The mean-variance model with capital controls and expectations formation. A test on German portfolio data

Abstract

The static mean-variance (MV) portfolio model is extended with capital controls and tested on quarterly data for German private net wealth. Investors have to learn about the forecasting model using the available information. Estimation takes place in two stages: for each period an up-to-date forecasting model is estimated to compute time-varying first and second moments, which are then used to estimate the degree of risk aversion and capital control effects. The MV-restrictions are strongly rejected. Poor measurement of expectations could partly explain this finding as expected next-period returns hardly affect asset demand. In contrast, a strong link exists between asset holdings and long-term expected returns.