Japan has experienced a number of large-scale disasters such as earthquakes and floods, so social demand for provisions, replacements and maintenance of infrastructures against the risks are requires.　However, both national and regional governments have difficulties to finance the needs because of the rapid aging of society, increasing social security costs, and burden of the government debt.　Therefore, public investment needs to be evaluated to account for efficient resource allocation. Benefits under uncertainty in practice are basically calculated in terms of the expected value for physical and human damages, while the use and estimates of monetary valuation of “anxiety for an uncertain situation” as a risk premium is an important issue for a cost-benefit analysis （MLIT, 2009）.　Asahi and Hagihara （2010 and 2013） showed the idea of vulnerability for infrastructure services and concluded the decision of public provision level needs to be considered from the viewpoint of household capacity for risk management and it is essential to understand the characteristic factors in decision making processes that household choice under uncertainty take into account. This paper investigates these factors using data from questionnaires on municipal water cutoff risk as a case study.　We focused on two decision making frameworks:the traditional decision making under risk and the prospect theory.　Whether risk averting behaviors and risk-cost tradeoff choices for water cut-off risk are affected by the risk aversion index, risk perception, and the reference point effect was examined with the logit regression and an interval regression analysis. Our main conclusion is that risk preference in terms of monetary scale does not affect the choice of a household in the case of water cut-off risk, whereas risk perception does.　The reference point effect in the combination of loss aversion and probability evaluation was observed in the household choice experiment concerning risk-cost trade-off in water pipe network investment.　These findings should be considered in an economic valuation model of vulnerability or risk premium for a better micro-based welfare measurement of public investment. JEL Classification:D61, D81, H54, H43