1 Introduction
1.2 Objective
From Choe and Eom (2009) and Li et al. (2013) studies, we learn that disposition effect vary across different types of traders in different countries’ futures market. As a result, the first purpose of this paper is to test whether the disposition effect vary across four types of investors (including domestic corporations, domestic individuals, domestic institutions, foreign institutions) in TAIFEX. We expect that domestic individuals would exhibit significant disposition effect.
From Kahneman and Tversky (1979), Shefrin and Statman (1985), and Odean (1998) studies, we conclude risk seeking makes both PGR and PLR decrease, while risk aversion makes both PGR and PLR increase. Besides, from Cao and Wei (2004) and Bassi et al. (2013) studies, we find lower temperature, less rain, and more sunshine make investors exhibit risk seeking, while higher temperature, more rain, and less sunshine make investors exhibit risk aversion. As a result, the second aim of this paper is to test how weather influences the disposition effect of four types of investors through its effect on risk attitude. We expect lower temperature, less rain, and more sunshine make investors exhibit risk seeking. Believe information has not reflected in prices and prices will continue to go up, which lead to both PGR and PLR decrease. Because of the asymmetric risk seeking degree of investors in gains and losses, the decreasing degree of PGR is larger than the decreasing degree of PLR, which leads to disposition effect decrease. On the other hand, higher temperature, more rain, and less sunshine make investors exhibit risk aversion. Believe information has reflected in prices and prices are too high now, leading to both PGR and PLR increase. Because of asymmetric risk aversion degree of investors in gains and losses, the increasing degree of PGR is larger than increasing degree of PLR, leading to disposition effect increase.
3
‧
people’s choices between different prospects, and they found that people’s behavior is inconsistent with expected utility theory. Hence, they developed an alternative model, called prospect theory, to demonstrate differences from expected utility theory, and to truly describe people’s decision making under risk. There are three important findings in prospect theory: First, the prospect theory uses the value function (figure 1), while the expected utility theory uses the utility function. Besides, instead of using simple probabilities in utility function, value function uses decision weights, which are function of probabilities. People care about gains or losses relative to a reference point measured in the value function, rather than the level of wealth measured in the utility function. Second, people exhibit risk aversion in gains, which means the value function is concave in the domain of gains; while exhibiting risk seeking in losses, meaning that the value function is convex in the domain of losses. Prospect theory can explain changes in risk attitude in the domain of gains and losses, however expected utility theory cannot do so. Third, because people feel more severely in losses than in gains of equivalent absolute value, the slope of value function is steeper for losses than for gains.Kahneman and Tversky (1979) also developed the weighting function (figure 2) to describe people’s behavior. People obviously overweight certain outcomes, which is defined as certainty effect, and overweight events with few probabilities. These cause slope of weighting function becoming relatively steep both in near certainty and zero.
The figures of both the value function and weighting function will be presented in next page.
4
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Figure 1 Value Function
Figure 2 Weighting Function
5
‧
Shefrin and Statman (1985) found investors tend to realize winners early and hold losers long, and they defined this phenomenon as disposition effect. They summarized four main elements: prospect theory, mental accounting, regret aversion and pride seeking, and self-control to explain this irrational behavior bias. First, prospect theory describes people care about gains or losses relative to a reference point measured in the value function. The value function is concave in gains, which means investors are risk aversion in gains and tend to sell winners early. Besides, the value function is convex in losses, meaning that investors are risk seeking in losses and tend to hold losers long.
Second, mental accounting describes investors can use tax swap to reduce tax liabilities, which means selling a losing stock and purchasing a substitute stock in order to realize loss for tax purposes while maintaining same risk exposure. However, investors tend to segregate different transactions into separate mental accounts. The tax swap needs to close an account at a loss, so investors are reluctant to do so. Third, regret aversion and pride seeking describes investors avoid realizing losses because they feel regretful to admit their previous judgment was wrong, while seeking gains realization because profits makes them feel pride of having made the decision correctly in the past. Finally, self-control describes investors are reluctant to realize losses but doing so in December in purpose of tax benefits. Investors postpone selling losers until the end of each year when they require themselves doing so before deadline as a measure of self-control.
Odean (1998) analyzed trading records for 10,000 discount brokerage house accounts during 1987-1993. He used the average purchase price as the reference point to determine gains and losses of every transaction. He considered there would be more winners stocks in the upward-moving market, so individual investors would tend to sell
6
‧
calculated proportion of gains realized (PGR) and proportion of losses realized (PLR) to test disposition effect of individual investors. There are three important findings:First, PGR is significantly higher than PLR, which means individual investors have high disposition effect. Second, disposition effect is not prevalent in December because of tax benefit. Individual investors with positive tax rates postpone realizing gains because of tax liabilities generated, while realizing losses in purpose of reducing tax liabilities. Third, Individual investors buy stocks because of favorable information, and sell them at high prices (winners) because they believe information has reflected in price. On the other hand, they may continue to hold stocks if prices go down (losers), believing information has not reflected. That is why individual investors tend to sell winners and hold losers because they believe current losers will outperform (mean revert) current winners in the future. However, research results indicate the average excess returns of winners sold continue to outperform losers held in subsequent months.
Locke and Mann (2000) analyzed transactions data during 1995 from the Chicago Mercantile Exchange (CME). The evidence indicated that professional futures floor traders have disposition effect.
Frino, Johnstone, and Zheng (2004) demonstrated the disposition effect in Sydney Futures Exchange (SFE). They found that after controlling the different trader characteristics, the disposition effect of on-floor professional futures traders (local traders) is stronger than that of the non-local traders. Local traders are more likely to turn paper losses into either paper gains or realized gains than non-local traders in the following transactions because they have privileged short-term information advantage on order flows, which allows them to accurately predict short-term price shifts.
7
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
Choe and Eom (2009) analyzed the Korean stock index futures trading database from January 2003 to March 2005 and found that the disposition effect of individual investors is stronger than that of the institutional and foreign investors.
Li et al. (2013) examined the disposition effect in Taiwan futures market. The result indicated that disposition effect of retail investors and foreign institutional investors are obvious, while that of proprietary investors is not significant. The phenomenon of mean reversion exists in the disposition effect of retail investors.
Chen, Kim, Nofsinger, and Rui (2007) analyzed the disposition effect in China stock market. They found that both individual investors and institutional investors exhibit significant disposition effect. Besides, individual investors exhibit stronger disposition effect than institutional investors. They also compared China stock market with U.S. stock market, and found that Chinese investors exhibit stronger disposition effect than U.S. investors.
Brown, Chappel, Rosa, and Walter (2006) investigated Australian Stock Exchange data for investors in IPO and index stocks between 1995 and 2000. The results indicated that the disposition effect exists across investor classes. Especially, disposition effect of individual investors is stronger than that of institutional investors, incorporated companies investors, and foreign investors.
From previous studies, we learn that disposition effect vary across different types of traders in different countries’ stock and futures markets. One thing is common in both stock and futures markets is that individual investors exhibit significant disposition effect in both financial markets.
8
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
2.3 Weather
Many researches indicate that weather conditions influence people’ mood, feelings, emotions, and risk attitude, which then affect people’ behavior and decision-making.
Mellers and McGraw (2001) found that greater anticipated good weather make people feel more optimism and risk seeking, whereas less anticipated good weather make people feel more pessimism and risk aversion. Cao and Wei (2004) found that lower temperature makes investors feel risk seeking, while higher temperature makes investors feel risk aversion. They further examined the relationship between temperature and the stock returns within eight international stock markets (including Taiwan), and found there existed negative correlation between temperature and returns.
Lower temperature is relative to higher stock returns and higher temperature is relative to lower stock returns. Bassi et al. (2013) implemented a lottery choice experiment to observe risk preference of 262 participants from March 2011 to September 2012. They employed a multiple price list method, and found that good weather and sunshine are associated with upbeat mood, which makes people feel risk seeking. Hirshleifer and Shumway (2003) examined the relationship between sunshine and stock market index returns across 26 countries (including Taiwan) from 1982 to 1997. The result indicated morning sunshine in the city of a country's leading stock exchange is significantly correlated with stock returns.
9
‧
We analyzed the trading data of Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) futures contracts traded on Taiwan Futures Exchange (TAIFEX) from January 2007 to December 2007. The underlying asset of TAIEX futures is TAIEX, which is value weighted index of all common stocks listed on Taiwan Stock Exchange (TWSE). The contract value of TAIEX futures is index points of TAIEX multiplied by 200 Taiwan Dollars (TWD). The trading time of TAIFEX is from 8:45 a.m. to 1:45 p.m.
Taiwan time Monday through Friday of regular business days of TWSE. The daily price limit was +/-7% of the previous trading day’s settlement price in 2007. The delivery months include the spot month, next calendar month, and next three quarterly months.
The last trading day is the third Wednesday of the delivery month of each contract. All futures contracts are automatically exercised on the expiration date and are settled with cash. Each investor should open a margin account and save the initial margin at the broker. When the margin is below the maintenance margin, the margin account should be refilled back to the initial margin level, otherwise the margin account will be closed automatically by TAIFEX. The database offers the detailed trading information. We chose the information of the trading date and time, trading price, trading quantity, trading direction (buy or sell), trader types, and trading account to do further analyses.
We chose the domestic corporations, domestic individuals, domestic institutions, and foreign institutions as our four types of investors, and the summarized information is presented in the Table 1. From Table 1, we can see that the total trading frequencies among four types of investors during 2007 were 11.7 million times, and the domestic individuals represent the most percentage (74.4%) in TAIFEX.
10
‧
studies, we chose temperature, rain, and sunshine as our weather variables to do further analyses. Because the TAIFEX is located in Taipei, we collected Taipei weather data of monthly averaged temperature, monthly accumulated rain, and monthly accumulated sunshine from the Taiwan Central Weather Bureau (TCWB) during 2007. Table 2 presents the summarized monthly weather information.Table 1 Trading frequency and percentage of four types of investors Types of Investors Trading Frequency Trading Percentage
domestic corporations 72,065 0.62%
domestic individuals 8,710,187 74.40%
domestic institutions 2,031,325 17.35%
foreign institutions 893,328 7.63%
Sum 11,706,905 100%
Table 2 Summary statistics of weather variables
11
Month Temperature(℃) Rain(mm) Sunshine(hr)
2007/01 17.3 111.4 94.7
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
3.2 Research Methodology
This dataset records the trades by second. We tracked every investor’s purchase quantity and average purchase price as well as selling quantity and average selling price of each trade. We defined each investor as holding a net long (short) position according to the first trading direction (buy or sell) during the sample period until the investor closes the position. We calculated the average cost of every transaction by volume weighted average price for each investor according to every trading direction. The average cost of the net long position is the average purchase price, while that of the net short position is the average selling price. At any time, an investor’s position is either net long or net short or no position (closes the position). When the investor of a net long position continues to buy or the investor of a net short position continues to sell, the average cost is recalculated. On the other hand, when the investor of a net long position sells or the investor of a net short position buys, the average cost remains the same.
If the investor has the position, there must be papers gains or paper losses. We then compare the trading price to the average cost to decide whether the position is traded at gains or at losses. When the trading price is higher (lower) than the average cost of the net long (short) position and the investor does not close the position, the position is counted as paper gains. In the meantime, the investor can close the position, and turn the paper gains into the realized gains. In contrast, when the trading price is lower (higher) than the average cost of the net long (short) position and the investor does not close the position, the position is counted as paper losses. If the investor closes the position, the paper losses will turn into the realized losses.
12
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
We employed the methodology developed by Odean (1998) to measure every investor’s disposition effect of each transaction. We first calculated every investor’s paper gains, paper losses, realized gains, and realized losses of each transaction in order to calculate the PGR and PLR. Then, we used PGR minus PLR to measure the disposition effect (DE) of each investor.
Proportion of gains realized (PGR) = Realized Gains
Realized Gains + Paper Gains (1)
Proportion of losses realized (PLR) = Realized Losses
Realized Losses + Paper Losses (2)
Because weather information is the monthly data (presented in the Table 2), we separately sum every investor’s PGR, PLR, and disposition effect (DE) of each transaction monthly according to four types of investors (including domestic corporations, domestic individuals, domestic institutions, and foreign institutions).
Table 3 presents the information of monthly PGR, PLR, and DE of four types of investors during 2007. From Table 3, we can see that domestic individuals have the highest disposition effect, which is consistent with the findings from Choe and Eom (2009) and Li et al. (2013).
13
‧
Table 3 Summary statistics of disposition effect (PGR-PLR) domestic
To the best of our knowledge, there have been very few researches to investigate the effect of weather on disposition effect. Besides, some studies examined disposition effect vary across different types of investors in different countries’ stock and futures markets (Locke and Mann, 2000; Frino et al., 2004; Brown et al., 2006; Chen et al., 2007; Choe and Eom, 2009 and Li et al., 2013). Others examined disposition effect, overconfidence, and representativeness bias (Chen et al., 2007), disposition effect and underreaction to news (Frazzini, 2006), disposition effect and momentum (Grinblatt and Hany, 2001). To the best of our knowledge, there have been very few researches to investigate the control variables of disposition effect. As a result, we only use weather variable as our independent variable to examine the effect of weather on disposition effect of four types of investors.
14
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
3.3 Hypotheses Development
According to the literature presented previously, we derive the following two hypotheses.
Hypothesis 1: Disposition effect may vary across four types of investors, and the domestic individuals would exhibit significant disposition effect.
From previous studies, we learn that the disposition effect exists among different types of investors in different countries’ futures market. Choe and Eom (2009) indicated that the disposition effect of individual investors is stronger than that of the institutional and foreign investors in the Korean stock index futures market. On the other hand, Li et al. (2013) found that the disposition effect of retail investors and foreign institutional investors are significant, while that of the proprietary investors is not obvious in Taiwan futures market. As a result, we extend the concept derived from the previous studies and consider that disposition effect may vary across four types of investors (including domestic corporations, domestic individuals, domestic institutions, and foreign institutions) in TAIFEX, and the domestic individuals would exhibit significant disposition effect.
15
‧
Hypothesis 2: Weather may influence the disposition effect of four types of investors through its effect on risk preference, which is measured by PGR and PLR. The lower temperature, less rain, and more sunshine make four types of investors exhibit risk seeking, which leads to PGR, PLR, and disposition effect decrease. On the other hand, higher temperature, more rain, and less sunshine make four types of investors exhibit risk aversion, leading to PGR, PLR, and disposition effect increase.
Kahneman and Tversky (1979), Shefrin and Statman (1985), and Odean (1998) proposed that disposition effect indicates investors exhibit risk aversion in gains and believe information has reflected in prices, which lead to sell winners early and proportion of gains realized (PGR) is higher. On the other hand, investors exhibit risk seeking in losses and believe information has not reflected in prices, which lead to hold losers long and proportion of losses realized (PLR) is lower. When PGR is significantly higher than PLR, which means individual investors have high disposition effect (DE).
As a result, we extend the concept derived from previous studies and consider that in gains, investors who exhibit risk seeking think information has not reflected in prices, so prices will continue to go up. They tend to hold on gains, which makes PGR decreases. In losses, investors who exhibit risk seeking also think information has not already reflected in prices, so prices will go up in the future. They tend to hold on losses, and believe losers will mean revert someday, which makes PLR decreases. Because of the asymmetric risk seeking degree of investors in gains and losses, the decreasing degree of PGR is larger than the decreasing degree of PLR, which leads to disposition effect decrease. On the other hand, in gains, investors who exhibit risk aversion think information has already reflected in prices, so prices are too high now. They tend to realize gains, which makes PGR increases. In losses, investors who exhibit risk aversion
16
‧
tend to realize losses in order to limit losses, which makes PLR increases. Because of the asymmetric risk aversion degree of investors in gains and losses, the increasing degree of PGR is larger than the increasing degree of PLR, leading to disposition effect increase. In summary, risk seeking makes PGR, PLR, and DE decrease, while risk aversion makes PGR, PLR, and DE increase.
Many researches indicate that weather conditions influence investors’ mood and
Many researches indicate that weather conditions influence investors’ mood and