A book-to-price score that is almost the same set of names next quarter
can still be a forecast that has already been paid. Score
autocorrelation and information-coefficient decay are different clocks.
The first says the rank persists. The second says the rank still lines
up with the next return. The retirement rule for a live strategy is the monitoring shell. This note
defines the two clocks for a characteristic portfolio and states which
one a rebalance is allowed to trust.
Learning objectives. By the end of this note you should be able to:
- define the horizon information coefficient and the lagged information coefficient, and give the December 31 example of each;
- explain why momentum information coefficients are described as decaying faster than value information coefficients, without quoting a half-life the extract does not give;
- separate overlapping-window bias in the standard error from genuine decay;
- use selection, sizing, and timing as the performance split that shows whether a decaying idea is still being sized as if it were fresh.
In this lesson
- Two correlations that share a forecast date
- What decay does inside a book that is already on
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Empirical Asset Pricing and Factor Investing syllabus.