Many studies have shown that consumption responds to the arrival of predictable income (excess sensitivity). This paper uses a buffer stock model of consumption to understand what causes excess sensitivity and to test which parameterization is consistent with empirical excess sensitivity estimates. Using high-frequency granular data from a personal finance app, I find that while liquidity constraints are a proximate cause, preferences are the ultimate cause of excess sensitivity. Furthermore, it finds that for feasible parameters, a quasi-hyperbolic version of the model is more consistent with the level of excess sensitivity relative to a standard exponential model.