Quant analytics: PCA for alpha generation
I am curious if anyone has had success applying principal component analysis in the context of alpha generation (and not in the context of risk, as PCA is typically applied). For instance, I like the straightforward “absorption ratio” introduced by Kritzman, et al. here:
I suspect that tracking the evolution of portfolio risk concentration may yield forward-looking clues on the portfolio’s performance. For instance, perhaps a factor mimicking portfolio may lose efficacy if its risk becomes highly concentrated during the life of the trade (in this hypothetical context, the “absorption ratio” is used as a timing parameter for factor rotation).
As an aside, outside of alpha generation, I think any “real-time” systemic risk metrics are highly relevant in today’s market. Particularly in the US equity space, where investors have enjoyed a respite (of sorts) from macro-driven Eurozone news since mid-Dec (coinciding with the ECB’s 3-yr LTRO announcement). The abrupt declines over the last hour of trading on May 22 (purportedly driven by comments from the former Greece PM regarding the country’s potential exit from the euro) is a sharp reminder to me just how quickly the market can switch back to the high correlation environments of 2H11. Monitoring systematic risk may give active investors sufficient time to adjust their level of activeness/leverage in open positions.
FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!