Transit Bus Validation Trials

Fifty-one drivers of Atlanta municipal buses (38 men, 13 women) were chosen by MARTA management for participation in a WayPoint validity study. Approximately half the drivers had ten or more years service without a preventable collision. The other half had at least two preventable collisions during the two years prior to the study.
Either the individual or group version of WayPoint was administered to the drivers. Information about collision history was not revealed to us either by management or by the drivers themselves until we had categorized drivers as high or low collision and submitted the employee's number to MARTA management. Drivers were told nothing about the purpose of the test until after they had finished taking the test.
Results. Comparing predicted with actual collision records, WayPoint was extremely accurate in classifying the drivers, as shown in the table at right:

39 out of 51 drivers (76%) were correctly classified as either High or Low Risk. This result was statistically significant and would occur by chance only 2 times in 10,000 attempts. The

figure below shows that the difference between WayPoint High and Low Risk groups was significant not only for "preventable collisions" (P) but also for "non-preventable" (NP) collisions. A middle group that MARTA recognizes as "probably preventable" (PS) fell just short of significance. This finding gave substance to a long-held impression by management that some drivers seem to have more "non-preventable" collisions than others.

Had WayPoint been used as a pre-employment screening test, -the "Accept" group (Low Collision) would have had only 32 of the 107 total collisions (30%) or 1.14 collisions `per operator. The "Reject" group (High Collision), on the other hand, had 70% of the total collisions or 3.26/operator. Replacing "Reject" operators with operators from the "Accept" group would have avoided 49 collisions and reduced the collision rate by 46%. Given that this study had an approximately 50/50 split in high and low collision drivers rather than the more usual 80/20 split in a population of drivers, a 25% reduction in fleet collisions is a more realistic expectation.
Subsequent to this study, we tested a MARTA driving instructor

Continue