The consilience of the Nobel to Shiller and IBM reporting earnings and IBM dropping 7 points from the close reminds one of how careful you have to be in developing systems based on past data.

How many things are wrong with Shiller's data. For one, they didn't report earnings in many of the time periods he uses. For another, they reported them 4 months late. For another, the earnings series are adjusted many years later. For another he's using average of 10 year earnings and 10 year prices, to come up with his p/e. Okay, it doesn't take account of expected earnings the next period. It doesn't take account of the movement up from the end of the year when reported earnings are better than expected and the move down when reported earnings are less than expected thereby giving a 100% change that the low p/e will show superior performance but certainly not feasible to implement. And of course the best estimate of the p/e is the current level not the level 10 years ago. Yes. Earnings and prices are a random walk in changes not levels. And yes, the use of averaging introduces spurious serial correlation as working pointed out. But yes, even for those who know that when the stock price has a tendency to go up in the period before the report, and it's statistically significant, you can't use it or make money with it. Because yet, like IBM, when the earnings are bad, they often report it earlier than the due date. So you're waiting for the report date, knowing that prices are going up, but then they report early, (perhaps the flexions acted), and you get caught with the down 20% price move even though the move up to the estimated reporting date shows positive correlation with the forward. I.e. you can't make a profit with it even though there's a extraordinary regularity in the past data.

It makes one wonder why the Nobel was given to a chronic bear when the reason the Nobel prize is so high is they put much of their endowment in the triumphal trios' drift upward in equities from 1$ in 1899 when they started to 30,000 today, etc. 

Richard Owen comments: 

Only the Chair could pack so many statistical anomalies into a handful of sentences!

Perhaps I have misunderstood the use of Shiller PE, but the traumas you describe apply to reasonably mediumish timeframes, whereas Shiller's work is typically discussed in the context of long range endowment allocations of Yale type money to broad indexes. Whether it is even then a useful tool is a different question. But hasn't a very low shiller PE typically coincided good prospective long range returns? But the issue being one would sometimes have to wait half a career or so to be proven right.

Ben Graham did similar work but discarded it as a means of allocating capital as the Baconian issue caught up to it. Grahams issue was the other side of it: exiting on a high shiller PE with a decade of bull market to go.


WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '8750' AND comment_approved = '1' ORDER BY comment_date




Speak your mind


Resources & Links