Voluntary carbon market price history: reading the 2021-2024 reported averages
Four fields show Transaction year, Report edition, Sample plus statistic, and Currency plus unit. A separate line states Historical average is not equal to current quote. The graphic identifies information needed to interpret a reported statistic; it does not show a numerical price series or recommend a purchase.
Diagram text description
Ecosystem Marketplace’s reported annual averages provide a bounded view of voluntary carbon market price history for 2021–2024. They show a higher reported average in 2022 than in 2021, followed by lower annual observations in 2023 and 2024. These are results from changing transaction samples, rather than the price path of one unchanged credit.
For a procurement or sustainability team, the useful task is to read each observation with its transaction year, report edition, unit and population. A historical average can provide context for a discussion. It cannot supply the price of an identified credit available for purchase now.
This guide therefore keeps the four annual observations separate from preliminary year-to-date data and from current quotations. It also retains a published disagreement over the 2024 value. The comparison worksheet later in the guide shows how to record what is known before deciding whether a historical statistic is relevant to a proposed purchase.
What the 2021-2024 observations show
The annual observations below come from two editions of Ecosystem Marketplace’s State of the Voluntary Carbon Markets reporting. The 2023 edition supplies 2021 and 2022; the 2025 edition supplies full-year 2023 and 2024. That edition boundary matters when reading the sequence.
| Transaction year | Report edition | Reported price: USD/tCO2e | Annual respondents | Source qualification |
|---|---|---|---|---|
| 2021 | 2023 edition | $4.04 | 151 | Volume-weighted reported average |
| 2022 | 2023 edition | $7.37 | 113 | Same report block as 2021 |
| 2023 | 2025 edition: new report block | $6.71 | 97 | Full-year observation |
| 2024 | 2025 edition | $6.34 in table/body; $6.37 in insight bullet | 82 | Published disagreement remains unresolved |
Table 1, shown one row at a time. Every field is retained.
Row 1
- Transaction year
- 2021
- Report edition
- 2023 edition
- Reported price: USD/tCO2e
- $4.04
- Annual respondents
- 151
- Source qualification
- Volume-weighted reported average
Row 2
- Transaction year
- 2022
- Report edition
- 2023 edition
- Reported price: USD/tCO2e
- $7.37
- Annual respondents
- 113
- Source qualification
- Same report block as 2021
Row 3
- Transaction year
- 2023
- Report edition
- 2025 edition: new report block
- Reported price: USD/tCO2e
- $6.71
- Annual respondents
- 97
- Source qualification
- Full-year observation
Row 4
- Transaction year
- 2024
- Report edition
- 2025 edition
- Reported price: USD/tCO2e
- $6.34 in table/body; $6.37 in insight bullet
- Annual respondents
- 82
- Source qualification
- Published disagreement remains unresolved
The 2024 discrepancy appears on the same report page: Table 1 and the surrounding narrative give $6.34 per tonne of carbon dioxide equivalent, while an insight bullet gives $6.37. The table above attributes both values. It does not average them, declare one corrected or assume an erratum has settled the difference.
The earlier edition also says its 2021 and 2022 figures can change as respondents submit additional data. Preserve the edition when reusing a number, because a value from another publication may reflect a different reporting snapshot.
Both source reports were consulted on 8 November 2026 (Asia/Shanghai). That consultation date describes access to those documents, rather than the date of the transactions.
Read the sequence descriptively: the reported average rises between the first two observations and is lower in the later annual observations. The figures do not establish investment returns, a like-for-like price change or the cause of a market movement. None is a 2025 or 2026 spot quotation.
Why no single price describes every carbon credit
An all-credit survey average describes the transactions included in that survey. It does not identify one project, methodology, vintage, delivery arrangement or credit that a seller will supply. Before comparing a quoted price with the historical table, ask what each number measures.
An identified-credit quotation is a different item: it needs the proposed units, quantity, price date, currency, delivery conditions and any fees or services. A vendor pricing model or index also needs its own method, inputs and category coverage. This guide has not verified numerical outputs from any vendor model.
A compliance allowance concerns a specified compliance programme and instrument. Do not substitute its price for the voluntary-credit observations simply because both are discussed as carbon prices. Establish the instrument and intended use first.
These distinctions explain why a broad historical average is a weak basis for deciding that a particular offer is expensive or cheap. The offer may concern a narrow category or vintage that the aggregate does not isolate. Conversely, a category label alone does not show that the offer and historical sample have matching terms.
Keep the comparison’s purpose explicit. For background context, a source-labelled annual observation may be useful. For a buying decision, the missing evidence is an actual, dated offer for identified credits and a reason those credits meet the buyer’s intended requirements. The historical table does not provide that evidence.
Transaction year is not publication year or vintage
Transaction year, report edition and credit vintage are separate fields. Transaction year describes when a trade took place. Report edition identifies the publication presenting the data. In this guide, the 2025 edition reports annual transaction observations for 2023 and 2024; its publication year does not turn those observations into 2025 prices.
The 2025 report’s glossary defines vintage by the year in which a project’s emissions reductions or removals were determined to occur, including estimated future occurrence where relevant. Vintage need not match issuance year: issuance can follow the mitigation, and some standards issue against future estimates. Neither date should be inferred from the year of a reported trade.
The upper row connects full-year 2023 transactions with the 2025 report edition that describes them. A separate lower band identifies vintage as the mitigation year, including estimated future occurrence where relevant under the report glossary. The diagram assigns no common vintage to the surveyed trades and excludes preliminary year-to-date data.
Diagram text description
Preliminary data need another boundary. The 2023 edition gives a preliminary 2023 year-to-date average of $6.97 per tCO2e from 74 respondents. Its footnote refers to transactions through 21 November 2023 but says they were reported to EM by 20 November 2023. Those cutoffs conflict. The same note says reporting frequency meant not all respondents were represented through that period.
Keep that preliminary observation outside the four annual rows. Comparing $6.97 with the later full-year $6.71 does not isolate a late-year price fall or establish a particular correction. Period coverage, respondent coverage and reporting snapshot differ.
When saving an extract for colleagues, label it with the transaction period and edition rather than a bare year. Record vintage separately only when the source or proposed credit identifies it. This prevents a report cover date from silently becoming a statement about the units being bought.
How sample and weighting affect the comparison
The reports use voluntarily supplied transaction data, typically from sell-side participants such as project developers, investors and intermediaries. This is a substantial reporting exercise, but the methodology expressly says it is not a complete census of trading activity. Participation and the transactions reported can change.
For the average price, EM uses transactions that report both price and volume, weighting prices by transaction volume. A larger reported transaction therefore contributes more to the average than a smaller one. To estimate total transaction value, the method applies that mean to total reported volume, including volume with no associated price. The value estimate is consequently not a complete independent sum of observed transaction prices.
A priced-trades input leads to the volume-weighted mean. Reported volume without a price sits on a separate branch. The mean and that volume feed an estimate of value. The diagram shows the reports’ method conceptually, without inventing transaction inputs or a new calculated market average.
Diagram text description
| Comparison issue | What the source establishes | Limit to retain |
|---|---|---|
| Price calculation | Volume weighting of trades with price and volume | No simple average of respondent quotes |
| Estimated total value | Mean extended to reported volume without prices | Some value is estimated using the mean |
| Annual participation | 151, 113, 97 and 82 respondents for the four years | Respondent composition changes |
| Two-year overlap | 61 respondents reported both 2023 and 2024 | Same respondents do not imply the same credit basket |
Table 2, shown one row at a time. Every field is retained.
Row 1
- Comparison issue
- Price calculation
- What the source establishes
- Volume weighting of trades with price and volume
- Limit to retain
- No simple average of respondent quotes
Row 2
- Comparison issue
- Estimated total value
- What the source establishes
- Mean extended to reported volume without prices
- Limit to retain
- Some value is estimated using the mean
Row 3
- Comparison issue
- Annual participation
- What the source establishes
- 151, 113, 97 and 82 respondents for the four years
- Limit to retain
- Respondent composition changes
Row 4
- Comparison issue
- Two-year overlap
- What the source establishes
- 61 respondents reported both 2023 and 2024
- Limit to retain
- Same respondents do not imply the same credit basket
Even the overlapping group can transact different projects, categories, vintages and quantities in different years. Its existence does not turn the full annual sequence into a constant-quality index. The report’s discussion of that group also contains ambiguously attached percentage changes, which are not used here.
Use the published mean rather than trying to reconstruct it from rounded volume and value totals. Preserve the disclosed $6.34/$6.37 disagreement instead of using rounded arithmetic to overrule the source. These checks keep the comparison tied to what was actually reported.
Price, turnover and retirements answer different questions
Price, traded volume, estimated transaction value and registry activity answer different questions. Their units may look compatible on a chart, yet their populations and meanings differ. The 2025 report separates respondent transaction data from issuance and retirement data drawn from ten listed registries.
| Measure | Unit | Population or basis | Question answered |
|---|---|---|---|
| Reported average price | USD/tCO2e | Respondent trades reporting price and volume | What weighted price did those reported trades show? |
| Reported traded volume | tCO2e, often displayed in millions | Transactions supplied by respondents | How much trading did respondents report? |
| Estimated transaction value | USD | Reported volume valued using the stated mean | What value does that estimation method produce? |
| Issuance and retirement | Credit quantities expressed in tCO2e | Separate registry records | What unit events did those registries record? |
Table 3, shown one row at a time. Every field is retained.
Row 1
- Measure
- Reported average price
- Unit
- USD/tCO2e
- Population or basis
- Respondent trades reporting price and volume
- Question answered
- What weighted price did those reported trades show?
Row 2
- Measure
- Reported traded volume
- Unit
- tCO2e, often displayed in millions
- Population or basis
- Transactions supplied by respondents
- Question answered
- How much trading did respondents report?
Row 3
- Measure
- Estimated transaction value
- Unit
- USD
- Population or basis
- Reported volume valued using the stated mean
- Question answered
- What value does that estimation method produce?
Row 4
- Measure
- Issuance and retirement
- Unit
- Credit quantities expressed in tCO2e
- Population or basis
- Separate registry records
- Question answered
- What unit events did those registries record?
A retirement is not itself a sale-price observation. The report’s glossary describes a credit passing through different holders, or being sold directly to its user, before retirement is recorded. A registry retirement total therefore cannot be treated as the survey’s traded volume.
Do not add registry retirements to respondent trades to create a larger market-transaction total. The records can concern related credits at different events, and the survey is not a census. Likewise, a high retirement count alone cannot establish a transaction price or a change in it.
For a briefing, select the measure that answers the actual question. Use the price statistic to describe the reported mean and a retirement record to describe retirement activity, while retaining the source and population of each.
Check comparability before using historical prices
The following worksheet is an explicitly fictional teaching case. An invented procurement team has a historical report extract and an incomplete seller note. No offer price or actual buyer decision is asserted. The team wants to know whether the annual average can establish the cost of its intended purchase.
| Comparison field | Historical evidence in the fictional file | Proposed-purchase evidence | Consequence or next request |
|---|---|---|---|
| Statistic | Reported volume-weighted annual average | Dated unit quotation absent | Procurement requests a complete offer |
| Period and edition | Full-year 2024, EM 2025 edition | Quote date and delivery period absent | Record both periods before comparison |
| Currency and unit | USD/tCO2e; $6.34 table/body, $6.37 bullet unresolved | Currency, unit basis and fees absent | Retain discrepancy; obtain terms |
| Population | 82 annual respondents; changing sample | Specific seller and units not established | Identify the actual offer |
| Category and vintage | Aggregate across reported credits | Category and vintage absent | Obtain matched credit identifiers |
| Geography | No single project geography isolated by the mean | Project location absent | Record the proposed project location |
| Source status | Published report with a disclosed inconsistency | Incomplete note, not an executable quote | Buyer comparison remains unresolved |
Table 4, shown one row at a time. Every field is retained.
Row 1
- Comparison field
- Statistic
- Historical evidence in the fictional file
- Reported volume-weighted annual average
- Proposed-purchase evidence
- Dated unit quotation absent
- Consequence or next request
- Procurement requests a complete offer
Row 2
- Comparison field
- Period and edition
- Historical evidence in the fictional file
- Full-year 2024, EM 2025 edition
- Proposed-purchase evidence
- Quote date and delivery period absent
- Consequence or next request
- Record both periods before comparison
Row 3
- Comparison field
- Currency and unit
- Historical evidence in the fictional file
- USD/tCO2e; $6.34 table/body, $6.37 bullet unresolved
- Proposed-purchase evidence
- Currency, unit basis and fees absent
- Consequence or next request
- Retain discrepancy; obtain terms
Row 4
- Comparison field
- Population
- Historical evidence in the fictional file
- 82 annual respondents; changing sample
- Proposed-purchase evidence
- Specific seller and units not established
- Consequence or next request
- Identify the actual offer
Row 5
- Comparison field
- Category and vintage
- Historical evidence in the fictional file
- Aggregate across reported credits
- Proposed-purchase evidence
- Category and vintage absent
- Consequence or next request
- Obtain matched credit identifiers
Row 6
- Comparison field
- Geography
- Historical evidence in the fictional file
- No single project geography isolated by the mean
- Proposed-purchase evidence
- Project location absent
- Consequence or next request
- Record the proposed project location
Row 7
- Comparison field
- Source status
- Historical evidence in the fictional file
- Published report with a disclosed inconsistency
- Proposed-purchase evidence
- Incomplete note, not an executable quote
- Consequence or next request
- Buyer comparison remains unresolved
The worksheet shows why the unsupported conclusion is “this purchase should cost the historical average”. A published mean and an incomplete seller note do not describe matching objects. Procurement owns the request for a dated offer; the sustainability reviewer owns the intended-use and credit-identity checks.
If the seller supplies the missing fields, the team can make a more specific comparison. It should still explain differences in category, vintage, quantity, delivery and services rather than expecting an aggregate historical mean to settle them. The report can remain background context while the current evidence does the work of evaluating the offer.
Keep unknowns visible in the handover. A blank quote date is not the report year, a missing project location is not global coverage, and an unspecified vintage is not the transaction year. Filling those gaps with assumptions would create a comparison the evidence does not support.
What this history cannot tell you about a purchase
This history can support a source-specific account of reported 2021–2024 averages. It cannot provide a current eligible-credit quotation, a complete all-time series, a fixed-basket price index, a forecast or a promised return. The four observations do not identify the quantitative cause of their changes.
The next decision
For a proposed purchase, obtain a dated offer identifying the credits, category, vintage, location, quantity, currency, unit basis, delivery and fees. Keep the intended-use review alongside that offer. Record what the historical statistic contributes and which parts of the purchase decision depend on other evidence.
Keep gross emissions separate from any optional credit selection. A price history does not calculate a buyer’s inventory, demonstrate a reduction in that inventory or supply assurance about a public claim.
If the work moves towards commercial reuse, the reports’ notices also need attention: they restrict commercial use of their data unless permitted by Forest Trends or applicable law. Attribution alone does not determine permission. The intended reuse remains unresolved; this guide grants no licence or purchase approval.
