Routinely randomise the display and order of items to estimate and adjust for biases in subjective reports
Recent psychological research has experienced a stark increase in the use of repeated subjective reports, such as online, smartphone-based daily diaries. This development holds great opportunities to study causal processes and developmental change, but it also brings new challenges. As is often the case in psychology, interest in specific substantive questions can outrun measurement research, so that many measurement and design decisions are made ad hoc and not evaluated rigorously. Recent work
... has added initial elevation bias to the list of common pitfalls that should be avoided when using subjective reports. Initial elevation bias refers to the phenomenon that negative states (i.e, thoughts, feelings, behaviors) in subjective reports are elevated when first assessed, as compared to later assessments. In this article, we employ a planned missingness design in a daily diary study of more than 1,200 individuals that were assessed over a period of up to 70 days to estimate and adjust for initial elevation bias. First, we found only a negligible bias related to initial presentation and item order: items were not answered differently depending on when and where they were shown. Second, we show that residualising these biases had minor effects. We conclude from our findings that the initial elevation bias may be more limited than previously reported and may only act at the level of the survey, not at the item level. We encourage researchers to make design choices that will allow them to routinely ascertain potential measurement reactivity biases empirically in their studies. Specifically, we advocate that researchers should routinely randomise item display and order in planned missingness designs, so that they can estimate biases affecting subjective reports. Another benefit of routinely randomizing item display is that it allows constructing brief survey instruments without compromising the construct breadth and the number of constructs covered.