國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
11
4. Know the best
Figure 1: Investors do not earn positive alpha on average
The fund industry on average seems doesn’t generate significant positive alpha.
Lewellen (2011) find institutional investors only generate insignificant 0.32% CAPM alpha annually. Dichev and Yu (2011) find for the hedge funds, the dollar-weighted return of the investors is even worse than the buy-and-hold return of the hedge funds by 3% to 7% annually. And also, no alpha generated, they just barely win over the risk-free rate. Dyck, Lins, and Pomorski (2013) find for actively managed funds in the U.S., they underperform 0.28% annually after fees.
Figure 2: Alpha is not ideal for skill measurement
Lack of alpha does not mean funds do not have skills on average. Berk and van Binsbergen (2015) challenge the fairness for using the alpha as a measurement. The
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
12
net alpha (alpha minus fees) may be the result of the competition of investors for skilled fund, so they are close to zero. Because big funds are generally harder to generate the same gross alpha as smaller funds. The return base measure as gross alpha (alpha before fees) may overlook the size effect of funds. They propose using fund value-added as a measure of funds skill. They find mutual funds do on average, add US$3.2 million annually. Kacperczyk, Nieuwerburgh, and Veldkamp (2014) also find the skill of mutual funds diminishing when a fund grows in size and age. Top 25% funds are average $400 million smaller and five years younger. In comparison, top managers are only slightly younger, one year on average, and 1.7 years of lesser experience. Roussanov, Ruan, and Wei (2018) also find 0.78% annually alpha decrease when funds have one standard deviation increase in log size. However, Ibbotson, Chen, and Zhu (2011) find bigger hedge funds take more risks and gain more rewards, thus there are no differences in performance compared with smaller funds.
Puckett and Yan (2011) mention a possible flaw in studying skills of funds. Most literature is based on quarterly data because of its availability. That may overlook the round-trip trades between quarters. They acquired complete transaction data for institutional investors, finds 0.27% to 0.34% return annually. When they mimicking the quarterly data with the same data set, their performances go negative.
Figure 3: Skill of funds exists and it is consistent, not just due to good lucks
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
13
As in any industry, some funds perform well while some are not. It is not fair to average out and make a conclusion all the funds do not have skills. Researches prove that skill of funds exists and it is consistent, not just due to good lucks. Kacperczyk, Nieuwerburgh, and Veldkamp (2014) construct a skill index, find after half-year, those on the top 20% outperform the bottom 20% by 3.0% to 6.3% annually, 0.6% to 2.1% annually after one year, based on different alpha calculation. Berk and van Binsbergen (2015) even find the skill can persistent about ten years. Sun, Wang, and Zheng (2018) show real skill is doing relatively well in recessions, not those doing well in booming. The top 20% hedge funds in recession outperform 7% than the bottom 20% in the next year, and their skill can persist for about three years. The top hedge funds in the booming market do not show persistent skill.
Table 5: How skilled funds riding the waves?
There is no one-size-fits-all solution for a market boom and recession. The best funds should be adjusted with the market cycles accordingly. How are skilled funds riding the waves? Kacperczyk, Nieuwerburgh, and Veldkamp (2014) find skilled mutual funds have better picking skill when market boom, and better stock-timing skill when in market recession. Previous job experience may contribute to the managers’ stock-picking skill or stock-timing skill. Chen et al. (2018) find those
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
14
mutual fund managers whose previous job is an industry analyst, they have better stock-picking skill. And those previous job is a microanalyst have better stock-timing skill. In different market sentiment environments also suit for different tactics. Smith et al. (2016) focus on hedge funds. They suggest because when sentiment is high, stocks have a higher chance to be overvalued. Thus, the technical analysis users have better performance than others by 5.3% in return, four-factor alpha 1.4%, and seven-factor alpha 1.3% annually. When the sentiment low, stocks have a higher chance to be undervalued. Thus, non-technical users like fundamental analysis can better capture those undervalue stocks. Non-technical users outperform by 2.4% in return, four-factor alpha 2.3%, and seven-factor alpha 0.6% (not significant) annually.
Knowledge is power. Newly found factors by academy may also widely use by investors. Mclean and Pontiff (2016) look into 97 predictors from 79 studies that able to predict cross-sectional stock returns. By comparing to out-of-sample results to rule out statistical bias, they estimate a 32% decline in post-publication returns are
contributed to investors learn from the academy. Though this study does not specify the investors’ type. However, Edelen, Ince, and Kadlec (2016) By divided into two periods, 1982-1996 and 1997-2011, they find the seven well-known anomalies do not disappear over time when those anomalies are much well known.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
15
Figure 4: Speed is the name of the game
Speed is the name of the game. Chordia, Roll, and Subrahmanyam (2005) find information measure by order imbalance, reacts in the range of 5 to 60 minutes. A specific case studied by Hu, Pan, and Wang (2017), they find high-frequency traders (HFTs) discover the price within 200 milliseconds. Many even jump the gun, suggest possible information leakage. Hendershott, Livdan, and Schürhoff (2015) find
evidence that the direction of institution trading can predict earnings surprises and unscheduled company crisis announcements. They begin to trading the stock ten days before the announcement. For macroeconomic news, they seem to lose prediction power except for the economic indicators. That suggests information leakage between companies and institutional investors. Kurov et al. (2019) find even macroeconomic news cannot immune to information leakage. They find 9 out of 20 macroeconomic have evidence for significant preannouncement movement. The price starts to move about 30 minutes before the announcement, that movement account for 30% to 67%
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043
16
total reacted price for the stock market, 16% to 55% for the bond market. Gao and Huang (2016) show that hedge funds may be using their connection with lobbyists to gain private political information. They see 6.7% to 11.1% annually difference by connected hedge funds investing in politically influenced stocks. Also, they find the performance drop significantly after Stop Trading on Congressional Knowledge (STOCK) Act, show the importance of regulations.
The golden standard to measure the performance is alpha. However, there are so many different kinds of alpha that often lead to widely different results. The reputable articles often list results from multiple kinds of alpha, including insignificant ones.
That diversity of alpha makes direct cross-comparison between the result of articles questionable. Many pieces of the research base on a select time frame or relatively short period, that result may change in the different financial environment. A well-conduct Mata-research to replicate the resulting from different articles and datasets has tremendous value. Such as Mclean and Pontiff (2016) reproduce the results of 97 predictors from 79 studies.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
DOI:10.6814/NCCU202001043