Public Opinion Quiz #3
The Gamble: Choice and Chance in the 2012 Presidential Election
• A typical forecasting model (for presidential elections) relies on a few factors, maybe three or four, that predate most of the general election campaign - factors such as presidential approve ratings or the rate of economic growth early in the election year. • Ideally a forecasting model does three things: o 1.) it draws from plausible theories about how voters make decisions in elections o 2.) it is parsimonious (that is, it attempts to predict elections with a few big factors rather than a lot of small factors that explain the isocracies in one elections, o 3.) It is accurate. • Some critics are incredulous that a model would rely on so few factors, none of which may measure campaign activities such as advertising. • Others say that since the model is only drawing form 1948 there could only be around 16 elections in it. • Others point out were models were wrong. • However, if constructed and interpreted correctly, they are crucial because they are based one defensible theory about how a crucial subset of voters makes up its mind by evaluating the "performance: of the incumbent and his party. o This is why forecasting models typically include facets like president approval, and economic growth. • The models also provide a baseline for how the incumbent is doing • The forecasting models, taken together, typically correctly predict the winner, even if they do not predict the margin of victory well • They model use a "forest not the trees" approach meaning that it cares about the hype sour rind the candidate not the candidate themselves. This method will most likely correctly predict the winner, but maybe the wrong margin of error
THE EVOLUTION OF ELECTION POLLING IN THE UNITED STATES D. SUNSHINE HILLYGUS*
• In this essay, I outline the evolution of polling as used for three different functions in U.S. presidential elections: forecasting election outcomes, understanding voter behavior, and planning campaign strategy. • Public Opinion Quarterly was founded in January 1937 on the heels of the advent of modern scientific polling in U.S. presidential elections. The first issue included an essay, ''Straw Polls in 1936,'' explaining how George Gallups quota-controlled survey of a few thousand triumphed over the Literary Digests straw poll of millions in correctly predicting the election outcome • a 900-percent increase in trial heat polls between 1984 and 2000 • There has also been a significant evolution in the nature of election polling. o polls were typically conducted by telephone, using live interviewers, on behalf of media organizations or political candidates. • Today, Internet surveys and IVR polls are increasingly common, and polls are often initiated by entrepreneurial pollsters conducting them not for a client, but for self-promotion • Journalists are no ling the gate keeper to determine if a given poll is sufficient quality to interest the public, with blogs and random polling websites popping up; • It also seems that we have seen a rise and fall in the credibility of polling since • Today, however, nonprobability samples—typically opt-in Internet surveys—are increasingly common, and probability samples are experiencing significant methodological challenges, such as increasing nonresponse and cell-phone-only households. • In this essay, I will briefly outline the evolution of polling as used for three different functions in U.S. presidential elections: forecasting election outcomes, understanding voter behavior, and planning campaign strategy • Forecasting Elections • Before polls, knowledgeable observers, political insiders, and bellwether states were the most commonly used election forecasts • today, and each new election cycle brings a wave of horserace polling numbers feeding the insatiable appetite of media, bloggers, and political junkies trying to predict the election outcomes. • Unlike most survey research topics, pre-election polls have a truth benchmark— the election results.4 • The reputation of survey firms rests in no small part on these accuracy assessments. • What deems a poll to be accurate? The margin of victory? The correct winner? The share of the vote they received? That is what makes it hard to tell. • ''Mosteller Measure 3,'', the average absolute error on all major candidates between the prediction and the actual results, and ''Mosteller Measure 5,'' the absolute value of the difference between the margin separating the two leading candidates in the poll and the difference in their margins in the actual vote. • Like any survey, the quality of predictions can be affected by sampling error and nonsampling errors, including coverage error, nonresponse error, measurement error, processing error, and adjustment erro • Election forecasts can go astray simply because they must predict future behavior. I • n other words, it is an unknown population to whom pollsters are trying to generalize because we do not know who will show up on Election Day. • Every survey firm has its own (often proprietary) method for defining likely voters, typically relying on self-reported measures of voter registration or vote history, but rarely do those models engage the most up-to-date scholarly research on political participation • Thus, while it is widely recognized that undecided respondents contribute to polling error, there is still no consensus about what to do with them. • Respondents are another source of error in pre-election polls. An accurate election prediction relies on respondents providing honest answers to the turnout and vote intention questions • Polling predictions can also be jeopardized by individuals changing their minds about their turnout and vote intention between the time of the survey interview and Election Day • Polling Aggregation • In recognition that individual poll results are subject to random sampling error and any potential biases introduced by a firms particular polling methodology, it has become popular to aggregate across many different polls. • Aggregating polls helps reduce volatility in polling predictions. Although there is the tendency for news organizations to focus great attention on every movement up or down in the polls, as Jackman (2005) noted, ''mediacommissioned polls employ sample sizes that are too small to reliably detect the relatively small day-to-day or week-to-week movements in voter sentiment we would expect to occur over an election campaign • Some aggregations simply take the average of all available poll numbers. Yet, naı¨vely pooling across polls ignores house effects, • house effects, all of the methodological decisions made by a particular survey firm • Also reflecting improvements in polling availability and accessibility, many polling predictions have recently shifted from national-level to state-level analyses. • 2 Not every state is polled consistently, especially in less competitive states and earlier in the campaign, • It seems clear that the future of poll-based election forecasts are aggregations of state-level polls to make Electoral College predictions, but there remains much to be learned about the be • On the one hand, state-level forecasts offer a much higher bar for assessing the accuracy of individual pollsters since there are 51 predictions to be made, rather than one. Certainly, we see much greater variability in state-level polls, which do not converge toward the end of the campaign to the same extent as national polls • Beyond Polling: Macro Econ Models • Beginning in the 1970s, academics developed macroeconomic statistical models using non-polling aggregate data to predict election outcome • Most statistical models include measures of government and economic performance, though there is debate as to the specific economic indicator to be used—whether GDP growth (Abramowitz 2004), job growth (Lewis-Beck and Tien 2004), inflation rate (Lewis-Beck 2005), or perceptions of personal finances (Holbrook 2004)—and about the inclusion of other variables, like polling numbers and war support, in the statistical model. • Election forecasters argue that statistical predictions should outperform other election predictions because they are rooted in a theory about voter behavior. • Problems: o One criticism of these models is that, like the national-poll-based forecasts, they typically predict the two-party popular vote rather than the Electoral College outcome o Another criticism is that once we account for the confidence intervals around the point estimate, it becomes evident that most models predict a wide range of possible outcomes, including victory by the opposing candidate (Lewis-Beck 2005). There are only a handful of presidential elections for which the necessary aggregate data are available to estimate the statistical models, so predictions are inherently imprecise • Another election-forecasting alternative is prediction markets, such as the Iowa Electronic Market. With such betting markets, people buy and sell candidate futures based on who they think will win the election • he key to an accurate poll-based prediction is a representative sample of likely voters and truthful responses to the vote choice question. • Understanding Voter Behavior • Survey research is the primary tool for answering questions about electoral behavior, including political participation, voter decision-making, public opinion, and campaign effects. • 1.) Voter turnout over reporting, is itself a classic example of the dangers of measurement error. • A second methodological evolution in voting behavior research is the recognition that causal relationships are exceptionally difficult to establish using surveys, especially cross-sectional surveys. T • Candidates and Polling • ng at the role of political polling by candidates, parties, and interest groups. Whereas polling was once the primary way to gauge the preferences of the public, today campaigns are increasingly relying on consumer and political databases instead of polling alone • . In these targeted communications, candidates are taking positions on more issues and more divisive issues than in broadcast messages. In 2004, for instance, the presidential candidates took positions on more than 75 different policy issues in their direct mail, including wedge issues like abortion, gay marriage, and stem cell research that were not mentioned in television advertising
The Transmission of Political Values from Parent to Child Author(s): M. Kent Jennings and Richard G. Niemi Source: The American Political Science Review, Vol. 62, No. 1 (Mar., 1968), pp. 169-184
• In understanding the political development of the pre-adult one of the central questions hinges on the relative and differentiated contributions of various socializing agents. • he importance of other agents, but he was neither the first nor the last observer to stress the preeminent position of the family • A recent major report about political socialization during the elementary years seriously questions the family's overriding importance. In contrast to the previously-held views that the family was perhaps preeminent or at least co-equal to other socializing agents stands the conclusion by Robert Hess and Judith Torney that "the public school is the most important and effective instrument of political socialization in the United States,"1 • The first and primary objective of the present article will be to assay the flow of certain political values from parent to child. • Using 12th graders for exploring the parental transmission of political values carries some distinct characteristics. In the first place, most of these pre-adults are approaching the point at which they will leave the immediate family • A second feature is that the formal civic education efforts of society, as carried out in the elementary and secondary schools, are virtually completed. • A final con- sideration is that while the family and the educational system have come to some terminal point as socializing agents, the pre- adult has yet to be much affected by actual political practice. • It should be emphasized that we are not necessarily searching for pat- terns of political rebellion from parental values. • Confronted with a number of political values at hand we have struck for variety rather than any necessary hierarchy of impor- tance. We hypothesized a range of correlations dependent in part on the play of factors assumed to alter the parent-student associa- tions (noted above). • Previous research has established party identification as a value dimension of considerable importance in the study of political behavior as well as a political value readily transmitted from parents to children • rlier studies. The substantial agreement between parent and student party affiliations is indicated by a tau-b (also called tau-beta) correlation of .47, a statistic nearly unaffected by the use of three, five, or all seven categories of the party identification spectrum generated by the question sequence." • The observed similarity between parents and students suggests that transmission of party preferences from one generation to the next is carried out rather successfully in the American context. However, there are also indications that other factors (temporarily at least) have weakened the party affiliations of the younger generation. • A number of factors might account for the lesser partisanship of the students, and we have only begun to explore some of them • On the one hand, the students simply lack their parents' long experience in the active electorate, and as a consequence have failed as yet to develop a similar depth of feeling about the parties.17 On the other hand, there are no doubt specific forces pushing students toward Independence • One way in which political values are expressed is through opinions on specific issues. However, as Con- verse has shown, many opinions or idea ele- ments not only tend to be bounded by systems of low constraint but are also quite unstable over relatively short periods of time among mass publics.'8 Hence in comparing student re- sponses with parent responses the problem of measurement may be compounded by attitude instability among both samples. • would be the normal case. To what extent is the family crucial in shaping the evaluations of social groupings and thus-at a further remove-the interpretation of questions of public policy • Moreover, the aggregate differences which do occur are not immediately explicable. For ex- ample, students rate Southerners slightly lower than parents, as we expected, but the difference in ratings of Negroes is negligible, which was unanticipated. Students rate Whites and Protestants somewhat lower than parents. This is not matched, however, by higher evalu- ations of the minority groups-Jews, for example. • ese relationships. As with opinions on specific issues, intra- pair correlations on group evaluations are at best moderately positive, and they vary appre- ciably as a result of socio-political visibility and, to a small degree, group membership characteristics • Previous research with young children sug- gests that sweeping judgments, such as the essential goodness of human nature, are formed early in life, often before cognitive develop- ment and information acquisition make the 22 Robert E. Agger, Marshall N. Goldstein, and Stanley A. Pearl, "Political Cynicism: Measure- ment and Meaning," Journal of Politics, 23 (August, 1961), p. 490; and Edgar Litt, "Political Cynicism and Political Futility," Journal of Politics, 25 (May, 1963), 312-323. evaluated objects intelligible. • Students on the whole are less cynical than parents; relative to other students, though, those with distrustful, hostile parents should themselves be more suspicious of the govern- ment, while those with trusting parents should find less ground for cynicism • These findings do not mean that parents fail to express negative evaluations in family inter- action nor that children fail to adopt some of the less favorable attitudes of their parents. What is apparently not transmitted is a gen- eralized cynicism about politics • preference congruity. We found that when we skipped from party identification to other sorts of political values the parent-student correlations decreased per- ceptibly. • We have found that the transmission of political values from parent to child varies re- markably according to the nature of the value. • Turning to the power relationships between parent and child we hypothesized two types of relationships: 1) the more "democratic" and permissive these relationships were the greater congruency there would be; and 2) the more satisfied the child was with the power relation- ships the greater would be the congruency. Where patterning appears it tends to support the first hypothesi • Attitude obj ects in the concrete, salient, rein- forced terrain of party identification lend sup- port to the model. But this is a prime excep- tion. The data suggest that with respect to a range of other attitude objects the corre- spondences vary from, at most, moderate sup- port to virtually no support. We have sug- gested that life cycle effects, the role of other socializing agents, and attitude instabilities help account for the very noticeable departures from the model positing high transmission.
• Erikson, Robert S and Tedin, Kent L. American public opinion: Its origins, content and impact. Routledge, 2015 **p. 81-86**
• People of all levels of spophisitication hold party identifidations. • It is easy to like one party better than the other while holding limited political information or giving even modest thought to an issue. • Party idetnfication is quite stabl o Only a small percentage of people switch parties • The source of one party identidication come from political values that were shared during childhood from famil. • Party identification is the best predictor of how poeole vote • Party identification can be asked and discovered in many ways • Over the years, the public has increasingly seen party difference on issues. • Democrats shifted more liberal as a result of the parties becoming more polarized at the elite level. • Gold water in 1964 on the Civil Rights act shifte republicans more conservative • Over the years the public has become more polarize don party lines o This only gets more true the more informed someone is. • Partisianship drives policy opinions more than the reverse.
• Erikson, Robert S and Tedin, Kent L. American public opinion: Its origins, content and impact. Routledge, 2015 **Chap. 2, p. 45-51**
• The one good test of how accurate polls are coming on election day • Election polling comes with the usual challenges to pollster's accuracy, plus more • For elections, pollsters need to keep track of who votes, and who does not vote. • The media tends to attribute more accuracy to pre-election polls than the best-designed polls do. • The first reason why perfection rolls are wrong is that the sentiment expressed refers to the tampered that it was taken in. o Horse race analogy (the horse that is ahead in the stretch doesn't always win. • There are also still House effects -varriations in survey results due to idiosyncratic ways in which survey organizations conduct their polling. o Houses vary for perfection results in the way of their frequency of callbacks to elicit response from refuses, how they putrefy and Wight their responses, and how they allocate the undecideds. • Polls will also inevitable vary because of ordinary sampling error. • Typical survey shows that 15 percent or more of the electorate to be undecided with 5 -8% being the number days away from voting. o Survey houses differ in the way they handle the undecideds • The most common-sense way would be to just list them as undecided, but the media often pressures them to see which way they are voting. o Sometimes they use scale questions to figure out where to place the votes. If there is a tie in the scale questions than party identification is used, or some polls just exclude undecided. • It is hard to decipher the American election day population o No one has determined a method with great accuracy of who will vote • Historically, the poor and the less educated are the least likely to vote, and these people tend to support democratic candidates. • Pollsters seem to do a good job of screening in the days leading up to the election when a voter interest in the election is fairly easy to ascertain. o But screening becomes a challenge to pollsters in the early days of the campaign. ♣ The danger is that shifts in the polls of likely votes are often due to who counts the vote rather than a shift. • To screen those who are vowing/ non-voting questions are used to see if they are registered and then to determine the chance of the people coming out to vote. Two methods o 1.) Cutoff Method - They divide the registered respondents into those more and less likely to vote and then count the preference of the most likely o 2.) Probable electorate method - the pollster weights registered respondents by their likelihood of voting, for instance if someone was 30% likely to vote, then they would be weighted .30 as much as a certain likely voter. • Early polls often give misleading views of what is going to happen, but polls conducted late in the campaign such as the final week show great accuracy. • Historically, the late polls show no bias in favor of one party candidate or the other. • The polls underestimated the size of Obamas lead by about 3%, because of last minutes support and the underestimation of minority turnout. • The exit poll is a century old and is an important and controversial poll. • Representative precincts are selected, and voters are interviewed immediately after leaving hot emoting booth. • Exit polls are commissioned by the news media to help them to project the winner soon after the polls close. • Questions involve demographic and altitudinal information and vote choice. • Although the National Election Poll gives data to the TV Stations, each network hires their own person and analyzes it themselves. • Exit polls play a large role, but not the whole thing. Stations get early returns from sample precinct and compare voting patterns from geographic areas • The major purpose of exit polls is to provide background data on which types of voters prefer which candidates. • Several reasons why raw exit data are unreliable o 1.) The raw data tend to overstate the democratic vote and understate the republican vote because democrats are more willing to cooperate with exit pollsters. o 2.) The large number of polling places that need to be monitored require that many interviews must be hired on a temporary basis. Which leads to inexperienced and poorly trained people. They are mostly young so old people ignore them which contributes to the democratic skew. • After the election outcomes are known, exit poll data can be properly weighted to adjust for factors and certain demographical • There are analytical advantages o 1.) The respondents only include voters, o 2.) Their response reflects their election day views
Election Forecasting - Forecasting: The Long View Michael S. LewisBeck and Charles Tien
• This article offers new way to evaluate the pros and cons of us preselediential elections, the long view versus the short view o Election forecasters who take the long view stress the elector theory and lead time, examinigng model performance over several contests ♣ Overarching goal is knowledge of how the electoral process works o In contrast, forecasters who take the short view stress accuracy exlusivley. ♣ Forecasts are made repeadltey, ecespcially as the election gets closer • The short view forgets the lesson that most variation in national election can be predicted even explained by political behavior. • Casual readers of the polls during election season will get the impression that American voters are fickle • Election forecasting models based on polls, often updated until election day itself, also reflect this impression These "short view forecasts have received considerable attention in the media in recent years • These "longview" forecasts receive more attention from and have more credibility among academics. • The forecast for long view is White House Part Vote Sharet = f(Politcst-6, Economict-6, Cyclest-1) • Election forecasters who take the long view tend to stress electoral theory in their models and frame their predictions a good distance in time from the election. • . It is easier to forecast an election the day before it rather than 180 days in advance. But by the day before, the forecast itself has little intrinsic interest, since the real outcome will be known the next day. Accuracy comes from better theory and an optimal lead time before the election itself. o Strong theory emerges from parsimony—that is, a few carefully selected explanatory variables, which also help prevent the model looking good simply by chance. • A concrete example of the longview approach comes from our Jobs Model (LewisBeck and Tien 2012), with the dependent variable of White House party share of the twoparty vote, and the independent variables of presidential popularity, economic growth × incumbent interaction, jobs creation, and incumbency advantage, all measured in advance. • The view, or perspective, of the Jobs Model can be regarded as "long" in the sense of adequate lead time before the specific contest (a precise forecast can be issued at the end of August of the election year), but also "long" in the sense of examining model performance over several contests • To move this systematic enterprise forward, the ability to replicate each and every method is paramount • Election forecasters who take the short view tend to stress accuracy and to downplay theory and lead. • The accuracy measures here tend to rely on the most recent poll(s) and can be a poor predictor of the final outcome. Since accuracy ranks as the sine qua non, forecasts are made repeatedly, especially as the election itself draws closer. • Formula for Short -> White House Party Share = f(vote Intentions, t-x) • In the short view, an overarching goal is uptodate reporting of the horse race between the candidates. Accumulation of knowledge about how the electoral process works ceases to be relevant. • To summarize, the major differences between longview and shortview forecasting models are the following. o First are the inputs or predictor variables. In longview models the predictor variables are usually measured at the national level and selected by strong theory learned from voting behavior studies. Shortview models, in contrast, use vote intention survey questions from multiple surveys gathered at the national or state levels, and these are updated frequently. o Second are transparency and replication. Longview models are explicit about how the forecast is derived, and the data are made available to others for replication. Shortview models often do not reveal exactly how their inputs are measured, making replication difficult if not impossible. Complicating the matter is that survey houses often do not spell out how they determine "likely voters" in their polls, a filter the shortview forecasters rely on • One consequence of short views is that they use almost similar survey and forcasing stratagey. • Another consequence of the short view for the art of forecasting is increased reliance on polls. The need for a current estimate, particularly as the election nears, requires that the polls be utilized increasingly, at the expense of other approaches • Even supposing a sound combinatory strategy, including proper weighting, all comes to naught if insufficient polls are available. Such a condition—absence of enough polls—may actually occur. • In addition to negative consequences for the forecasting enterprise, adherence to the short view poses negative consequences for the understanding of voting and elections. First, its stress on continuous updating of the horse race conveys the impression to students of politics, and to citizens generally, that election outcomes are fickle. • they take the long view or the short view. The former tends to rely on substantive but parsimonious national models that have considerable lead time, with onlyoccasional updating. The latter relies mostly on vote intention, measured repeatedly in polls across the campaign, where no lead time gains privilege. The short view of forecasting has gained favor, but notwithout negative consequences. Because it is ahistorical, the picture of how election forecasters andforecasts compare becomes distorted, in the direction of seeing too much homogeneity. Thus,disagreements and difficulties are underestimated. Further, because it must be ever newsworthy, shortviewforecasting depends more and more on the polls until, in the end, nothing else seems to matter. But this holds danger, for polls can go wrong, and do. •
