Minutes of Y's Men Meeting of January 5, 2017
Mr. Schwartz stated that the polls regarding the last election were not as bad as popularly assumed. He pointed out that polls had the popular vote going to Hillary by four points when she actually won by two points. In a pollster’s view the polls had it basically right.
Looking at the swing states’ polls, again the polls had it within the margin of error, but indicated that the margin between the two candidates was closing fast. Pennsylvania was an exception. There the Quinnipiac poll had Hillary ahead by five points. It ended up with Trump winning by only 40,000 votes, which is a virtual tie. The outcome was determined by last minute deciders going disproportionately for Trump.
Polls are just a snapshot of a moment in time. They can also point out motivations that are affecting opinion. But there is a great risk in aggregating poll results and drawing conclusions from those aggregates. Yet many people do this and the media acting as forecasters are especially prone to this, which can result in predictions like we saw that Hillary had an 80% chance of winning.
Other factors may also have played a role such as the discouragement of some voters from casting their ballets because they felt the outcome was certain.
Mr. Schwartz reiterated that polls do not provide pinpoint accuracy. There will always be sampling errors and thus the margins of error in the results.
Q. Is a 3% error rate consistent with the most economical polling practices and what checks are there with regard to sampling?
A. 3% is the most practical rate considering the expense and manpower it would take to reduce that number. Checks are provided in the form of the practices used in obtaining the samples. Samples are also checked against census results and prior exit polls as an additional safeguard.
Q. What have you learned with regard to future polling?
A. Mid-term elections are a totally different animal, but regarding 2020, I would like to see shorter field periods with more polls in fewer states.
Q. How do some polls sample a disproportionate number of people from one party?
A. Lots of people are not set in stone as to which party they will give their allegiance. Also there are polls of widely varying quality.
Q. What about people that don’t take polls and what about polls that show wide disparities in factualness?
A. We work very hard to get to our selected samples, but some people just can’t be reached or don’t want to participate. On factualness, people didn’t like either candidate and many didn’t believe either one.
Q. Who are your clients, and why do they use you?
A. We don’t have clients. But both the media and the populations are very interested in the polls. And independent polls are like a watchdog on the information put out by the candidates, which will always have spin.
Q. Can you evaluate a politician before their first 100 days in office?
A. In my opinion you can.
Q. Don’t polls create an atmosphere for self-fulfilling prophecies?
A. Polls can have an effect. Leading candidates tend to be able to raise more money for example. But if there were no polls reporters would make up information based on a few conversations because they need stories to tell an interested public. So polls have consequences, but they are better than the alternative.
Q. How do you get cell-phone numbers?
A. The same way you get land line numbers. There are certain prefixes for cell phones and then a computer generates calls to that prefix. Cell phones are critical because many people, especially the young, only have cell-phones. The problem is people in one state that have cell numbers from another state. You can’t reach them.
Q. Was there any group that surprised you?
A. Whites voted more heavily for Trump than expected. Also Trump picked up an unexpected number of Hispanic votes.
Q. What percentage of people refuses to participate?
A. This is a growing problem especially with regard to people you can’t reach. If you reach them a good number cooperate, but the numbers vary from state to state.
Q. What is the impact of money on opinion?
A. For presidential elections, it has very little effect in my opinion because people know the candidates and know that ads will just spin facts. It can have a greater effect in down ballet races where people don’t know the candidates as well.
Q. What kind of staff do you have?
A. We have between 200-300 interviewers and 200 call centers. They get paid $10.50 an hour. We have students and non-students. Many of the students are political science majors or sociology majors.
Q. Any revealing statistics on turn-out?
A. We got it mostly right. Some states had a lower black turn-out than anticipated.
Q. Using hindsight what would you advise Hillary now?
A. Spend more time in the Midwest