The Mutual Fund Industry Worldwide: Explicit and Closet Indexing, Fees, and Performance *

Transcription

1 The Mutual Fund Industry Worldwide: Explicit and Closet Indexing, Fees, and Performance * Martijn Cremers, University of Notre Dame mcremers@nd.edu Miguel A. Ferreira, Nova School of Business and Economics miguel.ferreira@novasbe.pt Pedro Matos, University of Virginia Darden School of Business matosp@darden.virginia.edu Laura Starks, University of Texas at Austin laura.starks@mccombs.utexas.edu This Version: February 2013 Abstract We examine the causes and consequences of indexing for active management by equity mutual funds worldwide. Explicit indexing is rare in countries with weaker regulation and less liquid stock markets, but many actively managed funds engage in closet indexing. We find that funds are more active and charge lower fees when they face more competitive pressure from explicitly indexed funds. Moreover, truly active funds generate higher risk-adjusted net returns than closet indexers. Overall, we conclude that explicit indexing improves the levels of competition and efficiency of the mutual fund industry in a country. Keywords: Mutual funds, Active management, Index funds, Exchange-traded funds, Competition, Fees, Performance JEL classification: G15, G18, G23 * We thank Wayne Ferson, Javier Gil-Bazo, Hao Jiang, Andrew Karolyi, Aneel Keswani, Borja Larrain, Lilian Ng, Henri Servaes, Michaela Verardo, and Jeffrey Wurgler; seminar participants at Arizona State University, Cass Business School, Cornell University, George Washington University, Imperial College, Instituto de Empresa, Rice University, State Street Global Advisors, Stockholm School of Economics-SIFR, Università Cattolica del Sacro Cuore, University of Colorado Boulder, University of Lugano, University of Mannheim, University of Melbourne, University of New South Wales, University of Sidney, University of Southern California, University of Technology Sidney, University of Utah, and University of Virginia-Darden School of Business; and conference participants at the 5th Biennial McGill Global Asset Management Conference, 5th Rotterdam School of Management Professional Asset Management Conference, 10th Rothschild Caesarea Center Conference, 2011 European Finance Association Meeting, 2011 Morningstar-Ibbotson Investment Conference, 2011 Morningstar Europe Conference, 2012 Inquire Europe Conference, 2012 SFS Cavalcade Conference, and 2012 American Finance Association Meeting for helpful comments. The authors acknowledge financial support from Inquire Europe and the European Research Council.

2 Practitioners and academics have long debated the societal benefits and degree of competition in the asset management industry, particularly among equity mutual funds. Beginning with Sharpe (1966) and Jensen (1968), academic researchers have debated the relative value of passive versus active management, and many have argued that indexed portfolios are superior investment vehicles. For example, Gruber (1996) questions the rapid growth in actively managed mutual funds, given evidence that their average performance is typically inferior to that of index funds. French (2008) and others argue that investors spend a significant amount on fees, expenses, and trading costs in actively managed funds that, on average, fail to beat market indices. Other studies, however, provide evidence that some mutual fund managers add value through active management. Grinblatt and Titman (1989, 1993) conclude that some fund managers are successful at beating their benchmarks before expenses; Kacperczyk, Sialm, and Zheng (2005) find that mutual funds with concentrated industry bets tend to outperform; and Cremers and Petajisto (2009) show that funds whose holdings are most different from their benchmarks (i.e., truly active as opposed to closet indexers) outperform their benchmarks net of fees. 1 Mutual funds have grown over the last half century to become one of the primary investment vehicles for households worldwide. Nearly 28,000 equity funds and $10.5 trillion in assets under management were recorded as of December 2010 (Investment Company Institute (2011)). The percentage invested in explicitly indexed funds has grown rapidly over the last decade from about 14% of assets under management in 2002 to about 22% in These explicitly indexed funds have thus become an increasingly important low-cost alternative to investment in stock markets as they have substantially lower fees than active funds in all countries. So what are the implications for active funds of the entry and growth of passively managed 1 For evidence on mutual funds, see also Wermers (2000), Bollen and Busse (2001), Avramov and Wermers (2006), Kosowski, Timmermann, Wermers, and White (2006), and Kacperczyk and Seru (2007). For evidence on pension plans, see Dyck, Lins, and Pomorski (2011) and Andonov, Bauer, and Cremers (2012). 1

3 funds? We hypothesize that if we hold the mutual fund customer base fixed, the introduction of explicitly indexed funds should lead to more competition; active funds should reduce their fees (the price ) or and differentiate their portfolios more (the product ). 2 Yet, market frictions, information asymmetries and less-than-perfectly competitive markets may allow active funds to maintain their market share and avoid having to lower fees or differentiate their products. Investors might base their decisions on fund attributes other than returns, such as fund family size, tax exposure, services, or investment strategies. Hortacsu and Syverson (2004) show considerable information/search frictions even for a homogeneous product such as an S&P 500 index fund. 3 In the case of active funds, investors may consider mutual funds to be heterogeneous products and find it difficult to distinguish between skilled and lucky managers. Berk and Green (2004) and Pastor and Stambaugh (2012) argue that fund managers may have skill and investors may want to invest in active funds even in the absence of ex post average positive alphas. Fund management companies can also create a perception among investors that they offer a differentiated product by expanding product offerings to preserve rents. Such actions could slow down investor learning about the benefits of indexed funds. 4 We study whether the presence of passively managed funds increases competition and efficiency in the asset management industry. We test whether the enhanced competitive pressure 2 For research on competition in the U.S. mutual fund industry, see, for example, Baumol, Goldfeld, Gordon, and Koehn (1990) and Coates and Hubbard (2007). 3 An expansion in fund choices could increase the population of investors using mutual funds for the first time. In this case, novice investors with high information/search costs could flock to the relatively more expensive funds. Hortacsu and Syverson (2004) find a rightward shift in the fee distribution of S&P 500 index funds in the late 1990s at the time of new fund introductions. In an experimental setting, Choi, Laibson, and Madrian (2010) find that not all investors pick S&P 500 index funds solely on the basis of fees. Collins (2005) explains how differences in services and characteristics of the index funds can influence their expense ratios. 4 Carlin and Manso (2011) refer to this practice as obfuscation. In addition, indexed products could reduce search costs and force active funds to differentiate by becoming more niche players (Bar-Isaac, Caruana, and Cunat (2012)). 2

4 created by the presence of passive funds (as proxied by the market share and fees of explicitly indexed funds) increases active funds product differentiation (as proxied by deviations of fund holdings from the benchmark) and reduces the price paid for active management (i.e., the fees charged). Ultimately, what investors care about should be risk-adjusted net returns that mutual funds deliver. Thus, we study whether the most active funds display better net performance. Using a new data set on open-end equity mutual funds and exchange-traded funds (ETFs) in more than 30 countries over , we find considerable cross-country variation in the prevalence of indexing. In the United States, the market share of index funds and ETFs (explicitly indexed funds) grew from 16% of assets under management in 2002 to 27% in In other countries, explicit indexing is less prevalent but growing faster, from 6% in 2002 to 13% in Furthermore, no explicitly indexed funds seem to be offered at all in some countries. Our worldwide sample provides a unique setting to test the effects of growth in explicitly indexed funds on competition in a mutual fund industry, because there is wide variation in the availability of indexed products and regulatory and market environments across countries. Despite the relatively small amount of explicit indexing, we find that closet indexing is frequent especially outside the United States. Using the Cremers and Petajisto (2009) active share measure, which captures the proportion of fund holdings that differs from a fund s benchmark, we find that outside the United States about 30% of the assets are managed by closet indexers (funds with an active share below 60%). This is twice the level of closet indexing in the United States. The relative levels of explicit and closet indexing vary across fund investment strategies. Country- and regional-focused funds tend to have higher levels of closet indexing than world funds, especially for non-u.s. funds. For example, 50% of the assets in domestic country funds 3

5 domiciled outside the United States are in closet index funds versus 14% for those domiciled in the United States. Cross-country regressions indicate that explicit indexing is more frequent in countries with stronger fund regulatory environments and with larger and more liquid stock markets. In contrast, we find that these factors have opposite effects on the frequency of closet indexing. We find that competition from explicitly indexed funds appears to affect active funds in several ways. First, actively managed funds have higher active shares in countries with more explicit indexing. This is consistent with the idea that explicit indexing forces active fund managers to differentiate themselves from low-cost passive alternatives. Second, the prevalence of explicitly indexed funds in a country is negatively related to fees charged to investors in active funds. In addition, active funds tend to charge higher fees in countries with more closet indexing. These findings support the idea that the pervasiveness of explicit indexing leads to more competition, while more closet indexing in a country s fund industry reflects a lack of competitive pressures. These results hold whether a fund s market of operation is defined according to its legal domicile or according to countries where the fund is sold. The results are robust to corrections for endogeneity of the presence of explicit and closet indexing in a country using instrumental variables methods. Finally, we find that active share predicts future fund performance. The most active mutual funds outperform their benchmarks after fees, while closet indexers underperform. The results hold for different measures of performance including benchmark-adjusted returns and four-factor alphas and are economically significant. A one standard deviation increase in active share is associated with an increase of about 1% per year in future benchmark-adjusted returns and 0.7% per year in future four-factor alpha. There is also some evidence that returns to active 4

6 management are higher in countries where low-cost passive alternatives are more popular. This is consistent with the idea that only skilled managers will survive in an environment with enhanced competitive pressure from index funds and ETFs. In contrast, the evidence does not support the idea that the growth of index-based investing could lower fund managers incentives to manage their portfolios actively (Wurgler (2011)). Overall, our findings suggest that the availability of explicit and closet indexing is a key determinant of competition in mutual fund industries around the world. We conclude that explicit indexing is associated with improved levels of competition and efficiency in a fund industry, while closet indexing is indicative of the reverse. Our work adds to the understanding of the structure of the asset management industry worldwide. The few papers analyzing mutual funds worldwide have so far focused on the determinants of industry size and fees across countries. Khorana, Servaes, and Tufano (2005, 2009), for example, find a link between the level of development of fund industries worldwide and a combination of legal, regulatory, and demand- and supply-side factors. To the best of our knowledge, we are the first to study the causes and consequences of explicit and closet indexing for the industrial organization of the mutual funds around the world. 5 I. Data and Variables Our analysis uses two primary databases: Lipper and FactSet/LionShares. The Lipper database provides a comprehensive sample of mutual funds offered and sold across a large number of countries. We focus exclusively on open-end domestic and international equity mutual funds and exchange-traded funds (ETFs) in the period. From this database we 5 Related work by Wahal and Wang (2011) shows that the entry of new mutual funds in the U.S. industry that are close substitutes for incumbents (as measured by overlap in portfolio holdings) forces incumbents to reduce fees. 5

7 obtain individual fund characteristics, such as fund name, domicile, investment style, sponsor, benchmark, monthly returns, total net assets (TNA), fees, and expenses. The Lipper database is survivorship bias-free, as it includes both active and defunct funds. Although multiple share classes are listed as separate observations in Lipper, they have the same holdings, the same manager, and the same returns before expenses. Thus, we keep as our unit of observation the share class that Lipper identifies as the primary share class and aggregate fund-level variables across the different share classes. 6 Table IA.I in the Internet Appendix provides summary statistics of all variables for the sample of open-end active funds in the period, and Table IA.II reports time series averages of country variables per country. Appendix B provides all variable definitions. The LionShares database covers portfolio equity holdings for institutional investors worldwide, including mutual funds and ETFs. Ferreira and Matos (2008) provide a detailed description of this database. We match the Lipper (fund characteristics and performance) and LionShares (fund holdings) databases by CUSIP, ISIN, or fund name. We identify funds nationalities by their legal domicile, which characterizes the relevant regulatory and legal system. The funds are domiciled in 32 countries across several regions: North America (Canada, the United States); Europe (Austria, Belgium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Poland, Portugal, Spain, Sweden, Switzerland, the United Kingdom); Asia-Pacific (Australia, China, Hong Kong, India, Japan, Malaysia, Singapore, Taiwan, Thailand); other regions (Brazil, Israel, South Africa); and off-shore (Ireland, Liechtenstein, Luxembourg). 7 Mutual funds, while taking a variety of names around the globe, 6 We run some robustness tests using individual share classes. 7 We require a minimum of 50 funds to include a country in the sample. 6

8 are fairly comparable investment vehicles worldwide (Khorana, Servaes, and Tufano (2005)). 8 Table I provides some key statistics for the sample by country of domicile as of December We report these statistics for the 20 major European and North American countries and combine the Asia-Pacific countries into one observation in Table I because the holdings data are limited for many funds in those countries. We also combine the statistics for other regions. Table I shows that the 24,492 funds in the sample have TNA totaling over $9.8 trillion across the 32 countries. 9 Equity mutual funds domiciled in the United States represent the majority of the TNA (over $5.7 trillion), but other markets are also sizable. The Lipper s coverage of funds can be compared with aggregate statistics on mutual funds from other sources. As of December 2010, Investment Company Institute (2011) reported a total number of equity mutual funds worldwide of 27,754 with a TNA of $10.5 trillion. We conclude that our sample is nearly comprehensive of the equity mutual fund universe. Table IA.III in the Internet Appendix provides details on the number of share classes by country of domicile and country of sale in While for a majority of funds the country of domicile corresponds to the single country of sale, some funds are registered for sale in multiple countries, so there is effectively competition across domiciles. Funds domiciled in countries such as Canada and the United States are mostly sold in their home markets, but funds domiciled in other markets such as France, Germany, the United Kingdom, and off-shore (Ireland, Luxembourg) are typically approved for sale in several markets. Because some funds have multiple share classes and are offered in more than one country, we can have multiple observations in a given year for the same fund. 8 The European Union, in an attempt to create a harmonized mutual fund industry, has adopted a common definition: UCITS (Undertakings for Collective Investment in Transferable Securities). In addition, the EU has adopted the European passport system (Directive 2001/107/EC), which facilitates cross-border marketing of UCITS. 9 Mutual funds held roughly 20% of world market capitalization as of December

9 Table I shows that detailed portfolio holdings are available from LionShares for 11,776 funds with TNA of approximately $7.9 trillion. In total, we have holdings data from the LionShares database for about 81% of the TNA in the Lipper database, but coverage varies across countries. For the 20 countries in North America and Europe, we have portfolio holdings for about 86% of the countries TNA in the Lipper database. This compares to portfolio holdings for only 32% of the TNA of funds domiciled in Asia-Pacific countries and 22% in the other regions. 10 Table I also reports the number of funds per country by funds declared investment type according to their prospectus: ETFs, index funds, and active funds. There are a total of 10,558 active funds in our sample as of 2010 (2,606 in the United States). Explicitly indexed funds include 561 ETFs and 657 index funds (308 and 239 in the United States). Assets under management amount to a total of $1.7 trillion in explicitly indexed funds and $6.2 trillion in active funds. Although passively managed funds have become increasingly popular, active funds still dominate the mutual fund industry both in the United States and elsewhere. II. Explicit and Closet Fund Indexing Around the World First, we document the availability of explicitly indexed funds across different countries and benchmark types. We use holdings-based measures to classify active funds into truly active or closet indexers and describe the level of fees charged to investors across countries. A. Explicit Indexing Table I shows that 22% of equity mutual fund assets under management worldwide are explicitly indexed as of The countries with the highest levels of explicit indexation are Switzerland (58%), Ireland (31%), the United States (27%), and France (25%). At the same time, 10 LionShares coverage of fund holdings is lower in some countries because disclosure is not mandatory. We obtain similar results when we exclude these countries, such that our results are not driven by selective disclosure. 8

10 some countries have almost no passively managed funds domiciled within the country. Investors in these countries which include Italy, the Netherlands, and Portugal, are still able to purchase index funds and ETFs that are offered in these countries but domiciled elsewhere. For example, large asset managers such as Vanguard and BlackRock manage index funds and ETFs domiciled in Ireland and offer these funds across several European markets. We conclude that there is substantial variation across countries in the degree to which fund assets are explicitly indexed. 11 B. Active Fund and Closet Indexing Although explicit indexing is rare in many countries, mutual fund managers classified as active may in fact practice a form of closet indexing (i.e., their fund holdings could be quite similar to the holdings of their benchmark index), while still marketing themselves and charging fees as if they engage in active management. To the extent that fund holdings overlap with index holdings, investors are effectively earning index-like gross returns, which they could obtain at lower fees through index funds or exchange-traded funds (ETFs). We measure the prevalence of closet indexing among active funds in a country using the active share measure developed by Cremers and Petajisto (2009). The active share of a fund represents the share of portfolio holdings that differs from the benchmark index holdings and is calculated as: Active Share,, (1) where w fund,i and w benchmark,i are the portfolio weights of stock i in the fund and its benchmark index, respectively, and the sum is taken over the universe of stocks. For a mutual fund that 11 Calculation of the market share of explicitly indexed funds does not require LionShares holdings data. To investigate the possibility of selection bias from using the sample of 11,776 funds with holdings data in LionShares, we calculate the market share of explicit indexing using the sample of 24,492 funds in Lipper (i.e., including those without a match to the holdings data). The degree of explicit indexation is similar to that reported in Table I. 9

11 never shorts stocks and never buys on margin, its active share will always lie between zero and 100%. Given our international setting, funds may hold different securities in the same company (e.g., common shares, depository receipts, and dual listings) that constitute fundamentally the same ownership stake in a company. We therefore sum all holdings in the same company as part of the same portfolio position. Our analysis of active management requires funds benchmarks and these benchmarks can be self-declared ( Fund Manager Benchmark ) or assigned by Lipper according to a fund s investment strategy ( Technical Indicator Benchmark ). We rely primarily on the Technical Indicator Benchmark both because the Fund Manager Benchmark is sparsely available and to avoid the concern that a fund manager may strategically choose a benchmark in order to rise in the performance rankings. 12 Appendix A lists the 88 specific benchmarks (Technical Indicator Benchmark) that Lipper assigns funds in our sample, which can be classified into three types: world (funds that invest worldwide), regional (funds that invest in a specific geographic region), and country (funds that invest in a specific country). Some of the world, regional, and country funds may have specific industry or investment styles. 13 We construct portfolio weights for the 88 different benchmark indices using the aggregate portfolio holdings of the explicitly indexed funds tracking each benchmark. 14 Use of the actual weights of explicitly indexed funds tracking each benchmark has the advantage that some of the weights in the official benchmark include stocks that in practice may not be fully investable by 12 Results using the Fund Manager Benchmark are consistent with all the results we report. 13 The sample includes only funds whose benchmarks have at least $10 billion of assets under management in The benchmark weights are calculated excluding synthetic ETFs that do not physically replicate the underlying benchmark index. In addition, for about 2% of the fund-year observations where there are not at least five explicitly indexed funds tracking a particular benchmark, we use as an alternative the aggregate portfolio of all active funds that track that benchmark. 10

12 mutual funds due to illiquidity or other constraints. In effect, the active share is measured in excess of explicitly indexed funds. On average (TNA-weighted), active funds in our sample have an active share of 69%, while passive funds have an active share of 16%. 15 The final columns in Table I provide the distribution of the active share measure for the sample of actively managed funds domiciled in each country as of We use an active share below 60% as the cutoff for an active fund to be classified as a closet indexer as in Cremers and Petajisto (2009). 16 All other funds with active share above 60% are classified as truly active. We find there is a wide range of closet indexing across countries. Although in the table we report active share statistics for 2010 only, we find that the active share of funds is an extremely persistent fund attribute over time (the pooled average serial correlation of active share at the fund-level is 0.95). Figure 1 compares the market share (as a percentage of TNA) of explicitly indexed funds, closet indexers and truly active funds across countries as of Countries are sorted, in ascending order, by the market share of truly active funds. The United States has one of the highest levels of explicit index management (with 27% of assets managed by index funds and ETFs), as well as the lowest level of closet indexing among all countries, at 15%. In countries with little explicit indexation, the actively managed funds domiciled there are relatively passive, as measured by their active shares. For example, Canada has a low level of explicit indexation at 7% but a high level of closet indexing at 37% of fund assets under management. Figure 2 shows the time series of the market shares of explicitly indexed funds, closet 15 The average active share of passive funds varies across benchmarks. For example, passive funds that track the S&P 500 index have an average active share of 4%. 16 An active share of 60% means that 40% of the fund portfolio weights overlap with the benchmark index weights. The 60% cutoff is somewhat arbitrary, but as, on average, half the holdings in any portfolio will beat the portfolio s average return, then an active fund (with a manager who tries to beat the benchmark) should have an active share of least 50%. In addition, the 60% threshold corresponds to classifying funds in the bottom tercile of the distribution of active share as closet indexers. 11

13 indexers, and truly active funds over the period. Panels A and B show the evolution in the United States and elsewhere, respectively. Explicit indexing has been increasing over time in both the United States and other countries. While assets under management by active funds dropped during the financial crisis (from $7 trillion in 2007 to $6.2 trillion at the end of 2010), explicitly indexed assets actually grew over the period (from $1.3 trillion to $1.7 trillion) with a big part coming from the growth of ETF assets (from $0.6 trillion to $0.8 trillion). The amount of closet indexing in the United States appears to have remained relatively stable over time with some decline during the financial crisis. Outside the United States, closet indexing has dropped significantly over the sample period. In robustness checks, we use two alternative methods to calculate active share. First, we construct the index weights based only on ETFs that undertake full physical replication of the indices. For the majority of the 88 benchmarks, we can identify a SPDR or ishares ETF that tracks these benchmarks. We call the active share measured against the ETF weights the pure- ETF active share. Second, to address any potential issues with Lipper s assignment of the Technical Indicator Benchmark to each fund, we construct the active share against all possible 88 benchmarks. We take the most representative benchmark for a fund at each particular time as the one with the lowest active share in the manner of Cremers and Petajisto (2009). We call this alternative measure the minimum active share. We find that in our sample both of these alternative measures are highly correlated with the original active share (correlation coefficients of 0.97 with the pure-etf active share and 0.94 with the minimum active share). For active funds, the average (TNA-weighted) pure-etf active share is 70% and the minimum active share is 65% which are very similar to the levels of our main measure of active share. 12

14 C. Explicit and Closet Indexing by Benchmark Type A fund s active share may depend on the opportunities and constraints provided by the investment opportunity set of the fund manager, which differ across benchmarks. Table II shows the total net assets (TNA) and the market shares of explicit and closet indexing per country as of December 2010 in the three different benchmark types: world, regional, and country funds. We further separate funds with a country benchmark into domestic (funds that invest in stocks of the same country where they are domiciled) and foreign (funds that invest in stocks of a country different from the one where they are domiciled). Table II also shows that the majority of equity mutual fund assets are invested domestically ($4.4 trillion). The next most prevalent type of fund is regional funds ($1.8 trillion), followed by world funds ($1.2 trillion), and foreign country funds ($437 billion). However, the proportions are not universal across countries. For example, domestic funds are predominant in the U.S, but world and regional funds are relatively more important in European countries. Table II presents the amount of explicit indexing per country according to fund benchmark type. The level of explicit indexing is highest for domestic country funds, where 27% of the fund assets are indexed. Explicit indexing is used somewhat less frequently for world funds (12%), regional funds (18%), and foreign country funds (18%). In addition, closet indexing is less common for funds pursuing global investment strategies (11%) than for regional funds (24%), domestic country funds (21%), and foreign country funds (27%). The diversity in the universe of stocks tracked by each benchmark has implications for the measurement of active share. For example, the SSgA World Index Equity Fund that tracks the MSCI World index (the most popular world index) includes over 1,600 stocks at the end of 2010, while the SPDR S&P 500 ETF Trust tracking the S&P 500 index (the most popular country 13

15 index) have 500 stocks. Some of the other explicitly indexed funds do not engage in full physical replication. For example, the average index or ETF fund tracking the S&P 500 has an average of 334 stocks. Thus, the definition of active versus the benchmark depends in part on the number of stock positions a fund needs to replicate its benchmark. We take into account this effect and other fund, benchmark, and country characteristics in the analysis of active share below. D. Fees We measure fees and expenses charged to mutual fund shareholders using the total expense ratio (TER) and loads. This category is broader than just management fees and includes all annual expenses that a fund charges its investors for investment management, administration, servicing, transfer agency, audit, and legal. TERs exclude certain distribution fees, such as frontend or back-end loads, as well as annual fees charged by distributors that are separate from the fund charges. Following Khorana, Servaes, and Tufano (2009), we calculate average total shareholder costs (TSC) per year for a fund investor, defined as TER plus one-fifth of the front-end load. This calculation assumes the typical investor holds a fund for five years, and that rear-end loads are waived if the fund is held for that length of time. The TERs are more comparable across countries than management fees, which may include payments for administration and distribution in some countries, but not others. If information on TER is not available (13% of the fund-year observations), we use the annual management fee instead, which constitutes a lower bound for the TER. 17 Table III reports average (TNA-weighted) TSC per country as of the end of The first column shows that the TSC varies considerably across countries, from an average of 0.99% per 17 The TSC ignores bid-ask spreads in the case of ETFs, which are typically narrow. For example, Morningstar (2012) reports that the Lyxor ETF Euro Stoxx 50 (the largest ETF on the Euro Stoxx 50 index) had a trailing 30-day average spread of 0.017% at the NYSE Euronext Paris as of February28,

16 year for funds in the United States to an average of 3.58% for funds in Poland; non-u.s. funds average 2.07%. Thus, consistent with Khorana, Servaes, and Tufano (2009) we find that shareholder costs in the United States are much lower, on average, than in other countries. The next columns in Table III show the average TSC separately for the three types of funds: explicitly indexed funds, closet indexers, and truly active. Figure 3 illustrates the range of the average TSC across the three types of funds in a country. Not surprisingly, explicitly indexed funds have the lowest costs, with an average TSC of 0.35% per year versus 1.64% for active funds. The costs for explicitly indexed funds are lower than for active funds across all countries, but again the United States stands out as having the lowest cost for explicitly indexed funds at 0.26% per year. In most countries, closet indexers are as costly as truly active funds, with an average TSC of 1.64% and 1.66% per year, respectively. These summary statistics confirm that explicitly indexed funds are a low-cost alternative to active funds worldwide and that closet indexers, which offer gross returns similar to passive funds, charge fees that appear indistinguishable from those of truly active funds. The final columns in Table III show average TSC across the different benchmark types: world, regional, and country. Most importantly, there are generally greater differences across countries than across benchmark types within countries. III. Determinants of Explicit and Closet Indexing Across Countries We hypothesize that explicitly indexed funds provide a low-cost substitute for active funds, which could lead to more intense competition in the fund industry. To test our hypothesis, we employ two measures of competition in a country s fund industry: the market share (percentage of total net assets, TNA) and the fees (average total shareholder cost, TSC) that explicitly indexed funds charge. Product differentiation is another way we can see competition in the 15

17 industry. Closet indexers offer a less differentiated product at high fees and may reflect a less competitive market. For these reasons, we use the market share of closet indexers as a measure of the lack of competition in a fund industry. We expect the levels and costs of explicit and closet indexing in a country to depend on several factors, most particularly the funds regulatory environment, the size of and competition in the fund industry, and the level of stock market development in the country. Khorana, Servaes, and Tufano (2005, 2009) show that stronger regulations and laws are related to relatively larger mutual fund industries and lower costs for fund investors. We expect to find more low-cost explicitly indexed products in environments where regulation is stronger, and specifically where mutual fund investor rights are better protected. Beside regulation, fund industry development should matter in the case of the breadth of fund offerings. We expect the overall size and concentration in an industry to positively affect the decision to introduce explicitly indexed products. Stock market conditions should also matter because they determine how easy it is to assemble passive products. Thus, explicitly indexed products are more likely to be offered in larger and more liquid stock markets. Table IV provides results on the determinants of explicit and closet indexing across countries over In Panel A, we focus on the country in which the fund is domiciled. We use two measures of explicit indexing as dependent variables. Explicit Indexing (% by country) is the market share of explicitly indexed funds as a percentage of the TNA in each country in a given year. Explicit Indexing (TSC by country) is the TNA-weighted average total shareholder cost (TSC) of explicitly indexed funds in each country in a given year. We also use a measure of closet indexing as dependent variable. Closet Indexing (% by country) is the market share of active funds with an active share below 60% as a percentage of the TNA in each country in a 16

18 given year. We examine the potential country-level determinants in separate regressions because there are a limited number of countries in our sample. In columns (1), (4), and (7) we test whether indexing is related to the country regulatory environment. Specifically, we use the extent to which regulatory approvals are required for fund management companies (approval) and the quality of a country s judicial system (judicial) as in Khorana, Servaes, and Tufano (2009). 18 We find that in countries with a more efficient judicial system and a stronger rule of law there is more and cheaper explicit indexing and less closet index. A more restrictive regulatory approval regime of new funds is positively associated with explicit indexing. Specifically, in column (1) we find a positive and significant coefficient on both the approval and judicial variables, which indicates that the market share of explicitly indexed funds increases with the quality of the regulatory environment. The economic impact of approval is considerable. For example, an increase in approval from one to two is associated with an increase of about 6% in the market share of explicitly indexed funds. A one standard deviation increase in judicial (7.01) is associated with an increase of about 3% in the market share of explicit indexing. Column (4) shows lower TSC for index funds in countries with more investor protection as measured by the judicial variable. The approval variable is not significantly related to fees. The judicial variable, however, is negatively associated with the level of closet indexing in a country as shown in column (7). We study the relation between indexing and characteristics of a country s fund industry. The results are shown in columns (2), (5) and (8) of Table IV. We use the size and level of 18 Approval is the sum of two dummy variables: (1) whether regulatory approval is required to start a fund and (2) whether the prospectus requires regulatory approval. Judicial system quality is the sum of five variables (all variables are scaled between 0 and 10): the efficiency of the judicial system, rule of law, corruption, risk of expropriation and risk of contract repudiation. 17

19 concentration of the fund industry in the domicile country and the gross domestic product (GDP) per capita as explanatory variables. Fund industry size is defined as the aggregate TNA of all equity mutual funds in the country, and the concentration of the industry is measured by the fund industry Herfindahl based on the market shares of fund families in the equity fund markets. Columns (2) and (5) show that the availability and cost of explicit indexing in a country are significantly related to the size of a country s fund industry. The larger the size of the fund industry, the greater the market share of explicitly indexed funds and the lower their cost. In addition, column (8) shows that the greater the size of the fund industry, the lower the level of closet indexing. These results are consistent with the idea that industry development and economies of scale make it easier to offer low-cost explicitly indexed products and, at the same time, to mitigate closet indexing among active funds. Industry concentration is significantly related only to the market share of explicitly indexed funds. Columns (3), (6), and (9) of Table IV examine the relation between indexing and stock market conditions in the domicile country as proxied by number of stocks, average stock volatility, and average Amihud (2002) illiquidity measure. The degree of explicit indexing increases with the number of stocks traded on a country s stock markets and with liquidity. Similarly, the cost of investing in explicitly indexed funds declined with the number of stocks, volatility, and liquidity of the market. Closet indexers have a larger market share in countries with smaller stock markets. The analysis so far assumes that a fund is offered for sale exclusively in the country of domicile, although some funds are offered for sale in multiple countries, and competition effectively occur across domiciles. We run all regressions in Panel A of Table IV with the market share and cost of explicitly indexed funds and the market share of closet indexers (as well 18

20 as other country characteristics) calculated by country of sale, rather than by country of domicile. The results reported in Panel B are in general consistent with those in Panel A. The results suggest that strong rule of law and larger mutual fund industries and stock markets promote explicit indexation, while discouraging closet indexing by active fund managers. We further refine the analysis by measuring explicit and closet indexing separately by benchmark type (world, regional, country-domestic, and country-foreign) in each domicile country or country of sale. The results (reported in Table IA.IV in the Internet Appendix) are robust to this alternative way of measuring the availability and the cost of explicitly indexed funds. Stock market conditions of the domicile country (or country of sale) directly affect the investment opportunities of domestic mutual funds but only indirectly in the case of foreigncountry, regional or world funds. In Table IA.V in the Internet Appendix, we show that the results are robust when we rely only on the sample of domestic funds to estimate the countrylevel indexing dependent variables. Overall, we conclude that explicit indexing in mutual funds is more prevalent and cheaper in countries with stronger regulation, larger fund industries, and deeper stock markets, while closet indexing is less prevalent. IV. Does Competitive Pressure by Explicitly Indexed Funds Affect Active Funds? We next study the impact of explicit indexing on active funds in terms of product differentiation (degree of activeness) and the price that investors pay for active management (total shareholder costs). The tests use the panel of active equity mutual funds for the period from 2002 to A. Product Differentiation Explicitly indexed funds are low-cost substitutes for the more expensive actively managed 19

21 funds. We expect higher market penetration by indexed products to create competitive pressure on active funds. One possible reaction is that the active fund managers will differentiate themselves by deviating from their benchmarks through more stock picking, sector bets, and market timing. We expect higher product differentiation among active funds, as proxied by active share, when investors are offered a more distinct choice between active and passive funds. To test this hypothesis, we estimate panel regressions in which the dependent variable is the yearly fund-level active share, and the main explanatory variables are the market share of explicit indexing or the average total shareholder costs (TSC) of explicitly indexed funds domiciled in the same country as the fund. We control for fund characteristics, fund benchmark characteristics (number of stocks, volatility, illiquidity) or fund benchmark dummies, dummies for particular types of funds (international, fund of fund, off-shore), and country characteristics. Regressions include year dummies, and standard errors are clustered by domicile-year or country of sale-year. Table V presents the results. Columns (1)-(4) show the results when we measure the indexing variables by country of domicile. Columns (5) and (6) show the results when we measure the indexing variable by country of sale to take into account that a fund can be offered for sale in multiple countries. Since a fund can have multiple share classes with diverse fee schedules offered to different investors, we run the regressions in columns (5) and (6) using the individual share class offered for sale in a country in a given year as a unit of observation. 19 We find that active funds have higher active shares in countries where the average TSC of explicitly indexed funds is lower. Columns (2), (4), and (6) show that the average TSC of 19 A fund with two share classes, each offered for sale in three countries, will have six different observations per year in this sample. In these tests, fund-level variables are measured at the individual share class level, and countrylevel variables are measured by the country of sale. The indexing variables - market share and cost of explicitly indexed funds and the market share of closet indexers - are measured by country of sale. 20

22 explicitly indexed funds coefficient is negative and significant, which suggests that funds choose higher active shares and engage in truly active management when faced with more competitive pressure from explicitly indexed funds. Economically, a standard deviation increase in the TSC of passive funds (0.53) is associated with an increase in average active share of almost 2%. Thus, results are consistent with the hypothesis that active funds perceive explicitly indexed funds as a competitive threat when the explicitly indexed products are a low-cost alternative. We find that active funds have higher active shares in countries where explicitly indexed funds have a higher market share, but this effect is statistically significant only when we define the explicit indexing variable by country of sale in column (5). Fund benchmark conditions are also important determinants of the level of active share. Funds that track benchmarks with more stocks and more liquid stocks tend to be more active. In addition, active share is related to a country s regulatory and fund industry environment. In countries with higher approval requirements and judicial quality, funds are more active, suggesting that stronger legal rules encourage active management. We also find that funds are more active in competitive fund industries such as larger and more contestable industries. The regressions also show that active share is higher for funds with higher tracking error, funds with higher TSC, smaller funds, younger funds, and funds in smaller fund families. More successful fund managers in terms of past performance and attracting more inflows have higher active shares. These coefficients are in line with those in Cremers and Petajisto (2009). The active share regression results are robust in several ways. In Table IA.VI in the Internet Appendix, we measure the indexing variables by country of domicile/benchmark type. In Table IA.VII in the Internet Appendix, we estimate the regression using the non-u.s. fund sample, as funds domiciled in the United States represent a large fraction of the observations. We also 21

23 consider alternative methods to estimate active share such as the pure-etf active share and the minimum active share. Finally, we estimate the regression model using weighted least squares, where the total net assets of the fund are employed as the weights, following Khorana, Servaes, and Tufano (2009). All these robustness checks produce results consistent with those in Table V. Finally, in Table IA.VIII in the Internet Appendix, we show that the results are robust when we rely only on the sample of domestic funds. In this sample where closet indexing is most prevalent, the effect of explicitly indexed funds on the active share of active funds is more pronounced. A one standard deviation increase in the TSC of passive funds is associated with an increase in average active share of almost 3%. We find one additional result: the market share of passive funds is positively associated with the degree of activeness of mutual funds. B. Total Shareholder Costs We study whether the availability of explicit and closet indexing in a country affects the price that investors pay for active management. Given that passive funds are a low-cost alternative to active funds, we expect fees charged by active funds to be negatively related to the market share of explicitly indexed funds and positively related to the average cost of these explicitly indexed products. More closet indexing represents less choice for investors among active funds, however, we expect a positive relation between the market share of closet indexers and fees charged by active funds. We estimate panel regressions of yearly fund-level total shareholder costs (TSC) using the sample of active funds. The main explanatory variables are the prevalence and cost of explicit and closet indexing in the country in which the fund is domiciled or offered for sale. We include fund-level active share as a determinant of the TSC, as well as the same fund and country characteristics used in Table V. We cluster standard errors by domicile-year or country of sale- 22

24 year. 20 Table VI presents the results. Columns (1)-(3) show results when we measure the indexing variables by country of domicile, and columns (4)-(6) show the results when we measure the indexing variables by country of sale. These last specifications are similar to that described in Table V, columns (5) and (6), where the dependent variable is the TSC of a given share class offered for sale in a given country, fund-level explanatory variables are measured at the individual share class level, and country-level explanatory variables are measured by country of sale. We find that total shareholder costs charged by active funds are higher in countries where explicitly indexed funds have less market share and are more expensive. The effect of the cost of explicitly indexed funds is statistically significant in all specifications as shown in columns (2) and (5). The estimates in column (2) show that a decline in the TSC of indexed funds of 50 basis points (the difference between U.S. and non-u.s. explicitly indexed funds TSC) is associated with 16 basis point lower fees charged by active funds. The effect of the market share of explicitly indexed funds is statistically significant only when we define the indexing variables by country of sale in column (4). Overall, the results suggest that investors pay a higher price for active management in markets where low-cost explicit indexed products exert less competitive pressure. Columns (3) and (6) show that active fund fees are higher in markets where closet indexing is more pervasive. According to estimates in column (3), an increase in the level of closet indexing of 15 percentage points (the difference between the United States and the rest of the world) is associated with an increase in total shareholder costs of about 5 basis points. This indicates that closet indexing is reflective of a less competitive fund industry. 20 We obtain similar estimates when we use double-clustered standard errors by country of domicile and year. 23

25 In all specifications, we find that a higher active share (i.e., more active portfolio management by a fund) is associated with funds charging higher fees. A one standard deviation increase in active share (22%) is associated with a TSC increase of about 13 basis points. While active management is not free, there is less of a cost difference between closet indexers and truly active funds compared to the typical cost difference between closet indexers and explicitly indexed funds (average TSC of 1.64% versus 0.35% in Table III). More active management as measured by a higher tracking error of a fund relative to its benchmark is also positively related to fees charged by active funds. Larger funds and younger funds charge investors lower fees. The regulatory and mutual fund industry environment in a country also seems to matter for fees, consistent with evidence in Khorana, Servaes and Tufano (2009). Fees are lower in stronger regulatory environments and when fund industries are larger and more contestable. In Table IA.IX in the Internet Appendix, estimates are consistent when we measure the indexing variables by domicile/benchmark type. Table IA.X in the Internet Appendix report additional robustness checks of the TSC regressions. Estimates are also consistent with those in Table V when we use the non-u.s. fund sample or weighted least squares. Results are robust if we use as alternative explanatory variables the pure-etf active share and the minimum active share. C. Endogeneity A potential concern with our findings is that the availability of explicitly indexed funds (and closet indexers) in a country is endogenously determined. Causality could run from level of fees charged by active funds to market share and cost of explicitly indexed funds, and not the other 24

26 way around. 21 Indeed, the fees charged to investors and the availability of explicitly indexed products are likely jointly determined. To address these potential endogeneity concerns, we need an instrumental variable that is correlated with the usefulness of explicitly indexed funds, but uncorrelated with the fees charged by active funds except indirectly through other independent variables. Our international setting is unique as we can use the variation in regulation and stock market conditions across countries as instruments. Our instrumental variables arguably affect the prevalence of explicit indexing but are (directly) unrelated to the cost of running an active fund: the size, volatility, and liquidity of the country s stock market. 22 The first instrument is the number of stocks in the fund s domicile country stock market, as more stocks available increases the opportunities for diversification and the attractiveness of assembling passive portfolios. Thus, in countries with more stocks we expect explicitly indexed products to be more readily available and cost less. The two other instruments are local stock volatility and liquidity of the local stock market. Higher average stock volatility makes passive funds more attractive for portfolio diversification, and more liquid stocks (proxied by the average Amihud (2002) illiquidity measure) minimize the trading costs for creating index funds. Thus, we the market share of the explicitly indexed funds to be higher and their cost lower in countries with more volatile stocks and greater liquidity. We expect the reverse relation for the market share of closet indexers. 23 Table VII reports estimates of two-stage least squares (2SLS) regressions of the total shareholder cost (TSC) of active funds using the instrumental variables methods. Panel A reports 21 In this case, we would expect that higher fees charged by active funds would lead to introduction of more explicitly indexed products, rather than fewer, as we find in Table VI. 22 We calculate the instruments using all stocks traded in a country and year available in the Datastream/Worldscope database. We only consider the primary listing of stock and stocks with valid returns. 23 The exclusion restriction may hold less in the case of the liquidity instrument, as active fund fees may be directly affected by stock market liquidity through trading costs. We obtain similar estimates when illiquidity is used as a control variable rather than as instrument in Table VII. Estimates are consistent using alternative proxies for stock market liquidity, such as share turnover and the percentage of stocks with zero returns. 25

27 estimates when indexing variables and other country-level variables are measured by domicile country, while Panel B reports estimates when indexing variables are measured by country of sale. Columns (1)-(3) report the first-stage regression estimates when the dependent variables are the market share of explicitly indexed funds, the average TSC of explicitly indexed funds, and the market share of closet indexers. Regressions include the other fund and country characteristics used in Table VI as well as year and benchmark dummies. The first-stage regression results support the view that the availability of explicitly indexed funds is positively related to the number of stocks and volatility (column (1)), while the cost of investing in explicitly indexed funds is negatively related to the number of stocks and volatility (column (2)). Column (3) suggests that the market share of closet indexers is also negatively related to the number of stocks and volatility. F-tests indicate the rejection of the hypotheses that instruments can be excluded from the first-stage regressions, which shows that the instruments are not weak. Furthermore, Hansen s overidentification tests confirm the validity of the instruments, suggesting that they are not related to active fund fees in any other way than through their impact on indexing. In the second-stage regressions, the dependent variable is the TSC of active funds. Columns (4) and (5) show that a higher market share of explicitly indexed funds and a lower cost of investing in explicitly indexed funds are associated with a reduction in the TSC of investing in active funds. These results suggest that the higher competitive pressure created by having passive funds in the market tends to lead their active fund rivals to charge lower fees. In addition, column (6) shows that the market share of closet indexers is positively related to the cost of active management. Panel B of Table VII reports estimates of the same regressions as shown in Panel A, but now 26

28 taking into account that a fund may have multiple share classes offered for sale in multiple countries as in Table VI, columns (7)-(9). The first-stage results by country of sale show that number of stocks is an important determinant of explicit indexing, while liquidity plays a more important role than volatility. Higher liquidity is correlated with more and less expensive explicitly indexed products. The second-stage results by country of sale are consistent with those by domicile in Panel A. The instruments in Table VII are measured according to stock market conditions of the domicile country or country of sale, which directly affect the investment opportunities of domestic mutual funds but only indirectly affect international funds. We find that instruments are significantly correlated with the endogenous explanatory variable even though they are noisy in the case of international funds. In Panel C of Table VII we rely only on the sample of domestic funds and find that both first-stage and second-stage regression results are statistically and economically stronger. In Table IA.XI in the Internet Appendix, passive management and instruments are measured in terms of country/benchmark type and the results are consistent with Table VII results. After we take into account the possibility that the extent of the availability of explicitly indexed funds is endogenous, we find evidence that the effect of more competitive pressure by explicitly indexed funds reduces fees that active funds charge. This evidence suggests a causal link from passive management to fees and competition among active funds. We conclude that passive management plays an important role in enhancing market contestability in the active management industry worldwide. 27

29 V. Returns to Active Management We have shown that competitive pressure from explicitly indexed funds makes active fund managers more likely to differentiate their products and charge lower fees. Next, we test whether these actions result in benefits to mutual fund investors, that is, whether investors benefit in terms of performance from investing in truly active funds and from the availability of explicitly indexed products. To measure the performance benefits we calculate the returns to active management through several approaches using batting averages and benchmark-adjusted returns. We first examine batting averages, defined as the percentage of fund-year observations with positive benchmark-adjusted returns (i.e., when a fund s net return exceeds the return on its benchmark). Panel A of Table VIII and Figure 4 show that the majority of active funds underperform their benchmarks between 2002 to On average, only in 45% of fund-years are there positive benchmark-adjusted returns. When we divide the funds by their active shares, we find that the truly active funds have higher batting averages than closet indexers in every year except in 2008 (i.e., during the financial crisis). Over the full period, truly active funds have a 47% batting average, which is significantly higher than the 41% average for closet indexers. Panel B of Table VIII shows the results when we divide the funds by benchmark type. There is variation in batting averages across benchmark types, with the lowest average for foreign country funds. In all benchmark types, we again find that the average odds of funds beating their benchmark go up with a fund s level of active share. Panel C of Table VIII reports average fund performance using benchmark-adjusted returns. The average benchmark-adjusted return for all active funds in our sample is approximately zero, which is consistent with results in other studies of mutual fund performance. Both equalweighted and value-weighted average performance show a similar pattern, and truly active funds 28

30 outperform closet indexers. Further, we find that the truly active funds are able to outperform their benchmarks on average, with stronger outperformance for larger funds (1.04% per year using value-weighting versus 0.12% per year using equal-weighting). We next examine whether active share predicts future fund performance. We estimate a panel regression of yearly fund risk-adjusted performance on active share and fund and country characteristics. All independent variables are measured with a one-year lag. The regressions also include benchmark and year dummies, and standard errors are clustered by domicile-year. Table IX reports the results. 24 To assess mutual fund performance, we use several risk-adjusted performance measures. Columns (1) and (2) use benchmark-adjusted returns. Columns (3) and (4) use benchmarkadjusted return four-factor alphas, and columns (5) and (6) use excess return (over U.S. T-bills) four-factor alphas. We estimate four-factor alphas using three years of past monthly fund returns with regional factors (Asia, Europe, North America, and Emerging Markets) or world factors in the case of world funds in the manner of Bekaert, Hodrick and Zhang (2009). 25 We then subtract the expected return from the realized fund return to estimate the fund abnormal return (alpha) in each year, which is measured as a sum of an intercept of the model and the residual as in Carhart (1997). Finally, columns (7) and (8) use the fund s information ratio defined as the four-factor benchmark-adjusted return alpha divided by the standard deviation of residuals. We find that funds with higher active share perform better. Thus, active share is a predictor of future fund performance across world markets, consistent with the Cremers and Petajisto (2009) results for U.S. domestic equity mutual funds. The effect of active share on future fund performance is both statistically and economically significant. A one standard deviation increase 24 We obtain similar estimates when we use double-clustered standard errors by country of domicile and year. 25 See Ferreira, Keswani, Miguel, and Ramos (2013) for details about the construction of the factors. 29

31 in active share is associated with a 1% increase in benchmark-adjusted returns and a 0.7% increase in four-factor alphas in the subsequent year. We control for alternative measures of active management in Table IX. The first measure is tracking error (i.e., volatility of the difference between a portfolio return and its benchmark index return), which is the most common measure of active management used in the fund industry. We find that tracking error is not consistently related to future fund performance and may even hurt performance as we see in the specifications that use alphas as performance measure. This suggests that the market rewards funds that are most active in stock selection, which is captured by active share, but does not reward factor bets, which is captured by tracking error. In columns (2), (4), (6), and (8) we further include R-squared as a measure of active management as proposed by Amihud and Goyenko (2012). R-squared is estimated using the four-factor model regression. Following Amihud and Goyenko (2012), we use a logistic transformation of R-squared, TR 2 = log[sqrt(r-squared)/(1 sqrt(r-squared))], as the dependent variable. A lower R-squared is indicative of more active management. We find that R-squared is insignificantly related to future fund performance in a regression that also includes active share and tracking error. Only active share is a statistically significant predictor of future fund performance. 26 The coefficients of the other fund characteristics are consistent with other findings for the U.S. mutual fund industry. Fund size is negatively related to performance, while family size is positively related (consistent with Chen, Hong, Huang, and Kubik (2004)). Total shareholder costs are negatively related to performance (Malkiel (1995), Carhart (1997), and Gil-Bazo and Ruiz-Verdu (2009)). 26 We find that the coefficients on tracking error and TR 2 are also insignificant in regressions that do not include active share. 30

32 In Table IA.XII in the Internet Appendix, estimates are also consistent with those in Table IX using alternative specifications. We estimate the regression using the sample of non-u.s. funds and weighted least squares. We also use as alternative explanatory variables the pure-etf active share and the minimum active share. Results are robust in all these alternative specifications. Results are again consistent results when we use the sample of domestic funds in Table IA.XIII in the Internet Appendix. We next test the implications of our hypothesis that the presence of passive funds affects the behavior of active fund managers. Specifically, we examine whether returns to active management are affected by the prevalence and cost of explicitly indexed funds as well as the prevalence of closet indexing. In these tests, we add the market shares of explicitly indexing and closet indexing and the average cost of explicitly indexing to the performance regressions and the interactions of these variables with active share. Table X presents the estimates of these regressions using four-factor alphas as a measure of risk-adjusted performance. In untabulated results, estimates are similar using the other measures of performance in Table IX. We find that average fund performance is negatively related to the average cost of explicit indexing in column (2), suggesting better average performance in countries where the total shareholder cost (TSC) of explicitly indexed funds is lower. This is consistent with the idea that enhanced competition by low-cost explicitly indexed funds spurs active funds to deliver better after-fee performance to investors. Moreover, the coefficient on the interaction between active share and the average TSC of explicitly index funds is negative and significant at the 10% level. Thus, there is some evidence of higher returns to active management in environments where cost explicitly indexed funds are cheaper. We do not find that the market share of explicit indexing or closet indexing has a significant impact on average fund performance. 31

33 VI. Conclusion We examine the causes and consequences of indexing in the equity mutual fund industry across 32 countries. Our results suggest that many investors worldwide face a limited opportunity set in their choice between active and passive mutual funds. In most countries, investors are not given the option of paying lower fees for passive management. While we find there is little explicit indexing in many countries, many active fund managers are effectively closet indexers when they choose portfolios that track their benchmarks closely. Thus, investors pay higher fees and obtain closet index management, rather than benefiting from truly active management (i.e., higher net returns). Where low-cost explicitly indexed funds are available, they affect the behavior of active funds, which tend to differentiate their products by deviating more from their benchmarks, charge lower fees and deliver higher returns. Our findings suggest that the growth of passively managed funds worldwide will likely increase competitive pressure on actively managed funds. This would improve the efficiency of the asset management industry. The growth of index-based investing could thus have broader implications for financial markets and asset prices that deserve future research. 32

34 References Amihud, Yakov, 2002, Illiquidity and stock returns: Cross-section and time series effects, Journal of Financial Markets 5, Amihud, Yakov, and Ruslan Goyenko, 2012, Mutual fund s R 2 as predictor of performance, Review of Financial Studies, forthcoming. Andonov, Aleksandar, Rob Bauer, and Martijn Cremers, 2012, Can large pension funds beat the market? Asset allocation, market timing, security selection and the limits of liquidity, Working paper, University of Notre Dame. Avramov, Doron, and Russ Wermers, 2006, Investing in mutual funds when returns are predictable, Journal of Financial Economics 81, Bar-Isaac, Heski, Guillermo Caruana, and Vicente Cunat, 2012, Design and market structure, American Economic Review 102, Baumol, William, Stephen Goldfeld, Lilli Gordon, and Michael Koehn, 1990, The Economics of Mutual Fund Markets: Competition versus Regulation (Kluwer Academic Publishers, Boston, MA). Bekaert, Geert, Robert Hodrick, and Xiayan Zhang, 2009, International stock return comovements, Journal of Finance 64, Berk, Jonathan, and Richard Green, 2004, Mutual fund flows and performance in rational markets, Journal of Political Economy 112, Bollen, Nick, and Jeffrey Busse, 2001, On the timing ability of mutual fund managers, Journal of Finance 56, Carhart, Mark, 1997, On persistence in mutual fund returns, Journal of Finance 52,

35 Carlin, Bruce, and Gustavo Manso, 2011, Obfuscation, learning, and the evolution of investor sophistication, Review of Financial Studies 24, Chen, Joseph, Harrison Hong, Ming Huang, and Jeffrey Kubik, 2004, Does fund size erode mutual fund performance? The role of liquidity and organization, American Economic Review 94, Choi, James, David Laibson, and Brigitte Madrian, 2010, Why does the law of one price fail? An experiment on index mutual funds, Review of Financial Studies 23, Coates, John, and Glenn Hubbard, 2007, Competition in the mutual fund industry: Evidence and implications for policy, Journal of Corporate Law 33, Collins, Sean, 2005, Are S&P 500 index mutual funds commodities? Investment Company Institute Perspective. Cremers, Martijn, and Antti Petajisto, 2009, How active is your fund manager? A new measure that predicts performance, Review of Financial Studies 22, Dyck, Alexander, Karl Lins, and Lukasz Pomorski, 2011, Does active management pay? New international evidence, Working paper, University of Toronto. Ferreira, Miguel, and Pedro Matos, 2008, The colors of institutions money: The role of institutional investors around the world, Journal of Financial Economics 88, Ferreira, Miguel, Aneel Keswani, Antonio Miguel, and Sofia Ramos, 2013, The determinants of mutual fund performance: A cross-country study, Review of Finance, forthcoming. French, Kenneth, 2008, Presidential Address: The cost of active investing, Journal of Finance 63,

36 Gil-Bazo, Javier, and Pablo Ruiz-Verdu, 2009, The relation between price and performance in the mutual fund industry, Journal of Finance 64, Grinblatt, Mark, and Sheridan Titman, 1989, Mutual fund performance: An analysis of quarterly portfolio holdings, Journal of Business 62, Grinblatt, Mark, and Sheridan Titman, 1993, Performance measurement without benchmarks: An examination of mutual fund returns, Journal of Business 66, Gruber, Martin, 1996, Another puzzle: The growth in actively managed mutual funds, Journal of Finance 51, Hortacsu, Ali, and Chad Syverson, 2004, Product differentiation, search costs, and competition in the mutual fund industry: A case study of S&P 500 index funds 119, Investment Company Institute, 2011, Investment Company Fact Book. Jensen, Michael, 1968, The performance of mutual funds in the period , Journal of Finance 23, Kacperczyk, Marcin, and Amit Seru, 2007, Fund manager use of public information: New evidence on managerial skills, Journal of Finance 62, Kacperczyk, Marcin, Clemens Sialm, and Lu Zheng, 2005, On the industry concentration of actively managed equity mutual funds, Journal of Finance 60, Khorana, Ajay, Henri Servaes, and Peter Tufano, 2005, Explaining the size of the mutual fund industry around the world, Journal of Financial Economics 78, Khorana, Ajay, Henri Servaes, and Peter Tufano, 2009, Mutual fund fees around the world, Review of Financial Studies 22,

37 Kosowski, Robert, Allan Timmermann, Russ Wermers, and Hal White, 2006, Can mutual fund stars really pick stocks? New evidence from a bootstrap analysis, Journal of Finance 61, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, 1998, Law and finance, Journal of Political Economy 106, Malkiel, Burton, 1995, Returns from investing in equity mutual funds 1971 to 1991, Journal of Finance 50, Morningstar, 2012, Total cost of ETF ownership: The bid-offer spread, February 28. Pastor, Lubos, and Robert Stambaugh, 2012, On the size of the active management industry, Journal of Political Economy 120, Sharpe, William, 1966, Mutual fund performance, Journal of Business 39, Wahal, Sunil, and Albert Wang, 2011, Competition among mutual funds, Journal of Financial Economics 99, Wermers, Russ, 2000, Mutual fund performance: An empirical decomposition into stock-picking talent, style, transactions costs, and expenses, Journal of Finance 55, Wurgler, Jeffrey, 2011, On the economic consequences of index-linked investing, in Gerald Rosenfeld, Jay Lorsch, and Rakesh Khurana, eds.: Challenges to Business in the Twenty- First Century: The Way Forward (American Academy of Arts and Sciences). 36

38 Table I Explicit and Closet Indexing by Country This table presents number of funds (first row) and TNA in billions of U.S. dollars (second row) per domicile country for the sample of open-end equity mutual funds taken from Lipper as of December The all funds column presents the number of funds and total net assets (TNA) for all funds in Lipper. All other columns present statistics for the sample of Lipper funds for which holdings are available in LionShares. TNA coverage is the percentage of TNA that funds in our sample represent out of the Lipper universe in each country. Explicit indexing is the percentage that index funds and exchange-traded funds represent of the TNA in a country. The final columns present the percentage of the TNA that active funds with an active share (AS) below 0.2, 0.4, 0.6, 0.8 and 1 represent of the TNA of funds with holdings in a country. Funds with Holdings TNA Coverage (%) Exchange Traded Funds Passive Funds Explicit Indexing (%) Active Funds Domicile All Funds Index Funds Active Funds AS<0.2 (%) AS<0.4 (%) AS<0.6 (%) AS<0.8 (%) AS<1 (%) Austria Belgium Canada 1, Denmark Finland France 1, Germany Ireland Italy Liechtenstein Luxembourg 3,017 2, , Netherlands Norway Poland Portugal Spain Sweden Switzerland U.K. 1, U.S. 4,173 3, ,606 5, , , Asia Pacific 6,163 1, , Other Regions 1, Total 20,319 8, ,952 (Non-U.S.) 4, , , Total 24,492 11, ,558 9, , ,

39 Table II Explicit and Closet Indexing by Country and Benchmark Type This table presents total net assets (TNA) and market shares of explicit and closet indexing per domicile country and benchmark type for the sample of open-end equity mutual funds taken from Lipper with portfolio holdings in LionShares as of December Explicit indexing is the percentage that index funds and exchange-traded funds represent of the TNA in a country. Closet indexing is the percentage of the TNA that active funds with active share measure below 0.6 represent of the TNA in a country. Funds are classified based on their benchmark as world funds, regional funds, country - domestic funds (funds investing in their domicile country) and country - foreign funds (funds investing in a country different from their domicile). Refer to Appendix A for a list of benchmarks. Domicile TNA ($ billion) World Funds Regional Funds Country - Domestic Funds Country - Foreign Funds Explicit Indexing (%) Closet Indexing (%) TNA ($ billion) Explicit Indexing (%) Closet Indexing (%) TNA ($ billion) Explicit Indexing (%) Closet Indexing (%) TNA ($ billion) Explicit Indexing (%) Closet Indexing (%) Austria Belgium Canada Denmark Finland France Germany Ireland Italy Liechtenstein Luxembourg Netherlands Norway Poland Portugal Spain Sweden Switzerland U.K U.S , Asia Pacific Other Regions Total (Non-U.S.) Total 1, , ,

40 Table III Average Total Shareholder Costs by Country This table presents the total net assets (TNA)-weighted average total shareholder cost in percentage points by country for the sample of open-end equity mutual funds taken from Lipper with portfolio holdings in LionShares as of December Total shareholder cost is the annual total expense ratio plus one-fifth of the front-end load. Explicit indexing includes index funds and exchange-traded funds. Closet indexing includes active funds with active share below 0.6. Truly active includes active funds with active share above 0.6. Funds are classified based on their benchmark as world funds, regional funds, country - domestic funds (funds investing in their domicile country) and country - foreign funds (funds investing in a country different from their domicile). Refer to Appendix A for a list of benchmarks. Total Shareholder Total Shareholder Cost of Indexed and Active Funds Total Shareholder Cost by Benchmark Type Domicile Cost Explicit Indexing Closet Indexing Truly Active World Regional Country - Domestic Country Foreign Austria Belgium Canada Denmark Finland France Germany Ireland Italy Liechtenstein Luxembourg Netherlands Norway Poland Portugal Spain Sweden Switzerland U.K U.S Asia Pacific Other Regions Total (Non-U.S.) Total

41 Table IV Determinants of Explicit and Closet Indexing at Country-Level This table presents estimates of yearly country-level regressions where the dependent variable is the percentage that index funds and exchange-traded funds represent of the TNA in a country (explicit indexing (% by country)), the TNA-weighted average total shareholder cost of index funds and exchange-traded funds in a country (explicit indexing (TSC by country)), and the percentage that active funds with active share measure below 0.6 represent of the TNA in a country (closet indexing (% by country)). The sample includes open-end equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to In Panel A the unit of observation is a domicile country j in year t. In Panel B the unit of observation is a country of sale k in year t. Regressions include year dummies. Refer to Appendix B for variable definitions. Robust t-statistics are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. Panel A: by Domicile Country Explicit Indexing (% by country) Explicit Indexing (TSC by country) Closet Indexing (% by country) (1) (2) (3) (4) (5) (6) (7) (8) (9) Approval ** (2.51) (1.54) (-0.40) Judicial ** *** *** (2.44) (-4.27) (-3.53) Fund industry size (log) *** *** *** (3.38) (-2.77) (-3.52) Fund industry Herfindahl *** (4.28) (0.20) (-0.16) GDP per capita (log) (-0.17) (-1.39) (-1.09) Number of stocks ( 10 3 ) *** *** *** (3.73) (-5.44) (-6.29) Volatility ** (0.12) (-2.31) (0.65) Amihud illiquidity *** *** (-2.81) (6.23) (0.32) Observations R-squared

42 Table IV (Continued) Panel B: by Country of Sale Explicit Indexing (% by country) Explicit Indexing (TSC by country) Closet Indexing (% by country) (1) (2) (3) (4) (5) (6) (7) (8) (9) Approval *** (4.69) (-0.38) (-0.06) Judicial *** *** *** (2.87) (-5.21) (-4.17) Fund industry size (log) *** *** *** (4.14) (-4.50) (-2.88) Fund industry Herfindahl ** *** (2.49) (0.63) (-2.89) GDP per capita (log) *** (-0.26) (-3.26) (-1.42) Number of stocks ( 10 3 ) *** *** *** (6.48) (-5.00) (-6.05) Volatility (-1.22) (-1.00) (0.44) Amihud illiquidity *** (-0.91) (6.22) (0.89) Observations R-squared

43 Table V Determinants of Active Management This table presents estimates of panel regressions where the dependent variable is the yearly active share, defined as the percentage of a fund s portfolio holdings that differ from the fund s benchmark. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to In columns (1)-(4) the unit of observation is a primary fund i domiciled in country j in year t. In columns (5) and (6) the unit of observation is a fund share class s offered for sale in country k in year t. Regressions include year dummies and bechmark dummies in some specifications. Refer to Appendix B for variable definitions. Robust t-statistics clustered by domicile-year (columns (1)-(4)) or country of sale-year (columns (5) and (6)) are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. by Domicile Country by Country of Sale (1) (2) (3) (4) (5) (6) Explicit indexing (% by country) ** (1.28) (0.47) (2.38) Explicit indexing (TSC by country) *** *** *** (-3.33) (-4.90) (-4.33) Number of stocks (benchmark) ( 10 3 ) *** *** (8.12) (8.04) Volatility (benchmark) (-0.14) (-0.34) Amihud illiquidity (benchmark) *** *** (-3.26) (-2.85) Tracking error *** *** *** *** *** *** (9.99) (9.50) (7.47) (7.16) (14.09) (13.81) Total shareholder cost *** *** *** *** *** *** (11.47) (12.14) (15.43) (15.71) (21.00) (20.52) Total net assets (log) *** *** *** *** ** ** (-10.17) (-9.56) (-10.87) (-10.16) (2.37) (2.38) Family total net assets (log) *** *** *** *** *** *** (-7.03) (-7.46) (-5.06) (-5.55) (-18.84) (-18.90) Fund age *** *** *** *** *** *** (-6.59) (-6.57) (-5.55) (-5.38) (-8.93) (-8.87) Flows *** *** *** *** *** *** (6.13) (6.03) (6.69) (7.05) (8.19) (8.22) Benchmark-adjusted return *** *** *** *** *** *** (2.91) (2.74) (5.87) (5.64) (11.69) (11.52) International fund dummy *** *** *** *** *** *** (5.11) (4.61) (-4.02) (-3.18) (-5.18) (-5.20) Fund of fund dummy *** *** *** *** *** *** (4.43) (4.80) (4.58) (5.01) (7.97) (8.02) Off-shore fund dummy * *** *** *** *** (1.00) (1.69) (2.86) (4.26) (6.31) (6.35) Approval * ** *** ** ** (1.57) (1.67) (1.98) (3.14) (-2.05) (-2.08) Judicial *** *** *** *** *** ** (4.31) (3.23) (4.70) (3.92) (4.89) (2.48) Fund industry size (log) *** *** *** ** *** *** (11.46) (10.08) (5.20) (2.50) (3.74) (3.23) Fund industry Herfindahl *** *** *** (-2.74) (-1.48) (-6.31) (-6.00) (-0.36) (0.96) GDP per capita (log) *** *** *** *** *** *** (-5.68) (-5.69) (-7.80) (-8.56) (-7.10) (-6.73) Benchmark dummies No No Yes Yes Yes Yes Observations 58,487 56,554 58,487 56, , ,797 R-squared

44 Table VI Determinants of Total Shareholder Costs of Active Funds This table presents estimates of panel regressions where the dependent variable is the yearly total shareholder cost, defined as total expense ratio plus one-fifth of the front-end load. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to In columns (1)-(6) the unit of observation is a primary fund i domiciled in country j in year t. In columns (4)-(6) the unit of observation is a fund share class s offered for sale in country k in year t. Regressions include year and benchmark dummies. Refer to Appendix B for variable definitions. Robust t-statistics clustered by domicile-year (columns (1)-(3)) or country of sale-year (columns (4)-(6)) are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. by Domicile Country by Country of Sale (1) (2) (3) (4) (5) (6) Explicit indexing (% by country) *** (-0.36) (-4.18) Explicit indexing (TSC by country) *** *** (7.44) (5.84) Closet indexing (% by country) ** ** (2.44) (2.29) Active share *** *** *** *** *** *** (13.96) (14.29) (13.93) (23.96) (23.56) (23.57) Tracking error *** *** *** *** *** *** (6.31) (6.61) (6.59) (9.87) (10.07) (10.35) Total net assets (log) *** *** *** *** *** *** (-20.27) (-22.74) (-20.53) (-18.01) (-17.83) (-17.78) Family total net assets (log) *** *** *** *** *** *** (2.82) (3.40) (2.87) (-5.06) (-4.62) (-4.64) Fund age *** *** *** *** *** *** (6.89) (6.30) (6.80) (13.18) (12.99) (13.02) Flows *** *** *** (0.90) (0.52) (0.72) (-6.59) (-6.51) (-6.51) Benchmark-adjusted return *** *** *** *** *** *** (-4.04) (-3.55) (-3.97) (-14.02) (-13.87) (-14.09) International fund dummy *** *** *** *** *** *** (14.99) (10.13) (13.60) (17.25) (19.34) (16.68) Fund of fund dummy *** *** *** *** *** *** (-6.90) (-6.78) (-7.04) (-13.99) (-13.89) (-13.98) Off-shore fund dummy *** *** *** *** (-0.43) (-3.09) (-0.36) (22.04) (21.67) (22.07) Approval *** *** *** (-8.55) (-11.81) (-8.63) (-0.11) (-0.51) (-0.61) Judicial ** (-1.59) (-0.68) (-1.50) (-1.16) (0.49) (-2.05) Fund industry size (log) *** *** *** *** *** (-3.39) (0.16) (-2.96) (-5.59) (-5.29) (-6.05) Fund industry Herfindahl *** *** *** (-4.27) (-4.27) (-4.58) (-0.40) (-1.18) (0.32) GDP per capita (log) ** *** *** *** *** *** (2.24) (3.14) (2.73) (2.80) (2.65) (2.61) Observations 58,487 56,554 58, , , ,103 R-squared

45 Table VII Instrumental Variables Regression of Total Shareholder Costs of Active Funds This table presents estimates of instrumental variables regressions of the yearly total shareholder cost, defined as total expense ratio plus one-fifth of the front-end load. The first-stage dependent variables are the percentage that index funds and exchangetraded funds represent of the TNA in a country (explicit indexing (% by country)), TNA-weighted average total shareholder cost of index funds and exchange-traded funds funds in a country (explicit indexing (TSC by country)), or percentage that active funds with active share measure below 0.6 represent of the TNA in a country (closet indexing (% by country)). The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to In Panel C the sample is restricted to domestic funds. In Panels A and C the unit of observation is a primary fund i domiciled in country j in year t. In Panel B the unit of observation is a fund share class s offered for sale in country k in year t. Regressions include the fund and country control variables used in Table VI as well as year and benchmark dummies. Refer to Appendix B for variable definitions. Robust t-statistics clustered by domicile-year (Panel A) or country of sale-year (Panel B) are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. Explicit Indexing (% by country) Panel A: by Domicile Country First Stage Regression Explicit Indexing (TSC by country) Second Stage Regression Closet Indexing (% by country) Total Shareholder Cost (1) (2) (3) (4) (5) (6) Explicit indexing (% by country) *** (-2.87) Explicit indexing (TSC by country) *** (3.67) Closet indexing (% by country) ** (2.11) Number of stocks ( 10 3 ) * *** *** (1.74) (-6.14) (-7.54) Volatility ** *** ** (2.59) (-3.50) (-2.15) Amihud illiquidity (-1.07) (-0.40) (0.16) Observations 58,487 56,554 58,487 58,487 56,554 58,487 R-squared Panel B: by Country of Sale Explicit Indexing (% by country) First Stage Regression Explicit Indexing (TSC by country) Second Stage Regression Closet indexing (% by country) Total Shareholder Cost (1) (2) (3) (4) (5) (6) Explicit indexing (% by country) *** (-4.37) Explicit indexing (TSC by country) *** (3.28) Closet indexing (% by country) *** (6.93) Number of stocks *** *** *** (7.07) (-2.95) (-11.85) Volatility *** (-2.91) (0.62) (0.47) Amihud illiquidity *** *** (3.10) (-2.69) (-0.85) Observations 423, , , , , ,103 R-squared

46 Explicit Indexing (% by country) Table VII (Continued) Panel C: by Domicile Country - Domestic Fund First Stage Regression Explicit Indexing (TSC by country) Second Stage Regression Closet Indexing (% by country) Total Shareholder Cost (1) (2) (3) (4) (5) (6) Explicit indexing (% by country) *** (-3.98) Explicit indexing (TSC by country) *** (4.65) Closet indexing (% by country) *** (2.62) Number of stocks ( 10 3 ) ** *** *** (1.97) (-6.72) (-5.66) Volatility *** *** (3.39) (-3.97) (-1.30) Amihud illiquidity *** (-4.68) (1.40) (1.52) Observations 24,825 24,156 24,825 24,825 24,156 24,825 R-squared

47 Table VIII Average Performance of Active Funds This table presents statistics on yearly benchmark-adjusted returns, defined as the difference between the fund return and its benchmark return. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to Closet indexers includes active funds with active share below 0.6. Truly active includes active funds with active share above 0.6. Panel A presents the percentage of fund-year observations with positive benchmark-adjusted returns by year. Panel B presents the percentage of fund-year observations with positive benchmark-adjusted returns for each benchmark type. Panel C presents the equal-weighted and value-weighted (TNA) average abnormal fund performance. Panel A: Percentage of Fund-Year Observations with Positive Benchmark-Adjusted Returns by Year Closet indexers 42% 31% 33% 37% 39% 40% 51% 44% 46% 41% Truly active 53% 46% 45% 54% 40% 53% 40% 50% 47% 47% All funds 49% 41% 41% 49% 40% 50% 43% 49% 46% 45% Panel B: Percentage of Fund-Year Observations with Positive Benchmark-Adjusted Returns by Benchmark Type World Regional Country - Domestic Country - Foreign Closet indexers 38% 40% 46% 33% Truly active 47% 48% 49% 40% All funds 46% 45% 48% 38% Panel C: Average Benchmark-Adjusted Returns Equal-Weighted Value-Weighted Closet indexers -0.35% 0.12% Truly active 0.12% 1.04% All funds -0.02% 0.77% 46

48 Table IX Determinants of Performance of Active Funds This table presents estimates of panel regressions where the dependent variable is the yearly benchmark-adjusted return (columns (1) and (2)), benchmark-adjusted return four-factor alpha (columns (3) and (4)), excess return four-factor alpha (columns (5) and (6)) and information ratio (columns (7) and (8)). Benchmark-adjusted return is the difference between the fund return and its benchmark return. Four-factor alphas are estimated using three-year of past monthly fund returns in U.S. dollars with regional factors (Asia, Europe, North America or Emerging Markets) or world factors in the case of global funds. The information ratio is the ratio of the alpha to the idiosyncratic volatility from the benchmark-adjusted return four-factor model. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to The unit of observation is a primary fund i domiciled in country j in year t. All explanatory variables are lagged by one period. Regressions include year and benchmark dummies. Refer to Appendix B for variable definitions. Robust t-statistics clustered by domicile-year are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. Benchmark-Adjusted Returns Benchmark-Adjusted Four-Factor Alphas Excess Return Four-Factor Alphas Information Ratio (1) (2) (3) (4) (5) (6) (7) (8) Active share *** *** *** *** ** ** *** *** (6.45) (6.51) (6.19) (6.21) (2.46) (2.43) (7.50) (7.53) Tracking error *** *** *** *** (0.56) (0.54) (-3.28) (-3.22) (-3.00) (-3.06) (0.83) (1.00) TR (-0.13) (0.62) (0.98) (-0.70) Total shareholder cost *** *** *** *** *** *** *** *** (-4.98) (-4.96) (-4.60) (-4.62) (-2.88) (-2.89) (-5.96) (-6.00) Total net assets (log) *** *** *** *** *** *** *** ** (-3.97) (-3.89) (-2.62) (-2.69) (-4.54) (-4.66) (-2.66) (-2.57) Family total net assets (log) *** *** *** *** *** *** *** *** (3.36) (3.36) (5.53) (5.54) (4.01) (4.01) (5.68) (5.69) Fund age (1.19) (1.16) (-0.26) (-0.22) (0.13) (0.20) (0.22) (0.17) Flows *** *** *** *** (-0.59) (-0.59) (2.63) (2.63) (0.03) (0.02) (4.24) (4.26) International fund dummy (-0.88) (-0.89) (-0.80) (-0.79) (-0.91) (-0.88) (-0.78) (-0.81) Fund of fund dummy ** ** *** *** * * *** *** (-2.51) (-2.52) (-2.69) (-2.72) (-1.70) (-1.74) (-2.79) (-2.77) Off-shore fund dummy (-0.21) (-0.22) (0.35) (0.39) (0.37) (0.41) (0.90) (0.86) Approval * * (-1.51) (-1.50) (-1.43) (-1.43) (-1.82) (-1.82) (-1.05) (-1.04) Judicial ** ** *** *** (1.49) (1.48) (2.33) (2.34) (0.61) (0.65) (2.65) (2.61) Fund industry size (log) (1.57) (1.57) (0.26) (0.25) (1.31) (1.30) (0.51) (0.52) Fund industry Herfindahl (-0.15) (-0.15) (-0.44) (-0.44) (1.49) (1.49) (-0.84) (-0.84) GDP per capita (log) (-0.26) (-0.26) (0.03) (0.01) (-0.57) (-0.59) (-0.28) (-0.25) Observations 50,925 50,925 51,570 51,570 51,570 51,570 51,570 51,570 R-squared

49 Table X Determinants of Performance of Active Funds: Effect of Explicit and Closet Indexing This table presents estimates of panel regressions where the dependent variable is the yearly benchmark-adjusted return fourfactor alpha. Benchmark-adjusted return is the difference between the fund return and its benchmark return. Four-factor alphas are estimated using three-year of past monthly fund returns in U.S. dollars with regional factors (Asia, Europe, North America or Emerging Markets) or world factors in the case of global funds. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to The unit of observation is a primary fund i domiciled in country j in year t. All explanatory variables are lagged by one period. Regressions include the fund and country control variables used in Table IX as well as year and benchmark dummies. Refer to Appendix B for variable definitions. Robust t-statistics clustered by domicile-year are reported in parentheses. *, **, *** reflects significance at the 10%, 5% and 1% levels. (1) (2) (3) Active share *** * ** (5.29) (1.73) (1.98) Explicit indexing (% by country) (0.43) Active share x Explicit indexing (% by country) (-0.95) Explicit indexing (TSC by country) ** (-2.34) Active share x Explicit indexing (TSC by country) * (1.75) Closet indexing (% by country) (-0.41) Active share x Closet indexing (% by country) (0.75) Observations 51,570 50,007 51,570 R-squared

50 Figure 1 Explicitly-Indexed, Closet Indexers and Truly Active Funds by Country This figure shows the percentage that index funds and exchange-traded funds represent of the TNA in a country (explicit indexing), the percentage that active funds with active share measure below 0.6 represent of the TNA in a country (closet indexing), and the percentage that active funds with active share measure above 0.6 represent of the TNA in a country (truly active) as of December The sample includes open-end equity mutual funds from Lipper for which holdings are available in LionShares. 49

51 Figure 2 Explicit and Closet Indexing by Year This figure shows the yearly percentage that index funds and exchange-traded funds represent of the TNA in a country (explicit indexing), the percentage that active funds with active share measure below 0.6 represent of the TNA in a country (closet indexing), and the percentage that active funds with active share measure above 0.6 represent of the TNA in a country (truly active). The sample includes open-end equity mutual funds from Lipper for which holdings are available in LionShares from 2002 to Panel A uses the sample of U.S. funds and Panel B the sample of non-u.s. funds. Panel A: U.S. Funds Panel B: Non-U.S. Funds 50

52 Figure 3 Average Total Shareholder Cost by Country This figure shows the TNA-weighted average total shareholder cost, defined as total expense ratio plus one-fifth of the front-end load, as of December The sample includes open-end equity mutual funds from Lipper for which holdings are available in LionShares. Explicit indexing includes index funds and exchange-traded funds. Closet indexing includes active funds with active share below 0.6. Truly active includes active funds with active share above

53 Figure 4 Average Performance of Active Funds by Year This figure shows the percentage of active fund-year observations with positive yearly benchmark-adjusted returns. A percentage of 50% means that half of the active funds in that group outperformed their respective benchmarks in a given year. The sample includes open-end active equity mutual funds taken from Lipper for which holdings are available in LionShares from 2002 to Closet indexing includes active funds with active share below 0.6. Truly active includes active funds with active share above