This study assesses the predictive power of multi-item scales designed to measure personal finance constructs compared to single items from the scales. We used a 2022–2023 data set of 993 adults in the United States collected by the National Endowment for Financial Education (NEFE) that includes a range of demographic variables, multi-item scales, and other variables capturing financial behaviors and experiences. For the present analysis, we selected three scales and, for each scale, a financial behavior variable that should theoretically be associated with it. Logistic regression was used to evaluate how well the full scale predicted the financial behavior compared to a single item selected from the same scale. Single items with the highest and lowest factor loadings and discriminatory power as evidenced through confirmatory factor analysis and Item Response Theory modeling were selected from the scales and used for the analysis. As expected, full scales explained the most variance in the outcomes. However, in contrast to our expectation, the lowest and highest loading/discriminating items tended to have similar predictive power. Our paper ends with four key insights learned from the study, which we hope will spark further research and conversation on this topic.
Borowiec et al. (Fri,) studied this question.