Senior Christine Dorband is studying Math and Actual Science and has been given a Pence-Boyce research grant funded by Olivet alumni. Her grant proposal, “Analysis of the 2016,” was written and her research was conducted with the help of her faculty mentor, Dr. Dale Hathaway. Dorband began researching in the middle of May and is in the process of looking at the data collected and writing her research paper.
Why were you interested in analyzing election polls?
With the elections coming up, there is a lot of information that is coming out because of the fact that both the Republican and the Democratic party have to nominate someone for the general election and we haven’t had that in several years.
You’re studying to see how accurate the polls are?
Yes. I wish I was comparing democrat versus republican. I’ve gotten a lot of questions about whether or not I’m predicting who’s going to win for the overall general election in the coming months. But ultimately with the data that I’ve collected, I’m separating democrat and republican into two separate analyses. Basically, looking to see if the polls that I’ve gathered and the independent variables that I’ve come up with play a role into what the primaries [show].
What conclusions do you have so far? Can you tell us?
I’m still coming up with that. We all know that there were a lot more contenders and candidates for the republican party whereas the democratic party only had two or two main ones. You would of thought that that would have played a role and been a significant role in how the outputs turned out … Basically, what I found is that the democrat and the republican party tend to have different strengths and weaknesses for those independent variables. Where the democrat party has strong say like population variable, the republican party wouldn’t necessarily have that. So there’s different things that affect both sides where one might think that they would both be affected by the same scenario.
Did the fact that there were so many republican contenders change anything on the republican side when you were looking at it?
It does and doesn’t. The data that I’ve collected is all different. If I’m comparing it to the democractic party, the outputs are kind of different, but I made sure that all the polls had the same information across the board to be consistent. In that way if I’m looking at it individually, it is the same.
What were the variables that you studied?
It’s predicting primary percentage. The poll percentage should have a correlation very close to the primary. Because you want it to have a strong relationship to it. So you have your primary which is your dependent variable. Your poll and then poll number–so how many people are in the poll. You have days which is the amount of time from the end of the poll to the primary or caucus. You have population—the different types of states. And then caucus that one was a dummy variable … So there were 7 variables that I studied and I wanted to add more in the sense of demographics, but a lot of the polls weren’t consistent enough in providing that much information.
Can you talk a little bit about how the polls work?
With the news you never know what is necessarily skewed, in one way or another. I wanted to make sure the data was accurate and not biased as well … I first started with which independent variables I should use. So I was looking at several polls in the beginning and trying to figure out which variables were consistent throughout the polls. Then gathering the other information—going to government poll websites. Whereas other people were doing research in the lab, most of my time was spent on the computer trying to find accurate data. Polls are accessible and so are their outcomes. Just like the primary results. They only show an accuracy of 97%.
— Nathan DiCamillo, News Editor