The statistically significant regression discontinuity estimates in Table 2 confirm the link between failing the HSEE and the increased probability of dropping out in early grades. The discontinuities based on the math score were larger than those based on the LAL scores. For example, the regression discontinuity estimates for the math score (−.011 with a standard error of .001) were twice as high as the regression discontinuity estimates for the LAL score (−.005 with a standard error of .002). This pattern held when looking at the observations in a smaller neighborhood closer to the cutoff score (columns 5, 6, 7 in Table 2). We noticed that the average passing rate in math was lower than the LAL rate for the HSPA exam (76% vs. 87%, as shown in Table 1). It is plausible that the HSPA math standards were set too high or that the students were poorly prepared for the math test. If either case is true today, students might become more discouraged when failing a more difficult test, or the perceived cost of preparing for (and passing) the retest might be higher when there is inadequate educational support for building up their math skills.