Empirical Result of the Variation of Risk Taking

5.2 Empirical Result of the Variation of Risk Taking

5.2.1 The Variation of Risk Taking for Overconfident Investors

We present the results that the difference in risk taking between prior and subsequent holding period for overconfident investors in all traders and different types of traders. Overconfident investors are defined as the investors that hold a net long (short) position in the definition period, trade gains from the position, and then have a net buy (sell) volume in the subsequent period. We test the hypothesis from Table 3 to Table 7 using following risk taking proxies: the number of trades, trade sizes, the number of orders, order size, and the ratio of the number of trades to the number of orders. For example, in Panel A of Table 3, overconfident behavior is defined that investors have a net long (short) position, accrue profit in previous 5-day definition period, and then tend to hold a net buy (sell) volume in subsequent 5-day definition period. Those overconfident investors for all traders execute 2.11 numbers of trades in the 20-day risk taking testing period after 5-day definition period more than in the 20-day risk taking testing period before prior 5-day definition period.

Moreover, the results indicate that overconfident investors take more risk than before.

In Panel A of Table 3, all of the t-statistics are significant at the 1% level; thus, we find that the risk taking for overconfident investors in later period is more than prior trade for different definition period and for different risk taking testing period.

In Panel B-E of Table 3, we examine the overconfident investors’ risk taking for foreign institutions, domestic institutions, futures proprietary firms and individual traders, respectively. We find that the results for all four types of traders except futures proprietary firms are as significantly as all traders. Table 4 to 5 present the similar results as in Table 3 that overconfidence affects investors to trade more trade size and number of orders than previous period. However, in Table 6 to 7, we find that the results for order size and trade-to-order ratio are only significantly positive by

‧

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

individuals, less significantly by foreign institutions and domestic institutions.

Consequently, from Table 3 to 7, overconfidence investors execute more trades and orders in subsequent than previous period for foreign institutions, domestic institutions, and individual traders. In the other words, they take more risk than before.

5.2.2 The Variation of Risk Taking for Non-overconfidence Investors

The results of empirical analysis of non-overconfidence investors’ risk-taking are presented from Table 8 to 12. We examine the difference in risk taking between prior and subsequent holding period for non-overconfident investors by all traders and different types of traders. We define overconfident investors as the investors that hold a net long (short) position in the definition period, trade gains from the position, and then have a net buy (sell) volume in the subsequent period. Non-overconfidence investors are defined as all investors except overconfident ones. The risk taking proxies from Table 8 to 12 are used the number of trades, trade sizes, the number of orders, order size, and the ratio of the number of trades to the number of orders, respectively. For example, in Panel A of Table 8, overconfident behavior is defined that investors have a net long (short) position, accrue profit in previous 5-day definition period, and then tend to hold a net buy (sell) volume in subsequent 5-day definition period. Non-overconfident investors are defined that investors do not

‧

is more than prior trade for different definition period and for different risk taking testing period.

In Table 8 and Table 10, we find that the results are almost statistically significant; thus, overconfidence affects investors to execute more number of trades and number of order than previous period. In Table 9 and Table 11, the results show that the statistics of the risk-taking proxies of trade size and order size are significantly positive for all types of traders except domestic institutions. Moreover, Table 12 presents that as the proxy is trade-to-order ratio, the results do not exist any consistency. From Table 8 to 12, non-overconfidence investors take more risk than before as they are foreign institutions, futures proprietary firms, and individual traders, while the statistics are less significantly for domestic institutions.

5.2.3 The Variation of Risk Taking Difference between Overconfident and Non-overconfident Investors

From Table 3 to 12, we could not provide that overconfidence will affect risk taking. Therefore, we examine whether the variation of risk taking difference between overconfident and non-overconfident investors.

In Table 13 to 17, we provide the results by all traders and four types of traders.

For example, in Panel A of Table 13, overconfident behavior is defined that the investors that hold a net long (short) position in the definition period, trade gains from the position, and then have a net buy (sell) volume in the subsequent period.

Non-overconfident investors are defined as investors do not be categorized in overconfident ones. Those overconfident investors for all traders execute 1.75 numbers of trades in the 20-day risk taking testing period after 5-day definition period more than in the 20-day risk taking testing period before 5-day definition period.

However, the non-overconfident investors execute 0.46 numbers of trades in the

‧

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

subsequent more than in risk taking testing period. Therefore, overconfident investors have a tendency to take more risk than non-overconfident ones. Moreover, all of the t-statistics in Panel A of Table 13 are significant at the 1% level, the results show that overconfident behavior affects the number of trades for different holding period and for different test holding period.

In Panel B-E of Table 13, we find that the relationship between overconfident behavior and the variations in number of trades are significantly positive only when the investors are domestic institutions and individual traders. Moreover, from Table 14 to 17, the results are similar to Table 13 as the risk-taking proxies are trade sizes, the number of orders, order size, and trade-to-orders ratio, respectively. According to our results, traders tend to trade more following overconfidence bias. The finding is consistent with our hypothesis overconfident investors take more risk than non-overconfident ones.

‧

The overconfidence hypothesis has been discussed in previous literatures (Barberis et al., 1998; Daniel et al., 1998; Gervais and Odean, 2001; Statman et al., 2006; Glaser and Weber, 2009; O’Connell and Teo, 2009). Gervais and Odean (2001) find that overconfident investors overestimate information from their prior outcome;

thus, they have a tendency to trade more after prior gains. In the paper, we use a dataset consist of account-level orders and transactions records from the TAIFEX to examine whether the behavioral bias is related to overconfidence hypothesis. The previous literatures have showed the relationship between overconfidence effect and trading volume. Therefore, in this paper, we examine whether the variation of risk-taking is following overconfidence effect, and we classify investors through the behavior of investors’ trading history not psychological assessment as prior literatures.

The variation of risk-taking is defined as the difference in the risk taking between the risk taking testing period and prior period. We follow Chou and Wang (2011) overconfidence hypothesis to separate investors. Therefore, we define overconfident investor as investor in definition period has a net long (short) position, trades gains from that position, and has a net buy (sell) volume in subsequent period.

Non-overconfident investors are defined as investors do not be categorized in overconfident ones. Moreover, we classify investors into four difference type:

individual traders, domestic institutional traders, futures proprietary firms and foreign institutional traders. Therefore, we also examine whether the trader types of investors will affect the relationship between overconfidence and behavioral bias.

First, we find that the risk taking is signification positive related to overconfidence effect. Secondly, our finding shows that as the investors are domestic institutions and individual traders, the overconfidence bias will affect investors to increase their risk