The impact of options use on mutual fund performance

Transcription

1 The impact of options use on mutual fund performance MARKUS NATTER 1, MARTIN ROHLEDER 2, DOMINIK SCHULTE 3, and MARCO WILKENS 4 University of Augsburg November 18, 2014 Abstract. Using a unique sample of 2,576 actively managed domestic equity funds from CRSP matched with information on options use available in mutual fund N-SAR filings with the SEC, we document superior performance and less risk of mutual funds using options. The annual risk-adjusted return of funds using options is 0.48 percentage points higher compared to their non-using peers. This is not due to the mechanical impact of options on measured performance but stems from superior skill of option users. We reveal that option users mainly employ covered call and protective put strategies to lower their market risk resulting in market betas which are 10 percentage points smaller than those of nonusers. JEL Classification: G11, G12 Keywords: Mutual fund performance, options. * We are grateful for helpful comments and suggestions by Matthias Bank, Oliver Entrop, Hendrik Scholz, and participants of the 2014 International doctoral seminar at the University of Innsbruck. We are grateful for financial support by the Research Center Global Business Management of the University of Augsburg. All remaining errors are our own. 1 Markus Natter, University of Augsburg, Chair of Finance and Banking, Universitaetsstr. 16, D Augsburg, Germany, phone: , markus.natter@wiwi.uni-augsburg.de. 2 Martin Rohleder, University of Augsburg, Chair of Finance and Banking, Universitaetsstr. 16, D Augsburg, Germany, phone: , martin.rohleder@wiwi.uni-augsburg.de. (Corresponding author) 3 Dominik Schulte, University of Augsburg, Chair of Finance and Banking, Universitaetsstr. 16, D Augsburg, Germany, phone: , dominik.schulte@wiwi.uni-augsburg.de. 4 Marco Wilkens, University of Augsburg, Chair of Finance and Banking, Universitaetsstr. 16, D Augsburg, Germany, phone: , marco.wilkens@wiwi.uni-augsburg.de.

2 The impact of options use on mutual fund performance Abstract. Using a unique sample of 2,576 actively managed domestic equity funds from CRSP matched with information on options use available in mutual fund N-SAR filings with the SEC, we document superior performance and less risk of mutual funds using options. The annual risk-adjusted return of funds using options is 0.48 percentage points higher compared to their non-using peers. This is not due to the mechanical impact of options on measured performance but stems from superior skill of option users. We reveal that option users mainly employ covered call and protective put strategies to lower their market risk resulting in market betas which are 10 percentage points smaller than those of nonusers. JEL Classification: G11, G12 Keywords: Mutual fund performance, options.

3 1 Introduction and Literature Overview In 2013, the dollar volume of options traded on the Chicago Board Options Exchange (CBOE) reached an amount of over $560 billion. On the over-the-counter (OTC) market, over $4,207 billion were traded in This shows the importance of options, which has led to extensive research on their potential costs and benefits. One benefit of using options could be lower transaction costs. Merton (1995), for example, argues that an investment in options only causes a fraction of the costs that occur for an investment in the respective underlying. In addition, options can be used to alter the risk-return profiles of portfolios more easily (Merton et al., 1978 and 1982). On the one hand, investors can use options to hedge against sharp declines in stock prices with a protective put strategy instead of selling the underlying assets. On the other hand, risk can be increased due to the option s embedded leverage. Early on, Black (1975) argues that investors trading options may have superior information because the potential profit of investing in these derivatives is greater than that from buying or selling the respective underlying due to the embedded leverage. More recently, Cao et al. (2005) support this argument and state that informed investors are attracted by option markets, especially in times of asymmetric information. Pan and Poteshman (2006) show that information is priced more efficiently on option markets compared to the markets of the underlying assets. They also assign this phenomenon to better informed investors participating in option markets. Additionally, buying and writing options is complex so that investors using options are potentially more sophisticated, may have specific skills, and access to superior information (Cici and Palacios, 2014). Furthermore, selling options can generate a steady income due to the earned option premia. Drawbacks of options are, for example, higher risks due to the leverage inherent in options. The collapses of Baring s Bank and Long Term Capital Management show that investing in options may lead to large losses. Furthermore, option use might also increase 1 See CBOE (2013) Market Statistics and BIS (2014). 1

4 costs as it requires complex risk management systems (Lynch-Koski and Pontiff, 1999). Another drawback, especially associated with option usage of mutual funds, might be based on the fact that options are complex instruments and require more sophisticated fund managers. These fund managers might charge higher salaries and increase overall costs of a fund (Chevalier and Ellison, 1999). Furthermore, Bollen and Whaley (2004) argue that due to buy pressure on put options by portfolio insurers these hedging instruments may be mispriced. If securities are not fairly priced, costs may arise and diminish returns of investors trading them. In light of these more general findings, there are several studies investigating how different investors use options. Bauer et al. (2009) show that private investors trading on option markets perform worse than investors who solely trade stocks. The authors attribute this underperformance to a lack of information and to the absence of timing skills. For hedge funds, Aragon and Martin (2012) as well as Chen (2011) find superior performance of managers using options and other derivatives, respectively. They show that hedge fund managers can exploit the potentially more efficient information pricing on options markets and generate higher performance as well as lower risk. The literature on mutual funds use of options so far has not offered clear evidence of benefits or limitations associated with these securities. Lynch-Koski and Pontiff (1999) were the first to examine mutual fund managers who use derivatives. They find no significant differences in performance and risk characteristics of users and nonusers. Their study is based on a telephone survey of a small sample of funds for the short period from 1992 to 1994, which might not be representative. Furthermore, in the 20 years since 1994 the capital market experienced dramatic growth and saw some major booms and crises to which regulators reacted with changes in the institutional framework. In addition, the authors admit that managers answers to the survey proved not to be reliable. Cao et al. (2011) find significantly higher raw returns of heavy derivative users, especially during the Russia crisis of August However, they measure performance without any risk-adjustment. Therefore, conclusions about hedging or speculative practices are not feasible. Furthermore, as the 2

5 Russia crisis is limited to only one month, their results are hardly representative for a general positive effect of derivatives use on fund performance. In a recent working paper, Cici and Palacios (2014) examine the impact of options on mutual fund performance using holdings data on individual exchange-traded option positions. They cannot assert significant differences between users of options and nonusers except for funds that excessively write puts. However, their results might be influenced by some limitations that are implicated by the methodology and data used. First, window dressing concerns could lead to an underestimation of the option usage amount as managers may decrease their holdings in high-risk securities prior to reporting in order to make their portfolios appear less risky (Musto, 1997 and 1999; Morey and O Neal, 2006). Second, they only focus on exchange-traded options although OTC-markets are more important for option trading than regular exchanges. Third, they have to rely on string searching algorithms to identify option positions. Due to these concerns, Cici and Palacios (2014) only identify 250 (10% of their sample) funds as option users whereas we identify 612 funds (24% of our sample) as option users (compare Table II). In contrast to previous studies, we employ actual mutual fund option use information from mandatory N-SAR filings obtained from the EDGAR database of the SEC. We match the N-SAR filings with CRSP Survivor-Bias-Free Mutual Fund Database resulting in a unique sample of 2,576 actively managed U.S. domestic equity funds during the period 1998 to To the best of our knowledge, this is the most comprehensive N-SAR/CRSP-dataset using regulatory information on fund options use to date. We extend the existing literature in several ways. First, to the best of our knowledge, we are the first to actually find significant cross-sectional performance differences between mutual funds that use options and their peers that do not use options. Specifically, funds that use at least one option of some kind during their existence outperform their non-using peers by 0.48 percentage points p.a. on average. The effect is economically and statistically significant. 3

6 Second, we contribute to the literature on fund manager skill by showing that the superior performance of option users is not purely mechanical but the result of managerial ability. In contrast to most existing studies on funds option use, we are able to integrate the time dimension in our analyses. We demonstrate that the outperformance of option using funds does not only exist during months in which a fund actually employs options. Rather, superior performance is also observable when a user fund decides not to employ options. Option users gain an outperformance of 0.72 percentage points during months in which they employ options and 0.6 percentage points p.a. in months when they tactically choose not to use options. This outperformance, even in times of nonuse, implies superior skill by option user funds. Third, in analyzing the outperformance of option user funds, we control for possible misspecifications of CAPM-based performance models when evaluating portfolios containing options. Since option-specific nonlinear payoff profiles lead to asymmetric return distributions, standard performance measurement approaches can be biased or even manipulated. Therefore, we estimate alphas following Leland (1999) and Bawa and Lindenberg (1977) to account for possible biases in our findings. In addition, we are the first to estimate a 5-factor Carhart model including an investable option factor to control for mutual fund portfolios containing options. Our investable option factor is based on the CBOE S&P 500 BuyWrite Index (BXM). Fourth, we show significantly lower market risk for option users. These differences in systematic risk are observable only during months in which funds actually employ options. They are thus based on a direct mechanical effect contrary to the permanent performance enhancing effect described above. Fifth, we infer the option strategies employed by portfolio managers from their returns. We regress the market beta of each fund on dummies indicating long and short positions in options. The loadings for long as well as short positions are negative and significant, i.e. both positions lead to lower systematic risk. This implies that funds use 4

7 protective put and covered call strategies to hedge against tail outcomes and to earn option premia. Our results are robust to different performance evaluation. Since strategies involving options bear risk characteristics that potentially differ from other investment practices, we calculate appropriate risk measures and our findings are unaffected by these changes. Moreover, our findings stay qualitatively unchanged for different time periods. The remainder of this paper is organized as follows. Section 2 gives an introduction to the institutional environment of mutual funds options use. Section 3 develops our research hypotheses. Section 4 defines our methodology. Section 5 describes the dataset used in this study. Our main empirical results are reported in Section 6. Section 7 describes robustness checks and further tests. Section 8 concludes. 2 Institutional Framework Any mutual fund registered in the United States is regulated by the SEC. The specifications regarding funds options use are codified in the Securities Act of 1933 and the Investment Company Act of 1940 (ICA). According to Section 18(f) of the ICA, mutual funds are generally prohibited from obtaining any kind of leverage. Uncovered written options can bear unlimited downside risk and are thus understood as leverage. Mutual funds nevertheless have the permission to sell options if they fulfill the SEC s asset coverage requirement, i.e. if the fund s total net assets plus the options market value divided by the options market value is greater than 300%. Since written options always have to be covered, there are three ways to short options: (i) selling an option on an underlying asset the fund already owns, (ii) selling an option on an underlying asset, for which the fund already owns an offsetting option position, (iii) holding highly liquid assets, e.g. cash, treasuries, corporate bonds, or liquid stocks, covering the option s market value in a segregated account. Long option positions are limited in their downside risk and therefore not treated as leverage. Nevertheless, mutual funds have to disclose their options use to the SEC in their semiannual N-SAR filing. 5

8 3 Hypotheses In our study we examine four major hypotheses which we develop in reference to the previous literature. Existing evidence is unclear regarding the performance impact of options use. Lynch-Koski and Pontiff (1999) as well as Cici and Palacios (2014) do not find any significant relations between option use and mutual fund performance. Cao et al. (2011) find higher raw returns for heavy user funds, but only during the Russia Crisis of Hedge fund performance on the other hand benefits from derivatives use (Aragon and Martin, 2012; Chen, 2011). These contradictory findings may be a result of different effects options use may have on performance. On the one hand options may lead to decreased performance due to associated costs. Bollen and Whaley (2004) point out that increased buy orders on put options from portfolio insurers has led to an increase in put prices. If a fund employs options for hedging purposes its performance might be lower because portfolio insurance is overpriced. Furthermore, hedging against tail risk demands premium payments without any measurable benefit if the insured event does not take place within the performance measurement period. On the other hand, options allow funds to use leverage. The literature on option markets (Black, 1975; Cao et al., 2005; Pan and Poteshman, 2006) suggest that investors on option markets are more sophisticated and have access to superior information which may lead to higher performance. In connection with leverage, this superior information might further enable them to generate higher risk-adjusted returns. Furthermore, transaction costs savings may facilitate stock picking and market timing strategies. In summary, there are more arguments in favor of a positive performance effect, such that our first hypothesis is that option user funds generate higher alphas than nonusers ( performance hypothesis ). Apart from the performance effect, mutual funds may also employ options to mitigate certain risks by hedging. Hence, our second hypothesis is that option users have lower market betas than nonusers ( risk hypothesis ). Our third hypothesis is that the performance effect is a result of superior skill ( skill hypothesis ), rather than a purely mechanical effect resulting from nonlinearities and 6

9 asymmetries associated with option returns. This can be grounded in the fact that investing in options requires funds to be more sophisticated than funds that avoid trading these securities (Cao et al., 2005). However, differences in risk and return profiles between option using and non-using funds may be an artificial result of using inappropriate models to measure fund performance. Among others, Leland (1999), Lhabitant (2000), Whaley (2002), and Goetzmann et al. (2007) show that performance measures of portfolios can be biased or even manipulated by using options. To confirm our skill hypothesis, we control for these biases in order to show that fund managers operating on option markets are more skilled than their non-using peers. Prominent examples like the fall of Barings Bank or Long Term Capital Management demonstrate potential hazards associated with derivatives in general and options in particular. Thus, of special interest to both regulators and investors are the option strategies employed by mutual funds. Our risk hypothesis suggests lower market betas for option users. A possible explanation for this is that funds follow hedging or risk reducing strategies. Simple option based hedging strategies are either protective put or covered call strategies according to Merton et al. (1978 and 1982). Hence, we develop a new methodology based on freely available regulatory data and easy implementable beta estimation to test our option type hypothesis that mutual funds mainly follow protective put and covered call strategies. 4 Methodology 4.1 PERFORMANCE MODELS To analyze our hypotheses we measure fund performance and risk using four different models. Our baseline model to determine fund performance is Carhart s (1997) 4-factor model as it is the widest spread model in the literature to date. For each fund we run the following regression: ER i,t = α i,4f + β i,mkt ER Mkt,t + β i,smb SMB t + β i,hml HML t (1) + β i,mom MOM t + ε i,t 7

10 where ER i,t is the return of fund i in month t in excess of the risk-free rate. ER Mkt,t is the market excess return, SMB t is the size factor, HML t is the value factor (Fama and French, 1993), and MOM t is the momentum factor (Carhart, 1997), respectively. The main variables of interest are the funds risk-adjusted returns, α i,4f, and their market betas, β i,mkt. Among others, Leland (1999), Lhabitant (2000), Whaley (2002), and Goetzmann et al. (2007) argue that standard CAPM-based alpha measures may be biased when assessing the performance of portfolios containing options. This is due to nonlinear payoffs and the neglect of the return distributions higher moments by the CAPM. If portfolio managers are aware of this, they may use options to manipulate CAPM-based performance measures. Specifically, Leland (1999) argues that skewness is positively valued by the CAPM. A covered call strategy (and analogously a protective put strategy) generates negatively (positively) skewed return distributions as skewness is sold (bought). Consequentially, alpha is positively (negatively) biased for portfolios with these strategies. Like Leland (1999), we control for any higher moments in return distributions by using the following model to measure performance: α L,i = E(r i ) B L,i [E(r Mkt ) r f ] r f, (2) where: B L,i = cov[r i, (1 + r Mkt ) b ] cov[r Mkt, (1 + r Mkt ) b ] with b = ln[e(1 + r Mkt)] ln(1 + r f ) var[ln(1 + r Mkt )] Here α L,i is Leland s alpha, E(r i ) is the expected net return of fund i and E(r Mkt ) is the market return. Nonlinear payoffs are another problem occurring when evaluating the performance of portfolios with options. To account for nonlinearity, we augment Carhart s (1997) 4-factor model with the CBOE S&P 500 BuyWrite Index (BXM) factor. 2 The BXM, developed by Whaley (2002), replicates a hypothetical passive total return covered call strategy. 3 In particular, the strategy is long the S&P 500 market portfolio and sells one-month near-the- 2 Results are not affected if we use the CBOE S&P 500 PutWrite factor instead of the CBOE S&P 500 BuyWrite factor. 3 Data is available from the CBOE 8

11 money call options on the S&P 500 every month. Thus, it does not use theoretically derived option premiums but market prices of actually traded options. This controls for possible outor underperformance that might arise if an option user trades potentially unfairly priced options. Moreover, the return distribution of the BXM is skewed to the left as well as nonlinear and should thus account for biases appearing when measuring the performance of portfolios containing options. 4 The performance regression is as follows: ER i,t = α i,5f + β i,mkt ER Mkt,t + β i,smb SMB t + β i,hml HML t + β i,mom MOM t + (3) β i,bxm BXM t + ε i,t Here, the BXM factor is defined as the return of the BXM index from month t to month t + 1 less the risk-free rate and thus represents an investable option factor. Furthermore, symmetric CAPM-based performance models may also be inadequate because options generate asymmetric payoff profiles. Bawa and Lindenberg (1977) argue that when considering asymmetric returns downside risk is more relevant. Therefore we also use the semi-variance instead of the symmetric variance when measuring fund performance: α BL,i = E(r i ) B BL,i [E(r i ) r f ] r f, (4) where: B BL,i = cov[r i, r Mkt r Mkt < 0] var[r Mkt r Mkt < 0] Here, α BL,i is the Bawa and Lindenberg (1977) alpha for fund i. As the models by Leland (1999) and Bawa and Lindenberg (1977) only consider market risk, we orthogonalize the fund s return as well as the market return on the Carhart (1997) factors to control for size, book-to-market, and momentum effects VARIABLE DEFINITIONS Our main explanatory variable Option_user i is a dummy variable which equals one if a fund is a user fund and zero otherwise. In our ensuing analyses, a specific fund is classified as a user fund if it uses options of some kind at least once during its existence. To analyze the investment in options at a more granular level, we specify additional indicator variables. In order to test the effect of actual option employment in the respective months on performance 4 The skewness of the return distribution of the BXM is and its kurtosis is Our results are not affected if we use unorthogonalized returns. 9

12 and risk we define the dummy variable Using i,t. This variable equals unity if a fund uses options in the respective month and zero otherwise. As shown in Table II, option users invest in options only 40% of the time. Hence, using options may be a tactical decision. To capture this effect, we define two further dummy variables to be used in combination. Using_user i,t equals one if a user fund employs options in a certain month, and zero otherwise. It thus measures the direct effect of option employment. Non_using_user i,t equals one if a user fund does not use options in a specific month, and zero otherwise. This variable thus measures the impact of a fund s active tactical decision to not employ options in the respective month, although it generally uses options. In addition to the binary information whether funds use options in a specific period or not, we also define two further dummy variables using balance sheet data on option dollar amounts. This data allows us to assign net long and net short positions to funds in specific periods by summing up reported dollar amounts in long options on equities and on futures and then subtracting the dollar amount reported for written options. If the result of this calculation is positive, the Long i,t dummy equals one, and in all other cases zero. Likewise, the Short i,t dummy equals one if this calculation is negative, and zero otherwise. These variables enable us to differentiate between long and short option strategies. To control for other fund characteristics besides differences in option use, we follow the existing literature on fund performance determinants (e.g., Almazan et al., 2004; Ferreira et al., 2012). As control variables, we use manager tenure, fund size represented by the natural logarithm of total net assets, annual turnover, monthly expense ratios, a dummy indicating if the fund charges loads, fund age, the percentage of assets held in cash, and the implied net flows calculated from TNA. 10

13 5 Data 5.1 SAMPLE CONSTRUCTION The data used in our study stems from two different sources. We obtain information on mutual fund returns and other fund characteristics from the CRSP Survivor-Bias-Free Mutual Fund Database. Data on option usage is obtained from semiannual N-SAR filings in the SEC s EDGAR database. 6 These filings provide information about the permission and the actual usage of different types of options. Namely, options on stocks (item 70B), debt securities (item C), stock indices (item D), futures (item G), and options on stock index futures (item H). Hence our option usage variables are based on all of these option types. 7 Furthermore, balance sheet data on options positions is given in the form of the dollar amount of overall market values for purchased equity options (Item 74G) and options on futures (Item 74H). Additionally, we have data on the amount of written options (Item 74R3) so we are able to distinguish between long and short positions in option contracts. Altogether we download 129,318 individual N-SAR text filings. The filings are then processed into a formatted dataset. In order to obtain the final dataset, we match a fund s N- SAR filing to its entries in the CRSP mutual fund database. Since there is no identifier that classifies the funds uniquely, we employ algorithmic string matching techniques to match funds by their names. Many fund names in N-SAR filings are incorrect and have to be adjusted manually to be matched to CRSP fund names. Errors that occur either because of potentially false entries or erroneous matches are removed by several screening techniques which we describe in the Appendix. Table A in the Appendix shows no significant deviations of our matched sample from the complete CRSP sample of actively managed domestic equity funds with respect to major fund characteristics. N-SAR filings are available electronically since However, in 1997 the short-short-rule was repealed with the Taxpayer Relief Act which has facilitated the use of derivatives for mutual funds. Thus, we limit our sample period to 1998 to The data is available at 7 In unreported results we find that our results hold when only looking at equity options. 11

14 The fund data obtained from CRSP are at the share class level. We aggregate them to fund level by value-weighing with the respective TNA for each share class. TNA, age, and load variables are exempt from this procedure. Fund level TNA is defined as the sum of the individual TNA of each share class, the fund s age is the age of the longest existing share class, and the load variable contains the load information of the largest share class. We exclude funds before they first surpass the threshold of 5 million in TNA as in Fama and French (2010). 8 Since we estimate performance measures via regression analysis, we also exclude funds that exhibit less than 24 fund monthly observations in order to maintain reliable results. 9 The final sample consists of 2,576 actively managed domestic equity funds with 234,679 monthly data points. For performance models and further tests, we obtain data from Kenneth French s homepage 10 and from Thomson Reuters Datastream. 5.2 DESCRIPTIVE STATISTICS Table I displays descriptive statistics of fund characteristics. Funds have negative performance measured by Carhart s alpha as well as by the other performance measures. The funds in our sample have return distributions that are slightly skewed to the left and have fatter tails than the normal distribution. The funds have a mean age of 10.3 years and the fund managers mean tenure is 5.8 years. Overall, this is in line with the existing literature. [Insert Table I here.] Table II reports descriptive statistics on the permission to use and on the actual usage of options for the mutual funds in our sample based on Item 70 of the N-SAR filings. 94% of funds are allowed to purchase and write options but only a fraction of them actually makes 8 The results remain qualitatively the same for thresholds of 15 and 50 million in TNA. 9 The results stay qualitatively unchanged for 48 fund months as minimum sample size per fund. 10 We thank Kenneth R. French for providing the data at 12

15 use of this permission. 11 Only 24% of all funds use some kind of option at least once. This is in line with findings of Almazan et al. (2004) who discover that mutual funds fixate permissions in their fundamental investment policies to ensure the greatest possible scope for investment practices, regardless of their inclination to use these investment practices. The underlying securities of our options are mainly stocks and stock indexes. This is not surprising because our sample consists solely of equity funds and, consequently, fixed income options are not as relevant. Deli and Varma (2002) and Chen (2011) interpret the suitability of the options to the respective investment style as evidence that funds try to mitigate transaction costs by using derivatives. In the third and fourth column of Panel A in Table II, we report the average percentage of time a fund uses options. It is obvious that option using fund managers are not engaged in options all the time. Hence, the timing of investing in options seems to be a tactical decision by fund managers. This is further supported by Panel B of Table II. In 89% of all monthly fund observations (208,441 fund-months) funds are permitted to use at least one kind of option. However, options are actually used in only 9% of all observations (22,012 fund-months). [Insert Table II here.] Table III reports summary statistics on mutual fund characteristics for option users and nonusers separately. 612 of our 2,576 actively managed domestic equity funds use some kind of option at least once during their existence. Option users are bigger (1,172 vs. 866 million TNA), older (12.2 vs. 9.7 years), and have higher annual turnover (137% vs. 89%). In addition, user funds tend to have lower excess returns in comparison with nonusers (0.36% vs. 0.42%). This could be due to option premia paid if funds use hedging strategies to protect against declines in security prices (Cao et al., 2011). Moreover, this could also be due to the fact that option users charge higher fees as their annual expense ratios are 22 basis points larger than the expense ratios of non-using funds. This is consistent with the results of Lynch- 11 If funds that have permission to use options differ severely from those funds that are not allowed to use options our results may be spurious. However, in unreported analyses our results are not affected by looking only at those funds that have permission to use options. 13

16 Koski and Pontiff (1999) for mutual funds and Chen (2011) for hedge funds. Higher fees could also be charged to compensate for the costs associated with options. Another reason might be that more sophisticated fund managers, who have the ability to successfully invest in options, demand higher salaries than other managers. Using options does not impact the overall risk of an investment fund, measured as the standard deviation of returns. Skewness and kurtosis are marginally higher for users of options. The return distributions of user funds are less negatively skewed and have slightly fatter tails. Options thus do not seem to be used for speculative purposes since their engagement does not lead to higher risks. Statistics regarding the funds alphas reveal first evidence for our performance hypothesis, namely that option users have higher risk-adjusted performance than nonusers (-0.72% vs % p.a.). Another interesting point is the amount of systematic risk. In line with the findings of Chen (2011) for hedge funds, option users face lower market risk exposure than nonusers. This may be the result of successful hedging strategies achieved through the implementation of protective put strategies. [Insert Table III here.] To assess how the tactical decision of time-variable option employment relates to fund characteristics, Table IV shows descriptive statistics on cross-sectional observations of using users and non-using users. User funds exhibit insignificantly lower excess returns in times when they are actually invested in options. Moreover, using option user funds have higher turnover ratios than temporary non-using users. In other respects, using users and non-using users do not differ measurably from each other. This indicates that eventual effects arise from the funds themselves rather than mechanically from the option. [Insert Table IV here.] 14

17 6 Empirical Results 6.1 PERFORMANCE, RISK, AND OPTIONS USE To test our performance hypothesis, we run the following cross-sectional regression: α i = β 0 + β 1 Option_user i + Σ J j=1β j Controls j + ε i (5) where α i is defined as fund i s risk-adjusted performance measured with the four different models described in Section 4. The indicator variable Option_user i takes on the value one if a fund is classified as a user fund and zero otherwise. Since user and nonuser funds differ from each other regarding central fund characteristics (see Table III), we follow the existing literature and include these characteristics as control variables. Table V reports the results for our four different alpha specifications. [Insert Table V here.] The Option_user i dummy has a positive and significant influence on Carhart s (1997) 4- factor alpha supporting our performance hypothesis. If a fund uses options at least once during its existence, it offers superior risk-adjusted performance on average compared to a nonuser fund. Signs on our control variables are largely in line with the literature. More experienced fund managers and larger funds generate significantly higher alphas. Both turnover and management fees have a negative impact on fund performance in line with results of Carhart (1997). The coefficients for loads and for net fund flows have positive signs, although only the latter is statistically significant. Older funds have a slightly lower riskadjusted performance and funds that hold more cash have higher alphas which is in line with Simutin s (2013) findings. The finding of higher alphas for option users may be driven by several biases arising with CAPM-based performance measures if the portfolio under investigation contains options. Therefore, Column (2) of Table V shows results for Equation (5) where α i is measured with the Carhart (1997) 4-factor model extended with an investable option factor based on the CBOE S&P 500 BuyWrite index. Likewise, Column (3) shows results for Leland s (1999) 15

18 alpha as our performance measure and Column (4) displays results for the Bawa and Lindenberg (1977) alpha as the dependent variable. The signs and values of the control variables are unaltered. The loadings of the option user indicator variable are qualitatively the same for all model specifications. Thus, our findings are not driven by technical biases of the employed performance measures. To test our risk hypothesis, we run a second cross-sectional regression where the dependent variable β i, Mkt is defined as the systematic risk of fund i according to the four performance models: β i, Mkt = β 0 + β 1 Option_user i + Σ J j=1β j Controls j + ε i (6) In line with our risk hypothesis, Table VI shows a significantly lower market risk exposure for option users compared to nonusers. This means that user funds manage and hedge their market risk exposure. For the BXM option factor model and the specifications of Leland (1999) and Bawa and Lindenberg (1977), the findings are the same, although insignificantly so for the Bawa and Lindenberg alpha. In all cases, there is a negative relationship between option use and market risk. This implies that user funds on average employ options to hedge and to lower their market risk exposure. The control variables have the expected signs. Fund managers with a higher degree of experience exhibit lower market betas in line with Chevalier and Ellison (1999). Interestingly, funds with higher management fees seem to have significantly higher betas. This could be due to higher wages drawn by better educated managers who manage higher beta funds (Chevalier and Ellison, 1999). Loads and net flow are negatively correlated with market risk. The loadings of cash positions are negative as cash, by definition, creates no market risk exposure. [Insert Table VI here.] 16

19 6.2 OPTION EMPLOYMENT AND SKILL Our cross-sectional results indicate that funds investing in options at least once during their existence offer higher risk-adjusted performance accompanied by lower market risk. The determinants for these findings, however, still remain unknown. On the one hand, option users outperformance could be caused directly by the options. On the other hand, the higher risk-adjusted performance of user funds could arise from superior skills that option using fund managers may possess. To uncover the source of option users superior performance, we test our skill hypothesis by running the following pooled panel regression, where fund performance is explained by different time-variable option employment dummy variables: α i,t = β 0 + β 1 OPT i, t + Σ J j=1β j Controls i,j,t + ε (7) i,t Here, α i,t is the monthly performance in month t measured via Carhart s (1997) alpha, the Carhart (1997) 4-factor model augmented with the BXM factor, Bawa and Lindenberg s (1977) alpha, and Leland s (1999) alpha. The monthly performance is estimated using daily fund returns from the CRSP mutual fund database. 12 OPT i,t is either the Using i,t dummy, indicating a fund that uses options in the respective month or is split up into two tactical decisions a user fund can make. Using_user i,t is a dummy indicating a user fund actually that employs options in the respective month. Non_using_user i,t indicates user funds that tactically refrain from using options in the respective month. If there is a purely mechanical relation between options and performance only Using_user i,t should have a positive impact on fund alphas. If managers of user funds have superior skill we would also expect a positive 12 Dimson (1979) argues that including leaded and lagged factor returns in daily return regressions helps circumventing biases arising from infrequent trading. As this should not be the case when considering mutual fund returns, we refrain from using this approach in our main analysis. In additional analyses not reported in this paper our results are unaffected by using Dimson s (1979) approach to estimate our daily performance measures. To control for sensitivity of our results regarding the granularity of the return data, we also calculate monthly alphas with monthly returns via rolling window regressions for 12 and 36 months window, both overlapping and non-overlapping. Results stay qualitatively the same. 17

20 performance impact of Non_using_user i,t. The control variables are the same as in our crosssectional analysis. Additionally, we include investment style and time fixed effects. Panel A of Table VII shows that funds using options in a given month generate an outperformance of 0.6 percentage points per year. This result holds for all of our four performance measures. Thus, it confirms the previous findings from our cross-sectional regressions and is further evidence in favor of our performance hypothesis. [Insert Table VII here.] To gauge if the superior performance of option users is due to the mechanics of options or if it can be attributed to special skills of the manager, the next specification directly tests how option users perform in months when they do not invest in options. Panel B of Table VII shows a positive and highly significant coefficient for the Using_user i,t dummy analog to the Using i,t dummy in Panel A. Accordingly, a fund which uses an option in a given month offers superior performance. In addition, non-using users also perform better than nonusers as indicated by a significant and positive Non_using_user i,t dummy coefficient. For the other model specifications, the coefficients for Using i,t, Using_user i,t, and Non_using_user i,t are basically the same and still remaining significant except for Bawa and Lindenberg s (1977) alpha. Hence, the superior performance of user funds has its roots in specific selection or timing skills of the fund manager and only partly in the mechanical effect of asymmetric and nonlinear option returns. This lends support to our skill hypothesis. 18

21 Table VIII includes the indicator variables Long i,t and Short i,t to assess the impact of predominant net long and net short positions in options on performance. [Insert 19

22 Table VIII here.] The first three columns show that only short positions in options generate higher performance. However, this could be due to CAPM-based performance measurement models being inappropriate for assessing the performance of managed portfolios incorporating options. Leland (1999) shows that by writing call or put options, an investor sells skewness resulting in a negatively skewed return distribution. Measuring the performance of such a strategy with CAPM-based models yields a positive alpha. Likewise, long option positions bias CAPM-based measures downwards. To control for all higher moments of the return distribution resulting from option strategies, we run Equation (7) with Leland s (1999) alpha as the dependent variable. Interestingly, the results for the Long i,t dummy change measurably, such that long positions in options also have a positive impact on performance. This is explained by the fact that Leland s (1999) model corrects alpha in the opposite direction. Consequentially, the impact of short positions is now less pronounced but nevertheless positive and statistically significant. In summary, both long and short option positions lead to higher risk-adjusted performance. Thus, performance, in addition to the skill effect, is also influenced by a mechanical effect. 6.3 OPTION EMPLOYMENT, MARKET RISK, AND OPTION STRATEGIES To understand hedging strategies with options employed by mutual funds we test our option type hypothesis by estimating the following pooled panel regression. β i,t,mkt = β 0 + β 1 OPT i, t + Σ J j=1β j Controls i,j,t + ε (8) i,t The dependent variable is the systematic market risk β i,t,mkt of fund i in month t measured with our four performance models based on daily returns. The option use and control variables are as defined in Section 6.2. Table IX displays the results. The employment of options in a certain month leads to a beta that is about 10 percentage points lower than in months without an engagement in options as shown by Panel A. [Insert Table IX here.] 20

23 In line with our previous findings of a higher alpha for user funds regardless of the actual employment of options, the lower systematic risk of user funds may also come from special skills of option using fund managers. However, results in Panel B repeal this expectation. Using_user i,t has a negative and significant impact on market risk, i.e. the actual implementation of options lowers a fund s beta. The factor loading of Non_using_user i,t is also negative but very small and statistically indistinguishable from zero. Thus, option user funds do not have lower market risk exposure in general, but only in times of actual option employment. The risk decreasing effects follow mechanically from the option strategies and not from special knowledge of the fund s manager. To test our option type hypothesis that mutual funds mainly follow protective put and covered call strategies, Table X shows loadings on the Long i,t and Short i,t dummy variables. [Insert Table X here.] If a fund is predominantly long options, its market beta computed according to Carhart (1997) is significantly lower ( ). The results are the same for the other market betas. To achieve a lower systematic risk via purchased options, an investor has to be long puts. By purchasing a put an investor indirectly sells some fraction of the respective option s underlying. It is now logical to assume that mutual funds use protective put strategies as introduced by Merton et al. (1982). The coefficient on the Short i,t dummy shows that if a fund is mainly writing options, its beta is also significantly lower ( ). Again this means that investors must be short the option s underlying. In case of written options this is only possible if calls are sold. As the SEC requires all short positions in options to be covered, mutual funds can only employ covered calls. Thus, in addition to the usage of protective put strategies mutual funds also employ covered calls to earn option premia. This is in line with our option type hypothesis. In addition, untabulated statistics show that option users are net long in 19% of the months they are invested in options and net short in 26% of the months invested. This means that covered call strategies are of higher importance than 21

24 protective puts. Moreover, this is supported by Cici and Palacios (2014). Using holdings data, they find that funds mainly follow covered call and protective put strategies. 6.4 OPTION TYPE To further analyze the origins of user fund outperformance we analyze the impact of Option_user i, Using_user i,t and Non_using_user i,t on performance and systematic risk for options on individual options and index options, separately. Funds buy or sell options on individual stocks due to information regarding the underlying. Furthermore, skill is only prevalent in individual stock option markets. Index option markets, do not offer those attributes. Hence, we expect our results regarding skilled fund managers to be stronger for individual stock options than for index options. In order to test this intuition, we aggregate information on equity options and futures options into individual options. Stock index options and stock index futures options are combined into index option. [Insert Table XI here.] Table XI shows the results using the Carhart (1997) 4-factor model plus an investable option factor based on the CBOE BuyWrite index (BXM) to calculate fund performance and market risk. 13 In the cross-section both option types have the expected signs on fund performance and market risk. However, only the users of individual options have a significant positive performance impact. This is in line with individual option using managers having stock picking skill. In our panel analysis only the use of individual options has a positive relation to fund performance. Moreover, Non_using_user i,t, is only significant for individual options as well, again implying superior skill for users of individual stock options only. Market risk, however, is hedged with individual options and to a much larger degree with index options. Thus, users of individual fund options seem to be skilled whereas the users of index options mainly follow hedging strategies. 13 Untabulated results using performance and market risk measures from Carhart (1997), Leland (1999), and Bawa and Lindenberg (1977) confirm this finding. 22

25 7 Robustness To strengthen our findings we perform several robustness checks. For brevity, these results are not presented in the paper, but are available from the authors upon request. First, we run our analyses without control variables and using different combinations of control variables. Additionally we also use the log of family TNA to control for family size. The results remain unchanged. Second, as our sample covers a long time span, we split our data set into two separate sub-periods. The first samples cover the years from 1998 to 2004 and from 2005 to 2013, respectively. While qualitatively the same during both sub-periods, our findings are stronger in the earlier period and weaker during in the second. 7.1 FURTHER PERFORMANCE MEASURES In addition, we estimate fund performance and risk with the CAPM and the Fama and French (1993) 3-factor model as our baseline model. Our findings remain the same. To control for publicly available information, we also recalculate performance and risk measures using a conditional model with factors proposed by Ferson and Schadt (1996) and our conclusions remain unaffected. As superior performance could be driven by premiums for illiquid securities, we augment the Carhart (1997) 4-factor model with the market illiquidity factor developed by Pastor and Stambaugh (2003). The results remain the same. In a recent study, Cremers et al. (2013) introduce a new approach to measure performance with benchmarks including transaction costs. Employing this methodology to estimate fund performance does not alter our results. An additional way of controlling for biases in performance measurement of portfolios containing options due to nonlinear and convex returns is to estimate the CAPM-based performance models with a supplemental squared market factor following the approach of Treynor and Mazuy (1966). Although, they use this to measure market timing, squared market returns not only measure nonlinearities resulting from market timing, but from other sources, such as asymmetric payoff profiles, as well. The results using the squared market factor are similar to our main results. 23