The Financial Review. Tailored versus Mass Produced: Portfolio Managers Concurrently Managing Separately Managed Accounts and Mutual Funds

Transcription

1 Tailored versus Mass Produced: Portfolio Managers Concurrently Managing Separately Managed Accounts and Mutual Funds Journal: Manuscript ID FIRE R Manuscript Type: Paper Submitted for Review Keywords: separately managed accounts, mutual funds, conflicts of interest, favoritism, concurrent management

2 Page of Tailored versus Mass Produced: Portfolio Managers Concurrently Managing Separately Managed Accounts and Mutual Funds Fan Chen a, Li-Wen Chen, Hardy Johnson b, Sabuhi Sardarli c a Meririmack College Girard School of Business Turnpike Street North Andover, MA 0 () -0 chenf@merrimack.edu b Kansas State University College of Business Administration 0 Business Building, 0 Lovers Lane Manhattan, KS 0 () - bhjohnson@ksu.edu c Kansas State University College of Business Administration 0 Business Building, 0 Lovers Lane Manhattan, KS 0 () - ssardarli@ksu.edu The authors thank Srini Krishnamurthy, (the editor), Robert Van Ness (the former editor), Bonnie Van Ness (the former editor), three anonymous referees, the seminar participants at University of Mississippi, University of Memphis, Sacred Heart University, National Central University, National Yunlin University of Science and Technology, and the 0 FMA annual meeting for comments and suggestions. Inaccuracies that may have been introduced in subsequent revisions are the sole responsibility of the authors.

3 Page of Tailored versus Mass Produced: Portfolio Managers Concurrently Managing Separately Managed Accounts and Mutual Funds Abstract Using a matched sample of separately managed accounts (SMAs) and mutual funds (MFs) with the same portfolio manager and investment style, we find that concurrently managed MFs consistently underperform their SMAs counterparts and generate more negative return gaps. Fund characteristics and liquidity betas fail to fully explain the underperformance. An eventstudy analysis finds that the weights placed into top (bottom)-performing stocks increase for existing SMAs (MFs) and negative return gaps increase for the MFs after the onset of concurrent management. We find that higher compensation collected by SMA fund managers are associated with more unseen managerial actions which positively contribute to the SMA return gap. Our results suggest that when managers concurrently manage both SMAs and MFs, they favor SMA performance over their MF performance. JEL Classifications: G0; G; G; G Keywords: Separately managed account, Mutual funds, Concurrent management, Favoritism

4 Page of Introduction For institutional investors and qualified individual investors, separately managed accounts (SMAs) are an alternative investment vehicle to mutual funds (MFs). These two investment options are similar in that a professional manager is in charge of making investment decisions on behalf of their clients while receiving compensation from managing fund assets. While MF clients investable assets are pooled, in SMA arrangements the portfolio manager buys securities on behalf of the client and places them in an account directly owned by the client. Many financial institutions offer both types of these funds for their qualified clients as the managerial skill required for these funds are transferable. SMAs and MFs by the same fund manager differs. We examine whether performance of Our second purpose is to identify potential reasons for this performance difference, should we find that the difference exists. To answer these questions, we analyze investment managers who concurrently control at least one SMA and one MF with the same investment strategy. Previous studies that have analyzed the performance of SMAs and MFs have not effectively controlled for managerial skill because those studies typically have matched funds at the managing firm level. Our pairing methodology at the fund manager level allows us to control for managerial skill, in addition to any firm level effects. A single manager in charge of both types of funds has an interest to have both funds perform well and would be expected to make similar investment decisions for both types, should the fund styles and strategies be identical. In fact, we find that on average, percent of the fund Throughout this study, we loosely refer to both SMAs and MFs as funds since they are different types of investment funds offered by financial institutions. The Wall Street Journal (Salisbury, Ian, SMAs beat funds in 00 ) and Morningstar report SMAs outperform MFs. Morningstar also reports that SMAs outperformed MFs in of stock and bond markets in 00 and outperformed in of categories from 00 to 00.

5 Page of holdings of SMAs and MFs that are managed by the same fund manager are identical. However, a manager has higher incentive to have a better performing SMA rather than a better performing MF. The first incentive arises from the relative size of SMAs as compared to MFs. In our paired sample of concurrently managed funds, an average SMA is almost three times larger than an average MF. Additionally, management fees (as a percent of assets under management) charged by SMAs are usually higher than those by MFs. Therefore, concurrent managers are better compensated by their SMA portfolios in comparison to their MF portfolios resulting in an incentive to favor SMAs and enhance their performance over the performance of MFs. The second incentive for favoring SMAs performance over MFs is due to differences in sensitivities of fund flows to performance between retail and institutional investors. Investment minimums are much higher for SMAs than those for MFs ranging from $00,000 to $ million. As a result, SMAs mostly cater to institutional investors and high net worth individuals. Previous literature (Del Guercio and Tkac, 00) finds that institutional investors are more sensitive to poor performance and abandon poorly performing funds while retail investors sensitivity is much lower. Therefore, a manager is likely to be more concerned about performance of his SMA than his MF as poor performance will likely result in higher fund outflows from the SMA relative to the MF. The mechanism of managerial favoritism or preferential treatment for SMAs can arise from transactions related to both homogenous and non-homogenous assets. For example, if a manager is buying the same security for both their MF and SMA, not all transactions will be executed at the same cost due to the price impacts of trading in large blocks of shares. Therefore, the manager has some discretion in assigning these shares with different cost bases across his MF and SMA portfolios. If the portfolio manager indeed has an incentive to enhance

6 Page of his SMA performance, then lower cost shares would be allocated to SMA accounts while higher cost shares would be allocated to MF accounts. Non-homogenous assets, such as illiquid securities, may also be used to enhance the performance of the SMA portfolios. Another channel for managerial favoritism is through attention to performance. Although managers have an interest in having well performing funds, more time and attention is likely to be diverted to SMAs in comparison to MFs, especially given that MF investors are less likely to abandon the fund in case of poor performance. In other words, when time and attention are limited, these resources are likely to be diverted to SMA portfolios. There are two alternative explanations for the performance difference between SMAs and MFs. The first explanation is potential differences in portfolio liquidity. Because MF clients assets are collectively invested, unexpected redemptions force the manager to liquidate some of the fund holdings, which potentially could hurt fund performance (Alexander, Cici, and Gibson, 00). To minimize the effect, in addition to holding cash and cash equivalents, MF managers may decide to hold more liquid positions to minimize redemption issues. However, in SMAs, redemptions by investors do not create liquidation events because SMAs are not collectively invested. In other words, the liquidity risk in SMA s is borne by the investor, while in MFs it is borne by the fund. This may allow an SMA manager to hold relatively less liquid assets in SMA portfolios and earn additional returns in the form of liquidity premium, in comparison to MF portfolios. The second alternative explanation for performance differences between SMAs and MFs is the potential differences in risk exposures even if funds are designated to follow the same investment strategy. If common risk factor loadings for SMA and MF returns are different, it could help to explain why raw SMA and MF performance (unadjusted for risk) are not similar.

7 Page of We construct our matched sample from Morningstar Principia and CRSP Mutual Fund databases. We identify investment managers that concurrently have at least one SMA and one MF with the same investment objective and strategy. For each manager we create matching pairs of SMAs and MFs in the period of All concurrently managed funds are paired as long as funds investment objectives are same. The final sample includes, pairs of fundyears with unique SMAs and unique MFs. When a manager has an unbalanced number of SMAs and MFs, some funds are used multiple times to create the pairs. Although the median SMA in our sample is much larger than the median MF ($ million vs. $ million total net assets), our matching methodology effectively controls for managerial skills, management company policies, investment strategies, market conditions, and managerial freedom. Additionally, any concerns of survivorship bias are addressed by our matching methodology, as this bias would equally affect both SMAs and MFs of the same manager over the same time horizon. This is particularly important since SMAs report their performance on a voluntary basis. Our results indicate that SMAs outperform MFs of the same manager by - bps, annually, when performance is measured by both net returns (net of investment fees) and riskadjusted four-factor alpha returns. By analyzing fund holdings, we find evidence consistent with preferential treatment of SMAs over MFs by investment managers which, we conclude, is the reason differences in performance exist. Specifically, we find that return gaps of SMAs are statistically and economically higher than those of MFs. Returns gap (the difference between the returns of the fund and those of a hypothetical portfolio that invests in previously disclosed fund holdings) is the result of manager s unobserved actions and can serve as a measure for manager s stock picking talent, hidden costs (including trading, commissions, and agency costs), value

8 Page of added (or subtracted) and has predictive power over a fund s future performance. Since many of these factors are constant for SMAs and MFs of the same manager, our research design of concurrent management allows us to assess the differences between the return gaps of SMAs and MFs. This, in turn, captures unobserved actions that are due to preferential treatment of one fund over the other, resulting in performance shifting between these funds. We find that managers show favoritism to SMA portfolios by placing higher weights on top performing stocks in SMA portfolios and lower weights on bottom performing stocks relative to MF portfolios. Additionally, we find that when SMA managers take over management of a comparable MF, their returns gaps are significantly larger in the SMA portfolios. Similarly, when MF managers take over management of an SMA, the return gap of the SMA portfolio is larger than that of their MF. After the onset of concurrent management, managers do a much better job of allocating the assets in their SMAs than they do in their MFs, by overweighting the top performing stocks and underweighting the poorly performing stocks in the SMA portfolios. We also find that our conflict of interest proxy is related to the differences in management fees between SMAs and MFs, especially when there are loads imposed on the mutual funds. We conclude that concurrent portfolio managers show preferential treatment to their SMA portfolios over their MF portfolios. We believe this is due to portfolio managers being better compensated by SMAs. We show that unseen actions taken by portfolio managers increase as the dollar amount of compensation from SMAs increase. Secondarily, we believe SMA portfolios are preferred due to their fund flows having greater sensitivity to performance than MFs. See Kacperczyk, Sialm and Zheng (00) for more in-depth discussion of return gap.

9 Page of We find no support for two potential explanations of differences in returns-risk exposure and liquidity of holdings. Risk exposures of SMAs and MFs of concurrent management with same investment objective are not meaningfully different. Additionally, we find that percent of holdings held in SMAs are identical to their MF counterparts, while the remaining percent of SMA holdings are less liquid in comparison, differences in liquidity betas fail to explain SMA outperformance. This study contributes to the sparse literature on SMA performance and how they compare to other investment vehicles. Assets under management of SMAs have been growing with a steady pace. As of beginning of 0, SMA portfolios controlled about $0 billion in assets and with an expected annual growth rate of - percent. As these investment vehicles become more important, our study analyzes their performance and potential issues, especially when they are concurrently managed with mutual funds. To our knowledge, this is also the only study that investigates potential reasons behind SMA outperformance by analyzing a sample of funds matched by an investment manager. This matching methodology provides us with an opportunity to compare SMA and MF performance while removing the effects of firm characteristics, managerial skill, and risk exposure from the analysis. Finally, this study contributes to the growing literature on the wide array of agency problems in professional wealth management functions. Managed Accounts Research, Cerulli Associates, Inc. Available at getcerullifile?filecid=producttest An important exception is Del Guercio, Genc, and Tran (0) which find that mutual funds whose manager is concurrently managing other types of portfolios underperform its peers without concurrent management. This finding is stronger when managers have a greater percentage of their assets under management outside of the fund industry or when their mutual fund clientele is less performance sensitive. Agency issues in wealth management functions have been documented by Chevalier and Ellison (, ), Elton, Gruber, and Blake (00), Zitzewitz (00), Cohen and Schmidt (00), Pool, Sialm, and Stefanescu (0), Doellman and Sardarli (0), among others.

10 Page of Industry Background and Prior Literature SMAs and MFs have many similarities. In both arrangements, clients hire professional money managers to make buy and sell decisions. Both offerings may specialize by asset class, investment objectives, style, etc., and are used by individual and institutional investors. However, there are important differences between these fund types. MF clients investable assets are pooled and the manager makes investment decisions for the pool. Clients own the shares of this pool while the MFs own the shares of the holdings. In SMA arrangements, managers buy and sell individual securities on behalf of the client and place them in an account directly owned by the client. The client s assets are not pooled or commingled with other clients assets and each account can, to some extent, be customized to reflect the customer s tax or social considerations. All SMA accounts with the same objective follow the SMA s model portfolio as determined by its management. When an SMA manager makes an investment decision, he does so for the model portfolio and, through arrangements with the brokerages, trades are executed for all clients of the SMA. Another important difference between SMAs and MFs is the target client base. Because SMAs generally require higher investment minimums (from $00,000 to $ million), they are mainly marketed to institutional investors or qualified individual investors (Peterson, Iachini, and Lam, 0). Despite the growing importance and assets under management of SMAs (more than 0 percent increase in assets under management during the 00-0 period, per Cerulli Associates, Inc.), there is limited academic research on their performance. In a randomly matched SMA and MF sample, Elton, Gruber, and Blake (0) find that SMAs outperform MFs by bps annually. However, when they match the sample according to management company and investment objective, they do not find a significant difference in performance. They

11 Page 0 of conclude that firms that exclusively cater to SMA clients might be driving the SMA outperformance. Another study comparing performance of these fund types, Bhargava, Gallo, and Swanson (00), finds that international SMAs outperform international MFs net of management fees. The advantage disappears when performance is compared using gross returns. Several papers analyze whether SMA returns are persistent, which would be evidence of managerial stock picking skills. Busse, Goyal, and Wahal (00) only find weak evidence that SMA managers are skilled over longer time horizons. Peterson, Iachini, and Lam (0) find that managerial skill persists in shorter horizon performance. They also document that a large fund size, more cash inflows, and a high number of accounts under management hinder performance. Ideally, a comparative study of SMAs and MFs would greatly benefit from transaction level information from managers that are making buy and sell decisions in the portfolios. Such data would allow us to see whether a manager is favoring one fund type over the other. However, due to unavailability of transactional data in this study, we rely on measures that capture managers unobserved actions. More specifically, we use return gap (developed by Kacperczyk, Sialm, and Zheng (00)), which is calculated as the difference between the fund returns and returns of a hypothetical portfolio mimicking the previously disclosed fund holdings while controlling for estimated trading costs (Keim and Madhavan, ). This return gap could capture managerial skill or hidden costs, including agency costs for the managers. Wermers (000) uses a similar measure to decompose performance into stock-picking talent, style selection, and transaction fees. Because we create a sample of funds matched on the investment manager, a difference in the return gaps of SMAs and MFs is unlikely to be due to managerial skill. It is more likely to reflect other hidden actions, such as the favoritism towards SMAs.

12 Page of This paper closely relates to Gaspar, Massa, and Matos (00) who study whether mutual fund families exhibit favoritism among their fund offerings. 0 More specifically, they find that high value funds (funds with higher expense ratios and past performance) outperform at the expense of lower value funds (funds with lower expense ratios and performance). performance shifting mechanism is through opposite trades and allocation of IPO shares (i.e., non-homogenous assets) to high value funds. In the context of our study, we expect SMA funds to be comparable to high value funds as managers are likely to collect higher compensation in comparison to their MF portfolios. Additionally, Del Guercio and Tkac (00) document that institutional investors are much more sensitive to poor fund performance in comparison to retail investors. Since SMA funds mostly cater to institutional investors and qualified individual investors, we expect a manager to favor SMA performance over MF performance As discussed above, SMAs and MFs differ in terms of their liquidity requirement. MF managers need to hold securities that are more liquid in the fund portfolio. Otherwise, they might be forced to make liquidity based trades due to investor redemption requests. Conversely, SMA managers can hold less liquid assets, make valuation-based trades (trades based on convictions about the value of the stock) rather than liquidity-motivated trades, and earn a liquidity premium since there is no redemption risk assumed by the fund. Wagner and Winter (0) find that MF managers prefer more liquid stocks in their portfolios. Additionally, Cao, Simin, and Wang (0) document that MF managers have the ability to time the market by increasing market exposure in periods leading up to more liquid markets and decrease exposure in periods preceding less liquid markets. A manager s ability to time liquidity can deliver superior returns. This Alexander, Cici, and Gibson (00) find that valuation-based trades Concurrent management and the potential agency issues have been previously investigated by Chen and Chen (00), Cici, Gibson, and Moussawi (00), and Nohel, Wang, and Zheng. (00), among others.

13 Page of significantly outperform liquidity-motivated trades. Coggin, Fabozzi, and Rahman () document that institutional pension plans, which also do not face redemption issues, exhibit outperformance as a result. These results in the literature suggest that liquidity exposure could potentially explain the SMAs outperformance over MFs, as they are more likely to hold less liquid assets and their managers are able to make more valuation-based trades. Consistent with liquidity arguments, Chordia () argues that MFs impose loads to discourage share redemptions and thereby enables managers to hold less liquid assets. Ivkovic and Weisbenner (00) find that investors redemption decisions are affected by loads. Additionally, Ippolito () and Nanda, Narayanan, and Warther (000) demonstrate that funds with loads outperform no-load funds. These findings in the literature indicate loads could be imposed to decrease the effect of the liquidity exposure and to make the flow-performance relationship less sensitive. This implies the potential favoritism is more likely when a concurrent manager is in control of a MF with a load.. Sample and Variable Construction The primary databases we use are Morningstar Principia and CRSP Mutual Fund Database (also known as CRSP Survivorship-Bias Free Mutual Fund Database). For MFs, fund level information including fund returns, characteristics, investment objectives, and management fees are obtained from the CRSP Mutual Fund Database. Although CRSP includes MF holding data, it is reported on a quarterly basis. Instead, we use Morningstar Principia to get MF portfolio holding information where holdings are usually reported on a monthly basis. For SMAs, we obtain all data items from Morningstar Principia which reports SMA fund characteristics and monthly holdings. The CRSP Mutual Fund Database does not include any

14 Page of SMA related information. When data matching is necessary across the two databases, we use both CUSIP and ticker to make accurate matches. To control for managerial skill, we include only domestic equity portfolio managers who concurrently manage at least one MF and one SMA. Using the name of the lead manager, we create a sample by pairing SMAs and MFs whose investments are managed by the same portfolio manager and with same investment objective. In cases of team managed funds, we use the identification of lead manager to create the fund pairs. We match our fund pairs based on investment objectives because risk composition, management fees, and holdings are likely to be different across funds with different strategies. This sample design allows us to eliminate the effects of managerial skill, management firm characteristics, risk composition, and management fees as we compare the performance of SMAs and MFs. We also obtain the portfolio holding information for paired SMAs and MFs in the dataset. Holding weights are used to calculate returns on hypothetical portfolios that mimic the funds in the sample and return gaps between fund returns and hypothetical portfolio returns. The primary advantage to using the holding data from Morningstar is that the data contains a higher frequency (usually monthly) of holdings than conventional databases that collect SEC filings on a quarterly or semi-annual basis (Elton, Gruber, Blake, Krasny, and Ozelge, 00). This allows us to construct a more accurate measurement of a return gap that does not rely on quarterly fund holdings. Holdings data is also matched to CRSP to obtain holdings performance. Our main sample includes both single-managed and team-managed funds. For team managed funds, we identify the lead manager and match SMAs and MFs based on the lead manager. Results are robust when we drop teammanaged funds. For additional robustness, we also compare performance of SMAs and MFs with four sub-sample scenarios using net returns, risk-adjusted alpha, and return gaps: () SMA is single-managed and MF is singlemanaged, () SMA is single-managed and MF is team-managed, () SMA is team-managed and MF is singlemanaged, and () SMA is team-managed and MF is team managed. The differences on performance between SMAs and MFs are all statistically significant; thus, we rule out the argument that our results could be driven by single- or team-management issues.

15 Page of MFs usually have multiple share classes while SMAs do not. We aggregate MF share class level variables (fees and returns) at the fund level by creating weighted-averages based on total net asset (TNA) of the share classes. Our matched SMA and MF sample spans from 00 to Although we focus on this time period for the matched sample, we collect return, TNA, and fee data before and after this period in order to calculate risk-adjusted returns (i.e., alpha returns), liquidity betas, and post-performance (in the event study analysis). More specifically, for risk-adjusted alpha returns and liquidity betas, we use prior -month and 0- month windows, respectively. In the event study analysis section of the study, we also analyze the performance over a -month window of post-concurrent management. In all, data used in this study covers the period of The matched sample includes domestic equity funds with more than $ million in total net assets, inception dates older than months, and return data without lapses in reporting. one of the fund counterparts eventually dies we remove that SMA-MF pair from our sample to further address potential survivorship bias. Additionally, our matching methodology remedies concerns of survivorship bias as survivorship would affect both SMAs and MFs similarly since we only analyze funds for which we have concurrent data. We also drop observations with missing variables such as holdings, fund returns, and fund and holding characteristics. Our final 0 Our sample funds are restricted to start in 00 for two reasons. First, although we obtain holding data from Morningstar for mutual funds, we cross-check it against CRSP holding data when available even if CRSP holdings are reported quarterly. Unfortunately, CRSP holding data is only available starting 00 and in order to calculate liquidity beta, we need the holding data months prior, so we restrict the sample to 00 and beyond. Second, the number of concurrently managed SMA and MF pairs with same investment objective prior to 00 is significantly smaller. In unreported results, we also follow the procedures of Jagannathan, Malakhov, and Novikov (00) to control for survivorship bias by eliminating funds that do not report return series without lapses and funds younger than years. We also truncate skewness in the return data. This alternative sample drops an additional pairs of sample fund years to correct for potential survivorship bias. Results are virtually identical with this alternative sample and are available upon request. If

16 Page of sample includes, pairs of fund years with unique SMAs and unique MFs. sample contains pairs of fund years in the large-cap, pairs in the mid-cap, and pairs in the small-cap fund category. It also represents 0 pairs of fund years in the growth, pairs in the blend, and pairs in the value-fund category, based on Morningstar style box fund categories... Fund performance measures We use two measures of fund performance: net returns and alpha returns. Net returns are fund level annual gross returns minus investment fees. As mentioned earlier, in cases of MFs, fund level returns are aggregated at the fund level using share class level data. Alpha returns ( ) are risk adjusted returns based on Carhart s () four-factor model:, = , () where r i,t is the return of fund i in month t, is risk-free rate in month t, is the excess return of the market portfolio in month t, and are the returns of factors mimicking portfolios for market equity (Small Minus Big) and book-to-market-equity (High Minus Low) in month t, and ε i,t is the residual return of portfolio i at time t. The benchmark returns of RMRF, SMB, and HML are obtained from Kenneth French s data library. Momentum portfolios are based on the funds average return over the past months. We calculate alphas from a window of past months to one month prior to the onset of concurrent management and roll this window for the following periods. Although our sample covers a matched dataset of SMA-MF In cases where a portfolio manager does not have the identical number of SMAs and MFs, we create the matched pairs from his existing funds. For example, if the manager has two SMAs and one MF with same investment objective, MF is used twice to create two distinct SMA-MF pairs. But, only the first concurrent fund (SMA or MF) is used in in the event study analysis. The

17 Page of pairs in the period of 00 to 00, we obtain return data prior to that period in order calculate risk-adjusted returns... Measure of unobserved actions and agency issues Following Kacperczyk, Sialm, and Zheng (00), we calculate return gaps as the difference between observed fund returns and the hypothetical portfolio mimicking the fund s previously disclosed holdings:, =,, where r i,t is the return of fund i in month t,,, () is previously disclosed holdings weight for fund i for security j, and, the return of security j in fund portfolio i in month t. Return gap captures the investment manager s unobserved actions which include managerial skills, trading costs, and agency costs. We adjust this equation to trading costs as estimated by following Keim and Madhavan ():, =.0+0., +0.0, 0.0, +.0,, =0.+0.0, +0., 0.0, +., where, and, are trading costs (in percentage of the trade value) of a given purchase and sale transaction of stock i in quarter t;, is the dollar value of a trade divided by the market capitalization of the stock,, ;, is the stock price; and, is a dummy variable equal to if the stock is traded on NASDAQ, zero otherwise. Monthly execution costs are obtained by dividing the quarterly costs equally over the three months. Further, in our research design, managerial skill is constant as we analyze concurrently managed SMA and MF. Therefore, return gap differences capture unobserved actions that are (a) (b)

18 Page of due to preferential treatment of one vehicle over the other and is a measure of agency related issues in concurrent management... Liquidity Exposure Measure Following Liu s (00) liquidity-augmented CAPM, we estimate the liquidity betas for both fund and portfolio holding levels. Liu s (00) liquidity-augmented CAPM is:, = + + +, () where, is the return of fund (or portfolio holdings) i in month t, is risk-free rate in month t, is the excess return of the market portfolio in month t, is monthly value of liquidity factor, is the estimate of market risk, and is the estimate of liquidity risk, also known as the liquidity beta. The liquidity betas from 00 to 00 are estimated at the end of each month using the prior 0-month window (-month minimum). Similar to the previous section, in calculating liquidity betas, we obtain data from the year 000 for estimation of equation (). For the liquidity beta of stock holdings, we use the proportion invested in each stock as a weight and calculate the weighted average liquidity beta for the fund. Stocks are split into two groups. One group (identical holdings) includes stocks held by both an SMA and a corresponding MF at the same time. The other group (non-identical holdings) includes stocks held either by the SMA or by the MF, but not both. Liquidity betas for fund i, at the holding level are calculated as follows: =,, (a) h =,,,,,, (b)

19 Page of h =,,,,,, (c) where, is holding weight of security j in the fund in month t,, is security j s liquidity calculated from equation (),,, is the holding weight of security j in SMA,,, is the holding weight of security j in MF, and I is a binary variable that takes on a value of one if the bracketed condition is true, and zero if not.. Empirical results.. Fund characteristics of SMAs and MFs Table reports means, medians, and differences in fund (Panel A) and holding (Panel B) characteristics for the concurrently managed SMA and MF pairs in our sample. The results show that all of the characteristics of the two sets of funds are economically and statistically different. First, we highlight the difference in the sizes of funds in our sample. An average SMA in the sample is about three times larger than an average MF (. times at the mean and. at the median). Second, MF expense ratios are higher than SMA expense ratios by basis points (mean) to basis points (median). This is consistent with Elton, Gruber, and Blake (0) who attribute the underperformance of MFs relative to SMAs to higher expense ratios. Managing MFs involves higher operating cost from advertising, auditing, custody, filing, and reporting to shareholders and regulators, as well as the costs associated with issuing and redeeming shares due to their open-ended nature. Despite their larger operating costs, managers charge basis points less in management fees on average for managing MFs than SMAs. [Insert Table here]

20 Page of SMA managers list their fee structures (including expense ratios and management fees ) in a regulatory filing (Form ADV Part ) and do not provide prospectuses that break down every category of expense like MF does. Although management fees (as a percent of managed assets) are subject to negotiation between an investor and fund manager, fees decrease as assets under management increase and clients reach various break points. Considering the larger size of assets under management for the SMAs in our sample, a concurrent manager earns $ million more ($,. million * 0.%) than when managing a MF counterpart. Large differences in compensation may provide an incentive for concurrent managers to favor the performance of SMAs over MFs. In Panel B, we compare the holding characteristics of our SMA and MF pairs. Differences in concentration of the two portfolios are modest measured by the total number of holdings and a higher percentage holding in stocks. At the median, pairwise differences in the number of securities in portfolios is one stock. Total weights of top 0 holdings in SMAs and MFs are also similar. Additionally, MF portfolios have slightly larger cash positions than SMA portfolios, although this is only statistically significant at the median. Larger cash positions in MFs are likely due to potential redemption needs, whereas the structure of SMAs negate the need for cash to be held to meet redemptions. If redemptions are requested by SMA investors, shares are sold at current market prices, potentially to the SMA investors detriment. However, if redemptions are requested by the MF investor, this could have a negative impact on the returns of the fund as a whole, thereby allocating liquidity risk to the fund and the fund manager... Comparison of SMA and MF returns The management fees disclosed in the Morningstar Database show fees decrease as AUM increases.

21 Page 0 of We start our analysis by examining the performance of SMAs and MFs managed by the same managers to document the outperformance of MFs by SMAs found by Elton, Gruber, and Blake (0) in our sample. In Table, we find that our matched sample SMAs outperform MFs after fees (net returns). Annualized net MF returns are bps (mean) and bps (median) lower than their SMA counterparts. In an untabulated analysis, we compare the performance of SMAs and MFs separately for each Morningstar investment style box. We confirm that this pattern holds for the vast majority investment styles, ranging from basis points to basis points in favor of SMAs ( of Morningstar investment style boxes for medians, and of investment style boxes for means). [Insert Table here] The next part of our analysis examines the potential reasons behind SMAs outperformance relative to their MF counterparts. One potential explanation for the outperformance is a variation in risk exposures between funds. However, when calculating riskadjusted alpha returns with equation () following Carhart (), we observe a similar pattern of outperformance of SMAs over MFs. Average annualized risk-adjusted returns of SMAs are bps ( bps at median) higher than those of MFs. Therefore, differences in risk exposures fail to explain the outperformance of SMAs. In Table, we also report risk factor loadings for SMAs and MFs and find that differences of the coefficients are not significantly different When measured using gross returns, the performance difference is smaller at around - bps. We should note that in this study we rely on net returns, risk-adjusted returns, and return gaps because direct and indirect trading costs, differences on portfolio turnovers, and operating cost, these measures capture the contents of the preferential treatment if it exists. These six investment styles are based on Morningstar style boxes: large, mid, small, growth, value, and blend. It might be helpful to provide an example to further explain how risk-adjusted alpha returns and four-factor loadings are reported in Table. To calculate fund i 's risk-adjusted alpha return for the month of Jan 00, we estimate equation () using monthly data from Jan 00-Dec 00 period. Factor loadings and alpha return are averaged across all sample funds and months, annualized and reported in Table.

22 Page of between the two groups, indicating outperformance is not due to different styles or common risk factors. This is not particularly surprising since matched pairs are also based on investment styles in addition to the identification of the portfolio manager. We conclude that SMA s outperformance is robust and is not due to size, market momentum, value versus growth, or other common risk factors. Table also reports return gaps for SMAs and MFs and their differences. Using equation (), return gap is calculated as the difference between actual fund net return and a hypothetical portfolio return that has the same holdings while adjusting for estimated trading costs calculated using equation () (Keim and Madhavan, ). We report that average SMA and MF funds have negative return gaps, which implies that unobserved actions of an average portfolio manager destroy value. However, we also find that return gaps of SMAs are larger than those of MFs. As Kacperczyk, Sialm, and Zheng (00) indicate, the return gap measures a fund s short-term performance due to unobserved actions and captures the manager s value-added relative to the previously disclosed holdings. Therefore, the return gap can be a proxy for managerial skills, trading costs, and agency costs. 0 With our matching methodology of concurrent management and investment style, the difference in return gaps is not affected by managerial skill or trading costs. Instead, the difference mainly captures managerial favoritism or preferential treatment in advantage of SMAs. Further, in untabulated analysis, we find that the difference in return gaps between SMAs and MFs is persistent for investment fund types. In four out of six Morningstar investment style boxes, return gaps are higher in SMAs. We attribute the Except for differences in medians, SMAs place higher weights on high book-to-market ratios and higher momentum equity portfolios. However, the difference is not economically significant to the extent that it would be driving the outperformance of SMAs. This finding is consistent with Kacperczyk, Sialm, and Zheng (00) who find weighted average of return gaps in in sample are negative.

23 Page of preferential treatment by portfolio managers to the asymmetric assets under management and management fees of the SMAs, which leads to higher compensation by SMAs. The results in Table demonstrate that SMAs outperform MFs when controlling for known determinants of returns. To test the hypothesis of whether SMA s outperformance is due to differences in the level of liquidity of holdings, we follow Liu s (00) liquidity-augmented CAPM model to estimate liquidity exposure according to equation () using fund level SMA and MF returns. In Panel A of Table, we report coefficients of liquidity alphas and betas at the fund level. Negative liquidity betas suggest both SMAs and MFs are fairly liquid. Although means and medians of liquidity betas are statistically significant, the differences between the coefficients are only marginally different in mean and not significantly different in median, a finding that is consistent with Wagner and Winter (0). 0 This finding serves as initial evidence that performance difference between SMAs and MFs is not driven by differences in liquidity. Consistent with this finding, liquidity alphas from Liu (00) model are higher for SMAs. Liquidity alphas can be interpreted as components of the net returns that are unexplained by liquidity exposure. The difference in alphas suggest that outperformance of SMAs is not driven by premiums earned from liquidity risk. [Insert Table here] Similarly, we analyze liquidity at the holding level and report the results in Panel B of Table. We split the holdings into identical and non-identical holdings after sorting each stock and its weight, place them into portfolios, and test for differences in liquidity betas between these In Liu s (00) model, smaller (or more negative) estimation coefficients of liquidity betas result from more liquid securities. 0 Mean differences are significant at the 0% level.

24 Page of holdings. First, we find that SMAs and MFs hold very liquid stocks, with both having negative liquidity betas. Second, although on average % of the holdings of SMAs and MFs are identical for concurrently managed funds, there is a difference in liquidity among the identical holdings betas which is likely driven by difference in weights. However, MFs are the ones with lower liquidity (i.e., lower betas), not SMAs. Lastly, for non-identical holdings, SMAs hold less liquid stocks with a liquidity beta of -0.0 in comparison to -0.0 for MFs. However, this difference is very small relative to the size of liquidity betas. Although the evidence in this univariate setting suggests that the non-identical holdings of SMAs are less liquid than MFs, overall the liquidity of the holdings of SMAs and MFs are only marginally different. The difference is not large enough to drive the outperformance of SMAs. In a subsequent section of our analysis, we control for liquidity when analyzing the SMA outperformance to test whether the differences in liquidity exposures are large enough to explain the outperformance. We also test the robustness of the liquidity alphas and liquidity betas (liquidity exposures) through other liquidity measures developed by Sadka (00) and Pastor and Stambaugh (00). The results using the Sadka (00) measure show that it is the permanent variable component of price impact (information-based liquidity risk factor) that explains the fund exposures in liquidity risk. The fixed transitory component of price impact (noninformation-based liquidity risk factor) barely explains the liquidity exposures at the fund level. Even though the coefficients for liquidity betas are significant at the variable level, the differences between SMAs and MFs are insignificant. At best, the results suggest the market wide liquidity risk explains the fund exposures in liquidity but could not explain why SMAs take on more risk than MFs for the concurrent managers. The results using the Pastor and Stambaugh These additional liquidity results are available upon request.

25 Page of (00) measure show both liquidity exposures of SMA and MF are insignificantly (low covariation) co-moving with aggregate liquidity factors. Given concurrent managers SMAs persistently outperform MFs, the abnormal returns are at least not from the market-wide liquidity exposures (risk). These results are consistent with Table in regard to differences in performance between SMAs and MFs which are not driven by the liquidity of their holdings. Differences in the liquidity alphas for both models across SMAs and MFs are all statistically significant at the % level. This confirms that the outperformance of SMAs is due to reasons besides liquidity exposure. Next, we compare the portfolio weights of SMAs and MFs and determine how weights relate to the outperformance of SMAs. More specifically, we test whether concurrent managers assign higher weights to top performing assets and lower weights in poorly performing assets in their SMA portfolios in comparison to MF portfolios. We pair the closest reporting dates of holdings for SMAs and MFs and sort each fund s holdings based on the past performance. For example, if the holdings are reported from the monthly basis, we sort based on past monthly returns and identify the top 0%, 0%, and 0% and the bottom 0%, 0% and 0% securities to examine the weights managers place in the holdings and the value-weighted returns for those holdings. Panel A of Table reports mean and median (in parenthesis) weights of the top and bottom 0%, 0%, and 0% performing stocks that are held in both SMA and MF portfolios. Differences in weights show that concurrent managers consistently overweight top performing stocks and underweight the bottom performing stocks in SMAs relative to MFs. In other words, even within same stock holdings, managers allocate better assets in SMA portfolios than they do in MF portfolios. Since managerial ability is controlled for by creating fund pairs, the difference suggests preferential treatment of SMA performance relative to MF performance. We also find

26 Page of that bottom performing stocks in MF portfolios that are being overweighted mostly belong to the identical stock holdings that are in SMA portfolios. Poorly performing stocks that belong to this group will hurt MF performance. [Insert Table here] By analyzing returns in Panel B of Table, we show that the favorable allocation of investable assets in SMAs contributes to their outperformance of MFs. Value-weighted alpha returns of the top and bottom 0%, 0%, and 0% of holdings are higher for SMAs in comparison to MFs. The overweighting of the best performing stocks and the underweighting of the poorly performing stocks in SMAs results in higher returns. These findings suggest performance differences between SMAs and MFs are at least partly due to holding weights placed into SMAs and MFs. The implication is that concurrent managers favor SMAs and better allocate fund assets, to which we again attribute higher compensation for the manager by the SMA... Event studies on concurrent management To separate the causality of return gap and concurrent management, we collect the inception dates for all concurrent lead managers in the sample period. We create an event study to test whether the return gaps increase or decrease after the onset of concurrent management. The event date is defined as the month portfolio manager starts concurrently managing SMA and MF portfolios. We calculate the holding period returns from month after to months after the event date for all reported holdings from SMAs and MFs, respectively, and match the closest reporting frequency of the two to precisely calculate the return gap. We denote existing MF managers that subsequently start SMA portfolios as MS, and existing SMA managers that subsequently start MF portfolios as SM. This subsample is smaller than our original sample

27 Page of because we remove concurrent managers with identical starting dates to assess differences in motivation between individuals who start as solely SMA or MF managers. Since we will test the performance gap surrounding the concurrent management, this differentiation is necessary for accurate conclusions on whether concurrent management generates preferential treatment. Our results in Table, Panel A, show that there is a basis point difference per year between the mean return gaps in the SM sample (the concurrent managers that started as SMAs managers). In other words, after SMA managers start MFs, their performance is better in the SMAs. Similarly, we find that there is an eight basis point difference per year between return gaps for MS sample, meaning MF managers also perform better in SMA portfolios after they start to concurrently manage both funds. Given that our methodology allows us to control for managerial skills, fund characteristics, and common risk exposures at fund and holding levels, it is difficult to explain why performance of SMAs would be higher for both of these subsamples. We believe these results exhibit favoritism for performance of SMAs on behalf of dual managers. We also believe that differences in fund size, compensation levels, and differences in behavior (flow-performance) of retail and institutional investors drive this preferential treatment. [Insert Table here] In Panel B of Table, we analyze return gaps for the window of months to month before concurrent management starts. Before this event date SM (MS) managers are in charge of SMAs (MFs) only. So, for comparison purposes we identify MFs (SMAs) controlled by other managers that become concurrently managed. We find that even before the concurrent management for SM sample, SMAs outperform MFs by bps. However, for MS sample, there is no difference in performance of these funds. These comparisons do not control for managerial skill as SMAs and MFs are in control of different managers before the concurrent