["Journal of Applied Econometrics, EarlyView. ", "\nABSTRACT\nThere is an ongoing debate about the buffer‐stock saving hypothesis. This paper develops a formal test of the hypothesis by examining the sub‐unity of the marginal propensity to consume out of permanent income (MPCP). Our test relies on a novel estimator of the average derivative of a nonparametric consumption model, using constructed measures of latent permanent income shocks that accommodate nonparametric income distributions. Following Li and Vuong (1998), we use a Kotlarski‐based approach to account for asymmetric errors, which is crucial for testing buffer‐stock behavior. Applied to the US Panel Study of Income Dynamics (PSID) data, our test rejects a unit MPCP, providing empirical support for buffer‐stock saving.\n"]