Of the returned surveys, 8 were
unusable, resulting in an effective response rate of 31%.2 To
ensure comparability, we deleted observations from the data
set in which complete sets of both primary questionnaire
data and secondary performance data were not available.