Seminar on “Understanding Pairs Trading
Date:24th December 2010
Speaker:Mr. Rohtas Kumar, CFA,
Deputy General Manager – Research, Cognizant
Venue: Hotel Kens, Srinagar Colony, Hyderabad
An evening seminar was organized on the topic “Understanding Pairs Trading”. The speaker was Mr. Rohtas Kumar, CFA, Deputy General Manager – Research, Cognizant.
Pairs of securities, which share common risk factors, should maintain a stable price ratio over time. The idea of pairs trading is to identify such pairs, whose prices appear to ‘move together’ (stable price ratio) over time and take simultaneous positions—long position in one stock and short position in another. Although the ratio may fluctuate considerably in the short term, it reverts towards a fairly stable price ratio. The objective of the pairs trading strategy is to identify such pairs and then identify trade opportunities arising out of temporary diversion of the price ratio from the historical average for those pairs. For example, when the price ratio goes significantly below the stable range i.e. more than two standard deviations from its historical mean, a profitable strategy would be to simultaneously buy the stock in the numerator of the price ratio and sell short the stock in the denominator. The main idea is to open a trade when it is likely that the long position will appreciate, relative to the short position, in the short term. The position would then be closed when the ratio has increased sufficiently i.e. within one standard deviation from its historical mean, either because the long stock has appreciated, or the short stock has depreciated, or both. The opposite positions would be taken when the ratio is particularly high.
The UBS Quantitative Equity Research team did a thorough analysis of six statistical methods that can be employed to identify stable and mean reverting pairs of stocks. The methods are the correlation of daily returns, the runs test, the KPSS test, the IKPSS test, the sum of squares, and the Adjusted DickeyFuller (ADF) test. Stationarity is the property that is fundamental to designing pairs screening methodologies. The results of the study indicate that statistical tests are successful in identifying pairs that display stable price ratios out of the sample. The sum of squares method and the ADF test appear to be the best performers. Contrary to popular view, using the correlation of returns as a metric to assess the suitability of a pair leads to very poor results. The correlation method is theoretically flawed. For example, if two stocks are strongly correlated to the market, but with different betas, they will not have the same sensitivity to market risk, but will appear to be good pairs.
The study also concluded that pairs trading does generate significant alpha. The ADF test and the sum of squares methods not only perform better in identifying stable and mean reverting pairs of stocks, but also in terms of profitability compared to other statistical methods. Finally, the study addressed issues of timing. Most of the gains to a pairs trading strategy accrue in the first five days after the divergence of the prices from their ‘equilibrium’. This conclusion has important implications for the interpretation of the profitability of pairs as well as the practical implementation of the strategy. The results also suggest that volatility is one of the likely conditioning variables that help predict the profitability of pairs trading: profits deteriorate during periods of low volatility.
