In constitutional states, representatives exchange information, discuss the budget direction and agree on a budget. The agreement is assumed to hold for one fiscal year. To test the validity of this constitutional assumption, we implemented an online panel survey with a randomized conjoint design three times over one year in Japan to track the direction of respondents’ preferences within a multidimensional public policy space. The policy space consisted of spending on education, infrastructure, health insurance, pensions, and poverty relief programs, as well as fiscal retrenchment. Providing information on the poverty rate in the first wave directed respondents’ preferences toward support for poverty relief programs by either increasing or reallocating the budget. The effects persisted in the second wave 5 months later across a diverse range of respondent backgrounds and political positions. By the third wave one year later, the effects had diminished. Once placed in a multidimensional space, information exchange might have a more extended scope than unidimensional approaches have shown. This finding, we believe, can broaden our capability to implement policies for poverty reduction.