HOW SAFE ARE MONEY MARKET FUNDS?

Marcin Kacperczyk and Philipp Schnabl

We examine the risk-taking behavior of money market funds during the

financial crisis of 2007–2010. We find that (1) money market funds experienced
an unprecedented expansion in their risk-taking opportunities; (2) funds had
strong incentives to take on risk because fund inflows were highly responsive to
fund yields; (3) funds sponsored by financial intermediaries with more money

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

fund business took on more risk; and (4) funds suffered runs as a result of their
risk taking. This evidence suggests that money market funds lack safety be-
cause they have strong incentives to take on risk when the opportunity arises
and are vulnerable to runs. JEL Codes: G21, G23, E44.

I. Introduction
Money market funds were at the center of attention during
the financial crisis of 2007–2010. Following the default of
Lehman Brothers in September 2008, a well-known fund—the
Reserve Primary Fund—suffered a run due to its holdings of
Lehman’s commercial paper. This run quickly spread to other
funds, triggering investors’ redemptions of more than $300 billion
within a few days after Lehman’s default. Its consequences ap-
peared so dire to financial stability that the U.S. government
decided to intervene by providing unlimited insurance to all

* We thank three anonymous referees and the editors: Larry Katz, Andrei
Shleifer, and Jeremy Stein for their guidance and extremely helpful suggestions.
We also thank Viral Acharya, Anat Admati, Ashwini Agrawal, Geraldo Cerqueiro,
Alex Chinco, Jess Cornaggia, Peter Crane, Martijn Cremers, Kent Daniel, Itamar
Drechsler, Darrell Duffie, Andrew Ellul, Mark Flannery, Itay Goldstein, Harrison
Hong, Ravi Jagannathan, E. Han Kim, Sam Lee, Patrick McCabe, Holger
Mueller, Stefan Nagel, Yihui Pan, Lasse Pedersen, Amiyatosh Purnanandam,
Uday Rajan, Alexi Savov, David Scharfstein, Amit Seru, Laura Starks, Philip
Strahan, and seminar participants at the NBER Summer Institute, AFA
Conference, CEAR Conference, CEPR Gerzensee, Chicago Booth/Deutsche Bank
Symposium, Chile Finance Conference, Napa Conference on Financial Markets
Research, Paul Wooley Conference on Financial Intermediation, Texas Finance
Festival, UBC Summer Finance Conference, Utah Winter Finance Conference,
Boston College, Harvard University, Imperial College, Indiana University, MIT
Sloan, National Bank of Poland, New York University, Oxford University,
Philadelphia Federal Reserve, Princeton University, University of Illinois at
Urbana-Champaign, University of Michigan, University of Nottingham,
University of Toronto, University of Warwick, University of Washington, and
Western Asset Management for helpful comments. This article was formerly dis-
tributed under the title ‘‘Implicit Guarantees and Risk Taking.’’
! The Author(s) 2013. Published by Oxford University Press, on behalf of President and
Fellows of Harvard College. All rights reserved. For Permissions, please email: journals
.permissions@oup.com
The Quarterly Journal of Economics (2013), 1073–1122. doi:10.1093/qje/qjt010.
Advance Access publication on April 5, 2013.

1073
1074 QUARTERLY JOURNAL OF ECONOMICS

money market fund depositors. The intervention was successful

in stopping the run, but it transferred the entire risk of the $3
trillion money market fund industry to the government.
This turmoil in the money fund industry came as a surprise
to many market participants. Prior to the run, investors regarded
money funds as a low-risk investment that was almost as safe as
cash. Indeed, for most of their history, money funds had invested
in safe assets and had generated yields similar to those of U.S.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Treasuries. However, during the early part of the financial crisis
the average fund yield increased relative to that of U.S.
Treasuries. As shown in Figure I, the yield differential was 15
basis points before August 2007 and increased to 90 basis points
after August 2007. Moreover, the cross-sectional dispersion in
fund yields increased from less than 30 to more than 150 basis
points. This sudden increase in the level and the dispersion of
fund yields suggests that the underlying asset risk of the funds
changed fundamentally during the financial crisis.1
In this article, we ask: How risky are money market funds?
The answer to this question is important for assessing risks to
financial stability in the United States. Money market funds are
the largest provider of short-term financing to financial institu-
tions, are similar in size to the entire sector of equity mutual
funds, and are also the largest provider of liquidity to corpor-
ations, issuing about the same amount of demand deposits as
the entire U.S. commercial banking sector. Money market
funds thus add a layer of financial intermediation between
issuers (mainly financial institutions) and investors (mainly cor-
porations). If money market funds have incentives to take on risk,
this additional layer of financial intermediation can weaken fi-
nancial stability through reducing market discipline on financial
institutions and making them more vulnerable to runs.2
Our analysis delivers four main results. First, money market
funds experienced an expansion in their risk-taking opportu-
nities starting in August 2007. Money market fund regulation

1. The crisis had also a wide-reaching impact on other parts of money markets,
such as the repo market (Gorton and Metrick 2009); unsecured and asset-backed
commercial paper (Brunnermeier 2009; Kacperczyk and Schnabl 2010; Acharya,
Schnabl, and Suarez 2013); Treasuries market (Krishnamurthy and Vissing-
Jorgensen 2010); and banks’ funding liquidity (Cornett et al. 2011).
2. Money funds have been discussed in Christoffersen (2001), Christoffersen
and Musto (2002), Kacperczyk and Schnabl (2010), McCabe (2010), and Strahan
and Tanyeri (2012).
 In basis points
-50

-100
0
50
100
150
200
250
300

Jan-02
Mar-02
May-02
Jul-02
Sep-02
Mean

Nov-02
Jan-03
Mar-03
5th percentile
95th percentile

May-03
Jul-03
Sep-03
Nov-03
Jan-04
Mar-04
May-04
Jul-04
Sep-04
Nov-04
Jan-05
Mar-05

FIGURE I
May-05
Jul-05
Sep-05
Nov-05
Jan-06
Mar-06

Dispersion in Money Market Fund Yields

May-06
Jul-06
Sep-06
Nov-06
Jan-07
Pre

Mar-07
May-07

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Jul-07
Sep-07
Post

Nov-07
Jan-08
Mar-08
May-08
Jul-08
1075 HOW SAFE ARE MONEY MARKET FUNDS?
 140
Repos Pre Post 1076
Deposits
120 Bank Obligations
FRNS

100 CP
ABCP

80

60

In basis points
40

20

0
QUARTERLY JOURNAL OF ECONOMICS

-20

FIGURE II
Spread by Money Market Instrument
We implement the regression model in Table III for the period from January 2005 to August 2008. Each point represents the three-
month average of coefficients on the interaction between month fixed effects and an indicator variable for repurchase agreements
(Repos), bank deposits (Deposits), bank obligations (Bank Obligations), floating rates notes (FRNS), commercial paper (CP), and asset-
backed commercial paper (ABCP), respectively. Each point represents the return relative to the omitted category (Treasuries and
agency debt) measured in basis points.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 HOW SAFE ARE MONEY MARKET FUNDS? 1077

requires funds to invest exclusively in highly rated, short-term

debt securities. As shown in Figure II, the spread between eligible
money market instruments and U.S. Treasuries was at most 25
basis points prior to August 2007, thus leaving little scope for risk
taking. However, after the run on asset-backed commercial paper
conduits in August 2007, many investors became aware that col-
lateral and liquidation values underlying money market instru-
ments had declined due to the U.S. subprime mortgage crisis,

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

which then prompted a repricing of risks in money markets. As
shown in Figure II, the spread between risky instruments, such
as bank obligations, and safe instruments, such as U.S. Treasu-
ries, increased from 25 up to 125 basis points after August 2007.
Hence, for the first time since their origin in the 1970s, money
funds had a choice of whether to invest in assets with a substan-
tial risk premium relative to safe government securities.3
Second, using weekly data on the universe of institutional
prime funds, we find that such funds had especially strong incen-
tives to take on risk.4 Our analysis reveals that fund flows are
highly responsive to current yields: a 1 standard deviation in-
crease in fund yields raises annualized fund assets by 43%.
This effect is economically large given that money funds charge
their investors a fixed share of assets under management. The
relationship is robust to including various controls, such as fund
age, expenses, assets size, fund flow volatility, fund family size,
and fund fixed effects. Also, the relationship is stronger after
August 2007 and coincides with the expansion in risk-taking
opportunities after the start of the financial crisis.
Third, we find that funds sponsored by financial intermedi-
aries with more money fund business took on more risk.
Specifically, a 1 standard deviation increase in institutional
prime money funds assets as a share of a sponsor’s total mutual
funds assets raises the share of risky fund assets by 3.7 percent-
age points, average maturity by 2.3 days, and fund yield by 3.0
basis points. This result is economically significant in that each

3. In general, the run on asset-backed commercial paper conduits marks the

start of the financial crisis (Acharya, Schnabl, and Suarez 2013). Historically, there
were other periods during which the yields of risky money fund instruments were
elevated for short periods. However, none of the episodes lasted for more than a few
weeks.
4. We focus on prime funds because we do not expect the subprime crisis to
have an economically meaningful effect on Treasury funds, which invest solely in
government securities.
1078 QUARTERLY JOURNAL OF ECONOMICS

respective effect accounts for 14.9%, 19.4%, and 19.2% of the

cross-sectional standard deviation of each risk measure.
Similarly, if a fund sponsor is a stand-alone asset manager, the
respective fund risks increase by 27.6%, 14.1%, and 46.7% rela-
tive to those of a fund whose sponsor is part of a financial con-
glomerate with non–money fund businesses. These results are
robust to including fund controls and fund fixed effects.
Our explanation for these results is that fund sponsors with

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

more non–money market fund business expect to incur large costs
if their money market funds fail. Such costs are typically reputa-
tional in nature, in that an individual fund’s default generates
negative spillovers to the fund’s sponsor other business. In prac-
tice, these costs are outflows from other mutual funds managed by
the same sponsor or a loss of business in the sponsor’s commercial
banking, investment banking, or insurance operations.
We conduct several tests to check the robustness of these
results. First, we show that our main findings remain the same
when we analyze stand-alone asset managers and control for a
sponsor’s credit rating, which suggests that the size of a sponsor’s
non–money fund business does not simply proxy for the sponsor’s
financial strength. Next, our results on risk taking do not hold for
the group of retail prime funds. Retail funds constitute a useful
placebo group because the same fund sponsors also offer funds to
retail investors, but retail investors are much less sensitive to
yield differentials than institutional investors are. This finding
suggests that our results are not driven by unobserved sponsor
characteristics, such as a sponsor’s risk aversion or the quality of
a sponsor’s risk management. Finally, our results on risk taking
gradually disappear after the government introduced unlimited
deposit insurance after Lehman’s default. Hence, it is unlikely
that our results are driven by the expectation of (implicit) gov-
ernment guarantees prior to Lehman’s default.
Fourth, we find that sponsor characteristics predict the
strength of runs in the immediate aftermath of Lehman’s default.
We focus on the one-week period after Lehman’s default because
the run stopped after the government provided unlimited deposit
insurance to all funds. We find that funds with more money fund
business and funds that took more risks before Lehman’s default
experienced larger runs. Moreover, funds sponsored by stand-
alone asset managers were less likely to provide financial support
during the run, even after controlling for risk taking before the
run. The run also triggered long-term adjustments in the money
 HOW SAFE ARE MONEY MARKET FUNDS? 1079

market fund industry. Sponsors with less money fund business

were more likely to exit the money market fund industry, consist-
ent with the notion that negative spillovers were larger for this
group. Among the funds that remained in business, some funds
changed their names to incorporate their sponsors’ names, prob-
ably to make their support more salient.
Overall, our main message is that money market funds are
risky. They add an additional layer of financial intermediation

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

between financial institutions and corporations, which exerts a
potentially destabilizing effect on financial markets. Funds have
strong incentives to chase yield (and risk), which reduces the
market discipline on financial institutions and makes them
more vulnerable to runs. Our results also shed light on the role
of demand deposits in limiting risk taking. While some theoret-
ical work argues that demand deposits mitigate intermediaries’
incentives to take risk through the threat of runs (e.g., Diamond
and Rajan 2000, 2001), our results suggest that in the context of
institutional prime funds, demand deposits may exacerbate risk
taking because of yield chasing by depositors.
The rest of the article proceeds as follows. Section II de-
scribes our research setting. Section III summarizes the data.
Section IV discusses the identification strategy and presents em-
pirical results. Section V concludes.

II. Institutional Setting: Money Market Funds

II.A. Primer on Money Market Funds
Money market funds emerged in the 1970s as an alternative
to bank deposits. At that time, Regulation Q limited the interest
that banks could pay on deposits to a rate that was much lower
than the yield on money market instruments. Money market
funds became an attractive alternative to investors because
they paid higher interest for taking on comparable risks. Even
though Regulation Q was eventually abolished, the size of the
fund industry grew steadily, reaching $2 trillion at the beginning
of 2006 (see Federal Reserve Flow of Funds Data).
An important characteristic of money funds is that, contrary
to bank deposits, money fund investments are not insured by the
government. But unlike other mutual funds, money funds have a
constant net asset value, typically $1 per share. This allows them
to offer demand deposits that can be purchased and redeemed on
1080 QUARTERLY JOURNAL OF ECONOMICS

demand and that are considered almost as safe as bank deposits.

Money funds maintain the stable net asset value by using histor-
ical cost accounting to assess the value of their holdings. They are
allowed to do so as long as the market value does not drop more
than 0.5% below the constant net asset value.
The down side of the constant net asset value is that it ex-
poses money funds to runs because the demand deposits are
backed by illiquid assets. If the market value of a fund’s holdings

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

is expected to drop below the constant net asset value, investors
tend to redeem their shares, which can exacerbate the market
value drop due to forced liquidation at fire-sale prices. Also,
funds may suffer losses on their investments because of changes
in interest rates or individual securities’ defaults, which can trig-
ger runs by investors wishing to redeem their shares before the
constant net asset value is suspended. As a result, money market
funds are susceptible to self-fulfilling runs as in Diamond and
Dybvig (1983).
To limit such risks, money market funds are regulated under
Rule 2a-7 of the Investment Company Act of 1940. This regula-
tion restricts fund holdings to short-term money market instru-
ments. Moreover, it requires short-term debt to be of high credit
quality. For example, it limits commercial paper holdings to those
that carry either the highest or second-highest rating from at
least two of the nationally recognized credit rating agencies.
Also, the regulation requires portfolio diversification: money
market funds must not hold more than 5% of their assets in
any instrument by a single issuer with the highest rating and
not more than 1% of their assets in any instrument of a single
issuer with the second-highest rating.
To provide an overview of the various money market instru-
ments held by money market funds, we use data provided by
iMoneyNet. These data are the most comprehensive data source
on money market funds’ holdings. We focus on taxable funds be-
cause nontaxable funds hold tax-exempt instruments issued by
state and municipal governments, which are not the focus of our
study. Taxable funds account for 84.5% of all assets under
management.
As of January 2006, there were 485 taxable funds, sponsored
by 148 companies, holding assets worth $1.67 trillion. Our study
focuses on prime funds, which can invest in debt other than gov-
ernment securities. There were 218 prime funds with total assets
under management of $1.26 trillion, 76.2% of total assets. The
 HOW SAFE ARE MONEY MARKET FUNDS? 1081

remainder of $396 billion, 23.8% of total assets, were held by 267

Treasury funds, which only hold government debt, government-
backed agency debt, and repurchase agreements. Among prime
funds, the largest asset class was commercial paper, accounting
for $325.3 billion, or 25.6% of total assets. The other sizable in-
vestments were in floating-rate notes ($265.9 billion), bank obli-
gations ($235.3 billion), asset-backed commercial paper ($186.3
billion), repurchase agreements ($151.1 billion), Treasuries and

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

government-backed agency debt ($62.5 billion), and bank de-
posits ($39.4 billion).
In terms of funding, institutional and retail investors
account for 57% and 43% of prime fund assets, respectively.
Most large prime funds are geared toward institutional investors.
As of January 2006, the 20 largest institutional prime funds ac-
counted for a total of $429 billion worth of assets.5 The largest
fund was the J.P. Morgan Prime Money Market Fund with assets
under management of $68.1 billion, followed by Columbia Cash
Reserves and BlackRock Liquidity, which were about half the
size. The smallest among the funds, Dreyfus Institutional Cash
Fund, still managed a considerable $12.6 billion. On average, the
funds were well diversified across instrument types but highly
exposed to risks in the financial industry as a whole. Assets ori-
ginated by the financial industry—measured as a total of finan-
cial commercial paper, structured securities, bank obligations,
and repurchase agreements—accounted for 91.4% of money
market fund assets.

II.B. The Role of a Fund Sponsor

A fund sponsor plays an important role in the money market
fund’s operation. Usually, the sponsor is a financial institution that
is either a stand-alone asset manager or a financial conglomerate.
The main tasks of the fund sponsor include managing the fund and
ensuring a smooth fund operation. Most important, the sponsor
chooses the fund portfolio and thus determines a fund’s risk.6

5. In our analysis, we classify a fund as institutional if it offers at least one

institutional class and as retail if it does not offer any institutional share classes.
6. In our tests, we assume that a fund sponsor can set its fund’s risk. In doing so,
we abstract from agency problems between the fund sponsor and fund manager. We
believe this assumption is plausible in the money fund industry because a fund’s
portfolio risk is observable and there is little scope for manager skill in portfolio choice.
1082 QUARTERLY JOURNAL OF ECONOMICS

In general, the sponsor’s incentives to take on risk depend on

the shape of the sponsors’s payoff function, which in turn is deter-
mined by the benefits and costs of risk taking. On the benefit side,
funds typically experience large inflows as a result of high yields.
This raises a sponsor’s incentive to invest in high-yielding assets,
especially if investors do not fully risk-adjust returns.
On the cost side, excessive risk taking can trigger a run on a
fund. If there is a run, fund sponsors have the option, but not the

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

obligation, to support failing funds. Even though fund sponsors
have no contractual obligation to support their funds during runs,
they may find it optimal to do so because of negative spillover
costs to other parts of the sponsors’ business.7 Such costs are
typically reputational in nature, in that an individual fund’s de-
fault could generate outflows from other mutual funds managed
by the same sponsor, or a loss of the sponsor’s general business. A
significant part of our empirical tests is devoted to estimating the
importance of such business spillovers for risk choices.
Our primary measure of business spillovers is Fund
Business, defined as the sponsor’s assets under management
excluding institutional prime funds divided by the sponsor’s
assets under management. For example, in January 2006,
Fidelity had $973 billion in assets under management, out of
which $59 billion was in institutional prime funds, which implies
Fund Business of 93.9%. The idea behind this measure is that
fund families with more assets under management in non–
money market funds have more at stake in case their money
market fund faces distress. Our second measure of business spill-
overs is Conglomerate—an indicator variable equal to 1 if the
fund sponsor is part of a financial conglomerate (i.e., commercial
bank, investment bank, or insurance company), and 0 otherwise.
This complements our first measure in that it captures a broader
idea of a franchise value at stake, as is the case for financial con-
glomerates. In our tests, we use both measures, bearing in mind

7. Many investors expect a sponsor to provide financial support to their fund if

the fund suffers a run. This expectation is evident in an investor alert by the
Financial Industry Regulatory Authority (FINRA), which states: ‘‘Typically,
there has been an expectation that when a money market fund reaches a point
where it might break the buck [i.e., suspend the constant net asset value], the
investment management firm that sponsors the fund will take action to infuse
the fund with cash so that the fund can maintain a stable NAV of $1.00 per
share’’ (FINRA 2010).
 HOW SAFE ARE MONEY MARKET FUNDS? 1083

that each one captures a slightly different type of cross-sectional

variation in the data.8

II.C. Money Market Funds during the Financial Crisis

1. Change in Risk-Taking Opportunities. Before August 2007,
the yields of money market instruments were almost the same as
the yields of U.S. Treasuries. Given that fund regulation re-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

stricted money funds to investing in money instruments, it effect-
ively prevented funds from differentiating their risks. As a result,
money funds invested in similar instruments and had similar
yields.
However, starting in August 2007, a number of events chan-
ged funds’ risk-taking opportunities. On August 9, 2007, the
French bank BNP Paribas halted withdrawals from three of its
mortgage-backed securities funds and suspended calculation of
their net asset values. Even though defaults on mortgages had
been rising throughout 2007, the suspension of withdrawals by
BNP had a profoundly negative effect on money market instru-
ments. Within one day, the interest rate spread of overnight
asset-backed commercial paper over the Fed funds rate rose
from 10 to 150 basis points, as investors became concerned
about the quality of collateral backing these instruments.
Money market funds suffered almost no direct losses from
impaired asset-backed commercial paper. The reason was that
these instruments were effectively insured by commercial
banks through the provision of liquidity guarantees that paid
off maturing asset-backed commercial paper in case of a run.9
However, going forward, it became clear to many investors that
liquidation values of money market instruments were lower and
that financial institutions issuing money market instruments
were riskier than previously thought. As a result, money
market fund instruments that were backed by financial

8. We also considered another measure of business spillovers that additionally

accounts for the size of the sponsor’s total assets by adding total sponsor assets to
total mutual fund assets, which is especially important for financial conglomerates.
This measure implicitly assumes that both fund assets and other types of assets
carry equal weight in the sponsor’s business activity, which is plausible but difficult
to measure. Our results remain qualitatively unchanged if we use this alternative
measure.
9. Acharya, Schnabl, and Suarez (2013) provide a detailed account of the struc-
ture of asset-backed commercial paper conduits and show that losses to outside
investors, such as money market funds, were small.
1084 QUARTERLY JOURNAL OF ECONOMICS

institutions—bank obligations, floating-rate notes, and commer-

cial paper—were perceived as risky and had to offer higher yields.
At the same time, the yields of safe instruments, such as U.S.
Treasuries, repurchase agreements, and bank deposits, remained
at much less elevated levels.
Figure II presents evidence of the sudden change in yields of
money market instruments.10 From January 2005 to July 2007,
all instruments had yields of about 15–25 basis points relative to

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

those of U.S. Treasuries and agency debt. However, starting in
August 2007, the yields of risky instruments started to increase
rapidly with a peak in March 2008 when spreads reached 125
basis points. After March 2008, the spreads started to decline
but were still at 60 basis points in August 2008. Over the same
period, the spreads of safe instruments remained constant at
around 20 basis points or even declined. In sum, the start of the
financial crisis in August 2007 provided money funds with the
opportunity to invest in riskier assets.11

2. Tale of Two Funds: Reserve Primary Fund and Fidelity

Institutional Prime. We illustrate the response to these risk-
taking opportunities with an example of two funds: the Reserve
Primary Fund (RPF) and the Fidelity Institutional Prime (FID).
The RPF was particularly well known in the industry because its
owner, Bruce Bent, was the founder of the first money market
fund. Similarly, the FID was well known because of its sponsor,
which is one of the largest asset managers. The funds were quite
similar before August 2007. Each managed about $25 billion in

10. The yields of individual instruments are not directly observable to us, but we
can impute them using fund-level data on yields and holdings. To this end, we
regress fund yields on interaction terms of indicator variables for each instrument
type and month-fixed effects plus standard controls. For each instrument type, the
corresponding interaction term captures the monthly yield relative to that of U.S.
Treasuries and agency debt.
11. The observed variation in yields of risky and safe assets coincided with key
events during the crisis. First, the expansion in risk-taking opportunities occurred
at the same time as the run on asset-backed commercial paper in August 2007.
Furthermore, the peak in yields of risky assets happened at the same time as the
near-bankruptcy of the investment bank Bear Stearns. Finally, the decline in
spreads prior to August 2008 and the sudden spike in September 2008 (not
shown in the figures) matched market conditions around the Lehman’s bank-
ruptcy. Indeed, common indicators of market distress during the crisis, such as
the LIBOR-OIS spread, exhibited similar time-series patterns, as did the yields
of risky instruments of money funds.
 HOW SAFE ARE MONEY MARKET FUNDS? 1085

assets and charged similar management fees. In what follows, we

present the evolution of each fund’s yields, assets, and holdings
over the period from August 2006 to August 2008.
In Figure III, we present the gross yields of both funds rela-
tive to the value-weighted industry average. Prior to August
2007, the yields of the two funds roughly matched the industry
average. However, starting in August 2007, the relative yields
diverged sharply: the yield of RPF increased by about 50 basis

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

points while the yield of FID stayed constant. The yield differen-
tial triggered significant differences in money flows. Relative to
the average asset growth of all institutional prime funds, by
August 2008 RPF increased its assets under management by
140%, whereas FID’s assets value grew only by 40%.
The observed differences in both yields and fund flows were
largely a consequence of the differences in the underlying fund
portfolios after August 2007. Figure IV shows that RPF increased
its holdings of risky assets from 0% to 60% while it reduced its
exposure to U.S. Treasuries and repurchase agreements from
40% to 10%. In contrast, the share of risky assets held by FID
remained steady in 2008.
We argue that the difference in risk taking between RPF
and FID can be largely attributed to the difference in their
sponsors’ non–money market fund business. While RPF was
managed by a stand-alone fund company with almost no other
funds under management, FID’s manager—Fidelity—is one of
the world’s largest assets managers with lots of other business.
As of January 2006, Fidelity sponsored 252 non–money market
mutual funds with $814 billion in assets under management.
These funds accounted for 93.9% of Fidelity’s assets under man-
agement and an even larger share of revenues because they
charge higher expense ratios than do money market funds.
Hence, Fidelity could suffer large losses from a failure of its
money market fund business if the run on FID triggered outflows
from its other funds.

3. Collapse of the RPF and Money Market Fund Runs. One of

the important instruments among RPF’s holdings was commer-
cial paper issued by Lehman Brothers. According to quarterly
Securities and Exchange Commission (SEC) filings, RPF had no
holdings of Lehman’s commercial paper prior to August 2007, but
by November 2007 the fund had purchased $375 million worth of
 Panel A: Reserve Primary Fund (RPF)
40 160%
1086
Spread
35 140%
Asset Growth
30 120%
25 100%
20 80%
15 60%
10 40%
Asset Growth

5 20%

Spread in basis points

0 0%
-5 -20%
-10 -40%

8/1/2006
9/1/2006
1/1/2007
2/1/2007
3/1/2007
4/1/2007
5/1/2007
6/1/2007
7/1/2007
8/1/2007
9/1/2007
1/1/2008
2/1/2008
3/1/2008
4/1/2008
5/1/2008
6/1/2008
7/1/2008
8/1/2008

10/1/2006
11/1/2006
12/1/2006
10/1/2007
11/1/2007
12/1/2007
FIGURE III
QUARTERLY JOURNAL OF ECONOMICS

Relative Performance and Assets: Reserve Primary versus Fidelity Institutional Prime
This figure plots weekly industry-adjusted spread and industry-adjusted asset growth of the Reserve Primary Fund (Panel A) and
the Fidelity Institutional Prime Money Market Fund (Panel B) from August 2006 to August 2008. The industry-adjusted spread is
computed as the difference between each individual fund’s spread and the value-weighted average spread of all institutional prime
funds. The industry-adjusted asset growth is computed as the fund’s asset growth deflated by total asset growth of all institutional
prime funds. We normalize asset growth to zero as of August 1, 2008.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 Spread in basis points
-10
0
5

-5
10
15
20
25
30
35
40

8/1/2006
9/1/2006
10/1/2006
11/1/2006
12/1/2006
Spread

1/1/2007
2/1/2007
Asset Growth

3/1/2007
4/1/2007
5/1/2007
6/1/2007
7/1/2007
8/1/2007
FIGURE III 9/1/2007
Continued
10/1/2007
11/1/2007
12/1/2007
1/1/2008
Panel B: Fidelity Institutional Prime (FID)

2/1/2008
3/1/2008
4/1/2008
5/1/2008
6/1/2008
7/1/2008
8/1/2008

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

0%
20%
40%
60%
80%

-40%
-20%
100%
120%
140%
160%
Asset Growth
1087 HOW SAFE ARE MONEY MARKET FUNDS?
 1088
Panel A: Reserve Primary Fund (RPF)
100%
ABCP
90% US + Repos
80% Other
70%
60%
50%

Holdings
40%
30%
20%
10%
0%

8/1/2006
9/1/2006
1/1/2007
2/1/2007
3/1/2007
4/1/2007
5/1/2007
6/1/2007
7/1/2007
8/1/2007
9/1/2007
1/1/2008
2/1/2008
3/1/2008
4/1/2008
5/1/2008
6/1/2008
7/1/2008
8/1/2008

10/1/2006
11/1/2006
12/1/2006
10/1/2007
11/1/2007
12/1/2007
QUARTERLY JOURNAL OF ECONOMICS

FIGURE IV
Assets Holdings: Reserve Primary versus Fidelity Institutional Prime
This figure plots weekly holdings of the Reserve Primary Fund (Panel A) and the Fidelity Institutional Prime Money Market Fund
(Panel B) from August 2006 to August 2008. U.S. + Repos is the share of assets invested in U.S. Treasuries, agency debt, and
repurchase agreements. ABCP is the share invested in asset-backed commercial paper. Other is the share invested in other securities.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 Holdings
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%

0%

8/1/2006
9/1/2006
10/1/2006
11/1/2006
12/1/2006
1/1/2007
2/1/2007
3/1/2007
4/1/2007
5/1/2007
6/1/2007
7/1/2007
8/1/2007
FIGURE IV
Continued
9/1/2007
10/1/2007
11/1/2007
12/1/2007
Panel B: Fidelity Institutional Prime (FID)

1/1/2008
2/1/2008
3/1/2008
4/1/2008
5/1/2008
6/1/2008

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Other
ABCP

7/1/2008
US + Repos
8/1/2008
1089 HOW SAFE ARE MONEY MARKET FUNDS?
1090 QUARTERLY JOURNAL OF ECONOMICS

Lehman’s paper. By May 2008, the fund further increased its

Lehman’s holdings to $775 million, which at the time accounted
for about 1% of its holdings.
On September 15, 2008, Lehman Brothers declared bank-
ruptcy. Its failure triggered a panic in financial markets and
led to a credit market freeze. As a result of the bankruptcy, the
net asset value of RPF fell below $1 per share. The revelation of
the fund’s exposure to Lehman’s risk caused an immediate run on

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

the fund. On September 16, 2008, the fund was forced to redeem
$10.8 billion in withdrawals and faced $28 billion of additional
withdrawal requests. The fund’s sponsor did not have sufficient
financial resources to guarantee payments and was forced to halt
redemptions. The run on RPF quickly spread to other money
funds. Within a week, institutional investors reduced their in-
vestments in money funds by more than $172 billion.12
Eventually, several funds became distressed and the conse-
quences of the redemptions became dire. To stop the run on funds,
on September 19, 2008, the U.S. Department of the Treasury
announced an explicit deposit insurance covering all money
fund investments made prior to Lehman’s default. This an-
nouncement stopped the run and redemption requests receded
shortly after. However, the announcement meant that the U.S.
government had effectively insured the risk of $3 trillion in fund
assets holdings.

III. Data and Summary Statistics

We collect data from five sources. First, we obtain data on the
universe of taxable money market funds from iMoneyNet, which
cover the period from January 2005 to September 2011 and in-
clude weekly fund-level data on yields, expense ratios (charged
and incurred), average maturity, holdings by instrument type,
and fund sponsor. Second, we complement the data with informa-
tion from the CRSP Mutual Fund Database, especially assets
under management of the fund sponsor. Third, we use
COMPUSTAT and companies’ websites for information on fund
sponsor characteristics. Fourth, we use S&P RatingsXpress,

12. This was the first industry-wide run in the history of money market funds.
Prior to Lehman’s default, only one fund ever broke the buck. In 1994, a small
money market fund, Community Bankers Money Fund, defaulted because of its
exposure to the Orange County bankruptcy.
 HOW SAFE ARE MONEY MARKET FUNDS? 1091

Lehman Brothers’ Bond Database, COMPUSTAT, and compa-

nies’ websites to collect data on credit ratings. Fifth, we collect
data on no-action letters issued by the SEC—an indication that a
sponsor provided financial support to its fund. Altogether, we
obtain a novel data set that, to the best of our knowledge, has
not been used in academic research before. Additional details on
the data collection are presented in the Appendix.
We conduct our analysis at the fund level. We therefore ag-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

gregate all share classes by fund and type of investor (retail, in-
stitutional). We compute fund characteristics (e.g., expense ratio)
as the weighted average with assets per share class as weights.
Some funds offer both retail and institutional share classes.
Institutional shares are generally larger; hence, we classify a
fund as institutional if it offers at least one institutional class
and as retail if it does not offer institutional share classes.13
Column (1) of Table I provides summary statistics for all in-
stitutional prime money funds as of January 2006. Our sample
includes 148 funds. The average fund size is $4.9 billion and the
average fund age is 10.6 years. We compute the spread as the
annualized gross yield (i.e., before expenses) minus the yield of
the one-month Treasury bill. The average spread is 6.9 basis
points and the average expense ratio is 32 basis points. In
terms of assets holdings, funds hold 32.0% in commercial paper,
19.8% in floating-rate notes, 13.5% in repurchase agreements,
13.4% in asset-backed commercial paper, 12.2% in bank obliga-
tions, 6.0% in U.S. Treasuries and agency-backed debt, and 3.2%
in deposits.
As discussed, a significant part of our empirical analysis ana-
lyzes the effect of a sponsor’s non–money market fund business on
risk taking. To examine whether funds differ in their characteris-
tics depending on a sponsor’s non–money market fund business,
we divide fund sponsors into two groups based on a sponsor’s non–
money market fund assets. Column (2) of Table I provides sum-
mary statistics for funds whose sponsors have Fund Business
(non–money market fund as a share of total mutual fund assets)
above the median value of 81.6% as of January 2006, and column
(3) shows summary statistics for funds whose sponsors have values

13. As a robustness check, we also estimate our regressions for funds that only
offer institutional shares. The coefficients are stable and remain statistically sig-
nificant (albeit standard errors widen slightly because of the reduction in
observations).
1092 QUARTERLY JOURNAL OF ECONOMICS

TABLE I
SUMMARY STATISTICS OF INSTITUTIONAL PRIME MONEY MARKET FUNDS

(1) (2) (3)

All High FB Low FB
Fund characteristics
Spread (bp) 6.93 6.60 7.28
(6.44) (7.54) (5.00)
Expense ratio (bp) 31.64 32.40 30.81

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

(19.10) (18.43) (19.90)
Fund size ($mil) 4,886 2,981 6,951***
(8,685) (4,833) (11,169)
Maturity (days) 34.32 35.12 33.45
(11.02) (12.48) (9.17)
Age (years) 10.61 10.43 10.81
(4.75) (5.53) (3.75)
Family size ($bil) 72.8 97.5 45.9**
(149.1) (200.9) (39.2)
Fund business 0.764 0.897 0.619***
(0.198) (0.064) (0.192)
Conglomerate (in %) 60.1 55.8 64.8
(49.1) (50.0) (48.1)
Portfolio holdings
U.S. Treasuries & agency 0.060 0.072 0.048
(0.109) (0.120) (0.095)
Repurchase agreements 0.135 0.142 0.126
(0.150) (0.169) (0.128)
Bank deposits 0.032 0.021 0.044**
(0.057) (0.039) (0.069)
Bank obligations 0.122 0.111 0.135
(0.126) (0.120) (0.132)
Floating-rate notes 0.198 0.192 0.204
(0.162) (0.168) (0.156)
Commercial paper 0.320 0.356 0.280**
(0.224) (0.252) (0.182)
Asset-backed CP 0.134 0.106 0.164**
(0.155) (0.151) (0.154)
Funds 148 77 71

Notes. This table shows summary statistics for all U.S. institutional prime money market funds as of
January 1, 2006. Fund Business (FB) is mutual fund assets other than institutional prime fund assets as a
share of total sponsor’s mutual fund assets. High (Low) FB includes all funds with Fund Business above
(below) the median value of Fund Business (81.6%). Fund characteristics are spread, expenses, fund size,
average portfolio maturity, age, family size (mutual funds assets other than money market fund assets),
and whether the fund sponsor is part of a conglomerate (in %). Holdings are the share of assets invested in
Treasuries and agency debt, repurchase agreements, bank deposits, bank obligations, floating-rate notes,
commercial paper, and asset-backed commercial paper. Cross-sectional standard deviations of the given
characteristics are in parentheses. ***, **, * represent 1%, 5%, and 10% statistical significance,
respectively.
 HOW SAFE ARE MONEY MARKET FUNDS? 1093

below the median. We find that both groups are quite similar in
terms of observable characteristics, such as spread, expense ratio,
maturity, age, and holdings. The main difference is that funds
sponsored by firms with high Fund Business are on average
more likely to be part of financial conglomerates. These results
suggest that the extent of a fund sponsor’s non–money fund busi-
ness was not chosen in anticipation of changes in risk-taking
opportunities in the money market fund sector.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

IV. Empirical Strategy and Results
In this section, we describe our empirical results. First, we
analyze how the financial crisis created opportunities to take risk
in the money market fund industry. We then estimate the flow–
performance relationship to study the risk-taking incentives.
Next, we analyze the importance of a sponsor’s non–money
fund business for the the risk-taking behavior of money funds.
Finally, we examine the ex post consequences resulting from the
collapse of RPF and the run on the money market fund industry.

IV.A. Expansion of Risky Investment Opportunities

We document the change in opportunities to take risk using
weekly data on fund holdings and fund yields in the following
regression model:
X
ð1Þ Spreadi, tþ1 ¼ i þ t þ j Holdingsi, j, t þ Xi, t þ "i, tþ1 ,
j

where Spreadi,t+1 is the gross yield of fund i in week t + 1 minus

the risk-free rate, Holdingsi,j,t denotes fund i’s fractional holdings
of instrument type j in week t, i denotes fund fixed effects, and t
denotes week fixed effects. The instrument types include repur-
chase agreements, bank deposits, bank obligations, floating-rate
notes, commercial paper, and asset-backed commercial paper.
The omitted category is Treasuries and government agency
debt. Xi,t is a vector of fund-specific controls that includes the
natural logarithm of fund size (Log(Fund Size)), fund expenses
(Expense Ratio), fund age (Age), and the natural logarithm of the
fund family size (Log(Family Size)). In all regressions, we allow
for flexible correlation of error terms within funds by clustering
standard errors at the fund level. Our coefficients of interest are
1094 QUARTERLY JOURNAL OF ECONOMICS

j, which measure the return on money market instrument j in

week t + 1 relative to that of Treasuries and agency debt.
We estimate the regression model separately for the post
period from August 2007 to August 2008 and the pre period
from January 2006 to July 2007. The post period starts with
the subprime crisis in August 2007 and ends right before the
market-wide run in September 2008. We do not include obser-
vations during the run and the period thereafter because subse-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

quent government interventions significantly altered risk-taking
incentives. Our estimation strategy is akin to estimating a stand-
ard difference-in-differences regression model.
Columns (1) and (2) of Table II report the post-period and
pre-period results. We find that risky instruments have signifi-
cantly larger yields in the post period relative to those in the pre
period, whereas safe instruments have similar yields during both
periods. For example, in the post period, the yield of a fund fully
invested in (risky) bank obligations would have been 88 basis
points higher than the yield of a fund fully invested in (safe)
Treasury and agency debt. The comparable differential in the
pre period would have only been 16 basis points. We find similar
effects for other risky instruments, such as floating-rate notes,
commercial paper, and asset-backed commercial paper. In con-
trast, the yield of a fund fully invested in (safe) repurchase agree-
ments would have been only 13–18 basis points higher than the
yield of a fund fully invested in Treasury and agency instru-
ments, both in the pre and post periods.14
One possible concern with the results is that funds with large
holdings of risky assets might also be riskier, along other unob-
served dimensions. For example, these funds may choose the
most risky assets within an asset class such that we would over-
estimate the impact of holding riskier assets. To address this con-
cern, we introduce fund fixed effects, which account for any
unobserved time-invariant fund characteristics within the pre
or post periods.
We find quantitatively and qualitatively similar results, as
reported in column (3). The yield of a fund fully invested in bank ob-
ligations would have been 91 basis points higher than the yield of a
fund fully invested in Treasury and agency instruments. In

14. The results on repurchase agreements are consistent with the findings in
Krishnamurthy, Nagel, and Orlov (2012).
 HOW SAFE ARE MONEY MARKET FUNDS? 1095

TABLE II
RETURNS BY INSTRUMENT TYPE

Spreadi,t+1

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

Post Pre Post Pre
Holdings
Repurchase Agreementsi,t 13.483 17.937*** 39.748** 11.470**

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

(8.638) (3.525) (17.739) (5.609)
Bank Depositsi,t 39.445** 17.550*** 50.392** 18.993***
(17.350) (4.044) (20.459) (6.906)
Bank Obligationsi,t 88.339*** 15.909*** 90.868*** 6.845
(7.929) (3.564) (19.307) (4.885)
Floating-Rate Notesi,t 81.953*** 22.709*** 89.466*** 9.761
(7.804) (3.556) (22.629) (6.514)
Commercial Papert 57.430*** 16.330*** 71.256*** 16.252***
(8.169) (3.370) (24.596) (5.694)
Asset-backed CPi,t 76.619*** 20.997*** 84.627*** 15.626**
(8.036) (3.230) (19.852) (6.190)
Fund characteristics
Log(Fund Size)i,t 0.513 0.259*** 3.864** 0.562
(0.349) (0.096) (1.946) (0.486)
Expense Ratioi,t 10.925*** 1.812* 87.327*** 53.502***
(2.995) (0.963) (25.457) (11.597)
Agei,t 1.18 0.394 0.576 0.621
(1.427) (0.505) (0.548) (0.457)
Log(Family Size)i,t 0.02 0.013 4.352 0.369***
(0.225) (0.067) (5.193) (0.139)
Constant 67.527*** 6.031 68.901 12.645
(10.849) (4.310) (68.832) (8.067)
Week fixed effects Y Y Y Y
Fund fixed effects N N Y Y
Observations 7,807 11,984 7,807 11,984
R-squared 0.94 0.79 0.95 0.80

Notes. The sample is all U.S. institutional prime money market funds. The dependent variable
Spread is computed as the annualized yield minus the Treasury bill rate. Holdings variables are the
share of assets invested in repurchase agreements, bank deposits, bank obligations, floating-rate notes,
commercial paper (CP), and asset-backed CP (omitted category is U.S. Treasury and agency). Fund char-
acteristics are natural logarithm of fund size, expense ratio, fund age, and natural logarithm of fund
family size. All regressions are at the weekly level and include week fixed effects. Columns (3) and (4)
include fund fixed effects. Columns (1) and (3) cover the period 8/1/2007–8/31/2008 (post period). Columns
(2) and (4) cover the period 1/1/2006–7/31/2007 (pre period). Standard errors are clustered at the fund
level. ***, **, * represent 1%, 5%, and 10% statistical significance, respectively.

contrast, the comparable differential would have only been 7 basis

points in the pre period. Hence, our findings are not driven by
unobserved time invariant fund characteristics.
Overall, these results suggest that money funds experienced
a large exogenous expansion in their risk-taking opportunities.
1096 QUARTERLY JOURNAL OF ECONOMICS

The expansion was economically significant in that the yields of

risky instruments, relative to safe ones, were five times larger
after August 2007, compared to before. Moreover, the expansion
was likely exogenous to money market funds as it was caused by
financial distress among issuers of money market instruments
and not by the funds themselves. Hence, starting in August
2007, funds were given a choice of whether to invest in risky or
safe assets.15

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

IV.B. The Flow–Performance Relationship
Funds can increase their revenues by attracting new flows.
This can happen by increasing risk and thus fund yields, which in
turn translates into greater fund inflows. Given that money funds
are paid fees as a fixed percentage of assets under management,
fund inflows lead to a higher income to the fund sponsor.16 We
assess the benefits of investing in riskier instruments by estimat-
ing the sensitivity of fund flows to past yields using the following
regression model:

ð2Þ Fund Flowi, tþ1 ¼ t þ Spreadi, t þ Xi, t þ "i, tþ1 ,

where Fund Flowi,t+1 is the percentage increase in a fund’s i size
from week t to week t + 1 accounting for earned interest, winsor-
ized at the 0.5% level; Spreadi,t and Xi,t are defined as in equation
(1). In addition, we include the volatility of fund flows, Flow
Volatilityi,t, measured as the standard deviation of weekly fund
flows over the previous quarter. We allow for correlation of error
terms within funds by clustering observations at the fund level.
Our coefficient of interest is , which measures the sensitivity of
fund flows to fund past yields.
Table III reports the results. Columns (1) and (2) show the
results separately for the post and pre periods. We find that a 1

15. Note that the overall issuance of riskier assets declined over this period. For
example, total asset-backed commercial paper outstanding dropped by almost 50%,
from $1.3 trillion in August 2007 to $700 billion in August 2008 (Acharya, Schnabl,
and Suarez 2013). Our focus is on the variation in holdings across funds. While the
majority of funds decreased their holdings of risky assets, some funds, such as the
RPF, increased them.
16. This model of competition has been documented in studies of equity mutual
funds. These studies find that past performance is one of the strongest predictors of
flows to equity funds (e.g., Chevalier and Ellison 1997).
 HOW SAFE ARE MONEY MARKET FUNDS? 1097

TABLE III
FLOW–PERFORMANCE RELATIONSHIP

Fund Flowi,t+1

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

Post Pre Post Pre Post Pre
Spreadi,t 0.013*** 0.003 0.024*** 0.000
0.026*** 0.009
(0.005) (0.005) (0.008) (0.004)
(0.009) (0.010)

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Fund Businessi,2006* 0.003 0.009
Spreadi,t (0.006) (0.009)
Conglomeratei,2006* 0.000 0.003
Spreadi,t (0.003) (0.005)
Log(Fund Size)i,t 0.120** 0.077*** 7.658*** 4.146*** 7.655*** 4.148***
(0.051) (0.029) (1.341) (0.720) (1.345) (0.720)
Expense Ratioi,t 0.549* 1.276*** 2.668 1.365 2.680 1.475
(0.320) (0.354) (5.965) (3.703) (5.915) (3.704)
Agei,t 0.159 0.078 0.015 0.715** 0.014 0.713**
(0.180) (0.149) (0.322) (0.323) (0.322) (0.323)
Flow Volatilityi,t 4.245* 2.476** 1.381 0.213 1.331 0.230
(2.322) (1.243) (3.175) (2.152) (3.166) (2.146)
Log(Family Size)i,t 0.026 0.032** 0.528 0.042 0.521 0.045
(0.023) (0.014) (1.238) (0.126) (1.245) (0.127)
Week fixed effects Y Y Y Y Y Y
Fund fixed effects N N Y Y Y Y
Observations 7,807 11,984 7,807 11,984 7,807 11,984
R-squared 0.022 0.017 0.085 0.052 0.085 0.052

Notes. The sample is all U.S. institutional prime money market funds. Columns (1), (3), and (5) cover the
period from 8/1/2007 to 8/31/2008 (post period). Columns (2), (4), and (6) cover the period from 1/1/2006 to 7/31/
2007 (pre period). The dependent variable is Fund Flow, computed as the percentage change in total net
assets from time t to time t + 1, adjusted for market appreciation and winsorized at the 0.5% level.
Independent variables are the weekly annualized spread from t to t  1, natural logarithm of fund size,
fund expense ratio, fund age, volatility of fund flows based on past 12-week fund flows, and natural logarithm
of fund family size. In columns (5) and (6), additional independent variables are the interactions of Spread
with Fund Business and Conglomerate. Fund Business (FB) is mutual fund assets other than institutional
prime fund assets as a share of total sponsor’s mutual fund assets. Conglomerate is an indicator variable
equal to 1 if the fund sponsor is affiliated with a financial conglomerate and 0 otherwise. All regressions are at
the weekly level and include week fixed effects. Columns (3) to (6) also include fund fixed effects. Standard
errors are clustered at the fund level. ***, **, * represent 1%, 5%, and 10% statistical significance,
respectively.

standard deviation increase in fund spread, equal to 65.9 basis

points, increases subsequent fund flows by 0.83% per week. This
effect is economically large because it implies that a fund could
increase its annual revenue by 43% by investing in riskier instru-
ments. Conversely, we find no significant effect of fund yields on
flows in the pre period. To rule out the possibility that our results
are driven by unobserved time-invariant fund-specific attributes
correlated with fund spreads, we estimate the model with fund
fixed effects. The flow–performance sensitivity, in columns (3)
and (4), is even larger. As before, we observe no impact on flows
in the pre period.
1098 QUARTERLY JOURNAL OF ECONOMICS

The flow–performance relationship may also vary by sponsor

characteristics. We are particulary interested in the effect of a
sponsor’s non–money market fund business. This characteristic
may affect the flow–performance relationship if it affects how
investors risk adjust fund returns or whether investors expect a
sponsor to support a fund in case of a run.
We test this hypothesis by estimating the flow-performance
relationship in equation (2) with additional controls for the spon-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

sor’s non–money market fund as a share of total mutual fund
assets (Fund Business) and whether a fund is affiliated with a
financial conglomerate (Conglomerate) and interactions of these
variables with fund spread. We present the results in columns
(5)–(6). For both subperiods, we find that the coefficients of the
interaction terms are statistically and economically insignificant
for both measures of business spillovers. Hence, conditional on
flow performance, there is no effect of sponsor characteristics on
fund flows. We also estimate all the regressions without adding
interaction terms. We find no statistically or economically signifi-
cant effect of Fund Business or Conglomerate on fund flows.
These findings suggest that investors do not risk adjust yields
based on sponsor characteristics.17

IV.C. Business Spillovers and Risk Taking

This section analyzes the risk-taking incentives in the cross-
section of money market funds. We examine whether fund spon-
sors’ concerns over their non–money fund business reduce their
funds’ risk.18
An important advantage of our setting is that the size of a
sponsor’s non–market fund business was not driven by risk-
taking opportunities in the money market fund industry.
Before August 2007, money market funds typically constituted

17. This evidence is consistent with theoretical models that show that the ex-
pected benefit of acquiring such information is low relative to the cost of learning
this information (Dang, Gorton and Holmstrom 2009). It is also consistent with
models in which investors neglect risks that are not salient to them given the ab-
sence of negative events from past data (Gennaioli and Shleifer 2010; Gennaioli,
Shleifer, and Vishny 2012).
18. We focus on non–money market fund business because this sponsor char-
acteristic is featured prominently in industry studies on money funds. However,
there might be other variables that affect the fund’s payoff function. Hence, we
interpret our analysis as an example of how variation in the payoff function affects
risk taking.
 HOW SAFE ARE MONEY MARKET FUNDS? 1099

a small part of larger fund families and the choice regarding the
fund family’s organization profile was likely independent of
money funds themselves. Money funds were considered a low-
fee, low-cost business with little scope for exploiting private in-
formation or superior managerial ability. They simply invested
in safe assets and were offered in conjunction with other, more
profitable funds. Hence, it is unlikely that sponsors actively
chose the size of their non–money fund businesses in anticipa-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

tion of changes in risk-taking opportunities in the money
market fund sector.
We analyze funds’ risk taking by estimating the following
regression model:

Riski, tþ1 ¼  þ t þ 1 Business Spilloversi, 2006

þ 2 Postt  Business Spilloversi, 2006 þ Xi, 2006 þ "i, tþ1 ,
ð3Þ
where Business Spilloversi,2006 is a generic name for either Fund
Business or Conglomerate. Post is an indicator variable equal to 1
for the post period and 0 for the pre period. Xi,2006 is a vector of
control variables similar to those in equation (2). Both business
spillovers variables and other controls are measured as of
January 2006, which mitigates the concern that fund risk choices
are driven by changes in fund characteristics due to investment
opportunity change.19 Our regression model also includes week
fixed effects (t), which account for any time differences in aggre-
gate fund flows or macroeconomic conditions driving the risk-
taking decisions of fund sponsors. Since Business Spillovers is a
fund sponsor attribute, risk taking within the same sponsor may
be correlated across its funds. To address this concern, we cluster
standard errors at the sponsor level.
We use three measures of risk (Risk), measured at a weekly
frequency. The first measure, Spread, is the gross yield minus the
Treasury bill rate.20 In the context of money market funds, spread

19. This is in contrast to specifications (1) and (2), which use time-varying con-
trols. As a result, the sample size is slightly smaller than in Tables II and III. The
results in Tables II and III are robust to using the smaller sample.
20. We use the spread instead of the yield for consistency with the previous
section. We note that this has no effect on our coefficient of interest because all
regressions include week fixed effects.
1100 QUARTERLY JOURNAL OF ECONOMICS

is a good measure of risk because it carries little scope for man-

agerial skill, which makes fund spread largely reflect fund port-
folio risk. A potential problem with using this measure is that it
may vary over time even though managers may not make any
active changes in the risk profile of their portfolios, only because
the yields of individual assets in the portfolio change.
To account for such mechanical changes in portfolio riski-
ness, we propose two other measures. Holdings Risk is the frac-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

tion of bank obligations net of repurchase agreements and
Treasuries in a fund portfolio. As reported in Table II, repos
and U.S. Treasuries are the safest instruments and bank obli-
gations are the riskiest instrument. Maturity Risk is the
weighted average maturity of assets in a fund portfolio. In gen-
eral, funds with longer maturities of their assets would be con-
sidered riskier.
We begin with a nonparametric analysis of the observed ef-
fects. For each month between January 2006 and August 2008,
we estimate the coefficient  from the cross-sectional regression
model (3) by interacting our main variable Fund Business with
month-fixed effects. Panel A of Figure V presents the estimates
for Holdings Risk. We find no visible differences in the impact of
Fund Business on portfolios’ risk prior to August 2007, but start-
ing from August 2007 the effect is negative and large. Panel B
reports the results for Maturity Risk, and Panel C for Spread.
Again, we observe similar patterns in loadings on Fund
Business as for Holdings Risk.
Next, we present the results from the difference-in-differ-
ences model corresponding to the nonparametric analysis. In col-
umns (1), (4), and (7) of Table IV, we show the results for
specification (3). For the post period, we find that a 1 standard
deviation increase in Fund Business reduces Holdings Risk by 3.7
percentage points, Maturity Risk by 2.3 days, and Spread by 3.0
basis points. The results are statistically and economically sig-
nificant: a 1 standard deviation increase in Fund Business cor-
responds to a 14.9% reduction in Holdings Risk relative to the
cross-sectional standard deviation of Holdings Risk. The respect-
ive quantities for Maturity Risk and Spread are 19.4% and 19.2%.
We obtain similar results for Conglomerate. In contrast, we do not
find any statistically significant impact of business spillovers on
risk in the pre period.
Our results might be also driven by unobserved time-invari-
ant differences among funds or fund sponsors that are correlated
 HOW SAFE ARE MONEY MARKET FUNDS? 1101

Panel A: Holdings Risk

0.2 Pre Post

0.1

-0.1
Coefficient

-0.2

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

-0.3

-0.4

-0.5

-0.6

-0.7
Oct-06

Oct-07
Jan-06
Feb-06

Apr-06
May-06
Jun-06
Jul-06
Aug-06
Sep-06

Nov-06
Dec-06
Jan-07
Feb-07

Apr-07
May-07
Jun-07
Jul-07
Aug-07
Sep-07

Nov-07
Dec-07
Jan-08
Feb-08

Apr-08
May-08
Jun-08
Jul-08
Aug-08
Mar-06

Mar-07

Mar-08
Panel B: Maturity Risk
30 Pre Post

20

10
Coefficient

-10

-20

-30
Dec-06

Dec-07
Jan-06
Feb-06

Apr-06
May-06
Jun-06
Jul-06
Aug-06
Sep-06
Oct-06
Nov-06

Jan-07
Feb-07

Apr-07
May-07
Jun-07
Jul-07
Aug-07
Sep-07
Oct-07
Nov-07

Jan-08
Feb-08

Apr-08
May-08
Jun-08
Jul-08
Aug-08
Mar-06

Mar-07

Mar-08

FIGURE V
Sponsor’s Fund Business Spillovers and Risk Taking
Each of the three panels plots interaction coefficients from an OLS regres-
sion. The dependent variable is one of the three risk measures: holdings risk
(A), maturity (B), and spread (C). The main independent variable is the inter-
action of the fund sponsor’s share of other mutual fund assets relative to all
total fund assets and monthly indicator variables. We include all control vari-
ables defined in Table IV.
1102 QUARTERLY JOURNAL OF ECONOMICS

Panel C: Spread
40 Pre Post
30
20
10
Coefficient

0
-10
-20

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

-30
-40
-50
-60
Jan-06
Feb-06
Mar-06
Apr-06

Jun-06
Jul-06
Aug-06
Sep-06
Oct-06
Nov-06
Dec-06
Jan-07
Feb-07
Mar-07
Apr-07

Jun-07
Jul-07
Aug-07
Sep-07
Oct-07
Nov-07
Dec-07
Jan-08
Feb-08
Mar-08
Apr-08

Jun-08
Jul-08
Aug-08
May-06

May-07

May-08
FIGURE V
Continued

with business spillovers. We address this issue by including spon-

sor fixed effects in columns (2), (5), and (8), and fund fixed effects
in columns (3), (6), and (9). In all these specifications, our results
are almost unchanged.

IV.D. Robustness
1. Do Unobserved Sponsor Characteristics Explain Risk
Choices? Our results suggest that a sponsor’s non–money
market business has a significant impact on its funds’ risk
taking. However, our effects may be driven by unobserved differ-
ences in investment styles or manager ability across fund
families, which in turn may be correlated with business spill-
overs. Specifically, a fund sponsor’s business spillovers may be
correlated with the quality of a sponsor’s risk management or a
sponsor’s risk aversion. To the extent that the variation in this
variable among funds is permanent, our estimator already ac-
counts for such differences. However, our empirical approach
might fail if the variation differentially affects risk taking in
the pre and post periods. For example, fund sponsors may differ
in their reactions to any changes in the quantity of risk, or in their
propensities to take risk when risk-taking opportunities arise,
because of these unobserved variables.
 TABLE IV
SPONSOR’S BUSINESS SPILLOVERS AND RISK TAKING

Holdings Riski,t+1 Maturity Riski,t+1 Spreadi,t+1

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Fund Businessi,2006*Postt 18.346* 21.289** 19.636** 11.449** 12.588** 11.902** 15.040** 15.473** 14.218*
(9.298) (8.705) (8.942) (5.335) (5.782) (5.910) (7.470) (7.443) (7.419)
Conglomeratei,2006*Postt 6.781** 5.580* 6.326** 1.665 1.520 1.703 7.263*** 7.215*** 7.321***
(3.017) (2.885) (2.962) (1.722) (1.750) (1.762) (2.429) (2.424) (2.428)
Fund Businessi,2006 18.126 5.398 2.764
(13.540) (5.310) (1.929)
Conglomeratei,2006 6.539 1.698 1.212*
(4.233) (1.938) (0.655)
Controlsi,2006 Y Y Y Y Y Y Y Y Y
Week fixed effects Y Y Y Y Y Y Y Y Y
Sponsor fixed effects N Y N N Y N N Y N
Fund fixed effects N N Y N N Y N N Y
Observations 19,096 19,096 19,096 19,096 19,096 19,096 19,096 19,096 19,096
R-squared 0.209 0.624 0.780 0.142 0.482 0.587 0.952 0.957 0.959
HOW SAFE ARE MONEY MARKET FUNDS?

Notes. The sample is all U.S. institutional prime money market funds for the period from 1/1/2006 to 8/31/2008. The dependent variables are: the fraction of assets held in risky
assets net of the riskless assets (Holdings Risk) in columns (1)–(3), average portfolio maturity (Maturity Risk) in columns (4)–(6), and the weekly annualized spread (Spread) in
columns (7)–(9). Fund Business is the sponsor’s share of mutual fund assets other than institutional prime money market funds in total sponsor’s assets. Conglomerate is an
indicator variable equal to 1 if the fund sponsor is affiliated with a financial conglomerate, and 0 otherwise. Post is an indicator variable equal to 1 for the period from 8/1/2007 to 8/
31/2008, and 0 otherwise. The other independent variables (Controls) are fund assets, expense ratio, fund age, and fund family size as of 1/1/2006, and interactions of these
variables with Post (coefficients not shown). All regressions are at the weekly level and include week fixed effects. Columns (2), (5), and (8) include sponsor fixed effects, and
columns (3), (6), and (9) include fund fixed effects. Standard errors are clustered at the sponsor level. ***, **, * represent 1%, 5%, and 10% statistical significance, respectively.
1103

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

1104 QUARTERLY JOURNAL OF ECONOMICS

Although we believe such differences are not obvious a

priori, we conduct a more direct test, in which we identify the
coefficients of interest based on the differences between institu-
tional and retail funds. This test is based on the observation
that most fund sponsors offer funds to both retail and institu-
tional clients. Hence, we expect that funds with the same spon-
sor should have similar levels of risk if their behavior is
governed by sponsor-specific characteristics, such as the quality

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

of the sponsor’s risk management or a sponsor’s risk aversion.
However, retail investors react much less to yield differentials
across funds than do institutional investors; therefore, we
expect a much smaller effect for retail funds if risk taking is
driven by business spillovers. Given that retail funds have a
different asset base, we adjust our previously used measure of
business spillovers accordingly and introduce a corresponding
measure, Retail Fund Business, computed as the sponsor’s
share of mutual fund assets other than retail prime money
market funds in total sponsor’s assets.
We begin our analysis with estimating the flow–performance
relationship for retail funds, separately for the pre and post per-
iods, with and without fund fixed effects. We restrict our sample
to sponsors that offer both retail and institutional prime funds.
Our results are robust to including sponsors that only offer retail
prime funds. Panel A in Table V presents the results in columns
(1) to (4). Although we observe a positive coefficient in the pre
period, we find that the flow–performance relationship is quite
weak in the post period. Thus, retail investors react less to yield
differentials across funds. Also, the effect is not driven by busi-
ness spillovers of fund sponsors as shown in columns (5) and (6).
Hence, the risk-taking incentives of retail funds are smaller than
those of institutional funds.
Building on this result, we further analzye risk taking of retail
funds using the setting of Table IV. We present the results in Panel
B of Table V and find striking differences relative those for institu-
tional funds in Table IV. Although we observe statistically and eco-
nomically significant coefficients for institutional funds, the
coefficients are insignificant for retail funds. If anything, the results
go in the opposite direction.
Overall, our results on risk taking across funds are unlikely
to be driven by differences among sponsors along some unob-
served characteristics, such as the quality of risk management
or risk aversion.
 HOW SAFE ARE MONEY MARKET FUNDS? 1105

TABLE V
EVIDENCE FROM RETAIL FUNDS

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

Fund Flowi,t+1

Post Pre Post Pre Post Pre

Panel A: Flow–performance relationship
Spreadi,t 0.002 0.008** 0.003 0.008** 0.008 0.021***

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

(0.002) (0.003) (0.003) (0.003) (0.005) (0.005)
Retail Fund 0.007 0.015**
Businessi,2006* (0.005) (0.007)
Spreadi,t
Conglomeratei,2006* 0.001 0.002
Spreadi,t (0.002) (0.004)
Controlsi,t Y Y Y Y Y Y
Week fixed effects Y Y Y Y Y Y
Fund fixed effects N N Y Y Y Y
Observations 3,724 6,004 3,724 6,004 3,724 6,004
R-squared 0.058 0.022 0.109 0.075 0.110 0.076

Holdings Riski,t+1 Maturity Riski,t+1 Spreadi,t+1

Panel B: Business spillovers and risk taking
Retail Fund 16.098 15.229 2.491 2.263 8.427 8.076
Businessi,2006* (14.417) (14.332) (6.324) (6.276) (14.702) (14.803)
Postt
Conglomeratei,2006* 7.072 6.726 4.045 3.989 4.017 3.973
Postt (6.125) (6.178) (2.529) (2.643) (5.455) (5.466)
Controlsi,2006 Y Y Y Y Y Y
Week fixed effects Y Y Y Y Y Y
Sponsor fixed effects Y N Y N Y N
Fund fixed effects N Y N Y N Y
Observations 9,492 9,492 9,492 9,492 9,492 9,492
R-squared 0.772 0.74 0.606 0.631 0.909 0.911

Notes. The sample is all U.S. retail prime money market funds for the period from 1/1/2006 to 8/31/
2008. In Panel A, we examine the flow–performance relationship for retail prime money market funds
(similar to Table III). In Panel B, we examine the relationship between business spillovers and risk for
retail prime money market funds (similar to Table IV).

2. Do Implicit Government Guarantees Explain Risk Choices?

Money market funds experienced a run after the default of
Lehman Brothers in September 2008. Because the likely conse-
quences of this run were severe, the government decided to bail
out the entire money fund industry and extended explicit guaran-
tees to all money funds and their investors. Effectively, for the
duration of the guarantee (which lasted over a year), this
1106 QUARTERLY JOURNAL OF ECONOMICS

intervention largely eliminated the risk of fund failure. Given that

the government did not rescue the RPF, this guarantee was likely
unexpected. However, we can test for the importance of implicit
guarantees because if the expectation of future government bail-
outs drives funds’ risk taking before Lehman’s default, then we
should find similar differences in risk taking after the introduction
of explicit guarantees.
To evaluate this hypothesis, we extend our analysis in Table IV

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

to three periods: January 2006–July 2007, August 2007–August
2008, and April 2009–December 2010. We do not include the data
for the two quarters immediately following Lehman’s default be-
cause funds need some time to adjust the risk of their funds.21 Also,
many financial markets were illiquid right after the default, so any
adjustment of risk by the funds was difficult to accomplish. Our
empirical strategy involves estimating the risk model, in which
Business Spillovers is interacted with two indicator variables:
Post, equal to 1 for the period August 2007–August 2008, and 0
otherwise; and Post-Lehman equal to 1 for the period April 2009–
December 2010, and 0 otherwise. We expect a zero effect of
Business Spillovers in the pre period, a negative effect in the post
period, and again a zero effect in the post-Lehman period.
We report the results in Table VI. Per our hypothesis, the
coefficient of the interaction term between Business Spillovers
and Post-Lehman is close to 0 for two risk measures. Hence, the
importance of business spillovers has become negligible once the
government rolled out an explicit support for all funds. This find-
ing indicates that our results cannot be explained by implicit gov-
ernment guarantees prior to Lehman’s default.22

3. Does Variation in Sponsor Financial Strength Explain Our

Results? Our empirical analysis so far reveals the importance of
business spillovers as a driver of risk-taking decisions of money
market funds. This result should be particularly strong if the

21. Duygan-Bump et al. (2013) and Kacperczyk and Schnabl (2010) discuss the
workings and exact timing of different government interventions.
22. Our findings stand in contrast with theoretical and empirical work on the
effect of bailouts on risk taking, which has examined the role of deposit insurance
and bank charter value (Keeley 1990; Freixas, Loranth, and Morrison 2007), man-
agerial control (Saunders, Strock, and Travlos 1990), monitoring by depositors
(Esty 1997), implicit guarantees provided by stakeholders (Panageas 2010), and
provision of systemic guarantees by the government (Kelly, Lustig, and Van
Nieuwerburgh 2011).
 TABLE VI
RISK TAKING AFTER GOVERNMENT GUARANTEE

Holdings Riski,t+1 Maturity Riski,t+1 Spreadi,t+1

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

Fund Businessi,2006*Postt 22.319** 19.636** 12.465** 11.902** 15.326** 14.218*
(8.529) (8.935) (5.625) (5.906) (7.429) (7.413)
Fund Businessi,2006*Post-Lehmant 22.999 9.592 9.419 5.162 2.965 1.057
(18.611) (19.229) (6.488) (6.283) (7.175) (7.502)
Conglomeratei,2006*Postt 5.278* 6.326** 1.399 1.703 7.123*** 7.321***
(2.899) (2.959) (1.728) (1.760) (2.413) (2.426)
Conglomeratei,2006*Post-Lehmant 0.860 3.569 4.081* 5.010** 3.441 4.044*
(5.894) (5.649) (2.159) (2.131) (2.168) (2.237)
Controlsi,2006 Y Y Y Y Y Y
Week fixed effects Y Y Y Y Y Y
Sponsor fixed effects Y N Y N Y N
Fund fixed effects N Y N Y N Y
Observations 28,409 28,409 28,409 28,409 28,409 28,409
R-squared 0.579 0.694 0.457 0.537 0.953 0.954
HOW SAFE ARE MONEY MARKET FUNDS?

Notes. The sample is all U.S. institutional prime money market funds for the period from 1/1/2006 to 12/31/2010. We estimate the same regression models as in Table IV for the
period from July 2006 to December 2010. We drop the month of the Lehman’s bankruptcy and the two quarters immediately after the Lehman’s bankruptcy to focus on risk taking
after a short adjustment period. We interact our main variables of interest with an indicator variable for the post period (July 2007–August 2008) and the post-Lehman period
(April 2009–December 2010). All regressions include the control variables specified in Table IV and interactions of the control variables with the post-Lehman indicator variable
(coefficients not shown). The regressions are at the weekly level and include week fixed effects. Columns (1), (3), and (5) include sponsor fixed effects, and columns (2), (4), and (6)
include fund fixed effects. Standard errors are clustered at the sponsor level. ***, **, * represent 1%, 5%, and 10% statistical significance, respectively.
1107

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

1108 QUARTERLY JOURNAL OF ECONOMICS

TABLE VII
CONTROLLING FOR FINANCIAL STRENGTH

Stand-alone asset managers

Holdings Riski,t+1 Maturity Riski,t+1 Spreadi,t+1

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

Fund Businessi,2006* 39.968*** 38.720*** 11.210** 10.029** 22.829*** 22.413***

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Postt (12.720) (12.656) (4.462) (4.403) (7.399) (7.470)
No Ratingi,2006* 8.901 8.273 2.121 1.689 7.858** 7.777**
Postt (5.394) (5.406) (1.437) (1.355) (3.082) (3.065)
Controlsi,2006 Y Y Y Y Y Y
Week fixed effects Y Y Y Y Y Y
Sponsor fixed effects Y N Y N Y N
Fund fixed effects N Y N Y N Y
Observations 7,645 7,645 7,645 7,645 7,645 7,645
R-squared 0.607 0.715 0.589 0.671 0.968 0.968

Notes. The sample is all U.S. institutional prime money market funds affiliated with an independent
asset manager for the period 1/1/2006–8/31/2008. The dependent variables, Fund Business and Post, are
defined in Table IV. No Rating is an indicator variable equal to 1 if the sponsor has no credit rating, and 0
otherwise. All regressions include the same control variables as in Table IV (coefficients not shown). They
are at the weekly level and include week fixed effects and fund fixed effects. Standard errors are clustered
at the sponsor level. ***, **, * represent 1%, 5%, and 10% statistical significance, respectively.

bailouts by fund sponsor are ex post optimal. However, the like-

lihood of a bailout also depends on a sponsor’s ability to do so,
which depends on the fund sponsor’s financial strength. If the
degree of business spillovers is correlated with financial strength,
then our main results may reflect the ability (rather than the
incentive) to bail out a fund. Even though these concepts are
closely related, we would like to distinguish empirically between
the effect of spillovers and the effect of financial strength.
To allow for such separation we refine our empirical design.
We analyze risk choices only for funds sponsored by stand-alone
asset managers. By analyzing stand-alone asset managers, we
can fix the financial strength margin while varying the busi-
ness spillovers margin. The key assumption underlying this
test is that stand-alone managers are highly constrained in
their ability to bail out funds. This assumption is plausible be-
cause stand-alone managers, unlike financial conglomerates, do
not have access to outside sources of funding in times of crises.
Table VII presents the results from estimating the same model
as in Table IV for stand-alone asset managers only. Consistent
with our earlier results, we find that the extent of non–money
market fund business (Fund Business) has an economically and
 HOW SAFE ARE MONEY MARKET FUNDS? 1109

statistically significant negative effect on all three risk

measures.
Although the premise of our test is similarity in financial
strength, we additionally control for any remaining variation in
a sponsor’s financial strength. Our measure of financial strength
is a fund sponsor’s credit rating. The reason credit rating might
be a good proxy for our purpose is that fund sponsors with good
credit standings may be more able to access short-term funding

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

markets and as such they may have more capacity to provide
support. We measure rating quality with an indicator variable
(No Rating) equal to 1 if the fund sponsor has no rating and 0
otherwise, that is, fund sponsors without a credit rating are
deemed to have lower financial strength.23
We find that the coefficient of the interaction term with No
Rating is negative for all risk measures. However, it is statistic-
ally significant only for one measure. This result is not entirely
surprising because our test was designed to eliminate much of the
cross-sectional variation in financial strength, and thus any un-
explained variation in financial strength is likely to have a low
statistical power to explain variation in risk.24 Overall, the re-
sults strengthen our interpretation that—conditional on financial
strength—the degree of business spillovers negatively affects risk
taking.

4. Additional Evidence. One possible explanation of our find-

ings could be that fund managers or fund management compa-
nies differ in their compensation levels; hence, they have
different incentives to take on risk. For example, if managers of
funds sponsored by companies with significant non–money
market fund business had lower compensation, such managers
could have greater incentives to take risk to increase their funds’
assets under management. We evaluate this possibility by relat-
ing the value of a fund’s compensation to Fund Business. We
compute total compensation as the product of fund size and ex-
pense ratio, which is the percentage fee charged by the fund on its
assets. The results from this estimation do not support the idea

23. We also explored the continuous version of ratings, while additionally as-
signing a low rating value for sponsors without rating. The results are similar.
24. The correlation between No Rating and Fund Business is relatively low a
32.4%. This provides further evidence that financial strength does not simply proxy
for business spillovers.
1110 QUARTERLY JOURNAL OF ECONOMICS

that differences in risk can be attributed to differences in man-

agerial compensation. If anything, we observe the opposite effect.
Funds with more non–money market fund business, on average,
have higher compensation levels.
In another test, we explore the importance of outliers. Money
funds in our sample exhibit a significant cross-sectional disper-
sion in their business spillovers. In fact, a few funds display par-
ticularly low levels, largely because they specialize in the money

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

market fund management. The presence of such cases raises the
possibility that our results might be driven by a few extreme ob-
servations. We inspect the data using various scattered plots and
find no good reason to believe that the outliers drive our results.
Furthermore, we exclude all fund observations with Fund
Business below 50% and reestimate the regression model in
Table IV. The results remain qualitatively unchanged and if any-
thing become quantitatively stronger.
The workings of money market funds often depend on the
size of the fund company. Anecdotally, large funds are considered
to be more involved in active risk choices, whereas smaller funds
are considered to be simple cash-parking vehicles that do not
engage in active risk-taking strategies. Hence, one would
expect our results to be stronger for large funds. To this end,
we estimate the regression model in Table IV for the subsample
of funds with assets under management over $1 billion, the value
often treated by practitioners as a cutoff for the fund to be con-
sidered large and important. We find that the risk effect indeed
becomes stronger, but not by much.
Finally, a possible concern with our results relates to our
motivating example. In particular, the case of the RPF consti-
tutes one of the most extreme risk-shifting behaviors among all
fund sponsors. To the extent that the RPF is sponsored by an
asset manager without business spillovers, our results might be
partially driven by the RPF. We exclude the fund from our sample
and reestimate the regression model in Table IV on the restricted
sample. We find no significant difference in magnitude of the co-
efficients of Business Spillovers.

IV.E. Post-Lehman Analysis

In this section, we assess the cross-sectional variation in the
response of fund investors and sponsors after Lehman’s default.
We focus on the one-week period after Lehman’s default because
 HOW SAFE ARE MONEY MARKET FUNDS? 1111

the run started with the redemption requests from RPF immedi-
ately after Lehman’s default and ended with the introduction of
government deposit insurance on money market funds.25 In our
first test, we assess the impact of risk taking and business spill-
overs on fund redemptions. To this end, we estimate the following
regression model:
Redemptionsi ¼  þ 1 Business Spilloversi þ 2 Spreadi þ Xi þ "i ,

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

ð4Þ
where Redemptions is the change in a fund size between
September 18 and September 25, 2008. Business Spillovers is
measured as before and Spread is the fund yield minus the one-
month Treasury bill rate in the week before the run. X is a vector
of controls that includes Log(Fund Size), Age, Expense Ratio, and
Log(Family Size) measured in the week prior to the run.
We present the results in column (1) of Table VIII. We find
that funds whose sponsors have greater business spillovers have
smaller redemptions: A 1 standard deviation increase in Fund
Business reduces redemptions by 3.4 percentage points, or by
32.0% of the average redemption. We find no statistically signifi-
cant effect of Conglomerate on redemptions. Importantly, this
effect is over and above the effect of prior risk taking as captured
by Spread. Funds with higher Spread also receive higher redemp-
tion requests: a 1 standard deviation increase in spread raises
redemption requests by 3.3 percentage points. Hence, these find-
ings are strongly supportive of our hypothesis that funds should
take into account runs when choosing their risk levels.
Subsequently, we assess the effect of business spillovers on
the likelihood of financial support. To assess financial support, we
collect data on no-action letters filed by the SEC. Funds typically
request such letters from the SEC before providing financial sup-
port to avoid charges of insider trading, and these requests are
routinely approved by the SEC. To illustrate the scale and scope
of the support, in Appendix Table A.1 we provide detailed infor-
mation about support arrangements established in the aftermath
of Lehman’s collapse. It is apparent that the support was not
limited to a few funds but was a common occurrence during
that period. In total, we observe 28 support events in the week
following Lehman’s default.26

25. Wermers (2011) also examines the relation between redemption requests
and fund characteristics in this period.
1112 QUARTERLY JOURNAL OF ECONOMICS

TABLE VIII
POST-LEHMAN RESULTS

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

Redemptions Support Exit Name change
Fund Businessi 0.159** 0.119 0.156 0.334*
(0.074) (0.366) (0.247) (0.174)
Conglomeratei 0.001 0.294** 0.243*** 0.018
(0.029) (0.133) (0.079) (0.101)

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Spreadi 0.001** 0.001 0.001 0.000
(0.000) (0.002) (0.001) (0.002)
Log(Fund Size)i 0.026*** 0.019 0.061** 0.023
(0.007) (0.015) (0.030) (0.018)
Expense Ratioi 0.191*** 0.019 0.188 0.216
(0.055) (0.185) (0.207) (0.504)
Agei 0.028 0.086 0.07 0.012
(0.033) (0.083) (0.059) (0.067)
Log(Family Size)i 0.015*** 0.03 0.028* 0.002
(0.004) (0.022) (0.016) (0.007)
Constant 0.041 0.273 0.166 0.369
(0.193) (0.444) (0.360) (0.389)

Observations 105 105 105 105

R-squared 0.381 0.189 0.135 0.053

Notes. The sample is all U.S. institutional prime money market funds that were active from 1/1/2006
until 10/1/2009. In column (1) the dependent variable is Redemptions defined as total value of redemptions
(fund outflows) in the week after Lehman’s bankruptcy (9/18–25/2008). In column (2) the dependent
variable is Support, an indicator variable equal to 1 if the fund’s sponsor filed a no-action letter with
the SEC in the week after the Lehman’s bankruptcy and 0 otherwise (20 funds declared support). In
column (3) the dependent variable is Exit, an indicator variable equal to 1 if the fund was closed in the
two years after the expiration of the government guarantee (10/1/2009), and 0 otherwise (16 out of 105
fund closures). In column (4) the dependent variable is Name, an indicator variable equal to 1 if the fund
name was changed to match the sponsor name, 0 if the name was unchanged, and equal to –1 if the fund
name was changed to be different from the sponsor name (eight name changes). All independent variables
are defined in Table III. In columns (1) to (3), the independent variables are measured as of the week
before the Lehman bankruptcy. In column (4), the independent variables are defined as of the end of the
government guarantee (10/1/2009). Standard errors are clustered at the sponsor level. ***, **, * represent
1%, 5%, and 10% statistical significance, respectively.

We analyze the effect of business spillovers on financial sup-

port using the following regression:
ð5Þ Supporti ¼  þ Business Spilloversi þ Xi þ "i ,
where Support takes a value of 1 if the fund sponsor offered fi-
nancial support to its fund and 0 otherwise.

26. Brady, Anadu, and Cooper (2012) use data from money market SEC filings
to determine the extent of financial support. They find that several funds received
sponsor support before September 2008. Importantly, these instances were trig-
gered by idiosyncratic events. In contrast, the post-Lehman run was the first in-
dustry-wide run in the history of money market funds.
 HOW SAFE ARE MONEY MARKET FUNDS? 1113

We present the estimation results in column (2) of Table VIII.

We find no effect of Fund Business on financial support. This
result suggests that funds incorporate the likelihood of providing
financial support in their ex ante risk taking. We find a positive
and statistically significant effect of Conglomerate on the prob-
ability of receiving financial support: funds affiliated with finan-
cial conglomerates are 29.4% more likely to receive financial
support in the week after Lehman’s default relative to stand-

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

alone asset managers. These findings provide strong support for
our earlier tests suggesting that stand-alone managers have lim-
ited ability to provide financial support.
Next, we evaluate the impact of business spillovers on fund
exit in the two years following October 1, 2009, which is the ex-
piration date of the government guarantee program. We identify
16 instances of fund closures during that period, which we want
to relate to reputation effects. We estimate the following regres-
sion model:

ð6Þ Exiti ¼  þ Business Spilloversi þ Xi þ "i ,

where Exit is an indicator variable equal to 1 if a fund exited the
market between October 1, 2009 and September 30, 2011, and 0
otherwise. Business Spillovers and X are defined as before.
The results in column (3) of Table VIII show that funds asso-
ciated with conglomerates are more likely to exit the market fol-
lowing the run on the industry. In contrast, we find no evidence of
such effect for Fund Business. The results suggest that in re-
sponse to adverse conditions in the industry sponsors with
greater business spillovers exit the market, possibly to shield
themselves from negative spillovers in the future.
Finally, we analyze the effect of business spillovers on fund
naming strategies. Prior to the run, some funds had names that
were distinctly different from the names of their fund sponsors.
However, in the aftermath of the run, some funds decided to
change their names in a way that would closely reflect the under-
lying sponsor name. For example, Bank of America used to offer a
fund named Columbia Cash Reserves, but this fund changed its
name to Bank of America Cash Reserves in November 2009. We
posit that funds with greater business spillovers might be more
likely to change their names because they want to signal to their
investors the potential safety of their operations. Our sample in-
cludes eight such name changes over the period of two years.
1114 QUARTERLY JOURNAL OF ECONOMICS

To evaluate the hypothesis, we estimate the following regres-

sion model:
ð7Þ NameChangei ¼  þ Business Spilloversi þ Xi þ "i ,
where Name Change is an indicator variable equal to 1 if a fund
changed its name to mimic its sponsor’s name between October 1,
2009, and September 30, 2011, equal to –1 if the fund changed its
name away from that of its sponsor,27 and 0 otherwise. Business

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Spillovers and X are defined as before and measured as of October
2009. The results in column (4) of Table VIII show that funds with
greater business spillovers, as measured by Fund Business, are
more likely to change their names following the run on the
industry.
Overall, our analysis suggests that concerns over possible
loss of business played an important role in the way fund in-
vestors and fund sponsors evaluated their funds in the aftermath
of the run on money market fund industry.

V. Concluding Remarks
We study the risk-taking incentives of money market funds.
We show that funds have strong incentives to take on risk be-
cause their flows are highly responsive to changes in fund yields.
We also find that observable fund characteristics predict funds’
risk taking. Using the change in relative risks of money market
instruments during the financial crisis of 2007–10 as an exogen-
ous shock to the funds’ risk-taking opportunities, we show that
funds sponsored by companies with more money fund business
took on more risk. We further show that funds whose sponsors
had less money fund business experienced smaller outflows, were
more likely to provide financial support during a market-wide
run in September 2008, and were more likely to exit the industry
or change their fund names.28
Our findings suggest that money market funds exert a po-
tentially destabilizing effect on financial markets. Money market

27. In our data, we observe only one instance of such a reverse name change.
28. The idea of tracing the impact of an exogenous crisis shock on money market
funds has been also used in a recent paper by Chernenko and Sunderam (2012).
They examine the ‘‘quiet run’’ on money market funds in 2011 and focus on the
impact on money market fund borrowers. Similar to our article, they find a strong
flow–performance relationship and thus offer further robustness to our findings.
 HOW SAFE ARE MONEY MARKET FUNDS? 1115

funds thus add a layer of financial intermediation between is-

suers (mainly financial institutions) and investors (mainly cor-
porations). Our results indicate that money market funds have
strong incentives to chase yields and are vulnerable to runs if the
risk materializes. Hence, this reduces the market discipline on
financial institutions and makes them more vulnerable to finan-
cial shocks.
In addition, our results provide new insights into the role of

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

short-term claimants in the modern theories of financial inter-
mediation (e.g., Diamond and Rajan 2000, 2001). Although
these theories argue that such claimants can mitigate intermedi-
aries’ incentives to take risk through the threat of runs, our re-
sults suggest that they may instead exacerbate risk-taking
incentives because of yield chasing by depositors.
Finally, although our explanation of risk taking mostly em-
phasizes the role of business spillovers, one could imagine other
explanations of our findings, such as ‘‘conglomerate bureau-
cracy.’’ In particular, stand-alone firms are known to respond
more aggressively to changes in industry Q than do the divisions
of a conglomerate. By the same token, an independent money
fund may respond more strongly to an opportunity to rapidly
grow its assets. Although the lack of precise data on internal de-
cision making inside fund organization makes it difficult to test
this theory directly, one could also argue that the bureaucracy
effect is related to business spillovers and arises endogenously to
protect fund sponsor’s reputation from risk-taking behavior of an
individual division.

Appendix: Data Construction

The main source of our data on money market funds is
iMoneyNet. The data are widely used across the money market
fund industry and represent the primary source of information on
money market funds. The iMoneyNet database covers the uni-
verse of money market funds. Every week all funds submit infor-
mation on total assets, yields, expense ratios, and holdings by
instrument. We confirm the full coverage by comparing the
iMoneyNet data with the list of all funds based on SEC data.
We also aggregate fund assets and compare them with total
fund assets reported by the SEC. Both tests confirm that
iMoneyNet covers the universe of money market funds. Most
 1116

APPENDIX TABLE A.1

DETAILED INFORMATION ON THE POST-LEHMAN SUPPORT ARRANGEMENTS

Support Value of distressed Support

Fund company Sponsor date Distress reason assets ralue Remarks

Dreyfus Cash Mgmt. BNY Mellon 10/20/2008 Lehman Brothers $97.2M Cash contribution necessary to maintain the fund value
Plus notes at 0.995
All Dreyfus funds BNY Mellon 10/20/2008 Distress of eligible CSA (Cash contribution necessary to maintain the fund
assets value at 0.995)
All Citizens funds BNY Mellon 10/20/2008 Distress of eligible CSA (Cash contribution necessary to maintain the fund
assets value at 0.995)
All General funds BNY Mellon 10/20/2008 Distress of eligible CSA (Cash contribution necessary to maintain the fund
assets value at 0.995)
Dreyfus Basic MMF BNY Mellon 10/20/2008 Lehman Brothers $45M Cash contribution necessary to maintain the fund value
notes at 0.995
Dreyfus LAP BNY Mellon 10/20/2008 Lehman Brothers $100M Cash contribution necessary to maintain the fund value
notes at 0.995
Dreyfus Worldwide BNY Mellon 10/20/2008 Lehman Brothers $20M Cash contribution necessary to maintain the fund value
Dollar MMF notes at 0.995
Russell MMF Northwestern 10/20/2008 The entire fund CSA (Cash contribution necessary to maintain the fund
Mutual Life Ins. value at 0.995)
USAA MMF USAA 10/22/2008 AIG notes $81.96M CSA USAA (Cash contribution necessary to maintain
QUARTERLY JOURNAL OF ECONOMICS

the fund value at 0.995)

Touchstone Advisors 10/22/2008

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 APPENDIX TABLE A.1
(CONTINUED)

Support Value of distressed Support

Fund company Sponsor date Distress reason assets ralue Remarks

Touchstone Invest. Morgan Stanley, MS ($5.06M), ST Cash contribution necessary to maintain the fund value
Trust Instit. MMF Southtrust Bank, ($1.4M), Wach. at 0.995 (LOC by Western and Southern Life
Wachovia notes ($6.08M) Insurance Company)
Touchstone Invest. Touchstone Advisors 10/22/2008 Morgan Stanley, MS ($5.1M), ST Cash contribution necessary to maintain the fund value
Trust MMF Southtrust Bank, ($1.6M), Wach. at 0.995 (LOC by Western and Southern Life
Wachovia notes ($4.07M) Insurance Company)
Touchstone Variable Touchstone Advisors 10/22/2008 Morgan Stanley, MS ($2.25M), Wach. Cash contribution necessary to maintain the fund value
Series MMF Southtrust Bank, ($1.5M) at 0.995 (LOC by Western and Southern Life
Wachovia notes Insurance Company)
Tamarack Prime Voyageur Asset 10/22/2008 The entire fund Cash contribution necessary to maintain the fund value
MMF Management at 0.995 (LOC by RBC)
Tamarack Instit. Voyageur Asset 10/22/2008 The entire fund Cash contribution necessary to maintain the fund value
Prime MMF Management at 0.995 (LOC by RBC)
RidgeWorth Prime SunTrust Banks 10/22/2008 Lehman Brothers $70M $70M Exchange of SunTrust Note for the Lehman note in the
Quality MMF notes amount of $70M
Principal MMF Principal Financial 10/22/2008 AIG notes CSA (Cash contribution necessary to maintain the fund
Group value at 0.995)
Principal Variable Principal Financial 10/22/2008 AIG notes CSA (Cash contribution necessary to maintain the fund
Contracts MMF Group value at 0.995)
Morgan Stanley Morgan Stanley 10/22/2008 The entire fund CSA (Cash contribution necessary to maintain the fund
funds value at 0.995)
HOW SAFE ARE MONEY MARKET FUNDS?

Active Assets funds Morgan Stanley 10/22/2008 The entire fund CSA (Cash contribution necessary to maintain the fund
value at 0.995)
Columbia MM Bank of America 10/22/2008 The entire fund CSA (Cash contribution necessary to maintain the fund
Reserves value at 0.995)
1117

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 1118

APPENDIX TABLE A.1

(CONTINUED)

Support Value of distressed Support

Fund company Sponsor date Distress reason assets ralue Remarks

ING LAP ING Groep N.V. 10/22/2008 AIG notes $46M CSA (Cash contribution necessary to maintain the fund
value at 0.995)
ING Classic MMF ING Groep N.V. 10/22/2008 AIG notes $28M CSA (Cash contribution necessary to maintain the fund
value at 0.995)
ING Instit. Prime ING Groep N.V. 10/22/2008 AIG notes $46M CSA (Cash contribution necessary to maintain the fund
MMF value at 0.995)
ING MMF ING Groep N.V. 10/22/2008 AIG notes; Lehman AIG ($8.5M), CSA (Cash contribution necessary to maintain the fund
notes Lehman ($2M) value at 0.995)
ING Brokerage Cash ING Groep N.V. 10/22/2008 AIG notes $8M CSA (Cash contribution necessary to maintain the fund
Reserves value at 0.995)
Western Asset Instit. Legg Mason 10/22/2008 Orion Finance. LLC $75M $20M CSA
MMF notes (SIV)
Western Asset Instit. Legg Mason 10/22/2008 The fraction of fund $452M CSA (Cash contribution necessary to maintain the fund
MMF value at 0.9975)
Russell MMF Northwestern 10/24/2008 Lehman Brothers $403M CSA (Cash contribution necessary to maintain the fund
Mutual Life notes value at 0.995)

Notes. The columns include the fund company offered support, the sponsor company providing support, the support date, the reason for support, the values of distressed
securities, the value of support, and additional remarks. CSA: Capital Support Agreement; POF: Prime Obligations Fund; DAP: Diversified Assets Portfolio; LAP: Liquid Assets
QUARTERLY JOURNAL OF ECONOMICS

Portfolio; LRP: Liquid Reserves Portfolio; POP: Prime Obligations Portfolio.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

 HOW SAFE ARE MONEY MARKET FUNDS? 1119

detailed and accurate information is available for the period from

January 2005 to September 2011, which is the period of our
analysis.
The data we obtain are reported at the share-class level. To
ensure precision of our tests, we first check that all share classes
are reported consistently throughout the data set (i.e., a share
class is reported every week after a share class enters the data
set and until it exits from the data set). We find that only 17 out of

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

1,820 share classes have some missing data. Almost all missing
data are from funds that report monthly for the first few months
of their existence and later switch to weekly reporting. We use
linear interpolation to generate weekly data for these funds.
We focus on prime money market funds. We have 236,335
total observations (and 1,027 share classes) for the period of our
analysis. Because our main analysis is at the fund level, we aggre-
gate the data across share classes. For that reason, we create a
unique fund identifier using information on total fund assets that
is provided together with assets per share class. The data also
provide an indicator variable equal to 1 for a fund’s main share
class and 0 for other share classes. We use this information to
double-check the construction of unique fund identifiers.
We perform several data checks to ensure that the construc-
tion of fund identifiers is accurate. Specifically, we verify that the
assets for all share classes add up to total fund size. These data
are available for most of our data set (174,706 observations). We
examine observations for which the difference between the two is
greater than $100,000 (the data are reported in $100,000 incre-
ments). There are only 201 observations for which the difference
exceeds $100,000. We further test whether reported asset hold-
ings add up to 100%. We only find 28 for which asset holdings do
not add up.
We construct fund-level variables by aggregating (weighted
by asset size) all institutional and retail share classes at the fund-
week level using the unique fund identifier. We obtain a total of
104,449 observations at the week-fund level. We label a fund as
institutional if the fund has at least one institutional share class
(47,959 observations). We label a fund as retail if there is no
institutional share class (56,490 observations). Most of our ana-
lysis focuses on institutional funds over the period from January
2006 to August 2008 (19,998 observations). The main analysis is
restricted to funds that remain in the data set throughout this
period (19,096 observations).
1120 QUARTERLY JOURNAL OF ECONOMICS

We merge the iMoneyNet data to the CRSP Survivorship

Bias Free Mutual Fund Database. The CRSP Mutual Fund
data are at the monthly level and we therefore match at that
frequency (any within-month variation is assumed to be con-
stant). To perform the match, we use the share class NASDAQ
identifier provided by iMoneyNet as our primary identifying
variable. If the NASDAQ identifier matches to more than one
observation in CRSP, we use the share class with the most

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

assets in CRSP. For a small number of observations,
iMoneyNet does not provide a NASDAQ identifier or the
NASDAQ identifier is not reported by CRSP. In that case, we
assign the same NASDAQ identifier based on other share
classes of the same fund. If no other share classes have a
valid NASDAQ identifier, we match the funds based on fund
name. If there is no entry in CRSP, we match directly to the
sponsor name based on the fund’s SEC filings in EDGAR. We
are able to match all fund observations.
For some of our analysis, we compute fund spreads by
deducting the one-month Treasury bill rate from fund yields.
We collect the one-month Treasury bill rate from Ken French’s
website at http://mba.tuck.dartmouth.edu/pages/faculty/ken.
french/data_library.html.
We assign the sponsor based on CRSP data. CRSP data pro-
vides detailed information about asset management companies
that sponsor the respective funds. Most funds have a fixed spon-
sor during our data period. However, in a few cases, fund spon-
sors might change, for example, due to mergers. If the sponsor
changes over the lifetime of a fund, we assign to the fund the
sponsor that was in charge of the fund as of the first week of
January 2006. For all sponsors, we collect information on
whether the sponsor is affiliated with a commercial bank, invest-
ment bank, insurance company, or is managed by a stand-alone
asset manager. We collect the information from COMPUSTAT,
company websites, EDGAR, SEC filings, and press reports. We
use at least two sources to ensure validity of this information. We
ensure that all data are as of January 2006.
We obtain sponsor ratings from several data sources. We first
match sponsor names to S&P RatingsXpress as of January 2006.
Next, we match any unmatched sponsors to the Lehman
Brothers’ Bond Database. We ensure that both data sets provide
the same information. We also double-check the information with
company websites and press releases.
 HOW SAFE ARE MONEY MARKET FUNDS? 1121

Finally, we collect information on financial support from the

SEC website. We collect all no-action letters posted in September
2008 or thereafter. We check with Peter Crane’s industry blog at
cranedata.com to ensure that we cover all sponsors that provided
bailouts. We collect information on the specifics of the bailouts
based on the no-action letter and press releases. We report
detailed information on sponsor bailouts in Appendix Table A.1.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

NYU Stern School of Business and NBER
NYU Stern School of Business, CEPR, and NBER

References
Acharya, Viral, Philipp Schnabl, and Gustavo Suarez, ‘‘Securitization without
Risk Transfer,’’ Journal of Financial Economics, 107, no. 3 (2013), 515–536.
Brady, Steffanie A., Ken E. Anadu, and Nathaniel R. Cooper, ‘‘The Stability of
Prime Money Market Mutual Funds: Sponsor Support from 2007 to 2011,’’
Federal Reserve Bank of Boston, 2012.
Brunnermeier, Markus, ‘‘Deciphering the Liquidity and Credit Crunch 2007–08,’’
Journal of Economic Perspectives, 23, no. 1 (2009), 77–100.
Chernenko, Sergey, and Adi Sunderam, ‘‘Frictions in Shadow Banking: Evidence
from the Lending Behavior of Money Market Funds,’’ (HBS Working Paper,
2012).
Chevalier, Judith, and Glenn Ellison, ‘‘Risk Taking by Mutual Funds as a
Response to Incentives,’’ Journal of Political Economy, 105, no. 6 (1997),
1167–1200.
Christoffersen, Susan E. K., ‘‘Why Do Money Fund Managers Voluntarily Waive
Their Fees?’’ Journal of Finance, 56, no. 3 (2001), 1117–1140.
Christoffersen, Susan E. K., and David K. Musto, ‘‘Demand Curves and the
Pricing of Money Management,’’ Review of Financial Studies, 15, no. 5
(2002), 1499–1524.
Cornett, Marcia, Jamie McNutt, Philip Strahan, and Hassan Tehranian,
‘‘Liquidity Risk Management and Credit Supply in the Financial Crisis,’’
Journal of Financial Economics, 101, no. 2 (2011), 297–312.
Dang, Tri Vi, Gary Gorton, and Bengt Holmstrom, ‘‘Opacity and the Optimality of
Debt for Liquidity Provision,’’ Yale University Working Paper, 2009.
Diamond, Douglas W., and Philip H. Dybvig, ‘‘Bank Runs, Deposit Insurance, and
Liquidity,’’ Journal of Political Economy, 91 (1983), 401–419.
Diamond, Douglas, and Raghuram Rajan, ‘‘A Theory of Bank Capital,’’ Journal of
Finance, 55 (2000), 2431–2465.
———, ‘‘Liquidity Risk, Liquidity Creation and Financial Fragility: A Theory of
Banking,’’ Journal of Political Economy, 109 (2001), 287–327.
Duygan-Bump, Burcu, Patrick M. Parkinson, Eric S. Rosengren, Gustavo
A. Suarez, and Paul S. Willen, ‘‘How Effective Were the Federal Reserve
Emergency Liquidity Facilities? Evidence from the Asset-Backed
Commercial Paper Money Market Mutual Fund Liquidity Facility,’’
Journal of Finance, 68, no. 2 (2013), 715–737.
Esty, Benjamin C., ‘‘Organizational Form and Risk Taking in the Savings and
Loan Industry,’’ Journal of Financial Economics, 44 (1997), 22–55.
Freixas, Xavier, Gyongyi Loranth, and Alan D. Morrison, ‘‘Regulating Financial
Conglomerates,’’ Journal of Financial Intermediation, 16 (2007), 479–514.
FINRA (Financial Industry Regulatory Authority), ‘‘Treasury’s Guarantee
Program for Money Market Mutual Funds: What You Should Know,’’
FINRA, 2010.
1122 QUARTERLY JOURNAL OF ECONOMICS

Gennaioli, Nicola, and Andrei Shleifer, ‘‘What Comes to Mind,’’ Quarterly Journal
of Economics, 125 (2010), 1399–1433.
Gennaioli, Nicola, Andrei Shleifer, and Robert Vishny, ‘‘Neglected Risks,
Financial Innovation, and Financial Fragility,’’ Journal of Financial
Economics, 104, no. 3 (2012), 452–468.
Gorton, Gary, and Andrew Metrick, ‘‘Haircuts,’’ Federal Reserve Bank of St. Louis
Review, 92 (2009), 507–519.
Kacperczyk, Marcin, and Philipp Schnabl, ‘‘When Safe Proved Risky: Commercial
Paper during the Financial Crisis of 2007–2009,’’ Journal of Economic
Perspectives, 24, no. 1 (2010), 29–50.
Keeley, Michael C., ‘‘Deposit Insurance, Risk, and Market Power in Banking,’’
American Economic Review, 80, no. 5 (1990), 1183–1200.

Downloaded from http://qje.oxfordjournals.org/ at University of Pittsburgh on January 11, 2015

Kelly, Bryan, Hanno Lustig, and Stijn Van Nieuwerburgh, ‘‘Too-Systemic-to-Fail:
What Option Markets Imply about Sector-Wide Government Guarantees,’’
(Working Paper, New York University, 2011).
Krishnamurthy, Arvind, Stefan Nagel, and Dmitry Orlov, ‘‘Sizing Up Repo,’’
Stanford University Working Paper, 2012.
Krishnamurthy, Arvind, and Annette Vissing-Jorgensen, ‘‘The Aggregate
Demand for Treasury Debt,’’ Working Paper, Kellogg School, 2010.
McCabe, Patrick, ‘‘The Cross Section of Money Market Fund Risks and Financial
Crises,’’ Working Paper, Federal Reserve Board, 2010.
Panageas, Stavros, ‘‘Bailouts, the Incentive to Manage Risk, and Financial
Crises,’’ Journal of Financial Economics, 95, no. 3 (2010), 296–311.
Saunders, Anthony, Elizabeth Strock, and Nickolaos G. Travlos, ‘‘Ownership
Structure, Deregulation, and Bank Risk Taking,’’ Journal of Finance, 45,
no. 2 (1990), 643–654.
Strahan, Philip, and Basak Tanyeri, ‘‘Once Burned, Twice Shy: Money Market
Fund Responses to a Systemic Liquidity Shock,’’ Working Paper, Boston
College, 2012.
Wermers, Russ, ‘‘Money Fund Runs,’’ Working Paper, University of Maryland,
2011.