1991issue C121-3
Construct an equity book by weakening residual co-movement, then setting a market slope
A book's expected return is the cost-weighted average of its holdings, but its variance depends on pairwise return association. After a small, sector-aware set has reduced residual co-movement, the cost-weighted average of holding slopes can be reset so systematic exposure stays inside a chosen tolerance.
- A book's expected return is the cost-weighted average of its holdings' expected returns, but its variance is not a simple average of those variances because pairwise association can amplify or offset fluctuations.
- Adding names is not enough. The goal is weak or negative return correlation, and a four-name book can be adequate if any two names are from different sectors, while 12 to 18 names are said to capture most available diversification.
- Relating each holding to one index splits risk into an undiversifiable market component and a residual that selection and diversification can reduce. The book's slope is the cost-weighted average of holding slopes.
- After residual risk is reduced, periodically resetting the book's market slope keeps systematic exposure inside a chosen tolerance. Comparable slope estimates should use one method, a recent window, and the investor's horizon.
Why variance is not an average
Building an equity book means combining holdings so total return is a cost-weighted average, while residual return variance depends on how those holdings move together. A book's expected return is the cost-weighted average of its holdings' expected returns. Its return variance is not a simple average of those holdings' variances, because pairwise return association can either amplify or offset fluctuations.
Pairwise return association is summarized by a correlation coefficient for direction and strength of co-movement and by covariance for the size of the combined fluctuations.
Diversify by association, not by name count
Construction is the practical sequence of choosing how many names to hold, checking pairwise return association, and then targeting a book-level market slope. Adding more names does not by itself produce adequate diversification. The construction goal is weak or negative return correlation across holdings.
Diversification limits leftover single-name risk by combining holdings whose returns are only weakly associated or move in opposite directions, rather than by adding names indiscriminately.
Replace pairwise covariance with one index slope
A market-sensitivity framework replaces exhaustive pairwise covariance work by relating each holding to one broad index, then splitting risk into an undiversifiable market component and a residual component that selection and diversification can reduce.
Market sensitivity is the slope of a regression of a holding's returns on index returns. Linear regression estimates that slope from a fitted line between a holding's returns and a chosen index over a recent sample whose length and frequency match the intended horizon. The index slope is 1. Cash-like instruments have slope 0. A slope of 1.4 is described as a 1.4 percent holding-return move per 1 percent index-return move.
Book-level market sensitivity is the cost-weighted average of holding slopes. Cash-like instruments do not change that average because their slope is 0, but they still enter the return calculation unless their yield is also 0.
Reset the book slope and keep estimates comparable
After residual risk is reduced through diversification, periodically resetting the book's market slope keeps systematic exposure inside a chosen tolerance and supports risk-adjusted return comparison. Use each holding's slope versus a market proxy, then the cost-weighted average of those slopes, to keep systematic exposure inside a stated tolerance and to judge results on a risk-adjusted basis.
Estimated slopes change with sample length, with daily versus weekly versus monthly subperiods, with the chosen index proxy, and over time for a single name. Comparable estimates should come from one method, use a recent window, and match the investor's horizon.
All readings on this track · 13 readings
- 1989Rank rotation opens the names; pairing keeps the hedge
- 1991Evaluating allocation sleeves against an index proxy
- 1991Construct an equity book by weakening residual co-movement, then setting a market slope
- 1992Sizing index futures to retarget portfolio beta
- 1992Portfolio beta overlays with index futures and puts
- 1997Construct a trailing stop from beta and price
- 1998An offense-defense matrix for regime-aware stock selection
- 1998Constructing a stock-book duration proxy from beta and bond markets
- 2007An index put as deductible insurance for an active stock book
- 2010Constructing option overlays for beta-adjusted portfolios
- 2014Long-only relative value with an index hedge and a crisis stop
- 2017Hedging a rising book with beta and ATR
- 2019Sizing micro index futures overlays for cash equity books