Asset Fire Sales in Equity Markets: Evidence

from a Quasi-Natural Experiment

Borja Larrain, Daniel Muñoz, and José Tessada

May 2016

Abstract

In November of 2007 a fire sale of Chilean stocks was triggered by a change in the
constraints that regulate pension fund portfolios. This regulatory shock provides a
cleanly identified fire sale unrelated to fundamentals. Stocks with more selling pressure
from pension funds lost approximately 4% in November compared to other stocks.
Although the selling pressure was temporary, prices reverted only after four to six
months. Pension funds initially mitigated the impact of the fire sale by selling large
index stocks in which demanding liquidity was less costly. Coordination across pension
funds increased during the fire sale. We find no significant evidence of real effects on
firm investment in the quarters after the fire sale.

Acknowledgements: We would like to thank comments and suggestions from an anonymous referee,
Murillo Campello (the editor), Jeff Pontiff (discussant at Finance UC), Giorgo Sertsios, Andrei Shleifer,
and seminar participants at the 7th International Conference of Finance UC, Universidad Adolfo Ibáñez,
Universidad de los Andes, and PUC-Chile. We thank Mounu Prem for sharing his data on quarterly
financial statements of Chilean firms. Larrain acknowledges funding from Proyecto Fondecyt Regular
#1141161. Tessada acknowledges funding from Proyecto Inserción de Capital Humano Conicyt
#79100017.
Larrain (*): Pontificia Universidad Católica de Chile, Escuela de Administración and Finance UC;
Muñoz: Pontificia Universidad Católica de Chile, Escuela de Administración;
Tessada: Pontificia Universidad Católica de Chile, Escuela de Administración and Finance UC.

(*) Corresponding author. Address: Avenida Vicuña Mackenna 4860, Macul, Santiago, Chile. Tel: (56
2) 2354-4025, e-mail: borja.larrain@uc.cl

0 
 A fire sale is a forced sale triggered by financial distress or by constraints that
affect investors’ portfolios. A fire sale is not driven by bad news about fundamentals,
but, since the sale is sudden, prices are discounted below these fundamentals or long
run values (Shleifer and Vishny, 2011). Price dislocations of this sort open up the door
for potential policy interventions. For example, direct purchases of securities by the
government can be justified if fire sales contribute to the fragility of the financial
system. Policy recommendations often hinge on having a precise knowledge of the inner
workings of fire sales and their price impact. Unfortunately, disentangling fire sales
from ordinary sales is not easy because, in practice, it is difficult to know whether price
discounts are motivated by a fire sale or by bad news.
In this paper we contribute to the fire sales literature by providing a cleanly
identified fire sale —unrelated to news about fundamentals— that occurred in an
emerging stock market. Identifying fire sales is particularly hard in the stock market
because the flow of information is continuous and liquidity in comparison to other
markets is high. Hence, price dislocations of the type produced by fire sales are, at
least in principle, rare in this market. On the other hand, as noted by Coval and
Stafford (2007), the stock market has the advantage of being a relatively transparent
market, where prices and trading volume are publicly known, and therefore where we
can trace the impact of a fire sale if there is one. We exploit a quasi-natural fire sale
to answer several open questions in the literature. First of all, how big are the price
dislocations caused by fire sales? Then, how long does it take for prices to revert to
pre-crisis levels? Do financial intermediaries try to mitigate the price impact of a fire
sale? If so, what mitigation mechanisms do they use? Does coordination among
intermediaries increase or decrease during fire sales?
The event that we study is a fire sale of Chilean stocks by the local private
pension funds. These institutional investors dominate the Chilean market since the
privatization of social security in the early 1980s. The regulator of the pension fund
industry sets limits on different asset classes according to the risk profile of each one
of the five types of funds that are offered. For example, type-C funds —in a scale from
A to E, where A represents the most risky fund and E the least risky fund– can invest
up to 40% in equities (domestic or international). Until October of 2007 these limits
were not enforced in a strict way, because of loopholes or a lax interpretation of the
law. In particular, some funds exceeded the investment limit in equities by almost 10%.

1 
 On October 26th, 2007 the regulator sent a memorandum to all pension funds
asking them to strictly obey the portfolio limits set by law. The memo specified that
excess investments had to be eliminated regardless of their cause. The memo did not
intend to be an opinion about the future of equity markets or a piece of investment
advice. For instance, the memo did not make any reference to market perspectives or
asset fundamentals. The reason for the memo was simply that, according to the
regulator, excess investments were not in line with the true spirit of the Chilean multi-
fund system established in 2002. The main purpose of the multi-fund system is to offer
a menu of funds with markedly different risk-return profiles. Excess investments
conspire against this by making some of the funds too similar to each other. Excess
investments can also lead investors to confusion and mistakes. For example, the level
of excess investments in type-C funds made them in practice look like type-B funds.
Investors who chose type-C funds were taking more risk than what they really wanted
to take according to their risk preferences.
Although the regulator gave a year to reduce excess investments, pension funds
immediately started selling domestic stocks, which represented close to half of their
equity portfolios at the time. In some sense the regulator shouted fire in a crowded
theater and nobody wanted to be the last one out the door. Whether this was the
optimal response from pension funds or a product of their lack of experience in similar
situations is something that we cannot judge with the data at hand. Competition and
regulatory incentives in the Chilean market probably induced in part the rush to sell:
sell as fast as possible, ahead of other funds if possible, in order to get an edge over
competitors who will be hit hard because of the poor performance induced by the same
sales. Perhaps pension funds could have followed other strategies such as smoothing
sales over time. For our purposes, however, more than counterfactual scenarios the
important issue to establish is the existence of a fire sale unrelated to asset
fundamentals.
Chilean pension funds have to disclose their portfolios at the monthly frequency
so we can measure precisely how they dealt with the regulatory change. In November
of 2007 pension funds sold shares that represented, for the average domestic stock in
their portfolios, 0.21% of total shares outstanding (almost 0.64% of the float). This
represented a quarter of the average monthly turnover for the typical stock, which is
quite high in comparison to other fire sales in the literature. For instance, the fire sales
studied by Coval and Stafford (2007) represent only 2% of average volume. During

2 
November pension funds reduced positions in 30% of the stocks they held and increased
positions in only 10%. For comparison, in the surrounding months the average number
of positions increased was about the same as the average number of positions reduced.
In short, the selling pressure felt in November was highly unusual. We find that foreign
investors absorbed a large fraction of this pressure. The local mutual fund industry
was too small to absorb a significant fraction of pension fund sales.
The price impact of the fire sale was substantial. In November 2007 the return
on stocks that were most affected by the selling pressure was close to 4% lower than
the return on other stocks. We focus on this cross-sectional effect since the market
wide effect is harder to identify given that we have only one event. Price differentials
did not revert until March-May of 2008, i.e., price impact lasted for 4 to 6 months.
Documenting the reversal of the initial effect is crucial for distinguishing the fire sale
from the alternative of information-based trading, which predicts no reversal. Also, the
pattern of reversal that we find is not consistent with a short-run liquidity effect,
because in that case the price impact should disappear immediately after the abnormal
selling pressure subsides. Instead, we find that prices only revert long after the selling
pressure of November is over. This reversal result points to an alternative model with
downward sloping demands (Shleifer, 1986) or slow-moving capital (Duffie, 2010).
The regulator mandated a reduction in equities, but it did not specify which
stocks to sell. The easiest thing to do was to reduce the portfolio proportionally, in
which case pressure would be related to the ownership stake that pension funds have
in each stock. Simply put, under the proportionality hypothesis, pension funds create
more pressure in segments where they are larger investors. We find that pension fund
ownership is indeed correlated with pressure during November, although not before or
after the event. We also find evidence consistent with the alternative hypothesis that
there was conscious selection of stocks during the fire sale. Similar to the finding of
Jotikasthira, Lundblad, and Ramadorai (2012), we find that pension funds
systematically tried to mitigate price impact by selling large index stocks, which have
lower costs of demanding liquidity. We also find some evidence that stocks with large
over- or under-weights in comparison to the market portfolio received more selling
pressure, which increased benchmarking with respect to the local equity index. We do
not find that past stock performance (i.e., momentum) is a predictor of pressure during
the fire sale, which one could expect if there is a preference for locking-in profits.

3 
 When looking at the cross-section of pension funds we find signs of increased
coordination during the fire sale. Equity portfolios became more alike since we find a
reduction of their active share (in the sense of Cremers and Petajisto, 2009) measured
with respect to the industry average, and we also find that the sensitivity of a fund’s
change in holdings with respect to changes in the other funds increased during the fire
sale.
Finally, we find a statistically insignificant reduction in investment by firms
during the quarters after the fire sale. This suggests that the real effects of the fire sale
were limited.
The rest of the paper is organized as follows. In Section 1 we relate our results
to the fire sales literature and highlight our contribution. In Section 2 we describe the
event and data in detail, including some background on the Chilean private pension
fund system. Section 3 shows the results on price impact. Section 4 explores mitigation
and amplification mechanisms of the fire sale. Section 5 gives suggestive evidence on
the real effects of the fire sale. In Section 6 we present our conclusions.

1. Relationship to the Fire Sales Literature

Our paper contributes first and foremost to the literature on fire sales in
financial markets. Shleifer and Vishny (2011) provide a summary of the theoretical
models. The typical cycle observed in a fire sale can be summarized as follows. First,
a negative shock (e.g., bursting of the U.S. housing bubble in 2007-8) hits the balance
sheet of financial intermediaries who start selling assets. Then, the price falls caused
by these initial sales produce distress in other intermediaries, which prompts even more
selling and further price falls.
We are focused on equity markets. Coval and Stafford (2007) develop a broad
measure of fire sales in equity markets derived from flows to U.S. mutual funds. They
show that stocks that experience flow-induced selling pressure have abnormally
negative returns. Several papers have followed their identification strategy (see
Edmans, Goldstein, and Jiang, 2012; Frazzini and Lamont, 2008; Jotikasthira,
Lundblad, and Ramadorai, 2012; and Lou, 2012). Other papers, such as Mitchell,
Pedersen, and Pulvino (2007), conduct event studies like we do.

4 
 Our framework is similar to Coval and Stafford (2007) in the sense that we
derive a measure of pressure from the portfolios of institutional investors. However, we
are focused on a particular event with the goal of providing an alternative identification
strategy that complements their broader approach. Identification is a primary concern
because fire sales are always hard to disentangle from bad news about fundamentals.
Also, clean identification is in general quite challenging since fire sales are often
amplification mechanisms for other shocks. In other words, it is hard to pin down the
pure fire sale component (i.e., the price effect solely motivated by distress or
constraints) in a downward spiral of prices that can be initiated by poor fundamentals.
The regulatory shock that we exploit provides an ideal experiment in this regard
because exogenous constraints rather than information initiate the selling.2
Although identification is key to get good estimates of the effect of the fire sale,
our use of a particular event comes at the potential cost of external validity. Mapping
our results to other fire sales, in other markets or asset classes, is not straightforward
precisely because ours is an unusual environment like many quasi-natural experiments.
Four things are important to keep in mind when extrapolating our results to other
settings. First, the shock occurs in late 2007 when some sings of the crisis of 2008 were
starting to appear. (Although in emerging markets the crisis was mostly felt after
September 2008, we focus on the cross-section of stocks in part because of this peculiar
timing). Second, Chile represents a relatively small stock market without the depth
and diversification of more developed markets. Third, pension funds are not (or are
less) affected by amplification mechanisms that are typical of other fire sales, such as
the use of leverage or a strong flow-performance sensitivity in investor flows. Finally,
we have just one event so it is hard to predict what would happen if similar shocks
were repeated over time and how would intermediaries or investors learn to handle
such shocks.
This last element is a disadvantage of having a single event, but it is also an
advantage from other perspective. Our experiment can shed light on the timing and
degree of price reversals after fire sales, which is an open issue in the literature. Coval
and Stafford (2007) find some reversal after fire sales induced by mutual fund flows;

                                                            
2
 Informationally-neutral events, although key for identification purposes, are hard to find in financial
markets. In this respect our paper is related to other informationally-neutral events such as index
rebalancing (Harris and Gurel, 1986; Shleifer, 1986; Kaul, Mehrotra, and Morck, 2000; Wurgler and
Zhuravskaya, 2002; Chen, Noronha, and Singal, 2004; Greenwood, 2005), or cosmetic changes in stock
names (Cooper, Dimitrov, and Rau, 2001). 

5 
although, the reversals they document are not complete even after 12 months.
Jotikasthira, Lundblad, and Ramadorai (2012) find evidence of full reversal in equity
indexes after approximately 3 months. There are multiple shocks in both of these
studies, so one concern with estimating reversal is event overlap as the horizon is
extended. In our case with a single event we find full reversal after a relatively long
horizon (four to six months), and with the advantage that there are no confounding
effects of other fire sales.
Beyond an arguably clean identification, we add to the literature in three
specific ways. First, we provide more evidence regarding the selection of assets and the
mitigation strategies that intermediaries use in fire sales. Coval and Stafford (2007)
find that intermediaries do not try to mitigate the costs of fire sales by trading
selectively in larger and more liquid stocks. In line with Jotikasthira, Lundblad, and
Ramadorai (2012), we find that intermediaries show a preference for trading larger
stocks in fire sales. Although intuitive at first, this mitigation strategy is not obvious.
As shown by Brown, Carlin, and Lobo (2010), an investor who needs to unwind a
portfolio may prefer to start by selling first the more illiquid assets if liquidity is
expected to get worse in the future. We also find evidence that stocks with large over-
or under-weights in comparison to the market portfolio received more selling pressure,
which increased benchmarking with respect to the local equity index. Finally, we study
the impact of new variables such as breadth of ownership and differences of opinion.
Second, we are able to measure the increasing coordination of intermediaries
during fire sales. The pension funds in our study show a preference for keeping their
portfolios close to peer funds as seen in a reduced active share compared to the industry
average. We also find that their portfolio trades become more correlated to the trades
of other peer funds during the fire sale. This increase in coordination can explain the
speed and depth that are typical of fire sales.
Third, we can add to the understanding of limits to arbitrage during fire sales.
The limits-to-arbitrage literature (Shleifer and Vishny, 1997) offers several mechanisms
to explain what prevents arbitrageurs from rapidly moving capital to the trading
opportunities that arise in a fire sale. First, there is specialization of intermediaries in
certain market niches, because, for example, of the costs of gathering information
(Merton, 1987). Specialization implies that the supply of capital in each market
segment is not perfectly elastic since there is idiosyncratic risk left undiversified that
arbitrageurs have to bear (DeLong, Shleifer, Summers, and Waldmann, 1990; see

6 
Pontiff, 2006, for a counterpoint). Emerging markets like Chile are partially segmented
from other markets because of capital controls or informational asymmetries (Bekaert,
Harvey, Lundblad, and Siegel, 2011). Hence, only a few specialized institutional
investors (mutual funds, hedge funds, closed-end funds, foreign institutions, etc.), with
limited funding capacity, represent the lion’s share of investors. Second, as Shleifer and
Vishny (1992, 1997) point out, the potential buyers in a fire sale (probably other
specialized investors) can also be constrained because they experience financial distress
or capital withdrawals at the same time. The timing of our event (4th quarter of 2007)
suggests that many of the foreign investors were experiencing financial distress or
capital outflows as the crisis of 2007-2008 was unraveling, and this probably
constrained their role as liquidity providers. Third, Duffie (2010) suggests that
inattention can lead to a pattern of slow moving capital that accounts for the price
impact of fire sales. The inattention hypothesis is less appealing in our case because
the regulatory change was featured extensively in the press. Finally, limits to arbitrage
can be self-imposed in the sense that potential buyers can choose to wait instead of
providing capital immediately if they anticipate a better deal in the future
(Brunnermeier and Pedersen, 2005). Although a plausible alternative, it is hard to test
it with our data since we do not observe the portfolios of investors that are not pension
funds.

2. Event Description

2.1. Background on the Chilean Pension Fund

Industry
The Chilean pension system was privatized in 1980 through the creation of
private pension funds that substituted the pay-as-you-go system run by the
government. By law all workers and employees have to contribute 10% of their taxable
income to individual retirement accounts. This obligation to contribute does not apply
to monthly incomes above a threshold of approximately 3,000 USD. Pension fund
administrators (AFPs from their acronym in Spanish: Administradores de Fondos de
Pension) charge a fee out of the contributions of the workers, but since 2008 they do

7 
not charge maintenance fees for assets under management (before 2008 the
maintenance fee was a small fixed amount per worker).
The pension fund industry has been instrumental for the development of the
local financial market. Since 1980 AFPs have accumulated a sizeable portion of Chilean
equity and fixed income. For example, the assets of the six existing AFPs at the start
of 2007 were close to 110 billion USD, or approximately 60% of Chilean GDP. Their
holdings of domestic equity represented about 9% of the local market capitalization
and almost 30% of free float.3
Since the year 2002 investors can choose from five types of funds that each
pension fund administrator is legally bound to offer. These five funds (A through E)
are balanced portfolios with a mix of equities, fixed income and other asset classes, in
order to serve the different risk profiles of investors. Fund A has the largest share of
equities among the five funds, and is considered to be the most risky fund. Fund E is
almost entirely invested in domestic fixed income. The largest type of fund is fund C,
which accounts for close to 40% of the assets in the pension fund system during our
sample period. Fund C was the only fund offered before 2002, hence its size. Funds A
and B account for approximately 20% of assets each, while funds D and E account for
less than 10% each.
Table 1 presents the set of legal limits for equity in the five types of funds.
Equity (domestic plus international) has to represent between 40% and 80% of fund
A, between 25% and 60% of fund B, and so on. The relative order has to be preserved
at all times, i.e., fund A has to invest more in equities than fund B, fund B more than
fund C, etc. Chilean equity is a significant fraction of the equity portfolio in any of the
funds (except for Fund E that has no equity). During 2007-2008 Chilean stocks
accounted for 40%-45% of total investment in equity of the pension fund system.
Pension funds are regulated by the Superintendencia de AFPs. (SAFP). The
SAFP’s mandate includes watching over investment limits, making sure that
information is disclosed to investors, and other administrative tasks. Chilean law sets
penalties for funds that perform poorly with respect to the industry average. This is
implemented by establishing a minimum yield that is equal to the previous 3-year
return of the average fund in each risk profile less a few percentage points. Together

                                                            
3
Chilean companies are typically controlled by large shareholders with an average concentration of 68%
of shares outstanding (Donelli, Larrain, and Urzua, 2013). This implies that the average free float is
approximately 30%.

8 
with other forces that lead to herding among fund managers, such as competition and
career concerns (Scharfstein and Stein, 1990), these penalties provide incentives not to
deviate too much from the investment decisions of other pension fund managers. In
practice, penalties have not been imposed since 1998 since the portfolios of pension
funds have stayed close to each other.
Pension funds disclose their portfolios on a monthly basis, and the SAFP makes
these portfolios available to the public on its website (www.safp.cl). This gives us a
unique opportunity to see exactly what securities pension funds hold at each point in
time. We also collect data on prices, trading volume, and accounting variables (e.g.
book value of equity) for domestic stocks from Economatica. We only cover stocks that
at least one pension fund held in its portfolio at some point during 2007-2008.

2.2. The Fire Sale of November 2007

As of October of 2007 the position in equities of most funds, except for funds
type-A, was above the legal limit. As seen in Figure 1 the excess investment at the
pension fund industry level was approaching 10% in type-C funds (the largest funds).
Figure 2 shows excess investments for the individual funds of each one of the six AFPs.
We see that all pension fund administrators had similar levels of excess investment
within each type of fund. For instance, all type-C and type-D funds had positive excess
investments in equity. All but one of the AFPs (Cuprum) had positive excess
investment in type-B funds.
There are two main reasons for these excess investments in equity. First, the
Chilean stock market experienced an impressive boom in 2003-2006 with a total return
of almost 170% (an annualized rate of return of 28%). In the year 2006 alone the
Chilean market had a return of 37%. Second, the interpretation of the law at the time
was lax or at least unclear. Under certain conditions the law allowed temporary
deviations of up to 5% from the legal limits, giving the pension fund the ability to
decide the best timing to eliminate those excess investments with the objective to
maximize the return for investors. In practice this was interpreted as giving an extra
5% to the limit and that only deviations above 5% had to be eliminated (and only
within three years). Furthermore, the interpretation of some pension fund managers
was that the law did not apply to excess investments caused by high returns on stocks
initially bought within the limits.

9 
 On October 26th of 2007 Ms. Solange Berstein, the Chilean regulator at the
time, sent a memorandum to all pension funds expressing her concern about the
situation.4 In the local market this memo is usually known as “Oficio Berstein”. Ms.
Berstein argued that the way in which excess investments had been handled in the
past was not in line with the spirit of the Chilean legislation. Excess investments led
investors to confusion and, potentially, to mistakes in their financial decisions. The
memo made clear that excess investments were not allowed, and it explicitly asked all
pension funds to eliminate them within a year.
The reaction among pension funds was mixed. Some of them supported the
decision of the regulator, even arguing that one year to reduce excess investments was
too much time. Others (e.g., AFP Habitat) opposed the decision, even going to court
to annul the requirement to eliminate excess investments. Courts ultimately sided with
the regulator. The extent to which the memorandum was anticipated is debatable.
There was talk about excess investments in the months leading up to October 2007,
but the decision of the regulator came mostly as a surprise. It is still possible that some
AFPs rebalanced their portfolios in anticipation of the memo, which would only bias
our tests against finding an effect in the months following the event.5
Irrespective of the different opinions about the memorandum, the result was a
clear and massive sale of domestic stocks held by pension funds as we now document.
Table 2, panel A, shows the fraction of positions reduced (sales), maintained, or
increased (purchases) in the domestic equity portfolios of pension funds during 2007.
Each stock held by each AFP (aggregating across its 4 funds with equity positions, A
through D) counts as one observation in a month. In each case we compute the change
in the number of shares held between the beginning and the end of the month. From
January 2007 through June 2008 the average fraction of reduced positions was 18%,
the average fraction of maintained positions was 66%, and the average fraction of
increased positions was 16%. In November 2007, the month following the
memorandum, 30% of positions were reduced and only 9% were increased.6 These
numbers are comparable to the trading behavior of mutual funds affected by large

                                                            
4
 Oficio Ordinario No. 16373 from the Superintendencia de AFPs, Friday October 26, 2007.
5
We explicitly test for differences during the months leading up to October 2007 and find no evidence
of anticipation. 
6
 A couple of trading days were left in October of 2007 after the memo was made public on Friday the
26th. AFPs probably started rebalancing their portfolios right away the next week. This only dampens
the effect we document for November since we consider changes between October 31st and November
30th as reported by the SAFP. 

10 
outflows in Coval and Stafford (2007, Table 2). During November the average
reduction in holdings was 11%, while during the entire sample period the average
reduction is only 5% (see Panel B). Simply put, there is a selling pattern in November
that clearly differs from the neighboring months. The selling pattern continues to a
lesser extent in December 2007, but it then disappears.
In the spirit of Coval and Stafford (2007) we define two measures of pressure:

∆ , ,
, , 1
∈ ,

∆ , ,
, , 2
∈ ,

where , represents total shares outstanding of stock s in month t-1, ,

represents shares in free float, and ∆ , , is the change in shares held by AFP i between
the end of month t-1 and the end of month t. The free float is computed by subtracting
the number of shares held by the controlling shareholder in each company (from
Donelli, Larrain, and Urzua, 2013). Positive pressure means that the pension fund
industry was a net buyer of a particular stock during the month. Negative pressure
means that the pension fund industry was a net seller.
Notice that we compute pressure applied to a particular stock by the aggregate
pension fund industry and not by a selection of individual funds, for example, only
those that had excess investments. Coval and Stafford (2007) condition their measure
of pressure on the fund being constrained. This makes sense in their setup since not
many funds are constrained at a given point in time. They also need this condition to
identify fire sales as opposed to information-based trading. In our setup identification
comes from the regulatory shock and the nature of the event itself. In fact, all AFPs
were constrained so the distinction between constrained and unconstrained has little
importance in practice for our results.
In our application the aggregate measure of pressure makes sense for two
reasons. First, we aggregate across the six pension fund administrators present in the
market because, as seen in Figure 2, they all behave very similarly. All administrators
closely follow each other, in large part due to the incentives provided by regulation, so
there is little cross-sectional variation in the existence of excess investment in equity.

11 
Second, we aggregate all funds (A through D) managed by each pension fund
administrator because the administrator can transfer assets between funds without
causing pressure in the market. For example, if fund C has excess investment in equity
it can directly transfer stocks to fund A of the same administrator if this fund has
slack. These transfers are legal (under the condition that they are made at the
prevailing market price) and they have the advantage of avoiding causing pressure in
the market. In Section 4.3 we show that these internal transfers were indeed used by
pension funds in the beginning. Hence, a measure of pressure that is net of internal
rebalancing is a more conservative assessment of the demand shock that the rest of
the market has to absorb. Another way to understand the need to aggregate across
funds is that, although the limits on excess investments are defined at the fund level,
the possibilities for internal rebalancing make these constraints only binding at the
fund administrator level. Since the funds with excess investments (B through D) where
larger than the funds without excess investments, in practice all pension fund
administrators were constrained when the memo was sent.
Table 2, Panel C, shows the average pressure across the approximately 100
stocks for which we have data during 2007. All numbers are reported as percentages
of shares outstanding or free float (i.e., numbers from equations (1) and (2) are
multiplied by 100). There is a clear selling pattern during November of 2007 with an
average total (free) pressure of -0.21% (-0.64%). Pressure during November is
significantly different from pressure in the neighboring months as illustrated by the
large t-statistics in the last line of Table 2. For example, total pressure in November
is 4 times larger (in magnitude) than pressure during December. There is no other
month in Table 2 with selling pressure of similar magnitude.
Table 2 also reports total pressure as a fraction of average turnover in the past
12 months for each stock.7 In November we find that pressure represents 24% of
average turnover, i.e., the selling activity of AFPs is close to a quarter of trading
volume in the market. Figure 3 shows the histogram of total pressure (multiplied by
100) and total pressure as a fraction of turnover in October and November of 2007.
We can see a shift of the distribution to the left, which means that AFPs become net

                                                            
7
Turnover is dollar volume divided by market capitalization at the end of the month. We compute the
average turnover in the previous 12 months excluding outliers such as equity issues and block sales. In
practice we do this by first winsorizing turnover at the 1% level in the full sample and then taking
averages across the previous 12 months for each stock.

12 
sellers of many stocks in November, and not just of a few stocks that drive down the
average of pressure. There are some big sales that represent more than 100% of the
average monthly turnover for these stocks. From now on we focus on total pressure as
our main measure of pressure since all measures are highly correlated and results are
similar irrespective of which one we use.
The last column in Table 2 reports the average total pressure computed over
the universe of local equity mutual funds instead of pension funds. The Chilean mutual
fund industry developed later than the pension fund industry; hence the assets under
management are an order of magnitude smaller than those of the pension fund
industry. The portfolios of mutual funds are reported at the monthly frequency in the
web page of another regulating agency, the Superintendencia de Valores y Seguros
(www.svs.cl), which is the Chilean equivalent to the SEC in the U.S. Although the
mutual fund industry could have been a liquidity provider in this event, i.e., buying
shares unloaded from the pension fund portfolios, we do not find that their buying
behavior is strong enough to compensate for the selling pressure of pension funds.
During November 2007 mutual funds put buying pressure on the average stock, but it
is simply too small to matter (0.000038%) in comparison to the selling pressure from
pension funds.
Foreign investors are better candidates than mutual funds to be the liquidity
providers in this event. We do not directly observe their portfolios; hence we cannot
be absolutely sure about their role, but we do have some suggestive evidence. Chilean
law requires that all firms list in annual reports their 12 largest shareholders, together
with the ownership stake of each shareholder at the end of the year. Pension funds
appear on this list for most Chilean stocks. Foreign investors do not appear directly,
but the custodian banks for foreigners and ADRs appear on the list. So, we use the
shares in the hands of custodian banks as a proxy for foreign ownership.8 We only
have access to the 12 largest shareholders, so it is possible that we do not observe some
pension funds or foreigners because they fall below this line.
In Table 3 we show the change in ownership between the end of 2006 and the
end of 2007 for pension funds and foreigners as two separate groups. The table covers
the large Chilean stocks for which we were able to make this decomposition. The

                                                            
8
 Citibank was the custodian bank for direct equity investments at the time (chapter XIV under Chilean
law). Depository banks for ADRs include Citibank, Bank of New York, and JP Morgan Chase among
others. 

13 
average change in pension fund ownership in these stocks is —1.21%, while the average
change in foreign ownership is 0.63%. The correlation between both changes is —0.33,
meaning that foreign ownership on average increased as pension fund ownership
decreased. This can be seen clearly in some of the Chilean blue chips such as Copec,
Endesa, and SQM-B. Overall, this evidence suggests that foreigners bought a
significant fraction of the shares dumped by pension funds in the fire sale.9
The existence of a fire sale in November of 2007 —even before documenting its
potential price impact– is hard to deny. As seen in Figure 1, excess investments come
down after October 2007, although this is likely a combination of both selling and price
effects since portfolio weights are affected by market prices. Eventually, all of the
remaining excess investments disappeared with the collapse of global equity markets
in September 2008. This made irrelevant the 12-month deadline (October 26, 2008)
initially imposed by the regulator.
Several market participants blamed the fire sale for the poor performance of the
Chilean stock market in comparison to other markets in the region (Latam). In
November the Chilean index (IPSA) fell more strongly than the Latam MSCI index (-
7.14% vs. -3.05%). In December 2007 both indexes had similarly poor returns (-4.76%
vs. -4.04%).10 These poor returns helped to eliminate some of the excess investments
of the pension funds. It is hard to pin down exactly how much (if any) of the aggregate
market return can be attributed to the fire sale. We focus on within market variation
in prices precisely to isolate our tests from the common shocks that could have affected
the entire Chilean market at the time of the fire sale.

3. Price Effects of the Fire Sale

We conduct an event study using raw returns and abnormal returns with similar
results. A challenge that we face to compute abnormal returns is that the fire sale is a
market-wide event, hence standard adjustments such as subtracting the market return,

                                                            
9
 Large controlling shareholders, which were also potential buyers during the fire sale, hardly changed
their ownership stakes during 2007. This is not surprising since the controlling blocks of Chilean
companies are extremely stable (Donelli, Larrain, and Urzua, 2013). 
10
 Returns on both of these indices are reported in Chilean pesos so currency fluctuations can also
explain some of the difference. 

14 
or the return from a market model, are not appropriate because the market itself is
affected by the event. Therefore, we first estimate a model for expected returns using
a pre-event panel (2002-2006) of monthly stock returns:

, , ,

, , , , 3

where , is the log of market equity at the end of the previous month,
, is the log of the book-to-market ratio measure with the timing of Fama
and French (1992),11 , is the return on the past 6 months (t-6 through t-1),
and , is the standard deviation of returns in the past 12 months (t-12
through t-1). We define the abnormal return as the difference between the raw return
and the predicted return obtained from the estimation of equation (3).12
Table 4 reports average raw returns (ARR), average abnormal returns (AAR),
their cumulative versions (CARR and CAAR), together with robust t-statistics. We
split the sample into “fire sale stocks” and “other stocks”. Fire sale stocks are defined
as those stocks with selling pressure above average during November 2007.13 Cross-
sectional averages of returns can be understood as the returns on equal-weighted
portfolios of each group of stocks. Both portfolios are hit hard during November 2007,
but the fire sale portfolio has a more negative return (-8.8% vs. -4.3%). The spread of
-4.4% is highly significant with a t-statistic of -3.43. The spread of -3.7% in terms of
abnormal returns is smaller, but still significant (t-statistic of -2.58). The spreads in
terms of CAR and CAAR during November are also large (-10.9% and -6.6%
respectively) and significant (t-statistics of -2.31 and -1.66 respectively).14

                                                            
11
 This means that returns from July of year t through to June of year t+1 are matched with the book-
to-market ratio from December of year t-1. This allows for delays in the availability of year-end book
values.
12
 The estimates of equation (3) are reported in Table 1 of the Online Appendix. 
13
 The fire sale portfolio has 58 stocks and other stocks portfolio has 35 stocks. As seen in Figure 3 the
cross-sectional distribution of pressure in November 2007 is negatively skewed. 
14
 Cumulative average returns on the right-hand-side of Table 4 differ slightly from the sum of average
returns on the left-hand-side of the same table because before accumulating returns throughout the
event window for a particular stock we replace missing monthly returns with zeros. Some small stocks
(most likely in the “other stocks” portfolio) have missing data. This is done to preserve the number of
observations instead of throwing away the entire time series of a stock because of one missing month.
Abnormal returns can also be missing because one of the variables in the expected return model is
missing for that stock. Our results are robust to simply dropping these stocks with missing returns. 

15 
 Dynamics are illustrated in Figure 4, which shows the spread in CARR and
CAAR between fire sale stocks and other stocks. The trough in CARR is precisely
during November 2007. The trough in CAAR is in January 2008. Up to February of
2008 price differentials stay relatively flat from the trough: the CARR (CAAR) in
February 2008 is still -10.1% (-6.2%) up from -10.9% (-6.6%) in November 2007, and
it is still statistically significant. In March the CARR (CAAR) is -7.9% (-3.9%), which
signals a start of reversal. This reversal is due to a significant return spread in March
in favor of fire sale stocks (3% in raw returns and 4.1% in abnormal returns). In
statistical terms the cumulative returns by March are not significantly different from
zero (see Table 4), so we can consider this as the month of reversal. However, in terms
of point estimates, by March prices have reverted only half-way through to their pre-
event levels. Full reversal in point estimates comes closer to May. Overall, we can say
that reversal takes between 4 to 6 months. These price dynamics resemble those
observed in Coval and Stafford (2007), although reversal is faster in our case.
The reversal of price differentials between fire sales stocks and other stocks
speaks against the informational hypothesis. Price effects should not reverse in the case
of a poor return based on new information regarding these stocks. The relatively slow
reversal also speaks against the short-run price pressure or liquidity hypothesis. Under
this hypothesis the pressure exerted by pension funds is accommodated by other
market participants by requiring a liquidity premium. But once the pressure
disappears, prices should revert to original levels. From Table 2 we can see that the
abnormal pressure from pension funds is specific to November of 2007 and not seen in
subsequent months. In particular, total pressure is not significantly different from
average in December ’07 (t-stat -1.45), February ’08 (t-stat 1.35), nor March ’08 (t-
stat -0.56), and even positive and above average in January ’08 (t-stat 2.26). The
challenge for the short-run price pressure hypothesis is then to explain why prices fail
to revert in December ’07 once the pressure from pension funds is over. The lack of an
immediate reversal points toward an explanation with downward sloping demands for
stocks (Shleifer, 1986) or slow-moving capital (Duffie, 2010) as suggested by the fire
sales literature.15

                                                            
15
 Since pension funds balanced their portfolios away from equity and into fixed income, we also look
for price effects in the local fixed income market. The sudden purchase (“fire purchase”) of bonds can
cause a fall in yields although we find no evidence in this regard, at least in the market of Chilean
government bonds that are heavily demanded by AFPs (see Figure 1 of the Online Appendix). The

16 
 4. Mitigation and Amplification Mechanisms during
the Fire Sale

4.1. Stock Selection

The regulator gave a general instruction to reduce the position in equities, but
allowed each pension fund administrator to pick what stocks to sell. In this section we
focus on their choices within domestic equities. The null hypothesis is that they
reduced all positions proportionally, i.e., there was no selection. For example, if pension
funds have to reduce the equity portfolio by 5%, the easiest thing to do is to reduce
every position by 5%. Under this hypothesis the cross-stock differences in pressure are
related to the ownership stake of pension funds in each stock. Simply put, if pension
funds reduce their portfolios proportionally, then they cause more pressure in those
assets where they have a larger investment relative to the amount standing. To see
this more clearly, recall our equation (1) and assume that all positions are reduced by
. Pressure in this case is:

1 , , , , , ,
, 4
∈ , ∈ ,

The last term in (4) is what we define as AFP ownership or the lagged total
stake of pension funds in a particular stock. Equation (4) shows that, under the null
hypothesis, cross-sectional differences in pressure should be solely related to AFP
ownership.
The alternative hypothesis is selection, i.e., pension funds choose to sell
relatively more of some stocks and less of others. We explore two main possibilities
here, although there may be others. First, pension funds can sell stocks according to
the costs of demanding liquidity in order to mitigate the price impact of their sales.
Depending on their expectations of future liquidity, it may be optimal to start selling

                                                            
fixed income market is deeper and bonds are more homogenous securities, which suggests smaller price
impact. Also, pension funds could have paced the purchases of bonds across a longer time period. 

17 
the more liquid securities or the less liquid securities (Brown, Carlin, and Lobo, 2010).
Second, pension funds can unwind previous bets (reduce over-weights or under-
weights) and reduce their distance to benchmarks and peer funds. A fire sale can be a
good excuse to close trades that did not work out; similarly, a fire sale can bring more
attention to the relative performance of pension fund portfolios and hence it is probably
not a good idea to deviate from the competition.
We explore several stock characteristics that are related to the selection
hypothesis. For instance, as suggested by Coval and Stafford (2007), the sale of large
stocks (measured by market capitalization) and stocks with high share turnover should
have relatively little price impact. Also, when the shareholder base is broad it is
probably less costly to sell a stock. In the spirit of Chen, Hong, and Stein (2002) we
compute the breadth of ownership as the fraction of pension fund administrators that
hold a particular stock. Similarly, when differences of opinion among investors are less
pronounced it is probably less costly to sell (Banerjee and Kremer, 2010). As a measure
of differences of opinion between pension funds we compute the cross-AFP standard
deviation of portfolio weights for a given stock. We normalize this standard deviation
by the mean weight of the stock across AFPs’ portfolios in the month; otherwise larger
stocks have a mechanically high level of disagreement.17
Breadth of ownership and differences of opinion can also be related to stock
selection in the second sense we suggest above. Stocks with low breadth and high
disagreement represent stronger bets within the pension fund industry: not all pension
funds have these stocks in their portfolios, or pension funds have varying levels of
investment in them. The local market or stock index is also a powerful reference point
for portfolio managers, hence we compute the overweight (OW) or underweight (UW)
of a given stock in the aggregate portfolio of pension funds in comparison to the free-
float adjusted market portfolio.18
In Table 5 we show results from cross-sectional regressions of the following type:

, ′ , , , 5

                                                            
17
 See Diether, Malloy, and Scherbina (2002) for related measures of differences of opinion. 
18
The main Chilean index or IPSA (Indice de Precio Selectivo de Acciones) uses weights that are
adjusted for free float.

18 
where pressure is measured in November 2007 and the characteristics are measured in
September of 2007.
The results in Table 5 show that, although AFP ownership is a very strong
predictor of pressure, and hence that there is evidence in favor of a proportional
reduction in pension fund portfolios, size is also significantly related to pressure no
matter what other controls we include in the regression. This suggests that there is
selection during the fire sale in favor of larger stocks, which can be expected since
larger stocks are more easily traded, and price impact in those stocks is much smaller.
Over-weights with respect to the market portfolio are not robust predictors of pressure
once we control for AFP ownership. The correlations between some of the variables in
these regressions are high (e.g., AFP ownership and breadth of ownership have a
correlation of 0.67), which suggests that it is hard to disentangle the null hypothesis
of proportional reduction against the alternative hypothesis of stock selection.22
In order to get a better understanding of our data we present pressure and other
stock characteristics for selected stocks in Table 6. These stocks represent some of the
largest and smallest stocks in the local index (IPSA) during 2007. Large stocks in Table
6 represent 54% of the free float of the Chilean market, which shows that this market,
like other emerging markets, is very concentrated in a few blue-chip stocks. AFP
ownership is higher in large stocks, but this correlation is far from mechanical since
ownership is related to the total shares outstanding of each stock, and not to the
weight of the stock in the market portfolio. For example, AFP ownership could be a
constant across stocks: if a portfolio manager wants to buy the market portfolio she
just needs to buy the same fraction, say 1%, of the shares outstanding of each stock.
Indeed, some of the stocks in Table 6 show that size and AFP ownership are not in a
one-to-one relationship. For example, AFPs own less of Copec than of Entel (8.8% vs.
24.8%), despite the fact that Copec is 4 times larger than Entel (12.6% vs. 2.8%). The
average pressure in the large stocks is -0.43%, which is more than twice in magnitude
the overall average pressure for the month of November (-0.21%). The average pressure
for small stocks is just -0.07%. As suggested by the multivariate analysis, some of these

                                                            
22
 Table 2 of the Online Appendix presents all the correlations between stock characteristics used in
this paper. Table 3 of the Online Appendix includes results for univariate regressions in the style of
equation (5) for several months before and after the fire sale. The main patterns are specific to the
event months. 

19 
differences in pressure can be attributed to AFP ownership, but also to selection
towards larger stocks during the fire sale.
These differences become apparent when we compare the changes in AFPs’
portfolios for large and small stocks. During November 2007 both small and large
stocks suffer selling pressure, but the pressure on large stocks is much stronger than
on small stocks (-0.34% v. -0.41%). The difference in pressure for small and large stocks
is significant during November (t-stat 2.36), but it is not significant in October or
December (t-stats 0.79 and 1.42 respectively.24
Taken as a whole we find evidence consistent with selection during the fire sale.
While it is true that as a first approximation pension funds reduced their portfolios
proportionally, there was also an attempt to mitigate the effects of the fire sale through
stock selection. Selling pressure was focused on larger stocks, in which price impact is
likely to be much smaller.

4.2. Other Mitigation Strategies

As seen in Figure 1 some funds had slack with respect to the legal investment
limits. In particular, in October of 2007 type-A funds where below the maximum
allowed for equity. An easy strategy to mitigate the need to sell was to transfer stocks
from the funds with excess investments (B, C, and D) to fund A. Regulation allowed
administrators to transfer assets between funds. It was only required that the trade
was marked at the market price observed at the time.
In Table 7 we find evidence that pension funds used this strategy early on
during the fire sale. Each month we report the correlation (in the cross-section of
stocks) between changes in shares held by fund A and by other funds. A negative
correlation means that type-A funds are buying stocks while the other funds are selling.
During October (remember that there were 3 trading days left in October after the
memo) we see a strong negative correlation of —0.80 between funds A and C, which
suggests that some of the equity in fund C was transferred to A. We do not see a
strong negative correlation during the previous month. In November all correlations
are high and positive. We view this as another sign that there was generalized selling
during November.

                                                            
24
 More detailed analysis is presented in Table 4 of the Online Appendix. 

20 
 This mitigation strategy faced, however, two limitations. First, type-A funds
were smaller than the other funds, in particular type-C funds. Second, the slack in
type-A funds was about 2%, while the excess investments in the other funds were much
larger (close to 10% in type-C funds). The measures of pressure that we compute in
equations (1) and (2) aggregate across all funds, and therefore are net of any mitigation
done through fund transfers.
In principle, pension funds could have made the adjustment required by law by
selling only foreign equity instead of domestic equity. We can only speculate about the
reasons why they decided to sell domestic equity so aggressively instead of selling only
foreign equity (they still probably sold some foreign equity).25 First, pension funds had
lobbied for an increase in the limits to foreign investment for a long time. In fact, these
limits were increased only months before the event. Selling foreign equity, which would
tilt portfolios even more towards domestic equity, was perhaps considered to be sub-
optimal in a mean-variance sense, or for reputational or political reasons regarding
their relationship with the regulator. Second, given the spectacular performance of the
Chilean stock market in the recent past, any preference for locking-in profits (Odean,
1998) would suggest selling domestic equity instead of foreign equity. Even if we
consider foreign equity as an alternative escape route, our results do not rely on
understanding why pension funds did not use it more aggressively. We simply focus
on the effects that the massive sales of equity by pension funds had on the cross-section
of local stocks.
Another mitigation strategy, or alternative strategy to what pension funds
actually did, was to wait and see instead of selling immediately. Sophisticated investors
such as pension funds could internalize the negative price impact of widespread selling
and avoid making things even worse. However, this ignores the strong incentives for
herding that exist in this market. Given the incentives derived from regulation, on top
of those provided by competition among funds, it was probably hard for any fund
manager to avoid running with the herd. Even worse, these incentives can make runs
a profitable strategy: sell as fast as possible, better if you manage to sell before others
can in order to accentuate the poor performance of competitors. If every fund manager

                                                            
25
 We do not know exactly how much foreign equity the pension funds sold. The data provided by the
regulator gives only dollar amounts invested in international assets (mutual funds, stocks, etc.), but not
the number of shares held as is reported for the domestic segment of the portfolio. Hence, we cannot
easily separate quantity and price effects as we did for domestic stocks. 

21 
acts like this, we will see a herd running as soon as the regulator’s memo is sent. In
the next section we explore the coordination that emerged across pension funds most
likely due to these incentives.

4.3. Amplification from Fund Outflows

A typical amplification mechanism in many fire sales is that poor initial returns
are followed by fund outflows, which prompt even more selling and even poorer returns.
As we see in Table 8, returns were indeed poor during November 2007, ranging between
-2.67% and -3.41% for fund C of each of the six pension funds in our sample. Although
we see some outflows in December 2007 and the following months, it is hard to pin
down these flows to cross-sectional differences in performance during the fire sale.26
For instance, the pension fund with the worst performance in November 2007 (Habitat)
had the second largest flow, in fact one of the only two positive flows, in December
2007 and among type-C funds. If anything, flows seem to mitigate more than amplify
the fire sale in this case. These results are confirmed using regressions of flows on past
return and past flows (to control for stickiness of flows as in Sialm, Stark, and Zhang,
2015), plus interactions of past returns and dummies for event months. We find that
although there is a small positive coefficient for past returns, during the event month,
and particularly in December 2007, the sensitivity is even smaller and almost zero.27
Overall, given the lack of amplification from flows as response to past returns,
our results represent a “pure” or “conservative” estimate of the price effects of a fire
sale.

4.4. Coordination Across Funds

As noted by Basak and Pavlova (2013), a high level of coordination can amplify
stock market volatility and the impact of a fire sale as investors act more highly
synchronized.

                                                            
26
Fund flows are computed from fund returns ( ) and assets under management ( ) as
1 / . 
27
 Results are available in Table 5 of the Online Appendix. 

22 
 We look at two different measures of coordination. First, we compute the
distance of pension fund i at time t with respect to the industry average in the following
way (see Cremers and Petajisto, 2009):

1
, , , , , 6
2
∈

Where , , is the weight of stock s in month t on the aggregate portfolio of

domestic equities of pension fund administrator i, and , is the cross-pension fund
average weight for that stock.29 Our second measure is sensitivity of the portfolio
changes of a given pension fund to the changes of other pension funds. We define for
each pension fund an average change in holdings as follows. Changes in holdings for a
particular stock are defined as in Table 2; we then take a value-weighted average of
these changes for a given pension fund in a given month. The value-weighted average
implies, for example, that a 10% reduction in a small stock with a portfolio weight of
1% is as important as a 1% reduction in a large stock with a portfolio weight of 10%.
Using simple regressions that allow us to estimate changes in active share and
sensitivity to other funds in the months around the event vis-à-vis baseline (normal
times) levels, we find that during the event there is a decrease in the active share, even
after controlling for fund fixed effects and for a linear time trend. The results for the
sensitivity to changes in other funds also indicate that during November 2007 there is
a significantly higher sensitivity to other funds’ changes and that this disappears in
the months after the event.30
Overall, it appears that the coordination of pension funds increased throughout
the fire sale, particularly early on in November 2007. It is noticeable that coordination
happens through informal interactions since it arises even before the portfolio of each
pension fund is made publicly available to its peers, which happens after more than a
month.

5. Real Effects of the Fire Sale

                                                            
29
 The Chilean regulation explicitly sets as benchmark for any fund the average of the pension fund
industry because penalties are defined in terms of underperformance with respect to this average. 
30
 Regression estimates are available in Tables 6 and 7 of the Online Appendix. 

23 
 Deviations of stock prices from fundamentals can have real effects in the sense
of affecting the investment decisions of firms (see, for instance, Baker, Stein, and
Wurgler, 2003; Campello and Graham, 2013; and Jotikasthira, Lundblad, and
Ramadorai, 2013). Although a reversal in stock prices of 4 to 6 months may be slow
from the perspective of market efficiency, it represents a relatively transitory
phenomenon from the perspective of real investment decisions. Still, in this section we
explore the potential real effects associated to our event.
We focus on quarterly investment given the transitory nature of the fire sale.32
Annual investment rates, in particular when comparing the years 2007 and 2008 would
be contaminated with the onset of the Lehman crisis in mid-2008. As proxies for
investment we study the growth rate of book assets and the change in property, plant,
and equipment (PPE) as a fraction of lagged assets. In a similar spirit as what Almeida
et al. (2011) do for the 2007 credit crisis in the U.S., we compare the last quarter of
2007 (the fire-sale quarter) with the first quarter of 2008 (the post-fire-sale quarter).
Given a reasonable lag for investment decisions, we do not expect to see any effect of
the fire sale before the first quarter of 2008.
In Table 9 we show the average investment for firms that are more and less
affected by the fire sale. We split firms into fire sale stocks and other stocks as in Table
4. We find small differences in investment in the last quarter of 2007 before the effects
of the fire sale are likely to be felt. In the first quarter of 2008 investment rates are
negative on average. More importantly, fire sale stocks present lower investment rates
than other stocks (-1.33% vs -0.01% in terms of asset growth; and -0.78% vs. -0.46%
in terms of changes in PPE). Hence, investment goes down precisely in those stocks
more affected by the fire sale. Unfortunately, none of these differences is statistically
significant at conventional levels, which in part can be explained by the relatively
small cross-section of firms. Overall, this evidence is only suggestive of real effects,
which is perhaps not surprising given the transitory nature of the shock, and it might
indicate that fire sales that are short-lived have limited impact on the real economy.
More permanent shocks, or repeated shocks, as those studied in Jotikasthira, Lundblad,
and Ramadorai (2013), can have larger and more statistically significant effects.

                                                            
32
Quarterly financial statements are obtained from Economatica. Not all firms in our sample report
quarterly financial statements.

24 
  

6. Conclusions

We study an asset fire sale triggered by regulatory constraints in an emerging

market. A change in the regulation of investment limits in Chilean private pension
funds caused massive sales of domestic equity in November of 2007. Stocks that
received more selling pressure from pension funds lost approximately 4% compared to
other stocks during that month. Prices reverted after four to six months. Our results
provide a cleanly identified price impact of fire sales, together with an estimation of
the horizon at which price reversals occur. In order to mitigate the impact of the fire
sale pension funds sold stocks in which demanding liquidity was less costly. In
particular, the sale was focused on large index stocks. We also find an increase in
coordination across the equity portfolios of pension funds.
When looking at firms’ investment decisions we find no significant evidence of
real effects of this (short-lived) fire sale, in spite of its marked impact on returns.

25 
References

Almeida, Heitor, Murillo Campello, Bruno Laranjeira, and Scott Weisbenner,

2011, Corporate debt maturity and the real effects of the 2007 credit crisis, Critical
Finance Review 1, 3-58.
Baker, Malcolm, Jeremy Stein, and Jeffrey Wurgler, 2003, When does the
market matter? Stock prices and the investment of equity-dependent firms, Quarterly
Journal of Economics, 118, 968-1005.
Banerjee, Snehal, and Ilan Kremer, 2010, Disagreement and learning: Dynamic
patterns of trade, Journal of Finance, 65(4), 1269-1302.
Basak, Suleyman, and Anna Pavlova, 2013, Asset prices and institutional
investors, American Economic Review, 103, 1728-1758.
Bekaert, Geert, Campbell Harvey, Christian Lundblad, and Stephan Siegel,
2011, What segments equity markets? Review of Financial Studies, 24(12) 3847-3890.
Brown, David, Bruce Carlin, and Miguel Lobo, 2010, Optimal Portfolio
Liquidation with Distress Risk, Management Science 56(11), 1997-2014.
Brunnermeier, Markus, and Lasse Pedersen, 2005, Predatory trading, Journal
of Finance 60(4), 1825-1863.
Campello, Murillo, and John R. Graham, 2013, Do Stock Prices Influence
Corporate Decisions? Evidence from the Technology Bubble, Journal of Financial
Economics, 107(1), 89-110.
Chen, Joseph, Harrison Hong, and Jeremy Stein, 2002, Breadth of ownership
and stock returns, Journal of Financial Economics, 66(2-3). 171-205.
Chen, Joseph, Greg Noronha and Vijay Singal, 2004, The price response to S&P
500 index additions and deletions: Evidence of asymmetry and a new explanation,
Journal of Finance, 59(4), August, 1901-1929.
Cooper, Michael, Orlin Dimitrov, and Raghavendra Rau, 2001, A rose.com by
any other name, Journal of Finance, 56(6), November, 2377-2388.
Coval, Joshua and Erik Stafford, 2007, Asset fire sales (and purchases) in equity
markets, Journal of Financial Economics, 86(2), November, 479-512.
Cremers, Martijn, and Antti Petajisto, 2009, How active is your mutual fund
manager? A new measure that predicts performance, Review of Financial Studies,
22(9), 3329-3365.

26 
 DeLong, Bradford, Andrei Shleifer, Lawrence Summers, and Robert Waldmann,
1990, Noise Trader Risk in Financial Markets, Journal of Political Economy 98, 703-
738.
Diether, Karl, Christopher Malloy, and Anna Scherbina, 2002, Differences of
opinion and the cross section of stock returns, Journal of Finance, 57(5), October,
2113-2141,
Donelli, Marcelo, Borja Larrain, and Francisco Urzua, 2013, Ownership
dynamics with large shareholders: An empirical analysis, Journal of Financial and
Quantitative Analysis, 48, 579-609.
Duffie, Darrell, 2010, Asset price dynamics with slow-moving capital, Journal
of Finance, 65(4), 1237-1267.
Edmans, Alex, Itay Goldstein, and Wei Jiang, 2012, The real effects of financial
markets: The impact of prices on takeovers, Journal of Finance 67(3), 933-971.
Fama, E., French, K., 1992. The cross section of expected stock returns. Journal
of Finance 47, 427–465.
Frazzini, Andrea, and Owen Lamont, 2008, Dumb money: Mutual fund flows
and the cross-section of stock returns, Journal of Financial Economics, 88(2), 299-322.
Greenwood, Robin, 2005, Short- and long- term demand curves for stocks:
Theory and evidence on the dynamics of arbitrage, Journal of Financial Economics,
75(3), March, 607-649.
Harris, Lawrence, and Eitan Gurel, 1986, Price and volume effects associated
with changes in the S&P 500 list: New evidence for the existence of price pressures,
Journal of Finance, 41(4), September, 815-829.
Jotikasthira, Chotibhak, Christian Lundblad, and Tarun Ramadorai, 2012,
Asset fire sales and purchases and the international transmission of funding shocks,
Journal of Finance, 67, 2015-2050.
Jotikasthira, Chotibhak, Christian Lundblad, and Tarun Ramadorai, 2013, How
do foreign investors impact domestic economic activity? Evidence from China and
India, Journal of International Money and Finance, 39, 89-110.
Kaul, Aditya, Vikas Mehrotra, and Randall Morck, 2000, Demand curves for
stocks do slope down: New evidence from and index weights adjustment, Journal of
Finance, 55, 893-912.
Lou, Dong, 2012, A flow-based explanation for return predictability, Review of
Financial Studies, 25(12), 3457-3489.

27 
 Merton, Robert, 1987, A simple model of capital market equilibrium with
incomplete information, Journal of Finance, 42(3), July, 483-510.
Mitchell, Mark, Lasse Pedersen, and Todd Pulvino, 2007, Slow moving capital,
American Economic Review 97(2), 215-200.
Pontiff, Jeffrey, 2006, Costly Arbitrage and the Myth of Idiosyncratic Risk,
Journal of Accounting and Economics 42(2), 35-52.
Scharfstein, David, and Jeremy Stein, 1990, Herd behavior and investment,
American Economic Review, 80(3), June, 465-479.
Shleifer, Andrei, 1986, Do demand curves for stocks slope down? Journal of
Finance, 41(3), 579-590.
Shleifer, Andrei, and Robert Vishny, 1992, Liquidation values and debt
capacity: A market equilibrium approach, Journal of Finance, 47(4), 1343-1366.
Shleifer, Andrei, and Robert Vishny 1997, The limits of arbitrage, Journal of
Finance 52, 35–55.
Shleifer, Andrei, and Robert Vishny, 2011, Fire sales in finance and
macroeconomics, Journal of Economic Perspectives, 25(1), 29-48.
Sialm, Clemens, Laura Starks, and Hanjiang Zhang, 2015, Defined contribution
pension funds: Sticky or discerning money? Journal of finance 70(2), 805-838.
Wurgler, Jeffrey, and Ekaterina Zhuravskaya, 2002, Does arbitrage flatten
demand curves for stocks? Journal of Business, 75(4), 583-608.

28 
 Table 1. Investment Limits set by Chilean Law on the Portfolios of Pension Funds 
The table contains the minimum and maximum portfolio weights for equity (domestic plus international) allowed in the five types of fund available in the Chilean 
pension fund system during 2007‐2008. Each pension fund administrator has to offer funds A through E. 
 
 
Limits on Equity as % of Total Fund
Type of Fund Min (%) Max (%)
Fund A 40 80
Fund B 25 60
Fund C 15 40
Fund D 5 20
Fund E 0 0  
 
   
 Table 2. Changes in Holdings of Domestic Equity in the Portfolios of Pension Funds and Pressure around November 2007 
Panel A shows the fraction of positions reduced, maintained or increased each month between May 2007 and August 2008 in the portfolios of pension fund 
administrators (AFPs). The monthly position in each stock in the aggregate portfolio of each pension fund represents one observation. Panel B shows the average 
change in holdings, conditional on reducing or increasing the position. The change in holdings is measured as the percentage change in the number of shares 
held of each stock. Panel C reports the average pressure across all stocks for which at least one AFP had holdings on that particular month. Total Pressure, as 
defined in equation (1) in the text, considers for each stock the aggregate buying and selling from pension funds during the month as a fraction of total shares 
outstanding for the company. Free Pressure, as defined in equation (2) in the text, adjusts shares outstanding for free float. Total pressure and free pressure are 
multiplied by 100. Total pressure/turnover divides total pressure by average turnover of the previous 12 months for the particular stock. The last column shows 
the same definition of total pressure in equation (1) but applied to the holdings of domestic equity by local mutual funds. In every panel the average reported at 
the bottom corresponds to the average for all months. The t‐test corresponds to a difference in means test, comparing the average of each variable for November 
2007 against the average for all other months pooled together. 
 
Panel A. Fraction Panel B. Average Panel C. Measures 
of positions change in holdings  of Pressure
Month Reduced Maintained Increased Reduced Increased Total Free Total/Turnover Mutual Funds
May‐07 0.13 0.67 0.20 ‐0.05 0.05 0.13 0.35 0.01 0.000054
Jun‐07 0.13 0.70 0.17 ‐0.03 0.21 0.04 0.34 0.11 0.000097
Jul‐07 0.17 0.71 0.13 ‐0.02 0.05 ‐0.01 ‐0.02 ‐0.01 0.000021
Aug‐07 0.17 0.67 0.17 ‐0.04 0.19 0.02 0.20 ‐0.06 ‐0.000009
Sep‐07 0.20 0.72 0.09 ‐0.04 0.13 ‐0.01 ‐0.06 ‐0.02 0.000046
Oct‐07 0.17 0.68 0.15 ‐0.05 0.09 ‐0.01 0.03 ‐0.09 ‐0.000015
Nov‐07 0.30 0.61 0.09 ‐0.11 0.07 ‐0.21 ‐0.64 ‐0.24 0.000038
Dec‐07 0.24 0.65 0.11 ‐0.04 0.05 ‐0.05 ‐0.27 ‐0.04 ‐0.000033
Jan‐08 0.21 0.60 0.19 ‐0.08 0.83 0.07 0.15 0.02 ‐0.000150
Feb‐08 0.18 0.65 0.17 ‐0.03 1.81 0.02 ‐0.07 0.01 ‐0.000039
Mar‐08 0.17 0.63 0.20 ‐0.04 5.92 ‐0.09 ‐0.45 ‐0.19 ‐0.000046
Apr‐08 0.12 0.67 0.21 ‐0.04 1.94 0.02 0.24 ‐0.05 ‐0.000100
May‐08 0.24 0.61 0.15 ‐0.07 0.04 ‐0.07 ‐0.57 ‐0.05 0.000014
Jun‐08 0.23 0.62 0.14 ‐0.05 0.24 ‐0.06 ‐0.43 ‐0.12 ‐0.000081
Jul‐08 0.20 0.64 0.16 ‐0.07 0.10 ‐0.05 ‐0.11 ‐0.09 ‐0.000070
Aug‐08 0.15 0.68 0.17 ‐0.03 0.34 0.04 0.06 0.03 ‐0.000041
Average 0.18 0.66 0.16 ‐0.05 0.82 ‐0.01 ‐0.08 ‐0.05 ‐0.000019
t‐stat test Nov 2007=average 3.62 1.48 4.17 2.40 1.77 4.71 3.91 2.83 1.29  
   
 Table 3. Changes (2007‐2006) in Pension Fund Ownership and Foreign Ownership of Selected Chilean Stocks 
For each stock we report the change in pension fund ownership and foreign ownership between December of 2006 and December in 2007 as reported in annual 
reports. Both pension funds and foreigners are reported as a group in this table. Foreign ownership includes direct holdings and ADRs, both identified through 
custodian banks. 

Stock Change in Pension  Change in Foreign

Fund Ownership (%) Ownership (%)
Enersis ‐1.05 ‐0.55
Cencosud ‐8.11 ‐1.02
Endesa ‐1.58 0.80
SQM‐B ‐2.52 5.75
Copec ‐0.45 0.60
Lan 0.00 ‐3.30
Colbun ‐0.02 0.57
Iansa ‐1.19 0.00
CTC‐A ‐4.10 4.80
Ripley ‐0.35 0.40
La Polar ‐4.82 3.16
D&S ‐8.16 1.77
Andina ‐0.90 ‐1.30
Antarchile 0.01 0.28
Banco Chile 0.15 ‐0.30
Banco Santander 0.29 0.38
CAP 0.27 ‐0.34
CCU ‐0.77 1.05
CMPC ‐0.99 1.76
CorpBanca 0.04 ‐3.78
Entel ‐0.09 1.90
Madeco 4.40 ‐0.86
Masisa ‐0.71 2.95
Vapores 1.63 0.32
Average ‐1.21 0.63
Correlation ‐0.33  
 
 
 Table 4. Event Study for Price Impact of Fire Sale 
This table presents the event study analysis for raw and abnormal returns. Panel A considers raw returns and Panel B uses abnormal returns. Abnormal returns 
are computed using the estimated coefficients in column 5 of Table 4. The first three columns in each panel correspond to average returns (AR), and the next 
three columns correspond to cumulative returns (cumulative average raw returns, CARR, in panel A, and cumulative average abnormal returns, CAAR, in panel 
B). Columns marked as “fire sale stocks” correspond to the average return for stocks with total pressure below average in November 2007. Columns marked as 
“other stocks” correspond to stocks with total pressure above average in November 2007. Columns marked “spread” correspond to the difference in average 
return between the two groups. Robust t‐stats are reported in parenthesis. Significance at *10%, **5%, ***1%. 
 
Panel A. Raw Returns (RR)
ARR CARR
Fire Sale Other Fire Sale Other
Month
Stocks Stocks Spread Stocks Stocks Spread
May‐07 0.027* 0.044*** ‐0.017 0.027* 0.044*** ‐0.017
(1.973) (3.424) (‐0.891) (1.970) (3.418) (‐0.911)
Jun‐07 0.078*** 0.060*** 0.018 0.104*** 0.099*** 0.005
(5.871) (5.407) (1.066) (5.573) (6.276) (0.209)
Jul‐07 ‐0.029*** 0.007 ‐0.036*** 0.073*** 0.106*** ‐0.032
(‐3.074) (1.041) (‐3.078) (3.522) (6.565) (‐1.227)
Aug‐07 ‐0.014 0.003 ‐0.018 0.059** 0.107*** ‐0.048
(‐1.230) (0.395) (‐1.224) (2.196) (5.896) (‐1.482)
Sep‐07 ‐0.012 0.016 ‐0.028* 0.047 0.122*** ‐0.074*
(‐1.334) (1.371) (‐1.899) (1.502) (5.201) (‐1.896)
Oct‐07 0.030** 0.020** 0.011 0.077** 0.142*** ‐0.065
(2.674) (2.256) (0.743) (2.130) (5.364) (‐1.447)
Nov‐07 ‐0.088*** ‐0.043*** ‐0.044*** ‐0.010 0.098*** ‐0.109**
(‐9.078) (‐5.037) (‐3.430) (‐0.269) (3.534) (‐2.309)
Dec‐07 ‐0.038*** ‐0.037*** ‐0.000 ‐0.048 0.059* ‐0.107**
(‐5.447) (‐5.568) (‐0.043) (‐1.276) (1.960) (‐2.222)
Jan‐08 ‐0.095*** ‐0.089*** ‐0.006 ‐0.147*** ‐0.039 ‐0.108**
(‐7.390) (‐8.068) (‐0.329) (‐3.572) (‐1.162) (‐2.049)
Feb‐08 0.039** 0.029** 0.010 ‐0.103** ‐0.002 ‐0.101*
(2.581) (2.557) (0.519) (‐2.635) (‐0.052) (‐1.866)
Mar‐08 0.005 ‐0.025** 0.030* ‐0.098** ‐0.019 ‐0.079
(0.358) (‐2.636) (1.796) (‐2.184) (‐0.497) (‐1.345)
Apr‐08 0.053*** 0.039*** 0.014 ‐0.045 0.014 ‐0.059
(3.373) (3.899) (0.742) (‐0.917) (0.355) (‐0.935)
May‐08 0.028 ‐0.003 0.031 ‐0.017 0.008 ‐0.025
(1.065) (‐0.280) (1.098) (‐0.280) (0.194) (‐0.341)
Jun‐08 ‐0.004 ‐0.004 0.000 ‐0.021 0.005 ‐0.026
(‐0.130) (‐0.184) (0.006) (‐0.245) (0.100) (‐0.261)
Jul‐08 ‐0.013 ‐0.011 ‐0.002 ‐0.034 ‐0.005 ‐0.028
(‐1.012) (‐0.962) (‐0.125) (‐0.413) (‐0.099) (‐0.290)
Aug‐08 ‐0.018 ‐0.023* 0.005 ‐0.052 ‐0.030 ‐0.021
(‐1.510) (‐1.977) (0.291) (‐0.638) (‐0.555) (‐0.217)  
 
 Table 4. Event Study for Price Impact of Fire Sale 
(cont.) 
 
 
Panel B. Abnormal Returns (AR)
AAR CAAR
Fire Sale Other Fire Sale Other
Month
Stocks Stocks Spread Stocks Stocks Spread
May‐07 0.005 0.021 ‐0.016 0.005 0.014 ‐0.009
(0.370) (1.456) (‐0.781) (0.371) (1.449) (‐0.574)
Jun‐07 0.058*** 0.034** 0.024 0.057*** 0.035*** 0.022
(3.993) (2.497) (1.217) (3.165) (3.385) (1.058)
Jul‐07 ‐0.048*** ‐0.021** ‐0.027** 0.014 0.021* ‐0.008
(‐4.832) (‐2.316) (‐2.031) (0.704) (1.900) (‐0.349)
Aug‐07 ‐0.031** ‐0.018 ‐0.013 ‐0.014 0.009 ‐0.023
(‐2.435) (‐1.648) (‐0.807) (‐0.564) (0.628) (‐0.800)
Sep‐07 ‐0.029*** ‐0.011 ‐0.018 ‐0.040 0.002 ‐0.041
(‐2.984) (‐0.769) (‐1.061) (‐1.364) (0.093) (‐1.218)
Oct‐07 0.018 ‐0.006 0.024 ‐0.024 ‐0.002 ‐0.022
(1.633) (‐0.551) (1.581) (‐0.729) (‐0.109) (‐0.570)
Nov‐07 ‐0.100*** ‐0.063*** ‐0.037** ‐0.112*** ‐0.046** ‐0.066*
(‐9.514) (‐6.631) (‐2.575) (‐3.330) (‐2.132) (‐1.664)
Dec‐07 ‐0.050*** ‐0.059*** 0.009 ‐0.157*** ‐0.089*** ‐0.068
(‐6.432) (‐6.839) (0.774) (‐4.626) (‐3.784) (‐1.661)
Jan‐08 ‐0.104*** ‐0.101*** ‐0.004 ‐0.253*** ‐0.168*** ‐0.086*
(‐7.954) (‐8.556) (‐0.215) (‐6.647) (‐5.639) (‐1.773)
Feb‐08 0.035** 0.013 0.022 ‐0.219*** ‐0.157*** ‐0.062
(2.082) (1.072) (1.058) (‐6.330) (‐4.923) (‐1.315)
Mar‐08 0.003 ‐0.038*** 0.041** ‐0.216*** ‐0.177*** ‐0.039
(0.208) (‐3.542) (2.337) (‐5.623) (‐5.405) (‐0.780)
Apr‐08 0.049*** 0.021* 0.029 ‐0.172*** ‐0.162*** ‐0.010
(2.888) (1.867) (1.397) (‐4.141) (‐4.988) (‐0.187)
May‐08 0.016 ‐0.012 0.028 ‐0.158*** ‐0.171*** 0.013
(0.557) (‐1.035) (0.919) (‐2.812) (‐5.412) (0.199)
Jun‐08 ‐0.008 ‐0.031 0.023 ‐0.165** ‐0.190*** 0.025
(‐0.256) (‐1.440) (0.615) (‐2.098) (‐4.952) (0.290)
Jul‐08 ‐0.022 ‐0.016 ‐0.006 ‐0.185** ‐0.200*** 0.015
(‐1.487) (‐1.281) (‐0.295) (‐2.495) (‐4.996) (0.173)
Aug‐08 ‐0.034** ‐0.044*** 0.010 ‐0.217*** ‐0.230*** 0.013
(‐2.640) (‐2.856) (0.513) (‐2.994) (‐5.349) (0.154)
 Table 5. Multivariate Regressions of Pressure on Stock Characteristics 
The table reports regressions of total pressure in November 2007 on stock characteristics measured in September 2007. Total pressure is defined in equation (1) 
in the main text. Robust standard errors are reported in parenthesis. *: significant at 10mm%, **: significant at 5%, ***: significant at 1%. 
 
Dependent variable: Total Pressure in Nov‐07
Characteristic in Sep‐07: (1) (2) (3) (4) (5)
AFP Ownership ‐1.931** ‐1.963** ‐1.729** ‐2.284
(0.749) (0.826) (0.860) (1.461)
Log Mkt Cap ‐0.042** ‐0.066*** ‐0.045** ‐0.075*
(0.017) (0.016) (0.018) (0.043)
OW/UW in AFP Portfolio ‐1.791 ‐6.839*** ‐2.229 ‐0.906
(1.905) (1.810) (2.064) (2.550)
Log B/M ‐0.037
(0.073)
Momentum 0.212
(0.194)
Volatility ‐1.294
(2.825)
Share turnover 0.556
(4.077)
Breadth of Ownership 0.115
(0.607)
Differences of opinion ‐0.051
(0.258)

Observations 90 85 87 85 67
R‐squared 0.179 0.152 0.099 0.176 0.162  
   
 Table 6. Pressure and Other Stock Characteristics for Selected Stocks in the Santiago Stock Exchange Index (IPSA) 
Market weight is the market capitalization of a stock divided by total capitalization of all stocks in the Chilean market. Free float market weight excludes stakes 
of controlling shareholders. Pension funds’ portfolio weight is the fraction of AFP portfolios invested in a particular stock. OW/UW is the over‐weight or under‐
weight computed as the difference between the two previous variables. AFP ownership is the total number of shares of a particular stock held by pension funds 
over total shares outstanding of that stock. Total and free pressure are computed as in equations (1) and (2) in the main text and are multiplied by 100. 
 
(1) (2)
Total Free Float Pension Funds'  AFP  Total Free 
Mkt Weight Mkt Weight Portfolio Weight OW/UW Ownership Pressure Pressure
Stock (Sept. 2007) (Sept. 2007) (Sept. 2007) (2)‐(1)  (Sept. 2007) (Nov. 2007) (Nov. 2007)
Large Stocks
COPEC 11.3% 12.6% 10.0% ‐2.6% 8.8% ‐0.24 ‐0.62
ENDESA 6.3% 7.2% 13.0% 5.9% 20.4% ‐0.59 ‐1.46
ENERSIS 6.2% 6.9% 11.7% 4.8% 18.4% ‐0.67 ‐1.72
CENCOSUD 4.2% 4.3% 5.0% 0.6% 12.0% ‐0.46 ‐1.28
CMPC 3.8% 5.0% 6.9% 1.9% 17.5% ‐0.60 ‐1.32
SQM‐B 2.9% 4.7% 3.1% ‐1.6% 11.9% ‐0.50 ‐0.89
LAN 2.9% 3.7% 0.9% ‐2.9% 3.2% ‐0.07 ‐0.16
CAP 2.3% 4.5% 4.0% ‐0.5% 18.9% ‐0.51 ‐0.73
ENTEL 2.2% 2.8% 5.5% 2.7% 24.5% ‐0.25 ‐0.56
COLBUN 1.8% 2.6% 2.0% ‐0.6% 10.6% ‐0.41 ‐0.79
Average 4.4% 5.4% 6.2% 0.8% 14.6% ‐0.43 ‐0.95

Small Stocks
QUINENCO 1.3% 0.6% 0.2% ‐0.4% 1.2% ‐0.04 ‐0.22
RIPLEY 1.2% 0.6% 0.3% ‐0.3% 2.7% ‐0.05 ‐0.28
ALMENDRAL 0.8% 0.5% 0.1% ‐0.4% 1.9% 0.00 0.00
IAM 0.7% 0.8% 0.3% ‐0.5% 4.0% ‐0.15 ‐0.35
INVERCAP 0.7% 0.9% 0.0% ‐0.9% 0.2% ‐0.02 ‐0.04
SONDA 0.6% 0.7% 0.1% ‐0.5% 2.7% ‐0.17 ‐0.42
BANMEDICA 0.6% 0.7% 0.1% ‐0.6% 2.8% 0.00 0.00
SECURITY 0.5% 0.3% 0.2% ‐0.1% 3.9% ‐0.07 ‐0.34
SK 0.4% 0.2% 0.2% ‐0.1% 4.6% ‐0.21 ‐0.99
EMBONOR‐B 0.4% 0.5% 0.0% ‐0.5% 0.1% 0.00 0.00
Average 0.7% 0.6% 0.2% ‐0.4% 2.4% ‐0.07 ‐0.26  
   
 Table 7. Correlations of Changes in Shares Held by Funds of Different Risk Profiles 
Each month (September through November 2007) we compute the correlation across stocks of changes in shares held by different types of funds (A through 
D). We first aggregate all funds of a given type into a single number of shares held of each stock by the pension fund industry, and then we take changes 
between consecutive months. 
 
Correlations between Changes in Shares held by Fund:
Sep‐07 A B C D
A 1.00
B 0.91 1.00
C 0.01 0.00 1.00
D ‐0.01 0.00 0.87 1.00

Oct‐07 A B C D
A 1.00
B 0.14 1.00
C ‐0.80 0.28 1.00
D 0.01 0.98 0.39 1.00

Nov‐07 A B C D
A 1.00
B 0.98 1.00
C 0.98 0.95 1.00
D 0.90 0.95 0.90 1.00  
 
   
 
Table 8. Pension Fund Returns and Flows around November 2007 
This table shows the monthly returns and flows to Fund C (the biggest fund) for each one of six pension fund administrators (AFPs) in the Chilean market between 
October 2007 and March 2008. Flows are reported as percentage of assets under management of the previous month. 
 
Pension Fund Administrator (AFP)
Bansander Cuprum Habitat Planvital Provida  Santa Maria
Returns (%)
Oct‐07 4.37 3.78 4.41 4.48 4.15 4.30
Nov‐07 ‐3.25 ‐2.67 ‐3.41 ‐3.40 ‐3.04 ‐3.23
Dec‐07 ‐0.67 ‐0.22 ‐0.39 ‐0.44 ‐0.55 ‐0.39
Jan‐08 ‐4.88 ‐4.14 ‐4.86 ‐5.31 ‐4.84 ‐5.15
Feb‐08 3.21 3.00 3.34 3.21 3.11 3.23
Mar‐08 ‐1.50 ‐1.32 ‐0.93 ‐1.24 ‐1.34 ‐1.27

Flows (%)
Oct‐07 ‐0.81 0.17 ‐0.09 ‐0.48 ‐0.29 ‐0.38
Nov‐07 ‐0.84 0.09 ‐0.13 ‐0.59 ‐0.39 ‐0.40
Dec‐07 ‐0.54 0.10 0.08 ‐0.42 ‐0.16 ‐0.32
Jan‐08 ‐0.14 1.13 0.07 ‐0.02 0.33 ‐0.04
Feb‐08 0.18 0.70 0.23 ‐0.03 ‐0.03 0.13
Mar‐08 ‐0.23 0.57 0.23 0.02 0.21 0.01  
   
  
Table 9. Real Effects of the Fire Sale 
We show the average book asset growth and change in property, plant, and equipment over lagged assets for fire sale stocks and other stocks as defined in Table 
5. We report averages for the last quarter of 2007 (fire‐sale quarter) and the first quarter of 2008 (post‐fire‐sale quarter). All data come from quarterly financial 
statements obtained from Economatica. 
 
 
 
Asset Growth Change PPE/ Assets # firms
2007 Q4 (fire‐sale quarter)
Fire Sale Stocks 2.98% 1.88% 29
Other Stocks 2.75% 1.89% 44

2008 Q1 (post‐fire‐sale quarter)
Fire Sale Stocks ‐1.33% ‐0.78% 29
Other Stocks ‐0.01% ‐0.46% 41  
 
   
 
Figure 1. Aggregate Excess Investment in Equity by Type of Fund  
Excess investment is measured as the difference between the share of equity in the aggregate portfolio of all AFPs and the legal limit. Numbers are computed 
separately for each one of the four types of funds (type‐E funds were not allowed to hold equity at the time). All numbers correspond to percentages of the total 
fund. The line “Oficio Berstein” marks the date of the memo sent by the regulator. All data on holdings come from data reported to the Superintendencia de 
Pensiones. 
 

 
   
  
 
Figure 2. Excess Investment in Equity by Pension Fund Administrator and Type of Fund  
Excess  investment  is  measured  as  the difference  between  the  share  of  equity  in  the  portfolio  of  each  one  of the  six  AFPs and  the  legal  limit.  Numbers  are 
computed separately for each one of the four types of funds (type‐E funds were not allowed to hold equity at the time). All numbers correspond to percentages 
of the total fund. All data on holdings come from data reported to the Superintendencia de Pensiones. 
 

Fund A Fund B

10
excess investment (%)

excess investment (%)

-8 -6 -4 -2 0 2

-5 0 5
2007m9 2007m11 2008m1 2008m3 2007m9 2007m11 2008m1 2008m3
Date Date

BANSANDER CUPRUM BANSANDER CUPRUM

HABITAT PLANVITAL HABITAT PLANVITAL
PROVIDA SANTA MARIA PROVIDA SANTA MARIA

Fund C Fund D

15
excess investment (%)

excess investment (%)

15

10
10

5
0 5

2007m9 2007m11 2008m1 2008m3 2007m9 2007m11 2008m1 2008m3

Date Date

BANSANDER CUPRUM BANSANDER CUPRUM

HABITAT PLANVITAL HABITAT PLANVITAL
PROVIDA SANTA MARIA PROVIDA SANTA MARIA

 
 
 
   
 
Figure 3. Cross‐sectional Distribution of Pension Fund Pressure in October and November of 2007 
This figures show the histogram of total pressure and pressure as a fraction of turnover in October and November of 2007. Each stock represents one observation 
in each month.  
 

100

100
80

80
60

60
Percent

Percent
40

40
20

20
0

0
-2 -1 0 1 2 3 -4 -2 0 2
Total Pressure Pressure/Turnover

Oct 2007 Nov 2007 Oct 2007 Nov 2007

 
 
   
 Figure 4. Spread in Cumulative Average Returns between Fire Sale Stocks and Other Stocks 
This figure shows the time series of the difference in cumulative average raw returns (CARR) and cumulative average abnormal returns (CAAR) between the fire 
sale stocks and other stocks around the event (November 2007). Abnormal returns are calculated using the model for expected returns in Table 4, column 5. 
 
0.04

0.02

‐0.02

CARR
‐0.04
CAAR

‐0.06

‐0.08

‐0.1

‐0.12