Portfolio
Ask most allocators to demonstrate that a portfolio is diversified and they will produce a correlation matrix. It is a reasonable place to start and a poor place to finish. Correlation describes how two return series happened to move together over a chosen window. It says nothing about why, and therefore nothing reliable about what happens next.
The failure mode
The pattern is familiar. A collection of strategies displays comfortably low pairwise correlations for several years. Markets then dislocate, financing tightens, and every strategy that depended on leverage or liquidity moves in the same direction at the same time. The correlations that mattered were not the ones in the matrix; they were the shared dependence on conditions that had never previously been tested in the sample.
This is not an argument against measurement. It is an argument that measurement must be accompanied by an understanding of cause.
Mapping return drivers
We describe every manager in terms of what actually produces their return, and we look for genuine variety across that map rather than across asset class labels. The broad categories we use are: risk premia harvested by bearing an unwanted exposure; structural inefficiencies created by regulation, index mechanics or mandate constraints; capacity-constrained niches too small for large capital; informational or analytical advantage in under-covered markets; and liquidity provision.
Two managers in completely different asset classes can sit in the same box — a credit manager and an equity manager may both be, fundamentally, selling liquidity to forced sellers. They will behave identically in the environment that matters. Conversely, two equity managers can genuinely diversify each other if one is harvesting a premium and the other exploiting an index-driven inefficiency.
Testing the map
We test the map against scenarios rather than history. For each of a defined set of conditions — a funding shock, an inflation surprise, a sustained low-volatility grind, a crowded unwind, a sovereign event — we estimate how each return driver behaves and check that the portfolio is not concentrated in drivers that all fail together.
The output is not a precise number. It is a list of conditions under which we would expect to lose money across several managers simultaneously, which is exactly the list we need in order to size positions, hold liquidity, and avoid being surprised by our own portfolio.
The cost of over-diversification
Diversification is not free and more is not always better. Every additional manager adds fees, operational risk, monitoring burden and the possibility of a fraud we did not detect. Beyond a certain point, added managers dilute conviction without meaningfully reducing risk — the incremental strategy is usually similar to something already held, because the genuinely distinct opportunities are scarce.
We would rather hold a smaller number of managers we understand deeply and can size with confidence than a long list that produces an impressive schedule and an average outcome.
Illiquidity is not a diversifier
A final caution. Assets that are marked infrequently display low measured correlation to public markets, and this is routinely presented as diversification. It is not. Smoothed valuations understate risk and correlation alike, and the underlying exposure often turns out to be the same economic risk in a less liquid wrapper. We adjust for this explicitly rather than accepting the reported series.