Are All Short-Term Institutional Investors Informed?

Abstract

AbstractWe examine whether being a hedge fund has any differential effect on the previously documented empirical relation between investment horizon and informativeness of institutional investors’ trades. We find that the positive and significant relation between short-term institutional demand and future stock returns exists only among hedge funds, while such relation does not exist for non–hedge fund institutions with short investment horizons. We also provide evidence that our results are not driven by (false) presumptions that hedge funds represent the majority of short-term institutional investors or that hedge fund demand constitute the lion’s share of the short-term institutional demand.Keywords: hedge fundsshort-term institutional investorsturnoverPL Credits: 2.0 Disclosure statementThe authors have no conflict of interest to declare. This article presents the authors’ opinions and not those of Northern Trust Corporation, its affiliates, clients, or employees. All errors are the sole responsibility of the authors.AcknowledgmentsWe thank Vikas Agarwal for sharing with us a comprehensive list of hedge funds that file 13F holdings. We also thank Kent Daniel and Kenneth French for making a large amount of data publicly available in their online data library.Notes1 In addition to these articles that provide evidence of a positive relation between short-term investment horizon and future stock returns, there are a few studies that show the contrary. For example, Cremers and Pareek (Citation2016) find that funds trading frequently generally underperform, drawing a negative cross-sectional relation between turnover and performance. Similarly, Chakrabarty, Moulton, and Trzcinka (Citation2017) show that majority of short-term institutional trades lose money.2 We differ from this literature in a significant way by working on a more refined sample in 13F. Specifically, we first divide the institutional investors as short-term and long-term institutional investors as in Yan and Zhang (Citation2009), and later we divide each group further into two groups as hedge funds and non–hedge funds. Thus, we analyze the performance of four groups of institutional investors’ trades: short-term hedge funds, short-term non–hedge funds, long-term hedge funds, and long-term non–hedge funds. Out of these four groups, we find that only the trades of short-term hedge funds predict future stock returns.3 In terms of the distribution of sources in identifying the 13F-filing hedge funds among all institutions in our sample, the union database covers approximately 90% of our 13F-filing hedge fund sample. An additional 6% of the sample comes from ADV filings. The other three criteria combined (industry publications, company websites, and news articles in Factiva) constitute the remaining 4% of our 13F-filing hedge fund sample.4 Agarwal, Fos, and Jiang (Citation2013), similarly detect 23% (1,199 out of 5,188) of all 13F-filing institutions as hedge funds during their sample period that ends in 2008 as well.5 In order to eliminate any concerns that our main findings are simply due this difference in CR values between short-term hedge funds and short-term non–hedge funds, in a separate analysis we decompose short-term non–hedge funds into two groups based on their CR values (short-term non–hedge funds with low CR values and short-term non–hedge funds with high CR values) and examine the predictive power of these two groups’ trades separately. We find that even though the short-term non–hedge funds with high CR values have an average CR value (30.40%) greater than the average CR value of the short-term hedge funds (27.34%), the trades of short-term non–hedge funds with high CR values still fail to predict the future stock returns in a significant way in risk-adjusted terms. Thus, this result rules out the possibility that the strong predictive power of short-term hedge funds’ trades is due to short-term hedge funds’ slightly higher CR values.6 In untabulated results, when we analyze the percentage of market capitalization by hedge funds in our sample over the years, we notice that it is very small in the initial years of our sample (less than 5%). However, this figure gradually increases over time and the average hedge-fund share of total market capitalization turns out to be around 10% during our sample period, 1994–2019.7 We define active share for institution i at quarter end t as follows: Active Shareit= 12∑j=1N|PWijt−MWjt|, where N represents the number of stocks in our sample (i.e., CRSP universe with our filters discussed in Section 2.1), PWijt denotes the weight of stock j in institution i’s portfolio at quarter end t, and MWjt is the weight of stock j in our sample of stocks at quarter end t. It should be noted that while the benchmark index of a mutual fund is usually readily identifiable, it is impossible to identify a benchmark index for an institutional investor since institutional investors may or may not include mutual funds alongside many other types of investment funds. For this reason, we simply use the most general equity market index possible (i.e., the CRSP value-weighted index) as the benchmark when calculating the active share of each institution.8 In untabulated results, we calculate the active share measure based on two alternative methods. In the first alternative method, the investment universe of the investor consists of the stocks held in the current portfolio. In addition, we determine the implied weights of the stocks in the portfolio based on the market weights of these stocks in the CRSP universe. Then, we employ the active share formula indicated in footnote 7 and calculate the active share. In the second alternative method, we employ a similar approach as Koijen and Yogo (Citation2019) and assume that the investment universe of an investor consists of the stocks that were held in the most recent 12 quarters (i.e., current quarter and past 11 quarters). Next, for each quarter, we calculate the proportion of stocks in the portfolio that are not in the investment universe of the investor and denote this proportion as active share. Using these two methods, we still find that average AS of short-term institutions is higher than that of long-term institutions. Similarly, hedge funds have higher AS values than non-hedge funds, and short-term hedge funds have higher AS values in comparison to short-term non-hedge funds.9 At the end of each quarter t, we calculate the HHI for an institution i as follows: HHIi,t=∑j=1J(Ni,j,tPi,j,t∑j=1JNi,j,tPi,j,t)2, where N and P represent the number of shares of stock j held by investor i and share price at the end of quarter t, respectively.10 Middle group includes middle three quintiles.11 We obtain a qualitatively and quantitatively similar alpha spread (0.38%, t-stat= 4.06) between short-term hedge funds’ heavy buy and heavy sell trades when we measure risk-adjusted returns using Fama and French’s (Citation2015) five-factor model. Similarly, we do not find a statistically significant five-factor alpha spread between short-term non–hedge funds’ heavy buy and heavy sell trades.12 As a robustness check, we replicate our value-weighted portfolio results with equal-weighted portfolios as well. Similar to our findings in Table 2, we see that the L/S portfolio for short-term hedge fund demand produces a monthly return of 0.37% (t-stat = 4.97) and a monthly four-factor alpha of 0.33% (t-stat = 4.19) in the subsequent quarter. In contrast, the trades of short-term non–hedge funds do not produce any statistically significant returns or alphas in the short-run. Furthermore, consistent with the prior literature and our findings in Table 2, neither the L/S portfolio for long-term hedge funds nor the L/S portfolio for long-term non–hedge funds generate statistically significant positive returns and alphas in the following quarter.13 Consistent with Yan and Zhang (Citation2009), we use two lags in computing Newey-West heteroscedasticity-adjusted t-statistics.14 A detailed description of the variables is provided in Table A1 of the appendix.15 As a robustness check, we also run the regression specification in Column 3 of Table 3 using weighted least-squares with each firm weighted by its natural logarithm of market capitalization (SIZE). We report our results from this regression in Column 2 of Table 4. We find that the average coefficient estimate on short-term hedge fund demand (ΔSIO_HF) is still positive and statistically significant (0.161 with a t-stat = 5.59) and the average coefficient estimate on short-term non–hedge fund demand (ΔSIO_NHF) is once again statistically not different from zero (0.017 with a t-stat = 0.67).16 For ease of comparison, we report original results from column 3 of Table 3 once again in column 1 of Table 4.17 In an alternative robustness test, following Lou (Citation2012), when we use total flows and regress the panel of each hedge fund’s trades on the total flows, we once again obtain similar results in the sense that our main findings are driven by informed trading of short-term hedge funds, not by persistent demand shocks and/or price pressure.18 We obtain similar results for the predictive power of short-term hedge funds’ trades when we conduct our portfolio tests for this restricted sample as well.19 We use median market capitalization and book-to-market ratio to split the sample into two.20 Following Cooper, Gutierrez, and Hameed (Citation2004), we define a down market as one where the previous 12-month cumulative market return is negative.21 We use four lags in computing Newey-West heteroscedasticity-adjusted t-statistics.22 When we exclude closed positions from the extensive short-term hedge fund demand, our findings remain qualitatively similar.Additional informationNotes on contributorsMustafa O. CaglayanMustafa O. Caglayan, is a Professor, Department of Finance, College of Business, Florida International University, Miami, FL.Umut CelikerUmut Celiker, is an Associate Professor, Department of Finance, Monte Ahuja College of Business, Cleveland State University, Cleveland, OH.Mete TepeMete Tepe, CFA, is a Vice President, Northern Trust Corporation, 50 S La Salle St, Chicago, IL.