Trends in Predictive Analytics in Higher Ed: Five Questions with Mike Johnson, Director of IR at Dickinson CollegeReading time: 4 minutes
In this segment of Five Questions with… we talk with Dr. Mike Johnson, the Director of Institutional Research at Dickinson College about trends in predictive analytics in higher education. Dr. Johnson has been a pioneer with his use of predictive analytics on his campus, investigating everything from enrollment likelihood to student success to researching gender bias.
What do you see as the next big trends in predictive analytics on college campuses in the next few years?
I can’t really put my finger on new areas but I will say that predictive modeling will be pervasive within the next few years. I think that in and of itself is new. As we’ve experienced here, once the decision makers see the benefits of the accurate models that they’re using to influence these important decisions and this happens in real time, I am sure that the number of requests for these types of analyses are almost certain to increase.
You’ve spoiled them in essence. You may start out with an enrollment model but soon you’ll be asked, as we have been, to product models for retention or GPA for certain subgroups or perhaps donor activity in addition to the things that you’re going to discover yourself. Those may be outright requests.
I am not sure of a new area but certainly a new way of implementing this into the decision making process.
Is that a big change from the direction of the last few years?
Speaking from my own experience, I would say yes. The way in which we make decisions will become increasingly reliant on the use of predictive modeling- more so in the years to come and it certainly is more prevalent on our campus today than it was couple of years ago.
I really do think that once people experience the strength of the analysis and how this really kind of puts them in their comfort zone: We can make this decision now and feel good about it. That is more common on our campus today and it will be even more common in the years to come.
Do you think people have gotten more used to the concept of data being a key part of decision making? Is the answer different depending on someone’s role or department in an institution?
People from across the institution certainly differ with regards to where they are coming from. However, I no longer assume that someone who has a background in the humanities or in a non-technical field that they aren’t technically savvy when it comes to statistics or data analysis or even predictive modeling. Looking at some of the folks on our campus that I really consider to be heavy users of data and data analysis are actually English professors and two have a background in History- one in Russian History and one in American History.
I think the concept of using the results of good data analysis is pretty easy to grasp regardless of your role, regardless of your background or even your institution. I believe the key part of my job is to try and convince them of this by playing an active role on campus and becoming more involved in areas where my office can provide some useful and applicable analyses.
Have trends in predictive analytics changed how you communicate the results of your analytic efforts with decision-makers at your school?
Absolutely. It is not unique to the field of Institutional Research but anytime you are conducting a briefing or communicating results it is critically important to understand your audience – your client or your boss, whatever the case may be. You need to quickly assess your experience with the technical aspects of the analysis whether that is simple descriptive statistics or whether you are doing hypothesis testing or regressions or whatever. Also, to what extent have they used data analysis for modeling to facilitate their decision-making process in the past? All of this is critically important if you want to be able to communicate- and I think I have learned that over the years.
Yes, we often have sophisticated results but you don’t have to communicate those results in a sophisticated manner. It is really just common sense and I do present the results differently depending on who I am talking to.
If there was one fun project in analytics you could work on right now, what would it be?
This time of year I look forward to creating my enrollment model every year- and that’s not new. We’ve been using Predict now going on our 8th entering cohort. We’ve discovered that there are new and exciting ways to improve the accuracy and efficiency of the models each and every year. It really never gets dull. The environment changes or the data elements or the data structure changes. It’s even possible that your priorities or your objectives can be adjusted. These things along with the fact that we create three different models for enrollment- one for early decision, one for early action and one for our regular admits, really does keep us on our toes. Even though that is not a new project, we tackle it every fall semester, but every year it’s new and I look forward to it.
I will say also that it’s fun when you find new ways to use predictive modeling. Not too long ago we used it to investigate whether gender bias existed with our faculty salary. It was a question that came from the floor at one of our faculty meetings. We modeled it and thankfully we discovered there was no gender bias but we were really able to dig into that. For us that was a new application of regression modeling.
We’ve become quite creative with some of our retention modeling and actually with some pretty good success where we compared the predicted vs. the actual first semester GPA. I actually got an email from not too long ago from a colleague who recently used it to investigate the inner workings of the U.S. News Rankings. You never really know when a new opportunity is going to surface but each time they do I love to jump on them because they are really fun.
Any thoughts on how the larger trends in predictive analytics interact with your work? Leave a comment below!