Wednesday, April 11, 2007

So Long Reversion

In an April 10th research report, Daiwa's numerical gurus Yoshino and Sagawa wrote the epithet for naive price reversion strategies in the Japanese equity market. That price reversion has been the mainstay of many a naive arbitrageur in Japan from the bubble to the depths and back is noteworthy, though official pronouncements a brokerage firm of anything that reduces customers' trading turnover is rare to say the least. But Mssr's Yoshino and Sagawa have - up until now - been doing nice work, so their research, if nothing else, should be weighed carefully.

Their preamble summarizing the literature and history is coherent enough. And their gross numbers, while pathetically inadequate DO mirror the general condition of naive reversion, highlighting in particular the dearth of return since 2002. And while they arrive at a possible replacement in the form of a simple sector reversion model, it's paltry return is in all likelihood not robust to transactions costs and security borrowing limitations.

But in my mind, they fail to address the most nagging question of all: Why should reversion have persisted in Japan, contravening the normative behaviour of return persistence patterns highlighted by Jegadeesh in the US, and replicated by others for virtually all other global equity markets? For in even weakly efficient markets, BOTH naive reversion AND naive momentum shouldn't exist. Nonetheless, it [reversion] did, and rather than mourn and eulogize its passing, researchers with any sense of curiousity should set to work on why it was, historically speaking, so omnipresent.


Sally said...

I know this is off-topic (sorry)but wanted to laud you for your December 2006 Wolfowitz prognostication!

Cassandra said...

Like Monty Python's "Knights of Nee", it may be just a flesh wound as the US still has the say over WB's stewardship. And one must be careful what one wishes for. Recall Wolfowitz was reshuffled and dealt away as he became a liability without throw to the dogs. If Wolfowitz goes, we could end up with Alberto Gonzalez running the World Bank!

Thanks for remembering!

Aaron Krowne said...

I think you meant "epitaph" near the beginning.

Also I was wondering if you could define "naive reversion". There seems to be nary a definition on the web, with this blog, and this very post on it, being the top search result (with none of the rest terribly promising).

Forgive us uninitiated finance outsiders!

"Cassandra" said...

[Price] "Momentum" and [Price}"reversion" (or return reversal) are two-sides of the same coin. The seminal research essentially ranked some universe of liquid stocks by their n-month historical return, formed portfolios based upon top (high momentum) and bottom (low momentum) decile return and tracked the portfolios performs for i-months calculating the resulting return spread.

The original research, and multitudes of subsequent studies found similar patterns of such decile spreads across most world markets: Short term (n=1mo & less) reversion was the rule and very profitable; intermediate term (where n=>1mo and n<15mo) momentum ruled; And longer-term (where n>>15mo indicated robust reversion returns). Interestingly, in Japan, momentum was not only absent at intermediate horizons, but the flip-side reversion, was reasonably profitable too.

Insofar as these portfolios are formed by a single factor and rote ranking method, I call term this "naive". This is true for momentum and reversion strategies.

There are of course many different flavour and levels of sophistication (with John Brush @Columbine being the doyen of momentum). Moreover, subsequent research (most newly minted PhD theses) has sliced and diced and analyzed momentum and reversion backwards and forwards to try to explain the persistence of persistence which complicate most efficient markets theories. Yet depsite all this research, "momentum" or "return persistence based upon prior return ranking", remains essentially unexplained outside the realm of behavioural conjectures. But by definition the same must hold true for abnormal returns to reversion.

So "naive reversion" is the simplest of phenomena, but yet remained present in Japan until 2002 or so. Of course, profitable momentum and reversal strategies live (in the real world), but they employ far more complex and discriminating analysis, be they purely numerical recipes to ferret out recurring robust opportunities, or multi-factor approaches that build and draw upon robust distributions that employ fundamental or technical, attributes. These live, but the simplistic or most naive
methodologies have been arbed away in the markets relentless drive towards something in the direction of greater efficiency.