CAIA® ESG screening: when does an impact claim hold up?

On this page

A fund can apply a credible ESG screen without showing that its investments caused a social or environmental improvement. A CAIA® candidate should first establish the rule that selected the holdings, then examine the claimed outcome, the people affected, the size and durability of the change, the investor’s contribution and the chance the reported result is wrong. A high screening pass rate answers a compliance question; it does not establish impact.

Start with the claim the manager actually makes

The CAIA Association’s 2023 Level II study guide listed negative and positive screening and impact investing within its ESG objectives. That older guide establishes the conceptual connection, while the Association says its current curriculum is updated annually and should govern the exam year. The useful skill is classification before calculation. Ask whether the manager promises to avoid particular exposures, select stronger issuers relative to peers, meet a conduct standard, incorporate financially material ESG information into analysis, or achieve a measurable outcome. Those propositions need different evidence. An investor may combine approaches, but a label alone does not make their evidence interchangeable.

A negative screen is a rule for exclusion. It needs a defined universe, a prohibited activity or threshold, a data source, an assessment date and a treatment for borderline cases. The reviewer can compare current holdings with the rule and record exceptions. A relative screen needs a peer group and a ranking method; a firm can score well within a weak sector while still having a large absolute footprint. A norms screen needs a named standard and a policy for allegations, confirmed breaches and remediation. Each is a selection procedure, not a direct measure of changes in the world.

ESG integration asks another question: did an analyst use material environmental, social or governance information in a risk, cash-flow or valuation judgement? A manager might decide that a governance weakness raises downside risk while still holding the issuer at a different valuation. There may be no automatic pass or fail. The joint responsible-investment definitions announced by CFA Institute distinguish screening, integration and impact investing as separate approaches. In an exam answer, identify the decision rule and the evidence for its use before attaching a category name.

Audit the screen without overstating it

Imagine an illustrative portfolio of 20 holdings. A manager says it excludes an activity entirely, but two holdings have that activity under the manager’s own stated definition. Eighteen holdings pass a count-based check, so the count pass rate is 18 divided by 20, or 90%. If those two holdings make up 15% of portfolio value, a value-weighted check gives 85%. These invented figures describe a compliance diagnostic. They do not measure avoided emissions, improved labour conditions or any other outcome. The difference between the two rates also shows why the denominator must be named.

Before declaring a breach, inspect the actual policy. Does it test the whole corporate group or only the listed entity? Is the threshold based on revenue, production, ownership or a data provider’s classification? Does the rule apply to holdings at acquisition, at each reporting date, or after an allowed transition period? A legitimate exception policy may exist. The examiner can give a clean rule and ask for its application; a real diligence note should preserve the uncertainties and request the missing dates and definitions.

The same holdings cannot prove that the manager changed issuer behaviour. Secondary-market ownership may be consistent with a values-based mandate, yet the issuer could have followed the same course without that investor. A pass-rate calculation is useful because it tests whether words match implementation. It is limited precisely because it stops at the portfolio boundary.

Follow the outcome through the causal chain

For an impact claim, write a short chain: investor action, financing or influence received by an enterprise, activity undertaken, output delivered and outcome experienced by people or the environment. A loan disbursed is an investor action. Equipment installed is an output. Reliable access to a service may be an outcome. The number attached to the enterprise’s entire activity is not automatically attributable to the investor. CFA Institute’s discussion of impact measurement stresses the move from intentions and funded activities towards demonstrable outcomes and transparent contribution.

The five dimensions offer a practical interrogation. What changed, including unintended harm? Who experienced the change, and were they already well served? How much occurred in scale, depth and duration? Contribution asks how the investor improved on the likely alternative. Risk asks how actual impact could differ from the forecast. A metric catalogue can help make an output consistently defined; it cannot answer the counterfactual on its own. When the source material refers to IRIS+ and these dimensions, treat them as an organising framework, not a single score that settles every case.

A worked impact example

Suppose an invented private credit fund reports that it lent to a water-treatment operator serving 10,000 households. The operator’s annual report says that all 10,000 received a service. The fund’s promotional line credits its loan with the full total. A candidate should ask what the loan financed and what would have happened otherwise. If an existing lender was ready to renew a facility on similar terms, continuation of all existing service is a weak case for incremental impact from the new fund. If the fund instead financed a documented extension reaching 2,000 previously unserved households, that provides a more focused outcome hypothesis. Neither number is a real observation; both are invented to show the reasoning.

The diligence file would still need dates, service-quality evidence, a baseline for the households, financing documents and a defensible comparison with the likely alternative. Some households might have obtained service through a different provider, while installed equipment might fail or service might be intermittent. The fund could have contributed alongside local government and other lenders. In that case, describing a contribution is more defensible than assigning every outcome exclusively to one investor. A transparent report should say what portion is measured, what is estimated, and what cannot be attributed.

Separate measurement error from attribution error

Measurement asks whether the recorded outcome occurred. Check the unit, reporting period, population, completeness, data collection process and independent assurance where available. A count of installed units can be accurate even if it tells little about whether people used them. Duration matters: a one-month result is weaker evidence for a lasting benefit than follow-up data. Risk includes operational failure, missing respondents, displacement of benefits elsewhere and adverse side effects. A manager may have a sound method and still face uncertainty; the task is to make that uncertainty visible.

Attribution asks whether the investor caused an improvement beyond the counterfactual. Compare the investment with plausible alternatives: conventional finance, other investors, expected regulation or a project already funded. For public-market holdings, the mechanism may be engagement rather than new financing, and the manager should show the intervention and the issuer response. Do not assume that every purchase of a qualifying issuer created additional output. Equally, do not dismiss all influence because it is difficult to quantify. Identify the mechanism and the evidence available for it.

If a manager supplies only a positive output total, the correct conclusion is usually narrower than either full endorsement or an accusation of misconduct. The evidence may support that portfolio companies delivered the output while leaving investor contribution unproved. That distinction is particularly useful in an exam vignette: a question may deliberately give a precise metric to tempt candidates into treating numerical detail as causation.

Put the two tests together

A concise diligence answer can use two columns mentally. For the portfolio rule, record the universe, rule, date, holdings and exceptions. For the impact assertion, record the baseline, beneficiary, outcome, investor mechanism, alternative explanation, time horizon and measurement risk. Resolve an inconsistency before making a strong claim. A fund could pass every exclusion yet have no demonstrable additional impact; another could have an impact thesis alongside a screening exception that needs disclosure. The two conclusions should remain separate.

Candidates who want to place this distinction in the wider curriculum can use the CAIA programme page for the programme context and the Omni methodology page for the study approach. In a timed answer, state what the available evidence establishes, name what remains unproved and propose the next document or observation that would change the judgement. That is a more useful assessment than repeating the manager’s preferred label.