Voting and Public Opinion Exam 1
What is snowball sampling?
Find the first person and the others becomes more assessable, starting point to the research
How should we study voting?
Journalistic approach and
When is question order less consequential? Mail survey or a telephone survey?
nonattitudes
People will answer questions about opinions whether they have them or not (People that just want the survey to carry on, if there's a question wording affect anything like that their the ones that are most likely to give random responses, or responses that are influenced by the wording of the question.)
What is simple random sampling?
Select where each number had equal chance of being selected
Screening
allows respondents to say I don't know or I haven't thought about it much (Standard ex: did you vote in the last election or were you out of town etc?)
Push Polls, aka PUGging Push poll
ask question to lead you to a particular answer
Branching question
ask question with decision tree off of it (ask people do you think of yourself as a democrat, republican, or independent, then if you say democrat it branches off to strong or not strong, etc)
journalistic approach
assume day to day events of campaigns influence voters.
Feeling thermometer
can tell how strongly people feel towards candidates on a scale
Middle answer
choosing 4 on a 1-7 scale, are they genuinely dead center middle, or they just choose the middle to look informed without being extreme
FRUGging
fundraising under the guise of research
Exit polls
interview with citizens just after they've voted
Attitudes vs. and opinions
more deep seeded, are more disposed to change
National survey organizations do Panel surveys-
same people are being surveyed
SUGging
selling under the guise of market research- pretend to do market \research but really trying to get you to buy a product. Ask preliminary questions about product but then sell something, "want to come answer questions about something and well give you a free dinner"
Index
take multiple items to measure a single concept (Measure different areas of same question using different measures)
Response stability/instability
whether people have genuine attitudes or firm, or susceptible to change
Dinkins affect
- mayor of new York city -refers to him doing worse in the election than in opinion polls, and people were hesitant to say they were oppose to him because he's African American. People chose screening questions and said they hadn't thought about race much because they didn't want to oppose him, or even went on to say they supported him when they didn't.
Do elections matter?
-1 vote doesn't matter, aggregation of votes matter -Politicians pay attention to us because we can vote, whether we do or don't -Policies that we get depend on who gets elected
Misinformation
-911 truthers- think bombs were placed in there, govt new, etc. -Obama not born here -Obama a muslim
One of the ways survey organizations try to get away with these issues is telephone survey
-Help us get around having to go see someone in person for the survey -Variety of ways in which we can draw our sample -One way is to get a phone directory -Poor people are more likely to have unlisted number because they cant afford to pay bill, they lose phone for a period of time, and they print the hard copy without their number. Well to do people can afford to pay fee to have unlisted number, but if you're poor you lose your phone you get it back, its unlisted and you might not even have the right phone number -Concern there would be a class bias with phone books -When George Gallup started doing surveys in the 30's phones were a luxury item, so the only ones with them were of higher classes, now a days not the case -Random digit dialers can do area code, the exchange, and the last 4 numbers for land lines, for cell phones its different, harder to get nationally representative sample using area codes because not accurate
What does the sampling error allow us to estimate?
-How off the sample can be with a certain level of confidence, when it says plus or minus a certain percent. -Tells us level of error from the range of values. Not a linear relationship, each improvement costs more and more -In order to reduce sampling error by 1 %, going from 4-3 costs more than going from 5-4 -+/- 3% is reasonable active **Cant ever be entirely 100% without error because we don't have an up to date list of the population, so even if census is used, its not up to date
Stratified sampling
-If were interested in small populations ex. House of reps, and were focusing on only senior level democrats -Over surveyed African Americans, to see if there is any variability between items, they stratified -We can look at surveys and lay it out
Methods of interviewing
-In person- highest response rate -Telephone -Mail -internet
Cluster Sample
-attempt to approximate simple random samplings -Instead of selecting people at random, survey organizations will divide the nation up into clusters, and then select the clusters at random -Apartment, city block, etc
-cheap -no interviewer effects -question exactly as you want it -has to be short -response rate low, don't plan on getting a return for a few weeks -don't know conditions for which they're sent out- diff conditions lead to diff results -don't know who filled out the survey
Internet surveys
-cheap -no interviewer effects -questions are exactly as you want them -biased sample -if general survey on web for people to find -or, if an email is sent to people
Commissioned polls
-customer service -FRUGging -SUGging -PUGging
In person
-expensive, even with clustered, still have to pay someone to go there hotels rental cars etc -interviewer effects -can be long -once you agreed, social desirability says less likely to say stop -less control over the interviewer as sponsor of survey
Interviewer Effects
-explanations -interpretations -interviewers are not suppose to offer explanations -one reason is because you want it to be reliable-everyone that's been asked the question to have the same experience -signs of approval/disapproval -general appearance -clothing -gender- middle aged women least threatening -race- answer is altered to please race of interviewer
Telephone interviews
-inexpensive -minimal interviewer effects -cant tell race over the phone -fairly short -answering machines
Citizens Views/Mixed
-polls are useful but pollsters are bad -people think politicians are weather vanes -losing candidates say the only poll that matters in on election day
Why should you be well informed on polls
-to avoid being manipulated -used in market research -compare yourself to others
What do we mean when a survey has been weighted
-used to compensate for bias in the sample to begin with, when its weighed you change how much on persons rate is worth -African Americans were triple counted, so each respondent was weighted one/third. change weight to make it representative of the population
Most samples contain roughly _____ people
1500
Compound question
2 questions embedded in one
Behavioral items
Attitudes describing their own behavior