Social Desirability Bias Against Admitting Anger: Bias in the Test-Taker or Bias in the Test?

The veracity of self-report is often questioned, especially in anger, which is particularly susceptible to social desirability bias (SDB). However, could tests of SDB be themselves susceptible to bias? This study aimed to replicate the inverse correlation between a common test of SDB and a test of anger, to deconstruct this relationship according to anger-related versus non-anger-related items, and to reevaluate factor structure and reliability of the SDB test. More than 200 students were administered the Marlowe-Crowne Social Desirability Scale Short Version [M-C1(10)] and the Anger Parameters Scale (APS). Results confirmed that anger and SDB scores were significantly and inversely correlated. This intercorrelation became nonsignificant when the 4 anger-related items were omitted from the M-C1(10). Confirmatory factor analyses showed excellent fit for a model comprising anger items of the M-C1(10) but not for models of the entire instrument or nonanger items. The first model also attained high internal consistency. Thus, the significant negative correlation between anger and SDB is attributable to 4 M-C1(10) anger items, for which low ratings are automatically scored as high SDB; this stems from a tenuous assumption that low anger reports are invariably biased. The SDB test risks false positives of faking good and should be used with caution.