ANNUAL

REVIEWS Further Limits of Arbitrage

Click here for quick links to
Annual Reviews content online,
including: Denis Gromb1,2 and Dimitri Vayanos2,3,4
• Other articles in this volume 1
INSEAD, Fontainebleau, 77305, France; email: [email protected]
• Top cited articles 2
• Top downloaded articles Center for Economic and Policy Research (CEPR), London EC1V 0DG,
• Our comprehensive search United Kingdom
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

3
London School of Economics, London WC2A 2AE, United Kingdom;
email: [email protected]
by University of New Hampshire on 05/27/14. For personal use only.

4
National Bureau of Economic Research (NBER), Cambridge, Massachusetts 02138

Annu. Rev. Financ. Econ. 2010. 2:251–75 Key Words

First published online as a Review in Advance on market anomalies, liquidity, financial constraints, financial
July 26, 2010
institutions
The Annual Review of Financial Economics is
online at financial.annualreviews.org Abstract
This article’s doi: We survey theoretical developments in the literature on the limits of
10.1146/annurev-financial-073009-104107
arbitrage. This literature investigates how costs faced by arbitra-
Copyright © 2010 by Annual Reviews. geurs can prevent them from eliminating mispricings and providing
All rights reserved
liquidity to other investors. Research in this area is currently evolv-
1941-1367/10/1205-0251$20.00 ing into a broader agenda, emphasizing the role of financial insti-
tutions and agency frictions for asset prices. This research has the
potential to explain so-called market anomalies and inform welfare
and policy debates about asset markets. We begin with examples of
demand shocks that generate mispricings, arguing that they can
stem from behavioral or from institutional considerations. We next
survey, and nest within a simple model, the following costs faced
by arbitrageurs: (a) risk, both fundamental and nonfundamental;
(b) short-selling costs; (c) leverage and margin constraints; and
(d) constraints on equity capital. We finally discuss implications
for welfare and policy and suggest directions for future research.

251
 1. INTRODUCTION
Standard models of asset pricing assume a representative agent who participates in all
markets costlessly. Equilibrium prices are tied to the representative agent’s consumption,
which coincides with the aggregate consumption in the economy. The relationship between
prices and consumption is summarized in the consumption CAPM, according to which an
asset’s expected return in excess of the risk-free rate is proportional to the asset’s covari-
ance with aggregate consumption. Intuitively, assets that correlate positively with con-
sumption add to the risk borne by the representative agent and must offer high expected
return as compensation.
The relationship between risk and expected return predicted by standard models
appears at odds with a number of stylized facts commonly referred to as market anomalies.
Leading anomalies include (a) short-run momentum, the tendency of an asset’s recent
performance to continue into the near future; (b) long-run reversal, the tendency of perfor-
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

mance measured over longer horizons to revert; (c) the value effect, the tendency of an
asset’s ratio of price to accounting measures of value to predict negatively future returns;
by University of New Hampshire on 05/27/14. For personal use only.

(d) the high volatility of asset prices relative to measures of discounted future payoff
streams; and (e) postearnings-announcement drift, the tendency of stocks’ earning sur-
prises to predict positively future returns. [See, for example, Jegadeesh & Titman (1993)
for evidence on short-run momentum, DeBondt & Thaler (1985) for long-run reversal,
Fama & French (1992) for the value effect, LeRoy & Porter (1981) and Shiller (1981) for
excess volatility, and Bernard & Thomas (1989) for postearnings-announcement drift. See
also the surveys by Fama (1991) and Schwert (2003).] Reconciling these anomalies with
standard models requires explaining variation in asset risk: For example, in the case of
short-run momentum, one would have to explain why good recent performance renders an
asset riskier and more positively correlated with aggregate consumption. A recent litera-
ture pursues explanations along these lines by introducing more general utility functions
for the representative agent. Yet, reconciling standard models with all the anomalies, and
in a way consistent with their quantitative magnitude, remains elusive.
The anomalies listed above concern the predictability of asset returns on the basis of
past prices and earnings. An additional set of anomalies concerns the relative prices of
assets with closely related payoffs. For example, (a) Siamese-twin stocks, with claims to
almost identical dividend streams, can trade at significantly different prices; (b) stocks of a
parent and a subsidiary company can trade at prices under which the remainder of the
parent company’s assets has negative value; and (c) newly issued on-the-run bonds can
trade at significantly higher prices than older off-the-run bonds with almost identical
payoffs. [See, for example, Rosenthal & Young (1990) and Dabora & Froot (1999) for
evidence on Siamese-twin stocks; Mitchell et al. (2002) and Lamont & Thaler (2003) for
the relative pricing of parent and subsidiary companies; and Amihud & Mendelson (1991),
Warga (1992), and Krishnamurthy (2002) for the on-the-run phenomenon.]
Anomalies concerning relative prices have been documented for a more limited set of
assets, partly because of the scarcity of asset pairs with closely related payoffs. These
anomalies are nonetheless important because they are particularly hard to reconcile with
standard models. Indeed, although standard models may offer slightly different predictions
as to how risk and expected returns are related, they all imply the law of one price: Assets
with identical payoffs must trade at the same price. In the previous examples, however,
differences in payoffs seem insignificant relative to the observed price differences.

252 Gromb
Vayanos

 Understanding why anomalies arise and persist requires a careful study of the process of
arbitrage: Who are the arbitrageurs, what are the constraints and limitations they face, and
why can arbitrage fail to bring prices close to the fundamental values implied by standard
models? This is the focus of a recent literature on the limits of arbitrage. This review
surveys important theoretical developments in that literature, nests them within a simple
model, and suggests directions for future research.
Limits of arbitrage are usually viewed as one of two building blocks needed to explain
anomalies. The other building block is demand shocks experienced by investors other than
arbitrageurs. Anomalies are commonly interpreted as arising because demand shocks push
prices away from fundamental values, and arbitrageurs are unable to correct the discrep-
ancies. Such nonfundamental shocks to demand are often attributed to investor irrational-
ity. In this sense, research on the limits of arbitrage is part of the behavioral finance agenda
to explain market anomalies on the basis of investors’ psychological biases. [Behavioral
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

explanations for the anomalies include Barberis et al. (1998), Daniel et al. (1998), Hong &
Stein (1999), and Barberis & Shleifer (2003). See also the survey by Barberis & Thaler
by University of New Hampshire on 05/27/14. For personal use only.

(2003).]
This review departs from the conventional view in two related respects. First, it
argues that research on the limits of arbitrage is relevant not only for behavioral expla-
nations of anomalies but also for the broader study of asset pricing. Indeed, non-
fundamental demand shocks can also arise because of institutional frictions relating to
contracting and agency, as illustrated in the next section. Research on the limits of
arbitrage characterizes how nonfundamental demand shocks, whether behavioral or
not, impact prices.
In the conventional view, nonfundamental demand shocks concern investors other than
arbitrageurs and therefore can be understood independently of the limits of arbitrage. Our
second departure is to argue that many nonfundamental demand shocks can be understood
jointly with limits of arbitrage within a setting that emphasizes financial institutions and
agency. Indeed, arbitrage is often performed by specialized institutions such as hedge funds
and investment banks, and the trading strategies of these institutions are constrained by
agency frictions. Simultaneously, financial institutions and agency frictions are the source
of many nonfundamental demand shocks. In this sense, financial institutions do not neces-
sarily correct anomalies but can also cause them. Research on the limits of arbitrage is
currently evolving into a broader agenda, emphasizing the role of financial institutions and
agency frictions for asset prices. This agenda has the potential to explain market anomalies
within a fully rational framework.
The emphasis on financial institutions and agency frictions is fruitful for the analysis of
welfare and public policy. Crises, including the recent one, show that government inter-
vention can be important for the smooth functioning of financial markets. In standard
models, however, there is no scope for such intervention because the equilibrium is Pareto
optimal. Research on the limits of arbitrage has the potential to deliver a more useful
framework for designing and assessing public policy. Indeed, this research takes a two-
tiered view of financial markets: A core of sophisticated arbitrageurs trade against
mispricings, and in doing so provide liquidity to a periphery of less sophisticated investors.
Under this view, the financial health of arbitrageurs is crucial for the smooth functioning of
markets and the provision of liquidity. Understanding how financial health is affected by
arbitrageurs’ trading decisions, and whether these decisions are socially optimal, can guide
public policy.

www.annualreviews.org
Limits of Arbitrage 253

 This review proceeds as follows: Section 2 presents examples of nonfundamental
demand shocks, emphasizing that they often stem from institutional considerations. Sec-
tion 3 surveys important theoretical developments in the literature on the limits of
arbitrage, and nests them within a simple model. It emphasizes the following costs faced
by arbitrageurs: (a) risk, both fundamental and nonfundamental; (b) costs of short selling;
(c) leverage and margin constraints; and (d) constraints on equity capital. [Risk is a cost
when arbitrageurs are not fully diversified and bear a disproportionate share of the risk of
arbitrage trades. Under-diversification is related to financial constraints, as we explain in
Section 3.4. In this sense, costs (a), (c) and (d) are related.] Although these are not the
only costs faced by real-life arbitrageurs, they are among the most important and have
received significant attention in the literature. Besides examining the implications of each
type of cost for asset-price behavior, Section 3 illustrates how these costs can be inte-
grated into richer models that incorporate multiple assets and dynamics. Such models
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

have the potential to address a variety of anomalies—both those concerning violations of

the law of one price and those concerning return predictability—and are the subject of a
by University of New Hampshire on 05/27/14. For personal use only.

rapidly growing literature. Finally, Section 4 discusses implications for welfare and public
policy.

2. DEMAND SHOCKS
Models of limited arbitrage typically assume that some investors experience demand
shocks that drive prices away from fundamental values. This section presents examples of
such shocks and their price effects. Shocks in the first example are probably best
interpreted as arising from behavioral considerations, whereas in the other examples they
more likely arise from institutional frictions.

2.1. Palm-3Com
Lamont & Thaler (2003) study the sale of Palm by its parent company 3Com and the
behavior of the two companies’ stock prices around this event. On March 2, 2000, 3Com
sold 5% of its stake in Palm through an initial public offering (IPO). 3Com also announced
that it would spin off its remaining stake in Palm to 3Com shareholders before the end of
the year. Under the terms of the spin off, 3Com shareholders would receive 1.525 shares of
Palm for each share of 3Com they owned.
The law of one price implies that prior to the spin off, 3Com shares should have been
trading at a price exceeding 1.525 times the price of Palm shares. This is because one 3Com
share was equivalent to 1.525 shares of Palm plus an equity claim on 3Com’s remaining
(non-Palm) assets, and the latter claim had nonnegative value because of limited liability.
The law of one price was, however, violated for a period of approximately two months
starting from the IPO. For example, on the day of the IPO, Palm closed at $95.06 per
share, implying a lower bound of 1.525  95.06 ¼ $145 for the share price of 3Com.
3Com, however, closed at $81.81 per share, having dropped from $104.13 on the previous
day. Under these prices, the implied value of 3Com’s non-Palm assets was $22 billion,
implying an economically significant violation of the law of one price.
The mispricing between Palm and 3Com is a challenge to standard asset-pricing
models. One must explain, in particular, why some investors were willing to buy Palm
shares for $95.06 when they could acquire them at a lower cost by buying 3Com shares.

254 Gromb
Vayanos

 The most plausible explanations are based on investors’ psychological biases and cogni-
tive limitations. Investors buying Palm were possibly not sophisticated enough to appre-
ciate the opportunity of buying it through 3Com. Moreover, because Palm was a
manufacturer of a relatively new product (handheld devices), it was possibly associated
with the new economy to a larger extent than 3Com was. This might have led investors,
overly optimistic about the new economy, to be willing to pay a disproportionately high
price for Palm.

2.2. Index Effects

Index effects stem from a stock’s addition to or deletion from prominent market indices.
Starting with Harris & Gurel (1986) and Shleifer (1986), a number of papers document
that addition to the Standard and Poor (S&P)’s 500 index raises the price of a stock,
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

whereas deletion lowers its price. For example, Chen et al. (2004) find that during 1989–
2000, a stock’s price would increase by an average 5.45% on the day of the announcement
by University of New Hampshire on 05/27/14. For personal use only.

that the stock would be added to the index, and a further 3.45% from the announcement
day to the day of the actual addition. Conversely, a stock’s price would decrease by an
average 8.46% on the day of the announcement that the stock would be deleted from the
index, and a further 5.97% from the announcement day to the deletion day. These effects
are economically significant.
Standard models can account for index effects only if additions and deletions convey
information about assets’ fundamental values. Even if one is to accept, however, that S&P
has an informational advantage relative to the market, it is hard to explain why this
advantage (a) grew larger after 1989, which is when index effects became the most signif-
icant; and (b) can be large enough to account for the observed index effects. A more
plausible explanation is that index additions and deletions trigger changes in the demand
by mutual funds. Indeed, passively managed mutual funds track indices mechanically, and
actively managed funds benchmark their performance against indices. If, therefore, a stock
is added to the S&P 500 index, funds that track or are benchmarked against the index are
eager to buy the stock, and this can raise the stock’s price. The institutional explanation is
consistent with the growth of index effects in recent decades, because this parallels the
growth of institutional investing, index tracking, and benchmarking.
Boyer (2007) provides further evidence consistent with the institutional explanation. He
focuses on the BARRA value and growth indices, which consist of value and growth stocks,
respectively. Unlike the S&P 500 index, BARRA indices are constructed using publicly
disclosed rules, so additions and deletions do not signal any private information. Boyer
finds that marginal value stocks, defined as those that recently switched from the growth
into the value index, comove significantly more with the value than with the growth index,
whereas the reverse is true for marginal growth stocks. These effects are hard to explain
within standard models because marginal value stocks have very similar characteristics to
marginal growth stocks. Instead, it is plausible that the effects arise from shifts in demand
by mutual funds. For example, inflows into funds tracking the value index trigger pur-
chases of all stocks in that index, and this can raise the prices of these stocks simulta-
neously. The institutional explanation is further strengthened by Boyer’s finding that the
effects appear only after 1992, which is when the BARRA indices were introduced. [Earlier
evidence linking index membership to comovement is in Vijh (1994) and Barberis et al.
(2005), which both focus on the S&P 500 index.]

www.annualreviews.org
Limits of Arbitrage 255

 2.3. Fire Sales by Mutual Funds
Mutual funds respond to outflows by selling stocks in their portfolios. Coval & Stafford
(2007) study the behavior of stock prices around sales driven by large outflows. They
define fire sales as the sales by the 10% of funds that experience the largest outflows
within a given quarter. For each stock, they compute the fraction of average volume
generated by fire sales, and they focus on the 10% of stocks for which this fraction is
largest. These stocks exhibit a V-shaped price pattern. During the fire-sale quarter and
the quarter immediately preceding, their average cumulative abnormal return is 7.9%.
This price decline is followed by a recovery: During the year following the fire-sale quarter,
the average cumulative abnormal return is 6.1%. This return rises to 9.7% during the
18 months following the fire-sale quarter. [Anton & Polk (2008), Jotikasthira et al. (2009),
Greenwood & Thesmar (2009), and Lou (2009) provide related evidence suggesting that
fund outflows have large price effects.]
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

The slow and predictable price recovery represents a challenge to standard models.
Indeed, these models can account for an increase in expected return only through an
by University of New Hampshire on 05/27/14. For personal use only.

increase in the covariance with aggregate consumption. Explaining why this covariance
increases for stocks sold by distressed mutual funds, and why such an increase can account
for an annual abnormal return of 6%, is difficult. A more plausible explanation is that
sales by distressed mutual funds generate price pressure, pushing prices below fundamental
values and raising expected returns going forward. This explanation is related to institu-
tional frictions: Distressed sales are likely to be triggered by investors who lose confidence
in the quality of managers running underperforming funds.

2.4. U.K. Pension Reform and the Term Structure

Demand shocks are likely to also affect prices outside the United States, and for assets
other than stocks. This is illustrated by our last example, which concerns the impact of the
U.K. pension reform on the term structure of interest rates, an episode described in greater
detail in Greenwood & Vayanos (2010a). A major objective of the U.K. pension reform
over the past 20 years has been to ensure the transparency and solvency of pension funds;
indeed, the reform was motivated partly by the failure of the Maxwell pension fund in the
early 1990s. The reform stipulated that pension funds had to meet a minimum ratio of
assets to liabilities. On the one hand, pension-fund assets, such as stocks and bonds, are
publicly traded and can be valued using market prices. On the other hand, pension-fund
liabilities are not traded, and their valuation requires a suitable discount rate. Under the
Pensions Act of 2004, this discount rate had to be the yield on long-term inflation-linked
government bonds, on the grounds that pension liabilities are long term and indexed to
inflation.
Pension funds responded to the reform by buying large amounts of long-term bonds.
Indeed, because long-term bonds were providing the discount rate to calculate the value of
pension liabilities, they were also the best hedge for these liabilities. Pension-fund pur-
chases had a significant impact on the term structure, especially at the long end. For
example, in late 2003, the inflation-indexed bonds maturing in 2016 and 2035 had
approximately the same yield. During 2004 and 2005, however, the yield of the 2035 bond
fell relative to that of the 2016 bond, with the spread reaching an all-time low of 0.49%
in January 2006. At that time, the 2035 and 2055 bonds had yields of 0.72% and 0.48%,

256 Gromb
Vayanos

 respectively, which are extremely low relative to the historical average of 3% of long real
rates in the United Kingdom. In 2005, in accordance with the generally held view that long
yields had decreased because of demand by pension funds, the government agreed to issue
bonds with maturities of up to 50 years, while also shifting the overall mix of maturities
toward the long term.
The inversion at the long end of the U.K. term structure is hard to rationalize within
standard representative-agent models. Indeed, in these models the interest rate for
maturity T is determined by the willingness of the representative agent to substitute
consumption between times 0 and T. Therefore, these models would attribute the drop
in the 30-year interest rate to the unlikely scenario that the pension reform signaled a
drop in aggregate consumption 30 years into the future. In a similar spirit, the expec-
tations hypothesis of the term structure would attribute the drop in the 2035–2016
yield spread to expectations about short-term interest rates past 2016 decreasing
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

sharply during 2004 and 2005. A more plausible explanation is that the reform trig-
gered high demand for long-term bonds by pension funds, and that this generated price
by University of New Hampshire on 05/27/14. For personal use only.

pressure.

2.5. Summary
The examples in this section describe a variety of demand shocks that had significant and
long-term price effects. A natural question is why arbitrageurs are unable to absorb such
shocks and bring prices back to fundamental values. For example, why don’t arbitrageurs
eliminate the abnormally high expected returns following fire sales by buying the stocks in
question? And why were arbitrageurs unable to eliminate the Palm-3Com mispricing by
shorting Palm and buying 3Com? Using a simple model, we next examine the constraints
faced by arbitrageurs and the implications for asset prices.

3. A SIMPLE MODEL

3.1. Cross-Asset Arbitrage and Intertemporal Arbitrage

We consider an economy in which assets are traded in Period 1 and pay off in Period 2.
There are two risky assets, A and B, paying off dA and dB, respectively. We denote by di
and si, respectively, the mean and standard deviation of di, i ¼ A, B; and by r the
correlation between dA and dB. For tractability, we assume that dA and dB are jointly
normal. We examine how shocks to the demand for asset A affect that asset’s equilibrium
price pA in Period 1. For simplicity, we take as exogenous both the price pB of asset B in
Period 1 and the riskless rate. We set the former equal to asset B’s expected payoff dB and
the latter to zero.
There are two types of agents: outside investors and arbitrageurs. Outside investors’
demand for asset A is inelastic and equal to u shares (with u positive or negative). We refer
to u as the demand shock: It is a constant for now but stochastic in Section 3.2. Arbitra-
geurs are competitive, risk averse, and maximize expected utility of wealth W2 in Period 2.
For tractability, we assume that utility is exponential. By possibly reinterpreting the
demand shock u as net demand, we normalize the supply of asset A to zero. This normal-
ization ensures that absent a demand shock (u ¼ 0), the price of asset A equals the asset’s
expected payoff dA .

www.annualreviews.org
Limits of Arbitrage 257

 A demand shock u 6¼ 0 can push the price of asset A away from the expected payoff.
Arbitrageurs trade to profit from this discrepancy. Doing so, they also provide liquidity to
outside investors. Suppose, for example, that u is positive, in which case outside investors
wish to buy asset A. Arbitrageurs provide liquidity because they take the opposite side of
this transaction, shorting the asset and limiting the price rise. Liquidity is high when the
demand shock’s price impact is small: If, for example, u is positive, high liquidity means
that the price rise is small.
The model has two interpretations, capturing different but closely related real-life
arbitrage situations. In the cross-asset arbitrage interpretation, assets A and B are different,
and arbitrageurs use asset B to hedge their position in asset A. In the intertemporal
arbitrage interpretation, arbitrageurs exploit discrepancies between the prices of the same
asset at different points in time. In that interpretation, the arbitrageurs’ positions in assets
A and B represent trades that arbitrageurs execute in the same asset in different periods.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

The two interpretations yield the same basic insight: The price effects of demand shocks
depend on the risk aversion of arbitrageurs and on the risk they cannot hedge away.
by University of New Hampshire on 05/27/14. For personal use only.

3.1.1. Cross-asset arbitrage. In Period 1, the arbitrageurs choose positions xA and xB in

assets A and B to maximize expected utility
E1 ½expðaW2 Þ ð1Þ
subject to the budget constraint
W2 ¼ W1 þ xA ðdA  pA Þ þ xB ðdB  pB Þ, ð2Þ
where a denotes the risk aversion of arbitrageurs and W1 denotes their wealth in Period 1.1
Substituting Equation 2 into Equation 1, and using normality and the assumption that
pB ¼ dB , we find that arbitrageurs maximize the mean-variance objective
a
xA ðdA  pA Þ  ðx2A s2A þ x2B s2B þ 2xA xB rsA sB Þ: ð3Þ
2
The optimal demand for asset B is xB ¼ ðrsA =sB ÞxA , i.e., arbitrageurs choose a position
in asset B to hedge that in asset A. Substituting the optimal xB into Equation 3, we can
write the arbitrageurs’ objective in a way involving only xA:
a
xA ðdA  pA Þ  x2A s2A ð1  r2 Þ: ð4Þ
2
The optimal demand for asset A is

dA  pA
xA ¼ : ð5Þ
as2A ð1  r2 Þ

Asset A being in zero supply, market clearing requires

xA þ u ¼ 0: ð6Þ

Combining Equations 5 and 6, we find the equilibrium price of asset A:

1
Our analysis of cross-asset arbitrage is related to Wurgler & Zhuravskaya (2002). As in our model, they take the
price of asset B to be exogenous and equal to the asset’s expected payoff. Unlike in our model, however, they restrict
arbitrageurs’ aggregate dollar investment in assets A and B to be zero.

258 Gromb
Vayanos

 pA ¼ dA þ as2A ð1  r2 Þu: ð7Þ

Equation 7 highlights a number of properties. First, larger demand shocks have a larger
price impact (@pA =@u > 0) because arbitrageurs must be compensated for bearing more
risk. Second, a given demand shock u has a larger price impact when arbitrageurs are more
risk averse (large a) and asset A has a more uncertain payoff (large sA), because in both
cases arbitrageurs require more compensation for bearing its risk. The shock’s impact is
also larger when the payoff of asset A is less correlated with that of asset B (small jrj)
because arbitrageurs are less able to hedge their position in asset A using asset B. Thus,
assets with higher idiosyncratic risk and fewer substitutes are more sensitive to demand
shocks. In the special case where assets A and B have identical payoffs (dA ¼ dB, i.e., r¼1),
the demand shock has no effect, and the two assets trade at the same price. Arbitrageurs
are able to align the price of asset A fully with that of asset B because they bear no risk in
exploiting price discrepancies between the two assets.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org
by University of New Hampshire on 05/27/14. For personal use only.

3.1.2. Intertemporal arbitrage. In cross-asset arbitrage, arbitrageurs exploit discrepancies

between the prices of different assets at a given point in time. We next consider
intertemporal arbitrage, in which arbitrageurs exploit discrepancies between the prices of
the same asset at different points in time. To interpret our model as one of intertemporal
arbitrage, we split Period 1 into subperiods 1A and 1B and assume that positions xA and xB
in assets A and B represent trades in the same asset in subperiods 1A and 1B, respectively.
The model so derived is a simplified version of Grossman & Miller (1988).
We denote by d the asset’s payoff in Period 2; by dA and dB the expectations of d as
of subperiods 1A and 1B, respectively; and by sA and sB the respective standard de-
viations of d. The expectation dB is random as of subperiod 1A if new information
arrives between subperiods 1A and 1B, and has variance s2A  s2B . (This follows from
Varðdj1AÞ ¼ E½Varðdj1BÞj1A þ Var½Eðdj1BÞj1A.)
Given that in subperiod 1B arbitrageurs can trade the asset at a price equal to its
expected payoff, they are not compensated for bearing risk. Therefore, their position in
subperiod 1B is zero, which means that their trade xB in subperiod 1B is the opposite of
their trade xA in subperiod 1A. Using dA ¼ dB ¼ d, pB ¼ dB and xA ¼ xB , we can write
Equation 2 as

W2 ¼ W1 þ xA ðdB  pA Þ: ð8Þ

Substituting Equation 8 into Equation 1, we find that arbitrageurs choose their position xA
in subperiod 1A to maximize the mean-variance objective
a
xA ðd  pA Þ  x2A ðs2A  s2B Þ: ð9Þ
2
This objective is identical to Equation 4, derived under cross-sectional arbitrage, provided
that we set r  sB =sA < 1. The parameter r, equal to the correlation between assets A and
B under cross-sectional arbitrage, reflects the informational similarity between subperiods
1A and 1B under intertemporal arbitrage: When, for example, r ¼ 1, the information
available in the two subperiods is identical. Maximizing Equation 9 yields the demand for
the asset (Equation 5), and imposing market clearing yields the asset’s equilibrium price
(Equation 7). As in the case of cross-sectional arbitrage, the demand shock has a larger
price impact when arbitrageurs are more risk averse (large a) and bear more risk. In the

www.annualreviews.org
Limits of Arbitrage 259

 case of intertemporal arbitrage, the relevant risk is that between subperiods 1A and 1B
and is measured by s2A  s2B ¼ s2A ð1  r2 Þ. This is because arbitrageurs offset their risky
position in subperiod 1A by an opposite position in subperiod 1B. In the extreme case
where subperiods 1A and 1B coincide, r ¼ 1, and the demand shock has no effect. As
Grossman & Miller (1988) emphasize, the time between subperiods 1A and 1B can be
interpreted as the time it takes for enough risk-bearing capacity to arrive in the market and
fully eliminate the effect of the demand shock.2
The two interpretations of our model correspond to different real-life arbitrage situa-
tions and, accordingly, to different strands of empirical studies. From an asset-pricing-
theory viewpoint, however, they are isomorphic. Henceforth, we focus on the cross-asset
arbitrage interpretation of the model and gradually enrich the basic model to illustrate
developments of the limits-of-arbitrage literature. The results we derive carry through to
the intertemporal arbitrage interpretation.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org
by University of New Hampshire on 05/27/14. For personal use only.

3.2. Nonfundamental Risk

In our baseline model, the risk borne by arbitrageurs is fundamental risk, i.e., the risk that
the two legs of the arbitrage do not yield the same payoff. An additional type of risk stems
from demand shocks when these affect prices. DeLong et al. (1990, henceforth referred to
as DSSW) label this type of risk noise-trader risk. We use instead the term nonfundamental
risk to emphasize that although demand shocks may be unrelated to asset payoffs, they can
arise from rational behavior, as Section 2 illustrates. To introduce nonfundamental risk in
our model, we assume that the variables dA , dB , and u, which are constant as of Period 1,
are random as of an initial Period 0. Fundamental risk arises in Period 0 because prices in
Period 1 depend on the realizations of dA and dB . Nonfundamental risk arises because asset
A’s price in Period 1 depends on the realization of u. We assume that fundamental risk in
Period 0 is the same as in Period 1, i.e., the standard deviation of di , i ¼ A, B, is si, and the
correlation between dA and dB is r. We denote the standard deviation of u by su, and
assume u to be independent of dA and dB . We maintain the notation pi for the price of
asset i in Period 1.
Equation 7 implies that as of Period 0, the standard deviation of asset A’s price in
Period 1 is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sA 1 þ a2 s2A ð1  r2 Þ2 s2u : ð10Þ

Equation 10 shows that nonfundamental risk su increases asset A’s volatility. The effect is
through the second term inside the square root, which is the ratio of nonfundamental
variance, a2 s4A ð1  r2 Þ2 s2u , to fundamental variance, s2A . This ratio is larger when the
demand shock is less predictable (large su) and when a given demand shock u has a larger
price impact in Period 1. Consistent with Equation 7, the price impact is larger when
arbitrageurs are more risk averse (large a), and the payoff of asset A is more uncertain
(large sA) and less correlated with the payoff of asset B (small jrj).

2
Recent literature derives the slow arrival of new investors from search costs. See, for example, Weill (2007), Duffie
et al. (2008), Duffie & Strulovici (2009), and Lagos et al. (2009). See also He & Xiong (2008), who derive capital
immobility and segmentation from agency frictions: Preventing traders from moving across assets can provide their
employer with a better signal of their effort.

260 Gromb
Vayanos

 Equation 7 and pB ¼ dB imply that as of Period 0, the correlation between the prices of
assets A and B in Period 1 is
r
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð11Þ
1 þ a2 s2A ð1  r2 Þ2 s2u

Equation 11 shows that nonfundamental risk lowers the correlation between assets
A and B. This is because it increases the volatility of asset A without affecting asset B.
Suppose next that there is a demand shock in Period 0, and an arbitrageur responds to
the shock by taking a position in assets A and B. Because nonfundamental risk increases
asset A’s volatility and lowers its correlation with asset B, it increases the volatility of the
arbitrageur’s return in Period 1. This volatility matters when the arbitrageur has a short
horizon and must close his position in Period 1. Price volatility caused by demand shocks in
Period 1 deters the arbitrageur from absorbing demand shocks in Period 0. DSSW build on
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

this idea to show that nonfundamental risk can be self-fulfilling. They assume a discrete-
time infinite-horizon economy with an exogenous riskless rate r and an asset paying a
by University of New Hampshire on 05/27/14. For personal use only.

constant dividend r in each period. Because one share of the asset yields the same payoff
as an investment of one dollar in riskless rate, the law of one price implies that the asset’s
price should be one. DSSW show, however, that when arbitrageurs have a one-period
horizon, an equilibrium exists in which this price is stochastic. Intuitively, if arbitrageurs
expect the price to be stochastic, demand shocks have an effect, and this renders the price
stochastic.
The stochastic equilibrium of DSSW hinges on a number of assumptions. One is that
arbitrageurs have short horizons: If they were infinitely lived, they would enforce the law
of one price through buy-and-hold strategies. Short horizons can be viewed as a reduced
form for financial constraints, as we show in Sections 3.4 and 3.5. A second critical
assumption is that of an infinite horizon: With a finite horizon, the law of one price would
hold. (Our model confirms this: If assets A and B have the same payoff, the correlation r is
one, and the nonfundamental risk in Equations 10 and 11 disappears.) Loewenstein &
Willard (2006) show that even under these two assumptions, the law of one price would
hold in DSSW if interest rates were endogenized, prices were precluded from becoming
negative, or borrowing limits were imposed. But although DSSW’s result on the failure of
the law of one price may not be robust, their broader point that nonfundamental risk is an
impediment to arbitrage remains important.3

3.3. Short-Selling Costs

In our analysis so far, the only cost arbitrageurs face is risk. Additional costs, however, stem
from the way arbitrageur positions are established and financed. Arbitrageurs often estab-
lish their positions in the repo market. For example, an arbitrageur wishing to establish a
long position in an asset can borrow some of the needed cash by posting the asset as
collateral—a repo transaction. Conversely, an arbitrageur wishing to establish a short

3
Related to nonfundamental risk, which stems from demand shocks, is a risk stemming from the actions of other
arbitrageurs. Abreu & Brunnermeier (2002, 2003) assume that arbitrageurs learn about a profitable investment
opportunity privately at different points in time, so the opportunity’s existence is not common knowledge. Arbitra-
geurs find it profitable to trade on the opportunity if they expect other arbitrageurs to trade on it in the near future.
Lack of common knowledge, however, prevents such coordination, and can cause the opportunity to persist even after
all arbitrageurs have become aware of its existence. For coordination frictions in arbitrage, see also Zigrand (2005).

www.annualreviews.org
Limits of Arbitrage 261

 position in an asset can borrow (to subsequently sell) the asset by posting cash as
collateral—a reverse repo transaction. The interest rate earned on the cash, known as the
repo rate, can differ across assets, and this can be a source of arbitrage cost. For example,
shorting an asset that carries a low repo rate relative to other assets is costly because the
cash collateral posted by the short seller earns a below-market interest rate.
Costs involved in establishing and financing positions are often referred to as holding
costs. Tuckman & Vila (1992, 1993) introduce exogenous holding costs and show that
they prevent arbitrageurs from eliminating mispricings. Arbitrageurs only trade against
mispricings that are large enough to compensate them for the holding costs they incur.
Dow & Gorton (1994) show that holding costs can have disproportionately large effects
when they are incurred by a sequence of short-horizon arbitrageurs.
Short-selling costs are holding costs associated with short positions. We introduce short-
selling costs in our model by assuming that shorting asset A involves a cost c per share.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Arbitrageurs maximize the objective

a
xA ðdA  pA Þ  ðx2A s2A þ x2B s2B þ 2xA xB rsA sB Þ  cjxA j1fxA <0g , ð12Þ
by University of New Hampshire on 05/27/14. For personal use only.

2
where 1S is equal to one if condition S is satisfied, and zero otherwise. Equation 12
is derived from Equation 3 by subtracting the cost cjxA j of holding a short position xA in
asset A. Solving for the optimal demand for asset A and combining with the market-
clearing condition (Equation 6), we find that the equilibrium price of asset A is given by

pA ¼ dA þ as2A ð1  r2 Þu if u  0, ð13aÞ

pA ¼ dA þ as2A ð1  r2 Þu þ c if u > 0: ð13bÞ

Short-selling costs affect the price only when the demand shock u is positive, because this is
when arbitrageurs hold a short position. The price increases by an amount equal to the
short-selling cost c, so that arbitrageurs are compensated for incurring c.
Short-selling costs have an effect even when the assets have identical payoffs, and, in
that case, they cause the law of one price to be violated. Setting r ¼ 1 in Equation 13b, we
find pA ¼ d þ c, where d denotes the expected payoff of the two assets (d  dA ¼ dB ).
Therefore, when the demand shock u is positive, the price of asset A exceeds that of the
identical-payoff asset B by an amount equal to the short-selling cost c. This analysis fits
well with the Palm-3Com example: c may be the cost of shorting Palm, and u may be the
demand by investors eager to hold Palm over 3Com. D’Avolio (2002) and Lamont &
Thaler (2003) report that c was large and prevented arbitrageurs from exploiting the
mispricing.
In our model, the short-selling cost c is an exogenous deadweight loss. A number of
papers seek to endogenize c on the basis of frictions in the repo market. In Duffie (1996),
short sellers have the choice between two assets with identical payoffs but different exog-
enous transaction costs. They prefer to short the low transaction cost asset, and their
demand to borrow that asset in the repo market lowers the repo rate, driving up the
short-selling cost. The friction in the repo market is that asset owners must incur an
exogenous transaction cost to lend their asset. Unlike in our model, the short-selling cost c
is not a deadweight loss, but accrues to asset owners. Therefore, it is an additional payoff
earned from holding the asset, and it increases the asset price—even in the absence of any
positive demand shock u. Krishnamurthy (2002) uses a similar model to show that a price

262 Gromb
Vayanos

 premium arising from short-selling costs can coexist with a premium arising from an asset’s
superior liquidity.
Duffie et al. (2002) model the repo market as a search market in which asset borrowers
and lenders negotiate bilaterally. The search friction generates a short-selling cost, which
increases in the demand for short selling. Vayanos & Weill (2008) show that the combina-
tion of a search spot market and a search repo market yields violations of the law of one
price—even in the absence of any exogenous differences in transaction costs. Short sellers
concentrate on the more liquid asset, and their activity is what renders the asset more
liquid. The more liquid asset carries a price premium because of both its superior liquidity
and the fact that demand for short selling drives up short-selling costs.
In the extreme case where short-selling costs are infinite, they amount to short-sale
constraints, whose implications for asset prices are examined in a number of papers. Miller
(1977) shows that when short sales are not allowed, pessimistic investors are unable to
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

trade and prices reflect the valuation of the most optimistic investors. Harrison & Kreps
(1978) show that with multiple trading periods, prices even exceed the valuation of inves-
by University of New Hampshire on 05/27/14. For personal use only.

tors who are currently the most optimistic. Indeed, these investors have the option to resell
the asset should other investors become more optimistic in future periods. Scheinkman &
Xiong (2003) value the resale option in a continuous-time model in which differences in
beliefs stem from overconfidence. Hong et al. (2006) show that overpricing and the value
of the resale option are highest for assets with low float. Diamond & Verrecchia (1987)
show that short-sale constraints do not cause overpricing when differences in beliefs stem
from private signals rather than an agreement to disagree. Indeed, the market adjusts
rationally for the fact that investors with negative private signals are unable to trade. Bai
et al. (2006) show that short-sale constraints can cause underpricing because they generate
uncertainty about the extent of negative private information. Hong & Stein (2003) show
that the occasional release of negative private information can be the source of market
crashes.

3.4. Leverage Constraints

In our analysis so far, there is no explicit role for arbitrageur capital: Although portfolio
decisions and asset prices depend on arbitrageur risk aversion, they are independent of
arbitrageur wealth. Wealth does not matter because of the simplifying assumption that
arbitrageurs have exponential utility, i.e., their coefficient of absolute risk aversion is
independent of wealth. Yet, the capital available to real-life arbitrageurs appears to be an
important determinant of their ability to eliminate mispricings and provide liquidity to
other investors.
The study of arbitrageur capital is related to that of financial constraints. Indeed, an
important theme in the literature on the limits of arbitrage is that arbitrageurs are sophis-
ticated traders, better able to identify mispricings than other, less sophisticated investors.
Because capital in the hands of arbitrageurs can earn higher return, other investors can gain
by investing their capital with arbitrageurs. If, however, arbitrageurs could access external
capital frictionlessly, they would be able to eliminate mispricings, and asset prices and
allocations would be as in standard models. Thus, although arbitrageurs can access exter-
nal capital, this access is limited by financial constraints. In limiting access to external
capital, financial constraints cause arbitrageurs to be under-diversified and bear a dispro-
portionate share of the risk of arbitrage trades.

www.annualreviews.org
Limits of Arbitrage 263

 In this section, we study constraints on arbitrageurs’ ability to lever up by raising
margin debt. We assume that assets A and B have identical payoffs. This isolates the effects
of leverage constraints from those of risk: Arbitrageurs bear no risk when exploiting a price
discrepancy between assets A and B, and such a discrepancy can arise solely because of
leverage constraints. To model leverage constraints, we focus on the mechanics of collat-
eral in the repo market, following Geanakoplos (1997, 2003). We also assume that arbi-
trageurs must collateralize their positions in each asset separately. The model so derived is
a simplified version of Gromb & Vayanos (2002, 2010).
Consider an arbitrageur wishing to establish a long position xi in asset i. The arbitra-
geur can borrow some of the needed cash by posting asset i as collateral. The borrowed
cash is typically less than the market value xipi of the asset collateral; otherwise, a drop in
the asset price would cause the value of the collateral to drop below that of the loan. To
determine the size of the loan, we assume, for simplicity, that margin loans have to be
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

riskless and that competitive lenders set the rate equal to the riskless rate (which is zero).
Because riskless loans are not feasible with normally distributed asset payoffs, we assume
instead that d is distributed symmetrically over the bounded support ½d  E, d þ E, where
by University of New Hampshire on 05/27/14. For personal use only.

d  E. An arbitrageur wishing to establish a long position xi in asset i can thus borrow a

maximum of xi ðd  EÞ and must pay the remainder

xi ðpi  d þ EÞ  xi mþ
i ð14Þ

out of his wealth. The parameter mþ i is the margin (or haircut) for a long position in asset i.
(An alternative to requiring margin loans to be riskless is to impose an upper bound on the
default probability. This would yield a constraint of the same form as Equation 14, which
would be interpreted as a value-at-risk constraint.)
Consider next an arbitrageur wishing to establish a short position xi in asset i. The
arbitrageur can borrow (and subsequently sell) asset i, posting cash as collateral. The cash
collateral typically exceeds the proceeds jxi jpi from the sale of the borrowed asset; other-
wise, an increase in the asset price would cause the value of the loan to rise above that of
the collateral. As for long positions, we assume that margin loans have to be riskless and
that competitive asset lenders pay the riskless rate on the cash collateral. (This assumption
eliminates the short-selling costs of Section 3.3.) An arbitrageur wishing to establish a short
position xi in asset i must post jxi jðd þ EÞ units of cash as collateral. Selling the asset yields
jxi jpi units, and the remainder

jxi jðd þ E  pi Þ  jxi jm

i ð15Þ

must be drawn from the arbitrageur’s wealth. The parameter m i is the margin (or haircut)
for a short position in asset i. The margin requirements mþ i and mi are increasing in the
parameter E, which is a measure of the volatility of asset i. Indeed, volatility increases the
maximum loss that a long or short position can experience.
From Equations 14 and 15, the positions xA and xB of an arbitrageur with wealth W1
must satisfy the following leverage constraint:
X
W1  jxi jðmþ 
i 1fxi >0g þ mi 1fxi <0g Þ: ð16Þ
i¼A,B

Arbitrageurs maximize expected utility (Equation 1) subject to their budget constraint

(Equation 2) and leverage constraint (Equation 16).

264 Gromb
Vayanos

 When can arbitrageurs eliminate price discrepancies between assets A and B, thus
enforcing the law of one price? The market-clearing condition (Equation 6) requires that
arbitrageurs absorb the demand shock u, i.e., take a position xA ¼u. If, in addition, assets
 then it is optimal for arbitrageurs
A and B trade at a price equal to their expected payoff d,
not to bear risk and hold an offsetting position xB ¼ u in asset B. Given that for
pA ¼ pB ¼ d the margins mþ 
i and mi are equal to E, the leverage constraint (Equation 16)
is satisfied if
W1  2jujE: ð17Þ

Arbitrageurs can enforce the law of one price if their wealth W1 is large relative to the
demand shock u and the margin requirement E. In that case, the demand shock has no
effect on the price of asset A and the market is perfectly liquid. Intuitively, arbitrageurs can
provide perfect liquidity when their wealth is large because the leverage constraint is not
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

binding. When, however, arbitrageur wealth is small, the leverage constraint is binding and
liquidity is imperfect.
by University of New Hampshire on 05/27/14. For personal use only.

Leverage constraints can give rise to amplification, whereby the effects of an exogenous
shock are amplified through changes in arbitrageur positions. Amplification can be derived
in our model by assuming that arbitrageurs enter Period 1 with a position from Period 0,
and that outside investors’ demand is elastic. If, for example, arbitrageurs enter Period 1
with a long position, then a negative demand shock u lowers the price pA, thus lowering
arbitrageur wealth W1 and tightening the leverage constraint (Equation 16). This can force
arbitrageurs to liquidate positions, amplifying the price drop. Because arbitrageurs drive
the price further away from fundamental value, their activity is price-destabilizing rather
than stabilizing. Moreover, because they liquidate positions following a negative demand
shock, they consume rather than provide liquidity. The liquidity providers are the outside
investors, and their demand must be assumed elastic so that they are willing to buy from
arbitrageurs.
Leverage constraints and amplification have been studied in macroeconomic settings,
starting with Bernanke & Gertler (1989) and Kiyotaki & Moore (1997). In these papers,
an adverse shock to economic activity depresses collateral values, and this can amplify the
drop in activity. Hart & Moore (1994, 1995) show that uncertainty about assets’ liquida-
tion values impairs agents’ ability to borrow—a close parallel to the result that the margins
in Equations 14 and 15 increase with volatility. Shleifer & Vishny (1992) endogenize
liquidation values and the ability to borrow in market equilibrium. Geanakoplos (1997,
2003) defines the concept of collateral equilibrium, in which margin contracts are endog-
enous, and shows that amplification can arise because margins increase following adverse
shocks.
In a financial market context, Gromb & Vayanos (2002) study how leverage constraints
affect the ability of arbitrageurs to eliminate mispricings and provide liquidity to outside
investors. They show, within a dynamic setting, that liquidity increases in arbitrageur
capital and that arbitrageurs can amplify exogenous shocks. [Weill (2007) derives a link
between arbitrageur capital and liquidity provision in a search model. See also Duffie &
Strulovici (2009), who use a search model to study the gradual flow of arbitrage capital
across trading opportunities.] Brunnermeier & Pedersen (2009) and Gromb & Vayanos
(2009a) extend this analysis to multiple assets in a static and a dynamic setting, respectively.
With multiple assets, leverage constraints generate not only amplification, but also conta-
gion, whereby shocks to one asset are transmitted to otherwise unrelated assets through

www.annualreviews.org
Limits of Arbitrage 265

 changes in arbitrageur positions. Pavlova & Rigobon (2008) derive a contagion result in an
international-economy model with portfolio constraints, of which leverage constraints are
a special case. Kondor (2009) shows that amplification can arise even in the absence of
exogenous shocks, purely as a consequence of arbitrage activity. Indeed, if a price discrep-
ancy between two assets were to remain constant or decrease over time, arbitrageurs would
exploit it and reduce it to a level from which it could increase. Garleanu & Pedersen (2009)
show that all else equal, assets with lower margin requirements can trade at higher prices.
Related results are derived in Cuoco (1997), Basak & Croitoru (2000, 2006), and
Geanakoplos (2003). Other dynamic models with leverage constraints include Aiyagari &
Gertler (1999), Allen & Gale (2000), Anshuman & Viswanathan (2005), Fostel &
Geanakoplos (2008), Adrian et al. (2009), Chabakauri (2009), Danielsson et al. (2009),
and Rytchkov (2009).4
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

3.5. Constraints on Equity Capital

by University of New Hampshire on 05/27/14. For personal use only.

In addition to constraints on margin debt, arbitrageurs often face constraints in raising

equity. For example, the equity of a mutual fund is determined by the flow of investors into
the fund and could be lower than the level at which the fund’s profitable investment
opportunities are fully exploited. Moreover, poor returns by the fund could trigger out-
flows by investors and so render the fund more constrained. Shleifer & Vishny (1997)
study the implications of constraints on equity capital for arbitrageurs’ ability to exploit
mispricings. They show that constraints give rise to amplification, via a mechanism akin
to that for leverage constraints. Suppose that arbitrageurs hold long positions in an asset.
A negative shock to the asset lowers its price and triggers outflows by investors in arbitrage
funds. As a consequence, arbitrageurs are forced to reduce their positions, amplifying
the price drop. These results can be derived in our model by assuming that (a) part of
arbitrageur wealth belongs to other investors, and is withdrawn following poor returns in
Period 1; (b) arbitrageurs cannot borrow; (c) arbitrageurs enter Period 1 with a position
from Period 0; and (d) outside investors’ demand is elastic.
Shleifer & Vishny (1997) show additionally that constraints have an effect not only
when they are binding, but also because of the possibility that they might bind in the
future. Indeed, suppose that arbitrageurs with ample capital in Period 0 identify an
underpriced asset. Investing heavily in that asset exposes them to the risk of large outflows
by investors were the underpricing to worsen in Period 1. This would deprive arbitrageurs
of capital when they need it the most, because this is when the underpricing is the
most extreme. As a consequence, arbitrageurs could refrain from investing heavily in
the underpriced asset in Period 0, keeping instead some capital in cash. Arbitrageurs’
strategy of seeking to match capital with profitable investment opportunities amounts
to risk management, as in Froot et al. (1993). Risk management by arbitrageurs is
also derived in models with leverage constraints: for example, Gromb & Vayanos (2002),

4
See also Yuan (2005) for a model in which leverage constraints hamper the revelation of private information;
and Grossman & Vila (1992), Liu & Longstaff (2004), Jurek & Yang (2007), and Milbradt (2009) for partial-
equilibrium models of portfolio choice by leverage-constrained investors. Amplification and contagion can also be
derived in models without explicit leverage constraints but where arbitrageur risk aversion depends on wealth. This
is done by Kyle & Xiong (2001) and Xiong (2001), who endow arbitrageurs with logarithmic utility, under which
the coefficient of absolute risk aversion decreases in wealth. Following adverse shocks, arbitrageurs reduce their
positions because they become more risk averse and not because they hit leverage constraints.

266 Gromb
Vayanos

 Liu & Longstaff (2004), Brunnermeier & Pedersen (2009), and Kondor (2009), and is
further emphasized in Acharya et al. (2009) and Bolton et al. (2009). Holmstrom & Tirole
(2001) explore the implications of risk management by firms for the cross-sectional pricing
of assets. They show that assets paying off in states where firms’ financial constraints bind
are more expensive than assets paying off in other states, because they provide capital
when it is needed the most.
A number of papers integrate constraints on the equity capital available to arbitrage
funds—and especially its dependence on past performance—into dynamic settings. In
Vayanos (2004), fund managers face the constraint that their fund will be liquidated
following poor performance, and this makes them unwilling to hold illiquid assets at times
of high volatility. In He & Krishnamurthy (2008, 2009), the capital available to fund
managers cannot exceed a fixed multiple of their personal wealth—a constraint derived
from optimal contracting under moral hazard. When managers underperform, the con-
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

straint becomes binding and causes volatility and risk premia to increase. In Vayanos &
Woolley (2008), investors withdraw capital following underperformance by fund man-
by University of New Hampshire on 05/27/14. For personal use only.

agers because they infer rationally that managers might be inefficient. If, in addition,
withdrawals are constrained to be gradual, they generate short-run momentum and long-
run reversal in asset returns. Dasgupta et al. (2008) derive short-run momentum and long-
run reversal from herding by fund managers, caused by reputation concerns. Guerrieri &
Kondor (2009) show that reputation concerns generate amplification effects. [Reputation
concerns are also explored in Dasgupta & Prat (2008) and Malliaris & Yan (2009). Other
papers studying general equilibrium implications of delegated portfolio management
include Cuoco & Kaniel (2009), Petajisto (2009), Basak & Pavlova (2010), and Kaniel &
Kondor (2010).]

3.6. Summary and Next Steps

Sections 3.1–3.5 show how an array of costs faced by arbitrageurs limit their ability to
eliminate mispricings and provide liquidity to outside investors. This section sketches what
we view as two important next steps in this research agenda: (a) derive the financial
constraints of arbitrageurs within an optimal contracting framework, and (b) develop
richer limits-of-arbitrage models that incorporate multiple assets and dynamics and that
can be used to address empirical puzzles.
As Sections 3.4 and 3.5 emphasize, arbitrageurs face financial constraints stemming
from agency problems with their providers of capital. Most of the papers referenced in
these sections do not fully endogenize the constraints. For example, papers on leverage
constraints typically rule out equity, and papers on equity constraints typically rule out
debt. Deriving financial constraints within an optimal contracting framework is an impor-
tant next step. Indeed, although the constraints studied in the literature take a variety of
forms, they tend to generate some similar results: for example, amplification, contagion,
and risk management by arbitrageurs. Endogenizing the constraints would help identify
whether the common results are driven by a single underlying friction, or whether the
constraints are fundamentally different. Identifying the frictions that underlie the con-
straints would also be useful for policy analysis, as it would clarify what a regulator can
and cannot do to alleviate the effects of the constraints.
Starting with Kehoe & Levine (1993), a macroeconomic literature explores the asset
pricing implications of financial constraints when these derive from limited commitment

www.annualreviews.org
Limits of Arbitrage 267

 by borrowers. Limited commitment is also the source of the constraints in Kiyotaki &
Moore (1997), whereas Holmstrom & Tirole (2001) derive the constraints from moral
hazard. Constrained agents in these papers are real-economy firms having access to pro-
ductive technologies not available to other investors. The interpretation of constrained
agents as financial firms is made explicit in He & Krishnamurthy (2008, 2009), who derive
constraints on fund managers uniquely able to invest in a subset of traded assets. The
constraints stem from moral hazard, and contracts are restricted to be static. [See also
Acharya & Viswanathan (2009), who derive constraints from asset substitution; and
Hombert & Thesmar (2009) and Stein (2009), who study arbitrageurs’ choice between
short- and long-term financing.] Extending this line of research to dynamic contracts, while
retaining the tractability that is necessary to compute asset prices in general equilibrium,
would be an important step forward. Recent literature on optimal dynamic financial
contracting in continuous time [e.g., Biais et al. (2007) and DeMarzo & Fishman (2007)]
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

could be useful in this respect. The result of that literature that investors punish
underperforming managers by scaling down their firms fits with the idea that investors
by University of New Hampshire on 05/27/14. For personal use only.

reduce their stakes in underperforming funds. [Bolton & Scharfstein (1990) and Gromb
(1994) derive this result in finite-horizon discrete-time settings, and Heinkel & Stoughton
(1994) derive the result in a two-period fund-management setting. See also Biais et al.
(2010), who show the converse result that investors reward overperforming managers by
scaling up their firms.]
Our understanding of the costs faced by arbitrageurs has greatly benefited from papers
exploring these costs in relatively simple and stylized settings. An important next step is to
integrate limits-of-arbitrage ideas more squarely into asset pricing theory by developing
tractable dynamic multiasset models that can address empirical puzzles. Work along these
lines could take constraints as given and so proceed in parallel with work on optimal
contracting—although an important objective should remain that the two lines of research
eventually merge.
A number of papers explore dynamic multiasset equilibrium settings under the assump-
tion that the only cost faced by arbitrageurs is risk. In Greenwood (2005) and Hau
(2009a), arbitrageurs absorb demand shocks of index investors following index redefini-
tions. In Gabaix et al. (2007), arbitrageurs are uniquely able to hold mortgage-backed
securities, and although they can hedge interest-rate risk in the government bond market,
they bear prepayment risk. In Garleanu et al. (2009), arbitrageurs absorb demand shocks
in the options market, and although they can hedge delta risk in the stock market, they
bear jump and volatility risk. In Vayanos & Vila (2009) and Greenwood & Vayanos
(2010b), arbitrageurs absorb shocks to the demand and supply of government bonds with
specific maturities and hedge by trading bonds with other maturities. A common theme in
these papers is that arbitrageurs transmit shocks to the demand for one asset to other
assets, with the effects being largest for assets that covary the most with the original asset.
This has implications for phenomena as diverse as index effects in the stock market, the
pricing of prepayment risk in the mortgage market, the behavior of implied volatility
surfaces in the options market, and the behavior of risk premia in the government bond
market. Hau (2009b), Jylha & Suominen (2009), and Plantin & Shin (2009) pursue similar
themes for the foreign exchange market, and Naranjo (2009) does so for the futures
market.
Other papers explore dynamic multiasset equilibrium settings under the assumption
that arbitrageurs face financial constraints. These papers are referenced in the last

268 Gromb
Vayanos

 paragraph of Section 3.4 for the case of leverage constraints and in Section 3.5 for the case
of constraints on equity capital. They constitute a rapidly growing literature which has the
potential to explain a variety of market anomalies on the basis of limits of arbitrage and
institutional frictions more broadly.

4. WELFARE AND POLICY

Financial crises, including the recent one, provide a painful reminder that government
intervention can be important for the smooth functioning of financial markets. Argu-
ments for intervention often center around the idea that failing financial institutions can
cause disruptions in asset markets and the effects can propagate to other institutions.
Standard models of asset pricing cannot be used to evaluate the merits of such arguments
because they abstract away from financial frictions and institutions. In these models
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

equilibrium is Pareto optimal, and there is no scope for intervention. [The same is true
for models in which arbitrageur wealth effects derive from wealth-dependent risk aver-
by University of New Hampshire on 05/27/14. For personal use only.

sion rather than explicit financial constraints (e.g., Kyle & Xiong 2001 and Xiong
2001).] Research on the limits of arbitrage has the potential to deliver a more useful
framework for designing and assessing public policy. Indeed, this research emphasizes the
role of specialized institutions in providing liquidity in asset markets. Understanding how
the financial health of these institutions is affected by their trading decisions, and
whether trading decisions are socially optimal, can guide public policy. Despite its rele-
vance, the welfare analysis of asset markets with limited arbitrage is still in its infancy. In
this section, we survey the existing work and highlight what we view as the main issues,
challenges, and promises.
We start by clarifying some methodological issues. Several papers study whether
constrained arbitrageurs stabilize or destabilize asset prices, i.e., decrease or increase vola-
tility. This is only indirectly a welfare question; assessing welfare by means of a utility-
based criterion, such as Pareto, is preferable. Given a welfare criterion, a natural question
is whether constraints lower efficiency, i.e., is unconstrained arbitrage better than
constrained arbitrage?5 This question, however, is of limited relevance for assessing policy.
Indeed, if arbitrageurs face constraints, one should not assume that a regulator could
remove or even relax them. A more suitable criterion is constrained efficiency: Can a
regulator increase welfare relative to the equilibrium by choosing new allocations that are
nevertheless subject to the constraints? At this juncture, the literature has taken two routes,
which we discuss next.
The first route is to retain a traditional dynamic equilibrium asset pricing model, but
not fully endogenize the constraints. This is done in Gromb & Vayanos (2002), who study
how leverage constraints affect arbitrageurs’ ability to provide liquidity. The main result in
terms of welfare is that equilibrium can fail to be constrained efficient, and a reduction in
arbitrageur positions can make all agents better off. The intuition is as follows: Following
an adverse shock, arbitrageurs incur capital losses and are forced to liquidate positions
because their leverage constraints tighten. As a result, they find themselves more
constrained and less able to provide liquidity—at a time when liquidity is low and its

5
A related question is whether the presence of arbitrageurs is beneficial, i.e., is constrained arbitrage better than no
arbitrage at all? One would expect the two questions to generally have a positive answer: Relaxing constraints
should be desirable because arbitrageurs provide liquidity.

www.annualreviews.org
Limits of Arbitrage 269

 provision profitable. Ex ante, arbitrageurs account for this possibility and engage in risk
management by keeping some capital in cash to exploit episodes of low liquidity. However,
they fail to account for the impact of their liquidations on other arbitrageurs during such
episodes. Indeed, liquidation by one arbitrageur hurts other arbitrageurs because it lowers
the price at which they can liquidate. This can hurt not only arbitrageurs but also outside
investors because of the reduced liquidity that they receive.
The second route has been followed by papers in macroeconomics and international
economics: for example, Caballero & Krishnamurthy (2001), Lorenzoni (2008), Acharya
et al. (2009), Hombert (2009), and Korinek (2009). These papers derive financial con-
straints endogenously from contracting frictions in three-period settings. The constraints,
however, concern real-economy firms having access to productive technologies, rather than
financial firms such as arbitrageurs. Inefficiencies arise because of a mechanism similar to
that mentioned in the previous paragraph: Firms do not account for the impact of their
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

liquidations on the prices at which other firms can liquidate. [See also Biais & Mariotti
(2009) for a similar inefficiency in a labor economics context and Nikolov (2009) for a
by University of New Hampshire on 05/27/14. For personal use only.

calibration exercise.]
The mechanism causing the inefficiencies is a pecuniary externality operating through
price changes and the redistribution of wealth that these generate. A redistribution of
wealth cannot be Pareto improving when markets are complete because marginal rates of
substitution across time and states of nature are identical for all agents. However, when
markets are incomplete, as is the case with financial constraints, marginal rates of substi-
tution differ, and Pareto improvements are possible. Geanakoplos & Polemarchakis (1986)
show this mechanism in a general incomplete-markets setting. [See also Stiglitz (1982),
who shows that in incomplete financial markets, the competitive equilibrium may fail to be
constrained efficient.]
This constrained inefficiency of limited arbitrage opens the door for an analysis of
policy [see Gromb & Vayanos (2002, 2009b) for a discussion]. Suppose, for example,
that arbitrageurs take excessive risk. A regulator might decrease their risk-taking incen-
tives by tightening their financial constraints with a risk-based capital requirement.
Alternatively, better risk management could be enforced directly by taxing arbitrageurs
in good times and possibly subsidizing them in bad times, in effect, managing part of
their resources for them. Subsidies in bad times can be implemented through lender-of-
last-resort policies or asset purchase programs. (For a discussion of such policies, see
Krishnamurthy 2009 and the references therein. Krishnamurthy argues that such poli-
cies are desirable because they can move the market to a more efficient equilibrium.)
Last, imperfect competition among arbitrageurs might lead them to internalize part of
the price effects of their investment decisions, and so adopt more socially desirable
investment policies.6 This research agenda has the potential to inform debates on systemic
risk, macro-prudential regulation, and lending of last resort, topics that are highly relevant
in the context of financial crises.

6
Much of the literature on the limits of arbitrage, including this review, assumes that arbitrageurs are perfectly
competitive. Imperfect competition constitutes an additional limit of arbitrage, given that arbitrageurs with market
power would require a larger price concession to absorb a demand shock. The industrial organization of arbitrage
activity is a largely unexplored area of research. It is related to financial constraints, given that that literature
assumes (often implicitly) that barriers to entry prevent the flow of new capital into the arbitrage industry. For
models with imperfectly competitive arbitrageurs see, for example, Attari & Mello (2006), Zigrand (2006), and
Oehmke (2009).

270 Gromb
Vayanos

 DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings
that might be perceived as affecting the objectivity of this review.

ACKNOWLEDGMENTS
We thank Suleyman Basak, Bruno Biais, Markus Brunnermeier, Bernard Dumas, Vincent
Fardeau, Robin Greenwood, Harald Hau, Peter Kondor, Arvind Krishnamurthy, Anna
Pavlova, Lasse Pedersen, Christopher Polk, Ana-Maria Santacreu, Yuki Sato, Andrei
Shleifer, Kari Sigurdsson, Pierre-Olivier Weill, and Jean-Pierre Zigrand for helpful com-
ments, and the Paul Woolley Centre at the London School of Economics for financial
support.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

LITERATURE CITED
by University of New Hampshire on 05/27/14. For personal use only.

Abreu D, Brunnermeier M. 2002. Synchronization risk and delayed arbitrage. J. Financ. Econ.
66:341–60
Abreu D, Brunnermeier M. 2003. Bubbles and crashes. Econometrica 71:173–204
Acharya V, Shin H, Yorulmazer T. 2009. A theory of slow-moving capital and contagion. Work. Pap.,
New York Univ.
Acharya V, Viswanathan V. 2009. Moral hazard, collateral and liquidity. Work. Pap., New York Univ.
Adrian T, Etula E, Shin H. 2009. Risk appetite and exchange rates. Work. Pap., Princeton Univ.
Aiyagari R, Gertler M. 1999. “Overreaction” of asset prices in general equilibrium. Rev. Econ. Dyn.
2:3–35
Allen F, Gale D. 2000. Bubbles and crises. Econ. J. 110:236–55
Amihud Y, Mendelson H. 1991. Liquidity, maturity, and the yield on U.S. Treasury securities.
J. Finance 46:479–86
Anshuman R, Viswanathan V. 2005. Costly collateral and liquidity. Work. Pap., Duke Univ.
Anton M, Polk C. 2008. Connected stocks. Work. Pap., Lond. Sch. Econ.
Attari M, Mello A. 2006. Financially constrained arbitrage in illiquid markets. J. Econ. Dyn. Control
30:2793–822
Bai Y, Chang E, Wang J. 2006. Asset prices under short-sale constraints. Work. Pap., Mass. Inst. Tech
Barberis N, Shleifer A. 2003. Style investing. J. Financ. Econ. 68:161–99
Barberis N, Shleifer A, Vishny R. 1998. A model of investor sentiment. J. Financ. Econ. 49:307–43
Barberis N, Shleifer A, Wurgler J. 2005. Comovement. J. Financ. Econ. 75:283–317
Barberis N, Thaler R. 2003. A survey of behavioral finance. See Constantinides et al. 2003, pp. 1053–125
Basak S, Croitoru B. 2000. Equilibrium mispricing in a capital market with portfolio constraints. Rev.
Financ. Stud. 13:715–48
Basak S, Croitoru B. 2006. On the role of arbitrageurs in rational markets. J. Financ. Econ. 81:143–73
Basak S, Pavlova A. 2010. Asset prices and institutional investors. Work. Pap., London Bus. Sch.
Bernanke B, Gertler M. 1989. Agency costs, net worth, and business fluctuations. Am. Econ. Rev.
79:14–31
Bernard V, Thomas J. 1989. Post-earnings-announcement drift: Delayed price response or risk pre-
mium? J. Account. Res. 27(Suppl.):1–48
Biais B, Mariotti T. 2009. Credit, wages, and bankruptcy laws. J. Eur. Econ. Assoc. 7:939–73
Biais B, Mariotti T, Plantin G, Rochet J-C. 2007. Dynamic security design: convergence to continuous
time and asset pricing implications. Rev. Econ. Stud. 74:345–90
Biais B, Mariotti T, Rochet J-C, Villeneuve S. 2010. Large risks, limited liability, and dynamic moral
hazard. Econometrica 78:73–118

www.annualreviews.org
Limits of Arbitrage 271

 Bolton P, Santos T, Scheinkman J. 2009. Outside and inside liquidity. Work. Pap., Columbia Univ.
Bolton P, Scharfstein D. 1990. A theory of predation based on agency problems in financial
contracting. Am. Econ. Rev. 80:93–106
Boyer B. 2007. Comovement among stocks with similar book-to-market ratios. Work. Pap., Brigham
Young Univ.
Brunnermeier M, Pedersen L. 2009. Market liquidity and funding liquidity. Rev. Financ. Stud.
22:2201–38
Caballero R, Krishnamurthy A. 2001. International and domestic collateral constraints in a model of
emerging market crises. J. Monet. Econ. 48:513–48
Chabakauri G. 2009. Asset pricing in general equilibrium with constraints. Work. Pap., Lond. Sch.
Econ.
Chen H, Noronha G, Singal V. 2004. The price response to S&P 500 index additions and deletions:
evidence of asymmetry and a new explanation. J. Finance 59:1901–29
Constantinides G, Harris M, Stulz R, ed. 2003. Handbook of the Economics of Finance: Financial
Markets and Asset Pricing. Vol. 1B. Amsterdam: North Holland
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Coval J, Stafford E. 2007. Asset fire sales (and purchases) in equity markets. J. Financ. Econ.
86:479–512
by University of New Hampshire on 05/27/14. For personal use only.

Cuoco D. 1997. Optimal consumption and equilibrium prices with portfolio constraints and stochas-
tic income. J. Econ. Theory 72:33–73
Cuoco D, Kaniel R. 2009. Equilibrium prices in the presence of delegated portfolio management.
Work. Pap., Univ. Penn.
Dabora E, Froot K. 1999. How are stock prices affected by the location of trade? J. Financ. Econ.
53:189–216
Daniel K, Hirshleifer D, Subrahmanyam A. 1998. A theory of overconfidence, self-attribution, and
security market under- and over-reactions. J. Finance 53:1839–85
Danielsson J, Shin H, Zigrand J-P. 2009. Risk appetite and endogenous risk. Work. Pap., Lond. Sch.
Econ.
Dasgupta A, Prat A. 2008. Information aggregation in financial markets with career concerns. J. Econ.
Theory 143:83–113
Dasgupta A, Prat A, Verardo M. 2008. The price impact of institutional herding. Work. Pap., Lond.
Sch. Econ.
D’Avolio G. 2002. The market for borrowing stock. J. Financ. Econ. 66:271–306
DeBondt W, Thaler R. 1985. Does the stock market overreact? J. Finance 40:793–805
DeLong B, Shleifer A, Summers L, Waldmann R. 1990. Noise trader risk in financial markets. J. Polit.
Econ. 98:703–38
DeMarzo P, Fishman M. 2007. Optimal long-term financial contracting. Rev. Financ. Stud.
20:2079–128
Diamond D, Verrecchia R. 1987. Constraints on short-selling and asset price adjustment to private
information. J. Financ. Econ. 18:277–311
Dow J, Gorton G. 1994. Arbitrage chains. J. Finance 49:819–49
Duffie D. 1996. Special repo rates. J. Finance 51:493–526
Duffie D, Garleanu N, Pedersen L. 2002. Securities lending, shorting, and pricing. J. Financ. Econ.
66:307–39
Duffie D, Garleanu N, Pedersen L. 2008. Valuation in over-the-counter markets. Rev. Financ. Stud.
20:1865–900
Duffie D, Strulovici B. 2009. Capital mobility and asset pricing. Work. Pap., Stanford Univ.
Fama E. 1991. Efficient capital markets: II. J. Finance 46:1575–617
Fama E, French K. 1992. The cross-section of expected stock returns. J. Finance 47:427–65
Fostel A, Geanakoplos J. 2008. Leverage cycles and the anxious economy. Am. Econ. Rev. 98:1211–44
Froot K, Scharfstein D, Stein J. 1993. Risk management: coordinating corporate investment and
financing policies. J. Finance 48:1629–58

272 Gromb
Vayanos

 Gabaix X, Krishnamurthy A, Vigneron O. 2007. Limits of arbitrage: theory and evidence from the
mortgage-backed securities market. J. Finance 62:557–95
Garleanu N, Pedersen L. 2009. Margin-based asset pricing and deviations from the law of one price.
Work. Pap., Univ. Calif. Berkeley
Garleanu N, Pedersen L, Poteshman A. 2009. Demand-based option pricing. Rev. Financ. Stud.
22:4259–99
Geanakoplos J. 1997. Promises, promises. In The Economy as an Evolving Complex System II, ed.
B Arthur, S Durlauf, D Lane, pp. 285–320. Reading, MA: Addison-Wesley
Geanakoplos J. 2003. Liquidity, default and crashes: endogenous contracts in general equilibrium. In
Advances in Economics and Econometrics: Theory and Applications II, Econometric Society
Monographs: Eighth World Congress, ed. M Dewatripont, L Hansen, S Turnovsky, 170–205.
Cambridge, UK: Cambridge Univ. Press
Geanakoplos J, Polemarchakis H. 1986. Existence, regularity, and constrained suboptimality of com-
petitive allocations when the asset market is incomplete. In Uncertainty, Information and Com-
munication, Essays in Honor of Kenneth J. Arrow, ed. W Heller, R Starr, D Starrett, 3:65–96.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Cambridge, UK: Cambridge Univ. Press

Greenwood R. 2005. Short- and long-term demand curves for stocks: theory and evidence on the
by University of New Hampshire on 05/27/14. For personal use only.

dynamics of arbitrage. J. Financ. Econ. 75:607–49

Greenwood R, Thesmar D. 2009. Stock price fragility. Work. Pap., Harvard Univ.
Greenwood R, Vayanos D. 2010a. Bond supply and excess bond returns. Work. Pap., Harvard Univ.
Greenwood R, Vayanos D. 2010b. Price pressure in the government bond market. Am. Econ. Rev.
100:585–90
Gromb D. 1994. Renegotiation in debt contracts. Work. Pap., INSEAD
Gromb D, Vayanos D. 2002. Equilibrium and welfare in markets with financially constrained arbitra-
geurs. J. Financ. Econ. 66:361–407
Gromb D, Vayanos D. 2009a. Financially constrained arbitrage and cross-market contagion. Work.
Pap., INSEAD
Gromb D, Vayanos D. 2009b. Leverage and liquidity dry-ups: a framework and policy implications.
Work. Pap., INSEAD
Gromb D, Vayanos D. 2010. A model of financial market liquidity based on arbitrageur capital.
J. Europ. Econ. Assoc. 8:456–66
Grossman S, Miller M. 1988. Liquidity and market structure. J. Finance 43:617–37
Grossman S, Vila J-L. 1992. Optimal dynamic trading with leverage constraints. J. Financ. Quant.
Anal. 27:151–68
Guerrieri V, Kondor P. 2009. Fund managers and defaultable debt. Work. Pap., Univ. Chic.
Harris L, Gurel E. 1986. Price and volume effects associated with changes in the S&P 500 list: new
evidence for the existence of price pressures. J. Finance 41:815–29
Harrison M, Kreps D. 1978. Speculative investor behavior in a stock market with heterogeneous
expectations. Q. J. Econ. 92:323–36
Hart O, Moore J. 1994. A theory of debt based on the inalienability of human capital. Q. J. Econ.
109:841–79
Hart O, Moore J. 1995. An analysis of the role of hard claims in constraining management. Am. Econ.
Rev. 85:567–85
Hau H. 2009a. Global versus local asset pricing: evidence from arbitrage of the MSCI index change.
Work. Pap., INSEAD
Hau H. 2009b. The exchange rate effect of multi-currency risk arbitrage. Work. Pap., INSEAD
He Z, Krishnamurthy A. 2008. Intermediary asset pricing. Work. Pap., Univ. Chic.
He Z, Krishnamurthy A. 2009. A model of capital and crises. Work. Pap., Univ. Chic.
He Z, Xiong W. 2008. Multi-market delegated asset management. Work. Pap., Univ. Chic.
Heinkel R, Stoughton N. 1994. The dynamics of portfolio management contracts. Rev. Financ. Stud.
7:351–87

www.annualreviews.org
Limits of Arbitrage 273

 Holmstrom B, Tirole J. 2001. LAPM: a liquidity-based asset pricing model. J. Finance 56:1837–67
Hombert J. 2009. Solvency shocks, liquidity shocks, and fire sales. Work. Pap., HEC
Hombert J, Thesmar D. 2009. Limits of limits of arbitrage: theory and evidence. Work. Pap., HEC
Hong H, Scheinkman J, Xiong W. 2006. Asset float and speculative bubbles. J. Finance 61:1073–117
Hong H, Stein J. 1999. A unified theory of underreaction, momentum trading, and overreaction in
asset markets. J. Finance 54:2143–84
Hong H, Stein J. 2003. Differences of opinion, short-sales constraints, and market crashes. Rev.
Financ. Stud. 16:487–525
Jegadeesh N, Titman S. 1993. Returns to buying winners and selling losers: implications for stock
market efficiency. J. Finance 48:65–91
Jotikasthira P, Lundblad C, Ramadorai T. 2009. Asset fire sales and purchases and the international
transmission of financial shocks. Work. Pap., Univ. Oxford
Jurek J, Yang H. 2007. Dynamic portfolio selection in arbitrage. Work. Pap., Princeton Univ.
Jylha P, Suominen M. 2009. Speculative capital and currency carry trades. J. Financ. Econ. In press
Kaniel R, Kondor P. 2010. A delegated Lucas tree. Work. Pap., Duke Univ.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Kehoe T, Levine D. 1993. Debt-constrained asset markets. Rev. Econ. Stud. 60:865–88
Kiyotaki N, Moore J. 1997. Credit cycles. J. Polit. Econ. 105:211–48
by University of New Hampshire on 05/27/14. For personal use only.

Kondor P. 2009. Risk in dynamic arbitrage: price effects of convergence trading. J. Finance 64:638–58
Korinek A. 2009. Systemic risk taking: amplification effects, externalities, and regulatory responses.
Work. Pap., Univ. Maryland
Krishnamurthy A. 2002. The bond/old-bond spread. J. Financ. Econ. 66:463–506
Krishnamurthy A. 2009. Amplification mechanisms in liquidity crises. Amer. Econ. J.: Macroecon.
2:1–30
Kyle A, Xiong W. 2001. Contagion as a wealth effect of financial intermediaries. J. Finance 56:1401–40
Lagos R, Rocheteau G, Weill P-O. 2009. Crises and liquidity in over the counter markets. Work. Pap.,
New York Univ.
Lamont O, Thaler R. 2003. Can the market add and subtract? Mispricing in tech stock carve-outs.
J. Polit. Econ. 111:227–68
LeRoy S, Porter R. 1981. The present value relation: tests based on implied variance bounds.
Econometrica 49:555–74
Liu J, Longstaff F. 2004. Losing money on arbitrage: optimal dynamic portfolio choice in markets with
arbitrage opportunities. Rev. Financ. Stud. 17:611–41
Loewenstein M, Willard G. 2006. The limits of investor behavior. J. Finance 61:231–58
Lorenzoni G. 2008. Inefficient credit booms. Rev. Econ. Stud. 75:809–33
Lou D. 2009. A flow-based explanation for return predictability. Work. Pap., Lond. Sch. Econ.
Malliaris S, Yan H. 2009. Nickels versus black swans: reputation, trading strategies and asset prices.
Work. Pap., Yale Univ.
Milbradt K. 2009. Trading and valuing toxic assets. Work. Pap., Mass. Inst. Tech.
Miller E. 1977. Risk, uncertainty and divergence of opinion. J. Finance 32:1151–68
Mitchell M, Pulvino T, Stafford E. 2002. Limited arbitrage in equity markets. J. Finance 57:551–84
Naranjo L. 2009. Implied interest rates in a market with frictions. Work. Pap., ESSEC
Nikolov K. 2009. Is private leverage excessive? Work. Pap., Lond. Sch. Econ.
Oehmke M. 2009. Gradual arbitrage. Work. Pap., Columbia Univ.
Pavlova A, Rigobon R. 2008. The role of portfolio constraints in the international propagation of
shocks. Rev. Econ. Stud. 75:1215–56
Petajisto A. 2009. Why do demand curves for stocks slope down? J. Financ. Quant. Anal. 44:1013–44
Plantin G, Shin H. 2009. Carry trades and speculative dynamics. Work. Pap., Toulouse Sch. Econ.
Rosenthal L, Young C. 1990. The seemingly anomalous price behavior of Royal Dutch/Shell and
Unilever N.V./PLC. J. Financ. Econ. 26:123–41
Rytchkov O. 2009. Dynamic margin constraints. Work. Pap., Temple Univ.
Scheinkman J, Xiong W. 2003. Overconfidence and speculative bubbles. J. Polit. Econ. 111:1183–219

274 Gromb
Vayanos

 Schwert W. 2003. Anomalies and market efficiency. See Constantinides et al. 2003, pp. 939–74
Shiller R. 1981. Do stock prices move too much to be justified by subsequent changes in dividends?
Am. Econ. Rev. 71:421–36
Shleifer A. 1986. Do demand curves for stocks slope down? J. Finance 41:579–90
Shleifer A, Vishny R. 1992. Liquidation values and debt capacity: a market equilibrium approach.
J. Finance 47:1343–66
Shleifer A, Vishny R. 1997. The limits of arbitrage. J. Finance 52:35–55
Stein J. 2009. Presidential address: sophisticated investors and market efficiency. J. Finance 64:1517–48
Stiglitz J. 1982. The inefficiency of the stock market equilibrium. Rev. Econ. Stud. 49:241–61
Tuckman B, Vila J-L. 1992. Arbitrage with holding costs: a utility-based approach. J. Finance
47:1283–302
Tuckman B, Vila J-L. 1993. Holding costs and equilibrium arbitrage. Work. Pap., Mass. Inst. Tech.
Vayanos D. 2004. Flight to quality, flight to liquidity and the pricing of risk. Work. Pap., Lond. Sch.
Econ.
Vayanos D, Vila J-L. 2009. A preferred-habitat theory of the term structure of interest rates. Work.
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Pap., Lond. Sch. Econ.

Vayanos D, Weill P-O. 2008. A search-based theory of the on-the-run phenomenon. J. Finance
by University of New Hampshire on 05/27/14. For personal use only.

63:1361–98
Vayanos D, Woolley P. 2008. An institutional theory of momentum and reversal. Work. Pap., Lond.
Sch. Econ.
Vijh A. 1994. S&P 500 trading strategies and stock betas. Rev. Financ. Stud. 7:215–51
Warga A. 1992. Bond returns, liquidity, and missing data. J. Financ. Quant. Anal. 27:605–17
Weill P-O. 2007. Leaning against the wind. Rev. Econ. Stud. 74:1329–54
Wurgler J, Zhuravskaya E. 2002. Does arbitrage flatten demand curves for stocks? J. Bus. 75:583–608
Xiong W. 2001. Convergence trading with wealth effects. J. Financ. Econ. 62:247–92
Yuan K. 2005. Asymmetric price movements and borrowing constraints: a REE model of crisis,
contagion, and confusion. J. Finance 60:379–411
Zigrand J-P. 2005. Rational asset pricing implications from realistic trading frictions. J. Bus. 78:871–92
Zigrand J-P. 2006. Endogenous market integration, manipulation and limits to arbitrage. J. Math.
Econ. 42:301–14

www.annualreviews.org
Limits of Arbitrage 275

 Annual Review of
Financial Economics

Volume 2, 2010 Contents

Portfolio Theory: As I Still See It

Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

Harry M. Markowitz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
by University of New Hampshire on 05/27/14. For personal use only.

Bayesian Portfolio Analysis

Doron Avramov and Guofu Zhou . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Cross-Sectional Asset Pricing Tests
Ravi Jagannathan, Ernst Schaumburg, and Guofu Zhou . . . . . . . . . . . . . . 49
CEO Compensation
Carola Frydman and Dirk Jenter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Shareholder Voting and Corporate Governance
David Yermack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
An Informal Perspective on the Economics and Regulation of
Securities Markets
Chester S. Spatt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Privatization and Finance
William Megginson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Asset Allocation
Jessica A. Wachter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Investment Performance Evaluation
Wayne E. Ferson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Martingale Pricing
Kerry Back. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Limits of Arbitrage
Denis Gromb and Dimitri Vayanos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Stochastic Processes in Finance
Dilip B. Madan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

vi
 Ambiguity and Asset Markets
Larry G. Epstein and Martin Schneider . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Risk Management
Philippe Jorion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

Errata
An online log of corrections to Annual Review of Financial Economics articles
may be found at http://financial.annualreviews.org
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org
by University of New Hampshire on 05/27/14. For personal use only.

Contents vii
 Annual Reviews
It’s about time. Your time. It’s time well spent.

New From Annual Reviews:

Annual Review of Statistics and Its Application
Volume 1 • Online January 2014 • http://statistics.annualreviews.org

Editor: Stephen E. Fienberg, Carnegie Mellon University

Associate Editors: Nancy Reid, University of Toronto
Stephen M. Stigler, University of Chicago
The Annual Review of Statistics and Its Application aims to inform statisticians and quantitative methodologists, as
Annu. Rev. Fin. Econ. 2010.2:251-275. Downloaded from www.annualreviews.org

well as all scientists and users of statistics about major methodological advances and the computational tools that
allow for their implementation. It will include developments in the field of statistics, including theoretical statistical
by University of New Hampshire on 05/27/14. For personal use only.

underpinnings of new methodology, as well as developments in specific application domains such as biostatistics
and bioinformatics, economics, machine learning, psychology, sociology, and aspects of the physical sciences.

Complimentary online access to the first volume will be available until January 2015.
table of contents:

• What Is Statistics? Stephen E. Fienberg • High-Dimensional Statistics with a View Toward Applications
• A Systematic Statistical Approach to Evaluating Evidence in Biology, Peter Bühlmann, Markus Kalisch, Lukas Meier
from Observational Studies, David Madigan, Paul E. Stang, • Next-Generation Statistical Genetics: Modeling, Penalization,
Jesse A. Berlin, Martijn Schuemie, J. Marc Overhage, and Optimization in High-Dimensional Data, Kenneth Lange,
Marc A. Suchard, Bill Dumouchel, Abraham G. Hartzema, Jeanette C. Papp, Janet S. Sinsheimer, Eric M. Sobel
Patrick B. Ryan • Breaking Bad: Two Decades of Life-Course Data Analysis
• The Role of Statistics in the Discovery of a Higgs Boson, in Criminology, Developmental Psychology, and Beyond,
David A. van Dyk Elena A. Erosheva, Ross L. Matsueda, Donatello Telesca
• Brain Imaging Analysis, F. DuBois Bowman • Event History Analysis, Niels Keiding
• Statistics and Climate, Peter Guttorp • Statistical Evaluation of Forensic DNA Profile Evidence,
• Climate Simulators and Climate Projections, Christopher D. Steele, David J. Balding
Jonathan Rougier, Michael Goldstein • Using League Table Rankings in Public Policy Formation:
• Probabilistic Forecasting, Tilmann Gneiting, Statistical Issues, Harvey Goldstein
Matthias Katzfuss • Statistical Ecology, Ruth King
• Bayesian Computational Tools, Christian P. Robert • Estimating the Number of Species in Microbial Diversity
• Bayesian Computation Via Markov Chain Monte Carlo, Studies, John Bunge, Amy Willis, Fiona Walsh
Radu V. Craiu, Jeffrey S. Rosenthal • Dynamic Treatment Regimes, Bibhas Chakraborty,
• Build, Compute, Critique, Repeat: Data Analysis with Latent Susan A. Murphy
Variable Models, David M. Blei • Statistics and Related Topics in Single-Molecule Biophysics,
• Structured Regularizers for High-Dimensional Problems: Hong Qian, S.C. Kou
Statistical and Computational Issues, Martin J. Wainwright • Statistics and Quantitative Risk Management for Banking
and Insurance, Paul Embrechts, Marius Hofert

Access this and all other Annual Reviews journals via your institution at www.annualreviews.org.

Annual Reviews | Connect With Our Experts

Tel: 800.523.8635 (us/can) | Tel: 650.493.4400 | Fax: 650.424.0910 | Email: [email protected]