Myopic Loss Aversion, the Equity Premium Puzzle, and GARCH

This paper concerns the distributional assumptions made on stock returns in the myopic loss aversion (MLA) proposed explanation to the equity premium puzzle. While Benartzi and Thaler (1995) assume temporal independence in these returns, we introduce a more realistic assumption incorporating conditional heteroskedasticity. This involves the work on temporal aggregation of GARCH processes of Drost and Nijman (1993). Using Swedish data, our estimation method produces an overall larger evaluation period than the one originally obtained by Benartzi and Thaler, e.g., over the sample period July 1961 through December 2003 the evaluation period increases from 12 to 17. This shows that MLA indeed can explain a large equity premium but, also, that the model is sensitive to the distributional assumption made on stock returns.