We present an alternative measurement method of investor overconfidence, using unique survey data on stock market predictions of investors. We apply the Parkinson estimate based on extreme bounds around the stock forecast to deduce investor confidence. The results support overconfidence.