but the final vote counts tally Harris at 72 million votes while Trump is at 75 million votes, and California, our hallmark massive blue state, is only three-fourths of the way to a complete report.




But it would also be plain wrong to state that polling was inaccurate. Polling has drastically improved compared to previous years. For example, according to ABC, pollsters moved away from random digit dialing, or RDD, which tended to affect polls unpredictably and cause oscillation. The exception to this was the Iowa Selzer poll which continued to use RDD, and Trump ended up winning Iowa by 13 points, constituting a 16 point error. However, polls underestimate Trump less than they did in 2016 and 2020, although there is still, to a degree, a left-leaning bias—again, not because of any secret conspiracy, but simply due to the nature of the polling mechanisms involved.



So the election is close. Like also discussed in class, the Electoral College has the potential to exacerbate results and leaves the decision down to a few swing states. It is true that Trump swept every single swing state, but the polling margin of error was lower than in previous years, but this margin of error was all Trump needed to pull a few million votes ahead of Harris and win a decisive Electoral College victory. There is inherent uncertainty in polling, even as polling techniques and mechanisms continue to improve.


US pollsters taking heat – again – for failing to predict Trump triumph | US elections 2024 | The Guardian

These Maps are Lying (if you let them)

Polls predicted a close race. Trump won handily. Pollster says one thing is 'very clear.'

2024 polls were accurate but still underestimated Trump - ABC News

5 comments:

Noah Shefer said...

From a few articles I have read from the New York Times and CNN, it seems like most pollsters agree that their polls did not accurately reflect the psychological effect of thinking that the US has a bad economy and wanting change. This caused the polls to be skewed towards Kamala Harris, and is further reinforced by Donald Trump's reputation as a businessman, even though he has bankrupted several businesses, including a casino.

A Financial Times article polled Americans about how they feel about the economy, and compared it to how it is actually performing. Its findings were pretty surprising: since the start of the Biden administration, peoples view on the economy has been much more negative than it actually is. Many people cited rising inflation and costs as to why the economy is struggling, but that is not true. Inflation has been lowered to record lows (2.4%), and wages are rising faster than the prices of consumer goods. However, 90% of the people the Financial Times article interviewed thought that they were worse off before Biden took office (and the COVID pandemic).

It is obvious that this negative perception of the economy helped Trump with his campaign, and while supporters might not have been struggling economically themselves, there was a "vibe" that the general population was struggling. Trump used this to his advantage by claiming that the economy was the worst its ever been, and was able to swing the few percent of votes he needed to win the election.

Darren Lo said...

The graph we looked at in class exemplifies this: the perception vs reality of inflation/the economy is radically misaligned. Combined with the fact that the economy improving doesn't directly translate to the average middle class family being able to afford eggs or new shoes, Trump was able to tap into that deep fear that was overlooked by Democrats even in issue-related polling.

Bridgette Martin said...

I think polling has definitely gotten better, but a lot of polling is just noise. I think many people who know who they’re gonna vote for are less interested in polls, the swing voters are gonna be the ones that whether or not they show up and who they vote for is gonna affect the vote. I don’t think the polls are lying or that the results of the election are lying. I think the Harris Campaign got too comfortable in the polls, that people had like Bidens administration or her position in his administration. As we saw in a poll in class, people who tend to read less news tend to be republicans so many of those people likely didn’t actually read up on Harris’s campaign, rather reading that she was a communist or that she was running on whatever Biden did the past 4 years which was not favorable by the American public who feel like their quality of life has gone down as taxes continue to go up. People who participate in polls are likely politically active viewers of media online skewing the amount of people who actually go to the ballot box on Election Day and vote.

Alex Zhao said...

I really like what Bridgette said -- polls aren't truth but rather the closest thing we have to truth (in the midst of election campaigning season). Anything can happen on election day.
Also, looking at simply the numbers, Kamala got more (popular) votes in Wisconsin than did Joe Biden. She lost, however, as Trump simply gained more votes her (Trump is ~+80,000 from 2020, while Kamala is ~+38,000). Although these gains can be attributed to a growing population in Wisconsin, in my opinion, these changes occurred as simply more people voted. Thus the polls are inaccurate as when they scaled their respondents' answers, they scaled disproportionately to the true election turnout.

Furthermore, I would argue that Trump supporters have a greater incentive to turn out than Democrat supporters -- since Trump has been out of office for 4 years, there is a greater need among Trump supporters for Trump to be in office rather than 'suffer' for another 4 years under Democratic rule.
Similarly, individuals leaning towards the Democratic party have slightly less of an incentive to vote -- life isn't going too bad, and while yes 2024 is (was) touted to be the most crucial election ever (up to this point), if the polls are all swinging in your favor day after day, maybe voting and standing in line for a few hours seems a little harder to do.

Overall pollsters are just doing their best to try to predict the actual "truth." And truth is ever changing. That's why they're inaccurate (sometimes wildly inaccurate).

Sora Mizutani said...
This comment has been removed by the author.