Is there a Dark Side to Exchange Traded Funds (ETFs)? An Information Perspective

Transcription

1 Is there a Dark Side to Exchange Traded Funds (ETFs)? An Information Perspective by Doron Israeli, Charles M. C. Lee, and Suhas Sridharan ** April 19, 2014 Preliminary and Incomplete (Please Do Not Quote) Abstract In this paper, we hypothesize and find that increased ETF ownership is accompanied by a decline in the pricing informativeness of the underlying component securities. Consistent with predictions from noisy rational expectations models with costly information, we first demonstrate that an increase in ETF ownership is associated with higher trading costs (measured as bid-ask spreads). Next, we show that an increase in ETF ownership is associated with: (1) an increase in the co-movement of firm specific returns with market and related industry returns ( price synchronicity ); (2) a decline in the predictive power of current firm specific returns for future earnings ( future earnings response coefficients ), and (3) a decline in the number of analysts covering the firm. Collectively, our evidence suggests increased ETF ownership can lead to lower benefits from information acquisition and a deterioration in firms information environment. JEL Classifications: G11, G14, M41 Keywords: Pricing informativeness; Exchange traded funds (ETFs); Institutional ownership ** Israeli (israelid@idc.ac.il) is at the Arison School of Business, the Interdisciplinary Center (IDC), Herzliya, Israel; Lee (clee8@stanford.edu) is at the Graduate School of Business, Stanford University, and Sridharan (suhas.sridharan@anderson.ucla.edu) is at the Anderson School of Management, UCLA. We gratefully acknowledge comments from seminar participants at UCLA, Emory University, Tel Aviv University, the Interdisciplinary Center (IDC) Herzliya, and research assistance from Woo Young Park and Padmasini Venkatachari. 1

2 Is there a Dark Side to Exchange Traded Funds (ETFs)? An Information Perspective I. Introduction Traditional noisy rational expectations models with costly information feature agents who expend resources to become informed. These informed agents earn a return on their information acquisition efforts by trading against the uninformed, and as they do so, the information they possess is incorporated into prices. 1 In these models, the supply of noisy traders adjusts to provide just sufficient reward for costly efforts in information acquisition and processing. The equilibrium between cost constraints faced by informed traders and gains from trading against the uninformed is reflected in the level of informational efficiency of security prices in the market. The inherent tension between the efficiency with which information is being incorporated into stock prices, and the incentives needed to acquire that information and disseminate it, is central to understanding the informational role of prices (e.g., Hayek 1945, Grossman 1989). In this paper, we use a natural experiment to examine the economic linkage between the market for firm-specific information and the market for individual securities and uninformed traders. Specifically, we study the impact of exchange-traded fund (ETF) ownership on the pricing informativeness of the individual component securities underlying the fund. In frictionless markets, a firm s ownership structure should have little to do with the informativeness of its share price. However the reality of market frictions arising from 1 See for example, Grossman and Stiglitz (1980), Admati (1985), Diamond and Verrecchia (1981), and Verrecchia (1982). 2

3 information acquisition costs leads us to posit that increased ownership by ETFs can be a significant economic event that has direct consequences for pricing informativeness. Our main hypothesis is that ETF ownership can influence a stock s pricing informativeness through its impact on the number of underlying securities available for trade and number of uninformed investors willing to trade. As ETF ownership grows, an increasing proportion of the outstanding shares for the underlying security becomes locked up (held in trust) by the fund sponsor. Although these shares are available for trade as part of a basket transaction at the ETFlevel, they are no longer available to traders who wish to transact on firm-specific information. In addition, ETFs offer an attractive investment alternative for uninformed traders who would otherwise trade the underlying component securities. These two effects create a steady siphoning of firm-level liquidity which in turn generates a disincentive for informed traders to expend resources to obtain firm-specific information. In particular, we first conjecture that as ETFs become larger holders of a firm s shares, the costs of transacting in the underlying security will increase. We further posit that these increased costs will serve as a deterrent to traders who would otherwise expend resources on information acquisition about that stock. In other words, the incentive for traders to seek out, acquire, and trade on firm-specific information will decrease, leading to a general deterioration in the pricing informativeness of the underlying securities held by ETFs. 2 To test these hypotheses, we conduct a series of tests using a cross-section of U.S. stocks that are held by ETFs between 2000 and Our first set of tests examines the effect of increased 2 The siphoning of liquidity from component securities can occur with other passive index products as well, such as open-end index funds. However, a key difference between ETFs and other index-linked open-end funds is that ETF shares are traded on organized exchanges throughout the day, while transactions with open-end funds occur only at the end of the day, and only at net asset value (NAV). In Section 2, we explain in detail the implications of this difference for our tests. 3

4 ETF ownership on firm liquidity (relative bid-ask spreads, or HLSPREAD). Recent studies provide evidence that increased ETF ownership is associated with an increase in intraday volatility (Ben-David et al. 2014) or a decrease in firm liquidity (Hamm 2013). Using a more extensive dataset and firm-level longitudinal data, we document a similar liquidity effect in our sample. Specifically, after controlling for firm size, book to market ratio, share turnover, and return volatility, we find that increased ETF ownership is associated with higher average daily bid-ask spreads. In our sample, a one percentage point increase in ETF ownership is associated with a 1.4% increase in the relative spread. This effect is strongest among firms with high ETF ownership and is unaffected by controlling for the overall level of institutional ownership. We also conduct a series of tests to evaluate the effect of ETF ownership on stock pricing informativess. Building on predictions from noisy rational expectations models with costly information, we expect pricing informativeness to decline as a consequence of increased ETF ownership. We examine the effect of ETF ownership on three metrics, each of which is measured over future time periods: (1) the extent to which variation in firm specific returns is explained by variation in market and related-industry returns (i.e., firms price synchronicity, or SYNCH); (2) the extent to which current firm specific returns forecast future earnings (i.e., firms future earnings response coefficient, or FERC); and (3) the extent to which firms are covered by analysts (i.e., analyst coverage, or ANALYST). To the extent that price synchronicity measure reflects lower firm-level pricing informativeness (Morck et al. 2000, Durnev et al 2003, and Roll 1988) we should expect a positive correlation between the level of ETF ownership and SYNCH, measured over the next fiscal quarter. 3 Using future earnings 3 We hasten to point out that prior evidence on whether the R-squared (price synchronicity) measure developed by Morck et al. (2000) captures pricing informativeness is decidedly mixed. In this study we adopt the view offered by the majority of studies of stock price synchronicity (Roll 1988, Wurgler 2000, 4

5 response coefficients measured over the next fiscal year as an alternate measure of pricing informativeness, we expect a negative correlation between our ETF holdings variable and the firm s FERC. Finally, we also predict a negative correction between ETF holdings and the number of analysts covering the stock (ANALYST). Our findings are consistent with these hypotheses. First, after controlling for size, book to market ratio, systematic and idiosyncratic risks, return skewness, and institutional ownership, we find that firms with higher levels of ETF ownership experience higher levels of stock return synchronicity. We observe a nonlinear positive relationship between ETF ownership and the extent to which variation in firm specific returns is explained by variation in market wide and related- industry returns. We also find that ETF ownership is related to the magnitude of future earnings response coefficients in a manner consistent with our hypothesis. In our sample, a one percent increase in ETF ownership is associated with a five percent reduction in the magnitude of the future earnings response coefficient. This effect is strongest among firms with particularly high levels of ETF ownership; firms with more than three percent ETF ownership experience a twenty-two percent reduction in the future earnings response coefficient. These results are robust to the inclusion of controls for institutional ownership, asset growth, firm size, future returns, and a loss indicator. Our findings suggest that an increase in ETF ownership is associated with a decrease in price informativeness in the form of weaker incorporation of next period s earnings into current stock prices. Our empirical analyses also indicate that firms with high ETF ownership suffer from reduced information gathering activities by market participants. After controlling for firm size, book to Durnev et al. 2003, DeFond and Hung 2004, and Durnev et al 2004). These studies suggest that lower R- squared firms (i.e. firms with higher idiosyncratic volatility) have more informative prices. However, recent studies by Chan and Chan (2014) and Li et al. (2014) find contrary evidence. To the extent that the former studies misinterpret the economic meaning of synchronicity, so does this part of our study. 5

6 market ratio, return and earnings volatilities, share turnover, and levels of intangible assets and research and development expenses, we observe that higher levels of ETF ownership are associated with lower levels of analyst coverage. Specifically we find that each eight percent increase in ETF ownership is associated with the loss of an additional analyst. These results are also strongest for firms with more than three percent of outstanding shares being held by ETFs. Overall, our results suggest that, after controlling for a host of other variables, increased ETF ownership has a negative effect on stock pricing informativeness and the corporate information environment. These findings contribute to a growing literature on the economic consequences of indexlinked products. The rapid increase in index-linked products in recent years has attracted the attention of investors, regulators, and financial researchers. 4 While the benefits of ETFs to investors are well understood, far less is known about other economic consequences they may bring to financial markets. A number of prior studies suggest that trading associated with the ETF-arbitrage mechanism can improve intraday price discovery for the underlying stocks (Hasbrouck 2003, Yu 2005, Chen and Strother 2008, Fang and Sanger 2012, and Ivanov et al. 2013). Other studies highlight concerns related to the pricing and trading of these instruments, including the more rapid transmission of liquidity shocks, and elevated return volatility, particularly in times of market stress (Wurgler 2010). Empirically, the rise of ETFs has been linked with higher return correlations (Da and Shive 2013, Sullivan and Xiong 2012), greater 4 Sullivan and Xiong (2012) note that while passively managed funds represent only about one-third of all fund assets, their average annual growth rate since the early 1990 s is 26 percent, double that of actively managed assets. Much of the increase in passively managed assets has been in the form of ETFs. According to Madhavan and Sobczyk (2014) as of June, 2014 there were 5,217 global ETFs representing $2.63 trillion in total net assets. 6

7 systemic risk (Ramaswamy 2011), and increased return volatility both for the component stocks and for the overall market (Broman 2013, Krause et al. 2013). Our analysis adds an informational perspective to this debate. We argue that the formation of ETFs has the effect of siphoning noise traders from component securities, and transferring much of this liquidity to the ETF instrument itself. As the ratio of noise traders to informed traders declines at the firm-level, the incentives for firm-level information acquisition will likewise decline, leading to deterioration in firms information environment. Our results are broadly consistent with the existence of such a link in the cross-section of U.S. equity stocks. These findings help to highlight a potentially serious economic consequence of ETFs. Our evidence also provides direct empirical evidence for a long-standing prediction from the rational expectations literature. According to this theory, when information is costly to acquire and process, market prices will only be informationally efficient to the extent that there are incentives for information acquisition. Using the emergence of ETFs, we link the siphoning of firm-level liquidity to a reduction in the incentive for information acquisition, and lower market pricing informativeness. Lee and So (2014) argue that the study of market efficiency involves the analysis of a joint equilibrium in which both these markets need to be cleared simultaneously. Our findings highlight the close relationship between the market for equities, the market for noise traders, and the market for information about these equities. The remainder of our study is organized as follows. In the next section, we provide some institutional details on ETFs. In Section 3, we develop our main hypotheses and outline our research design. Section 4 reports our empirical findings, and Section 5 concludes. II. Exchange-traded Funds (ETFs) 7

8 In the United States, ETFs are registered under the Investment Company Act of 1940 and are classified as open-ended funds or as unit investment trusts (UITs). Like open-end index funds, in a typical ETF, the underlying basket of securities is defined with the objective of mimicking the performance of a broad market index. But ETFs differ in some important respects from traditional open-ended funds. For example, unlike open-ended funds, which can only be bought or sold at the end of the trading day for their net asset value (NAV), ETFs can be traded throughout the day much like a closed-end fund. In addition, ETFs do not sell shares directly to investors. Instead, they only issue them in large blocks called creation units to authorized participants ( AP s) who effectively act as market-makers. Only the ETF manager and designated APs participate in the primary market for the creation/redemption of ETF shares. At the inception of the ETF, APs buy an appropriate basket of the predefined securities and deliver them to the ETF manager, in exchange for a number of ETF creation units. Investors can then buy or sell individual shares of the ETF from APs in the secondary market on an exchange. Shares of the ETF trade during the day in the secondary market at prices that can deviate from their net asset value (NAV), but the difference is kept in line through an arbitrage mechanism in the primary market. For example, when an ETF is trading at a premium to an AP s estimate of value, the AP may choose to deliver the creation basket of securities in exchange for ETF shares, which in turn it could elect to sell or keep. Notice that the creation/redemption mechanism in the ETF structure allows the number of shares outstanding in an ETF to expand or contract based on demand from investors. As Mahavan and Sobczyk (2014) observe, this creation-redemption mechanism means that liquidity can be accessed through primary market transactions in the underlying assets, beyond the visible secondary market. This additional element of liquidity means that trading costs of 8

9 ETFs are determined by the lower bound of execution costs in either the secondary or primary markets, a factor especially important for large investors. (p.3). In other words, unlike openend funds, traders interested in accessing the assets represented by the ETF can now choose to trade either in the secondary ETF market (buy/sell the ETF shares directly), or in the primary market (buy/sell the basket securities). An interesting corollary to the foregoing discussion is that ETFs are most likely to be successful when the underlying securities are relatively less liquid or difficult to borrow (thus creating an equilibrium demand for the ETF shares, with its lower trading costs). For example, the highly popular small-cap ETF, IWM, is based on the Russell 2000 index. While the underlying securities are typically less liquid (i.e. they represent the 2,000 stocks in the Russell Index that are below the largest 1,000), IWM itself is over $26 billion in size and trades at extremely low costs. We conjecture that particularly in these settings, uninformed ( noise ) traders will gravitate towards ETFs and away from the underlying stocks, with attendant consequences for the trading costs of the underlying basket of securities and the underlying component securities pricing informativeness. III. Hypothesis Development and Research Design The primary goal of this study is to investigate whether an increase in the proportion of firm shares held by ETFs is negatively associated with the quality of its information environment or its stock pricing informativeness. In addressing this question we identify three central dimensions of a firm s information environment that have bearings on its pricing informativeness: (1) transactions costs; (2) the extent to which stock prices reflect firm specific information, and (3) firm-specific information gathering activities. Accordingly, we test the effects of ETF ownership on each of these three dimensions. Based on the characteristics of ETF 9

10 securities and the Milgrom and Stokey (1982) no-trade theorem, we posit that ETF securities serve as attractive substitutes to the underlying securities for noise traders who will gravitate towards ETFs and away from the underlying stocks. As noise traders shift towards trading ETFs and away from trading the underlying securities directly, we expect to observe higher transactions costs for trading the underlying component securities (Mahavan and Sobczyk 2014). We further expect that these increased transactions costs will deter market participants from engaging in firm-specific information gathering activities and will lead to less informative prices (Grossman and Stiglitz 1980; Admati 1985). Based on the reasoning outlined above and building on predictions from noisy rational expectations models with costly information, we raise the following two hypotheses: H1: An increase in ETF ownership is associated with higher trading costs for the underlying security H2: An increase in ETF ownership is associated with a reduction in the extent to which the prices of underlying securities reflect firm specific information and the extent to which market participants gather firm-specific information. This manifests itself in three ways: A. Firms with higher levels of ETF ownership experience higher stock return synchronicity due to less firm-specific information being impounded in price. B. The returns of firms with higher levels of ETF ownership contain less information about future earnings. C. Firms with higher levels of ETF ownership experience lower levels of analyst coverage. 10

11 Our first hypothesis is that an increase in ETF ownership will be associated with higher trading costs (i.e., will have a negative effect on the liquidity of the underlying component security). To test this hypothesis, we estimate the following regression: = + + _ (1) + _ + Our dependent variable,, is the Corwin and Schultz (2012) monthly high-low measure of bid-ask spread for firm i in the first month of quarter t. We use this measure as a proxy for trading costs because it is much less time and data-intensive to calculate than intraday bid-ask spread measures, and Corwin and Schultz (2012) demonstrate that it outperforms the Roll (1984), Lesmond et al. (1999), and Holden (2009) techniques for measuring spreads. Our main variable of interest,, is the percentage of firm i s shares held by ETFs as of the end of quarter t-1. ETF ownership is highly correlated with overall institutional ownership, and prior research suggests there might be a relation between institutional ownership and bid-ask spreads. 5 To isolate the effect of ETF ownership on stock liquidity and make sure our results are not confounded by the relation of ETF ownership to institutional ownership, we adopt two approaches. The first approach involves the inclusion of (defined as the percentage of firm i s shares held by all institutions as of the end of quarter t-1) directly in Eq. (1) as an additional control variable. However, given the high correlation between ETF and INST, 5 Prior research on the relation between bid-ask spreads and institutional ownership is mixed. Glosten and Harris (1988) suggest that higher levels of concentrated institutional ownership will increase bid-ask spreads, while higher levels of dispersed institutional ownership might encourage competition that reduces bid-ask spreads. Consistent with this notion, Agarwal (2011) documents a non-monotonic (Ushaped) relation between spreads and institutional ownership. 11

12 estimations involving both variables potentially suffer from multicollinearity. To address this concern, we generate an estimate of ETF ownership that is orthogonal to the magnitude of institutional ownership. This new variable, ETF_ORTH, is defined as the residual ( ) from the following regression, estimated over the pooled sample: = + + (2) By construction, ETF_ORTH captures the dimension of ETF ownership that is uncorrelated with INST. As an alternative to including INST as an additional control variable in Eq. (1), we also use ETF_ORTH alone as a measure of ETF ownership that already accounts for the firm s level of institutional ownership. In Eq. (1), represents a number of control variables. One of them is the log of market value of equity (LN(MVE)) as of the end of quarter t-1, because we expect larger firms to exhibit smaller bid-ask spreads. Copeland and Galai (1983) demonstrate that bid-ask spreads increase with the return volatility and decrease with the share turnover of the underlying security. Correspondingly we include as controls the annualized standard deviation of daily returns during quarter t-1 (STD(RET)) and average share turnover during quarter t-1 (TURN). To control for the effects of financial distress or growth opportunities, we also include book to market ratio (BTM) at the end of quarter t-1 (Fama and French 1992, Lakonishok et al. 1994). Finally, to control for time and industry trends in bid-ask spreads, we include industry and year-quarter fixed effects in our estimation of Eq. (1). Our first hypothesis predicts that the coefficient will be positive, indicating that, ceteris paribus, firms with higher levels of ETF ownership experience higher bid-ask spreads. Our second hypothesis is that price informativeness of a firm s equity decreases as ETF ownership increases. This hypothesis involves three predictions, each of which we test 12

13 separately. First, we examine the relation between ETF ownership and stock return synchronicity, a measure of the extent to which firm-specific return variation is explained by variation in market wide and related-industry returns variation. Roll (1988) posits that greater levels of firm-specific information being impounded into price drive low levels of return synchronicity. Wurgler (2000), Durnev et al. (2003), DeFond and Hung (2004), and Durnev et al (2004) provide support for this hypothesis in a variety of settings. Following these studies, we expect stock return synchronicity to be inversely related to the amount of firm specific information embedded in stock price. Since we hypothesize that equity price informativeness decreases as ETF ownership increases, we expect stock return synchronicity and ETF ownership to be positively associated. We follow the methodology outlined by Durnev (2003) to estimate firm specific stock return synchronicity ( ). First, we obtain the adjusted coefficient of determination (adjusted ) from a two-factor return model using the following equation: = (3) In Eq. (3), RET it is firm i s return on day k, MKTRET k is the value-weighted market return on day k, and INDRET k is the return of firm i s industry (defined using the Fama-French 48 classifications) on day k. 6 Eq. (3) is estimated separately for each firm-quarter, using daily returns for firm i over the trading days in quarter t. We only estimate Eq.(3) for firm-quarters in which there are at least 30 days of data available. We then calculate as the logarithmic transformation of the adjusted from Eq. (3) to create an unbounded continuous measure of 6 We adopt this model of returns to measure firm specific adjusted R 2 (and, consequently, synchronicity) because it is the most frequently used in the literature (e.g., Piotroksi and Roulstone 2004, Hutton et al 2010, Chan and Chan 2014). To ensure our inferences are not affected by the method chosen to estimate firm specific adjusted R 2 we also estimate synchronicity using the measures outlined in Crawford et al (2012) and Li et al. (2014). Our inferences are unchanged by these alternate measurement techniques. 13

14 synchronicity (Morck 2000, Crawford et al. 2012, Hutton et al. 2010) 7 : =. High values of this synchronicity measure indicate that a greater fraction of variation in firm s stock returns are explained by the variation in market and related-industry returns. Using this synchronicity measure, we estimate the following panel regression to test the first prediction of our second hypothesis: = + + _ (4) + _ ++ remains as defined in Eq. (1) above. Our second hypothesis predicts that the coefficient will be positive, indicating that, ceteris paribus, firms with higher levels of ETF ownership experience more synchronous stock returns. In estimating Eq. (4), we include several control variables ( ) that prior research suggests are associated with stock return synchronicity. Piotroksi and Roulstone (2004) find a positive relation between institutional holding and stock return synchronicity. To address this, we control for institutional holding in the two ways discussed previously: using INST directly and also using ETF_ORTH. Following Jin and Myers (2006) we control for the skewness of firm i s returns over quarter t-1 (SKEW) Since Li et al. (2014) show that synchronicity is often confounded with systematic risk, we include CAPM beta as a control for systematic risk. As additional controls, we include the log of market value of equity as of the end of quarter t-1 (LN(MVE)), book-to-market ratio as of the end of quarter t-1 (BTM), average share turnover during quarter t-1 (TURN), and year-quarter and 7 In computing Synch it we exclusively use adjusted values. Following Crawford et al. (2012), we truncate the sample of adjusted values at

15 industry fixed effects The second prediction of H2 concerns the relation between ETF ownership and the extent to which stock prices or current returns reflect firm-specific future earnings. To test this prediction we follow Kothari and Sloan (1992) and Collins et al. (1994) and estimate the following equation: = (5) _ + _ + In Eq. (5), we regress current period fiscal-year returns ( ) on prior period earnings, ( ) contemporaneous earnings ( ), and future earnings ( ). We define earnings as net income before extraordinary items scaled by lagged market value of equity. The coefficient measures the relation between current returns and future earnings; prior research terms this coefficient the future earnings response coefficient (FERC) and offers it as a measure of the extent to which current returns reflect/predict future earnings (Ettredge et al. 2005). To address our research question we include as explanatory variables the level of ETF ownership ( ) at the end of year t - 1 as well as the interaction between the level of ETF ownership and past, current, and future earnings ( ± ). Our hypothesis predicts that the coefficient on the interaction of ETF ownership with future earnings will be negative, indicating that FERCs are lower for firms with higher ETF ownership. As in previous regressions, represents a number of control variables as suggested by prior research. We control for the effect of institutional ownership at the end of 15

16 year t-1 in two ways to distinguish between the impact of ETF ownership on FERCs and that of institutional ownership: by including INST directly as a control and by introducing and orthogonalized measure of ETF ownership (ETF_ORTH). Following Collins et al. (1994), we include future returns ( ) as an additional explanatory variable to address the measurement error induced by using actual future earnings as a proxy for expected future earnings. We control for the growth in assets from year t-1 to year t+1 (ATGROWTH), to control for the effect of a firm s growth on the ability of its stock returns to reflect future earnings. Because firms experiencing losses are expected to have lower FERCs we control for this effect by including an indicator (LOSS t+1 ) that equals one if the firm experienced a loss in year t+1 ( <0). We control for the natural logarithm of market value of equity at the end of year t as prior research indicates that larger firms have greater FERCs. Finally, we include year and industry fixed effects to capture time and industry unobserved factors that might impact the return-earnings relation. Our second hypothesis also posits that high levels of ETF ownership deter market participants from gathering firm specific information. Consequently, we expect that firms with higher levels of ETF ownership will experience lower analyst coverage. To test this hypothesis, we estimate the following equation = + (6) + _ + _ + In Eq. (6), is the number of unique analysts on I/B/E/S providing forecasts of firm i s quarter t earnings. We use this measure as a natural proxy for the extent to which market participants gather firm-specific information. As before, represents a number of 16

17 control variables suggested by prior literature. Barth et al. (2001) demonstrate that firms with large research and development expenses or intangible assets experience greater analyst coverage. To control for this effect, we include the proportion of research and development expenses relative to total operating expenses (RD_F) and the proportion of intangible assets relative to total assets (INTAN_F) as controls. Eq. (6) also includes controls for size and book to market ratio. We use the natural logarithm of market value of equity as a proxy for size as it allows us to capture potential nonlinearities between size and analyst coverage. Following Lang and Lundholm (1996), we also control for earnings volatility (STD(EARN), the standard deviation of earnings before extraordinary items over the prior 6 quarters) and return volatility (STD(RET)). Our second hypothesis predicts analyst coverage declines with increasing ETF ownership, so we predict the coefficient will be negative. Appendix A provides detailed descriptions of all the variables used in our empirical tests. To control for potential time-series correlation between firm-specific measures of ETF ownership as well as other variables, we base our inferences from Eq. (2), (4), and (6) on t- statistics calculated based on standard errors clustered by firm. Because in Eq. (5) the dependent variable is annual future returns, we base our inferences from this equation on t-statistics calculated based on standard errors clustered by year (Gow et al 2010). IV. Empirical Analyses a. Data and sample selection We determine ETF ownership by first using CRSP, Compustat, and OptionMetrics to identify all ETFs traded on the major U.S. exchanges. For each ETF, we obtain quarterly equity holdings from the Thomson Financial S12 database. Then, for every stock being held, we define ETF as the total number of shares held by any ETF divided by total shares outstanding in that 17

18 quarter. We repeat this process for every firm quarter between 2000 and 2011 to construct our quarterly panel. Our sample begins in 2000 because it is the first year with sufficient variation in ETF ownership to conduct our analyses. Our sample ends in 2011 due to data availability constraints. All firm-quarters not held by any ETF in the sample period are included in the sample with ETF = 0. Figure 1 reports the average ETF ownership across firms for each year of our sample. The figure reveals a significant increase in average ETF ownership over our sample period, from 0.2% in 2000 to 3.6% in We obtain market-related data on all US-listed firms from CRSP and accounting data from Compustat. To be included in our sample, each firm-quarter observation must have information on stock price, shares outstanding, and book value of equity. We also require sufficient data to calculate the standard deviation of daily returns and average share turnover within each firm quarter. We restrict our analyses to firms with non-negative book to market ratios in every quarter of our sample period. This results in a sample of 130,930 firm-quarters and 6,089 unique firms. In some of our analyses, we also require data on firm-level return synchronicity and analyst coverage. In the synchronicity (analyst coverage) analyses, our sample size is reduced to 106,248 (75,729) firm-quarter observations. Panel A of Table 1 presents descriptive statistics for the main variables used in the analyses. Of particular interest for our analyses is the level of ETF holding, measured as a percentage of total shares outstanding. The mean (median) percentage ETF ownership is 1.64 (0.80) percent. This is substantially lower than the level of institutional ownership, which has a mean (median) of (50.65) percent. The distributional statistics of both ETF and institutional ownership in our sample are consistent with prior research (Hamm 2013, Jiambalvo 18

19 2002). Panel A also reveals that the average share price of our sample firms is $21.90, mitigating concerns that our sample is constructed of penny stock firms. Panel B of Table 1 presents Pearson and Spearman correlation coefficients between the key variables in our empirical analysis. In our sample, ETF exhibits a significant positive Pearson (Spearman) correlation of (0.454) with INST. This is consistent with prior research documenting the strong relation between ETF and institutional ownership (Hamm 2013). ETF is also positively correlated with firm size (Pearson coefficient = 0.391) and turnover (Pearson coefficient = 0.349). Panel B reveals that ETF is negatively correlated (Pearson coefficient = ) with our central measure of trading costs, HLSPREAD, although it is important to note that this unconditional correlation does not control for the impact of related factors such as firm size. b. ETF ownership and transactions costs Table 2 presents the summary statistics from the estimation of Eq. (1), which is designed to test our first hypothesis. Column 1 reveals that the measure of bid-ask spread, HLSPREAD, exhibits the expected relations with our control variables. HLSPREAD is decreasing in firm size measured as the natural logarithm of market value of equity (coefficient = , t-statistic = ) and share turnover (coefficient = , t-statistic = -6.76). In addition, Column 1 of Table 2 reveals that the level of ETF ownership is significantly positively related to bid-ask spreads (coefficient = 0.014, t-statistic = 2.79). In column 2 we observe that this positive relation not only persists but is slightly strengthened after controlling for the level of institutional ownership (coefficient = 0.015, t-statistic = 4.10). The coefficient on institutional ownership itself is negative ( ) but indistinguishable from zero (t-statistic = -0.29). 19

20 In Column 3 of Table 2 we focus our analysis on firms with particularly high levels of ETF ownership. We define and include in our estimation a new indicator variable ETF_HI which equals one for a particular firm quarter if the level of ETF ownership for that firm quarter exceeds three percent. The results in Column 3 of Table 2 reveal that the positive relation between ETF ownership and bid-ask spreads is concentrated among firms with more than three percent ETF ownership, as the coefficient on ETF_HI is positive (0.143) and significant (tstatistic = 11.23) while the coefficient on ETF is negative (-0.006) and insignificant (t-statistic =- 1.63). Columns 2 and 3 of Table 2 present estimations in which both ETF and INST are included as independent variables. We re-estimate columns 2 and 3 of Table 2 using ETF_ORTH instead of ETF and INST and present the regression summary statistics in columns 4 and 5, respectively. Column 4 reveals that the coefficient on ETF_ORTH (0.014) is positive and highly significant (t-statistic = 4.62). Column 5 confirms that this effect is concentrated among firms with high levels of ETF ownership, as the coefficient on ETF_HI (0.130) is significantly positive (t-statistic = 4.62) when controlling for ETF_ORTH. Overall, the evidence presented in Table 2 indicates that, after controlling for the effect of institutional ownership and a host of other variables, firms with high ETF ownership experience higher bid-ask spreads. This result supports our hypothesis that firms with high ETF ownership experience greater transactions costs. c. ETF ownership and price informativeness Table 3 presents the summary statistics from the estimation of Eq. (2), which is designed to examine the relation of ETF ownership to stock return synchronicity. Column 1 reveals that, consistent with prior research, our measure of synchronicity (SYNCH) is increasing in firm size 20

21 measured as the natural logarithm of market value of equity (coefficient = 0.504, t-statistic = 61.86) and decreasing in share turnover (coefficient = , t-statistic = -2.15). Our measure of systematic risk, BETA, also exhibits a significantly positive coefficient (0.629, t-statistic = 43.73), supporting the findings of Li et al (2014) that synchronicity and systematic risk coincide. Column 2 of Table 3 reveals that the level of ETF ownership is significantly positively related to synchronicity (coefficient = 0.278, t-statistic = 30.48) after controlling for the level of institutional ownership, which also exhibits a positive relation with synchronicity (coefficient = 0.004, t-statistic = 7.98). In Column 3 of Table 3 we again introduce the indicator variable ETF_HI and find that it carries a significant positive coefficient (0.347, t-statistic = 8.52). This indicates that positive relation between ETF ownership and synchronicity is particularly concentrated in firms with high levels of ETF ownership. As before, to assuage concerns about multicollinearity between ETF and INST, we also examine how ETF_ORTH relates to stock return synchronicity in columns 4 and 5. Column 4 reveals that the coefficient on ETF_ORTH (0.262) is positive and highly significant (t-statistic = 29.22). Column 5 confirms that this result is stronger for firms with high levels of ETF ownership. The coefficient on ETF_HI (0.734) is significantly positive (t-statistic = 20.28) while the coefficient ETF_ORTH declines from to 0.163, though remaining significantly positive (t-statistic =16.61). Overall, the results presented in Table 3 indicate that firms with high ETF ownership experience higher stock return synchronicity. This relation is robust to the inclusion of controls for systematic risk, firm size, book-to-market ratio, turnover, return skewness, and institutional ownership, as well as industry and time effects. To the extent that higher return synchronicity reflects lower levels of firm-specific information 21

22 being impounded in returns, this result supports our hypothesis that firms with high ETF ownership experience reduced pricing informativeness. To examine the economic significance of our findings, in Figure 3 we plot the inferred value of adjusted R 2 as a function of ETF ownership using the parameter estimates from Column 2 of Table 3. We calculate inferred adjusted R 2 by inverting our expression for SYNCH and using the predicted values of SYNCH from our estimation of Eq. (2): = 1+ The advantage of this approach for analyzing the relation between ETF ownership and adjusted R 2 is that it allows ETF ownership to vary while holding all other explanatory variables constant at their median values. Figure 3 reveals that inferred adjusted R 2 and ETF ownership have a convex relationship. The curve steepens as ETF ownership increases, consistent with the relation between ETF ownership and adjusted R 2 being strongest for firms with the highest levels of ETF ownership adjusted R 2 more than quadruples, from.02 to.09, from the bottom to the top of the ETF ownership distribution. We also test our hypothesis regarding ETF ownership and price informativeness by examining the coefficient estimate on the interaction between ETF ownership and future earnings. The results of this analysis are presented in Table 4. As the dependent variable in table 4 is annual future returns, we present t-statistics from standard errors clustered by year (Gow et al 2010). Consistent with prior literature, we observe a significantly positive future earnings response coefficient across columns 1 to 5 (0.575 to 0.566, t-statistics = 4.61 to 4.54). Columns 1 and 2 also reveal that the interaction of future earnings with ETF ownership carries a significantly negative coefficient ( and , t-statistics = and -2.73). This 22

23 suggests that FERCs are lower for firms with higher levels of ETF ownership. We also find, consistent with Jiambalvo et al. (2002), that the coefficient on the interaction of INST with future earnings is positive (.001). However this coefficient is not statistically different from zero (tstatistic = 0.38) To further explore this result, we include the indicator ETF_HI and its interactions with past, current, and future earnings in the estimation summarized in column 3 of Table 4. We observe a negative and significant coefficient on the interaction of ETF_HI with future earnings, reinforcing our inference that FERCs are lower for firms with higher levels of ETF ownership. We also estimate the regressions summarized in columns 2 and 3 using ETF_ORTH in place of ETF and INST. We find that the interaction between ETF_ORTH and future earnings carries a significantly negative coefficient (-0.035, t-statistic = -2.58). The coefficient on ETF_HI in column 5 is also significantly negative (-0.163, t-statistic = -2.61) when controlling for ETF_ORTH. Taken together, the results presented in Table 4 indicate that firms with high ETF ownership experience a weaker association between current returns and future earnings. This supports our hypothesis that the stock prices of firms with high ETF ownership are impounding less firm-specific information. d. ETF ownership and information gathering Our second hypothesis posits that the higher transactions costs associated with ETF ownership costs will reduce the incentives for market participants to engage in costly information acquisition activities. As a result, we predict that firms with high levels of ETF ownership will experience reduced levels of analyst coverage. We test this hypothesis by estimating Eq. (4). The summary statistics from this estimation are presented in Table 5. Consistent with Barth et al. (2001) we observe a significant positive coefficient (2.156 to 2.169, 23

24 t-statistic = 5.52 to 5.57) on RD_F across all specifications. We also find a positive coefficient on STD(RET), indicating that analyst coverage is higher for firms with higher return volatility. Consistent with our predictions, ETF exhibits a significant negative coefficient across all specifications. In column 2, where we control for INST as well, the coefficient on ETF is (t-statistic = -5.72). The magnitude of this coefficient suggests that for each percentage point increase in ETF ownership, analyst coverage (defined as number of analysts) declines by.16. Alternatively, every 6.25 percent increase in ETF ownership would be associated with a firm having one less analyst. This is the opposite of the relationship we observe between institutional ownership and analyst coverage. INST exhibits a positive coefficient (0.009) that is also significantly different from zero (t-statistic = 3.50), suggesting that firms with higher levels of institutional ownership have more analysts covering them. We again observe that the relation between ETF ownership and analyst coverage is concentrated amongst firms with high levels of ETF ownership; column 3 of Table 5 reveals that the coefficient on ETF_HI is negative (-0.612) and highly significant (t-statistic = -6.56). When ETF_HI is included the coefficient on ETF is almost halved, from in column 2 to in column 3, though it remains statistically different from zero (t-statistic = -2.43). The results are also robust to the alternate specification using ETF_ORTH rather than ETF and INST in simultaneity. The coefficient on ETF_ORTH is negative (-0.164) and significant (t-statistic =- 5.89) and declines less dramatically than ETF when we include ETF_HI as an additional control variable. Overall, the results in table 5 suggest that ETF ownership and analyst coverage are negatively related; when a large percentage of a firm s shares outstanding are held by ETFs, they are less likely to receive analyst coverage. This supports our hypothesis that high levels of ETF ownership dampen information-gathering activities from market participants such as analysts. 24

25 V. Conclusion Our paper examines the linkage between the market for firm-specific information and the market for individual securities and uninformed traders. We assume the market consists of agents who expend resources to become informed and noise traders against whom these informed agents earn a return on their information acquisition efforts. As informed traders trade against noise traders, information is incorporated into prices. The availability of uninformed traders is essential to the informativeness of price in these models; without uninformed traders, informed traders lose their incentives to engage in costly information gathering activities and consequently prices will cease to impound information. ETFs offer a unique setting in which to examine the implications of traditional noisy rational expectations models. The market for ETFs has grown dramatically in the past decade and ETFs now constitute an attractive investment alternative, especially for those interested in diversified strategies. By focusing on the natural growth of exchange-traded funds over the past decade, we study how changes in the composition of a firm s investor base impacts its share price informativeness. Specifically, we hypothesize and find that increased ETF ownership is accompanied by a decline in the pricing informativeness of the underlying component securities. We first demonstrate that an increase in ETF ownership is associated with an increase in firms bid-ask spreads. This is consistent with the idea of uninformed traders exiting the market of the underlying security in favor of the ETF. As uninformed traders exit and bid-ask spreads rise, we posit that price informativeness will decline. Consistent with this prediction, we find that higher levels of ETF ownership are associated with stronger price synchronicity and a reduction in future earnings response coefficients. We also observe a negative association between ETF ownership and the number of 25

26 analysts covering the firm. Collectively, the evidence presented in this paper suggests increased ETF ownership can lead to weaker information environments for firms. 26

27 References Agarwal, P Institutional Ownership and Stock Liquidity. Working paper. Admati, A. R A noisy rational expectations equilibrium for multi-asset securities markets. Econometrica 53: Barth, M.E., R. Kasznik, and M. McNichols Analyst Coverage and Intangible Assets. Journal of Accounting Research 39: Ben-David, I., F. Franzoni, and R. Moussawi Do ETFs Increase Volatility? Working Paper. Ohio State University, University of Lugano and the Swiss Finance Institute, and University of Pennsylvania, January. Broman, M. S Excess Co-movement and Limits-to-Arbitrage: Evidence from Exchange- Traded Funds. Working paper, York University. Chan, K. and Y.C. Chan SEO Underpricing, Stock Price Synchronicity and Analyst Coverage. Journal of Financial Economics, forthcoming. Chen, G. and T. S. Strother On the contribution of index exchange traded funds to price discovery in the presence of price limits without short selling. Working paper. University of Otago, New Zealand. February. Collins, D., S. Kothari, J. Shanken, and R. Sloan Lack of Timeliness versus Noise as Explanations for Low Contemporaneous Return-Earnings Association. Journal of Accounting & Economics 18: Copeland, T and D. Galai Information Effects on the Bid-Ask Spread. The Journal of Finance 38: Corwin, S. and P. Schultz A Simple Way to Estimate Bid-Ask Spreads from Daily High and Low Prices. The Journal of Finance 67: Crawford, S. S., D. T. Roulstone, and E. C. So, Analyst Initiations of Coverage and Stock Return Synchronicity. The Accounting Review 87: Da, Z. and Shive, S Exchange-Traded Funds and Equity Return Correlations. Working paper, University of Notre Dame. DeFond, M. and M. Hung, Investor protection and corporate governance: Evidence from worldwide CEO turnover. Journal of Accounting Research 42: Diamond, D. W., and R. E. Verrecchia, 1981, Information aggregation in noisy rational expectations model. Journal of Financial Economics 9: Durnev, A., R. Morck, and B. Yeung Value-enhancing capital budgeting and firm-specific stock return variation. The Journal of Finance 59: