Brokers

and Order Flow Leakage:

Evidence from Fire Sales1
Andrea Barbon Marco Di Maggio Francesco Franzoni Augustin Landier

October 2018

Abstract

Using trade-level data, we study whether brokers play a role in spreading order flow information in the
stock market. We focus on large portfolio liquidations, resulting in temporary price drops and identify
the brokers that intermediate these trades. These brokers’ clients are more likely to predate on the
liquidating funds than to provide liquidity. Predation leads to profits of about 25 basis points over ten
days and increases the liquidation costs for the distressed fund by 40%. This evidence suggests a role
of information leakage in exacerbating fire sales.

1
Barbon: USI Lugano and Swiss Finance Institute (email: andrea.barbon@usi.ch ); Di Maggio: Harvard Business School and
NBER (email: mdimaggio@hbs.edu); Franzoni: USI Lugano and Swiss Finance Institute (email: francesco.franzoni@usi.ch);
Landier: Toulouse School of Economics (email: augustin.landier@tse-fr.eu). We thank Malcolm Baker, John Campbell,
Laurent Frésard, Slava Fos (discussant), Joel Hasbrouck, Terry Hendershott (discussant), Gary Gorton, Owen Lamont, Jongsub
Lee (discussant), Andrew Lo, Toby Moskowitz, Abhiroop Mukherjee (discussant), Cameron Peng (discussant), Erik Stafford,
and seminar/conference participants at Aalto University, the Becker Friedman Institute CITE conference on New Quantitative
Models of Financial Markets, Bocconi University, EIEF, FCA, FINRA Market Structure conference, FIRS, LAEF 2nd OTC
Markets and Securities Workshop, INSEAD, Macro-Prudential Conference ECB, New York Federal Reserve Bank, AQR, UC
Dublin, WFA for helpful comments.
1 Introduction
Large institutional orders are typically split in smaller amounts over time to avoid market-impact
(see Garleanu and Pedersen, 2013, Di Mascio et al., 2016). One concern when executing an order
slowly over time is that other traders might anticipate the intent to trade the stock in the near future
and take advantage by trading in the same direction to benefit from the future price impact. This
problem is particularly pronounced in the case of fire sales, during which the seller is forced to
bring to the market a large quantity of assets in a limited amount of time (Coval and Stafford,
2007; Ellul, Jotikasthira, and Lundblad, 2011). Moreover, if the liquidation occurs at times of
market stress, predatory trading can make the market more illiquid and amplify adverse shocks
(Greenwood, Landier, and Thesmar, 2015). Given this possibility, some observers suggest that
reducing the frequency of portfolio disclosure can be a desirable measure to prevent predatory
behavior (Brunnermeier and Pedersen, 2005).
However, market participants may possess information about forced liquidations thanks to their
close relationship with the liquidating managers. Among all actors in the market, brokers are in
the privileged position of observing the daily trades of a fund. In the case of hedge funds, prime
brokers operate also as lenders and risk managers, so that they are aware whether the fund is about
to breach some risk limit and deleverage its portfolio. They can also infer the trading habits of their
clients, such as whether they tend to cut trades in small orders over several days when executing a
large order. Thanks to this information, brokers are best placed to predict the future trades of their
clients.
Brokers may decide to spread the news that a client’s large trade is likely to extend over time
to other market participants. They may have an incentive to do so in order to establish a reputation
as a source of valuable information and attract new business. Other investors can use this
information to predate on the distressed fund. On the other hand, brokers may be reluctant to foster
predatory trading against a client, as it may harm their reputation. Rather, according to this
argument, they should invite other traders to provide liquidity and take the other side of the slow
trade. It remains, therefore, an open empirical question whether brokers foster predatory trading
or liquidity provision in case of slow trading by a client. The paper aims to address this question.
Forced liquidations of portfolio holdings offer an ideal setting to investigate these issues.1
Accordingly, we exploit proprietary trade-level data and identify asset managers that sell a
significant fraction of their portfolio during a relatively short amount of time. We restrict attention
to asset managers whose order flow is abnormally negative for at least five days in a row.
Moreover, we focus on managers that liquidate multiple stocks (on average about 20 stocks) at a
significantly faster pace than usual. We identify about four hundred of these events in the period
between 1999 and 2014. We verify that the stock price movements resulting from this sale are only
temporary, consistent with the notion that we identify liquidity events. Price impact would have to
display a permanent component, if sales were motivated by fundamental reasons.
Not all brokers employed by the liquidating fund are going to be aware that the fund is in
distress. The liquidating fund has little incentive to disclose its intention to liquidate a large fraction
of its portfolio; in fact, it is likely to use multiple brokers to minimize price impact and info leakage
(on average 29). Hence, we label as aware only the brokers that intermediate a large enough
fraction of volume. Our first result is that there is a significantly higher probability of predatory
behavior for orders executed through aware brokers. Specifically, the clients of the aware brokers
are much more likely to execute sell trades in the same stocks with the same broker over the same
period. While liquidity provision also takes place among clients of aware brokers, this activity
does not appear to be as prevalent as predatory trading.
Next, we explore the heterogeneity across the different clients of the aware brokers. If the
brokers are spreading information about order flow, they are more likely to do so with their best
clients, from which the brokers can extract the highest rents. As a proxy for the strength of the

1
We decide to focus on large liquidations (which we label “fire sales” for convenience), and do not include large purchases in our
analysis, because we aim to have a clean identification of liquidity-motivated trades. First, in our data, the majority of institutional
investors are long-only (about 90%). Hence, it is somewhat less likely for a sale to be information motivated (as the manager would
need to have the stock already in the portfolio) than for a buy transaction. Second, large cash inflows can be allocated slowly over
time and are, therefore, less likely to impose a concentrated liquidity demand on the market than large outflows. Moreover, fire
sales can pose a systemic threat if they cause a propagation of idiosyncratic shocks to the balance sheets of other investors. Hence,
studying the effect of information leakage on fire sales is especially relevant, including from the regulatory perspective.

2
investor-broker relation, we use the trading volume and the commissions generated by a client.2
The main result of this analysis is that the best clients of the aware brokers are significantly more
likely than other clients to sell the stocks that the liquidating manager is offloading during the fire
sale with respect to immediately before the fire sale.3 Hence, the evidence suggests that predation
is more likely than liquidity provision among the best clients of the brokers that intermediate fire
sales. The magnitude is economically significant as the net probability of predation more than
doubles for the best clients of aware brokers relative to the small clients of the aware brokers.
Consistent with predatory trading, we find that a significant fraction of positions that are sold by
other managers than the distressed fund during the fire sale period, ranging from 30% to 42%, are
bought back in the ten days following the fire sale.
We provide an array of robustness checks to rule out the possibility that the originator of the
fire sale and the followers are trading as a response to the same information signal. For instance,
we exclude from our sample all the events that occurred during recessions and the results are
unaffected. We also exclude all events occurring around earning announcements, changes in
analyst recommendations, or any other type of negative news as reported by the press and classified
by the data provider Ravenpack. We also exclude stocks with negative momentum and high short
interest to address the concern that selling managers follow similar trading strategies founded on
a negative signal on the stock.
To strengthen the identification of fire sale events, we focus on a natural experiment in which
some mutual funds were forced to liquidate their holdings. Specifically, as a consequence of the
late-trading scandal of 2003, twenty-seven fund families experienced significant outflows. Anton
and Polk (2014) use these outflows to identify an exogenous driver of mutual funds’ selling
activity. Kisin (2011) estimates that funds of implicated families lost 14.1% of their capital within
one year and 24.3% within two years. Crucial for our purposes, the brokers of the liquidating funds

2
We find that these relations are extremely persistent, consistent with the findings in Goldstein, Irvine, Kandel, and Wiener (2009),
corroborating the hypothesis that brokers might have an incentive to nurture such relations.
3
We control for time, manager, event, stock, and broker fixed effects. Hence, differences across stocks, such as their liquidity, or
across brokers, such as their ability to execute, cannot explain our results. We also provide a specification in which we control for
broker-manager relationship fixed effects, which controls for the matching between asset managers and brokers. These results are
in the Internet Appendix.

3
were aware of the specific stocks that were being sold and of the timing of these liquidations. We
show that the clients of the relevant brokers were significantly more likely to liquidate the same
stocks after the scandal broke out on the same days on which the implicated funds are also selling.
This test reassures us that, even when we consider plausibly exogenous variation in the source of
the liquidation, we find very similar behavior.
One of the contributions of the paper concerns the value of the order-flow information. We
compute the profits that the asset managers make during the fire sales and show that the best clients
of the aware broker, who are the most-likely investors to benefit from information leakage, are
able to generate an additional 25 bps in the few days of the fire sale. Given average fund
performance, these results suggest that being able to predict fire sales can be quite profitable.
We also provide evidence on the externalities arising from the previous findings, i.e. the losses
incurred by managers exposed to predation. We focus on the execution shortfall, computed as the
volume-weighted percentage difference between the execution price and a benchmark price. We
find that price impact is higher by about 40% when the trades are executed through brokers that
are aware of the liquidations. We interpret this spread as the cost of predation. Also important, our
evidence highlights one important amplification mechanism for asset price fluctuations.
We conclude by addressing another important question: Do brokers gain from leaking order
flow information? We compute the brokers’ commissions and show that the best clients who take
advantage of the order flow information by preying on the liquidating funds pay, relative to the
other clients of the brokers and relative to the period before the fire sale, 16% higher commissions
in standard deviation units, confirming that brokers get rewarded for the information they provide.
Overall, our findings point out an important trade-off between slow trading execution meant to
reduce price impact, e.g. as in Kyle (1985), and leakage of order flow information. The latter
becomes more likely when the asset managers trade in the same direction over an extended period
of time. This consideration is not confined to fire sales events. In fact, we find that the
autocorrelation among large trades in our data, i.e. those larger than 1% of average daily volume,
is about 35%. Hence, as a rule, managers tend to trade in the same direction over multiple days,
which opens the possibility for the brokers to predict future order flow.

4
 Our paper bridges two strands of the literature. First, there is a vast literature on fire sales.4
Second, there is a growing number of studies investigating the importance of the network of
relations among market participants in various domains, e.g. Li and Schürhoff, 2018; Di Maggio,
Franzoni, Kermani, Sommavilla, 2018; Di Maggio, Kermani, and Song, 2017; Hollifield,
Neklyudov, and Spatt, 2016; Afonso, Kovner, and Schoar, 2013; Hendershott, Li, Livdan, and
Schürhoff, 2016. Our novel contribution is to highlight the key role played by brokers during fire
sales, which might be amplified due to brokers leaking order flow information.5
Evidence that brokers leak valuable information to selected clients is present in Irvine, Lipson
and Puckett (2006) regarding future analyst recommendations, in McNally, Shkilko and Smith
(2015) regarding brokers passing on information about firm insiders’ order flow, and in Di
Maggio, Franzoni, Kermani, and Sommavilla (2018) regarding informed order flow.
Closest to our work, a recent paper by van Kervel and Menkveld (2018) studies the behavior of
HFTs around large orders of institutional investors. The authors find that HFTs provide liquidity
if the order is short-lived (below seven hours), but they back run on the order if it lasts for several
hours within a day, that is, HFTs trade in the same direction of information-motivated orders. The
latter behavior increases the trading costs for the institution, as predicted by the theory of Yang
and Zhu (2016). Similar to van Kervel and Menkveld (2108), we also study the interplay among
institutional investors and we detect a trading behavior by other investors that is harmful to the

4
Theoretically, Shleifer and Vishny (1992, 1997), and Kiyotaki and Moore (1997) suggest that fire sales occur when the natural
buyers are unable to purchase the assets due, for instance, to agency problems. Brunnermeier and Pedersen (2005) and Di Maggio
(2016) show that the market might become illiquid exactly when liquidity is needed most due to unconstrained arbitrageurs taking
advantage of the temporary price pressure by selling and then buying back the asset only after the fire sale has ended.See Shleifer
and Vishny (2011) for a survey of this literature. A complete list of works on fire sales and price dislocations in financial markets
is beyond the scope of the paper, but it includes among others Allen and Gale (1994), Gromb and Vayanos (2002), Brunnermeier
and Pedersen (2009), Acharya, Gale, and Yorulmazer (2011), Garleanu and Pedersen (2011). Recently, Yang and Zhu (2016)
provided a two-period Kyle (1985) model of “back-running,” where in addition to informed and noise traders there is an investor
who learns from the order-flow generated by the informed speculator after the order is filled.
5
Our findings also relate to a growing literature examining the way in which information spreads in financial markets due, for
instance, to information percolation (Duffie, Malamud, and Manso, 2009, 2014), or network effects (Babus and Kondor, 2016 and
Walden, 2016). We contribute to this literature by providing empirical support to the notion that information can be readily
disseminated through interactions between intermediaries and market participants. Furthermore, our results can also inform the
theoretical developments of this literature as we point out that this information dissemination is strategic, a feature currently missing
in the existing theoretical literature and a driver of network formation in financial markets. Also related to our paper, Farboodi and
Veldkamp (2017) provide a long-run growth model where traders have the option to extract information from order flow data
mining and study the implication for price informativeness and market liquidity.

5
initiator of a larger order. Our evidence differs and complements their results in several
dimensions. First, we focus explicitly on liquidity-motivated orders (i.e. fire sales) and show that
predation occurs also in these circumstances and not just around information-motivated trades.
Second, we show that predatory behavior characterizes also traditional asset managers, not just
HFTs.6 Third, we identify institutional brokers as instrumental in spreading order flow information
and fostering predation. Finally, we highlight the systemic threat caused by predatory trading as it
can amplify price dislocations during fire sales.7
The remainder of the paper is organized as follows. Section 2 describes the data sources and
summary statistics and Section 3 discusses our main results on the behavior of asset managers and
the role of brokers during fire sales. Section 4 presents the results on the value of order flow
information, while Section 5 concludes.

2 Data and summary statistics

In order to analyze whether and how brokers leak order flow information during fire sales, one
needs a detailed trade-level dataset that also reports information on the institutional investors and
brokers involved in each trade. Abel Noser Solutions, formerly Ancerno Ltd. (we retain the name
‘Ancerno’ for simplicity), responds to these requirements. Ancerno performs transaction cost
analysis for institutional investors and makes these data available for academic research under the
agreement of non-disclosure of institutional identity.

6
Our paper does not focus on high-frequency predation because HFTs are not present in our data. Yet, the question of liquidity
provision vs. predation, which we address, has received special attention in the HFT literature. Moreover, the destabilizing effect
of predation during fire sales that we document also finds a counterpart in the studies focusing on the impact of HFTs on market
efficiency and volatility. Biais and Foucault (2014), O’Hara (2015), and Menkveld (2016) provide surveys of the rapidly growing
HFT literature. Several empirical studies find that HFT activity is beneficial in that it reduces transaction costs (Hendershott, Jones,
and Menkveld, 2011; Hasbrouck and Saar, 2013; Menkveld, 2013; Brogaard et al., 2015; van Kervel, 2015) and it improves price
efficiency (Boehmer, Fong, and Wu, 2014; Brogaard, Hendershott, and Riordan, 2014). The evidence on the relation between HFTs
and short-term volatility and crashes is mixed. Some studies document a negative relation (Hasbrouck and Saar, 2013; Chaboud et
al., 2014; Hasbrouck, 2015) whereas others document a positive relation (Gao and Mizrach, 2013; Ye, Yao, and Gai, 2013;
Boehmer, Fong, and Wu, 2014; Kirilenko et al., 2017).
7
Our results are also consistent with Chung and Kang (2016), who use monthly hedge fund returns to document comovement in
the returns of hedge funds sharing the same prime broker.

6
 We have access to identifiers for managers that initiate the trades and brokers that intermediate
those trades from 1999 to 2014.8 There are several advantages to this dataset. First, clients submit
this information to obtain objective evaluations of their trading costs, and not to advertise their
performance, suggesting that the data should not suffer from self-reporting bias. Furthermore,
Ancerno collects trade-level information directly from hedge funds and mutual funds when these
use Ancerno for transaction cost analysis. However, another source of information derives from
pension funds instructing the funds they have invested in to release their trading activities to
Ancerno for an independent check. Third, Ancerno is free of survivorship biases as it includes
information about institutions that were reporting in the past but at some point terminated their
relationship with Ancerno.
Previous studies, such as Puckett and Yan (2011), Anand, Irvine, Puckett, and Venkataraman
(2012, 2013), have shown that the characteristics of stocks traded and held by Ancerno institutions
and the return performance of the trades are comparable to those in 13F mandatory filings.
Furthermore, Goldstein, Irvine, Kandel, and Wiener (2009), using an earlier version of our data,
provide a useful description of the institutional brokerage industry. They show that institutions
value long-term relations with brokers. Also, consistent with our results, the best institutional
clients are compensated with the allocation of superior information around changes of analyst
recommendations.
Ancerno information is organized on different layers. At the trade-level, we know: the
transaction date and time at the minute precision (only for a subset of trades), the execution price;
the number of shares that are traded, the side (buy or sell) and the stock CUSIP. Our analysis is
carried out at the ticket level, i.e. we aggregate all trades on the same stock, on the same side of
market (buy or sell), by the same manager, executed through the same broker, on the same day.
Next, we provide the definition of a fire sale event. Our goal is to identify liquidity-motivated
sales that attract brokers’ attention and are likely to generate a significant but temporary price

8
Relative to the standard release of Ancerno that is available to other researchers, we managed to obtain manager and broker
identifiers also for the latest years (that is, after 2011), under the agreement that no attempt is made to identify the underlying
institutional names.

7
impact. Hence, we impose two requirements. For a given manager, the selling amount needs to
exceed the manager’s standard trading volume for a protracted period.9 At the stock level, the
liquidation volume needs to make a sufficient fraction of total trading volume.
In more detail, to identify liquidating funds we start by computing the signed volume Z-score
for each manager m on day t as
&'()*+ – - &'()*+ (1)
Z"# = ,
. &'()*+

where &'()*+ is the portfolio level dollar volume traded by manager m on day t, and its mean and
standard deviation are estimated over a rolling window of 120 trading days ending one week before
day t. Then, for a given manager, we require that during a fire sale event Z"# is below -0.25 for at
least five trading days in a row. This requirement ensures that the sale is taking place on a
sufficiently long period of time for the broker to realize about the fire sale and for it to represent a
significant event in the life of the fund. Given this condition, all the fire sales that we identify
corresponds to events in which the order imbalance at the fund level is negative, as evident from
Table A8 in the Internet Appendix.
In addition, we impose a filter at the stock level to ensure that the sale volume is large enough
to generate price pressure. For stock j to be part of the fire sale event, we require that the volume
traded by the manager is at least 1% of the CRSP volume on day t for at least four out of the five
fire sale days.
We decide to keep events in which at least 10 stocks are involved in a fire sale. The goal is to
reduce the probability that liquidating funds are selling as a consequence of stock-specific

9
Our level of analysis is at the manager code level; hence, at the level of the management company. Our decision to focus on the
management company level is founded on several arguments. First, our definition of fire sales selects events that are particularly
large for an asset manager. In this sense, it is more likely that fire sales arise when multiple clients withdraw their funds from a
management company. Focusing on a specific client-manager relationship then has the potential to miss these larger events. Second,
if only one fund in the company was in distress, or just a few, other funds could help by providing liquidity. Specifically, the
healthier funds could relieve the distressed fund of some of its assets by engaging in cross-trading, a practice that is described in
Gaspar, Massa, and Matos (2006), and more recently in Eisele, Nefedova, Parise, and Peijnenburg (2017) using Ancerno data. The
possibility of intra-family subsidization, then, motivates us to focus on events that involve the entire family of funds. Finally, the
choice to focus on the management company, as opposed to specific funds within the family, is also dictated by data availability.
In the version of Ancerno that is available to us, the alphanumeric identifier for the specific fund (manager) is often missing or not
meaningful.

8
information. Focusing on liquidations of a large number of stocks makes it less likely that the sales
are information driven.
Next, we distinguish between aware and unaware brokers. Intuitively, we define a broker as
aware that a stock is subject to fire sale pressure if it intermediates a sufficiently large volume on
that stock arising from the originator. In detail, the variable Aware is a dummy, defined at the
event-broker-stock-day level, indicating that the broker is aware of the fire sale happening on a
given stock-day. That is, for broker b, stock j on day t, and event e, the aware dummy 012345,6,*,7
equals one if the volume on stock j originated by the liquidating fund that is intermediated by
broker b on day t is above 2% of the average daily volume (ADV) for that stock. Note that this
does not require the broker to have knowledge of the overall size of the liquidation, but just
realizing that the distressed fund is responsible for a significant fraction of the daily volume. In
Table A1 of the Internet Appendix we demonstrate the robustness of our main results to several
changes in the ADV-related threshold used to identify aware brokers (for thresholds from 1% to
5%). Further, we show that our results are robust to the additional requirement that the broker
needs to intermediate a large volume of at least N stocks in the fire sale basket to become aware
of the fire sale event (for N = 1, 5, 10).
Panels A and B of Table 1 provide the summary statistics for the key variables in our analysis.
We identify a total of 385 fire sale events over the 1999-2014 period, each lasting at least 5 days
and with the liquidating funds selling on average $377 million worth of stock (median: $177
million). Figure A2 displays the distribution of events over our sample period. It shows that the
events are evenly distributed over time; in fact, even during the recessions marked with red
squares, the number of events does not spike. This confirms that our methodology identifies funds
subject to idiosyncratic shocks rather than market-wide events.10

10
The lack of clustering of fire sale events during crisis periods results as an intentional feature of our definition of fire sales. In
computing the Z-score, at the numerator, we subtract from a given day’s order flow the average daily order flow over the prior six
months. Hence, if the order flow is negative over a protracted period, such as during the crisis, at some point the Z-score will cease
to identify fire sales. The desirability of this feature is that we do not generate a sample of fire sales in which the crisis is overly
represented.

9
 We can compute the fraction of the liquidated portfolio that the liquidation volume represents.
In particular, we estimate the liquidating funds’ portfolios by cumulating their trades over the two
years prior to the fire sale. Then, we divide the total volume of sold stocks by the reconstructed
portfolio size. We find this fraction to be sizeable at 9.16%, on average. Arguably, this
methodology tends to underestimate the liquidating managers’ actual portfolio because we do not
know their positions at the beginning of the estimation period, so that the fraction provides an
upper bound. In any event, this evidence suggests that these large sales are unlikely to be inspired
by stock-specific information.
On average, 22 stocks are heavily sold during a fire sale event, with about $17.2 million sold in
each stock, which indicates that these events involve more than just isolated stocks. Figure A1
shows the distribution of these events as a function of the number of stocks, from events involving
10 to 50 stocks, as well as the distribution of the volume of trades by the liquidating fund that can
even reach more than two billion dollars in some cases.
Fire sales are intermediated by an average of 29 brokers, while the number of aware brokers
per event is on average 1.7. Furthermore, the price of the stocks sold in the fire sale declines by 83
basis points on average during the first five days of the event (Table 1, Panel C), but there is
significant variation. In fact, for the bottom quartile, the price drops by more than 3%.
Using TAQ data, we can report that the fire sale volume is on average 50% of the TAQ order
imbalance (median 10%) and it is on average 27% of TAQ sell volume (median 19%). We
conclude that the liquidating fund imbalances constitute a sizeable fraction of the TAQ imbalance
for the fire sale stocks.
Finally, we provide evidence on the type of stocks the liquidating managers are selling. For
each stock in the fire sale, we compute the fraction of the total volume in the fire sale that it
represents. Panel E of Table 1 shows the results from regressions of the fraction of the fire sale
that stock j represents on its weight in the selling manager’s reconstructed portfolio, market
capitalization, volatility, the Amihud (2002) ratio, and various measures of past performance at
different horizons. We find that, after controlling for the quantity held by the manager (i.e.
portfolio weight), the funds tend to sell the larger, more liquid, and less volatile stocks in their

10
portfolio. Also, asset managers tend to sell the stocks with higher past performance. These findings
resonate with the predictions of theoretical models studying the liquidation strategies in case of
distress (Scholes 2000, Brown, Carlin, and Lobo 2010).
Corroborating our identification strategy for fire sales, the highly significant positive coefficient
on the reconstructed portfolio weight suggests that the liquidating funds are not building short
positions; rather, they are selling positions that are already present in their portfolio.
In addition to the extensive margin, we have also investigated the sequence of the sales by
changing the dependent variable to “first day in which the stock is sold”, defined as the number of
business days from the first day of the fire sale in which a particular stock is sold the first time.
The results are reported in Panel F of Table 1 and show that the most liquid and less volatile stocks
are sold earlier.

3 Main Results
This section starts by discussing our empirical strategy and then presents the main evidence on the
role of brokers in spreading order flow information during fire sale events.

3.1 Fire Sales

We start our analysis by characterizing the fire sale events. Figure 1 plots the average (across
stocks and events) daily signed volume (i.e. order imbalance) for the liquidating fund during the
event window, where the zero is defined as the first day of the five-day window over which we
identify the fire sale. The large negative volume before day 0 is due to the fact that, while
liquidations likely start earlier, we impose stringent criteria for them to be defined a fire sale. We
note that, although the daily order imbalance is smaller in magnitude after five days, it is still
negative even after fifteen days. This is important, because it highlights the nature of the sale: the
liquidating fund does not repurchase the stocks back (even when we extend the horizon further
out). Hence, this fact weakens the possibility that the liquidating fund is short selling the stock
because it expects the price to decline, and then buys the stock back.

11
 Figure 2, instead, plots the average DGTW adjusted cumulative returns for the stocks included
in the fire sales across all the events. The returns are mostly flat pre-event and then start
precipitating quite rapidly while the liquidating fund (for simplicity, the originator) is selling most
intensely, i.e. during the five-day interval [0,4], then to slowly recover over time. Specifically, we
find that after about twenty days they are back to the pre-event levels. This is a faster reversal than
what is found in the existing literature on fire sales (Coval and Stafford, 2007). On average, the
price drops by almost 1% during the five-day event-time interval [0, 4], which we label liquidation
period. Importantly, the fact that we observe a reversal over such a short horizon tends to rule out
the possibility that the liquidation and the price decline are due to negative fundamental news on
the stock. On the contrary, the price path is strongly consistent with price pressure following
liquidity motivated trades.

3.2 Predation or Liquidity Provision?

The theoretical literature makes mixed predictions on whether other market participants that
anticipate a large liquidity order will predate upon it or, instead, provide liquidity. Brunnermeier
and Pedersen (2005) and Di Maggio (2016) predict that investors that become aware of a
liquidation will predate on the distressed fund and deteriorate market quality. On the other hand,
Admati and Pfleiderer (1991), in their “sunshine trading” model, argue that investors credibly
announcing their intention to transact for non-fundamental reasons attract natural liquidity
providers to the market. The empirical work by Bessembinder, Carrion, Tuttle, and Venkataraman
(2016) provide evidence that is consistent with this prediction in the context of predictable roll
trades of oil futures contracts by a large ETF. Therefore, a priori, the type of behavior that other
market participants adopt vis-à-vis a liquidating fund remains an open empirical question.
To disentangle whether brokers foster predatory trading or liquidity provision we estimate the
following specification

849 :34;29<(=>,<,?,9,4 = @1 01234 + C>,<,?,9,4 , (2)

<,?,9,4

12
where 01234D,6,*,7 is a dummy equal to one if broker b executing the trades is aware of the fire
sale on stock i on day t of event e, i.e., it is defined at the event-broker-stock-day level. The
dependent variable, 849 :34;29<(=+,D,6,*,7 , is constructed as the difference between the
probability of predation and the probability of liquidity provision. In turn, the probability of
predation is a dummy equal to one if the client m of broker b trades in the same direction as the
originator, i.e. demanding liquidity, on a stock i on day t of event e. The dummy equals zero if the
client provides liquidity by trading in the opposite direction of the originator or the client does not
trade on that stock-day.11 Symmetrically, the probability of liquidity provision is 1 if the client
trades in the opposite direction of the fire sale, and 0 otherwise. We also estimate specifications in
which the dependent variable is defined as the net predation variable multiplied by the ratio of
dollar volume of the broker’s clients to the market capitalization of the stock (this variable is
standardized by subtracting the mean and dividing by standard deviation). The sample includes
trades executed by all managers with all brokers in the database on the fire sale stocks.
These specifications rely on heterogeneity across brokers for identification: some brokers are
more exposed to order flow information as they intermediate a higher fraction of the order flow by
the liquidating fund.12 Standard errors are clustered at the broker level. In the Internet Appendix,
we report equivalent results with clustering at the broker and stock level, and the broker and day
level (Table A6).
We present the results in Table 2, Panel A. Columns (1)-(2) focus on Net Predation, i.e. the
difference between the predation and the liquidity provision dummies, while columns (3)-(4)
present the results for the volume-weighted version of the dependent variable. For each dependent
variable, we modify the baseline specification by adding day-by-stock fixed effects to the existing
set of fixed effects. In particular, manager and broker fixed effects ensure that our estimates are
not driven by unobservable broker or manager characteristics. Day-by-stock fixed effects aid in
ruling out two alternative explanations. First, asset managers might sell the stock due to stock-

11
To identify non-trading clients, we consider all the managers that traded with the broker in that stock over the previous 20
business days.
12
We find that the subset of brokers that are deemed aware during our sample period amounts to roughly 10% of the brokers present
in Ancerno.

13
specific public news, then, day-stock fixed effects would capture this potentially important
confounding factor. Second, predation might be driven by information about prices and trades,
rather than by information leakage. For instance, the liquidating managers create price impact and
abnormally high volume in the market, which is a source of public information that asset managers
can use to spot trading opportunities, without relying on brokers.
We find that trades executed by aware brokers have between 11% and 20% higher difference
in the probability of predation relative to the probability of liquidity provision.13 The analysis of
volume in columns (3)-(4), confirms the finding.
In Panels B and C, we separately study the relation between broker awareness and the
probabilities of predation and liquidity provisions, respectively. While there is a significant
relation between broker awareness and predation, the effect on liquidity provision is not clear and
it is positive and significant only in the specifications where liquidity-provision volume is the
dependent variable. Even there, the effect is smaller than for predatory volume.
Another way of investigating predation and liquidity provision is to compare the cumulative
order imbalance from the start of the fire sale, where order imbalance is the difference between
buy and sell trades divided by the sum of the two. We report the series for the aware and the
unaware brokers with standard errors bands during the events in Figure 3. Confirming the
regression evidence in Table 2, the imbalances through the aware brokers are negative during the
first several days of the liquidation and are significantly lower than those of the unaware brokers.
Overall, the results show that the brokers who are more likely to realize that the fund is engaged
in a large liquidation are also more likely to intermediate trades that are consistent with predatory
trading. Instead, the evidence that aware brokers facilitate liquidity provision is scanty at best.

13
In Internet Appendix Table A9, we report results from specifications without the fixed effects. In these specifications, the
constant, i.e. the level for unaware brokers, is virtually zero, while the slope on the aware dummy is 23%. Hence, the economic
magnitude is substantial.

14
3.3 Best Clients and Predatory Trading
To sharpen our identification, we focus on the aware brokers and test yet another implication of
our information leakage hypothesis. If the aware brokers provide information about order flow
from liquidating managers, and if the information rents can be dissipated by leaking to too many
traders, we should expect this disclosure to be selective and to allow the broker to extract the
highest rents. Thus, we should expect the brokers to favor their best clients.
To proxy for the strength of the manager-broker relationship, we use information about both
the volume and the commissions generated by manager m with broker b in a window of 6 months
ending one month before the fire sale event and construct two measures of relationship strength
(the ‘Best Client’ proxies). The first variable is defined as the volume generated by the client as a
fraction of the total volume intermediated by the broker and expressed in decimal units. The other
measure is computed in a similar fashion, but the dollar volume is replaced by the dollar trading
commissions generated by the manager. Summary statistics are reported in Panel D of Table 1.
These variables are highly persistent, with an autocorrelation of 90% at the monthly frequency.
This fact suggests that brokers might have an incentive to nurture these relationships over time and
that the heterogeneity across clients of the same broker might be a relevant source of variation for
identifying the effect of interest.
To study the role of broker-client relationships in information diffusion, we estimate the
following specification

849 :34;29<(=+,5,6,*,7 = @E F4G9 H)<4=9+,6,* × J<KL<;29<(= :43<(;*,7 + (3)

@M F4G9 H)<4=9+,6,* + @N J<KL<;29<(= :43<(;*,7 + C+,5,6,*,7 ,

where index j runs over stocks, index b over brokers, index e over events, index m over managers
and index t over days. As for equation (2), our main dependent variable is the difference between
the probability of predation and the probability of liquidity provision. As in the case of Table 2,
we also use a version of this variable that is multiplied by the volume of the trade. The dummy
Liquidation period indicates the first five days of the fire sale, that is, for the period of most intense

15
liquidation by the fund in distress. The reference period is the time before the beginning of the fire
sale. All specifications include time (at the monthly frequency), manager, event, stock and broker
fixed effects. We conservatively double-cluster the standard errors at both the stock and manager
level, which allows for arbitrary correlation within trades in the same stock and by the same
manager.14
Table 3 presents the results. We find that the asset managers in a closer relationship with the
fire-sale-aware broker are significantly more likely to sell their holdings of the fire-sale stock with
the same broker during the liquidation period. The magnitude is economically significant as the
net probability of predation more than doubles for the best clients of aware brokers (top decile)
relative to the small clients of the aware brokers (bottom decile).15
A more stringent identification strategy exploits variation across managers as well as across
brokers. That is, we compare the difference between the behavior of the best clients of the brokers
that are aware of the fire sale and the behavior of the best clients of the brokers that are unaware,
relative to the non-best clients of both types of brokers. Formally, Panel B of Table 3 reports the
results from the following specification

849 :34;29<(=+,5,6,*,7 = @E F4G9 H)<4=9+,* × 012345,6,7 × J<KL<;29<(=:43<(;*,7

+ @M F4G9 H)<4=9+,* × 012345,6,7
+ @N F4G9 H)<4=9+,* × J<KL<;29<(= :43<(;*,7 & (4)
+ @P 012345,6,7 × J<KL<;29<(= :43<(;*
+ @Q F4G9 H)<4=9+,* + @R J<KL<;29<(= :43<(;*,7 + @S 012345,6,7
+ C+,5,6,*,7 .

In this specification, we define 012345,6,7 at the event-broker-stock level by collapsing

awareness on the time dimension by taking the max, i.e. to each broker b which eventually becomes

14
In robustness tests provided in Panel B of Internet Appendix Table A6, we cluster standard errors along alternative multiple
dimensions: Event, Stock, and Manager; Event, Stock, and Day; Event, Stock, and Broker level. The results remain significant.
15
To reach this conclusion we use the estimates from the regressions without fixed effects in Table A10, Panel B. In particular, the
net probability of predation moves from 1% in the bottom decile to 2.1% in the top decile of best clients.

16
aware of the fire sale event e on a stock j we assign 012345,6,7 = 1. Results from this specification
are reported in Panel B of Table 3 and confirm that clients with stronger ties to the aware broker
are significantly more likely to sell the stock involved in the liquidation than the best clients of the
other brokers involved in the liquidation. Results are robust across the two ‘Best Client’ measures.
One interesting question at this point is whether there is significant persistence in the set of
asset managers that predates and in those that get predated. We find that more than 60% of the
victims were predated only once. The median is thus 1, while the average is 3.13 times. This
suggests that the liquidations we are focusing on are unlikely to happen frequently enough for the
funds to become aware of that and potentially punish the broker. In fact, even among those funds
who are predated more than 2 times, the average time between two consecutive events is 2.86
years. It is thus difficult for a manger to learn about the brokers’ leakage, given that, from a
manger’s perspective, predation happens rarely and inference is very noisy.16
From the perspective of the predators, we also show that this is a concentrated activity. In fact,
among all the predators in our sample, 30% of them predate on more than 10 events during our
sample period. Predatory behavior is persistent: conditional on having predated at time t, the
probability of the same manager predating again in t+1 is more than twice as large. Figure A3 in
the Internet Appendix presents the result for different time horizons. Therefore, the evidence
suggests that, consistent with our hypothesis, the brokers leak their information to a restricted
number of clients that are likely to take advantage of this information.
One additional dimension that we can explore is when the predators start trading in the same
direction. Intuitively, if the predator starts on the first day of liquidation, it is potentially much
more harmful than if the predator starts on the last day of the liquidation. We examine this question
in Internet Appendix Table A12. We find that the best clients of the aware brokers are significantly
faster in predation. In particular, the average predator starts predating on the third day of the
liquidation, while best clients of aware brokers start already on the second day, on average. This

16
We also find that, among the funds involved in a fire sale, 40% also acted as a predator at least once.

17
is interesting because suggests that the best clients of the aware brokers are rewarded through early
access to information.
Further, we can test whether brokers give a preferential treatment to their best clients when they
need to liquidate. In Internet Appendix Table A10, we find there is less predation when the fund
in distress is one of the broker’s best clients. We do not find significantly more liquidity provision,
but the results are confirmed when we look at the difference between predatory and liquidity
provision volumes. Overall, clients with closer ties to the brokers do enjoy an advantage when they
need to liquidate.
Finally, using reconstructed portfolios, we find that predators appear to short the fire-sale
stocks in 43% of the cases. We caveat, however, that this estimate is exposed to large measurement
error due to the approximation of the true portfolio. From a theoretical point of view, we do not
have a strong prior as to whether predation should occur with stocks that are already in the
predator’s portfolio or stocks that the predator needs to short. Empirically, given that the stocks
that are most likely to be predated tend to be the largest and most liquid stocks in the market, it is
somewhat more likely that these stocks are already in the predators’ portfolios. This fact can
explain why a slight majority of predatory trades consists of sales of existing positions.17

3.4 Robustness to Aggregate and Stock-Specific News

Having established that the best clients of the aware brokers are more likely to sell the same stock
as the distressed fund during the liquidation period, we examine whether the results can be driven
by other factors than information leakage by the broker. The main alternative hypothesis that might
explain these results is that asset managers are responding to the same common shock occurring
during the same event windows. This might occur for two reasons. First, there might be a common

17
Table A2 in the Internet Appendix examines the characteristics of the stocks that are more subject to predation. We split the
sample of fire sale stocks by the median of the amount of predation. In turn, this quantity is the number of manager-days in which
a client of an aware broker trades in the same direction as the liquidating fund. Then, for different variables, we compute the average
for stocks that are liquidated in events above the median (More Predation) and below the median (Less Predation). The overall
evidence is that the events with stronger predatory activity involve larger, more liquid, and less volatile stocks.

18
disruption in the market that leads funds to offload their positions. Alternatively, news about the
specific stocks might be released, triggering the funds’ trading behavior.
As already discussed, some of our prior evidence (e.g. the fact that we control for stock-by-day
fixed effects in Table 2) already helps to rule out these hypotheses. Nonetheless, we provide
several other tests to rule out these alternative explanations. The first step to ensure that the
correlation among traders is not due to general disruption in the market is to exclude the two
recessions in our sample, i.e. the tech crunch and the financial crisis. Panel A of Table A3 of the
appendix presents this analysis. The results are robust to this change in the estimation sample, with
both the economic and statistical significance being unaffected.
Next, we test if negative stock-specific news might explain our baseline results. To do so, we
collect information about earnings announcements and changes in analyst recommendations.
Intuitively, earning announcements might work as a catalyst, and a negative surprise might trigger
a series of liquidations. We exclude ten trading days around the announcements. Another important
piece of fundamental information that might drive funds’ behavior is changes in analyst
recommendations. One might reasonably expect that multiple liquidations might follow a
downgrade, especially an unexpected one. Therefore, we also exclude these events from our
sample. Earnings announcements and analyst recommendations are not the only news that might
trigger a coordinated response from market participants. In order to have the most comprehensive
information about stock-specific news, we use the data provided by Ravenpack. The dataset is
generated as the result of a comprehensive analysis of all types of information from newswires
about each stock, from lawsuit to mergers and acquisitions. A machine learning algorithm is then
employed to classify the news in good and bad on a scale from 0 to 100, where 50 is the cutoff
below which news are identified as bad. Even in the restricted sample excluding bad news, we
confirm in Panel B of Table A3 of the appendix that the best clients of aware brokers are more
likely to predate on the liquidating manager.
Another instance in which fund managers might find themselves trading in the same direction
is when the stocks belong to the same strategy, e.g. momentum, which might be commonly adopted
by multiple funds. Furthermore, asset managers might be liquidating underperforming stocks.

19
Then, as an additional robustness check, in Panel C of Table A3 we exclude from our sample all
stocks exhibiting negative momentum. Specifically, we compute the returns of the stocks sold
during the fire sale and exclude those with negative returns in the week preceding the fire sale.
The results are unaffected.
To check whether our results could be driven by changes in investors’ expectations about the
stocks, Panel D of Table A3 also considers short selling data from Markit (formerly DataEx
database). Intuitively, stocks with high short interest might be subject to correlated sales across
funds, which might be triggered by company specific events or investors’ common beliefs about
the stock performance, rather than by the desire to take advantage of a liquidating fund. Then, we
show the robustness of our results to the exclusion of events where the liquidated stocks exhibit a
significant level of short interest, defined as a utilization ratio (i.e. shares on loan divided by shares
available to lend) in the top quartile.
As an additional test to rule out the alternative hypothesis that funds are responding to similar
shocks rather than deliberately taking advantage of the fire sale, we explore the number of stocks
that are affected by the predatory behavior of the aware broker’s clients. The idea is that if investors
are simply responding to a common shock to a stock, we might find that their sales are concentrated
on that particular stock. On the other hand, if multiple stocks out of the 20 that are involved on
average in a fire sale are sold by the best clients of the aware broker, predation on the liquidating
fund seems more likely. To test this conjecture, Table 4 reports results where the outcome variable
is the number of fire-sale stocks for which the manager sells its holdings in columns (1)-(2), and
the fraction of stocks involved in the fire sales for which we observe predatory behavior in columns
(3)-(4). We find that top-decile clients of the aware brokers tend to sell about 8 more stocks than
bottom-decile clients do (column (1)), and to predate about 33% more of the stocks involved in
the fire sale (column (3)).18

18
The bottom decile of the Best Client proxy based on volume is 0, while the value for the top decile 0.58 is (median point between
the 90th percentile and the max of the distribution). Hence, we get an increase of 14.5×0.58 = 8.4 stocks, while the increase in the
fraction of predated stocks is: 57.8×0.58 = 33.5%.

20
3.5 Evidence of Trade Reversion
To corroborate the hypothesis that our results are driven by predatory behavior by the asset
managers who are able to acquire order flow information via the broker, we test whether these
same asset managers are also likely to cover their positions by repurchasing the stock in the
following days.
To this purpose, we compute the fraction of a manager’s negative position that is subsequently
reversed. In detail, the percentage of position reversed for manager m during event e for stock j is
defined as the ratio U4V7,+,5 = F(LWℎ9F2YZ7,+,5 / \();],#,^ , where \();],#,^ is the dollar sum of
all sell orders in that period, and F(LWℎ9F2YZ7,+,5 is the dollar sum of buy orders during the
period, where we sum only over he buy orders that are preceded by a negative cumulative order
flow. Our motivation is to avoid counting as reversals the buy orders that occur before sales have
taken place. We compute this measure around each fire sale event, for the ten days before and after
the fire sale. We then compare the percentage of position reversed across clients of the aware
brokers before and after the fire sale events. The liquidating funds are excluded from the sample.
In Table 5, we find that a significant fraction of the predating managers’ positions is covered in
the ten days following the fire sale. We interpret this evidence as strong indication that the
predating managers are motivated by the prospect of short-term gains at the expense of the
liquidating fund.19

19
To give a sense of the magnitude if the trade reversal, we compare this unwinding activity to the reversal of sell trades by the
same group of predatory managers taking place over a random sample of five-day intervals that do not include a fire sale (a placebo
sample). Figure A4 in the Internet Appendix compares the unwinding of predators’ trades after the fire sale to trades on placebo
days, where predators are managers trading in the same direction of the fire sale and placebo days are days are in intervals in which
no fire sale takes place. It is evident that reversal is significantly higher after the fire sales. In particular, already one day after the
fire sale 30% of the sell positions are bought back, while the number is closer to 3% in the placebo sample. After one month the
reversal plateaus at about 50% of the positions, while it is below 25% in the placebo sample. We conclude that the evidence of
trade reversal after fire sales is economically significant. The fact that not all sell trades are reversed suggests that either some
investors already intended to sell the stock and took the opportunity of a price decline to do it, or that some investors mistook the
price drop for a negative signal on the stock and decided to drop it from the portfolio.

21
3.6 Late-Trading Scandal as a Natural Experiment
We can envisage another alternative interpretation to the proposed view that order flow leakage
by brokers explains our evidence. Specifically, the intermediating broker can be the original source
of the information about the liquidated stocks, which then triggers the large sale as well as smaller
sales by other managers in the same direction.
To further rule out this alternative, we identify an exogenous determinant of fire sales. In
particular, we need a driver of liquidations that is manager-specific, i.e. it is not inspired by the
broker, and which does not depend on the identity of the liquidated stocks or the composition of
the manager’s portfolio.
Anton and Polk (2014) use the liquidations triggered by outflows following the late-trading
scandal as a natural experiment to identify exogenous selling activity (also see, Kisin 2011). We
follow these authors and focus on the mutual fund scandal that erupted in September 2003. At the
time, the New York Attorney General Eliot Spitzer announced the discovery of illegal late trading
activities and market timing practices on the part of several hedge fund and mutual fund
companies. The scandal had a significant impact on the 27 fund families involved: they
experienced significant outflows as they lost 14.1% of their capital within one year and 24.3%
within two years (Kisin, 2011). This is an ideal experiment for our purposes because it allows us
to identify stocks that for exogenous reasons are subject to selling pressure. Although market
participants were aware that these fund families were experiencing investors’ outflows, the
brokers’ vantage point allows them to pin down when these funds were liquidating and which
stocks were involved in the liquidation. Both pieces of information are crucial in making the
predation profitable and they are not publicly available.
To test whether even in this case, the brokers are responsible for leaking information about the
stocks that are liquidated and the timing of these liquidations, we manually match the identity of
the fund families included in Spitzer’s complaint with our trade-level dataset, in order to identify
the sales trades of these fund families and the brokers through which they execute them.20

20
A complete list of the fund families involved in the scandal arising from Spitzer’s complaint can be found on the webpage:
https://en.wikipedia.org/wiki/2003_mutual_fund_scandal#List_of_implicated_fund_companies.5B4.5D.5B5.5D. Out of the 27

22
Corroborating the validity of our matching procedure, we find that the matched managers rank in
the top quartile by sales in the two-year period following the breakout of the scandal.
Then, we focus on daily transactions of the managers that are not involved in the scandal for a
period of four years centered on the month of the announcement of the complaint by Spitzer
(September 2003) and define a dummy :(G9 \Y2=;2)* , indicating the two years after the
complaint broke out. Next, we define a broker-stock-day level dummy variable, \4))<=W6,5,* ,
indicating that at least one of the charged funds is selling stock j on day t through broker b. Then,
we define the dependent variable Probability of Predation as a dummy variable that equals 1 if a
non-charged manager is selling stock j on day t through broker b. The dependent variable equals
0 if a non-charged manager trades on a different day, or on a different stock, or with a different
broker. In a difference-in-differences setting, we regress the probability of predation on the
interaction between \4))<=W6,5,* and the dummy :(G9 \Y2=;2)* .
Table 6 reports the estimates. Consistently with the previous baseline results, we find that the
clients of the brokers employed by the funds involved in the scandal were significantly more likely
to liquidate the same stocks after the scandal broke out. For example, in Column (1), there is a
6.2% higher probability of non-charged managers to trade in the same direction as a charged
manager on the same day through the same broker relative to the pre-scandal period (i.e. 8.7% -
2.5%).
These results corroborate the interpretation that the clients of the aware brokers adopt predatory
trading strategies to take advantage of temporary price movements due to fire sales, and that these
results cannot be explained away by shocks to the market or to the single stocks as well as by a
common response to the release of public information, given that the timing of the sales and the
identity of the stocks that are sold is information to which only the intermediating brokers have
access. Moreover, the interpretation relying on the idea that brokers are generating stock-specific

families that are involved, we are able to find a match in our dataset for 19 of them. These 19 managers are responsible for 7 out
of the 31 fire-sale events in the two-year-period after the scandal broke out, i.e. they late-trading scandal families generate 23% of
the fire sales. Importantly, the implicated funds represent only about 2.1% of the managers in the database (i.e. 19/900). Hence, the
implicated families weigh about 10 times more than the other managers in generating fire sales in those two years.

23
trading ideas seems implausible, given that there is no reason for this activity to increase after the
breakout of the scandal or for stocks liquidated by the implicated funds.

3.7 Heterogeneity
We should expect the most active managers in the sample to be the ones more willing and
capable of taking advantage of the liquidating funds’ trades. To proxy for these characteristics, we
can investigate whether the results differ for hedge funds and other institutions. Intuitively, hedge
funds are more likely to have the ability to promptly react to information released by the brokers
than mutual funds or pension funds. We manually identify the hedge funds in Ancerno following
the procedure in Franzoni and Plazzi (2015).
Table A4 in the Appendix reports the estimates of Equation (3) for hedge funds and other
institutions. The results clearly show that the hedge funds are the main culprits of predation. The
statistical significance, as well as the economic significance, is weaker for non-hedge funds. This
evidence corroborates the hypothesis that the behavior we observe is a deliberate attempt by the
smart money to take advantage of temporary price fluctuations.
To investigate whether predation is even more prominent in periods of financial distress, we
split our sample based on the value of the VIX during each fire sale event. The results in Appendix
Table A7 show that predation through aware brokers is somewhat stronger during periods of
financial distress. This finding may result from the fact that liquidations are more significant during
times of market stress. Additionally, the price impact of liquidations is also larger because the
market is more illiquid. This fact creates additional room for predators to profit from the
liquidation.

4 The Value of Order Flow Information

4.1 Profitability of Predatory Strategies
An important question at this point is whether the asset managers that receive the information
from the broker are able to generate higher abnormal returns.

24
 To address this question, we compute the profits that asset managers generate during the fire
sales. In particular, starting from the first day of the liquidation (day 0), at the close of each day
we compute the marked-to-market value of the net position in a given stock and subtract from this
value the net cash amount that was necessary to build that position over the period. To express
these profits as a fraction of capital at risk, we divide them by the absolute value maximum dollar
outlay over the period in which the profits are computed.21
We start by showing in the left panel of Figure 5 the profits of managers that are best clients
(defined as those generating more than 5% of the volume intermediated by the broker in the
previous semester) of aware and unaware brokers at the daily frequency after the start of the fire
sale. Intuitively, if as shown in Table 2 the trades executed by unaware brokers are significantly
less likely to be predatory, we should find that their clients are also less likely to profit from these
fire sales events. Indeed, the figure shows that the clients of aware brokers are able to capture
significant returns after the start of the liquidation, while the trades of the clients of unaware
brokers do not generate significant profits. The profits for the best clients peak at about 25 bps on
the ninth day from the start of the fire sale. Importantly, most of the profits are generated in the
first five days, i.e. while the price of the fire-sold stocks is still decreasing and it makes sense to
predate. The profits that are generated later in the period are instead consistent with liquidity
provision.
Next, to provide more systematic evidence from regression analysis, we estimate the following
specification:

:3(_<9G+,5,6,* = @E F4G9 H)<4=9+ ×:(G9[0,9]* + @M F4G9 H)<4=9+ (5)

+ @N :(G9[0,9]* + C+,5,6,* ,

which tests whether a manager m’s profits are significantly higher in the ten days after the start of
the fire sale relative to the prior ten days, as a function of clients’ proximity to aware brokers.

21
To be clear, we subtract stocks that are sold from stocks that are bought to compute the net position, which can end up being
negative, as in a short sale. The net cash amount to build the position can also be negative if the sell transactions exceed in dollar
value the buy transactions. This fact implies that when we compute the maximum exposure, we need to use the absolute value.

25
Intuitively, as with the estimation of the predation probability, we are comparing the behavior of
managers that should be aware of the fire sale, given their relationship with the broker, with those
who are likely not, before and after the beginning of the fire sale. We choose a ten-day window to
allow managers the time to close the predatory short positions that they likely accumulate during
the first five days of the fire sale, which is the period over which on average the stock price declines
(see Figure 2).22
Table 7 reports the results showing that aware brokers’ best clients exhibit significantly higher
profits than other managers during the period under consideration. Clients in the top decile of our
relationship metric based on trading volume, in the ten days following the beginning of the fire
sale, are able to outperform by more than 50 basis points on average relative to the managers in
the bottom decile trading on the same stocks in the same period (i.e. (136.56-48.57)×0.58=51 bps,
where 0.58 is the value of the Best Client proxy in the top deciles, while it is zero in the bottom
decile). Considering the low average performance of institutional asset managers (see, among
others, Busse, Goyal, and Wahal, 2010) these returns are, indeed, highly economically significant.
One might wonder whether the clients of aware brokers are always able to generate higher
profits than the clients of unaware brokers. Although we already control for manager-fixed effects,
we also directly test for this possibility in the right panel of Figure 5, which provides a placebo
test for the left panel of Figure 5. The figure reports the profits for the two groups of managers,
but for a random sample of event windows other than the ones included in our fire-sale analysis.
We find that the two groups are indistinguishable in terms of their performance during these other
times.
We can provide more details regarding the profitability of managers that are strictly defined as
predators, that is, the aware brokers’ clients that trade in the same direction as the liquidating fund
during the five days of the fire sales. Considering all the predated stocks, the average predatory
position of a predator during a fire sale event is $10 million (median $6 million). The profits arising

22
Of course, the positions could be closed before day 5 and still be profitable. Our methodology for computing profits is flexible
enough to allow for all such possibilities.

26
from these positions amount on average to $280,449 per event-manager (median: $126,420).23
Next, these predatory profits correspond to 2.1 bps of the portfolio value (median: 0.9 bps). This
percentage profit is generated over 10 trading days on a predatory trade involving about 4 stocks
on average. Thus, it seems a significant source of returns, given the limited amount of capital that
is required to carry it out.24
We can also quantify the commissions generated by the predators’ trades. Each of the aware
brokers earns an average of $40,407 per fire sale event, considering only predatory trades by their
best clients on the fire sale stocks, which corresponds to roughly 50% of the total commission
these brokers earn on those stocks during the liquidation period (excluding those generated by the
liquidating funds). This said, we expect most of the benefits for the brokers to originate from the
future business that the improvement in relationship with tipped predators brings about. In
particular, in Section 4.4, we show that in the two years following the fire sale event, the predators
pay higher commissions per dollar traded to the tipping broker than other clients.

4.2 Price Impact

Having established that the predatory traders are able to capture significant returns, we investigate
the dark side of predation. The conjecture is that predatory volume causes stock prices to decline
significantly more than what they would do in the absence of predation. In turn, this steeper decline
in prices leads the liquidating fund to achieve lower returns on its sale trades.
Testing this conjecture requires the specification of a counterfactual. Fortunately, we can
identify fire sales events for which there are no aware brokers. These are 29 events (i.e. 7.5%) out
of a total of 385 events. In these situations, no broker observes a large enough fraction of the

23
The estimate of 32 bps for additional profits during predation period (Table 7) is computed averaging over all best clients of the
aware brokers (to avoid hardwiring the result). Additionally, it results from difference-in-difference regressions including fixed
effects. For this reason, this estimate is lower than the estimate of 280 bps (= $280,449/ $10 million) that we obtain for the average
profits of the predators only.
24
Because we compute profits using transaction prices, the price impact component of transaction costs is already accounted for.
To account for the explicit component of transaction costs, we can use the estimates provided in Ross, Israel, Moskowitz and
Serban (2017), based on AQR’s internal data. For US equity trades they report that commissions, fees, and taxes erode about 0.3
bps of the notional per trade. Therefore, assuming the trades involve roundtrip transactions, accounting for explicit costs would
reduce our estimated average trade-level profit in Table 7 (column 1) from 32 bps to 31.4 bps for the best clients of aware broker.
Instead, for the subset of predators, the profits as a fraction of the value of the open position drop from 280 to 279.4 bps.

27
liquidation to be deemed aware according to the criteria specified in Section 2. According to our
identification strategy, no information leakage occurs on these events. More realistically, the
information leakage is expected to be significantly lower.
Based on this strategy, we run regressions of price impact onto the broker awareness dummy.
In this case, the broker awareness dummy denotes situations in which there is at least one aware
broker for that stock-event. The price impact is computed as execution shortfall, i.e. the volume-
weighted percentage difference between the execution price and a benchmark price (e.g. Keim and
Madhavan, 1997).
We use three different benchmarks to show that our results do not crucially depend on a single
measure. Specifically, we use the price at the placement time of the first fire sale trade, the open
price of the day of the first fire sale trade, and finally the transaction price of the first fire sale
trade. In all specifications, we control for the volume in the fire sale, the volume of the following
trades (i.e. the trades in the same direction over the same five-day window), and the liquidity of
the stock (Amihud, 2002, Illiquidity Ratio), as they are all potentially important drivers of price
impact. In more detail, for each benchmark price we compute the implementation shortfall at the
ticket-level for the sales by the liquidating funds during the liquidation period as

d32=G2Y9<(=:3<Y4 − F4=Yℎ>23Z:3<Y4 (6)

.
F4=Yℎ>23Z:3<Y4

We average this quantity at the event-stock-broker level, using as weights the volume of each
transaction, to obtain an event-stock-broker level measure of price impact. We then further
collapse on the broker dimension to study the price impact at the event-stock level.
Results are reported in Table 8. In specifications (1)-(3), we regress the event-stock level price
impact measures on the dummy denoting events in which there is at least one aware broker.
Consistently across measures, we show that the price impact costs borne by the liquidating funds
are higher when at least one broker is aware of the liquidation event, and statistically significant
in two cases out of three. The estimates are also economically significant as the price impact
increases by at least 14 bps and up to 36 bps, which amounts to a two-fold increase in the baseline

28
average price impact. In specifications (4)-(6), we exploit the granularity of our data and run a
similar specification in an event-stock-broker level sample. In this case, for the same stock-event,
we can have aware and unaware brokers. We can then include broker fixed effects to control for
the possibility that heterogeneity in price impact results from difference in broker execution
quality. The results remain significant and the magnitude decreases only slightly.25
We can also reduce concerns that our findings are driven by expectation of larger price impact
affecting awareness rather than the opposite. The literature on block trading in equities (e.g. Seppi,
1990) suggests that this alternative interpretation might be due to the fact that brokers generally
frown upon clients breaking up a large order if this large order is likely to create meaningful price
impact. Therefore, the liquidating manager on this particular large order may pick very few brokers
if he anticipates price impact. This is a plausible alternative channel, given the argument that the
clients might refrain from splitting large orders if they are likely to generate a bigger price impact.
However, we can address this issue empirically by including in our main specification the number
of brokers used by the liquidating fund. With this control, the identification arises from comparing
aware and unaware brokers conditional on the number of brokers used in the liquidation. This
specification alleviates the concern that broker awareness is an endogenous variable, where the
endogeneity emerges because liquidating funds choose to use fewer brokers when they expect
higher price impact. On the contrary, our results show a positive (although not statistically
significant) correlation between price impact and the number of brokers used in the liquidation.
Finally, we can provide a graphical description of the difference in price paths between the case
in which brokers have the possibility to leak (aware brokers) and the case in which brokers do not
have information about the liquidation (unware brokers). Figure 4 plots the cumulative return of
the fire sale stocks during fire sale events. The red line with squares represents the cumulative

25
To further fix ideas in terms of orders of magnitude, we can look at Ross, Israel, Moskowitz and Serban (2017), based on AQR’s
internal data: the paper documents price-impact for a long-only momentum fund with USD 1.6B under management as of 2016.
They report that during 2009-2016, market impact was in the range of 6-11 bps per dollar traded. By comparison, in our sample of
liquidations, the average execution shortfall is 52 bps per dollar traded by the liquidating funds (median 40 bps). This puts us far
away from typical price-impact range. One might also be concerned whether the liquidations might be too small to attract arbitrage
capital. However, Table 4 shows that predators are able to trade multiple stocks from a given liquidation event, which through
diversification increases the Sharpe ratio of this type of trade.

29
return averaged across these stocks and events for the aware brokers. The green line with circles
is an estimation of the counterfactual cumulative return, based on unaware brokers. The series
draw on estimates from a regression specification similar to the one reported in column (3) of
Table 8, but run on daily observations starting on day 0. More precisely, the vertical distance
between the two series is the estimate of the aware broker dummy for a specific day of the interval.
Figure 4 is a useful way to show that the transaction cost of the liquidating funds significantly
increases in the presence of predatory trading. At the trough of price impact, fifth day of the fire
sale, the cumulative return is about -105 bps with aware brokers and about -75 bps in case of
unaware brokers, i.e. the case in which we conjecture that no leakage occurs, a 40% increase.
These results speak to the role of information leakage in exacerbating fire sales.26

4.3 Persistence in Broker-Manager Relationships

One could wonder why the liquidating funds do not better hide their trades to avoid this higher
price impact. There are several non-mutually exclusive explanations. First, the evidence suggests
that in fact they try to hide their trades as they tend to employ an average of about 29 brokers to
intermediate these trades. Second, the funds are most likely in a rush to liquidate, which makes
them prioritize execution speed over price impact. For the same reason, they are likely to rely on
familiar brokers, as opposed to search for other brokers, which can take time. Third, there is a
significant amount of stickiness in the trading relationships between brokers and their clients. The

26
In Appendix Table A11, using the full Ancerno sample, we compute the characteristics of managers that during fire sale events
trade with aware/unaware brokers. The table shows that managers trading with aware brokers during a fire sale, in general, generate
more trading volume (row 1). Accordingly, they have more relations with brokers (row 2). However, they use a significantly lower
number of brokers per million dollars that they trade (row 3). This fact suggests that, typically, the managers that face predation
(i.e. those that turn to aware brokers) are less in the habit of splitting their volume across multiple brokers. Moreover, we look at
the commissions that are generated by the managers. Given that they trade more, the managers that deal with aware brokers pay in
general more commissions (row 4). In terms of the repartition of their commission to different brokers, we find that the first category
of managers pay more commissions to aware brokers than to unaware brokers, even outside of fire sale events, both per dollar
traded (row 5) and in total dollar amounts (row 6). Using Ancerno, Goldstein, Irvine, Kandel, and Wiener (2009) classify broker-
manager relationships into premium and discount. The evidence in the table suggests that managers that are predated are more
likely to have a premium relationship with aware brokers. Overall, it appears that managers that end up being predated have a more
focused relationship with brokers. In particular, they appear to be premium clients who entrust the aware brokers with a larger
fraction of their traders. Possibly, these predated clients are trading off the risk of being predate against the advantages in terms of
information generation, IPO allocation, etc., which originate from being a premium customer of a broker (Goldstein, Irvine, Kandel,
and Wiener, 2009).

30
autocorrelation of our relationship measures at the monthly frequency is above 90%. Table A13
reports the autocorrelation of various measures of concentration, such as the number of brokers
and the Herfindahl index, both on average and during fire sales events. The relevant finding is that
indeed there is significant persistence in the concentration of asset managers’ trades among
brokers. This result holds both when managers are seeking liquidity, i.e. during fire sales, and
when they are not. It appears, therefore, that managers find it difficult to start interacting with new
brokers, i.e. building new relationships with brokers, at the time when a timely execution of their
trades is needed to meet investors’ redemptions demands.

4.4 Quid Pro Quo

Another natural question is whether brokers gain from leaking order flow information about their
clients. One might argue that it would be in their best interest to build a reputation as a loyal trading
partner by keeping the order flow information private. On the other hand, brokers have an incentive
to increase the volume they intermediate as they are paid on commissions. We can address this
question by exploiting the granularity of our data and testing whether best clients tend to reward
the brokers by channeling more trades to them. In Table 9 we regress the average
H(>><GG<(= f43 ;())237,+,6,* paid by manager m to broker b during month t, defined as the ratio
of the total amount in dollars paid in commissions and the total dollar volume traded by manager
m and intermediated by broker b in that month, on the interaction of the dummy variable
identifying the two years following the fire sale event with each of our Best Clients proxies. We
find that the clients who are more likely to receive order flow information tend to increase their
commissions to the brokers, which strongly suggests a quid pro quo between these parties.27

27
To get a sense of magnitude, a top-decile client (for which the Best Client proxy based on volume is 0.58) after the fire sale event
pays 16% of a standard deviation higher commissions per dollar relative to smaller clients. In Table A14 of the Internet Appendix
we show that this estimate increases to 24% when we restrict to mangers which generated the highest profits during the event.

31
5 Conclusion

This paper studies whether brokers’ incentives to attract and retain clients crucially induce sharing
of order flow information with other market participants. The evidence suggests that brokers tend
to reveal the occurrence of a fire sale to their best clients, allowing them to generate significant
profits by predating on the liquidating fund. Furthermore, this information leakage comes at the
expense of higher price impact, and leads to a more costly liquidation for the fire sale originator.
These findings have implications for academics, practitioners, and policy makers. First, our
results indicate an important cost associated with slow execution. Slow execution has been widely
advocated by academics as a way to minimize price impact since Kyle (1985) and routinely
implemented by practitioners. In fact, according to our results, executing large trades over multiple
days allows the brokers to forecast order flow and to trigger predatory behavior by other market
participants. This might adversely affect price impact.
Information leakage might be a source of concern for regulators as well, since it might
exacerbate the costs associated with fire sales, especially at times of scarce liquidity. Regulations
are unclear on what type of information the brokers can and cannot share with their clients. As
pointed out by Fox, Glosten, and Rauterberg (2015), a broker has a legal duty not to use knowledge
of a client’s order to its own advantage. Hence, a broker trading on its own behalf, before the
client’s order reaches the market, violates such duty.28 By extension of this notion, even if a broker
receives an indirect benefit from leaking order flow information to a third party, the broker is in
violation of the aforementioned legal duty.29
This said, brokers can always argue that information leakage occurs while they are still
respecting the fiduciary duty of best execution vis à vis their clients. In particular, information may
leak while the brokers search for counterparties for a large trade. Irrespective of the fraudulent

28
This situation describes front running, which is prohibited under common law (e.g. Opper v. Hancock), federal law (Section
10(b) of the Exchange Act and Rule 10b-(5)), and industry self-regulatory standards (FINRA Rule 570(a)).
29
Moreover, a broker passing information concerning a customer order to a third party is violating its agency duties of
confidentiality, as well as provisions of Regulation ATS (if the broker is an operator of an Alternative Trading System, such as a
dark pool), and probably its own marketing material. See RESTATEMENT (THIRD) OF AGENCY (2006) Section 8.05(2); and
Rule 301(b)(10) of Regulation ATS.

32
intentions of the broker, the evidence that we present in the paper shows that liquidity provision is
lower among aware brokers than other brokers. Yet, a regulatory attempt to stop information
leakage is likely to be challenging, because it will have to deal with the brokers’ need to operate
as deal-makers, as well as with the reluctance of many asset managers to disclose more information
about their trading activities.
Finally, our results shed light on a recent debate over the exchanges’ and brokers’ use of their
access to market data to sell data products.30 Critics maintain that institutional investors, who
routinely need big trades to be executed anonymously, can be negatively impacted as these
products could be used to “reverse-engineer” their strategies and lead to front-running. Our
findings show that, even in the absence of such supplemental information, a number of large
investors, who entertain a strong business relation with brokers, are able to exploit order flow
information at the expense of those seeking liquidity provision. Our estimates might serve as a
benchmark, and probably a lower bound, for the costs associated with releasing such data
products.31
A fruitful avenue for further research is to build upon the insights of this paper towards a more
articulated theory of how the relationship between asset managers and intermediaries, such as
brokers, affects trading behavior and asset prices. Specifically, one could structurally estimate how
the flow of information diffuses among market participants and address questions about the
efficiency of such strategic behavior by the brokers for price discovery and asset allocation, as
well as providing insights into the counterfactual results in the presence of new regulations aimed
at curbing this practice.

30
The most recent dispute involves NASDAQ seeking the SEC’s approval for an options-data service called the “Intellicator
Analytic Tool.” This new service would provide market color to subscribers by revealing whether a trade was initiated by a small
investor or a big money manager. This story was reported in a recent WSJ article “Wall Street Fears Nasdaq Proposal Would
Expose Trading Secrets” (available at https://www.wsj.com/articles/could-the-intellicator-spill-the-markets-secrets-
1510223403?tesla=y#comments_sector).
31
Our results also highlight the importance of the fiduciary duty between broker-dealers and their clients. A few states in the U.S.
are moving in the direction of tightening such duty for brokers. For instance, Nevada is considering an expanded interpretation of
fiduciary duty in which the brokers would be required to “disclose to a client, at the time advice is given, any gain [the broker] may
receive, such as profit or commission, if the advice is followed.”

33
References
Acharya, Viral V., Douglas Gale, and Tanju Yorulmazer, 2011, Rollover risk and market freezes, Journal
of Finance 66(4): 1177-1209.
Admati, Anat R., and Paul Pfleiderer, 1991, Sunshine trading and financial market equilibrium, Review of
Financial Studies 4.3 (1991): 443-481.
Afonso, G., Kovner, A. and Schoar, A., 2013. Trading partners in the interbank lending market, FRB of
New York Staff Report, (620).
Allen, Franklin, and Douglas Gale, 1994, Limited Market Participation and Volatility of Asset Prices,
American Economic Review, 84(4): 933–55.
Amihud, Yakov, 2002, Illiquidity and stock returns: cross-section and time-series effects, Journal of
Financial Markets, 5(1): 31-56.
Anand, Amber, Paul Irvine, Andy Puckett, and Kumar Venkataraman, 2012, Performance of institutional
trading desks: an analysis of persistence in trading costs, Review of Financial Studies, 25, 557-598.
Anand, Amber, Paul Irvine, Andy Puckett, and Kumar Venkataraman, 2013, Institutional trading and stock
resiliency: evidence from 2007-2009 financial crisis, Journal of Financial Economics, 108, 773-793.
Anton, Miguel, and Christopher Polk, 2014, Connected Stocks, Journal of Finance 69(3), 1099–1128.
Babus, A. and Kondor, P., 2016, Trading and information diffusion in over-the-counter markets, Working
Paper, London School of Economics.
Bessembinder, H., Carrion, A., Tuttle, L., & Venkataraman, K. (2016). Liquidity, resiliency and market
quality around predictable trades: Theory and evidence. Journal of Financial Economics, 121(1), 142-
166.
Biais, Bruno, and Foucault, Thierry, HFT and market quality, Bankers, Markets & Investors 128.1 (2014):
5-19.
Brown, David B., Bruce Ian Carlin, and Miguel Sousa Lobo, 2010, Optimal Portfolio Liquidation with
Distress Risk. Management Science 56, 1997–2014.
Brunnermeier, Markus K., and Lasse Heje Pedersen, 2005, Predatory trading, Journal of Finance 60(4),
1825-1863.
Brunnermeier, Markus, and Lasse H. Pedersen, 2009, Market Liquidity and Funding Liquidity, Review of
Financial Studies, 22(6): 2201–38.
Busse, Jeffrey A., Amit Goyal, and Sunil Wahal, 2010, Performance and persistence in institutional
investment management, Journal of Finance 65(2): 765-790.
Coval, J. and Stafford, E., 2007, Asset fire sales (and purchases) in equity markets, Journal of Financial
Economics, 86(2), pp.479-512.
Chung, J.W. and Kang, B.U., 2016. Prime Broker-Level Comovement in Hedge Fund Returns: Information
or Contagion?, Review of Financial Studies 29.12 (2016): 3321-3353.
Di Maggio, Marco, 2016, Market turmoil and destabilizing speculation, Columbia Business School
Research Paper No. 13-80.
Di Maggio M., Kermani A., Song Z., 2017, The value of trading relations in turbulent times, Journal of

34
 Financial Economics, Volume 124, Issue 2, Pages 266-284.
Di Maggio, Marco, Franzoni, Francesco, Kermani, Amir, and Sommavilla, Carlo, 2018, The relevance of
broker networks for information diffusion in the stock market, Journal of Financial Economics,
forthcoming.
Di Mascio, Rick and Lines, Anton and Naik, Narayan Y., 2016, Alpha decay and Strategic trading, Working
Paper, Columbia Business School.
Duffie, Darrell, Semyon Malamud, and Gustavo Manso, 2009, Information percolation with equilibrium
search dynamics. Econometrica, 77(5): 1513-1574.
Duffie, Darrell, Semyon Malamud, and Gustavo Manso, 2014. Information percolation in segmented
markets, Journal of Economic Theory, 153, pp.1-32.
Eisele, Alexander, Nefedova, Tamara, Parise, Gianpaolo, and Peijnenburg, Kim, 2017, Trading Out of
Sight: An Analysis of Cross-Trading in Mutual Fund Families, Centre for Economic Policy Research
Working Paper.
Ellul, Andrew, Chotibhak Jotikasthira, and Christian T. Lundblad, 2011, Regulatory pressure and fire
sales in the corporate bond market, Journal of Financial Economics 101(3): 596-620.
Farboodi, Maryam, and Laura Veldkamp, 2017, Long run growth of financial technology, No. w23457,
National Bureau of Economic Research.
Fox, Merritt B., Lawrence R. Glosten, and Gabriel V. Rauterberg, 2015, The New Stock Market: Sense and
Nonsense, Duke LJ 65 (2015): 191.
Franzoni, Francesco, and Alberto Plazzi, 2015, Do hedge funds provide liquidity? Evidence from their
trades, Swiss Finance Institute Working Paper.
Garleanu, Nicolae, and Lasse Heje Pedersen, 2011, Margin-based asset pricing and deviations from the law
of one price, Review of Financial Studies 24(6): 1980-2022.
Garleanu, Nicolae, and Lasse Heje Pedersen, 2013, Dynamic trading with predictable returns and
transaction costs, Journal of Finance 68(6): 2309-2340.
Gaspar, Jose-Miguel, Massimo Massa, and Pedro Matos, 2006, Favoritism in mutual fund families?
Evidence on strategic cross-fund subsidization, Journal of Finance 61, 73-104.
Greenwood, Robin, Augustin Landier, and David Thesmar, 2015, Vulnerable banks, Journal of Financial
Economics 115(3): 471-485.
Gromb, Denis, and Dimitri Vayanos, 2002, Equilibrium and welfare in markets with financially constrained
arbitrageurs, Journal of Financial Economics, 66(2–3): 361–407.
Goldstein, Michael A., Paul Irvine, Eugene Kandel, and Zvi Wiener, 2009, Brokerage commissions and
institutional trading patterns, Review of Financial Studies 22(12), 5175-212.
Hendershott, T., Li, D., Livdan, D. and Schürhoff, N., 2016, Relationship trading in OTC markets, Swiss
Finance Institute Working Paper.
Hollifield, Burton, Neklyudov, Artem and Spatt, Chester S., 2016, Bid-Ask spreads, trading networks and
the pricing of securitizations, Review of Financial Studies, forthcoming.
Irvine, Paul, Marc Lipson, and Andy Puckett, 2006, Tipping, Review of Financial Studies 20(3): 741-768.
Keim, Donald B., and Ananth Madhavan, 1997, Transactions costs and investment style: an inter-exchange

35
 analysis of institutional equity trades, Journal of Financial Economics 46(3): 265-292.
van Kervel, Vincent and Menkveld, Albert J., 2016. High-Frequency trading around large institutional
orders, Journal of Finance, forthcoming.
Kisin, R. 2011, The impact of mutual fund ownership on corporate investment: Evidence from a natural
experiment, Working Paper, Washington University in St. Louis.
Kyle, Albert S., 1985, Continuous auctions and insider trading. Econometrica, 1315-1335.
Kiyotaki, Nobuhiro, and John Moore, 1997, Credit cycles, Journal of political economy 105.2 (1997): 211-
248.
Li, Dan and Schürhoff, Norman, 2018, Dealer Networks, Journal of Finance, forthcoming.
McNally, William J., Andriy Shkilko, and Brian F. Smith, 2015, Do brokers of insiders tip other clients?
Management Science 63(2): 317-332.
Menkveld, Albert J., 2016, The economics of high-frequency trading: Taking stock, Annual Review of
Financial Economics 8 (2016): 1-24.
O’Hara, Maureen, 2015, High frequency market microstructure, Journal of Financial Economics 116.2
(2015): 257-270.
Puckett, Andy, and Xuemin (Sterling) Yan, 2011, The interim trading skills of institutional investors,
Journal of Finance, 66, 601-633.
Ross, Adrienne and Moskowitz, Tobias J. and Israel, Ronen and Serban, Laura, 2017, Implementing
Momentum: What Have We Learned?
Scholes, Myron 2000, Crisis and risk management, American Economic Review 90,17–21.
Seppi, D.J., 1990, Equilibrium block trading and asymmetric information, Journal of Finance, 45.1, 73-94.
Shleifer, Andrei, and Vishny, Robert W., 1992, Liquidation values and debt capacity: A market equilibrium
approach, Journal of Finance 47.4 (1992): 1343-1366.
Shleifer, Andrei, and Vishny, Robert W., 1997, A survey of corporate governance, Journal of Finance,
52(2), 737-783.
Shleifer, Andrei, and Vishny, Robert W., 2011, Fire sales in finance and macroeconomics, Journal of
Economic Perspectives 25(1), 29-48.
Walden, Johan, 2016, Trading, profits, and volatility in a dynamic information network model, Working
Paper, University of California, Berkeley.
Yang, Liyan, and Haoxiang Zhu, 2016, Back-running: Seeking and hiding fundamental information in
order flows, Rotman School of Management Working Paper 2583915.

36
Figure 1: Liquidation Volume and Price Pattern. The figure plots the average daily signed volume (i.e. order imbalance) of the
fire sale originator on the fire sale stocks, expressed in Million Dollars.

Figure 2: Price Pattern. The figure plots the average DGTW-adjusted cumulative returns for the stocks sold during the fire sales
along with 95% confidence bands.

37
Figure 3: Order Flow through Aware Brokers. The figure plots the cumulative order imbalance of the transactions intermediated
by the aware brokers (green solid line) and unaware brokers (red dashed line) on the fire sale stocks, excluding those generated by
the liquidating funds. The daily order imbalance computed as the difference of the volume of buy and sell orders divided by the
total absolute volume. The measure is then averaged across fire sales in event time.

Figure 4: Price Paths with and without Information Leakage. The figure plots the cumulative return of the fire sale stocks
during fire sale events involving at least one aware broker. The red line with squares represents the cumulative return averaged
across stocks and events in which aware brokers are present. The green line with circles represents the cumulative return averaged
across stocks and events in which no aware brokers are present. The series are based on estimates from a regression specification
similar to the one reported in Table 8, but run on daily observations.

38
Figure 5: Profitability of Predatory Trades: Liquidation Events (left) and Placebo Sample (right). The left panel of the figure
plots the profits of the managers that are best clients of the aware (green solid line with circles) and unaware (red dashed line with
squares) brokers during the fire sale events. Best clients of a broker are defined as the managers generating more than 5% of the
volume intermediated by that broker in the previous semester. The right panel repeats the exercise for random event windows other
than the actual fire sales employed in the analysis.

39
 Table 1
Summary Statistics
In Panels A, B, and C, we report summary statistics for the volume Z-score and the 385 fire sale events identified by our
methodology. In Panel D we report summary statistics for the manager-broker relationship proxies employed in the paper,
expressed in percentage units. To identify fire sale events, we start by computing the signed volume Z-score g*+ for each manager
m on day t as Z"# = ( &'()*+ – - &'()*+ )/. &'()*+ , where &'()*+ is the portfolio level dollar volume traded by manager m
on day t, and its mean and standard deviation are estimated over a rolling window of 120 trading days ending one week before day
t. Then, at the portfolio level, we define manger > as liquidating if g*+ is below -0.25 for at least 5 trading days in a row. Next, we
impose a filter at the stock level: for stock j to enter the fire sale basket we require that the volume traded by the manager is above
1% of the CRSP daily volume for at least 4 of the fire sale days. Finally, we keep events in which at least 10 stocks are sold by the
liquidating fund. Standard errors are clustered at the event level. In Panel E we regress the amount sold of each stock as a fraction
of the total fire sale volume on a set of stocks characteristics, while in Panel F we regress the first day in which each stock is sold
the first time by the liquidating fund, in event time, on the same set of stocks characteristics. T-statistics are reported in parentheses.
Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel
PanelA:
A -Volume Z-Score
Volume Z-Score
Obs Mean S.D. Min 0.25 0.5 0.75 Max
All Managers-Days 941219 -0.035 3.249 -41.714 -0.369 0.027 0.394 35.889
Fire Sales Days 2210 -2.075 4.518 -41.714 -1.768 -1.038 -0.616 -0.251
Fire Sales Events 385 -2.002 3.410 -37.818 -1.672 -1.172 -0.878 -0.344

Panel B: Fire Sale Events

Unit Obs Mean S.D. 25% 50% 75% 90%
Dollar Volume Million Dollars 385 -377.062 534.635 -503.571 -177.461 -50.544 -18.244
Fraction of Portfolio Percentage 385 9.164% 23.921% 1.224% 2.274% 5.879% 15.828%
Number of Stocks 385 21.917 10.090 13 18 29 38
Event Length Trading Days 385 5.766 1.439 5 5 6 7
Number of Brokers 385 28.803 16.095 18 27 39 52
Number of Aware Brokers 385 1.694 0.968 1 2 2 3

Pure
Panelanel C:Sale
C: Fire FireStocks
Sale Stocks
Unit Obs Mean S.D. 25% 50% 75% 90%
Dollar Volume Million Dollars 8438 -17.204 20.305 -23.401 -11.246 -3.542 -1.366
CRSP volume ratio Percentage 8438 -14.576% 16.000% -18.749% -9.922% -4.585% -2.409%
Price Decrease in [0,4] Percentage 8438 0.831% 4.613% -1.904% 0.666% 3.388% 7.131%
Number of Brokers 8438 5.737 5.039 2 4 8 13
Number of Aware Brokers 8438 0.522 0.603 0 0 1 1

Panel D: Manager-Broker Relashionship Proxies

Obs Mean S.D. Min 25% 50% 75% 90% Max

Ranking based on Volume 501568 0.035 0.079 0.000 0.000 0.004 0.031 0.101 0.965

Ranking based on Commission Paid 501568 0.032 0.071 0.000 0.000 0.005 0.032 0.088 0.924

40
Panel D: Within Fire Sale Basket Panel E: Fire Sale Stocks Selection
Dependent Variable Amount Sold as a Fraction of the Fire Sale

(1) (2) (3) (4) (5) (6)

Portfolio Weight 1.863*** 1.830*** 1.319*** 1.805*** 1.301*** 1.318***

(6.522) (6.427) (5.875) (6.540) (5.815) (5.842)
Amihud Ratio -0.691*** -0.486*** -0.506***
(-8.419) (-6.579) (-6.775)
Market Cap 2.614*** 2.427*** 2.441***
(11.580) (10.926) (10.977)
Volatility -6.698*** -3.838*** -3.394***
(-12.549) (-7.296) (-6.438)
One Month Return 0.112
(0.981)
Six Months Return 0.209*
(1.741)
One Year Return 0.340***
(2.783)

Observations 7,948 7,948 7,948 7,948 7,948 7,948

R-squared 0.134 0.142 0.237 0.164 0.253 0.257
Time FE Yes Yes Yes Yes Yes Yes
Manager FE Yes Yes Yes Yes Yes Yes
Event FE Yes Yes Yes Yes Yes Yes

Panel E: Timing
Panel F: Fire Sale Stocks Timing
Dependent Variable First Day In Which The Stock Is Sold

(1) (2) (3) (4) (5) (6)

Portfolio Weight -0.034*** -0.033*** -0.026*** -0.033*** -0.025*** -0.025***

(-3.862) (-3.732) (-3.218) (-3.831) (-3.128) (-3.184)
Amihud Ratio 0.028*** 0.025*** 0.025***
(4.008) (3.515) (3.530)
Market Cap -0.040*** -0.036*** -0.037***
(-5.736) (-5.233) (-5.319)
Volatility 0.113*** 0.055 0.050
(2.982) (1.443) (1.311)
One Month Return 0.011
(1.228)
Six Months Return 0.002
(0.189)
One Year Return -0.018*
(-1.785)

Observations 7,948 7,948 7,948 7,948 7,948 7,948

R-squared 0.209 0.211 0.213 0.211 0.215 0.215
Time FE Yes Yes Yes Yes Yes Yes
Manager FE Yes Yes Yes Yes Yes Yes
Event FE Yes Yes Yes Yes Yes Yes

41
 Table 2
Predatory Behavior and Broker Awareness
The table reports results on the likelihood of a broker to attract predatory trades. The regressions are run at the ticket-level, excluding
trades by managers originating the fire-sale of interest or another overlapping fire-sale. In Columns (1)-(2) of Panel A the dependent
variable is the difference between a dummy indicating predation and a dummy indicating liquidity provision, i.e. it takes value one
when the trade is in the same direction of the volume by the liquidating fund for that stock on that day (i.e. it is a sell trade), it
equals negative one if the trade is in the opposite direction (i.e. a buy trade) and equals zero if the manager is not trading that stock
on that particular day. In Columns (3)-(4) of Panel A we multiply the above described dependent variable by the volume of the
trade as a fraction of market capitalization, standardized. The independent variable Aware is a dummy, defined at the event-broker-
stock-day level, indicating that the broker is aware of the fire sale happening on the traded stock on that day. Precisely, this means
that for broker B, stock j on day t of the fire sale event e broker b intermediates transactions on stock j from the liquidating fund
originating e amounting to a volume which is above 2% of the average daily volume of stock j. In Panel B we focus only on
predation, using the above defined predation dummy as dependend variable in columns columns (1)-(2) and its volume-weighted
counterpart in columns (3)-(4). In Panel C we focus only on liquidity provision, using the above defined liquidity provision dummy
as dependend variable in columns columns (1)-(2) and its volume-weighted counterpart in columns (3)-(4). All specifications
include date, manager, and broker fixed-effects. We further add broker-stock fixed effects in odd-numbered columns and day-stock
fixed effects in even-numbered columns. Standard errors are clustered at the broker level. T-statistics are reported in parentheses.
Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel A: Net Predation

Probability of Predation - Predatory Volume -
Dependent Variable
Probability of Liquidity Provision Liquidity Provision Volume

(1) (2) (3) (4)

Aware Dummy 0.202*** 0.113*** 0.039*** 0.027**

(7.142) (5.199) (3.080) (2.323)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes
Brokers ⨉ Stock FEs Yes Yes
Day ⨉ Stock FEs Yes Yes

Observations 487,605 462,841 487,605 462,841

R-squared 0.203 0.229 0.136 0.159

42
 Panel B: Predation
Dependent Variable Probability of Predation Predatory Volume

(1) (2) (3) (4)

Aware Dummy 0.103*** 0.124*** 0.062*** 0.049***

(7.469) (11.399) (4.578) (4.065)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes
Brokers ⨉ Stock FEs Yes Yes
Day ⨉ Stock FEs Yes Yes

Observations 487,605 462,841 487,605 462,841

R-squared 0.360 0.274 0.189 0.193

Panel C: Liquidity Provision

Dependent Variable Probability of Liquidity Provision Liquidity Provision Volume

(1) (2) (3) (4)

Aware Dummy -0.099*** 0.011 0.022** 0.024**

(-6.806) (0.716) (2.483) (2.448)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes
Brokers ⨉ Stock FEs Yes Yes
Day ⨉ Stock FEs Yes Yes

Observations 487,605 462,841 487,605 462,841

R-squared 0.369 0.263 0.155 0.173

43
 Table 3
Probability of Predation and Broker-Client Relationship Strength
The table presents evidence of the effect of broker-client relationship strength on the probability of predatory behavior. The
regressions are run at the stock-day-manager-broker level, excluding trades by managers originating the fire-sale of interest or
another overlapping fire-sale. In all specifications the dependent variable is the difference between a dummy indicating predation
and a dummy indicating liquidity provision, i.e. it takes value one when the trade is in the same direction of the volume by the
liquidating fund for that stock on that day (i.e. it is a sell trade), it equals negative one if the trade is in the opposite direction (i.e. a
buy trade) and equals zero if the manager is not trading that stock on that particular day. In Panel A we regress the dependent
variable on the continuous variable Best Client, measuring the strength of the manager-broker relation, the dummy Liquidation
Period, indicating the first 5 days of the fire sale, and the interaction of the two. The relationship strength variables are definined
as follows: (i) RVol#,n," is the fraction of dollar volume intermediated by broker b in the semester preceeding day t which is
generated by manager m and (ii) RCom#,n," is the fraction of dollar commissions earned by broker b in the semester preceeding
day t which is generated by manager. Both variables are expresse in decimal units and thus take values in the interval [0,1]. The
dummy Liquidation Period is zero in the five days before the fire sale. We consider all trades on stock j intermediated by brokers
65
that eventually become aware that the stock is subject to fire sale pressure, i.e. brokers b for which >2q*∈ s,P (012F3(* ) = 1,
65
where 012F3(* is defined as above. The regression is run on a sample that includes five days before the fire sale and five days
from the start of the fire sale, defined as the first day in which our liquidation measure crosses the threshold. In Panel B we regress
the dependent variable on the triple interaction of the following variables: Aware Broker indicating that the broker is aware, Best
Client, and Liquidation Period indicating the first 5 days of the fire sale. Time fixed effects are at the monthly frequency. Standard
errors are clustered at the event-stock-manager level. T-statistics of the differences are reported in parentheses. Asterisks denote
significance levels (***=1%, **=5%, *=10%).

Panel A: Difference in Differences

Probability of Predation Predatory Volume
Dependent variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)
Ranking based on Ranking based on Ranking based on Ranking based on
Best clients proxy
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.055*** 0.081*** 0.124*** 0.127***

(3.181) (4.182) (3.236) (2.806)
Best Client 0.023 0.048** -0.001 0.003
(1.427) (2.500) (-0.038) (0.096)
Liquidation Period 0.006*** 0.005*** 0.002 0.003
(5.683) (4.942) (0.618) (0.661)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 501,567 501,567 501,567 501,567

R-squared 0.046 0.046 0.013 0.013

44
 Panel B: Triple Interaction
Dependent variable Probability of Predation - Probability of Liquidity Provision
(1) (2) (3) (4)
Best clients proxy Ranking based on Volume Ranking based on Ranking based on Volume Ranking based on
Commissions Paid Commissions Paid

Aware Broker ⨉ Best Client ⨉ Liquidation Period 9.814*** 11.589*** 9.184*** 10.794***
(4.527) (4.927) (4.356) (4.756)
Best Client ⨉ Liquidation Period 1.752*** 1.287*** 1.554*** 1.157***
(5.664) (5.220) (4.989) (4.634)
Aware Broker ⨉ Liquidation Period 0.011*** 0.011*** 0.011*** 0.011***
(10.252) (9.552) (10.196) (9.578)
Best Client ⨉ Aware Broker 7.466*** 9.082*** 7.500*** 8.949***
(3.893) (4.018) (3.755) (3.750)
Best Client 3.747*** 2.969*** 3.664*** 2.931***
(18.986) (14.932) (13.956) (10.602)
Aware Broker 0.006*** 0.005*** 0.003*** 0.003***
(7.138) (5.686) (4.035) (3.550)
Liquidation Period 0.012*** 0.012*** 0.011*** 0.011***
(39.642) (41.789) (34.415) (36.251)
Constant 0.021*** 0.022***
(99.348) (100.764)

Time Fixed Effects Yes Yes

Manager Fixed Effects Yes Yes
Event Fixed Effects Yes Yes
Stock Fixed Effects Yes Yes
Broker Fixed Effects Yes Yes

Observations 4,226,877 4,226,877 4,128,803 4,128,803

R-squared 0.005 0.004 0.022 0.022

45
 Table 4
Evidence of Predation on Multiple Stocks
The table reports results on the number of stocks experiencing predatory pressure. For each fire sale event, we consider the basket
of liquidated stocks, and for each manager actively trading at least one stock in the basket we count the number of stocks traded in
the same direction of the fire sale originator. In the first two specifications, we consider event-manager observations and we regress
the number of predated stocks on best client proxies. These are constructed by interacting the original best client proxies with the
broker awareness dummy at the ticket-level, and then by taking the maximum value at the event-manager level. In other words the
relationship strength assigned to each manager is the value of the best relationship across the aware brokers in the fire sale event.
Then, the number of predated stocks is calculated considering all of the fire sale stocks predated by the manager across all brokers.
In specification (3)-(4), we repeat the exercise by adopting as dependent variable the fraction of predated stocks relative to the
stocks in the fire sale basket. Event, manager and day fixed effects are included in the regressions and standard errors are double
clustered at the manager and event level. T-statistics are reported in parentheses. Asterisks denote significance levels (***=1%,
**=5%, *=10%).

Dependent variable Number of Predated Stocks Fraction of Predated Stocks

(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client 14.551*** 15.066*** 57.855*** 59.791***

(4.864) (4.516) (5.148) (4.855)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes

Observations 28,168 28,168 28,168 28,168

R-squared 0.390 0.386 0.465 0.461

46
 Table 5
Predators’ Position Reversal
The dependent variable is the fraction of sales in a given stock that a given manager subsequently reverses. In detail, in a given
time period, either before or after the beginning of the fire sale, the percentage of position reversed for manager m during event e
for stock j is defined as the ratio U4V7,+,5 = F(LWℎ9F2YZ7,+,5 / \();],#,^ , where \();],#,^ is the dollar sum of all sell orders in
that period, and F(LWℎ9F2YZ7,+,5 is the dollar sum of buy orders during the period, where we sum only the buy orders that are
preceded by a negative cumulative order flow. We compute this measure around each fire sale event, for the event time periods
:34 = [−10, −1] and :(G9 = [0,9], considering all trades on stock j intermediated by brokers who eventually become aware that
the stock is subject to fire sale pressure. We then compare the percentage of position reversed across clients with different
relationship strength with the aware brokers before (:34) and during (:(G9) the fire sale events. Liquidating funds are excluded
from the sample. In columns (1)-(2) we present results for the specifications without fixed effects, while in columns (3)-(4) we
report results with time, stock, and manager fixed effects. Standard errors are clustered at the manager level. T-statistics are reported
in parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Dependent variable Percentage of Positions Reversed

(1) (2) (3) (4)

Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Post[0,9] 25.091*** 24.110*** 23.352*** 20.676***

(11.788) (7.413) (5.404) (4.151)
Best Client 5.128*** 6.448*** -7.022** -1.939
(3.441) (2.791) (-2.010) (-0.526)
Post[0,9] 11.427*** 12.764*** 16.287*** 18.320***
(17.298) (19.318) (14.256) (15.494)
Constant 2.723*** 2.878***
(5.788) (6.116)

Time Fixed Effects Yes Yes

Stock Fixed Effects Yes Yes
Manager Fixed Effects Yes Yes

Observations 37,276 37,276 31,000 31,000

R-squared 0.028 0.023 0.258 0.256

47
 Table 6
Evidence from the 2003 Mutual Fund Scandal
We first match the list of 27 mutual fund families involved in the 2003 late-trading scandal with managers in our dataset and mark
them as charged. We focus on daily transactions of the managers that are not involved in the scandal for a period of four years
centered on the month of the announcement of the complaint by Spitzer (September 2003) and define a dummy :(G9 \Y2=;2)* ,
indicating the two years after the complaint broke out. Next, we define a broker-stock-day level dummy variable, \4))<=W6,5,* ,
indicating that at least one of the charged funds is selling stock j on day t through broker b. Then, we define the dependent variable
Probability of Predation as a dummy variable that equals 1 if a non-charged manager is selling stock j on day t through broker b.
The dependent variable equals 0 if a non-charged manager trades on a different day, or on a different stock, or with a different
broker. We then regress the probability of predation minus the probability of liquidity provision (defined as in Table 3) on the
interaction between \4))<=W6,5,* and the dummy :(G9 \Y2=;2)* . Standard errors are clustered by manager-stock to and T-statistics
are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Dependent variable Probability of Predation - Probability of Liquidity Provision

(1) (2) (3) (4) (5)

Selling ⨉ Post Scandal 0.087*** 0.097*** 0.069*** 0.060*** 0.046***

(11.406) (12.800) (9.261) (8.220) (6.342)
Selling 0.147*** 0.141*** 0.148*** 0.152*** 0.179***
(23.040) (22.135) (22.406) (23.537) (28.281)
Post Scandal -0.025*** 0 0 0 0
(-9.289) 0 0 0 0

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes
Stock Fixed Effects Yes Yes
Broker Fixed Effects Yes

Observations 12,087,004 12,087,004 12,087,001 12,086,863 12,086,781

R-squared 0.001 0.013 0.068 0.076 0.082

48
 Table 7
Profitability of Predatory Trades
The table reports results on the profitability of trades by predators around the fire sales events. We divide each event into a pre-fire
sale period [−10, −1] and a post-fire sale period [0,9], where zero denotes the day on which the fire sale starts. We then compute
the profitability of trades by manager m on stock j over the window t = 9s , 9E , which denotes either the pre or post fire sale
period. The profitability measure which we use as dependent variable in all spefications is defined by the following formula
:3(_<92?<)<9u+,5,v = w23Zd(w23Z49+,5,v − H2Gℎx)(1G+,5,v /-qf(GL34+,5,v .
Here, w23Zd(w23Z49+,5,v is the marked-to-market dollar value of the position at time 9E , defined as the product of the share
position cumulated from 9s to 9E with the market price of stock j on day 9E . H2Gℎx)(1G+,5,v is the dollar amount spent to build the
position, i.e. the opposite of the dollar volume of each transaction in the stock (based on execution prices) from from 9s to 9E .
-qf(GL34+,5,v is the maximum dollar outlay over the relevant period, defined as max "∈{ H2Gℎx)(1G+,5, *| ,* . We relate the
profitability (expressed in basis points) of trades executed by aware brokers to our relationship strength proxies (i.e. the fraction of
the volume intermediated by the broker over the previous semester generated by the manager, expressed in decimal units, as well
as the fraction of the commissions) in the pre- and post- fire sale periods, using event-manager-stock level observations. In rows
(1)-(2) we present results for the specifications without fixed effects, while in rows (3)-(4) we report results with time, stock and
manager fixed effects. Standard errors are clustered at the manager level. T-statistics are reported in parentheses. Asterisks denote
significance levels (***=1%, **=5%, *=10%).

Dependent variable Return on Capital (basis points)

(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Post[0,9] 136.558** 144.821** 147.847** 159.582**

(2.355) (2.235) (2.110) (1.966)
Best Client -48.574 -61.826 -78.883** -108.693***
(-1.145) (-1.303) (-2.414) (-2.929)
Post[0,9] -7.160*** -7.102*** -7.719** -7.665**
(-2.783) (-2.761) (-2.520) (-2.503)
Constant 8.646*** 8.697***
(4.651) (4.679)

Time Fixed Effects Yes Yes

Manager Fixed Effects Yes Yes

Observations 263,346 263,346 263,211 263,211

R-squared 0.000 0.000 0.034 0.034

49
 Table 8
Price Impact and Broker Awareness
This table reports results on the price impact experienced by the fire sale originators. Considering all trades by fire sale originators
from the beginning of each fire sale event (t=0) to the last day of the fire sale (i.e. the last day on which the criteria for a fire sale
definition are satisfied), we construct the following price impact measures: (i) the execution shortfall based on the first placement
price, (ii) the execution shortfall based on the first open price, (iii) the execution shortfall based on the first transaction price. We
aggregate the measures taking their volume-weighted average across transactions and express them in basis points. In specifications
(1)-(3) we regress the price impact measures on a dummy indicating the presence of at least one aware broker at the event-stock
level and the total volume of other managers relative to the stock market capitalization. We control for the originator volume
relative to the stock market capitalization and the Amihud ratio of the stock, estimated on the previous six months. Time and stock
fixed effects are added to the regression. In specifications (4)-(6), we repeat the exercise at the event-stock-broker-level and also
add broker fixed effects. Continuos explanatory variables are standardized and standard errors are clustered by event. T-statistics
are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Dependent variable Price Impact (basis points)

Granularity Stock Level Broker-Stock Level

(1) (2) (3) (4) (5) (6)

Benchmark Price First Placement Market Open First Transaction First Placement Market Open First Transaction
Price Price Price Price Price Price

Aware Broker Dummy 25.176* 36.194** 14.250 11.901*** 11.320** 8.970**

(1.849) (2.503) (1.442) (2.764) (2.217) (2.496)
Followers Volume 23.801*** 24.286*** 8.520* 4.882** 4.898* 2.457
(2.787) (2.710) (1.680) (2.020) (1.815) (1.259)
Generator Volume 6.996 8.560 0.520 21.760*** 20.890*** 11.607**
(0.646) (0.706) (0.067) (3.691) (3.293) (2.449)
Amihud Ratio -19.080 -20.435 -18.598 -12.067 -6.532 -8.238
(-1.070) (-1.101) (-1.382) (-1.262) (-0.703) (-1.437)
Number of Brokers -3.489 1.193 -1.938 6.359 9.710 4.731
(-0.515) (0.152) (-0.373) (1.137) (1.509) (1.334)

Time Fixed Effects Yes Yes Yes Yes Yes Yes

Stock Fixed Effects Yes Yes Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes

Observations 6,291 6,291 6,291 28,265 28,265 28,265

R-squared 0.430 0.430 0.415 0.323 0.338 0.265

50
 Table 9
Commissions Paid to Aware Brokers
The table presents evidence on the post-event increase of commissions paid by predators to aware brokers. For each month t on a
window starting two years before and ending two year after each fire sale event e, we define the average
H(>><GG<(=_f43_;())237,+,6,* paid by manager m to broker b as the ratio H(>>7,+,6,* /&'()7,+,6,* , where H(>>7,+,6,* is the
total amount in dollars paid in commissions by manager m to broker b during month t and &'()7,+,6,* is the total dollar volume
traded by manager m and intermediated by broker b in that month. For each event, we consider brokers which are marked as Aware
on at least one of the fire sale stocks and managers whose trades are intermediated by at least one of these broker in the ten trading
days around the event. We then regress H(>><GG<(=_f43_;())237,+,6,* on the interaction of the dummy variable :(G97,* ,
indicating the two years following the fire sale event, with each of our Best Clients proxies. Standard errors are clustered by event-
broker-manager to account for time-series autocorrelation in commissions paid. T-statistics are reported in parentheses. Asterisks
denote significance levels (***=1%, **=5%, *=10%).

Dependent variable Commissions per dollar (basis points)

(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Post 1.659*** 0.909*** 1.608*** 0.863***

(6.977) (3.657) (7.986) (3.872)
Best Client -13.032*** -9.626*** -5.306*** -1.701***
(-31.531) (-24.702) (-20.886) (-6.460)
Post -0.381*** -0.345*** -0.565*** -0.530***
(-14.035) (-12.774) (-22.084) (-20.803)
Constant 6.296*** 6.126***
(198.908) (195.180)

Time Fixed Effects Yes Yes

Manager Fixed Effects Yes Yes
Broker Fixed Effects Yes Yes

Observations 1,168,535 1,168,535 1,168,521 1,168,521

R-squared 0.029 0.014 0.303 0.301

51


Brokers and Order Flow Leakage:

Evidence from Fire Sales
Andrea Barbon Marco Di Maggio Francesco Franzoni Augustin Landier

INTERNET APPENDIX
 Number of Fire Sale Stocks Total Volume by Originators (in m$)

Figure A1: Number of Stocks and Liquidation Volume. The left panel shows the histogram of events with different number of
stocks involved in the fire sale. The right panel shows the distribution of the total volume executed by the liquidating funds.

Figure A2: Fire Sale Events. The figure plots the number of fire sales events by month. Hollow red squares identify events
happening during the two NBER recessions in our sample period.

1
Figure A3: Predators Persistence. The figure compares the unconditional probability of predation with the probability of
predation conditional on having predated at least once in the previous period. Subsequent periods are defined over weekly, monthly,
quarterly and yearly horizons.

Figure A4: Unwinding of Predatory Positions. The figure plots the average percentage amount of predatory positions built during
the fire sales by predators (managers who trade in the same direction of the liquidating fund during the fire sales) after one day,
one week and one month after the end of the fire sale (green, solid line). The red dotted line displays results from the same exercise
applied to a placebo sample of trades, i.e. sell trades by the same group of predators taking place over a random sample of five-day
intervals that do not include any fire sale.

2
 Table A1
Robustness: Broker Awareness Measures
The table shows the robustness of our main results with respect to different definitions of the broker awareness
measure. We recall that the broker Awareness dummy is defined at the event-broker-stock-day level, indicating that
the broker is aware of the fire sale happening on a given stock-day. We now generalize the definition given in the text,
by requiring the following two conditions to hold for broker b, stock j on day t, event e and given numbers ~ and 8:
(i) the liquidation volume on stock j intermediated by broker b on day t is above ~% of the average daily volume
(ADV) for stock j; (ii) broker b satisfies condition (i) of at least 8 stocks in the fire sale basket. The table presents our
main results for ~ ∈ {1,2,5} and 8 ∈ {1,5,10}, reporting only the estimate and t-statistics for the main coefficient of
interest in the regression (i.e. the one appearing in the first row of the first column in the original table). Standard
errors are clustered as in the corresponding table. T-statistics are reported in parentheses. Asterisks denote significance
levels (***=1%, **=5%, *=10%).

N X=1% X=2% X=5%

Predatory Behavior and Broker 0.203*** 0.195*** 0.168***
1
Awareness (6.414) (5.595) (4.092)
Table 2 - Panel A 0.161*** 0.157*** 0.144*
5
(5.177) (3.951) (1.951)
0.148*** 0.166** 0.191**
10
(2.919) (2.398) (2.088)
Predatory Behavior and Broker 0.038*** 0.043** 0.054**
1
Awareness (2.651) (2.505) (2.242)
Table 2 - Panel A 0.056*** 0.074*** 0.117***
5
(2.982) (2.972) (2.636)
0.060* 0.074 0.131*
10
(1.769) (1.532) (1.836)
Probability of Predation and 0.040*** 0.055*** 0.060***
1
Broker-Client Relationship Strength (2.631) (3.181) (2.690)
Table 3 - Panel A 0.052* 0.090*** 0.165***
5
(1.948) (2.868) (3.678)
0.068* 0.135*** 0.086
10
(1.949) (3.202) (1.313)
Predators’ Position Reversal 11.815*** 11.594*** 10.981***
1
Table 5 (6.763) (6.104) (4.781)
10.888*** 10.107*** 7.045*
5
(4.252) (3.459) (1.707)
10.795*** 13.945*** 12.260**
10
(2.933) (3.108) (1.968)

3
 N X=1% X=2% X=5%
Excluding Negative News 0.044*** 0.052*** 0.050**
1
Table A3 - Panel B (2.657) (2.801) (2.087)
0.050* 0.089*** 0.166***
5
(1.834) (2.804) (3.642)
0.075** 0.137*** 0.092
10
(2.108) (3.166) (1.364)
Excluding Negative Momentum Stocks 0.040** 0.070*** 0.092***
1
Table A3 - Panel C (1.979) (3.016) (3.081)
0.062* 0.103** 0.181***
5
(1.691) (2.383) (2.873)
0.092* 0.227*** 0.201**
10
(1.946) (4.119) (2.297)
Excluding High Short Interest Stocks 0.036** 0.050*** 0.061***
1
Table A3 - Panel D (2.280) (2.849) (2.658)
0.054** 0.101*** 0.177***
5
(2.002) (3.187) (3.915)
0.070** 0.138*** 0.084
10
(1.997) (3.252) (1.282)
Excluding NBER Recessions Periods 0.039** 0.052*** 0.059**
1
Table A3 - Panel A (2.357) (2.816) (2.509)
0.077*** 0.117*** 0.190***
5
(2.745) (3.562) (4.053)
0.085** 0.157*** 0.084
10
(2.340) (3.603) (1.231)

4
 Table A2
Characteristics of Predated Stocks
The table reports characteristics of the fire sale stocks, partitioned into two groups based on degree of predation they are subject
to. More precisely, for each fire sale stock event, we record the number of best clients (defined as those generating at least 5% of
the volume intermediated by the broker over the previous semester) of aware broker P divided by the number N of managers
actively trading during the fire sale event on that stock. The ratio P/N is then used to split the set of fire sale stocks into two parts,
using the median of this variable as cutoff. For each of the two groups we take the average of the following quantities, computed
at the event-stock level: (i) the dollar volume liquidated during the fire sale; (ii) the volume liquidated during the fire sale as a
fraction of the volume recorded in CRSP for that stock; (iii) the first day in which the stock is sold during the fire sale, in event
time; (iv) the weight of the stock in the portfolio of the liquidating fund, reconstructed based on previous transactions; (v) the
Amihud illiquidity ratio of the stock, computed using data from the semester preceding the fire sale; (vi) the market capitalization
of the stock; (vii) the daily return volatility of the stock, estimated using data from the semester preceding the fire sale; (viii) the
cumulating return of the stock during the month preceding the fire sale. For each quantity, we report the averages of the two groups
and their difference. T-statistics of the differences are reported in parentheses. Asterisks denote significance levels (***=1%,
**=5%, *=10%).
Fire Sale Stocks Characteristics
More Predation Less Predation Difference t-stat
Liquidation volume (million $) 18.595 11.056 7.539*** (32.747)
Volume / CRSP volume (%) 12.812 23.633 -10.821*** (-48.060)
First day sold 0.285 0.454 -0.170*** (-25.569)
Portfolio weight (%) 0.409 0.293 0.116*** (15.475)
Amihud ratio 0.019 0.277 -0.258*** (-85.176)
Market Cap (million $) 7.844 0.362 7.482*** (37.504)
Daily Return Volatility (%) 0.425 0.606 -0.181*** (-65.154)
Past month performance (%) -0.098 -1.659 1.561*** (10.006)

5
 Table A3
Robustness: Excluding Bad News and Underperforming Stocks
The table reports results on a first set robustness checks on the results presented in Table 3. In the specifications of Panel A we
exclude the fire-sale events happening during NBER recession periods, which in our sample include the burst of the dot-com bubble
(March 2001 – November 2001) and the global financial crisis (December 2007 – June 2009). In the specifications of Panel B we
exclude stocks subject to negative fundamental news in a window of 5 days before and after the start of the fire-sale event, as
proxied by (i) negative earning surprises, (ii) Raven Pack news index in the bottom quartile, (iii) negative analyst reccomendation
changes. In Panel C we exclude stocks expriencig negative returns in a window of 10 days preceding the start of the fire-sale event.
In Panel D, we exclude stocks with high short interest in the 2 weeks preceding the fire sale event, as proxied by a value of utilization
ratio, computed using data from Markit as shares on loan / shares available from lending, in the top quartile of the cross-sectional
distribution in the CRSP universe. We cluster standard errors at the event-stock-manager level. T-statistics are reported in
parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel A: Excluding NBER Recessions Periods

Probability of Predation Predatory Volume
Dependent variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)
Ranking based on Ranking based on Ranking based on Ranking based on
Best clients proxy
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.052*** 0.077*** 0.115*** 0.113**

(2.816) (3.740) (2.814) (2.289)
Best Client 0.027 0.055*** -0.001 0.008
(1.498) (2.600) (-0.029) (0.226)
Liquidation Period 0.006*** 0.005*** 0.003 0.003
(5.113) (4.459) (0.690) (0.776)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 428,314 428,314 428,314 428,314

R-squared 0.048 0.048 0.016 0.016

Panel B: Excluding Negative News

Probability of Predation Predatory Volume
Dependent variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)
Ranking based on Ranking based on Ranking based on Ranking based on
Best clients proxy
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.052*** 0.076*** 0.102*** 0.110**

(2.801) (3.600) (2.748) (2.295)
Best Client 0.017 0.046** 0.011 0.013
(0.966) (2.217) (0.350) (0.349)
Liquidation Period 0.006*** 0.006*** 0.003 0.003
(5.531) (4.874) (0.779) (0.755)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 447,504 447,504 447,504 447,504

R-squared 0.047 0.048 0.013 0.013

6
 Panel C: Excluding Negative Momentum Stocks
15 Probability of Predation Predatory Volume
Dependent variable
15 - Probability of Liquidity Provision - Liquidity Provision Volume
20 (1) (2) (3) (4)
Ranking based on Ranking based on Ranking based on Ranking based on
40 Best clients proxy
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.070*** 0.107*** 0.125*** 0.148***

(3.016) (4.062) (3.170) (2.601)
Best Client 0.034* 0.061** -0.019 -0.012
(1.717) (2.551) (-0.578) (-0.280)
Liquidation Period 0.004*** 0.003** 0.001 0.001
(3.159) (2.425) (0.327) (0.227)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 289,082 289,082 289,082 289,082

R-squared 0.051 0.051 0.025 0.025

Panel D: Excluding High Short Interest Stocks

15 Probability of Predation Predatory Volume
Dependent variable
15 - Probability of Liquidity Provision - Liquidity Provision Volume
20 (1) (2) (3) (4)
Ranking based on Ranking based on Ranking based on Ranking based on
40 Best clients proxy
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.050*** 0.078*** 0.090*** 0.101**

(2.849) (3.915) (2.735) (2.345)
Best Client 0.025 0.053*** 0.013 0.017
(1.511) (2.645) (0.448) (0.517)
Liquidation Period 0.005*** 0.005*** 0.002 0.002
(4.991) (4.229) (0.522) (0.476)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 477,463 477,463 477,463 477,463

R-squared 0.043 0.043 0.013 0.013

7
 Table A4
Hedge Funds vs. Other Institutions
The table reports results on the heterogeneity of the predatory behavior with respect to the characteristics of the clients. We run
stock-level regressions with the same specification as in the baseline version of Table 3, but restricting to managers identified as
hedge funds in Panel A and to the complementary set of other institutions in Panel B. We cluster standard errors at the event-stock-
manager level. T-statistics are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Dependent variable Probability of Predation - Probability of Liquidity Provision

Subsample Hedge Funds Other Institutions

(1) (2) (3) (4)
Ranking based on Ranking based on
Best clients proxy Ranking based on Volume Ranking based on Volume
Commissions Paid Commissions Paid

Best Client ⨉ Liquidation Period 0.065*** 0.083*** 0.034 0.069**

(3.189) (3.692) (1.201) (2.068)
Best Client 0.034 0.027 0.003 0.062*
(1.467) (1.027) (0.114) (1.921)
Liquidation Period 0.006*** 0.006*** 0.006*** 0.005***
(4.182) (3.796) (3.978) (3.287)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes

Observations 230,780 230,780 270,784 270,784

R-squared 0.055 0.055 0.061 0.062

8
 Table A5
Robustness: Broker-Manager Fixed Effects
The table presents evidence of the effect of broker-client relationship strength on the probability of predatory behavior.
The regressions are run at the stock-day-manager-broker level, excluding trades by managers originating the fire-sale
of interest or another overlapping fire-sale. In all specifications, the dependent variable is the difference between a
dummy indicating predation and a dummy indicating liquidity provision, i.e. it takes value one when the trade is in
the same direction of the volume by the liquidating fund for that stock on that day (i.e. it is a sell trade), it equals
negative one if the trade is in the opposite direction (i.e. a buy trade) and equals zero if the manager is not trading that
stock on that particular day. We regress the dependent variable on the proxies for the manager-broker relationship
strength, a dummy indicating the first 5 days of the fire sale, and the interaction of the two variables. The liquidation
period dummy equals zero for the five days before the fire sale. We consider all trades on stock j intermediated by
brokers that eventually become aware that the stock is subject to fire sale pressure, i.e. brokers B for which
É5 É5
>2q*∈ s,P (012F3(* ) = 1 where 012F3(* is defined as above. The regression is run on a 5 days window centered
at the beginning of the fire sale (t=0), defined as the first day in which our liquidation measure crosses the threshold.
In the first two specifications we include broker×manager fixed effects and in the last two specifications we include
broker×originator fixed effects, where the originator is the manager initiating the fire sale. Standard errors are clustered
by event-stock-manager and T-statistics are reported in parentheses. Asterisks denote significance levels (***=1%,
**=5%, *=10%).

Dependent variable Probability of Predation - Probability of Liquidity Provision

(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.059*** 0.088*** 0.056*** 0.083***

(3.402) (4.475) (3.203) (4.261)
Best Client -0.027 0.003 0.028* 0.049***
(-1.136) (0.111) (1.749) (2.735)
Liquidation Period 0.009*** 0.008*** 0.009*** 0.008***
(3.854) (3.512) (3.403) (3.122)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Event Fixed Effects Yes Yes Yes Yes
Stock Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes Yes Yes
Broker-Manager Fixed Effects Yes Yes
Broker-Originator Effects Yes Yes

Observations 501,562 501,562 501,567 501,567

R-squared 0.081 0.081 0.062 0.063

9
 Table A6
Alternative Clustering of Standard Errors
The table presents robustness tests on the results in Tables 2, Panel A, and Table 3, Panel A, based on alternative ways
to cluster the standard errors. In particular, Panel A presents robustness for Table 2, Panel A, and Panel B reports
robustness for Table 3, Panel A. T-statistics are reported in parentheses. Asterisks denote significance levels (***=1%,
**=5%, *=10%).

Panel A: Predatory Behavior and Broker Awareness

Probability of Predation Predatory Volume
Dependent Variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4) (5) (6)

Aware Broker 0.195*** 0.195*** 0.195*** 0.043** 0.043*** 0.043**

(5.595) (5.551) (5.509) (2.505) (2.626) (2.519)

Time FE Yes Yes Yes Yes Yes Yes

Manager FE Yes Yes Yes Yes Yes Yes
Broker FE Yes Yes Yes Yes Yes Yes

Clustering broker broker-stock broker-day broker broker-stock broker-day

Observations 496,555 496,555 496,555 496,555 496,555 496,555

R-squared 0.065 0.065 0.065 0.017 0.017 0.017

Panel B: Probability of Predation and Broker-Client Relationship Strength

Dependent Variable Probability of Predation - Probability of Liquidity Provision

(1) (2) (3)

Best clients proxy Ranking based on Volume Ranking based on Volume Ranking based on Volume

Best Client ⨉ Liquidation Period 0.059*** 0.059*** 0.059***

(3.404) (4.068) (3.458)
Best Client 0.023 0.023** 0.023
(1.464) (1.979) (1.459)
Liquidation Period 0.008*** 0.008*** 0.008***
(3.588) (4.212) (3.562)

Time FE Yes Yes Yes

Manager FE Yes Yes Yes
Stock FE Yes Yes Yes
Broker FE Yes Yes Yes

Clustering Event-Stock-Manager Event-Stock-Date Event-Stock-Broker

Observations 501,567 501,567 501,567

R-squared 0.053 0.053 0.053

10
 Table A7
Predation Conditional on VIX Levels
The table presents evidence of a higher level of predatory activity during periods of market turmoil. We first compute
the average level of the VIX Index during each fire sale events, by tanking the average across the fire sale days. We
then use the median of the distribution of the event-level VIX to split the sample of fire sale events into two groups.
We re-run the regression specifications of columns (3)-(4) and (7)-(8) of Table2, Panel A, separately for each of the
two subsamples and we report the results in Panel A for events with VIX level above median and in Panel B for events
with VIX level below median. Standard errors are clustered at the broker level. T-statistics are reported in parentheses.
Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel A: VIX Above Median

Probability of Predation Predatory Volume
Dependent Variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)

Aware Dummy 0.217*** 0.216*** 0.043 0.045

(6.265) (6.478) (1.542) (1.586)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes

Observations 211,876 211,750 211,876 211,750

R-squared 0.058 0.064 0.008 0.018

Panel B: VIX Below Median

Probability of Predation Predatory Volume
Dependent Variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)

Aware Dummy 0.176*** 0.165*** 0.042*** 0.041***

(4.409) (4.111) (2.800) (2.708)

Time Fixed Effects Yes Yes Yes Yes

Manager Fixed Effects Yes Yes Yes Yes
Broker Fixed Effects Yes Yes

Observations 284,761 284,654 284,761 284,654

R-squared 0.071 0.077 0.019 0.022

11
 Table A8
Order Imbalance of Liquidating Funds
The table presents summary statistics on the imbalance of liquidating funds during the fire sale periods, including both
the volume generated on the fire sale stocks and the other stocks traded by the liquidating fund in that period. We
report the net signed dollar volume and the relative order imbalance, defined as the ratio between the net signed share
volume and the absolute share volume. In Panel A we aggregate the imbalance measures at the event-level by taking
the average across the liquidation days of each fire sale, while in Panel B we report the statistics at the event-day-
level, computing the imbalance measures for each day of the fire sale.

Fire sale days: Panel A: Day-level

Count Mean S.D. Min 5% 10% 25% 50% 75% 90% 95% Max
Dollar Volume (million $) 1920 -83.17 170.16 -4597.74 -274.68 -187.53 -99.22 -42.63 -15.15 -3.98 -1.14 -0.03
Order Imbalance 1920 -0.30 0.27 -1.00 -0.94 -0.74 -0.42 -0.22 -0.11 -0.05 -0.01 -0.01

Fire sale events: Panel B: Event-level

Count Mean S.D. Min 5% 10% 25% 50% 75% 90% 95% Max
Dollar Volume (million $) 385 -414.79 603.80 -7522.75 -1180.31 -866.21 -548.34 -263.30 -104.67 -41.93 -17.73 -9.74
Order Imbalance 385 -0.30 0.24 -1.00 -0.92 -0.66 -0.39 -0.24 -0.14 -0.08 -0.06 -0.04

12
 Table A9
Regressions without Fixed Effects
The table revisits our main results in Tables 2 and 3, Panel A, using specifications without fixed effects. In Panel A
we report results for a specification similar to that of Panel A of Table 2, but without fixed effects. In Panel B we
report results for a specification similar to that of Panel A of Table 3, but without fixed effects. In Panel A we cluster
standard errors at the broker level, while in Panel B we cluster standard errors at the event-stock-manager level. T-
statistics are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel A: Predatory Behavior and Broker Awareness

Probability of Predation Predatory Volume
Dependent Variable
- Probability of Liquidity Provision - Liquidity Provision Volume

(1) (2)

Aware 0.232*** 0.054***

(5.175) (3.078)
Constant -0.022 -0.002
(-1.119) (-1.166)

Observations 496,729 496,729

R-squared 0.002 0.000

Panel B: Probability of Predation and Broker-Client Relationship Strength

Probability of Predation Predatory Volume
Dependent Variable
- Probability of Liquidity Provision - Liquidity Provision Volume
(1) (2) (3) (4)
Best clients proxy Ranking based Ranking based on Ranking based Ranking based on
on Volume Commissions Paid on Volume Commissions Paid

Best Client ⨉ Liquidation Period 0.065*** 0.092*** 0.138*** 0.140***

(3.763) (4.697) (3.611) (3.081)
Best Client 0.055*** 0.068*** 0.058*** 0.067**
(4.024) (4.339) (2.715) (2.512)
Liquidation Period 0.006*** 0.005*** 0.002 0.003
(6.008) (5.255) (0.642) (0.719)
Constant 0.004*** 0.004*** -0.006*** -0.006***
(5.701) (5.243) (-3.108) (-2.915)

Observations 501,568 501,568 501,568 501,568

R-squared 0.001 0.001 0.000 0.000

13
 Table A10
Differential Treatment for Best Clients
The table presents evidence of a differential treatment by aware brokers when the liquidating fund is one of their best
clients. In details, we first define a dummy J<KL<;29<=WxL=;F4G9H)<4=97,+,6 which is equal to one if the liquidating
fund f originating the fire sale event e is among the best clients of broker b, i.e. the manager generated at least 5% of
the volume intermediated by the aware broker in the previous semester. We then interact this dummy with the broker
awareness dummy defined in Table 2 and run regression at the event-manager-broker-stock level with three different
specifications, where the dependent variable is respectively (i) the predatory volume (i.e. the product of the predation
dummy defined in Table 3 multiplied by the volume of the transaction as a fraction of the market capitalization of the
traded stock); (ii) the liquidity provision volume (i.e. the product of the liquidity provision dummy defined in Table 3
multiplied by the volume of the transaction as a fraction of the market capitalization of the traded stock); (ii) the
difference between the predatory volume and the liquidity provision volume. We include manager, broker and day
fixed effects. Standard errors are clustered at the event-stock-manager level and T-statistics are reported in
parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

(1) (2) (3)

Predatory Volume - Liquidity
Dependent Variable Predatory Volume Liquidity Provision Volume
Provision Volume

Aware Dummy ⨉ Liquidating Fund Best Client -0.049** 0.001 -0.041*

(-2.131) (0.077) (-1.796)
Aware Dummy 0.108*** 0.045*** 0.064***
(4.432) (2.910) (2.855)
Liquidating Fund Best Client -0.002 -0.010** 0.004
(-0.280) (-2.297) (0.798)

Time Fixed Effects Yes Yes Yes

Manager Fixed Effects Yes Yes Yes
Broker Fixed Effects Yes Yes Yes

Observations 496,555 496,555 496,555

R-squared 0.026 0.017 0.017

14
 Table A11
Characteristics of Liquidating Funds facing Aware Brokers
The table presents characteristics of the liquidating funds, partitioned into two groups based on the number of aware
brokers they face. More precisely, for each fire sale event we record the number of aware brokers A divided by the
number of fire sale stocks N. We then average the ratio A/N across all the fire sale events generated by each liquidating
fund. The cross-sectional distribution of the manager-level variable is then used to split the set of liquidating funds
into two parts, using the median as cutoff. For each of the two groups we take the average of the following quantities,
computed at the manager-level using the entire Ancerno dataset: (i) the total dollar volume generated by the fund; (ii)
the number of broker relations, defined as the number of brokers which intermediated at least one transaction of the
fund; (iii) the ration between the number of broker relations and the total dollar volume in dollar millions; (iv) the
total dollar amount paid in commissions by the manager to all the connected brokers; (v) the ratio between the average
commission per dollar paid to aware brokers (i.e. those brokers which are tagged as aware at least once in one fire sale
originated by the manager) and the average commission per dollar paid to the complementary set of brokers (unaware);
(vi) the ratio between the total dollar commission paid to aware brokers and the total dollar commission paid to
unaware brokers; (vii) the ratio between the total volume intermediated by the aware broker and the volume
intermediated by the unaware brokers. For each quantity, we report the averages of the two groups and their difference.
T-statistics for the differences are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%,
*=10%).

Liquidating Managers Characteristics

Aware Brokers Unaware Brokers Difference t-stat

Total dollar volume (billion $) 507.333 110.696 396.637*** (5.656)

Total brokers relations 217.925 188.547 29.377** (2.443)

Number of brokers per million $ 0.010 0.026 -0.016** (-2.432)

Total commissions paid (million $) 153.199 47.783 105.416*** (4.795)

Commission per dollar ratio (aware / unaware) 0.915 0.819 0.096** (2.559)

Total dollar commission ratio (aware / unaware) 2.663 0.574 2.089** (2.438)

Volume ratio (aware / unaware) 0.184 0.076 0.108 (1.278)

15
 Table A12
Timing of Predation
The table presents evidence of the effect of broker awareness on the timing of predation. The regressions are run at
the event-manager-broker-stock level, focusing on the liquidation period, excluding trades by managers originating
the fire-sale of interest or another overlapping fire-sale. In all specifications, the dependent variable is a number
counting the number of days after the beginning of the fire sale in which the first predatory trade occurred. Predation
and the AwareBorker dummy are defined as in Table 3. In rows (1)-(2) we present results for the specifications without
fixed effects, while in specifications (3)-(4) we report results with day, manager, and stock fixed effects. Standard
errors are clustered at the broker level and t-statistics are reported in parentheses. Asterisks denote significance levels
(***=1%, **=5%, *=10%).

Dependent variable First Day of Predation

(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based on Ranking based on
Volume Commissions Paid Volume Commissions Paid

Best Client ⨉ Aware Broker -1.202*** -1.379*** -0.697** -0.787**

(-7.725) (-7.941) (-2.364) (-2.447)
Best Client -0.357*** -0.275*** -0.254*** -0.173***
(-11.381) (-6.911) (-5.002) (-2.773)
Aware Broker 0.118*** 0.129*** -0.023 -0.015
(6.004) (6.611) (-0.800) (-0.566)
Constant 2.203*** 2.187***
(304.624) (305.812)

Time Fixed Effects Yes Yes

Stock Fixed Effects Yes Yes
Manager Fixed Effects Yes Yes

Observations 98,771 98,771 98,411 98,411

R-squared 0.003 0.002 0.237 0.236

16
 Table A13
Persistence of Broker Concentration
This table reports results on the concentration of brokers employed by asset managers in our sample. We construct three proxies of
broker concentration: (i) the Herfindahl Index (HHI) of the trading volumes at the monthly frequency, (ii) the normalized Herfindahl
Index (HHI) of the trading volumes at the monthly frequency - defined as (Ñ − 1/8)/(1 − 1/8) where N is the number of brokers
in our sample and H is the usual Herfindahl Index - and (iii) the number of brokers intermediating at least one trade of the manager
in the given month. In Panel A, we regress each proxy on their one-month, six-months and one-year lags using observations at the
manager-month level. In Panel B, we repeat the same exercise restricting to the sample to fire sale events. All the specifications
include month fixed effects. T-statistics are reported in parentheses. Asterisks denote significance levels (***=1%, **=5%,
*=10%).

Panel A - Unconditional Panel A: Unconditional Brokers Concentration

Dependent Variable HHI HHI Normalized HHI Normalized HHI Number of Brokers Number of Brokers

(1) (2) (3) (4) (5) (6)

One Month Lag 0.592*** 0.398*** 0.388*** 0.279*** 0.961*** 0.756***

(193.897) (104.549) (111.338) (70.742) (908.650) (270.680)
Six Months Lag 0.220*** 0.179*** 0.144***
(55.312) (43.966) (43.850)
One Year Lag 0.172*** 0.156*** 0.084***
(44.943) (39.187) (30.395)

Month Fixed Effects Yes Yes Yes Yes Yes Yes

Observations 70,284 60,839 70,284 60,839 70,284 60,839

R-squared 0.362 0.433 0.161 0.215 0.922 0.931

Panel B - Fire Sales Panel B: Brokers Concentration during Fire Sale Events
Dependent Variable HHI HHI Normalized HHI Normalized HHI Number of Brokers Number of Brokers

(1) (2) (3) (4) (5) (6)

One Month Lag 0.260*** 0.222*** 0.203*** 0.170*** 1.038*** 1.011***

(12.642) (9.749) (10.445) (6.796) (55.184) (18.047)
Six Months Lag -0.001 -0.001 0.021
(-0.292) (-0.227) (0.351)
One Year Lag 0.027** 0.032* 0.008
(2.180) (1.813) (0.203)

Month Fixed Effects Yes Yes Yes Yes Yes Yes

Observations 322 284 322 284 322 284

R-squared 0.654 0.734 0.590 0.670 0.958 0.957

17
 Table A14
Commissions Paid to Aware Brokers
The table presents evidence on the post-event increase of commissions paid by predators to aware brokers. For each
month t on a window of two years around each fire sale event e, we define the average H(>><GG<(=_f43_;())237,+,6,*
paid by manager m to broker b as the ratio H(>>7,+,6,* /&'()7,+,6,* where ratio H(>>7,+,6,* is the total amount in
dollars paid in commissions by manager m to broker b during month t and &'()7,+,6,* is the total dollar volume traded
by manager m and intermediated by broker b in that month. For each event, we consider brokers which are marked as
Aware on at least one of the fire sale stocks and managers whose trades are intermediated by at least one of these
broker in the ten trading days around the event. We then regress H(>><GG<(=_f43_;())237,+,6,* on the interaction of
the dummy variable :(G97,* , indicating the two years after the fire sale event, with each of our Best Clients proxies.
In Panel A we look at the clients that are more likely to predate on that stock in that event, which we identify as those
that are above the median of the distribution of profitability in the ten-day window after the event. In Panel B we run
the same analysis focusing only on managers that trade in the same direction as the liquidating fund during the
liquidation periods. We add event, manager, and brokers fixed-effects to the regression and we cluster standard errors
by event-broker-manager to account for time-series autocorrelation in commissions paid. T-stats are reported in
parentheses. Asterisks denote significance levels (***=1%, **=5%, *=10%).

Panel A: Highest Predatory Profits

Dependent variable Commissions per dollar (basis points)
(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based Ranking based on
Volume Commissions Paid on Volume Commissions Paid

Best Client ⨉ Post 2.493*** 1.537*** 2.234*** 1.327***

(6.923) (4.254) (7.317) (3.982)
Best Client -13.523*** -10.269*** -5.990*** -2.456***
(-21.484) (-18.022) (-14.568) (-5.932)
Post -0.418*** -0.376*** -0.595*** -0.556***
(-10.372) (-9.355) (-15.528) (-14.495)
Constant 6.395*** 6.246***
(138.850) (136.728)

Time Fixed Effects Yes Yes

Manager Fixed Effects Yes Yes
Broker Fixed Effects Yes Yes

Observations 531,527 531,527 531,516 531,516

R-squared 0.027 0.014 0.304 0.302

Panel B: Predators Only

Dependent variable Commissions per dollar (basis points)
(1) (2) (3) (4)
Best clients proxy Ranking based on Ranking based on Ranking based Ranking based on
Volume Commissions Paid on Volume Commissions Paid

Best Client ⨉ Post 1.629*** 1.086*** 1.598*** 1.093***

(6.463) (4.062) (7.167) (4.406)
Best Client -10.619*** -7.594*** -4.681*** -1.334***
(-25.839) (-19.411) (-17.674) (-4.662)
Post -0.477*** -0.441*** -0.641*** -0.607***
(-14.362) (-13.385) (-20.418) (-19.469)
Constant 5.881*** 5.685***
(149.915) (146.372)

Time Fixed Effects Yes Yes

Manager Fixed Effects Yes Yes
Broker Fixed Effects Yes Yes

Observations 706,703 706,703 706,690 706,690

R-squared 0.030 0.013 0.321 0.319

18