Predicting Academic Success
Overview
Every year, millions of students enroll in college, yet approximately 40% don't complete their degrees. With such high attrition rates, universities desperately need better ways to predict student success. But can a single test score really tell us who will thrive in higher education?
Details
Every year, millions of students enroll in college, yet approximately 40% don’t complete their degrees. With such high attrition rates, universities desperately need better ways to predict student success. But can a single test score really tell us who will thrive in higher education? I investigated whether SAT scores alone can reliably predict cumulative college GPA—a question with major implications for admissions policies nationwide.
The Challenge
The College Board claims SAT scores strongly predict college performance, but independent research often contradicts this, showing that high school GPA is more predictive.
Unlike most studies that use multiple variables, I deliberately limited my analysis to SAT scores alone to test whether this single metric that universities invest so heavily in holds predictive power on its own.
If SAT scores aren’t predictive, universities are making life-changing admission decisions based on flawed metrics.
My Approach
I employed simple linear regression analysis on 84 student records, focusing exclusively on the relationship between SAT scores and cumulative GPA. This minimalist approach was intentional—stripping away all other variables to isolate the true predictive power of standardized testing.
Key Analytical Steps
Correlation Analysis Used the Pearson correlation coefficient to measure the linear relationship between variables. Result: 0.637 correlation—suggesting a moderate positive relationship.
Ordinary Least Squares (OLS) Regression Built a predictive model to quantify exactly how much GPA changes for each point increase in SAT score. The model showed statistical significance (p < 0.001), but with important caveats.
Model Validation Examined residuals and diagnostic statistics to assess model reliability:
R-squared: 0.406 (SAT explains about 41% of GPA variance) Normality tests revealed some deviation, suggesting model limitations
Visual Analysis Created scatter plots with regression lines to reveal the story beyond the numbers—where predictions worked and where they failed spectacularly.
Key Findings
There’s a relationship. SAT scores and GPA move together in the same direction—higher SAT scores generally correlate with higher GPAs.
But it’s not strong enough: With only 41% of GPA variance explained, SAT scores leave 59% of academic success unexplained. That’s a massive gap.
The scatter plot reveals the truth: While many students cluster near the prediction line, substantial outliers exist—students with high SAT scores who struggle, and students with lower scores who excel.
What This Means for Universities SAT scores alone are insufficient for predicting academic success. The model’s limitations suggest that:
Other factors (motivation, study habits, support systems) play enormous roles
A single test score can’t capture the complexity of student potential
Current admission practices may be excluding qualified students while admitting others likely to struggle