Why Do Fund Families Release Underperforming Incubated Mutual Funds?

Why Do Fund Families Release Underperforming Incubated Mutual Funds? Sara E. Shirley and Jeffrey R. Stark Although the average incubated mutual fund outperforms nonincubated funds by up to 3.41% annually, a large number of released funds underperform during incubation. We find that launching underperforming incubated mutual funds is associated with objectives that attract large inflows and lower relative risk. These findings are consistent with the use of incubation to maximize fee revenue through means other than the flow-to-performance relationship. We also find that underperforming incubated funds are incubated longer suggesting that families release funds opportunistically to take advantage of outperformance when it is observed. The value of a mutual fund to its fund family is its ability to generate revenue. This revenue comes from a fund s capacity to increase its assets under management, which, in turn, increases the fees collected by the fund and results in additional revenue for the fund family. Evans (2010) discusses four methods that are used by fund families to increase revenue, one of which is fund performance. It is well established in the literature that the mechanism by which performance leads to increased revenue is the relationship between past performance and current period inflows (Ippolito, 1992; Chevalier and Ellison, 1997; Sirri and Tufao, 1998). As a fund outperforms, investors chase performance, which increases assets under management and provides additional fee revenues for the family. However, competing on the basis of performance is a difficult task given the documented underperformance of mutual funds (Jensen, 1968; Carhart, 1997; French, 2008; Fama and French, 2010). Nevertheless, outperformance does occur and may be the result of favorable treatment of a specific fund within a family (Guedj and Papastaikoudi, 2004; Gaspar, Massa, and Matos, 2006) or incubation (Evans, 2010). The process of mutual fund incubation begins when fund families internally launch small funds with low levels of total net assets. After a period of evaluation, the family determines which mutual funds to publicly launch and which funds to discontinue. By taking advantage of the incubation process and the established relationship between past performance and current period inflows, mutual fund families are able to selectively launch incubated mutual funds that, on We thank an anonymous referee, Marc Lipson (Editor), Scott Gilbert, Jason Greene, James Musumeci, Edward Nowlin, Mark Peterson, David Rakowski, and Wanli Zhao for helpful comments and discussions. We are also thankful to Frank Hatheway from NASDAQ for providing the data on mutual fund ticker creation dates and to seminar participants at Roger Williams University. The authors are responsible for any remaining errors. Sara E. Shirley is an Assistant Professor of Finance at Roger Williams University in Bristol, RI. Jeffrey R. Stark is an Assistant Professor of Finance at Bridgewater State University in Bridgewater, MA. Financial Management Fall 2016 pages 507 528

508 Financial Management Fall 2016 average, have generated a positive objective alpha. However, outperformance is difficult to come by. Evans (2010) suggests that incubation is not used to identify skilled managers, but to generate a misleading track record of outperformance as the Securities and Exchange Commission (SEC) allows incubation period performance to be advertised once a fund has been released, but does not require that the performance of discontinued funds be documented. The process of mutual fund incubation raises two unanswered questions: 1) what factors determine how long a mutual fund is incubated? and 2) why do a large portion of incubated funds that are released to the public underperform during the incubation process? These questions are the focus of the present study. We examine a sample of 3,610 US domestic equity mutual funds from 1999 through September 2014 and find that 22.63% (817) of the funds are incubated. We determine that of the 817 incubated funds in our sample, 354 or 43.33% have negative incubation period performance when measured by an equally weighted objective alpha. Our results indicate that the release of underperforming funds is significantly related to an objective that is attracting large inflows. This finding helps explain the launch of underperforming incubated funds and suggests that the need to fill family demands may be a major driving force in how mutual fund families use incubation. We also find that the probability of a quicker release of a fund is positively associated with incubation period performance, objective inflows, and family inflows. These findings support the view that fund families take advantage of outperforming incubated mutual funds as they appear in order to avoid any spuriously outperforming mutual funds from reverting back to average performance. In addition, these results support fund families launching incubated mutual funds faster when there is increased demand for the fund family s mutual funds or the incubated mutual fund s objective, suggesting that fund families seek to take advantage of spillover effects (Nanda, Wang, and Zheng, 2004). This paper is organized as follows. Section I details the data, while Section II examines the distribution of incubation period performance and its relationship with flows. Section III provides the empirical analysis of incubation length. Section IV presents the empirical analysis of the decision to release an underperforming mutual fund, while Section V provides our conclusions. I. Data and Sample Creation Incubated mutual funds are identified based on information from two databases. Mutual fund data (excluding ticker creation dates) are obtained from the Survivor-Bias-Free US Mutual Fund Database hosted by the Center for Research in Security Prices (CRSP). Other variables of interest from the CRSP database include mutual fund and fund family characteristics. 1 Ticker creation dates, or the date when a mutual fund is made available to the public, are obtained from a database hosted by the National Association of Securities Dealers (NASD). The NASD database provides a snapshot of all active mutual funds for each month from January 1999 to September 2014 and includes data for the ticker creation date, the fund s ticker, and the fund s name. Variables from 1 CRSP Style codes were utilized to create our sample of domestic equity mutual funds. Codes of EDSG, EDSH, EDSF, EDSN, EDSR, EDST, EDSU, EDSG, EDSS, EDSI, EDSM, EDSA, EDS, EDCL, EDCM, EDCS, EDCI, EDYG, EDYB, EDYH, EDYS, and EDYI were retained. For a complete description of the objectives used, see the Appendix (Section I and Table A1).

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 509 the CRSP database are obtained for each month from 1999 to the third quarter of 2014 to coincide with the NASD data. Before merging the CRSP data with the NASD data to identify incubated mutual funds, the data are filtered as in Evans (2010). 2 The NASD database includes funds that were created prior to 1999 if they survived past that time. Accordingly, any fund created prior to 1999 is omitted from the sample. In addition, fund name observations that contain the word TEST are removed from the NASD data. The merged databases identify incubated mutual funds by ticker. However, since tickers can be recycled over time, matches are checked by comparing fund names for CRSP data and NASD data. Observations with mismatched fund names are removed from the sample. After merging the two databases, we obtain the first date for which each share class has a valid return observation and group together all share classes belonging to the same mutual fund. The start date of the mutual fund is set to the start date of its earliest share class. After obtaining a mutual fund s start date and ticker creation date, we identify incubated funds by calculating the difference in time between the fund s ticker creation date and the date of the fund s first return. A negative difference between these two dates indicates that the mutual fund has been incubated and has generated returns prior to a ticker being assigned to the fund. To account for lags in a fund applying for a ticker from the NASD, observations with negative differences of less than six months are not considered to be incubated and are grouped accordingly. Observations with a positive difference of greater than three months (i.e., more than three months from the ticker creation date to the first returns) are attributed to errors in the data and are removed from the sample. 3 The final data sample consists of 3,610 US domestic equity mutual funds that were launched between 1999 and September 2014. Of this sample, 817 funds (22.63%) are incubated. 4 Descriptive statistics for all of the mutual funds in the sample are displayed in Table I. The sample of postincubation and nonincubated mutual funds include up to the first 36 months of a fund s life once it is made available to the public. These data indicate that during the incubation process, mutual funds are relatively small, with an average size of $29.50 million. However, after incubation, the average size rises to $107.43 million. This pattern is consistent with the established function of incubation, which is to allow a mutual fund to operate privately and with a small amount of capital while establishing a track record of performance. In addition, data in Table I indicate that the incubated funds are smaller than their nonincubated counterparts both during and postincubation. The expense ratio, front load, and rear load of incubated mutual funds does not change between the incubation period and the postincubation period, although nonincubated mutual funds charge lower expense ratios than incubated mutual funds (1.27% vs. 1.48%, respectively). 2 Evans (2010) provides detailed instructions for combining the databases in the Supplements & Datasets section of the Journal of Finance website at http://www.afajof.org/details/page/3626901/supplements.html. 3 The nature of incubation data inherently includes a sample bias due to the unobserved incubated mutual funds that do not survive the incubation process and is, therefore, not unique to our analysis (Evans, 2010; Gibson and Finke, 2014). Although this bias does determine the composition of the funds in our sample, the expected impact on our conclusions are minimal due to the similarities between our proxy sample for the unobservable funds and the sample of incubated funds used throughout our analysis. See the Appendix Section II for a detailed discussion of the potential impact of this bias. 4 The 22.63% found between 1999 and September 2014 is similar to the 23.1% reported by Evans (2010) in a sample of funds from 1999 to 2006 (242 of 1,048).

510 Financial Management Fall 2016 Table I. Sample Descriptive Statistics This table contains the descriptive statistics of our samples of incubated and nonincubated funds. The sample of incubated mutual funds is separated into two subsamples: 1) an incubation period and 2) a postincubation period. The mean and median of total net assets of a fund and family, the expense ratio of a fund, the average front and rear load of a fund, and the number of funds that were and were not incubated are reported below by subsample. Postincubation and nonincubated performances are calculated over a period of up to the first three years (36 months) a fund is available for public investment. Incubated Funds During Incubation Postincubation Nonincubated Funds Variable Mean Median Mean Median Mean Median Total net assets ($MM) $29.50 $4.50 $107.43 $25.80 $120.16 $28.10 Family size ($MM) $40,701 $5,360 $55,116 $7,475 $96,446 $11,865 Expense ratio 1.48% 1.42% 1.50% 1.45% 1.27% 1.25% Average front load 2.00% 2.80% 1.99% 2.79% 2.28% 2.96% Average rear load 0.73% 0.40% 0.72% 0.38% 0.67% 0.33% Number of funds 817 2,793 II. Incubation Period Performance and Flows A. Methodology Prior literature primarily focuses on the use of incubation to generate a track record of outperformance that creates increased inflows to the incubated fund once it has been released to the public. Consistent with the prior literature, Table II demonstrates that the incubated mutual funds in our sample outperform nonincubated funds when measured by an equally weighted and value-weighted objective alpha and a four-factor alpha during incubation by up to 3.41% risk adjusted per year. 5 This outperformance does not last though with incubated and nonincubated mutual funds performing equally well during postincubation. Further analysis of incubated mutual funds and their incubation period performance reveals that a large portion of incubated funds (354 of 817 or 43.33%) are released to the public with negative objective alphas, as illustrated in Figure 1. To better understand the motivation for incubating and releasing a mutual fund, we create subsamples based on an equally weighted objective alpha measure of fund performance during incubation. As demonstrated in Table III, during incubation, both outperforming and underperforming mutual funds are relatively close in size. However, postincubation, the outperforming funds grow faster than the underperforming funds, increasing in size from $25.46 million to $132.06 million, while the underperforming funds increase in size from $27.16 million to $71.08 million. Table IV reports the difference in performance from incubation to postincubation for outperforming and underperforming funds. During incubation, the annualized difference in average equally weighted objective alpha performance between the subgroups of outperforming and 5 Objective alpha is defined as the raw returns of a mutual fund minus the total equally-weighted average returns of all other funds within the same fund objective. This measure of performance is a proxy for fund performance relative to a benchmark, similar to the performance advertised by many mutual funds (http://www.lipperweb.com/research/indexservice.aspx). For robustness, a value-weighted approach is also presented with all of the results.

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 511 Table II. Performance of Incubated and Nonincubated Mutual Funds This table examines performance measures for incubated and nonincubated funds. Following the analysis of Evans (2010), Panel A compares the performance of incubated mutual funds during their incubation period to the performance of mutual funds that were not incubated, while Panel B compares the performance of incubated mutual funds during their postincubation period to the performance of nonincubated mutual funds. Asterisks in Columns 1 and 2 indicate the significance levels for a t-test of annualized performance from zero, and the asterisks in Column 3 provide the significance of the difference in annualized performance between the two samples. Performance measures examined are an equal-weighted and a value-weighted objective alpha, a four-factor alpha (Carhart, 1997) and raw returns. Incubation period performance is calculated over the entirety of the incubation period. Postincubation and nonincubated performances are calculated over a period up to the first three years (36 months) a fund is available for public investment. Panel A. Incubation Period Performance Incubated Funds Nonincubated Funds (817 Funds) (2,791 Funds) Annualized Variable t = Incubation Period t = 1 to 36 Months Difference Four-factor alpha 2.89% 0.52% 3.41% EW objective alpha 3.53% 0.47% 3.06% VW objective alpha 2.66% 0.44% 3.10% Raw returns 12.84% 5.75% 7.10% Panel B. Postincubation Period Performance Incubated Funds Nonincubated Funds (817 Funds) (2,791 Funds) Annualized Variable t = 1 to 36 Months t = 1 to 36 Months Difference Four-factor alpha 0.13% 0.52% 0.39% EW objective alpha 0.84% 0.47% 0.37% VW objective alpha 0.24% 0.44% 0.68% Raw returns 4.12% 5.75% 1.64% Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level. underperforming incubated mutual funds is 17.16% per year [10.52% ( 6.64%)]. However, both the positive and negative performance subgroups generate returns close to 1% postincubation. This is consistent with the findings of Table II, confirming that incubation period performance does not predict postincubation performance. Similar to Evans (2010), we employ multivariate regression to examine the impact of incubation on the flows received by a mutual fund after launch. We extend the work of Evans (2010) by creating subsamples based on a positive or negative objective alpha performance during incubation. The relationship is formally examined with the following linear regression equation: Flow i,t = α i + β 1 Incubated i + β 2 Fund TNA i,t + β 3 Family TNA i,t + β 4 Fund Flow i,t 1 + β 5 Objective Flow i,t + β 6 Fees i,t + β 7 Multiple i,t + β 8 Performance i + ε i, (1)

512 Financial Management Fall 2016 Figure 1. Distribution of Incubation Period Performance of Individual Mutual Funds This figure displays the distribution of incubation period fund cumulative performance as measured by an equally weighted objective alpha. Table III. Sample Descriptive Statistics by Incubation Period Performance This table contains the descriptive statistics of incubated mutual funds. The sample is separated into a group of mutual funds that have a positive incubation period equally weighted objective alpha and a negative incubation period equally weighted objective alpha. The mean and median of total net assets of a fund and family, the expense ratio of a fund, the average front and rear load of a fund, and the number of funds with positive and negative incubation period performance are reported below. Positive EW Incubation Objective Alpha Negative EW Incubation Objective Alpha During Incubation Postincubation During Incubation Postincubation Variable Mean Median Mean Median Mean Median Mean Median Total net assets ($MM) $25.46 $6.92 $132.06 $40.55 $27.16 $5.68 $71.08 $19.16 Family size ($MM) $54,102 $9,407 $54,988 $8,959 $51,152 $6,729 $54,694 $6,764 Expense ratio 1.47% 1.45% 1.47% 1.44% 1.54% 1.46% 1.52% 1.50% Average front load 1.96% 2.76% 1.97% 2.79% 2.06% 2.83% 2.03% 2.80% Average rear load 0.68% 0.40% 0.70% 0.40% 0.73% 0.33% 0.74% 0.33% Number of funds 463 354

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 513 Table IV. Performance of Incubated Mutual Funds This table presents the annualized equally weighted objective alpha performance of incubated mutual funds during incubation and for up to 36 months postincubation by incubation period performance. Also presented is the annualized difference in performance between incubation and postincubation period returns. Asterisks in Columns 1 and 2 indicate the significance levels for a t-test of annualized performance from zero, while the asterisks in Column 3 provide the significance of the difference in annualized performance between the incubation and postincubation samples. Incubation Period Postincubation Annualized Obj. Alpha Annualized Obj. Alpha Annualized Difference Positive obj. alpha 10.52% 0.93% 9.59% Negative obj. alpha 6.64% 1.21% 7.85% Significant at the 0.01 level. where Flow i,t is the growth in assets of fund i from time t 1tot after accounting for the growth of assets attributed to the performance of fund i, and is measured as Flow = TNA i,t TNA i,t 1 (1 + R i,t ), (2) where TNA represents the total net assets of fund i at time t (TNA i,t )ort 1(TNA i,t 1 ) and R i,t represents the returns of fund i during period t. Flow is calculated for 12-month periods to create a measure of annual fund flows. For funds that are incubated, flow measurements begin postincubation. As in Evans (2010), monthly cross-sectional ranks are used rather than dollar flow due to the large variation in mutual fund size among startup funds. Incubated is a dummy variable that takes a value of one if a mutual fund has been incubated and zero otherwise. The remaining variables in the regression equation are defined as follows. Fund_TNA i,t is the log of TNA for fund i at time t. Family_TNA i,t is the log of TNA for the management company of fund i at time t, excluding the TNA of the fund of interest (fund i or the dependent variable). All other mutual funds within the family are included in the calculation, not just those that have been incubated. Fund_Flow i,t-1 is a rank from 0.01 to 1.0 of inflows received by fund i over the prior 12-month period. Objective_Flow i,t is a rank from 0.01 to 1.0 of fund flows to the objective for fund i during period t, excluding the flows attributed to the fund of interest (fund i or the dependent variable). Fees i,t represents the expense ratio for fund i at time t. Multiple i,t is a dummy variable indicating whether the incubated fund s family currently has additional funds within an objective class and Performance i is the lifetime performance of fund i as measured by raw returns and includes incubation period performance for incubated mutual funds. B. Results Table V presents the results of regressing flow (the rank of net dollar flows of a fund, as calculated in Equation (2)) on incubation (whether a fund has been incubated or not). Year fixed effects are included to account for any variation from year to year. For incubated mutual funds, the start of the inflow period is the first postincubation month. The dependent variable, Flow, is measured over the second 12-month period (Months 13 to 24) for both incubated and nonincubated mutual funds so that lagged flows can be calculated over the first 12-month period for each fund.

514 Financial Management Fall 2016 Table V. Relationship between Incubation and Mutual Fund Flows This table provides the coefficient estimates from the regression listed below of fund flows on a dummy variable identifying whether a mutual fund was incubated, as well as fund and family characteristics. Flow i,t = α i + β 1 Incubated i + β 2 Fund TNA i,t + β 3 Family TNA i,t + β 4 Fund Flow i,t 1 + β 5 Objective Flow i,t + β 6 Fees i,t + β 7 Multiple i,t + β 8 Performance i + ε i. The sample used here contains all of the funds identified in the sample (both incubated and nonincubated) sorted into subgroups based on a positive or negative incubation period equally weighted objective alpha. The dependent variable is the rank of dollar flows to the fund and takes on a fractional value between 0.01 and 1.0 based on net dollar flows that year. Observations of flow begin after a mutual fund has been open to the public for 12 months, allowing for lagged flows to be calculated. Independent variables include a dummy variable identifying incubated mutual funds, the log of a fund s total net assets, the log of the family s total net assets, a rank measure of the flows to a fund over the prior 12 months taking a value between 0.01 and 1.0, the rank of flows to the investment objective over the period the dependent variable is calculated and ranked between 0.01 and 1.0, the fees charged by the fund, a dummy variable indicating whether a fund family already had a mutual fund in the incubated fund s objective class, and the total raw returns of a fund. Annual fixed effects are included. One annual observation is retained for each fund that has the necessary 24 months of observations. Asterisks indicate statistical significance of the coefficients and t-statistics are in parentheses. Fund Flow Rank Variable Positive Negative Intercept 0.38 0.37 (17.81) (18.22) Incubated 0.00 0.03 (0.39) (4.31) Fund TNA 0.01 0.01 (3.32) (3.14) Family TNA 0.00 0.00 (1.27) (1.68) Fund Flow t 1 0.25 0.25 (28.15) (28.84) Objective Flow 0.03 0.03 (3.69) (3.57) Fees 0.02 0.02 ( 4.12) ( 3.98) Multiple 0.00 0.00 ( 0.42) ( 0.49) Performance 0.06 0.06 (7.64) (7.69) Fixed effects Yes Yes Adjusted R 2 10.27% 10.77% Observations 12,004 11,854 Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level. Our sample includes both incubated and nonincubated funds, and the analysis covers the period from the ticker creation date up to 60 months after the fund is made available to the public. Fund performance extends back to the fund s inception. One annual observation is retained for each year that a mutual fund is in our sample. We separate our sample of incubated funds into two

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 515 subsamples: 1) funds that generate a positive equally weighted objective alpha during incubation and 2) those that generate a negative equally weighted objective alpha during incubation. Nonincubated mutual fund performance is measured in the same way, beginning at the date of inception. As demonstrated in Table V, for the subsample of funds that generate positive objective alphas during incubation, no significant relationship is found between flow and the incubated variable. These results indicate that the differences in postincubation flow among mutual funds that generate positive objective alphas during incubation is explained by performance, objective flows, expense ratio, fund size, and lagged fund flows. For the subsample of mutual funds that generate negative objective alphas during incubation, the relationship between flow and incubation remains positive and significant with an incubated coefficient of 0.03. These results indicate that the postincubation flows from these underperforming mutual funds cannot be explained by the fund s performance or other control variables. These findings suggest that underperforming incubated mutual funds serve the fund family in ways that are unrelated to the establishment of a track record of outperformance. III. Incubation Length A. Methodology Evidence in the literature regarding mutual fund manager skill reveals that few mutual funds outperform (Jensen, 1968; Malkiel, 1995; Carhart, 1997; Fama and French, 2010). The findings suggest that, on average, outperformance by a mutual fund, even during incubation, cannot be attributed to manager skill. Accordingly, there is no guarantee that a fund s performance will persist into the future. Given the potentially fleeting nature of mutual fund outperformance and the increased inflows a fund receives for having positive incubation period performance, it follows that fund families will take advantage of outperforming funds by launching them before returns become negative. According to this logic, the duration of incubation for a mutual fund should be inversely related to the mutual fund s performance. Another explanation for the timing of mutual fund launches is voiced by Robert Puff, Chief Investment Officer at Twentieth Century Funds, The industry will keep introducing whatever has been working well lately. 6 In other words, funds are released to the public when an objective s or family s funds are in demand. We use a Cox proportional hazard regression model (Hosmer and Lemeshow, 1999; Fox, 2002) to explore the determinants of a fund s incubation release. The use of this model allows us to examine the impact of our covariates on the decision to release an incubated fund in a given period rather than continuing on with the incubation process. The regression is defined as Incubation Release i = β 1 Incubation Performance i + β 2 Objective Flow i + β 3 Family Flow i + β 4 Family Performance i + β 5 Objective TNA i + β 6 Family TNA i + β 7 Fund TNA i + β 8 Incubation Risk i + β 9 Fees i + β 10 Front i + β 11 Rear i + β 12 Multiple i + β 13 FamilyInc i + ε i, (3) where Incubation_Release i is measured as the month that mutual fund i is released from incubation. Incubation_Performance i measures the objective alpha performance or four-factor alpha of 6 See the article by Jason Zweig in the July 1996 edition of Money.

516 Financial Management Fall 2016 an incubated fund over the entirety of its time spent in incubation. Family_Flow i measures the flows of the fund family for fund i over the previous 12 months, excluding the flows attributed to the fund of interest (fund i or the dependent variable) and takes on a rank value of 0.01 to 1.0. Family_Performance i measures performance over the previous 12 months of the incubated fund s family, excluding the performance of the incubated fund of interest (fund i or the dependent variable). Objective_TNA i is the log of total net assets within the objective class for fund i at the time the incubated mutual fund is released to the public, excluding the TNA of the fund of interest (fund i or the dependent variable). Incubation_Risk i is the standard deviation of fund performance during incubation, Front i is the front end load that fund i charges over its incubation period, and Rear i is the rear load that fund i charges over its incubation period. FamilyInc i,t-1 is a dummy variable that takes a value of one if the family of fund i released an incubated mutual fund over the previous 12 months and zero otherwise. The variables of Objective_Flow, Family_TNA, Fund_TNA, Risk, Fees, and Multiple are defined as in Equation (1). B. Results Earlier results that establish the outperformance of mutual funds during incubation as contrived indicates uncertainties surrounding the timing of outperformance (i.e., when will it occur?) and the duration (i.e., how long will it last?). The inability of fund families to control the performance of their incubated mutual funds and the positive inflow response to incubation period performance provide a strong motivation for fund families to launch incubated mutual funds at the first sight of outperformance. In Table VI, we divide the 817 incubated mutual funds in our sample into subgroups based on the length of their incubation: 243 funds are incubated from 7 to 12 months, 320 funds are incubated from 13 to 24 months, and 254 funds are incubated for 25 months or more. Table VI provides a breakdown of performance for each subgroup. The annualized equally weighted objective alpha measure of performance (four-factor alpha) decreases as the incubation period increases, from a high of 6.85% (4.67%) for funds incubated from 7 to 12 months to a low of 0.10% (0.34%) for funds incubated for 25 or more months. In addition, it is the mutual funds that are incubated for the shortest time that attract the most money postincubation. When the 817 incubated mutual funds are further divided into subgroups based on incubation period performance, the results are unchanged. As illustrated in Table VII, there is no pattern to the postincubation performance. However, the funds that outperform during incubation show much greater growth postincubation than the funds that underperform. Taken together, these data indicate that regardless of the length of incubation or the level of incubation period performance, on average, incubated mutual funds tend to revert to an equally weighted objective alpha close to 1% postincubation. Table VIII presents the results of multivariate analysis of the impact of performance and other fund and family characteristics on the likelihood that a mutual fund is released from incubation rather than continuing the incubation process using Equation (3). We examine incubation period performance and the impact of objective and family-level flows. The dependent variable measures the decision to launch an incubated fund to the public in a given month. Consistent with the expectation that fund families take advantage of market conditions to maximize inflows, our results indicate that funds are more likely to be launched when fund performance is high (Incubation Performance), when there is the greatest likelihood of the fund experiencing spillover benefits from a hot family or objective (Objective Flow and Family Flow), and when a fund is charging a higher expense ratio (Fund Expense Ratio). Also of interest is the significant relationship with

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 517 Table VI. Descriptive Statistics of Incubation Period Performance and Length of Incubation This table provides the mean and median of incubation period performance and fund characteristics by the length of a fund s incubation. Incubation length is divided into three subgroups of 7 to 12, 13 to 24, and 25+ months. All statistics are presented as during incubation in Panel A and postincubation for a period measuring up to 36 months in Panel B. Performance measures reported include monthly raw returns, an equally weighted objective alpha, a value-weighted objective alpha, and a four-factor alpha (Carhart, 1997). Also reported is the number of funds incubated for each subgroup of length, the average expense ratio, and the total net assets of a fund by incubation length. Postincubation measures are calculated over a period up to the first 36 months a fund is available for public investment. Panel A. Incubation Period Performance by Incubation Length Incubation Length 7 to 12 Months 13 to 24 Months 25+ Months Variable Mean Median Mean Median Mean Median Number of funds 243 320 254 Raw returns 15.77% 15.27% 15.06% 15.30% 10.48% 11.82% EW objective alpha 6.85% 2.34% 3.63% 1.64% -0.10% 0.48% VW objective alpha 6.48% 1.64% 2.46% 0.53% 1.17% 0.15% Four-factor alpha 4.67% 0.66% 3.57% 0.68% 0.34% 0.36% Expense ratio 1.52% 1.45% 1.52% 1.50% 1.45% 1.31% Total net assets ($MM) $31.47 $6.48 $23.05 $4.63 $33.49 $3.90 Panel B. Postincubation Period Performance by Incubation Length Number of funds 243 320 254 Raw returns 6.19% 11.98% 6.49% 12.05% 6.79% 13.02% EW objective alpha 0.05% 0.15% 0.52% 0.17% 0.14% 0.08% VW objective alpha 0.75% 0.80% 0.23% 0.41% 0.84% 0.50% Four-factor alpha 0.99% 1.04% 0.66% 0.41% 0.30% 1.13% Expense ratio 1.55% 1.44% 1.51% 1.50% 1.43% 1.35% Total net assets ($MM) $135.75 $38.50 $94.86 $24.90 $96.02 $17.65 Family Incubated in Prior Year, which indicates that fund families that are more familiar with the incubation process are willing to launch funds earlier on in their incubation cycle. The results presented in Tables VI, VII, and VIII largely support the view that fund families launch mutual funds quickly once outperformance is achieved. The manufactured nature of incubation period outperformance increases the motivation for fund families to release mutual funds that outperform before their outperformance reverses. The findings relating family and objective flows to incubation length are consistent with the view that mutual funds seek to maximize inflows by continuing to offer investors what they want. Our results indicate that the main determinants of incubation length are revenue driven, specifically the performance of the mutual fund and the demand for the fund s family or objective class at that time. IV. Underperforming Incubated Funds A. Methodology To examine the motivation for releasing an underperforming incubated mutual fund to the public, we examine the determinants for launching a new mutual fund. Based on Chevalier and Ellison (1997), we study the motivations for launching a mutual fund that are based on fulfilling family needs. The decision to launch an incubated fund can be viewed in the same way as the

518 Financial Management Fall 2016 Table VII. Descriptive Statistics of Incubation Period Performance and Length of Incubation and Incubation Period Performance This table provides the mean and median of incubation period and postincubation period performance and fund characteristics by the length of a fund s incubation and by incubation period performance. Incubation length is divided into three subgroups of 7 to 12, 13 to 24, and 25+ months. These subgroups are divided further based on a fund s generated incubation period performance. All statistics are presented as during incubation in Panel A and postincubation in Panel B. Performance measures reported include monthly raw returns, an equal and value-weighted objective alpha, and a four-factor alpha (Carhart, 1997). Also reported is the average expense ratio and the total net assets of a fund by incubation length. Postincubation measures are calculated over a period up to thefirst 36 months a fund is available for public investment. Panel A. Incubation Period Performance by Incubation Length Positive Incubation Performance by Length Negative Incubation Performance by Length 7 to 12 13 to 24 25 to 36 7 to 12 13 to 24 25+ Months Months Months Months Months Months Variable Mean Median Mean Median Mean Median Mean Median Mean Median Mean Median Raw returns 54.36% 44.04% 54.48% 41.28% 53.51% 45.08% 54.72% 34.13% 57.42% 28.80% 46.43% 27.60% EW objective alpha 14.88% 8.82% 10.29% 7.50% 6.60% 4.55% 8.18% 5.90% 6.83% 4.56% 5.63% 3.08% VW objective alpha 31.68% 20.53% 27.91% 19.20% 26.87% 17.04% 21.60% 12.42% 23.08% 12.90% 24.24% 15.00% Four-factor alpha 17.28% 10.26% 10.80% 6.66% 6.48% 3.72% 10.30% 6.36% 5.04% 3.48% 4.56% 2.94% Expense ratio 1.49% 1.49% 1.49% 1.49% 1.47% 1.35% 1.79% 1.53% 1.49% 1.50% 1.40% 1.31% TNA ($MM) $28.80 $9.97 $19.22 $6.37 $12.33 $5.55 $34.80 $6.20 $29.06 $6.48 $42.32 $4.49 Panel B. Postincubation Period Performance by Incubation Length Raw returns 10.30% 13.32% 10.21% 13.16% 9.60% 9.83% 6.96% 5.22% 7.00% 14.06% 0.24% 1.00% EW objective alpha 0.72% 0.50% 1.43% 1.03% 0.44% 0.50% 2.16% 1.01% 2.64% 0.96% 1.92% 0.52% VW objective alpha 0.00% 0.00% 0.84% 0.38% 0.16% 0.30% 2.96% 1.84% 2.11% 0.03% 1.61% 0.23% Four-factor alpha 0.22% 0.63% 1.61% 0.00% 0.29% 1.26% 2.61% 2.22% 0.86% 1.44% 0.32% 1.01% Expense ratio 1.49% 1.44% 1.50% 1.49% 1.42% 1.35% 1.59% 1.50% 1.58% 1.50% 1.42% 1.35% TNA ($MM) $171.07 $68.48 $105.35 $37.81 $92.76 $31.43 $79.34 $31.53 $80.60 $19.53 $98.73 $16.15

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 519 Table VIII. Determinants of Mutual Fund Incubation Length This table reports the coefficients from the Cox proportional hazard regression model below of incubation length on fund, family, and objective characteristics. Incubation Release i = β 1 Incubation Performance i + β 2 Objective Flow i + β 3 Family Flow i + β 4 Family Performance i + β 5 Objective TNA i + β 6 Family TNA i + β 7 Fund TNA i + β 8 Incubation Risk i + β 9 Fees i + β 10 Front i + β 11 Rear i + β 12 Multiple i + β 13 FamilyInc i + ε i. The dependent variable measures the decision to launch an incubated mutual fund to the public in a given month. Independent variables include incubation period performance measured as an equally weighted objective alpha, a value-weighted objective alpha, or a four-factor alpha (Carhart, 1997), the rank from 0.01 to 1.0 of net dollar flows to the concurrent investment objective over the prior 12 months, the rank from 0.01 to 1.0 of net dollar flows of the family over the prior 12 months, the excess returns of the incubated fund s family over the prior 12 months, the log of the investment objective s total net assets at the time a fund is released from incubation, the log of the incubated fund family s total net assets at the time a fund is released from incubation, the log of the incubated fund s total net assets at the time it is released from incubation, the standard deviation of the incubated mutual fund s monthly returns during incubation, the average expense ratio of the incubated fund over the course of its incubation period, the average front load of the incubated fund over the course of its incubation period, the average rear load of the incubated fund over the course of its incubation period, a dummy variable indicating whether a fund family already had a mutual fund in the incubated fund s objective class, and a dummy variable taking a value of one if the fund family launched an incubated mutual fund over the prior 12 months. Asterisks indicate statistical significance of the coefficients and p-statistics are in parentheses. EW Objective Alpha VW Objective Alpha 4 Factor Alpha Variable 1 2 3 Incubation Performance 0.455 0.867 11.340 (0.014) (0.000) (0.000) Objective Flow 0.579 0.536 0.555 (0.007) (0.013) (0.010) Family Flow 0.697 0.694 0.695 (0.000) (0.000) (0.000) Family Performance 0.404 0.539 18.649 (0.107) (0.029) (0.286) Objective TNA 0.035 0.032 0.034 (0.300) (0.338) (0.321) Family TNA 0.008 0.007 0.005 (0.488) (0.556) (0.658) Fund TNA 0.025 0.028 0.025 (0.266) (0.218) (0.275) Incubation Risk 0.383 0.876 1.203 (0.810) (0.580) (0.451) Fees 0.206 0.206 0.194 (0.000) (0.000) (0.000) Front 0.036 0.037 0.033 (0.192) (0.178) (0.222) Rear 0.110 0.096 0.104 (0.033) (0.063) (0.044) Multiple 0.140 0.150 0.129 (0.106) (0.082) (0.135) FamilyInc 0.684 0.685 0.666 (0.000) (0.000) (0.135) AIC 9,286.84 9,269.20 9,282.13 Number of obs. 817 817 817 Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level.

520 Financial Management Fall 2016 decision to exercise an American call option on a dividend paying stock. In other words, a fund family will retain the option to exercise until the fund and market conditions are conducive to maximizing inflows to the fund or the fund s family. This perspective is supported by evidence in the literature indicating that new mutual funds are launched into hot objectives (Ippolito, 1992; Sirri and Tufano, 1998; Khorana and Servaes, 1999) and by existing mutual funds taking advantage of market conditions to maximize inflows (Cooper, Dimitrov, and Rau, 2001; Cooper, Gulen, and Rau, 2005). Accordingly, we do not expect to find underperforming incubated mutual funds launched unless the fund s objective is particularly strong (i.e., attracting large inflows). The motivation for launching an outperforming fund is clear. Inflows follow outperformance. This motivation prompts managers of incubated mutual funds to compete with each other (both within a family and outside) in an attempt to generate top performing funds (Brown, Harlow, and Starks, 1996; Kempf and Ruenzi, 2008; Massa and Patgiri, 2009). It follows that outperforming incubated mutual funds are those that have taken the greatest amount of risk. The relationship between risk and performance during the incubation period may also indicate that certain funds are incubated in an attempt to generate outperformance, such as those that take on additional risk, while other funds are incubated for the ex ante purpose of fulfilling family needs, resulting in decreased risk taking by removing the need to compete on the basis of performance. We perform a logistic regression analysis to examine the existence of underperforming incubated mutual funds. The regression equation is defined as Incubation Under Perform i,t = α i + β 1 Fund TNA i,t + β 2 Family TNA i,t + β 3 Objective TNA i,t + β 4 Incubation Risk i + β 5 Fees i,t + β 6 Incubation Length i + β 7 Market t 1 + β 8 FamilyInc i,t 1 + β 9 Objective Flow i + β 10 Family Flow i,t +β 11 Multiple i,t +ε i, (4) where Incubation_Under_Perform i,t takes a value of one if the mutual fund s incubation period performance is negative and zero otherwise, as measured by objective alphas and a four-factor alpha. By assigning a value of one to negative performance, the regression coefficients represent their relative impact on the likelihood of releasing an underperforming incubated mutual fund. Market t-1 is equal to the S&P 500 market index returns over the previous 12 months. All other independent variables are measured as in Equations (1) and (3). 7 B. Results The results presented earlier in Figure 1 suggest that incubation must serve a purpose in addition to generating a track record of outperformance as 43.33% of all incubated mutual funds in our sample underperform during their time in incubation when measured by an equally weighted objective alpha. The benefits of outperforming during incubation and the increased inflows that follow are already established, both here and in the existing literature, raising the question as to why underperforming incubated mutual funds are released. To identify the determinants of launching an underperforming mutual fund, we examine the relationship between fund, family, and market characteristics and the likelihood of releasing an 7 Control variables were motivated by Reinganum (1985), Sirri and Tufano (1998), Khorana and Servaes (1999), and Evans (2010).

Shirley & Stark Why Do Fund Families Release Underperforming Incubated Mutual Funds? 521 Table IX. Determinants of Releasing an Underperforming Incubated Mutual Fund This table reports the coefficients from the logistic regression below of the decision to release an underperforming mutual fund from incubation. Incubation Under Perform i,t = α i + β 1 Fund TNA i,t + β 2 Family TNA i,t + β 3 Objective TNA i,t + β 4 Incubation Risk i + β 5 Fees i,t + β 6 Incubation Length i + β 7 Market t 1 + β 8 FamilyInc i,t 1 + β 9 Objective Flow i + β 10 Family Flow i,t + β 11 Multiple i,t + ε i. The event examined is the decision to release an underperforming fund and coefficients represent the impact of the dependent variable on the likelihood of this occurring. Column 1 measures incubation period performance with an equally weighted objective alpha, Column 2 measures performance with a value-weighted objective alpha, and Column 3 measures performance with a four-factor alpha (Carhart, 1997). Independent variables include the log of fund total net assets, the log of a fund family s total net assets, the log of a fund s objective total net assets, the standard deviation of the incubated mutual fund s monthly returns during incubation, the average fund expense ratio, the length of a fund s incubation period in months, the cumulative returns of the S&P 500 index over the prior 12 months, a dummy variable taking a value of one if the fund family launched an incubated mutual fund over the prior 12 months, the rank of net flows to the concurrent investment objective over the prior 12 months, the rank of net flows of the family over the prior 12 months, and a dummy variable indicating whether a fund family already had a mutual fund in the incubated fund s objective class. Annual fixed effects are included. Asterisks indicate statistical significance of the coefficients and p-statistics are in parentheses. EW Objective Alpha VW Objective Alpha 4 Factor Alpha Variable 1 2 3 Intercept 1.219 0.097 0.791 (0.251) (0.926) (0.449) Fund TNA 0.180 0.136 0.059 (0.011) (0.090) (0.247) Family TNA 0.009 0.031 0.061 (0.828) (0.438) (0.125) Objective TNA 0.086 0.097 0.050 (0.170) (0.114) (0.414) Incubation Risk 0.172 0.216 0.278 (0.000) (0.000) (0.000) Fees 0.159 0.054 0.037 (0.150) (0.599) (0.719) Incubation Length 0.011 0.013 0.016 (0.063) (0.057) (0.048) Market 0.062 0.440 0.215 (0.938) (0.566) (0.778) FamilyInc 0.001 0.005 0.002 (0.643) (0.111) (0.441) Objective Flow 0.682 0.957 0.556 (0.017) (0.001) (0.045) Family Flow 0.131 0.263 0.128 (0.667) (0.379) (0.668) Multiple 0.028 0.222 0.238 (0.886) (0.250) (0.216) Fixed effects Yes Yes Yes Adjusted R 2 12.44% 14.17% 13.32% Number of obs. 817 817 817 Significant at the 0.01 level. Significant at the 0.05 level. Significant at the 0.10 level.

522 Financial Management Fall 2016 underperforming fund using a logistic regression, as in Equation (4). We retain one observation for each incubated mutual fund on the date that it applies for a ticker, as indicated in the NASD database. As demonstrated in Table IX, the regression coefficient for the Objective Flow variable is positive and significant. This finding is consistent with Ippolito (1992) and Khorana and Servaes (1999) and indicates that fund families will launch new funds into an objective if that objective is likely to generate sufficient increased inflows to benefit the family, regardless of fund performance. The negative coefficient for Incubation Period Risk indicates that underperforming incubated mutual funds are associated with less risk taking during the incubation period. This finding is consistent with the literature on mutual fund tournaments. Given that top performing funds have an increased likelihood of being launched, a mutual fund will take on additional risk in an attempt to reach the top levels of performance. Since underperforming incubated mutual funds do not take on the same level of risk as their outperforming counterparts, some underperforming funds may be incubated for the ex ante purpose of filling family needs rather than performance needs and, therefore, take on less risk from their inception. The results presented in Table IX are consistent with the view that underperforming incubated mutual funds are released from incubation due to the additional inflows that the mutual fund can generate for the fund family through market demand for the fund s objective rather than through outperformance. V. Conclusion In this paper, we examine how mutual fund incubation is used to benefit fund families. We begin by reexamining some of the results in Evans (2010) using an updated sample that extends from 1999 through September 2014. Our sample consists of 3,610 newly launched US domestic equity mutual funds, of which 817 (22.63%) are incubated. Consistent with Evans (2010), we find that, on average, incubated mutual funds generate increased performance during their incubation period, outperforming nonincubated funds by up to 3.41% risk adjusted annually. However, this performance reverses in the postincubation period, which supports the view that incubation is not used to identify skilled managers, but to establish a track record of artificial outperformance. If fund families are not using incubation to identify skilled managers, the benefit to the family from the incubation period outperformance must come in the form of additional inflows. By building on Evans (2010) analysis of the relationship between incubation and subsequent period inflows, we find that investors respond positively to mutual fund incubation and their response is driven by mutual funds with the highest incubation period performance. The recognition that outperformance of an incubated fund is not necessarily based on skill provides increased motivation for fund families to launch outperforming funds before this performance reverses. This view is supported by results from our investigation indicating that outperforming mutual funds are released sooner. The finding that fund families are more likely to introduce new funds into objectives that are doing well and attracting large inflows in an attempt to capture spillover benefits helps to explain incubation length and supports the use of incubation as a way to maximize revenue for the fund family through means in addition to performance. Finally, we look at the existence of underperforming incubated mutual funds. If the primary benefit to incubation is the establishment of a track record of outperformance (Evans, 2010), then underperforming mutual funds should not survive the process. However, in our sample of 3,610 mutual funds, 43.33% of released funds show negative objective alphas for incubation period performance. Our investigation establishes that underperforming funds are being released for