Window dressing in the Active Share scores in publicly reported portfolios

Andreu, Laura; Forner, Carlos; Sarto, José Luis Article Window dressing in the Active Share scores in publicly reported portfolios BRQ Business Research Quarterly Provided in Cooperation with: Asociación Científica de Economía y Dirección de Empresas (ACEDE), Madrid Suggested Citation: Andreu, Laura; Forner, Carlos; Sarto, José Luis (2024) : Window dressing in the Active Share scores in publicly reported portfolios, BRQ Business Research Quarterly, ISSN 2340-9444, Sage Publishing, London, Vol. 27, Iss. 3, pp. 324-345, https://doi.org/10.1177/23409444211024645 This Version is available at: https://hdl.handle.net/10419/327047 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ https://doi.org/10.1177/23409444211024645 Business Research Quarterly 2024, Vol. 27(3) 324 –345 © The Author(s) 2021 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/23409444211024645 journals.sagepub.com/home/brq Creative Commons Non Commercial CC BY-NC: This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 License (https://creativecommons.org/licenses/by-nc/4.0/) which permits non-commercial use, reproduction and distribution of the work without further permission provided the original work is attributed as specified on the SAGE and Open Access page (https://uk.sagepub.com/aboutus/openaccess.htm). Introduction The existence of window dressing (WD) in portfolio management has attracted increasing interest among academics and practitioners because of its managerial consequences. Previous literature describes WD as a management strategy intended to include attractive assets (hot stocks) and/or prevent unattractive assets in the reported portfolios to draw higher money flows. Different measures have been proposed to capture this phenomenon, depending on the reasons for WD behavior (see Agarwal et al., 2014). In this study, we analyze a different type of WD where managers attempt to increase the difference in the portfolio weights between their reported portfolio and the benchmark. Actively managed funds offer portfolios that attempt to outperform a reference benchmark by actively deviating from the benchmark weights overweighting (underweighting) the stocks that the manager considered good (bad) investments. Active mutual funds charge higher fees for this active management. Moreover, investors could consider large deviations from the benchmark a proxy for skill and consequently indicate higher probabilities for better future performance. Therefore, we suggest that fund managers could be tempted to artificially increase the deviation of their portfolio weights from the benchmark weights in the months when their portfolio holdings are made public to give an extra image of activeness. With this WD strategy, managers could better justify their high fees and also portray a higher image of skill to potential investors, which could consequently increase their money flows. We use the Active Share (AS) measure proposed by Cremers and Petajisto (2009) as an aggregate measure of how fund holdings deviate from the benchmark. This measure adds the absolute value of all the fund stock weight deviations from the benchmark weights. Although the work of Cremers and Petajisto (2009) was published after our sample period (2000–2006), the AS metric is so Window dressing in the Active Share scores in publicly reported portfolios Laura Andreu1, Carlos Forner2 and José Luis Sarto1 Abstract Using a unique database that includes publicly disclosed fund holdings at the end of the quarter as well as the holdings in all non-publicly disclosed months, we found that some funds could alter their portfolios in publicly disclosed months to artificially increase their Active Share scores and consequently appear more active and take advantage of the positive relationship between Active Share and money flows. We show how, consistent with non-informed trades, these funds erode their future performance. However, these funds reach their objective of increasing future money flows. Moreover, we find that window-dresser funds can be identified by controlling the level of tracking error. The funds with high Active Share scores and low tracking errors have the highest levels of Active Share window dressing and the worst future returns. However, compared with less active funds, they are able to capture higher money flows. JEL CLASSIFICATION: G23; G11 Keywords Active Share, monthly fund portfolio holdings, tracking error, window dressing 1 Department of Accounting and Finance, University of Zaragoza, Zaragoza, Spain 2 Department of Accounting and Finance, University of Alicante, Alicante, Spain Corresponding author: Laura Andreu, Department of Accounting and Finance, University of Zaragoza, Gran Vía 2, 50005 Zaragoza, Spain. Email: [email protected] 1024645BRQ0010.1177/23409444211024645Business Research QuarterlyAndreu et al. research-article2021 Regular Paper Andreu et al. 325 intuitive that we consider it a good proxy of how investors compute aggregate fund holding deviations from the benchmark in the sample period. “Hot stocks” and “AS” WD practices refer to portfolio holding movements before the fund portfolio becomes public with the same final objective of increasing money flows. However, they are different. Note that when “hot stocks” WD managers buy (sell) attractive (unattractive) stocks, which are overweighted (underweighted) in the portfolio, this trade will necessarily increase stock deviation from benchmark, and therefore increase AS levels. However, when “hot stocks” WD managers buy (sell) attractive (unattractive) stocks which are underweighted (overweighted) in the portfolio, this trade could reduce the stock deviation from benchmark, and therefore reduce AS levels. To compute AS WD, it is necessary to have a database that includes fund holdings both in the months during which the fund portfolios are made public to investors and the months during which these fund portfolios are not made public. We employ a unique database for the Spanish mutual fund market that includes both publicly disclosed fund holdings at the end of the quarter (March, June, September, or December) and the fund holdings in all nonpublicly disclosed months (non-quarter-end months). This information is not publicly available to investors as the Spanish official supervisor provides non-quarter-end fund holdings for research purposes only. This makes this database especially useful for our aim. Moreover, this database is free of any selection bias, given that this information is available to all mutual funds and is provided by the official supervisor rather than a private supplier. The quality of the information available in the Spanish market allows us to detect the potential WD in the AS levels in publicly reported months, as we can compute how the AS in public quarter-end months differs from the AS in the two previous and the following two non-public months. We refer to this AS WD proxy as AS_difference. This study cannot be extended to other eurozone markets.1 However, our results are relevant to the worldwide mutual fund industry. Cremers and Petajisto (2009) argue that AS scores based on reported fund holdings at the end of a quarter are unlikely to suffer any potential WD because the portfolio distortion supposes a high turnover cost. However, evidence of AS WD in the Spanish mutual fund market would mean that fund managers worldwide could have enough incentives to alter their portfolios despite the turnover costs. Moreover, the Spanish domestic mutual fund market exhibits some characteristics (small size of the Spanish Stock Exchange) that make it difficult for managers to deviate from the benchmark. If Spanish managers find enough incentives to alter their AS despite the difficulties of deviating from the benchmark, managers from other countries with fewer difficulties could have even more incentives to window dress the level of AS. This study’s main contribution is testing, for the first time in literature, whether mutual funds alter their holdings in publicly disclosed months to appear more active and how this WD strategy influences future fund performance and money flows. We also contribute to the WD literature by examining the undisclosed information. Few studies (e.g., Elton et al., 2010; Morey & O’Neal, 2006; Musto, 1997; Ortiz et al., 2012) are able to use both disclosed and undisclosed information. However, in most cases, there exists some reporting selection bias. In our study, the incentives to alter AS in publicly reported months depend on whether the AS score is positively related to the future performance of fund (manager skills) and future money flows. In the first part of the study, we demonstrate that the higher the AS level, the higher the future fund returns and money flows. Moreover, we also investigate whether this relation is robust for the inclusion of tracking error (TE) as an alternative measure of fund activeness. If some managers alter their AS levels, they will increase this measure before publicly disclosed months and revert to their previous levels afterward. In this case, managers deviate the portfolio holdings from the benchmark only to offer a temporal higher level of activeness when they are going to make the information public, and not because the manager is overweighting (underweighting) stocks that, according to the manager’s skill, are going to be winners (losers). This AS increase is artificial and not related to skill. In addition, if AS is a proxy for management skills, the skills should not differ between publicly and non-publicly disclosed months. Therefore, this WD strategy will erode fund performance because of the increase in turnover and expense ratios. However, if some managers decide to modify the AS level despite its expected negative impact on fund performance, it would be because this practice allows for future money flows to increase. Consistent with this idea, we find that high levels of AS in publicly reported months regarding the AS levels in the two previous and the following two non-publicly reported months (AS_difference) are related to worse future fund performance and better future money flows. Our analysis also focuses on detecting funds that are more likely to distort the AS level. Given that TE cannot be modified easily, we suggest that funds with high AS but low TE levels are more likely to alter their AS levels. We consistently find that contrary to the other categories of funds, on average, funds with high AS but low TE levels increase their AS levels before their holdings are made public and reduce them afterward. They therefore show significant and positive AS_difference values. Moreover, this category of funds shows the worst future performance but attracts higher money flows than funds with similar levels of TE. Finally, given the findings of Golez and Marin (2015), who observe that Spanish bank-affiliated mutual funds systematically increase their holdings in the controlling 326 Business Research Quarterly 27(3) bank stock around some events, we also analyze the possible AS WD strategies for subsamples of non-independent funds (funds managed by management companies that are integrated in a financial group conglomerate) and independent funds. We consistently find that the negative impact of AS WD at quarter-ends on fund performance is exclusive to non-independent funds. This negative impact is not observed in independent funds. The rest of this article is organized as follows. The “Literature review and hypotheses” section presents the literature review and the hypotheses of the study. The “Data and measures” section describes the data and measures used in this study. The “Empirical models and methods” section describes the empirical models and methods used to test the hypotheses of this study. The “Main results” section presents the main findings of our hypotheses. The “Characteristics of stocks traded by AS window dressers: Is AS WD a practice different from ‘hot stock’ WD?” section examines the characteristics of stocks traded by AS window dressers. The section “Examining the funds that are more prone to window dress the AS level” examines the characteristics of funds that are more likely to alter the AS level. The “Robustness analyses” section presents some robustness analyses. The “Conclusion” section concludes the study. Literature review and hypotheses The information content in publicly disclosed fund holdings has attracted the interest of many studies. One important line of research, which is related to the existence of a conflict of interest between fund managers and investors, is the analysis of WD in fund holdings. Haugen and Lakonishok (1988) and Ng and Wang (2004) consider WD in fund holdings as the main factor that drives high returns for small and recent loser stocks after the year-end. Most studies examine the influence of this practice on price anomalies, such as the January effect, rather than examining the existence of this practice and its underlying motivations (see, for example, Agarwal et al., 2014). Specifically, Agarwal et al. (2014) highlight that there is limited understanding of the incentives for managers to engage in WD. They affirm that if WD is rewarded with higher flows, one might wonder why not all fund managers follow such activities. To disentangle this problem, these authors suggest that WD is a risky decision because its success in attracting higher flows depends on fund performance in the delay period between the end of the month when the portfolio is publicly disclosed and the date when this information is publicly disclosed to investors.2 Hence, the main motivation for this practice is the perception of managers that portfolio disclosures significantly influence the opinions of investors regarding their professional skills. Managers are tempted to improve the appearance of their reported portfolio (i.e., including/deleting attractive/unattractive assets) before presenting them to clients or shareholders to attract larger money flows. The existence of WD is problematic for investors who seek reliable information on fund holdings that will help them allocate their money more appropriately. Another line of research related to the information content in public fund holding disclosures uses these data to measure the level of active management and its relationship with fund performance. The first study to analyze this relation was Wermers (2003) which does not measure the level of active management using holding disclosures but uses return information through the TE. More recently, other studies have started analyzing active management using portfolio holdings information, which “a priori” provides a more comprehensive image of the activeness of portfolio managers. Kacperczyk et al. (2005) examined active management using portfolio concentration. More recently, Cremers and Petajisto (2009) analyzed how fund holdings deviate from benchmarks through the AS measure. These studies found a positive relationship between the level of active management and fund performance. The AS measure has attracted substantial attention both in academia and in the management industry. Moreover, more active equity funds and institutional money managers currently report their AS. An example of this impact is the article published by Frazzini et al. (2016) which reviews the great impact of AS on the investment community. They indicate that institutional investors are more focused on asset managers with a high AS and that some institutional investors have even embedded a high AS requirement in their investment guidelines. Another example of the impact of AS in the fund community is Caquineau et al. (2016) from Morningstar, who advise investors not to rely solely on AS when selecting funds. In this study, we link both lines of research (WD and active management) by examining a different practice of WD, that is, the report of a high level of differentiation from the benchmark on disclosure dates (AS WD) in the Spanish equity market.3 Given that the incentives to distort the AS level depend on whether high AS levels in publicly reported portfolios send a positive signal to potential investors, we first analyze whether the US evidence of the effect of AS on future fund returns is observed in the Spanish market. Hence, our baseline hypothesis is as follows: Hypothesis 1 (H1): AS level positively influences subsequent fund performance. We then contribute to the literature examining the following novel research hypotheses: Hypothesis 2 (H2): AS level positively influences subsequent money flows. Hypothesis 3 (H3): AS WD negatively influences subsequent fund performance. Andreu et al. 327 Hypothesis 4 (H4): AS WD positively influences subsequent money flows. Data and measures To analyze WD involving how fund holdings deviate from benchmarks, we use the AS measure proposed by Cremers and Petajisto (2009), which compares the holdings of a mutual fund with the holdings of its benchmark index. Specifically, the AS of a given mutual fund p in a given month t is defined as follows AS ww pt pi bi i N ,, , =− = ∑ 1 21 (1) where wpi, and wbi, are the portfolio weights of asset i in mutual fund p and benchmark b.4 We also compute the traditional TE measure of fund activeness to be used as a control variable in our analysis of AS WD. The TE measure is commonly defined as the time-series standard deviation of the difference between a fund return () , Rpt and its benchmark return () , Rbt . The TE of a given mutual fund p in a given month t is calculated as follows TE StdevR R pt pt bt ,, , =−    (2) The TE at the end of a given calendar month was calculated using the time series of the previous 120 daily returns. It was later annualized. As expected, the AS and TE measures were positively correlated (.5046). Alternatively, we also calculate the TE by regressing the excess fund returns over the risk-free rate on excess benchmark returns, as follows5 RR RR pt ft pbtftp pt,, ,, , −=+−   + αβ ε (3) TE Stdev pt beta pt ,, =   ε (4) To compute the AS of mutual funds, we require data on the portfolio composition of mutual funds, as well as their benchmarks. The mutual fund holdings database used in this study comprises the monthly portfolio reports of all actively managed Spanish domestic equity funds. This information differs from the previous literature that analyzes quarterly holdings, which implies the loss of interim trades; therefore, it is impossible to test a change in the fund activity level in these interim months. The regulator of the Spanish market, the Spanish Securities and Exchange Commission (CNMV), supplied us with this monthly database until December 2006, after which the information is provided on a quarterly basis (calendar quarter-ends).6 The database is free of survivorship bias because it also includes dead funds. In addition, the database overcomes any possible problem of reporting selection bias that may be present in the scarce research using high-frequency portfolio information, where mutual funds voluntarily supply their portfolio holdings to private data providers (see, for example, Elton et al., 2010). Our main objective is to test the possible existence of WD at the level of AS shown by equity mutual funds on publicly reported dates (quarter-end months). We therefore focus our study on the time period in which the portfolio holdings of all mutual funds are reported on a monthly basis (1999–2006).7 The initial sample included 169 Spanish domestic equity mutual funds. From this sample, we eliminate passive management funds, such as index funds or exchange traded fund (ETFs), and a couple of funds oriented to Small or Mid-Caps. We begin our analysis in December 2000 because of the requirement of data intended to estimate the momentum factor. The final sample comprises 137 funds from December 2000 to December 2006. The CNMV mutual fund database provides information on the portfolio holdings and investment vocations of Spanish mutual funds. It also contains other characteristics of mutual funds such as daily and monthly returns, total net assets (TNA), number of investors, annual management fees, fund age, and management company. Panel A of Table 1 reports the number of funds examined annually, the average TNA of the funds, the average number of investors, the average number of stocks held by mutual funds, the average number of management companies examined, and the average annual turnover ratio. Specifically, the average TNA of the funds in our sample is €68 million, and the average number of investors is approximately 2,800. Furthermore, the average number of securities held by a fund is 40. Panel B of Table 1 reports the share of fund portfolios in the main types of securities (stocks, fixed-income securities, other mutual fund units, and cash and cash equivalents) across the years studied. As expected, according to the investment vocation of the funds examined, the main investment is in domestic stocks and, more precisely, in domestic stocks belonging to the Ibex35 index. This panel also shows that the percentage invested in fixed income and other mutual fund units is relatively small. In conclusion, the low percentage of non-controlled securities (less than 1% of the portfolios) reinforces the quality of our database. We include the Ibex35 index as a benchmark for all funds because it is the most important large-cap index for the Spanish stock market and is, therefore, the most popular self-reported benchmark for our sample of Spanish domestic equity funds (these funds invest, on average, more than 61% of the portfolio in Ibex35 stocks, as shown in Table 1). For all funds, we used the most relevant benchmark for Spanish large-cap equity instead of using the benchmark self-reported by the manager in the fund 328 Business Research Quarterly 27(3) prospectus because managers can report a misleading benchmark that is easily beaten (Sensoy, 2009). For robustness, we also conducted two analyses: (1) considering the index that produced the lowest AS between Ibex35 index and IGBM (Índice General de Bolsa de Madrid), as in Petajisto (2013) and Cremers and Petajisto (2009), and (2) limiting the analysis to funds that self-reported Ibex35 or IGBM as the benchmark.8 We have historical month-end constituents for the Ibex35 index and IGBM index, and the daily and monthly returns of their constituents from DataStream. This historical information allows us to consider new benchmark constituents and deletions from the benchmarks. All stock holdings, for both mutual funds and benchmarks, were matched with the stock returns through the international securities identification numbering system (ISIN) code for each security. Table 2 shows some descriptive statistics on the distribution of funds along the two dimensions of active management (AS and TE). At each calendar quarter-end (publicly reporting dates), the funds are independently sorted by their AS and TE. We used 20%, 40%, 60%, and 80% AS cuts and 4%, 8%, and 12% TE cuts (% per year). Panel A reports the minimum, average, and maximum number of funds in this bivariate distribution for the sample period and in the univariate marginal distributions along the AS and TE dimensions. The Spanish univariate AS distribution differs slightly from that observed in US equity funds. Most of the Spanish domestic equity funds (85% of the funds on average) have an AS lower than 60%, which is the opposite of the distribution observed in the US market. For example, in 2002, only 457 of 1,678 funds (27%) had AS levels lower than 60% in the United States (Table 1, Cremers & Petajisto, 2009) and 82% in Spain. Therefore, the majority of domestic equity funds in Spain fall in the category that Cremers and Petajisto (2009) refer to as closet indexers, that is, funds with low AS levels that claim to be active. Regarding the percentage of assets under management, the figures from the US and Spanish markets also differ significantly. For example, in December 2006, in the United States, less than 40% of total assets were managed by funds with AS levels lower than 60% (Figure 5, Petajisto, 2013); these accounted for 81% in the Spanish market. Similarly, European Securities and Markets Authority (ESMA, 2016) analyzes the Undertakings for the Collective Investment in Transferable Securities (UCITS) equity funds domiciled in European Union (EU) member states for the period from 2005 to 2012 and finds that 15% of funds have AS levels less than 60% and TE levels less than 4%. However, this type of fund represented 43% of the Spanish sample for the same period.9 ESMA (2016) also measures the percentage of funds with an AS Table 1. Summary statistics of Spanish domestic equity funds over time. 2000 2001 2002 2003 2004 2005 2006 Panel A No. of funds 101 101 97 97 95 102 100 TNA (in thousand euros) 72,819 59,723 50,258 45,241 66,630 79,617 86,839 No. of investors 2,848 2,749 2,715 2,582 2,861 2,837 3,020 Average number of stocks held 43 39 37 36 39 41 41 No. of management companies 55 55 53 53 51 55 53 Turnover 58.52% 42.84% 40.44% 37.35% 35.67% 40.70% 44.64% Panel B Stocks 81.69% 78.32% 78.36% 77.21% 76.52% 79.56% 82.24% Spanish 71.40% 69.39% 70.70% 71.19% 70.62% 74.60% 77.93% Ibex 61.42% 58.92% 58.68% 60.62% 59.73% 62.73% 65.13% Non-Ibex 9.99% 10.48% 12.02% 10.57% 10.89% 11.88% 12.80% European 8.76% 8.08% 7.13% 5.78% 5.74% 4.81% 4.23% Other 1.53% 0.85% 0.53% 0.24% 0.16% 0.15% 0.08% Fixed income 4.97% 5.38% 4.16% 3.23% 3.24% 3.67% 2.13% Other mutual fund units 0.39% 0.20% 0.13% 0.16% 0.09% 0.06% 0.21% Cash and cash equivalents 12.01% 15.10% 16.88% 18.92% 19.72% 16.31% 15.09% Non-controlled securities 0.95% 1.00% 0.47% 0.48% 0.42% 0.39% 0.33% Total 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% 100.00% TNA: total net assets. Note: The table is split into Panel A and Panel B. Panel A reports the annual characteristics of mutual funds, such as the number of funds analyzed, the TNA, the number of investors, the average number of stocks held by the portfolios, the number of management companies involved each year, as well as the annual turnover. Panel B reports the portfolio share of fund portfolios for the main types of securities across the years. The assets invested by funds are classified as follows: stocks (Spanish, European, and others), fixed income, other mutual fund units, cash and cash equivalents, and non-controlled securities. Spanish stocks are split into Ibex35 stocks and non-Ibex35 stocks. The data correspond to the average of the portfolios that report each year. Andreu et al. 329 Table 2. Descriptive statistics depending on the level of active management. Active Share Tracking error (% per year) Low (<4%) 4%–8% 8%–12% High (>12%) All Panel A: Minimum/average/maximum number of mutual funds High (>80%) 0/0/0 0/1/4 0/0/4 0/2/6 2/3/6 60%–80% 0/0/4 0/4/9 0/1/7 0/3/11 7/10/13 40%–60% 0/9/20 1/8/16 0/2/14 0/1/6 15/21/29 20%–40% 9/33/51 0/7/22 0/2/16 0/0/4 37/44/51 Low (<20%) 5/9/18 0/1/5 0/0/2 0/0/1 6/11/19 All 14/52/79 13/22/36 0/7/30 0/7/23 Panel B: Median TNA (in thousand euros) High (>80%) 31,959.25 31,692.60 60%–80% 23,510.54 12,022.00 3,634.73 13,239.92 40%–60% 40,934.90 22,147.63 16,898.50 3,344.75 26,154.98 20%–40% 54,646.38 38,221.38 19,844.08 44,307.14 Low (<20%) 52,448.58 15,099.00 44,019.52 All 54,462.62 24,408.86 18,555.32 6,395.82 32,321.16 Panel C: Average annual management fee (%) High (>80%) 1.62 1.61 60%–80% 1.73 1.81 1.58 1.69 40%–60% 1.83 1.86 1.84 2.04 1.84 20%–40% 1.92 1.75 1.91 1.88 Low (<20%) 2.06 1.84 2.03 All 1.94 1.79 1.84 1.66 1.85 Panel D: Average annual turnover (%) High (>80%) 48.02 47.91 60%–80% 44.97 63.53 58.49 55.07 40%–60% 44.08 54.39 48.24 87.60 49.96 20%–40% 30.16 37.46 32.21 32.55 Low (<20%) 15.76 13.05 15.19 All 29.32 42.84 41.79 51.01 36.78 Panel E: Average percentage of independent funds High (>80%) 56.67 54.00 60%–80% 58.50 53.33 65.50 56.14 40%–60% 24.47 29.24 8.15 90.00 26.03 20%–40% 5.67 18.15 47.38 11.71 Low (<20%) 7.40 20.00 12.89 All 7.70 32.72 37.20 63.73 22.70 Panel F: Percentage of funds that charge fees over performance High (>80%) 0.00 3.33 60%–80% 5.06 3.57 13.56 7.86 40%–60% 21.83 12.53 24.91 36.67 20.35 20%–40% 9.59 17.58 18.17 12.30 Low (<20%) 5.33 20.00 5.15 All 10.27 14.05 14.06 17.05% Panel G: Net return adjusted to control portfolios by TNA, flows, and turnover (50th percentile) (%) High (>80%) 2.96 2.36 60%–80% −0.06 0.62 0.88 0.26 40%–60% −0.15 −0.19 −0.70 −0.11 −0.16 20%–40% −0.26 −0.12 0.04 −0.13 Low (<20%) −0.32 −0.12 −0.25 All −0.26 −0.08 −0.21 1.48 TNA: total net assets. Note: This table reports the descriptive statistics of Spanish domestic equity mutual funds, depending on their level of active management, measured through the AS and TE. At each calendar quarter-end (publicly reporting dates), the funds are independently sorted as per their AS and TE. We used 20%, 40%, 60%, and 80% AS cuts and 4%, 8%, and 12% TE cuts (% per year). Panel A reports the minimum, average, and maximum number of mutual funds; Panel B shows the median TNA for each calendar quarter-end and partition with at least five funds; Panel C shows the average annual management fees charged by the funds; Panel D shows the average annual turnover; Panel E shows the average percentage of funds managed by an independent management company (as opposed to bank-affiliated funds); Panel F shows the percentage of funds that charge fees over performance; and Panel G shows TNA, flows, and turnover control portfolio adjusted returns. The sample period is December 2000 to December 2006. 330 Business Research Quarterly 27(3) level of less than 50% and a TE level of less than 3%, as these cuts could be more indicative in member states with relatively small equity markets. This report finds that only 7% of funds fall into this category. However, these funds represent 32% of the funds in our sample. The high fraction of closet indexers in the Spanish mutual fund market can be explained by the smaller size of the Spanish Stock Exchange in comparison to other exchanges, such as those in the United States. The average low level of activeness in the Spanish mutual fund industry can also be explained by the low levels of competitiveness. According to Cremers et al. (2016), markets with less explicit indexing funds are less competitive, and active funds have lower AS levels. Therefore, the low supply of explicit indexing funds in Spain (only 9% of TNA as of December 2010 as stated by Cremers et al., 2016) is another characteristic of the Spanish industry. The bivariate distribution of funds shows the expected positive relationship between the two active management dimensions. The number of funds with low AS and high TE and vice versa is very low. Nevertheless, as in Cremers and Petajisto (2009), both variables show substantial dispersion in the other dimension. Therefore, both active management measures could have additional explanatory power in terms of fund performance. This finding confirms that it is important to distinguish between the two dimensions of active management. Panel B of Table 2 reports the median TNA of funds in the AS–TE bivariate distribution. The size of mutual funds tends to decrease monotonically when shifting from the less active to the most active funds. However, this tendency disappears for funds with extreme AS values (more than 80%). This relatively large size of Spanish mutual funds with AS values higher than 80% explains why the percentage of Spanish closet indexers is less extreme when assets under management are considered instead of the number of funds. However, mutual funds with low TE values are the largest group represented in the sample, as in Cremers and Petajisto (2009). Panel C of Table 2 reports the average annual management and custodial fees charged by Spanish equity mutual funds depending on the bivariate distribution. Contrary to what one could expect, Panel C shows that Spanish domestic equity funds with the lowest AS values charge higher annual management and custodial fees than the remaining funds. This puzzling fact is somehow similar to that found in Gil-Bazo and Ruiz-Verdú (2009), where funds with poor before-fee performance charge higher fees. This finding can be explained by the presence of investors with different levels of sophistication and degrees of sensitivity to past performance. Therefore, we can conclude that the Spanish fund industry is characterized by closet indexers that charge high management fees providing only index-like returns. Panel D reports an average portfolio turnover of 37% per year, which is lower than the 95% reported by Cremers and Petajisto (2009). The funds with low levels of AS and TE are those with less turnover (lower than 20%). However, we do not find the expected positive relation along the entire distribution. Panel E shows the average percentage of funds managed by independent management companies. This percentage of funds managed by independent management companies increases with the AS and TE levels. This finding is related to the work of Golez and Marin (2015), who find that Spanish bank-affiliated mutual funds systematically increase their holdings in the controlling bank stock during certain events, such as price drops, indicating that management company ownership matters. Díaz-Mendoza et al. (2014) found that performancebased fee funds perform significantly better than the other risky funds considered in the Spanish mutual fund industry. Panel F shows the percentage of funds that charge fees based on performance. The results do not support a relationship between fund activeness and this variable. Finally, Panel G compares the performance mean of the more active funds versus the more passive funds controlling for TNA, flows, and turnover via “control portfolios.” Specifically, net returns are adjusted considering the excess fund net return in a given month in comparison with the “expected” net return according to its TNA, flows, and turnover. Each month, we construct two categories for each of the three above-mentioned variables using the 50th percentile to obtain a sufficient number of portfolios in each of the eight categories. We compute the equally weighted average net returns for each category. We then match each mutual fund month with the TNA– flow–turnover category to which the fund belongs in that month and calculate its adjusted net return as the difference in their net returns. We find that the adjusted returns of control portfolios are positively related to both measures of activeness (AS and TE) and that the funds with the highest levels of AS and TE have the best performance. To compute the money flows of mutual funds, we follow the absolute and relative money flow measures proposed in previous literature (see, for example, Guercio & Tkac, 2002; Sirri & Tufano, 1998) Absolute Flow TNATNA R it it it it,,, =− ⋅+ () −11 (5) RelativeFlow TNATNA R TNA it it it it it , ,, , , =−⋅+ () − − 1 1 1 (6) where TNAi,t is the total net assets of fund i in month t, and Ri,t is the return of fund i in month t. Although the absolute flows are relevant from a management perspective because the vast majority of the funds Andreu et al. 331 examined charged their management fees based on TNA, the previous literature has also recommended the use of relative flows when the size of mutual funds is quite dispersed. To avoid potential biases derived from disparities among the average size of the funds according to their level of AS (see Panel B of Table 2), we additionally propose an approach in which money flows are adjusted considering the excess flow attracted by a fund in a given quarter in comparison with the “expected” flow according to its size. Specifically, each month, we construct the TNA deciles and compute their equally weighted average flows. We then match each mutual fund month with the TNA decile in which the fund belongs in that month and calculate its adjusted money flow as the difference in the absolute flows attracted. Finally, as a proxy to detect AS WD, we compute AS_difference, which is the difference between the AS level in the publicly reported month (end of quarter) and the average AS level in the surrounding non-publicly reported months AS differenceASAS AS AS AS tt tt tt _=− +++ −− ++ 21 12 4 (7) Empirical models and methods The influence of active management on future fund performance Previous studies show that the average mutual fund slightly outperforms the market return before expenses and fails to outperform the market return after expenses (see, for example, Daniel et al., 1997; Fama & French, 2010; Jensen, 1968). Recent studies examining mutual fund trades have demonstrated the superior ability of certain active mutual funds. Chen et al. (2000), Alexander et al. (2007), Baker et al. (2010), and Jiang et al. (2014) find support for the hypothesis of the trading skills of mutual fund managers.10 Outperformance can only arise from active management; therefore, there should be cross-sectional differences in fund performance depending on the level of AS. Actively managed funds offer a portfolio that attempts to outperform a reference benchmark by actively deviating from the benchmark weights by overweighting (underweighting) the stocks that the manager considered good (bad) investments. Active mutual funds charge higher fees for active management. Therefore, information regarding the level of active management (AS) is important. It is impossible to beat a benchmark without some level of AS, that is, the lower the level of AS, the lower the possibility of outperforming the benchmark. However, a high level of AS is necessary but is not a sufficient condition to beat the benchmark; this depends on the skill of the manager. Accordingly, Cremers and Petajisto (2009) demonstrate that the AS level in the US market is positively related to fund performance (manager skill). Given that the incentives to window dress the AS level depend on whether high levels of AS in publicly reported portfolios send a positive signal to potential investors, we first analyze whether the US evidence of the positive effect of AS on subsequent fund performance is observed in the Spanish market (H1). To study the relationship between AS and future fund performance, we run the following pooled panel regression AdjRetur nAST E AdjReturn control it it it it _ _ ,, , , +=+ + + + 10 12 3 αα α α ssit it,, + ε (8) where the dependent variable is the gross or net benchmark adjusted return of mutual funds in month t + 1, while the independent variables are the AS and TE levels of mutual funds, the Adj_Return, and some control variables measured in month t. Specifically, we use the annual turnover, annual management fees, log(TNA), log(TNA)2, number of stocks, fund age, number of investors, and annual relative money flows (flows/TNA) as control variables. Year fixed effects are included in all the specifications. The influence of active management on future money flows If AS influences future fund performance (H1), we should analyze whether this positive signal given by reporting high levels of AS makes investors more likely to invest in these funds and therefore influences the level of future money flows captured by mutual funds (H2). To test this hypothesis, we examine the relationship between the AS level and the future money flows received by each mutual fund, including some control variables that could also determine the magnitude of the investment flows, as shown in the following regression Flow AS TE AdjReturncontrols it it it it it ,,, ,, _ +=+ + ++ + 1012 3 αα α αε iit, (9) where the dependent variable is the fund money flow in quarter t + 1, defined as the absolute, relative, and adjusted money flows (see equations (5) and (6)). Adjusted money flows are flow deviations from their respective TNA decile flows. The independent variables are the levels of AS and TE of the mutual funds at the end of quarter t, the Adj_ Return, and the control variables explained in equation (8). All the explanatory variables are measured at the end of quarter t (publicly disclosed months). Year fixed effects are included in all the specifications. This pooled regression has been conducted on a quarterly basis because the portfolio holdings in the months with no public disclosure are not known to investors. Therefore, they cannot affect future money flows. 338 Business Research Quarterly 27(3) ∆ AS DPublic controls it it it it ,,,, _=+ ++ αα ε 01 (14) where the dependent variable is the change in AS in month t, α0 is a constant, and D_Public is a dummy variable that equals 1 for publicly disclosed months (calendar quarterends), and 0 otherwise. We run this regression for the two AS categories and for each of the four AS–TE fund categories. Once a fund is ranked in a given category in each calendar quarter-end, it remains in the same category during the following 2 months. The control variables are TE, Adj_ Return, and those defined in equation (8). Therefore, the constant coefficient measures the average ∆AS in non-publicly reported months, and the coefficient of the dummy measures how the average ∆AS in publicly reported months deviates from the former. Table 10 shows the results. Table 9. AS level at quarter-end months and the surrounding months. Panel A TE AS Low (80%) High (20%) All Low (80%) Min/aver/max no. 66/69/73 3/4/8 69/73/79 High (20%) Min/aver/max no. 3/4/8 12/14/16 17/18/20 All Min/aver/max no. 69/73/79 17/18/20 Panel B TE AS Low (80%) High (20%) All Low (80%) AS_difference (%) −0.42*** (.00) −1.12*** (.01) −0.46*** (.00) Month – 1 AS (%) 31.54*** (.01) 35.30*** (.01) 31.77*** (.00) Ranking AS (%) 31.18 33.41 31.30 Month + 1 AS (%) 31.60** (.02) 33.38 (.71) 31.71** (.02) Month + 2 AS (%) 31.63* (.07) 33.99 (.67) 31.79* (.06) High (20%) AS_difference (%) 2.04*** (.00) −0.08 (.60) 0.37** (.04) Month – 1 AS (%) 58.45*** (.00) 73.48 (.13) 70.21 (.11) Ranking AS (%) 60.76 73.22 70.46 Month + 1 AS (%) 59.88 (.29) 73.55 (.59) 70.60 (.90) Month + 2 AS (%) 58.30*** (.01) 73.07 (.32) 69.94* (.06) All AS_difference (%) −0.29** (.02) −0.35** (.04) Month – 1 AS (%) 33.06* (.08) 65.03*** (.00) Ranking AS (%) 32.84 64.33 Month + 1 AS (%) 33.11 (.15) 64.93 (.36) Month + 2 AS (%) 33.05 (.44) 64.74 (.81) AS: Active Share; TE: tracking error. Note: This table contains two panels. Panel A shows the minimum, average, and maximum number of funds in each category, depending on the level of active management measured through the AS and TE. Panel B shows the AS level at calendar quarter-end months when portfolio holdings are publicly reported (ranking AS) and the AS level in the previous month (month – 1) and the next non-publicly reported months (month + 1 and month + 2) for the sample period. In addition, the table shows the AS_difference, defined as the difference between the AS level of the public month and the average AS in the 4 non-public months surrounding that month. Every 3 months, the funds were independently ranked according to their AS and TE levels. Then, we construct two portfolios for each variable using the 80th percentile and four portfolios from their intersections. The p values (in parentheses) report the significance of the paired test of means between the AS in the ranking month and that in the surrounding months. *, **, and *** indicate significance at .1, .05, and .01 levels, respectively. Andreu et al. 339 Table 10 shows that, on average, Spanish equity mutual funds tend to reduce the AS level in publicly reported months in relation to the average AS change in non-publicly reported months (a statistically significant −0.41%). This finding is consistent with the evidence provided in Table 8 and with the results provided in Table 9. This result can be partially explained by the fact that most of the funds are classified as low AS and low TE. Furthermore, these funds attempt to appear less active in quarter-ends, probably because they are oriented to conservative investors looking for low active management. When the different fund categories according to their active management are analyzed, we observe that this tendency to reduce AS in publicly reported months is exclusive to low AS funds. The changes in the AS levels in the high AS and low TE funds are the opposite. These funds tend to increase the AS level in publicly reported months relative to the average AS change in non-publicly reported months (a statistically significant 3.28%). Finally, as expected, funds with both high AS and TE do not exhibit different behavior in their monthly AS changes because this category gathers these mutual funds with real active management captured by both metrics, that is, AS and TE. The results are consistent when the control variables and time fixed effects are included. In summary, this study provides evidence of a certain type of WD to appear more active conducted by mutual funds with a high level of AS and a low level of TE when portfolios are publicly available to financial market participants. Robustness analyses High AS–low TE funds and future performance In the previous section, we found that funds with high AS but low TE levels exhibit higher levels of AS_difference. In this section, we attempt to provide additional evidence by analyzing the future performance of the four fund categories. If this measure actually measures AS WD, then we should expect the high AS and low TE fund category to exhibit worse performance than the funds with a similar level of AS, but with no distortion, that is, high AS and high TE funds. Table 11 shows the annualized values of the benchmark adjusted returns (Panel A) and the four-factor alphas (Panel B) in quarter t + 1 for each of the four categories of funds and for both gross and net returns.19 Funds with high AS and low TE levels not only exhibit worse performance than funds with high AS and TE, they are the fund Table 10. AS level in publicly disclosed months versus non-publicly disclosed months: pool regression. Fund categories Low AS Low TE Low AS High TE High AS Low TE High AS High TE Low AS High AS All Panel A: Change in AS in publicly disclosed months Constant 0.19*** (.00) 0.63 (.21) −1.10*** (.00) −0.20 (.12) 0.22*** (.00) −0.39*** (.00) 0.10** (.04) Dpublic −0.57*** (.00) −2.50* (.10) 3.28*** (.00) −0.15 (.39) −0.67*** (.00) 0.61** (.02) −0.41*** (.00) n4,908 274 275 992 5,182 1,267 6,586 R2 (%) 0.60 4.49 6.44 0.04 0.77 0.45 0.27 Adjusted R2 (%) 0.58 4.14 6.10 −0.06 0.75 0.37 0.25 Panel B: Change in AS in publicly disclosed months with control variables and time fixed effects Dpublic −0.53*** (.00) −1.32 (.12) 3.05*** (.00) −0.15 (.35) −0.57*** (.00) 0.58** (.02) −0.33*** (.01) Control variables Yes Yes Yes Yes Yes Yes Yes Time fixed effects Yes Yes Yes Yes Yes Yes Yes n4,695 238 255 965 4,933 1,220 6,184 R2 (%) 1.97 9.60 11.59 1.71 1.69 1.78 1.27 Adjusted R2 (%) 1.62 2.61 5.25 −0.06 1.35 0.39 0.99 AS: Active Share; TE: tracking error. Note: This table shows the impact of the public reporting dates on the change in the AS level of Spanish domestic equity mutual funds depending on the level of activeness through a pooled regression. At calendar quarter-ends, funds are ranked according to their AS and TE levels into four categories using the 80th percentile. Once a fund is ranked in a given category in each calendar quarter-end, it remains in the same category during the next 2 months. Specifically, Panel A reports the results of estimating a pooled regression where the dependent variable is the change in the level of AS of each mutual fund in month t, DAS, while the independent variable is a constant and a dummy variable, DPublic, that equals 1 for publicly disclosed months (March, June, September, and December). Panel B reports the estimation results, including the following control variables: the TE, the fund adjusted return, and the control variables defined in Table 3 for month t. The control variables with annual frequency remained constant in all months of the calendar year. The p values (in parentheses) are based on standard errors clustered by the fund. *, **, and *** indicate significance at .1, .05, and .01 levels, respectively. 340 Business Research Quarterly 27(3) category with the worst performance. The funds included in this category fail to reach the expected levels of good performance related to their high level of AS. Table 11 also shows other interesting relations that are different from those observed in the US market. First, we found that neither AS nor TE can provide a positive and Table 11. Annual performance according to the level of active management. Panel A TE AS Low (80%) High (20%) All High–low Low (80%) Gross adjusted return (%) −0.76 (.15) −0.29 (.92) −0.76 (.21) 0.47 (.85) Net adjusted return (%) −2.66*** (.00) −2.20 (.45) −2.67*** (.00) 0.46 (.86) High (20%) Gross adjusted return (%) −3.28** (.05) 5.81 (.11) 3.43 (.26) 9.08*** (.01) Net adjusted return (%) −5.10*** (.00) 4.16 (.26) 1.75 (.56) 9.26*** (.01) High–low TE High–low AS All Gross adjusted return (%) −0.93* (.09) 4.24 (.22) 5.17* (.10) 4.19* (.10) Net adjusted return (%) −2.83*** (.00) 2.53 (.46) 5.37* (.08) 4.41* (.08) Mix H-L High-low Gross adjusted return (%) −2.52* (.09) 6.10*** (.00) 6.56** (.05) Net adjusted return (%) −2.44* (.10) 6.36*** (.00) 6.82** (.04) Panel B TE AS Low (80%) High (20%) All High–low Low (80%) Gross alpha 4 (%) −1.09*** (.00) −2.67* (.10) −1.23*** (.00) −1.57 (.30) Netalpha 4 (%) −2.98*** (.00) −4.57*** (.01) −3.12*** (.00) −1.59 (.30) High (20%) Gross alpha 4 (%) −4.56** (.02) 1.88 (.34) 0.17 (.93) 6.44*** (.00) Netalpha 4 (%) −6.38*** (.00) 0.25 (.90) −1.49 (.43) 6.63*** (.00) High–low TE High–low AS All Gross alpha 4 (%) −1.30*** (.00) 0.66 (.71) 1.96 (.24) 1.40 (.42) Netalpha 4 (%) −3.19*** (.00) −1.02 (.57) 2.16 (.20) 1.63 (.35) Mix H-L High-low Gross alpha 4 (%) −3.47** (.04) 4.54*** (.00) 2.97 (.11) Netalpha 4 (%) −3.40** (.05) 4.82*** (.00) 3.23* (.08) TE: tracking error; AS: Active Share; Mix H-L: Mix High–Low. Note: Panels A and B report the annual performance (benchmark adjusted returns and four-factor alphas). At calendar quarter-ends (publicly reporting date), the funds are independently ranked according to their AS and TE. Then, we construct two categories for each variable using the 80th percentile and four categories from their intersection. The funds for each category were held 3 months after the formation date. The “Mix High–Low” strategy shows the difference between “High AS and TE” funds and “Low AS and TE” funds. The p values (in parentheses) were based on White’s standard errors. *, **, and *** indicate significance at .1, .05, and .01 levels, respectively. Andreu et al. 341 statistically significant performance. This finding differs from that of Cremers and Petajisto (2009), who show that AS allows investors to select funds with good performance. However, our results are consistent with Cremers et al. (2016), who state that the average alphas generated by active management are lower in markets where closet indexing is generalized. Second, we also find that the AS level predicts fund performance among high TE funds (the difference in the four-factor alpha between high AS and low AS is above 4%, a figure positive and statistically significant for both gross and net). Similarly, the TE level predicts the fund performance among high AS funds (the difference in the four-factor alpha is also above 4%). This last finding differs from that of Cremers and Petajisto (2009), who do not observe that TE predicts fund performance within any AS category. The differences in the findings may be because we independently rank funds according to both metrics to obtain appropriate orthogonal portfolios, while these authors sort funds into AS first and then into TE quintiles. Sorting funds first into AS and then into TE tests whether TE has predictive power after controlling for AS, but not the contrary. To analyze the predictive power of AS controlling for TE, funds should first be sorted into TE. Third, we can perform a better selection if we focus on the funds that simultaneously have high levels of both active measures. Although funds with both high AS and TE levels do not yield significant positive performance, it must be highlighted that, in contrast to the other funds, on average, they are the only category able to generate enough value to, at least, compensate for their fees.20 High AS–low TE funds and future money flows In this section, we attempt to provide extra support to H4 by analyzing the relationship between the different categories of active management based on AS and TE and their money flows. Table 12 shows the monthly average adjusted flow in the 3 months after the formation date (holding period) for each of the four portfolio categories. We find that despite the worst performance of window dressers (high AS–low TE funds), these funds do not exhibit significantly lower levels of money flows than the category of funds with both high AS and TE. Furthermore, they attract significantly higher money flows than funds with both low AS and TE. Therefore, window-dresser funds reach the objective of attracting money flows. This evidence supports H4. AS WD and future performance: Monthly basis regression and independent funds subsample In addition to the quarterly based regression of equation (11), we also analyzed the impact of AS WD in publicly disclosed months on future performance in a monthly basis regression, as shown in the following equation AdjReturns DPublic AS DPublic AS it it it it __ _ ,, ,, + =+ ++ × 101 23 αα αα iit it it controls , ,, ++ ε (15) where the dependent variable is the gross and net adjusted returns of mutual funds in the next month t + 1, while the Table 12. Adjusted investment flows according to the level of active management. AS TE Low (80%) High (20%) All High–low Low (80%) −300.82*** (.00) 509.38*** (.00) −262.58*** (.00) 810.20*** (.00) High (20%) 697.60* (.06) 1,094.71*** (.00) 996.74*** (.00) 397.11 (.36) High–low TE High–low AS All −259.57*** (.00) 980.46*** (.00) 1,240.03*** (.00) 1,259.32*** (.00) Mix H-L High–low 998.42*** (.01) 585.33*** (.01) 1,395.53*** (.00) TE: tracking error; AS: Active Share; Mix H-L: Mix High–Low; TNA: total net assets. Note: The table shows the monthly average money flows adjusted by TNA decile portfolios in the 3 months after the formation date (holding period) obtained by Spanish domestic equity mutual funds, depending on the level of activeness measured through the AS and TE measures. At calendar quarter-ends (publicly reporting date), the funds are independently ranked according to their AS and TE levels. Then, we construct two categories for each variable using the 80th percentile and four categories from their intersection. The “Mix High–Low” strategy shows the difference between “High AS and TE” funds and “Low AS and TE” funds. The p values (in parentheses) were based on White’s standard errors. *, and *** indicate significance at .1, and .01 levels, respectively. 342 Business Research Quarterly 27(3) independent variables are as follows: DPublic_ is a dummy variable that equals 1 for publicly disclosed months, and 0 otherwise; AS of mutual funds in month t; and the interaction between DPublicit _, and ASit,. The control variables are TE, Adj_Returns, and those defined in equation (8). Hence, the interaction variable, DPublic AS it it _ ,, ×, measures deviations of AS in publicly disclosed months from non-publicly disclosed months and, therefore, could be considered a proxy for changing the holdings to appear more active. If the practice of artificially increasing the level of AS in publicly disclosed months deteriorates fund performance, then we should expect negative coefficients for this variable. Moreover, we suggest that the negative performance impact of changing portfolio holdings in publicly disclosed months depends on how this practice is conducted. Accordingly, Golez and Marin (2015) find that Spanish bank-affiliated mutual funds (non-independent funds) make holding changes based on the bank’s interest. Based on this previous evidence, we hypothesize that independent funds are, on average, less prone to performing WD, and if they increase the AS level in publicly reported months (high levels of AS_difference), it will more likely be related to informed transactions rather than AS WD. Therefore, they will not damage performance. In contrast, non-independent funds are, on average, more prone to performing WD (subject to the interest of the bank they are affiliated with). If they increase the AS level in publicly reported months, it will probably be related to WD rather than to informed trading. Therefore, they will damage performance. Table 13 shows the results of running equation (15) for the full sample and for independent and non-independent Table 13. AS WD and future fund returns: independent versus bank-affiliated funds. Gross adjusted returnst+1 Net adjusted returnst+1 Full sample Independent Bank-affiliated Full sample Independent Bank-affiliated Publict0.3414*** (.01) −0.0925 (.80) 0.3785*** (.01) 0.3416*** (.01) −0.1025 (.78) 0.3850*** (.01) ASt1.1595*** (.00) 2.5667*** (.00) 0.2679 (.41) 1.2047*** (.00) 2.4965*** (.00) 0.3721 (.25) Dpublic AS it t _,×−1.0311*** (.01) −0.3323 (.65) −1.0925** (.02) −1.0376*** (.01) −0.3105 (.67) −1.1204** (.02) n6,311 1,361 4,950 6,311 1,361 4,950 R2 (%) 4.98 8.26 5.37 5.26 8.54 5.74 Adjusted R2 (%) 4.69 6.93 4.99 4.98 7.21 5.37 AS: Active Share; WD: window dressing; TE: tracking error. Note: This table reports the results of the analysis that examines the performance impact of increasing the AS level in publicly disclosed months. The dependent variables are the gross and net benchmark adjusted returns of mutual funds in month t + 1, while the independent variables are as follows: the AS levels of mutual funds in month t; D_Public, which is a dummy variable that equals 1 for publicly disclosed months (March, June, September, and December), and 0 otherwise; the interaction between AS and D_public; TE; the fund adjusted returns; and the control variables explained in Table 3 measured in month t. In addition, the table shows the results of the full sample split into independent funds and nonindependent funds (bank-affiliated mutual funds). Year fixed effects are included in all the specifications. The p values (in parentheses) are based on standard errors clustered by the fund. **, and *** indicate significance at .05, and .01 levels, respectively. (bank-affiliated) subsamples. The results for the full subsample show that increasing the level of activity at quarterends negatively influences fund performance (coefficient of D_publicxAS), which confirms H3. Moreover, this negative impact is not observed in the independent funds subsample and only the bank-affiliated funds subsample shows a statistically significant coefficient on D_publicxAS. This result supports our intuition that the increase of AS in publicly disclosed months is probably related to informed trades for the independent funds subsample and probably related to WD for the non-independent funds subsample. Conclusion In this study, we use a unique database that includes nonpublicly disclosed portfolios to test whether mutual funds alter their holdings in publicly disclosed months to appear more active by reporting a high level of AS to attract future money flows. No study in the literature has examined whether the level of deviation from the benchmark can be altered by portfolio managers (AS WD). We compute how fund holdings deviate from the benchmarks using the AS metric of Cremers and Petajisto (2009). We examine the relationship between AS and future fund performance, and the relationship between AS and money flows, and demonstrate that the higher the AS level, the higher the future fund returns and flows. Hence, we hypothesize that certain portfolio managers have incentives to show a high level of AS when the portfolios are publicly available. Therefore, we study the existence of AS WD and how this practice influences future fund performance and money flows. We define the AS_difference as Andreu et al. 343 the difference in the level of AS in publicly reported months as opposed to the level in the surrounding months and use it as a proxy for AS WD. We find that high AS_difference values are related to poor future fund performance but high future money flows. Finally, we analyze funds that are more prone to conducting AS WD. We find that, contrary to the other categories of funds, on average, funds with high AS and low TE levels increase their AS levels before their holdings are public and reduce them afterward, hence showing a significant and positive AS_difference. Moreover, this category of funds exhibits the worst future performance but attracts higher money flows than funds with similar levels of TE. This finding also indicates that this strategy of deviating more from benchmarks when holdings are public makes economic sense because the higher the money flows are, the higher the incomes to the management companies, given that most Spanish mutual funds charge their fees based on TNA. However, we should not forget that these investment strategies conducted by certain funds erode investor fund performance, particularly in the case of bank-affiliated funds. We believe that the results presented here are compelling enough to warrant further analysis. The above-mentioned results provide evidence of the implications of this research to policymakers who should consider the necessity of monthly portfolio holdings instead of quarterly holdings to limit the discretion of portfolio managers. Although we are aware of the difficulties in obtaining both publicly and non-publicly available portfolio holdings for all funds in a given investment category, future research in this area could extend this study to other countries in the eurozone and the US market. The existence of publicly and non-publicly available holdings for all mutual funds in the different eurozone countries will allow researchers to examine whether the AS WD phenomenon is just a characteristic of the Spanish market or if it is a widespread phenomenon. Another line of research could involve a more detailed analysis of the investors’ reactions, to examine whether they are aware of this managerial practice. Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Spanish government and the European Union FEDER funds RTI2018-093483-B-I00. ORCID iD Laura Andreu https://orcid.org/0000-0003-3661-5448 Notes 1. Morningstar, for instance, provides monthly portfolio holdings for certain funds in other eurozone countries. However, there is a selection bias because this information is not always available for all mutual funds. In addition, this information is publicly available for other market competitors, investors, and so on. This fact prevents extending the analysis to other eurozone countries. Similarly, the official supervisor in Spain does not provide monthly portfolio holdings from 2007 onward. Therefore, this limits our time period of study from 2000 to 2006. 2. If the manager performs well during the delay period, investors are less likely to attribute portfolio movements to window dressing (WD) and more likely to improve security selection. Consequently, subsequent to the delay period, managers may benefit from incrementally higher flows than that justified by the fund’s performance. Contrastingly, if the performance during the delay period is bad, investors are more likely to attribute the portfolio movements to WD and therefore cause the manager to incur the cost of lower flows. 3. Some studies on the Spanish mutual fund industry examine WD (in equity and bond funds) and the January effect and confirm that investors react to portfolio holdings information in this market (see, for example, Ortiz et al., 2010, 2012, 2015, among others). There are also studies that examine other research questions in the Spanish industry that show that investors use information based on disclosed portfolio holdings to assess managerial ability in conjunction with past performance (see, for example, Álvarez et al., 2014; Andreu et al., 2017). 4. The sum is taken over the universe of equity positions only, as we apply the measure exclusively to all-equity portfolios. 5. The correlation coefficient between AS (Active Share) and TEbeta is .3775. We present the results for tracking error (TE; with higher correlation with AS) in the text as we aim to analyze whether AS has explanatory power beyond TE measures. The results using TEbeta are available upon request. 6. The monthly portfolio holdings provided by Spanish Securities and Exchange Commission (CNMV) contain officially disclosed and non-officially disclosed portfolios because management companies in Spain must report to investors on a quarterly basis, which is more frequent than the European Union’s requirement of semi-annual portfolio reports. Because the fiscal year of Spanish management companies is the calendar year, reports are mandatory at the end of each calendar quarter. Consequently, from 1999 to 2006, the database was compiled from official CNMV filings, as well as non-official disclosures provided from CNMV to the authors exclusively for research purposes. Therefore, fund managers could not anticipate the release of these undisclosed intra-quarter portfolios. 7. The quarter-end disclosed holdings officially provided by CNMV from March 2007 onward could be complemented with monthly portfolio holdings provided by Morningstar. However, this unofficial information is not available for all months and for all mutual funds, implying a selection bias. Moreover, the monthly portfolio holdings provided by Morningstar are publicly available, and the goal of our 344 Business Research Quarterly 27(3) test is to check the differences between publicly and nonpublicly available holdings. 8. The results obtained when using the Spanish benchmark that produces the lowest AS are quite robust along the different analyses. Similarly, when we limit the analysis to the 93 funds that self-report Ibex35 or IGBM (Índice General de Bolsa de Madrid), the results are robust. The results of both of the robustness analyses are not reported in the study due to brevity reasons. These results are, however, available from the authors upon request. 9. Although we only present results for the period from December 2000 to December 2006 because of the availability of monthly portfolio holdings to focus our study on the analysis of WD, we have quarterly AS and TE data until June 2014 for all the mutual funds of the sample. 10. This fact has attracted the attention of researchers who compare passive funds and actively managed funds (see, for example., Detzler & Wiggins, 1997; Doshi et al., 2015; Farooqi et al., 2019; Pástor et al., 2015; Pástor & Stambaugh, 2002). 11. The return gap is computed as the difference between the actual returns and the hypothetical returns based on the previous holdings (see, for example, Kacperczyk et al., 2008). 12. We use the returns of the Ibex35 Total Return Index as the equity benchmark and the returns of one-day Spanish Treasury bill repos as the risk-free rate. The factors of size, book to market, and momentum have been calculated following the same procedure detailed on the website of Kenneth French considering the stocks traded in the Spanish stock market (see, for example, http://mba.tuck.dartmouth. edu/pages/faculty/ken.french/data_library.html). 13. The results are robust when TE is not included in the equation. The results are available from the authors upon request. 14. It is difficult to put the size of our results of AS WD in perspective to other evidence of WD because we are analyzing a different kind of WD. However, if we compare our variable AS_difference dummy value higher >3% with the BHRG10% dummy (defined as 1 if ‘backward holding return gap’ is in the top 10th percentile) of Agarwal et al. (2014), our –2.06% AS_difference dummy coefficient for net Carhart alphas in our Table 6 is higher than the –0.58% BHRG10% dummy coefficient of Table 12(3) in Agarwal et al. (2014). This comparison should be considered with caution because our control variables, although similar, are not exactly the same as those of Agarwal et al. (2014). 15. The disposition effect, that is, the tendency to hold onto losses and to sell winners too soon, has been documented in the US market (see, for example, Cici, 2012; Odean, 1998) and in the Spanish mutual fund industry (see, for example, Andreu et al., 2020). 16. See Bandarchuk and Hilscher (2013) for an example of how conditional ranks can drive false conclusions. 17. The main conclusions, although with less significance, hold with the 50th percentile. These results are available from the authors upon request. 18. As a robustness analysis, we tried to refine the portfolio that consists of mutual funds characterized by a high value of AS and a low value of TE. To do that, we use other percentiles different from the 80th percentile to classify a mutual fund as one with “low TE.” The problem is that even with the 60th percentile, the high AS–low TE portfolio was empty in many time periods. 19. We have also checked the robustness of Table 11 using the total net assets (TNA), flows, and turnover control portfolio adjusted returns used in Panel G of Table 2. The results support our evidence that funds with high AS and low TE levels have the worst performance. The authors can provide these results upon request. 20. Non-reported tests also show that the performance difference between high and low active funds is significantly higher (at the 1% level) when the two active measures (AS and TE) are used (Mix High–Low) than when only one active measure is used (high–low AS or high–low TE). References Agarwal, V., Gay, G. D., & Ling, L. (2014). Window dressing in mutual funds. The Review of Financial Studies, 27, 3133– 3170. https://doi.org/10.1093/rfs/hhu045 Alexander, G. J., Cici, G., & Gibson, S. (2007). Does motivation matter when assessing trade performance? An analysis of mutual funds. The Review of Financial Studies, 20, 125– 150. https://doi.org/10.1093/rfs/hhl008 Álvarez, J., Andreu, L., Ortiz, C., & Sarto, J. L. (2014). A nonparametric approach to market timing: Evidence from Spanish mutual funds. Journal of Economics and Finance, 38, 119–132. https://doi.org/10.1007/s12197-0129228-9 Andreu, L., Mateos, L., & Sarto, J. L. (2017). The value added by trading based on valuation criteria. International Review of Finance, 17, 327–352. https://doi.org/10.1111/irfi.12097 Andreu, L., Ortiz, C., & Sarto, J. L. (2020). Disposition effect in fund managers. Fund and stock-specific factors and the upshot for investors. Journal of Economic Behavior & Organization, 176, 253–268. https://doi.org/10.1016/j. jebo.2020.04.002 Baker, M., Litov, L., Wachter, J. A., & Wurgler, J. (2010). Can mutual fund managers pick stocks? Evidence from their trades prior to earnings announcements. Journal of Financial and Quantitative Analysis, 45, 1111–1131. https://doi.org/10.1017/S0022109010000426 Bandarchuk, P., & Hilscher, J. (2013). Sources of momentum profits: Evidence on the irrelevance of characteristics. Review of Finance, 17, 809–845. https://doi.org/10.1093/ rof/rfr036 Caquineau, M., Möttöla, M., & Schumacher, J. (2016). Active Share in European equity funds: The activeness of largecap European fund managers through the lens of Active Share. Morningstar. Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of Finance, 52, 57–82. https://doi. org/10.1111/j.1540-6261.1997.tb03808.x Chen, H. L., Jegadeesh, N., & Wermers, R. (2000). The value of active mutual fund management: An examination of the stockholdings and trades of fund managers. Journal of Financial and Quantitative Analysis, 35, 343–368. https:// doi.org/10.2307/2676208 Cici, G. (2012). The prevalence of the disposition effect in mutual funds’ trades. Journal of Financial and Quantitative Analysis, 47, 795–820. https://doi.org/10.1017/S002210901 2000348 Cremers, K. J. M., Ferreira, M. A., Matos, P. P., & Starks, L. T. (2016). Indexing and active fund management: International Andreu et al. 345 evidence. Journal of Financial Economics, 120, 539–560. https://doi.org/10.1016/j.jfineco.2016.02.008 Cremers, K. J. M., & Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. The Review of Financial Studies, 22, 3329–3365. https:// doi.org/10.1093/rfs/hhp057 Daniel, K., Grinblatt, M., Titman, S., & Wermers, R. (1997). Measuring mutual fund performance with characteristicbased benchmarks. The Journal of Finance, 52, 1035–1058. https://doi.org/10.1111/j.1540-6261.1997.tb02724.x Detzler, M. L., & Wiggins, J. B. (1997). The performance of actively managed international mutual funds. Review of Quantitative Finance and Accounting, 8, 291–313. https:// doi.org/10.1023/A:1008219106132 Díaz-Mendoza, A. C., López-Espinosa, G., & Martínez, M. A. (2014). The efficiency of performance-based fee funds. European Financial Management, 20, 825–855. https://doi. org/10.1111/j.1468-036X.2012.00654.x Doshi, H., Elkamhi, R., & Simutin, M. (2015). Managerial activeness and mutual fund performance. The Review of Asset Pricing Studies, 5, 156–184. https://doi.org/10.1093/ rapstu/rav005 Elton, E. J., Gruber, M. J., Blake, C. R., Krasny, Y., & Ozelge, S. O. (2010). The effect of holdings data frequency on conclusions about mutual fund behavior. Journal of Banking & Finance, 34, 912–922. https://doi.org/10.1016/j.jbankfin.2009.10.002 European Securities and Markets Authority. (2016). Supervisory work on potential closet index tracking. https://www.esma. europa.eu/sites/default/files/library/2016-165_public_statement_-_supervisory_work_on_potential_closet_index_ tracking.pdf Fama, E. F., & French, K. R. (2010). Luck versus skill in the cross-section of mutual fund returns. The Journal of Finance, 65, 1915–1947. https://doi.org/10.1111/j.15406261.2010.01598.x Farooqi, J., Jory, S., & Ngo, T. (2019). Active fund managers and earnings management at portfolio companies. Review of Accounting and Finance, 19, 48–82. https://doi. org/10.1108/RAF-11-2017-0209 Frazzini, A., Friedman, J., & Pomorski, L. (2016). Deactivating Active Share. Financial Analysts Journal, 72, 14–21. https://doi.org/10.2469/faj.v72.n2.2 Gil-Bazo, J., & Ruiz-Verdú, P. (2009). The relation between price and performance in the mutual fund industry. The Journal of Finance, 64, 2153–2183. https://doi.org/10.1111/j.15406261.2009.01497.x Golez, B., & Marin, J. M. (2015). Price support by bank-affiliated mutual funds. Journal of Financial Economics, 115, 614–638. https://doi.org/10.1016/j.jfineco.2014.10.008 Guercio, D. D., & Tkac, P. A. (2002). The determinants of the flow of funds of managed portfolios: Mutual funds vs. pension funds. Journal of Financial and Quantitative Analysis, 37, 523–557. https://doi.org/10.2307/3595011 Haugen, R. A., & Lakonishok, J. (1988). The incredible January effect. Dow Jones-Irwin. Jensen, M. C. (1968). The performance of mutual funds in the period 1945–1964. The Journal of Finance, 23, 389–416. https://doi.org/10.1111/j.1540-6261.1968.tb00815.x Jiang, H., Verbeek, M., & Wang, Y. (2014). Information content when mutual funds deviate from benchmarks. Management Science, 60, 2038–2053. https://doi.org/10.1287/mnsc.2013 .1847 Kacperczyk, M., Sialm, C., & Zheng, L. (2005). On the industry concentration of actively managed equity mutual funds. The Journal of Finance, 60, 1983–2011. https://doi.org/10.3386/ w10770 Kacperczyk, M., Sialm, C., & Zheng, L. (2008). Unobserved actions of mutual funds. The Review of Financial Studies, 21, 2379–2416. https://doi.org/10.1093/rfs/hhl041 Morey, M. R., & O’Neal, E. S. (2006). Window dressing in bond mutual funds. Journal of Financial Research, 29, 325–347. https://doi.org/10.1111/j.1475-6803.2006.00181.x Musto, D. K. (1997). Portfolio disclosures and year-end price shifts. The Journal of Finance, 52, 1563–1588. https://doi. org/10.1111/j.1540-6261.1997.tb01121.x Ng, L., & Wang, Q. (2004). Institutional trading and the turn-ofthe-year effect. Journal of Financial Economics, 74, 343– 366. https://doi.org/10.1016/j.jfineco.2003.05.009 Odean, T. (1998). Are investors reluctant to realize their losses? The Journal of Finance, 53, 1775–1798. https://doi. org/10.1111/0022-1082.00072 Ortiz, C., Ramírez, G., & Vicente, L. (2010). Quarterly return patterns in the Spanish stock market. Applied Financial Economics, 20, 1829–1838. https://doi.org/10.1080/09603 107.2010.528366 Ortiz, C., Ramírez, G., & Vicente, L. (2015). Mutual fund trading and portfolio disclosures. Journal of Financial Services Research, 48, 83–102. https://doi.org/10.1007/s10693-014-0198-2 Ortiz, C., Sarto, J. L., & Vicente, L. (2012). Portfolios in disguise? Window dressing in bond fund holdings. Journal of Banking & Finance, 36, 418–427. https://doi.org/10.1016/j. jbankfin.2011.07.017 Pástor, L., & Stambaugh, R. F. (2002). Investing in equity mutual funds. Journal of Financial Economics, 63, 351–380. https://doi.org/10.1016/S0304-405X(02)00065-X Pástor, L., Stambaugh, R. F., & Taylor, L. A. (2015). Scale and skill in active management. Journal of Financial Economics, 116, 23–45. https://doi.org/10.1016/j.jfineco.2014.11.008 Petajisto, A. (2013). Active Share and mutual fund performance. Financial Analysts Journal, 69, 73–93. https://doi. org/10.2469/faj.v69.n4.7 Sensoy, B. A. (2009). Performance evaluation and self-designated benchmark indexes in the mutual fund industry. Journal of Financial Economics, 92, 25–39. https://doi. org/10.1016/j.jfineco.2008.02.011 Sirri, E. R., & Tufano, P. (1998). Costly search and mutual fund flows. The Journal of Finance, 53, 1589–1622. https://doi. org/10.1111/0022-1082.00066 Wermers, R. (2003). Are mutual fund shareholders compensated for active management bets? [Working paper]. University of Maryland.