Adding ACT scores to models with HSGPA contributes little to the prediction of college graduation and does not reduce between-school variability¶

Claim · All claims
Evidence · 1 study (2 entries) · associational r3 · q2
Currency · newest study 2019 · checked against its sources 2026-10-20 (automated, 2 of 2 entries) · not reviewed by a person · overdue for a recheck since 2022: tool and model accuracy changes quickly

Subclaims¶

q2 i? When HSGPA and ACT are entered together, the linear ACT coefficient shrinks from 0.129 to a nonsignificant -0.016 (a quadratic ACT term, -0.108, remains in the model) while the HSGPA coefficient is essentially unchanged (0.708 vs 0.703). → Allensworth 2019 q2 i? School-level variance in the combined model (0.622) is similar to the HSGPA-only model (0.603), so individual ACT scores do not reduce high school variability in predicting graduation. → Allensworth 2019 (2)

q2 i? In combined models the linear ACT coefficient shrinks to a nonsignificant -0.016 while the HSGPA coefficient is essentially unchanged, and school-level variance is similar (0.622 vs 0.603). → Allensworth 2019

Evidence¶

Allensworth 2019¶

Allensworth, Elaine M., & Clark, Kallie. (2019). Are GPAs an Inconsistent Measure of College Readiness across High Schools? Examining Assumptions about Grades versus Standardized Test Scores. University of Chicago Consortium on School Research. https://consortium.uchicago.edu/publications/are-gpas-inconsistent-measure-achievement-across-high-schools-examining-assumptions

q2 · i? · associational · r3

Joint hierarchical model entering standardized HSGPA and ACT with quadratic terms (n=17,753). The linear ACT coefficient becomes nonsignificant, though a quadratic ACT term (-0.108) remains, so the authors describe ACT as contributing little to the prediction rather than nothing. The ACT slope still varies significantly by high school (0.213), ranging from negative to positive across schools.

"The main HSGPA coefficient does not change substantially relative to the model without ACT scores in Table 2 (0.708 vs 0.703), but the main ACT coefficient shrinks considerably from the model without HSGPA (from 0.129 to a nonsignificant -0.016)."

Allensworth 2019 (2)¶

Allensworth, Elaine M., & Clark, Kallie. (2019). Are GPAs an Inconsistent Measure of College Readiness across High Schools? Examining Assumptions about Grades versus Standardized Test Scores. University of Chicago Consortium on School Research. https://consortium.uchicago.edu/publications/are-gpas-inconsistent-measure-achievement-across-high-schools-examining-assumptions

q2 · i? · associational · r3

Variance-component comparison between the combined HSGPA-plus-ACT model and the HSGPA-only model. The article states plainly that individual-level ACT scores do not reduce between-school variability.

"Because the ACT score contributes little to the prediction, there is a similar amount of school-level variance in the combined model (0.622) as the model that includes HSGPA alone (0.603, from Table 1). ACT scores used at the individual student level do not reduce the variability by high school in predicting who will graduate college."

Discussion¶

Related Claims¶

• ACT scores weakly predict college graduation, and the school-level variance in the ACT slope (0.192) exceeds the average slope (0.129), so school effects introduce more noise than the ACT signal — related
• ACT scores show a weak relationship with college graduation, with slope variation across high schools larger than the average signal — related
• School average ACT scores predict college graduation for students with the same individual HSGPA and ACT score, reducing between-school variance by 42 percent, while school poverty is not significant once average ACT is included — related
• For four-year (rather than six-year) graduation, ACT scores are slightly predictive but their coefficient is one-sixth the HSGPA coefficient — related
• Combining ACT scores with HSGPAs reduces high school variance in college graduation rates relative to HSGPA-only models — reports the opposite
• HSGPA has a strong, positive relationship with college graduation that does not vary significantly in its linear slope across high schools — related
• ACT scores show significant subgroup misprediction by race, ethnicity, and gender that HSGPA models do not, and school-level variance is smaller among students with the same ACT score than the same HSGPA — related
• Both HSGPA and ACT strongly predict four-year college enrollment and enrollment in colleges with strong graduation rates — related
• GPA combined with test scores was the best prediction of college degree attainment for all English Learner groups, but the test-score relationship partly reflects school composition — reports the opposite
• Both HSGPA and ACT scores leave large high school effects on college graduation unexplained, which the study recommends measuring and publishing directly — related
• Students with the same HSGPA or the same ACT score graduate from college at different rates depending on which high school they attended — related
• HSGPAs have a strong, consistent relationship with college graduation that is larger than school effects, while ACT scores have a weak relationship smaller than high school effects with slopes varying by school — related
• Among Chicago graduates who enter four-year colleges, high school GPA is the strongest predictor of graduating within six years — related
• The linear slope of the HSGPA–college graduation relationship does not vary significantly across high schools, so HSGPA provides a consistent signal across schools — related
• College graduation rates vary significantly across high schools among students with the same HSGPA, and about one-fourth of that school-level variation is induced by comparing students with similar HSGPAs — related