The overall effect in the entire database is an 8.0 percent fatality reduction (from 48.67 to
44.80), which is statistically significantly at the two-sided .05 level (Z=2.12). Even though the
overall effectiveness of 8.0 percent is statistically significant there is a caution flag, namely, the
wide difference between the effectiveness for pickup trucks (-4.9%) and SUVs (+17.5%).
Intuitively, there is no obvious reason for inconsistency between pickup trucks and SUVs.
Perhaps other factors are causing SUVs to have a larger fatality reduction than pickup trucks.
One possible factor could be a vehicle-age effect, due to the certified vehicles in the
database being newer than the pre-certified vehicles. The age effect might be different for
pickup trucks and SUVs. Table 4 limits the preceding analyses to just ±1 or ±2 MY before and
after certification. Reducing the database to just ±1 MY eliminates most of the age effect
because the Pre and Post vehicles are nearly the same age.