## onlineannexes

## Source details

**Canonical URL:** [onlineannexes](https://www.imf.org/-/media/files/publications/esr/2022/english/onlineannexes.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/esr/2022/english/onlineannexes.pdf.md)
- [Structured JSON version](/-/media/files/publications/esr/2022/english/onlineannexes.pdf.json)

---

### Conclusion from the EBA CA Model Analysis (global risk aversion interactions and reserve currency status dropped)
- Model refinements and variable selection:
  - Global risk aversion with its interactions and reserve currency status were dropped from the refined EBA CA model specification.
  - In the REER-Level model:
    - Global risk aversion with its interactions is non-robust and excluded from the refined specification.
    - The capital stock variable is excluded from the refined specification.
    - The credit gap variable was excluded from the refined specification after updated-data estimation found results either not statistically significant or counterintuitive.
  - Parameter estimates are generally comparable across previous and refined specifications, with some noteworthy differences described below.

- Key estimation differences and statistical findings:
  - Interest rate differentials and foreign exchange intervention coefficients:
    - Estimates for the interest rate differentials and foreign exchange intervention coefficients are larger with the new specification that uses the FARI capital control index.
  - VAT revenue:
    - The VAT revenue coefficient is larger in the refined specification compared to the previous specification and becomes statistically significant.
  - Demographics and reserve currency status:
    - The effect of demographic variables becomes larger and with stronger statistical significance in the refined specification.
    - The impact of the reserve currency status variable is smaller in absolute value and has weaker statistical significance in the refined specification.
  - Dropping non-robust variables does not materially alter model fit.

- CA–REER semi-elasticities: definitions and updated estimates
  - Definition:
    - The semi-elasticity of the CA-to-GDP ratio with respect to the REER is η_CA = η_TB + η_IB, where:
      - η_TB = ∆(TB/GDP) / ∆(REER/REER) is the trade balance semi-elasticity-to-GDP.
      - η_IB = ∆(IB/GDP) / ∆(REER/REER) is the income balance semi-elasticity-to-GDP.
    - For a given CA gap, the corresponding REER gap (in percent) is REER_gap = CA_gap / η_CA.
  - Trade balance semi-elasticities methodology:
    - η_TB = η_X s_X − η_M s_M, where η_X and η_M are export and import elasticities w.r.t. the REER, and s_X and s_M are nominal shares of exports and imports to GDP.
    - Export and import elasticities are estimated from dynamic panel reduced-form equations (ln X and ln M equations) with country and time fixed effects, using nominal USD series and an eleven-year moving average to smooth export/import shares.
  - Updated medium-run export and import elasticity estimates (panel, EBA sample up to 2019Q4):
    - Exports long-run elasticity: -0.15 (previous: -0.11).
    - Imports long-run elasticity: 0.65 (previous: 0.56).
    - Only the long-term imports elasticity is significant at the 5 percent level.
    - Robustness: excluding outliers based on Cook’s distance leaves the long-run imports elasticity unchanged; the long-run exports elasticity increases somewhat but remains imprecisely estimated.

  - Income balance semi-elasticities: methodology and estimates
    - Method:
      - η_IB = η_IC s_IC − η_ID s_ID, where η_IC and η_ID are elasticities of income credit and debit flows (ratios to GDP) w.r.t. the REER, and s_IC and s_ID are their GDP shares.
      - Elasticities estimated using a panel with annual data 1986–2019, including country and time fixed effects, following Colacelli and others (2020) specification.
      - Long-term elasticities computed as η_F = (∑ β_j^F) / (1 − δ_1^F), for F = {IC, ID}.
    - Preferred estimates (Colacelli and others (2020) preferred specification, EBA sample):
      - η_IC = -0.25 (income credits), statistically significant at the 1 percent level.
      - η_ID = -0.13 (income debits), statistically significant at the 5 percent level.
      - Negative signs reflect valuation effects and currency composition; income credit and debit responses tend to partially offset each other.
      - Gross income flows are significantly smaller than gross trade flows; hence the income balance response to REER is relatively modest.

  - Updated CA–REER semi-elasticities and implications
    - Updated CA elasticities construction:
      - Trade balance elasticities use economy-specific averages of exports- and imports-to-GDP ratios between 2016-2026, using WEO forecasts where needed.
      - Income balance elasticities use data for 2016-2026 for income credit- and debit-to-GDP ratios from the WEO database.
    - Findings:
      - Trade balance response to REER movements is larger than the income balance response.
      - The relation between the CA and the REER mainly reflects exports and imports movements following REER changes.
      - In most cases, income balance elasticities are close to zero and smaller in magnitude than trade elasticities.
    - Aggregate numeric results and interpretation:
      - On a GDP-weighted basis, the average CA elasticity moves from -0.17 to -0.20.
      - A 10 percent REER appreciation is on average associated, other things equal, with a 2 percent of GDP fall in the CA balance.

### COVID-19 and 2021–early 2022 adjustors to current accounts
- Shipping costs and transport services balances:
  - Supply bottleneck pressures eased somewhat in early 2022, but the war in Ukraine led to another increase in shipping costs in mid-2022.
  - As a result, transport services balances have fluctuated sharply in a few countries in 2021.
  - Changes in this item in the current account were sizable (more than ½ percent of GDP in absolute value) in a number of cases such as France, Korea, and Thailand.
  - The impact of the shipping costs increase on the transport services balance and the current account were treated as temporary and expected to dissipate over the medium term.

- Household consumption composition shift (method and findings):
  - Pandemic-driven shift: household consumption moved from services toward durables and other consumer goods.
  - Advanced economies: shift toward both durable and nondurable consumer goods.
  - Emerging market and developing economies: smaller shift away from services and a larger share of nondurables consumption.
  - Adjustor focuses on shift in composition rather than overall level of consumption (the latter declined in 2020 and is reflected in the standard EBA cyclical adjustment).
  - Estimation method for import impact:
    - Compare (1) counterfactual 2021 levels of durables, nondurables, and services consumption assuming 2019 shares and 2021 total private consumption; (2) actual 2021 levels of these categories; and (3) import content of durables, nondurables, and services consumption.
    - Import-content data: United States estimate from Hale and others 2019; other economies scaled using percentage of foreign value added in domestic demand from the OECD’s TiVA data set relative to the United States.
    - Quarterly data available for 14 advanced economies and 7 emerging market and developing economies.
    - For economies with missing data, shifts in 2021 vs 2019 shares are based on advanced-economy and emerging-market-and-developing-economy averages.
  - Impact on exports: based on (1) sum of import impacts across economies for durables, nondurables, and services household consumption; and (2) each economy’s share of total world exports.
    - Durable goods share uses UN Comtrade BEC 61 (“Consumer goods not elsewhere specified, durable”).
    - Nondurable goods share and services import content use data for trade in goods excluding oil.
  - Avoidance of double-counting:
    - Excludes foreign travel share from services consumption (to avoid overlap with travel adjustor).
    - Excludes pharmaceutical and other medical products from nondurables (to avoid overlap with medical goods adjustors).
  - Net estimated current account impacts (largest in US dollars): China (largest net positive) and the United States (largest net negative) (Online Annex Figure 1.2.2).

- Trade in medical products (method and findings):
  - COVID-19 medical emergency produced unusual exports and imports of medical products, affecting current account movements in 2020 and 2021.
  - Data sources and coverage:
    - Export and import data from UN Comtrade and national customs offices for COVID-19 related medical products listed in the WTO (2020) information note “Trade in Medical Goods in the Context of COVID-19” and the “Joint Indicative List of Critical COVID-19 Vaccine Inputs for Consultation” (WTO and others, 2021).
    - Values calculated at HS six-digit subheading level for 145 separate subheadings covering pharmaceuticals, medical supplies, medical equipment, PPE, and vaccine inputs.
  - Adjustment for embedded foreign content:
    - Gross export data adjusted to subtract embedded foreign content (intermediate good imports) using Asian Development Bank Multiregional Input-Output Tables for 2019.
    - Resulting foreign intermediate goods imports are added to imports for each economy considered.
    - Intermediate imports allocated to exporters based on total goods trade export shares for respective years from the April 2022 World Economic Outlook database; shares computed to add to 100 percent for the sample of economies included in the EBA and/or ESR exercises.
    - Reallocation of intermediate goods to exporters has only a modest influence on results.
  - Outcome:
    - Analysis computes associated change in net exports in 2021 compared with 2019 for economies with necessary UN Comtrade or customs data.
    - Estimated positive net effect on the current account is particularly large for medical goods exporters such as China and Malaysia (Online Annex Figure 1.1.1).

- Other idiosyncratic COVID-19 factors:
  - Remittances: fluctuations deemed important in Mexico and Malaysia.
  - Income balance: economies with large foreign direct investment liabilities saw increases in income and current account balances due to lower dividend payments to foreign investors on external liabilities (example: South Africa).
  - Net compensation of employees: temporarily higher in Australia, affecting the income balance.
  - Mineral exports: increased global demand for gold and other minerals not included in the EBA terms-of-trade cyclical component (such as rhodium and palladium) led to temporary increases in mineral exports (example: South Africa).
  - Aeronautics: lower demand for travel services had knock-on effects in the aeronautics sector (example: France).

- Regression evidence on travel and current account relation (Online Annex Table 1.2.1):
  - Dependent variable: Change in Cyclically Adjusted Current Account Balance (Percent of GDP).
  - Reported coefficients and statistics:
    - Change in Travel Balance/GDP: 0.80***, 0.72***, 0.74***, 0.75*** (standard errors reported in parentheses: (0.25), (0.23), (0.26), (0.23)).
    - Lags of Change in CA: 3 (included in all specifications).
    - Output Gap and TOT: NO / YES / NO / YES across the four columns.
    - Economy and Year FE: YES for all columns.
    - Observations: 1,308; 1,301; 1,299; 1,299.
    - R-squared: 0.14; 0.14; 0.11; 0.13.
  - Note: Table reports point estimates and robust standard errors in parentheses. ***, **, and * indicate statistical significance at the 1,5, and 10 level, respectively. 2SLS denotes two-stage-least-squares estimates with the change in travel services exports relative to GDP instrumenting the change in the travel services balance relative to GDP, respectively. GMM denotes generalized method of moments estimates with instruments listed in the table note. All regressions include a dummy variable for China after 2014 to reflect accounting changes in the travel services balance. CA = current account; FE = fixed effect; TOT = terms of trade.

- Summary of COVID-19 adjustors to current accounts (Online Annex Table 1.2.2 — select economy totals, in percent of GDP):
  - Argentina: 0.0
  - Australia: 0.2
  - Belgium: -0.2
  - Brazil: -0.2
  - Canada: 0.1
  - China: -1.6
  - Euro Area: 0.1
  - France: 0.0
  - Germany: -0.6
  - Hong Kong SAR: -0.5
  - India: 0.7
  - Indonesia: 0.6
  - Italy: 0.3
  - Japan: 0.4
  - Korea: -0.9
  - Malaysia: -0.8
  - Mexico: 0.1
  - The Netherlands: -0.4
  - Poland: -0.6
  - Russia: -0.6
  - Saudi Arabia: 1.1
  - Singapore: -1.4
  - South Africa: -2.7
  - Spain: 1.7
  - Sweden: -0.5
  - Switzerland: -0.6
  - Thailand: 5.6
  - Türkiye: -0.2
  - United Kingdom: 0.3
  - United States: 0.8
  - Note: Table reports adjustors to the current account, which equal the estimated impact on the current account but with the opposite sign.

*Source: Online annexes — EBA CA model analysis refinements and CA–REER elasticity estimates; External Sector Report Chapter 1, IMF | July 2022.*

### conclusion from the EBA CA model analysis, where global risk aversion with its interactions and reserve

### Conclusion from the EBA CA Model Analysis (global risk aversion interactions and reserve currency status dropped)

### Model refinements and variable selection
- Global risk aversion with its interactions and reserve currency status were dropped from the refined EBA CA model specification.
- In the REER-Level model:
  - Global risk aversion with its interactions is non-robust and excluded from the refined specification.
  - The capital stock variable is excluded from the refined specification.
  - The credit gap variable was excluded from the refined specification after updated-data estimation found results either not statistically significant or counterintuitive.
- Parameter estimates are generally comparable across previous and refined specifications, with some noteworthy differences described below.

### Key estimation differences and statistical findings
- Interest rate differentials and foreign exchange intervention coefficients:
  - Estimates for the interest rate differentials and foreign exchange intervention coefficients are larger with the new specification that uses the FARI capital control index.
- VAT revenue:
  - The VAT revenue coefficient is larger in the refined specification compared to the previous specification and becomes statistically significant.
- Demographics and reserve currency status:
  - The effect of demographic variables becomes larger and with stronger statistical significance in the refined specification.
  - The impact of the reserve currency status variable is smaller in absolute value and has weaker statistical significance in the refined specification.
- Dropping non-robust variables does not materially alter model fit.

### CA–REER semi-elasticities: definitions and updated estimates
- Definition:
  - The semi-elasticity of the CA-to-GDP ratio with respect to the REER is η_CA = η_TB + η_IB, where:
    - η_TB = ∆(TB/GDP) / ∆(REER/REER) is the trade balance semi-elasticity-to-GDP.
    - η_IB = ∆(IB/GDP) / ∆(REER/REER) is the income balance semi-elasticity-to-GDP.
  - For a given CA gap, the corresponding REER gap (in percent) is REER_gap = CA_gap / η_CA.
- Trade balance semi-elasticities methodology:
  - η_TB = η_X s_X − η_M s_M, where η_X and η_M are export and import elasticities w.r.t. the REER, and s_X and s_M are nominal shares of exports and imports to GDP.
  - Export and import elasticities are estimated from dynamic panel reduced-form equations (ln X and ln M equations) with country and time fixed effects, using nominal USD series and an eleven-year moving average to smooth export/import shares.
- Updated medium-run export and import elasticity estimates (panel, EBA sample up to 2019Q4):
  - Exports long-run elasticity: -0.15 (previous: -0.11).
  - Imports long-run elasticity: 0.65 (previous: 0.56).
  - Only the long-term imports elasticity is significant at the 5 percent level.
  - Robustness: excluding outliers based on Cook’s distance leaves the long-run imports elasticity unchanged; the long-run exports elasticity increases somewhat but remains imprecisely estimated.

### Income balance semi-elasticities: methodology and estimates
- Method:
  - η_IB = η_IC s_IC − η_ID s_ID, where η_IC and η_ID are elasticities of income credit and debit flows (ratios to GDP) w.r.t. the REER, and s_IC and s_ID are their GDP shares.
  - Elasticities estimated using a panel with annual data 1986–2019, including country and time fixed effects, following Colacelli and others (2020) specification.
  - Long-term elasticities computed as η_F = (∑ β_j^F) / (1 − δ_1^F), for F = {IC, ID}.
- Preferred estimates (Colacelli and others (2020) preferred specification, EBA sample):
  - η_IC = -0.25 (income credits), statistically significant at the 1 percent level.
  - η_ID = -0.13 (income debits), statistically significant at the 5 percent level.
  - Negative signs reflect valuation effects and currency composition; income credit and debit responses tend to partially offset each other.
  - Gross income flows are significantly smaller than gross trade flows; hence the income balance response to REER is relatively modest.

### Updated CA–REER semi-elasticities and implications
- Updated CA elasticities construction:
  - Trade balance elasticities use economy-specific averages of exports- and imports-to-GDP ratios between 2016-2026, using WEO forecasts where needed.
  - Income balance elasticities use data for 2016-2026 for income credit- and debit-to-GDP ratios from the WEO database.
- Findings:
  - Trade balance response to REER movements is larger than the income balance response.
  - The relation between the CA and the REER mainly reflects exports and imports movements following REER changes.
  - In most cases, income balance elasticities are close to zero and smaller in magnitude than trade elasticities.
- Aggregate numeric results and interpretation:
  - On a GDP-weighted basis, the average CA elasticity moves from -0.17 to -0.20.
  - A 10 percent REER appreciation is on average associated, other things equal, with a 2 percent of GDP fall in the CA balance.

*Source: Online annexes — EBA CA model analysis refinements and CA–REER elasticity estimates (External Sector Report Chapter 1, IMF | July 2022).*

### 2021. These supply bottleneck pressures eased somewhat in early 2022, but the war in Ukraine has led to

### onlineannexes - 2021. These supply bottleneck pressures eased somewhat in early 2022, but the war in Ukraine has led to

### Shipping costs and transport services balances
- Supply bottleneck pressures eased somewhat in early 2022, but the war in Ukraine led to another increase in shipping costs in mid-2022.
- As a result, transport services balances have fluctuated sharply in a few countries in 2021.
- Annex Table 1.2.2 reports changes in the transport service balances (as percent of GDP) in 2021 compared to 2020.
- Changes in this item in the current account were sizable (more than ½ percent of GDP in absolute value) in a number of cases such as France, Korea, and Thailand.
- The impact of the shipping costs increase on the transport services balance and the current account were treated as temporary and expected to dissipate over the medium term.

### Household consumption composition shift (method and findings)
- Pandemic-driven shift: household consumption moved from services toward durables and other consumer goods.
- Advanced economies: shift toward both durable and nondurable consumer goods.
- Emerging market and developing economies: smaller shift away from services and a larger share of nondurables consumption.
- Adjustor focuses on shift in composition rather than overall level of consumption (the latter declined in 2020 and is reflected in the standard EBA cyclical adjustment).
- Estimation method for import impact:
  - Compare (1) counterfactual 2021 levels of durables, nondurables, and services consumption assuming 2019 shares and 2021 total private consumption; (2) actual 2021 levels of these categories; and (3) import content of durables, nondurables, and services consumption.
  - Import-content data: United States estimate from Hale and others 2019; other economies scaled using percentage of foreign value added in domestic demand from the OECD’s TiVA data set relative to the United States.
  - Quarterly data available for 14 advanced economies and 7 emerging market and developing economies.
  - For economies with missing data, shifts in 2021 vs 2019 shares are based on advanced-economy and emerging-market-and-developing-economy averages.
- Impact on exports: based on (1) sum of import impacts across economies for durables, nondurables, and services household consumption; and (2) each economy’s share of total world exports.
  - Durable goods share uses UN Comtrade BEC 61 (“Consumer goods not elsewhere specified, durable”).
  - Nondurable goods share and services import content use data for trade in goods excluding oil.
- Avoidance of double-counting:
  - Excludes foreign travel share from services consumption (to avoid overlap with travel adjustor).
  - Excludes pharmaceutical and other medical products from nondurables (to avoid overlap with medical goods adjustors).
- Net estimated current account impacts (largest in US dollars): China (largest net positive) and the United States (largest net negative) (Online Annex Figure 1.2.2).

### Trade in medical products (method and findings)
- COVID-19 medical emergency produced unusual exports and imports of medical products, affecting current account movements in 2020 and 2021.
- Data sources and coverage:
  - Export and import data from UN Comtrade and national customs offices for COVID-19 related medical products listed in the WTO (2020) information note “Trade in Medical Goods in the Context of COVID-19” and the “Joint Indicative List of Critical COVID-19 Vaccine Inputs for Consultation” (WTO and others, 2021).
  - Values calculated at HS six-digit subheading level for 145 separate subheadings covering pharmaceuticals, medical supplies, medical equipment, PPE, and vaccine inputs.
- Adjustment for embedded foreign content:
  - Gross export data adjusted to subtract embedded foreign content (intermediate good imports) using Asian Development Bank Multiregional Input-Output Tables for 2019.
  - Resulting foreign intermediate goods imports are added to imports for each economy considered.
  - Intermediate imports allocated to exporters based on total goods trade export shares for respective years from the April 2022 World Economic Outlook database; shares computed to add to 100 percent for the sample of economies included in the EBA and/or ESR exercises.
  - Reallocation of intermediate goods to exporters has only a modest influence on results.
- Outcome:
  - Analysis computes associated change in net exports in 2021 compared with 2019 for economies with necessary UN Comtrade or customs data.
  - Estimated positive net effect on the current account is particularly large for medical goods exporters such as China and Malaysia (Online Annex Figure 1.1.1).

### Other idiosyncratic COVID-19 factors
- Categories and notable country-specific impacts:
  - Remittances: fluctuations deemed important in Mexico and Malaysia.
  - Income balance: economies with large foreign direct investment liabilities saw increases in income and current account balances due to lower dividend payments to foreign investors on external liabilities (example: South Africa).
  - Net compensation of employees: temporarily higher in Australia, affecting the income balance.
  - Mineral exports: increased global demand for gold and other minerals not included in the EBA terms-of-trade cyclical component (such as rhodium and palladium) led to temporary increases in mineral exports (example: South Africa).
  - Aeronautics: lower demand for travel services had knock-on effects in the aeronautics sector (example: France).

### Regression evidence on travel and current account relation (Online Annex Table 1.2.1)
- Dependent variable: Change in Cyclically Adjusted Current Account Balance (Percent of GDP).
- Reported coefficients and statistics:
  - Change in Travel Balance/GDP: 0.80***, 0.72***, 0.74***, 0.75*** (standard errors reported in parentheses: (0.25), (0.23), (0.26), (0.23)).
  - Lags of Change in CA: 3 (included in all specifications).
  - Output Gap and TOT: NO / YES / NO / YES across the four columns.
  - Economy and Year FE: YES for all columns.
  - Observations: 1,308; 1,301; 1,299; 1,299.
  - R-squared: 0.14; 0.14; 0.11; 0.13.
- Note: Table reports point estimates and robust standard errors in parentheses. ***, **, and * indicate statistical significance at the 1,5, and 10 level, respectively. 2SLS denotes two-stage-least-squares estimates with the change in travel services exports relative to GDP instrumenting the change in the travel services balance relative to GDP, respectively. GMM denotes generalized method of moments estimates with instruments listed in the table note. All regressions include a dummy variable for China after 2014 to reflect accounting changes in the travel services balance. CA = current account; FE = fixed effect; TOT = terms of trade.

### Summary of COVID-19 adjustors to current accounts (Online Annex Table 1.2.2 — select economy totals, in percent of GDP)
- Argentina: 0.0
- Australia: 0.2
- Belgium: -0.2
- Brazil: -0.2
- Canada: 0.1
- China: -1.6
- Euro Area: 0.1
- France: 0.0
- Germany: -0.6
- Hong Kong SAR: -0.5
- India: 0.7
- Indonesia: 0.6
- Italy: 0.3
- Japan: 0.4
- Korea: -0.9
- Malaysia: -0.8
- Mexico: 0.1
- The Netherlands: -0.4
- Poland: -0.6
- Russia: -0.6
- Saudi Arabia: 1.1
- Singapore: -1.4
- South Africa: -2.7
- Spain: 1.7
- Sweden: -0.5
- Switzerland: -0.6
- Thailand: 5.6
- Türkiye: -0.2
- United Kingdom: 0.3
- United States: 0.8
- Note: Table reports adjustors to the current account, which equal the estimated impact on the current account but with the opposite sign.

*External Sector Report Chapter 1, IMF | July 2022*

---


_Source: https://www.imf.org/-/media/files/publications/esr/2022/english/onlineannexes.pdf_