This paper investigates precautionary saving under liquidity constraints in Pakistan using household panel data. In particular, while we estimates Kimball's (1990) prudence parameter, we deviate from Dynan's (1993) framework by explicitly considering liquidity constraints, as in Zeldes (1989). By doing so, we attempt to differentiate the standard precautionary saving caused by uncertainty from the one due to liquidity constraints. Furthermore, endogenous liquidity constraints are considered to resolve issues of selection biases. In this study, we document substantial evidence of the presence of precautionary saving in Pakistan. More specifically, the estimated prudence is significantly higher for liquidity-constrained households as compared with unconstrained ones. The results support the emerging view that facilitating saving may often be more important than finding better ways of lending to the poor.