New Frontier presented new research on the portfolio monitoring problem at the JOIM fall conference in Boston.  "Dynamic Portfolio Monitoring" is an important extension of our current patented technology for rigorously deciding when to trade.  The new procedures address the limitations of the ad hoc techniques in current practice.  The algorithms are statistically, financially, and computationally state-of-the-art.  This work continues our corporate policy commitment of developing and applying the most advanced asset management technology for practice available today.

Abstract
Richard O. Michaud, David N. Esch, Robert O. Michaud

Professional asset management requires effective portfolio monitoring.  A manager needs to know when the current portfolio is sufficiently different from target to recommend trading.  But rebalancing rules are typically ad hoc and suboptimal.  Trading on a fixed calendar schedule or using simple heuristics characterizes much practice.  Such proposals ignore the essential character of the monitoring decision, which depends on the statistical similarity between two portfolios.  Of the few proposed statistical procedures almost all are based on unrealistic assumptions in order to use familiar null distributions.  Real-world portfolio management demands inequality constraints on portfolio weights and targeted risk levels on an efficient frontier.  Practical decision rules require compute-intensive methods which would be intractable using traditional analytical techniques.

The first practical test for mean-variance optimality is the Michaud rebalancing rule, which tests a current portfolio for statistical deviation from a targeted portfolio on the Michaud Resampled Efficient Frontier.[1]  However, the rule does not consider the case when some information defining the existing portfolio may have been used in the target.  This partial input match results in an overly conservative rebalance signal.  This issue is important because a manager wants to know when to trade effectively as soon as possible but no sooner.  We develop new algorithms that allow overlapping data in the Michaud test.  The distribution extends the critical range for the Michaud rule and boosts its power.  We give two procedures, one for purely historical data and the other for the more general case of managed risk-return estimates.[2]  Both algorithms are illustrated with examples.  Applications include large-scale, automatable, dynamic monitoring customizable to manager styles and other investment considerations.

[1] See Michaud and Michaud (Efficient Asset Management, Ch. 7, 2008) for an extensive discussion.  The Michaud optimization and rebalancing rules are protected by U.S. patents and patents pending.
[2] Patent pending.