On 21 April 2021, before a single company had filed a sustainability statement under the new European rules, the European Commission put a price on them. Its impact assessment said that the option it preferred "would add EUR 3,6 billion in annual incremental costs compared to the previous option." Five years later, that sentence is still the most precise euro figure in the public record about what sustainability reporting costs. It is a forecast. It was written before the law existed, about a version of the law that has since been rewritten twice.
This is the opening piece of the sustainability chapter of the CareerOn Industry Atlas. The chapter asks where the work and the money are in corporate sustainability reporting, the regime Europe calls the CSRD, and which of those jobs a candidate can rehearse. We start with the money because it is the question every employer, adviser and graduate asks first, and because it is the question the public record answers least well. We print what each document measures beside what it says. We keep a proposal, an adopted law and a supervisor's finding apart. And we add nothing across documents that measure different things.
Every euro figure the public record prints about sustainability reporting is a Commission forecast made before reports were filed, and inside those forecasts the largest single line is the auditor, not the data; the supervisors who have since read real reports count examinations and samples, never money, and the 2026 law that now governs the regime was adopted without a published count of the companies it covers.
What the documents can and cannot count
We read nine dated documents across four families, and opened every one of them live on 2 October 2026. Three are adopted European law: the CSRD itself, Directive (EU) 2022/2464 of 14 December 2022; the so-called stop-the-clock Directive (EU) 2025/794 of 14 April 2025; and the Omnibus I Directive (EU) 2026/470 of 24 February 2026, which rewrote the scope. Three are Commission analysis: the 2021 impact assessment, the Omnibus proposal COM(2025) 81 of 26 February 2025, and the staff working document SWD(2025) 80 published with it. Two come from the European Securities and Markets Authority, ESMA, which coordinates the national market supervisors. The last is a study by the French market authority, the AMF, of October 2025.
The EU legal texts and Commission analyses are served by EUR-Lex, which shows a screen to automated readers, so we read those six in a real browser and checked each quoted passage word for word. The ESMA and AMF documents are PDFs, which we fetched directly.
The family matters more than the date. An adopted act tells you what a company must do. A proposal tells you what the Commission wanted a company to do. An impact estimate tells you what the Commission expected that to cost, using averages, before anyone did it. A supervisory finding tells you what an examiner saw in reports that were actually filed. These are four different kinds of truth. Most commentary on the cost of the CSRD quietly treats them as one.
Here is the first finding, and it is a negative one. None of the nine documents prints a measured figure for what companies have spent on sustainability reporting. Not the law, which defines the work and not its price. Not the supervisors, who read the output. Not the French study, which says on its own pages that it "Does not reflect AMF supervisory actions" and describes a sample it does not claim is representative. In the documents we admitted, the only euros are forecasts. Absence in nine documents is not proof that no measurement exists anywhere; it is proof that none of the public documents a reader is most likely to be quoted has one.
The price came before the law
Read the five legal and Commission documents in date order and the sequence is unusual. The cost was estimated in 2021. The law was adopted in 2022. In 2025 the Commission proposed to shrink it, and the co-legislators froze part of it. In 2026 they adopted a smaller version. Each step is a different legal state, and the figures attached to them do not travel from one step to the next.
Five documents, four legal states: the price came before the law
| Date | Document | Legal state | What it settles |
|---|---|---|---|
| 21 Apr 2021 | SWD(2021) 150 | Impact estimate | EUR 3,6 billion a year in added costs for one policy option, before any report was filed |
| 14 Dec 2022 | Directive (EU) 2022/2464 | Adopted act | Limited assurance means fewer tests than reasonable assurance |
| 26 Feb 2025 | COM(2025) 81 | Proposal | Companies in scope reduced by about 80%, by the proposal’s own estimate |
| 14 Apr 2025 | Directive (EU) 2025/794 | Adopted act | Reporting postponed by two years for the companies concerned |
| 24 Feb 2026 | Directive (EU) 2026/470 | Adopted act | Scope above EUR 450 000 000 turnover and 1 000 employees; reasonable assurance removed; no count of companies printed |
Sources 4 European Commission · 1 EUR-Lex (Publications Office of the European Union) · 5 European Commission · 2 EUR-Lex (Publications Office of the European Union) · 3 EUR-Lex (Publications Office of the European Union)
The 2022 act set the terms of the work. One sentence in it carries more economic weight than it seems: "In a limited assurance engagement, the auditor performs fewer tests than in a reasonable assurance engagement." Limited assurance was the starting point, and the law left open a later step to the heavier standard. That sentence defines the auditor's workload, and so its fee, without naming a number.
In April 2025, the co-legislators decided that for the companies due to report in the following waves, "the sustainability reporting requirements for those undertakings should be postponed by two years." The act says nothing about how many companies that was or what they saved.
The 2025 proposal is where the most widely repeated number in this debate comes from. It says: "The number of undertakings subject to mandatory sustainability reporting requirements would be reduced by about 80%." That is the proposal's own estimate of its own effect, dated 26 February 2025. The adopted law of February 2026 did not keep the proposal's thresholds. It limits the regime to "undertakings with a net turnover exceeding EUR 450 000 000 and an average of more than 1 000 employees during the financial year." It prints thresholds. It does not print a count of the companies above them.
So the honest reading is this: the 80% belongs to a text that was never adopted in that form. Anyone who credits the 2026 law with an 80% cut in scope is quoting a proposal as if it were a statute.
Inside one forecast, the auditor is the largest line
The working document published with the 2025 proposal is the most detailed cost picture available, and it is worth reading closely, because its figures are often quoted without their base. It estimates that "some 35 000 out of the 45 000 companies would be exempted from CSRD reporting, and 10 000 remain in scope." It then adds: "these undertakings would save total audit costs around EUR 2 billion per year."
Inside the 2025 forecast, the auditor is the largest line
| Line | As printed | Base of the figure |
|---|---|---|
| Companies in scope before the revision | 45 000 | Estimate, SWD(2025) 80 |
| Companies exempted | 35 000 | Estimate against the 2025 proposal |
| Companies left in scope | 10 000 | Estimate against the 2025 proposal, not the 2026 law |
| Audit costs saved | around EUR 2 billion per year | Estimate for the exempted companies only |
| What the adopted law says about assurance | Reasonable assurance removed to avoid higher assurance costs | Adopted act, Directive (EU) 2026/470 |
Sources 6 European Commission · 3 EUR-Lex (Publications Office of the European Union)
Three things about those figures should travel with them. First, they are estimates for the proposal's scope, not the adopted law's. Second, they are built from average cost estimates taken from an earlier study, applied to a population, with an assumed 30% rate of subsidiaries reporting. Third, the audit saving applies to the exempted companies only. None of this makes the figures wrong. It makes them a forecast with stated assumptions, and it means they rank lines inside a forecast rather than measure spending.
Within that forecast, though, the ranking is clear: the single largest saving the Commission named was the auditor's. The adopted law confirms that the co-legislators read it the same way. In removing the step to the heavier standard, the 2026 directive gives its reason in plain words: "To avoid an increase in the costs of assurance for undertakings, the requirement to adopt reasonable assurance standards should be removed." Of all the lines in the cost of reporting, assurance is the one the law itself moved to control the price.
That is the mechanism under the money. The cost of sustainability reporting is not mainly a cost of collecting data, in the documents we read. It is a cost of having that data checked by someone independent, to a standard the law defines. When Europe wanted to make the regime cheaper, it did not only remove companies. It capped how hard the checker has to look.
What the supervisors counted instead of money
If the forecasts are the only euros, what do we know from reports that were actually filed? The supervisors have now read the first wave, and they counted carefully. They did not count money.
The supervisors who read real reports counted reports, not euros
| What the supervisor counted | Measure | Scope | Publisher, date |
|---|---|---|---|
| 263 examinations of CSRD sustainability statements; 25% were prepared voluntarily | Examination count | EEA, listed issuers | ESMA, May 2026 |
| Materiality disclosures read in a sample of 91 issuers from 23 Member States | Sample study | EEA, listed issuers | ESMA, October 2025 |
| First French statements read; the study states it does not reflect supervisory actions | Sample study, not representative | France, listed companies | AMF, October 2025 |
Sources 7 European Securities and Markets Authority (ESMA) · 8 European Securities and Markets Authority (ESMA) · 9 Autorité des marchés financiers
ESMA's report on 2025 enforcement, published on 7 May 2026, records the first full year in which national supervisors examined CSRD statements: "Of the 263 examinations of sustainability statements prepared in accordance with the CSRD/ESRS, 25% related to statements prepared voluntarily by issuers who were not yet subject to the ESRS." A quarter of the reports examined were filed by companies that were not yet required to file them. That is a signal of how much effort companies put in before the law reached them, and it is still not a cost.
On 14 October 2025, ESMA published a fact-finding study of how companies explained materiality, the judgement at the centre of the regime about which sustainability topics matter enough to report. "The analysis was performed on a sample of 91 issuers from 23 Member States." A sample of listed companies, read for quality. No euros.
The AMF's study of the first French statements reads the same way. It offers observations on what French listed companies published and how, it says it does not reflect supervisory actions, and it does not claim statistical representativeness. It is useful for what good looks like. It is silent on what good costs.
Taken together, the supervisory record tells a reader where scrutiny is going. It is going to the judgement behind the report, especially materiality, and to whether a statement can be defended. That is the same place the forecast put the money.
The counter-case, kept in the piece
We tested three hypotheses before writing, and one of them failed. We expected the adopted 2026 law to come with an official count of companies in scope. It does not. The only counts in the record, the 80% and the 10 000, belong to the 2025 proposal. We print them with that date and that label, and we have not invented an adopted figure to replace them.
The claim that the auditor is the largest line also needs its limits. It rests on one estimate, for the exempted companies only, under a stated assumption about subsidiaries. A company that already had strong data systems may find that assurance is its largest bill. A company that had none may find that building the data is. The documents we admitted cannot settle that for any single company, and we do not claim they can.
Finally, the absence of measured spend is a finding about the public record, not about the world. Consultancies, audit firms and industry bodies may hold such figures. Until a dated document prints one, we will not.
What this means for your career
If you are a candidate, the money evidence points at two kinds of named, paid work. The first is assurance: the independent check the law defines and the Commission priced as the largest line. That is the work of a statutory auditor or independent assurance provider, and the 2026 law has kept it at the limited standard, which shapes how much of it there will be. The second is the judgement supervisors examine most closely, materiality, and the ability to defend it in front of an auditor and an examiner. The supervisors also point at a third role, the supervisory enforcement examiner who reads the statements, which is a public-sector career with its own path.
CareerOn has two simulations for this field today. ESG analyst: CSRD reporting and double materiality rehearses the materiality judgement. ESG analyst: non-financial reporting under CSRD rehearses the reporting work itself. We have to be exact about what this piece can claim for them. The supervisory study names materiality disclosures as what ESMA read, but none of the nine documents we admitted describes double materiality as a task a company performs, and none describes data collection or gap analysis before assurance. So in this piece both simulations are unplaced: we offer them as practice for the regime, and we do not say that the money evidence places them. The third piece of this chapter, on the work itself, reads the supervisory findings in full and will test both placements again.
If you are an employer, there is no public benchmark of what reporting should cost you. Any budget you set is your own, and a figure from a 2021 or 2025 forecast is not a quote. If you are hiring, the scarce skill in these documents is not filling templates. It is making a judgement about what matters and being able to defend it.
CareerOn does not yet simulate the statutory auditor or independent assurance provider, or the supervisory enforcement examiner. Both are recorded as catalogue gaps. That is our gap, and we name it.
What we could not prove
We found no admitted document that measures what companies have spent on sustainability reporting. We found no official count of the companies the 2026 law covers. The cost figures we print are forecasts, made for a proposal that was not adopted in that form, with averaged inputs and an assumed subsidiary rate. The supervisory counts are of listed issuers only, and the French study is not representative. If a dated document measures the spend or counts the scope of the adopted law, we will add it and say what it changes.
The CareerOn view, which the desk signs: read every cost forecast as a forecast, and read the 2026 law for what it prints, which is thresholds. The work that the money, the law and the supervisors all point to is the same work: an independent check, and a judgement that can survive it. The next piece of this chapter follows that work into the firms that do it.