In contrast to providing standard reminders about remembering household budgets, does asking survey respondents about their discretionary income and its use affect their voting responses in a national advisory referendum survey? We explore this question using U.S. household data from a unique set of multi-mode random samples (telephone and Internet surveys), and an advisory referendum concerning the Kyoto Protocol. The contingent valuation method is applied to estimate household willingness to pay (WTP) for a split-sample treatment: respondents who only received a standard reminder of household budgets (control group) versus respondents who received two mental accounting-type questions on discretionary income and its uses (treatment group). Results indicate that the treatment significantly influences voting responses and lowers estimated household WTP.