America’s fight with deflation has brought to the surface much debate about Japan’s lost 2-decades, but for many savvy investors following a particular type of market neutral strategy called mean reversion pair-trading, Japan has been and always will be a gold-mine quite unlike any other in the world.
I have identified more than 50 pairs among Japan’s largest capitalization stocks in which a simple mean-reversion model yields an average compound average annualized return of 33% in a 3-year out-of sample test. Many of these pairs have had no losing trades if held through to the closing signal is triggered, and the maximum negative excursions average only 15%. These trades all have effectively zero-betas, near zero currency exposure, and usually zero industry sector exposure. The average holding period is 51 days, so it is not a difficult strategy to execute. It makes no difference whether monetary policy is enlightened or pathetic. It makes no difference whether the yen is over or under-valued. And it makes no difference whether the global economy is expanding or retracting. For a strategy with such low risk, the prospect of 33% annual returns is extraordinary.
A similar search in the US market yields similar results for only a handful of pairs. What makes Japan the pair trader’s paradise? First let me explain the strategy and the types of pairs that I look for. The most important characteristic of a good pair is that the stock prices must be highly correlated. This is not a strategy in which I look for a good company to buy and a bad company to short. That is a high risk strategy, in my view, because good companies can always get worse and bad companies can always improve. I am looking for companies that get better or worse at the same time so that the spread between the two companies moves within a well-defined range. Contrary to popular belief, this is not a low return strategy. Very few strategies outperform this one on any kind of a consistent basis. One of the reasons for the strong performance is that losses are extremely rare.
I require an R-Squared of at least .75. In Japan, almost every one of the largest 50 stocks by market cap has at least 3 potential pairs when this rule is applied. In the United States, more than half of the top 50 have no potential pairs at all. The second requirement is that the ratio of the two stock prices actually returns to mean on a frequent basis. Stocks can be highly correlated and still drift apart at a constant rate, which would be detrimental to a mean-reversion strategy. This second requirement eliminates almost all of the remaining potential trades in the US, but leaves me with many spectacular opportunities in Japan.
In my view, these trading characteristics of the two markets reflect fundamental differences in business culture and industrial structure that are unlikely to go away any time soon. Shusaku Endo compared Japan to a swamp in his 1967 novel Silence. "This country is a more terrible swamp than you can imagine. Whenever you plant a sapling in this swamp the roots begin to rot." For the pair trader, there is nothing quite like two saplings rotting together.
One of the unique aspects of Japanese industrial structure is the much greater level of direct competition in almost every industry. For example, there are 20 listed drug store chains, 20 private railway companies, and 12 electric power companies. There are 14 auto assemblers, and 129 listed parts suppliers. Of course, there are many companies in the US who compete in the same industries, but they tend to pursue strategies of differentiation, while Japanese companies have a stronger tendency to pursue cost leadership strategies. The result is that they invest in the same projects at roughly the same times and none ever achieve any remarkable differentiation from their rivals. If a leading retailer decides to open a store in China, within six months, ten other Japanese retailers will be all over China. If a Railway operator opens a hotel, 8 other railway operators will get into the hotel business over night. An unexpected but delightful result of this pattern is that even companies in different businesses tend to be highly correlated because all companies become much more geared to the same macro-economic inputs. Endo was right. It's a giant swamp.
Another aspect of Japanese culture that helps the pair trader and frustrates the long-only investor is the lack of any serious merger and acquisition activity. There is a widespread and highly mistaken view that Japanese managers do not care about their shareholders. They care about their shareholders a lot, but only in the way one cares about a migraine. The real purpose of running a business is to provide stable and meaningful employment. To do that well, you must find a way to placate shareholders, but merging with another company for the purpose of eliminating redundant positions is something that will never make any sense in a Japanese business context.
Every now and then, there is a wave of mergers in Japan that ignites speculation that Japan is going to change. Don’t ever be fooled by this. There is nothing in the performance of western economies that is likely to convince any Japanese manager that there is anything wrong with their approach. Mergers happen in Japan, more so in recent years than in the past, but almost always as a matter of last resort and almost always as measures of desperation. The ultimate purpose of such mergers is almost always to save as many jobs as possible.
As a pair-trader you have to fall in love with this custom because the odds of waking up in the morning and finding one of your shorts has been bid up 30% before the open is almost zero. Although I’m not aware of it happening recently, it used to be common for Japanese acquisitions to be arranged at prices that were well below the prevailing market. We called these happy situations the take-under. Imagine being short and hoping that your position gets a bid!
Finally, there is the dirth of venture capital in Japan which makes it so difficult for new companies to emerge. Amongst the largest 100 US corporations, a dozen were start-ups less than two decades ago. In Japan, only 3 companies were start-ups in any living person’s memory. Two others were spin-offs. It is incredibly difficult to fund a new idea in Japan unless you work through existing companies. It is not a coincidence that these account for most of the companies in the top 50 of Japan for which we cannot find good pairs. This does not mean there are never any new ideas. Quite the contrary, in fact. Usually, new ideas are embraced by all existing companies at exactly the same time and reach saturation points far more rapidly than in the West, but no manufacturer ever gets a significant head start on its rivals. This keeps the playing field level and prevents any of the pair relationships that we’ve identified from ever changing significantly.
A Note on Execution
A trade is signaled when the ratio of the two stock prices moves past one standard deviation from the mean and the trade is closed when it returns to the mean. As long as the long-term relationship is stationary, a single standard deviation signifies that 97.5% of the future prices will be at levels that are profitable but it is important to realize that on 2.5% of occasions, losses, sometimes in the double-digits can easily occur. Do not cut losses when this happens. The old trader’s maxim of cutting losses early and running with winners will only destroy this strategy. Don’t do it. Especially if a pair ratio falls to 2-standard deviations, this will signify the point where 98.75% of future prices are likely to be at levels that are profitable to the trade. Cutting losses at this point would be ridiculous. The only reason you would cut any of these trades before the closing signal occurs is if there is a total game-changing event. Examples of this might include a pair of food processors in which one of them suddenly decides to acquire a very large hazardous waste disposal business. That would probably change the relationship between the two companies.
Separated at Birth
A recent example that illustrates some of these points is the East Japan Rail vs West Japan Rail spread. These two companies were both part of the National Railway before they were split up and sold to the public. They are identical twins with 75% of their revenues from transporting commuters and 25% from selling goods and services to these commuters in station buildings and hotels and shopping facilities along the way. The probability of one significantly outperforming the other over the long run is practically nil. Even their quarterly earnings are practically identical as can be seen in chart 2. R2 of the underlying instruments is 95%.
Nevertheless, in the past 4 months, West has outperformed East by about 30%, bringing the spread to within 2 standard deviations away from norm. Unless something has happened to change the long-term relationship dramatically, this means there is less than 2% chance of losing money on this trade. This is unlikely. It does not cost more to build railways in the East than it does in the West, nor can one charge much more than the other for its services.
Buy JR East/Sell JR West Trade Model with signals
Buy JR East/Sell JR West Trade Model with signals
Another current favorite of mine is the Denso/Honda trade. Although Denso is a supplier of high-tech parts to Honda’s rival Toyota, it’s hard to find many differences in the performance of the two companies, and the R-squared of the stock pair is .85. One difference we think will prove important is that Denso earns 54% of its profits from Asia while Honda gets 67% from the US. Denso is also increasing its share of sales to US assemblers - something Honda can’t do. Investor aversion to Toyota has skewed the relationship of these two companies more than one standard deviation in favor of Honda. How do you imagine a US assembler sustaining outperformance against an Asian parts supplier?
Buy Denso/Sell Honda Trade Model with Signals
Buy Denso/Sell Honda Trade Model with Signals
Remember, no M&A
Recently, we added the Long Sharp / Short Sony trade when specualtion about an Apple acquistion of Sony got crazy. The story that ran in Tokyo went something like "Apple has 51mn to spend and Sony has many things Apple needs." Doesn't anyone do the math anymore? Apple has $24bn, but Sony would cost at least $36bn even at the current market price and has nothing that Apple needs. It is a bureucratic and impossible to managed behomoth that wouldn't even be on Apple's long list. If you are smart enough to have $24bn, you would not be stupid enough to spend $36 of it to create nothing but headaches.
Buy Sharp/Sell Sony trade and Model Signals
Disclosure: No Positions