The point estimate for the cholesterol test is
largest, but also has a substantial standard error. The results in columns (4) to (6) of Table 3
show the marginal eects from probit estimations. They estimate the impact of the cholesterol test to be somewhat larger, and marginally signicant, at least in our preferred specication
with the tightest controls in place. Thus, if anything, the cholesterol test treatment increased
the fraction of rejected donations, but not the fraction of usable donations.