## _wp1104

## Source details

**Canonical URL:** [_wp1104](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1104.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1104.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1104.pdf.json)

---

### 3.1 Portfolio Holdings: Summary Statistics — Data, Scope, Objectives
- Data and coverage:
  - Coordinated Portfolio Investment Survey (CPIS) data for 2001 to 2007 covering investors from 75 countries.
  - Equity and debt security holdings summarized in Table 4 and Table 5; bilateral exposures toward U.S. and Japan in Table 6.
- Objectives:
  - Examine investor preferences for domestic, regional, and global assets.
  - Test home bias and regional bias in Asia relative to Latin America, Eastern Europe, and a group of industrialized countries.
  - Complement measures of integration based on convergence in equity premia and interest rates.

### 3.1 Portfolio Holdings: Key Empirical Findings (summary)
- Asia appears slightly more “home biased” than other regions with respect to equity holdings.
- Regional tendency likely reflects strong country-level home bias toward domestic stock markets rather than explicit preference for other Asian assets.
- CPIS analysis covers equity and debt securities and bilateral holdings; summary statistics and bilateral patterns presented in referenced tables.

### 3.1 Portfolio Holdings: Selected magnitude statistics (2007 year-end; equity and debt holdings in mill US$)
- Equity holdings — Size of holdings:
  - Japan: 479,145.3
  - Korea, Republic of: 79,073.8
  - Hong Kong, China: 105,047.0
  - Singapore: 92,625.6
  - Malaysia: 47,150.8
  - Thailand: 2,209.7
  - Philippines: 115.4
  - Indonesia: 507.4
  - India: 236.7
  - Pakistan: 25.0
- Equity within-Asia investment shares:
  - Simple average within-Asia: 0.3845
  - Weighted average within-Asia: 0.2139
  - Industrialized countries weighted average within-group: 0.7508
  - Latin America weighted average: 0.0123
  - Eastern Europe weighted average: 0.0168
- Debt holdings — Size of holdings:
  - Japan: 1,363,105.9
  - Korea, Republic of: 45,986.6
  - Hong Kong, China: 205,591.0
  - Singapore: 168,745.9
  - Malaysia: 2,006.7
  - Thailand: 10,659.2
  - Philippines: 4,880.9
  - Indonesia: 1,350.0
  - India: 5.7
  - Pakistan: 214.7
- Debt within-Asia investment shares:
  - Simple average within-Asia: 0.1506
  - Weighted average within-Asia: 0.0662
  - Industrialized countries weighted average (debt): 0.9372
  - Latin America weighted average (debt): 0.0678
  - Eastern Europe weighted average (debt): 0.0206
- U.S. and Japan exposure in Asian portfolios (ratios):
  - Equity toward United States — Average (Asia): 0.2598; Weighted average (Asia): 0.3705
  - Equity toward Japan — Average (Asia): 0.0241; Weighted average (Asia): 0.0252
  - Debt toward United States — Average (Asia): 0.3370; Weighted average (Asia): 0.3875
  - Debt toward Japan — Average (Asia): 0.0132; Weighted average (Asia): 0.0087
  - Country-specific notable values (Equity / Debt):
    - Philippines: United States 0.7682 / 0.4671; Japan 0.0019 / 0.0170
    - Japan: United States 0.4631 / 0.4339; Japan 0.00000 / 0.0000
    - Korea, Republic of: United States 0.3175 / 0.6010; Japan 0.0555 / 0.0107
    - Hong Kong, China: United States 0.1745 / 0.2540; Japan 0.0851 / 0.0597
- Text summary: The US looms large in Asian portfolio allocation relative to Japan; for equity holdings, only Hong Kong, China, Singapore and Korea have >5% of portfolios in Japanese assets; for debt only Hong Kong, China’s holdings exceed 5%.

---

### II. Convergence in Interest Rates and Equity Premia — Methodology and Descriptive Findings
- Methodology:
  - Construct σ(t), cross-country standard deviation of equity premia, using monthly equity prices and short-term government yield data for 52 countries (Table 1).
  - Equity premium = annualized rate of change in equity prices − annualized short-term government yield (one to three months).
  - Sample period: January 1984 to March 2009 (countries enter sample at different points).
  - Regression: σ(t) = A0 + A1 σ(t-1) + A2 σ(t-2) + A3 Z(t) + A4 TREND where Z(t) includes VIX, CRISIS (Laeven and Valencia (2008)), and cross-section MEAN.
- Descriptive findings:
  - Dispersion in equity premia averaged about 75 percent over full period with sharp spikes.
  - No overall linear downward trend; marked decrease in dispersion during 2002 to mid-2007 reversed by the financial crisis.
  - Within-group patterns:
    - Asia: after the late-1990s crisis, regional convergence accelerated; pace almost as fast as any other region.
    - Eastern Europe: converged faster than any other group pre-crisis.
  - Dispersion in interest rates (all economies):
    - Averaged 5% over 1984-94; elevated during emerging market crises; since 2002 declined below pre-crisis average.
    - Industrialized countries: decline from about 5% to about 1%.
    - Asia: averaged about 4%; declined to about 2% since 2002 before rising during the crisis.
    - Latin America: decline from about 16% to 4%.
    - Eastern Europe: decline from 35% to 1.5% before increasing with the crisis.

### II. Convergence: Selected regression results and implications
- Dispersion in Equity Premia (Table 2) — selected coefficients (dependent: Log of Dispersion in Equity Premia):
  - All countries (column (1)):
    - Log Dispersion in EqPrem(t-1): 0.186*** (0.058)
    - Log Dispersion in EqPrem(t-2): 0.105* (0.057)
    - VIX: 0.0116*** (0.0025)
    - Cross-section Mean Equity Premium: -0.000462 (0.00028)
    - Crisis: 0.0242*** (0.0052)
    - Time trend: -0.0000806 (0.00022)
    - Constant: 2.620*** (0.27)
    - Observations: 279; R-squared: 0.41; Sample Period: Jan. 1986 to Mar. 2009
  - Asia (column (3)):
    - Log Dispersion in EqPrem(t-1): 0.194** (0.081)
    - VIX: 0.0162*** (0.0042)
    - Crisis: 0.0371** (0.015)
    - Time trend: -0.00231*** (0.00080)
    - Constant: 2.912*** (0.41)
    - Observations: 159; R-squared: 0.44; Sample Period: Jan. 1996 to Mar. 2009
  - Time trend negative indicates convergence; fastest estimated trend decline in Eastern Europe, then Asia, Industrialized, Latin America.
- Dispersion in Interest Rates (Table 3) — selected coefficients (dependent: Log of Dispersion in Interest Rates):
  - All countries (column (1)):
    - Log Dispersion in Int Rates (t-1): 0.992*** (0.060)
    - Log Dispersion in Int Rates (t-2): -0.120** (0.057)
    - VIX: 0.00153 (0.00095)
    - Cross-section Mean Interest Rate: 0.0383*** (0.0058)
    - Crisis: -0.00465*** (0.0017)
    - Time trend: 0.00126** (0.00050)
    - Time trend squared: -0.00000328* (0.0000019)
    - Constant: -0.159*** (0.050)
    - Observations: 279; R-squared: 0.97; Sample Period: Jan. 1986 to Mar. 2009
  - Asia (column (3)):
    - Log Dispersion in Int Rates (t-1): 0.347*** (0.069)
    - VIX: 0.00572*** (0.0013)
    - Cross-section Mean Interest Rate: 0.101*** (0.011)
    - Crisis: -0.0173*** (0.0064)
    - Time trend: 0.00650*** (0.0013)
    - Constant: -0.353*** (0.100)
    - Observations: 159; R-squared: 0.94; Sample Period: Jan. 1996 to Mar. 2009
- Interpretations:
  - VIX is consistently significant for equity premia dispersion; CRISIS significant for All countries and Asia samples.
  - Cross-section mean interest rate strongly predicts dispersion in interest rates.
  - A quadratic time trend improves fit for interest rate dispersion regressions.

---

### 3.2 Portfolio Holdings: Gravity Model Estimates — Specification and Main Findings
- Gravity model:
  - Dependent variable: bilateral investment holding (pooled 2001–2007).
  - Time-varying regressors: product of real GDP (investor × issuer), product of real per capita GDP (investor × issuer).
  - Time-invariant regressors: log distance, Common border, Island, Regional Trade Agreement, bilateral trade 1998 (log).
  - Additional regressor: Real Exchange Rate Volatility (log).
- Main gravity-estimation findings (Table 7):
  - Product of GDPs (log) coefficients:
    - Equity (col (1)): 0.488*** (0.11)
    - Equity with region dummies (col (2)): 0.490*** (0.11)
    - Bonds (col (3)): 0.462*** (0.083)
    - Bonds with region dummies (col (4)): 0.430*** (0.091)
  - Product of per capita GDP (log):
    - Equity: 0.904*** (0.11); 0.772*** (0.14)
    - Bonds: 0.679*** (0.072); 0.517*** (0.11)
  - Distance (log):
    - Equity: -0.187 (0.17); 0.146 (0.15)
    - Bonds: -0.379*** (0.12); -0.256* (0.13)
  - Bilateral Trade (log):
    - Equity: 0.510*** (0.099); 0.504*** (0.10)
    - Bonds: 0.328*** (0.072); 0.377*** (0.071)
  - Real Exchange Rate Volatility (log):
    - Equity: -0.661** (0.30); -0.294 (0.27)
    - Bonds: -0.890*** (0.22); -0.891*** (0.21)
  - Sample size and fit:
    - Observations: 10815 (equity specs), 11545 (bond specs)
    - R-square overall: 0.571, 0.597 (equity); 0.596, 0.607 (bonds)
- Regional dummy findings:
  - Asia dummies (Asia_Single, Asia_Pair) insignificant in both equity and debt equations (cols (2) and (4)).
  - Industrialized_Pair positive and significant for equities and bonds (0.761* (0.44) equity; 0.979** (0.43) bonds).
  - LatAm_Single negative and significant in equity (−1.523*** (0.55)).
  - EEurope_Pair positive and significant in equity (1.557** (0.78)).

### Group-specific gravity regressions — Selected coefficients (Tables 8 and 9)
- Equity regressions (Table 8) — Product of GDPs (log):
  - Asia: 0.149 (0.22)
  - Industrialized: 0.642*** (0.089)
  - Latin America: 0.628* (0.33)
  - E. Europe: 0.269 (0.21)
- Equity regressions — Product of per capita GDP (log):
  - Asia: 1.027*** (0.14)
  - Industrialized: 0.847*** (0.078)
- Equity regressions — Distance (log):
  - Asia: -0.664*** (0.25)
- Debt regressions (Table 9) — Product of GDPs (log):
  - Asia: 0.242 (0.19)
  - Industrialized: 0.760*** (0.10)
  - Latin America: 0.283** (0.12)
  - E. Europe: 0.349*** (0.12)
- Debt regressions — Product of per capita GDP (log):
  - Asia: 0.810*** (0.18)
  - Industrialized: 0.597*** (0.11)
- Debt regressions — Distance (log):
  - Industrialized: -0.626*** (0.14)
  - Latin America: -0.839*** (0.24)
- Notes on reported observations and fit (as printed in source):
  - Equity R-square: Asia 0.603; Industrialized 0.709; Latin America 0.423; E. Europe 0.302
  - Debt R-square: Asia 0.607; Industrialized 0.677; Latin America 0.312; E. Europe 0.314
  - Some printed observation counts in source appear anomalous (as printed).

### Gravity-model interpretation and suggested further research
- Product of per capita GDPs and bilateral trade are consistently significant determinants of bilateral holdings across many subgroup regressions.
- Distance matters more for debt than equity in many specifications.
- Suggested further research for Asia:
  - Role of foreign direct investment.
  - Effects of presence of financial centers.
  - Impact of China‘s gravitational pull.
  - Role of the overseas Chinese community.

---

### Home Bias — Definition, Statistics, and Determinants
- Definitions:
  - Home bias for country i: HB_i = 1 - (w_i / w_i*)
  - Bilateral home bias for i toward j: HB_ij = 1 - (w_ij / w_j*)
  - Interpretation: Full international diversification → HB = 0; No holdings of j by i → HB = 1.
- Home bias summary (Table 10) — Equity (Panel A) selected entries (Average 2001-2007):
  - Hong Kong, China: Industrialized 0.933; Asia 0.896; Latin America 1.000; Eastern Europe 1.000; Other 1.000
  - India: Industrialized 1.000; Asia 1.000; Latin America 0.999; Eastern Europe 1.000; Other 1.000
  - Singapore: Industrialized 0.754; Asia 0.100; Latin America 0.952; Eastern Europe 0.977; Other 0.926
  - Asia Average: Industrialized 0.926; Asia 0.841; Latin America 0.993; Eastern Europe 0.997; Other 0.943
  - Industrialized Average: Industrialized 0.617; Asia 0.782; Latin America 0.963; Eastern Europe 0.950; Other 0.941
- Home bias — Debt (Panel B) selected entries (averages; Hong Kong, China and Singapore averages are 2001-06):
  - Hong Kong, China (*): Industrialized 0.613; Asia 0.347; Latin America 0.884; Eastern Europe 0.828; Other 0.779
  - Singapore (*): Industrialized 0.649; Asia 0.918; Latin America 0.982; Eastern Europe 0.965; Other 0.888
  - Asia Average (*): Industrialized 0.768; Asia 0.791; Latin America 0.921; Eastern Europe 0.915; Other 0.944
  - Industrialized Average: Industrialized 0.512; Asia 0.956; Latin America 0.965; Eastern Europe 0.958; Other 0.984
- Interpretation:
  - Asian economies tend to be very home-biased toward the industrialized group in equities (exceptions: Singapore and the Philippines).
  - Most Asian economies are slightly less home-biased toward other Asian economies than toward industrialized ones (exceptions: Japan and the Philippines).
  - Debt home bias tends to be below equity home bias; Asia is the second least home-biased region in bond holdings (0.791), after Industrialized (0.512).

### Determinants of average bilateral home bias (Table 11) — Selected coefficients (average 2001–2007)
- Product of GDPs (log):
  - Equities (col (1)): 0.0341*** (0.010)
  - Bonds (col (3)): 0.0781*** (0.013)
- Product of per capita GDP (log):
  - Equities: -0.0669*** (0.0072)
  - Bonds: -0.0829*** (0.0092)
- Distance (log):
  - Equities: 0.0865*** (0.015)
  - Bonds: 0.0766*** (0.020)
- Common Language:
  - Equities: -0.0757** (0.034)
  - Bonds: -0.0874** (0.042)
- Bilateral Trade (log):
  - Equities: -0.0293*** (0.0090)
  - Bonds: -0.0452*** (0.012)
- Real Exchange Rate Volatility (Log):
  - Equities: 0.0902*** (0.022)
  - Bonds: 0.136*** (0.027) — effect larger for bonds than equities
- Regional dummies (selected):
  - Asia_Single — Equities (col (2)): 0.0853*** (0.033); Bonds (col (4)): 0.193*** (0.041)
  - Asia_Pair — Equities (col (2)): 0.138** (0.057); Bonds (col (4)): -0.0658 (0.086)
  - Industrialized_Pair — Bonds (col (4)): 0.125** (0.060)
  - EEurope_Pair — Equities (col (2)): -0.204*** (0.077); Bonds (col (4)): -0.257*** (0.092)
- Model fit and observations:
  - Observations: 2840 (equities); 2871 (bonds)
  - F-Statistics (Prob>F): equities 50.74 (.000); equities+region 29.30 (.000); bonds 64.45 (.000); bonds+region 38.18 (.000)
- Interpretation:
  - Real exchange rate volatility raises home bias, larger effect for bonds.
  - Bilateral equity home bias is higher if either or both economies are in Asia.
  - Bond home bias increases when one economy in the pair is in Asia (Asia_Single), but not when both are (Asia_Pair insign./negative).

---

### IV. Risk-sharing and Financial Stability — Theoretical expectations and evidence
- Theoretical expectations:
  - Financial integration should raise cross-country consumption correlations and lower cross-country output correlations; should reduce correlation between domestic saving and domestic investment.
- Literature overview:
  - Mixed evidence: Imbs (2006), Kose, Prasad and Terrones (2007), Bekaert, Harvey and Lundblad (2006) yield heterogeneous conclusions on whether financial integration materially raises risk sharing, with benefits concentrated in industrialized countries in some studies.
- Empirical evidence in this chapter:
  - Figure 3 (1971-2007) indicates a decline in the dispersion of consumption growth — some evidence of increased risk sharing — but also a decline in dispersion in output growth.
  - Regression evidence (Table 12) — selected findings:
    - Panel A (Standard Deviation of GDP Growth): Time trend negative and significant in All countries and subgroups (e.g., Time trend: -0.0236*** (0.0085) All countries).
    - Panel B (Standard Deviation of Consumption Growth): Time trend negative and significant across groups (e.g., Time trend: -0.0453*** (0.012) All countries).
    - Per Capita World Consumption Growth has negative association with cross-country consumption dispersion in some samples (e.g., -0.318** (0.13) All countries).
  - Interpretation: Time trends indicate declines over time in cross-country dispersions of GDP and consumption growth; but convergence in GDP growth as well as consumption growth complicates standard risk-sharing interpretations.

---

### 4.2 Financial Integration and Crises — Stability, investor behavior, and liabilities
- Potential benefits of local/regional market development:
  - Higher scale and liquidity; richer instrument choice; ability to issue long-term local-currency debt.
- Debate on whether regional/local investor base reduces crisis likelihood:
  - Arguments for reduced crisis likelihood include regional bias and familiarity of local investors plus more long-term local-currency instruments.
  - Counterarguments emphasize weak corporate governance, limited informational advantage of local investors, and deeper structural drivers of portfolio structure.
  - Key empirical question: are global investors more prone to panics than local/regional investors?
- Single-currency market as stabilizing channel:
  - Evidence from European bond markets after the euro: borrowing costs fell for countries with previously high spreads; euro-denominated bond class broadened investor base and liquidity.
  - Implication: benefits from Asian financial integration would be enhanced by a currency union.
- Evidence from past crises and investor behavior:
  - Mexico (1994): domestic investors turned pessimistic before international investors; domestic funds’ NAVs fell sharply pre-devaluation.
  - Korea (1997): foreign investors followed “positive feedback strategies” but trades did not have lasting price impact in late 1997.
  - International mutual funds (1996–1999): tendency to herd small (7% more concentrated); funds cut exposure in the month before crises, extent modest.
  - Equity market liberalization did not increase instability (Bekaert, Harvey and Lundblad (2004)).
  - Bank lending: Japanese banks reduced East Asian loan exposure by about 30% in 1998-99 vs. UK and US banks 13% and 17%; Japanese loans failed to recover to pre-crisis levels nearly 10 years later.
- Investor behavior during the subprime crisis: Evidence from Brazil
  - Foreign holdings in Brazil:
    - Below $50 billion during previous decade; surpassed $250 billion by mid-2008.
    - 2007 net inflows < $34 billion; asset holdings of foreign investors increased by over $112 billion (capital inflows + price appreciation).
    - International investors’ holdings ~80% stocks in 2007.
    - Stock market gained over 40% in 2007; the real appreciated by 16% in 2007.
  - Crisis-period adjustments (September-November 2008):
    - Net outflows amounted to about $13 billion; asset holdings declined by over $116 billion.
    - Compound loss of 55% of the stock market index (the Bovespa) in US dollar terms.
    - Gross inflows about $25 billion during September-November (gross flows in both directions).
  - Liability-structure implications:
    - “Original sin” world: foreign-currency debt → sudden stops cause large depreciations and solvency stress.
    - “New liability” world: local-currency debt and large foreign-held local equity → depreciation spreads losses to foreign investors and reduces dollar value of liabilities (improves solvency) but allows immediate security sales producing liquidity and exchange-rate pressures.
  - Additional Brazil observations:
    - Outflows by Brazilian investors and firms ~ just under $6 billion in September-November (excludes foreign currency facilities and swaps from central bank).
    - Losses by firms on derivative contracts tied to currency movements noted.
  - Preliminary conclusion: Brazil case does not suggest a clear-cut distinction between foreign and local investors’ impact on local financial markets.

---

### Cross-border holdings, home bias, capital controls, and policy implications — Bottom-line assessment
- Cross-border holdings:
  - Magnitude and pattern: cross-border holdings among Asian countries are significant and increased over the prior decade, but pattern not strikingly different from Latin America and Eastern Europe and below industrialized benchmarks.
  - U.S. assets dominate Asian portfolios relative to Japanese assets.
- Gravity and home bias:
  - Gravity model describes cross-border equity and bond holdings well; no evidence Asia has escaped gravity effects.
  - Real exchange rate volatility is an important determinant of home bias; it raises home bias (effect larger for bonds).
  - Product of per capita GDPs reduces home bias across country groups and instrument types.
  - Little evidence that Asia no longer conforms to standard home-bias models.
- Capital controls:
  - An index measure of capital controls (1995-2005 averages) yields no systematic effect of controls on asset holdings in main regressions.
  - In home bias regressions, capital controls increased home bias for bond holdings in all-countries and several regional samples — but not for Asia.
- Prices, quantities, and convergence:
  - Convergence evidence: equity premia and interest rate dispersions show episodic convergence with crises interrupting the process; Asia experienced convergence but not faster than other regions.
  - Consumption and GDP growth dispersions fell over time, indicative of increased convergence; no evidence risk-sharing improvements are faster within Asia than other groups.
- Local investor base and stability:
  - Surveyed evidence invites caution about claims that local investors inherently provide greater stability.
  - Case studies and empirical results show mixed outcomes; foreign and local investors both can contribute to market stress during crises.
- Policy implication:
  - Asian regional financial integration is increasing but Asian economies remain more integrated with global financial markets than with regional neighbors.
  - Stronger ties to global markets are not necessarily undesirable; gains from global diversification may exceed those from purely regional integration.
  - Policymakers should weigh benefits of deeper regional integration against the predominance and potential advantages of global diversification.

*Source: IMF Working Paper content unit _wp1104 (sections and tables as presented in the supplied text).*

### 3.1 Portfolio Holdings: Summary Statistics................................

### 3.1 Portfolio Holdings: Summary Statistics

### Data and scope
- Portfolio holdings analysis uses Coordinated Portfolio Investment Survey (CPIS) data for 2001 to 2007.
- CPIS provides estimates of international portfolio holdings by investors from 75 countries.
- Summary statistics for equity and debt security holdings are reported in Table 4 and Table 5; bilateral exposures toward major markets (U.S. and Japan) are reported in Table 6.

### Objectives of this section
- Directly examine international investors‘ preferences for domestic, regional, and global assets.
- Test the extent of home bias and regional bias in Asia, benchmarking against Latin America, Eastern Europe, and a group of industrialized countries.
- Complement earlier sections that measured financial integration via convergence in equity premia and interest rates.

### Key empirical findings (as reported)
- Asia, as a region, appears slightly more “home biased” than other regions with respect to equity holdings.
- The observed regional tendency may reflect strong home bias at the country level toward domestic stock markets rather than an explicit preference for other Asian assets.
- The CPIS-based analysis covers both equity and debt securities and includes bilateral holdings; summary statistics and bilateral patterns are presented in the referenced tables.

### Interpretation within the paper’s framework
- The CPIS evidence is used to assess whether increasing financial integration is associated with reduced home bias and larger cross-border portfolio positions.
- Findings are interpreted alongside earlier measures of integration: convergence in equity premia and in interest rates (sections 2 and accompanying figures/tables).

*Source: 3.1 Portfolio Holdings: Summary Statistics (from the provided IMF content).*

### Section 2 Section 3 Section 4

### _wp1104 - Section 2 Section 3 Section 4

### II. CONVERGENCE IN INTEREST RATES AND EQUITY PREMIA
- Methodology
  - Construct σ(t), the cross-country standard deviation (dispersion) of equity premia, using monthly data on equity prices and the yield on short-term government securities for the 52 countries listed in Table 1.
  - Equity premium defined as annualized rate of change in equity prices minus annualized yield on government securities at maturities ranging from one to three months.
  - Sample period: January 1984 to March 2009 (some countries enter sample at different points).
  - Regression specification:
    - σ(t) = A0 + A1 σ(t-1) + A2 σ(t-2) + A3 Z(t) + A4 TREND
    - Z(t) includes VIX (implied volatility of S&P 500 index options), CRISIS (number of banking, currency and sovereign debt crises per Laeven and Valencia (2008)), and the cross-section MEAN of the dependent variable.

- Descriptive findings
  - Over the full period, dispersion in equity premia averaged about 75 percent, with sharp spikes.
  - No discernible linear downward trend overall, but a marked decrease in dispersion during 2002 to mid-2007 that was reversed by the onset of the financial crisis.
  - Within-group patterns:
    - Asia: after the late-1990s crisis, regional convergence accelerated; pace almost as fast as any other region.
    - Eastern Europe: converged to each other faster than any other group, though the present crisis disrupted convergence.
  - Dispersion in interest rates (top panel, all economies):
    - Averaged 5% over 1984-94.
    - Elevated during emerging market crises; since 2002 declined to below pre-crisis average.
    - Industrialized countries: decline from about 5% to about 1%.
    - Asia: averaged about 4%; declined to about 2% since 2002 before rising during the present crisis.
    - Latin America: decline from about 16% to 4%.
    - Eastern Europe: decline from 35% to 1.5% before increasing with the crisis.

### Empirical results — Dispersion in Equity Premia (Table 2)
- Main regression insights
  - VIX is significant and positively associated with dispersion in equity premia for All countries and all four regional groups.
  - Cross-section mean equity premium is not significant in any regression.
  - CRISIS is significant in the All countries and Asia samples [columns (1) and (3)] but not for other groups.
  - Time trend (A4) negative indicates convergence; estimated trend decline in dispersion is fastest in Eastern Europe, then Asia, Industrialized countries, and Latin America.
- Selected coefficient estimates and statistics (dependent variable: Log of Dispersion in Equity Premia)
  - All countries (column (1)):
    - Log Dispersion in EqPrem(t-1): 0.186*** (0.058)
    - Log Dispersion in EqPrem(t-2): 0.105* (0.057)
    - VIX: 0.0116*** (0.0025)
    - Cross-section Mean Equity Premium: -0.000462 (0.00028)
    - Crisis: 0.0242*** (0.0052)
    - Time trend: -0.0000806 (0.00022)
    - Constant: 2.620*** (0.27)
    - Observations: 279
    - R-squared: 0.41
    - Sample Period: Jan. 1986 to Mar. 2009
  - Industrialized (column (2)):
    - Log Dispersion in EqPrem(t-1): 0.153** (0.059)
    - VIX: 0.0182*** (0.0033)
    - Time trend: -0.00162*** (0.00028)
    - Constant: 2.742*** (0.27)
    - Observations: 279
    - R-squared: 0.39
    - Sample Period: Jan. 1986 to Mar. 2009
  - Asia (column (3)):
    - Log Dispersion in EqPrem(t-1): 0.194** (0.081)
    - VIX: 0.0162*** (0.0042)
    - Crisis: 0.0371** (0.015)
    - Time trend: -0.00231*** (0.00080)
    - Constant: 2.912*** (0.41)
    - Observations: 159
    - R-squared: 0.44
    - Sample Period: Jan. 1996 to Mar. 2009
  - Latin America (column (4)):
    - Log Dispersion in EqPrem(t-1): 0.0875 (0.080)
    - VIX: 0.0131** (0.0050)
    - Time trend: -0.00148* (0.00084)
    - Constant: 3.301*** (0.43)
    - Observations: 159
    - R-squared: 0.11
    - Sample Period: Jan. 1996 to Mar. 2009
  - E. Europe (column (5)):
    - Log Dispersion in EqPrem(t-1): 0.122 (0.079)
    - VIX: 0.0173*** (0.0049)
    - Time trend: -0.00278*** (0.00076)
    - Constant: 3.469*** (0.45)
    - Observations: 159
    - R-squared: 0.27
    - Sample Period: Jan. 1996 to Mar. 2009
- Note: *, ** and *** indicate significance at, or below the 10%, 5% and 1% levels respectively. Robust standard errors in parentheses.

### Empirical results — Dispersion in Interest Rates (Table 3)
- Main regression insights
  - VIX significant only for Asia and Eastern Europe groups.
  - Cross-section mean interest rate is very significant and positively associated with dispersion in all regressions.
  - A quadratic time trend provided a better fit than a linear trend.
  - Results suggest a drop in dispersion in Asia and Eastern Europe over the last decade; for industrialized countries and Latin America, no decline in dispersion once controlling for shifts in the cross-section mean.
- Selected coefficient estimates and statistics (dependent variable: Log of Dispersion in Interest Rates)
  - All countries (column (1)):
    - Log Dispersion in Int Rates (t-1): 0.992*** (0.060)
    - Log Dispersion in Int Rates (t-2): -0.120** (0.057)
    - VIX: 0.00153 (0.00095)
    - Cross-section Mean Interest Rate: 0.0383*** (0.0058)
    - Crisis: -0.00465*** (0.0017)
    - Time trend: 0.00126** (0.00050)
    - Time trend squared: -0.00000328* (0.0000019)
    - Constant: -0.159*** (0.050)
    - Observations: 279
    - R-squared: 0.97
    - Sample Period: Jan. 1986 to Mar. 2009
  - Industrialized (column (2)):
    - Log Dispersion in Int Rates (t-1): 0.842*** (0.060)
    - VIX: -0.000891 (0.00062)
    - Cross-section Mean Interest Rate: 0.0150*** (0.0045)
    - Crisis: -0.0187*** (0.0064)
    - Time trend: 0.000123 (0.00056)
    - Time trend squared: 0.0000000826 (0.0000014)
    - Constant: -0.0484 (0.077)
    - Observations: 279
    - R-squared: 0.98
    - Sample Period: Jan. 1986 to Mar. 2009
  - Asia (column (3)):
    - Log Dispersion in Int Rates (t-1): 0.347*** (0.069)
    - VIX: 0.00572*** (0.0013)
    - Cross-section Mean Interest Rate: 0.101*** (0.011)
    - Crisis: -0.0173*** (0.0064)
    - Time trend: 0.00650*** (0.0013)
    - Time trend squared: 0.0000000826 (0.0000068)
    - Constant: -0.353*** (0.100)
    - Observations: 159
    - R-squared: 0.94
    - Sample Period: Jan. 1996 to Mar. 2009
  - Latin America (column (4)):
    - Log Dispersion in Int Rates (t-1): 0.823*** (0.080)
    - VIX: 0.000787 (0.00099)
    - Cross-section Mean Interest Rate: 0.00992** (0.0045)
    - Crisis: 0.00229 (0.0048)
    - Time trend: 0.001070 (0.0011)
    - Time trend squared: -0.00000622 (0.0000054)
    - Constant: 0.0999 (0.11)
    - Observations: 159
    - R-squared: 0.95
    - Sample Period: Jan. 1996 to Mar. 2009
  - E. Europe (column (5)):
    - Log Dispersion in Int Rates (t-1): 0.612*** (0.076)
    - VIX: 0.0109*** (0.0018)
    - Cross-section Mean Interest Rate: 0.0410*** (0.0049)
    - Crisis: -0.00694 (0.0077)
    - Time trend: 0.0118*** (0.0024)
    - Time trend squared: -0.0000716*** (0.000012)
    - Constant: -0.401** (0.17)
    - Observations: 159
    - R-squared: 0.98
    - Sample Period: Jan. 1996 to Mar. 2009
- Note: *, ** and *** indicate significance at, or below the 10%, 5% and 1% levels respectively. Robust standard errors in parentheses.

### III. PORTFOLIO HOLDINGS AND HOME BIAS — Summary Statistics
- Data and coverage
  - Use Coordinated Portfolio Investment Survey (CPIS) to study cross-border portfolio holdings (foreign equity and debt securities), valued at market prices, classified by issuer residence.
  - CPIS expanded in 2001; annual data available since 2001. Data used: 2001 to 2007.
  - Tables summarize 2007 year-end holdings for 10 Asian economies in sample, showing shares invested in groups: Industrialized, Asia, Latin America, Eastern Europe, and Other; last column shows total holding in millions of US$.

- Equity holdings (Table 4) — Highlights for Asia (ratios of total invested by investor economy)
  - Size of holdings:
    - Japan: 479,145.3 (mill US$)
    - Korea, Republic of: 79,073.8 (mill US$)
    - Hong Kong, China: 105,047.0 (mill US$)
    - Singapore: 92,625.6 (mill US$)
    - Malaysia: 47,150.8 (mill US$)
    - Thailand: 2,209.7 (mill US$)
    - Philippines: 115.4 (mill US$)
    - Indonesia: 507.4 (mill US$)
    - India: 236.7 (mill US$)
    - Pakistan: 25.0 (mill US$)
  - Within-Asia investment shares (simple and weighted averages)
    - Simple average within-Asia: 0.3845
    - Weighted average within-Asia: 0.2139
  - Industrialized countries benchmarks
    - Industrialized countries weighted average investment within-group: 0.7508
  - Other regional weighted averages (as issuer groups)
    - Latin America weighted average: 0.0123
    - Eastern Europe weighted average: 0.0168

- Debt holdings (Table 5) — Highlights for Asia (ratios of total invested by investor economy)
  - Size of holdings:
    - Japan: 1,363,105.9 (mill US$)
    - Korea, Republic of: 45,986.6 (mill US$)
    - Hong Kong, China: 205,591.0 (mill US$)
    - Singapore: 168,745.9 (mill US$)
    - Malaysia: 2,006.7 (mill US$)
    - Thailand: 10,659.2 (mill US$)
    - Philippines: 4,880.9 (mill US$)
    - Indonesia: 1,350.0 (mill US$)
    - India: 5.7 (mill US$)
    - Pakistan: 214.7 (mill US$)
  - Within-Asia investment shares (simple and weighted averages)
    - Simple average within-Asia: 0.1506
    - Weighted average within-Asia: 0.0662
  - Industrialized countries benchmarks
    - Industrialized countries weighted average investment within-group (debt): 0.9372
  - Other regional weighted averages (as issuer groups)
    - Latin America weighted average (debt): 0.0678
    - Eastern Europe weighted average (debt): 0.0206

- U.S. and Japan exposure in Asian portfolios (Table 6)
  - Equity holdings toward United States (ratios)
    - Average (Asia): 0.2598
    - Weighted average (Asia): 0.3705
  - Equity holdings toward Japan (ratios)
    - Average (Asia): 0.0241
    - Weighted average (Asia): 0.0252
  - Debt holdings toward United States (ratios)
    - Average (Asia): 0.3370
    - Weighted average (Asia): 0.3875
  - Debt holdings toward Japan (ratios)
    - Average (Asia): 0.0132
    - Weighted average (Asia): 0.0087
  - Country-specific notable values (Equity / Debt)
    - Philippines: United States 0.7682 / 0.4671; Japan 0.0019 / 0.0170
    - Japan: United States 0.4631 / 0.4339; Japan 0.00000 / 0.0000
    - Korea, Republic of: United States 0.3175 / 0.6010; Japan 0.0555 / 0.0107
    - Hong Kong, China: United States 0.1745 / 0.2540; Japan 0.0851 / 0.0597
  - Summary statement from text:
    - The US looms large in the portfolio allocation of Asian countries relative to Japan. For equity holdings, only Hong Kong, China, Singapore and Korea have >5% of portfolios in Japanese assets; for debt only Hong Kong, China’s holdings exceed 5%.

### Summary conclusion on regional integration and home bias
- The extent of regional financial integration within Asia, measured by cross-border portfolio investments, is broadly similar to that in Latin America and Eastern Europe but far below the industrialized country average.
- Convergence evidence:
  - Equity premia dispersion: episodic convergence within regions with interruptions during volatility and crises; Eastern Europe showed the fastest convergence pace prior to the crisis.
  - Interest rate dispersion: substantial declines in Asia and Eastern Europe in recent years; industrialized countries exhibit much lower dispersion levels than emerging regions.

*Source: _wp1104 - Section 2 Section 3 Section 4*

### 3.2 Portfolio Holdings: Gravity Model Estimates

### 3.2 Portfolio Holdings: Gravity Model Estimates

### Gravity model specification and variables
- Dependent variable: bilateral investment holding (pooled data for 2001 to 2007).
- Time-varying independent variables:
  - Product of real GDP (investor × issuer).
  - Product of real per capita GDP (investor × issuer).
- Time-invariant independent variables (from Rose (2005)):
  - Distance between countries (log).
  - Dummy variables: Common border, Island, Regional Trade Agreement.
  - Bilateral trade in 1998 (log).
- Additional variable (following Fidora, Fratzcher and Thimann (2007)):
  - Real Exchange Rate Volatility (log).

### Main gravity-estimation findings (Table 7)
- Bilateral equity holdings (columns (1) and (2)) and bond holdings (columns (3) and (4)) yield expected signs for most determinants and many statistically significant estimates.
- Selected coefficient estimates (standard errors in parentheses):
  - Product of GDPs (log):
    - Equity (col (1)): 0.488*** (0.11)
    - Equity with region dummies (col (2)): 0.490*** (0.11)
    - Bonds (col (3)): 0.462*** (0.083)
    - Bonds with region dummies (col (4)): 0.430*** (0.091)
  - Product of per capita GDP (log):
    - Equity: 0.904*** (0.11); 0.772*** (0.14)
    - Bonds: 0.679*** (0.072); 0.517*** (0.11)
  - Distance (log):
    - Equity: -0.187 (0.17); 0.146 (0.15) — distance less important for equity in some specs
    - Bonds: -0.379*** (0.12); -0.256* (0.13) — distance matters more for debt holdings
  - Common Language:
    - Equity: 0.909*** (0.33); 0.894*** (0.28)
    - Bonds: 0.0506 (0.21); 0.0888 (0.24)
  - Island:
    - Equity: 0.564* (0.33); 0.487 (0.31)
    - Bonds: 0.802*** (0.22); 0.968*** (0.31)
  - Regional Trade Agreement:
    - Equity: 0.537 (0.39); 0.683** (0.32)
    - Bonds: 1.095*** (0.28); 0.591** (0.25)
  - Bilateral Trade (log):
    - Equity: 0.510*** (0.099); 0.504*** (0.10)
    - Bonds: 0.328*** (0.072); 0.377*** (0.071)
  - Real Exchange Rate Volatility (Log):
    - Equity: -0.661** (0.30); -0.294 (0.27)
    - Bonds: -0.890*** (0.22); -0.891*** (0.21)
- Sample size and fit:
  - Observations: 10815 (equity specs), 11545 (bond specs)
  - R-square overall: 0.571, 0.597 (equity); 0.596, 0.607 (bonds)
- Regional dummy findings from Table 7:
  - Asia dummies (Asia_Single, Asia_Pair) are insignificant in both equity and debt equations (columns (2) and (4)).
  - Industrialized_Pair is positive and significant for both equities and bonds (Industrialized_Pair: 0.761* (0.44) in equity; 0.979** (0.43) in bonds).
  - LatAm_Single is significantly negative in equity (LatAm_Single: -1.523*** (0.55)) in that specification.
  - EEurope_Pair is positive and significant in equity (EEurope_Pair: 1.557** (0.78)).

### Group-specific gravity regressions (Tables 8 and 9)
- Approach: estimate gravity model separately by investor economy group (Asia, Industrialized, Latin America, E. Europe).
- Equity holdings by investor group (Table 8) — selected coefficients (standard errors in parentheses):
  - Product of GDPs (log):
    - Asia: 0.149 (0.22)
    - Industrialized: 0.642*** (0.089)
    - Latin America: 0.628* (0.33)
    - E. Europe: 0.269 (0.21)
  - Product of per capita GDP (log):
    - Asia: 1.027*** (0.14)
    - Industrialized: 0.847*** (0.078)
    - Latin America: 0.406* (0.23)
    - E. Europe: 0.413* (0.23)
  - Distance (log):
    - Asia: -0.664*** (0.25)
    - Industrialized: 0.0246 (0.14)
    - Latin America: -0.440 (0.62)
    - E. Europe: -0.312 (0.20)
  - Bilateral Trade (log):
    - Asia: 0.651*** (0.24)
    - Industrialized: 0.622*** (0.12)
    - Latin America: 0.540** (0.26)
    - E. Europe: 0.352* (0.20)
  - Real Exchange Rate Volatility (Log):
    - Asia: -0.647 (0.60)
    - Industrialized: -0.218* (0.12)
    - Latin America: -0.953 (0.96)
    - E. Europe: -0.216 (0.44)
  - Observations:
    - Asia: 150; Industrialized: 457; Latin America: 299; E. Europe: 121416 (as printed)
  - R-square overall: Asia 0.603; Industrialized 0.709; Latin America 0.423; E. Europe 0.302
- Debt holdings by investor group (Table 9) — selected coefficients (standard errors in parentheses):
  - Product of GDPs (log):
    - Asia: 0.242 (0.19)
    - Industrialized: 0.760*** (0.10)
    - Latin America: 0.283** (0.12)
    - E. Europe: 0.349*** (0.12)
  - Product of per capita GDP (log):
    - Asia: 0.810*** (0.18)
    - Industrialized: 0.597*** (0.11)
    - Latin America: 0.754*** (0.21)
    - E. Europe: 0.140 (0.19)
  - Distance (log):
    - Asia: -0.0673 (0.41)
    - Industrialized: -0.626*** (0.14)
    - Latin America: -0.839*** (0.24)
    - E. Europe: -0.334* (0.20)
  - Bilateral Trade (log):
    - Asia: 0.499*** (0.19)
    - Industrialized: 0.281** (0.12)
    - Latin America: 0.170 (0.11)
    - E. Europe: 0.242** (0.094)
  - Real Exchange Rate Volatility (Log):
    - Asia: -0.463 (0.34)
    - Industrialized: -0.199 (0.16)
    - Latin America: -1.570** (0.69)
    - E. Europe: -1.102*** (0.29)
  - Observations:
    - Asia: 166; Industrialized: 257; Latin America: 701; E. Europe: 1501492 (as printed)
  - R-square overall: Asia 0.607; Industrialized 0.677; Latin America 0.312; E. Europe 0.314
- Interpretation:
  - Product of per capita GDPs and bilateral trade remain consistently significant across many subgroup regressions (exceptions noted in the source for certain debt regressions).
  - Regional Trade Agreement impact on Asian equity and debt holdings is positive but not precisely measured in subgroup results.

### Suggested directions for further research (from authors)
- For Asia, potential determinants to investigate:
  - Role of foreign direct investment.
  - Effects of presence of financial centers.
  - Impact of China‘s gravitational pull.
  - Role of the overseas Chinese community.

### Home Bias: definition and measurement
- Home bias (HB) for country i:
  - HB_i = 1 - (w_i / w_i*)
- Bilateral home bias (country i toward country j):
  - HB_ij = 1 - (w_ij / w_j*)
- Interpretation:
  - Full international diversification: w_ij = w_j* → HB = 0.
  - No holdings of j by i → HB = 1.

### Home bias summary statistics (Table 10)
- Panel A: Equity Investment Home Bias, Average 2001-2007 (selected entries)
  - Hong Kong, China: Industrialized 0.933; Asia 0.896; Latin America 1.000; Eastern Europe 1.000; Other 1.000
  - India: Industrialized 1.000; Asia 1.000; Latin America 0.999; Eastern Europe 1.000; Other 1.000
  - Singapore: Industrialized 0.754; Asia 0.100; Latin America 0.952; Eastern Europe 0.977; Other 0.926
  - Asia Average: Industrialized 0.926; Asia 0.841; Latin America 0.993; Eastern Europe 0.997; Other 0.943
  - Industrialized Average: Industrialized 0.617; Asia 0.782; Latin America 0.963; Eastern Europe 0.950; Other 0.941
  - Latin America Average: Industrialized 0.781; Asia 0.853; Latin America 0.907; Eastern Europe 0.970; Other 0.909
  - Eastern Europe Average: Industrialized 0.628; Asia 0.865; Latin America 0.881; Eastern Europe 0.755; Other 0.872
- Panel B: Debt Investment Home Bias, Average (note: for Hong Kong, China and Singapore averages are 2001-06; Asia Average accounts for this)
  - Hong Kong, China (*): Industrialized 0.613; Asia 0.347; Latin America 0.884; Eastern Europe 0.828; Other 0.779
  - Indonesia: Industrialized 0.524; Asia 0.633; Latin America 0.691; Eastern Europe 0.996; Other 1.000
  - Philippines: Industrialized 0.501; Asia 0.421; Latin America 0.841; Eastern Europe 0.545; Other 0.882
  - Singapore (*): Industrialized 0.649; Asia 0.918; Latin America 0.982; Eastern Europe 0.965; Other 0.888
  - Asia Average (*): Industrialized 0.768; Asia 0.791; Latin America 0.921; Eastern Europe 0.915; Other 0.944
  - Industrialized Average: Industrialized 0.512; Asia 0.956; Latin America 0.965; Eastern Europe 0.958; Other 0.984
  - Latin America Average: Industrialized 0.800; Asia 0.978; Latin America 0.851; Eastern Europe 0.971; Other 0.977
  - Eastern Europe Average: Industrialized 0.677; Asia 0.939; Latin America 0.895; Eastern Europe 0.879; Other 0.963
- Interpretation from source text:
  - Asian economies tend to be very home-biased toward the industrialized group in equities (exception: Singapore and the Philippines).
  - Most Asian economies are slightly less home-biased toward other Asian economies than toward industrialized ones (exceptions: Japan and the Philippines).
  - Regional averages: Asia equity home bias .841 vs. Latin America .907; Asia not as low as Eastern Europe .755 or Industrialized .617.
  - Debt home bias tends to be below equity home bias (many Panel B entries lower than Panel A).
  - Asia is the second least home-biased region in bond holdings (.791), after Industrialized (.512), ahead of Latin America (.851) and Eastern Europe (.879).

### Determinants of average bilateral home bias (Table 11)
- Dependent variable: average bilateral home bias over 2001-2007 for equities (cols (1),(2)) and bonds (cols (3),(4)).
- Selected coefficient estimates (standard errors in parentheses):
  - Product of GDPs (log):
    - Equities (col (1)): 0.0341*** (0.010)
    - Equities with region dummies (col (2)): 0.0297*** (0.011)
    - Bonds (col (3)): 0.0781*** (0.013)
    - Bonds with region dummies (col (4)): 0.0605*** (0.013)
  - Product of per capita GDP (log):
    - Equities: -0.0669*** (0.0072); -0.0547*** (0.0089)
    - Bonds: -0.0829*** (0.0092); -0.0877*** (0.012)
  - Distance (log):
    - Equities: 0.0865*** (0.015); 0.0647*** (0.020)
    - Bonds: 0.0766*** (0.020); 0.0306 (0.024)
  - Common Language:
    - Equities: -0.0757** (0.034); -0.111*** (0.037)
    - Bonds: -0.0874** (0.042); -0.0661 (0.045)
  - Island:
    - Equities: -0.0724*** (0.024); -0.130*** (0.025)
    - Bonds: -0.125*** (0.029); -0.168*** (0.031)
  - Bilateral Trade (log):
    - Equities: -0.0293*** (0.0090); -0.0400*** (0.0097)
    - Bonds: -0.0452*** (0.012); -0.0481*** (0.013)
  - Real Exchange Rate Volatility (Log):
    - Equities: 0.0902*** (0.022); 0.118*** (0.026)
    - Bonds: 0.136*** (0.027); 0.218*** (0.031) — effect larger for bonds than equities
  - Regional dummies:
    - Asia_Single:
      - Equities (col (2)): 0.0853*** (0.033)
      - Bonds (col (4)): 0.193*** (0.041)
    - Asia_Pair:
      - Equities (col (2)): 0.138** (0.057)
      - Bonds (col (4)): -0.0658 (0.086)
    - Industrialized_Pair:
      - Equities (col (2)): -0.0824 (0.054)
      - Bonds (col (4)): 0.125** (0.060)
    - EEurope_Pair:
      - Equities (col (2)): -0.204*** (0.077)
      - Bonds (col (4)): -0.257*** (0.092)
- Observations and model fit:
  - Observations: 2840 (equities); 2871 (bonds)
  - F-Statistics (Prob>F): equities 50.74 (.000), equities+region 29.30 (.000); bonds 64.45 (.000), bonds+region 38.18 (.000)
- Interpretation from source text:
  - Real exchange rate volatility raises home bias, with a larger effect for bonds than for equities.
  - Bilateral equity home bias is higher if either or both economies are in Asia (equities column (2)).
  - For bond home bias (col (4)), an increased home bias is detected when one of the economies in the pair is in Asia (Asia_Single), but not when both are (Asia_Pair insignificant/negative).
  - Pairs of Latin American or Eastern European economies seem less bilaterally home-biased in bond holdings (LatAm_Pair and EEurope_Pair reported negative/insignificant patterns in bonds).

### Risk-sharing and financial stability (section IV, 4.1)
- Theoretical expectations:
  - Financial integration should raise cross-country consumption correlations and lower cross-country output correlations; should reduce correlations between domestic saving and domestic investment.
- Literature summary:
  - Imbs (2006): financial integration (CPIS 2001) raises cross-country consumption correlations (increased risk sharing) but also raises cross-country output correlations.
  - Kose, Prasad and Terrones (2007): increases in international risk sharing modest and largely benefit industrialized countries.
  - Bekaert, Harvey and Lundblad (2006): financial liberalization associated with lower consumption growth volatility, effect weaker for emerging markets.
  - Asia-specific literature: mixed results (Kim and Lee, 2008; Kim, Kim and Wang, 2006; Kim, Oh and Jeong, 2005).
- Empirical evidence in this chapter:
  - Cross-country dispersions of output and consumption growth used to assess risk sharing.
  - Figure 3 (1971-2007) indicates a decline in the dispersion of consumption growth — some evidence of increased risk sharing — but also a decline in dispersion in output growth.
- Regression evidence (Table 12):
  - Dependent variables: standard deviation of real GDP growth across economies (Panel A) and standard deviation of consumption growth (Panel B).
  - Common covariates: lagged dependent variable, per capita world GDP (or consumption) growth, CRISIS indicator, time trend.
  - Selected coefficient estimates and interpretations:
    - Panel A (Standard Deviation of GDP Growth):
      - Standard Deviation of GDP Growth (t-1): 0.106 (0.17) for All countries; 0.303* (0.16) Industrialized; 0.162 (0.15) Asia; -0.101 (0.15) Latin America
      - Crisis: 0.0776** (0.034) for All countries; 0.0403 (0.096) Industrialized; 0.280 (0.20) Asia; 0.141 (0.086) Latin America
      - Time trend: -0.0236*** (0.0085) All countries; -0.0282*** (0.0079) Industrialized; -0.0294** (0.014) Asia; -0.0317* (0.018) Latin America
      - Observations: 37 for all panels in Table 12; R-squared varying by group (All countries 0.40; Industrialized 0.59; Asia 0.32; Latin America 0.20)
    - Panel B (Standard Deviation of Consumption Growth):
      - Per Capita World Consumption Growth: -0.318** (0.13) All countries; 0.0589 (0.14) Industrialized; -0.333 (0.23) Asia; -0.705** (0.29) Latin America
      - Crisis: 0.0693 (0.043) All countries; 0.336** (0.14) Industrialized; 0.357 (0.33) Asia; 0.0336 (0.11) Latin America
      - Time trend: -0.0453*** (0.012) All countries; -0.0469*** (0.012) Industrialized; -0.0395* (0.021) Asia; -0.0553** (0.024) Latin America
      - Observations: 35 for consumption panels; R-squared: All countries 0.38; Industrialized 0.50; Asia 0.20; Latin America 0.34
  - Time trend is very significant in all cases, indicating declines over time in cross-country dispersions of GDP and consumption growth.

*Source: _wp1104 - 3.2 Portfolio Holdings: Gravity Model Estimates*

### 4.2 Financial Integration and Crises

### 4.2 Financial Integration and Crises

### Benefits of fuller development of local financial markets
- Achievement of higher scale, which makes markets more liquid and prices more meaningful.
- Provides firms with a richer choice of instruments to manage their capital structure, including bond debt and equity.
- Enhancement of the ability to issue long-term, local currency-denominated debt, which improves the firms‘ ability to manage debt more efficiently.
- These issues have been studied extensively in the context of the Asian Bond Markets initiative (see ADB, 2008).

### Debate: can local/regional market development reduce financial crises?
- Arguments for reduced crisis likelihood:
  - Asian investors may display a “regional bias” given familiarity with Asian companies and assets and a more favorable perception of risk, making them less likely to engage in panic-driven sales.
  - Development of regional financial markets would provide a better set of instruments (more long-term, local currency debt) which could make the region less crisis-prone even if investors are equally fickle.
- Arguments against:
  - Non-transparent corporate governance and accounting in Asia undermines any informational advantage of local investors.
  - No solid evidence that Asian investors would have a stronger preference for Asian financial assets than global investors.
  - Portfolio structure is related to deeper reasons such as risk diversification and macroeconomic fundamentals, and may not change with improved market infrastructure and liquidity.
- Key question: Are global hedge funds or non-Asian investors more prone to panic and withdraw funding on a broad scale than local or regional investors?

### Single-currency market as a stabilizing channel
- Evidence from European bond markets after the introduction of the euro:
  - Borrowing costs fell sharply for countries that had higher yield spreads before the common currency.
  - Effects included gains in liquidity and broadening of the investor base facilitated by a market denominated in a major currency like the euro.
  - Investors perceived bonds issued by countries in the euro group as belonging to the same asset class, creating a much larger asset class of euro-denominated bonds that helped achieve more favorable borrowing conditions and a deeper, more stable market.
- Implication: Benefits from Asian financial integration would be enhanced in the presence of a currency union, even if the institutional framework were not as extensive as in the case of the European Union.

### Evidence from studies of investor behavior in past crises
- Mexico (December 1994 peso crisis):
  - Frankel and Schmukler (1996): Mexican investors turned pessimistic on exchange rate sustainability before international investors. Net asset values of country funds declined sharply relative to the value of country funds two weeks before the devaluation.
  - Frankel and Schmukler (1998): Domestic investors lead market trends over foreign investors in a study of 61 country (and multi-country) funds based in New York City.
- Korea (1997 Asian crisis):
  - Choe, Kho and Stulz (1998): Foreign investors tended to follow “positive feedback strategies” (selling stocks that fell and buying those that appreciated), but there was no evidence their trades had a lasting impact on market prices, particularly during the last three months of 1997.
- International mutual funds (1996 to 1999):
  - Borensztein and Gelos (2003): Funds have a tendency to “herd,” but the extent is small. On average, international mutual funds were 7% more concentrated on one side of the market (either buying or selling).
  - Funds tended to cut exposure in the month before crises, but the extent of sales was relatively modest.
- Equity market liberalization:
  - Bekaert, Harvey and Lundblad (2004): Did not find evidence of increased market instability after liberalization to international equity investment.
- Regional foreign investors (bank lending during Asian crises):
  - Siregar and Choy (2008): Japanese banks reduced loan exposure to East Asian markets by about 30% in 1998-99; UK and US banks cut back by only 13 and 17%, respectively.
  - Loans by Japanese banks failed to recover to pre-crisis levels nearly 10 years after the Asian crisis.
  - Econometric analysis suggests higher sensitivity to financial risk (an index including variables such as external debt and international reserves) explains Japanese banks‘ behavior.

### Investor behavior during the subprime crisis: Evidence from Brazil
- Change in foreign holdings in Brazil:
  - Foreign holdings of local securities hovered below $50 billion during the previous decade, then grew sharply and surpassed $250 billion by mid-2008 (figures reported by the Securities Commission of the Brazilian government).
- Drivers of increase in foreign holdings:
  - Capital inflows and asset value appreciation (booming stock market and appreciating real).
  - In 2007: net inflows amounted to less than $34 billion, but asset holdings of foreign investors increased by over $112 billion.
  - International investors‘ holdings comprised mostly stocks, around 80% of total asset value in 2007.
  - The stock market gained over 40% in 2007; the real appreciated by 16% in 2007.
- Crisis-period adjustments (September-November 2008):
  - Net outflows in September-November amounted to about $13 billion, contrasting with a decline of over $116 billion in asset holdings.
  - Compound loss of 55% of the value of the stock market index (the Bovespa) in US dollar terms.
  - Gross inflows amounted to about $25 billion during the September-November period (gross flows ran in both directions).
- Implications of the new liability structure versus “original sin”:
  - Old “original sin” structure: most liabilities were foreign currency-denominated debt issued in international markets; sudden stops produced large exchange rate depreciations that increased foreign-currency debt burdens.
  - New liability structure: large part of debt liabilities denominated in local currency and local equity positions are a large fraction of foreign investors‘ portfolios.
    - Advantage: a sudden stop with depreciation spreads losses to foreign investors and reduces the dollar value of liabilities, bolstering solvency of governments, firms and banks with local-currency debts.
    - Disadvantage: liquidity pressures can be stronger because global investors can sell securities immediately and put pressure on the exchange rate as they repatriate positions.
  - In the “original sin” world, attempts to liquidate positions largely remained within international investors‘ realm and did not have a direct effect on the country‘s foreign exchange market.
- Additional Brazil observations:
  - Available data suggest outflows by Brazilian investors and firms amounted to just under $6 billion in the September-November period (does not include foreign currency facilities and currency swaps provided by the central bank).
  - Notorious cases of losses by Brazilian firms from derivative contracts designed to protect against appreciation of the real, which included multiple liability for a depreciation beyond a certain level and ultimately caused heavy losses.
- Preliminary conclusion from Brazil: the case does not suggest a clear-cut distinction between foreign and local investors in terms of their impact on local financial markets.

### Selected conclusion (from paper’s survey)
- 1. There is a process of convergence in equity premia and interest rates among Asian countries, particularly over the period 2000 to 2006 (see section 2 of the paper, Tables 2 and 3, and Figures 1 and 2). However, with some exceptions, much the same process has occurred in Latin America and Eastern Europe. Convergence has tended to be faster in the Eastern European case than in Asia. In all regions, crises interrupt the process of convergence.

*Source: _wp1104 - 4.2 Financial Integration and Crises*

### 2.  Cross-border  holdings  among  Asian  countries  are  significant  in  many  cases  and  have

### _wp1104 - 2.  Cross-border  holdings  among  Asian  countries  are  significant  in  many  cases  and  have

### Cross-border holdings: magnitude and pattern
- Cross-border holdings among Asian countries are significant in many cases and have increased over the last decade.
- The pattern of Asian cross-border holdings is not strikingly different from that in other regional groups (Latin America; Eastern Europe) and is not near industrialized country "benchmarks" (section 3.1; Tables 4 and 5).
- U.S. assets continue to play a dominant role relative to Japanese assets in the portfolio allocations of Asian countries (Table 6).

### Gravity model and determinants of holdings
- The gravity model continues to successfully describe cross-border equity and bond holdings (section 3.1; Tables 7 to 9).
- No evidence was found that financial integration in Asia has reached a level that counteracts the effects of gravity variables (section 3.1; Tables 7 to 9).
- Attempts to detect modifications to gravity-variable effects by regional integration (e.g., interacting variables with Asia_Pair) did not appreciably change magnitudes, signs, or significance; interaction terms were largely not significant (endnote 9).
- The importance of global financial markets (US and UK) is large in the data; a dummy-variable approach shows the Global coefficient is very significant, while interactions for regional groups were generally not significant (endnote 8).

### Home bias and real exchange rate volatility
- Several Asian economies display departures from regional home bias comparable to Latin American countries but lower than Eastern Europe or industrialized countries (section 3.2; Table 10).
- A gravity model explains home bias well; adding real exchange rate volatility is an important augmentation to the standard model (section 3.2; Table 11).
- Little evidence suggests Asia no longer conforms to the standard home-bias model (section 3.2; Table 11).
- The product of per capita GDPs reduces home bias across all country groups and for both equities and bonds; other gravity-variable performance is mixed (endnote 12).
- Real exchange rate volatility raises home bias, though the effect is not significant for industrialized countries for both equities and bonds and not significant for Asian countries in the case of bonds (endnote 12).

### Prices, quantities, and consumption/GDP convergence
- Trends in "prices" (equity premia, interest rates) and "quantities" (cross-border equity and bond holdings) are reflected in convergence in consumption growth rates among Asian countries (section 4.1; Figure 3 and Table 12).
- There is no evidence that increased risk sharing within Asia is occurring at a faster rate than in other country groups.
- The study finds convergence not only in consumption growth but also in GDP growth, a puzzling result from the perspective of standard economic theory yet consistent with previous studies (section 4.1; Figure 3 and Table 12).
- Risk-sharing improvements may be long-term phenomena driven by output-growth-rate convergence related to trade in ideas, technologies, and diffusion of institutions (endnote 13).

### Local investor base and stability
- The stability of a local investor base is often cited as a benefit of regional financial integration (section 4.2).
- The survey of the evidence invites caution about the claim that local investors provide greater stability (section 4.3).
- A case study of Brazil‘s experience with the ongoing financial crisis does not indicate a clear distinction between foreign and local investors in terms of their impact on local financial markets (section 4.4; Figures 4 and 5).
- Foreign investors fleeing domestic markets can contribute to steep market declines and currency depreciation, complicating assessments of solvency risk versus market volatility in a post-"original sin" world (endnote 15).

### Capital controls and asset holdings
- An index-based measure of capital controls (1995-2005 averages of outflow restrictions by investor and inflow restrictions by issuer) was used to investigate effects on cross-border asset holdings and home bias (endnote 11).
- No systematic or significant effect of controls on asset holdings was found for equities and bonds using the main regressions.
- In home bias regressions, capital controls increased home bias for bond holdings in the all-countries sample and for industrialized countries, Latin America, and Eastern Europe — but not for Asia (endnote 11).

### Methodological and data notes
- The study uses VIX series from 1990 onwards and the older VXO series for 1986 through 1989 (endnote 2).
- Outliers: Data for Bermuda, the Cayman Islands, Ireland, Luxembourg and Panama were excluded from regressions due to being huge outliers; some 2007 bond holdings data for Hong Kong, China and Singapore were treated as outliers and for those series 2001 to 2006 data were used where appropriate (endnote 4).
- Standard errors in Tables 7, 8 and 9 account for clustering by investor economy; main conclusions are not sensitive to this choice, nor to estimating separate regressions for each year rather than pooling the data (endnote 6).

### Bottom-line assessment and policy implication
- Asian regional financial integration is clearly increasing, at least in tranquil times, but Asian economies remain more integrated with global financial markets than with regional neighbors.
- From an economic standpoint, stronger ties to global markets are not necessarily undesirable: Imbs and Mauro (2007) suggest gains from global diversification may be higher than those from purely regional integration.
- Policy implication: While promoting deeper regional financial integration can have benefits, policymakers should weigh regional integration efforts against the continued predominance and potential advantages of global diversification.

*Source: IMF Working Paper content unit _wp1104 (sections and endnotes as presented in the supplied text).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1104.pdf_