## 1. Priority Recommendations

## Source details

**Canonical URL:** [1. Priority Recommendations](https://www.imf.org/-/media/files/publications/cr/2022/english/1vnmea2022002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2022/english/1vnmea2022002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2022/english/1vnmea2022002.pdf.json)

---

### Mission purpose and context
- A technical assistance (TA) mission was conducted remotely with the General Statistics Office (GSO) during November 30–December 4, 2020 to prepare the next rebase of national accounts with focus on: better incorporation of non-observed activities; development of supply and use tables (SUTs) and input-output tables (IOTs) for the new benchmark year. The mission was funded by the IMF’s Data for Decisions project.
- The GSO plans to compile a new national benchmark for 2020. The existing base year 2010 needs updating to incorporate recent structural changes.
- Vietnam’s economy recorded growth of 2.91 percent in volumes in 2020 compared to 2019; however, COVID-19 impacts produce unusual patterns in 2020 national accounts (notably tourism-related activities).

### Key findings
- Using 2020 as the benchmark year is not ideal because 2020 exhibits unusual characteristics due to the pandemic, affecting representativeness and intra-annual information (further analysis in Appendix 1).
- Despite atypical 2020 conditions, postponing the benchmark is undesirable because long-term data collection programs are already initiated or ongoing (e.g., economic census planned March–August 2021).
- Adoption of chain-linked volume estimates will reduce risk of distortions relative to fixed-base year measures by updating annual economic structures; expected rebase completion: 2023.
- The shift to chain-linking entails significant compilation challenges, requires training and software, and will represent a major change for users (including loss of additivity).
- Comprehensive inclusion of non-observed activities is essential: exclusion underestimates economy size, biases GDP growth rates, and hampers international comparisons.
- SUTs and IOTs are central to improving the benchmark, identifying data gaps, and incorporating non-observed activities; their compilation requires training, balancing software, and organizational arrangements.
- Data and metadata accessibility in English is limited; quarterly GDP estimates are published in Vietnamese only; more detailed metadata and timely indicators are needed.
- A monthly indicator of economic growth requires further TA; IMF follow-up options to be explored.

### Priority recommendations (Table 1)
- March 2021: List and prioritize non-observed activities and identify possible indicators. — Responsible: GSO
- August 2021: Develop methodologies to improve the coverage of the non-observed economy. — Responsible: GSO
- June 2022: Compile SUTs and IOTs for 2020. — Responsible: GSO

### Detailed action plan and milestones (Table 2)
- Outcome: Data are compiled and disseminated using the concepts and definitions of the latest manual/guide.
- H (High priority) — List and prioritize the non-observed activities and identify possible indicators. Target Completion Date: March 2021.
- H (High priority) — Develop methodologies to improve the coverage of the non-observed economy. Target Completion Date: August 2021.
- M (Medium priority) — Finalize the data collection for the new benchmark year 2020. Target Completion Date: December 2021.
- M — Populate the SUTs for 2020 incorporating non-observed activities and new source data including from administrative sources, the economic census, the agriculture, forestry and fishery survey, the household living standard survey, and the capital investment survey. Target Completion Date: March 2022.
- M — Develop methodologies and procedures to compile chain-linked volume estimates for annual and quarterly GDP estimates. Target Completion Date: March 2022.
- M — Balance the SUTs for 2020. Target Completion Date: May 2022.
- H — Finalize the SUTs and IOTs for 2020. Target Completion Date: June 2022.
- M — Develop a strategy and methodologies to backcast national accounts estimates aligned with the new benchmark. Target Completion Date: August 2022.
- M — Organize outreach events to users to communicate on the rebase and new methodologies. Target Completion Date: September 2022.
- M — Backcast national accounts estimates. Target Completion Date: November 2022.
- M — Start pilot-testing the compilation and preparation of quarterly national accounts using chain-linking. Target Completion Date: December 2022.
- M — Publish rebased national accounts and backcasted series. Target Completion Date: March 2023.

### Methodological guidance and implications
- Chain-linking advantages and challenges:
  - Improves quality by constantly updating weights in volume estimates; involves compiling estimates in previous years’ prices and chaining them.
  - For some countries differences between fixed-base and chain-linked GDP growth can be limited; nevertheless, given pandemic impacts on 2020, chaining reduces distortion risks.
  - Chain-linking increases compilation complexity, especially for quarterly estimates when benchmarking to annual estimates (quarterly or annual overlap approaches).
  - Software should be used to limit compilation errors.
  - Loss of additivity from chaining can confuse users; outreach is required. Options include re-referencing estimates every year or every two years to maintain additivity for recent years.
  - When compiling SUTs in volumes annually, previous years’ prices should be used for volume measures.
- Further compliance with the System of National Accounts 2008 should be envisaged, but priority should be given to improving coverage of non-observed activities and compiling SUTs/IOTs.

### Non-observed activities: strategy and compilation
- Non-observed activities include activities currently excluded due to statistical deficiencies, household production for own-final consumption, underground activities, informal activities, and illegal activities; distinctions are not always clear.
- Recommended actions:
  - Conduct a systematic review of all non-observed activities to select appropriate data sources and indicators.
  - Pay special attention to activities enabled by digital platforms (examples cited: Uber-equivalent taxi services; Airbnb-equivalent accommodation services).
  - Use results from the economic census (planned March–August 2021) and household living standard survey to improve coverage.
  - IMF can provide further TA as needed.

### Developing SUTs and IOTs for 2020
- SUTs at detailed product level and balancing will identify data gaps, inconsistencies, and incoherence and will strengthen GDP measure coherence.
- Requirements and recommendations:
  - Training for GSO compilers on SUT/IOT concepts and compilation strategies (training provided during the mission; more may be needed).
  - Use software for balancing; manual balancing is necessary for large imbalances, but software (e.g., proportional iterative fitting) is required to handle numerous small adjustments and ensure accuracy.
  - Implement clear organizational structures for workflow, version control, documentation, and coordination across expert teams.
  - IMF can assist with implementation of balancing procedures and automation to derive IOTs from SUTs.

### Dissemination and metadata
- Current dissemination issues:
  - Quarterly GDP estimates published in Vietnamese only.
  - GSO publishes infra-annual estimates in English for the first nine months of the year; quarterly English publication would facilitate international access.
  - Metadata in English are limited to outlines; detailed metadata on data sources and methods are lacking.
  - Online extraction tools provide aggregated estimates (example: only the total for Agriculture, forestry and fishing available for 2018).
- Recommendations:
  - Improve data and metadata access in English, including more detailed methodological documentation and data granularity.
  - IMF can provide advice on dissemination best practices and on broader macroeconomic statistics dissemination in the context of dissemination standards assessments.

### Monthly indicator of economic growth
- Further technical assistance is needed to develop a monthly indicator of economic growth; the IMF will explore options following TA provided in 2019. Development will also support improvement of quarterly indicators.

### Appendix I — Is 2020 a Good Choice for a New Benchmark Year for GDP?
- Recommendation on benchmark timing:
  - The IMF recommends that countries produce benchmark estimates of GDP every five to ten years (with five years the preferred interval).
  - For some countries, their next scheduled benchmark year is 2020. It is recommended that countries choose an alternative year to produce GDP benchmarks.
  - Countries that planned to produce benchmark estimates of GDP for the year 2021 are advised to consider adjusting their plans and likewise choose an alternate benchmark year if local conditions warrant.
- Reason 1 — representativeness of the benchmark year:
  - A benchmark year should be representative and reflect “normal economic activity” for a given country.
  - Developing benchmark estimates in a year when there is an acute or ongoing economic shock is unadvisable because the shock may cause significant temporary shifts in production and consumption patterns.
  - Benchmark estimates are used as weights to aggregate detailed indexes to more aggregated indexes; an unrepresentative benchmark can distort these aggregations.
  - Benchmark estimates also serve as interpolation points for time series; if one benchmark is an outlier and poor interpolation techniques are used, compilers could introduce a smoothing effect that:
    - does not reflect economic reality;
    - causes distortions in measurement of the business cycle;
    - tends to mute the rapidity of declines and the subsequent recovery.
  - Good interpolation techniques, however, maintain turning points and reflect historical trends regardless of outlier observations.
- Reason 2 — data quality and collection challenges in 2020:
  - The COVID-19 pandemic has had a major impact on data collection activities; source data used by compilers may be of inferior quality.
  - Specific data-collection problems observed:
    - NSOs were sometimes unable to collect data because respondents were shut down or working from home and difficult to contact.
    - Provision of data to NSOs took a back seat to more pressing priorities.
    - The type of transactions during this period are unusual and more susceptible to misrecording.
  - Using these “inferior” data sources to establish benchmarks undermines the goal of benchmarking, since governments invest heavily in data operations in a benchmark year.
  - If there are widespread challenges in collecting or processing data during the benchmark year, the investment would be better spent in another period.
- Reason 3 — implications for quarterly series and intra-annual information:
  - Data collection activities for 2020 are more severely affected for intra-annual information, so even if annual information is adequate, estimation of quarterly “benchmark” series may be flawed.
  - Major disadvantages of poor quarterly data in the benchmark year include:
    - These poor benchmark quarterly series will not be revised until the next benchmark year.
    - The first annual GDP estimate is generated from an extrapolation of quarterly series, so there may be a bias in the first signal and thus large revisions when the first annual compilation is produced.
    - Severe distortions in the quarterly series will hamper backwards linking, seasonal adjustment, and analyses.
- Specific case: GSO plans and chain-linking:
  - Plans by the GSO to develop a new benchmark for 2020 are now too advanced to select another benchmark year.
  - The implementation of chain-linking will correct possible biases.
  - A detailed analysis of data collected for 2020 relating to activities affected by the pandemic and future related source data is recommended.

### Appendix II — compilation of chain volume measures (numerical example)
- Example scope: two items (Bananas, Pineapple) and three periods (PERIOD 0, PERIOD 1, PERIOD 2).
- Table values and derived totals (preserved exactly):
  - Bananas:
    - P0 Q0 V0: 1 5 5
    - P1 Q1 V1: 2 8 16
    - P2 Q2 V2 (Annually rebased): 8 3 13 39 26
  - Pineapple:
    - P0 Q0 V0: 4 3 12
    - P1 Q1 V1: 4 5 20
    - P2 Q2 V2 (Annually rebased): 20 5 10 50 40
  - Totals (current): 17 (PERIOD 0), 36 (PERIOD 1), 89 (PERIOD 2)
  - Totals (annually rebased): 28 (PERIOD 1), 66 (PERIOD 2)
  - Index: 100 (PERIOD 0), 164.7 (PERIOD 1), 183.3 (PERIOD 2)
  - Chain volume index: 100 (PERIOD 0), 164.7 (PERIOD 1), 302.0 (PERIOD 2)
  - Chain volume estimate: 17 (PERIOD 0), 28.0 (PERIOD 1), 51.3 (PERIOD 2)
  - Constant prices: 17 (PERIOD 0), 28 (PERIOD 1), 53 (PERIOD 2)
  - Constant prices growth: 64.7% (PERIOD 1 vs PERIOD 0), 89.3% (PERIOD 2 vs PERIOD 1)
  - Chain volume growth: 83.3% (PERIOD 2 vs PERIOD 1)
- Explanatory calculations and notes (preserved):
  - Annually rebased estimates are values in previous years’ prices. Example: in period 1, the value of bananas is P0 x Q1 = 1 x 8 = 8. The current price estimate is P1 x Q1 = 2 x 8 = 16.
  - Annually rebased estimates across different periods are not priced using the same pricing period and are not time series. Time series are obtained by chaining the indices.
  - Volume indices are obtained as the ratio of the annually rebased value by the current value of the previous period multiplied by 100:
    - Period 1: 28 / 17 x 100 = 164.7
    - Period 2: 66 / 36 x 100 = 183.3
  - Chaining volume indices are obtained by multiplying indices. For period 2 the chain volume index, referenced to period 0, is 164.7 x 183.3 / 100 = 302.
  - The chain volume estimate is obtained by multiplying the chain volume index by the current value of the first period: 17 x 302 / 100 = 51.3.
- Additivity and re-referencing:
  - Chain volume estimates are no longer additive after two periods.
  - Chain volume estimate referenced to period 0 for bananas is 13 for period 2 and for pineapples this value is 40. The sum of these two items is 53, which equals the constant price estimate using period 0 as the pricing period and is higher than the chain volume estimate of both commodities (51.3).
  - One option to maintain additivity in recent periods is to re-reference every year or every two years; however, re-referencing adds complexities to the compilation process.

*Source: IMF technical assistance mission report, “1. Priority Recommendations” (content unit: 1vnmea2022002).*

### 1. Priority Recommendations _____________________________________________________________________________ 5

### 1. Priority Recommendations

### Mission purpose and context
- A technical assistance (TA) mission was conducted remotely with the General Statistics Office (GSO) during November 30–December 4, 2020 to prepare the next rebase of national accounts with focus on: better incorporation of non-observed activities; development of supply and use tables (SUTs) and input-output tables (IOTs) for the new benchmark year. The mission was funded by the IMF’s Data for Decisions project.
- The GSO plans to compile a new national benchmark for 2020. The existing base year 2010 needs updating to incorporate recent structural changes.
- Vietnam’s economy recorded growth of 2.91 percent in volumes in 2020 compared to 2019; however, COVID-19 impacts produce unusual patterns in 2020 national accounts (notably tourism-related activities).

### Key findings
- Using 2020 as the benchmark year is not ideal because 2020 exhibits unusual characteristics due to the pandemic, affecting representativeness and intra-annual information (further analysis in Appendix 1).
- Despite atypical 2020 conditions, postponing the benchmark is undesirable because long-term data collection programs are already initiated or ongoing (e.g., economic census planned March–August 2021).
- Adoption of chain-linked volume estimates will reduce risk of distortions relative to fixed-base year measures by updating annual economic structures; expected rebase completion: 2023.
- The shift to chain-linking entails significant compilation challenges, requires training and software, and will represent a major change for users (including loss of additivity).
- Comprehensive inclusion of non-observed activities is essential: exclusion underestimates economy size, biases GDP growth rates, and hampers international comparisons.
- SUTs and IOTs are central to improving the benchmark, identifying data gaps, and incorporating non-observed activities; their compilation requires training, balancing software, and organizational arrangements.
- Data and metadata accessibility in English is limited; quarterly GDP estimates are published in Vietnamese only; more detailed metadata and timely indicators are needed.
- A monthly indicator of economic growth requires further TA; IMF follow-up options to be explored.

### Priority recommendations (Table 1)
- March 2021: List and prioritize non-observed activities and identify possible indicators. — Responsible: GSO
- August 2021: Develop methodologies to improve the coverage of the non-observed economy. — Responsible: GSO
- June 2022: Compile SUTs and IOTs for 2020. — Responsible: GSO

### Detailed action plan and milestones (Table 2)
- Outcome: Data are compiled and disseminated using the concepts and definitions of the latest manual/guide.
- H (High priority) — List and prioritize the non-observed activities and identify possible indicators. Target Completion Date: March 2021.
- H (High priority) — Develop methodologies to improve the coverage of the non-observed economy. Target Completion Date: August 2021.
- M (Medium priority) — Finalize the data collection for the new benchmark year 2020. Target Completion Date: December 2021.
- M — Populate the SUTs for 2020 incorporating non-observed activities and new source data including from administrative sources, the economic census, the agriculture, forestry and fishery survey, the household living standard survey, and the capital investment survey. Target Completion Date: March 2022.
- M — Develop methodologies and procedures to compile chain-linked volume estimates for annual and quarterly GDP estimates. Target Completion Date: March 2022.
- M — Balance the SUTs for 2020. Target Completion Date: May 2022.
- H — Finalize the SUTs and IOTs for 2020. Target Completion Date: June 2022.
- M — Develop a strategy and methodologies to backcast national accounts estimates aligned with the new benchmark. Target Completion Date: August 2022.
- M — Organize outreach events to users to communicate on the rebase and new methodologies. Target Completion Date: September 2022.
- M — Backcast national accounts estimates. Target Completion Date: November 2022.
- M — Start pilot-testing the compilation and preparation of quarterly national accounts using chain-linking. Target Completion Date: December 2022.
- M — Publish rebased national accounts and backcasted series. Target Completion Date: March 2023.

### Methodological guidance and implications
- Chain-linking advantages and challenges:
  - Improves quality by constantly updating weights in volume estimates; involves compiling estimates in previous years’ prices and chaining them.
  - For some countries differences between fixed-base and chain-linked GDP growth can be limited; nevertheless, given pandemic impacts on 2020, chaining reduces distortion risks.
  - Chain-linking increases compilation complexity, especially for quarterly estimates when benchmarking to annual estimates (quarterly or annual overlap approaches).
  - Software should be used to limit compilation errors.
  - Loss of additivity from chaining can confuse users; outreach is required. Options include re-referencing estimates every year or every two years to maintain additivity for recent years.
  - When compiling SUTs in volumes annually, previous years’ prices should be used for volume measures.
- Further compliance with the System of National Accounts 2008 should be envisaged, but priority should be given to improving coverage of non-observed activities and compiling SUTs/IOTs.

### Non-observed activities: strategy and compilation
- Non-observed activities include activities currently excluded due to statistical deficiencies, household production for own-final consumption, underground activities, informal activities, and illegal activities; distinctions are not always clear.
- Recommended actions:
  - Conduct a systematic review of all non-observed activities to select appropriate data sources and indicators.
  - Pay special attention to activities enabled by digital platforms (examples cited: Uber-equivalent taxi services; Airbnb-equivalent accommodation services).
  - Use results from the economic census (planned March–August 2021) and household living standard survey to improve coverage.
  - IMF can provide further TA as needed.

### Developing SUTs and IOTs for 2020
- SUTs at detailed product level and balancing will identify data gaps, inconsistencies, and incoherence and will strengthen GDP measure coherence.
- Requirements and recommendations:
  - Training for GSO compilers on SUT/IOT concepts and compilation strategies (training provided during the mission; more may be needed).
  - Use software for balancing; manual balancing is necessary for large imbalances, but software (e.g., proportional iterative fitting) is required to handle numerous small adjustments and ensure accuracy.
  - Implement clear organizational structures for workflow, version control, documentation, and coordination across expert teams.
  - IMF can assist with implementation of balancing procedures and automation to derive IOTs from SUTs.

### Dissemination and metadata
- Current dissemination issues:
  - Quarterly GDP estimates published in Vietnamese only.
  - GSO publishes infra-annual estimates in English for the first nine months of the year; quarterly English publication would facilitate international access.
  - Metadata in English are limited to outlines; detailed metadata on data sources and methods are lacking.
  - Online extraction tools provide aggregated estimates (example: only the total for Agriculture, forestry and fishing available for 2018).
- Recommendations:
  - Improve data and metadata access in English, including more detailed methodological documentation and data granularity.
  - IMF can provide advice on dissemination best practices and on broader macroeconomic statistics dissemination in the context of dissemination standards assessments.

### Monthly indicator of economic growth
- Further technical assistance is needed to develop a monthly indicator of economic growth; the IMF will explore options following TA provided in 2019. Development will also support improvement of quarterly indicators.

*Source: IMF technical assistance mission report, “1. Priority Recommendations” (content unit: 1vnmea2022002).*

### Appendix I. Is 2020 a Good Choice for a New Benchmark Year

### Appendix I. Is 2020 a Good Choice for a New Benchmark Year for GDP?

### Recommendation on benchmark timing
- The IMF recommends that countries produce benchmark estimates of GDP every five to ten years (with five years the preferred interval).
- For some countries, their next scheduled benchmark year is 2020. It is recommended that countries choose an alternative year to produce GDP benchmarks.
- Countries that planned to produce benchmark estimates of GDP for the year 2021 are advised to consider adjusting their plans and likewise choose an alternate benchmark year if local conditions warrant.

### Reason 1 — representativeness of the benchmark year
- A benchmark year should be representative and reflect “normal economic activity” for a given country.
- Developing benchmark estimates in a year when there is an acute or ongoing economic shock is unadvisable because the shock may cause significant temporary shifts in production and consumption patterns.
- Benchmark estimates are used as weights to aggregate detailed indexes to more aggregated indexes; an unrepresentative benchmark can distort these aggregations.
- Benchmark estimates also serve as interpolation points for time series; if one benchmark is an outlier and poor interpolation techniques are used, compilers could introduce a smoothing effect that:
  - does not reflect economic reality;
  - causes distortions in measurement of the business cycle;
  - tends to mute the rapidity of declines and the subsequent recovery.
- Good interpolation techniques, however, maintain turning points and reflect historical trends regardless of outlier observations.

### Reason 2 — data quality and collection challenges in 2020
- The COVID-19 pandemic has had a major impact on data collection activities; source data used by compilers may be of inferior quality.
- Specific data-collection problems observed:
  - NSOs were sometimes unable to collect data because respondents were shut down or working from home and difficult to contact.
  - Provision of data to NSOs took a back seat to more pressing priorities.
  - The type of transactions during this period are unusual and more susceptible to misrecording.
- Using these “inferior” data sources to establish benchmarks undermines the goal of benchmarking, since governments invest heavily in data operations in a benchmark year.
- If there are widespread challenges in collecting or processing data during the benchmark year, the investment would be better spent in another period.

### Reason 3 — implications for quarterly series and intra-annual information
- Data collection activities for 2020 are more severely affected for intra-annual information, so even if annual information is adequate, estimation of quarterly “benchmark” series may be flawed.
- Major disadvantages of poor quarterly data in the benchmark year include:
  - These poor benchmark quarterly series will not be revised until the next benchmark year.
  - The first annual GDP estimate is generated from an extrapolation of quarterly series, so there may be a bias in the first signal and thus large revisions when the first annual compilation is produced.
  - Severe distortions in the quarterly series will hamper backwards linking, seasonal adjustment, and analyses.

### Specific case: GSO plans and chain-linking
- Plans by the GSO to develop a new benchmark for 2020 are now too advanced to select another benchmark year.
- The implementation of chain-linking will correct possible biases.
- A detailed analysis of data collected for 2020 relating to activities affected by the pandemic and future related source data is recommended.

### Appendix II — compilation of chain volume measures (numerical example)
- The example explains calculations to chain-link annual volumes estimates with two items (Bananas, Pineapple) and three periods (PERIOD 0, PERIOD 1, PERIOD 2).
- Table values (PERIOD 0 / PERIOD 1 / PERIOD 2; P = price, Q = quantity, V = current value):
  - Bananas:
    - P0 Q0 V0: 1 5 5
    - P1 Q1 V1: 2 8 16
    - P2 Q2 V2 (Annually rebased): 8 3 13 39 26
  - Pineapple:
    - P0 Q0 V0: 4 3 12
    - P1 Q1 V1: 4 5 20
    - P2 Q2 V2 (Annually rebased): 20 5 10 50 40
  - Totals (current): 17 (PERIOD 0), 36 (PERIOD 1), 89 (PERIOD 2)
  - Totals (annually rebased): 28 (PERIOD 1), 66 (PERIOD 2)
  - Index: 100 (PERIOD 0), 164.7 (PERIOD 1), 183.3 (PERIOD 2)
  - Chain volume index: 100 (PERIOD 0), 164.7 (PERIOD 1), 302.0 (PERIOD 2)
  - Chain volume estimate: 17 (PERIOD 0), 28.0 (PERIOD 1), 51.3 (PERIOD 2)
  - Constant prices: 17 (PERIOD 0), 28 (PERIOD 1), 53 (PERIOD 2)
  - Constant prices growth: 64.7% (PERIOD 1 vs PERIOD 0), 89.3% (PERIOD 2 vs PERIOD 1)
  - Chain volume growth: 83.3% (PERIOD 2 vs PERIOD 1)
- Explanatory calculations and notes:
  - Annually rebased estimates are values in previous years’ prices. Example: in period 1, the value of bananas is P0 x Q1 = 1 x 8 = 8. The current price estimate is P1 x Q1 = 2 x 8 = 16.
  - Annually rebased estimates across different periods are not priced using the same pricing period and are not time series. Time series are obtained by chaining the indices.
  - Volume indices are obtained as the ratio of the annually rebased value by the current value of the previous period multiplied by 100:
    - Period 1: 28 / 17 x 100 = 164.7
    - Period 2: 66 / 36 x 100 = 183.3
  - Chaining volume indices are obtained by multiplying indices. For period 2 the chain volume index, referenced to period 0, is 164.7 x 183.3 / 100 = 302.
  - The chain volume estimate is obtained by multiplying the chain volume index by the current value of the first period: 17 x 302 / 100 = 51.3.
- Additivity and re-referencing:
  - Chain volume estimates are no longer additive after two periods.
  - Chain volume estimate referenced to period 0 for bananas is 13 for period 2 and for pineapples this value is 40. The sum of these two items is 53, which equals the constant price estimate using period 0 as the pricing period and is higher than the chain volume estimate of both commodities (51.3).
  - One option to maintain additivity in recent periods is to re-reference every year or every two years; however, re-referencing adds complexities to the compilation process.

*Source: Appendix I and Appendix II, 1vnmea2022002 - Appendix I. Is 2020 a Good Choice for a New Benchmark Year for GDP?*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2022/english/1vnmea2022002.pdf_