Friendly Response to CXOAG's March 4, 2010 blog

In his March 4, 2010 blog, Steve LeCompte (webmaster of CXOAG) examines the daily data I provide to estimate the value of the S&P 500 using RYT, since July 2009. His conclusion is that RYT has little or no predictive power from one day to the next. I think Steve LeCompte provides a great service to the Finance community by surveying key discoveries in academia and distilling their essence and usefulness to practitioners. I also do not disagree with Steve's conclusions. However, I find them a bit misplaced. Let me explain. During the Vancouver Olympics you would not expect the scoring system for PAIRS skating to be used for judging ice dancing. Wikipedia's definition states that "Ice dance differs from pair skating by having different requirements for lifts, requiring spins to be performed as a team in a dance hold, and by disallowing throws and jumps." Of course, I'd love to see ice dancers do triple lutzes...

The data I publish on my website is meant to represent the CURRENT state of the valuation of the S&P 500 index not where it is likely to be the NEXT DAY. What I argue is that in order to predict you first have to have a thorough understanding of the mechanisms underlying market valuation. As Steve points out, on average my estimate has had a 0.19% percentage error from the actual index over the period, on a DAILY basis. I do not know of ANY other market valuation model out there that can come close to matching that. Right now, my model is NOT a predictive model.

Now, we teach in Finance departments that if a stock is undervalued then you (the investor) should buy because it is likely to rise in the FUTURE. Of course, it is true under two conditions: 1) the valuation takes a middle of the road view of long-term growth assumptions and future cash-flows, broadly correct rather than accurately wrong and 2) that the company and the economy do NOT experience drastic changes in terms of unforeseen events with BIG consequences (I am not necessarily referring to Black Swans here. For example, a sudden jump in investment opportunities suffices).

My technique is trying to pinpoint as ACCURATELY as possible why the S&P 500 is valued the way it is. In fact, the information as it is currently displayed is NON-actionable. Why? Because I use updated estimates of EPS, inflation risk premium, government yields, which change everyday. I do not make middle of the road assumptions about future growth etc... This fact is actually perfectly consistent with Efficient Market Hypothesis (semi-strong form) that all available information (in particular forecasts) is already included in the pricing of stocks. My discovery in publishing the daily data was actually how strong this result is, given that the flow of new information seems to significantly disturb prices from one day to the next. In other words, contemporaneous DAILY mismatches are not informative of next day's direction.

I understand that Steve is looking for exploit-ability of this techniques, as in the end this is where the rubber meets the road. We actually had a friendly e-mail exchange about it months ago. But I think Steve is asking too much of the model as it is right now. Before GPS was invented, people had a vague notion of their location as they were driving around looking at a map. GPS brought accuracy with respect to that information. This service is useful. Now, you wouldn't say that GPS failed because it wasn't giving driving direction when it first came out.

The next step for RYT IS prediction. I take Steve's analysis seriously and interpret it as a call to get moving in that direction. Agreed. For example, I am currently working on a paper that shows that daily movements in Treasury yields are strongly correlated with S&P 500 price movements during the financial crisis. This and other predictive models for the inputs in the RYT formula will lead to a testable model for trading.

Albany, March 7, 2010