From polls to votes to seats: Forecasting the 2010 British general election

This paper outlines and discusses techniques for three stages in forecasting parliamentary seats from British opinion polls: adjusting and aggregating published vote-intention figures from across different polls; forecasting how public opinion might change before election day; and predicting the seat totals from the forecasted election-day vote shares. Specifically, we consider a state-space model for opinion polls which correct for house effects and other sources of survey error, the estimation of the historical relationship between polls and the election-day share of the vote, and a probabilistic approach to predicting the winner in each constituency.     

Forecasting the 2015 general election using aggregate local election data

Local government election results are used to estimate a national equivalent vote that provides the basis for a general election forecast. By-elections provide the means for calculating weighted quarterly averages of national support. These show trends similar to those obtained by national polls. By-election results in the three month period leading to the general election contribute towards the national vote share calculation and seat distributions are determined by uniform national swing. Additionally, results from the main council elections from 2011 onwards are aggregated to the parliamentary constituency level and used to estimate each party's relative performance in key target seats. This information is used to fine-tune the final seat forecast which suggests a hung parliament is the most likely outcome.     

Collective failure? Lessons from combining forecasts for the UK's referendum on EU membership

Research on election forecasting suggests there are benefits from combining different sources of information. This paper discusses the experience of a combined forecasting method that was developed for the UK’s EU referendum in 2016. The sources included opinion polls, vote expectation surveys, prediction and betting markets, expert and volunteer forecasts, and various forecasting models based on polling and other kinds of data. Averages of sources within each of these categories all, in our final forecast, suggested Remain was more likely to win but with varying degrees of certainty. Combining them produced a forecast that beat some but not others. Opinion polls and citizen forecasts came closest to the true outcome. Betting and prediction market participants and volunteer forecasters were the most overconfident that the UK would vote Remain. This may have been because they were distrustful of the polls following the 2015 general election miss and had too strong an expectation of a late swing towards the status quo similar to those in Scotland in 2014 and Quebec in 1995.     

Using state polls to forecast U.S. Presidential election outcomes

This paper uses pre-election polls to forecast U.S. Presidential election outcomes in the states and the Electoral College. The approach is notable in three ways. First, we employ state-level polls to predict voting outcomes in the states; second, we associate probabilities with alternative election outcomes, and third, we identify states most likely to be pivotal in the Electoral College. Using information available on the day before the election, we estimated that the probability of a Republican victory in the Electoral College in the 2004 election was 47.27%.     

Do local elections predict the outcome of the next general election? Forecasting British general elections from local election national vote share estimates

This article outlines a method for forecasting British general elections from national level vote shares at local elections. Although local elections are notionally ‘local’, the evidence suggests that they at least partly mirror national electoral fortunes. A simple general election vote share on local election vote share regression model that accounts for partisan differences and incumbency effects fits past data with reasonable accuracy. Based on the results of the 2013 and 2014 local elections, the model forecast a 56% probability of hung parliament, with a 78% probability of the Conservatives receiving the largest share of the vote.     

“Wisdom of crowds”? A decentralised election forecasting model that uses citizens’ local expectations

Many studies report the “wonders of aggregation” and that groups (often) yield better decisions than individuals. Can this “wisdom of crowds”-effect be used to forecast elections? Forecasting models in first-past-the-post systems need to translate vote shares into seat shares by some formula; however, the seat–vote ratio alters from election to election. To circumvent this problem, this paper proposes citizen forecasting, which aggregates citizens’ local expectations to directly forecast constituencies. Using data from the 2010 British Election Study, this paper finds (1) that groups are better forecasters than individuals, (2) that citizen forecasting correctly predicts a hung parliament, and (3) that marginality and group size are important predictors for “getting it right”.     

Piecing it all together and forecasting who governs: The 2015 British general election

This paper discusses a new probabilistic forecasting method that was designed for the 2015 British general election. It proceeds in a series of steps from opinion poll averaging, forecasting national-level vote shares and uncertainty estimates, and subsequent simulation of hypothetical election results, through modelling of constituency polls and survey data to identify and adjust for patterns in the constituency-level variation in party performance, and finally to probabilistic forecasting of seat outcomes and of different combinations of parties commanding relevant governing majorities in parliament.     

Prognosemodell auf dem Prüfstand: Die Bundestagswahl 2005

Our forecast model for German elections came within three-tenth of a percent in predicting the vote of the governing parties (SPD and Greens, 42.3% combined) in the 2005 Bundestag election. The predictors of the vote were (1) the popularity of the incumbent chancellor, (2) the long-term electoral support for the incumbent parties, and (3) a discount for the governing coalition’s tenure in office. Given the recent formation of a new party left of the SPD, the popularity of Chancellor Schröder was adjusted (assuming that supporters of this new party favored him as chancellor over the CDU challenger). Our forecast gave the governing parties enough of a vote to rule out victory for the center-right alternative (CDU/CSU and FDP) so long as the new left party did as well as polls suggested.     

Opinion Polls,Coalition Signals and Strategic Voting: Evidence from a Survey Experiment

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.     

The calculus of voting: An empirical test

Abstract. According to the rational choice model, the calculus of voting takes the form of the equation R = BP − C, where the net rewards for voting (R) are a function of the instrumental benefits from the preferred outcome compared to others (B) and the probability (P) of casting the decisive vote that secures these benefits, minus the costs of becoming informed and going to the polls (C). Here, we provide a systematic test of this model. The analysis relies on two surveys, conducted during the 1995 Quebec referendum and the 1996 British Columbia provincial election, in which very specific questions measured each element of the model. As well, this study incorporates two other factors that can affect the propensity to vote — Respondents' level of political interest and their sense of duty. We find that B, P, and C each matter, but only among those with a relatively weak sense of duty. The feeling that one has a moral obligation to vote is the most powerful motivation to go to the polls. We conclude that the rational choice model is useful, but only in explaining behaviour at the margins of this important norm.     

Polls,coalition signals and strategic voting: An experimental investigation of perceptions and effects

Polls and coalition signals can help strategic voters in multiparty systems with proportional representation and coalition governments to optimise their vote decision. Using a laboratory experiment embedded in two real election campaigns, this study focuses on voters' attention to and perception of polls and coalition signals. The manipulation of polls and coalition signals allows a causal test of their influence on strategic voting in a realistic environment. The findings suggest that active information acquisition to form fairly accurate perceptions of election outcomes can compensate for the advantage of high political sophistication. The theory of strategic voting is supported by the evidence, but only for a small number of voters. Most insincere vote decisions are explained by other factors. Thus, the common practice to consider all insincere voters as strategic is misleading.     

The Timeline of Elections: A Comparative Perspective

How do voter preferences come into focus over the electoral cycle in different countries? Do they evolve in patterned ways? Does the evolution vary across countries? This article addresses these issues. We consider differences in political institutions and how they might impact voter preferences over the course of the election cycle. We then outline an empirical analysis relating support for parties or candidates in pre‐election polls to their final vote. The analysis relies on over 26,000 vote intention polls in 45 countries since 1942, covering 312 discrete electoral cycles. Our results indicate that early polls contain substantial information about the final result but that they become increasingly informative over the election cycle. Although the degree to which this is true varies across countries in important and understandable ways given differences in political institutions, the pattern is strikingly general.     

Party over Policy? Scottish Nationalism and the Politics of Independence

The Scottish National Party (SNP) won control of Scotland's devolved government in the 2007 election yet opinion polls show no majority for its objective of independence in Europe. While the party is adept at exploiting short‐term political opportunity structures in the wider British context, as well as appealing to the ‘opinion electorate’, it appears less successful at persuading a majority of Scottish voters to agree with its core ideology. Helpful parallels can be drawn between 2007 and the last time the party polled over 30 per cent of the popular vote in Scotland at the 1974 (October) British General Election—then, as now, the Scottish voter appears to be willing to distinguish between party and policy.     

Vote Compass in British Columbia: insights from and about published polls

ABSTRACT

Published polls in the 2013 British Columbia provincial election conspicuously failed. This paper uses the Vote Compass Voter Advice Applications to shed light on the prediction failure. At the same time, it considers whether the dynamics of expressed vote intention in the Compass can be deployed as a rolling cross-section – in particular, whether the dynamics of participant self-selection are driven by temporally differentiated bias. Analysis will include selection effects within the Compass, as participants choose whether or not to proceed beyond the “voter advice” component to the political perception and intention batteries.     

The influence of local and national economic conditions on French legislative elections

This article's aim consists in building and estimating a model which explains and forecasts the outcomes of the French legislative elections by department. This model, which constitutes the first attempt for such a geographical level, emphasises the role of the economic and political factors in the explanation of the legislative vote. The model seems to be very accurate in forecasting the elections of the past at the local and national level. Furthermore, its behaviour for the 2002 election was very satisfactory. This model is therefore a reliable alternative to the vote intention polls as an electoral forecasting instrument.     

Forecasting the 2010 general election using aggregate local election data

The paper presents a revised method for estimating national vote shares using aggregate data from local government by-elections. The model was originally developed to forecast the annual outcome of local elections but was adapted in time to provide an accurate forecast of Labour’s landslide victory at the 1997 general election. However, over the past decade the changing pattern of party competition which has seen parties becoming more selective about which elections to contest has led to more elections being excluded from the modelling because they failed to meet the exacting criteria that all three major parties, Conservative, Labour and Liberal Democrats, had contested both the by-election and the previous main election, normally held in May. Relaxing these criteria, although increasing the number of available cases would adversely affect the forecast, over- or under-estimating party votes. Instead, the revised method overcomes the problem of differential competition by estimating vote shares for parties that contest one but not both elections. A further innovation is the calculation of a weighted moving quarterly average which takes account of the number of days elapsed between the by-election date and the date of forecast. Using the new method we provide estimates for likely party shares for the 2010 general election.     

Momentum in the polls raises electoral expectations

Voters rely on opinion polls to help them predict who is going to win elections. But they are regularly exposed to different polling results over time. How do changes in the polls affect their expectations? I show that when the polls indicate that a party’s support has increased, voters’ expectations for that party’s performance will be higher than they would be at the same vote share but without such evidence of growth, because the party appears to have momentum. Across six survey experiments in Britain (total N > 14,000), I find that this effect persists even when changes in vote share are well within the margin of error, when comparing a small change in vote share to consistently polling at the larger vote share, when the change makes little difference to a party’s objective probability of victory, and when voters have strong preferences that might colour their interpretation of the polls. In short, the appearance of momentum in the polls robustly raises voters’ expectations that a party will win an election. This finding has major implications for any area of research in political science where expectations feature, for theoretical understandings of how people perceive the future, and for salient policy debates about the regulation of opinion polls.     

Voting May Be Habit-Forming: Evidence from a Randomized Field Experiment

Habit is a frequently mentioned but understudied cause of political action. This article provides the first direct test of the hypothesis that casting a ballot in one election increases one's propensity to go to the polls in the future. A field experiment involving 25,200 registered voters was conducted prior to the November general election of 1998. Subjects were randomly assigned to treatment conditions in which they were urged to vote through direct mail or face-to-face canvassing. Compared to a control group that received no contact, the treatment groups were significantly more likely to vote in 1998. The treatment groups were also significantly more likely to vote in local elections held in November of 1999. After deriving a statistical estimator to isolate the effect of habit, we find that, ceteris paribus, voting in one election substantially increases the likelihood of voting in the future. Indeed, the influence of past voting exceeds the effects of age and education reported in previous studies.     

Yes, Prime Minister: The key to forecasting British elections

We use our “PM and Pendulum” Model to forecast the outcome of the 2010 General election. The vote function of the model, aside from a cyclical dynamic, relies on approval of the prime minister as the sole predictor. We find that PM Approval predicts the vote (and vote intention between elections) more accurately than does Government Approval. Turning to the forecasting of seats, we examine the accuracy of the autoregressive model of the vote-seat translation against the uniform-swing model, which is widely used by pollsters and the media. Testing the alternatives on election data since 1910, our autoregressive vote-seat translation model proves superior to the uniform-swing model.     

Forecasting the 2015 British election through party popularity functions

To forecast the May 7, 2015 British General Election, we develop party popularity models based on Continuous Monitoring Survey (CMS) data from April 2004 to February 2015. Our models predict party vote shares three months prior to the election, using previous support levels, national economic evaluations, macro-partisanship and political measures. Our Seemingly Unrelated Regression (SUR) methodology allows us to predict support for the Tories, Labour, Liberal Democrats and “other” parties, separately, yet simultaneously, by constraining total support for all parties to 100%. Our model, estimated with data from February 2015, predicts that Labour will win the highest vote share in Great Britain, but that no party will win a majority of seats in parliament.