As colleges put together attrition and completion models, they make choices about what data to analyze and include. In discussions, I occasionally hear institutions say they do not want a model that is centered on survey data. Usually, the rationales are focused on not being able to get survey responses from all of their students or thinking they can create a solid risk model without survey data. I understand their concerns, and I would not recommend an approach that is based solely on survey data. But, fundamentally, the value of survey data in risk prediction is unmistakable to me. Let me explain a little further:
First, some information is simply not available in existing student records. Institutional records have data about pre-college experiences, enrollment patterns, overall academic performance, and financial aid. Learning management systems have data about some on-line course behaviors. But neither has information about student goals or commitment levels. I’ve talked before aboutnon-cognitive factors that affect success, and surveys are a solid method for collecting non-cognitive information. Student factors such as self-confidence, grit, college knowledge, self-perceptions, and motivation are important. And what about early adjustment to college? Surveys can help us understand whether students feel like they fit and how they are changing at various times throughout their college career. Surveys can help us spot simple behaviors including study time, employment hours, and involvement that will help or hurt their chances for success. With any of this information, we can provide students with individualized feedback in real time.
Second, surveys are valuable for their speed and efficiency. Students can conveniently complete surveys online in a short period of time, sometimes even from their mobile devices, and a system can collect responses from hundreds or thousands of students in just a few days. So surveys are fast and efficient, allowing us to gather critical information before faculty or staff have a chance to observe or record behaviors or grades.
Third, and maybe most importantly, survey responses are related to attrition and completion. In fact, there are easily more than 50 years’ worth of articles, books, and other peer-reviewed sources that use surveys to predict attrition and completion. This research is filled with retention studies that confirm the importance of goals, commitments, non-cognitive variables, social and academic integration, behaviors, and more in predicting college student success. Many of these studies include survey data. They also control for information that would come from existing college records, including demographics, test scores, and enrollments. In other words, these studies show that survey data adds to our ability to predict risk.
Our research also confirms the power of survey data. At many campuses, issues such as homesickness, academic resiliency, and even intent to leave are predictors of college retention even when we control for other campus data. Whether we use correlations, regression models, or more sophisticated analysis such as path models, survey data has real power. It relates to both academic performance and retention in both the immediate semester as well as beyond the current academic year.
Surveys are an important tool for understanding and predicting student retention, completion and success. Why wouldn’t survey data be part of our work?
For a closer look at how survey data can give us valuable insights into the mindsets of our students,check out our research note, “First Year Students Who Plan to Transfer: Characteristics and Implications.” Did you know that 74% of students who plan to transfer decideafterthey enter their institutions?
Next up, Dr. Sherry Woosley tackles why we needmorethan surveys to predict college student attrition risk!