## wpiea2019132

## Source details

**Canonical URL:** [wpiea2019132](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019132.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019132.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019132.pdf.json)

---

### A Conceptual Framework
- Framework based on standard balance of payments identities and accounting equations.
- Key stock identity for nominal net foreign asset position B_c:
  - B_c = ∑_{j∈J} S_cj [ Q_j (A^EQ_cj − L^EQ_cj) + (A^D_cj − L^D_cj) ]
- Balance of payments identity (CA = FA) linking net acquisition of foreign assets to trade balance and investment income:
  - TB_c + ∑_{j∈J} IB_cj = ∑_{j∈J} S_cj [ Q_j (dA^EQ_cj − dL^EQ_cj) + (dA^D_cj − dL^D_cj) ]

### Stock and Flow Identities (normalized by nominal GDP)
- Change in net foreign asset position as percent of GDP, d(b_c), decomposed:
  - d(b_c) = −(π_c + g_c) b_c + [ tb_c + ∑_{j∈J} ib_cj ] + ∑_{j∈J} [ \dot{s}_{cj} b_{cj} + \dot{q}_j b^{eq}_{cj} ] + oc_c
- Notation definitions (preserved exactly):
  - b_c = B_c / (P_c GDP_c) NFA, GDP share
  - π_c = dP_c / P_c Change in the CPI of country c
  - g_c = dGDP_c / GDP_c Real GDP growth rate
  - \dot{s}_{cj} = dS_{cj} / S_{cj} Rate of change in value of currency j
  - \dot{q}_j = dQ_j / Q_j Change in equity prices of country j
  - b^{eq}_{cj}, b^d_{cj}, tb_c, ib_c, ca_c, v_c, oc_c defined as in source

### Returns on Debt and Equity — statistical recording principles
- Statistical income is nominal and transaction-based (IMF (2009)); records coupon/interest and distributed dividends but not unpaid (retained) returns.
- Return on debt (extended Fisher equation):
  - i^d_j = r_j + π^e_j + λ^e_j + rp^e_j  for all j∈J
    - r_j real interest rate; π^e_j inflation expectations; λ^e_j expected default rate; rp^e_j risk premium
  - Real-exchange-rate relation:
    - \dot{s}_{cj} = π_c − π_j + \dot{ρ}_{cj}
    - \dot{ρ}_{cj} = \dot{ρ}^e_{cj} + \tilde{ρ}_j
    - Leading to: \dot{s}_{cj} = π^e_c + \tilde{π}_c − π^e_j − \tilde{π}_j + \dot{ρ}^e_{cj} + \tilde{ρ}
- Return on portfolio equity:
  - Dividend yield on portfolio equity: div^peq_j = ge_j − re_j
  - Statistical income on portfolio equity equals distributed dividends: i^peq_j = div_j
  - Portfolio equity price change decomposition:
    - \dot{q}^{peq}_j = π_j + re_j + \hat{q}_j
    - \hat{q}_j = \hat{q}^e_j + \tilde{q}_j
- Return on direct equity:
  - Statistical income equals total earnings: i^fdi_j = ge_j
  - Direct equity price change:
    - \dot{q}^{fdi}_j = π_j + \hat{q}_j

### Statistical Versus Economic Measures of the Current Account
- Statistical income balance:
  - i b^{stat}_c = ∑_{j∈J} ( ge_j b^{fdi}_{cj} + (ge_j − re_j) b^{peq}_{cj} + (r_j + π^e_j + λ^e_j + rp^e_j) b^d_{cj} )
- Decomposition consistent with statistical definition:
  - d(b_c) = −g_c · b_c + ca^{stat}_c + v^{stat}_c + oc_c
  - ca^{stat}_c = tb_c + ∑_{j∈J} ( ge_j b^{eq}_{cj} − re_j b^{peq}_{cj} + (r_j + π^e_j + λ^e_j + rp^e_j) b^d_{cj} )
  - v^{stat}_c = ∑_{j∈J} [ ( \tilde{ρ} + \dot{ρ}^e_{cj} ) b_{cj} + ( \tilde{q}_j + \hat{q}^e_j ) b^{eq}_{cj} + re_j b^{peq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j + π^e_j + λ^e_j ) b^d_{cj} ]
- Economic (real-income) measure proposed:
  - ca^{econ}_c = tb_c + ∑_{j∈J} ( ge_j b^{eq}_{cj} + (r_j + rp^e_j) b^d_{cj} )
  - v^{econ}_c = ∑_{j∈J} [ ( \tilde{ρ} + \dot{ρ}^e_{cj} ) b_{cj} + ( \tilde{q}_j + \hat{q}^e_j ) b^{eq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j ) b^d_{cj} ]
- Measurement-distortion relations:
  - ca^{stat}_c = ca^{econ}_c − M^{econ}_c
  - v^{stat}_c = v^{econ}_c + M^{econ}_c + μ_c
  - d(b_c) = −g_c · b_c + ca^{econ}_c + μ_c + oc_c
  - M^{econ}_c = ∑_{j∈J} [ re_j b^{peq}_{cj} − (π^e_j + λ^e_j) b^d_{cj} ]
  - μ_c = ∑_{j∈J} [ \tilde{ρ} b_{cj} + \tilde{q}_j b^{eq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j ) b^d_{cj} ]

### Measurement Distortions — characterization and implications
- General:
  - Measurement issues M_c are associated with one-for-one shifts between the income balance and NIIP valuation changes and cancel out if income and valuation changes are consolidated.
  - Primary drivers: financing of new equity (retained earnings vs distributed/dividend-financed), recording of nominal vs real returns, and treatment of default risk.

- The Inflation Distortion
  - Statistical recording of nominal interest overstates real returns to the extent of expected inflation compensation.
  - Inflation distortion IE_c:
    - IE_c = ∑_{j∈J} π^e_j b^d_{cj}
  - Effects depend on net debt positions by currency and bilateral inflation differentials.
  - Two otherwise identical economies with different exposures to currencies with differing expected inflation (but equal real rates) will have different statistical current accounts and valuation changes due to this distortion.

- The Retained Earnings Distortion
  - Statistical income records distributed dividends but not retained earnings on portfolio equity; retained earnings increase asset market value and show up in NIIP valuation changes.
  - Retained earnings distortion RE_c:
    - RE_c = ∑_{j∈J} −re_j b^{peq}_{cj}
  - Larger gross portfolio equity positions and higher retention rates increase distortion magnitude.
  - Direction:
    - Large gross foreign portfolio equity asset positions → statistical income underestimates true returns (statistical current account underestimates economic current account).
    - Large gross portfolio equity liabilities suggest current account overestimation.
  - No distortion for direct investment retained earnings (treated as new FDI capital): retained earnings on FDI recorded as financial account transaction, dA^{FDI}_j = re_j A^{FDI}_j.
  - Investment fund shares treated as FDI (no retained-earnings distortion).

- The Expected Default Rate Distortion
  - Expected default compensation is recorded in nominal interest; associated expected capital losses are reflected in NIIP valuation changes, producing a distortion analogous to inflation.
  - Default distortion DI_c:
    - DI_c = ∑_{j∈J} λ^e_j b^d_{cj}
  - Debt risk premia related to higher moments (rp_j) need not produce measurement issues because there is no corresponding offset in NIIP valuation changes.

### Definitional Measurement Issues (Section 2.5)
- Broader definitional measurement issues relate to treatment of expected valuation changes, including expected equity price changes (ˆq^e_j) and expected real exchange rate changes ( ̇ρ^e_cj).
- The line between income on assets and pure valuation changes is conceptually ambiguous and requires further analysis (see Adler and Garcia-Macia (2018)).

### Empirical estimation strategy (Section 3.1)
- Two measurement distortions empirically estimated: retained earnings distortion of income on portfolio equity, and inflation distortion of debt income.
- Data limitations prevent reasonable estimation of the distortion related to expected default rates.
  - Estimating ex-ante compensation for default risk requires an empirical measure of the default risk component in interest rates on debt held cross border, which is not currently available.
  - Credit default swaps (CDS) and comparisons between effective interest rates and interest rate swaps are problematic measures for this purpose.
- Data sources and definitions summarized in Appendix Table 1 (in the source).

- Retained earnings estimation methods (three approaches; results averaged):
  - Flow approach:
    - Uses recorded income streams and country dividend-yield (dy_c) and price-earnings (PE_c) ratios.
    - Estimated unrecorded retained earnings:
      - RE_c = re^W iA_PEQ_c − re_c iL_PEQ_c  (Equation (30) in source)
      - re_c = 1/(dy_c/PE_c − 1)
  - Stock approach:
    - Uses gross portfolio positions and PE ratios to estimate total earnings; retained earnings is the difference between total earnings and distributed dividends.
    - Estimated retained earnings:
      - RE'_c = rep^W A_PEQ_c − rep_c L_PEQ_c  (Equation (31) in source)
      - rep_c = 1/(PE_c − dy_c)
  - Hybrid approach:
    - Uses international portfolio equity income flows for distributed dividends and portfolio stock positions with PE data for total earnings.
    - Estimated retained earnings:
      - RE''_c = A_PEQ_c / PE^W − L_PEQ_c / PE_c − (iA_PEQ_c − iL_PEQ_c)  (Equation (32) in source)
  - The paper uses the average across the three estimation methods.

- Inflation distortion methodology:
  - Inflation income in country c:
    - π_income_c = Σ_j π_j ND_cj  (Equation (33) in source)
      - π_j is inflation rate in currency j; ND_cj is country c’s net foreign debt position denominated in currency j (including domestic currency c among j).
  - Annual inflation rates from International Financial Statistics.
  - Expected inflation approximated by a 5-year moving average of past realized inflation rates.
  - Currency weights from country authorities when available; otherwise estimates from Benetrix et al (2015).
  - Gross foreign debt positions from External Wealth of Nations database (Lane and Milesi-Ferretti (2007)).

### Estimated distortions and current account imbalances — main findings (Section 3.2)
- Sample: 28 largest economies, representing around 85 percent of world GDP, period 1994-2016. (Sample excludes Ireland due to data limitations.)
- Main findings:
  - Accounting for the two measurement distortions does not fundamentally change the overall picture of current account imbalances in the sample, but the distortions may account for a non-trivial share of imbalances.
  - On average across countries in any given time period, the distortions appear to contribute to an overestimation of imbalances (i.e., true imbalances may be smaller than suggested by existing current account statistics).
  - Example: in 2016, about 22 percent of overall current account surpluses and deficits appear to be related to mismeasurement according to the estimates.
  - Importance of the distortions has grown over time, consistent with increased global integration in trade and finance.
- Relative importance:
  - Both retained earnings and inflation distortions increase over time, although the inflation distortion appears to be a significantly larger source of mismeasurement.
- Cross-section and country-specific patterns (average 2012-16):
  - Retained earnings distortion:
    - Leads to apparent overvaluation of the current account in: Korea, Russia, Switzerland.
    - Leads to apparent undervaluation of the current account (up to several percentage points of GDP) in: Hong Kong SAR, Singapore.
  - Inflation distortion:
    - Particularly large in financial center economies: Hong Kong SAR, Singapore, followed by Switzerland and Belgium.
    - In these countries inflation distortions lead to overestimation of the current account.
  - In Hong Kong SAR and Singapore, the two distortions are large in absolute value and partly offset each other, but the inflation distortion dominates, implying an upward bias in current account measures.
- Overall: consistently large total distortions primarily in economies considered financial centers with large foreign investment positions (notably Hong Kong SAR, Singapore, Switzerland, and to a lesser extent Belgium and the UK).
- Uncertainty and caveats:
  - Large uncertainty surrounds point estimates.
  - Only two measurement distortions that can be meaningfully estimated for a large set of countries are accounted for; other sources could be influential.
  - Some counterparties to cross-border operations are missing from the sample, causing parts of estimated distortions not to cancel out across sampled countries.
  - Refined country-specific estimates could use more detailed data on currency composition of foreign investment positions where available.

### Merchanting and profit-shifting — conceptual conclusions (Section 4)
- Merchanting can affect how flows are divided between income and trade balance, but it is not conceptually clear that merchanting per se distorts the current account measure.
- Two forms of merchanting in statistics:
  - True merchanting (domestically-owned company): recorded as exports of intermediation services; if properly compiled and reported, should not affect measurement of the current account; contributes positively to the current account with corresponding financial account entry.
  - Profit shifting (multinationals using low-tax jurisdictions): intra-company imports/exports recorded as merchanting in the low-tax country, contributing positively to that country’s trade balance; the subsidiary is owned by the foreign parent so the net merchanting balance is offset by an income balance entry (dividends or retained earnings) to the parent. Except for tax components, profit-shifting affects composition between trade and income balances, but not necessarily the overall current account balance.
- While profit shifting can take many complex forms, an unequivocal conceptual argument that profit shifting distorts the measurement of the current account is lacking.
- Merchanting activities may correlate with the measurement distortions explored (notably retained earnings); retained earnings distortion is a relevant candidate to explain related statistical patterns.

### Box 1.2 (IMF (2017)) — major findings, limitations, and policy relevance
- Major findings:
  - The paper proposes a unified conceptual framework for understanding the role of inflation, retained earnings, default risk, and other aspects of investment income in measurement distortions of the income balance.
  - Based on estimates of the retained earnings and inflation distortions, measurement distortions can be empirically non-negligible drivers of observed current account imbalances.
  - Distortions tend to be more important in economies with large gross and net external investment positions.
- Data and empirical limitations:
  - Data limitations constrain more precise estimation and consideration of additional distortions such as default risk.
  - Specific constraints: limited information on stock-flow reconciliation in external accounts; limited information about currency composition of gross positions.
  - Need to strengthen data collection on cross-border transactions and cross-border collaboration of statistical authorities.
- Policy relevance and recommendations:
  - While practical and data limitations may prevent adoption of different statistical measures, it is central for good policy analysis that measurement distortions are taken into account in comprehensive assessments of countries’ external imbalances.
  - The framework and estimates represent a small step toward quantifying potential distortions; continued work and better data are required.
  - A better conceptual understanding of multinational profit shifting and its effects on external accounting is needed as global value chain integration proceeds.
- Data sources referenced (selection):
  - IMF Balance of Payments Database; IMF World Economic Outlook Database; Philip R. Lane and Gian Maria Milesi-Ferretti, The External Wealth of Nations Database; IMF Coordinated Portfolio Investment Survey; Thomson Reuters Datastream; Bénétix, Lane, and Shambaugh (2015) database.

*Source: wpiea2019132 (PDF chapter/section content provided in the prompt)*

### 3.  Section 4 discusses other empirical issues related to measurement, including the impli-

### wpiea2019132 - 3.  Section 4 discusses other empirical issues related to measurement, including the impli-

### A Conceptual Framework
- Framework based on standard balance of payments identities and accounting equations.
- Key stock identity for nominal net foreign asset position B_c:
  - B_c = ∑_{j∈J} S_cj [ Q_j (A^EQ_cj − L^EQ_cj) + (A^D_cj − L^D_cj) ]
- Balance of payments identity (CA = FA) links net acquisition of foreign assets to trade balance and investment income:
  - TB_c + ∑_{j∈J} IB_cj = ∑_{j∈J} S_cj [ Q_j (dA^EQ_cj − dL^EQ_cj) + (dA^D_cj − dL^D_cj) ]

### Stock and Flow Identities (normalized by nominal GDP)
- Change in net foreign asset position as percent of GDP, d(b_c), decomposed as:
  - d(b_c) = −(π_c + g_c) b_c + [ tb_c + ∑_{j∈J} ib_cj ] + ∑_{j∈J} [ \dot{s}_{cj} b_{cj} + \dot{q}_j b^{eq}_{cj} ] + oc_c
  - Notation definitions preserved exactly, including:
    - b_c = B_c / (P_c GDP_c) NFA, GDP share
    - π_c = dP_c / P_c Change in the CPI of country c
    - g_c = dGDP_c / GDP_c Real GDP growth rate
    - \dot{s}_{cj} = dS_{cj} / S_{cj} Rate of change in value of currency j
    - \dot{q}_j = dQ_j / Q_j Change in equity prices of country j
    - b^{eq}_{cj}, b^d_{cj}, tb_c, ib_c, ca_c, v_c, oc_c defined as in source

### Returns on Debt and Equity — statistical recording principles
- Statistical income is nominal and transaction-based (IMF (2009)); generally records recorded coupon/interest and distributed dividends but not unpaid (retained) returns.
- Return on debt (extended Fisher equation):
  - i^d_j = r_j + π^e_j + λ^e_j + rp^e_j  for all j∈J
    - r_j real interest rate; π^e_j inflation expectations; λ^e_j expected default rate; rp^e_j risk premium
  - Real-exchange-rate relation:
    - \dot{s}_{cj} = π_c − π_j + \dot{ρ}_{cj}
    - \dot{ρ}_{cj} = \dot{ρ}^e_{cj} + \tilde{ρ}_j
    - Leading to: \dot{s}_{cj} = π^e_c + \tilde{π}_c − π^e_j − \tilde{π}_j + \dot{ρ}^e_{cj} + \tilde{ρ}
- Return on portfolio equity:
  - Dividend yield on portfolio equity: div^peq_j = ge_j − re_j
  - Statistical income on portfolio equity equals distributed dividends: i^peq_j = div_j
  - Portfolio equity price change decomposition:
    - \dot{q}^{peq}_j = π_j + re_j + \hat{q}_j
    - \hat{q}_j = \hat{q}^e_j + \tilde{q}_j
- Return on direct equity:
  - Statistical income equals total earnings: i^fdi_j = ge_j
  - Direct equity price change:
    - \dot{q}^{fdi}_j = π_j + \hat{q}_j

### Statistical Versus Economic Measures of the Current Account
- Statistical income balance:
  - i b^{stat}_c = ∑_{j∈J} ( ge_j b^{fdi}_{cj} + (ge_j − re_j) b^{peq}_{cj} + (r_j + π^e_j + λ^e_j + rp^e_j) b^d_{cj} )
- Decomposition consistent with statistical definition:
  - d(b_c) = −g_c · b_c + ca^{stat}_c + v^{stat}_c + oc_c
  - ca^{stat}_c = tb_c + ∑_{j∈J} ( ge_j b^{eq}_{cj} − re_j b^{peq}_{cj} + (r_j + π^e_j + λ^e_j + rp^e_j) b^d_{cj} )
  - v^{stat}_c = ∑_{j∈J} [ ( \tilde{ρ} + \dot{ρ}^e_{cj} ) b_{cj} + ( \tilde{q}_j + \hat{q}^e_j ) b^{eq}_{cj} + re_j b^{peq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j + π^e_j + λ^e_j ) b^d_{cj} ]
- Economic (real-income) measure proposed:
  - ca^{econ}_c = tb_c + ∑_{j∈J} ( ge_j b^{eq}_{cj} + (r_j + rp^e_j) b^d_{cj} )
  - v^{econ}_c = ∑_{j∈J} [ ( \tilde{ρ} + \dot{ρ}^e_{cj} ) b_{cj} + ( \tilde{q}_j + \hat{q}^e_j ) b^{eq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j ) b^d_{cj} ]
- Measurement-distortion relations:
  - ca^{stat}_c = ca^{econ}_c − M^{econ}_c
  - v^{stat}_c = v^{econ}_c + M^{econ}_c + μ_c
  - d(b_c) = −g_c · b_c + ca^{econ}_c + μ_c + oc_c
  - M^{econ}_c = ∑_{j∈J} [ re_j b^{peq}_{cj} − (π^e_j + λ^e_j) b^d_{cj} ]
  - μ_c = ∑_{j∈J} [ \tilde{ρ} b_{cj} + \tilde{q}_j b^{eq}_{cj} − ( \tilde{π}_j + \tilde{λ}_j ) b^d_{cj} ]

### Measurement Distortions — characterization and implications
- General:
  - Measurement issues M_c are associated with one-for-one shifts between the income balance and NIIP valuation changes and cancel out if income and valuation changes are consolidated.
  - Primary drivers: financing of new equity (retained earnings vs distributed/dividend-financed), recording of nominal vs real returns, and treatment of default risk.

- The Inflation Distortion
  - Statistical recording of nominal interest overstates real returns to the extent of expected inflation compensation.
  - Inflation distortion IE_c:
    - IE_c = ∑_{j∈J} π^e_j b^d_{cj}
  - Effects depend on net debt positions by currency and bilateral inflation differentials.
  - Two otherwise identical economies with different exposures to currencies with differing expected inflation (but equal real rates) will have different statistical current accounts and valuation changes due to this distortion.

- The Retained Earnings Distortion
  - Statistical income records distributed dividends but not retained earnings on portfolio equity; retained earnings increase asset market value and show up in NIIP valuation changes.
  - Retained earnings distortion RE_c:
    - RE_c = ∑_{j∈J} −re_j b^{peq}_{cj}
  - Larger gross portfolio equity positions and higher retention rates increase distortion magnitude.
  - Direction: large gross foreign portfolio equity asset positions → statistical income underestimates true returns (statistical current account underestimates economic current account); large gross portfolio equity liabilities suggest current account overestimation.
  - No distortion for direct investment retained earnings (treated as new FDI capital): retained earnings on FDI recorded as financial account transaction, dA^{FDI}_j = re_j A^{FDI}_j.
  - Investment fund shares treated as FDI (no retained-earnings distortion).

- The Expected Default Rate Distortion
  - Expected default compensation is recorded in nominal interest; associated expected capital losses are reflected in NIIP valuation changes, producing a distortion analogous to inflation.
  - Default distortion DI_c:
    - DI_c = ∑_{j∈J} λ^e_j b^d_{cj}
  - Debt risk premia related to higher moments (rp_j) need not produce measurement issues because there is no corresponding offset in NIIP valuation changes.

*Source: wpiea2019132 (PDF chapter/section content provided in the prompt)*

### 2.5    Definitional Measurement Issues

### 2.5    Definitional Measurement Issues

### Key definitional issues
- Broader definitional measurement issues relate to treatment of expected valuation changes, including expected equity price changes (ˆq e j) and expected real exchange rate changes ( ̇ρ e cj).
- The line between income on assets and pure valuation changes is conceptually ambiguous and requires further analysis (see Adler and Garcia-Macia (2018)).

### Empirical application: scope and limitations
- Two measurement distortions are empirically estimated: retained earnings distortion of income on portfolio equity, and inflation distortion of debt income.
- Data limitations prevent reasonable estimation of the distortion related to expected default rates.
  - Estimating ex-ante compensation for default risk requires an empirical measure of the default risk component in interest rates on debt held cross border, which is not currently available.
  - Credit default swaps (CDS) and comparisons between effective interest rates and interest rate swaps are problematic measures for this purpose.
- All data sources and definitions are summarized in Appendix Table 1 (in the source).

### 3.1 Empirical estimation strategy

#### 3.1.1 The retained earnings distortion — overview
- Retained earnings on portfolio equity are estimated using:
  - international investment income and stock positions,
  - stock market data on average earnings and dividends for equities listed in national stock markets,
  - bilateral country exposures to portfolio equity.
- Assumption: portfolio breakdown of stocks in cross-border portfolio equity investments is similar to the national average.
- Data on equity stocks and flows are adjusted to be net of stock and income on investment fund shares (share assumed 1 for missing country-years).
- Three estimation approaches are used and averaged: flow approach, stock approach, hybrid approach.

Flow approach
- Uses recorded income streams on foreign portfolio equity positions to reflect distributed dividends.
- Applies country dividend-yield (dy_c) and price-earnings (PE_c) ratios to recorded investment income on portfolio equity assets (iA_PEQ_c) and liabilities (iL_PEQ_c).
- Estimated unrecorded retained earnings:
  - RE_c = re^W iA_PEQ_c − re_c iL_PEQ_c (Equation (30) in source).
  - re_c = 1/(dy_c/PE_c − 1)  [notation as in source: re_c = 1 dy_c PE_c −1].
- Dividend yield and price earnings ratios are based on averages for country j stock markets retrieved from Datastream.

Stock approach
- Uses gross portfolio investment positions and stock market PE ratios to estimate total earnings; multiplies earnings by dividend yield to estimate distributed dividends; retained earnings is the difference.
- Estimated retained earnings:
  - RE'_c = rep^W A_PEQ_c − rep_c L_PEQ_c (Equation (31) in source).
  - rep_c = 1/(PE_c − dy_c)  [notation as in source: rep_c = 1 PE − dy_c].

Hybrid approach
- Uses international portfolio equity income flows to capture distributed dividends and portfolio stock positions with PE data to estimate total earnings.
- Estimated retained earnings:
  - RE''_c = A_PEQ_c / PE^W − L_PEQ_c / PE_c − (iA_PEQ_c − iL_PEQ_c) (Equation (32) in source).
- Advantage: maximizes use of observed external income and stock position data; drawback: earnings and distributed dividends rely on different sources and may be less consistent.

- For the paper’s results, the average across the three estimation methods is used.

#### 3.1.2 The inflation distortion — methodology
- Inflation distortion estimated using:
  - actual data from authorities or estimates of currency composition of international debt positions,
  - inflation rates.
- Inflation income in country c:
  - π_income_c = Σ_j π_j ND_cj (Equation (33) in source),
    - where π_j is inflation rate in currency j and ND_cj is country c’s net foreign debt position denominated in currency j (including domestic currency c among j).
- Annual inflation rates are from International Financial Statistics.
- Expected inflation approximated by a 5-year moving average of past realized inflation rates.
- Currency weights in international debt positions use country authorities’ data when available; otherwise estimates from Benetrix et al (2015).
- Total exposure (including off-balance sheet) would be the relevant measure, but data on off-balance sheet currency exposures in cross-border positions are not available.
- Data on gross foreign debt positions are from the External Wealth of Nations database (Lane and Milesi-Ferretti (2007)).

### 3.2 Estimated distortions and current account imbalances — main findings
- Retained earnings and inflation distortions computed for the 28 largest economies, representing around 85 percent of world GDP, for the period 1994-2016.
  - The sample excludes Ireland due to data limitations.
- Accounting for the two measurement distortions does not fundamentally change the overall picture of current account imbalances in the sample, but:
  - The distortions may account for a non-trivial share of imbalances.
  - On average across countries in any given time period, the distortions appear to contribute to an overestimation of imbalances (i.e., true imbalances may be smaller than suggested by existing current account statistics).
  - Example: in 2016, about 22 percent of overall current account surpluses and deficits appear to be related to mismeasurement according to the estimates.
  - Importance of the distortions appears to have grown over time, consistent with increased global integration in trade and finance.
- Relative importance:
  - Both retained earnings and inflation distortions increase over time, although the inflation distortion appears to be a significantly larger source of mismeasurement.
- Cross-section and country-specific patterns (average 2012-16):
  - Retained earnings distortion:
    - Countries where retained earnings distortion leads to overvaluation of the current account: Korea, Russia, Switzerland.
    - Countries where retained earnings distortion leads to undervaluation of the current account (up to several percentage points of GDP): Hong Kong SAR, Singapore.
  - Inflation distortion:
    - Particularly large in financial center economies: Hong Kong SAR, Singapore, followed by Switzerland and Belgium.
    - In these countries inflation distortions lead to overestimation of the current account.
  - In Hong Kong SAR and Singapore, the two distortions are large in absolute value and partly offset each other, but the inflation distortion dominates, implying an upward bias in current account measures.
- Overall: consistently large total distortions primarily in economies considered financial centers, with large foreign investment positions (notably Hong Kong SAR, Singapore, Switzerland, and to a lesser extent Belgium and the UK).
- Figure-based insight (as described in source):
  - Figure 1: accounting for distortions does not fundamentally change global CA balances; distortions have grown over time.
  - Figure 2: inflation distortion larger than retained earnings distortion on average 2001-16.
  - Figures 3–6: cross-country variation with financial centers standing out; absolute measurement distortions plotted against gross foreign assets in percent of GDP.
- Uncertainty and caveats:
  - Large amount of uncertainty surrounds point estimates.
  - Only two measurement distortions that can be meaningfully estimated for a large set of countries are accounted for; other sources of distortion could be influential.
  - Some counterparties to cross-border operations are missing from the sample, which can cause parts of estimated distortions not to cancel out across sampled countries.
  - Refined country-specific estimates could use more detailed data on currency composition of foreign investment positions where available.

### 4 Merchanting and profit-shifting — conceptual conclusions
- Merchanting can affect how flows are divided between income and trade balance, but it is not conceptually clear that merchanting per se distorts the current account measure.
- Two forms of merchanting in statistics:
  - True merchanting (domestically-owned company): should be recorded as exports of intermediation services and, if properly compiled and reported, should not affect measurement of the current account; it contributes positively to the current account with a corresponding financial account entry.
  - Profit shifting (multinationals using low-tax jurisdictions): intra-company imports/exports recorded as merchanting in the low-tax country, contributing positively to that country’s trade balance; because the subsidiary is owned by the foreign parent, the net merchanting balance is offset by an income balance entry (dividends or retained earnings) to the parent. Except for tax components, profit-shifting affects composition between trade and income balances, but not necessarily the overall current account balance.
- While profit shifting can take many complex forms (references in source: Lipsey (2010), Fatih Guvenen and Ruhl (2017), Lane (2017)), an unequivocal conceptual argument that profit shifting distorts the measurement of the current account is lacking.
- Merchanting activities may correlate with the measurement distortions explored (notably retained earnings); retained earnings distortion is a relevant candidate to explain related statistical patterns.

*Source: IMF working paper section 2.5 and associated empirical results as presented in the supplied content.*

### Box 1.2 in IMF (2017) discusses the case of Ireland.

### wpiea2019132 - Box 1.2 in IMF (2017) discusses the case of Ireland

### Major findings on measurement distortions
- The paper proposes a unified conceptual framework for understanding the role of different measurement distortions in the income balance of the current account.
- The framework shows that the statistical treatment of:
  - inflation,
  - retained earnings,
  - default risk, and
  - other aspects of investment income
  can have material implications for the measurement of external accounts.
- Based on estimates of the retained earnings and inflation distortions, the paper finds that measurement distortions can be empirically non-negligible drivers of observed current account imbalances.
- These distortions tend to be more important in economies with large gross and net external investment positions.
- As global integration proceeds, merchanting and profit shifting by multinational corporations can affect external accounting; a more granular understanding is needed to assess associated distortions in external accounts.

### Data and empirical limitations
- Data limitations remain an important constraint on:
  - more precise estimation of specific distortions, and
  - taking into account new possible distortions such as those related to default risk.
- For the inflation and retained earnings distortions, particular constraints include:
  - limited information on stock-flow reconciliation in external accounts (statistics to reconcile flow information from balance of payments statistics with stock information from the international investment position), and
  - limited information about the currency composition of gross positions.
- There is a need to strengthen data collection efforts on cross-border transactions and cross-border collaboration of statistical authorities on such efforts.

### Policy relevance and recommendations
- While conceptual, practical and data limitations may prevent the adoption of different statistical measures, it is central for good policy analysis that measurement distortions are taken into account in comprehensive assessments of countries’ external imbalances.
- The framework and specific estimates presented represent a small step toward quantifying potential distortions; continued work and better data are required.
- A better conceptual understanding of how multinational corporations shift profits cross borders between subsidiaries, and how this affects external accounting, will be needed as integration in global value chains proceeds.

### Data sources referenced (selection)
- IMF Balance of Payments Database: investment income on equity and investment fund shares, credit/debit balances in USD; equity and investment fund shares in USD.
- IMF World Economic Outlook Database: percentage change in Consumer Prices (period average); Current Account Balance in USD; Nominal GDP in USD.
- Philip R. Lane and Gian Maria Milesi-Ferretti, The External Wealth of Nations Database: Foreign Portfolio Investment Positions in Debt Instruments, in USD; Total Reserve Assets, in USD.
- IMF Coordinated Portfolio Investment Survey: bilateral country exposure weights for portfolio equity assets/liabilities.
- Thomson Reuters Datastream: Dividend-Yield Ratio; Price-Earnings Ratio.
- Bénétix, Agustin S., Lane, Philip R., and Shambaugh, Jay C., 2015 database: currency weights of international debt positions (shares of US Dollar, Euro, GB Pound, Japanese Yen, Swiss Franc and domestic currency).

*Source: Box 1.2 in IMF (2017), as included in wpiea2019132.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019132.pdf_