MSc Finance Master Thesis
Feedback Mechanism in the case of growth funds:
Evidence from past returns and sentimental fund flows.
University of Groningen
Faculty of Economics and Business
Abstract
Limited scholarly attention has been given to the impact of investor sentiment on fund flows. In order to extend the theoretical knowledge on sentiment sensitivities of funds, we measure them directly on the basis of fund flows. We cite different investor bases and behaviors as the reason for varying sentiment sensitivities among funds. Based on the sensitivities, we categorize funds as chasing or contrarian. Our hypotheses suggest that sentiment-chasing fund flows channel negative effects on fund returns. Our analyses show that fund flows affect fund performances and that variations in these flows can be explained by investor sentiment. However, these results do not necessarily imply that fund flows channel negative sentiment effects on fund returns. We find only negligible performance differences between sentiment-chasing and sentiment-contrarian funds. Sentiment-contrarian investors increase their own returns through timing skills. We contribute to the literature with a sensitivity analysis of fund flows by explaining their variations through investor sentiment and fund specific past returns. We find past return-chasing fund flows in U.S. mutual growth funds. These flows hamper fund performances, indicating a negative flow-performance relation and feedback mechanism.
Author: Tim-Fabian Ecklebe Student Number: S4124332 Supervisor: Dr. Auke Plantinga
Keywords: flow-performance relation, return predictability, investor sentiment JEL Classification: G11; G12; G17; G41
1. Introduction
During the 1970s the efficient markets theory supported by Fama and others reached its height of its dominance. Commonly, efficient markets are defined as markets in which security prices fully reflect all relevant and available information about the fundamental value of the securities. Motivated by discovered anomalies suggesting that changes in security prices occur for no fundamental reason, an academic discussion emerged (Shiller, 2003). A lot of the discussion shifted towards developing models of human psychology to explain irrational investor behavior: Kahneman and Tversky (1979, 1984) have established the prospect theory and the effect of loss aversion, Thaler (1980) has described the phenomenon of mental accounting in different buckets, Shefrin and Statman (1985) have discovered the disposition effect whereas Fischhoff et al. (1977) and Odean (1998) have examined overconfidence. The field of Behavioral Finance evolved with Shiller being one of the prominent researchers contradicting the efficient markets theory. He defines Behavioral Finance as ‘finance from a broader social science perspective including psychology and sociology’ (Shiller, 2003, p.83).
Recent strands in Behavioral Finance literature discover that stock prices are affected by sentiment. Investors tend to be influenced from their state of mind, or sentiment when deciding about investment decisions (Lucey and Dowling, 2005). Baker and Wurgler (2006) define investor sentiment as optimism or pessimism of an investor about future stock market activity and investment risks that are not founded on facts. They claim sentiment to create cross-sectional variations in returns across stocks. Massa and Yadav (2015) find that low-sentiment-beta funds significantly outperform their high-sentiment-low-sentiment-beta counterparts.
Furthermore, Behavioral Finance literature investigates feedback mechanisms such as feedback loops between fund flows and fund returns. Peng and Wang (2019) find that positive feedback traders stimulate initial momentum and following reversals, leading to advantages for first-movers. Zhong et al. (2018) discover self-reinforcing feedback loops caused by emerging herd behaviors. In the field of flow-performance relations, two controversial opinions emerged. The first opinion claims investors to be rational whereas the second opinion posits that investors are subject to behavioral biases and sentiment swings (Bailey et al. (2011), Wilcox (2003), Capon et al. (1996)). Frazzini and Lamont (2008) describe fund flows as “dumb-money”, where retail investors reduce their wealth in the long run.
investor flows. By measuring sentiment sensitivities on stock level and aggregating them on fund level, they build a proxy for the degree of a fund’s sentiment beta (FSB).
To expand the theoretical knowledge of sentiment sensitivities of funds, we want to measure them directly by means of fund flows. So far, we know little about the impact of investor sentiment on fund flows and whether these sentimental flows impact subsequent returns. There is reason to believe that sentimental flows impact fund performances (Jiang and Yüksel, 2019). Motivated by Massa and Yadav (2015), we want to find evidence for an adverse effect of sentiment sensitivities on fund returns. We claim fund flows to channel these negative sentiment impacts.
Whereas Massa and Yadav (2015) claim portfolio strategies as responsible for fund’s sentiment betas, we expect different investor bases and subsequent behavior as a reason for varying sentiment sensitivities among funds. It can be thought of Momentum investors who expand their investments in case of a positive sentiment and reduce their investments if a negative sentiment is prevailing. In contrast to that it can also be thought of Contrarian investors who act in the opposite way. In the end, these behaviors lead to chasing or sentiment-contrarian fund flows. We hypothesize funds experiencing sentiment-sentiment-contrarian fund flows to outperform their sentiment-chasing counterparts.
This study consists of four sections. First, the literature on flow-performance relations, feedback mechanism and investor sentiment are reviewed, and we present our theoretical framework, and the research questions this study wants to investigate on. Second, the data and employed methods will be explained before third, we present and discuss the results. Finally, we conclude the main findings, point out limitations and discuss recommendations for future research.
2. Literature Review and Hypotheses
2.1. Flow-Performance Relation
different mutual funds. Teo and Woo (2004) also find evidence in favor of the “dumb-money” effect. Concludingly, it needs to be distinguished between the ability of investors to identify skilled fund managers and trend chasing from an investor sentiment point of view.
The most recent study by Jiang and Yüksel (2019) concentrates on the second opinion as explanation for empirical patterns of mutual fund flows. They investigate the tendency of investors to chase past fund performance and the impacts of fund expenses and sheer visibility while controlling for high and low sentiment periods. They test the well-documented positive relation between fund flows and future fund performances but find that the relation is significant only during high sentiment periods. Massa and Yadav (2015) show that mutual funds employ portfolio strategies based on the prevailing market sentiment. Within their model, they build a proxy for the degree of a fund’s sentiment beta (FSB). They investigate the sensitivities of individual stocks with respect to investor sentiment and aggregate these betas on the respective fund level. They define a sentiment contrarian strategy which predicts that mutual funds invest in low-sentiment beta stocks in order to generate high future performance. Overall, they suggest low-FSB funds to outperform their high-FSB counterparts.
The idea of a negative relation between fund flows and future fund performances can also be supported by the work from Blanchett (2010). He argues that while a mutual fund experiences large inflow’s the respective portfolio manager needs to determine how to invest these new assets. In many cases, he or she sticks to the existing investment strategy and purchases additional shares of the existing holdings. This in turn drives up the prices, resulting in a short-term performance boost - the “smart-money” effect. However, Blanchett finds that this boost vanishes over time, resulting in a “dumb-money” effect. This finding suggests that the new inflows have been invested in overbought and therefore overvalued securities. In a later publication, Blanchett (2012) measures the total future return costs associated with the
emerging inflows. He finds that the costs of fund flows are approximately accounting for -0.46% bps for large-cap mutual funds, -1.21% bps for mid-cap mutual funds, and -1.52% bps
for small-cap mutual funds.
gather a better understanding from what has to do with investment skills and what is motivated by investor sentiment.
2.2. Feedback Mechanism in Behavioral Finance
Feedback trader try to discover trends in past stock prices and base their decisions on the expectation of continuing trends (Bohl and Reitz, 2004). De Long et al. (1990) describe positive feedback trading as the tendency of buying past winners and selling past losers. Therefore, feedback trading can be thought of as responsible for generating momentum and reversal. Its underlying trades cause temporary price pressures in stocks, stimulating initial momentum and eventually leading to stock price reversals (Peng and Wang 2019). Barberis et al. (1998), Daniel et al. (1998), Hong and Stein (1999) and Barberis et al. (2015) also study this effect. Thereby, positive feedback trading can be related to several underlying mechanisms. De Long et al. (1990) describe momentum trading or front-running, Hong and Stein (1999) find bounded-rational behavior as responsible whereas Barberis et al. (1998) investigate underlying psychological behavior as for example the effect of representativeness.
Frictional behavior is also thought of as responsible for feedback trading. Peng and Wang (2019) study in this context mutual fund benchmarking. They investigate the impacts of positive feedback trading on stock price dynamics and subsequently the performance of mutual funds. They divide their fund sample in high and low feedback trading and claim that the caused price pressures from this trading are the most plausible explanation for enhanced cross-sectional momentum and excess volatility. In terms of stocks held by positive feedback funds, Peng and Wang (2019) find that they exhibit much stronger momentum which almost doubles the returns from a simple momentum strategy.
Another type of feedback loops are self-reinforcing feedback loops. The study from Zhong et al. (2018) indicates that momentum traders impact the change of price fluctuations and the equilibrium between trend-following and trend-rejecting strategies. In case of low market impact the existence of momentum traders has only a slight effect on the change of price fluctuations but outbalances the equilibrium between trend-following and trend-rejecting strategies. As a result, the trend-following herd behaviors become dominant ensuring the existence of a self-reinforcing positive feedback loop. This stays in contrast to the case of high market impact, where momentum traders cause high price fluctuations followed by suppressed trend-following strategies. As a result, Zhong et al. (2018) expect a negative feedback loop.
To summarize, Behavioral Finance claims feedback trading as responsible for feedback mechanisms such as feedback loops between flows and returns. Positive feedback traders cause initial momentum and following reversals, leading to advantages for first-movers. Moreover, emerging herd behaviors give rise to self-reinforcing feedback loops.
2.3. Investor Sentiment
In 1979, Kahneman and Tversky develop the prospect theory, which claims that each individual is subject to a loss aversion which results in an asymmetric perception of gains and losses. During their studies they observe that people tend to be more risk-averse when facing a risky choice leading to potential gains, but on the other hand risk-seeking when confronted with a risky choice leading to potential losses. This belief is contrary to the expected utility theory that postulates a decision making that perfectly rational agents would conduct. Based on these results, Lucey and Dowling (2005) find that in addition investors tend to be influenced from their state of mind, or sentiment in their decision-making process, especially under conditions of uncertainty. Investor sentiment can thereby be defined as optimism or pessimism of an investor about future stock market activity and investment risks that are not founded on facts (Baker and Wurgler, 2006). Here, high investor sentiment corresponds to optimism whereas low investor sentiment relates to pessimism. In other words, influenced by their previous experiences and expectations, and due to more or less risky choices in the context of stock trading, investors are anticipating potential losses or gains similar to the assumptions of prospect theory.
Different ways of measuring investor sentiment have been explored. They range from surveys, which examine the psychological dimensions of sentiment and market variables, to biases like overconfidence, conservatism or representativeness which can also be used to explain investors over- or underreaction to past returns (Baker and Wurgler, 2006). All these single measures are imperfect because of potential confounding factors. Surveys for example are highly subjective and since an investor’s mood fluctuates on a day-to-day basis, those measures that predict fund returns on the long term are especially useful. The solution to control for such confounding factors is to combine multiple imperfect measures (Baker and Wurgler, 2006).
Baker and Wurgler (2006, 2007) therefore develop the most widely used and sophisticated investor sentiment index based on a macroeconomic view, which consists of six proxies: trading volume, dividend premium, closed-end fund discount, the number and first-day returns on IPOs and the equity share in new issues. Using this index, they show an effect of investor sentiment on the stock market, in particular on such markets that are difficult to value and costly to arbitrage. According to Baker and Wurgler (2006, pp. 1645-1646) “mispricing is the result of an uninformed demand shock in the presence of a binding arbitrage constraint” thus “newer, smaller, more volatile, unprofitable, non-dividend paying, distressed or with extreme growth potential, and firms with analogous characteristics - are likely to be more affected by shifts in investor sentiment”. In line with these claims, they figure when investor sentiment is high, stocks of such firms that are especially attractive to speculators in contrast to arbitrageurs tend to achieve relatively low returns while low investor sentiment shows a contrasting pattern (Baker and Wurgler, 2006). Beer and Zouaoui (2013) posit similar conclusions using an index that includes different measurements for investor sentiment.
2.4. Hypotheses
While engaging with mutual funds, investors need to form their beliefs not only about future performances of the funds but also about the general optimism or pessimism in the activity of the stock market. Therefore, investor sentiment influences the decision-making process of investors as market-wide optimism or pessimism needs to be taken into account. Subsequently, there is reason to believe that in- and outflows of mutual funds tend to be influenced by investor sentiment. Indro (2004) finds support for this assumption while he studied the relationship between aggregated equity fund flows and investor sentiment using weekly flow data. He finds evidence for a strong relationship even after accounting for the effects of risk premium and inflation.
Massa and Yadav (2015) investigate sentiment sensitivities of individual stocks and aggregate these on fund level to obtain fund’s sentiment beta. They show that mutual funds employ portfolio strategies to obtain certain sentiment beta. Moreover, they define a sentiment-contrarian strategy which predicts that mutual funds invest in low-sentiment beta stocks in order to generate high future performance. Overall, they find low-sentiment beta funds to outperform their high-sentiment beta counterparts.
In contrast to the idea of portfolio strategies as a reason for sentiment sensitivities of funds, our study suggests certain investor bases as responsible. It can be thought of Momentum investors who expand their investments in case of a positive sentiment and reduce their investments if a negative sentiment is prevailing. This behavior could also be interpreted as trend-chasing. In contrast to that it can also be thought of Contrarian investors who sell their funds if sentiment is high and increase their positions if sentiment is low. These Contrarian investors use funds as mean reversion instruments or a way of exercising control over their fund managers. For example, restraining the managers investment appetite during high sentiment periods. Ultimately, these behaviors lead to sentiment-chasing fund flows in case of Momentum investors and sentiment-contrarian fund flows if Contrarian investors dominate.
We set our Hypotheses as following:
Hypothesis 1a: Fund flows impact fund performance.
Hypothesis 1b: Variations in fund flows can be explained by investor sentiment. Hypothesis 2a: Sentiment-chasing fund flows negatively impact fund performance. Hypothesis 2b: Sentiment-contrarian fund flows positively impact fund performance.
While Hypothesis 1a concentrates on the relation between fund flows and fund performances,
Hypothesis 1b targets the relation between fund flows and investor sentiment. Hypotheses 2a
and 2b combine both to make specific assumptions about the impact of investor sentiment on fund performances. The following figure 1 provides an overview about the conceptual model.
Figure 1. Conceptual model.
According to Baker and Wurgler (2006), markets that are difficult to value and costly to arbitrage face stronger effects of investor sentiment and are likely to be more affected by shifts in it. These markets are dominated by firms who are (Baker and Wurgler 2006, pp.1645-1646) “newer, smaller, more volatile, unprofitable, non-dividend paying, distressed or with extreme growth potential, and firms with analogous characteristics”. We use these conclusions and focus within our study on a well-chosen proxy to account for the described firm characteristics. We decide for Mutual Growth Funds as they come close to Baker and Wurgler’s (2006) definition. Moreover, we think of pessimistic investors, who start selling their growth funds first as these funds are the ones suffering the most from declining markets. In addition, optimistic investors will start buying growth funds first as they profit from growing markets the most. Hence, we posit that these investor behaviors strengthen the sentiment impacts within growth funds.
Fund performance Sentiment-chasing Fund Flows
+ _ Investor sentiment
Sentiment-contrarian Fund Flows
3. Data and Methodology
3.1. Data
Mutual fund data
The analysis of this study will be conducted on the basis of a sample of U.S. equity-based growth funds. The data is collected from Lipper, using the MLGE classification which offers exposure to U.S. Growth funds and belongs to the Holdings-Based Fund Classifications. The funds report in U.S. Dollar. Since investor sentiment tends to be cyclical, different investor sentiment periods – high and low – need to be represented in the analysis. Therefore, the mutual fund sample should be extensive enough to allow capturing the different dimensions of investor sentiment. The chosen sample period contains monthly data from 10 years – January 2009 until January 2019.
First, the fund universe and classification components were identified with the help from the Eikon Fund Screener. Second, the actual data was pulled with Thomson Reuters Datastream. In particular, monthly Total Net Assets (TNA) and Total Return Index (RI) levels for the funds were obtained. In the course of data preparation and cleaning those funds were deleted where data availability restrictions were present. From 374 funds in the classification 154 funds could be used for the further analysis. We do not control our fund sample for closed funds. Therefore, our dataset may face a survivorship bias. After cleaning the mutual fund sample, the fund data needs to be prepared for the analysis. First, the discrete fund specific returns need to be calculated from the Total Return Index data. This was done by using the following formula. Return of fund 𝑖 at the end of month 𝑡 is calculated as:
𝑅!,# = %𝑅𝐼!,#$%− 𝑅𝐼!,#( / 𝑅𝐼!,# (1)
where 𝑅𝐼!,#$% and 𝑅𝐼!,# are the monthly Total Return Index levels of fund i at the end of month 𝑡 + 1 and 𝑡 respectively.
Hereafter, the fund in- and outflows need to be aggregated in a monthly normalized fund flow. The fund flows were calculated in percent to avoid influences from different fund sizes. The calculation was done by using the following formula. Flow of fund 𝑖 at the end of month 𝑡 is calculated as:
where 𝑇𝑁𝐴!,# and 𝑇𝑁𝐴!,#&% are the monthly total net assets of fund i at the end of month 𝑡 and
𝑡 − 1 respectively, and 𝑅!,# is the monthly return of fund i at the end of month t.
Sentiment data
This study intends to measure investor sentiment based on the methodology from Baker and Wurgler (2006, 2007). They describe the most widely used investor sentiment proxies and continuously publish their sentiment data on their website. The obtained sentiment proxies cover the US level and therefore do not need to be adjusted for this study. Moreover, the sentiment proxies from Baker and Wurgler (2006, 2007) already have been orthogonalized with respect to a set of six macroeconomic indicators. Hence, the majority of macroeconomic influences in the sentiment index can be avoided.
Factor data
To assess fund performances and conduct risk-adjusted comparisons, U.S. factor data is needed. The data was obtained from the Kenneth R. French Data Library which provides Fama-French (1993) research factors including Market Risk Premium, Small-minus-Big (SMB) and High-minus-Low (HML) portfolio returns. Factor and return data are reported in U.S. Dollar and therefore consistent to the fund data.
The following table 1 illustrates the summary statistics for the analyzed data.
Summary Statistics Mean Standard deviation Min Max
Total Net Assets ($ millions) 687,52 2.463,25 0,10 30.889
Monthly Total Return (%) 1,19 5,18 -90,39 167,48
Monthly Net Cash Flows (%) 3,05 183,95 -99,97 21.397,21
Investor Sentiment -0,15 0,28 -0,89 0,38
We can see that the Total Net Assets from our fund sample range between below 1 and above 30.000 $ millions. This large range of assets can be explained by new funds entering the chosen fund classification and longer existing funds. Moreover, it is worth noting that the Monthly Total Returns range from -90% to +167%. These extreme high and low return rates can be
Table 1.
also striking that the Monthly Net Cash Flows vary consistently. In most of the extreme cases we can regard institutional share classes which explain the high in- as well as outflows.
3.2. Correlation Analysis
The objective within this part is to determine the predictive power of sentiment correlations on future fund performances. Hence, the question might be: Can correlations from the first half predict alphas in the second half. To answer this question, correlations within the first half of the sample period will be regressed against the alphas belonging to the second half. This can be done by employing a linear regression. By doing so, we can establish whether there is a significant, predictive relation between both.
Correlations
To get a first idea about the relation between investor sentiment and fund flows, correlations between both are going to be investigated. An interesting question in this case is to what extend does past correlation between sentiment and flows for a particular fund predict its future correlation as it might have an impact of the portfolio choice of an investor. Moreover, by means of correlations we are able to draw conclusions about the investor base. In case of positive correlations, we claim Momentum investors as investor base and in case of negative correlations we suggest the investor base to be dominated by Contrarian investors. Again, it would be interesting to see whether these funds have always these investors. To assess for consistency, we calculate correlations for different time-series. The sample period will be divided in two parts: the first five years and the second five years. In the following we refer to these time-periods as first half and second half. Furthermore, correlations will also be computed for the full sample period.
Estimating Fama French Alphas
Second, the funds will be introduced in an asset pricing model to estimate fund specific alphas. Instead of employing a panel regression, which would result in one alpha for the whole sample, our choice was made for the French (1993) Three-factor model. With the help of Fama-French’s (1993) model we can calculate alphas one by one on fund level. The mutual fund’s performances will be adjusted for the common risk factors Market Risk Premium, SMB and HML to assess whether they outweigh the factor-related returns. The following regression equation will be employed:
𝑅!,#− 𝑅𝑓# = 𝛼!,#+ 𝛽!,#'()𝑀𝐾𝑇 + 𝛽
where 𝑅𝑓# is the risk-free rate at the end of month 𝑡, 𝛼!,# is the constant term of fund 𝑖 at the
end of month 𝑡, 𝛽!,# represents the beta for the respective risk factor of fund 𝑖 at the end of
month 𝑡, 𝑀𝐾𝑇, 𝑆𝑀𝐵, 𝐻𝑀𝐿 are the risk factors and 𝜀!,# is the mean disturbance term.
Again, these calculations will be done for the full sample and for both sub periods to assess consistency and allow for further analyses. In case of excess returns, positive alphas can be achieved.
3.3. Fund Flows and Fund Returns
The fundamental idea of this study is that sentiment induced fund flows impact fund performances. Before analyzing this combined statement, we need to separately establish whether fund flows generally have an impact on fund performances. Therefore, interdependencies with investor sentiment will not be considered in this section.
We explain fund returns with an asset pricing model extended by a FLOW factor. We add fund specific flows from the previous month to the three Fama-French (1993) factors. Thus, we can estimate exposures to the Fama-French (1993) factors as well as to the FLOW factor. Concludingly, cross-sectional regressions of fund returns as a function of Fama-French (1993) factors and fund specific flows from the previous month will be employed. The following regression equation will be applicable:
𝑅!,#− 𝑅𝑓# = 𝛼!,#+ 𝛽!,#'()𝑀𝐾𝑇 + 𝛽
!,#*'+𝑆𝑀𝐵 + 𝛽!,#,'-𝐻𝑀𝐿 + 𝛽!,#.-/0𝐹𝐿𝑂𝑊!,#&%+ 𝜀!,# (4)
where 𝛽!,#.-/0represents the beta for the introduced risk factor 𝐹𝐿𝑂𝑊 of fund 𝑖 at the end of
month 𝑡 and 𝐹𝐿𝑂𝑊!,#&% are the fund specific flows from fund 𝑖 at the end of the previous month 𝑡 − 1.
To review Hypothesis 1a: Fund Flows impact Fund Performance, we conduct a one-sample t-test over our results to show whether the mean of flow coefficients is significantly different from zero. In the course of the t-test, a T-value with corresponding p-value as well as a critical T-value at a significance level of p<0.05 will be determined.
3.4. Sensitivity Analysis
chase past fund performances in fund flows, we add fund specific past returns as explaining variable in our model.
A cross-sectional regression model will be employed. Within this model, fund specific flows will be described as a function of investor sentiment and past fund returns. More precisely, the specific monthly flow percentage will present the dependent variable and the investor sentiment index as well as the past returns will stand for the underlying, independent variables. The following regression equation will be used:
𝐹𝐿𝑂𝑊!,# = 𝛼!,# + 𝛽!,#*12)𝑆𝐸𝑁𝑇#+ 𝛽!,#3𝑅!,#&%+ 𝜀!,# (5)
where 𝛽!,#*12)represents the beta for the Baker and Wurgler (2006, 2007) sentiment index 𝑆𝐸𝑁𝑇
at the end of month 𝑡 and 𝛽!,#3 the beta for the return 𝑅 of fund 𝑖 at the end of the previous month
𝑡 − 1.
Similar to the procedure of the correlation analysis in 3.2., the sensitivity analysis will be conducted for all three sub periods: First half, second half and full sample. By doing so, we can examine whether fund sensitivity changes over time and whether a certain investor base consistently buys certain funds.
As outcome of the cross-sectional regressions, flow coefficients for investor sentiment and past returns will be obtained. In case of positive sentiment coefficients, we conclude that the fund flows are chasing whereas reverse or negative coefficients lead to sentiment-contrarian fund flows. Based on the respective coefficients, we then establish a sentiment score for each fund with which help we sort the funds ranging from the highest positive to the most negative score. Eventually, the funds will be divided in the following two groups: Sentiment-chasing and sentiment-contrarian funds. The sentiment-Sentiment-chasing group will consist of the 20 funds with the highest positive score whereas the sentiment-contrarian group will be composed out of the 20 funds with the most negative score. Similarly, we compose two groups of past return-chasing and past return-contrarian funds based on the past return coefficients.
3.5. Performance Evaluation
The returns of the funds within the two groups of sentiment-chasing and sentiment-contrarian funds will be evaluated using Fama-French (1993) alphas, arithmetic returns as well as Time-weighted and Money-Time-weighted returns. Moreover, the funds can be investigated for their investor’s timing ability. Eventually, we can assess Hypothesis 2a: Sentiment-chasing Fund
Flows negatively impact Fund Performance and Hypothesis 2b: Sentiment-contrarian Fund Flows positively impact Fund Performance for their validity.
First, our fund sample needs to be checked for their excess returns over the Fama-French (1993) factors. On the basis of alphas, we can test whether there is any prediction of sentiment sensitivity on future fund performance. As we already have introduced our fund sample in the Fama-French (1993) model in part 3.2., we obtain fund specific alphas from these results. Second, we will summarize the realized returns in subsequent periods by means of arithmetic, Time-weighted and Money-weighted returns. Based on these calculations we can review funds and test to predict the performance in relation to their sentiment sensitivity.
The Time-weighted return represents the return of a buy-and-hold strategy and is also described as geometric mean return. The measure focusses solely on the return of an investment while disregarding the impact of fund flows. Each time a fund flow occurs, the portfolio is valued before and after the flow. The monthly return over the whole sample period is equal to the geometric average of the returns over all intervals. The following formula will be applicable: Time-weighted return of fund 𝑖 is calculated as
𝑇𝑊𝑅! = (∏) (1 + 𝑅!,#))
#4%
!
"− 1 (6)
According to Dichev (2007), Money-weighted returns describe the investor’s return rather than the investment’s return. Therefore, we can elaborate the investor’s actual returns and determine how well they time the market. Related to our Hypotheses 2a and 2b we can examine whether Momentum investors hurt their own performance and whether Contrarian investors increase their outcomes. Specifically, we can review the ability of investors to invest and withdraw money from the stock market prior to positive and negative market returns. Hence, Money-weighted returns assign more weight to periods with larger exposures to the stock market. The most common measure to account for Money-weighted returns is the internal rate of return. The following procedure describes the calculation of Money-weighted or Dollar-weighted returns. Money-weighted return of fund 𝑖 is obtained by solving the following equation for 𝑀𝑊𝑅:
𝑇𝑁𝐴) / (1 + 𝑀𝑊𝑅!)) = 𝑇𝑁𝐴
where 𝑇𝑁𝐴5 and 𝑇𝑁𝐴) are the total net assets of fund i at the end of month 0 and 𝑇 respectively.
Similar to Dichev (2007), we analyze investor’s timing ability by calculating the difference between the internal rate of return and the buy-and-hold return. Friesen and Sapp (2007) interpret this difference as performance gap, and we intend to follow their approach: Performance gap of fund 𝑖 is calculated as
𝐺𝐴𝑃! = 𝑇𝑊𝑅!− 𝑀𝑊𝑅! (8)
Furthermore, we will conduct a performance evaluation for past chasing and past return-contrarian funds. We want to examine whether return-chasing behavior or hot hands hurt or increase future returns. Based on the results, we can assess potential negative or positive feedback mechanisms. The performance evaluation will be carried out similarly as done for the sentiment-chasing and sentiment-contrarian groups.
4. Results
4.1. Correlation Analysis
The following table 2 illustrates our results:
Funds Correlations First Half Alphas Second Half
1st quintile 0.1924 -0.00096
5th quintile -0.1693 -0.00111
Correlations / Alphas Full Sample First Half Second Half
Full Sample 0.8788
First Half 0.9750 0.4207
Second Half 0.5131
In the first part of the table, we can regard sentiment and fund flow correlations within the first half of our sample period and Fama French (1993) alphas from the second half. The results are displayed for the 1st quintile with the highest positive correlations and for the 5th quintile with the most negative correlations. In our computation, we can regard correlations varying from 0.2933 to -0.4399 for the full sample period. We conclude that the relations between investor sentiment and fund flows vary among the funds. If we compare the alphas from the 1st and 5th quintiles, we already can see that the alphas do not vary significantly.
We perform linear regressions to examine the relations in detail. The second part of the table visualizes their results. To assess the predictive power of the correlations, we regress them from the first half against the alphas from the second. With a significance level of 0.4207 we find no significant relation, supporting our first impression that there is no predictive relation between both. To test whether there is a relation at all, we regress correlations and alphas belonging to the same time periods. We can regard a significant relation between both within the first half – the significance level amounts for 0.9750. During the second half, this level vanishes to 0.5131, resulting in a level of 0.8788 for the full sample period. Therefore, we claim a marginally significant relation between correlations and excess returns, even if we do not find predicting power in the out-of-sample analysis.
Table 2.
4.2. Fund Flows and Fund Returns
Within this section we analyze whether fund flows generally have an impact on fund performances. We explain fund returns with the Fama-French (1993) model extended by a FLOW factor. We obtain exposure estimates for the factors on fund level and aggregate them over all 154 funds. Moreover, we sort the estimates by the FLOW coefficients and divide the sample in five quintiles. We display the 1st and 5th quintiles and the aggregate over all funds in the following table 3:
Fund returns Market SMB HML FLOW Alpha
Aggregated 1.1216 0.0070 -0.2533 0.0139 -0.0014
1st quintile 1.1221 -0.0055 -0.2440 0.0431 -0.0014
5th quintile 1.0940 0.0750 -0.2621 -0.0130 -0.0015
Based on the aggregated estimates, we conclude significant relations between fund returns and the factors. Noticeable, the market factor with an aggregated coefficient of 1.1216 explains the majority of return variations in our fund sample followed by HML which has a significant negative relation of -0.2533. The mean of flow coefficients amounts for 0.0139 with a standard deviation of 0.0209. Overall, there is only a small property of variations left which we cannot explain by either the factors Market, SMB or HML nor by the FLOW factor. On fund level, we can regard significant coefficients for our FLOW factor ranging from 0.0828 to -0.0572.
Eventually, we perform a one-sample t-test over the mean of flow coefficients to review
Hypothesis 1a. The result shows that the mean of flow coefficients is with 0.0139 at a T-value
of 8.2374 with a corresponding p-value of 7.308E-14 significantly different from zero. Moreover, the critical T-value of 1.9756 at a significance level of p<0.05 is far exceeded. Therefore, we can affirm Hypothesis 1a: Fund Flows impact Fund Performance and conclude that fund flows generally have a significant impact on fund performances.
4.3. Sensitivity Analysis
The sensitivity analysis of our study investigates the property of fund flows. Specifically, we examine fund flow parts for their sentiment and past return sensitivity. We employ a cross-sectional regression model in which monthly fund flow percentages present the dependent
Table 3.
variable and the Baker and Wurgler (2006, 2007) sentiment index as well as fund specific past returns the underlying, independent variables. Similar to the correlation analysis, we perform regressions for all three time-series: First half, second half and full sample.
In the course of the analysis, we obtain flow coefficients for investor sentiment and fund specific past returns. We interpret fund flows with positive sentiment coefficients as sentiment-chasing and fund flows with negative coefficients as sentiment-contrarian. Results are displayed for our two created groups: Sentiment-chasing and sentiment-contrarian funds and sorted by the sentiment coefficients. We aggregate sentiment, past return and constant estimates for both groups on the basis of the 20 funds with the highest positive score and the 20 funds with the most negative score. Within the aggregation of past return coefficients for the sentiment-contrarian group, we exclude one fund due to the extreme finding of a past return coefficient of -5.070. This extreme value would highly drive our aggregate estimate, as our sentiment-contrarian group consist of only 20 funds. The following table visualizes our results:
Sentiment-chasing funds Sentiment Past Return Constant
Aggregated 0.1711 0.9280 0.0614
1st 10 funds 0.2496 0.9812 0.0941
2nd 10 funds 0.0927 0.8749 0.0287
Sentiment-contrarian funds Sentiment Past Return Constant
Aggregated -0.1036 1.0776 -0.0100
1st 10 funds -0.1676 1.1431 -0.0054
2nd 10 funds -0.0395 1.0120 -0.0145
We obtain sentiment coefficients varying from 0.686 to -0.480 over the whole fund sample. Within the group of sentiment-chasing funds, the sentiment coefficients vary from 0.686 to 0.0814 and the aggregate amounts for 0.1711. Within the group of sentiment-contrarian funds, the sentiment coefficients vary from -0.482 to -0.0269 and the aggregate accounts for -0.1036. We conclude that some of the funds experience stronger sentiment sensitivities than others and that fund flows vary from chasing to contrarian. We can regard sentiment-chasing and -contrarian flows on fund level as well as on group level. A one-sample t-test shows that the mean of sentiment coefficients is with 0.0467 at a T-value of 2.7874 with a
Table 4.
corresponding p-value of 0.0060 significantly different from zero. The critical T-value of 1.9756 at a significance level of p<0.05 is exceeded. Hence, we can affirm Hypothesis 1b:
Variations in Fund Flows can be explained by Investor Sentiment. The standard deviation of
the mean of sentiment coefficients amounts for 0.2078.
Furthermore, we can regard astonishing high coefficients for fund specific past returns, irrelevant of the fund’s sentiment sensitivity. We claim that past returns explain and drive a vast majority of fund flows. This finding supports literature in the area of past return-chasing behavior (Jiang and Yüksel, 2019). Based on our results, we are able to assess flow-performance relations. If past fund flow-performance is high, investors will invest further money in these funds, leading to altering fund performances. The upcoming performance evaluations in part 4.4 will shed a light on potential performance impacts.
Next, we assess whether fund sensitivities change over time enabling us to make assumptions about whether a certain investor base consistently buys certain funds. Again, we aggregate sentiment coefficients similarly as done in table 4. Within the aggregation of sentiment-contrarian coefficients, we exclude one fund due to an extreme value which would dominantly drive our aggregated estimates.
Sentiment-chasing funds Full Sample First Half Second Half
Aggregated 0.1711 0.1576 0.3838
1st 10 funds 0.2496 0.2300 0.6113
2nd 10 funds 0.0927 0.0852 0.1562
Sentiment-contrarian funds Full Sample First Half Second Half
Aggregated -0.1036 -0.0939 0.0243
1st 10 funds -0.1676 -0.1542 0.0130
2nd 10 funds -0.0395 -0.0336 0.0356
We can regard that sentiment sensitivities within the sentiment-chasing group are somewhat consistent over time, being consistently in the positive area. The funds within the sentiment-contrarian group exhibit changing sensitivities, facing vanishing negative coefficients in the second half. Therefore, we conclude a consistent investor base in case of sentiment-chasing funds and altering investors in case of sentiment-contrarian funds. This finding is in line with
Table 5.
Sentiment Consistency: Table summarizes sentiment coefficients of 462 regressions. Results sorted for Sentiment score/coefficient. Two groups:
Sentiment-chasing and Sentiment-contrarian funds.Observations (months): 119
Momentum investors consistently chasing investor sentiment and Contrarian investors who use funds as mean reversion instruments or ways of restraining or encouraging fund managers in high or low sentiment periods.
4.4. Performance Evaluation
In the following table 6 we compare Fama-French (1993) alphas, arithmetic returns as well as Time-weighted and Money-weighted returns from the two groups of sentiment-chasing and sentiment-contrarian funds. By means of Time-weighted and Money-weighted returns we can assess investor’s timing ability which we call GAP. Results are sorted by sentiment score/coefficient and aggregated for the whole 20 funds of each group and for the first 10 and second 10 funds in relation to their sentiment scores/coefficients.
Sentiment-chasing Sentiment Alpha AR TWR MWR GAP
Aggregated 0.1711 -0.00131 1.06% 0.97% 3.67% -2.70%
1st 10 funds 0.2496 -0.00126 1.10% 0.98% 6.00% -5.02%
2nd 10 funds 0.0927 -0.00135 1.03% 0.97% 1.09% -0.12%
Sentiment-contrarian Sentiment Alpha AR TWR MWR GAP
Aggregated -0.1036 -0.00175 1.06% 0.91% 4.84% -3.93%
1st 10 funds -0.1676 -0.00167 1.07% 0.92% 8.46% -7.55%
2nd 10 funds -0.0395 -0.00184 1.06% 0.90% 1.23% -0.32%
We can regard that the Fama-French (1993) Alphas as well as the arithmetic returns and Time-weighted returns do not vary significantly among our two groups. Fama-French (1993) alphas are overall slightly negative, suggesting performance cuts caused by management or performance fees of the funds. Arithmetic returns account for slightly above 1% per month on average, which yields are high yearly return. The Time-weighted returns, which represent the returns of a buy-and-hold strategy, account for slightly under 1% per month on average. We can see that the Time-weighted returns for sentiment-chasing funds are marginally higher than those from the sentiment-contrarian ones. Only the Money-weighted returns vary significantly among our fund groups. The results show with 4.84% to 3.67% higher Money-weighted returns for sentiment-chasing funds than for their sentiment-contrarian counterparts. Moreover, we can see that Money-weighted returns with 6.00% and 8.46% are higher for the 1st 10 funds than for the 2nd 10 funds with 1.09% and 1.23% in both groups. The results for performance gaps show
Table 6.
similar results, suggesting that investors in sentiment-contrarian funds experience better timing abilities than those in sentiment-chasing ones.
According to Dichev (2007), Money-weighted returns describe the investor’s return rather than the investment’s return. Thus, Money-weighted returns show the investor’s actual returns and determine how well they time the market. Based on the results, we conclude that investors in sentiment-contrarian funds have better timing abilities than those from sentiment-chasing ones. Contrarian investors increase their own returns by adjusting their investment behavior leading to timing abilities. As we only have indirect evidence for this statement, survey data would be needed for a validation.
In table 7 we compare Fama-French (1993) alphas, arithmetic returns as well as Time-weighted and Money-weighted returns from the two groups of past chasing and past return-contrarian funds. Results are sorted by past return coefficients and aggregated similarly as done for table 6.
Past return-chasing Past return Alpha AR TWR MWR GAP
Aggregated 1.3267 -0.00154 1.23% 1.12% 2.75% -1.64%
1st 10 funds 1.5006 -0.00145 1.22% 1.11% 4.33% -3.22%
2nd 10 funds 1,1528 -0.00162 1.24% 1.12% 1.18% -0.06%
Past return-contrarian Past return Alpha AR TWR MWR GAP
Aggregated 0.2768 -0.00041 1.36% 1.11% 2.65% -1.54%
1st 10 funds -0.0242 -0.00021 1.41% 1.06% 1.04% 0.02%
2nd 10 funds 0.5779 -0.00061 1.32% 1.18% 4.45% -3.27%
While both groups only show slight differences when comparing their Time-weighted and Money-weighted returns, we find significant performance differences when comparing Fama-French (1993) alphas and arithmetic returns. The alphas from the past return-contrarian group are with -0.00041 for the whole group higher than -0.00154 for the whole group of past return-chasing funds. Moreover, the arithmetic returns are with 1.36% to 1.23% higher for the past return-contrarian finds. Based on these results, we claim return-chasing behavior or hot hands to hurt fund performances while return-contrarian behavior increases returns. Moreover, these results indicate a feedback mechanism between fund returns and fund flows. Last month’s returns drive current fund flows and these flows drive current fund returns. As the impact of
Table 7.
return-chasing behavior turns out to hamper fund performances, we posit a negative flow-performance relation.
To summarize, we only find negligible performance differences between sentiment-chasing and sentiment-contrarian funds. In the course of our study, we suggest that sentiment-chasing investors hurt subsequent fund performances while sentiment-contrarian investors increase them. Our results posit that these claims cannot be validated. Therefore, we need to reject both
Hypotheses 2a and 2b. However, we find significant performance differences between past
return-chasing and past return-contrarian funds. Therefore, we claim return-chasing behavior to hurt fund performances, while return-contrarian behavior increases returns. These results indicate a feedback mechanism between fund returns and fund flows resulting in a negative flow-performance relation.
5. Conclusion
Overall, we find that investor sentiment determines fund flows and that fund flows determine fund returns. However, these findings do not necessarily imply that fund flows channel negative sentiment impacts on fund returns. We only find negligible performance differences between sentiment-chasing and sentiment-contrarian funds. Furthermore, we show that investors from sentiment-contrarian funds have better timing abilities than those from sentiment-chasing ones. Hence, Contrarian investors increase their own returns by adjusting their investment behavior leading to timing advantages. Regarding the consistency of sentiment sensitivities, we only find a consistent investor base in case of chasing funds but altering investors in sentiment-contrarian funds. Our correlation analysis finds no significant, predictive relation between sentiment correlating fund flows and future fund performances. We can only regard marginally significant relations between correlations and performances within the same time periods.
While conducting our sensitivity analysis we find astonishing high coefficients for fund specific past returns, regardless of the fund’s sentiment sensitivity. This finding coincides with the work from Jiang and Yüksel (2019). Also, our performance evaluation shows that this return-chasing behavior hurts fund performances. This result indicates a feedback mechanism between fund returns and fund flows. Last month’s returns drive current fund flows and these flows drive current fund returns. Thus, a feedback loop or mechanism is present. As the impact of return-chasing behavior turns out to hamper fund performances, we posit a negative flow-performance relation.
introduce a rather simple approach with which we are able to measure fund’s sentiment sensitivities directly on fund level.
This study has limitations that could be addressed with future research. While conducting the sensitivity analysis, we do not control the explaining variables for correlations and interdependencies. A piecewise regression could investigate isolated and unique contributions within the variance explanation, and therefore could extend the findings presented here. It should also be noted that our fund data received from Thomson Reuters is not controlled for closed funds. Thus, our analyzed dataset may face a survivorship bias. Moreover, although we show that certain funds consistently exhibit consistent sentiment sensitivities, we have only indirect evidence for a corresponding investor base. Survey data would be needed for final validation.
References
Asur, S., Huberman, B.A., 2010. Predicting the future with social media, in: In Proceedings of the 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology - Volume 01, WI-IAT ’10. IEEE Computer Society, pp. 492–499. Bailey, W., Kumar, A., Ng, D., 2011. Behavioral biases of mutual fund investors. Journal of
Financial Economics 102, 1–27. https://doi.org/10.1016/j.jfineco.2011.05.002
Baker, M., Wurgler, J., 2006. Investor Sentiment and the Cross-Section of Stock Returns. The Journal of Finance 61, 1645–1680. https://doi.org/10.1111/j.1540-6261.2006.00885.x Baker, M., Wurgler, J., 2007. Investor Sentiment in the Stock Market. Journal of Economic
Perspectives 21, 129–152. https://doi.org/10.1257/jep.21.2.129
Barberis, N., Greenwood, R., Jin, L., Shleifer, A., 2015. X-CAPM: An extrapolative capital asset pricing model. Journal of Financial Economics 115, 1–24.
https://doi.org/10.1016/j.jfineco.2014.08.007
Barberis, N., Shleifer, A., Vishny, R., 1998. A Model of Investor Sentiment. Journal of Financial Economics 49, 307–343.
Beer, F., Zouaoui, M., 2013. Measuring Stock Market Investor Sentiment. JABR 29, 51–68. https://doi.org/10.19030/jabr.v29i1.7555
Blanchett, D.M., 2012. Fund Flows, Momentum, and Mutual Fund Performance. The Journal of Investing 21, 83–91. https://doi.org/10.3905/joi.2012.21.2.083
Blanchett, D.M., 2010. Do Passive or Active Investors Make Better Asset Allocation Decisions? The Journal of Index Investing 1, 74–80.
https://doi.org/10.3905/jii.2010.1.2.074
Bohl, M., Reitz, S., 2006. Do Positive Feedback Traders Act in Germany’s Neuer Markt. Quarterly Journal of Business and Economics 45. https://doi.org/10.2307/40473411 Bohl, M.T., Reitz, S., 2004. The Influence of Positive Feedback Trading on Return
Autocorrelation: Evidence for the German Stock Market, in: Geberl, S., Kaufmann, H.-R., Menichetti, M.J., Wiesner, D.F. (Eds.), Aktuelle Entwicklungen Im
Finanzdienstleistungsbereich. Physica-Verlag HD, Heidelberg, pp. 221–233. https://doi.org/10.1007/978-3-7908-2651-7_15
Brown, G.W., Cliff, M.T., 2005. Investor Sentiment and Asset Valuation. The Journal of Business 78, 405–440.
Capon, N., Fitzsimons, G.J., Alan Prince, R., 1996. An individual level analysis of the mutual fund investment decision. J Finan Serv Res 10, 59–82.
https://doi.org/10.1007/BF00120146
Collins, S., Plantier, C., 2014. Are Bond Mutual Fund Flows Destabilizing? Examining the Evidence from the “Taper Tantrum” (SSRN Scholarly Paper No. ID 2510666). Social Science Research Network, Rochester, NY. https://doi.org/10.2139/ssrn.2510666 Coval, J., Stafford, E., 2007. Asset fire sales (and purchases) in equity markets. Journal of
Financial Economics 86, 479–512.
Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under- and overreactions. Journal of Finance 53, 1839–1885.
De Long, J.B., Shleifer, A., Summers, L., Waldmann, R., 1990. Noise Trader Risk in Financial Markets. Journal of Political Economy 98, 703–38.
De Long, J.B., Shleifer, A., Summers, L., Waldmann, R., 1990. Positive Feedback Investment Strategies and Destabilizing Rational Speculation. Journal of Finance 45, 379–395. Dichev, I.D., 2007. What Are Stock Investors’ Actual Historical Returns? Evidence from
Dollar-Weighted Returns. The American Economic Review 97, 386–401.
Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3–56. https://doi.org/10.1016/0304-405X(93)90023-5
Feroli, M., Kashyap, A.K., Schoenholtz, K.L., Shin, H.S., 2014. Market Tantrums and Monetary Policy (SSRN Scholarly Paper No. ID 2409092). Social Science Research Network, Rochester, NY. https://doi.org/10.2139/ssrn.2409092
Fischhoff, B., Slovic, P., Lichtenstein, S., 1977. Knowing with certainty: The appropriateness of extreme confidence. Journal of Experimental Psychology: Human Perception and Performance 3, 552–564. https://doi.org/10.1037/0096-1523.3.4.552
Frazzini, A., Lamont, O., 2008. Dumb money: Mutual fund flows and the cross-section of stock returns. Journal of Financial Economics 88, 299–322.
Friesen, G.C., Sapp, T.R.A., 2007. Mutual fund flows and investor returns: An empirical examination of fund investor timing ability. Journal of Banking & Finance 31, 2796– 2816.
Goldstein, I., Jiang, H., Ng, D.T., 2017. Investor flows and fragility in corporate bond funds. Journal of Financial Economics 126, 592–613.
https://doi.org/10.1016/j.jfineco.2016.11.007
Gruber, M.J., 1996. Another Puzzle: The Growth in Actively Managed Mutual Funds. The Journal of Finance 51, 783–810. https://doi.org/10.1111/j.1540-6261.1996.tb02707.x Hong, H., Stein, J.C., n.d. A Unified Theory of Underreaction, Momentum Trading, and
Overreaction in Asset Markets. The Journal of Finance 42.
Indro, D.C., 2004. Does Mutual Fund Flow Reflect Investor Sentiment? Journal of Behavioral Finance 5, 105–115. https://doi.org/10.1207/s15427579jpfm0502_4
Jiang, G.J., Yüksel, H.Z., 2019. Sentimental mutual fund flows. Financial Review 54, 709– 738. https://doi.org/10.1111/fire.12201
Kahneman, D., Tversky, A., 1984. Choices, values, and frames. American Psychologist, 39(4), 341-350. https://doi.org/10.1037/0003-066X.39.4.341
Kahneman, D., Tversky, A., 1979. Prospect Theory: An Analysis of Decision under Risk. Econometrica 47, 263–291. https://doi.org/10.2307/1914185
Keswani, A., Stolin, D., 2008. Which Money Is Smart? Mutual Fund Buys and Sells of Individual and Institutional Investors. The Journal of Finance 63, 85–118.
https://doi.org/10.1111/j.1540-6261.2008.01311.x
Lacalle, D., 2019. Are the Effects of Unconvential Monetary Policy on Financial Markets causing Bubbles? JBAFP 1, 1–1. https://doi.org/10.26870/jbafp.2018.01.003
Li, H., Guo, Y., Park, S.Y., 2017. Asymmetric Relationship between Investors’ Sentiment and Stock Returns: Evidence from a Quantile Non-causality Test. International Review of Finance 17, 617–626. https://doi.org/10.1111/irfi.12120
Liu, S., 2015. Investor Sentiment and Stock Market Liquidity. Journal of Behavioral Finance 16, 51–67. https://doi.org/10.1080/15427560.2015.1000334
Lou, D., 2012. A Flow-Based Explanation for Return Predictability. Rev Financ Stud 25, 3457–3489. https://doi.org/10.1093/rfs/hhs103
Lucey, B.M., Dowling, M., 2005. The Role of Feelings in Investor Decision-Making. Journal of Economic Surveys 19, 211–237. https://doi.org/10.1111/j.0950-0804.2005.00245.x Makrehchi, M., Shah, S., Liao, W., 2013. Stock Prediction Using Event-Based Sentiment
Analysis. 2013 IEEE/WIC/ACM International Joint Conferences on Web Intelligence (WI) and Intelligent Agent Technologies (IAT). https://doi.org/10.1109/WI-IAT.2013.48 Massa, M., Yadav, V., 2015. Investor Sentiment and Mutual Fund Strategies. Journal of
Financial and Quantitative Analysis 50, 699–727.
Odean, T., 1998. Volume, Volatility, Price, and Profit When All Traders Are Above Average. The Journal of Finance 53, 1887–1934. https://doi.org/10.1111/0022-1082.00078
Oh, C., Sheng, O., 2011. Investigating Predictive Power of Stock Micro Blog Sentiment in Forecasting Future Stock Price Directional Movement, in: ICIS.
Piccoli, P., Chaudhury, M., 2018. Overreaction to extreme market events and investor sentiment. Applied Economics Letters 25, 115–118.
https://doi.org/10.1080/13504851.2017.1302052
Renault, T., 2017. Intraday online investor sentiment and return patterns in the U.S. stock market. Journal of Banking & Finance 84, 25–40.
Sapp, T., Tiwari, A., 2004. Does Stock Return Momentum Explain the “Smart Money” Effect? The Journal of Finance 59, 2605–2622. https://doi.org/10.1111/j.1540-6261.2004.00710.x
Shefrin, H., Statman, M., 1985. The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence. The Journal of Finance 40, 777–790.
https://doi.org/10.2307/2327802
Shiller, R.J., 2003. From Efficient Markets Theory to Behavioral Finance. Journal of Economic Perspectives 17, 83–104. https://doi.org/10.1257/089533003321164967 Teo, M., Woo, S.-J., 2004. Style effects in the cross-section of stock returns. Journal of
Financial Economics 74, 367–398. https://doi.org/10.1016/j.jfineco.2003.10.003 Thaler, R., 1980. Toward a positive theory of consumer choice. Journal of Economic
Behavior & Organization 1, 39–60. https://doi.org/10.1016/0167-2681(80)90051-7 Wilcox, R.T., 2003. Bargain Hunting or Star Gazing? Investors’ Preferences for Stock Mutual
Funds. The Journal of Business 76, 645–663. https://doi.org/10.1086/377034
Yang, C., Zhou, L., 2015. Sentiment approach to underestimation and overestimation pricing model. Economic Modelling 51, 280–288.
Zheng, L., 1999. Is Money Smart? A Study of Mutual Fund Investors’ Fund Selection Ability. Journal of Finance 54, 901–933.
