DERIVATIVES, SHORT SELLING AND U.S. EQUITY AND BOND MUTUAL FUNDS. Current Version September 2014

Transcription

1 DERIVATIVES, SHORT SELLING AND U.S. EQUITY AND BOND MUTUAL FUNDS by Kaveh Moradi Dezfouli a and Lawrence Kryzanowski b Current Version September 2014 a Dezfouli is a Ph.D. Candidate, John Molson School of Business, Concordia University, 1455 De Maisonneuve Blvd W., Montreal, Quebec, Canada H3G 1M8, kmoradi@jmsb.concordia.ca b Kryzanowski (corresponding author) is a Professor and Senior Concordia University Research Chair in Finance, John Molson School of Business, Concordia University, Montreal, QC, Canada, H3G 1M8, +1 (514) , x2782, lawrence.kryzanowski@concordia.ca Financial support from Senior Concordia University Research Chair in Finance, IFM2 and SSHRC is gratefully acknowledged. The usual disclaimer applies.

2 DERIVATIVES, SHORT SELLING AND U.S. EQUITY AND BOND MUTUAL FUNDS ABSTRACT This paper uses a large unique database of U.S. open-end equity and bond funds using the intersection of data from all semi-annual N-SAR filings and Morningstar Direct for the 38, six-month periods from 1994 to We find that transaction-cost efficiencies are the primary determinants of both having and exercising permission to transact in derivatives and short selling for U.S. open-end equity and bond funds, and that both choices are strongly associated with fund and family sizes, and fund turnover. Our inferences are robust to the use of critical t-values that are adjusted for large sample sizes. Keywords: Mutual funds, derivatives, trade cost efficiencies. JEL Classification: G11, G23.

3 DERIVATIVES, SHORT SELLING AND U.S. EQUITY AND BOND MUTUAL FUNDS 1. INTRODUCTION The use of derivatives by mutual funds has increased since 1997 (Koski and Pontiff, 1999; Bae and Yi, 2008) due to the passage of the The Tax Payer Relief Act of 1997 that repealed the short-short rule. This rule discouraged funds from participating in short-term investments (including derivatives) by imposing higher tax rates on funds that earned more than 30% of their gross profits from short-term trading. 1 Derivative usage potentially has both costs and benefits for mutual fund investors. Fund investors may benefit when derivatives are used for risk management or to minimize transaction costs (Merton, 1995; Scholes, 1981; Deli and Varma, 2002). Fund investors may lose when the use of derivatives enables the managers to alter fund risk to serve manager-specific incentives and thus increase agency problems (Jensen and Meckling, 1976; Chevalier and Ellison, 1997; Deli and Varma, 2002). Deli and Varma (2002) examine the permission to invest in derivatives in the context of US mutual funds and conclude that transaction cost efficiencies are the main motivations for funds to obtain and exercise permission to transact in derivatives. We observe that the proportion of funds that are permitted to invest in derivatives has risen significantly in the past decade. While about 65% of the funds in the sample examined by Deli and Varma (2002) were permitted to transact in at least one type of derivative security, our sample shows a relatively higher proportion of funds (85.7%) have gained such permission. Deli and Varma (2002) examine the relation between permission to transact in derivatives and fund characteristics such as fund or family size (Deli and Varma, 2002). Nevertheless, only about 23% of the funds in our sample with such permission actually transact in derivatives. The determinants and implications of this significant difference between the proportion of funds that are allowed to and those that actually do is an interesting empirical question that has not yet been addressed in the literature. We conduct our tests on a large unique database of funds using the intersection of data from all the semi-annual N-SAR filings and Morningstar Direct for the 38 six-month periods from 1994 to Less than 1% (1.2%) of our sample report not having permission to invest in derivatives (having permission to short sell) after previously reporting that they have such permission. Furthermore, most of these funds report having permission to invest in the following report, which suggests a filing clerical error. 1 Our results are unchanged if the 172 observations from 1994 to 1997 are removed from the 125,000 observations in the total sample. 3

4 Our first objective is to study the possible determinants of the likelihoods of U.S. domestic equity and bond funds having and exercising permission to transact in derivatives. Many of the possible determinants are grounded in the theoretical benefits attributed to transacting in derivatives for mutual funds. We find that transaction-cost efficiencies are the primary determinants of not only having but also exercising permission to transact in derivatives for U.S. open-end equity and bond funds, and that both choices are strongly associated with fund and family sizes, and fund turnover. Thus, our first major finding is consistent with the hypotheses that scale economies and the capabilities of (only bond fund) advisors are important determinants of both of these choice decisions. The second objective of this paper is to extend the literature by testing if the likelihoods of U.S. domestic equity and bond funds having and exercising permission to transact in derivatives are related to the quality of their trade executions as proxied by brokerage fees. We find that these likelihoods are mostly higher for funds that pay lower levels of brokerage fees which is in support of the hypotheses that derivative transactions are used to achieve cost efficiencies in trade execution. Our study also examins the role that derivative transactions play in fund risk shifting (e.g., lottery-like and mid-year tournament behaviours), and the determinants of derivatives investment activity (relative and absolute) to gain further insight into the derivative strategies used by mutual funds. To do such, we will examine the dollar holding in options and in futures, and the impact of fund performance on these choice decisions and vice versa. The remainder of this paper is organized as follows. In the next section, we review the related literature. Section 3 describes our sample and data. We develop and report tests of hypotheses dealing with the likelihood of having and using permission to transact in derivatives and cost efficiencies in section 4 and trade execution performance in section 5. We develop and report tests of hypotheses dealing with the relation between fund performance and the use of derivatives in section 6. Section 7 concludes the paper. 2. REVIEW OF THE LITERATURE ON THE POSSIBLE BENEFITS OF MUTUAL FUND USAGE OF DERIVATIVES A mutual fund can use derivatives to improve performance by lowering its trade and holding costs and/or altering its level of risk. 4

5 2.1 Derivative Usage to Minimize Trade Costs, Liquidity-based Trading and Cash Holdings 2 The first potential source of cost efficiencies is the use of derivatives to achieve a particular risk level with potentially lower transaction or trade costs (Koski and Pontiff, 1999; Deli and Varma, 2002) and with a lower investment amount as compared to the case of solely relying on the underlying assets (Cao, Ghysels and Hatheway, 2011). The second potential source of cost efficiencies is derivative usage to minimize cash holdings in order to avoid liquidity motivated trading. Positive (negative) liquidity shocks resulting from high (low) flows increase (decrease) the cash holdings of a fund and can cause the manager to adjust the portfolio to bring it back to the target level of risk (Bloomfield, Leftwich, and Long, 1977). After a period of superior performance, a fund is likely to receive a higher inflow which reduces the fund s targeted risk exposure and imposes an opportunity cost of holding excessive cash (Deli and Varma, 2002). Conversely, poor performance periods are likely to increase the outflows and deviations from targeted risk levels and force the fund to use cash reserves, liquidate holdings or borrow to meet redemptions (Koski and Pontiff, 1999). Since these liquidity motivated trades are not initiated by information, they are considered costly for investors as in equilibrium informed traders benefit at the expense of liquidity traders (Grossman and Stiglitz, 1980; Chordia, 1996). Although funds may respond slowly to cash-flow shocks after periods of poor or superior performance (Koski and Pontiff, 1999), a fund s adjustment of its cash flow holdings is mainly by trading. Hence, this imposes a significant indirect cost to the fund due to this liquidity-motivated trading (Edelen, 1999; Nanda, Narayanan and Warther, 2000; Alexander, Cici and Gibson, 2006). Derivative use can potentially make these issues less severe and less costly. For example, excessive cash holdings can be invested in derivatives whose underlying assets are similar to the fund s portfolio holdings (Deli and Varma, 2002) to minimize opportunity and trade costs (Cao, Ghysels and Hatheway 2011). Using a sample of hedge funds, Chen (2011) shows that derivative users with low cash holdings are less likely to liquidate holdings in bearish markets. Frino, Lepone and Wong (2009) report lower returns for non-derivative users that experience shocks to their fund flows. 2.2 Managerial Incentives and Derivative Usage to Change a Fund s Risk Since mutual fund advisors are generally compensated as a percentage of assets under management (AUM), advisors benefit from an increase in a fund s net flows which are affected by its performance. Money is transferred into funds that recently showed superior performance (Ippolito, 1992; Sirri and Tufanno, 1998), and even when the superior performance does not persist (Berk and Green, 2004). 2 This is based on the discussion in Deli and Varma (2002). 5

6 Obviously, both investors and advisors of high performing funds are better off but the superior performance of a fund may result from excessive risk taking that differs from the risk preferences of its shareholders. While fund managers are expected to maintain a certain level of risk previously communicated to shareholders and to abide by other restrictions while managing the fund s portfolio (Almazan, Brown, Carlson and Chapman, 2004), managers are likely to deviate from this communicated risk level if their incentives are not aligned with those of the fund investors. 3 If a manager of an underperforming fund fears job loss (Kempf, Ruenzi and Thiele, 2009) or lower fees due to lower fund flows (Edelen, 1999), the manager may decide to increase the risk level by using derivatives or investing in lottery-like stocks (Kumar, 2009). In some extreme cases, managers may even decide to go for broke (Deli and Varma, 2002). Various studies find that managers of poorly performing funds tend to increase the fund s risk level in order to improve its already poor performance (Brown, Harlow, and Starks, 1996; Chevalier and Ellison, 1997; Koski and Pontiff, 1999; Cao, Ghysels and Hatheway, 2011). In contrast, a manager of a good performing fund may decide to play it safe by lowering the fund s risk level in order to avoid any potential losses that may be detrimental to the fund s performance to date. Thus, a better performing fund may simply avoid certain investment opportunities as being too risky for its managers but suitable for its shareholders. Consistent with these conjectures, managers of underperforming funds demonstrate tournament behaviour by increasing mid-year portfolio risk (Edelen, 1999; Taylor, 2003; Goetzmann, Ingersoll, Spiegel and Welch, 2007; Elton et al., 2010; Schwarz, 2012; Huang, Sialm and Zhang, 2011). This risk adjustment even occurs within fund families where managers compete to win the within-family tournament (Kempf and Ruenzi, 2008). This self-serving, risk-changing behavior by managers also occurs for hedge funds (Aragon and Nanda, 2011). Although derivatives are widely used for managing risk and market timing by mutual funds (Koski and Pontiff, 1999; Bae and Yi, 2008), derivatives are also an ideal instrument for high-risk speculation with high potential rewards (Aragon and Martin, 2012) without necessarily turning over a considerable portion of the fund s portfolio (Brown, Harlow, and Starks, 1996). Koski and Pontiff (1999) and Deli and Varma (2002) also consider the possibility of derivative usage to alter fund risk. Benson, Faff and Nowland (2007) identify different risk-shifting patterns for funds based on derivative usage. Using a sample of NSAR filings, Cao, Ghysels and Hatheway (2011) report that specialized equity funds and global funds, unlike funds in general, do use derivatives extensively, and that the timing of their derivative usage is based on previous returns in order to increase fund risk. 3 Sensoy (2009) finds a mismatch between the benchmark selected and fund characteristics for almost a third of all U.S. equity fund managers. 6

7 3. SAMPLE AND DATA 3.1 Sample Selection Our data from the Semi-Annual reports filed with the SEC (referred to as N-SAR forms hereafter) and the Morningstar database are for all US open-end funds from 1994 (first NSAR filing date) to This resulted in the collection of data for 55,448 N-SAR filings for 1,626 registrants with unique CIK identifiers. N-SAR forms are filed at the registrant level for individual funds or for a series of funds and contain information in item 7 at the series level (i.e., aggregated over all the classes of shares for the same fund). To obtain the intersection of the two databases, we matched by name since N-SAR filings only identify the funds in each filing by their names at the time of filing while Morningstar only reports the most recent name for a fund. To control for name changes, we followed all the funds through their lifetime and recorded all the possible name changes. This involved an extensive search for SEC filings that report name changes for the funds in the N-SAR filings. The Morningstar database has information for 12,601 funds once aggregated over classes of shares for each fund or for 50,762 separate share classes. For our final sample, we identified and matched 10,308 funds aggregated over classes or 39,577 share classes between the N-SAR filings and the Morningstar database. In other words, our sample includes 82% of the funds aggregated over share classes or 78% of the share classes available in Morningstar. The sample begins with 228,074 fund reports, where a fund report is a semi-annual filing of an individual fund that may be reported individually or along with other funds (as a part of a series) by a registrant. We were able to match 179,202 of these fund reports with the funds in Morningstar. To ensure the accuracy of our hand matching, we used the method of Edelen, Evans, and Kadlec (2012). We compared the monthly purchase and redemption values reported in N-SAR with the corresponding values reported by Morningstar. A N-SAR filing should (and does) report all monthly purchase and redemption values for the six month period of the report so in total we have twelve values to use for our check. We only kept a N-SAR observation if all the twelve values reported therein was matched to the last digit of the values reported by Morningstar. Our choice to only keep the observations with twelve exact matches may seem overly conservative. However, we argue that such a choice is necessary for two reasons. First, we aim to reduce the likelihood of any mismatch. Second, N-SAR filings typically contain erroneous data points particularly in value items, as many items are to be reported in thousands of dollars while some filers mistakenly report these values in dollars or millions of dollars (Christoffersen, Evans, and Musto, 2013). If we assume sales and repurchases entries represent reporting accuracy, then we are only keeping the observations that do not even have slight deviations. This step 7

8 reduced our sample size to 125,599 fund-reports. Furthermore, to ensure outlier values are not included in our sample, we winsorized the top and bottom 1% of the sample based on size and turnover values. Correlations between size and turnover values from the N-SAR and Morningstar databases are 98.31% and 88.83%, respectively. It is worth noting that minor discrepancies between the information available in Morningstar and that extracted from N-SAR filings may not necessarily mean an error in matching or reporting. Funds generally provide information to Morningstar earlier than they do to the SEC using the N-SAR form, since the latter is only filed every six months. This enables the funds to update the information for SEC submissions and report more accurate values. Moreover, N-SAR contains audited information whereas funds may not necessarily update the information submitted to Morningstar to control for updates or audits. Consequently, we expect minor discrepancies between entries in Morningstar and N-SAR. 3.2 Data Extraction Deli and Varma (2002) investigate the effects of a fund s permission to transact in derivatives. We build on their results by also studying whether funds did transact in derivatives and their determinants by examining N-SAR item 70. Items 70B through 70H contain the answers to the questions of whether a fund was permitted to transact in a certain group of derivatives and If permitted by investment policies, [the fund] engaged in [such transactions] during the reporting period. The derivative categories listed in the N-SAR form are: 70B, Writing or investing in options on equities; 70C, Writing or investing in options on debt securities; 70D, Writing or investing in options on stock indices; 70E, Writing or investing in interest rate futures; 70F, Writing or investing in stock index futures; 70G, Writing or investing in options on futures; and 70H, Writing or investing in options on stock index futures. We are able to identify if a fund primarily invests in domestic or foreign equities using N-SAR item 68B, its investment objectives using items 62, 66, and 69 of form N-SAR. A bond fund is identified as one that primarily invests in debt securities (N-SAR item 62A). We classify the investment objective of a bond fund as the type of debt security that has the largest value in its holdings using N-SAR item 62 (as in Deli and Varma, 2002). An equity fund is one that usually invests in equity securities (N-SAR item 66A) and an equity (bond) index fund is identified in N-SAR item 69 and 66 (62). A small number of the funds in our sample (1,523 observations) primarily invest in debt securities according to N-SAR item 62A but usually invest in equity securities as well according to N-SAR item 66A. While these funds are placed in the sub-samples of both debt and equity funds, the results are robust to their exclusion. 3.3 Summary Statistics for Derivative Permission and Usage 8

9 As reported in Panel A of Table 1, our sample consists of 125,599 fund-report observations: 85,627 equity funds, 35,561 bond funds, 6331 index-equity, and 670 index-bond funds. Domestic funds or funds that primarily invest in the US account for 105,791 (84.23%) of the funds. Foreign funds only account for 16,659 (13.26%) of our sample. Since the foreign funds are located in the United States, we include these funds in our study to provide better comparability of our results with those of Deli and Varma (2002). [Please insert table 1 about here.] A substantial majority of the funds are allowed to transact in derivatives, and as we subsequently show this percentage has increased significantly in the past decade when compared to the percentages reported by Deli and Varma (2002). While most funds have permission to transact in derivatives, a relatively smaller percentage of the funds engage in such transactions. Specifically, 107,636 (85.70% of our) sample funds were allowed to transact in at least one type of derivative and only 28,958 (23.06% of our) funds did so. Compared to equity funds, a greater percentage of bond funds transact in at least one type of derivative (31.84% versus 20.73%, respectively). Of the 5,912 (93.38% of the) index-equity funds that are permitted to transact in derivatives, the majority of 3,475 (54.89% of the) index-equity funds do. Their main usage is of equity derivatives for index-equity replication strategies. Equity and bond funds favor investing in same-asset-class derivatives with 17,174 (20.05% of the) equity funds investing in equity derivatives and only 1,908 (2.23%) investing in debt derivatives. Similarly, 10,325 (29.03% of the) bond funds transact in debt derivatives and only 3,765 (10.59%) transact in equity derivatives. The annual progression of the total number of funds and the relative proportions of the funds in our four samples that had or exercised permission to transact in derivatives are reported in table 2. We observe that the proportions are less than 0.1% for both the debt and equity samples from 1994 to 1998, and that the proportions with permission to transact in derivatives increased markedly from 1995 to Furthermore, we find that index-equity and index-bond funds are the most likely to have permission and to transact in derivatives. [Insert table 2 about here] The N-SAR form only reports investment amounts for two types of derivatives: Options on equities (item 74G of N-SAR form) and Options on all futures (item 74H of N-SAR form) in the balance sheet part of the form (item 74). Thus, assuming that the amounts reported in the balance sheets at the end of each report are a good proxy of derivative engagement, the proportions are obtained by adding up these two amounts for all the funds in a sample, and then dividing their sum by the total TNA of all the funds in that sample. While the annual proportions are quite low, they under represent the relative usage of 9

10 derivatives by the mutual funds. A dollar investment in derivatives is not comparable to a dollar investment in the underlying due to the leverage effect associated with the small margin associated with derivative transactions. Equity funds exhibit the highest proportions, and index-bond funds exhibit the lowest proportions between 2005 and 2012 inclusive. The most likely reason is that options on equities or futures are investment choices in line with the investment objectives of equity funds. Table 3 reports correlations between different pairs of derivatives types in the lower (upper) diagonal for which they have (exercise) the permission to transact. Transaction choices generally show lower correlations. Not surprising given their similarity, the three highest correlations for having permission to transact are for stock index futures with options on stock index futures (0.9299), equity options with stock index options (0.8994), and stock index options with options on stock index futures (0.8726). Furthermore, funds that transacted in Interest rate futures or Options on futures are very likely to have transacted in Options on debt securities, and funds that transacted in Options on futures are highly likely to have transacted in Interest rate futures. [Please insert table 3 about here.] 4. COST-RELATED DETERMINANTS OF LIKELIHOODS OF HAVING AND EXERCISING PERMISSIONS FOR TRANSACTIONS IN DERIVATIVES In this section, we first formulate hypotheses for the impact of three potential determinants for the likelihood of having and exercising permission by funds to transact in derivatives. We then test these hypotheses by estimating logistic regressions for samples of equity and bond mutual funds. 4.1 Hypotheses We expect size to have a positive effect on either having obtained permission to transact in derivatives or having transacted in derivatives. Transacting in derivatives requires a start-up cost and extra human capital from the advisor s side (Deli and Varma, 2002). Investors of larger funds are more likely to grant permission to transact in derivatives as the extra cost is spread over a greater investor base. Hence, we expect larger funds to be able to get the permission from their shareholders to transact in derivatives and to use their economies of scale to engage more frequently in derivatives transactions. Furthermore, funds that belong to larger families are able to use the human capital within the family which can make the investors more likely to be permitted to transact in derivatives. Managers of these funds are also more likely to choose to transact in derivatives if they obtain the support from their affiliated family. Thus, we test the following two hypotheses in their alternative forms: 10

11 H a 1 : The likelihood of having and exercising permission to transact in derivatives is greater for larger funds. H a 2 : The likelihood of having and exercising permission to transact in derivatives is greater for funds than belong to larger fund families. Edelen (1994) and Deli and Varma (2002) suggest that funds with better-informed advisors have higher portfolio turnovers. These funds find it more costly to trade based on liquidity needs so these funds tend to transact in derivatives more frequently to avoid liquidity motivated trading. Consequently, there should be a positive relationship between turnover and transacting in derivatives. We argue that this positive relationship exists for both having and exercising permission to transact in derivatives. Therefore, we test the following hypothesis in its alternative form: H a 3 : The likelihood of having and exercising permission to transact in derivatives is greater for funds with higher turnover ratios. Funds may use load fees to discourage redemptions (Chordia, 1996) or maintain a certain level of stability in cash flows. Hence, load fees may reduce the likelihood of forced trading because of liquidity needs. Consequently, load fees may facilitate a fund s objective to minimize the need for liquidity trading so that funds with higher load fees may not find it necessary to transact in derivatives. On the other hand, load fees may signal concerns about liquidity trading costs by advisors which can be a motivation for the advisors to ask for permission to transact in derivatives and to do so. Moreover, the level of load fees charged may play a role. Assuming a fund charges load fees, the amount of load fees collected by funds, if sufficient, may make it unnecessary for the manager to transact in derivatives. Thus, we test the following two hypotheses in their alternative forms: H a 4 : The likelihood of having and exercising permission to transact in derivatives is greater for funds with load fees. H a 5 : The likelihood of having and exercising permission to transact in derivatives is greater for funds with higher load fees. 4.2 Methodology We use the following logistics regression model to test the first three hypotheses: Derivatives m i,j,t = a + b 1 FNDSIZE i,j,t + b 2 RGSTSIZE j,t + b 3 TURNOVER i,j,t + b 4 FrontDum i,j,t 11

12 +b 5 BackDum i,j,t + b 6 FrontAmtPer i,j,t + b 7 BackAmtPer i,j,t + b 8 Rank6Month i,j,t + b 9 Rank6Month i,j,t 1 + b 10 NetFlow i,j,t + b 11 NetFlow i,j,t 1 + b 12 Perf. Flow i,j,t + b 12 Perf. Flow i,j,t 1 + k DebtStrat i,j,t,k + l EquityStrat i,j,t,l + ε i,j,t (1) Where, Derivatives m i,j,t is a dummy variable for fund i of registrant j in semi-annual period t that takes a value of 1 if it has characteristic m in that period and zero otherwise, where characteristic m either captures whether or not the fund has permission to transact in derivatives or whether or not the fund transacted in derivatives for that period; FNDSIZE i,j,t is the natural log of the size of fund i of registrant j in semi-annual period t in millions of dollars (item 74T of form N-SAR or Net assets of common shareholders ) to test if economies of scale affect a fund s likelihood of having or exercising permission to transact in derivatives. The top and bottom 1% of the values are winsorized to control for outliers. RGSTSIZE j,t is the natural log of the aggregated sizes in millions of dollars of all the funds reported by registrant j in semi-annual period t in one N-SAR filing, which is also used to test if economies of scale affect a fund s likelihood of having or exercising permission to transact in derivatives. This variable is obtained by aggregating all the reported Net assets of common shareholders (items 74Ts) in one deposited form N-SAR. When filing a N-SAR report, each registrant includes the information for all of its funds (series) that share a common year-end date in that filing. 4 Hence, aggregating fund sizes over a N-SAR file can be considered a relatively strong proxy for registrant size. While a registrant is not necessarily considered the family, registrant size is used as a proxy for family size in our study as we use family size as a measure of funds network and family support. TURNOVER i,j,t is the turnover ratio of fund i of registrant j in semi-annual period t in percentages from item 71D in form N-SAR to test if a fund s likelihood of having or exercising permission to transact in derivatives is higher for funds with greater turnover. This item is calculated using the lesser of purchases (item 71A) or sales (item 71B) divided by monthly average value of portfolio (item 71C) as calculated by Deli and Varma (2002), Christoffersen, Evans, and Musto (2013) and Edelen, Evans, and Kadlec (2012). The top and bottom 1% of the values are winsorized to control for outliers. FrontDum i,j,t is a dummy variable for fund i of registrant j in semi-annual period t that takes a value of one if a front-end sales load [was] deducted from any share sales during the reporting period (N registrant in our sample of 1,626 have more than one year-end date for their funds and hence are subject to multiple filings in every year. 12

13 SAR item 29) and zero otherwise. This variable is included to test if a fund s likelihood of having or exercising permission to transact in derivatives is related to whether or not a fund charged a front-end load during the semi-annual period. BackDum i,j,t is a dummy variable for fund i of registrant j in semi-annual period t, which takes a value of one if N-SAR item 34 reports an amount for deferred or contingent deferred sales load during the reporting period and zero otherwise. This variable is included to test if a fund s likelihood of having or exercising permission to transact in derivatives is related to whether or not a fund charges a backend load during the semi-annual period. FrontAmtPer i,j,t is the total front-end sales loads collected in thousands of dollars for fund i of registrant j in semi-annual period t as reported in N-SAR item 30A divided by total NAV of shares sold for fund i of registrant j in semi-annual period t as reported in N-SAR items 28G01 and 28G03. This variable represents the average front load charge in percentages and is included to test if a fund s likelihood of having or exercising permission to transact in derivatives is related to the relative size of front-end load fees during a semi-annual period. BackAmtPer i,j,t is the total deferred or contingent deferred sales loads in thousands of dollars for fund i of registrant j in semi-annual period t as reported in N-SAR item 35 divided by total NAV of shares redeemed and repurchased in thousands of dollars for fund i of registrant j in semi-annual period t as reported in N-SAR items 28G04. This variable represents the average back load charge in percentages and is included to test if a fund s likelihood of having or exercising permission to transact in derivatives is related to the relative size of back-end load fees during a semi-annual period. Rank6Month i,j,t is the six month after-tax return percentile rank within each Morningstar Category for fund i of registrant j in semi-annual period t where the best and worst performance ranks are 1 and 100, respectively. Morningstar ranks are one of the main factors used by Morningstar to rate funds (Guercio and Tkac, 2008), and are quite popular measures of performance for investors (Sharpe, 1998). Since Morningstar provides performance ranks for each share class, we use a weighted-average rank over the classes of shares to better capture differences in share-class mixes based on their size. This variable is included to test if a fund s likelihood of having or exercising permission to transact in derivatives is related to the short-term relative performance of the fund. NetFlow i,j,t is the net flow for fund i of registrant j in semi-annual period t as a percentage of TNA. Net flow is calculated as the difference between total inflows and total outflows divided by the TNA at the 13

14 beginning of semi-annual period t. 5 Total inflows are the sum of New Sales (item 28G01 from form N-SAR), Reinvestments (item 28G02 from form N-SAR), and Other Sales (item 28G03 from form N- SAR). Total Outflows are given by item 28G04 from form N-SAR or Total NAV of Shares Redeemed or Repurchased. Perf. Flow i,j,t is the interaction variable of Rank6Month with NetFlow for fund i of registrant j in semi-annual period t. DebtStrat i,j,t,k is a dummy variable for bond fund i of registrant j in semi-annual period t for investment strategy k that takes the value of one if investment strategy k has the highest percentage allocation in N-SAR and zero otherwise. The nine investment strategies considered for bond funds are U.S. Treasury - Short-term (N-SAR item 62B), U.S. Treasury - Long-term (N-SAR item 62M), U.S. Government Agency - Short-term (N-SAR item 62C), U.S. Government Agency - Long-term (N-SAR item 62N), State & Municipal Tax-Free - Short-term (N-SAR item 62E), State & Municipal Tax-Free - Long-term (N-SAR item 62O), Corporate Bond - Long-term (N-SAR item 62P), Commercial Paper (N-SAR item 62I), and Time Deposit (N-SAR item 62J), respectively. This variable is included to test if a fund s likelihood of having or exercising permission to transact in derivatives over the semi-annual period t is related to the primary investment strategy utilized by the bond fund over the semi-annual period. EquityStrat i,j,t,l is a dummy variable for fund i of registrant j in semi-annual period t for investment strategy l that take the value of one if investment strategy l has the highest percentage allocation in N- SAR and zero otherwise. The six investment strategies considered for equity funds are Aggressive Capital Appreciation (N-SAR item 66B), Capital Appreciation (N-SAR item 66C), Growth (N-SAR item 66D), Growth & Income (N-SAR item 66E), Income (N-SAR item 66F), and Total Return (N- SAR item 66G ). This variable is included to test if a fund s likelihood of having or exercising permission to transact in derivatives over the semi-annual period t is related to the primary investment strategy utilized by the equity fund over the semi-annual period. 4.3 Summary Statistics for the Independent Variables Summary statistics for equity and index-equity funds are presented in Panel A of Table 4, and for bond and index-bond funds in Panel B of Table 4. While 87.74% (73.35%) of the equity funds report having permission to transact in equity (debt) derivatives, only 20.06% (2.23%) report transacting in equity (debt) derivatives % (43.84%) of the equity fund observations indicate the existence of front (back) 5 According to Sirri and Tufano (1998), the flow measures are not sensitive to whether the TNA is measured at the end, beginning or over the period. 14

15 load fees which account for an average charge relative to fund TNA of 4.57% (4.16%). Our sample of equity funds follows various investment strategies % of our equity fund observations indicate Capital Appreciation as the main investment strategy followed by 24.3% indicating Growth. Bond funds have the second largest number of observations. While 83.37% (81.31%) of the bond funds report having permission to transact in debt (equity) derivatives, only 29.03% (10.59%) report transacting in derivatives % (55.5%) of the bond fund observations report the existence of front (back) load fees which account for an average 8.85% (16.68%) load fees relative to fund TNA. Of our bond fund observations, 11.40% and 9.45% indicate Long-Term Tax-Exempt and Long-Term Corporate bond as their main investment strategies. [Please insert table 4 about here.] With regard to index-equity funds, 92.58% (53.7%) report having permission to transact in equity (debt) derivatives, and only 53.17% (2.13%) report doing so. The percentages differ for index-bond funds where 83.28% (89.70%) have permission to transact in debt (equity) derivatives with 16.72% (15.07%) actually exercising such permission. For the two samples of index fund observations, 25.71% (26.48%) of those for index-equity funds indicate the existence of front (back) load fees which account for an average 0.45% (0.56%) front (back) load charge relative to their TNA, and 32.62% (31.95%) of those for indexbond funds indicate the existence of front (back) load fees which account for an average 2.37% (0.5%) front (back) load charge relative to their TNA. The two most prevalent investment strategies are Capital Appreciation (28.38%) and Growth (20.31%) for the index-equity funds, and Long-Term U.S Government Agencies (21.04%) and Long-Term U.S Treasuries (19.55%) for the index-bond funds. Based on untabulated results for the correlations between various potential determinants of having and exercising the permission to transact in derivatives, 6 fund and registrant sizes have significant correlations of , , , and for our samples of equity, bond, index-equity, and index-bond funds, respectively. 7 However, based on their Variance Inflation Factors (VIFs) and the correlation matrices of Estimated Variance-Covariance Matrix of Fitted Coefficients, both variables are retained in our regressions. Not surprisingly, the correlations for the dummy variables for front and back end loads are highly correlated at , , and for our equity, bond, index-equity and indexbond funds samples, respectively. Their VIFs also indicate that the inclusion of both of these variables is a likely source of multicollinearity. Similarly, for each of these four samples, the correlations for net flows with Rank-Flows are , , and , respectively, and for lagged net flows and Rank-Flows are , , and , respectively. Since they also have somewhat high 6 Appendices A1, A8, A15, and A22. 7 All untabulated results referred to in this paper are available from the authors. 15

16 VIFs (>4), only one of each highly correlated pair of independent variables is retained to minimize any impact from multicollinearity. Following Deli and Varma (2002), we include investment strategy dummy variables in our models to test if different investment objectives affect a fund s decisions with regard to derivatives. N-SAR form reports investment strategies for equity funds (item 66 of form N-SAR) and reports bond fund s percentage of net assets [invested] in each type [of debt security] investment. We assign a certain investment strategy to a bond fund if that particular strategy had the highest percentage of investment across all debt holdings (Deli and Varma, 2002). The strategy dummy variables are a possible source of multicollinearity. Based on the correlations and VIFs for the Growth and Capital Appreciation dummies for the equity funds, we drop the Capital Appreciation dummy variable to avoid multicollinearity. Similarly based on the VIFs, we drop S.T. U.S. Treasuries, S.T. Tax-Exempt, and Time Deposits in all six regressions for the sub-sample of bond funds. Before proceeding, we need to consider the impact of using samples that contain a large number of observations on drawing statistical inferences. Since confidence intervals becomes smaller with larger sample sizes, a null hypothesis is increasingly more likely (almost certain) to be rejected for an increasingly large (approaching infinity) sample size for a given level of significance (Leamer, 1978, Ch. 4; Shanken, 1987; Connolly, 1989). However, the posterior level of belief in the prescribed value (under the null hypothesis) would be close to certainty since the estimated value would be very close to the actual value as the sample size approaches infinity. Of the various approaches implicitly or explicitly used in the literature to address this paradox, we use more appropriate (lower) critical levels of significance for drawing inferences (namely, 0.05, 0.01 and 0.001) as in, for example, Aitken and Frino (1996) Empirical Findings The ordinary (binary-outcome) logistic regression results presented in Tables 5, 6, 7 and 8 for the samples of equity, debt, index-equity, and index-bond funds support our first five alternative hypotheses dealing with the relation between both derivatives permission and usage with fund size and turnover. 9 Based on Tables 5 and 6, equity and bond fund and registrant sizes are positively associated with the likelihood of having and exercising the permission to transact in all types of derivatives with one exception. Specifically, larger equity funds are less likely to transact in debt derivatives. Unlike equity 8 This is often referred to as Lindley's paradox. Other approaches include the use of (i) a large-sample posterior odds ratio drawn from Zellner (1984) and used by, e.g., Griffiths and White (1993) and Kryzanowski and Zhang (2002) to find the t-statistic needed to generate a posterior odds ratio of 20:1, and (ii) sample size-adjusted critical t-values drawn from Leamer (1978) and used by, e.g., Connolly (1989). 9 As expected, the untabulated results for the undifferentiated full sample are similar to those for the equity funds. All untabulated results are available from the authors. 16

17 funds, index-equity funds tend to be permitted and to transact less in equity derivatives with increasing size. Index-equity funds that belong to larger registrants on the other hand, are more likely to be permitted and transact in both types of derivatives probably because more expertise at lower cost due to scale economies are available for larger registrants and their affiliated funds.. Larger index-bond funds and index-funds that belong to larger registrants are more (less) likely to have permission (invest) to invest in debt derivatives and less likely to invest in equity derivatives [Please insert Tables 5, 6, 7, 8, and 9 about here.] The turnover ratios for both equity and debt funds are positively and significantly associated with the likelihood of a fund having and exercising the permission to transact in all types of derivatives, which is consistent with our third alternative hypothesis (Edelen 1994) with one exception. Equity fund with higher turnover are only more likely to have permission to invest in equity derivatives. In contrast, indexequity funds with higher turnovers are more (less) likely to be permitted to invest (transact) in equity derivatives. Moreover, high turnover index-equity funds are less (more) likely to be permitted to invest (transact) in debt derivatives. We find that most of the likelihoods of having and exercising permissions to transact in derivatives are significantly and positively related to the existence of a back-end load fee for both equity and bond funds (see Tables 5 and 6, respectively). However, index-equity and index-bond funds with such fees are less likely to be permitted to transact in debt derivatives and index-equity funds with back-end load fees are more likely to transact in debt and equity derivatives. 10 Equity and bond funds that charge higher front-end load fees are more likely to transact in both debt and equity derivatives. 4.5 Determinants of Likelihoods of Having and Exercising Permissions to Short Sell We use an alternative to model (1) that replaces the dependent variable Derivatives m i,j,t Short m i,j,t to test for the determinants of having and exercising the permission to short sell as reported in N-SAR item 70R. This new dummy variable for fund i of registrant j in semi-annual period t, where characteristic m either captures if the fund has permission to short sell or whether it is engaged in short selling for that period. The results reported in table 9 indicate that short-selling is more (less) likely to be permitted for larger bond and index-bond (equity) funds while larger equity and bond (index-equity) funds are more (less) likely to engage in short selling. Equity, bond, and index-equity (index-bond) funds that belong to larger registrants are more (less) likely to be permitted to short sell while equity and index-bond (bond) funds of a larger registrant are less (more) likely to engage in short selling. Existence of a back-end load fee with 10 Our results are not significantly altered if back-load dummy variable is replaced with a front-load dummy. 17

18 increases the likelihood to be permitted and to engage in short selling. Higher front-load fees increase the likelihood to be permitted to short sell for bond and index-bond funds, and also increase (decrease) the likelihood to engage in short selling for equity (index-equity) funds. Higher back-end load fees increase (decrease) the likelihood for bond (index-bond) funds to be permitted to short sell and decreases the likelihood of equity funds to engage in short-selling activities. 5. RELATION BETWEEN FUND PERFORMANCE, FLOWS AND DERIVATIVE PERMISSION AND USAGE 5.1 Hypotheses and Methodology Derivatives are widely used for managing risk and market timing by mutual funds (Koski and Pontiff, 1999; Bae and Yi, 2008). As derivatives are believed to be cheaper for maintaining or moving towards or remaining at a certain level of risk exposure, we expect that funds that use derivatives have better performance compared to non-derivative-user funds, all else held equal. This leads to our sixth hypothesis in its alternative form: H 6 a : The performances of mutual funds are higher if they have and exercise permission to transact in derivatives. Various studies find that managers of poorly (better) performing funds tend to increase (decrease) risk levels in order to improve (maintain) their current performances (Brown, Harlow, and Starks, 1996; Chevalier and Ellison, 1997; Koski and Pontiff, 1999; Cao, Ghysels and Hatheway, 2011). If derivatives are used for risk repositioning by fund managers as reported in these studies, performance and derivative usage should be related. We propose that managers of poorly performing funds choose to transact in derivatives to improve fund performance. Thus, our seventh hypothesis in its alternative form is: H 7 a : Funds are more likely to transact in derivatives following a period of poor performance. 5.2 Empirical Findings To test hypotheses six and seven, we use both rank and Lag rank variables or the fund performance in the current and previous 6-month period and the level of exercising permission to transact in derivatives. We find that better performing equity funds (funds with lower ranks) in the most recent sixmonth period are more likely to have transacted only in debt derivatives (see Tables 5 and 6). In contrast, poor (good) performing bond funds in the most recent six-month period are significantly more likely to transact in equity (debt) derivatives. Superior performing index-equity funds over the most recent (previous) six months are more likely to transact in equity (debt) derivatives. Index-bond funds with 18

19 superior performance in both six-month periods are more likely to have transacted in debt derivatives. Fund flows in both the current and previous six-month periods have no significant effect on having or exercising permission to engage in all types of derivatives for both equity and bond funds. While Evans, Ferreira, and Prado (2014) report a negative performance for funds engaging in short selling, we find that only bond funds that perform poorly in the previous six-month period are more likely to engage in short selling in the current six-month period. The increased likelihood of an index-equity (index-bond) fund having permission to transact and transacting in equity (bond) derivatives with lower net flows in the previous six-month period suggests that index-funds in particular use derivatives as an alternative to replicate indexes when they face lower cash flows. Index-equity (index-bond) funds transact in equity (debt) derivatives instead of buying the constituents in the index due to greater cost effectiveness when the change in net investable amount or net flow is low. 6. RELATIONSHIP BETWEEN AMOUNTS INVESTED IN DERIVATIVES AND FUND CHARACTERISTICS 6.1 Hypotheses So far, our study has mainly focused on funds having and exercising permissions to transact in derivatives. Our results identified transaction cost efficiencies as the main factor affecting these choice decisions, and that the likelihood of acquiring and exercising such permissions increases with increasing fund size. In this section, we conjecture that there are diminishing net benefits or a capacity constraint from increased participation in derivatives with fund size. Hence, we propose the following two hypotheses in their alternative forms: H a 8a : A fund s portfolio allocation to derivatives does not grow with its size. H a 8b : Derivative-using funds increase their relative investment allocation to derivatives following a period of poor performance. 6.2 Methodology and Empirical Findings Hypotheses seven and eight are tested using equation (1) where Derivatives% m i,j,t is the natural log of the sum invested in two types of derivatives (form N-SAR, Options on Equities and Options on all Futures ) by fund i of registrant j in semi-annual period t 19

20 divided by TNA of fund i of registrant j in semi-annual period t; and all the other variables are as defined in section 4. [Please insert Tables 10 and 11 about here.] Based on Table 10, we find support for hypothesis H a 8a since the portion of TNA invested in derivatives is lower with increasing fund size for the equity and index-equity funds, and with increasing registrant size for equity funds only. We also find mixed support for H a 8b. The portfolio weight of derivatives following a period of poor performance increases for equity and bond funds. Investment strategies also play a role in the amount invested in derivatives. Bond funds with a Long-Term U.S Government Agencies investment strategy invest a larger percentage of their TNA in derivatives but equity funds with Aggressive Capital Appreciation, Growth & Income, and Total Return strategies invest a smaller percentage of their TNA in derivatives. 6.3 Amount Invested in Short-selling Activities We replace the dependent variable Derivatives% m i,j,t with Short% m i,j,t in model (1) to identify the factors that affect the amounts that funds invest in short selling. This new variable is the natural log of the amount invested in short selling (form N-SAR item 74-R02, Other liabilities-short sales ) by fund i of registrant j in semi-annual period t divided by its corresponding TNA; and all the other variables are as defined in section 4. The results reported in table 11 are for equity and bond funds only as the number of observations are too small for the samples of index-equity and index-bond funds (22 and 12 respectively). We observe that larger equity and bond funds and bond funds belonging to larger registrants allocate a lower percentage of their TNA to short selling, and that short selling allocations increase with higher turnovers, poorer performance during the prior six months (equity funds only) and the existence of backend load fees (bond funds only). 7. RELATIONSHIP BETWEEN CASH HOLDINGS AND THE TWO DERIVATIVE CHOICES 7.1 Hypothesis and Methodology If periods of superior (poor) performance are followed by positive (negative) cash flows (Koski and Pontiff, 1999) which may shift a fund s risk to an undesired level, then mutual funds can be expected to respond to excess (insufficient) cash holdings by trading. This can result in an indirect cost from liquiditymotivated trading (Edelen, 1999; Nanda, Narayanan and Warther, 2000; Alexander, Cici and Gibson, 20