Non-technical summary

Research Question

Previous studies on macroeconomic expectations suggest that the rounding of survey responses can be interpreted as a signal of subjective uncertainty. However, the actual motivations behind survey participants’ rounding decisions have rarely been investigated directly. This paper uses self-reported data to examine whether private households in Germany round their probabilistic inflation expectations – where respondents assign probabilities to different inflation outcomes – in order to express their uncertainty.

Contribution

We use data from the Bundesbank’s Survey on Consumer Expectations (BOP-HH) and rely on a special question included in one wave of the survey that asks respondents to indicate the primary reason behind their decision to round their probability forecasts. This allows us to directly observe how individuals explain their own rounding behavior. Our paper contributes to the literature that uses rounding as a proxy for uncertainty and to broader discussions about how to measure economic uncertainty in household surveys. In particular, we provide evidence on whether rounding reflects uncertainty or arises for other reasons, and whether the frequency of different reasons for rounding varies across socioeconomic groups.

Results

We find that although rounding is positively correlated with uncertainty, only 14 % of respondents report that they round their probabilistic expectations due to uncertainty. Most households state that their rounded probabilities either reflect their true expectations or help to simplify the response process. Older, female, and lower-income individuals are more likely to report rounding due to uncertainty. However, there are no significant differences in actual uncertainty levels between households who round due to uncertainty and those who round for other reasons. Rounding due to unfamiliarity with inflation or probability concepts is associated with above-average uncertainty, while rounding to reflect true expectations is associated with below-average uncertainty. We also find no meaningful differences in how frequently households update their rounding choices, their implicit expectation uncertainty, or their point forecasts depending on their reported reason for rounding. Overall, our findings suggest that researchers should be cautious in interpreting rounding in responses to probabilistic question formats as a signal of uncertainty in household surveys.