While it is known that access to physical infrastructure enhances household welfare, there are very few micro-econometric studies that analyze its role in mitigating chronic and transientpoverty. This paper aims to bridge this gap in the existing literature by evaluating the impact of a large-scale irrigation infrastructure project implemented in Sri Lanka. It identifies the treatment effect of irrigation access by exploiting a natural experimental situation where the government used lotteries to randomly distribute irrigated plots. We extend the seasonal consumption smoothing model of Paxson (1993) by introducing endogenous credit constraints. By using unique household level monthly panel data over a period of two years, it is shown that the average income increases, and the probability of binding credit constraint declines with irrigation accessibility through which transient poverty is mitigated. These empirical results suggest that irrigation infrastructure has a positive impact on reducing both chronic and transient poverty directly and indirectly by improving income and relaxing credit constraints. The structural estimation results support the validity of our theoretical framework.