Stephen Peplow. Image from stephenpeplow.com
“I did it for myself, as a modelling exercise. The university published their own set of results, but it seems a little elementary with graphs and numbers, and I just wanted to dig a little deeper.”
Stephen Peplow is a data analyst and professor at Kwantlen Polytechnic University. He teaches statistics and economics at various levels. He recently published “Predicting the Uptake of a University’s Offers of Places,” which notes that universities mail out many offers of enrollment to potential students who never reply. The paper draws from a dataset of just under 13,000 students, which Peplow notes is “not a lot.”
“I wasn’t working for the university, but I was using the university’s dataset. I published it because I thought people might be interested, more in the method and the technique than the actual results, but the results are quite good,” he says. “It’s an assessment tool, to try and save time, but that’s not why I did it.
“I could calculate the chance of someone taking up an offer [from KPU] with about 80 per cent accuracy.”
Some of the findings were interesting, such as the low response from 18-year-old students, and the high responses from mature students. Younger students, just coming out of high school, likely send out university applications to many universities, then pick from the ones they get offers from. Mature students might be coming to KPU for specific courses or trades.
Peplow notes that with his model, one might be able to link it with a spreadsheet, and attach a response probability next to students’ names.
“This is just people who’ve actually been offered a place at Kwantlen. They’ve been through the application procedure and received the offer letter. It would be interesting, but I don’t have the data for people who come to Kwantlen, what type are more likely to come,” he said. “If I had more data at an early stage in the process, then I could run the same thing. This is not rocket science, universities in the United States do that. It’s very similar to predicting how people are going to vote in elections.”
According to the data, the more physically close you are to Kwantlen, the more likely you are to accept the offer. “I also tried other things, like whether or not they lived close to public transport. The beauty of this sort of thing is if you can get in things like bus routes. Nobody likes to pay parking,” Peplow said.
“What I found very striking was whether people were going into trades or undergraduate was a huge divider. And so, people who’re going into trades hardly ever drop an offer. I suppose that’s because Kwantlen is very good for trades, and there aren’t a lot of places teaching that in the Surrey or Langley area.
“Someone who’s accepted into the school of design, it’s almost unheard of them to drop an offer, because it’s quite hard to get into and it’s very good,. You have a lot of money going into the Chip and Shannon School of Design.
“I just want to make a point of, this is just an example. The new word is ‘predictive analytics’ or ‘predictive modelling.’ With more data and stuff like that, you can make it much more accurate. Let’s say I was sitting in the registrar’s office with a computer, and every time someone takes up an offer, drops this, drop that, then the model could be live, and I could make real-time predictions. That’s what big data does.”
Peplow’s report can be read here.
