Wednesday, November 2, 2016

Quinnipiac a strong example of university-based polling

It’s election season, and with elections come, inevitably, polls. Elon University is only one of many universities conducting studies on people’s attitudes about the current U.S. political climate. Simply put, the Elon Poll is one drop in a massive bucket.
A particular university I’ve found interesting this month is Quinnipiac University in Connecticut. The university poll takes a random sample of adults 18 and older through random digit dialing (where random phone numbers are generated by a computer), and typically only reports on registered voters. The people calling are a mix of university students and non-students, not computers, and states included in the poll are Connecticut, Colorado, Florida, Iowa, New Jersey, New York, Ohio, Pennsylvania and Virginia.
Methodology is especially important because of how much it reveals about a poll’s strengths and weaknesses. People being polled are less likely to be honest about controversial issues if they’re speaking to a real person than if they’re doing a written survey. However, random digit dialing is a good way to get a truly random sample, as every number has an equal chance of being called. It’s also important to note that Quinnipiac is the one sponsoring this poll, meaning that there isn’t an outside company paying for this poll to happen – there’s no outside influence on the poll’s results.
When elections are imminent, political polls help people all over gauge what national attitudes are towards the presidential candidates.


by Megan Garnache

RAND Panel unique in polling

One way to track changes in political opinion is with a panel, where the same group of people is asked to respond periodically during a campaign season. Each election season, the RAND Corporation conducts a Presidential Election Panel Survey, or PEPS, using the RAND American Life Panel.
 According to the website, “RAND American Life Panel (ALP) is a nationally representative, probability-based panel of over 6000 members ages 18 and older who are regularly interviewed over the internet for research purposes.” Other polls use panels, but RAND differs in a few ways:
 RAND American Life Panel
·      The poll gives the technology needed to participate to the panel members. This way the panel is more similar to the U.S. adult population as a whole
·      This panel is continuous, so that responses from participants can be tracked over time. This gives RAND an interesting perspective into true changes in public opinion (this approach forecasted the 2012 U.S. presidential election)
·      Survey results are available to researchers online, which links new data to old data
 Advantages
1.     Researchers  can track participant answers over time, making it possible to measure change.
2.     Probabilistic Polling: rather than simply asking if the participants are voting, PEPS asks the participant how likely they are to vote overall, and then how likely they are to vote for each candidate
 RAND’s methodology is an accredited system, sponsored by: nonprofit organizations, US government agencies, international organizations, US state and local governments, colleges and universities and other professional associations.


By Lily Mandell

CBS News and Random Digit Dialing

CBS News Poll Methodologies
CBS News conducts surveys by telephone of anyone 18 years or older living in the continental United States. The sample of participants is random and is not chosen based on race, age or any other characteristic. In order to conduct the surveys, CBS gathers telephone numbers through a random search done entirely by the computer. This process is a probability sample called “random-digit dialing”, which gives every home an equal chance of being called by their interviewers, provided they have a landline. In general, CBS talks to 1,000 people out of the random sample they have determined.

Strengths
•  All people in target population have an equal chance of being interviewed
•  You can easily calculate the margin of sampling error
•  Most accurate form of survey research
Weaknesses
•  Non-response bias can be a problem
•  Response rate calculations can be imprecise
•  Not everyone has a landline. People who just have a cell phone can be systematically excluded.

Why is this methodology important?

Random-digit dialing is important in election opinion polling because it is has the best chance of capturing the real opinions of others in the population. The respondents are not chosen based on their political views or any other characteristics, which is important to understand when dealing with political polls. Finally, CBS pollsters weight their sample according to Census demographics to try to get an accurate, representative sample.
by Lauren Cuddy

Gravis Marketing Partners with One America News Network For National Polling

Nonpartisan research firm Gravis Marketing conducted a national poll on October 20 for One America News Network (OANN), a conservative cable news television channel.

Bias in the Numbers
Analysts should proceed with caution when using this data. FiveThirtyEight’s pollster ratings of historical accuracy, statistical bias, and methodology of research found that Gravis has a bias towards Republican candidates, in line with sponsor OANN’s political bias. In the poll, 40 percent of voters said they represented the Democratic party, 33 percent represented the Republican party, and 27 percent represented the Independent party.

Methodology Considerations
Gravis used automated telephone calls to reach about 2,100 registered voters. Pollsters try for randomly selected samples, but older Americans tend to respond better to automated calls. More older Americans use landline phones, which are easier to use for responding to surveys. Older voters were prevalent in the recent poll.
Other concerns include losing a younger audience that will not listen to an automated recording. In an effort to combat this potential skew in the results, Gravis weights the polling participants by anticipated voting demographics. Also, automated recordings get more declines overall, and this is hard to account for with weighting.
There are some benefits to automated polling, though.

The absence of an interviewer also has its perks. The questions asked will remain consistent across interviews. In addition, research shows that people will even respond more honestly when they do not have to share their answers with a person directly.
by Katrine Ryan