## _wp04134

## Source details

**Canonical URL:** [_wp04134](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04134.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04134.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04134.pdf.json)

---

### Introduction: motivation and scope
- Purpose: understand the geography of international portfolio equity investment to reveal frictions in global economic integration and explain bilateral asymmetries in portfolios.
- Data: geographical breakdown of international portfolio holdings at end-2001 by 67 source countries and up to 218 destination countries/territories.
- Key conceptual points:
  - Bilateral portfolio asymmetries affect cost of equity capital, wealth exposures to country-specific shocks, risk-sharing, and comovements in macro/financial variables.
  - Research questions: roles of product-market frictions, cultural/informational factors, and domestic vs. international financial development.

### Theoretical framework
- Approaches summarized:
  - Obstfeld and Rogoff (2001): product-market (trade) frictions generate home bias even with complete financial markets; extended to N-country model here.
  - Martin and Rey (2003): transactional frictions in asset markets (iceberg costs, endogenous market capitalization) and informational frictions.
  - Davis, Nalewaik, and Willen (2001): incomplete markets with country-specific labor-income risk; cross-country heterogeneity in holdings due to differing correlations between assets and labor income.
- N-country Obstfeld–Rogoff generalization (benchmark):
  - Country j’s share of country i’s equity is decreasing in bilateral trading cost τij relative to i’s average trading costs and increasing in relative preference for good i (ωij).
  - Theoretical allocation link: log(x∗ij) corresponds one-to-one to log(IMPij) (equation (12) log(x∗ij)=log(IMPij)).
  - Full empirical specification allows deviations via aggregate and bilateral financial frictions: log(xij) = log(IMPij) + Fij + source/host fixed effects + error (equation (14)).

### Empirical strategy
- Cross-sectional observation: single year, end-2001.
- Goal: isolate contributions of (i) bilateral factors, (ii) source-country factors, and (iii) host-country factors.
- Bilateral regressors include:
  - log(1+imports), log distance, time zone difference, common language dummy, colonial relationship dummy, currency union dummy, investment tax treaty dummy, common legal origin dummy, correlations (stock returns, GDP growth, growth vs. host stock returns).
- Controls and methods:
  - Source and host fixed effects capture aggregate country characteristics (size, GDP per capita, stock market capitalization, capital controls, financial center status, historical Sharpe ratio, beta, volatility).
  - Estimation: panel OLS with fixed effects, IV, and Tobit to account for many zero bilateral observations.
  - Dependent variable: log(1+equity).

### Data: description and key measurement issues
- CPIS coverage: end-1997 and end-2001; 2001 CPIS includes 67 source countries.
- Total recorded portfolio equity investment in CPIS (end-2001): US$5.16 trillion.
- Notable non-participating jurisdictions: British Virgin Islands, China, Kuwait, Saudi Arabia, Taiwan Province of China, UAE.
- Under-reporting and custodial/third-party reporting differences (examples: Cayman Islands, Bahamas, Germany; U.S. uses mix of methods).
- Persistence: correlation of bilateral positions for end-2001 and end-2002 is 0.99 for pairs with data for both years (preliminary end-2002).
- Offshore/financial center role:
  - Offshore centers and small financial centers reported assets > US$700 billion (incomplete coverage caveat).
  - Derived equity liabilities to offshore and small financial centers: US$870 billion.

### Stylized facts (aggregate and bilateral)
- Aggregate figures:
  - Total CPIS equity assets = US$5.16 trillion.
  - Largest asset holders (absolute): United States US$1,613 billion; United Kingdom US$558 billion; Germany US$381 billion; Luxembourg US$319 billion; Switzerland US$247 billion.
  - Largest asset holders (ratio to GDP): Luxembourg 16.6; Jersey 14.8; Guernsey 13.5; Isle of Man 9.6; Bermuda 8.6.
- Derived liabilities (top destinations; bracketed numbers are IIP-reported liabilities): United States 1000 (1533) US$ billion; United Kingdom 711 (768); France 387 (416); Luxembourg 376 (N.A.); Japan 330 (376).
- Stock market structure (2001 highlights):
  - World shares: United States domestic stock market cap 48.9 percent of world; Euro area 15.9; Japan 9.3; United Kingdom 8.9.
  - Percent of domestic stock market owned by foreign portfolio investors: Euro area 36.5 percent; United Kingdom 35.6 percent; United States 12.9 percent; Japan 16.7 percent.
- Allocation vs. benchmark:
  - United Kingdom invests much more in the euro area than the United States.
  - Japan’s foreign equity investment aligns most closely with a market-cap benchmark.

### Empirical findings: bilateral equity holdings
- Trade–equity link (robust):
  - Table 5 (All countries) OLS panel estimate for Avg. imports, 1997–2001 = 0.32 (t = 17.93)**.
  - Tobit estimates larger in many specifications (e.g., Table 5 Tobit (1) 0.73 (t = 21.82)**; Tobit (5) 0.52, t = 5.13)**).
  - Additional controls reduce but do not eliminate coefficient (e.g., column (2) ~0.22, t = 8.42)**; narrower sample column (3) 0.21 (t = 3.58)**.
  - Simple pooled regression: elasticity 0.39 (t-statistic 22.2), R-squared 0.18.
- Informational / cultural proximity effects:
  - Distance: doubling physical distance reduces equity holdings by 61 percent in broad OLS (Table 5 column (2) log distance = -0.61, t = 9.17)**; effect weakens in samples with stock-market controls.
  - Common language: raises equity holdings by about 40 percent (Table 5 column (2) coefficient 0.33, t = 3.63)**.
  - Colony dummy: large effect for emerging-market sources (Table 7 column (2) colony = 1.55, t = 2.98)**.
  - Currency union dummy: large and significant in full-sample OLS (Table 5 column (2) 1.33, t = 7.64)**; OECD subsample with stock-market variables (Table 6 column (3) currency union = 0.61, t = 2.69)**.
  - Common legal origin: raises investments by 30–40 percent (Table 5 column (3) = 0.28, t = 2.93)**.
- Financial correlation measures:
  - Correlation in GDP growth rates often positive and significant (e.g., Table 5 column (2) correl. in growth rates = 0.61, t = 3.66)**.
  - Correlation in stock returns positive and significant in some specs (Table 5 column (3) correl. in stock returns = 2.08, t = 4.52)**.
- Interpretation and caveats:
  - Trade variables remain robust after informational proxies are included.
  - Imports coefficient below theoretical unity; possible reasons: measurement error in imports, alternative hedging channels, index composition mismatch, commodity composition.
  - Distance appears particularly predictive of zero versus nonzero investment (incidence) in some samples.

### Empirical findings: determinants of aggregate portfolio equity assets
- Dependent-variable approaches: source fixed effect from bilateral regressions and measured stock (max of CPIS and IIP).
- Key correlates (Table 8):
  - Log GDP per capita: positive and significant.
    - Column (1) fixed effect: log GDP per capita coefficient 0.58 (t = 2.90)**.
    - Column (4) Max(CPIS,IIP): coefficient 1.35 (t = 4.35)**.
  - Log domestic stock market capitalization: elasticity around 0.5.
    - Column (1) 0.56 (t = 3.20)**; Column (4) 0.58 (t = 2.11)*.
  - Volatility measures: SD of “beta residual” (1980–96) negative in some specs.
    - Column (3) = -6.48 (t = 1.94)†; Column (6) = -11.49 (t = 2.18)* — suggesting higher volatility reduces outward equity investment.
  - Capital controls index: negatively correlated in bivariate sense but loses significance when controlling for GDP per capita.

### Empirical findings: determinants of aggregate portfolio equity liabilities
- Dependent-variable approaches: host fixed effect and measured liabilities (max of CPIS derived liabilities and IIP).
- Key correlates (Table 9):
  - Log domestic stock market capitalization: robust positive correlate with elasticity below unity.
    - Column (4) Max(CPIS,IIP) coefficient 0.71 (t = 5.20)**.
    - Fixed-effect regressions show coefficients 0.46–0.53 (t-statistics 4.93–5.18)**.
  - Financial center dummy: highly significant positive effect.
    - Fixed-effect column (1) 0.63 (t = 2.20)*.
    - Max(CPIS,IIP) columns (5)-(6) coefficients 0.75 (t = 2.23)* and 0.82 (t = 2.35)*.
  - Middle East dummy: highly significant negative coefficients in some specs (e.g., column (4) -1.73, t = 4.53)** — reflects underestimation of liabilities in CPIS due to nonparticipation by large regional holders.
  - Log GDP per capita: significant in some specifications (columns (4)-(6)) but not in fixed-effect regressions.
  - Capital controls index, Sharpe ratio, beta, and volatility measures: generally insignificant in multivariate liability regressions.
- Foreign ownership share:
  - Stock market capitalization coefficients imply the log share of domestic stock market owned by foreigners is well explained (last-row fit reported as R2 ~ 0.44–0.54 for share regressions in Table 9).

### Conclusions and directions for future research
- Main conclusions:
  - Bilateral portfolio equity holdings are strongly correlated with bilateral trade (imports) and informational/cultural proximity (common language, legal origin); trade is a key determinant consistent with the theoretical benchmark.
  - Aggregate outward equity positions increase with GDP per capita and domestic stock market size; higher volatility reduces outward holdings.
  - Host attractiveness for liabilities is driven strongly by domestic stock market size and financial-center status.
- Important caveats: CPIS measurement limitations (incomplete coverage, under-reporting, third-party holdings) and the special role of offshore/financial centers.
- Suggested future work:
  - Allocate, even approximately, equity invested in offshore centers to ultimate destinations.
  - Use bilateral position asymmetries to compute differential wealth effects of shocks and implications for bilateral exchange-rate responses and risk-sharing.
  - Comparative analysis across asset classes (portfolio debt, bank loans and deposits, FDI).
  - Explore whether financial remoteness or investor-base composition affects aggregate investment, cost of capital, and stability of capital flows.

*Source: _wp04134 - 1. Largest Holders of Portfolio Equity Assets (IMF working paper content).*

### 1. Largest Holders of Portfolio Equity Assets......................................................................... 3

### _wp04134 - 1. Largest Holders of Portfolio Equity Assets......................................................................... 3

### Main sections (section titles and page numbers)
- 1. Largest Holders of Portfolio Equity Assets ......................................................................... 30
- 2. Largest Holders of Portfolio Equity Liabilities ................................................................... 30
- 3. Summary Statistics on Stock Market Size and Foreign Ownership (2001) ........................ 31
- 4. Foreign Portfolio Equity Investment: Actual and Predicted Shares.................................... 31
- 5. Bilateral Portfolio Equity Holdings, All Countries: Panel Regressions with Fixed Source and Host Effects ...................................................................................................... 32
- 6. Bilateral Portfolio Equity Holdings, OECD Source Countries: Panel Regressions with Fixed Source and Host Effects .................................................................................... 33
- 7. Bilateral Portfolio Equity Holdings, Emerging Market Source Countries: Panel Regressions with Fixed Source and Host Effects................................................................ 34
- 8. Determinants of Aggregate Portfolio Equity Assets (2001 end-of-period stock) ............... 35
- 9. Determinants of Aggregate Portfolio Equity Liabilities (2001 end-of-period stock) ......... 36

### Appendices and supplementary tables
- A1. Sample Size, Number of Observations=0, and Total Equity Holdings ............................ 37
- A2. Portfolio Equity Investment: The Role of Financial and Offshore Centers ..................... 38

### Figures (list and page numbers)
- Figure 1. Degree of Foreign Equity Bias ............................................................................................ 39
- Figure 2. Scatter of 2001 Equity Holdings versus 2001 Imports........................................................40

*Source: _wp04134 - 1. Largest Holders of Portfolio Equity Assets (PDF chapter/section).*

### References .............................................................................................................

### _wp04134 - References .............................................................................................................

### Introduction: motivation and scope
- Purpose: understand the geography of international portfolio equity investment to reveal frictions in global economic integration and explain bilateral asymmetries in portfolios.
- Data: geographical breakdown of international portfolio holdings at end-2001 by 67 source countries and up to 218 destination countries/territories.
- Key conceptual points:
  - Bilateral portfolio asymmetries affect cost of equity capital, wealth exposures to country-specific shocks, risk-sharing, and comovements in macro/financial variables.
  - Research questions include roles of product-market frictions, cultural/informational factors, and domestic vs. international financial development.

### Theoretical framework
- Approaches summarized:
  - Obstfeld and Rogoff (2001): product-market (trade) frictions generate home bias even with complete financial markets; extended to N-country model here.
  - Martin and Rey (2003): transactional frictions in asset markets (iceberg costs, endogenous market capitalization) and informational frictions.
  - Davis, Nalewaik, and Willen (2001): incomplete markets with country-specific labor-income risk; cross-country heterogeneity in holdings due to differing correlations between assets and labor income.
- N-country Obstfeld–Rogoff generalization (benchmark):
  - Country j’s share of country i’s equity is decreasing in bilateral trading cost τij relative to i’s average trading costs with others and increasing in relative preference for good i (ωij).
  - Theoretical allocation link between bilateral equity holdings x∗ij and imports: log(x∗ij) corresponds one-to-one to log(IMPij) (equation (12) log(x∗ij)=log(IMPij)).
  - Full empirical specification allows deviations from benchmark via aggregate and bilateral financial frictions: log(xij) = log(IMPij) + Fij + source/host fixed effects + error (equation (14)).

### Empirical strategy
- Observation: single cross-sectional year, end-2001.
- Aim: isolate contributions of (i) bilateral factors (trade, information, financial frictions), (ii) source-country factors, and (iii) host-country factors.
- Bilateral regressors include: log(1+imports), log distance, time zone difference, common language dummy, colonial relationship dummy, currency union dummy, investment tax treaty dummy, common legal origin dummy, correlations (stock returns, GDP growth, growth vs. host stock returns).
- Source and host fixed effects used to capture aggregate country characteristics (size, GDP per capita, stock market capitalization, capital controls, financial center status, historical Sharpe ratio, beta, volatility).
- Estimation methods: panel OLS with fixed effects, IV (to address potential endogeneity of imports and correlations), and Tobit (to account for many zero bilateral observations). Dependent variable: log(1+equity).

### Data: description and key measurement issues
- CPIS coverage: end-1997 and end-2001; 2001 CPIS includes 67 source countries (expanded coverage vs. 1997).
- Total recorded portfolio equity investment in CPIS (end-2001): US$5.16 trillion.
- Measurement issues:
  - Incomplete country coverage (notable non-participants include British Virgin Islands, China, Kuwait, Saudi Arabia, Taiwan Province of China, UAE).
  - Under-reporting by participants due to incomplete institutional coverage (examples: Cayman Islands, Bahamas, Germany).
  - Third-party holdings and custodial reporting differences (U.S. uses mix of methods).
  - Collection method variability for first-time participants.
- Persistence: correlation of bilateral positions for end-2001 and end-2002 is 0.99 for pairs with data for both years (preliminary end-2002 data).
- Offshore/financial center role:
  - Offshore centers and small financial centers reported assets > US$700 billion (incomplete coverage caveat).
  - Derived equity liabilities to offshore and small financial centers: US$870 billion.

### Stylized facts (aggregate and bilateral)
- Aggregate:
  - Total CPIS equity assets = US$5.16 trillion.
  - Largest asset holders (absolute): United States US$1,613 billion; United Kingdom US$558 billion; Germany US$381 billion; Luxembourg US$319 billion; Switzerland US$247 billion.
  - Largest asset holders (ratio to GDP): Luxembourg 16.6; Jersey 14.8; Guernsey 13.5; Isle of Man 9.6; Bermuda 8.6 (table 1 exact entries preserved).
- Derived liabilities (top destinations): United States 1000 (1533) US$ billion; United Kingdom 711 (768); France 387 (416); Luxembourg 376 (N.A.); Japan 330 (376) (table 2; bracketed numbers are IIP-reported liabilities).
- Stock market structure (Table 3 highlights):
  - World shares: United States domestic stock market cap 48.9 percent of world; Euro area 15.9; Japan 9.3; United Kingdom 8.9.
  - Percent of domestic stock market owned by foreign portfolio investors: Euro area 36.5 percent; United Kingdom 35.6 percent; United States 12.9 percent; Japan 16.7 percent.
- Allocation vs. benchmark (Table 4):
  - Example deviations: United Kingdom invests much more in the euro area than the United States; Japan’s foreign equity investment aligns most closely with market-cap benchmark.

### Empirical findings: bilateral equity holdings
- Main robust result: strong positive relation between bilateral imports and bilateral equity holdings.
  - Table 5 (All countries): OLS panel estimate for Avg. imports, 1997–2001 = 0.32 (t = 17.93)**; IV and Tobit show variations (Tobit column (5) coefficient 0.52, t = 5.13)**; when additional controls added coefficient ~0.22 (column (2), t = 8.42)**; narrower sample with stock-market variables column (3) coefficient 0.21 (t = 3.58)**.
  - Tobit estimates typically larger: e.g., Table 5 Tobit (1) 0.73 (t = 21.82)** in simpler specification, reflecting censoring impact.
  - Simple pooled regression reported in text: elasticity 0.39 (t-statistic 22.2), R-squared 0.18 (scatter Figure 2).
- Informational / cultural proximity significant:
  - Distance: doubling physical distance reduces equity holdings by 61 percent in broad OLS (Table 5 column (2) log distance = -0.61, t = 9.17)**; distance effect weakens or loses significance in narrower samples with stock-market controls.
  - Common language: raises equity holdings by about 40 percent (Table 5 column (2) coefficient 0.33, t = 3.63)**.
  - Colony dummy: especially large for emerging-market sources (Table 7 column (2) colony = 1.55, t = 2.98)** (interpretable as large multiplicative effect).
  - Currency union dummy: large and very significant in full-sample OLS in some specifications (Table 5 column (2) 1.33, t = 7.64)**; stronger effect in OECD subsample when stock-market variables included (Table 6 column (3) currency union = 0.61, t = 2.69)**.
  - Common legal origin: countries invest 30–40 percent more in hosts with similar legal background (Table 5 column (3) = 0.28, t = 2.93)**.
- Financial correlations and diversification proxies:
  - Correlation in GDP growth rates often positive and significant for full sample and OECD (counter to simple diversification prediction): e.g., Table 5 column (2) correl. in growth rates = 0.61, t = 3.66)**.
  - Correlation in stock returns positive and significant in some specifications (Table 5 column (3) correl. in stock returns = 2.08, t = 4.52)**.
  - For emerging-market sources, correlation and some distance/time-difference effects vary; sample reduction affects significance.
- Interpretation:
  - Trade variables remain robust after controlling for gravity-type informational proxies.
  - The imports coefficient is below theoretical value of unity; possible reasons: measurement error in imports, alternate hedging channels (investing in multinational firms at home), index composition mismatch, commodity composition.
  - Distance appears to help predict whether investment occurs at all (zero vs. nonzero) more than investment size in some samples.

### Empirical findings: aggregate asset positions
- Dependent variable approaches: source fixed effect (from bilateral regressions) and actual stock (max of CPIS and IIP).
- Key correlates (Table 8):
  - Log GDP per capita: positive and significant. Examples:
    - Column (1) fixed effect: log GDP per capita coefficient 0.58 (t = 2.90)**.
    - Column (4) Max(CPIS,IIP): coefficient 1.35 (t = 4.35)**.
  - Log domestic stock market capitalization: elasticity around 0.5. Examples:
    - Column (1) 0.56 (t = 3.20)**; Column (4) 0.58 (t = 2.11)*.
  - Beta generally insignificant; SD of beta residual (volatility) negative and, in some specs, significant: e.g., SD of “beta residual” (1980–96) in column (3) = -6.48 (t = 1.94)† and column (6) = -11.49 (t = 2.18)* — suggests higher volatility reduces outward equity investment.
  - Capital controls index negatively correlated in bivariate sense but loses significance when controlling for GDP per capita (collinearity).
- Conclusion: level of development and depth of domestic financial market (stock market size, lower volatility) are associated with larger gross foreign equity positions.

### Empirical findings: aggregate liability positions
- Dependent variable: host fixed effect and measured liabilities (max of CPIS derived liabilities and IIP).
- Key correlates (Table 9):
  - Log domestic stock market capitalization: robust positive correlate; elasticity well below unity. Example: Column (4) Max(CPIS,IIP) coefficient 0.71 (t = 5.20)**; fixed-effect regressions show 0.46–0.53 (t-statistics 4.93–5.18)**.
  - Financial center dummy: highly significant positive effect. Example: fixed-effect column (1) 0.63 (t = 2.20)*; Max(CPIS,IIP) columns (5)-(6) coefficients 0.75 (t = 2.23)* and 0.82 (t = 2.35)*.
  - Middle East dummy: highly significant negative coefficients in some specifications (e.g., column (4) -1.73, t = 4.53)** — reflects underestimation of liabilities in CPIS due to large regional holders not participating.
  - Log GDP per capita significant in some specifications (columns (4)-(6)), but not in fixed-effect regressions.
  - Capital controls index negatively correlated but statistically insignificant in multivariate regressions.
  - Sharpe ratio, beta, and volatility measures generally insignificant for liabilities.
- Result on foreign ownership share: regressions (noted in text) show stock market capitalization coefficients imply the log share of domestic stock market owned by foreigners is well explained (last-row fit reported as R2 ~ 0.44–0.54 for share regressions in Table 9).

### Conclusions and directions for future research
- Main conclusions:
  - Bilateral portfolio equity holdings are strongly correlated with bilateral trade (imports) and informational/cultural proximity (common language, legal origin); trade emerges as a key determinant consistent with the theoretical benchmark.
  - Aggregate outward equity positions increase with GDP per capita and domestic stock market size; volatility reduces outward holdings.
  - Host attractiveness (liabilities) is driven strongly by domestic stock market size and financial-center status.
- Important caveats: CPIS measurement limitations (incomplete coverage, under-reporting, third-party holdings) and the special role of offshore/financial centers.
- Suggested future work:
  - Allocate, even approximately, equity invested in offshore centers to ultimate destinations.
  - Use bilateral position asymmetries to compute differential wealth effects of shocks and implications for bilateral exchange-rate responses and risk-sharing.
  - Comparative analysis across asset classes (portfolio debt, bank loans and deposits, FDI).
  - Explore whether financial remoteness or investor-base composition affects aggregate investment, cost of capital, and stability of capital flows.

*Source: _wp04134 - References (IMF Working Paper content: introduction, theoretical framework, empirical strategy, data, empirical results, conclusions, tables, and references as supplied).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp04134.pdf_