«Abstract Using a comprehensive sample of trades by Schedule 13D ﬁlers, who possess valuable private information when they accumulate stocks of ...»
Do prices reveal the presence of informed trading?$
Pierre Collin-Dufresne1, Vyacheslav Fos2
First Version: November 2011
This Version: July 22, 2013
Using a comprehensive sample of trades by Schedule 13D ﬁlers, who possess valuable
private information when they accumulate stocks of targeted companies, this paper
studies whether several measures of adverse selection reveal the presence of informed
trading. The evidence suggests that on days when Schedule 13D ﬁlers accumulate shares, both high-frequency and low-frequency measures of stock liquidity and adverse selection indicate higher stock liquidity and lower adverse selection, even though prices
are positively aﬀected. We document three channels that help explain this phenomenon:
(a) informed traders select times of higher liquidity when they trade, (b) liquidity increases in response to informed traders’ trades, (c) informed traders use limit orders.
Keywords: Informed Trading, Liquidity, Adverse Selection, Activist Shareholders $ We thank Azi Ben-Rephael, Terry Hendershott, Gur Huberman, Wei Jiang, Robert Korajczyk (AFA discussant), Norman Schuerhoﬀ, and, especially, Yakov Amihud and Larry Glosten for many helpful comments. We also thank seminar participants at the University of Illinois at Urbana-Champaign, Copenhagen Business School, Tsinghua University, Columbia University, and participants at the AFA 2013 Annual Meeting and IDC Summer Conference for their helpful comments and suggestions. Virginia Jiang, Xinran Li, Urvi Maru, Hana Na, Shan Qiao, Soﬁya Teplitskaya, and Tong Tong provided excellent research assistance.
Email addresses: firstname.lastname@example.org (Pierre Collin-Dufresne), email@example.com (Vyacheslav Fos) 1 Carson Family Professor of Finance, Columbia University, EPFL & SFI, and NBER 2 College of Business, University of Illinois at Urbana-Champaign Introduction An extensive body of theory suggests that stock liquidity, as measured by the spread between the bid and ask quotes and by the price impact of trades, should be informative about the magnitude of asymmetric information between market participants (Copeland and Galai, 1983; Glosten and Milgrom, 1985; Kyle, 1985; Easley and O’Hara, 1987;
Admati and Pﬂeiderer, 1988). For example, in his seminal contribution, Kyle (1985) shows how an informed trader hides his private information and optimally accumulates shares at a rate inversely proportional to his price impact,3 Kyle’s lambda, which measures the dollar change in price due to a dollar change in order ﬂow. The theory predicts that Kyle’s lambda, which can be estimated from a regression of price change on order ﬂow, should be higher for stocks with more informed trading (relative to noise trading).
Following that literature there have been many attempts to measure trading costs empirically, and to decompose such costs into diﬀerent components such as a adverse selection, order processing cost, and inventory cost (e.g., early papers include Glosten, 1987; Glosten and Harris, 1988; Stoll, 1989; Hasbrouck, 1991a; Amihud, 2002). An extensive empirical literature relies on adverse selection measures assuming they capture information asymmetry (e.g., Barclay and Hendershott, 2004; Vega, 2006; Duarte et al., 2008; Bharath et al., 2009; Kelly and Ljungqvist, 2011). Most of these measures rely on some empirical estimate of price impact and its persistence to identify the amount of private information in trades. While there is an obvious endogeneity issue with this approach (since we do not actually know if the trades are informed), it is natural to think that such price impact measures correlate positively with the informational motivation of trades. For example, in their well-known survey of the micro-structure literature, Biais, Glosten, and Spatt (2005) describe the empirical relation between adverse selection and 3 The informed trader’s optimal trading strategy is to trade as a linear function of the diﬀerence between his signal and the current price, at a rate inversely proportional to his price impact, and that increases as maturity approaches so that all the information eventually makes it into prices.
Unconditionally, the expected trading rate of the informed trader is constant (in his own ﬁltration).
2 eﬀective spread, denoted by a price impact measure λ, as follows: “As the informational motivation of trades becomes relatively more important, λ goes up.” (page 232).
But, do these empirical measures of adverse selection actually capture information asymmetry?
To test this one would want to separate informed from uninformed trades ex-ante and measure their relative impact on price changes. Unfortunately, since we generally do not know the traders’ information sets, this is hard to do in practice. As a result it is often assumed that some types of investors are informed. For example, Boulatov et al.
(2009) use the institutional order ﬂow as a proxy for informed trading.
In this paper we use a novel data set of trades by investors we can identify as having substantial private information to study whether proposed liquidity measures reveal the presence of informed trading. Speciﬁcally, we exploit a disclosure requirement to identify trades that rely on valuable private information. Rule 13d-1(a) of the 1934 Securities Exchange Act requires investors to ﬁle with the SEC within 10 days of acquiring more than 5% of any class of securities of a publicly traded company if they have an interest in inﬂuencing the management of the company. In particular, Item 5(c) of Schedule 13D requires the ﬁler to “... describe any transactions in the class of securities reported on that were eﬀected during the past sixty days or since the most recent ﬁling of Schedule 13D, whichever is less.” Thus, Schedule 13D ﬁlings reveal the date and price at which all trades by the Schedule 13D ﬁler were executed during the 60 days that precede the ﬁling date.4 We hand collect a comprehensive sample of trades from the Schedule 13D ﬁlings. We view this sample as an interesting laboratory to study the liquidity and the price impact of informed trades. First, an average Schedule 13D ﬁling in our sample is characterized by a positive and signiﬁcant market reaction upon announcement. For example, the cumulative return in excess of the market is about 6% in the (t-10,t+1) window around 4 As we explain in Section 3, our sample includes original Schedule 13D ﬁlings only, i.e., amendments to previously submitted ﬁlings are excluded from the sample.
3 the ﬁling date and about 3% in the (t-1,t+1) window around the ﬁling date. Second, we calculate several measures of proﬁts made by Schedule 13D ﬁlers and show that these proﬁts are economically signiﬁcant. For example, an average Schedule 13D ﬁler gains $0.8 million on a $22 million stake in a $293 million market cap company. To summarize, the evidence implies that Schedule 13D ﬁlers’ information is valuable. Therefore, we can classify the pre-announcement trades by Schedule 13D ﬁlers as informed trades. It is also important to realize that, by its very nature, the information held by Schedule 13D ﬁlers is likely to qualify as ‘private information.’5 Our main empirical result is that standard measures of adverse selection and stock liquidity do not reveal the presence of informed traders. Speciﬁcally, we ﬁnd that several measures of adverse selection are lower on days on which Schedule 13D ﬁlers trade, which suggests that adverse selection is lower and the stock is more liquid when there is signiﬁcant informed trading in that stock. For example, on an average day when Schedule 13D ﬁlers trade, the measured price impact is almost 30% lower relative to the sample average. Both high- and low-frequency measures suggest that liquidity is higher when Schedule 13D ﬁlers trade. For example, Amihud’s (low-frequency) illiquidity measure decreases by almost 10% when Schedule 13D ﬁlers increase their position. Importantly, we show that days when Schedule 13D ﬁlers trade are characterized by positive and signiﬁcant market-adjusted returns, which suggests that informed trades do impact prices. Liquidity measures, however, fail to detect that price impact.
To summarize, the evidence constitutes a serious challenge to the argument that standard measures of stock price liquidity, and in particular of the adverse selection component, capture the presence of informed trading (at least not trading based on the long-lived type of information that Schedule 13D ﬁlers hold).
5 Collin-Dufresne and Fos (2013) develop a theoretical model in which activist shareholders can expend eﬀort and change ﬁrm value. In that model the market price depends on the market maker’s estimate of the activist’s share-ownership, since the latter determines the eﬀort level of the informed trader, and hence the liquidation value of the ﬁrm. This model shows that a signiﬁcant part of the valuable private information pertains to the activist’s own holdings, which by deﬁnition is information known only to him.
4 We consider three possible mechanisms that could explain this result.
First, and consistent with the the theoretical model presented in Collin-Dufresne and Fos (2012), Schedule 13D ﬁlers might select the time at which they trade and step in when the market and/or the target stock happen to be liquid.
Second, Schedule 13D ﬁlers might attract additional uninformed volume. In this case, informed traders also trade when the stock is more liquid. But the diﬀerence is that the informed trades are causing the increase in liquidity.
Third, standard liquidity measures are based on models that assume that informed traders mostly demand immediacy, i.e., use market orders. Schedule 13D ﬁlers, however, possess relatively long-lived information and therefore might place limit orders instead (e.g. Kaniel and Liu, 2006). Thus, informed investors with long-lived information might improve stock liquidity (and receive the spread rather than pay it).
We perform several tests that indicate that all three mechanisms contribute to our ﬁndings.
For a sub-sample of trades where we can identify the individual ‘time-stamped’ trades, we ﬁnd clear evidence that Schedule 13D use limit orders. Further, we ﬁnd that before their ownership crosses the 5% threshold Schedule 13D are more likely to use limit orders than after, when they have only ten days left to trade.
We also construct a proxy for usage of market orders based on the average Schedule 13D trader’s buy price relative to the VWAP and show that when Schedule 13D ﬁlers are more likely to use market orders the impact of their trading on measures of adverse selection measures is less negative.
However, using two placebo tests that exploit reforms implemented by NASDAQ and NYSE, we show that the limit order mechanism cannot be the sole explanation. Indeed, we ﬁnd that even in samples where limit orders were not (or less) available to informed traders, there is no signiﬁcant positive relation between informed trades and liquidity measures.
Instead, there is very strong statistical evidence that the pattern in abnormal volume 5 observed on (and around) days when insiders trade is not random (both comparing the target ﬁrms’ abnormal volume to its own past history or to a matched sample of ﬁrms). This clearly shows that either 13D ﬁlers select the days when they trade based on available liquidity or that their trades generate abnormal patterns in the stock’s liquidity.
Consistent with the ‘selection mechanism,’ we show that Schedule 13D ﬁlers trade more aggressively not only when the stock they are purchasing is more liquid, but also when market-wide conditions change. For example, a high aggregate volume and a low market return positively aﬀect the likelihood of a trade by Schedule 13D ﬁlers on a given day.
Overall, we conclude that Schedule 13D ﬁlers are likely (a) to trade when stock liquidity is high for exogenous as well as for endogenous reasons, and (b) to use limit orders, which leads to an inverse relation between standard empirical measures of adverse selection and the informational motivation of trades.
The rest of the paper is organized as follows. Section 1 discusses related literature.
Section 2 provides an overview of the institutional background. Section 3 describes the data. The magnitude of information asymmetry is analyzed in Section 4. Section 5 shows that when Schedule 13D ﬁlers trade, stock prices increase. Section 6 describes liquidity measures used in the analysis. Section 7 presents the main evidence on the eﬀect of informed trading on liquidity measures. Section 8 studies mechanisms that are consistent with the inverse relation between adverse selection measures and informed trading. Finally, Section 9 concludes.
1. Related Literature
This paper is related to several strands of literatures.
First, this paper contributes to the empirical literature that relies on liquidity measures as a proxy for information asymmetry (e.g, Barclay and Hendershott, 2004;
Vega, 2006; Duarte et al., 2008; Bharath et al., 2009; Kelly and Ljungqvist, 2011). Our 6 evidence suggests that empirical measures of information asymmetry might not reveal the presence of informed traders. Therefore, empirical researchers should be cautious when relying on a liquidity measure as a proxy for information asymmetry.
Second, our paper is related to the large literature on the estimation of the asymmetric information component of transaction costs (e.g., Easley and O’Hara, 1987;
Glosten and Harris, 1988; Stoll, 1989; Hasbrouck, 1991a; Lin et al., 1995). In contrast to this literature, our paper does not rely on time-series properties of stock prices to identify informed trades, but uses well-identiﬁed trades executed by informed traders to study the impact of asymmetric information on stock price liquidity measures.
Third, our paper is related to the empirical literature that studies the impact of informed trading on stock liquidity. One strand of this literature studies the impact of share repurchases on stock liquidity and ﬁnds mixed results (Barclay and Smith, 1988;