4. Empirical Results
4.3 Robustness checks
4.3.1 The Analysis of Robustness for all investors
From this section, we start to do some robustness analyses in order to check the objectivity of empirical results above. We divide our data into two samples from the middle of time. In Table 7, panel A involves the fore half of sample periods, from January 2004 to June 2006, with 620 trading days while panel B includes the remaining periods, from July 2006 to December 2008, with 621 trading days. Here, three groups are sampled by absolute return method and trading activities are the average of trade size and trading numbers on following 5 days of reference day.
We can see that no matter which sample period is, the results are coherent with our expectations. Investors in group 1 trade more rather than in group 2 and group 3 due to their risk-seeking attitude resulted from loss numbness. On the other hand, investors in group 3 are the most conservative toward making trades since they are sensitive to the next loss. All P-value are under 0.0001 which indicates the
significance in statistic. According to this robustness analysis, we verify the empirical results we provide in section 4 are unbiased. The degree of prior loss plays an
important role in making subsequent trades for all investors.
4.3.2 The Analysis of Robustness by trader types
We further sort data by trader types to develop robustness test, which provides clarification that whether prior loss degree have an impact on following trading activities for each trader type. Trading data of each trader types was also split into samples by two time periods, the earlier 2.5 years and another 2.5 years afterward, for 620 and 621 trading day respectively. We follow absolute return method again and the results of robustness checks for three different trader types are as follows.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
For domestic institutions, we provide the result in Table 8. Panel A is the trading data on previous 2.5 years and panel B shows it for the remaining 2.5 years. On the first 2.5 years, the difference of mean between three groups is not significant because P-value is almost higher than 5%. On the back 2.5 years, the average on trade size and trading numbers of group 1 are greater than group 2, and P-value is under 0.0001 showing the difference between group 1-2 is significant. However, the trading
activities of group2 do not dominate over group 3 in this time period, which is against our expectations. Overall, it is not clear that the subsequent trading activities of domestic institutions are influenced by previous loss degree.
For foreign institutions, we can grab some information from Table 9, which shows the data on the prior and the back 2.5 years in panel A and panel B respectively.
To sum up the analysis of robustness, we can tell that foreign institutions trade with behavior bias since all P-value are much smaller than 5%. It is evident that prior loss degree is an essential concern toward trading for foreign institutions.
Finally, we discuss individual investors in Table 10. As the same format we provide above, panel A and panel B present trading information of individual investors on the fore 2.5 years and the back 2.5 years. The statistical results support our hypotheses because all P-value is under 0.0001, which is an extremely small number showing the difference of trading activities within three groups is strongly significant. Combining this robustness analysis with our empirical results in section 4.2.3, we draw conclusion that there are irrational behaviors among individual investors and the evidence is obviously objective. That is, the degree of previous loss has a great effect upon the next trading actions for individual investors.
‧
Robustness checks for all investors. In this table, we check the robustness of our results above.
We divide all investors to two samples by the middle of year. In robustness analysis here, we sample the data by absolute return only. We examine the difference between group 1-2 and group 2-3 on both the first and the second period by T-test with 5% significant level. Panel A presents the analysis of the first 2.5 years (from January 2004 to June 2006) while panel B shows the remaining 2.5 years (from July 2006 to December 2008).
Trade Size Number of Trades
N Mean St. Dev. Mean St. Dev.
Panel A: trading activities on the fore 2.5 years
Group 1 100.2 323.5 39.0888 96.2409 5,427
Group 2 24.1940 62.8730 14.0544 26.8915 30,981
Group 3 10.6438 24.9345 8.0119 12.9770 259,214
Difference between Group 1 and Group 2 (P-value)
76.0194 (<0.0001)
25.0344 (<0.0001) Difference between Group 2
and Group 3 (P-value)
13.5502 (<0.0001)
6.0425 (<0.0001)
Panel B: trading activities on the back 2.5 years
Group 1 114.9 283.6 53.4819 128.8 7,083
Group 2 36.2805 94.9097 22.6493 43.2797 39,968
Group 3 18.5388 49.6869 14.4867 30.4251 275,835
Difference between Group 1 and Group 2 (P-value)
78.5716 (<0.0001)
30.8325 (<0.0001) Difference between Group 2
and Group 3 (P-value)
17.7417 (<0.0001)
8.1627 (<0.0001)
‧
Robustness checks for domestic institutions. In this table, we check the robustness for
domestic institutions. We divide domestic institutions to two samples by the middle of year. In robustness analysis here, we sample the data by absolute return only. We examine the
difference between group 1-2 and group 2-3 on both the first and the second period by T-test with 5% significant level. Panel A presents the analysis of the first 2.5 years (from January 2004 to June 2006) while panel B shows the remaining 2.5 years (from July 2006 to December 2008).
Trade Size Number of Trades
N Mean St. Dev. Mean St. Dev.
Panel A: trading activities on the fore 2.5 years
Group 1 845.2 845.3 274.2 252.1 57
Group 2 479.8 382.3 222.1 306.0 18
Group 3 421.7 445.5 176.1 190.5 13
Difference between Group 1 and Group 2 (P-value)
365.3 (0.0135)
52.1345 (0.4702) Difference between Group 2
and Group 3 (P-value)
58.1410 (0.6994)
46.0342 (0.6359)
Panel B: trading activities on the back 2.5 years
Group 1 942.5 1178.3 412.7 615.8 116
Group 2 283.6 497.4 104.8 147.9 66
Group 3 423.2 1409.0 203.4 799.9 89
Difference between Group 1 and Group 2 (P-value)
658.9 (<0.0001)
307.9 (<0.0001) Difference between Group 2
and Group 3 (P-value)
-139.7 (0.3887)
-98.5717 (0.2585)
‧
Robustness checks for foreign institutions. In this table, we check the robustness for foreign institutions. We divide foreign institutions to two samples by the middle of year. In robustness analysis here, we sample the data by absolute return only. We examine the difference between group 1-2 and group 2-3 on both the first and the second period by T-test with 5% significant level. Panel A presents the analysis of the first 2.5 years (from January 2004 to June 2006) while panel B shows the remaining 2.5 years (from July 2006 to December 2008).
Trade Size Number of Trades
N Mean St. Dev. Mean St. Dev.
Panel A: trading activities on the fore 2.5 years
Group 1 141.9 195.8 42.1233 46.7663 146
Group 2 65.9457 119.6 26.4706 31.2526 221
Group 3 35.6890 71.5643 16.4147 25.0202 762
Difference between Group 1 and Group 2 (P-value)
75.9721 (<0.0001)
15.6527 (0.0005) Difference between Group 2
and Group 3 (P-value)
30.2567 (0.0004)
10.0559 (<0.0001)
Panel B: trading activities on the back 2.5 years
Group 1 172.8 243.2 74.1912 94.3637 204
Group 2 78.7887 138.2 39.8979 63.4061 284
Group 3 42.4991 84.6936 25.0434 41.5655 1,060
Difference between Group 1 and Group 2 (P-value)
93.9956 (<0.0001)
34.2933 (<0.0001) Difference between Group 2
and Group 3 (P-value)
36.2897 (<0.0001)
14.8545 (0.0002)
‧
Robustness checks for individual investors. In this table, we check the robustness for individual investors. We divide individual investors to two samples by the middle of year. In robustness analysis here, we sample the data by absolute return only. We examine the difference between group 1-2 and group 2-3 on both the first and the second period by T-test with 5% significant level. Panel A presents the analysis of the first 2.5 years (from January 2004 to June 2006) while panel B shows the remaining 2.5 years (from July 2006 to December 2008).
Trade Size Number of Trades
N Mean St. Dev. Mean St. Dev.
Panel A: trading activities on the fore 2.5 years
Group 1 90.9194 306.2 36.4382 90.9365 5,224
Group 2 23.6270 60.5484 13.8433 25.3677 30,742
Group 3 10.5492 24.2681 7.9787 12.7964 258,439
Difference between Group 1 and Group 2 (P-value)
67.2924 (<0.0001)
22.5948 (<0.0001) Difference between Group 2
and Group 3 (P-value)
13.0778 (<0.0001)
5.8646 (<0.0001)
Panel B: trading activities on the back 2.5 years
Group 1 98.9089 216.2 46.6963 91.5219 6,763
Group 2 35.5638 91.8134 22.3889 42.5636 39,618
Group 3 18.3148 41.9530 14.3846 26.5693 274,684
Difference between Group 1 and Group 2 (P-value)
63.3451 (<0.0001)
24.3074 (<0.0001) Difference between Group 2
and Group 3 (P-value)
17.2490 (<0.0001)
8.0042 (0.0002)
‧
Behavioral finance provides another explanations and predictions of people’s decision, which lays emphasis on psychology and social sciences rather than human rationality. One of the substantial theories is hedonic editing, which argues that people would choose to integrate or segregate multiple outcomes to realize the highest
perceived value. However, we know that people would not follow this mental framing rule in some cases, in particular, for losses case. Therefore, quasi-hedonic editing has taken place of hedonic editing hypothesis. We focus on losses case in this thesis. The empirical results show that when facing the equivalent amount of losses, the higher prior loss is, the more willing investors would trade. In contrast, investors are more unwilling to trade due to the smaller prior loss. It is because investors are sensitized by smaller loss and numbed by greater ones previously. Smaller loss makes investors sensitive to the second loss so that they are risk averse and trade less. Comparatively, investors are insensitive to the next loss due to the numbness by greater loss, as a consequence, they make more trades. It is the evidence for the failure of hedonic editing hypothesis. Further, we examine whether this behavioral bias is an explanation for each trader types. We categorize our dataset and test three types of trader
separately. Generally, individual investors exhibit the strongest significant bias on quasi-hedonic editing while domestic institutions and foreign institutions have no obvious significance on this issue. It follows our predictions that professional trading experience and well-disciplined investment strategy are helpful in alleviating
behavioral bias. Compared to other trader types, most individual investors lack training in investment, which is the reason for their irrational trading behaviors. In conclusion, our study proposes an exception for hedonic editing hypothesis and supports quasi-hedonic editing hypothesis.
‧
1. Barber, B. M., and Odean, T., 2000. Trading is hazardous to your wealth: The common stock investment performance of individual investors. Journal of Finance 55, 773-806.
2. Barber, B. M., and Odean, T., 2001. Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics 116, 261-292.
3. Barber, B. M., Lee, Y. T., Liu, Y. J., and Odean, T., 2008. Just how much do individual investors lose by trading. Review of Financial Studies 22, 609-632.
4. Chou, R. K., Wang, Y. Y., and Yan, X, 2013. Trader types and consequences of the disposition effect. Working paper, National Chengchi University.
5. Kahneman, D., and Tversky, A., 1979. Prospect theory: An analysis of decision under risk. Econometrica 47, 263-291.
6. Linville, P. W., and Fischer, G. W., 1991. Preferences for separating or combining events. Journal of Personality and Social Psychology 60(1), 5-23.
7. Locke, P. R., and Mann, S. C., 2005. Professional trader discipline and trade disposition. Journal of Financial Economics 76, 401-444.
8. Lim, S. S., 2006. Do investors integrate losses and segregate gains? Mental accounting and investor trading decisions. Journal of Business 79, 2539-2574.
9. Lehenkari, M., 2009. The hedonic editing hypothesis: evidence from the Finnish stock market. The Journal of Behavioral Finance 10, 9-18.
10. Liu, Y. J., Tsai, C. L., Wang, M. C., and Zhu, N., 2010. Prior consequences and subsequent risk taking: New field evidence from the Taiwan Futures Exchange.
Management Science 56, 606-620.
11. Read, D., Loewenstein, G., and Rabin, M., 1999. Choice Bracketing. Journal of Risk and Uncertainty 19, 171-197.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
33
12. Shiller, R. J., 2003. From efficient markets theory to behavioral finance. Journal of Economic Perspectives 17, 83-104.
13. Thaler, R. H., 1985. Mental accounting and consumer choice. Marketing Science 4, 199-214.
14. Thaler, R. H., and Johnson E. J., 1990. Gambling with the house money and trying to break even: The effects of prior outcomes on risky choice. Management Science 36, 643-660.
15. Thaler, R. H., 1999. Mental accounting matters. Journal of Behavioral Decision Making 12, 183-206.