If you select for careful, risk-averse populations, you'll see correlary effects of your selection and theirs, irrellevant to the measure you're taking.