In this paper we develop a stochastic model for household liquidity. In the model, the optimal liquidity policy takes the form of a liquidity range. Subsequently, we use the model to calibrate the upper bound of the predicted liquidity range. Equipped with knowledge about the relevant control barriers, we run a series of empirical tests on a panel data set of Dutch households covering the period 1992-2007. The results broadly validate our theoretical predictions that households (i) exhaust most of their short-term liquid assets prior to increasing net debt, and (ii) reduce outstanding net debt at the optimally selected upper liquidity barrier. However, a small minority of households appear to act sub-optimally. Poor and vulnerable households rely too frequently on expensive forms of credit (such as overdrafts) hereby incurring substantial amounts of fees and fixed borrowing costs. Elderly households and people on social benefits tend to accumulate too much liquidity. Finally, some households take on expensive short-term credit while having substantial amounts of low-yielding liquid assets.