<p>
If you consider all available stats instead of just math SAT, math SAT I adds little improvement to predicting college GPA beyond the other stats. For example, the Geiser UC study found the following regression coefficients for cumulative 4th-year college GPA among math/science majors:</p>
<p>Math/Physics GPA – HS GPA: 0.35, SAT V: -0.01, SAT M: 0.02, SAT W: 0.11, SAT II M: 0.12, SAT II 2nd: 0.09</p>
<p>HS GPA was by far the most influential stat. Math & verbal SAT I had a negligible contribution. SAT II tests and SAT writing had a significant contribution, but notably less than HS GPA. When considering all of these stats as well as education, income, and API; Geiser was only able to account for 26% of the variance in college GPA. With just M+V SAT, he was only able to account for 13%. Grad rate followed a similar pattern.</p>
<p>The Duke study considered even more variables, including things like the application LORs and essay. It found the following variables had the greatest contribution to dropping out of an engineering major, ranked from most significant to least significant. + indicates increases changes of dropping out. - indicates decreases chance of dropping out. Note that SAT scores had a lower contribution than all measured variables except for having desirable personal qualities (desirable person qualities increased chance of dropping out of engineering).</p>
<ol>
<li>Being female (+)</li>
<li>HS curriculum (-)</li>
<li>How easy/harshly college classes graded (-)</li>
<li>Application essays (-)</li>
<li>HS achievement (-)</li>
<li>Being an URM (+)</li>
<li>Being Asian (-)</li>
<li>Application LORs (-)</li>
<li>SAT score (-)</li>
<li>Application desirable personal qualities (+)</li>
</ol>