A new framework referred to as the mean impact curve (MIC), is used to model and test for nonlinear mean reversion in stock returns. The MIC mea- sures how news is incorporated into the conditional mean of asset returns, thereby providing an extension of the news impact curve of Engle and Ng (1993) which measures the relationship between news and conditional volatility. An explicit parametric specification of the MIC is derived from a model where investors follow heterogeneous trading strategies.

The empirical results show strong evidence of significant nonlinearities in the Finnish stock index and some of its components. The shape of the MIC for these stocks shows that there is mean aversion in the conditional mean for ‘small’ shocks, but mean reversion for ‘large’shocks.