Biases in retrospective self-reports of time use: An empirical study of computer users

Research on information systems (IS) use has often relied upon retrospective self-reports. One example is when the amount of time spent using a system is reported, often as an indication of user acceptance. In this study, self-assessments of computer usage are compared with computer-monitored interactive use and conned time for 401 managers and professionals. When self-assessed use was compared with logged interactive use, there was a 32% difference in the average amount of use (3.9 vs. 2.7 hours/day). When the self-assessment was compared with total conned time, the averages were similar to each other (3.9 vs. 4.0 hours). In both comparisons, though, there were considerable differences between individual self-assessments and logged time (with a median absolute percentage difference of 47% when compared with logged connect time). Individual estimates of use tended to regress toward the population's mean use. Those whose use of the system was relatively light overestimated their use, while heavy users underestimated theirs. When a test of the relationship between user satisfaction and amount of use was conducted using self-assessments, the relationship was not statistically significant. When the same test was conducted using logged connect time, the relationship was significant. These results suggest that fare should be exercised in operationalizing the user acceptance construct through self-reports of time use.