How Often Do Betting Odds or Polls Accurately Predict the Presidential Election?

Scott Kacsmar
June 21, 2024

To the lengths that American politics have gone to be covered like sporting events, there is no surprise that betting on the presidential election is a popular thing to do now even if it must be done through an offshore sportsbook. We also have more focus on polling data than ever before thanks to websites like FiveThirtyEight.

With so many betting odds and so much polling data out there, are top-rated sportsbooks able to accurately predict these presidential elections?

The 2024 U.S. presidential election is less than 5 months away in November. There should be more than 150 million voters choosing between incumbent Joe Biden, former president Donald Trump, and an independent like Robert F. Kennedy Jr.

Let’s look at the pros and cons of the top prediction methods of betting markets and polling data, where the methods differ, and if one is more accurate than the other at predicting presidential elections in the U.S.

2024 U.S. Presidential Election Odds

Betting Markets: Surprisingly Strong at Predicting Elections

Betting markets are nothing new for U.S. presidential elections. In fact, it was very common in the late 1800s and early 1900s for Americans to bet on the election and for newspapers to publish the odds. As the years went by, people began betting on horse racing and sports, and the internet and offshore oddsmakers eventually limited betting on such markets.

However, in the absence of scientific polling, betting markets were surprisingly strong at predicting elections. According to Keith Romer of NPR’s Planet Money, out of 15 presidential elections between 1884 and 1940, the markets were only wrong once.

One valuable aspect of the betting market has always been to apply real-time information and updates on the fly to the prices. When James Comey of the FBI reopened an investigation into Hillary Clinton’s emails just before the 2016 election, that had a negative impact on her chances to win. When something like COVID-19 happened and the lockdown was in place, that changed the 2020 election between Donald Trump and Joe Biden.

Should something shocking happen to Trump in July when he is sentenced for his hush money case in Manhattan, something like a multi-year prison sentence, then you can bet the market will move in price swiftly and significantly to account for that.

Another positive, especially back in the day, is that people would put their money where their mouth is. If they place a cash wager on a candidate to win, then they have the opportunity to vote for them and do their part to make it happen in November.

Scientific Polls Find Their Place in Elections

While betting odds did a fine job of gauging public sentiment, there was still a need for something more scientific that could be studied, tracked, and repeated in future elections. At the time, the Literary Digest was a popular magazine and precursor to American polling.

Still, it was not very accurate. The Literary Digest predicted that Alfred Landon would defeat Franklin Roosevelt in the 1936 presidential election, which was wrong. This gave an opportunity to George Gallup to modernize and create scientific political polling of the masses in America in the 1930s. He started his company in 1935 to collect data that would gauge public interest in candidates and the most pressing political issues of the time, and he accurately predicted a Roosevelt victory in 1936.

Gallup’s Legacy

That was the beginning of the Gallup Poll and polling at a nationwide level. From 1936 to 2008, the Gallup Poll was accurate in every presidential election except for 1948 and 1976. In 1948, pretty much every method missed the upset of Harry S. Truman defeating Thomas Dewey. In fact, the Chicago Daily Tribune infamously published the “Dewey Defeats Truman” headline when that was not the case in an upset win.

As for 1976, the Gallup Poll had Gerald Ford edging out a close win over Democrat candidate Jimmy Carter, who won the election that year. Otherwise, Gallup was good for the better part of 70 years.

Accuracy Dilemma

Polls can be very useful to campaigns to judge where they are weak or strong in states. They can also tell where they may need to campaign more. Not to mention how voters may change their preferences over time as the election gets closer. They also can drive up voter interest to make sure these people polled show up to vote in November.

However, polls can run into many accuracy issues. They can be oversampling certain demographics, not interviewing enough minority groups, and outdated methods of contacting people. For instance, many people have moved away from a home phone number to only using a cellphone number, or maybe they don’t answer phone calls at all and prefer to do all of their communication via text and online sites.

There’s also the unpredictability of modern voters. People with no social media footprint and never taken part in any polling could always come out in large numbers to vote in November. Voters who are registered as independents can always be wild cards for pollsters as they could easily vote for either party or even a 3rd party.

The truth is people also lie to pollsters. Purposely misleading someone in an era where tribalism is on constant display is even more likely today than in past decades.

2012 Election: Betting vs. Polls

The 2012 presidential election was between Democratic incumbent Barack Obama and Republican Mitt Romney. The betting markets favored Obama.

In fact, British-based betting firm Spreadex offered an Electoral College spread of 296-300 for Obama to 239-243 for Romney. The result ended up being much more decisive, as Obama won 332 electoral votes to 206 for Romney, who lost all but 1 battleground state.

Bettors becoming voters can be a good thing, but where betting money can become a problem is market manipulation.

In the 2012 election, someone referred to as a “whale” in betting circles wagered anywhere from $4 million to $7 million on Romney to defeat Obama, which obviously did not happen. Obama won his reelection bid in one of the least dramatic elections of the 21st century. Some experts estimated that lone bettor accounted for a third of the money bet on Romney at that site.

In 2012, Gallup’s final survey had Romney at 49% and Obama at 48%, so it predicted a Romney win, which was false. Nate Silver found that Gallup was the least accurate of the 23 polling firms he analyzed in his work for FiveThirtyEight. Silver became popular in political circles after his model correctly predicted 49 of 50 states for the 2008 presidential election.

Obama actually trailed after the 1st debate, then the polls had it as a tie in October. RealClear Politics had the final numbers at 48.8% for Obama and 48.1% for Romney. It ended up being 51.01% for Obama and 47.15% for Romney.

We would give the edge to betting markets on this one.

2016 Election: Betting vs. Polls

We should start by saying that 2016 is one of the biggest upsets in U.S. election history, so most methods, experts, and pundits getting that one wrong is not a surprise. Donald Trump, who was a 200-to-1 underdog early in the process, essentially beat the favored Hillary Clinton by a margin of tens of thousands of votes from 3 battleground states in Michigan, Wisconsin, and Pennsylvania.

Most polling services had Trump losing by a fair margin to Clinton. The LA Times/USC were one of the only polls that had a Trump victory, which was a poll that was criticized for months leading up to the election.

Nate Silver’s model also gave Trump a 28.6% chance of winning in 2016, which was still better than most of the other pollsters that year when Trump did indeed pull off the upset win. Silver also defended the LA Times for publishing their polling data instead of placing too much confidence in the masses saying Trump was going to lose.

When Polls Crumbled

Polling results were closer for the popular vote breakdown, which Clinton won. The final polling numbers had Clinton at 46% and Trump at 42%. The actual numbers were 48.18% for Clinton and Trump at 46.09%. However, Trump’s electoral college victory was a surprise to most, including many bettors even though the markets were a bit more bullish on Trump’s chances in 2016.

Again, reacting to the news on the fly is a built-in advantage for betting odds over polls for presidential elections. With 2 weeks left leading up to the election, Clinton had -550 odds to +375 for Trump according to Odds Shark.

But after the FBI announced an investigation into Hillary’s emails, her odds fell to -300 while Trump improved to +200. That was about a 10-point swing in percentage points for Trump to win, going from 23.5% to 33%. But on election night, things quickly turned for the oddsmakers to favor Trump even if people were still heavily on Clinton to win. Just before 8:00 p.m. on election night, Clinton had -700 live odds to win. Barely 2 hours later, she was +145, and just before midnight, Clinton was a +1000 underdog as Trump was the heavy favorite at -2000.

The 2016 election was such an upset that some thought polling should be done away with, but Silver defended it as most of the state-level polls were still close to the actual outcomes. It was just Trump pulling off that sweep of Pennsylvania, Wisconsin, and Michigan by tiny margins that gave him the electoral win.

It will always go down as one of the biggest upsets in election history.

2020 Election: Betting vs. Polls

In the 2020 election, Joe Biden was a modest favorite over incumbent Donald Trump. This was always going to be a unique election as mail-in ballots took more prominence than ever before due to COVID-19.

In fact, the counting of votes allowed for some incredible value in live betting, as Trump had taken an early lead on Tuesday night before the mail-in ballots largely were counted overnight and for several days that week, allowing Biden to catch up and ultimately win. There were many hours where Biden was a +600 live underdog at sportsbooks as Republicans struggled to understand why the mail-in ballots, which were largely Democratic votes, boosted Biden so much.

Biden was ultimately able to turn Wisconsin, Pennsylvania, and Michigan blue again to secure his win in the popular vote and electoral college.

The final polling numbers had Biden at 51% and Trump at 44%. The actual result was very close, as it was Biden at 51.31% and Trump at 46.86%. Trump thought polls would underestimate him again but that was not the case.

RealClear Politics had a final average in betting odds of Biden at 63.8% and Trump at 35.4%, which meant about -177 odds for Biden. So, it was not as close as we were expecting 2024 to be between these candidates. But according to Forbes, over 70% of money bets were on Trump, the loser in 2020.

Which Is More Accurate at Predicting the Election?

While no method is 100% at predicting U.S. presidential elections, both betting markets and polls can be strong predictors and should be viewed together to fully gauge the prospects of an election.

Over time, betting markets arguably have the better track record as they performed great before polling became big in the 1930s, and they have done a solid job since. Both can run into accuracy issues with manipulative bettors and untruthful responders. Also, the markets have largely aligned with what the polling data has been saying for decades, so it’s harder today to see a large disagreement between the methods.

Still, betting odds do have an edge in being able to update for real-time events while polls, which take time to compile, can be outdated by the time they are released. Given their ages, if Biden or Trump were to have a stroke tomorrow, that is something you’d look for to have an instant impact on the betting odds while polling would be slow to capture that new information.

The added benefit of people putting their money where their mouth is with bets can also make it easier to trust the betting prices more than the polls. Remember, those people are likely going to vote in November.