We develop a model of stock return volatility that includes a positive risk-return relation, a significant volatility feedback effect and explains a rich set of empirical phenomena. It explains the negative correlation between returns and volatility we observe in stock data. We find that including volatility feedback dramatically strengthens the risk-return relation. Contrary to some previous research we find that volatility feedback is economically significant, explaining around 13 percent of daily, and 28 percent of monthly, stock return volatility. We demonstrate that previous studies have found an economically insignificant feedback effect because of their choice of either empirical methodology or model specification.