where TTPij represents the trading preferences of investor j towards stock i, and G denotes one investor group, such as male, female, and so on.
We will run the regression by different groups to examine whether investors in one specific group will show preferences for certain stock characteristics. Each group in the study will be implemented the regression respectively. Hence, we can compare the preference differences between various types of investors.
6. Data Mining
(1) Self Organizing Map
In our thesis, we apply the tool of SPSS Clementine 7.0, which is used for data mining solutions to run the procedures of SOM. To find the appropriate number of seeds and better performance of SOM, we try two combinations of seeds in the network. We set the network of SOM as 3×3 and 4×4 respectively, and include all the variables of stock characteristics
and investor dimensions in the procedure. For the network of 3×3 and 4×4, it can produce 9 and 16 clusters at most, but the numbers of total clusters it generate finally are decided by the network. The data we used in the process are the 6,993,052 trading records for 501 firms in our sample.
The investor dimensions included are gender, wealth levels, regions, trading frequency, and trading experience for the institutional investors and individual investors. As to the nine stock characteristics, we divide each variable into three ranks: low, middle, and high. It is difficult to interpret the result if the stock variables contain their own values. By ranking the values of stock characteristics, we can focus and condense the trading preferences of investors over stock characteristics. After the procedure of SOM, it will present the result of cluster distributions. Then, we compare the distribution to examine whether there are proportionately large clusters, which contain the most trade records in the sample. The greater the differences of the proportion between each cluster, the better the performance of the network because it means that there are apparent diversity between the features in the sample. Thus, according to the results, we will select the network with better performance to analyze the trading preferences of investors. Then, we apply the decision tree to further find the rule of our sample.
(2) The Decision Tree
After selecting the network of SOM, we extract the data of the main clusters to do the analyses. The sample distributed in the same cluster own similar features. If the number of sample in one cluster is large, it means the preferences, which inferred from the result are significant. Therefore, we focus the analysis on the primary clusters. We apply SPSS Clementine 7.0 as the tool to run the procedure of the decision tree. Then, we find out the rules of each level in the tree, and select the most significant in each cluster. Hence, we can examine whether there are trading preferences for stock characteristics among the investors. The results by using regression and data mining will be compared to examine whether there are similarities and differences between these two results.
7. The Diagram of Process
The process of our research methodology is as follows:
Original
Data Data
Set
Regression
SOM
Decision Tree
Result 1
Result 2 filtering
comparing
Empirical Results and Analysis from Statistical Regressions 1. Trading Volume vs. Investor Characteristics
Table XI shows the average annual trading amount of the investor accounts under each dimension in our sample. In addition, we calculate the average annual value of total trading amount divided by wealth of the investor accounts. It is easier for wealthier investors to have greater total trading volume. Therefore, we set up another measurement of dividing the total trading amount by the wealth of the investor. This ratio indicates the relative trading amount, and it is more meaningful than the total amount.
The larger the value of trading amount and trading amount divided by wealth, the greater the willingness of the investors to make stock investments. Nevertheless, for the relative trading amount, it must be interpreted in comparison with other groups, instead of looking at only the absolute value. In order to confirm whether the investor dimensions have the potential influences on stock trading, we also apply the UNIANOVA to do the examination. From Table XI, we analyze whether various types of investors will present significant differences on the trading amount of stocks and the ratio of the amount divided by their wealth.
For institutional and individual investors, the annual trading volume is much greater for institutional investors, because they usually have large amount of property. As for the relative trading amount, the value may be overestimated for institutional investors. We use the largest trading amount of investors’ records as the proxy for their wealth. However, it may be much wealthier for institutional investors because this proxy is not very objective for them. For male and female investors, it shows that male investors trade more than female investors. Besides, it indicates that in the dimension of wealth levels, wealthier investors significantly make more stock trading than other investors. However, the same with institutional investors, it is not obvious to measure the wealth level of the investors in group 5, and the value of their wealth may be underestimated. Thus, their relative trading amount (40.44) will be overestimated. We also find that investors in the North region trade more than those in other regions.
Nevertheless, investors in other regions like to make stock investments more than the North investors, because they have larger relative trading amount. Moreover, investors with the greatest trading frequency trade
significantly much more than investors in other groups. There are no significant differences between total trading amounts among investors with different trading experiences, but more experienced investors are more likely to make stock trading.
< Table XI Inserted Here>
2. Trading Volume vs. Stock Characteristics
Table XII shows the average annual trading amount of stocks in relation to the stock characteristics in our thesis. As to the stock characteristics, we partition the stocks into three groups (low, middle, and high) for each stock variable. In this way, we can find what kind of stock characteristics will induce investors to make more trading. However, this value is not very objective because stocks with high prices tend to obtain greater total trading volumes. We establish another measurement by calculating the relative amount of the average annual total trading amount divided by the size of the stock. If the stock characteristic has larger relative trading amount, it means that the stock characteristic are more attractive to investors.
It indicates that stocks with high beta are more likely to be traded by investors. Therefore, investors may have greater preferences for stocks with high beta than low beta. Stocks with high prices have the largest average annual trading volume, but the least relative trading amount. Thus, stocks with lower prices are favorable for investors. Besides, stocks with higher rate of return are traded more than those with low rate of return because of better past performance. Stocks with greater sizes are also more attractive to investors.
< Table XII Inserted Here>
3. Stock Preferences of Various Investors
We investigate the stock preferences of investors to analyze whether the stock characteristics will have impacts on the stock trading behavior of various types of investors. First, we examine the preferences of investors for diverse stock characteristics. From Table XIII, we find that in general, investors tend to choose the stocks with shorter listed periods, higher beta, lower dividend yield, higher EPS, higher market-to-book, lower prices,
lower P/E ratios, and larger sizes. In the following, we will analyze their preferences in various investor dimensions, and try to figure out the possible reasons.
< Table XIII Inserted Here>
(1) Institutional Investors vs. Individual Investors
Based on the results of Table XIV, we find that both institutional and individual investors show favor to more “youthful” stocks, which indicates that the stock has shorter listed periods. For stocks that are just listed in short periods, it is easier for them to have price fluctuation in a wide range.
Therefore, investors can obtain greater gains and losses in this period.
Besides, they like to trade stocks with larger sizes. In general, firms with large sizes are more famous and less risky for investors. Therefore, it is reasonable for investors to choose stocks with greater market capitalization.
Besides, both institutional and individual investors prefer stocks with lower prices, but institutional investors are less sensitive. Lower prices let investors with less wealth also capable of making stock investments because they are not able to afford the high trading amount of the stocks if the stock prices are too high. Moreover, it is more difficult for investors to sell stocks with high prices when needed because of the large trading amount. In the dimension of institutional investors, the result does not show many significant preferences for stocks characteristics. A possible reason is that the numbers of institutional investors and their trade records in our sample only account for small percentage. Most of investors in the sample are individual investors.
As for individual investors, we also find that they have the preferences for stocks with higher beta, which indicates that the prices of the stocks will be more volatile than the market. High volatility of the stocks can cause more gains or losses. Thus, investors in Taiwan may have the tendency towards risky investments. Furthermore, it shows that individual investors prefer to trade stocks with lower dividend yield. Stocks with larger dividend yield usually have lower price volatility, and investors mainly earn the dividend revenues of the investments. On the other hand, although lower dividend yield incurs fewer dividends, the stocks usually bear higher price volatility. Hence, the investors primarily gain from the price differences of the stocks. It also means that investors tend to invest in
more risky stocks because they will have high opportunities to cause larger gains. The other possible reason for choosing stocks with lower dividend yield is that the dividend revenues have to be taxed in Taiwan. However, gains from the stock exchange are tax exempt. Hence, investors tend to earn money from the trading of stocks instead of from the receiving of dividend revenues. In addition, individual investors are more likely to invest in stocks with higher EPS because it is easier for stocks with high EPS to come into notice. Individual investors also show favor to stocks having larger market-to-book. Stocks with higher market-to-book are
“growth stocks” because they have larger market values than their book values. Hence, it indicates that the investments will be more risky. On the other hand, if the stocks bear lower market-to-book, it means that the market values match their book values. They are called “value stocks.”
Thus, in conformity with the results of beta and dividend yield, we can find that investors are more likely to regard stock trading as risky investments.
In addition, individual investors show preferences for stocks with lower P/E ratios. In general, a higher P/E ratio suggests that investors are expecting higher earnings growth in the future compared to companies with a lower P/E ratio. For investors, the price of a stock represents the cost of investing in one share of the stock, and EPS indicates the profit of investing in one share of the stock. Therefore, a P/E ratio is the reciprocal of the rate of return of the investments. Thus, a lower P/E ratio usually means that the potential return of the stock investments is larger.
< Table XIV Inserted Here>
(2) Gender
Table XV shows that both male investors and female investors tend to invest in stocks with shorter listed periods, higher beta, lower dividend yield, higher EPS, higher M/B, lower P/E ratios, and larger sizes. Thus, both male and female investors like to trade stocks which may cause larger price fluctuation because it is possible for them to get greater gains. They also prefer the stocks with high public attention. The preferences among male and female investors in Taiwan do not differentiate a lot. However, as for the rate of return of stocks, male investors have the tendency to trade stocks with lower rate of return, but female investors like those with larger rate of return. Stocks with larger past rate of return represent that they have better past performance and usually own stable and higher stock prices. On
the other hand, stocks bearing lower past rate of return often have lower stock prices, but they possess more opportunities to get a big price rise.
Besides, although both male and female investors prefer the stocks with lower prices, female investors are more sensitive than male investors.
Therefore, in conformity with the existing studies (Dwyer et al. (2002)), we can infer that male investors are more adventurous, overconfident and more likely to gamble on the stock exchange. On the contrary, female investors are relatively more conservative and seek for reliable source of making money.
< Table XV Inserted Here>
(3) Wealth Level
We decompose investors into five different wealth groups and implement the regression analysis for each group. From Table XVI, we find that individual investors with different wealth levels all have the tendency to trade stocks with shorter listed periods, higher beta, lower dividend yield, lower P/E ratios, and larger sizes. Stocks with high EPS are more likely to attract the attention of the investors; therefore, most of the investors with different wealth levels prefer the stocks having high EPS. It shows that from group 1 to group 4, the sensitivity for beta is decreasing, but it reverses to rise up for group 5. Thus, investors that are less wealthy are more likely to make risky stock investments. Besides, we find that the preferences of less wealthy investors are significantly positive with market-to-book. Nevertheless, investors with higher wealth levels appear to prefer stocks with lower market-to book. Again, it shows that less-wealthy investors tend to be more adventurous. The possible reason is that they want to make a fortune from the stock trading. However, high market-to-book will create more risk for investors; therefore, investors with less wealth may tend to trade the stocks in short period instead of holding them in longer periods. On the contrary, wealthier investors prefer the “value stocks” and they exhibit the tendency towards investing in the stocks in longer periods. Moreover, investors in different wealth levels all show preferences for stocks having lower P/E ratios. It means that investors like stocks which the prices correspond with their earnings. They are unwilling to spend too many costs for stocks. Stocks with greater sizes are less risky for investors because the firm is large-sized. Therefore, all of the investors like the stocks having large sizes, but wealthier investors are less sensitive
because they have more property.
< Table XVI Inserted Here>
(4) Region
We partition the investors into four groups based on the region they live, and then run the regression for each group. From Table XVII, the result shows that the trading preferences among investors from different regions in Taiwan are almost the same. They are more likely to invest in stocks with shorter listed periods, higher beta, lower dividend yield, higher EPS, lower P/E ratios, and larger sizes. Nevertheless, as for return, investors of the North region have significant preferences for stocks with lower rate of return, which is opposite to investors in other regions.
Besides, the North investors also exhibit favor to stocks with high market-to-book, and they are very sensitive to the stocks with shorter listed periods in comparison with other investors. Hence, investors lived in the North region of Taiwan are more likely to take the risk and seek for the opportunities to obtain larger amount of money.
< Table XVII Inserted Here>
(5) Trading Frequency
On the basis of trading frequency, investors in our sample are divided into five groups. Group 1 is investors with least trading frequency, and group 5 is those with greatest frequency. Table XVIII indicates that investors with different trading frequencies all show favor to stocks with higher beta, higher EPS, higher return, and larger sizes. Investors in group 1 to group 4 all prefer stocks with shorter listed periods and lower dividend yield, but the sensitivity is decreasing if the investors become making the stock trading more frequent. Similarly, with the trading frequency increasing, investors become less sensitive to the beta of stocks. Moreover, investors in group 1 and group 2 significantly show preferences for stocks having higher market-to-book. Thus, investors with less trading frequency have the tendency to be gamble on the stock trading because they may want to try their luck to make gains. Furthermore, investors with different trading frequencies all like the stocks with better past rate of return, and investors making less trading put even more emphases on the past performance of the firms. Besides, investors with less trading frequency are
more sensitive to the sizes of the stocks.
< Table XVIII Inserted Here>
(6) Trading Experience
We divide the investors into three groups by using the years of opening accounts as the proxy for trading experiences. Investors in group 1 are the least experienced, and those in group 3 are the most experienced.
The result of Table XIX shows that investors with different trading experiences all exhibit preferences towards stocks with shorter listed periods, higher beta, lower dividend yield, higher EPS, lower prices, lower P/E ratios, and larger sizes. However, the most experienced investors are less sensitive to the listed periods of stocks relative to other investors.
Besides, with increasing experiences, the sensitivity to dividend yield declines. Investors in group 1 and group 2 tend to choose stocks with higher market-to-book, and the least experienced investors put even greater significance. On the other hand, the most experienced investors appear to like stocks with lower market-to-book, which are less risky and belong to
“value stocks.” As for return, investors that are most experienced like stocks with higher rate of return; however, the investors that are less experienced prefer those with lower rate of return. Therefore, the results show that investors that are less experienced tend to be more adventurous and like risky investments.
< Table XIX Inserted Here>
4. Summary
Table XX summarizes the results of the trading preferences among various investors for stock characteristics. In general, it shows that
Table XX summarizes the results of the trading preferences among various investors for stock characteristics. In general, it shows that