## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

08-03

First Draft: 1une 1998

This Version: February 2006

Pairs Trading: Performance of a Relative

Value Arbitrage Rule

Evan Gatev, Boston CoIIege

WiIIiam N. Goetzmann, YaIe SchooI of Management, InternationaI

Center for Finance

K. Geert Rouwenhorst, YaIe SchooI of Management, InternationaI

Center for Finance

This paper can be downloaded without charge from the

Social Science Research Network Electronic Paper Collection:

http://ssrn.com/abstract÷141615

1

Pairs Trading: Performance of a Relative Value Arbitrage Rule

Evan Gatev

Assistant Professor

Boston College

William N. Goetzmann

Edwin J. Beinecke Professor of Finance and Management Studies

Yale University

K. Geert Rouwenhorst

Professor

Yale University

First draft: June 1998

This version: February 2006

Abstract

We test a Wall Street investment strategy, “pairs trading,” with daily data over 1962-2002.

Stocks are matched into pairs with minimum distance between normalized historical prices. A

simple trading rule yields average annualized excess returns of up to 11 percent for self-

financing portfolios of pairs. The profits typically exceed conservative transaction costs

estimates. Bootstrap results suggest that the “pairs” effect differs from previously-documented

reversal profits. Robustness of the excess returns indicates that pairs trading profits from

temporary mis-pricing of close substitutes. We link the profitability to the presence of a

common factor in the returns, different from conventional risk measures.

We are grateful to Peter Bossaerts, Michael Cooper, Jon Ingersoll, Ravi Jagannathan, Maureen O’Hara, Carl

Schecter and two anonymous referees for many helpful discussions and suggestions on this topic. We thank the

International Center for Finance at the Yale School of Management for research support, and the participants in the

EFA’99 Meetings, the AFA’2000 Meetings, the Berkeley Program in Finance and the Finance and Economics

workshops at Vanderbilt and Wesleyan for their comments.

2

Introduction

Wall Street has long been interested in quantitative methods of speculation. One popular

short-term speculation strategy is known as “pairs trading.” The strategy has at least a twenty-

year history on Wall Street and is among the proprietary "statistical arbitrage" tools currently

used by hedge funds as well as investment banks. The concept of pairs trading is disarmingly

simple. Find two stocks whose prices have moved together historically. When the spread

between them widens, short the winner and buy the loser. If history repeats itself, prices will

converge and the arbitrageur will profit. It is hard to believe that such a simple strategy, based

solely on past price dynamics and simple contrarian principles, could possibly make money. If

the U.S. equity market were efficient at all times, risk-adjusted returns from pairs trading should

not be positive.

In this paper, we examine the risk and return characteristics of pairs trading with daily

data over the period 1962 through December 2002. Using a simple algorithm for choosing pairs,

we test the profitability of several straightforward, self-financing trading rules. We find average

annualized excess returns of about 11 percent for top-pairs portfolios. While pairs strategies

exploit temporary components of stock prices, we show that our profits are not caused by simple

mean reversion as documented in the previous literature. We examine the robustness of our

results to a wide variety of risk factors – including not only the widely used factors in the

empirical literature but also potential low-frequency institutional factors such as bankruptcy risk.

In addition, we explore the robustness of our results to microstructure factors such as the bid-ask

bounce, short-selling costs, and transactions costs. While some factors such as short-selling and

transactions costs affect the magnitude of the excess returns, pairs trading remains profitable for

reasonable assumptions over the sample period of study, as well as over a true out of sample test

3

of four years. We interpret the results of our analysis as evidence in favor of profitable arbitrage

in expectations that may accrue to market participants who possess relatively low transactions

costs and the ability to short securities. We also find evidence that points to a systematic factor

that influences the profitability of pairs trading over time. This unidentified latent risk factor has

been relatively dormant recently. The importance of this risk factor is correlated with the returns

to pairs trading, which is consistent with the view that the profits are a compensation to

arbitrageurs for enforcing the “Law of One Price.”

We argue that our results reveal something about the mechanism and performance of

relative-price arbitrage activities in practice. This is potentially useful to researchers, because,

despite considerable theory about market efficiency, economists have little empirical information

about how efficiency is maintained in practice. In addition, despite the fact that hedge funds

have attracted an increasing amount of investment capital over the past decade, the study of

hedge fund strategies is in its infancy in the financial economics literature. This paper examines

the risk and return characteristics of one widely practiced active trading strategy.

One natural question to ask is whether our results imply a violation of equilibrium asset

pricing. While the documented profitability of the pairs trading rule is a robust result, it is not

inconsistent with all pricing models. Indeed the reversion in relative values we find is consistent

with a pricing model in prices developed and tested by Bossaerts (1988). Thus our paper at the

very least suggests that this class of models merits further empirical investigation.

The remainder of the paper is organized as follows. Section 1 provides some background

on pairs trading strategy. The next section describes our methodology of constructing pairs and

calculating returns. The empirical results are described in section 3, and section 4 provides

conclusions and directions for future research.

4

1. Background of Pairs Trading

1.1 History

In the mid-1980's, the Wall Street quant Nunzio Tartaglia assembled a team of physicists,

mathematicians and computer scientists to uncover arbitrage opportunities in the equities

markets. Tartaglia's group of former academics used sophisticated statistical methods to develop

high-tech trading programs, executable through automated trading systems, that took the

intuition and trader's “skill” out of arbitrage and replaced it with disciplined, consistent filter

rules. Among other things, Tartaglia's programs identified pairs of securities whose prices

tended to move together. They traded these pairs with great success in 1987 – a year when the

group reportedly made a $50 million profit for the firm. Although the Morgan Stanley group

disbanded in 1989 after a couple of bad years of performance, pairs trading has since become an

increasingly popular "market-neutral" investment strategy used by individual and institutional

traders as well as hedge funds. The increased popularity of quantitative-based statistical

arbitrage strategies has also apparently affected profits. In a New York Times interview, David

Shaw, head of one of the most successful modern quant shops and himself an early Tartaglia’s

protégé, suggests that recent pickings for quant-shops have become slim – he attributes the

success of his firm, D.E. Shaw, to early entry into the business. Tartaglia's own explanation for

pairs trading is psychological. He claims, that “…Human beings don't like to trade against

human nature, which wants to buy stocks after they go up not down.

1

” Could pairs traders be the

disciplined investors taking advantage of the undisciplined over-reaction displayed by individual

investors? This is at least one possible – albeit psychological – explanation for our results,

1

Hansell, S., 1989, “Inside Morgan Stanley's Black Box,” Institutional Investor, May, p.204.

5

which is consistent with Jegadeesh and Titman's (1995) finding that contrarian profits are in part

due to over-reaction to company-specific information shocks rather than price reactions to

common factors.

1.2 Data Snooping and Market Response

In our study we have not searched over the full strategy space to identify successful

trading rules, but rather we have interpreted practitioner description of pairs trading as

straightforwardly as possible. Our rules follow the general outline of first “find stocks that move

together,” and second “take a long-short position when they diverge and unwind upon

convergence.” A test requires that both of these steps must be parameterized in some way. How

do you identify “stocks that move together?” Need they be in the same industry? Should they

only be liquid stocks? How far do they have to diverge before a position is put on? When is a

position unwound? We have made some straightforward choices about each of these questions.

We put positions on at a two-standard deviation spread, which might not always cover

transactions costs even when stock prices converge. Although it is tempting to try potentially

more profitable schemes, the danger in data-snooping refinements outweighs the potential

insights gained about the higher profits that could result from learning through testing

2

.

As with all filter rules using historical asset pricing data, data-snooping is a potential

concern. One approach to the data snooping issue is to test the results out-of-sample. We

completed and circulated the first draft of the working paper in 1999, using data through the end

of 1998. The time lag between the first analysis and the present study gives us an ideal hold-out

sample. Using the original model, but the post 1988 data, we found that over the 1999-2002

2

Froot and Dabora (1999) consider “twin” stocks that trade in different international markets to examine issues of market

integrationon.

6

period, the excess return of the fully invested portfolio of the top twenty pairs averaged 10.4

percent per annum, with an annual standard deviation of 3.8% and a large and significant

Newey-West-adjusted t-statistic of 4.82 – consistent with the long-term, in-sample results of our

original analysis. We were careful not to adjust our strategy from the first draft to the current

draft of the paper, to avoid data-snooping criticisms. Not only does this additional four-year

sample suggest that the results were not simply an artifact of the earlier sample period, over

which pairs-trading was known to be popular, but it also suggests that the public dissemination

of the results has apparently not affected the general risk and return characteristics of the

strategy, despite curiosity from the professional sector.

1.3 Relative Pricing

Asset pricing can be viewed in absolute and relative terms. Absolute pricing values

securities from fundamentals such as discounted future cash flow. This is a notoriously difficult

process with a wide margin for error. Papers by Bakshi and Chen (1997) and Lee et al. (1997),

for example, are heroic attempts to build quantitative value investing models. Relative pricing is

only slightly easier. Relative pricing means that two securities that are close substitutes for each

other should sell for the same price – it does not say what that price will be. Thus, relative

pricing allows for bubbles in the economy, but not necessarily arbitrage or profitable speculation.

The Law of One Price [LOP] – and a “near-LOP” is applicable to relative pricing – even if that

price is wrong.

Ingersoll (1987) defines the LOP as the “proposition ... that two investments with the

same payoff in every state of nature must have the same current value.” In other words, two

securities with the same prices in all states of the world should sell for the same amount. Chen

7

and Knez (1995) extend this to argue that “closely integrated markets should assign to similar

payoffs prices that are close.” They argue that two securities with similar, but not necessarily

matching payoffs across states should have similar prices. This is of course a weaker condition,

and subject to bounds on prices for unusual states, however it allows the examination of “near-

efficient” economies, or in Chen and Knez’ case, near integrated markets. Notice that this

theory corresponds to the desire to find two stocks whose prices move together as long as we can

define states of nature as the time-series of observed historical trading days.

We use an algorithm to choose pairs based on the criterion that they have had the same or

nearly the same state prices historically. We then trade pairs whose prices closely match in

historical state-space, since the LOP suggests that in an efficient market, their prices should be

nearly identical. In this framework, the current study can be viewed as a test of the LOP and

near-LOP in the U.S. equity markets, under certain stationarity conditions. We are effectively

testing the integration of very local markets – the markets for specific individual securities. This

is similar in spirit to Bossaerts’ (1988) test of co-integration of security prices at the portfolio

level. We further conjecture that the marginal profits to be had from risk arbitrage of these

temporary deviations are crucial to the maintenance of first-order efficiency. We could not have

the first effect without the second.

1.4 Co-integrated Prices

The pairs trading strategy may be justified within an equilibrium asset-pricing framework

with non-stationary common factors like Bossaerts and Green (1989) and Jagannathan and

Viswanathan (1988). If the long and short components fluctuate with common non-stationary

factors, then the prices of the component portfolios would be co-integrated and the pairs trading

8

strategy would be expected to work. Evidence of exposures to common non-stationary factors

would support a non-stationary factor pricing framework.

The space of normalized, cum-dividend prices, i.e. cumulative total returns with

dividends re-invested, is the basic space for the pairs trading strategies in this paper. The main

observation about our motivating models of the CAPM-APT variety is that they are known to

imply perfect collinearity of prices, which is readily rejected by the data. On the other hand,

Bossaerts (1988) finds evidence of price co-integration for the US stock market. We would like

to keep the notion of the empirically observed co-movement of prices, without unnecessarily

restrictive assumptions, hence we proceed in the spirit of the co-integrated prices literature.

More specifically, our matching in price space can be interpreted as follows. Suppose that prices

obey a statistical model of the form,

p

it

= ! "

il

p

lt

+ #

it

, k < n (1)

where #

it

denotes a weakly dependent error in the sense of Bossaerts (1988). Assume also that p

it

is weakly dependent after differencing once. Under these assumptions, the price vector p

t

is co-

integrated of order 1 with co-integrating rank r = n-k , in the sense of Engle and Granger (1987)

and Bossaerts (1988). Thus, there exist r linearly independent vectors {$

q

}

q=1.. r

such that z

q

=

$

q

`p

t

are weakly dependent. In other words, r linear combinations of prices will not be driven by

the k common non-stationary components p

l

. Note that this interpretation does not imply that

the market is inefficient, rather it says that certain assets are weakly redundant, so that any

deviation of their price from a linear combination of the prices of other assets is expected to be

temporary and reverting.

9

To interpret the pairs as co-integrated prices, we need to assume that for n » k, there are

co-integrating vectors which have only two nonzero coordinates. In that case the sum or

difference of scaled prices will be reverting to zero and a trading rule could be constructed to

exploit the expected temporary deviations. Our strategy relies upon exactly this conclusion. In

principle one could construct trading strategies with trios, quadruples, etc. of stocks, which

would presumably capture more co-integrated prices and would yield better profits.

The assumption that a linear combination of two stocks can be weakly dependent may be

interpreted as saying that a co-integrating vector can be partitioned in two parts, such that the

two corresponding portfolios are priced within a weakly dependent error of another stock.

Given the large universe of stocks, this statement is always empirically valid and provides the

basis of our formation procedure

3

. However, it is important to recognize the possibility of

spuriously correlated prices, which are not de-facto co-integrated.

1.5 Bankruptcy Risk

The risk of bankruptcy is one reason why the returns on individual securities cannot be

taken as stationary. Sensitivity of the pairs trading to the default premium suggests that the

strategy may work because we are pairing two firms, the first of which may have a constant or

decreasing probability of bankruptcy (short end), while the second may have a temporarily

increasing probability of bankruptcy (long end). The “surprise improvements” in the short end

are then followed by improvement in the long end if that stock survives. In other words, the

source of the profit is the improving ex-post (non)realization of idiosyncratic bankruptcy risk in

the long (loser) stock. In such case, we would expect to have asymmetry in the profits from the

3

Note that the case n » k corresponds to the standard finance paradigm where in the large universe of n stocks, expected returns

are driven by a few, namely k common factors. This paradigm is supported by existing empirical work, e.g. see Connor and

10

long and the short components, with most of the profits coming from the long end

4

. We test long

and short positions separately to see if this is driving our results.

2. Methodology

Our implementation of pairs trading has two stages. We form pairs over a twelve-month

period (formation period) and trade them in the next six-month period (trading period). Both

twelve months and six months are chosen arbitrarily and have remained our horizons since the

beginning of the study.

2.1 Pairs Formation

In each pairs-formation period, we screen out all stocks from the CRSP daily files that

have one or more days with no trade. This serves to identify relatively liquid stocks as well as to

facilitate pairs formation. Next, we construct a cumulative total return index for each stock over

the formation period. We then choose a matching partner for each stock by finding the security

that minimizes the sum of squared deviations between the two normalized price series. Pairs are

thus formed by exhaustive matching in normalized daily “price” space, where price includes re-

invested dividends.

We use this approach because it best approximates the description of how traders

themselves choose pairs. Interviews with pair traders suggest that they try to find two stocks

whose prices “move together.” In addition to “unrestricted” pairs, we will also present results

by sector, where we restrict both stocks to belong to the same broad industry categories defined

by Standard & Poors: Utilities, Transportation, Financial and Industrials. Each stock is assigned

Korajczyk (1993) for references, which generally finds less than ten common non-stationary components.

eree for this example.

11

to one of these four groups, based on the stock’s SIC code. The minimum-distance criterion is

then used to match stocks within each of the groups.

2.2 Trading Period

Once we have paired up all liquid stocks in the formation period, we study the top 5 and

20 pairs with the smallest historical distance measure, in addition to the 20 pairs after the top 100

(pairs 101-120). This last set is valuable because most of the top pairs share certain

characteristics, which will be described in detail below. On the day following the last day of the

pairs formation period, we begin to trade according to a pre-specified rule.

Figure 1 illustrates the pairs trading strategy using two stocks, Kennecott and Uniroyal in

the six-month period starting in August of 1962. The top two lines represent the normalized

price paths with dividends re-invested, and the bottom line indicates the opening and closing of

the strategy on a daily basis. It is clear why these two firms paired with each other. They

generally tended to move together over the trading interval. Notice that the position first opens

in the seventh trading day of the period and then remains open until day 36. Over that interval,

the spread actually first increased significantly before convergence. The prices remain close

during the period and cross frequently. The pair opens five times during the period, however not

always in the same direction. Neither stock is the “leader.” In our example, convergence

occurs in the final day of the period, although this is not always the case.

We select trading rules based on the proposition that we open a long-short position when

the pair prices have diverged by a certain amount, and close the position when the prices have

reverted. Following practice, we base our rules for opening and closing positions on a standard

deviation metric. We open a position in a pair when prices diverge by more than two historical

12

standard deviations, as estimated during the pairs formation period. We unwind the position at

the next crossing of the prices. If prices do not cross before the end of the trading interval, gains

or losses are calculated at the end of the last trading day of the trading interval. If a stock in a

pair is delisted from CRSP, we close the position in that pair, using the delisting return, or the

last available price

5

. We report the payoffs by going one dollar short in the higher-priced stock

and one dollar long in the lower-priced stock.

2.3 Excess Return Computation

Because pairs may open and close at various points during the six-month trading period,

the calculation of the excess return on a portfolio of pairs is a non-trivial issue. Pairs that open

and converge during the trading interval will have positive cash flows. Because pairs can re-

open after initial convergence, they can have multiple positive cash flows during the trading

interval. Pairs that open but do not converge will only have cash flows on the last day of the

trading interval when all positions are closed out. Therefore, the payoffs to pairs trading

strategies are a set of positive cash flows that are randomly distributed throughout the trading

period, and a set of cash flows at the end of the trading interval which can either be positive or

negative. For each pair we can have multiple cash flows during the trading interval, or we may

have none in the case when prices never diverge by more than two standard deviations during the

trading interval. Because the trading gains and losses are computed over long–short positions of

one dollar, the payoffs have the interpretation of excess returns. The excess return on a pair

5

The profits are robust with respect to this delisting assumption. A potential problem arises if inaccurate and stale prices

exaggerate the excess returns and bias the estimated return of a long position in a plummeting stock. To address this potential

concern we have re-estimated our results under the extreme assumption that a only a long stock experiences a -100% return when

it is delisted. This zero-price extreme includes among other things, the possibility of non-trading due to lack of liquidity. Since

selective loss on the long position always harms the pair profit, this extreme assumption biases the results against profitability.

However, pairs trading remains profitable under this alternative: for example, the average monthly return on the top-20 pairs

portfolio is 1.32% with a standard deviation of 1.9%.

13

during a trading interval is computed as the reinvested payoffs during the trading interval

6

. In

particular, the long and short portfolio positions are marked-to-market daily. The daily returns

on the long and short positions are calculated as value-weighted returns in the following way,

!

i%P

w

i,t

r

i,t

r

P,t

= -------------- (2)

!

i%P

w

i,t

w

i,t

= w

i,t-1

(1+r

i,t-1

) = (1+r

i,1

)… (1+r

i,t-1

) (3)

where r defines returns and w defines weights, and the daily returns are compounded in order to

obtain monthly returns. This has the simple interpretation of a buy-and-hold strategy.

We consider two measures of excess return on a portfolio of pairs: the return on

committed capital and the fully-invested return, i.e. the return on actual employed capital. The

former scales the portfolio payoffs by the number of pairs that are selected for trading, the latter

divides the payoffs by the number of pairs that open during the trading period. The former

measure of excess return is clearly more conservative: if a pair does not trade for the whole of

the trading period, we still include a dollar of committed capital as the cumulative return in our

calculation of excess return. It takes into account the opportunity cost of hedge funds of having

to commit capital to a strategy even if the strategy does not trade. To the extent that hedge funds

are flexible in their sources and uses of funds, computing excess return relative to the actual

capital employed may give a more realistic measure of the trading profits.

We initiate the pairs strategy by trading the pairs at the beginning of every month in the

sample period, with the exception of the first twelve months, which are needed to estimate pairs

for the strategy starting in the first month. The result is a time series of overlapping six-month

trading period excess returns. We correct for the correlation induced by overlap by averaging

6

This is a conservative approach to computing the excess return, because it implicitly assumes that all cash earns zero interest

rate when not invested in an open pair. Because any cash flow during the trading interval is positive by construction, it ignores

the fact that these cash flows are received early, and understates the computed excess returns.

14

monthly returns across trading strategies that start one month apart as in Jegadeesh and Titman

(1993). The resulting time series has the interpretation of the payoffs to a proprietary trading

desk, which delegates the management of the six portfolios to six different traders whose

formation and trading periods are staggered by one month.

3. Empirical Results

3.1 Strategy Profits

Table 1 summarizes the excess returns for the pairs portfolios that are unrestricted in the

sense that the matching stocks do not necessarily belong to the same broad industry categories.

In section 3.5 we will consider sector-neutral pairs strategies. Panel A summarizes the excess

returns of pairs strategies when positions are opened at the end of the day that prices diverge, and

closed at the end of the day of price convergence. The first row shows that a fully-invested

portfolio of the five best pairs earned an average excess monthly return of 1.31% (t-statistic =

8.84), and a portfolio of the twenty best pairs 1.44% per month (t = 11.56). Using the more

conservative approach to computing excess returns, using committed capital, gives excess

returns of 0.78% and 0.81% per month respectively. Either way, these excess returns are large in

an economical and statistical sense, and suggest that pairs trading is profitable.

The remainder of Panel A provides information about the excess return distributions of

pairs portfolios. There are diversification benefits from combining multiple pairs in a portfolio.

As the number of pairs in a portfolio increases, the portfolio standard deviation falls. The

diversification benefits are also apparent from the range of realized returns. Interestingly, as the

number of pairs in the strategy increases the minimum realized return increases, while the

maximum realized excess return remains relatively stable. During the full sample period of 474

months, a portfolio of 20 pairs experienced 71 monthly periods with negative payoffs, compared

15

to 124 months for a portfolio of 5 pairs. The decrease in the standard deviation, and the increase

of the lower end of the return distribution is also reflected in an increased skewness coefficient.

Since pairs trading is in essence a contrarian investment strategy, the returns may be

biased upward due to the bid-ask bounce (Jegadeesh (1990), Jegadeesh and Titman (1995),

Conrad and Kaul (1989)). In particular, our strategy sells stocks that have done well relative to

their match and buys those that have done poorly. Part of any observed price divergence is

potentially due to price movements between bid and ask quotes: conditional on divergence, the

winner’s price is more likely to be an ask quote and the loser’s price a bid quote. In Panel A we

have used these same prices for the start of trading and our returns may be biased upward due to

the fact that we are implicitly buying at bid quotes (losers) and selling at ask quotes (winners).

The opposite is true at the second crossing (convergence): part of the drop in the winner’s price

can reflect a bid quote, and part of the rise of the loser’s price an ask quote.

To address this issue, Panel B of Table 1 provides the excess returns when we initiate

positions in each pair on the day following the divergence and liquidate on the day following the

crossing. The average excess returns on the fully-invested portfolios and on committed capital

drop by about 30-55 and 20-35 basis points respectively. While the excess returns remain

significantly positive, the drop in excess returns suggests that a non-trivial portion of the profits

in Panel A may be due to bid-ask bounce. It is difficult to quantify which portion of the profit

reduction is due to bid-ask bounce and which portion stems from true mean reversion in prices

due to rapid market adjustment. None-the-less, this difference raises questions about the

economic significance of our results when we include transactions costs. We will return to a

detailed discussion of this issue in section 3.3. Unless stated otherwise, the remainder of the

paper will report results for pairs strategies that open (close) on the day following divergence

(convergence).

16

3.2 Trading Statistics and Portfolio Composition

Table 2 provides a summary of the trading statistics and composition of the pairs

portfolios. What are the characteristics of the stocks that are matched into pairs? How often

does a typical pair trade? Because pairs trading is an active investment strategy, it is important

to evaluate the profitability relative to the trading intensity of the portfolios. As mentioned

before we use a two standard deviation trigger to open a pairs position. The second line of panel

A in Table 2 reports the average price deviation of the 2 standard deviation trigger. For the top

five pairs, the position typically opens when prices have diverged by 4.76% or more. This is a

relatively narrow gap in prices

7

. The trigger spread increases with the number of pairs in the

portfolio, because the standard deviation of the prices increases as the proximity of the securities

in price space decreases. The next lines of Panel A also shows that on average almost all pairs

open during the six-month trading period, and on average more than once. Of the top 5 pairs, on

average 4.81 open during the trading period, and the average number of round trips per pair is

2.02. The average duration of an open position is 3.75 months. This indicates that pairs trading

– implemented according to the particular rules we chose – is a medium-term investment

strategy.

Panel B of Table 2 describes the composition of the pairs in terms of market

capitalization and industry membership. In terms of size, the average stock in the top 5 and top

20 pairs belongs to the second and third deciles from the top; 74% of the stocks in the top 20

pairs belong to the top three size deciles using CRSP breakpoints, and 91% come from the top

five size deciles. About two-thirds of the pairs combine stocks from different size deciles (i.e.

“size mixed pairs”), and the stocks in mixed pairs differ on average by a single decile.

The remainder of Panel B gives a breakdown of the pairs by industry composition. On

7

The optimal trigger point in terms of profitability may actually be much higher than 2 standard deviations, although we have

17

average, 71% of the stocks in the top 20 pairs are utility stocks, despite the fact that Utilities

represent a fairly small proportion of the stocks in the whole sample. This is not surprising

perhaps, because utility stocks tend to have lower volatility and tend to be correlated with

interest rate innovations. The strategy does not always match stocks within sectors. The

percentage of mixed sector pairs ranges from 20 per cent for the top 5 pairs to 44 per cent for

pairs 101-120. Given the predominance of utilities among the top pairs, it is fair to ask whether

the profitability of pairs trading profitability is limited to the utility sector, or whether pairs

strategies are also profitable in other sectors of the market. We address this question in section

3.5.

3.3 Transactions costs

Table 1 shows that the average monthly excess return of unrestricted pairs strategies falls from

1.44%, for the top-20 portfolio, to 0.90% per month if we postpone the trades to the day

following the crossing. This drop in the excess returns implies an estimate of the average bid-ask

spread and hence the transactions costs of trading in the sample. While actual transactions costs

may be different, it is informative to know whether the trading profits are large enough to

survive this estimate of transactions costs.

Suppose the extreme case where the prices of the winner at the first crossing (divergence)

are ask prices and the loser are bid prices. If the next day prices are equally likely to be at bid or

ask the delaying trades by one day will reduce the excess returns on average by half the sum of

the spreads of the winner and the loser. If at the second crossing (convergence) of the pairs the

winners is trading at the bid, and the loser at the ask, waiting one day will reduce the excess

returns on average again by one half of the sum of the bid-ask spreads of both stocks. In this

not experimented to find out.

18

extreme case, waiting a day before trading reduces the return on each pair by the round-trip

transactions costs in that pair. Because we trade each pair on average 2 times during the six-

month trading interval, the drop in the excess returns of 324 basis points per six-months by

waiting one day reflects the cost of 2 round-trips, which implies a transactions costs of 162 bp

per pair per round-trip. This may be interpreted as an estimated effective spread of 81 bp. The

effective spread for the all-pairs portfolio is 70 bp. This indirect estimate is higher than the

transactions costs reported by Peterson and Fialkowski (1994), who find that the average

effective spread for stocks in the CRSP database in 1991 was 37 basis points, and is consistent

with the trading costs estimated by Keim and Madhavan (1997). Since 91% of the stocks in the

top 20 pairs belong to the top 5 deciles of CRSP stocks, it is possible that the effective spread is

even lower that 37 bp.

Do our trading strategies survive these transactions costs? The profits on our trading

strategies in Table 1 range from 437 to 549 basis points over a six-month period. If the prices

used to compute these excess returns are equally likely to be at bid or ask, which seems a

reasonable assumption, we have to correct these excess returns to reflect that in practice we buy

at the ask and sell at the bid prices. In other words, we have to subtract the round-trip trading

costs to get an estimate of the profits after transactions costs. Our conservative estimate of

transactions costs of 162 bp times 2 rounds trips per pair results in an estimate of 324 bp

transactions cost per pair per six-month period. This gives average net profits ranging from 113

to 225 bp over each six-month period. Comparing these profits to the reported standard errors,

we conclude that they are both economically and statistically significant.

Further analysis is required to get more precise estimates of influence of transactions

costs of pairs trading strategies. An important question in this context is whether the trading rule

that we have used to open and close pairs can be expected to generate economically significant

19

profits even if pairs trading works perfectly. Because we use a measure of historical standard

deviation to trigger the opening of pairs, and since this estimated standard deviation is the

smallest among all pairs, it is likely to underestimate the true standard deviation of a pair. As a

consequence, we may simply be opening pairs “too soon” and at a point that we cannot expect it

to compensate for transactions costs even if the pair subsequently converges. Results that are not

reported here suggest that this is indeed the case for some of our pairs.

There is a second reason why our trading strategies require “too much” trading. We open

pairs at any point during the trading period when the normalized prices diverge by two standard

deviations. This is not a sensible rule towards the end of a trading interval. For example, suppose

a divergence occurs at the next to last day of the trading interval. The convergence has to be

substantial in order to overcome the transactions cost that will be incurred when we close out the

position on the next day (the last day of the trading interval). Unreported results suggest that this

is also an important source of excess trading.

3.4 Pairs trading by industry group

The pairs formation process thus far has been entirely mechanical. A computer stock has

the opportunity to match with a steel firm, and a utility with a bank stock. This does not mean

that these matches are likely. As shown in Table 2, the fraction of mixed pairs is typically well

below 50 percent. Common factor exposures of stocks in the same industry will make it more

likely to find a match within the same sector. Also, firms that are in industries where cross-

sectional differences in factor exposures are small or return variances are low are more likely to

end up among the top ranking of pairs. For this reason it is perhaps not surprising that many of

the top pairs match two utilities. Are the profits to pairs trading consistent across sectors? We

examine the returns on pairs trading where stocks are matched only within the four large sector

20

groupings used by Standard and Poor’s: Utilities, Transportation, Financials, and Industrials.

The results are summarized in Table 3. As in Table 1, the pairs are traded with a one-day delay

before opening and closing a position in order to minimize the effect of the bid-ask bounce on

trading. The monthly excess returns for the top 20 pairs are the largest in the Utilities sector, with

1.08% (Newey-West t =10.26). The profits for the other industry groups are somewhat lower,

but all statistically significant, with the average Transportation, Financials, and Industrials top 20

pairs earning 0.58% (NW t = 4.26), 0.78% (NW t = 7.60) and 0.61% (NW t = 6.93) respectively

over a one-month period.

Table 3 also gives a more detailed picture of the return distributions and trading

characteristics of the pairs trading strategies by sector. It shows that the excess return

distributions of the sector pairs portfolios are generally skewed right and exhibit positive excess

kurtosis relative to a normal distribution. The conclusion from these Tables is that pairs trading is

profitable in every broad sector category, and not limited to a particular sector.

3.5 The Risk Characteristics of Pairs Trading Strategies.

To provide further perspective on the risk of pairs trading, Table 4 compares the risk

premium of pairs trading to the market premium (SP500), and reports the risk-adjusted returns to

pairs trading using two different models for measuring risk. Table 5 summarizes value-at-risk

measures for pairs portfolios.

The top part of Table 4 compares the excess return to pairs trading to the excess return on

the SP500. Between 1963 and 2002, the average excess return to pairs trading has been about

twice as large as the excess return of the SP500, with only one half to one third of the risk as

measured by standard deviation. As a result, the Sharpe Ratios of pairs trading are between 4 and

6 times larger than the Sharpe Ratio of the market. Goetzmann et al. (2002) show that Sharpe

21

Ratios can be misleading when return distributions have negative skewness. This is unlikely to

be a concern for our study, because our Table 1 showed that the returns to pairs portfolios are

positively skewed, which – if anything – would bias our Sharpe ratios downward.

In order to explore the systematic risk exposure of the pairs portfolios, we regress their

monthly excess returns on the three factors of Fama and French (1996), augmented by two

additional factors. The motivation for the additional factors is that pairs strategies invest based

on the relative strength of individual stocks. It is therefore possible that pairs trading simply

exploits patterns in returns that are known to earn significant profits. For example, Jegadeesh

(1990) and Lehmann (1990) show that reversal strategies that select stocks based on prior one-

month return earn positive abnormal returns. We control for this possibility by constructing a

short-term reversal factor measured as the excess return of stocks in the top 3 deciles of prior-

month return minus the return on stocks in the bottom three deciles

8

. If pairs strategies sell

short-term winners, and buy short-term losers, we expect the exposures of pairs portfolios to be

positive to the reversal factor. The second additional factor controls for exposure to medium-

term return continuation (Jegadeesh and Titman (1993)). To the extent that pairs trading sells

medium-term winners, and buys medium-term losers, the pairs excess returns will be negatively

correlated with momentum. To examine this possibility, we include a momentum factor in our

risk model that is constructed along the lines of Carhart (1997).

Table 4 shows that only a small portion of the excess returns of pairs trading can be

attributed to their exposures to the five risk factors. The intercepts of the regressions show that

risk-adjusted returns are significantly positive, and lower than the raw excess returns by about

10-20 bp per month. Because pairs strategies are market-neutral, the exposures to the market are

small and with one exception insignificant. Exposures to the other two FF factors, the difference

8

The construction is similar to Carhart’s (1997) momentum factor, but the performance-sorting horizon here is one month.

22

between small and big stocks (SMB), and the difference between value and growth stocks

(HML), are not significant and the point estimates alternate in sign. The exposures to

momentum and reversals have the predicted signs, and more than half are statistically significant.

As can be expected, some of the winner stocks that a pairs strategy shorts are short-term winners

while others are medium-term winners. Similarly, the losers are evenly divided between short-

term and medium-term losers. Overall the exposures are not large enough to explain the average

returns to pairs trading. The significance of the risk-adjusted returns indicates that pairs trading is

fundamentally different from simple contrarian strategies based on reversion (Jegadeesh (1990)

and Lehmann (1990)). The next sections provide further evidence to support this view.

The bottom of Table 4 shows the results of regressing the pairs returns on an alternative

set of risk factors suggested by Ibbotson: the excess return on the S&P 500, the U.S. small stock

premium, the U.S. bond default premium and the U.S. bond horizon premium. Although the risk

exposures are generally not significant, the signs of coefficients are positive across all portfolios.

Thus, when corporate bonds increase in price relative to government bonds, the pairs portfolios

make money. There are a range of possible explanations for this pattern. Relatively cheaper

borrowing rates by arbitrageurs may force stock prices closer to equilibrium values, or common

factors affecting convergence in both stock and bond markets may be responsible. The

portfolios also appear to be sensitive to shifts in the yield curve, i.e. when long-term spreads

decrease, pairs trading is more profitable. At first glance, the sensitivity to term-structure

measures may be explained by the presence of interest rate sensitive Utility stocks in many of the

top pairs. However, interest rate movements also seem to matter for the more broadly diversified

pairs portfolios. In sum, the pairs portfolios seem to have low exposure to various sources of

systematic risk. The R-squared of most regressions is low, indicating that the portfolios are

nearly factor-neutral. This may be expected since they are constructed in a way that should

23

essentially match up economic substitutes.

Table 5 reports both monthly and daily value at risk [VAR] measures to summarize the

quantiles of the empirical distributions of the pairs’ excess returns. Panel A shows that the worst

monthly loss over the almost 40-year sample period was 12.6 % for the top five pairs portfolio,

and 8.2% for the top 20 portfolio. On average, only once in every hundred months did these

portfolios lose more than 4.32 % and 1.94 % respectively. Panel B shows that on its worst day,

the top five portfolio recorded a 10.08 % loss, compared to a 6.72% loss of the top-20 portfolio.

On average, once every hundred days did these portfolios lose more than 1.24 % and 0.65 %

respectively.

The VAR is useful because it provides a gauge to the potential leverage that could be

applied to these strategies. A five-to-one leverage ratio applied to the top 5 pairs would appear to

have been adequate to cover the worst monthly loss in the 40- year period. Although the lessons

of recent history have taught us not to rely too heavily on historical VAR measures for gauging

capital needs for exploiting convergence strategies, the pairs portfolios seem to be exposed to

relatively little risk.

Figure 2 shows the monthly performance of the top 20 pairs, based upon the next-day

trading rule. Pairs trading was very profitable in the 1970's and 1980's, and then had a span of

more modest performance, when the returns were sometimes negative. Figure 3 compares the

cumulative excess returns of the top 20 one-day-waiting strategy the cumulative excess returns

of investment in the S&P 500 index. The smooth index of the pairs trading portfolio contrasts

dramatically with the volatility of the stock market. Pairs trading performed well over difficult

times for U.S. stocks. When the U.S. stock market suffered a dramatic real decline from 1969

through 1980, the pairs strategy experienced some of its best performance. By contrast, in the

mid-90s the market performed exceptionally well, but pairs trading profits were relatively flat.

24

Perhaps after its discovery in the early 1980's by Tartaglia and others, competition decreased

opportunity. On the other hand, pair trading might simply be more profitable in times when the

stock market performs poorly.

Two additional explanations may account for the temporal variation in profitability. The

first is the long-term trend in transactions costs over the history of our sample. The early part of

the analysis represents a period with high fixed commissions. These frictions might have

prevented the rapid convergence of relative prices. The secular decrease in transactions costs

may have attracted more relative value equity arbitrage. This is becoming increasingly so with

the introduction of more sophisticated trading technologies and networks. The second

explanation may be the rise in hedge funds in the period since 1989. The TASS hedge fund

database reports that hedge fund assets rose from $4 billion in 1977 to $137 billion in 2000, with

the most dramatic growth after 1992. While this database does not capture proprietary trading

operations of investment banks, it does suggest that risk arbitrage activities have grown

significantly in the last decade of our sample period. Of the $137 billion in hedge fund assets in

2000, $119 billion employed “Market Neutral” “Relative Value” or “Arbitrage” strategies, all of

which may potentially impact pairs trading operations. Thus, it is possible that lower

transactions costs and large inflows of investment capital to relative-value arbitrage have

together decreased the rate of return on pairs trading, at least in the simple form we test in this

paper. In section 3.8 we will revisit this issue, and show that in contrast to the raw returns, the

risk-adjusted return to pairs trading has been relatively stable over time.

3.6 Pairs trading and contrarian investment

Because pairs trading bets on price reversals, it is an example of a contrarian investment

strategy. The results of Table 4 showed that the returns to pairs are positively correlated with –

25

but not explained by – short-term reversals documented by Lehmann (1990) and

Jegadeesh(1990). In this section we further explore whether our pairs trading strategies are

merely a disguised way of exploiting these previously documented negative autocorrelations. In

particular, we conduct a bootstrap where we compare the performance of our pairs to random

pairs. The starting point of the bootstrap is the set of historical dates on which the various pairs

open. In each bootstrap we replace the actual stocks with two random securities with similar

prior one-month returns as the stocks in the actual pair. Similarity is defined as coming from the

same decile of previous month’s performance. The difference between the actual and the

simulated pairs returns provides an indication of the portion of our pairs return that is not due to

reversion. We bootstrapped the entire set of trading dates 200 times. The results are summarized

in Table 6. On average we find that the returns on the bootstrapped pairs are well below the true

pairs returns. In fact, the excess returns to the simulated pairs are slightly negative, and the

standard deviations of the returns are large relative to the true pairs, which is a reflection of the

fact that the simulated pairs are poorly matched. The conclusion from the simulations confirms

the conclusion from the factor regressions, that the pairs strategy does not merely reflect one-

month mean reversion. In addition, in the next section, we will show that the long and short

portfolios that make up a pair do not provide equal contributions to the profitability of the

strategy. These three findings combined strongly suggest that our pairs trading strategy seems to

capture temporal variation in returns that is different from simple mean reversion.

3.7 Risk and Return of the Long and Short Positions

There are at least three reasons to separately examine the returns to the long and short

portfolios that make up a pairs position. First, the separate returns provide further insight into the

question of mean reversion. If pairs trading simply exploits mean reversion, one would expect

26

the abnormal returns to the long and short positions to be equal because the opening of a pair is

equally likely to be triggered by either stock. Second, if the excess returns are predominantly

driven by the short position, it becomes important to examine whether short-sale considerations

might prevent arbitrageurs from competing away the profits. Finally, the risk exposures of the

two portfolios can provide further clues at to the reason for the profitability of pairs – or example

the possibility that the long and short portfolios have different exposures to common non-

stationary risk factors such as bankruptcy risk

9

. The returns and the risk exposures of the

component portfolios are summarized in Table 7.

The Table shows that much of the pairs risk-adjusted excess return comes from the short

portfolio, which contains the stocks that have increased in value relative to their counterparts

prior to opening of the pair. By contrast, the alphas of the long portfolio containing the stocks

that decreased in value relative to their counterparts are smaller, and insignificantly different

from zero for the top-5 and top-20 portfolios. The asymmetry of the results provides further

evidence that the returns to pairs are not due to simple one-month mean reversion. And because

much of the abnormal return comes from the short position, which experienced an increase in

relative value prior to opening, it is unlikely that the returns are driven by a reward for unrealized

bankruptcy risk. The robustness of our results to the cost of short sales will be discussed in

section 3.9.

3.8 Sub-period Analysis and the Presence of a Dormant Risk Factor

Table 8 summarizes the profitability of pairs trading when we spit the sample period at

the end of 1988. A comparison between the two top halves of the two panels shows a drop in the

raw excess returns to pairs trading. For example, the excess return of the top-20 strategy drops

9

We thank a referee for suggesting this explanation.

27

from 118 bp per month to about 38 bp per month. Has increased hedge fund activity arbitraged

away the anomalous behavior of pairs since the pre-1989 period? Inspection of the risk adjusted

returns shows that this is not the case: the average risk-adjusted return of the top-20 portfolio

drops by about a third from 67 to 42 basis points per month, but remains significantly positive in

both sub-periods (t = 4.41 and 3.77 respectively). Changes in the factor exposures and factor

volatilities can explain only part of the lower returns in the early part of the sample, but not the

risk-adjusted returns to pairs trading.

Are the positive risk-adjusted returns to pairs trading a general failure of our risk model?

In other words, are there reasons to believe that the risk adjusted pairs returns are a compensation

for an omitted (latent) risk factor? Inspection of the correlation between disjoint pairs portfolios

provides some support for this view. Moreover, this latent risk factor seems to have been

relatively dormant over the second half of our sample, which can account for the lower recent

profitability of pairs trading. The full-sample correlation between the excess returns of the top-

20 and the 100-120 pairs portfolios is 0.48. Because there is no overlap between the positions of

these portfolios, the correlation indicates the presence of a common factor to the returns.

Moreover, the correlation is 0.51 over the profitable pre-1988 period but much lower (0.18) over

the lower excess return post-1988 period. These correlations are not driven by the 5

“systematic” risk factors we considered, because the correlations of the residuals from the factor

regressions are very similar to the raw correlations. In particular, the correlation between the

top-20 Fama-French-Momentum-Reversal (FFMR) residuals and the top 100-120 FFMR

residuals is 0.41. The respective sub-period FFMR correlations are 0.42 and 0.20. This is

further illustrated in Figure 4, which shows that the rolling 24-month correlation between the two

portfolios was especially high during the pre-1989 period.

These results suggest that there is common component to the profits of pairs portfolios

28

that is not captured by our conventional measures of systematic risk. The common component

was stronger during the first half of our sample, which is consistent with higher profits (abnormal

returns) than in the second half of our sample when the factor was more dormant. These results

are consistent with the view that the abnormal returns documented in this paper are indeed a

compensation for risk, in particular the reward to arbitrageurs for enforcing the “Law of One

Price”.

3. 9 Robustness to Short-selling costs

Having identified and back-tested a filter rule, and subjected it to a range of controls for

risk, the question of the nature of pairs profits remains. Why do prices of close economic

substitutes diverge and converge? The convergence is easier to understand than the divergence,

given the natural arbitrage motivation, and the documented existence of relative-value

arbitrageurs in the U.S. equity market. However, this does not explain why prices drift away

from parity in the first place. One possible explanation is that prices diverge on random liquidity

shocks that cannot be exploited by professional arbitrageurs due to short-selling costs

10

.

First, there are explicit short-selling costs in the form of specials. Second, D’Avolio

(2002) argues that short recalls are potentially costly because they may deprive arbitrageurs of

their profits. This opportunity cost is reinforced by D’Avolio’s findings that short recalls are

more common as the price declines. For example, if the short stock is recalled when it starts to

converge downward, then the pair position is forced to close prematurely and the arbitrageur

does not capturing the profit from the pair convergence. We perform two tests for robustness of

the profits that corresponds to these two types of short-selling costs.

The first test is motivated by the findings of D’Avolio (2002) and Getczy et al. (2002)

10

We are grateful to the editor Maureen O’Hara for pointing out this plausible explanation and for suggesting the way to test it.

29

that specials have minimal effect of large stocks. Correspondingly, we test for robustness of

profits by trading pairs that are formed using only stocks in the top three size deciles

11

. The

results, given in Panel A of Table 9, can be directly compared to those in Panel B of Table 1. The

comparison shows that the profits of the top-20 strategy drop by about 2 basis points per month,

but increase for the top 5 portfolio. Overall, the profits change little and remain highly

significant. This shows that the pairs trading profits are not driven by illiquid stocks that are

likely to be on special.

The second test is motivated by the evidence in Chen et al. (2002) and D’Avolio (2002)

that short recalls are driven by dispersion of opinion. We use high volume as proxy for

divergence of opinion and perform pairs-trading under recalls, where we simulate recalls on the

short positions, and subsequent closing of the pair position, on days with high volume. High

volume days are defined as days on which daily volume exceeds average daily volume over the

18months (both split-adjusted) by more than one standard deviation. Panel B of Table 9 shows

the profits of pairs trading with the high volume recalls. The profits decline slightly by 4-13bp

per month, yet they remain large and positive. For example, the top 20 pairs portfolio earns an

average of 85bp per month with short recalls, with a Newey-West t-statistic of 9.07. The results

in Panel B can be interpreted as an estimate of the opportunity cost of short recalls.

Overall, the small effects confirm that the profits persist when trading pairs of large

stocks as well as when shorts are recalled. These results show that pairs trading profits are

robust to short-selling costs. For better-positioned investors, e.g. large institutions and hedge

funds, the pairs trading profits are likely to remain essentially unaffected by potential shorting

costs. Geczy et al. (2002) argue that for large traders, who have better access to most stocks at

“wholesale prices,” direct shorting costs of the rebate rate on short sales are low (4 to 15

11

Using large liquid stocks also mitigates the problem of stocks that are hard to short and can be overvalued, see e.g. Jones and

30

bp/year). The main implicit shorting cost stems from limited availability and is relevant mostly

for general retail investors. The impact of such potential short sales constraints on the

profitability of pairs trading by large investors is mitigated by our use of liquid stocks that trade

every day over a period of one year.

4. Conclusion

We examine a hedge fund equity trading strategy based on the notion of co-integrated

prices in a reasonably efficient market, known on Wall Street as pairs trading. Pairs are stocks

which are close substitutes according to a minimum distance criterion using a metric in price

space. We find that trading suitably formed pairs of stocks exhibits profits, which are robust to

conservative estimates of transaction costs. These profits are uncorrelated to the S&P 500,

however they do exhibit low sensitivity to the spreads between small and large stocks and

between value and growth stocks in addition to the spread between high grade and intermediate

grade corporate bonds and shifts in the yield curve. In addition to risk and transactions cost, we

rule out several explanations for the pairs trading profits, including mean-reversion as previously

documented in the literature, unrealized bankruptcy risk, and the inability of arbitrageurs to take

advantage of the profits due to short-sale constraints.

One view of the lower profitability of pairs trading in recent year is that returns are

competed away by increased hedge fund activity. The alternative view, taken in this paper, is that

abnormal returns to pairs strategies are a compensation to arbitrageurs for enforcing the “Law of

One Price”. We present two pieces of empirical evidence that supports this view. First, while raw

returns have fallen the risk-adjusted returns have continued to persist despite increased hedge

fund activity. Second, our results suggest that the change in risk-adjusted returns of pairs trading

Lamont (2002).

31

is accompanied by the diminished importance of a common factor that drives the returns to pairs

strategies. A further examination of the nature of this common factor and its link to the

profitability of pairs trading is an important question for future research.

32

References:

Bakshi, G. and Z. Chen, 1997, “Stock Valuation in Dynamic Economies,” working paper, Ohio

State University.

D’Avolio, G., 2002, “The Market for Borrowing Stock,” Journal of Financial Economics, 66,

271-306.

Bossaerts, P., 1988, “Common Nonstationary Components of Asset Prices,” Journal of

Economic Dynamics and Control, 12, 347-364.

Bossaerts, P. and R. Green, 1989, “A General Equilibrium Model of Changing Risk Premia:

Theory and Evidence,” Review of Financial Studies, 2, 467-493.

Chen, J., H. Hong and J. Stein, 2002, “Breadth of Ownership and Stock Returns,” Journal of

Financial Economics, 66, 171-205.

Chen, Z. and P. Knez, 1995, “Measurement of Market Integration and Arbitrage,” Review of

Financial Studies, 8, 287-325.

Carhart, M., 1997, “On Persistence in Mutual Fund Performance,” Journal of Finance, 52, 57-

82.

Connor, G. and R. Korajczyk, 1993, “A Test for the Number of Factors in an Approximate

Factor Model,” Journal of Finance, 48, 1263-1291.

Conrad, J. and G. Kaul, 1989, “Mean Reversion in Short-horizon Expected Returns,” Review of

Financial Studies, 2, 225-240.

Engle, R. and C. Granger, 1987, “Co-integration and Error Correction: Representation,

Estimation and Testing,” Econometrica, 55, 251-276.

Fama, E. and K. French, 1996, “Multifactor Explanations of Asset Pricing Anomalies,” Journal

of Finance 51, 131-155.

Froot, K. and E. Dabora, 1999, “How are Stock Prices Affected by the Location of Trade?,”

Journal of Financial Economics, 53, 189-216.

Geczy, C., D. Musto and A. Reed, 2002, “Stocks are Special Too: An Analysis of the Equity

Lending Market,” Journal of Financial Economics 66, 241-269.

Goetzmann W., et al., “Sharpening Sharpe Ratios”, working paper, Yale School of Management.

Hansell, S., 1989, “Inside Morgan Stanley's Black Box,” Institutional Investor, May, 204.

Ingersoll, J., Jr., 1987, Theory of Financial Decision-Making, Rowman and Littlefiled, New

Jersey.

33

Jagannathan, R. and S. Viswanathan, 1988, “Linear Factor Pricing, Term Structure of Interest

Rates and the Small Firm Anomaly,” Working Paper 57, Northwestern University.

Jegadeesh, N., 1990, “Evidence of Predictable Behavior of Security Returns,” Journal of

Finance, 45, 881-898.

Jegadeesh, N., and S. Titman, 1993, “Returns to Buying Winners and Selling Losers:

Implications for Stock Market Efficiency,” Journal of Finance, 48, 65-91.

Jegadeesh, N., and S. Titman, 1995, “Overreaction, Delayed Reaction, and Contrarian Profits,”

Review of Financial Studies, 8, 973-93.

Jones, C. and O. Lamont, 2002, “Short-sale Constraints and Stock Returns,” Journal of

Financial Economics, 66, 207-239.

Keim, D. and A. Madhavan, 1997, “Transactions Costs and Investment Style: An Inter-exchange

Analysis of Institutional Equity Trades,” Journal of Financial Economics, 46, 265-292.

Lee, C., et al., 1997, “What is the Intrinsic Value of the Dow?,” working paper, Cornell

University.

Lehmann, B., 1990, “Fads, Martingales and Market Efficiency,” Quarterly Journal of

Economics, 105, 1-28.

Peterson, M. and D. Fialkowski, 1994, “Posted versus Effective Spreads: Good Prices or Bad

Quotes?,” Journal of Financial Economics, 35, 269-292.

34

Table 1: Excess Returns of Unrestricted Pairs Trading Strategies

Summary statistics of the monthly excess returns on portfolios of pairs between July 1963 and December 2002 (474

observations). We trade according to the rule that opens a position in a pair at the end of the day that prices of the

stocks in the pair diverge by two historical standard deviations (Panel A). The results in Panel B correspond to a

strategy that delays the opening of the pairs position by one day. All pairs are ranked according to least distance in

historical price space. The “top n” portfolios include the n pairs with least distance measures, and the portfolio

“101

Pairs Portfolio top 5 top 20 101-120 All

Panel A: Excess Return Distribution (No Waiting)

Average excess return (Fully Invested) 0.01308 0.01436 0.01081 0.01104

Standard error (Newey-West) 0.00148 0.00124 0.00094 0.00099

t-statistic 8.84 11.56 11.54 11.16

Excess return distribution

Median 0.01194 0.01235 0.00955 0.00728

Standard deviation 0.02280 0.01688 0.01540 0.01670

Skewness 0.62 1.39 1.34 3.42

Kurtosis 7.81 10.54 10.30 25.25

Minimum -0.10573 -0.06629 -0.03857 -0.02721

Maximum 0.14716 0.13295 0.12684 0.17178

Observations with excess return < 0 26% 15% 21% 17%

Average excess return on committed capital 0.00784 0.00805 0.00679 0.00614

Panel B: Excess Return Distribution (1 day Waiting)

Average monthly return (Fully Invested) 0.00745 0.00895 0.00795 0.00715

Standard error (Newey-West) 0.00119 0.00096 0.00085 0.00090

t-statistic 6.26 9.29 9.40 7.92

Excess return distribution

Median 0.00699 0.00690 0.00694 0.00411

Standard deviation 0.02101 0.01527 0.01438 0.01577

Skewness 0.34 1.45 0.98 3.32

Kurtosis 10.64 16.13 7.78 25.66

Minimum -0.12628 -0.08218 -0.04266 -0.02951

Maximum 0.14350 0.13490 0.10464 0.16325

Observations with excess return < 0 35% 23% 28% 32%

Average excess return on committed capital 0.00463 0.00520 0.00503 0.00396

35

Table 2: Trading Statistics and Composition of Pairs Portfolios

Trading statistics and portfolio composition of portfolios of pairs portfolios between July 1963 to December 2002

(474 months). Pairs are formed over a 12 month period according to a minimum distance criterion, and then traded

over the subsequent 6 month period. We trade according to the rule that opens a position in a pair on the day

following the day on which the prices of the stocks in the pair diverge by two historical standard deviations. The

“top n” portfolios include the n pairs with least distance measures, and the portfolio “101-120" studies the twenty

pairs after the top 100. Panel A summarizes the trading characteristics of a pairs strategy. Pairs are opened when

prices diverge by two standard deviations. Average deviation to trigger opening of pair is the cross-sectional

average of two standard deviations which panel B contains information about the size and industry membership of

the stocks in the various pairs portfolios.

Pairs Portfolio top 5 top 20 101-120 All

Panel A: Trading Statistics

Average price deviation trigger for opening pairs 0.04758 0.05284 0.07560 0.16888

Average number of pairs traded per 6-month period 4.81 19.30 19.41 1944.22

Average number of round-trip trades per pair 2.02 1.96 1.78 1.62

Standard deviation of number of round-trips per pair 0.62 0.40 0.27 0.16

Average time pairs are open in months 3.75 3.76 3.98 3.97

Standard dev. of time open, per pair, in months 0.80 0.45 0.38 0.17

Panel B: Pair Portfolio Composition

Average size decile of stocks 2.54 2.71 3.41 4.57

Average weight of stocks in top 3 size deciles. 0.78 0.74 0.58 0.40

Average weight of stocks in top 5 size deciles 0.91 0.91 0.79 0.62

Average weight of pairs from different deciles 0.66 0.69 0.75 0.82

Average decile difference for mixed pairs 0.97 0.97 0.97 0.98

Average sector weights

Utilities 0.72 0.71 0.32 0.08

Transportation 0.02 0.02 0.02 0.03

Financials 0.11 0.13 0.26 0.16

Industrials 0.15 0.14 0.40 0.73

Mixed sector pairs 0.20 0.22 0.44 0.33

36

Table 3: Industry Sector Pairs Trading

Summary statistics for the excess monthly return distributions for pairs trading portfolios by sector. We trade

according to the “wait 1 day” rule described in the text. The average number of stocks in the industry groups is as

follows: 156 Utilities, 61 Transportation, 371 Financials, and 1729 Industrials. There is no “20 after 100" portfolio

for the Transportation industry group. The t-statistic of the mean is computed using Newey-West standard errors

with six lags.

Portfolio top 5 top 20 20 after 100 All

Panel A: Utilities

Mean Excess Return 0.00905 0.01084 0.009256 0.01036

t-statistic (NW) 7.37 10.26 6.11 10.51

Median 0.00829 0.00938 0.00665 0.00969

Standard Deviation 0.02154 0.01645 0.02640 0.01472

Skewness 0.44 0.76 0.66 1.39

Kurtosis 13.67 12.38 5.52 12.74

Minimum -0.12347 -0.08750 -0.07868 -0.03519

Maximum 0.16563 0.12730 0.12133 0.12878

Observations with excess return < 0 28% 19% 35% 18%

Panel B: Transportation

Mean Excess Return 0.00497 0.00577 0.00440

t-statistic (NW) 2.98 4.26 3.07

Median 0.00594 0.00547 0.00339

Standard Deviation 0.03892 0.02942 0.02871

Skewness -0.50 -0.11 0.53

Kurtosis 4.98 5.00 5.89

Minimum -0.19570 -0.12467 -0.12526

Maximum 0.10961 0.11099 0.13560

Observations with excess return < 0 44% 42% 44%

Panel C: Financials

Mean Excess Return 0.00678 0.00775 0.00854 0.00726

t-statistic (NW) 4.79 7.60 7.12 7.62

Median 0.00557 0.00615 0.00660 0.00511

Standard Deviation 0.02598 0.01792 0.02358 0.01808

Skewness 0.81 0.61 1.98 2.64

Kurtosis 6.37 6.48 18.27 21.21

Minimum -0.081988 -0.07810 -0.100477 -0.04452

Maximum 0.149291 0.10638 0.195797 0.17211

Observations with excess return < 0 40% 33% 32% 34%

Panel D: Industrial

Mean Excess Return 0.00490 0.00607 0.00664 0.00715

t-statistic (NW) 4.11 6.93 6.80 7.52

Median 0.00292 0.00547 0.00550 0.00397

Standard Deviation 0.02361 0.01601 0.01825 0.01672

Skewness 0.40 0.43 0.94 2.97

Kurtosis 4.94 4.75 7.65 20.95

Minimum -0.10016 -0.04786 -0.05412 -0.03431

Maximum 0.10546 0.08022 0.11619 0.16072

Observations with excess return < 0 42% 36% 35% 34%

37

Table 4: Systematic Risk of Pairs Trading Strategies

Monthly risk exposures for portfolios of pairs formed and traded according to the “Wait One Day” rule discussed in

the text, over the period between June 1963 and December 2002. The five actors are the three Fama-French

factors, Carhart’s Momentum factor, and the Reversal factor discussed in the text. Returns for the portfolios are in

excess of the riskless rate. S&P 500 returns are calculated in excess of Treasury bill returns. The Ibbotson factors

are from the Ibbotson EnCorrr analyzer: The U.S. Small stock premium is the monthly geometric difference

between small company stock total returns and large company stock total returns. U.S. bond default premium is the

monthly geometric difference between total return to long-term corporate bonds and long term government bonds.

The U.S. bond horizon premium is the monthly geometric difference between investing in long term government

bonds and U.S. treasury bills. t-statistics are in parentheses below the coefficients and are computed using Newey-

West standard errors with 6 lags.

top 5 top 20 20 aft. top 100 All Eq. Pre.

Wait One Day Portfolio Performance

Mean Excess Return 0.00745 0.00895 0.00795 0.00715 0.00410

Standard Deviation 0.02101 0.01527 0.01438 0.01577 0.04509

Sharpe Ratio 0.35 0.59 0.55 0.45 0.09

Monthly Serial Correlation 0.14 0.24 0.19 0.12 0.05

Factor Model: FF, Momentum, Reversal

Intercept 0.00545 0.00764 0.00714 0.00512

(3.81) (7.08) (8.66) (5.30)

Market -0.06661 -0.03155 -0.07697 -0.14520

(-1.03) (-0.64) (-1.77) (-3.10)

SMB -0.04233 0.00111 -0.02333 -0.07079

(-0.71) (0.02) (-0.50) (-1.66)

HML 0.05740 0.04514 -0.01724 -0.05403

(1.37) (1.45) (-0.59) (-1.82)

Momentum -0.02804 -0.04817 -0.10312 -0.18077

(-0.94) (-2.45) (-5.83) (-8.50)

Reversal 0.10192 0.07237 0.09459 0.20077

(1.50) (1.27) (2.24) (4.34)

R-square 0.05 0.09 0.18 0.54

Factor Model: Ibbotson Factors

Intercept 0.00716 0.00857 0.00766 0.00651

(6.32) (9.25) (9.39) (7.77)

Market -0.00182 0.01377 0.01642 0.06466

(-0.07) (0.74) (0.90) (1.98)

Small Stock Premium 0.04120 0.05227 0.03646 0.07608

(1.32) (2.22) (1.66) (1.93)

Bond Default Premium 0.14593 0.15989 0.16811 0.30571

(1.11) (1.38) (1.81) (2.82)

Bond Horizon Premium 0.07997 0.06818 0.04034 0.03422

(1.55) (1.64) (1.04) (0.77)

R-square 0.02 0.05 0.04 0.15

38

Table 5: Value at Risk of Pairs Trading

Monthly and daily value-at-risk percentiles of pairs trading strategies between July 1963 to December 2002 (474

months). Pairs are formed over a 12 month period according to a minimum distance criterion, and then traded

over the subsequent 6 month period. We trade according to the rule that opens a position in a pair on the day

following the day on which the prices of the stocks in the pair diverge by two historical standard deviations. The

“top n” portfolios include the n pairs with least distance measures, and the portfolio “101-120" studies the twenty

pairs after the top 100. The average number of pairs in the all-pair portfolio is 2057.

top 5 top 20 20 aft. 100 All

Panel A: Monthly VAR

Mean Excess Return 0.00745 0.00895 0.00795 0.00715

Standard Deviation 0.02101 0.01527 0.01438 0.01577

Serial Correlation 0.14 0.24 0.19 0.12

Value at Risk

1% -0.04320 -0.01943 -0.02236 -0.01994

5% -0.02142 -0.01002 -0.01293 -0.00877

10% -0.01516 -0.00577 -0.00756 -0.00614

25% -0.00460 0.00054 -0.00145 -0.00146

Probability of return below 0 0.35 0.23 0.28 0.32

Min. historical observation -0.12628 -0.08218 -0.04266 -0.02951

Panel B: Daily VAR

Mean Excess Return 0.00033 0.00040 0.00035 0.00027

Standard Deviation 0.00492 0.00296 0.00277 0.00169

Serial Correlation -0.12 -0.06 0.00 0.35

Value at Risk

1% -0.01236 -0.00647 -0.00653 -0.00327

5% -0.00710 -0.00398 -0.00400 -0.00202

10% -0.00504 -0.00293 -0.00288 -0.00149

25% -0.00239 -0.00133 -0.00130 -0.00071

Probability of return below 0 0.47 0.44 0.45 0.46

Min. historical observation -0.10079 -0.06723 -0.01987 -0.01069

39

Table 6: Returns to Random Pairs Sorted on Prior One Month Return

Bootstrap of random pairs traded according to the rule which opens a position in a random pair when the stocks in

the true pair diverge by two historical standard deviations and closes the position after the next crossing of prices.

The random stocks are selected within the same last-month performance decile on the day the position is opened.

The top panel gives summary statistics of the monthly excess returns on value-weighted portfolios of n pairs of

stocks where the position is opened immediately. The last column is a portfolio of all 120 top pairs. The bottom

panel summarizes the performance with 1 day waiting before the position is opened. The statistics are computed

over 200 replications of the bootstrapped sample.

Portfolio Top 5 Top 20 Top 100-120 All 120

Fully Invested

Mean Excess Return -0.00137 -0.00111 -0.00105 -0.00113

Standard deviation 0.05521 0.02295 0.02264 0.01200

Median -0.00192 -0.00153 -0.00156 -0.00162

Committed Capital

Mean Excess Return -0.00083 -0.00077 -0.00089 -0.00091

Standard deviation 0.02635 0.01358 0.01443 0.00760

Median -0.00123 -0.00100 -0.00123 -0.00122

Panel B: Wait 1 day

Fully Invested

Mean Excess Return -0.00177 -0.00004 -0.00154 -0.00156

Standard deviation 0.05966 0.05310 0.02404 0.01213

Median -0.00241 -0.00172 -0.00206 -0.00201

Committed Capital

Mean Excess Return -0.00103 -0.00094 -0.00112 -0.00112

Standard deviation 0.02811 0.01363 0.01449 0.00761

Median -0.00152 -0.00118 -0.00148 -0.00139

40

Table 7: Returns to Long and Short Components of Pairs

Monthly risk profile for the long and short positions of the pairs portfolios formed and traded according to the “Wait One Day” rule discussed in the text. The

returns in the bottom half of the table are in exccess of the 30-day T-bill returns. The risk adjustment includes the Fama-French factors, as well as Momentum

and the Reversal factors discussed in the text. The t-statistics are computed using Newey-West standard errors with 6 lags. Absolute kurtosis is reported.

Portfolio top 5 top 20 20 after top 100 All

Long Short Long Short Long Short Long Short

Portfolio Performance

Mean monthly return 0.01245 0.00501 0.01330 0.00435 0.01458 0.00663 0.01623 0.00908

Standard error (NW) 0.00183 0.00177 0.00179 0.00174 0.00170 0.00165 0.00254 0.00231

Standard deviation 0.03875 0.03259 0.03653 0.03161 0.03437 0.03146 0.05157 0.04601

Monthly serial corr 0.06 0.11 0.09 0.12 0.13 0.12 0.17 0.16

Regression on Market, SMB, HML, Momentum and Reversal Factors

Intercept 0.00103 -0.00442 0.00243 -0.00521 0.00287 -0.00426 -0.00101 -0.00613

t-statistic 0.47 -2.74 1.27 -3.35 2.85 -4.26 -1.32 -5.22

U.S. Equity Risk Prem 0.37415 0.44075 0.47617 0.50772 0.48520 0.56217 0.58571 0.73091

t-statistic 4.60 6.19 6.68 8.44 10.23 9.74 14.99 12.13

SMB -0.16764 -0.12532 -0.06506 -0.06616 0.01120 0.03453 0.22192 0.29271

t-statistic -1.74 -2.00 -0.80 -1.22 0.27 0.95 3.45 3.15

HML 0.48401 0.42661 0.52539 0.48025 0.39451 0.41175 0.25732 0.31134

t-statistic 7.12 6.92 9.36 9.47 9.76 11.54 7.49 8.11

Momentum -0.05430 -0.02625 -0.04456 0.00361 -0.10901 -0.00589 -0.08832 0.09245

t-statistic -1.48 -0.65 -1.21 0.10 -3.53 -0.21 -3.53 2.81

Reversal 0.15854 0.05662 0.09601 0.02365 0.17288 0.07830 0.39208 0.19131

t-statistic 1.83 1.13 1.26 0.55 4.33 1.90 10.44 2.98

R-square 0.39 0.41 0.48 0.49 0.78 0.75 0.97 0.93

41

Table 8: Subperiod Analysis

Panel A: pre-1989

Monthly risk profile for portfolios of pairs formed and traded according to the “Wait One Day” rule discussed in

the text, over the two sub-periods 7/1963-12/1988 (Panel A) and 1/1989- 12/2002 (Panel B). The “top n” portfolios

include the n pairs with least distance measures, and the portfolios” 20 after top 100" has the pairs after the top

100 pairs. The average number of pairs in the all-pair portfolio is 2057. The t-statistics are computed using

Newey-West correction with six lags for the standard errors.

Portfolio top 5 top 20 20 aft. top 100 All Factor sd

Wait One Day Portfolio Performance

Mean Excess Return 0.01034 0.01181 0.01052 0.00992

Standard Deviation 0.02259 0.01689 0.01527 0.01651

Regression on Fama-French Factors

Intercept 0.00353 0.00670 0.00710 0.00446

t-statistic 1.72 4.41 6.54 4.19

U.S. Equity Risk Premium -0.43395 -0.31200 -0.28946 -0.43429 0.04580

t-statistic -4.29 -3.57 -3.80 -7.26

SMB: Small minus Big -0.44181 -0.33193 -0.27508 -0.40184 0.02923

t-statistic -3.75 -2.86 -2.86 -5.76

HML: High minus Low Book to Market 0.03568 0.03162 -0.06840 -0.06983 0.02597

t-statistic 0.64 0.73 -1.84 -1.97

Momentum 0.01291 -0.01630 -0.07689 -0.15848 0.03506

t-statistic 0.29 -0.50 -2.92 -5.01

Reversal 0.43575 0.33274 0.28765 0.45222 0.07228

t-statistic 4.44 3.62 4.05 7.98

R-square 0.19 0.23 0.26 0.70

Panel B: post-1988

Portfolio top 5 top 20 20 aft. top 100 All Factor sd

Wait One Day Portfolio Performance

Mean Excess Return 0.00217 0.00375 0.00327 0.00212

Standard Deviation 0.01660 0.00987 0.01121 0.01295

Regression on Fama-French Factors

Intercept 0.00337 0.00417 0.00363 -0.00065

t-statistic 2.12 3.77 3.01 -0.56

U.S. Equity Risk Premium 0.07339 0.04804 0.00241 -0.06958 0.04390

t-statistic 1.68 1.81 0.06 -2.43

SMB: Small minus Big -0.00400 0.02888 0.02332 -0.06063 0.03856

t-statistic -0.10 1.32 0.69 -2.83

HML: High minus Low Book to Market 0.03441 0.01412 0.01830 -0.08202 0.03641

t-statistic 0.61 0.45 0.50 -2.65

Momentum -0.00424 -0.02266 -0.08840 -0.12670 0.04926

t-statistic -0.13 -1.29 -4.17 -7.59

Reversal -0.06259 -0.01727 0.02404 0.18808 0.04448

t-statistic -1.76 -0.67 0.54 5.56

R-square 0.02 0.04 0.19 0.64

42

Table 9: Robustness to Short-selling Costs

Summary statistics of the monthly excess returns on portfolios of pairs. We trade according to the rule that opens a

position in a pair when the prices of the stocks in the pair diverge by two historical standard deviations. Panel B

reports the summary statistics for the rule that waits one-day before opening and closing the position. The “top n”

portfolios include the n pairs with least distance measures, and the portfolios ”20 after top 100" has the pairs after

the top 100 pairs. The average number of pairs in the all-pair portfolio is 2057. There are 474 monthly

observations, from 7/1963 until 12/2002. The t-statistics are computed using the Newey-West standard errors with

6-lag correction. Absolute kurtosis is reported.

Portfolio top 5 top 20 20 after top 100 All

Panel A: Top 3 deciles

Mean monthly return (Fully Invested) 0.00835 0.00914 0.00728 0.00690

Standard error (Newey-West) 0.00112 0.00096 0.00075 0.00064

t-statistic 7.47 9.48 9.66 10.85

Excess return distribution

Median 0.00800 0.00793 0.00614 0.00508

Standard deviation 0.01904 0.01464 0.01579 0.01294

Kurtosis 0.02 0.98 0.80 1.77

Skewness 7.01 8.20 6.46 11.67

Minimum -0.10894 -0.05646 -0.03803 -0.02680

Maximum 0.10144 0.10003 0.10642 0.10478

Observations with excess return < 0 32% 26% 33% 27%

Mean excess return on committed capital 0.00493 0.00514 0.00462 0.00401

Panel B: Short Recalls on High Volume

Mean monthly return (Fully Invested) 0.00619 0.00854 0.00665 0.00585

Standard error (Newey-West) 0.00115 0.00094 0.00085 0.00098

t-statistic 5.39 9.07 7.81 5.95

Excess return distribution

Median 0.00592 0.00703 0.00531 0.00312

Standard deviation 0.02210 0.01488 0.01505 0.01555

Skewness 0.15 0.48 1.03 2.82

Kurtosis 5.39 6.00 7.39 18.98

Minimum -0.09611 -0.06594 -0.04408 -0.03465

Maximum 0.10504 0.07255 0.10084 0.14173

Observations with excess return < 0 39% 26% 34% 37%

Mean excess return on committed capital 0.00304 0.00347 0.00296 0.00202

43

Figure 1

Daily normalized prices: Kennecott and Uniroyal (pair 5)

Trading period August 1963 - January 1964

44

Figure 2

Monthly excess returns of Top 20 pairs portfolio

May 1963 - December 2002

45

Figure 3

Cumulative excess return of Top 20 pairs and S&P500

May 1963 - December 2002

46

Figure 4

Top 20 pairs and 101-120 pairs portfolios

Rolling 24-month correlation of 5-factor risk-adjusted excess returns

**Pairs Trading: Performance of a Relative Value Arbitrage Rule
**

Evan Gatev Assistant Professor Boston College William N. Goetzmann Edwin J. Beinecke Professor of Finance and Management Studies Yale University K. Geert Rouwenhorst Professor Yale University First draft: June 1998 This version: February 2006

Abstract We test a Wall Street investment strategy, “pairs trading,” with daily data over 1962-2002. Stocks are matched into pairs with minimum distance between normalized historical prices. A simple trading rule yields average annualized excess returns of up to 11 percent for selffinancing portfolios of pairs. The profits typically exceed conservative transaction costs estimates. Bootstrap results suggest that the “pairs” effect differs from previously-documented reversal profits. Robustness of the excess returns indicates that pairs trading profits from temporary mis-pricing of close substitutes. We link the profitability to the presence of a common factor in the returns, different from conventional risk measures.

We are grateful to Peter Bossaerts, Michael Cooper, Jon Ingersoll, Ravi Jagannathan, Maureen O’Hara, Carl Schecter and two anonymous referees for many helpful discussions and suggestions on this topic. We thank the International Center for Finance at the Yale School of Management for research support, and the participants in the EFA’99 Meetings, the AFA’2000 Meetings, the Berkeley Program in Finance and the Finance and Economics workshops at Vanderbilt and Wesleyan for their comments.

1

Introduction Wall Street has long been interested in quantitative methods of speculation. One popular short-term speculation strategy is known as “pairs trading.” The strategy has at least a twentyyear history on Wall Street and is among the proprietary "statistical arbitrage" tools currently used by hedge funds as well as investment banks. The concept of pairs trading is disarmingly simple. Find two stocks whose prices have moved together historically. When the spread between them widens, short the winner and buy the loser. If history repeats itself, prices will converge and the arbitrageur will profit. It is hard to believe that such a simple strategy, based solely on past price dynamics and simple contrarian principles, could possibly make money. If the U.S. equity market were efficient at all times, risk-adjusted returns from pairs trading should not be positive. In this paper, we examine the risk and return characteristics of pairs trading with daily data over the period 1962 through December 2002. Using a simple algorithm for choosing pairs, we test the profitability of several straightforward, self-financing trading rules. We find average annualized excess returns of about 11 percent for top-pairs portfolios. While pairs strategies exploit temporary components of stock prices, we show that our profits are not caused by simple mean reversion as documented in the previous literature. We examine the robustness of our results to a wide variety of risk factors – including not only the widely used factors in the empirical literature but also potential low-frequency institutional factors such as bankruptcy risk. In addition, we explore the robustness of our results to microstructure factors such as the bid-ask bounce, short-selling costs, and transactions costs. While some factors such as short-selling and transactions costs affect the magnitude of the excess returns, pairs trading remains profitable for reasonable assumptions over the sample period of study, as well as over a true out of sample test

2

of four years. We interpret the results of our analysis as evidence in favor of profitable arbitrage in expectations that may accrue to market participants who possess relatively low transactions costs and the ability to short securities. We also find evidence that points to a systematic factor that influences the profitability of pairs trading over time. This unidentified latent risk factor has been relatively dormant recently. The importance of this risk factor is correlated with the returns to pairs trading, which is consistent with the view that the profits are a compensation to arbitrageurs for enforcing the “Law of One Price.” We argue that our results reveal something about the mechanism and performance of relative-price arbitrage activities in practice. This is potentially useful to researchers, because, despite considerable theory about market efficiency, economists have little empirical information about how efficiency is maintained in practice. In addition, despite the fact that hedge funds have attracted an increasing amount of investment capital over the past decade, the study of hedge fund strategies is in its infancy in the financial economics literature. This paper examines the risk and return characteristics of one widely practiced active trading strategy. One natural question to ask is whether our results imply a violation of equilibrium asset pricing. While the documented profitability of the pairs trading rule is a robust result, it is not inconsistent with all pricing models. Indeed the reversion in relative values we find is consistent with a pricing model in prices developed and tested by Bossaerts (1988). Thus our paper at the very least suggests that this class of models merits further empirical investigation. The remainder of the paper is organized as follows. Section 1 provides some background on pairs trading strategy. The next section describes our methodology of constructing pairs and calculating returns. The empirical results are described in section 3, and section 4 provides conclusions and directions for future research.

3

He claims. pairs trading has since become an increasingly popular "market-neutral" investment strategy used by individual and institutional traders as well as hedge funds. David Shaw.E. Tartaglia's programs identified pairs of securities whose prices tended to move together. In a New York Times interview.1” Could pairs traders be the disciplined investors taking advantage of the undisciplined over-reaction displayed by individual investors? This is at least one possible – albeit psychological – explanation for our results. the Wall Street quant Nunzio Tartaglia assembled a team of physicists. p. 4 . that took the intuition and trader's “skill” out of arbitrage and replaced it with disciplined. Although the Morgan Stanley group disbanded in 1989 after a couple of bad years of performance. mathematicians and computer scientists to uncover arbitrage opportunities in the equities markets. executable through automated trading systems. Among other things. Background of Pairs Trading 1.1 History In the mid-1980's. suggests that recent pickings for quant-shops have become slim – he attributes the success of his firm.204. head of one of the most successful modern quant shops and himself an early Tartaglia’s protégé. S. Tartaglia's own explanation for pairs trading is psychological. The increased popularity of quantitative-based statistical arbitrage strategies has also apparently affected profits.” Institutional Investor. 1989. that “…Human beings don't like to trade against human nature. “Inside Morgan Stanley's Black Box. consistent filter rules.1. Tartaglia's group of former academics used sophisticated statistical methods to develop high-tech trading programs. which wants to buy stocks after they go up not down. to early entry into the business.. May. They traded these pairs with great success in 1987 – a year when the group reportedly made a $50 million profit for the firm. 1 Hansell. D. Shaw.

which might not always cover transactions costs even when stock prices converge. using data through the end of 1998. but rather we have interpreted practitioner description of pairs trading as straightforwardly as possible. we found that over the 1999-2002 2 Froot and Dabora (1999) consider “twin” stocks that trade in different international markets to examine issues of market integrationon. Our rules follow the general outline of first “find stocks that move together. As with all filter rules using historical asset pricing data. but the post 1988 data. We put positions on at a two-standard deviation spread. the danger in data-snooping refinements outweighs the potential insights gained about the higher profits that could result from learning through testing2. 5 .” and second “take a long-short position when they diverge and unwind upon convergence.2 Data Snooping and Market Response In our study we have not searched over the full strategy space to identify successful trading rules. How do you identify “stocks that move together?” Need they be in the same industry? Should they only be liquid stocks? How far do they have to diverge before a position is put on? When is a position unwound? We have made some straightforward choices about each of these questions. Although it is tempting to try potentially more profitable schemes. We completed and circulated the first draft of the working paper in 1999. Using the original model. data-snooping is a potential concern. One approach to the data snooping issue is to test the results out-of-sample. 1. The time lag between the first analysis and the present study gives us an ideal hold-out sample.which is consistent with Jegadeesh and Titman's (1995) finding that contrarian profits are in part due to over-reaction to company-specific information shocks rather than price reactions to common factors.” A test requires that both of these steps must be parameterized in some way.

Chen 6 . despite curiosity from the professional sector. 1. Ingersoll (1987) defines the LOP as the “proposition . two securities with the same prices in all states of the world should sell for the same amount.82 – consistent with the long-term. Absolute pricing values securities from fundamentals such as discounted future cash flow. over which pairs-trading was known to be popular. This is a notoriously difficult process with a wide margin for error. but not necessarily arbitrage or profitable speculation. Relative pricing is only slightly easier. Papers by Bakshi and Chen (1997) and Lee et al. but it also suggests that the public dissemination of the results has apparently not affected the general risk and return characteristics of the strategy. Thus... We were careful not to adjust our strategy from the first draft to the current draft of the paper. Relative pricing means that two securities that are close substitutes for each other should sell for the same price – it does not say what that price will be.” In other words. to avoid data-snooping criticisms. in-sample results of our original analysis. (1997).4 percent per annum. The Law of One Price [LOP] – and a “near-LOP” is applicable to relative pricing – even if that price is wrong.period. relative pricing allows for bubbles in the economy. that two investments with the same payoff in every state of nature must have the same current value. Not only does this additional four-year sample suggest that the results were not simply an artifact of the earlier sample period. for example.8% and a large and significant Newey-West-adjusted t-statistic of 4. the excess return of the fully invested portfolio of the top twenty pairs averaged 10.3 Relative Pricing Asset pricing can be viewed in absolute and relative terms. with an annual standard deviation of 3. are heroic attempts to build quantitative value investing models.

or in Chen and Knez’ case. We then trade pairs whose prices closely match in historical state-space. and subject to bounds on prices for unusual states. then the prices of the component portfolios would be co-integrated and the pairs trading 7 . Notice that this theory corresponds to the desire to find two stocks whose prices move together as long as we can define states of nature as the time-series of observed historical trading days. the current study can be viewed as a test of the LOP and near-LOP in the U. We use an algorithm to choose pairs based on the criterion that they have had the same or nearly the same state prices historically. since the LOP suggests that in an efficient market. In this framework.S. but not necessarily matching payoffs across states should have similar prices. We could not have the first effect without the second. under certain stationarity conditions. This is similar in spirit to Bossaerts’ (1988) test of co-integration of security prices at the portfolio level. near integrated markets. their prices should be nearly identical.4 Co-integrated Prices The pairs trading strategy may be justified within an equilibrium asset-pricing framework with non-stationary common factors like Bossaerts and Green (1989) and Jagannathan and Viswanathan (1988). equity markets. We are effectively testing the integration of very local markets – the markets for specific individual securities. 1. however it allows the examination of “nearefficient” economies. This is of course a weaker condition.” They argue that two securities with similar. We further conjecture that the marginal profits to be had from risk arbitrage of these temporary deviations are crucial to the maintenance of first-order efficiency.and Knez (1995) extend this to argue that “closely integrated markets should assign to similar payoffs prices that are close. If the long and short components fluctuate with common non-stationary factors.

Assume also that pit is weakly dependent after differencing once. More specifically. Thus. 8 . We would like to keep the notion of the empirically observed co-movement of prices. r such that zq = $q`pt are weakly dependent. rather it says that certain assets are weakly redundant. is the basic space for the pairs trading strategies in this paper. without unnecessarily restrictive assumptions. Suppose that prices obey a statistical model of the form. On the other hand. in the sense of Engle and Granger (1987) and Bossaerts (1988). The main observation about our motivating models of the CAPM-APT variety is that they are known to imply perfect collinearity of prices. Under these assumptions. there exist r linearly independent vectors {$q}q=1. In other words. r linear combinations of prices will not be driven by the k common non-stationary components pl. which is readily rejected by the data. k < n (1) where # it denotes a weakly dependent error in the sense of Bossaerts (1988). the price vector pt is cointegrated of order 1 with co-integrating rank r = n-k . p it = ! " il p lt + # it .strategy would be expected to work. Bossaerts (1988) finds evidence of price co-integration for the US stock market. Evidence of exposures to common non-stationary factors would support a non-stationary factor pricing framework. hence we proceed in the spirit of the co-integrated prices literature.. cumulative total returns with dividends re-invested. cum-dividend prices.e. our matching in price space can be interpreted as follows. Note that this interpretation does not imply that the market is inefficient. i. so that any deviation of their price from a linear combination of the prices of other assets is expected to be temporary and reverting. The space of normalized.

the source of the profit is the improving ex-post (non)realization of idiosyncratic bankruptcy risk in the long (loser) stock. However. Our strategy relies upon exactly this conclusion. which are not de-facto co-integrated. In such case. of stocks. expected returns are driven by a few. In principle one could construct trading strategies with trios. while the second may have a temporarily increasing probability of bankruptcy (long end). quadruples. see Connor and 9 . Given the large universe of stocks. e. we need to assume that for n » k. In that case the sum or difference of scaled prices will be reverting to zero and a trading rule could be constructed to exploit the expected temporary deviations. there are co-integrating vectors which have only two nonzero coordinates. etc. The “surprise improvements” in the short end are then followed by improvement in the long end if that stock survives. The assumption that a linear combination of two stocks can be weakly dependent may be interpreted as saying that a co-integrating vector can be partitioned in two parts. namely k common factors.g. Sensitivity of the pairs trading to the default premium suggests that the strategy may work because we are pairing two firms. In other words. the first of which may have a constant or decreasing probability of bankruptcy (short end). This paradigm is supported by existing empirical work.To interpret the pairs as co-integrated prices. we would expect to have asymmetry in the profits from the 3 Note that the case n » k corresponds to the standard finance paradigm where in the large universe of n stocks. which would presumably capture more co-integrated prices and would yield better profits. this statement is always empirically valid and provides the basis of our formation procedure3.5 Bankruptcy Risk The risk of bankruptcy is one reason why the returns on individual securities cannot be taken as stationary. 1. it is important to recognize the possibility of spuriously correlated prices. such that the two corresponding portfolios are priced within a weakly dependent error of another stock.

Financial and Industrials. we construct a cumulative total return index for each stock over the formation period. where we restrict both stocks to belong to the same broad industry categories defined by Standard & Poors: Utilities. We use this approach because it best approximates the description of how traders themselves choose pairs. we screen out all stocks from the CRSP daily files that have one or more days with no trade.” In addition to “unrestricted” pairs. 10 .1 Pairs Formation In each pairs-formation period. This serves to identify relatively liquid stocks as well as to facilitate pairs formation. 2. Methodology Our implementation of pairs trading has two stages.long and the short components. 2. Next. where price includes reinvested dividends. eree for this example. We test long and short positions separately to see if this is driving our results. Each stock is assigned Korajczyk (1993) for references. Interviews with pair traders suggest that they try to find two stocks whose prices “move together. we will also present results by sector. Transportation. We form pairs over a twelve-month period (formation period) and trade them in the next six-month period (trading period). Pairs are thus formed by exhaustive matching in normalized daily “price” space. We then choose a matching partner for each stock by finding the security that minimizes the sum of squared deviations between the two normalized price series. with most of the profits coming from the long end4. Both twelve months and six months are chosen arbitrarily and have remained our horizons since the beginning of the study. which generally finds less than ten common non-stationary components.

Figure 1 illustrates the pairs trading strategy using two stocks. Over that interval. On the day following the last day of the pairs formation period. 2. we base our rules for opening and closing positions on a standard deviation metric. and close the position when the prices have reverted. we begin to trade according to a pre-specified rule. Following practice. The minimum-distance criterion is then used to match stocks within each of the groups. We open a position in a pair when prices diverge by more than two historical 11 . This last set is valuable because most of the top pairs share certain characteristics. Kennecott and Uniroyal in the six-month period starting in August of 1962. in the seventh trading day of the period and then remains open until day 36. although this is not always the case. The prices remain close during the period and cross frequently. We select trading rules based on the proposition that we open a long-short position when the pair prices have diverged by a certain amount.to one of these four groups. and the bottom line indicates the opening and closing of the strategy on a daily basis. based on the stock’s SIC code. we study the top 5 and 20 pairs with the smallest historical distance measure. It is clear why these two firms paired with each other. convergence occurs in the final day of the period. They Notice that the position first opens generally tended to move together over the trading interval. which will be described in detail below. The top two lines represent the normalized price paths with dividends re-invested.2 Trading Period Once we have paired up all liquid stocks in the formation period. Neither stock is the “leader. however not always in the same direction. The pair opens five times during the period. the spread actually first increased significantly before convergence.” In our example. in addition to the 20 pairs after the top 100 (pairs 101-120).

Therefore. We report the payoffs by going one dollar short in the higher-priced stock and one dollar long in the lower-priced stock. or the last available price5. we close the position in that pair. To address this potential concern we have re-estimated our results under the extreme assumption that a only a long stock experiences a -100% return when it is delisted.32% with a standard deviation of 1. this extreme assumption biases the results against profitability. as estimated during the pairs formation period. they can have multiple positive cash flows during the trading interval. The excess return on a pair The profits are robust with respect to this delisting assumption. We unwind the position at the next crossing of the prices. the payoffs to pairs trading strategies are a set of positive cash flows that are randomly distributed throughout the trading period.standard deviations.3 Excess Return Computation Because pairs may open and close at various points during the six-month trading period. Because the trading gains and losses are computed over long–short positions of one dollar. For each pair we can have multiple cash flows during the trading interval. Pairs that open and converge during the trading interval will have positive cash flows. A potential problem arises if inaccurate and stale prices exaggerate the excess returns and bias the estimated return of a long position in a plummeting stock. or we may have none in the case when prices never diverge by more than two standard deviations during the trading interval. 2. pairs trading remains profitable under this alternative: for example. Since selective loss on the long position always harms the pair profit. If a stock in a pair is delisted from CRSP. If prices do not cross before the end of the trading interval. However. This zero-price extreme includes among other things. the calculation of the excess return on a portfolio of pairs is a non-trivial issue.9%. the payoffs have the interpretation of excess returns. gains or losses are calculated at the end of the last trading day of the trading interval. and a set of cash flows at the end of the trading interval which can either be positive or negative. the average monthly return on the top-20 pairs portfolio is 1. using the delisting return. Because pairs can reopen after initial convergence. the possibility of non-trading due to lack of liquidity. 5 12 . Pairs that open but do not converge will only have cash flows on the last day of the trading interval when all positions are closed out.

t ri. the long and short portfolio positions are marked-to-market daily.t !i%P wi. and the daily returns are compounded in order to obtain monthly returns.during a trading interval is computed as the reinvested payoffs during the trading interval6. rP.t = -------------!i%P wi. because it implicitly assumes that all cash earns zero interest rate when not invested in an open pair.t (2) wi. i. we still include a dollar of committed capital as the cumulative return in our calculation of excess return. which are needed to estimate pairs for the strategy starting in the first month.t-1(1+ri.e. and understates the computed excess returns. The daily returns on the long and short positions are calculated as value-weighted returns in the following way.t = wi. the return on actual employed capital. Because any cash flow during the trading interval is positive by construction. In particular. To the extent that hedge funds are flexible in their sources and uses of funds. We consider two measures of excess return on a portfolio of pairs: the return on committed capital and the fully-invested return. The former measure of excess return is clearly more conservative: if a pair does not trade for the whole of the trading period. with the exception of the first twelve months. 13 .t-1) = (1+ri. We initiate the pairs strategy by trading the pairs at the beginning of every month in the sample period. The former scales the portfolio payoffs by the number of pairs that are selected for trading. computing excess return relative to the actual capital employed may give a more realistic measure of the trading profits. It takes into account the opportunity cost of hedge funds of having to commit capital to a strategy even if the strategy does not trade.1)… (1+ri. We correct for the correlation induced by overlap by averaging 6 This is a conservative approach to computing the excess return.t-1) (3) where r defines returns and w defines weights. it ignores the fact that these cash flows are received early. the latter divides the payoffs by the number of pairs that open during the trading period. The result is a time series of overlapping six-month trading period excess returns. This has the simple interpretation of a buy-and-hold strategy.

There are diversification benefits from combining multiple pairs in a portfolio. Interestingly. and a portfolio of the twenty best pairs 1.5 we will consider sector-neutral pairs strategies. as the number of pairs in the strategy increases the minimum realized return increases.31% (t-statistic = 8. the portfolio standard deviation falls.84). while the maximum realized excess return remains relatively stable.monthly returns across trading strategies that start one month apart as in Jegadeesh and Titman (1993). gives excess returns of 0. The resulting time series has the interpretation of the payoffs to a proprietary trading desk.78% and 0.81% per month respectively. The first row shows that a fully-invested portfolio of the five best pairs earned an average excess monthly return of 1. using committed capital.1 Strategy Profits Table 1 summarizes the excess returns for the pairs portfolios that are unrestricted in the sense that the matching stocks do not necessarily belong to the same broad industry categories. compared 14 . During the full sample period of 474 months. In section 3. Panel A summarizes the excess returns of pairs strategies when positions are opened at the end of the day that prices diverge. a portfolio of 20 pairs experienced 71 monthly periods with negative payoffs. The diversification benefits are also apparent from the range of realized returns.44% per month (t = 11. which delegates the management of the six portfolios to six different traders whose formation and trading periods are staggered by one month. and closed at the end of the day of price convergence. and suggest that pairs trading is profitable. 3. Using the more conservative approach to computing excess returns. these excess returns are large in an economical and statistical sense. Empirical Results 3.56). The remainder of Panel A provides information about the excess return distributions of pairs portfolios. Either way. As the number of pairs in a portfolio increases.

While the excess returns remain significantly positive. The average excess returns on the fully-invested portfolios and on committed capital drop by about 30-55 and 20-35 basis points respectively. The decrease in the standard deviation. and the increase of the lower end of the return distribution is also reflected in an increased skewness coefficient. It is difficult to quantify which portion of the profit reduction is due to bid-ask bounce and which portion stems from true mean reversion in prices due to rapid market adjustment. the drop in excess returns suggests that a non-trivial portion of the profits in Panel A may be due to bid-ask bounce. In particular. the winner’s price is more likely to be an ask quote and the loser’s price a bid quote. 15 . the returns may be biased upward due to the bid-ask bounce (Jegadeesh (1990).3. The opposite is true at the second crossing (convergence): part of the drop in the winner’s price can reflect a bid quote. Jegadeesh and Titman (1995). the remainder of the paper will report results for pairs strategies that open (close) on the day following divergence (convergence). Panel B of Table 1 provides the excess returns when we initiate positions in each pair on the day following the divergence and liquidate on the day following the crossing. Since pairs trading is in essence a contrarian investment strategy. We will return to a detailed discussion of this issue in section 3. and part of the rise of the loser’s price an ask quote. Conrad and Kaul (1989)). this difference raises questions about the economic significance of our results when we include transactions costs. To address this issue. Unless stated otherwise. Part of any observed price divergence is potentially due to price movements between bid and ask quotes: conditional on divergence. None-the-less. our strategy sells stocks that have done well relative to their match and buys those that have done poorly. In Panel A we have used these same prices for the start of trading and our returns may be biased upward due to the fact that we are implicitly buying at bid quotes (losers) and selling at ask quotes (winners).to 124 months for a portfolio of 5 pairs.

This indicates that pairs trading – implemented according to the particular rules we chose – is a medium-term investment strategy. and 91% come from the top five size deciles. and on average more than once. On 7 The optimal trigger point in terms of profitability may actually be much higher than 2 standard deviations. The second line of panel A in Table 2 reports the average price deviation of the 2 standard deviation trigger. What are the characteristics of the stocks that are matched into pairs? How often does a typical pair trade? Because pairs trading is an active investment strategy. As mentioned before we use a two standard deviation trigger to open a pairs position.75 months. Panel B of Table 2 describes the composition of the pairs in terms of market capitalization and industry membership. 74% of the stocks in the top 20 pairs belong to the top three size deciles using CRSP breakpoints. it is important to evaluate the profitability relative to the trading intensity of the portfolios. The next lines of Panel A also shows that on average almost all pairs open during the six-month trading period. on average 4.2 Trading Statistics and Portfolio Composition Table 2 provides a summary of the trading statistics and composition of the pairs portfolios.e. This is a relatively narrow gap in prices7. The trigger spread increases with the number of pairs in the portfolio. and the average number of round trips per pair is 2. The average duration of an open position is 3. the average stock in the top 5 and top 20 pairs belongs to the second and third deciles from the top. The remainder of Panel B gives a breakdown of the pairs by industry composition. For the top five pairs.02. “size mixed pairs”). In terms of size. About two-thirds of the pairs combine stocks from different size deciles (i. because the standard deviation of the prices increases as the proximity of the securities in price space decreases.81 open during the trading period. Of the top 5 pairs. the position typically opens when prices have diverged by 4. and the stocks in mixed pairs differ on average by a single decile.76% or more.3. although we have 16 .

While actual transactions costs may be different. 3. or whether pairs strategies are also profitable in other sectors of the market.44%. Given the predominance of utilities among the top pairs. 71% of the stocks in the top 20 pairs are utility stocks. and the loser at the ask.5. The percentage of mixed sector pairs ranges from 20 per cent for the top 5 pairs to 44 per cent for pairs 101-120. This drop in the excess returns implies an estimate of the average bid-ask spread and hence the transactions costs of trading in the sample.average. In this not experimented to find out. it is informative to know whether the trading profits are large enough to survive this estimate of transactions costs. This is not surprising perhaps. If the next day prices are equally likely to be at bid or ask the delaying trades by one day will reduce the excess returns on average by half the sum of the spreads of the winner and the loser. If at the second crossing (convergence) of the pairs the winners is trading at the bid. Suppose the extreme case where the prices of the winner at the first crossing (divergence) are ask prices and the loser are bid prices.3 Transactions costs Table 1 shows that the average monthly excess return of unrestricted pairs strategies falls from 1. it is fair to ask whether the profitability of pairs trading profitability is limited to the utility sector. 17 . to 0. for the top-20 portfolio. waiting one day will reduce the excess returns on average again by one half of the sum of the bid-ask spreads of both stocks. We address this question in section 3. because utility stocks tend to have lower volatility and tend to be correlated with interest rate innovations.90% per month if we postpone the trades to the day following the crossing. The strategy does not always match stocks within sectors. despite the fact that Utilities represent a fairly small proportion of the stocks in the whole sample.

the drop in the excess returns of 324 basis points per six-months by waiting one day reflects the cost of 2 round-trips. This gives average net profits ranging from 113 to 225 bp over each six-month period. If the prices used to compute these excess returns are equally likely to be at bid or ask.extreme case. we have to subtract the round-trip trading costs to get an estimate of the profits after transactions costs. and is consistent with the trading costs estimated by Keim and Madhavan (1997). we have to correct these excess returns to reflect that in practice we buy at the ask and sell at the bid prices. Since 91% of the stocks in the top 20 pairs belong to the top 5 deciles of CRSP stocks. Comparing these profits to the reported standard errors. who find that the average effective spread for stocks in the CRSP database in 1991 was 37 basis points. waiting a day before trading reduces the return on each pair by the round-trip transactions costs in that pair. This indirect estimate is higher than the transactions costs reported by Peterson and Fialkowski (1994). which implies a transactions costs of 162 bp per pair per round-trip. The effective spread for the all-pairs portfolio is 70 bp. Because we trade each pair on average 2 times during the sixmonth trading interval. Further analysis is required to get more precise estimates of influence of transactions costs of pairs trading strategies. An important question in this context is whether the trading rule that we have used to open and close pairs can be expected to generate economically significant 18 . This may be interpreted as an estimated effective spread of 81 bp. we conclude that they are both economically and statistically significant. In other words. which seems a reasonable assumption. it is possible that the effective spread is even lower that 37 bp. Our conservative estimate of transactions costs of 162 bp times 2 rounds trips per pair results in an estimate of 324 bp transactions cost per pair per six-month period. Do our trading strategies survive these transactions costs? The profits on our trading strategies in Table 1 range from 437 to 549 basis points over a six-month period.

profits even if pairs trading works perfectly. As a consequence. The convergence has to be substantial in order to overcome the transactions cost that will be incurred when we close out the position on the next day (the last day of the trading interval). For example. Are the profits to pairs trading consistent across sectors? We examine the returns on pairs trading where stocks are matched only within the four large sector 19 . Results that are not reported here suggest that this is indeed the case for some of our pairs. We open pairs at any point during the trading period when the normalized prices diverge by two standard deviations. the fraction of mixed pairs is typically well below 50 percent. we may simply be opening pairs “too soon” and at a point that we cannot expect it to compensate for transactions costs even if the pair subsequently converges. firms that are in industries where crosssectional differences in factor exposures are small or return variances are low are more likely to end up among the top ranking of pairs. There is a second reason why our trading strategies require “too much” trading. This is not a sensible rule towards the end of a trading interval. For this reason it is perhaps not surprising that many of the top pairs match two utilities. and since this estimated standard deviation is the smallest among all pairs. and a utility with a bank stock. Common factor exposures of stocks in the same industry will make it more likely to find a match within the same sector. A computer stock has the opportunity to match with a steel firm. Because we use a measure of historical standard deviation to trigger the opening of pairs. suppose a divergence occurs at the next to last day of the trading interval. Unreported results suggest that this is also an important source of excess trading. Also. As shown in Table 2. it is likely to underestimate the true standard deviation of a pair. 3.4 Pairs trading by industry group The pairs formation process thus far has been entirely mechanical. This does not mean that these matches are likely.

61% (NW t = 6.26). Goetzmann et al. The top part of Table 4 compares the excess return to pairs trading to the excess return on the SP500. As in Table 1. Table 4 compares the risk premium of pairs trading to the market premium (SP500). the average excess return to pairs trading has been about twice as large as the excess return of the SP500.08% (Newey-West t =10. The monthly excess returns for the top 20 pairs are the largest in the Utilities sector. and not limited to a particular sector. Financials. (2002) show that Sharpe 20 . the pairs are traded with a one-day delay before opening and closing a position in order to minimize the effect of the bid-ask bounce on trading. with 1.78% (NW t = 7. To provide further perspective on the risk of pairs trading. with the average Transportation. 0. Financials.60) and 0. 3. The results are summarized in Table 3. and Industrials top 20 pairs earning 0. the Sharpe Ratios of pairs trading are between 4 and 6 times larger than the Sharpe Ratio of the market.58% (NW t = 4. The profits for the other industry groups are somewhat lower. Table 3 also gives a more detailed picture of the return distributions and trading characteristics of the pairs trading strategies by sector. It shows that the excess return distributions of the sector pairs portfolios are generally skewed right and exhibit positive excess kurtosis relative to a normal distribution. with only one half to one third of the risk as measured by standard deviation.groupings used by Standard and Poor’s: Utilities. and reports the risk-adjusted returns to pairs trading using two different models for measuring risk. As a result. but all statistically significant. Transportation. The conclusion from these Tables is that pairs trading is profitable in every broad sector category. Between 1963 and 2002. Table 5 summarizes value-at-risk measures for pairs portfolios. and Industrials.5 The Risk Characteristics of Pairs Trading Strategies.26).93) respectively over a one-month period.

but the performance-sorting horizon here is one month. the exposures to the market are small and with one exception insignificant. we expect the exposures of pairs portfolios to be positive to the reversal factor. This is unlikely to be a concern for our study. In order to explore the systematic risk exposure of the pairs portfolios. the pairs excess returns will be negatively correlated with momentum. To examine this possibility. and buy short-term losers. Exposures to the other two FF factors. The second additional factor controls for exposure to mediumterm return continuation (Jegadeesh and Titman (1993)). Table 4 shows that only a small portion of the excess returns of pairs trading can be attributed to their exposures to the five risk factors. 21 . because our Table 1 showed that the returns to pairs portfolios are positively skewed. For example.Ratios can be misleading when return distributions have negative skewness. Because pairs strategies are market-neutral. we include a momentum factor in our risk model that is constructed along the lines of Carhart (1997). which – if anything – would bias our Sharpe ratios downward. augmented by two additional factors. and lower than the raw excess returns by about 10-20 bp per month. The motivation for the additional factors is that pairs strategies invest based on the relative strength of individual stocks. the difference 8 The construction is similar to Carhart’s (1997) momentum factor. It is therefore possible that pairs trading simply exploits patterns in returns that are known to earn significant profits. We control for this possibility by constructing a short-term reversal factor measured as the excess return of stocks in the top 3 deciles of priormonth return minus the return on stocks in the bottom three deciles8. To the extent that pairs trading sells medium-term winners. we regress their monthly excess returns on the three factors of Fama and French (1996). If pairs strategies sell short-term winners. Jegadeesh (1990) and Lehmann (1990) show that reversal strategies that select stocks based on prior onemonth return earn positive abnormal returns. and buys medium-term losers. The intercepts of the regressions show that risk-adjusted returns are significantly positive.

or common factors affecting convergence in both stock and bond markets may be responsible. the U. when long-term spreads decrease. the signs of coefficients are positive across all portfolios. the pairs portfolios seem to have low exposure to various sources of systematic risk. The significance of the risk-adjusted returns indicates that pairs trading is fundamentally different from simple contrarian strategies based on reversion (Jegadeesh (1990) and Lehmann (1990)). interest rate movements also seem to matter for the more broadly diversified pairs portfolios. As can be expected. Although the risk exposures are generally not significant. bond horizon premium. bond default premium and the U. This may be expected since they are constructed in a way that should 22 .S. The exposures to momentum and reversals have the predicted signs. some of the winner stocks that a pairs strategy shorts are short-term winners while others are medium-term winners. There are a range of possible explanations for this pattern.S.between small and big stocks (SMB). Thus. In sum.S. Similarly. indicating that the portfolios are nearly factor-neutral. At first glance.e. The bottom of Table 4 shows the results of regressing the pairs returns on an alternative set of risk factors suggested by Ibbotson: the excess return on the S&P 500. i. when corporate bonds increase in price relative to government bonds. The portfolios also appear to be sensitive to shifts in the yield curve. However. pairs trading is more profitable. the losers are evenly divided between shortterm and medium-term losers. The R-squared of most regressions is low. and more than half are statistically significant. and the difference between value and growth stocks (HML). Relatively cheaper borrowing rates by arbitrageurs may force stock prices closer to equilibrium values. the U. The next sections provide further evidence to support this view. small stock premium. Overall the exposures are not large enough to explain the average returns to pairs trading. the pairs portfolios make money. are not significant and the point estimates alternate in sign. the sensitivity to term-structure measures may be explained by the presence of interest rate sensitive Utility stocks in many of the top pairs.

when the returns were sometimes negative.essentially match up economic substitutes. and then had a span of more modest performance. the pairs portfolios seem to be exposed to relatively little risk. 23 . Panel A shows that the worst monthly loss over the almost 40-year sample period was 12.08 % loss. Pairs trading was very profitable in the 1970's and 1980's.32 % and 1.S. the top five portfolio recorded a 10. the pairs strategy experienced some of its best performance.2% for the top 20 portfolio.24 % and 0. Figure 3 compares the cumulative excess returns of the top 20 one-day-waiting strategy the cumulative excess returns of investment in the S&P 500 index. The smooth index of the pairs trading portfolio contrasts dramatically with the volatility of the stock market. only once in every hundred months did these portfolios lose more than 4. in the mid-90s the market performed exceptionally well. stocks. A five-to-one leverage ratio applied to the top 5 pairs would appear to have been adequate to cover the worst monthly loss in the 40. compared to a 6. Table 5 reports both monthly and daily value at risk [VAR] measures to summarize the quantiles of the empirical distributions of the pairs’ excess returns.6 % for the top five pairs portfolio. but pairs trading profits were relatively flat. On average.72% loss of the top-20 portfolio. once every hundred days did these portfolios lose more than 1. Pairs trading performed well over difficult times for U.65 % respectively.94 % respectively. By contrast. On average. When the U. stock market suffered a dramatic real decline from 1969 through 1980. based upon the next-day trading rule.year period. The VAR is useful because it provides a gauge to the potential leverage that could be applied to these strategies. and 8.S. Figure 2 shows the monthly performance of the top 20 pairs. Panel B shows that on its worst day. Although the lessons of recent history have taught us not to rely too heavily on historical VAR measures for gauging capital needs for exploiting convergence strategies.

it is an example of a contrarian investment strategy. Thus. The TASS hedge fund database reports that hedge fund assets rose from $4 billion in 1977 to $137 billion in 2000. Of the $137 billion in hedge fund assets in 2000.Perhaps after its discovery in the early 1980's by Tartaglia and others. and show that in contrast to the raw returns. $119 billion employed “Market Neutral” “Relative Value” or “Arbitrage” strategies.8 we will revisit this issue. While this database does not capture proprietary trading operations of investment banks. The early part of the analysis represents a period with high fixed commissions. The second explanation may be the rise in hedge funds in the period since 1989. These frictions might have prevented the rapid convergence of relative prices. it is possible that lower transactions costs and large inflows of investment capital to relative-value arbitrage have together decreased the rate of return on pairs trading. competition decreased opportunity. all of which may potentially impact pairs trading operations. The first is the long-term trend in transactions costs over the history of our sample. 3. at least in the simple form we test in this paper. In section 3. Two additional explanations may account for the temporal variation in profitability. the risk-adjusted return to pairs trading has been relatively stable over time. with the most dramatic growth after 1992. it does suggest that risk arbitrage activities have grown significantly in the last decade of our sample period. pair trading might simply be more profitable in times when the stock market performs poorly. The results of Table 4 showed that the returns to pairs are positively correlated with – 24 . This is becoming increasingly so with the introduction of more sophisticated trading technologies and networks. The secular decrease in transactions costs may have attracted more relative value equity arbitrage. On the other hand.6 Pairs trading and contrarian investment Because pairs trading bets on price reversals.

In fact. In particular. we will show that the long and short portfolios that make up a pair do not provide equal contributions to the profitability of the strategy. The conclusion from the simulations confirms the conclusion from the factor regressions. We bootstrapped the entire set of trading dates 200 times. the excess returns to the simulated pairs are slightly negative. 3. which is a reflection of the fact that the simulated pairs are poorly matched. and the standard deviations of the returns are large relative to the true pairs. Similarity is defined as coming from the same decile of previous month’s performance. If pairs trading simply exploits mean reversion.7 Risk and Return of the Long and Short Positions There are at least three reasons to separately examine the returns to the long and short portfolios that make up a pairs position. On average we find that the returns on the bootstrapped pairs are well below the true pairs returns. we conduct a bootstrap where we compare the performance of our pairs to random pairs. These three findings combined strongly suggest that our pairs trading strategy seems to capture temporal variation in returns that is different from simple mean reversion. In each bootstrap we replace the actual stocks with two random securities with similar prior one-month returns as the stocks in the actual pair. First. In addition. The starting point of the bootstrap is the set of historical dates on which the various pairs open. in the next section. The results are summarized in Table 6. The difference between the actual and the simulated pairs returns provides an indication of the portion of our pairs return that is not due to reversion. one would expect 25 . the separate returns provide further insight into the question of mean reversion. In this section we further explore whether our pairs trading strategies are merely a disguised way of exploiting these previously documented negative autocorrelations. that the pairs strategy does not merely reflect onemonth mean reversion.but not explained by – short-term reversals documented by Lehmann (1990) and Jegadeesh(1990).

8 Sub-period Analysis and the Presence of a Dormant Risk Factor Table 8 summarizes the profitability of pairs trading when we spit the sample period at the end of 1988.9. Second. it is unlikely that the returns are driven by a reward for unrealized bankruptcy risk. and insignificantly different from zero for the top-5 and top-20 portfolios. Finally. the excess return of the top-20 strategy drops 9 We thank a referee for suggesting this explanation. which experienced an increase in relative value prior to opening. the alphas of the long portfolio containing the stocks that decreased in value relative to their counterparts are smaller. By contrast. 26 . it becomes important to examine whether short-sale considerations might prevent arbitrageurs from competing away the profits. the risk exposures of the two portfolios can provide further clues at to the reason for the profitability of pairs – or example the possibility that the long and short portfolios have different exposures to common nonstationary risk factors such as bankruptcy risk9. The Table shows that much of the pairs risk-adjusted excess return comes from the short portfolio. The asymmetry of the results provides further evidence that the returns to pairs are not due to simple one-month mean reversion. A comparison between the two top halves of the two panels shows a drop in the raw excess returns to pairs trading. For example. The returns and the risk exposures of the component portfolios are summarized in Table 7. And because much of the abnormal return comes from the short position. if the excess returns are predominantly driven by the short position. The robustness of our results to the cost of short sales will be discussed in section 3. which contains the stocks that have increased in value relative to their counterparts prior to opening of the pair. 3.the abnormal returns to the long and short positions to be equal because the opening of a pair is equally likely to be triggered by either stock.

which shows that the rolling 24-month correlation between the two portfolios was especially high during the pre-1989 period.51 over the profitable pre-1988 period but much lower (0.41 and 3. the correlation is 0.77 respectively). which can account for the lower recent profitability of pairs trading. These correlations are not driven by the 5 “systematic” risk factors we considered.42 and 0.41.from 118 bp per month to about 38 bp per month. The respective sub-period FFMR correlations are 0. Because there is no overlap between the positions of these portfolios. These results suggest that there is common component to the profits of pairs portfolios 27 . because the correlations of the residuals from the factor regressions are very similar to the raw correlations. the correlation between the top-20 Fama-French-Momentum-Reversal (FFMR) residuals and the top 100-120 FFMR residuals is 0. but not the risk-adjusted returns to pairs trading. Changes in the factor exposures and factor volatilities can explain only part of the lower returns in the early part of the sample.20. The full-sample correlation between the excess returns of the top20 and the 100-120 pairs portfolios is 0. Are the positive risk-adjusted returns to pairs trading a general failure of our risk model? In other words.18) over the lower excess return post-1988 period. Moreover.48. are there reasons to believe that the risk adjusted pairs returns are a compensation for an omitted (latent) risk factor? Inspection of the correlation between disjoint pairs portfolios provides some support for this view. Has increased hedge fund activity arbitraged away the anomalous behavior of pairs since the pre-1989 period? Inspection of the risk adjusted returns shows that this is not the case: the average risk-adjusted return of the top-20 portfolio drops by about a third from 67 to 42 basis points per month. This is further illustrated in Figure 4. this latent risk factor seems to have been relatively dormant over the second half of our sample. In particular. but remains significantly positive in both sub-periods (t = 4. the correlation indicates the presence of a common factor to the returns. Moreover.

However. then the pair position is forced to close prematurely and the arbitrageur does not capturing the profit from the pair convergence.that is not captured by our conventional measures of systematic risk. in particular the reward to arbitrageurs for enforcing the “Law of One Price”. and subjected it to a range of controls for risk. 28 . there are explicit short-selling costs in the form of specials. First. These results are consistent with the view that the abnormal returns documented in this paper are indeed a compensation for risk. D’Avolio (2002) argues that short recalls are potentially costly because they may deprive arbitrageurs of their profits. We perform two tests for robustness of the profits that corresponds to these two types of short-selling costs. One possible explanation is that prices diverge on random liquidity shocks that cannot be exploited by professional arbitrageurs due to short-selling costs10. 9 Robustness to Short-selling costs Having identified and back-tested a filter rule. this does not explain why prices drift away from parity in the first place. This opportunity cost is reinforced by D’Avolio’s findings that short recalls are more common as the price declines. For example. and the documented existence of relative-value arbitrageurs in the U. Second. equity market. The common component was stronger during the first half of our sample. (2002) 10 We are grateful to the editor Maureen O’Hara for pointing out this plausible explanation and for suggesting the way to test it. if the short stock is recalled when it starts to converge downward. given the natural arbitrage motivation. 3. the question of the nature of pairs profits remains. which is consistent with higher profits (abnormal returns) than in the second half of our sample when the factor was more dormant. The first test is motivated by the findings of D’Avolio (2002) and Getczy et al. Why do prices of close economic substitutes diverge and converge? The convergence is easier to understand than the divergence.S.

with a Newey-West t-statistic of 9.g. These results show that pairs trading profits are robust to short-selling costs. yet they remain large and positive. For example. (2002) argue that for large traders. large institutions and hedge funds. and subsequent closing of the pair position. The profits decline slightly by 4-13bp per month. can be directly compared to those in Panel B of Table 1. given in Panel A of Table 9.g. the small effects confirm that the profits persist when trading pairs of large stocks as well as when shorts are recalled. Geczy et al. see e. the profits change little and remain highly significant. the top 20 pairs portfolio earns an average of 85bp per month with short recalls. but increase for the top 5 portfolio. The comparison shows that the profits of the top-20 strategy drop by about 2 basis points per month. the pairs trading profits are likely to remain essentially unaffected by potential shorting costs. The second test is motivated by the evidence in Chen et al.that specials have minimal effect of large stocks. Overall. Jones and 29 . e. For better-positioned investors. (2002) and D’Avolio (2002) that short recalls are driven by dispersion of opinion. We use high volume as proxy for divergence of opinion and perform pairs-trading under recalls. High volume days are defined as days on which daily volume exceeds average daily volume over the 18months (both split-adjusted) by more than one standard deviation. This shows that the pairs trading profits are not driven by illiquid stocks that are likely to be on special. where we simulate recalls on the short positions.07.” direct shorting costs of the rebate rate on short sales are low (4 to 15 11 Using large liquid stocks also mitigates the problem of stocks that are hard to short and can be overvalued. Overall. Panel B of Table 9 shows the profits of pairs trading with the high volume recalls. The results in Panel B can be interpreted as an estimate of the opportunity cost of short recalls. Correspondingly. on days with high volume. The results. we test for robustness of profits by trading pairs that are formed using only stocks in the top three size deciles11. who have better access to most stocks at “wholesale prices.

4. These profits are uncorrelated to the S&P 500. We present two pieces of empirical evidence that supports this view. including mean-reversion as previously documented in the literature. known on Wall Street as pairs trading. One view of the lower profitability of pairs trading in recent year is that returns are competed away by increased hedge fund activity.bp/year). The alternative view. Pairs are stocks which are close substitutes according to a minimum distance criterion using a metric in price space. and the inability of arbitrageurs to take advantage of the profits due to short-sale constraints. however they do exhibit low sensitivity to the spreads between small and large stocks and between value and growth stocks in addition to the spread between high grade and intermediate grade corporate bonds and shifts in the yield curve. Second. We find that trading suitably formed pairs of stocks exhibits profits. The main implicit shorting cost stems from limited availability and is relevant mostly for general retail investors. In addition to risk and transactions cost. Conclusion We examine a hedge fund equity trading strategy based on the notion of co-integrated prices in a reasonably efficient market. is that abnormal returns to pairs strategies are a compensation to arbitrageurs for enforcing the “Law of One Price”. while raw returns have fallen the risk-adjusted returns have continued to persist despite increased hedge fund activity. which are robust to conservative estimates of transaction costs. The impact of such potential short sales constraints on the profitability of pairs trading by large investors is mitigated by our use of liquid stocks that trade every day over a period of one year. 30 . we rule out several explanations for the pairs trading profits. our results suggest that the change in risk-adjusted returns of pairs trading Lamont (2002). First. taken in this paper. unrealized bankruptcy risk.

is accompanied by the diminished importance of a common factor that drives the returns to pairs strategies. A further examination of the nature of this common factor and its link to the profitability of pairs trading is an important question for future research. 31 .

J. French. 1995. 131-155. 66. E. C. New Jersey. S.” Journal of Financial Economics 66.” Journal of Finance 51. May. Chen. 32 .. 5782. G. 204. and C. Green. 48.” Review of Financial Studies. 271-306.. “Mean Reversion in Short-horizon Expected Returns. M. “A General Equilibrium Model of Changing Risk Premia: Theory and Evidence. “Multifactor Explanations of Asset Pricing Anomalies. “Measurement of Market Integration and Arbitrage. Rowman and Littlefiled. “Stock Valuation in Dynamic Economies. “Breadth of Ownership and Stock Returns. 2002. “Stocks are Special Too: An Analysis of the Equity Lending Market.. P. Ingersoll.. Froot. Fama.. Hansell. Jr. Dabora.. Engle. 1997.References: Bakshi. Theory of Financial Decision-Making. J. et al. Kaul. Chen. 55. 52.. 287-325. and K.” Journal of Finance. 1989.” Institutional Investor. “On Persistence in Mutual Fund Performance. R. 1996.” Journal of Finance. 12.. K. 1997. Korajczyk.” Review of Financial Studies. “Co-integration and Error Correction: Representation. and Z. Hong and J. 241-269. Estimation and Testing. 1999. D. 8. G.” Journal of Financial Economics. 171-205. 53. “A Test for the Number of Factors in an Approximate Factor Model. 66. 467-493. Connor. 1993. G. 189-216. Bossaerts.” working paper.. and R.” Journal of Economic Dynamics and Control. Conrad.” Journal of Financial Economics. and R. “Sharpening Sharpe Ratios”. 347-364. Stein. Bossaerts. “How are Stock Prices Affected by the Location of Trade?. Goetzmann W. H. 2.. 1989. “The Market for Borrowing Stock. Geczy. “Common Nonstationary Components of Asset Prices. “Inside Morgan Stanley's Black Box. 1987. 1987. D’Avolio. P. and G. and E. 1988. Carhart. Ohio State University. 251-276. 225-240. Knez. 1989. and P. Reed. Granger. working paper. Musto and A. 2. 2002. J. Yale School of Management. Z.” Review of Financial Studies. 2002. 1263-1291. Chen.” Econometrica.” Journal of Financial Economics.

. 8. N.” Journal of Financial Economics. “Overreaction. 1997. N. 1990. and Contrarian Profits. 973-93.” Journal of Finance. “Fads. Delayed Reaction. 66. et al... Lehmann. “Posted versus Effective Spreads: Good Prices or Bad Quotes?. Lamont. 265-292. M.Jagannathan. R. C. “Evidence of Predictable Behavior of Security Returns. and O. Jegadeesh. Peterson. 105. Madhavan.” Journal of Financial Economics. 33 . 65-91. Keim. “What is the Intrinsic Value of the Dow?. Fialkowski. C. Lee. 35. 207-239. Term Structure of Interest Rates and the Small Firm Anomaly. and A.” Review of Financial Studies.. “Linear Factor Pricing. “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Martingales and Market Efficiency. 1994.” Working Paper 57.. 1990. 1995. Viswanathan.” Quarterly Journal of Economics. 881-898. B. 48. D. N. Cornell University. 2002.” working paper. “Transactions Costs and Investment Style: An Inter-exchange Analysis of Institutional Equity Trades.” Journal of Finance. and S. and S. 45. 269-292. Jones. “Short-sale Constraints and Stock Returns. Jegadeesh.. Jegadeesh. 1993. 1997. 46. 1988. and D. Titman.” Journal of Financial Economics. 1-28. Titman. and S. Northwestern University.

02101 0.14716 26% 0.66 -0.54 All 0.00096 9.06629 0.00503 0.00148 8.00728 0.01688 1.34 10.00119 6.01104 0.29 0.00955 0.13295 15% 0.00679 0.01235 0.Table 1: Excess Returns of Unrestricted Pairs Trading Strategies Summary statistics of the monthly excess returns on portfolios of pairs between July 1963 and December 2002 (474 observations).25 -0.16325 32% 0.00784 0.16 Average excess return (Fully Invested) Standard error (Newey-West) t-statistic Excess return distribution Median Standard deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Average excess return on committed capital 0.01194 0.81 -0. The “top n” portfolios include the n pairs with least distance measures.98 7.00520 0.00396 34 .01670 3.34 10.00715 0.54 -0.00795 0. We trade according to the rule that opens a position in a pair at the end of the day that prices of the stocks in the pair diverge by two historical standard deviations (Panel A).01438 0.00094 11.00124 11.00805 0.56 101-120 0.00099 11.13490 23% 0.04266 0.10573 0.26 0.00614 Panel B: Excess Return Distribution (1 day Waiting) Average monthly return (Fully Invested) Standard error (Newey-West) t-statistic Excess return distribution Median Standard deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Average excess return on committed capital 0.12628 0.00463 0.01436 0. and the portfolio “101 Pairs Portfolio top 5 top 20 Panel A: Excess Return Distribution (No Waiting) 0.42 25.00699 0.03857 0.00090 7.01577 3.00690 0.12684 21% 0.00694 0.62 7.02951 0.39 10.01308 0.00411 0.64 -0.01081 0.00895 0. The results in Panel B correspond to a strategy that delays the opening of the pairs position by one day.00085 9.30 -0.08218 0.02721 0. All pairs are ranked according to least distance in historical price space.01540 1.01527 1.17178 17% 0.40 0.92 0.78 -0.32 25.14350 35% 0.13 -0.10464 28% 0.02280 0.00745 0.84 0.45 16.

91 0.41 0.32 0. in months Panel B: Pair Portfolio Composition Average size decile of stocks Average weight of stocks in top 3 size deciles.07560 19.Table 2: Trading Statistics and Composition of Pairs Portfolios Trading statistics and portfolio composition of portfolios of pairs portfolios between July 1963 to December 2002 (474 months).13 0.03 0.97 0.78 0.33 35 .82 0.91 0.05284 19.62 3.30 1.58 0.97 0.57 0.14 0.40 0.80 top 20 0.02 0.69 0.79 0.78 0.22 1.38 All 0.71 0.62 0.76 0.02 0.11 0.73 0.02 0.44 4.20 2.26 0.16 3.74 0. and then traded over the subsequent 6 month period.40 0. of time open.08 0. Pairs are formed over a 12 month period according to a minimum distance criterion.40 3.16 0.27 3.97 0.97 0.75 0.16888 1944.71 0. The “top n” portfolios include the n pairs with least distance measures. per pair. Panel A summarizes the trading characteristics of a pairs strategy. Average weight of stocks in top 5 size deciles Average weight of pairs from different deciles Average decile difference for mixed pairs Average sector weights Utilities Transportation Financials Industrials Mixed sector pairs 2.75 0.02 0.17 Average price deviation trigger for opening pairs Average number of pairs traded per 6-month period Average number of round-trip trades per pair Standard deviation of number of round-trips per pair Average time pairs are open in months Standard dev.41 1.98 0.72 0. Average deviation to trigger opening of pair is the cross-sectional average of two standard deviations which panel B contains information about the size and industry membership of the stocks in the various pairs portfolios.98 0. Pairs Portfolio top 5 Panel A: Trading Statistics 0.66 0.81 2.96 0. and the portfolio “101-120" studies the twenty pairs after the top 100.22 3.54 0.04758 4.45 101-120 0. We trade according to the rule that opens a position in a pair on the day following the day on which the prices of the stocks in the pair diverge by two historical standard deviations.62 0. Pairs are opened when prices diverge by two standard deviations.15 0.

75 -0.00665 0.98 5.02154 0. There is no “20 after 100" portfolio for the Transportation industry group.12347 -0.10638 40% 33% Panel D: Industrial 0.93 0.02640 0.07 0.00547 0.43 4.01472 1.00615 0.16563 0. 371 Financials.27 -0.79 7.00497 0. Portfolio Mean Excess Return t-statistic (NW) Median Standard Deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Mean Excess Return t-statistic (NW) Median Standard Deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Mean Excess Return t-statistic (NW) Median Standard Deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Mean Excess Return t-statistic (NW) Median Standard Deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 top 5 top 20 Panel A: Utilities 0.00339 0.37 10.00 -0.60 0.00557 0.01601 0.19570 -0.12730 28% 19% Panel B: Transportation 0.04786 0.62 0.01825 0.11 0.00397 0.37 6.00775 4.53 5.00678 0.00726 7.00292 0.17211 34% 0.149291 0.44 0.94 7.48 -0.Table 3: Industry Sector Pairs Trading Summary statistics for the excess monthly return distributions for pairs trading portfolios by sector.00854 7.98 18.98 4.97 20.52 0. and 1729 Industrials.50 -0.21 -0.39 12.40 0. 61 Transportation.081988 -0.195797 32% 0.00594 0.12 0.03519 0.07868 0.65 -0.02942 -0.13560 44% 0.12467 0.67 12.08022 42% 36% 0. The t-statistic of the mean is computed using Newey-West standard errors with six lags.00969 0.00938 0.64 21.07810 0.00607 4.10546 0.00664 6.01036 10.95 -0.89 -0. We trade according to the “wait 1 day” rule described in the text.81 0.08750 0.02598 0.16072 34% 36 .100477 0.61 6.02358 1.26 0.12526 0.12878 18% 0.00829 0.10016 -0.66 5. The average number of stocks in the industry groups is as follows: 156 Utilities.02871 0.51 0.11 6.74 -0.03892 0.80 0.00577 2.04452 0.00547 0.01808 2.03431 0.11099 44% 42% Panel C: Financials 0.00550 0.05412 0.00440 3.01084 7.38 -0.11619 35% 20 after 100 0.26 0.009256 6.52 -0.02361 0.11 4.01672 2.76 13.00905 0.01792 0.00511 0.10961 0.94 4.00715 7.12133 35% All 0.00660 0.00490 0.01645 0.

The Ibbotson factors are from the Ibbotson EnCorrr analyzer: The U.81) 0.14 0.45) -0.S.06818 (1.16811 (1.00766 (9.09459 (2.03) -0.82) -0.01438 0.30571 (2.06466 (1.00182 (-0.07) 0.00857 (9.04 0.24 0.03422 (0.98) 0.04509 0.06661 (-1.15989 (1.04233 (-0. Small stock premium is the monthly geometric difference between small company stock total returns and large company stock total returns.00895 0.34) 0.00410 0. Reversal Intercept Market SMB HML Momentum Reversal R-square Factor Model: Ibbotson Factors Intercept Market Small Stock Premium Bond Default Premium Bond Horizon Premium R-square 0.81) -0.24) 0.02) 0.14593 (1.01577 0.22) 0. bond default premium is the monthly geometric difference between total return to long-term corporate bonds and long term government bonds.02101 0. Returns for the portfolios are in excess of the riskless rate.01377 (0. bond horizon premium is the monthly geometric difference between investing in long term government bonds and U.77) -0.09 20 aft.04) 0.00716 (6.55 0.09 0.27) 0.07237 (1.00745 0.77) 0.00714 (8.10) -0. top 5 Wait One Day Portfolio Performance Mean Excess Return Standard Deviation Sharpe Ratio Monthly Serial Correlation Factor Model: FF.50) 0.82) 0.03155 (-0.S.04034 (1.66) -0.66) 0. Carhart’s Momentum factor.00512 (5.S.45 0.14520 (-3.37) -0.35 0. The five actors are the three Fama-French factors.66) -0. 0.74) 0. S&P 500 returns are calculated in excess of Treasury bill returns.Table 4: Systematic Risk of Pairs Trading Strategies Monthly risk exposures for portfolios of pairs formed and traded according to the “Wait One Day” rule discussed in the text.02333 (-0.05 top 20 0.01527 0.03646 (1.55) 0.50) -0.00715 0.54 Eq.19 0. and the Reversal factor discussed in the text.05227 (2.32) -0. U.00111 (0. over the period between June 1963 and December 2002.05403 (-1.30) -0.00651 (7.04514 (1. top 100 0. Momentum.02 0.10312 (-5.18 All 0.39) 0.32) 0.12 0.08) -0.01642 (0.15 0.90) 0.59 0.07697 (-1.00764 (7.38) 0.07997 (1.01724 (-0.S.00795 0.07608 (1.05 0. Pre.00545 (3.71) 0.77) 0.07079 (-1.05740 (1.04120 (1.10192 (1.45) 0.02804 (-0. t-statistics are in parentheses below the coefficients and are computed using NeweyWest standard errors with 6 lags.64) 0.50) 0.64) 0.25) 0.20077 (4. The U.94) 0.59) -0. treasury bills.18077 (-8.93) 0.11) 0.83) 0.05 37 .04817 (-2.

28 -0.45 -0.00745 0.00895 0.35 -0.44 -0.00647 -0.02142 -0.Table 5: Value at Risk of Pairs Trading Monthly and daily value-at-risk percentiles of pairs trading strategies between July 1963 to December 2002 (474 months).00504 -0.01438 0.24 -0.00277 0. top 5 Panel A: Monthly VAR Mean Excess Return Standard Deviation Serial Correlation Value at Risk 1% 5% 10% 25% Probability of return below 0 Min. and then traded over the subsequent 6 month period.00577 0.00296 -0.01002 -0.00130 0.00288 -0.00239 0.00149 -0.00398 -0.00877 -0.06723 0.01516 -0.14 -0.00169 0.00756 -0.01527 0. We trade according to the rule that opens a position in a pair on the day following the day on which the prices of the stocks in the pair diverge by two historical standard deviations.00492 -0.00400 -0.19 -0.01069 top 20 0. The average number of pairs in the all-pair portfolio is 2057.01987 0.02101 0.23 -0.01943 -0.10079 0.00 -0. and the portfolio “101-120" studies the twenty pairs after the top 100.00710 -0.00054 0.00027 0.08218 20 aft. historical observation 0.00133 0.00293 -0.01293 -0.02236 -0.04320 -0.12 -0.00040 0.47 -0. historical observation 0.35 -0.00033 0.01994 -0.02951 38 .00614 -0. The “top n” portfolios include the n pairs with least distance measures.00202 -0. Pairs are formed over a 12 month period according to a minimum distance criterion.00795 0.00071 0. 100 0.32 -0.01236 -0.46 -0.12628 Panel B: Daily VAR Mean Excess Return Standard Deviation Serial Correlation Value at Risk 1% 5% 10% 25% Probability of return below 0 Min.00460 0.00653 -0.04266 All 0.00146 0.00035 0.06 -0.12 -0.01577 0.00145 0.00327 -0.00715 0.

00103 0.00156 -0.01363 -0.00091 0.02264 -0.00154 0.00112 0.00083 -0. Portfolio Fully Invested Mean Excess Return Standard deviation Median Committed Capital Mean Excess Return Standard deviation Median Fully Invested Mean Excess Return Standard deviation Median Committed Capital Mean Excess Return Standard deviation Median Top 5 -0.00156 0.00089 0.00148 All 120 -0.02295 -0.00137 0.00077 0.05521 -0.01443 -0.00760 -0.00123 -0.00162 -0.01358 -0.00177 0.00113 0.01213 -0.00123 -0.00201 -0.02635 0.00761 -0.00152 -0.00206 -0. The statistics are computed over 200 replications of the bootstrapped sample.01449 -0.01200 -0.00122 -0. The random stocks are selected within the same last-month performance decile on the day the position is opened.Table 6: Returns to Random Pairs Sorted on Prior One Month Return Bootstrap of random pairs traded according to the rule which opens a position in a random pair when the stocks in the true pair diverge by two historical standard deviations and closes the position after the next crossing of prices. The bottom panel summarizes the performance with 1 day waiting before the position is opened.00111 0.00004 0.02811 -0.02404 -0.00094 0. The last column is a portfolio of all 120 top pairs.00118 39 .05310 -0.00153 Top 100-120 -0.00192 Top 20 -0.00241 -0.00112 0.00139 -0. The top panel gives summary statistics of the monthly excess returns on value-weighted portfolios of n pairs of stocks where the position is opened immediately.00172 -0.00105 0.05966 -0.00100 Panel B: Wait 1 day -0.

39 top 5 Short Long top 20 Short 20 after top 100 Long Short All Long 0.74 -0.09 0.95 0.80 -1. Equity Risk Prem t-statistic SMB t-statistic HML t-statistic Momentum t-statistic Reversal t-statistic R-square 0.55 4.44 10.03437 0.02625 -0.00170 0.09245 2.01623 0.12532 -0.11 0.00435 0.00589 -0.60 -0.48520 0.09601 0.75 40 .12 0.26 0.47 0.04601 0.07830 1.74 1.00521 0.00254 0.13 0.06506 -0. Momentum and Reversal Factors -0.22 0. The returns in the bottom half of the table are in exccess of the 30-day T-bill returns.31134 8.08832 -3.36 9.42661 0.17288 0.00287 -0.39451 0.49 -0. Absolute kurtosis is reported.03146 0.58571 14.05430 -1.44075 0. Portfolio Long Portfolio Performance Mean monthly return Standard error (NW) Standard deviation Monthly serial corr Intercept t-statistic U.85 -4.00177 0.01458 0.01245 0.S.90 0.16 -0.25732 7.26 0.00426 -2.01120 0.00174 0.03259 0.93 0.21 0.15854 1.00231 0.03453 -2.05157 0.41175 6.00243 -0.00908 0.11 0.48 0.03161 0.00501 0.19 6.00101 -1. The risk adjustment includes the Fama-French factors. HML.21 0.37415 4.98 0.00361 -0.15 0.00183 0.27 0.47 9.13 0.22 0.10 -3.68 8.00103 0.54 -0.12 -0.44 0.47617 0.00179 0.16764 -1. SMB.73091 12.12 Regression on Market.00663 0.49 0.33 1. as well as Momentum and the Reversal factors discussed in the text.19131 2.97 Short 0.83 0.48025 0.45 0.53 -0.00613 -5.05662 0.39208 10.76 11.27 -3.92 9.48 0.41 0.81 0.74 0.48401 7.01330 0.Table 7: Returns to Long and Short Components of Pairs Monthly risk profile for the long and short positions of the pairs portfolios formed and traded according to the “Wait One Day” rule discussed in the text.50772 0.78 0.17 -0.29271 3.00 -0.99 0.52539 0.00442 0.22192 3.04456 0.00165 0.65 -1.53 0. The t-statistics are computed using Newey-West standard errors with 6 lags.03653 0.06616 0.23 9.35 2.06 0.03875 0.10901 -0.02365 0.56217 6.32 0.13 1.

04390 0.98 0.00987 0.29 -3.83 -0.03568 0.02888 1.64 Factor sd 0.56 -0.05 0.06840 -1.62 0.15848 -5.43429 -7.43 -0.00363 3.44 3.45 -0.00217 0.00424 -0.06063 -2.84 -0.41 -0.00065 -0.01527 0. The t-statistics are computed using Newey-West correction with six lags for the standard errors.06983 -1.97 -0.04448 41 .12670 -7.50 -0.02404 0.33193 -3.00992 0.06958 -2. over the two sub-periods 7/1963-12/1988 (Panel A) and 1/1989.75 -2.12 0.65 -0.70 All 0.73 0.07228 Factor sd 0.40184 -5.29 -0.54 0.67 0.01052 0.00353 0.01727 -0.33274 4.45222 7.03641 0.06259 -1.04804 1.19 0.59 0.00241 0. The “top n” portfolios include the n pairs with least distance measures. top 100 0.72 4.02597 0.43395 -0.02332 0.68 -0.03162 0.61 -0.19 -0.26 -0.26 20 aft. The average number of pairs in the all-pair portfolio is 2057.03856 0.Table 8: Subperiod Analysis Panel A: pre-1989 Monthly risk profile for portfolios of pairs formed and traded according to the “Wait One Day” rule discussed in the text.19 All 0.17 0.S.86 -0.00212 0.00446 4.31200 -4.07689 -2.02266 -1.04580 0.03441 0.92 0.00417 3.02 0.28765 4.80 -0.64 0.01291 -0.03506 0.18808 5.32 0.00375 0.12/2002 (Panel B).01630 0. Portfolio Wait One Day Portfolio Performance Mean Excess Return Standard Deviation Regression on Fama-French Factors Intercept t-statistic U. and the portfolios” 20 after top 100" has the pairs after the top 100 pairs.00327 0.06 0.00400 -0.86 0.01 0.00337 2.69 0.04926 0.76 -0.01412 0.01660 0.54 -0. Equity Risk Premium t-statistic SMB: Small minus Big t-statistic HML: High minus Low Book to Market t-statistic Momentum t-statistic Reversal t-statistic R-square Portfolio Wait One Day Portfolio Performance Mean Excess Return Standard Deviation Regression on Fama-French Factors Intercept t-statistic U.04 0.S.44181 -0.56 0.01830 0.50 0.01181 0.27508 -2.01689 20 aft. top 100 0.43575 0.01651 0.76 0.01034 0. Equity Risk Premium t-statistic SMB: Small minus Big t-statistic HML: High minus Low Book to Market t-statistic Momentum t-statistic Reversal t-statistic R-square top 5 0.57 -0.01 0.01295 -0.00710 6.07339 1.08202 -2.77 0.00670 1.10 0.23 Panel B: post-1988 top 5 top 20 0.02923 0.08840 -4.81 0.13 -0.28946 -3.01121 0.02259 top 20 0.29 -0.

00493 0.20 -0.01294 1. Portfolio Mean monthly return (Fully Invested) Standard error (Newey-West) t-statistic Excess return distribution Median Standard deviation Kurtosis Skewness Minimum Maximum Observations with excess return < 0 Mean excess return on committed capital top 5 top 20 Panel A: Top 3 deciles 0.01555 2.10504 39% 0.00514 0.00347 0.00112 7.67 -0.05646 0.00793 0.00854 0.09611 0.77 11.00835 0.00619 0.07255 26% 0.06594 0. The average number of pairs in the all-pair portfolio is 2057.80 6.10642 33% 0.66 All 0.00508 0.15 5.00800 0.95 0.00 -0.00531 0.00401 Panel B: Short Recalls on High Volume Mean monthly return (Fully Invested) Standard error (Newey-West) t-statistic Excess return distribution Median Standard deviation Skewness Kurtosis Minimum Maximum Observations with excess return < 0 Mean excess return on committed capital 0.00585 0.02680 0.81 0.14173 37% 0.10894 0.10084 34% 0.98 8.10478 27% 0. There are 474 monthly observations.01505 1.02 7. Panel B reports the summary statistics for the rule that waits one-day before opening and closing the position.00064 10.01488 0.02210 0.00592 0.03 7.01904 0.00098 5.01579 0.47 0.00665 0.48 6.39 -0. The “top n” portfolios include the n pairs with least distance measures.00296 0.00075 9. Absolute kurtosis is reported. We trade according to the rule that opens a position in a pair when the prices of the stocks in the pair diverge by two historical standard deviations. and the portfolios ”20 after top 100" has the pairs after the top 100 pairs.03803 0.98 -0.10003 26% 0.00312 0.00096 9.00462 0.00085 7.00304 0.00115 5.00728 0.01464 0.82 18.04408 0.00914 0.00202 42 . The t-statistics are computed using the Newey-West standard errors with 6-lag correction.39 -0.01 -0.00094 9.Table 9: Robustness to Short-selling Costs Summary statistics of the monthly excess returns on portfolios of pairs.03465 0.07 0.85 0.00703 0.00614 0.46 -0. from 7/1963 until 12/2002.48 20 after top 100 0.10144 32% 0.00690 0.39 0.

January 1964 43 .Figure 1 Daily normalized prices: Kennecott and Uniroyal (pair 5) Trading period August 1963 .

Figure 2 Monthly excess returns of Top 20 pairs portfolio May 1963 .December 2002 44 .

December 2002 45 .Figure 3 Cumulative excess return of Top 20 pairs and S&P500 May 1963 .

Figure 4 Top 20 pairs and 101-120 pairs portfolios Rolling 24-month correlation of 5-factor risk-adjusted excess returns 46 .

- Pair Trading
- Pairs Trading
- Pairs Trading
- Pairs Trading
- Pair Trading
- Algo trading_Quant Congress USA 2011
- Pairs Trading g Gr
- Pairs Trading
- How to Start a Hedge Fund Quant Trading Business by Ernie Chan
- High Frequency Trading & Flash Crash
- Pairs Trading
- Pairs Trading
- Pairs Trading
- SSRN - Pairs Trading
- High Freq Pairs
- Algorithmic Trading Workshop
- Quantitative Trading Strategies in R Part 3 of 3
- The Application of Pairs Trading to Stock Markets
- Quantitative Trading Strategies in R Part 2 of 3
- The Secret to Finding Profit in Pairs Trading
- Pairs Trading Using Cointegration Approach (1)
- Duration,Votality,Convexity
- Pairs Trading by Cointegration
- NSEL Arbitrage
- Multi-Asset Pairs Trading
- Anatomy of Pairs Trading_EngelbergGaoJag_31August2008
- ADR Aribitrage Opportunities for Dummies
- Quantitative Trading Strategies in R Part 1 of 3
- Copula-Based Pairs Trading Strategy
- Ivashina Stein QJE

Skip carousel

- Animating Mr. Market
- fcic_outline_20100604.pdf
- Check Register FY 08-09 Woodlands RUD
- GMO-Q2-2014
- rev_frbsf_1978no2.pdf
- UT Dallas Syllabus for fin7330.001 06s taught by Huibing Zhang (hxz054000)
- 1973_07109.pdf
- S.E.C. Order Against Carlson Capital
- 68434_1970-1974
- Citadel Letter Dec 22
- Falco v. Donner Foundation, Inc., 208 F.2d 600, 2d Cir. (1953)
- The Fed is a Crappy Reinsurer IV
- FreddieandFannieValue Investing Congress West 2011 Kao VIL
- Did Enron Pillage California?, Cato Briefing Paper No. 72
- 1973_07107.pdf
- Rev Frbclev 197204
- 64334_1965-1969
- 3rd
- Latency Arbitrage
- frbsf_let_19930718.pdf
- frbsf_let_19930618.pdf
- Paul Sonkin Investment Strategy
- Securities & Exchang v. Cochran, 214 F.3d 1261, 10th Cir. (2000)
- frbsf_let_19830114.pdf
- 38studiosassessment
- 1970_6535
- A Short-Run Analysis of Exchange Rates and International Trade with an Application to Australia, New Zealand, and Japan
- United States of America Ex Rel. Michael Lissack v. Sakura Global Capital Markets, Inc. And Mitsui Taiyo Kobe Global Capital, Inc., 377 F.3d 145, 2d Cir. (2004)
- GMO QtlyLetter 2Q14
- Lewis County Board of Legislators Agenda Sept. 1, 2015

Sign up to vote on this title

UsefulNot usefulClose Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Close Dialog## This title now requires a credit

Use one of your book credits to continue reading from where you left off, or restart the preview.

Loading