The Effects of Investor Sentiment on Returns and Idiosyncratic Risk in the Japanese Stock Market

Transcription

1 International Research Journal of Finance and Economics ISSN Issue 6 () EuroJournals Publishing, Inc. The Effects of Investor Sentiment on Returns and Idiosyncratic Risk in the Japanese Stock Market Mei-Chen Lin Associate Professor of Business and Administration National Taipei University, Taipei, 37 Taiwan meclin@mail.ntpu.edu.tw Abstract This paper examined the relationships among investors sentiment, returns and risk in the Japanese equity market. Results showed that subsequent to higher investor sentiment, volatile, small-sized, low-dividend-paying and value firms had higher returns and potential idiosyncratic risk. Low sentiment attenuated these conditional cross-sectional patterns. In comparison to prior findings, similar to those with US samples, volatile, small, and low dividend-paying stocks were more affected by shifts in sentiments. Different from the US market, value stocks rather than growth stocks were affected by sentiment-based demands in the Japanese stock market. When the effects from SMB, HML and MOM are controlled, sentiment did not exhibit predictive power concerning future stock returns. However, in US market, sentiment indeed had significant predictive power for firm-characteristics portfolios. Keywords: Investor sentiment; Risk; Returns JEL Classification Codes: G; G4. Introduction Baker and Wurgler (6) documented that, since mispricing was a result of both an uninformed demand shock and a limit on arbitrage, investor sentiment might affect the cross-section of stock returns when sentiment-based demands or arbitrage constraints vary across stocks. Further theoretical and empirical research showed that, due to their high idiosyncratic risk and unfeasibility to trade, arbitrage was relatively risky and costly for smaller, highly volatile and distressed firms, but with extreme growth potential (Wurgler and Zhuravskaya, ; Amihud and Mendelsohn, 986; D Avolio, ; Geczy, Musto, and Reed, ; Jones and Lamont, ; Duffie, Garleanu, and Pedersen, ). Overall, investor sentiment should have a relatively greater impact on small and highly volatile stocks and firms in distress, with extreme growth potential but without dividends. The above research was based on US data. However, social and cultural differences play an important role in the propensity to engage in speculation and arbitrage. For example, Slovic (996) presented some of the earliest evidence for the social/cultural argument with gender differences in risk taking. Olsen and Cox () documented that cultural factors had non-trivial effects on how assets were managed; investment managers made different decisions based on identifiable cultural differences, even with equivalent training, experience and information. The differences between Asian and Western cultures have been examined in much research. For example, Hofstede (98) showed that in an Asian culture, when a person experiences a disastrous loss, his family or friends will

2 International Research Journal of Finance and Economics - Issue 6 () 3 intervene to support him. In Western cultures, however, a person is expected to sustain the adverse outcomes of his decisions on his own. Since a catastrophic loss affects Asian and Western societies differently, risk attitude and the propensity to speculation and arbitrage may also differ. Similarly, Chui, Titman and Wei (8) suggested that cultural differences might play a role in the relative strength of behavioral biases between countries. The Japanese equity market is the second largest in the world. It behaves differently from the U.S. market, both in terms of the historical returns and the market aggregate idiosyncratic volatility. Therefore, the Japanese market presents an ideal opportunity to study sentiment on cross-sectional effects under different market conditions. Though some studies have discussed the impacts of sentiment on the Japanese stock markets (e.g., Shiller, Kon-Ya and Tsutsui, 996; Brown, et al., 5), to the best of my knowledge, few papers have addressed the cross-sectional differences of sentiments on Japanese stock returns. Furthermore, although Lee et al. () examined the impacts of changes in investor sentiment on the conditional volatilities of the DJIA, S&P5, and NASDAQ indices, no paper has examined whether cross-sectional effects of sentiment on volatility do exist. Intuitively, if noise trading affects prices, the noisy signal is sentiment, the risk noise causes volatility; sentiment is then positively associated with volatility (Brown, 999; Graham and Harvey, 996; Lee, Jiang, and Indro, ). Danthine and Moresi (993) provided similar arguments that both information arrival and noise trading interact to generate price volatility. If a stock attracts more noise trading or less rational traders, its price volatility would be more salient. Along this line of reasoning, stocks that are vulnerable to sentiments are associated with higher volatility; there would be a cross section of price volatility with sentiment. Consequently, aside from a cross-section of stock returns, the paper also examined the relationship between investor sentiments and cross-section of volatilities in the Japanese market. The main results of the paper are, when sentiment was high, volatile, small, low dividendpaying and value stocks tended to earn higher returns in the subsequent month. On the other hand, low sentiment attenuated these conditional cross-sectional patterns. Volatile, small, low dividend-paying, and value stocks had higher future idiosyncratic risk when sentiment was high, but the difference in future idiosyncratic risk across book-to-market stocks was insignificant when sentiment was low. In comparison to the findings, similar to those with US samples, volatile, small, and low dividend-paying stocks are more affected by shifts in sentiments. Surprisingly, value stocks, rather than growth stocks, are more affected by sentiment-based demands in the Japanese stock market. This reflects that the US and Japanese investors behave differently when sentiment drives the speculative demand. This difference can also be seen from the finding that, when the effects from SMB, HML and MOM are controlled, sentiment did not exhibit predictive power concerning future stock returns; however, in the US market, sentiment has significant predictive power for firm-characteristics portfolios. In other words, sentiment has stronger predictive power in US market. The remainder of the paper proceeds as follows. Section shows data resources and the empirical approach, and Section 3 reports the empirical results. Some conclusions are presented in the last section.. Data and Empirical Approach.. Characteristics and Returns The firm returns, risk and security characteristics pertaining to firm size, profitability, dividends, asset tangibility, and growth opportunities and/or distress, were obtained from the Pacific-Basin Capital Markets Research Center (PACAP) database compiled by the University of Rhode Island. The sample includes common stocks listed on the Tokyo Stock Exchange between 975 through 5, but excluding financial companies and stocks with negative earnings or negative book equities. The majority of Japanese firms have March as the end of their fiscal year, and practically all of the companies publish their financial statements within three months after the end of the fiscal year.

3 3 International Research Journal of Finance and Economics - Issue 6 () Accordingly, the portfolios used herein were based upon the fundamental variables known to investors as of the end of June. Size characteristic includes market equity (ME) from June of year t, measured as price times shares outstanding. Std is the standard deviation of monthly returns from July of year t to June of year t +. Profitability characteristics include the return on equity (E+/BE), which is positive for profitable firms and zero for unprofitable firms. Earnings (E) are income before extraordinary items, and E > if earnings are positive; book equity (BE) is shareholders equity. Dividend characteristics include dividends to equity (DIV/BE), which signifies dividends per share times shares outstanding divided by book equity. Asset tangibility characteristics are measured by property, plant and equipment over assets PPE/A. Characteristics indicating growth opportunities, distress or both, include book-to-market equity BE/ME, whose elements are defined above. External finance (EF/A) is the change in assets minus the change in retained earnings divided by assets. Sales growth (GS) is the change in net sales divided by prior-year net sales. All variables are Winsorized at 99.5 and.5 percent... Sentiment Proxy The sentiment variables were measured at monthly frequencies. These variables are based on recent market performance, type of trading activity, and other sentiment proxies. A. Market Performance The first groups are variables based on recent market performance. Among them, one variable is the ratio of the number of advancing issues to declining issues (ADV/DEC). ARMS index is a modification of ADV/DEC, which incorporates trading volumes. This measure is the ratio of the Advt AdvVolt number of advances to declines standardized by their respective volumes: ( ARMSt = ). Dect DecVolt The number of new high prices in relation to new low prices (HI/LO) over the past one year is also designed to reflect the relative strength of the market (Brown and Cliff, 4). B. Type of Trading Activity A second category of variables is related to particular types of trading activity. Turnover ratio (TURN) was computed as the ratio of trading volume to the number of outstanding shares; it measures the market liquidity and can serve as a type of sentiment (Baker and Stein, 4). In a market with shortsales constraints, irrational investors participate and thus add liquidity in a market when investors sentiment is positive and when the price is overvalued. By contrast, when irrational investors are pessimistic, the short-sales constraint keeps them out of the market. Hence, high liquidity tends to coincide with overvaluation (Bris, Goetzmann, and Zhu, 7). This to some degree explains why permanent cross-firm differences in liquidity are associated with permanent cross-firm differences in expected returns. C. Other Sentiment Proxies Following Baker and Stein (4), the share of equity issues in total equity issues and debt issues is used as another measure of financing activity that may reflect sentiment. The equity share (EQU) is defined as gross equity issuance divided by gross equity plus gross long-term debt issuance. Moreover, the IPO market is sensitive to investor enthusiasm sentiment, and high first-day returns on IPOs are often cited as a measure of investor sentiment. As a result, the project takes the number of IPOs A series of papers has examined closed-end fund discounts as a measure of sentiment. For example, Lee et al. (99), Swaminathan (996), and Neal and Wheatley (998) claimed closed-end fund discounts as a proxy for investor sentiment, while Chen et al. (993) and Elton et al. (998) provided evidence to the contrary. Due to the above debate and data availability, this paper does not include closed-end fund discounts as one of sentiment proxies.

4 International Research Journal of Finance and Economics - Issue 6 () 3 (NIPO) and the average first-day returns (RIPO) as two variables of investors sentiment. The last sentiment proxy is the dividend premium, P D-ND, the log difference of the average market-to-book ratios of payers and nonpayers (Brown and Cliff, 4; Baker and Wurgler, 6). 3 Since there are no perfect or uncontroversial proxies for investor sentiment, this paper follows Baker and Wurgler (6) to form a composite sentiment index based on the first principle component of a number of proxies: ADV/DEC, ARMS, HILO, TURN, EQU, NIPO, RIPO and P D-ND. 4 Prior to the principle components analysis, to distinguish between a common sentiment component and a common business-cycle component, each of the eight raw proxies were regressed on five macroeconomic variables proposed by Chen, Roll, and Ross (986). These five include: growth rate in industrial production, unanticipated inflation, the change in chosen growth rate in industrial production, unanticipated inflation, the change in expected inflation, risk premium, and interest rate term structure. The growth rate in industrial production is defined as: ln(ip t /IP t- ), where IP t is the industrial production at time t. Unanticipated inflation is denoted as UI(t)=I(t) E[I(t) t-] where I(t) is the realized monthly first difference in the logarithm of the Consumer Price Index for period t (I t = ln CPI t - ln CPI t- ), E [ I t t- ] = ln E [CPI t ]-ln CPI t-. The series of expected consumption index (E [CPI t]) is obtained by the method of Fama and Gibbons (984). The change in expected inflation (DEI t) is defined as: E [It + t] - E [It t-]. Risk premium (UPR) is denoted as Average Long-Term Interest Rates on Loans and Discounts for All Banks (JAM37) minus year Yield to Maturity of Government Bonds. Interest rate term structure (UTS) is defined as Yield to Maturity, -year Government Bonds minus one-month Gensaki Rate. 5 After the regression, the first principle components of the eight residuals are estimated from these regressions, labeled with a superscript O, and their lags, which produced a first-stage index with sixteen loadings, one for each of the current and lagged proxies. Then, the correlations between the first-stage index and the current and lagged values of each proxy were computed. Finally, Sentiment is defined as the first principle component of the correlation matrix of these current or lagged variables, which had higher correlations with the first-stage index, but with rescaling of the coefficients so that the index has zero mean and unit variance. This procedure leads to a parsimonious sentiment index as follows: Sentiment O O t =.48 ADV DEC +.66 t ARMS O t HILO O t -.94 TURN O t +.47 EQU O t -. RIPO O t NIPO O t D ND, o Pt () Table summarizes and correlates the sentiment measures. It can be found that the maximum sentiment index is 4.88 and the minimum one is -.476, this indicates a asymmetric extreme sentiment index during the sample period. In addition, numbers of initial public issues at previous period (NIPO t- ) had the highest correlation with sentiment index (.954); the second one was equity share (.584) Though price changes in the Tokyo exchange market are subject to price limits, they are very wide for first day trading of new-listed stocks. For example, in the Tokyo stock exchange, the upper limit of the first-day trading is Offering price 4 + price limit (depending on the level of the price), and the lower limit is Offering price /4 - price limit (depending on the level of the price). Therefore, it rarely hits limits on the first day for new listed stocks. Baker and Wurgler (4) used this variable to proxy for relative investor demand for dividend-paying stocks. Due to data availability, this paper did not include discount in closed-end funds as sentiment variable. Because the one-month Gensaki Rate was not available before December 977, the call money rate was used as a proxy from January 975 to November 977.

5 33 International Research Journal of Finance and Economics - Issue 6 () Table : Summary statistics and their correlations for variables forming sentiment index This table summarized the average values (Mean), standard deviation (Std), maximum value (Max), and minimum value (Min) for orthogonalized sentiment variables across the sample period. The first variable is the ratio of the number of advancing issues to declining issues (ADV/DEC). ARMS index is the ratio of the number of advances to declines standardized by their respective volumes. The equity share (EQU) is defined as gross equity issuance divided by gross equity plus gross long-term debt issuance. The number of new high prices to new low prices (HILO) over the past one year is also employed to capture the relative strength of the market. The further sentiment proxy is the dividend premium, P D-ND, the log difference of the average market-to-book ratios of payers and nonpayers. Moreover, the average first-day returns (RIPO) and the number of IPOs (NIPO) are also taken as two of the investors sentiment variables. Turnover ratio (TURN) is computed as the ratio of trading volume to the number of outstanding shares. Country ADV/DEC t- ARMS t HILO t- TURN t EQU t RIPO t NIPO t P D_ND, t Sentment t Panel A: Summary Statistics Mean Std Max Min Panel B: Correlation ADV/DEC O t-. ARMS O t HILO O t TURN O t EQU O t RIPO O t NIPO O t P D_ND,O t Sentiment O t Sorting To determine whether sentiment has cross-sectional effects on stock returns, stocks were first sorted into eight portfolios based on the following characteristics: risk (Std), firm size (ME), positive-earningto-book value (E+/BE), dividend-to-book value (DIV/BE), fixed assets over total assets(ppe/a), bookto-market ratio (BE/ME), the ratio of external financing to asset (EF/TA) and sales growth rate (GS). Then, each of the eight portfolios was further separated into two groups according to the level of sentiment at the end of the previous month. The value-weighted monthly return for each group was calculated and patterns sought. As a result, the cross-sectional effects resulting from the conditional difference of average returns across characteristics- based portfolios can be identified. 3. Empirical Results 3.. Investor Sentiment and Cross-section Returns Baker and Wurgler (6) addressed the theoretical effects of sentiment on a cross-section of returns. They argued that investor sentiment might affect the cross-section of stock prices through two channels: sentiment-based demands and arbitrage constraints. In the first channel, sentiment drives the relative demand for speculative investments, and so causes cross sectional effects even if arbitrage constraints are the same across stocks. The more subjective in their valuations, the more vulnerable the stock is to broad shifts in the propensity to speculate. By contrast, there is much less subjective judgment on the value of a firm with tangible assets and stable dividends; therefore, it is likely to be less affected by fluctuations in the propensity to speculate. This channel suggests that investors demand stocks with some salient characteristics that are compatible with their sentiment. Investors

6 International Research Journal of Finance and Economics - Issue 6 () 34 have low propensities to speculate on safe firms, like profitable, dividend paying stocks. Likewise, the salient characteristics no earnings and no dividends imply that the stocks are speculative. In the second channel, a body of theoretical and empirical research shows that arbitrage tends to be particularly risky and costly for small, unprofitable extreme-growth or distressed stocks (Wurgler and Zhuravskaya, ; Amihud and Mendelsohn, 986; D Avolio, ; Geczy, Musto and Reed, ; Jones and Lamont, ; Duffie, Garleanu and Pedersen, ; Lamont and Thaler, 3; Mitchell, Pulvino and Stafford, ; Brunnermeier and Pedersen, 5). The same stocks that are the hardest to arbitrage also tend to be the most difficult to value (Baker and Wurgler, 6), and are vulnerable to shifts to sentiments. These two channels lead to quite similar predictions and have somewhat overlapping effects. Table presents conditional characteristics effects, and Figure plots the results. It shows the average portfolio returns over months where anthologized sentiment from the previous month end includes the positive, negative, and the difference between the two averages. First, the positive differences on the first Panel STD indicates that, regardless of firms riskiness, stocks had higher average returns when following a positive sentiment index than when following a negative one. Similar patterns were also found when conditioning on other criteria: firm size (ME), positive-earning-to-book value (E+/BE), dividend-to-book value (DIV/BE), fixed assets over total assets (PPE/A), book-tomarket ratio (BE/ME), the ratio of external financing to asset (EF/TA) and sales growth rate (GS). Table : Future returns by sentiment and firm characteristics For each month, ten portfolios were formed according to total risk (STD), firm size (ME), earnings-book ratio for profitable firms (E/BE), dividend-book ratio for dividend payers (DIV/BE), fixed assets over total assets (PPE/A), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth (GS). I then report average portfolio returns over months where SENTIMENT from the previous month end includes: positive, negative, and the difference between the two averages. (small) (large) STD Positive Negative Difference ME Positive Negative Difference E/BE Positive Negative Difference DIV/BE Positive Negative Difference PPE/A Positive Negative Difference BE/ME Positive Negative Difference EF/A Positive Negative Difference GS Positive Negative Difference

7 35 International Research Journal of Finance and Economics - Issue 6 () Figure : Two-way Sorts: Future Returns by Sentiment Index and Firm Characteristics For each month, we form ten portfolios according to standard deviation (STD), firm size (ME), earnings-book ratio for profitable firms (E/BE), dividend-book ratio for payers (DIV/BE), fixed assets (PPE/A), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth (GS). I also calculate portfolio returns for unprofitable, nonpaying, zero PP&E, and zero R&D firms. The dark solid line is returns following positive SENTIMENT periods, and the dark dashed line is returns following negative sentiment periods. The thin line is the difference..5 Panel A: STD.5 Panel B: ME Panel C: E/BE Panel D: DIV/BE Panel E: PPE/A 4 Panel F: BE/ME Panel G: EF/A Panel H: GS In the context of cross section, the STD panel of Table indicates that the cross-sectional effect of return volatility was conditional on sentiment. As expected, high-risk stocks (measured by standard deviation) appeared to come into favor when sentiment was high; they subsequently earned returns of.6% per month over the next month, which was higher than low risk stocks return of.7%.

8 International Research Journal of Finance and Economics - Issue 6 () 36 However, the cross-sectional effect of risk was fully reversed in low sentiment conditions. In particular, the average returns of low risk stocks (.55%) were larger than those of high-risk stocks (.4%). The top STD stocks have greater differences in returns between positive and negative sentiments than do the bottom ones. This confirms Baker and Wurgler s (6), conclusion that highly volatile stocks are especially prone to fluctuations in sentiment because they are relatively hard to value, which deters arbitrage activity. The ME panel revealed that the size effect of Banz (98) appeared in both periods. For example, when sentiment was high (low), returns averaged.99% (.67%) per month for the bottom ME group and.7% (.3%) for the top ME group, with the former higher than the latter. In terms of the difference in returns between positive and negative periods, the bottom ME stock returns are greater than the top ME ones. As shown in Panel B of Fig., the difference in unconditional average returns between bull and bear sentiments across ME deciles vary with ME. This indicates that cross sectional effects of market value are related to sentiment; this is similar to the US where smaller firms are more affected by sentiments. As for profitability and dividend, these characteristics again displayed intriguing patterns. No matter whether sentiments are high or low, monthly returns over the next month are higher on less profitable than is the case for more profitable firms. 6 Likewise, low-dividend paying companies had greater monthly returns regardless of the sentiment. Except for the extreme groups, there are also some differences in average returns across levels of dividend payments and profitability. However, the difference in average returns between bull and bear sentiments across E/BE deciles is dynamic. The fourth and ninth groups top the highest difference and the tenth group reaches the lowest values. The results for difference in average returns between bull and bear sentiments across DIV/BE deciles are clearer. To be more specific, the lowest dividend paying companies have the highest difference in returns between bull and bear market sentiments. These, to some degree, reflect that less profitable and less dividend-paying firms are more vulnerable to sentiment shifts. These findings are also consistent with Baker and Wurgler (6) that unprofitable and nonpaying firms are generally both harder to value and harder to arbitrage. As shown, value (high book-to-market value) firms exhibited higher subsequent returns regardless of sentiment. Similarly, when the other definition of growth, GS, is used, low GS groups have larger cross sectional returns. This is consistent with Lakonishok, Shleifer, and Vishny (994). The cross sectional difference in average returns between high and low sentiments is higher for high BE/ME firms than for low BE/ME firms, indicating that value firms are more sensitive to sentiments. Kumar and Lee (6) provided supporting evidence, namely that shifts in the buy-sell imbalance of retail investors are more positively correlated with returns of value stocks rather than with growth stocks. The remaining variables: asset tangibility characteristics and external finance, also displayed intriguing patterns and some unconditional explanatory power. Future returns are generally higher for low PPE/A stocks and low EF/A stocks. The results were reminiscents of Loughran and Ritter (995) and Spiess and Affleck-Graves (995, 999). The higher return difference of the bottom PPE/A group (.65%) also confirms the view that firms with more intangible assets may be difficult to value and may be overvalued. But, though there are some variations in the return difference across EF/A groups, the return difference between the top EF/A group (.66%) and the bottom EF/A group (.67%) is almost the same. Overall, no matter what the sentiment is at the beginning of the month, subsequent returns are relatively high on high volatile stocks, small stocks, less-dividend-paying stocks, and value stocks. In terms of difference in returns, volatile, small, and low-dividend-paying stocks are vulnerable to shifts in sentiments. 6 The only exception is the second to the highest profitable group after a positive sentiment.

9 37 International Research Journal of Finance and Economics - Issue 6 () 3.. Investor Sentiment and Return Volatility Intuitively, investor sentiment will affect volatility through shocks from speculative demand, and there exist many supporting papers. Brown (999) found that noise traders sentiment was positively associated with stock volatility. Shiller (98) and Leroy and Porter (98) found stock market volatility to be far greater than could be justified by changes in dividends, which was usually labeled as excess volatility of stock prices. The volatility of price would also change over time for such reasons as the rate of arrival of information about the firm, the firm s leverage, and changes of noise trading, etc (Black, 986; Danthine and Moresi, 993). DDSW (99) also concluded that unconditional price variance increased as investor sentiment persisted; then, stocks that were prone to be speculative objects would be more volatile. In the context of arbitrage costs on volatility, Shleifer and Vishny (997) theoretically documented that, for a given noise trading process, volatile securities exhibit greater mispricing and a higher average return to arbitrage in equilibrium. The volatility, in the short run, would also expose arbitrageurs to risk of losses and the need to liquidate the portfolio under pressure from fund investors. This arbitrage cost would keep arbitrageurs from trading and stabilizing these stocks. The greater the arbitrage costs are, the fewer the arbitrageurs trade in, and the larger the volatility of the stock. This leads to cross sectional volatility concerning sentiment. Specialized arbitrageurs are concerned more about idiosyncratic volatility since it cannot be hedged and arbitrageurs are not diversified (Shleifer and Vishny, 997). Therefore, this section examines the effects of sentiments on the cross section of idiosyncratic volatility. The monthly standard deviation through daily returns was first calculated. By running monthly returns against monthly market returns, the monthly idiosyncratic risk was obtained. Then each monthly idiosyncratic risk was placed into a bin according to the decile rank belonging to that characteristic at the beginning of the month, and then according to the level of Sentiment of the previous month. I then computed the average idiosyncratic risk for each decile and looked for patterns. Table 3 reports the conditional characteristic effects of the idiosyncratic risk, and Figure plots the results. An overview shows that there is a higher idiosyncratic risk following a positive sentiment than following a negative sentiment. The STD panel of Table 4 shows the effect of price volatility, as measured by past one-year standard deviations of monthly returns, conditional on sentiment. Stocks with past higher total risk are accompanied by a higher idiosyncratic risk following a positive sentiment, but this phenomenon attenuates in a negative sentiment period. Apart from this, the difference in the idiosyncratic risk is higher for high STD groups. This is consistent with the above finding that noise traders appear to like high volatile stocks when sentiment is high. Higher returns and subsequent higher idiosyncratic volatility then follow. Table 3: STD ME E/BE Future idiosyncratic risk by sentiment and firm characteristics For each month, ten portfolios were formed according to total risk (STD), firm size (ME), earnings-book ratio for profitable firms (E/BE), dividend-book ratio for dividend payers (DIV/BE), fixed assets over total assets (PPE/A), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth (GS). I then report average portfolio returns over months where SENTIMENT from the previous month end includes: positive, negative, and the difference between the two averages. (small) (large) Positive Negative Difference Positive Negative Difference Positive Negative Difference

10 International Research Journal of Finance and Economics - Issue 6 () 38 DIV/BE PPE/A BE/ME EF/A GS Positive Negative Difference Positive Negative Difference Positive Negative Difference Positive Negative Difference Positive Negative Difference Figure : Two-way Sorts: Future Idiosyncratic Risk by Sentiment Index and Firm Characteristics For each month, we form ten portfolios according to standard deviation (STDEV), firm size (ME), earnings-book ratio for profitable firms (E/BE), dividend-book ratio for payers (DIV/BE), fixed assets (PPE/A), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth (GS). I also calculate portfolio returns for unprofitable, nonpaying, zero PP&E, and zero R&D firms. The dark solid line is returns following positive SENTIMENT O periods, and the dark dashed line is returns following negative sentiment periods. The thin line is the difference..5 Panel A: STD 3 Panel B: ME Panel D: DIV/BE Panel C: E/BE Panel E: PPE/A 4 Panel F: BE/ME Panel G: EF/A Panel H: GS

11 39 International Research Journal of Finance and Economics - Issue 6 () The conditional effect of size shown in the ME panel is more prevalent. The average idiosyncratic risk in both high and low sentiment periods decreases with the ME decile. This size effect of idiosyncratic risk exists in both positive and negative sentiment periods, while the magnitude of the latter is less. Thus, not only do small stocks perform better than do large stocks (see Table ); however, small stocks also have more idiosyncratic risk than large stocks. Besides, the difference in idiosyncratic risk reveals that small stocks are more sensitive than large stocks. However, Panel E/BE reveals that there is no consistent difference in idiosyncratic risk for higher and lower profitability firms. Similar patterns are also found for stocks sorted by asset intangibility (see Panel PPE/A). When comparing the very lowest dividend paying one with other groups (see Panel DIV/BE), the lowest dividend paying companies have the most idiosyncratic risk following a positive sentiment period. The difference in idiosyncratic risk is also larger for low dividend-paying stocks. This reveals that the lowest dividend-paying companies are out of noise traders favor. The remaining variables: book-to-market, external finance and sales growth, display the effect of growth opportunity and distress. Likewise, external finance and sales growth did not present any consistent pattern of idiosyncratic risk subsequent to either positive or negative sentiments. Nevertheless, value stocks appear to have higher idiosyncratic risk after a positive sentiment, and this pattern is not strong after a negative sentiment period. Intuitively, growth firms appear to come in favor of noise traders and thereby display higher risk. This confirms De Long, Shleifer, Summers, and Waldmann s (99) study, namely that arbitrage is not necessary to reduce price volatility. In summary, higher sentiment is always followed by higher idiosyncratic risk, whether the sentiment is positive or negative. Future idiosyncratic risk is relatively higher for volatile stocks, small stocks, low dividend-paying stocks, and value stocks. Volatile stocks, small stocks, and low dividendpaying stocks are sensitive to sentiments. Although high BE/ME stocks have higher future idiosyncratic risk, the difference in future idiosyncratic risk across BE/ME stocks is insignificant after a negative sentiment period. There is also no significant and systematic difference in future idiosyncratic risk across portfolios formed on profitability and distress Regression Tests with long-short Portfolios based on Firm Characteristics Investor sentiment may be a worthwhile short-term momentum indicator, and sentiment indicators are often viewed as short-term market timing tools. Then, it is of interest to explore the usefulness of sentiment in regard to forecast portfolios that are long on a stock with high values of a characteristic and short on stocks with low values such as SMB and HML. This provides a robust test regarding whether the effect of investor sentiment on stock returns differs with stock characteristics. The dependent variable is the monthly return on a long-short portfolio, such as SMB, and the monthly returns at time t are regressed on the lagged sentiment proxy. This is useful for explicitly examining whether sentiment affects portfolios like SMB and HML, because they are often used to proxy for systematic risks; if it delivers similar results, then it indicates that the results are not driven by changes in the cross-sectional differences of firm characteristics. The following regression is run: R Xit=High t - R xit=low,t = c + d SENTIMENT it- + ε it, () where R Xit=High,t - R xit=low,t is the monthly returns on a long-short portfolio. A multivariate regression including the well-known Carhart four-factors is also adopted to provide robust results: R Xit=High,t -R xit=low,t =c+dsentiment it- +rrm t +ssmb t +hhml t +mmom t + ε it, (3) where R mt is the excess return of the value-weighted market return over the risk-free rate; SMB is the return on portfolios for small market value minus big market value while separated from returns on HML; HML stands for returns of portfolios for high book to market value, minus low book to market value separated from returns on SMB.

12 International Research Journal of Finance and Economics - Issue 6 () 4 It measures the historic excess returns of small caps and value stocks over the market as a whole. 7 MOM is the return on high-momentum stocks minus the return on low-momentum stocks, where momentum is measured over months [-, -]. 8 SMB and HML are excluded from the regressors when they are put on the left side to be forecasted. Furthermore, extreme growth opportunities effects are separated from distress effects by constructing High, Medium, and Low portfolios based on the top three, middle four, and bottom three decile breakpoints, respectively. The growth and distress variables are broken into high minus medium and medium minus low portfolios. Regarding the GS and EF/A variables, high minus medium portfolios are assigned to the growth opportunity panel, and medium minus low portfolios are assigned to the distress panel. By contrast, in the case of BE/ME variable, medium minus low portfolios are regarded as growth opportunity, and high minus medium portfolios are regarded as distress. Panels A and B of Table 4 show the results without and with controlling for the effect from SMB, HML and MOM factors, respectively; either way, there is no significant relationship between beginning-of-month sentiments and subsequent returns for various types of stocks. One exception occurs in the case of the GS variable under the growth opportunity panel. Specifically, when sentiment is high, subsequent returns on high sales-growth firms are high relative to returns on medium sales-growth firms, when controlling for the effects from SMB, HML and MOM. The results support predictions that, in the Japanese market, sentiment did not provide strong predictions regarding future stock returns, except for stocks that have growth in sales. In comparison with prior findings of Baker and Wurgler (6), in US market, higher sentiment forecasts relatively higher returns on young, low volatile, payers, and profitable firms. Sentiment also has marginal predictive power for the EF/A and GS portfolios, with high sentiment associated with relatively high future returns on medium-ef/a and GS stocks. This shows that the magnitude of the predictive power of sentiment in these two markets is different, and sentiment has stronger predictive power in US market. Table 4: Time Series Regressions of Portfolio Returns This table showed regression results of long-short portfolio returns on lagged Sentiment, the market risk premium, the Fama-French factors (HML and SMB), and a momentum factor (MOM). The long-short portfolios are formed based on firm characteristics (X): total risk (STD), firm size (ME), earnings-book ratio for profitable firms (E+/BE), dividend-book ratio for dividend payers (DIV/BE), fixed assets over total assets (PPE/A), book-to-market ratio (BE/ME), external finance over assets (EF/A), and sales growth (GS). High is defined as a firm in the top quintile; low is defined as a firm in the bottom quintile. The sentiment is the first principle components of the one period lagged monthly return (R m ), the ratio of the number of advancing issues to declining issues (ADV/DEC), the ratio of the number of advances to declines standardized by their respective volumes (ARMS), the number of new high prices to new low prices over the past one year (HI/LO), turnover ratio (turnover ratio), the share of equity issues in total equity issues and debt issues (Equity share), the number of IPOs (NIPO) and the average first-five-day returns (RIPO), dividend premium 7 As described in Fama and French (993), the portfolios constructed at the end of each June are the intersections of two portfolios formed on size (market equity, ME) and three portfolios formed on the ratio of book equity to market equity (BE/ME). The size breakpoint for year t is the median market equity at the end of June of year t. BE/ME for June of year t is the book equity for the last fiscal year end in t- divided by ME for December of t-. The BE/ME breakpoints are the 3th and 7th NYSE percentiles. SMB (Small Minus Big) is the average return on the three small portfolios minus the average return on the three big portfolios, HML (High Minus Low) is the average return on the two value portfolios minus the average return on the two growth portfolios. 8 Following Fama and French, we use six value-weight portfolios formed on size and prior (-) returns to construct MOM. The portfolios, formed monthly, are the intersections of two portfolios formed on size (market equity, ME) and three portfolios formed on prior (-) return. The monthly size breakpoint is the median market equity. The monthly prior (-) return breakpoints are the 3 th and 7 th NYSE percentiles. MOM is the average return on the two high prior return portfolios minus the average return on the two low prior return portfolios.

13 4 International Research Journal of Finance and Economics - Issue 6 () (P D-ND ). Average monthly returns are matched to Sentiment from the previous month. P- values are in braces. STD ME E+/BE DIV/BE PPE/A BE/ME EF/A GS BE/ME EF/A GS BE/ME EF/A GS Profitability and Growth Opportunity and Risk Size Tangibility Growth Opportunity Distress Dividend Policy Distress H-L S-B H-L H-L H-L H-L H-L H-L M-L H-M H-M H-M M-L M-L Panel A: R Xit=High,t - R xit=low,t= c + d SENTIMENT O it-+ ε it Intercept Sentiment R Xit=High,t-R xit=low,t=c+dsentiment O it-+rrm t+ssmb t+hhml t+mmom t+ ε it Intercept Sentiment Rm-Rf SMB HML MOM Conclusion Finding the same sentiment effects working in different markets could strengthen our confidence in the robustness of the sentiment s role. Moreover, the issue of how the dynamics of volatility vary in the cross-section is yet unexplored. Therefore, a different history of stock returns, like securities traded in the Japanese market, was used to provide more insights concerning the effects of sentiment on the stock market. The main empirical finding is that for some characteristics-based stocks, the cross-section of future stock returns is conditional upon the beginning-of-month sentiment. In particular, when sentiment is high, volatile, small, low dividend paying and value stocks tend to earn higher returns. However, low sentiment significantly attenuated these conditional cross-sectional patterns. When controlling the effects from SMB, HML and MOM, sentiment does not provide strong predictive power regarding future stock returns, except stocks that have growth in sales. In addition to the cross sectional effects of sentiment on future stock returns, conditional volatilities also vary in the cross-section. An optimistic sentiment is followed by a higher idiosyncratic risk for volatile, small stocks, low dividend paying and value stocks. Nevertheless, a pessimistic sentiment did not lead to a strong pattern for most firm characteristics. In comparison to prior findings with the US samples, the Japanese investors are prone to hold value stocks rather than growth stocks when sentiment is high. The difference in investors behavior reveals that sentiment drives the different demand for speculative investments across the US and Japanese stock market. However, when controlling for the effect from SMB, HML and MOM factors, in the Japanese market, sentiment did not provide strong predictions regarding future stock returns, except for stocks that have growth in sales. This is different from those in US market that higher sentiment forecasts relatively higher returns on young, low volatile, payers, and profitable firms. Sentiment also has marginal predictive power for portfolios formed based on age, volatility, dividendpaying, profitable, growth and distress. This shows that the magnitude of the predictive power of sentiment in these two markets is different, and sentiment has stronger predictive power in US market. Acknowledgment The author acknowledges financial support from the National Science Council of Taiwan (Grant number NSC96-46-H-39-6).

14 International Research Journal of Finance and Economics - Issue 6 () 4 References [] Amihud, Y. and H. Mendelson (986) Asset Pricing and the Bid-Ask Spread Journal of Financial Economics, Vol. 7, pp [] Antoniou, A. and P. Holmes (995) Futures Trading, Information and Spot Price Volatility: Evidence for the FTSE- Stock Index Futures Contract Using GARCH Journal of Banking & Finance, Vol. 9, pp [3] Baker, M. and J. C. Stein (4) Market Liquidity as a Sentiment Indicator Journal of Financial Markets, Vol. 7, No. 3, pp [4] Baker, M. and J. Wurgler (6) Investor Sentiment and the Cross-Section of Stock Returns Journal of Finance, Vol. 6, No. 4, pp [5] Banz, R. (98) The Relationship between Return and Market Value of Common Stocks Journal of Financial Economics, Vol. 9, pp [6] Black, F. (986) Noise Journal of Finance, Vol. 4, pp [7] Bris, A., W. N. Goetzmann, and N. Zhu (7) Efficiency and the Bear: Short Sales and Markets Around the World Journal of Finance, Vol. 6, No. 3, pp [8] Brorsen, B. W. (99) Futures Trading, Transaction Costs, and Stock Market Volatility Journal of Futures Markets, Vol., pp [9] Brown, G. W. (999) Volatility, Sentiment, and Noise Traders Financial Analysts Journal, Vol. 55, pp [] Brown, G. W. and M. T. Cliff (4) Investor Sentiment and the Near-Term Stock Market Journal of Empirical Finance, Vol., pp. -7. [] Brown, S. J., W. N. Goetzmann, T. Hiraki, and N. W. M. Shiraishi (5) Investor Sentiment in Japanese and U.S. Daily Mutual Fund Flows Working paper. [] Brunnermeier, M. and L. Pedersen (5) Predatory Trading Journal of Finance, Vol. 6, pp [3] Chen, N. F., R. Roll, and R. Ross (986) Economic Forces and the Stock Markets Journal of Business, Vol. 59, pp [4] Chui, A. C. W., T. Sheridan, and K. C. J. Wei (8) Individualism and Momentum Around the World AFA 6 Boston Meetings Paper. [5] Danthine, J. P. and S. Moresi (993) Volatility, Information, and Noise Trading European Economic Review, Vol. 37, pp [6] D Avolio, G. () The Market for Borrowing Stock Journal of Financial Economics, Vol. 66, pp [7] De Long, J. B., A. Shleifer, L. G. Summers, and R. J. Waldmann (99) Noise Trader Risk in Financial Markets Journal of Political Economy, Vol. 98, No. 4, pp [8] Duffie, D., N. Garleanu, and L. H. Pedersen () Securities Lending, Shorting, and Pricing Journal of Financial Economics, Vol. 66, No [9] Fama, E. F. and M. R. Gibbons (984) A Comparison of Inflation Forecast Journal of Monetary Economics, Vol. 3, pp [] Geczy, C. C., D. K. Musto, and A. V. Reed () Stocks are Special too: An Analysis of the Equity Lending Market Journal of Financial Economics, Vol. 66, pp [] Graham, J. and C. Harvey (996) Market Timing Ability and Volatility Implied in Investment Newsletter Asset Allocation Recommendations Journal of Financial Economics, Vol. 4, No. 3, pp [] Harris, L. (989) S&P 5 Cash Stock Price Volatilities Journal of Finance, Vol. 44, pp [3] Hofstede, G. (98) Culture Consequences: International Differences in Work-related Values Sage Publications in London. [4] Jones, C. and O. Lamont () Short Sale Constraints and Stock Returns Journal of Financial Economics, Vol. 66, pp