## Improving the Analytical Usefulness of the IMF’s COFER Data (tnmea2025014)

## Source details

**Canonical URL:** [Improving the Analytical Usefulness of the IMF’s COFER Data (tnmea2025014)](https://www.imf.org/-/media/files/publications/tnm/2025/english/tnmea2025014.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tnm/2025/english/tnmea2025014.pdf.md)
- [Structured JSON version](/-/media/files/publications/tnm/2025/english/tnmea2025014.pdf.json)

---

### Executive summary and introduction
- COFER (Currency Composition of Foreign Exchange Reserves) is a quarterly IMF survey tracking the currency breakdown of the world’s foreign exchange reserves.
- At present, 149 entities report data.
- Participation is voluntary except for Special Data Dissemination Standard Plus adherents, who are required to complete the survey and disclose their participation.
- COFER data are kept strictly confidential: the IMF does not publish country-level data or breakdowns below the global total, and the names of COFER reporters remain confidential unless the relevant authorities grant explicit consent.

### New methodology: imputing the “unallocated” share
- Objective: impute the current “unallocated” portion of COFER so published currency compositions (in dollars and shares) account for 100 percent of world foreign exchange reserves.
- Rationale:
  - Previously, aggregate totals included an “unallocated” category to account for gaps from nonreporting, incomplete submissions, or discrepancies with the IMF’s International Liquidity database.
  - Imputing provides a complete currency composition and eliminates the need for users to assume the composition of the “unallocated” portion.
- Key facts:
  - The new COFER series—including the imputations—will be published in December 2025, alongside the release of COFER data for the third quarter of 2025.
  - The revised COFER series will include updates from the first quarter of 2000 through the second quarter of 2025, as well as new data for the third quarter of 2025.
  - Beginning with the release of the third quarter of 2025, IMF staff will publish the share of total imputed reserves.

### Methodology: core steps and imputation approaches
- Major imputation approaches used by IMF staff:
  - Stratified mean imputation
  - Carry-forward imputation
  - Mixed method (combination of carry-forward and stratified mean)
- Imputation workflow (steps):
  1. Estimate a country’s total foreign exchange reserves.
     - For COFER reporters: use COFER-reported totals.
     - For countries not reporting to COFER: use the International Liquidity database to estimate total reserves.
     - For countries reporting to neither: use an external official source; if unavailable, carry forward the best available last-known estimate.
     - The previous practice of including the difference between International Liquidity and COFER in “unallocated” has been discontinued.
  2. Place reporting countries into stratified subgroups based on reported currency composition.
     - First definitional split: issuers of currencies in the reserve currency basket (“Reservia”) versus non-Reservia.
     - Reservia consists of 27 countries that, by definition, do not hold their own currency as foreign reserves; these countries account for approximately 50 percent of global foreign exchange reserve holdings.
     - Non-Reservia consists of 164 countries/entities accounting for the other 50 percent.
  3. For non-Reservia reporting countries, algorithm stratifies them into six subgroups based on deviations from the reported global average currency allocation (Subgroup 1 through Subgroup 6), with an explicit example tied to global averages (e.g., 56 percent US dollars, 21 percent euros, and 23 percent other currencies).
  4. Compute unweighted average (mean) currency allocation of each subgroup each quarter and apply that mean to nonreporting countries’ total reserves to impute values.
- Placement of never-reporting countries:
  - Countries that have never reported to COFER are placed into stratified subgroups by IMF staff using public data on trade and financial linkages, geopolitical relationships, qualitative factors, and judgement.
  - The largest part of the unallocated total is attributable to countries that have never reported to COFER; their entire reserves were previously “unallocated” and will now be allocated using stratified mean imputation.
- Confidentiality safeguards and special handling:
  - IMF staff divide countries into large reserve holders and small reserve holders:
    - Small holders: those with foreign exchange reserve holdings less than 2 percent of global foreign exchange reserves.
    - For small holders, carry-forward imputation or mixed method used as needed; changes unlikely to materially affect aggregates.
  - Carry-forward imputation rules:
    - For previous reporters who stop reporting: use most recent reported shares for subsequent missing values.
    - For nonreporting countries who were previous COFER reporters: impute based on most recently reported shares for the first four periods after they stop reporting.
  - Small reserve holders:
    - First four quarters after stopping reporting: carry-forward imputation.
    - Fifth quarter: allocated based on subgroup average (stratified mean imputation), where subgroup mean excludes the nonreporter’s most recently reported shares.
  - Large reserve holders:
    - First four quarters after stopping reporting: carry-forward imputation.
    - Fifth quarter: stratified mean imputation with the nonreporting country’s most recently reported shares included in the group average computation.
  - Large holders with unusual previously reported skew:
    - Use a mixed method (weighted average of group allocation and the large holder’s previous allocation) over a period longer than four quarters to smooth exit and reduce disclosure risk.
    - Example mixed-method weighting: group allocation (20 percent) and the allocation of the large holder in the previous period (80 percent), applied iteratively so the imputed value converges to the subgroup average over time.
  - IMF staff will continuously incorporate external quantitative and qualitative data (e.g., national data publications) to improve imputed values for nonreporters.

### Results summary and key statistics
- IMF staff imputed currency allocations for every country for every quarter from 2000:Q1 to 2025:Q2.
- High-level effects:
  - Total reserves decrease slightly on average; allocated reserves now equal total reserves; unallocated reserves are eliminated.
  - Claims in each currency increase in dollar terms as previously unallocated amounts are distributed across currencies.
  - Changes in share terms are modest and vary by currency and over time.
- Selected statistics and table entries as published in the source (verbatim):
  - The unallocated share declined substantially over time; in 2025:Q2 the “unallocated” portion was $919.3 billion (that is, 7.1 percent of the $12.9 trillion in foreign exchange reserves held globally in that quarter).
  - Table 1. Currency Composition of Foreign Exchange Reserves, 2025:Q2 (as published vs new method)
    - Monetary levels (millions)
      - Total foreign exchange reserves: As Published 12,94 4,765 — New Method 12,893,887 — Difference –50,878 — Difference –0.39%
      - Allocated reserves: As Published 12,025,455 — New Method 12,893,887 — Difference 868,432 — Difference 7. 2 2 %
      - Claims in US dollars: As Published 6,773,338 — New Method 7, 3 6 4 , 6 815 — Difference 591,343 — Difference 8.73%
      - Claims in euros: As Published 2,540,427 — New Method 2,593,936 — Difference 53,509 — Difference 2 .11%
      - Claims in Chinese renminbi: As Published 255,371 — New Method 255,323 — Difference – 47 — Difference –0.02%
      - Claims in Japanese yen: As Published 670,074 — New Method 742,4 4 072,365 — Difference 72,291 — Difference 10.80%
      - Claims in pounds sterling: As Published 580,230 — New Method 606,711 — Difference 26,481 — Difference 4.56%
      - Claims in Australian dollars: As Published 250,935 — New Method 263,808 — Difference 12,873 — Difference 5.13%
      - Claims in Canadian dollars: As Published 313, 828 — New Method 335,563 — Difference 21,734 — Difference 6.93%
      - Claims in Swiss francs: As Published 19, 5 4 42 — New Method 2,137 — Difference 2,593 — Difference 13. 27%
      - Claims in other currencies: As Published 621,708 — New Method 709, 289 — Difference 87, 5 811 — Difference 14.09%
      - Unallocated reserves: As Published 919, 310 — New Method 0 — Difference –919, 310 — Difference –100.00%
    - Share terms (percent and percentage points)
      - Total foreign exchange reserves: As Published 100.00% — New Method 100.00% — Difference 0.00 — Difference 0.00%
      - Shares of allocated reserves: As Published 92.90% — New Method 100.00% — Difference 7.10 — Difference 7. 6 4 %
      - Shares of US dollars: As Published 56.32% — New Method 57.12% — Difference 0.79 — Difference 1.41%
      - Shares of euros: As Published 21.13% — New Method 20 .12% — Difference –1.01 — Difference –4.77%
      - Shares of Chinese renminbi: As Published 2.12% — New Method 1.98% — Difference – 0 .14 — Difference –6.75%
      - Shares of Japanese yen: As Published 5.57% — New Method 5.76% — Difference 0.19 — Difference 3.34%
      - Shares of pounds sterling: As Published 4.83% — New Method 4.71% — Difference – 0 .12 — Difference –2.48%
      - Shares of Australian dollars: As Published 2.09% — New Method 2.05% — Difference –0.04 — Difference –1.95%
      - Shares of Canadian dollars: As Published 2. 61% — New Method 2.60% — Difference – 0.01 — Difference –0.28%
      - Shares of Swiss francs: As Published 0 .16% — New Method 0 .17% — Difference 0.01 — Difference 5.64%
      - Shares of other currencies: As Published 5 .17% — New Method 5.50% — Difference 0.33 — Difference 6.40%
      - Shares of unallocated reserves: As Published 7.10% — New Method 0.00% — Difference –7.10 — Difference –100.00%
- Changes to total reserves:
  - The new Total Foreign Exchange Reserves series is virtually unchanged compared with the published series.
  - Across the series, the average quarterly value of total foreign exchange reserves for the world is approximately 0.1 percent lower than previously published.
  - Previously published figures aligned COFER totals with the International Liquidity database by including differences in “unallocated”; under the new methodology IMF staff discontinue automatic linking and select the most reliable data source (COFER, International Liquidity, external official sources) for each country.
- Changes to allocated reserves:
  - Measured in US dollars, allocated reserves to each currency increase substantially because previously unallocated amounts have been distributed and added to allocated totals.
  - In the new series, “allocated” reserves equal “total” reserves since all reserves are allocated.
  - IMF staff will begin publishing the share of total reserves that has been imputed starting with COFER data for 2025:Q3.

### Changes over time and by currency
- Overview:
  - Previously unallocated amounts have been distributed across currencies using an imputation/stratified method; overall changes to aggregate shares are modest.
  - The revised series closely mirrors the previously published series because countries for which data were imputed represent a small share of the total over most of the period.
- Unallocated reserves:
  - Unallocated reserves, which grew steadily until 2018 and then began to decline, are now eliminated as part of the COFER dataset.
  - The unallocated amount falls to zero across the time series, as measured in US dollars and as a share of total foreign exchange reserves.
  - There will be no “unallocated” portion in any of the revised data because 100 percent of foreign exchange reserves will be allocated across currencies using the methods described.
  - IMF staff will publish the share of total reserves that have been imputed beginning with data from the third quarter of 2025.
- Historical aggregate statistics (as reported):
  - The share of countries with imputed data peaked at 46.8 percent of total foreign exchange reserves in the fourth quarter of 2013.
  - That share declined to as low as 6.13 percent by the second quarter of 2019.
- By currency (series coverage noted verbatim):
  - US dollars: New US dollar shares are slightly lower than previously published from 2000 to 2017 and slightly higher than previously published from 2018 onward. Figures cover the first quarter of 2000 to the second quarter of 2025.
  - Euros: New euro shares are slightly lower than previously published across most of the time series. Figures cover the first quarter of 2000 to the second quarter of 2025.
  - Chinese renminbi: New renminbi shares begin in 2016:Q4 and are largely unchanged, though shares in the most recent quarters are lower than previously published. Note: The IMF separately identified the Chinese renminbi in its official foreign exchange reserves database starting October 1, 2016. Figures cover 2016:Q4—2025:Q2.
  - Japanese yen: New yen shares are higher in the early years of the series and less changed in the more recent period. Figures cover the first quarter of 2000 to the second quarter of 2025.
  - Pounds sterling: New pounds sterling shares are higher than previously published in the early part of the series and lower than previously published more recently. Figures cover the first quarter of 2000 to the second quarter of 2025.
  - Australian dollar: Shown separately starting 2012:Q4; new shares are slightly higher than previously published from 2012 to 2022 and little changed thereafter. Figures cover 2012:Q4—2025:Q2.
  - Canadian dollar: Shown separately starting 2012:Q4; new shares are higher than previously published from 2012 to 2022 and little changed thereafter. Figures cover 2012:Q4—2025:Q2.
  - Swiss franc: New Swiss franc shares are slightly higher in the early years of the series and less changed in the recent period. Figures cover the first quarter of 2000 to the second quarter of 2025.
  - Other currencies: New “other currencies” shares are generally higher than previously published across the series. Figures cover the first quarter of 2000 to the second quarter of 2025 and include Canadian and Australian dollars until 2012:Q4 and Chinese renminbi until 2016:Q4.

### Analytical usefulness, confidentiality, and next steps
- Analytical improvements:
  - Imputing missing values across the time series makes underlying trends in the currency composition of foreign exchange reserves clearer and removes the need for users to assume unallocated reserves mirror the allocated average.
  - Eliminating the unallocated portion reduces distortions from mechanically smoothing reporting countries into and out of the dataset over time.
  - Imputing missing data in every period produces a more consistent and analytically robust time series of global reserve composition.
- Confidentiality:
  - The new approach mitigates disclosure risks when countries join or leave the reporting sample, thereby strengthening confidentiality.
  - COFER remains a strictly confidential data collection; IMF staff remain fully committed to safeguarding submissions.
  - The IMF continues not to publish breakdowns below the “world” total level to preserve confidentiality.
- Publication schedule and next steps:
  - The new COFER series—including the imputations—will be published in December 2025, alongside the release of COFER data for the third quarter of 2025.
  - The revised COFER series will include updates from the first quarter of 2000 through the second quarter of 2025 as well as new data for the third quarter of 2025.
  - IMF staff will publish the share of total reserves that have been imputed beginning with data from the third quarter of 2025.

*Source: Executive Summary from tnmea2025014 - Improving the Analytical Usefulness of the IMF’s COFER Data*

### Executive Summary 1

### Executive Summary 1

### Introduction
- COFER (Currency Composition of Foreign Exchange Reserves) is a quarterly IMF survey tracking the currency breakdown of the world’s foreign exchange reserves.
- At present, 149 entities report data, consisting of IMF member countries, nonmember countries or economies, and other foreign exchange reserve holding entities.
- Participation is voluntary except for Special Data Dissemination Standard Plus adherents, who are required to complete the survey and disclose their participation.
- COFER data are kept strictly confidential: the IMF does not publish country-level data or breakdowns below the global total, and the names of COFER reporters remain confidential unless the relevant authorities grant explicit consent.

### New Methodology: Imputing the Unallocated Share
- The new methodology imputes the current “unallocated” portion of COFER to provide a complete currency composition—expressed in both dollars and shares—that accounts for 100 percent of the world’s foreign exchange reserves.
- Previously, aggregate totals included an “unallocated” category to account for gaps from nonreporting, incomplete submissions, or discrepancies with the IMF’s International Liquidity database.
- Staff use a multipronged statistical approach, leveraging the underlying currency composition in COFER to estimate each country’s total foreign exchange reserves.
- As a result, the new COFER dataset will no longer include an “unallocated” component; all foreign exchange reserves will be allocated across currencies.

### Analytical Improvements and Confidentiality Benefits
- Imputing missing values across the time series:
  - Makes underlying trends in the currency composition of foreign exchange reserves clearer.
  - Eliminates the need for users to assume the composition of the “unallocated” portion (for example, presuming it reflects the average composition of the “allocated” portion).
  - Reduces distortions in aggregate trends caused by mechanical inclusion or exclusion of countries over time.
- The new approach mitigates disclosure risks when countries join or leave the reporting sample, thereby strengthening confidentiality.
- COFER remains a strictly confidential data collection; IMF staff remain fully committed to safeguarding submissions.

### Publication Timeline and Dataset Coverage
- The new COFER series—including the imputations—will be published in December 2025, alongside the release of COFER data for the third quarter of 2025.
- The revised COFER series will include updates from the first quarter of 2000 through the second quarter of 2025, as well as new data for the third quarter of 2025.
- Beginning with the release of the third quarter of 2025, IMF staff will also publish the share of total imputed reserves.

*Source: Executive Summary from tnmea2025014 - Improving the Analytical Usefulness of the IMF’s COFER Data*

### Introduction

### Introduction

### Overview
- The Currency Composition of Foreign Exchange Reserves (COFER) dataset tracks how much of the world’s foreign exchange reserves are held in different currencies and is widely used by central banks, other official institutions, and the private sector.
- Reserves in non-dollar currencies are converted into US dollars for COFER reporting, so exchange rate movements can change reported shares even when no buying or selling occurs.
- Historically COFER published totals broken into “allocated” and “unallocated” components; “allocated” reserves are broken down into eight named currencies plus “other currencies.”
- The IMF does not maintain a formal list of reserve currencies, but currencies separately identified in the COFER survey are widely regarded as reserve currencies.

### Problem Posed by an “Unallocated” Component
- Missing data (nonreporting, incomplete reporting, or discrepancies between COFER and the International Liquidity database) were previously included in COFER as “unallocated.”
- The International Liquidity database includes nearly 100 percent of the world’s foreign exchange reserves.
- An “unallocated” share can obscure or exaggerate trends if users assume it mirrors the average composition of the “allocated” portion but the reserves of nonreporting countries differ significantly from that average.
- Countries move in and out of “allocated” and “unallocated” status (starting/stopping reporting, incomplete reports, or reporting differences with the International Liquidity database), complicating time-series interpretation.
- Past statistical smoothing to protect confidentiality when countries entered/exited reporting could obscure underlying allocation patterns or exaggerate trends, especially when large or unusually skewed reserve holders were smoothed.

### Policy Change
- The COFER dataset will no longer include an “unallocated” component.
- IMF staff will impute the “unallocated” portion so published COFER currency compositions (in dollars and shares) cover 100 percent of world foreign exchange reserves.
- The change aims to simplify interpretation of published currency shares and limit residual disclosure risks when countries change reporting status.

---

### Methodology: Imputing the Unallocated Share of COFER

- The unallocated share declined substantially over time; in 2025:Q2 the “unallocated” portion was $919.3 billion (that is, 7.1 percent of the $12.9 trillion in foreign exchange reserves held globally in that quarter).
- IMF staff imputed the currency composition for the unallocated portion and extended the imputation back to 2000:Q1 (with caveats for earlier periods where larger imputations were necessary).

Major imputation approaches used by IMF staff:
- Stratified mean imputation
- Carry-forward imputation
- Mixed method (combination of carry-forward and stratified mean)

Steps in the imputation workflow:
1. Estimate a country’s total foreign exchange reserves.
   - For COFER reporters, IMF staff will use the information reported to COFER as the total reserves for reporting countries.
   - For countries not reporting to COFER, IMF staff will use the International Liquidity database to estimate total reserves.
   - For countries reporting to neither, IMF staff will use an external official source (e.g., central bank website); if unavailable, IMF staff will carry forward the best available last-known estimate.
   - The previous practice of including the difference between International Liquidity and COFER in “unallocated” has been discontinued.

2. Place reporting countries into stratified subgroups based on their reported currency composition.
   - First definitional split: issuers of currencies in the reserve currency basket (“Reservia”) versus non-Reservia.
   - Reservia consists of 27 countries that, by definition, do not hold their own currency as foreign reserves; these countries account for approximately 50 percent of global foreign exchange reserve holdings.
   - Non-Reservia consists of 164 countries/entities accounting for the other 50 percent.

3. For non-Reservia reporting countries, an algorithm stratifies them into six subgroups based on patterns relative to the reported global average currency allocation:
   - Subgroup 1: Skewed toward the euro; away from both the US dollar and a combination of all other currencies.
   - Subgroup 2: Skewed toward the US dollar; away from both the euro and a combination of all other currencies.
   - Subgroup 3: Skewed toward the US dollar and a combination of all other currencies; away from the euro.
   - Subgroup 4: Skewed toward a combination of all other currencies; away from both the US dollar and euro.
   - Subgroup 5: Skewed toward the euro and a combination of all other currencies; away from the US dollar.
   - Subgroup 6: Skewed toward both the US dollar and the euro; away from a combination of all other currencies.
   - Example: If global averages are 56 percent US dollars, 21 percent euros, and 23 percent other currencies, a reporting country with euro allocation >21 percent, US dollar allocation <56 percent, and allocation to other currencies <23 percent would be placed in subgroup 1.

4. Compute unweighted average (mean) currency allocation of each subgroup each quarter and apply that mean to nonreporting countries’ total reserves to impute values.

- Countries that have never reported to COFER are placed into stratified subgroups by IMF staff using public data on trade and financial linkages, geopolitical relationships, qualitative factors, and judgement.

- The largest part of the unallocated total is attributable to countries that have never reported to COFER; their entire reserves were previously “unallocated” and will now be allocated using stratified mean imputation.

Confidentiality safeguards and special handling:
- IMF staff divide countries into large reserve holders and small reserve holders:
  - Small holders: those with foreign exchange reserve holdings less than 2 percent of global foreign exchange reserves.
  - For small holders, changes are unlikely to materially affect aggregates; either carry-forward imputation or mixed method will be used as needed.
- Carry-forward imputation:
  - For previous reporters who stop reporting, the most recent reported shares are used to impute subsequent missing values.
  - For nonreporting countries who were previous COFER reporters, their currency composition is imputed based on their most recently reported shares for the first four periods after they stop reporting.
- Small reserve holders:
  - First four quarters after stopping reporting: allocated based on most recently reported shares (carry-forward imputation).
  - Fifth quarter of nonreporting: allocated based on subgroup average (stratified mean imputation). Subgroup mean excludes the nonreporter’s most recently reported shares.
- Large reserve holders:
  - First four quarters after stopping reporting: carry-forward imputation.
  - Fifth quarter of nonreporting: stratified mean imputation is used with the nonreporting country’s most recently reported shares included in the group average computation.
- Large reserve holders with an unusual previously reported skew:
  - IMF staff use a mixed method (weighted average of group allocation and the large holder’s previous allocation) over a period longer than four quarters to smooth exit and reduce disclosure risk.
  - Example mixed-method weighting: group allocation (20 percent) and the allocation of the large holder in the previous period (80 percent), with weights applied iteratively so the imputed value converges to the subgroup average over time.

- IMF staff will continuously incorporate external quantitative and qualitative data (e.g., national data publications) to improve imputed values for nonreporters.

---

### Results (Summary and Key Statistics)

- IMF staff imputed currency allocations for every country for every quarter from 2000:Q1 to 2025:Q2.
- Effects of imputation:
  - Total reserves decrease slightly on average; allocated reserves now equal total reserves; unallocated reserves are eliminated.
  - Claims in each currency increase in dollar terms as previously unallocated amounts are distributed across currencies.
  - Changes in share terms are modest and vary by currency and over time.

Table 1. Currency Composition of Foreign Exchange Reserves, 2025:Q2 (as published vs new method)
- Monetary levels (millions)
  - Total foreign exchange reserves: As Published 12,94 4,765 — New Method 12,893,887 — Difference –50,878 — Difference –0.39%
  - Allocated reserves: As Published 12,025,455 — New Method 12,893,887 — Difference 868,432 — Difference 7. 2 2 %
  - Claims in US dollars: As Published 6,773,338 — New Method 7, 3 6 4 , 6 815 — Difference 591,343 — Difference 8.73%
  - Claims in euros: As Published 2,540,427 — New Method 2,593,936 — Difference 53,509 — Difference 2 .11%
  - Claims in Chinese renminbi: As Published 255,371 — New Method 255,323 — Difference – 47 — Difference –0.02%
  - Claims in Japanese yen: As Published 670,074 — New Method 742,4 4 072,365 — Difference 72,291 — Difference 10.80%
  - Claims in pounds sterling: As Published 580,230 — New Method 606,711 — Difference 26,481 — Difference 4.56%
  - Claims in Australian dollars: As Published 250,935 — New Method 263,808 — Difference 12,873 — Difference 5.13%
  - Claims in Canadian dollars: As Published 313, 828 — New Method 335,563 — Difference 21,734 — Difference 6.93%
  - Claims in Swiss francs: As Published 19, 5 4 42 — New Method 2,137 — Difference 2,593 — Difference 13. 27%
  - Claims in other currencies: As Published 621,708 — New Method 709, 289 — Difference 87, 5 811 — Difference 14.09%
  - Unallocated reserves: As Published 919, 310 — New Method 0 — Difference –919, 310 — Difference –100.00%

- Share terms (percent and percentage points)
  - Total foreign exchange reserves: As Published 100.00% — New Method 100.00% — Difference 0.00 — Difference 0.00%
  - Shares of allocated reserves: As Published 92.90% — New Method 100.00% — Difference 7.10 — Difference 7. 6 4 %
  - Shares of US dollars: As Published 56.32% — New Method 57.12% — Difference 0.79 — Difference 1.41%
  - Shares of euros: As Published 21.13% — New Method 20 .12% — Difference –1.01 — Difference –4.77%
  - Shares of Chinese renminbi: As Published 2.12% — New Method 1.98% — Difference – 0 .14 — Difference –6.75%
  - Shares of Japanese yen: As Published 5.57% — New Method 5.76% — Difference 0.19 — Difference 3.34%
  - Shares of pounds sterling: As Published 4.83% — New Method 4.71% — Difference – 0 .12 — Difference –2.48%
  - Shares of Australian dollars: As Published 2.09% — New Method 2.05% — Difference –0.04 — Difference –1.95%
  - Shares of Canadian dollars: As Published 2. 61% — New Method 2.60% — Difference – 0.01 — Difference –0.28%
  - Shares of Swiss francs: As Published 0 .16% — New Method 0 .17% — Difference 0.01 — Difference 5.64%
  - Shares of other currencies: As Published 5 .17% — New Method 5.50% — Difference 0.33 — Difference 6.40%
  - Shares of unallocated reserves: As Published 7.10% — New Method 0.00% — Difference –7.10 — Difference –100.00%

- Changes to Total Reserves
  - The new Total Foreign Exchange Reserves series is virtually unchanged compared with the published series.
  - Across the series, the average quarterly value of total foreign exchange reserves for the world is approximately 0.1 percent lower than previously published.
  - Previously published figures aligned COFER totals with the International Liquidity database by including differences in “unallocated”; under the new methodology IMF staff discontinue automatic linking and select the most reliable data source (COFER, International Liquidity, external official sources) for each country.

- Changes to Allocated Reserves
  - Measured in US dollars, allocated reserves to each currency increase substantially because previously unallocated amounts have been distributed and added to allocated totals.
  - In the new series, “allocated” reserves equal “total” reserves since all reserves are allocated.
  - IMF staff will begin publishing the share of total reserves that has been imputed starting with COFER data for 2025:Q3.

*Source: COFER; and IMF staff calculations.*

### 1. Trillions of US Dollars2. Shares of Total Foreign Exchange Reserves

### 1. Trillions of US Dollars2. Shares of Total Foreign Exchange Reserves

### Overview of the revision
- Previously unallocated amounts in the COFER dataset have been distributed across currencies using an imputation/stratified method; overall changes to aggregate shares are modest.
- The revised series closely mirrors the previously published series because countries for which data were imputed represent a small share of the total over most of the period.
- The shares shown are aggregate allocations of countries in the COFER dataset; aggregate results are essential for identifying broad trends but can obscure heterogeneity of individual country allocations.

### Changes to unallocated reserves
- Unallocated reserves, which grew steadily until 2018 and then began to decline, are now eliminated as part of the COFER dataset.
- The unallocated amount falls to zero across the time series, as measured in US dollars and as a share of total foreign exchange reserves.
- There will be no “unallocated” portion in any of the revised data because 100 percent of foreign exchange reserves will be allocated across currencies using the methods described in this technical note.
- IMF staff will publish the share of total reserves that have been imputed beginning with data from the third quarter of 2025.

### Key historical aggregate statistics (as reported)
- The share of countries with imputed data peaked at 46.8 percent of total foreign exchange reserves in the fourth quarter of 2013.
- That share declined to as low as 6.13 percent by the second quarter of 2019.

### Changes, by currency
- General note: Changes reflect allocation of previously unallocated reserves for countries whose imputed shares differ from the global average; the group of imputed countries is dynamic over time.

- US dollars
  - New US dollar shares are slightly lower than previously published from 2000 to 2017 and slightly higher than previously published from 2018 onward.
  - Figures cover the first quarter of 2000 to the second quarter of 2025.

- Euros
  - New euro shares are slightly lower than previously published across most of the time series.
  - Figures cover the first quarter of 2000 to the second quarter of 2025.

- Chinese renminbi
  - The new renminbi shares begin in 2016:Q4 and are largely unchanged, though shares in the most recent quarters are lower than previously published.
  - This reflects allocation for countries whose imputed shares of renminbi holdings are lower than the global average.
  - Note: The IMF separately identified the Chinese renminbi in its official foreign exchange reserves database starting October 1, 2016.
  - Figures cover 2016:Q4—2025:Q2.

- Japanese yen
  - New yen shares are higher in the early years of the series and less changed in the more recent period.
  - Figures cover the first quarter of 2000 to the second quarter of 2025.

- Pounds sterling
  - New pounds sterling shares are higher than previously published in the early part of the series and lower than previously published more recently.
  - Pattern similar to the yen (higher in early years, less changed recently) but opposite to the US dollar pattern.
  - Figures cover the first quarter of 2000 to the second quarter of 2025.

- Australian dollar
  - Shown separately starting 2012:Q4; new shares are slightly higher than previously published from 2012 to 2022 and little changed thereafter.
  - Figures cover 2012:Q4—2025:Q2.
  - Note: On June 28, 2013, the IMF published the quarterly COFER data with an expanded currency range separately identifying the Australian dollar and the Canadian dollar.

- Canadian dollar
  - Shown separately starting 2012:Q4; new shares are higher than previously published from 2012 to 2022 and little changed thereafter.
  - Figures cover 2012:Q4—2025:Q2.

- Swiss franc
  - New Swiss franc shares are slightly higher in the early years of the series and less changed in the recent period; pattern similar to pounds sterling and yen.
  - Figures cover the first quarter of 2000 to the second quarter of 2025.

- Other currencies
  - New “other currencies” shares are generally higher than previously published across the series.
  - This reflects allocation for countries whose imputed shares of other currency holdings are higher than the global average.
  - Figures cover the first quarter of 2000 to the second quarter of 2025 and include Canadian and Australian dollars until 2012:Q4 and Chinese renminbi until 2016:Q4.

### Improving analytical usefulness and strengthening confidentiality
- The new imputation method improves the analytical usefulness of COFER by making underlying trends in the currency composition of foreign exchange reserves clearer and removing the need for users to assume unallocated reserves mirror the average composition of allocated reserves.
- Eliminating the unallocated portion reduces distortions from mechanically smoothing reporting countries into and out of the dataset over time.
- Imputing missing data in every period produces a more consistent and analytically robust time series of global reserve composition.
- Confidentiality: Imputing the unallocated portion helps mitigate residual disclosure risks when countries enter or exit the reporting sample; COFER remains a strictly confidential dataset and IMF staff remain committed to safeguarding confidentiality.
- The IMF continues not to publish breakdowns below the “world” total level to preserve confidentiality.

### Next steps and publication schedule
- The new COFER series—including the imputations—will be published in December 2025, alongside the release of COFER data for the third quarter of 2025.
- The revised COFER series will include updates from the first quarter of 2000 through the second quarter of 2025 as well as new data for the third quarter of 2025.
- IMF staff will publish the share of total reserves that have been imputed beginning with data from the third quarter of 2025.

### Annex I — Technical background and dataset history (highlights)
- COFER is conducted quarterly by the IMF’s Statistics Department and disseminated at the end of each quarter for a reference date of the end of the previous quarter.
- Participation is voluntary; Special Data Dissemination Standard Plus adherents are required to participate.
- At present, 149 reporters share data with the IMF to be aggregated in the COFER dataset.
- COFER data are strictly confidential; the names of reporters are confidential unless authorities consent to disclosure; the IMF does not publish individual country data or breakdowns below the world total level.
- Concept of reserves aligns with BPM6 (unchanged in BPM7): external assets readily available to and controlled by monetary authorities for balance of payments financing needs, exchange market intervention, and related purposes (BPM6, paragraph 6.64).
- COFER covers claims such as foreign banknotes, bank deposits, treasury bills, short- and long-term government securities, and other claims usable in balance of payments needs; excludes monetary gold, special drawing rights holdings, and reserve position in the IMF.
- Data are reported in millions of US dollars.
- Historical timeline highlights:
  - Annual data from 1995 to 1998; quarterly data from 1999 onward.
  - Australian dollar and Canadian dollar separately identified starting in data published June 28, 2013 (expanded currency range).
  - IMF published the list of COFER reporters on September 30, 2015.
  - IMF stopped publishing an “advanced economies” / “emerging and developing economies” breakdown starting with 2015:Q2 to avoid residual disclosure.
  - Chinese renminbi separately identified effective October 1, 2016; change reflected in the survey for 2016:Q4.

*Source: COFER; and IMF staff calculations.*

### ANNEX II. Analytical Usefulness

### ANNEX II. Analytical Usefulness of the COFER Dataset

### Role and importance of reserves and COFER
- Central banks’ foreign exchange reserves provide insurance against external shocks and facilitate monetary and exchange rate policy.
- The currency composition of reserves—US dollars, euros, yen, sterling, renminbi, and a growing set of “other” currencies—matters for global financial stability and for the international roles of currencies.
- The IMF’s Currency Composition of Official Foreign Exchange Reserves (COFER) survey is the standard dataset for tracking these patterns.

### Determinants of currency composition (findings from literature)
- Classic and modern studies converge on determinants for a country’s currency composition of foreign exchange reserves:
  - Transaction and invoicing links.
  - Currency pegs.
  - Financial integration.
- Safety, liquidity, and market depth are important because reserve managers require markets that can absorb large flows with minimal price impact and minimal legal and operational risk.
- Network effects are a powerful determinant of persistent currency dominance, although they do not preclude meaningful diversification across several leading currencies.
- Active diversification and valuation effects play a key role in the currency composition of reserves.

### Empirical applications and key results using COFER
- COFER has facilitated empirical inquiries around reserve diversification, currency dominance, and market intervention strategies.
- Landmark and follow-up studies using COFER include:
  - Lim (2007): investigated whether reserve portfolios respond to exchange rate fluctuations through rebalancing behaviors.
  - Wong (2007): examined international reserve diversification using quantity shares derived from COFER data.
  - Studies on the euro’s challenge to the dollar and historical reserve currency compositions (Eichengreen and Mathieson 2000; Lim 2006).
  - Arslanalp and Simpson-Bell (2021): investigated the decline of the dollar share of currency reserves over a 25-year period while maintaining its dominance.
  - Goldberg and Hannaoui (2024): argued that changes in dynamic portfolio allocations by a small handful of countries—rather than a systematic retreat from the dollar in official reserve portfolios—explain the declining dollar share in the overall COFER.
  - Arslanalp, Eichengreen, and Simpson-Bell (2022, 2024): examined the role active diversification in relation to valuation effects plays in the currency composition of foreign exchange reserves.

### Limitation of COFER and alternative data efforts
- The primary limitation of COFER is confidentiality: individual country data remains restricted, accessible to only four IMF staff in charge of compiling the COFER aggregates.
- Researchers have developed alternative datasets to enable disaggregated analysis:
  - Ito and McCauley (2021) developed a new disaggregated dataset of individual central bank reserve compositions for a subset of countries (covering 63–73 countries depending on dataset), allowing more nuanced, country-level analysis.

*Source: ANNEX II. Analytical Usefulness of the COFER Dataset, TNM/2025/14*

---


_Source: https://www.imf.org/-/media/files/publications/tnm/2025/english/tnmea2025014.pdf_