The Hedonic Editing Hypothesis: Evidence from the Finnish Stock Market

Abstract Prospect theory-based hedonic editing hypothesis posits that people integrate or segregate multiple outcomes so as to achieve the highest perceived value. We test the hypothesis in an actual decision-making context by investigating how stock market investors time their sales of stocks when realizing gains and losses. If the principles that guide investors' behavior are those suggested by the hedonic editing hypothesis, we should observe investors integrating losses more frequently than gains and integrating smaller losses with larger gains rather than the other way around. Our results do not conclusively support either of these assumptions but suggest that the relationship between prospect theoretic preferences and investor behavior is not as general as it might seem. Keywords: Hedonic editing hypothesisProspect theory ACKNOWLEDGMENTS The author is indebted to her dissertation advisor Jukka Perttunen for numerous constructive discussions and for his encouragement. Thanks are also due to Matti Keloharju, Hersh Shefrin, and seminar participants at the GSF workshop 2005 for helpful comments and suggestions. Appreciation is also expressed to Finnish Central Securities Depository for provision of the data. This research was financially supported by the OKO Bank Group Research Foundation, the University of Oulu Foundation, the Finnish Foundation for Economic Education, the Finnish Foundation for Share Promotion, the Finnish Konkordia Fund, the Finnish Cultural Foundation, The Finnish Savings Banks Group Research Foundation, the Tauno Tönning Foundation, and the Foundation for the Promotion of Finnish Securities Market. The support is gratefully acknowledged. Notes 1. To be precise, the term 'hedonic editing' is from CitationThaler and Johnson 1990. 2. As CitationThaler 1985 notes, this analysis is generalizable to cases with several outcomes. 3. Individual investors are initially identified by their social security number. For security reasons, each original identification number is replaced with a specific running number. 4. It is possible that investors have other reference points instead of or in addition to the purchase price. For example, Heath, Huddard, and Lang 1999 consider several potential reference points and find that the highest stock price reached over the prior year has the strongest effect on the decision of an investor to exercise stock options. Grinblatt and Keloharju 2001b, in turn, report that monthly highs and lows affect the propensity of investors to sell a stock. 5. A notable exception is the research conducted by CitationGrinblatt and Keloharju 2000, 2001a, 2001b, and Citation2004, which covers various topics related to the trading behavior of investors. 6. For the most part, results remain qualitatively unchanged if mean value is used instead of median. 7. Another test is conducted (with fairly similar results) by excluding investors who do not execute a purchase within 10 days of the sale date. 8. We thank Matti Keloharju for inviting us to consider this point and for directing us to the relevant reference. 9. On the other hand, one should recognize the possibility that the causality may run from preferences to order strategies. That is, the preferences of investors to integrate losses (segregate gains) may in fact lead them to use market (limit) orders instead of limit (market) orders. 10. See, for example, Dyl 1977 and CitationGrinblatt and Keloharju 2004. 11. According to Table 3, there are six cases where the association between multiple sale and losses appears to be insignificant, after adjusting for the effect of the third variable. It is important to note that the Mantel-Haenszel statistics have low power in detecting associations when, as in the present cases, the pattern of association in one stratum is in the opposite direction of the pattern displayed in the other stratum. Consequently, an insignificant Mantel-Haenszel statistic either indicates that there is no association or suggests that neither pattern of association has enough strength to dominate the other. We believe the latter to be true in this instance, as the stratum-specific odds ratios are statistically significant in all six cases. 12. We thank Hersh Shefrin for pointing this out to us.