Skip to content

New to Omni?·Get 20% off your first purchase with code WELCOME20 at checkout.See plans

ESG screening providers: when does a data gap change the answer?

A screening score can be useful and still conceal missing or estimated inputs. Learn how to compare providers, test coverage, and explain a defensible decision.

ConceptOmni Curriculum Team6 min read

Two unequal stacks of blank papers on a desk with a magnifying glass and navy notebook.
On this page

An ESG screening result is a provider's constructed answer to a defined question, using the information it can obtain. When data are missing, estimated or old, a neat score may appear more comparable than its inputs. A candidate should identify the provider's objective, trace the most important inputs, and test whether an apparent difference between issuers survives those limitations.

Start with the question the provider answers

A screen can flag conduct against a stated standard, exclude a category of activity, or rank companies on a chosen measure of ESG risk. A rating can then combine indicators into a summary. These operations are related but they need not answer the same question. A conduct screen asks whether a condition is met. A financially focused risk assessment asks how exposure and management of an issue might affect the issuer. A footprint measure describes an activity or outcome. Treating all three as interchangeable turns a provider choice into an unnoticed change in the mandate.

CFA Institute places ESG data providers alongside analysts' own research and information from non-profit organisations as possible sources for investment analysis. Its certificate curriculum includes ESG analysis, valuation and integration, which is why a candidate needs to understand the construction of a data product rather than simply memorise that data products exist. In an exam explanation, name the purpose of the measure before evaluating a score. If the purpose is unstated, ask what a higher score is supposed to mean and whether the comparison is within a sector, across sectors or against a fixed rule.

A useful provider comparison starts with scope. Does the service cover public issuers only, or also smaller or private entities relevant to the task? Does it supply a company-wide score, indicator-level observations, or specific controversy flags? Next inspect input provenance: company disclosure, regulatory filing, external report, estimate or provider judgment. Finally inspect how the service weights indicators, assigns peers and refreshes results. A polished output cannot repair a mismatch between the required decision and the provider's method.

Why apparently complete databases have holes

Missingness has several forms. A field may be blank because an issuer did not disclose it, because the provider could not map the disclosure to its taxonomy, or because the provider does not cover that issuer. An estimate fills an empty cell but does not make it observed. A stale observation has a value but may describe an earlier operating structure. A methodology revision can alter a score even when the underlying company information has not changed. Each condition asks for a different response; a single label such as poor data quality loses that distinction.

CFA Institute's discussion of climate data describes practical checks: compare third-party data with original documents, consider another provider, and understand estimation methods where available. It also notes that ratings are unique to providers and should not simply be placed on a common scale. That is a reason to retain the underlying indicator and its provenance when explaining a decision. If one supplier reports an actual disclosed figure and another supplies an estimate, the analyst should not average the two and call the result verified.

Coverage can also create a selection effect. Imagine a screen requiring a reported metric. If larger issuers have more complete disclosure, the eligible set may become tilted towards them even when the mandate never asked for a size preference. The screen may have selected disclosure capacity as well as the ESG characteristic of interest. An analyst can explain that consequence without asserting that undisclosed issuers perform well or badly. Missing is an information state, not a favourable or unfavourable performance grade.

A worked comparison with invented numbers

Consider an entirely illustrative mandate that needs to compare two issuers on an environmental operating measure and governance controls. Both indicators use a hypothetical scale from zero to one hundred, and the mandate gives each an equal weight. Issuer A has a disclosed environmental value of 80 and a documented governance value of 60. Issuer B has no reported environmental value; Provider One estimates it at 70 and records a governance value of 70. On those inputs, A's simple average is 70 and B's simple average is also 70. A superficial comparison calls them equivalent.

Now imagine Provider Two estimates B's missing environmental value at 40 because it uses a different peer set. Its result for B becomes 55, while A remains 70. None of those invented numbers is a market observation or a description of a real provider. The point is that B's apparent standing changes solely because the analyst accepted a different estimate for one unobserved cell. A sound answer states the two calculated results, identifies the estimated input, and explains that the decision is sensitive to a provider assumption. It does not announce that B's true environmental performance is 40, 70 or anything between.

The equal weights themselves are another assumption. If the mandate gives more weight to the environmental indicator, the uncertainty in B's missing value matters more. The candidate should say which assumption drives the result before changing a threshold or advising a decision. When a screen is meant to apply a strict data-availability rule, the right treatment may be to mark B for further investigation rather than to insert either estimate. When estimates are allowed, document the chosen method and show how the answer changes under a plausible alternative.

A repeatable decision check

First write the mandate in operational language: the unit being assessed, the issue, the comparison group and the required threshold. Then identify whether the provider's measure captures financial exposure, issuer conduct or an outcome. Do not infer the definition from a friendly label. Read the methodology and ask whether a higher number always means the same thing in the rows being compared. Sector-relative ranks and absolute measures answer different comparisons even if both are called ratings.

Second trace the decisive fields. A complete audit of every cell may be impossible under time pressure, so begin with fields that change the decision near a cutoff. Record whether each is reported, externally sourced, estimated or unavailable, and whether the measurement period is still useful. If the input is a company statement, inspect whether the underlying document actually says what the vendor field claims. If a controversy flag appears, check its date and whether it refers to the issuer under review. This is verification of the specific decision, not a claim that one source is always superior.

Third run a sensitivity check. Replace a material estimate with another defensible assumption, or set it aside, and see whether the ranking or pass/fail result changes. If it does, describe the decision as conditional on that input. If it does not, say the result is stable under the tested assumptions, without pretending all uncertainty is gone. Where a methodology changed, compare the method versions before attributing a score movement to issuer behaviour. These checks turn a database export into an explained conclusion.

What exam answers often miss

A common error is to interpret disagreement as proof that one provider has made a mistake. Providers may select different issues, peer groups, weights or estimation rules, and may be pursuing different objectives. Disagreement is a prompt to inspect construction. Another error is to treat a disclosed value and an imputed value as equally evidenced because they occupy the same column. A third is to confuse a screen's pass result with a comprehensive assessment of a company. The screen answers its rule; it does not settle every ESG question.

For revision, practise explaining an output in this order: objective, input, method, missingness and decision sensitivity. This sequence works for a provider comparison and for a single provider whose rating changes. The ESG programme page places this concept within the certificate, while the methodology page describes Omni's study approach. The transferable exam skill is to make the hidden assumptions visible before drawing a conclusion from the score.

Sources

  1. Sustainable Investing Certificate Curriculum | CFA Institute · cfainstitute.org · Retrieved
  2. Environmental, Social, and Governance (ESG) Considerations in Investment Analysis | CFA Institute · cfainstitute.org · Retrieved
  3. Climate Data in the Investment Process: Challenges, Resources, and Considerations | CFA Institute · blogs.cfainstitute.org · Retrieved

Explore ESG study resources

Educational content, not investment advice. This article is written to help candidates prepare for professional exams. It is not a recommendation to buy, sell or hold any security, and it does not take your personal circumstances into account.

Omni Finance Academy is an independent exam-preparation provider and is not affiliated with or endorsed by CFA Institute, GARP, CAIA Association or CFP Board. CFA®, FRM®, CAIA® and CFP® are trademarks of their respective owners. Full disclaimer.