XTR – Correlation

In probability theory and statistics, correlation indicates the strength and direction of a linear relationship between two random variables. This statistical parameter can not only help in stock selection but also create structured products based on market neutralizing.
Suggesting a hedging strategy, a week before market fell more than 8% was nothing short of timing. Whether the markets run up this week or not, March futures expiry should make your hedged portfolios richer. And the way we see thing at Orpheus, market neutral strategies should become more than a bread earner in the time ahead, especially if volatility continues to rise.
Building on where we left last time, the primary aim of creating correlation matrices is to formulate high integrity pairs. For this we have studied correlations over various time frames (historical, few years, quarterly, weekly and daily). The current XTR illustrates some of the historical correlation matrices.
The highest correlation pair stands at 0.985 between SIF 1 and SIF2. But SIF2 (SIF2.BX) and SIF5 (SIF5.BX) make a better active pair, owing to their tradability in futures. A long – short strategy can be a continuous market neutral strategy between SIF2 and SIF 5. And just like the SIF pairs, TLV (BATR.BX) and BRD (BRDX.BX) (Romanian banking majors) have the highest correlation at 0.95. This means irrespective of their separate fundamental drivers BRD and TLV make a good pair. And from a statistical view, exposure to one bank would have compensated for the exposure to the other or both the stocks.
To illustrate this case further, we can compare the price performance of the banking majors. Though on a short term basis say July 07 to Feb low BRD fell 13% more than TLV. From Jan 07 to July high the comparisons stood at 68%, 48% for BRD and TLV respectively. And if you look at overall performance starting Jan 2006, both the stocks returned similar price performances at around 45%. This is what a good correlation pair does. Annualized returns for good pairs are similar irrespective of the underlying events, which drive the respective stocks. It’s also to do with the large cap hypothesis we talked about. Two blue chips in a high correlation pair, which are also sector leaders can’t deliver divergent price performances. And a difference, if any can only be in the short term, over a few weeks to a few months like we have illustrated in our case above.
And as markets mature even these short term price inefficiencies (example BRD returning 13% less than TLV since July 07) can be taken care by arbitrageurs or strategists who play between such highly correlated pairs (Short BRD, Long TLV and vice versa). This is where market neutral comes in. We at Orpheus have highlighted these market neutral strategies between TLV vs. BRD, SIF2 vs. SIF5, SNP (SNPP.BX) vs. OIL (BRT-) on prior occasions. A few market neutral pairs (long one, short other) out of the many we illustrated in Orpheus market letters failed to give consistent results. One of them was the SNP-RRC (ROMP.BX) pair, which as we illustrated in the correlation matrices last time (XTR.170308) is a low correlation pair at 0.59, the very reason a continuous running long – short strategy is not very successful in the respective case. This is despite the fact the both stocks are from the same sector. Sector grouping can enhance a pair integrity, but it does not guarantee it.
And correlations are not just local, they are global in their scalability. We talk about them all the time. Like for example the correlation of OIL (Brent) with SNP. Many of our readers have expressed surprise at SNP an OIL relationship. How SNP fell despite the net rise in
OIL prices from $ 50 to $ 100? How BETFI fell despite DOW rise? And how is Euro Ron connected to BET or BETFI? Over the long term OIL and SNP has a 0.85 correlation (this makes it a better pair than the local RRC – SNP pair). However, owing to the long termism of the correlation indicator, price inefficiencies between stock and its underlying commodity (in this case, OIL and SNP) can extend for more than a few months. This is what happened, OIL rose and SNP fell (We will explain SNP-OIL pair more in our next issue on Intermarket). A similar relationship is assumed to be in DOW Jones Industrial and local market indices (BETFI, BET, BETC). However, the correlations are weaker when we consider the DOW and BETFI. And if you are really looking at CAC or DAX for trading BETFI, you are on a heading towards a losing streak. The DOW and BETFI correlation is higher when markets fall and extremely poor and sometimes negative. This means that we had many occasions when local indices were rising while the DOW was falling. The correlation has never reached 0.9 between DOW and BET.
Hungarian BUX (slide 14) on the other hand (has the highest correlation with BET) and can tell us more about the Romanian indices than DOW, which is more of a mass psychology play (herding). Panic is a bigger motivator than greed. No wonder when markets go down, the correlations increase all over, and not just with DOW. But with a host of global indices including Russia, Nikkei, Shanghai, India, Brazil etc. It’s the DOW link, which makes more news, the reason it get’s anchored in mass psychology.
Correlation cannot only help identify pairs, but also assist is stock selection. When we need to assume market risk (beta) we can chose highly correlated pair, but when markets are at euphoric levels, stock picking can be based on negative correlation stocks. Globally, negative correlation is a much desired strategy today. Hence the significance of understanding the correlation matrix cannot be undermined. In the previous issue of XTR (100308) we showcased the XTR 21 beta. This week we have shown a correlation matrix (slide 13) between component stocks. As you can see there are many negatively correlated pairs, no wonder XTR 21 falls lesser than BETFI every time we have a negative week. Since inception XTR 21 never fell more than BETFI (week over week). We need more history to validate this, but both portfolio beta and negatively correlated pair among XTR components confirm our view.
Negative correlation has better predictability than low correlation. DOW has a low correlation with local Romanian indices, but EURRON (slide 14)has a negative correlation with BET over the last four years. This also suggests the local currency is a better indicator than the global benchmark. And last but not least all these pair components lead and lag in performance against each other. This is why SNP underperformance against OIL is cyclical. And we might not be far away from the time when OIL corrects and SNP rises. Next week we extend this pair formulation strategy to sectors, to identify outperforming sectors from the ones that are set to underperform.
On the XTR 21 we had another week of drawdown, as the broad market corrected. The benchmark is down 10% from inception. But we are still positive for the coming months and continue to consider these low risk entry points. XTR 21 outperformed BETFI yet again. But since the fall was sizeable most market indices were down. From the late economic cycle sectors, it was the utilities, staples and materials which witnessed marginal losses.
To read the latest issue of XTR.240308 write to us for a free trial today or download the report from REUTERS KNOWLEDGE, YAHOO FINANCE, THOMPSON ONE or THOMPSON RESEARCH.