The Bradley effect is a theory explaining a discrepancy between voter polls and election outcomes, where minority candidates perform worse on election day than pre-election polls suggested. It occurs because some voters tell pollsters they intend to vote for a non-white candidate to avoid appearing biased, but ultimately vote for the white opponent in the secrecy of the voting booth.

First identified in the early 1980s, this phenomenon has sparked decades of intense debate among political scientists, pollsters, and campaigns. Understanding whether the Bradley effect is a real, persistent psychological quirk or simply a convenient excuse for poor polling methodology is critical for anyone trying to accurately read political data today.

The Origin: The 1982 California Gubernatorial Race #

The term “Bradley effect” (sometimes called the “Wilder effect”) was coined after the 1982 California gubernatorial election. Tom Bradley, the highly popular, five-term African American Mayor of Los Angeles, ran as the Democratic nominee for governor against Republican George Deukmejian, who was white.

Going into election day, virtually every major poll showed Bradley with a comfortable lead. Some polls had him ahead by as many as 7 to 10 percentage points. Even exit polls conducted on election day led several media outlets to prematurely project Bradley as the winner.

However, when the official ballots were counted, Deukmejian won the election by a razor-thin margin of about 1.2 percentage points (just over 93,000 votes out of nearly 8 million cast).

The sudden and unexpected discrepancy shocked observers. Analysts and political strategists searched for an explanation. They theorized that a significant segment of white voters lied to pollsters. Fearing they would be perceived as racially biased if they admitted they were voting against a Black candidate, these voters claimed they were either undecided or planning to vote for Bradley. Once they entered the private voting booth, they cast their ballots for Deukmejian.

This theory gained further traction after several other high-profile races in the late 1980s and early 1990s:

• The 1989 Virginia Gubernatorial Election: Douglas Wilder, a Black Democrat, ran against Republican Marshall Coleman. Final pre-election polls showed Wilder leading by 9 to 11 percentage points. On election day, Wilder won, but by less than half a percentage point (0.5%), becoming the first elected Black governor in U.S. history.
• The 1989 New York City Mayoral Election: David Dinkins, a Black Democrat, held a massive lead in the polls against Republican Rudy Giuliani. Dinkins won the race, but by a much narrower margin than the double-digit lead polls had predicted.
• The 1990 North Carolina Senate Election: Harvey Gantt, a Black Democrat, ran against the white incumbent Republican Jesse Helms. Late polls showed Gantt ahead, but Helms won the election by a decisive 53% to 47% margin.

The Psychology of Social Desirability Bias #

At its core, the Bradley effect is a manifestation of social desirability bias. This is a well-documented psychological phenomenon where survey respondents answer questions in a way they believe will be viewed favorably by others.

In political polling, social desirability bias occurs when a respondent feels that admitting their true voting preference violates a social norm or could make them look prejudiced to the person interviewing them.

Several factors influence the strength of social desirability bias in polling:

1. The Interview Medium: Social desirability bias is strongest in face-to-face interviews or live-telephone calls. Talking to a real person—especially one who sounds professional and polite—increases the pressure on the respondent to conform to perceived social norms.
2. The Salience of the Issue: If race, gender, or controversial social stances are heavily emphasized in the campaign, voters are more likely to self-censor their true opinions.
3. The “Undecided” Cushion: Rather than outright lying and saying they support the minority candidate, biased voters often tell pollsters they are “undecided.” This allows them to avoid making a socially sensitive choice during the interview, even though they have already decided to vote for the white candidate.

Does the Bradley Effect Still Exist? #

For years, the Bradley effect was accepted as a standard polling hazard. However, its relevance in modern American politics has been heavily debated, particularly after the historic election of Barack Obama.

The 2008 and 2012 Presidential Elections #

The ultimate test for the Bradley effect came in 2008, when Senator Barack Obama became the first African American presidential nominee of a major party. Many analysts predicted that the Bradley effect would cause Obama to underperform his polls on election day.

Instead, the opposite happened. The final pre-election polling averages were remarkably accurate. In some states, Obama actually outperformed his final polling numbers, largely due to unprecedented turnout among young and minority voters. The same accuracy was observed in 2012 when Obama won re-election.

Most political scientists concluded that by 2008, the Bradley effect had largely vanished or had been mitigated by other factors, such as:

• Declining Racial Prejudice: Over time, social norms changed, and support for non-white candidates became highly normalized.
• Improved Polling Models: Pollsters developed better ways to identify “likely voters” and adjust for non-response bias.
• The “Reverse” Bradley Effect: Some researchers hypothesized a reverse effect, where some voters privately supported Obama but were hesitant to admit it to their conservative peers or family members.

The “Shy Trump” Voter: A Modern Variant #

While the original racially motivated Bradley effect may have faded, a modern cousin emerged during the 2016 and 2020 presidential elections: the “Shy Trump” voter hypothesis.

In 2016, Donald Trump consistently underperformed in pre-election state-level polls but went on to win the Electoral College. Many strategists theorized that some educated, suburban, or moderate voters supported Trump but refused to admit it to live pollsters because of the intense social controversy surrounding his candidacy.

To see how these dynamics play out in real time across different cycles, you can track the latest state and national data on this interactive polling dashboard.

While the “Shy Trump” theory is still debated, most polling analysts argue that the 2016 errors were actually caused by a failure to weight polls by education levels and a high number of late-deciding voters who broke heavily for Trump in the final days, rather than systematic lying by respondents.

How Modern Pollsters Combat Social Desirability Bias #

Because human psychology is constantly evolving, modern polling organizations have developed sophisticated methodologies to minimize social desirability bias and ensure their data reflects reality.

1. Shifting to Automated and Online Polling #

One of the most effective ways to eliminate social desirability bias is to remove the human interviewer entirely.

• Interactive Voice Response (IVR): These are automated “robocalls” where respondents press buttons on their keypad to answer.
• Online Panels: Respondents fill out surveys on their computer or smartphone in complete privacy.

Because there is no human on the other end of the line, voters feel much safer admitting controversial or non-mainstream preferences. To compare current trends and check how pollsters are adjusting their margins today, explore the comprehensive election data on Election Tracker, which compiles the latest presidential, senate, and house surveys.

2. Sophisticated Weighting and Modeling #

Instead of taking raw responses at face value, pollsters weight their sample to match the actual demographics of the voting population (such as age, race, gender, and education). If a pollster suspects certain groups are underreporting their support for a candidate, they can look at historical turnout data and past polling errors to adjust their “likely voter” algorithms.

3. Indirect Questioning #

Sometimes, pollsters will ask indirect questions to gauge a respondent’s true feelings. For example, instead of asking, “Do you support Candidate A?”, they might ask, “Who do you think your neighbors are voting for?” Often, respondents project their own private beliefs onto their community, giving pollsters a more accurate picture of the local political climate.

Frequently Asked Questions #

What is the difference between the Bradley Effect and the Shy Trump voter? #

While both are rooted in social desirability bias, the Bradley effect specifically refers to voters hiding their opposition to a minority candidate to avoid appearing racially biased. The “Shy Trump” voter theory refers to voters hiding their support for a populist or controversial candidate to avoid social judgment or political arguments.

Has the Bradley Effect been scientifically proven? #

The Bradley effect has always been a theory rather than a universally proven scientific law. While the polling discrepancies in 1982 (Tom Bradley) and 1989 (Douglas Wilder) are well-documented, many political scientists argue these errors were caused by traditional polling flaws—such as failing to count late-deciding voters or miscalculating voter turnout—rather than widespread, coordinated deception by voters.

How do online polls reduce social desirability bias? #

Online polls reduce this bias by offering complete anonymity. Without a live human interviewer on the phone, the pressure to conform to socially desirable answers is removed, allowing respondents to be more honest about their actual voting intentions. If you want to keep an eye on how these shifting methodologies affect upcoming political landscapes, you can easily monitor the newest approval ratings and trend charts to stay ahead of the curve.

Why do polling errors still happen if the Bradley Effect is dead? #

Polling errors still occur, but they are rarely due to the Bradley effect. Modern polling errors are usually caused by:

• Non-response bias: Certain groups of voters (like working-class men or young people) are harder to reach and less likely to answer surveys.
• Turnout modeling errors: Pollsters miscalculating exactly who will show up to vote on election day.
• Late-deciding voters: A significant block of voters making up their minds in the final 48 hours of a campaign.