Ph.D in Biotechnology who has joined a eCommerce startup in Hong Kong
Polls struggle to match election results, is the error an inherent bias or are polls a 'black art'?
Led Astray On the evening of Nov. 8, 2016 voters were aghast with disbelief, as the Clinton Trump election results rolled in. Nate Silver's FiveThirtyEight's projections - who have correctly forecast the winner of nearly every state, since 2008 - 2012 handed a 72% probability for a Clinton victory. The New York Times model gave Clinton a 85% chance, whilst the Princeton Election Consortium placed Clinton to win at 99%. How did the polls struggle to match election results? As the beautifully crafted poll simulation pointed out reporting on polling results need to account for a 'margin of error' and explain any inherent bias that skews the data.
Margin Of Error Nate Cohn does a deep dive into statistical weights or bias of polls and how one man in Illinois is distorting National Polling Averages. The unnamed person is one of 3000 panelist of the U.S.C. Dornsife/LA Times Daybreak poll, who is determined to vote Trump. According to the article, in some polls, he's weighted as much as 30 times above the average respondent, or 300 times more than the least weighted respondent.
What Happened? Polls aggregate information about voter preferences, studies have shown that when the electorate is small and voting costs are negligible, an equilibrium exists - voters report their true political alliances. When the electorate is large or voting costs are significant, poll respondents possess incentives to influence voting behaviour of others by misreporting their true preferences.
Abstract: Do polls simply measure intended voter behavior or can they affect it and, thus, change election outcomes? Do candidate ballot positions or the results of previous elections affect voter behavior? We conduct several series of experimental, three-candidate elections and use the data to provide answers to these questions. In these elections, we pay subjects conditionally on election outcomes to create electorates with publicly known preferences. A majority (but less than two-thirds) of the voters are split in their preferences between two similar candidates, while a minority (but plurality) favor a third, dissimilar candidate. If all voters voted sincerely, the third candidate — a Condorcet loser — would win the elections. We find that pre-election polls significantly reduce the frequency with which the Condorcet loser wins. Further, the winning candidate is usually the majority candidate who is listed first on the poll and election ballots. The evidence also shows that a shared history enables majority voters to coordinate on one of their favored candidates in sequences of identical elections. With polls, majority-preferred candidates often alternate as election winners.
Pub.: 01 Jul '93, Pinned: 19 Apr '17
Abstract: Larger margins of victory impart mandates that pull government policy toward a winner's platform. Voters with centrist preferences then may find pre-election polls useful. Centrists wish to moderate mandates and may abstain rather than vote for the nearest candidate. If polls are known to elicit voting intentions, then races will tend to be closer than predicted, and turnout will be highest in races predicted to be tight. However, voters at the extremes will respond to polls with guile – indeed all voters will. As a result, centrists cannot rely on pre-election polls, and poll results have no effect on voting.
Pub.: 01 Jan '98, Pinned: 19 Apr '17
Abstract: We examine the ability of pre-election polls to aggregate information about voter preferences. We show that if the electorate is small and voting costs are negligible, then an equilibrium exists in which citizens report their true political preferences. If the electorate is large or voting costs are significant, however, then no such equilibrium exists because poll respondents possess incentives to influence the voting behavior of others by misreporting their true preferences. We find that when a truthful equilibrium does exist, a poll can raise expected welfare by discouraging turnout among members of the minority.
Pub.: 06 Jun '08, Pinned: 19 Apr '17
Abstract: Although there is a large literature on the predictive accuracy of pre-election polls, there is virtually no systematic research examining the role that a candidate’s gender plays in polling accuracy. This is a surprising omission given the precipitous growth of female candidates in recent years. Looking at Senate and Gubernatorial candidates from 1989 to 2008 (more than 200 elections in over 40 states), we analyze the accuracy of pre-election polls for almost the complete universe of female candidates and a matched sample of white male cases. We demonstrate that pre-election polls consistently underestimate support for female candidates when compared to white male candidates. Furthermore, our results indicate that this phenomenon—which we dub the Richards Effect, after Ann Richards of Texas—is more common in states which exhibit traits associated with culturally conservative views of gender issues.
Pub.: 16 Sep '10, Pinned: 19 Apr '17
Abstract: Analysis of data from the American Life Panel shows that in the presidential election of 2008 and in multiple statewide elections in 2010, citizens exhibited large differences in their expectations of election outcomes. Expectations were strongly positively associated with candidate preferences, persons tending to believe that their preferred candidate is more likely to win the election. Committed supporters of opposing candidates regularly differed by 20-30% in their assessments of the likelihood that each candidate would win. These findings contribute evidence on the false consensus effect, the empirical regularity that own preferences tend to be positively associated with perceptions of social preferences. We used unique measures of preferences and perceptions that enabled respondents to express uncertainty flexibly. We studied a setting that would a priori seem inhospitable to false consensus--one where persons have little private information on social preferences but substantial common knowledge provided by media reports of election polls.
Pub.: 23 Feb '12, Pinned: 19 Apr '17
Abstract: In this paper we assess opinion polls, prediction markets, expert opinion and statistical modelling over a large number of US elections in order to determine which perform better in terms of forecasting outcomes. In line with existing literature, we bias‐correct opinion polls. We consider accuracy, bias and precision over different time horizons before an election, and we conclude that prediction markets appear to provide the most precise forecasts and are similar in terms of bias to opinion polls. We find that our statistical model struggles to provide competitive forecasts, while expert opinion appears to be of value. Finally we note that the forecast horizon matters; whereas prediction market forecasts tend to improve the nearer an election is, opinion polls appear to perform worse, while expert opinion performs consistently throughout. We thus contribute to the growing literature comparing election forecasts of polls and prediction markets. Copyright © 2015 John Wiley & Sons, Ltd.
Pub.: 11 Nov '15, Pinned: 19 Apr '17
Abstract: The Issues and Leaders model predicts the national popular two-party vote in US presidential elections from people’s perceptions of the candidates’ issue-handling competence and leadership qualities. In previous elections from 1972 to 2012, the model’s Election Eve forecasts missed the actual vote shares by, on average, little more than one percentage point and thus reduced the error of the Gallup pre-election poll by 30%. This research note presents the model’s forecast prior to the 2016 election, when most polls show that voters view Republican candidate Donald Trump as the stronger leader but prefer the Democrat’s nominee Hillary Clinton when it comes to dealing with the issues. A month prior to Election Day, the model predicts that Clinton will win by four points, gaining 52.0% of the two-party vote.
Pub.: 04 Nov '16, Pinned: 19 Apr '17
Abstract: Accounting for undecided and uncertain voters is a challenging issue for predicting election results from public opinion polls. Undecided voters typify the uncertainty of swing voters in polls but are often ignored or allocated to each candidate in a simplistic manner. Historically this has been adequate because first, the undecided tend to settle on a candidate as the election day draws closer, and second, they are comparatively small enough to assume that the undecided voters do not affect the relative proportions of the decided voters. These assumptions are used by poll authors and meta-poll analysts, but in the presence of high numbers of undecided voters these static rules may bias election predictions. In this paper, we examine the effect of undecided voters in the 2016 US presidential election. This election was unique in that a) there was a relatively high number of undecided voters and b) the major party candidates had high unfavorability ratings. We draw on psychological theories of decision making such as decision field theory and prospect theory to explain the link between candidate unfavorability and voter indecisiveness, and to describe how these factors likely contributed to a systematic bias in polling. We then show that the allocation of undecided voters in the 2016 election biased polls and meta-polls in a manner consistent with these theories. These findings imply that, given the increasing number of undecided voters in recent elections, it will be important to take into account the underlying psychology of voting when making predictions about elections.
Pub.: 28 Mar '17, Pinned: 19 Apr '17
Abstract: Recent studies find that defection from one's most preferred party to some other party is as common under proportional representation (PR) as it is in plurality systems. It is less elaborated how election-specific contextual factors affect strategic vote choice under PR. This study looks at the impact of two potentially important contextual factors: parties’ coalition signals about cooperation with other parties (referred to as ‘pre-electoral coalitions’) and polling information, which vary from one election to the next. The focus is strategic voting for smaller parties at risk of falling below an electoral threshold. The hypothesis is that parties that are included in well-defined coalitions will benefit from strategic ‘insurance’ votes if the polls show that they have support slightly below the threshold. However, smaller parties that do not belong to a coalition would be less likely to benefit from insurance votes. Extensive survey experiments with randomized coalition signals and polls give support to the idea that a voter's tendency to cast an insurance vote depends on whether the polls show support below or above the threshold and whether the party is included in a coalition or not.
Pub.: 28 Nov '16, Pinned: 19 Apr '17
Abstract: Political candidates often believe they must focus their campaign efforts on a small number of swing voters open for ideological change. Based on the wisdom of opinion polls, this might seem like a good idea. But do most voters really hold their political attitudes so firmly that they are unreceptive to persuasion? We tested this premise during the most recent general election in Sweden, in which a left- and a right-wing coalition were locked in a close race. We asked our participants to state their voter intention, and presented them with a political survey of wedge issues between the two coalitions. Using a sleight-of-hand we then altered their replies to place them in the opposite political camp, and invited them to reason about their attitudes on the manipulated issues. Finally, we summarized their survey score, and asked for their voter intention again. The results showed that no more than 22% of the manipulated replies were detected, and that a full 92% of the participants accepted and endorsed our altered political survey score. Furthermore, the final voter intention question indicated that as many as 48% (±9.2%) were willing to consider a left-right coalition shift. This can be contrasted with the established polls tracking the Swedish election, which registered maximally 10% voters open for a swing. Our results indicate that political attitudes and partisan divisions can be far more flexible than what is assumed by the polls, and that people can reason about the factual issues of the campaign with considerable openness to change.
Pub.: 18 Apr '13, Pinned: 25 Apr '17