Water problem in rural areas of Ethiopia is two-fold: low coverage levels and poor quality that require urgent attention to reduce associated health and social implications. Women and children spend hours a day collecting water: time that would be better spent in education or productive employment. Cognizant of this fact, the government and donor organizations are currently performing a number of activities to improve the coverage and quality of water supply with partial cost recovery systems. Hence, the affordability and willingness of the consumers that are supposed to be served need to be examined. The primary objective of this study is, therefore, to estimate willingness to pay (WTP) of rural households for improved water service provision and identify its determinant by employing contingent valuation method (CVM) in Haramaya district. The study used primary data obtained from a survey conducted on randomly selected rural households. We used double bounded dichotomous choice elicitation method administered by in-person interview. The data was analyzed using descriptive statistics and bivariate probit model. Response to the hypothetical scenario revealed that sampled households expressed their WTP with a mean WTP of 27.30 cents per 20 liters jerrycan. The results of bivariate probit model revealed that household income, education, sex, time spent to fetch water, water treatment practice, quality of water and expenditure on water have positive and significant effects on WTP for improved water service provision, while age of the respondent has a negative and significant effect.