Methods Of Forecasting American Election Outcomes

Author : Philippe Mongrain
Publisher :
ISBN 13 :
Total Pages : 0 pages
Book Rating : 4.:/5 (137 download)

DOWNLOAD NOW!

Download or read book Citizens as Election Forecasters written by Philippe Mongrain and published by . This book was released on 2022 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Forecasting is a daily activity. We are constantly making judgments about the future. This is not surprising since forward thinking is an essential condition for the survival and creativity of humankind. Prediction is also one of the primary functions of the scientific enterprise. The idea that there may be factors or conditions that improve the predictive skills of individuals is controversial. Indeed, there is some confusion between prediction as a scientific goal and the more folkloric manifestations of prediction akin to prophecies or divination. However, a growing body of work demonstrates that the quality and accuracy of predictive inferences can be explained at least in part by the information individuals possess. Over the last few decades, political science has seen a renewed interest in electoral forecasting. Citizen forecasting is one of the most successful methods for predicting election outcomes. This method consists, as the name suggests, in surveying citizens about their electoral expectations. These predictions are then aggregated or incorporated into models to estimate the most likely election outcome. The goal of this dissertation is to identify and estimate the effect of different variables on citizens' predictive ability as well as to determine the consequences of unfulfilled electoral expectations on satisfaction with democracy and perceptions of electoral integrity. The empirical demonstration relies on a variety of electoral studies and consists of four articles, each of which forms a chapter. Overall, more than 730,000 citizens' election forecasts from 11 different countries were collected for this dissertation. The first article examines the role of political sophistication, and particularly of individuals' factual political knowledge, on their ability to predict election outcomes. This paper also investigates whether an individual's level of political knowledge can exert a moderating effect on the relationship between partisan preferences and electoral expectations. To do so, data from the Making Electoral Democracy Work project covering regional/national and district elections in three countries (i.e., Canada, France, and Germany) are used. The results support the hypothesis that highly knowledgeable voters are better forecasters. However, high levels of political knowledge do not appear to be sufficient to reduce the influence of partisan bias on electoral expectations. The first paper also introduces a new measure based on the Brier score to assess the quality of probabilistic forecasts in the context of elections. The second paper proposes a weighting strategy favouring the predictions of individuals with a high predictive ability potential. According to the "miracle of aggregation" principle, in the absence of systematic biases, errors in individual judgments within a population should cancel each other out and lead to a correct decision at the aggregate level. This phenomenon reflects the idea of collective intelligence (or "wisdom of crowds"). Aggregation would thus have epistemic properties. Therefore, the second article seeks to take advantage of the varying levels of competence within a group in order to maximize the informational input of the most competent members without completely ignoring the contribution of less competent members. Two measures were chosen to assess potential competence, namely, an index of respondents' factual political knowledge and an index of past forecasting performance. This paper mobilizes item response theory and panel data to create these indices. The analysis of data from district-level elections in Canada, France, Germany, and Great Britain shows that weighted aggregation is a superior alternative to unweighted aggregation. The third article analyzes the relationship between social interactions and predictive accuracy. A limited body of work suggests that certain characteristics of social networks may generate enough relevant information to provide a fairly accurate picture of the electorate and its preferences. Specifically, the frequency of political discussion within a network, its level of heterogeneity, its size, and the degree of political expertise of its members should contribute to predictive accuracy. This article uses the meta-analytic approach to analyze the relationship between these four variables and the predictive accuracy of citizens. The analyses are based on a dataset including observations from nearly 70 election surveys conducted in 10 different countries. In general, the nature of an individual's interpersonal interactions does not appear to have much impact on his or her ability to predict election outcomes. The fourth and final article shifts the attention to the consequences of the electoral predictions that citizens make. Specifically, this chapter evaluates how unfulfilled electoral expectations affect political attitudes. The impact of unmet expectations on post-election attitudes remains relatively understudied. According to the expectancy-disconfirmation model, voters who incorrectly predicted the victory of their preferred party or candidate should show lower levels of satisfaction with the way democracy works than voters who correctly anticipated the defeat of their preferred party or candidate. These voters should also be less confident in the integrity of the electoral process. In contrast, "surprised" winners should be more satisfied and confident than winners who correctly anticipated the outcome. Data from more than 20 regional and national elections in four countries (i.e., Canada, France, the United Kingdom, and the United States) suggest that voters' expectations have little impact on their level of satisfaction with democracy. In a number of cases, however, unexpected losers were more likely to question the integrity of the electoral process than expected losers. This is especially true of the 2020 U.S. presidential election, which witnessed an unusual number of fraud claims and malpractice accusations. In highly polarized contexts, unmet expectations can potentially worsen the consequences of defeat. In the conclusion, the main results of each of the four papers are reviewed. Implications for research in the field of election forecasting are also discussed. The contributions of this dissertation are of both theoretical and practical interest. From a theoretical standpoint, the results reinforce the notion of "forecasting ability" by establishing a positive association between factual political knowledge and forecasting performance across different levels of elections and electoral systems. The results also cast doubt on explanations in terms of social interactions or, at the very least, invite a rethinking of how we measure the characteristics of social networks. This dissertation is also one of the few works to consider the influence of citizens' electoral expectations on their attitudes toward democracy and the electoral process following an unexpected defeat or victory. Finally, from a practical standpoint, the results suggest that weighting individual judgments by competence is a viable strategy for improving the measurement of public opinion with regards to projected election outcomes.