In households surveys, quantities of interest are frequently elicited using categorized (range-card) formats rather than open-ended questions. One advantage of this format is that is typically reduces item non-response. Unfortunately, results from research in social psychology suggest that the choice of bracket values in range-card questions is likely to influence responses. As yet, there is not much known about the effects of bracketing bias on the measurement of economic quantities and regression analysis. This paper reports evidence on existence and size of bracketing bias based on data from controlled survey experiments. I also discuss strategies for avoiding bracketing bias in household surveys.