The Composite That Only Contains Winners
An equity boutique launched a small-cap strategy four years ago and funded fourteen accounts. Three have since closed: two after performance-driven redemptions, and one because the client's IPS prohibited tobacco holdings and the PM declined to carve them out. For marketing, the PM proposes a 'since inception' record built from the eleven surviving accounts, plus her six-year record from a previous employer (a similar but not identical strategy, different platform, different capacity), plus a pro forma showing the new institutional share class at 55bp instead of 85bp. The draft headline reads '22.1% annualised over 10 years, 11 accounts, no down years.' The firm claims compliance with GIPS and has no written policy on composite construction, portability, or hypothetical performance.
1. Identify each distinct issue in the draft and say what the applicable principle requires for each.
Read the worked answer
1. Removing terminated accounts from the composite. The composite must include all actual, fee-paying, discretionary portfolios managed to the strategy, and terminated portfolios must remain in the composite for the periods during which they were managed. Dropping the two performance-driven closures is survivorship bias injected into a live track record - the most serious issue in the draft, because it is a misrepresentation of actual results rather than a presentation error.
2. The tobacco-restricted account. Client-imposed restrictions that prevent implementation of the strategy make a portfolio non-discretionary for that strategy, so exclusion from the composite is correct. The defect is that the exclusion was not governed by a documented policy applied consistently - the same test must apply to every account, in advance, not case by case after the fact.
3. The prior-firm record. Portability is permitted only where the decision-makers and the decision-making process moved to the new firm, the strategy is substantially the same, and the supporting records are available. 'Similar but not identical', different platform and different capacity is a direct challenge to the 'substantially the same' condition. As drafted, linking the two is not supportable.
4. The pro forma share class. A lower-fee class with no history produces hypothetical performance. It may be shown only under a written policy, clearly labelled as hypothetical, and it must not be linked to or blended with actual composite returns in a single headline figure.
5. '22.1% over 10 years' and 'no down years'. The firm has four years of history; the ten-year figure is an arithmetic blend of actual, prior-firm and hypothetical results. 'No down years' over four years with eleven accounts is a selected statistic; required disclosures include benchmark, composite and firm assets, number of portfolios, dispersion, and three-year ex-post standard deviation net of fees.
2. What can still be presented, and how, once these problems are corrected?
Read the worked answer
Presentable:
- A properly defined composite covering all fourteen accounts for the full four years, including the two terminated accounts through their last full measurement period, and excluding the restricted account under a documented non-discretionary policy.
- Gross- and net-of-fee returns for that composite, against a stated small-cap value benchmark, with composite and firm assets, portfolio count, internal dispersion, and three-year annualised ex-post standard deviation.
- The prior-firm record as supplemental information, clearly labelled, separately presented, not linked arithmetically to the current composite, and only if the firm can support the portability conditions - otherwise it should be described in narrative terms with no performance figure.
- The 55bp share class as a clearly labelled hypothetical illustration, under a written policy, shown alongside the actual 85bp net returns rather than as a headline.
Not presentable under any labelling: the blended '22.1% over 10 years' headline, or any figure that adds actual, prior-firm and hypothetical results into one number. A reader cannot un-blend it, and the label does not cure the misimpression.
Process fix before anything is published: adopt written policies on composite construction (including termination and non-discretionary treatment), portability, error correction, and hypothetical/pro forma performance; retain the records supporting every figure; and have compliance sign off on the final deck. A firm claiming GIPS compliance without these policies has a compliance problem independent of any single number.
3. The PM argues: 'The accounts that left did so voluntarily - including them would misrepresent what my current team can do.' How do you respond?
Read the worked answer
The argument confuses why an account left with whether it was managed to the strategy. The composite tests the second question only. Both redemption-driven closures were fee-paying, discretionary, and managed exactly like the surviving eleven for as long as they were open. Removing them because they did badly is precisely the selection the standard exists to prevent - and it is worse in a four-year sample where three of fourteen accounts is over a fifth of the record.
It is also self-defeating on its own terms. If the two closures genuinely reflected market conditions rather than the process, including them costs little and makes the record credible. If including them changes the headline materially - as it will, given the deck's dependence on 'no down years' - then the number the PM wants to publish is the misrepresentation, not the correction.
What I would offer as a middle path: include the accounts in the composite as required, and additionally disclose in the presentation that two accounts were closed following redemptions, with the composite's dispersion showing the range of client outcomes. Transparency about client turnover is legitimate context; deleting the accounts is not. The same logic applies to the tobacco-restricted account: exclusion is fine, silence about the exclusion is not.