Self-Reports of Academic Performance: Response Errors Are Not Well Behaved

Exploratory data analysis (EDA) is used to study errors in self-reports of lest scores and grades from a survey sample of college students. Both response and non-response are found to be systematically biased, with unfortunate effects in combination. Errors are not normally distributed, and would be better modeled as contaminated distributions made up of two or more simple distributions. Errors are correlated with each other and with other variables, leading to spuriously inflated as well as deflated intervariable correlations. These findings may be typical of survey data in general; hence, more realistic error models and robust estimation methods are desirable.