WHAT IS POLLING?Based on the polls who did you think was going to win the Presidential election?Polling is -> Surveying a variety of Americans to gain a snapshot of what public opinion is on candidates, thoughts, and issues at a particular point in timeA poll only reflects opinion at the time it was donePolling uses a sample of the population as a microcosm of a larger populationHow? - > Through representative samples and weightingRepresentative Sample: A sample that accurately reflects the demographic characteristics of the target populationIt can be difficult to get a representative sample just through survey queriesResponse rates for telephone surveys – 10.20%Response rates have gone down a lot over time, even in last 10-15 yearsHow do Pollsters still get Representative Samples?WEIGHTING: Weighting is using the % of demographics in the population (often using the census) to fix the effects of skewed demographics in sample through corrective multiplicationEX of Weighting: To get the actual views of the county, Multiply each male respondent by 2/3 (2/3*75%=50%) and each female respondent by 2 (2*25%=50%)WHOSE OPINION COUNTS?Different groups want different kinds of populations’ opinionsThe desired target population mattersMany non-academic surveys looks at 3 different groupsALL ADULTS: SURVEYS OFDesirable for getting the opinions of the general American populationLess useful for predicting/explaining electionsAll adults include some people who cant vote (illegal immigrants, felons, etc)Also includes people who wont or probably wont vote (low knowledge, not registered)For weighting, its easy to compare samples of all adults to the censusSamples of all adults are less likely to have defined preferences on complex issuesREGISTERED VOTERS: SURVEYS OFInclude primarily the opinions of people who are politically consequentialMore useful for predicting/explaining elections, than all adults“Registered Voters” is still a defined target populationCan be compared to state voter polls and pollster listsCan be weighted according to themThese surveys still include many people who wont voteLIKELY VOTERS: SURVERYS OFGet the opinions of the most politically important people for electionsAre the BEST for predicting/explaining electionsBUT are NOT a defined populationPollsters have to apply a likely voter “screen” to determine who is likely to voteSurvey researchers disagree about best way to screenSurveys of likely voters are MUCH harder to weight, b/c there is no defined population of likely votersThe closer a poll is to election day, the more the population of likely voters looks like registered voters, b/c registered voters are mobilized by politicians and interest groups over the course of campaigns. The further away from Election Day a poll is, the LESS the population of likely voters looks like registered votersTo weight surveys of likely voters, Pollsters have to make assumptions about the demographics of voters on Election DayHow do Pollsters make these assumptions?THEY USE:The demographic turnout of last presidential electionCounts of registered votersMeasures of demographic group election enthusiasm based on survey questions (how enthusiastic are you this elections to vote compared to last election)Also, when they make assumptions about voter demographics to be weight survey samples of likely voters, pollsters also make assumptions about the PID breakdown of likely votersAs in, what % of voters on election day will identify as democrat & what % turnout advantage is this over the Republican votersEX: If 37% of Voters ID as dem, and 33% ID as rep, and 31% ID as Independent, this would be a 4 percentage-point advantage for DemocratsDemocrats have averaged roughly around 3 to 4 percentage point advantage in the last 5-6 presidential electionsPOLLING AND THE 2012 PRESIDENTIAL ELECTIONPollsters made WILDY different assumptions about PID and turnout and this led to EXTENSIVE variations in poll support or Romney and ObamaAssumptions I saw ranged from R+1% Advantage to D+16% Advantage (WHEN IT WAS ACTUALLY D+6% ADVANTAGE)This was very FRUSTRATINGPollster assumptions about turnout generally fell in 2 camps:Those who thought PID turnout would give Dems an advantage like 2008 (D+7) or betterThose who thought PID turnout would give dems considerably less advantage than 2008 (D+5 or less)Actual voter breakdown in 2012 by PID: D+6BOTH groups were off, but b/c 1st group was more right they were correct in predicting an Obama winREASONS why Pollsters who assumed more of a democratic advantage were rightObamas GOVT operation was better than expectedRomneys GOVT operation was much worse than expectedRepublicans didn’t have the enthusiasm advantage that early polling found, especially in key swing statesHOUSE EFFECTS:A tendency of a particular polling companies (OR HOUSES HENCE THE TERM) polls to skew a particular directionThese are largely determined by variations in survey methodology, including likely voter screensOFTEN, A HOUSE effect might be a slight skew toward the DEMS or GOP (REP), but tend to have more variation over time than other companies pollsThe existence of House Effects is the 1 reason, that it is IMPORTANT:To compare same pollsters polls over timeTo Aggregate poll resultsOtherwise, the knowledge that you gain from any single poll or company’s polls can be misleadingGOOD SOCIAL SCIENCE: You would always rather have more data than less, and data over time rather than just a single point.KINDS OF POLLSStraw polls/pseudo-polls: The survey sample is determined by whoever self-selects inEX: “TEXT/CALL in now to give your opinion” (POPULAR ON AMERICAN IDOL)Often presented as a scientific representation of a more general population, but NOTWhat these polls do is called SELECTING on your DEPENDENT VARIABLEIn the case of Straw polls/Pseudo-Polls, they reportedly represent the opinions of adults or people who watched a political debate or some other group, but in actuality they represent the people whoEXIT POLLS:Surveys conducted as people as leaving their polling place on election dayAN interviewer approaches people (every 10th person, for example) and asks them to participateResults based on the people who responds to an interviewers approachPROBLEMS WITH EXIT POLLSVoters may not be sampled randomly (approaching every 10th person isn’t always done correctly, interviewers may be less likely to approach certain people)Voters do not vote at random times (some kinds of people