Going pair-shaped?

We are currently seeing something of a minor crisis in the pair-trade space, with many practitioners complaining that the pair-trade alpha of old is no longer as easy to extract. At the same time we find widespread interest in expanding the pair-trade strategy space in new directions. In this article we briefly review some explanations for the crisis, and some directions for the future (see box below). We then argue that a necessary pre-requisite for pushing pair trading beyond its traditional limits is a clear understanding of the very different forms that pair trading can take, and an understanding of the implications of these differences. The differences are often obscured by the blanket use of the term “pair trade” to cover a range of disparate strategies, as well as by the accommodating nature of the standard Bollinger-band type analysis1.

Roots of the Crisis

Various factors, of varying degrees of importance, have changed the nature of the equity pair-trade space in South Africa over the last 10 years. Among the many factors mentioned
in the literature we will focus here on changes in market structure, on overtrading and on market regime changes

Market structure changes

A prime example here would be the gradual removal of the old pyramid and N-share structures from the Johannesburg Stock Exchange, as it has come into line with international standards.

Trades based on suchrelationships were structurally close to beingarbitrage trades2, and as such necessarilywell-behaved. The only requirements were to identify the standard discount that applied, and to avoid being caught on the wrong side of a corporate decision to unwind the structure. Remarkably, such pairs often displayed sufficient variability to allow for profitable trading. But that era of relatively easy money is over.


Even after the removal of many of these structural pairs, there remained a substantial number of obvious pairs which could be assumed to be driven by common factors and thus behave in similar ways (Anglos vs Billiton would be an obvious example). This kind of statistical pair trading lends itself to the standard analytic approaches, such as Bollinger Bands and other technical tools, and
it is these trades that have been the heart of the classical statistical arbitrage pair trade.

Regardless of the exact trading style used, this approach to trading was clearly successful for
a large number of market participants. However, the increase in the number of equity hedge funds over the last decade has naturally led to an exponential increase in the amount of time and money allocated to exploiting this limited set of standard pair trades.

The South African equity market is, of course, not especially large, and the number of sufficiently liquid pairs of this type is rather small, so that the standard candidates are known to, and traded by, most participants in the area. This has meant that the increase in hedge fund activity has closed down the profitability of the more well-known trades very rapidly.

Market regime changes

Since statistical pair trading must use historical data in some or other way to develop its trading rules, changes in the relative behaviour of stocks can fundamentally affect the profitability of trading. Such changes can come in various ways:

  • Changes in market volatility, in particular cross-sectional volatility Pair trading requires an adequate amount of variability in returns within the population of candidate stocks (as best measured by cross-sectional volatility – the standard deviation of stock returns on a single day, or other holding period, across the universe of stocks).Thus where market moves are highly correlated the opportunity for profitable relative trades may simply not exist. However, the market tends to cycle through periods of greater and lower cross-sectional volatility, so the problem is not permanent.
  • Changes in market regimes pre- and post-2008 It is possible that certain general relationships which held pre-2008 no longer hold, and that this may be a short-term, medium term or permanent change. For example, it is clear that the market post the 2008 crash was much more aware of credit and debt issues in pricing securities than before.
  • Shorter-term switching in stock valuation regimes in the current period An interesting phenomenon has been the problems confronted by market-neutral funds, and thus presumably the pair trading style, in 2010. Although the market as a whole this year has swung wildly and somewhat unpredictably between extremes of optimism and pessimism, a naïve assumption would regard market-neutral funds (or trades) as immune to this volatility. However, it would seem that these switches also affect the basis on which the market makes relative stock valuations. In particular it is clear that when bearish sentiments predominate, attributes such as low debt ratios and solid dividend expectations are uppermost in stock valuations, whereas during bullish periods growth attributes come to the fore.


    For the full report, View or Download the pdf by clicking here. [File Size: 592 KB]