Using data from the New Jersey HSPA test, this paper finds statistically significant evidence that students who barely failed the exam, especially the math exam, were more likely to drop out than students who barely passed it. While the raw dropout rate of the students who barely failed the initial test was 5–6%, the regression discontinuity estimates indicated a 1 percentage point increase in the exit probability for barely failers compared to barely passers. In the retest the regression discontinuity estimates indicated a 1–1.5 percentage point increase of exit probability for barely failers, whose actual exit rate was 2–3%. The regression discontinuity estimates indicated a 14% increase in the raw exit probability for LEP students who barely failed the first LAL test. The regression discontinuity estimates in dropout propensity were larger and more significant for Black and Hispanic students as well as economically disadvantaged students.