A questionnaire used in student evaluations of interdisciplinary courses during six semesters contained two Likert items stated in a direct negative mode which were embedded in a questionnaire (14-18 items) in which the remaining items were phrased in a direct positive mode. In the seventh semester and thereafter, the two negative items were restated as direct positive stems. Item-analysis demonstrated that in the direct negative mode, the two items had low item-to-total correlations and that the internal consistency reliability of the sum score could be improved by eliminating the two negatively phrased items. Also, the two negatively worded items defined a separate factor. After they were reworded into a direct positive mode, these two items showed markedly improved item-to-total correlations. Moreover, the unique factor disappeared, which suggests that it was a methodological artefact probably attributable to respondent carelessness. Including a few negative items in an otherwise positively stated questionnaire leads to ambiguity of results rather than controlling for response sets. We therefore recommend against the practice.

No abstract

No abstract

No abstract

A convenience sample of 1,741 Internet users completed a 12-item financial risk-tolerance questionnaire. They also rated themselves on their tolerance for financial risk using a 4-point rating scale. The 12-item summated rating score was used to predict the self-rating. The residual between actual and predicted self-rating was compared by sex. The residual for males was positive, indicating that men tended to overestimate their proclivity for taking risks. Conversely, the residual for females was negative, suggesting that women underestimate their tolerance for risk. The relationship held when controlling for other factors linked to risk tolerance, i.e., age, household income, marital status, and education. It was also noted that risk tolerance was overestimated by younger respondents and those with a graduate education.