Abstract: | The excess smoothness puzzle is explored using a simple version of the permanent income hypothesis. The new feature is that consumers do not know the observed data‐generating process for income. Instead they estimate the income process every period using the past income data and update their income forecasts as new data arrive. Two scenarios are examined: first, where the income has a linear deterministic trend and second, where the income has a constant trend. There is a misspecification bias in the estimate of the marginal propensity to consume (MPC). This bias is of second‐order importance in the first scenario while it is of first‐order importance in the second. We conclude that the second scenario, which may be relevant for less developed countries, may offer a potential solution to the excess smoothness puzzle. Copyright © 2001 John Wiley & Sons, Ltd. |