정치 여론조사의 진화와 과제

### 여론조사의 진화와 도전 과제: 신뢰에서 의심으로
* 📉 2016년 힐러리 클린턴에게 유리한 예측이 나오면서 여론조사의 정확성에 대한 의문이 제기되면서 **여론조사의 신뢰도가 최저치**를 기록했습니다.
* 🗳 2020년 선거에서도 비슷한 부정확성이 나타났으며, 미국 역사상 가장 접전으로 기록될 **2024년 대선**을 앞두고 여론조사업체들은 주저하는 모습을 보였습니다.
* 📊 조지 갤럽의 선구적인 여론조사 방법은 **80여 년 전 과학적으로 선정된 5,000명의 미국인 표본**을 사용하여 여론을 측정함으로써 존경받는 산업을 구축했습니다.
* 📰 1936년 알프 랜던의 승리를 잘못 예측한 리터러리 다이제스트의 실패는 **정확한 표본 추출의 중요성**을 부각시켰고, **갤럽은 이를 성공적으로 해결**했습니다.
* 📞 1970년대 전화 여론조사의 등장으로 접근성이 개선되었지만, 90년대 들어 **응답률이 하락**하고 **온라인 설문조사가 부상**하면서 환경이 복잡해졌습니다.
* 🔍 네이트 실버는 2008년 대선 당시 데이터 분석을 통해 여론조사를 활성화했지만, 이후 선거에서 지속적인 문제가 드러났으며, 특히 **2016년에는 여론조사가 트럼프의 지지율을 과소평가한 것**으로 나타났습니다.
* 📉 2020년 여론조사는 **바이든의 승리를 정확하게 예측**했지만, **트럼프의 유권자층**을 고려하지 못하면서 여론조사 방법과 대중의 신뢰에 대한 재평가로 이어졌습니다.
* 🔄 여론조사 업계는 변화하는 커뮤니케이션 선호도와 유권자들의 회의론 등 현대의 도전에 적응하면서 **신뢰성과 정확성을 회복**해야 하는 중요한 과제에 직면해 있습니다.

최근 선거에서 여론조사 기관의 예측이 자주 빗나가면서 신뢰의 위기에 직면해 있습니다.

That says it all. Narrator: Trust in polls and pollsters hit a new low in 2016, when most showed that Hillary Clinton was more likely to win. How did the polls lead us astray? It just came out of nowhere. Narrator: And then again in 2020. The polling was wrong. It was wrong again. Narrator: Fast forward to 2024, and few are daring to make a call. I mean, this is the closest race in American history. It is really neck and neck. Narrator: In fact, there is only one perfect poll, and that's called an election. Believe it or not, the polls are not becoming less accurate over time. So how did a once-respected industry that grew to a value of $20 billion come to this? And can it win that trust back? TV presenter: Famed statistician Dr. George Gallup tells Washington from week to week.

what the nation is thinking. Narrator: It's been more than 80 years since a man named George Gallup conducted his first poll. He used a simple but effective technique to survey the opinions of the American public. TV presenter: 5,000 men and women mathematically chosen from every walk of American life to make up an accurate microcosm of the entire US population. Narrator: Before this time, surveys had been a lot less scientific. Pollster: How are you going to vote? I'm going to vote for Roosevelt, because I think he's going to bring back the bar. Narrator: The most reputable forecasting of the early 20th century was conducted by the Literary Digest. It asked readers to mail in sample ballots. In 1936, though, the Literary Digest backed Alf Landon to win the presidential election. But Roosevelt ended up winning by a landslide. TV presenter: The great American public has its say at the polls,

and the result is a Roosevelt victory of amazing proportion. Narrator: George Gallup had predicted that result by surveying a wider cross-section of Americans. TV presenter: Today, the true cross-section survey has become a trusted political barometer. Narrator: The Literary Digest went bankrupt, and Gallup's polls became a hit. The Gallup surveys of the 1930s ushered in a new era for polling and for voter confidence in it. Back then, Americans were not only willing but excited to talk to pollsters. It used to be such an honor to be included as one of the special households who got to speak for thousands of Americans. Narrator: In 1948, Time magazine even described Gallup as the Babe Ruth of the polling profession. That same year, Gallup backed the wrong candidate in the presidential race, but his reputation survived. We know that no better method has yet been found for measuring political trends.

Jay: Those were the days when polling was revered. And technology had to catch up with polling, which it finally did in the 1970s. Narrator: Once telephones were widespread in people's homes, pollsters didn't have to go door to door anymore. But there was a downside. Telephone response rates began to fall in the '90s with the invention of caller ID. Some companies eventually transitioned to online surveys that voters can respond to whenever they want. And by the early 2000s, there were more than 5,000 polling organizations in the US. That's when a baseball statistician named Nate Silver analyzed a number of polls and fed them into models that simulated the elections. He used them to come up with forecasts and give the public a glimpse of what Election Day in 2008 could look like. And may God bless the United States of America. Narrator: Silver was able to correctly predict the winner

in every state except for one. Basically, he made polls cool again. But in the 2016 elections, all of this changed. News anchor: Look at how close he is. Right now, he has 257 electoral votes. He needs 13 more. Narrator: It wasn't just a tough night for the news industry, but for pollsters. I had chills when it became clear, whatever it was, 9, 10 o'clock Eastern Time, that this was no layup for Hillary Clinton. Narrator: It was a huge shock. Dr. Wang, you tweeted recently that you were so sure of the result that you'd, put it up on the screen, eat a bug. A lot of people were wrong, but nobody else made the promise I did. Narrator: The polls basically got trashed. But that wasn't totally fair. Polls are not meant to be an exactly perfect prediction of what the eventual vote breakdown is going to be.

And it's normal for the polls to be three to four points off. Narrator: The polls did miss, but not by an unusual amount. They were off by about 4.8 percentage points. That's within half a point of the average error for election polls over the last 50 years. On Election Day, pollsters said that Clinton was leading by about 3.2 percentage points. Clinton did end up winning the popular vote, but by a smaller margin of 2.1 points. And as we all know, she lost the election. What happened was that Trump was able to win the electoral votes in a lot of states where Clinton was leading by a small margin. But fine margins can make a huge difference in the electoral college. Part of the problem was that pollsters were too focused on talking to those who seemed like likely voters based on their past voting history. The people who ended up voting for him are perhaps low-propensity, low-turnout voters.

who would not pass a likely voter screen or just not be reached by a poll at all but did ultimately turn out to vote for him. Narrator: They also didn't factor in education levels. College-educated voters were more likely to vote for Clinton, and they were also more likely to respond to surveys. So by not adjusting the sample, pollsters ended up with a forecast that overestimated support for Clinton. You can end up in a situation where you just miss a big part of the electorate and end up underestimating a campaign's support. Narrator: Ultimately, the results didn't match the scenario the media had hyped up the most. But the reputational damage seemed to be done, and the public started to lose faith. CNN sucks! CNN sucks! Narrator: Pollsters tried to implement a couple of changes in the 2018 midterm elections, like factoring in educated voters, and the forecasts turned out to be correct.

But when 2020 came around, lack of trust was still one of Jay's biggest challenges. Voters were reluctant to answer his questions. How do we draw a sample that actually is random? And we as a profession are struggling right now with how to do that because people do not want to be bothered. They have ways of communicating that are invisible to us. Narrator: Trump voters are less willing to participate in the polls, but not because they're shy. If you have been conditioned for four years by tweet after tweet that says, That poll that shows me, Donald Trump, trailing is a fake poll, you may be reluctant to talk to a pollster. And any poll that the fake news gives you because the real numbers are higher, but any poll they give you is a fake. Narrator: Pollsters did correctly predict that Joe Biden would win the 2020 election, but they again underestimated support for Trump.

One of the states that first started reporting its votes was Florida, where polls had Biden up by a few points but was still expected to be a very close race. As results started coming in, we saw that Trump was over-performing the polls again. Narrator: News organizations chose to represent their predictions in new ways. The New York Times published a table that included a column that read, If polls are as wrong as they were in 2016. And even 538 showed a number of maps to tease the idea of multiple scenarios. Overall, the polls did all right. Grace: Between response rates, coming up with a representative sample, and accounting for the pandemic, given that, it is pretty impressive that polls did predict the right winner. Narrator: They were off by four points. And that's normal by historical standards. More than 100 years after the first poll was conducted, the industry has had to reinvent itself all over again.

Would you mind telling us who you're going to vote for and why? I'm going to vote for Gov. Roosevelt because I think he'll bring back prosperity. I think Hoover deserves another chance. I'm sure he's the best man for that office.

How Election Polls Are Actually Getting Better | Insider Docs | Business Insider

What you want to know about business. 

Visit our homepage for the top stories of the day: https://www.businessinsider.com 
Business Insider on Facebook: https://www.facebook.com/businessinsider
Business Insider on Instagram: https://www.instagram.com/insiderbusiness
Business Insider on Twitter: https://www.twitter.com/businessinsider
Business Insider on Snapchat: https://www.snapchat.com/discover/Business_Insider/5319643143
Business Insider on TikTok: https://www.tiktok.com/@businessinsider

If you have a permissions or licensing request, please connect with our partners at PARS International who will be able to provide you with an accurate quote for the overall usage rights that you’re interested in. A representative from PARS will contact you shortly after filling out this request form: https://www.parsintl.com/publications/insider/


Business Insider

2016년과 2020년에 부정확한 선거 예측으로 여론조사에 대한 신뢰도가 최저치를 기록했습니다.

여론조사 산업의 가치는 200억 달러로 성장했지만, 지속적인 오류로 인해 평판이 나빠졌습니다.

1930년대 조지 갤럽의 과학적 접근 방식은 미국 인구를 정확하게 대표하고자 했지만, 리터러리 다이제스트의 결함이 있는 방법으로 인해 큰 예측 실패로 이어졌습니다.

1930년대 조지 갤럽의 설문조사는 리터러리 다이제스트의 파산 이후 여론조사 및 유권자 신뢰도에 혁명을 일으켰습니다.

1970년대에 전화가 널리 보급되면서 여론조사 방식이 바뀌었지만, 90년대에 응답률이 떨어지면서 온라인 설문조사로 전환하게 되었습니다.

2000년대 초반 네이트 실버의 여론조사 분석은 2008년 대통령 선거의 승자를 비롯한 정확한 선거 예측을 제공했습니다.

2016년 선거가 투표에 미친 영향

2016년 대선은 힐러리 클린턴의 승리에 대한 예측이 부정확한 것으로 드러나면서 여론조사 기관에 전환점이 된 선거였습니다.

8% 포인트 정도 빗나갔지만, 이는 선거 여론조사의 일반적인 오차 범위 내에 있는 수치였습니다.

트럼프가 저성향, 저투표율 유권자의 지지를 얻었기 때문에 과거 투표 이력을 기반으로 한 투표 가능성이 높은 유권자에 초점을 맞춘 것이 오차의 원인이 되었습니다.

여론조사 기관은 설문조사에 응답하지 않거나 교육 수준이 낮은 유권자 등 특정 유권자 인구 통계를 고려하지 않아 클린턴에 대한 지지도가 과대평가되었습니다.

언론의 과장된 시나리오는 선거 결과와 일치하지 않았고, 여론 조사에 대한 대중의 신뢰를 잃게 만들었습니다.

2018년 중간선거에 대한 변화에도 불구하고 여론조사 기관들은 여전히 무작위 표본 추출과 특히 트럼프 유권자들의 신뢰를 얻는 데 어려움을 겪었고, 이는 정확한 예측에 대한 지속적인 도전으로 이어졌습니다.

투표가 선거 결과에 미치는 영향

플로리다주는 가장 먼저 개표 결과를 보고한 주 중 하나로, 트럼프가 예상보다 높은 득표율을 기록했습니다.

뉴욕 타임즈와 538은 새로운 예측을 내놓는 등 언론사들도 예측을 수정했습니다.

4포인트 차이에도 불구하고 여론조사는 궁극적으로 트럼프의 당선을 예측했으며, 이는 역사적 기준으로 볼 때 정상적인 것으로 간주됩니다.