This note introduces a new method for attributing to the signals and constraints that drive quantitative, signal-driven strategies. Quantitative asset managers and their clients seek this type of attribution because signals, signal weights, and constraints are often the central daily decisions driving performance.

The attribution landscape is broad and varied,² but few systems attribute to these decisions. Examples that do include Grinold and Easton (1998), Grinold (2006) and Menchero and Poduri (2008). The more recent two were strong advances but are regression-based. In practice, such systems often leave half or more of the risk and return unexplained, and deliver ‘explained’ contributions that PM’s can see are incorrect. Attempting improvement by performing additional regressions reduces residuals but increases model error in contribution estimates. Segmenting the investment process into steps typically introduces ambiguity due to order-dependence.

Complete Attribution (CAtt) solves these problems and delivers complete and error-free attribution to quantitative management decisions. It does not redistribute performance and has no plug. By exploiting the manager’s complete knowledge of the quantitative process³, it also avoids residuals, model error in contribution estimates and multi-collinearity.

CAtt then gives asset managers and their clients a reliable transparency into their portfolios that they have not previously had. It answers questions such as which signals are doing well and which badly in this portfolio, have constraints helped or hurt, is the turnover about right, is the risk budget right, is that new signal having an impact yet, which individual signal performances are trending up or down? These answers benefit asset managers in two ways. First, the superior understanding of the impact of past decisions promotes better future returns. Second, it raises the level of conversation with prospects and clients, further enhancing asset gathering and asset retention.

These are lofty claims, and CAtt is a new approach. We therefore mention several grounds for confidence that Complete Attribution behaves as we claim.

Expert, fully informed endorsements

Ron Kahn, Global Head of BlackRock Scientific Equity Research, and Dan diBartolomeo, owner of Northfield, have both examined the mathematics and understand and agree with the approach.

Examples, examples

In a longer document, we demonstrate CAtt’s behavior in multiple investment scenarios. These include: adding and dropping signals in portfolios with high and low turnover; industry and factor constraints; market neutral vs. strongly and weakly binding long-only constraints; leverage constraints; factor timing; quadratic penalties; transformations in the combining of signals; and trade overrides. In each case, prospective clients can see CAtt capturing all the behavior one expects in these scenarios.

Rigorous daily checking

CAtt separately determines the risk and return contributions of each signal and each linear constraint, and the collective contributions of the non-core features. Each day, for each portfolio, CAtt performs a simulation and checks that these contributions add to their performance when acting together.

Independent sanity checks

PMs can, independently of CAtt, run simulations of individual signals and compare the IR rankings to the return contributions reported by CAtt.

Live demonstrations

Mayfield and Northfield provide live demonstrations of CAtt.

Head-to-head comparisons

We provide direct comparisons of CAtt results with those of Grinold’s holdings-based attribution, and traditional risk-factor-based attribution.

Customized tests

We invite prospective clients to design their own scenario in which they would like to see CAtt perform. We customize demonstrations to your interests and needs.

Free trial

We offer a trial for clients to run CAtt in their own environment, on their own portfolios, and completely under their control.