Arbitrage opportunities exist when profits can be earned from spread trading, the simultaneous purchase and sale of similar securities. A trader buys the one that she believes is relatively cheap while shorting the relatively expensive security in the hope that the spread will narrow. Assuming that her views are well-founded, losses can still result, in part because of what the academic finance literature puts under the rubric “limits to arbitrage.” Assuming reasonable liquidity levels, these limits stem from two sources of risk: fundamental risk and noise trader risk. The first exists if the long and short sides of the spread are not perfect substitutes (as is almost always the case). For example, if you think Ford is cheap relative to GM, a long Ford-short GM arbitrage could founder because unanticipated negative Ford-specific (i.e., unrelated to the broad auto industry) news is announced.
Noise trader risk, defined to be the risk that mispricing becomes worse in the short run, has been famously documented by Lamont and Thaler (2003) with the Royal Dutch/Shell case. At one time Royal Dutch, headquartered in the Netherlands, and Shell, based in the U.K., were two totally distinct companies. Their association began in 1907, but this association fell short of a full merger. Using what is known as a dual-listed company structure, Royal Dutch and Shell merged all of their operations, agreeing on an ownership of 60% for Royal Dutch and 40% for Shell for their subsidiaries. All after-tax cashflows, including dividend payments, were electively split in the proportion of 60:40. Despite the fact that this was common knowledge to all shareholders, their share prices, amazingly, often were well out of synch with this ratio, with researchers finding persistent and large price deviations of up to 35% away from parity. Since the dual-listed status made it clear that Royal Dutch and Shell were perfect substitutes, the mispricing that was observed was logically the result of noise trader risk.[1]
This case notwithstanding, in most realistic arbitrage situations both fundamental risk and noise trader risk are factors. Say we believe that two “similar” securities, though not perfect substitutes for each other, are out of alignment. Assembling zero-cost portfolios of such perceived mispriced pairs of securities for diversification purposes may be the solution. In this context Gatev, Goetzmann and Rouwenhorst (2006) first documented the efficacy of pairs trading, a quantitative strategy that has existed on Wall Street for over 25 years. While it made sense that Royal Dutch and Shell should move together, what if we don’t worry about obvious commonalities between securities but rather let the data do the talking? Pairs trading involves the exhaustive (computer-assisted) search for pairs of securities that have moved closely together (say) over the last year. On the expectation that this pattern will continue to hold for at least the next several months, a pairs trade is put into practice when a divergence of sufficient magnitude is observed. The cheap security is purchased while the dear security is shorted in the expectation that the two stocks will move back together. Amazingly this simple strategy has worked quite well in the U.S., with a simple trading rule yielding 11%/year for self-financed portfolios of pairs.
Jacobs and Weber (2012) have recently extended this work in two important directions. First, they show that pairs trading is more profitable when market participants are likely to be inattentive. Attention is a scarce resource: it is not possible for everyone to be focusing on all information at the same time. People naturally prioritize, paying attention to what seems most important at the moment before shifting attention elsewhere. When executing a left turn a wise driver will postpone the philosophical debate he is having with his passenger until the turn is completed. It has been argued elsewhere that investors may reduce their attention to individual securities when an abundance of market-wide or sector-specific information arrives (Peng and Xiong (2006)). Jacobs and Weber develop a “distraction” metric based on this notion, and then go on to show that on days when distraction is highest pairs trading is most profitable.
Second, they extend their empirical analysis to the international realm by investigating the efficacy of pairs trading in Japan, the U.K., France, Germany, Switzerland, Italy, the Netherlands and Hong Kong. These markets represent the eight largest non-North American stock markets by capitalization. They find that pairs trading is profitable in all eight, with profits ranging from 6%/year (Italy) to 13%/year (Germany and France). Further, with the exception of Japan, distraction predicts the profitability of pairs trading as was true with U.S. data. The purpose of the present paper is to explore whether pairs trading is an effective return-enhancement strategy in Canada. To preview, the answer is in the affirmative, especially when markets are likely to be inattentive. Download the full paper.
[1] In 2004 Royal Dutch/Shell announced plans for a full merger (which it followed through with in 2005), partly as a result of persistent parity deviations. See de Jong, Rosenthal and Van Dijk (2009).