## _wp0413

## Source details

**Canonical URL:** [_wp0413](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0413.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0413.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0413.pdf.json)

---

### I. Introduction — research questions and approach
- Central questions:
  - How high are average returns to emerging market lending—particularly sovereign lending—over long periods?
  - How volatile are they over shorter periods?
  - How correlated are they with returns on other assets, such as U.S. equities or U.S. high‑yield bonds?
  - After debt crises and restructurings, how much do private creditors actually lose?
  - How have average returns evolved over time—from the 1980s to the 1990s?
- Data and method:
  - Main data source: World Bank’s Global Development Finance (GDF) database (available annually since 1970).
  - Adjust published GDF data to capture all cash flows associated with initial disbursements and vice versa using available World Bank supplemental data.
  - Compute internal rates of return using adjusted flows; outstanding debt stocks—and initial debt stocks when computing returns over subperiods—are valued at secondary‑market debt prices.
- Scope and contributions:
  - First paper to study returns on postwar emerging‑market debt over 1970–2000 including the 1980s debt crisis and the 1990s boom‑bust cycle.
  - Two alternative approaches to recover flows from World Bank data (direct and indirect) to enhance robustness.
  - Use secondary‑market price data since 1986 to construct average year‑end debt prices and characterize volatility and covariance with other asset classes.

### II. Key empirical findings (long run and aggregate)
- Long‑run (1970–2000) returns and premiums:
  - During 1970–2000, returns averaged about 9 percent per annum, about the same as that on a 10–year U.S. government bond.
  - Aggregate long‑run finding summarized: average long‑run returns to emerging‑market private creditors from 1970 to 2000 were about 8–10 percent.
  - Ex post risk premiums over the entire 1970–2000 period were close to zero.
  - Aggregate numeric example (aggregate row summaries from Table 3): All countries: actual return 8.4; comparator returns shown include US3y 8.4, US10y 9.1, UScb 10.4.
- Country dispersion:
  - Most major debtor countries have rates of return clustered in the range between 8 percent and about 10.5 percent.
  - 11 out of 22 countries, accounting for “77 percent” of the face value of debt in 1985 (“74 percent” in 2000), had average rates of return within 100 basis points of the corresponding 3 year U.S. treasury bond.
  - 15 out of 22 countries, making up “91 percent” of the debt in 1985 and “95 percent” in 2000, had average rates of return within 200 basis points of the corresponding 3 year U.S. treasury bond.
- Comparisons with other assets:
  - Returns were about the same or slightly higher than returns on 3 year U.S. treasury bonds.
  - Returns were about the same or slightly lower than returns on 10 year U.S. treasury bonds.
  - Returns were about “100-200 basis points” lower than the return on U.S. corporate bonds.

### III. Time decomposition and heterogeneity (subperiods)
- Broad decomposition of 1970–2000:
  - 1970–89: cycle with negative or very low returns.
  - 1989–1993: period of very high returns.
  - Post‑1993: period of lower but positive ex post spreads.
- Subperiod numeric patterns:
  - Dividing sample at 1989 (the trough): aggregate ex post spreads are highly negative for 1970–89 (examples reported in the order of -330 to -500 basis points for some computations) and strongly positive for 1989–2000.
  - Dividing at 1992: aggregate 1970–92 spreads are negative but smaller (order of -110 to -180 basis points); 1992–2000 spreads substantially positive but lower than 1989–2000 averages.
  - Three‑way split (1970–1989, 1989–1994, 1994–2000):
    - Exceptional recovery and high returns in 1989–1994 (rebound coincident with Brady deals).
    - Strong returns 1994–1997 driven mainly by countries not directly hit by the Asian crisis.
    - Returns decline after 1997 (Russian crisis) but remain slightly positive in aggregate.
- Brady deal countries:
  - Brady‑deal era analysis shows creditor outcomes improved markedly in many cases; Table 7 indicates bank spreads (indirect approach) mostly above -300 basis points for major creditors, with Mexico and Venezuela having spreads basically zero or positive in some computations.

### IV. Methodology: measuring flows, valuation, and direct vs. indirect approaches
- Measurement complications and categories:
  - Primary focus on privately held long term debt (public and publicly guaranteed (PPG) and private non‑guaranteed (PNG)).
  - GDF net transfers = disbursements − (interest payments + principal repayments), with conventions that capture buybacks and interest‑arrears clearance as principal/interest repayments.
  - Cross‑asset restructurings (debt‑equity swaps, short‑to‑long consolidations, loan→bond Brady exchanges) create biases if raw GDF net transfers are used by category.
- Constructed modified net transfer series (creditor perspective):
  - Define ntr* = –ntr + des + x, where:
    - ntr are GDF net transfers from debtor perspective,
    - des = market value of debt swapped for equity,
    - x = sum of quasi‑repayments (positive) and quasi‑disbursements (negative) attributable to cross‑category stock adjustments.
  - For total debt x = 0; for privately held long term debt x = – sc (sc = increase in long term debt stock due to short term consolidation).
- Two implementation approaches:
  - Direct approach: measure components of x directly (use World Bank data on debt‑equity swaps and short‑term consolidations). For privately held long term debt: ntr* = –ntr + des – sc.
  - Indirect approach: use stock‑flow identities and back out x as a residual via ∆D = d – r – dfr + ccv + ic – x + u, where u captures measurement error.
- Method choice and robustness:
  - Both approaches are applied; long‑run rates of return and overall conclusions are robust to method choice.
  - Indirect method generally more trustworthy in many cases, but not uniformly superior; both methods presented because method choice can materially affect some country‑level and short‑period results.

### V. Valuation, sample coverage, and survival bias
- Secondary‑market prices and valuation conventions:
  - Secondary‑market debt prices used to value outstanding debt stocks and initial stocks when computing subperiod returns.
  - Secondary‑market prices available only for some instruments and countries and not prior to 1986; end‑1970 debt is valued at par for the full 1970–2000 sample.
- Sample and representativeness:
  - End‑2000 debt prices available for 22 countries: Algeria, Argentina, Brazil, Chile, Colombia, Côte d'Ivoire, Ecuador, Indonesia, Jordan, Korea, Lebanon, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, Philippines, Thailand, Turkey and Venezuela.
  - The 22‑country group accounted for roughly: “around 80 percent in the 1970s, about 70 percent in the mid‑1980s, about 60 percent in the early 1990s, rising and currently at about 65 percent” of privately held developing country debt.
  - Long‑term debt comprises between “65 and 85 percent” of privately held external debt in these 22 countries; at end‑2000 it stood at “79 percent”.
- Duration and floating interest structure:
  - World Bank data indicate average maturity about 10 years.
  - Large fraction of debt carried floating interest: “about 50 percent in the 1970s, rising to two‑thirds in the early 1980s before falling back to about 50 percent in the 1990s.”
  - EMBIG sample average durations observed between 3 and 4.2 years; authors take the bottom of that range as a best guess for average duration over the whole period.
- Survival bias assessment:
  - To assess survival bias authors: (a) report results for the 2000 sample without adjustment; (b) compute results for a 1991‑defined sample assuming a zero 2000 price for countries with no secondary market price in 2000 (1991‑based assumption yields a lower bound for returns).
  - Example sensitivity: using 2001 rather than 2000 prices made little difference in aggregate returns, but Argentina’s average long‑run return fell from “8.8 percent” to “3.7 percent” due to a collapse in Argentine debt prices (88 cents → 29 cents).

### VI. Volatility, comovement, and asset‑allocation implications
- Data constraints:
  - Annual returns for volatility/covariance available from secondary‑market sample starting in 1986; hence up to 14 annual observations for many series.
- Volatility facts (aggregation summary):
  - Volatility of annual emerging debt returns since 1986 is higher than that of returns on U.S. corporate or high‑yield bonds, but far lower than the volatility of emerging‑market‑equity returns.
  - Example: for τp–2000 (typically 1986/1987–2000) standard deviation of aggregate emerging‑market debt returns ~15 (Table reports St. dev. 15.9 for “All countries”).
- Betas and correlations:
  - Emerging‑market debt returns exhibited little correlation with the U.S. and world stock markets during this period; estimated betas with respect to U.S. and world stock markets are insignificantly different from zero (sometimes slightly negative).
  - Emerging‑market debt returns exhibit a positive and significant correlation with emerging‑market equity returns; estimated betas with respect to emerging‑market equities lie in the “0.15 – 0.3” range.
- Interpretation for risk premia:
  - Low ex post spreads for emerging‑market debt may be consistent with CAPM‑like pricing if emerging‑market debt returns have low or negative covariance with world or U.S. equity markets; emerging‑market debt may provide diversification benefits.

### VII. Appendix: pricing data, aggregation rules, and alternative asset constructions
- Secondary‑market price construction and aggregation (Appendix Table 2 and notes):
  - Loan prices: LDCDR provides one end‑of‑year loan price per country; used directly when available.
  - Brady and non‑Brady bond prices: weighting by issue amounts using Bloomberg data, World Debt Tables, bradynet.com, and other sources; when amount issued missing, simple average of issues used.
  - Aggregation rules for “All” prices: when multiple instrument types exist use weighted averages by face/outstanding amounts or GDF ratios of bank vs bond debt as weights.
  - J.P. Morgan EMBI inclusion rules summarized (daily prices, minimum $500 million outstanding, maturity requirements, US$ denomination).
- Representativeness:
  - Representativeness rates generally in the 50 to 100 percent range; some rates >100 percent explained by use of “amount issued” rather than “amount outstanding.”
- Alternative asset comparators and computation (Appendix III):
  - Construct comparators (e.g., 3 year U.S. treasury zero coupon bonds, 10 year U.S. treasury zero coupon bonds, U.S. corporate bond index) by treating disbursements to emerging markets as invested in the comparator and computing IRRs for the artificial flow stream.
  - Rationale for 3‑year U.S. bond comparator: given estimated average duration, the 3 year U.S. bond is most appropriate default‑risk comparator though the 10 year is also shown.
  - Recursive construction examples provided for τ = 3 and for return‑indexed comparators (equations and sample recursions given).

### VIII. Returns over one‑year holding periods and country‑level volatility (Appendix IV highlights)
- One‑year holding period returns (Appendix Table 8):
  - Series presented for All countries, All Brady countries, All non‑Brady countries, regional groups and many country‑series across 1986–87 to 1999–2000.
  - Examples of annual swings in one‑year returns for select countries (as reported): Argentina one‑year series includes values such as -43.2, -17.9, 50.4, 72.5, 64.1, 38.4, 10.9, -25.9, 27.7, 23.7, 11.5, 4.3, 7.1, 4.8; Brazil includes -22.8, 8.2, -19.2, 146.1, -34.0, 12.3, 181.3, -28.0, 17.3, 30.9, 9.1, -17.6, 36.8, 11.4; Indonesia and other countries show large year‑to‑year variation.
- Country‑level returns, volatilities and betas (Appendix Table 7):
  - Country examples (debt returns refer to PPG debt, private creditors, average of "direct" and "indirect" approaches):
    - Argentina: return 8.8; volatility entries and betas reported in table with standard errors in parentheses.
    - Brazil: return 8.2; country‑specific volatility and beta statistics provided.
    - Chile: return 10.2; Mexico: return 9.7; Venezuela: return 9.1; Indonesia: return 10.4; Korea: return 9.4; Malaysia: return 9.9; Philippines: return 6.6; Thailand: return 10.8; Turkey: return 9.3 (full table includes standard errors and multiple comparator betas).
  - Appendix Table 7 notes: statistics are gross U.S. dollar returns based on annual returns.

### IX. Conclusions and interpretation
- Core conclusion:
  - Investors earned about the same over 1970–2000 on emerging‑market private credit as they would have on medium‑term U.S. government bonds; average ex post risk premiums for 1970–2000 were close to zero.
- Two plausible interpretations emphasized by authors:
  1. The long‑run ex post result reflects realized outcomes worse than anticipated for the 1970s/1980s (mispricing or bad luck), with the 1980s debt crisis producing very negative returns that were only partly offset by the 1990s recovery.
  2. The true long‑run risk premium on emerging‑market debt may be low because emerging‑market debt returns are not strongly correlated with world or U.S. stock returns—hence low required premia in equilibrium.
- Policy‑relevant observations:
  - Formal debt restructuring under international rules (Brady deals) can benefit creditors as well as debtors and contributed to the 1989–1994 rebound in ex post returns.
  - The persistently low long‑run ex post spreads in the Brady group relative to non‑Brady countries remains a puzzle; history and timing of restructurings materially affect long‑run averages over thirty years.

*Source: _wp0413 - IMF Working Paper (excerpts from PDF chapter/section).*

### 1.      Average Duration of EMBIG Debt in Our Sample..........................................14

### 1.      Average Duration of EMBIG Debt in Our Sample..........................................14

### Major Sections
- 1.      Average Duration of EMBIG Debt in Our Sample..........................................14
- 2. Comparison of Assumptions......................................................................................... 17
- 3. Countries with Secondary-Market Prices in 2000: Ex Post Returns 
 to Private Creditors, 1970
–2000 ................................................................................... 19
- 4. Countries with Secondary-Market Prices in 1991: Aggregate Ex Post  
 Returns to Private Creditors, 1970
–2000 ...................................................................... 22
- 5. Countries with Secondary-Market Prices in 2000: Price Sensitivity of Ex Post 
 Sovereign Returns to Private Creditors......................................................................... 23
- 6. The “Eighties” Versus the “Nineties”: Ex Post Spreads, Sovereign Debt,  
         Private         Creditors ........................................................................................................... 26
- 7. Brady Deal Countries: Ex Post Sovereign Spreads up to Brady Deal Year ................. 28
- 8. Decomposing Spreads, 1970
–89, 1989–94, and 1994–2000: Ex Post Sovereign  
 Spreads, Private Creditors............................................................................................. 29
- 9. Spreads over Three-Year Periods:  Ex Post Spreads, Public and Publicly  
 Guaranteed Debt, Private Creditors .............................................................................. 31
- 10.    Average Long-Run Spreads over an Increasing Horizon: Ex Post Spreads,  
 Public and Publicly Guaranteed Debt, Private Creditors.............................................. 33
- 11.    Returns, Volatility, and Betas: Results for Aggregate Flows ....................................... 35

### Appendix Tables (titles and page references)
- 1. Aggregated Secondary-Market Prices, 1986–2001 ...................................................... 45
- 2. Secondary-Market Loan Prices, 1986
–2001 ................................................................. 46
- 3. Secondary-Market Brady Bond Prices, 1986–2001...................................................... 47
- 4. Secondary-Market Non-Brady Bond Prices, 1986
–2001 ............................................. 48
- 5. Representativeness of Prices in All Outstanding Privately Held Debt ......................... 52
- 6. Representativeness of Prices in All Public and Publicly Guaranteed  
 Privately Held Debt....................................................................................................... 53
- 7. Returns, Volatility, and Betas: Results for Individual Countries.................................. 56
- 8. Returns over One-Year Holding Periods ...................................................................... 57

### Figures (titles and page references)
- 1. External Debt of Developing Countries, 1970
–2000.................................................... 16
- 2. Net Transfers on Debt to 22 Emerging Market Countries, 1970
–90 ............................ 18
- 3. Countries with Secondary-Market Prices in 2000: Average Returns and Ex Post 
 Spreads for Public and Publicly Guaranteed Debt, 1970–2000.................................... 20

*Source: _wp0413 - 1.      Average Duration of EMBIG Debt in Our Sample (PDF chapter/section).*

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

### _wp0413 - References.........................................................................................................................

### I. Introduction — research questions and approach
- Central questions:
  - How high are average returns to emerging market lending—particularly sovereign lending—over long periods?
  - How volatile are they over shorter periods?
  - How correlated are they with returns on other assets, such as U.S. equities or U.S. high-yield bonds?
  - After debt crises and restructurings, how much do private creditors actually lose? Ex post, have creditors of countries that did default fared much worse than creditors of other emerging market countries that did not?
  - How have average returns evolved over time—in particular, from the 1980s to the 1990s?
- Data and broad method:
  - Use the World Bank’s Global Development Finance (GDF) database (available annually since 1970) as the main source for debt flows between creditors and debtor countries.
  - Adjust published GDF data to capture all cash flows associated with initial disbursements and vice versa, using available (although generally unpublished) World Bank data.
  - Compute internal rates of return using adjusted flows; outstanding debt stocks—and initial debt stocks when computing returns over subperiods—are valued at secondary-market debt prices.
- Scope and contributions:
  - First paper to study returns on postwar emerging-market debt over a long horizon including the 1980s debt crisis and the 1990s boom-bust cycle.
  - Two alternative approaches to recover flows from World Bank data enhance robustness.
  - Analysis includes evolution of returns over shorter horizons and subperiods; secondary-market price data since 1986 are used to construct average year-end debt prices.
  - Provides rudimentary characterization of volatility of annual returns since the late 1980s and covariance with emerging market equities, U.S. equities, and U.S. high-yield bonds.
  - Results presented for aggregate flows and separately for main debtor countries.

### Key empirical findings (summary)
- Long-run (1970–2000) returns:
  - During 1970–2000, returns averaged about 9 percent per annum, about the same as that on a 10–year U.S. government bond.
  - Ex post risk premiums over the entire 1970–2000 period were close to zero.
  - Most major debtor countries have rates of return clustered in the range between 8 percent and about 10.5 percent.
- Time decomposition and heterogeneity:
  - 1970–2000 decomposes into:
    - 1970–89: cycle with negative or very low returns.
    - 1989–1993: period of very high returns.
    - Post-1993: period of lower but positive ex post spreads.
  - The 1990s boom is driven substantially by developments in the 1989–93 period, which preceded large-scale crisis lending.
- Volatility and correlations (since 1986):
  - Volatility of annual emerging debt returns since 1986 is higher than that of returns on U.S. corporate or high-yield bonds, but far lower than the volatility of emerging-market-equity returns.
  - Emerging-market debt returns, like emerging-market equities (but unlike U.S. corporate and high-yield bonds), exhibited little correlation with the U.S. and world stock markets during this period.

### II. Literature context (concise)
- Historical comparisons:
  - Eichengreen and Portes (1986): interwar sovereign bonds yielded nominal returns near 3.3–5.4 percent, close to risk-free government rates of that era.
  - Lindert and Morton (1990): excess returns on foreign government bonds in 1850–1914, 1915–45 and post-1954 were small (about 0.4 percentage points).
  - These mirror the paper’s finding that excess returns on emerging-market debt over 1970–2000 are small.
- 1980s debt-crisis studies:
  - Lindert (1989), Cohen (1992), Klingen (1994) used World Bank data to compute ex post returns for 1970s–1980s; Klingen found average dollar rates of 8 percent nominal and 2 percent real for 1970–92 privately held debt.
  - Methodological debates center on whether to use reported flows or to impute flows from changes in debt stocks and how to adjust for currency valuation effects, restructurings, and measurement error.
- 1990s asset-class literature:
  - Studies of emerging-market equity and bond indices (e.g., Bekaert and Harvey; EMBI/EMBI+ indices) find higher average returns, higher volatility, and low correlations with developed markets for equities; bond-index studies focus on 1990s liquid traded instruments and differ in scope from this paper.
- Corporate finance parallel:
  - Elton et al. (2001) decompose corporate spreads into expected default loss, tax-related premium, and residual economic risk premium; analogous decomposition is considered here by examining residual spreads after accounting for defaults on emerging market debt.
- BOP/IIP literature:
  - Lane and Milesi-Ferretti (2002, 2003) estimate returns on assets/liabilities for industrial countries; differs from this paper as it focuses only on privately held emerging-market debt.

### III. Methodology and data — measuring returns and net transfers
- Overall approach:
  - Compute internal rates of return from payment flows between private creditors and emerging market debtor countries, applying the investment-project IRR analogy.
  - Two complications:
    - Constructing an appropriate debt flow series from World Bank GDF data due to debt restructuring operations and measurement issues.
    - No final repayment at the end of the sample period; outstanding debt stocks must be valued at secondary-market prices to treat beginning and end stocks as market transactions when computing subperiod returns.
- Debt categories and focus:
  - Focus mainly on broad debt categories to keep cross-category restructuring accounting manageable.
  - Primary focus on privately held long term debt (both public and publicly guaranteed, and private non-guaranteed), and privately held public and publicly guaranteed debt (essentially sovereign debt held by private creditors).
  - Total external debt (which would include short term debt) is not the primary focus because:
    - Virtually all emerging market debt traded in secondary markets has original maturity > 1 year (long term).
    - Except for interest arrears, GDF does not break down short term debt by public/private creditors.
    - World Bank debtor reporting is not considered reliable for short term privately issued debt.
- Measuring “net transfers”:
  - GDF provides net transfers defined as disbursements minus the sum of interest payments and principal repayments for various external debt types and creditor classes.
  - GDF net transfer concept captures standard disbursements and debt service and—except in certain cross-asset restructuring cases—reflects debt restructurings as desired.
  - Important GDF conventions:
    - Cash payments from buybacks are recorded as “principal repayments” and included in net transfers on long term debt.
    - Clearance of interest arrears on long term debt is included in net transfers on long term debt (even though accumulated interest arrears are recorded as short term debt).
- Problem of cross-asset restructurings and required adjustment:
  - When restructurings exchange instruments across asset classes, GDF records a stock operation in one debt category but the counterpart may be in another category or not recorded (e.g., debt-equity swaps), biasing raw net transfers in either direction depending on the category tracked.
  - Examples:
    - Debt-equity swaps: debt stock declines and subsequent debt service declines, but no cash repayment is recorded; from investor perspective receipt of equities constitutes a claim on a new payments stream—ignoring this underestimates return on debt.
    - Consolidation of short term debt into long term debt: short term debt falls without repayment while long term debt increases without disbursement; long term debt flows would overestimate returns on long term debt and short term flows would underestimate returns on short term debt.
    - Exchange of loans for bonds (Brady deals): loan stock and loan debt service decline without repayment while bond stock increases without disbursement; reported net transfers mechanically used by category would underestimate returns on loans and overestimate returns on bonds.
- Construction of modified net transfer series:
  - Define ntr* (from creditor perspective) as:
    - ntr*  =  –ntr + des + x    (1)
    - where:
      - ntr are net transfers from the debtor perspective as recorded in the GDF,
      - des stands for the market value of debt swapped for equity,
      - x is the sum of quasi-repayments (positive sign) and quasi-disbursements (negative sign) attributable to operations involving stock adjustments across debt categories.
  - Interpretation of x:
    - For total debt, x = 0 (only debt-equity swaps need to be added to –ntr).
    - For privately held long term debt, x = – sc, where sc is the increase in the long term debt stock resulting from short term debt consolidation (a quasi-disbursement).
    - For more specific categories, additional components (e.g., conversions of private non-guaranteed debt into public debt) may be relevant.
- Two implementation approaches:
  - Direct approach:
    - Attempt to measure components of x directly (e.g., use World Bank data on debt-equity swaps and short term debt consolidations).
    - For privately held long term debt: ntr* = –ntr + des - sc.
    - World Bank supplied data on debt-equity swaps and short term debt consolidations make the direct approach feasible for broad debt categories.
  - Indirect approach (Lindert-style “back out”):
    - Use identity relating debt stocks to flows and back out x as a residual.
    - Identity used:
      - ∆D = d – r – dfr + ccv + ic – x + u    (2)
      - where:
        - ∆D denotes the measured change in the long term debt stock (in U.S. dollars) for the relevant creditor class,
        - d denotes (recorded) disbursements,
        - r denotes (recorded) principal repayments,
        - dfr stands for “debt forgiveness or reduction” (reduction in face value through operations other than principal repayments),
        - ccv (cross-currency valuation) denotes changes in measured dollar debt stocks due to exchange rate movements between debt denomination currency and the U.S. dollar,
        - ic denotes interest and interest arrears capitalized,
        - x is the placeholder for net cross-category debt restructurings other than debt-equity swaps,
        - u denotes a residual attributable to measurement error in any identity category.
      - All items in (2) except x and u are observable; if u = 0, x can be computed as the residual.
- Choice and reliability considerations:
  - Both direct and indirect approaches are applied for the two broadest debt categories in the paper.
  - In absence of measurement error u, both approaches should give identical results; in practice u ≠ 0 and differences arise for some countries and periods.
  - Long-run rates of return and overall conclusions are robust to the choice of method in this study.
  - Which method is preferable depends on the source of measurement error:
    - If measurement error is mostly in reported net flows d–r and in x as recorded directly, the direct approach may mismeasure ntr* and the indirect method may be preferable.
    - If measurement error primarily affects adjustments dfr, ccv, and ic needed for the indirect approach, the direct method may be preferable.
- Valuation and secondary-market prices:
  - Secondary-market debt prices are used to value outstanding debt stocks and initial stocks when computing subperiod returns.
  - Secondary-market prices are available only for some instruments and countries and not available prior to 1986, requiring assumptions for earlier periods and exposing analyses of subperiods to possible survival bias.

*Source: _wp0413 - References (excerpts provided from the PDF chapter).*

### Section 1 of the published GDF) and adding “net change in interest arrears” (Section 8 of the

### _wp0413 - Section 1 of the published GDF (excerpt)

### Methodology: direct vs. indirect stock-flow reconciliation
- The paper constructs net transfer series using two methods: an “indirect” approach (adds unexplained change in the debt stock to recorded net flows) and a “direct” approach (uses flows as reported). Appendix I explains the formal relationship between the two.
- Systematic sources of residual error identified:
  - World Bank debtor reporting on short-term debt consolidation is unreliable for the 1980s and is not published for that reason.
  - Countries sometimes “discovered” previously unrecorded debt or broadened external debt definitions (e.g., to include state and provincial debt), producing upward stock adjustments with no corresponding recorded disbursement. The indirect method treats these as a “quasi-disbursement” by adding unexplained stock changes to flows.
  - World Bank cross-currency valuation (ccv) series are computed only for public and publicly guaranteed (PPG) debt, not for private non-guaranteed (PNG) debt; differences between end-period and average exchange rates can introduce residuals and errors in net transfers computed via the indirect approach.
- Conclusion on methods: the indirect method is generally more trustworthy but not uniformly superior; results from both methods are presented because the choice can materially affect some country-level and short-period results while not altering overall paper findings.

### Valuing debt stocks and addressing survival bias
- Valuation approach:
  - Privately held long-term debt outstanding (in the final period inclusive of interest arrears) is multiplied by a weighted average of secondary market debt prices, weighted by face value outstanding of each instrument.
  - For mid 1990s–2000: instrument-level prices mainly from J.P. Morgan’s “EMBI Global” bond index; for late 1980s/early 1990s: representative country-specific secondary market loan prices for the pre-Brady period.
- Data limitations and assumptions:
  - No secondary market price data prior to 1986; therefore rates of return cannot be computed for subsamples beginning before 1986. For the 1970–2000 sample, end-1970 debt is valued at par (face value = market value).
  - Average maturity: World Bank data indicates average maturity of emerging market debt in the sample was about 10 years. A large fraction of debt carried floating interest: “about 50 percent in the 1970s, rising to two-thirds in the early 1980s before falling back to about 50 percent in the 1990s.”
  - Effective duration for floating-rate debt is about zero; average effective duration lies between 6 months and 10 years. EMBIG sample average durations observed between 3 and 4.2 years; the authors take the bottom of that range as a best guess for average duration over the whole period.
- Sample coverage and survival bias:
  - End-2000 debt prices available for 22 countries: Algeria, Argentina, Brazil, Chile, Colombia, Côte d'Ivoire, Ecuador, Indonesia, Jordan, Korea, Lebanon, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, Philippines, Thailand, Turkey and Venezuela.
  - The 22-country group accounted for roughly: “around 80 percent in the 1970s, about 70 percent in the mid-1980s, about 60 percent in the early 1990s, rising and currently at about 65 percent” of privately held developing country debt.
  - Long-term debt comprises between “65 and 85 percent” of privately held external debt in these 22 countries; at end-2000 it stood at “79 percent”.
  - To assess survival bias the authors: (a) report results for the 2000 sample without adjustment; (b) compute results for a 1991-defined sample assuming a zero 2000 price for countries with no secondary market price in 2000. The 1991-based assumption yields a lower bound for returns; differences between the 2000 and 1991 samples bound potential survival bias.

### Computing alternative-asset returns (comparators) and duration considerations
- Alternative assets considered include: 3 year U.S. treasury zero coupon bonds, 10 year U.S. treasury zero coupon bonds, U.S. corporate bond index, and in later sections U.S. high-yield bonds and equities.
- Method for comparators: construct a stream of flows that treats disbursements to emerging markets as if they had been invested in the alternative asset (bond held to maturity or one-year instruments via return indices) and compute internal rates of return for these artificial flows (see Appendix III).
- Rationale for 3-year U.S. bond comparator: given estimated average duration, the 3 year U.S. bond is the most appropriate default-risk comparator though the 10 year is also shown because it is a common benchmark.

### Main quantitative results: long-run returns (1970–2000)
- Aggregate long-run finding: average long-run returns to emerging-market private creditors from 1970 to 2000 were about 8–10 percent.
- Relative comparisons (aggregate):
  - Returns were about the same or slightly higher than returns on 3 year U.S. treasury bonds.
  - Returns were about the same or slightly lower than returns on 10 year U.S. treasury bonds.
  - Returns were about “100-200 basis points” lower than the return on U.S. corporate bonds.
- Aggregate numeric examples (aggregate row summaries from Table 3):
  - All countries: actual return 8.4; comparator returns shown include US3y 8.4, US10y 9.1, UScb 10.4 (table presents multiple method variants).
- Country-level:
  - 11 out of 22 countries, accounting for “77 percent” of the face value of debt in 1985 (74 percent in 2000), had average rates of return within 100 basis points of the corresponding 3 year U.S. treasury bond.
  - 15 out of 22 countries, making up “91 percent” of the debt in 1985 and “95 percent” in 2000, had average rates of return within 200 basis points of the corresponding 3 year U.S. treasury bond.
- Robustness checks:
  - Results robust across: (i) method to construct net transfers (direct vs. indirect); (ii) debt concept (all privately held vs. sovereign only); (iii) country sample definition (1991 vs. 2000 sample); (iv) across major debtor countries; (v) valuation of end-stock of debt (sensitivity to using 2001 prices, exclusion of collateralized Brady bonds, uniform price shifts of ±15% and ±25%); and (vi) inclusion of bond-issue fees (sensitivity: assumed fees of 1 percent of disbursement raise average long-run return by about 3-5 basis points).
  - Price-sensitivity examples (Table 5): using 2001 rather than 2000 prices generally made little difference in aggregate returns, but Argentina’s average long-run return fell from “8.8 percent” to “3.7 percent” due to a collapse in Argentine debt prices (88 cents → 29 cents).

### Changes in returns over time and subperiod analysis
- Two boom–bust cycles identified in net flows (Figure 2):
  - First build-up: mid-1970s followed by collapse 1983–late 1980s with trough in 1988 or 1989.
  - Second long boom: 1989–1998, followed by a sharper collapse ending 2000.
- Subperiod decomposition results (Tables 6–10):
  - If the sample is divided at 1989 (the trough), aggregate ex post spreads are highly negative for 1970–89 (e.g., aggregate spreads in the order of -330 to -500 basis points reported for some computations) and strongly positive for 1989–2000.
  - If the sample is divided at 1992 (end of many Brady restructurings), aggregate 1970–92 spreads are negative but much smaller (order of -110 to -180 basis points), with 1992–2000 spreads substantially positive but lower than 1989–2000 averages.
  - A three-way split (1970–1989, 1989–1994, 1994–2000) shows:
    - Exceptional recovery and high returns in 1989–1994 (a rebound coincident with Brady deals).
    - Strong returns during 1994–1997 driven mainly by countries not directly affected by the Asian crisis (1994–1997 yields ex post spreads exceeding 1300 basis points in Brady deal countries for some three-year windows, per Table 9).
    - Returns decline after 1997 (Russian crisis) but remain slightly positive in aggregate.
- Brady-deal country banks:
  - Table 7 (up to Brady deal year): bank spreads (indirect approach) mostly above -300 basis points for major creditors; Mexico and Venezuela had spreads basically zero or positive; Jordan and Nigeria had worse outcomes.

### Volatility, comovement, and asset-allocation implications
- Data and constraints:
  - Annual returns available from secondary market prices sample starting in 1986; maximum of 14 annual observations for volatility and covariance estimates.
- Volatility and betas (Table 11 summary):
  - For τp–2000 (typically 1986/1987–2000) standard deviation of aggregate emerging-market debt returns ~15 (table reports example: St. dev. 15.9 for “All countries”).
  - Emerging-market debt volatility was about as high as U.S. or world stock markets and slightly higher than U.S. high-yield bond returns, but significantly lower—about half—than emerging-market equity volatility.
  - Estimated betas of emerging-market debt returns with respect to U.S. and world stock markets are insignificantly different from zero (sometimes slightly negative). By contrast, U.S. corporate and U.S. high-yield bonds have significant positive betas with respect to the U.S. stock market.
  - Emerging-market debt returns exhibit a positive and significant correlation with emerging-market equity returns; estimated betas with respect to emerging-market equities lie in the “0.15 – 0.3” range.
- Interpretation: the low ex post spreads for emerging-market debt may be consistent with CAPM-like pricing if emerging-market debt returns have low or negative covariance with world or U.S. equity markets; equivalently, equilibrium risk premia could be small if emerging-market debt provides diversification benefits relative to global equities.

### Conclusions and interpretation
- Core finding: investors earned about the same over 1970–2000 on emerging-market private credit as they would have on medium-term U.S. government bonds; average ex post risk premiums for 1970–2000 were close to zero.
- Two plausible interpretations:
  1. The long-run ex post result reflects realized outcomes worse than anticipated for the 1970s/1980s (mispricing or bad luck), with the 1980s debt crisis producing very negative returns that were only partly offset by the 1990s recovery.
  2. The true long-run risk premium on emerging-market debt may be low because emerging-market debt returns are not strongly correlated with world or U.S. stock returns—hence low required premia in equilibrium.
- Role of the 1990s:
  - The 1990s boom is largely concentrated in the early 1990s (1989–1994 recovery), driven by successful Brady restructurings rather than large-scale official bailouts.
  - The 1994–1997 period also produced very high ex post returns for some groups (Brady countries), potentially reflecting an era of liquidity/boom and possibly an investor-friendly official policy stance; however, alternative explanations (e.g., broader financial market boom) are plausible.
- Implications:
  - Formal debt restructuring under international rules (Brady deals) can benefit creditors as well as debtors.
  - The persistently low long-run ex post spreads in the Brady group relative to non-Brady countries remains a puzzle and suggests history matters—long-run averages over thirty years still reflect the impact of the 1980s crisis.

*Source: _wp0413 - Section 1 of the published GDF*

### APPENDIX I

### APPENDIX I

### I. GDF DATA ISSUES

- Purpose: Describe what the “net transfers” concept in the GDF exactly comprises, based on GDF documentation (including Debtor Reporting System questionnaires and instructions) and conversations with Bank staff who maintain the data.

- Context: Since Lindert (1989) there has been debate whether reported “net transfers” (reported disbursements minus repayments minus interest payments) must be adjusted when computing internal rates of return from the creditor perspective.

- Lindert (1989) claims two biases in reported “net transfers”:
  - Exclusion of “involuntary refinancing” (roll-overs in crisis situations) which would omit large shares of gross new lending from disbursement records while debt service on old and new loans may be reported.
  - Ignoring consolidation of short term debt into long term debt, which can cause underestimation of returns in the category written down (short term debt) and overestimation in the category where stocks go up (long term debt).

- Findings on Lindert’s points:
  - Consolidation of short term into long term debt is a real source of bias; it is a cross-category stock operation that affects measured returns.
  - “Involuntary refinancing” is treated by the Bank like a rescheduling: no flows are recorded and no stocks change. Thus, no net-bias arises from involuntary refinancing because neither a repayment nor a disbursement is recorded in that operation; classifying all refinancing as “voluntary” would also leave reported “net transfers” unchanged (repayments and disbursements would cancel).

- Klingen (1994) suggested three additional adjustments to reported net transfers:
  1. Treat debt-equity swaps as repayments.
  2. Treat cash buy-backs as repayments.
  3. Reflect clearance of interest arrears as repayments.
  - Assessment:
    - Only (1) is necessary for the authors' purposes (a special case of accounting for cross-category debt stock changes).
    - Cash payments associated with buy-backs are already captured as repayments in the GDF.
    - The Bank attempts to reflect clearance of interest arrears on long term debt as interest payments on long term debt, even though accumulation of such arrears are counted as part of the short term debt stock.

### Stock-Flow Reconciliations (conceptual identities and interpretation)

- Equation reproduced from text for reference:
  - ∆D ≡ d – r – dfr + ccv + ic + u’      (2)
    - residual u’ defined to include all unobservable items, i.e. u’ = – x + u in earlier notation.
    - when applied to all long term debt, x equals short term debt consolidation into long term debt.

- Stock-flow identity implicit in Section 8 of published GDF country tables:
  - ∆D ≡ nfl + nia - dfr + ccv + ic + u’’      (3)
    - nfl = “Net flows on (total) debt”
    - nia = “Net change in interest arrears”
    - u’’ is the published residual in the GDF stock-flow reconciliation for total debt (short term + long term + IMF debt).

- Key facts about GDF definitions and coverage:
  - The GDF data on cross-currency valuation (ccv) applies exclusively to PPG and IMF. No data on currency composition of either short term debt or PNG long term debt; ccv for these categories is implicitly assumed to be zero.
  - Except for ad hoc adjustments in exceptional cases, GDF “Net flows on debt” are defined as:
    - (Disbursements on long term debt and by the IMF) – (Repayments on long term debt and to the IMF) + Change of the short term debt stock – Net change in interest arrears + Consolidation of short term debt.
    - Because of this definition, consolidation of short term into long term debt leaves “net flows on debt” unaffected: a positive entry in the consolidation term is canceled by a reduction in the short term debt stock.

- Decomposition and algebra (text reproduction and interpretation):
  - Writing (3) in long-hand after decomposing into the four main categories (short term, IMF, PPG, PNG) and reflecting facts above yields identity (4) in the source:
    - (4) TDTDIMFPPGTDTD
      STDIMFIMFPNGPNGPPGPPG
      STDIMFPNGPPGTD
      uicccvccvdfdsrnia
      stcniaDrdrdrd
      DDDDD
      ′′
      ++++−−+
      +−∆+−+−+−
      ≡∆+∆+∆+∆≡∆
      )()()() 
    - This simplifies to identity (5) in the source:
      - (5) TDTDIMFPPGTDTDIMFIMFPNGPNGPPGPPG
        IMFPNGPPG
        uicccvccvdfdsrstcrdrdrd
        DDD
        ′′
        ++++−−+−+−+−
        ≡∆+∆+∆
        )()()(

- Comparison of identities (5) and (2a) (equation (2) applied to public and publicly guaranteed long term (PPG) debt):
  - Two primary differences:
    1. Short term debt consolidation:
       - Published residual u’’ (from identity (5)) is unaffected by short term debt consolidation because the GDF definition of nfl offsets the consolidation with short term stock change.
       - The residual u’ for the PPG-only reconciliation (equation (2a)) will reflect short term debt consolidation because PPG ∆D changes without an offset in measured right-hand-side items.
    2. Cross-currency valuation term:
       - In (5) ccv refers only to sub-items (PPG and IMF) of total debt; it does not capture ccv for PNG debt (implicitly assumed zero).
       - Thus published residual u’’ will contain unaccounted currency valuation effects for PNG debt, whereas the residual in (2a) would not.

- Summary interpretation:
  - The primary economic interpretation of the residual published in the GDF (u’’) is unaccounted currency valuation effects for PNG debt.
  - The primary economic interpretation of the residual backed out for PPG (u’) is short term debt consolidation.
  - Both residuals can also be driven by measurement error or discovery of measurement error (revisions to debt stock without corresponding flows), so the two residuals are expected to be positively correlated — and empirically they are.

- Extension for indirect method including PNG:
  - When equation (2) is applied to both PPG and PNG to implement an “indirect method” for a broader debt concept including PNG, the authors attempt to estimate a ccv term for PNG rather than assuming it zero.
  - They do this by assuming PNG has the same currency composition as PPG for each country.
  - Caveat: This assumption may be incorrect; the residual may include differences between true and assumed ccv for PNG.

### Computation of Cross-Currency Valuation Adjustments (ccv)

- For PPG debt owed to private creditors, the World Bank computes cross-currency valuation adjustments according to the definition (equation (6) in source):

  )(**
  11
  1
  1
  1−−
  −
  −
  −
  −=
  −
  ≡
  tt
  d
  t
  t
  tt
  tt
  eeD
  e
  ee
  Dccv      (6)

  - Notation:
    - t e denotes the end-period exchange rate between the U.S. dollar and the denomination currency (or currency basket) of the loan (i.e. US$ per denomination currency unit).
    - t D is the debt stock outstanding at the end of year t in U.S. dollars.
    - d t D is the debt stock outstanding in denomination currency units.

- Analytical demonstration (text summary):
  - Rewriting t D in terms of denomination currency stock and ccv shows that the ccv definition gives rise to a non-zero residual in the debt stock-flow reconciliation (equation (2)).
  - Substituting into the definition of t u (equation (2)) and assuming dfr, ic and x are zero to simplify algebra yields:
    - u ≡ (r_t – r~_t) – (d_t e_t – d_t e~_t) etc. (text shows derivation; final compact expression:)
    - u ≡ ( ~ e_t – e_t ) (r_t / e_t)  (expressed in source as: )( ~ ) tt t t rd e ee u − − ≡ )
  - Interpretation:
    - The ccv definition (6) produces non-zero residuals to the extent average and end-period exchange rates differ in any year.
    - As a percent of net dollar flows in a year, the residual equals the percentage difference between average and end-period exchange rates in the same period.
    - Residual is likely small if denomination currency is a major currency with limited fluctuation against the dollar.

- Practical implication:
  - Currency valuation adjustments computed with end-period exchange rates versus flows measured using period-average exchange rates naturally generate residual terms in stock-flow reconciliations; these residuals should be interpreted in light of the exchange rate measurement convention.

*Source: APPENDIX I (extracted text).

### Appendix Table 2. Secondary-Market Loan Prices, 1986–2001.

### Appendix Table 2. Secondary-Market Loan Prices, 1986–2001.

### Content and coverage
- Table reports country-level secondary-market loan prices for 1986–2001.  
- Note: "Taken directly from datasources, without weighting."
- Many country-year cells are "n.a." where data are not available.

### Data sources
- Sources listed: LDC Debt Report, J.P. Morgan, Bloomberg.
- Additional source mentions in notes: World Debt Tables (WDT), Bonds, Equities and Loan (BEL) database, bradynet.com, Salomon Brothers (SB), NMB Bank, New York (NMB)/Latin Finance, Financial Flows and the Developing Countries (FFDC).

### Methodology for price construction and aggregation
- Loans: The LDCDR provides one end-of-year loan price per country, used directly.
- Bradys:
  - Whenever Brady or pre-Brady bonds are recorded in the LDCDR, they are weighted by their issue amounts using Bloomberg data.
  - Brady bond prices from the LDCDR were used for the early 1990s for countries with early Brady deals.
- Non‑Brady bonds:
  - LDCDR lists select bonds for early years; World Debt Tables, BEL, Bloomberg, and bradynet.com are used to assign amount issued/outstanding for weighting.
  - If amount issued is unavailable, a simple average of bond issues is used.
- Aggregation rules ("All"):
  - When only one type of issue is available (e.g., only loans or only Bradys), that value is used for the "all" section.
  - When loan and Brady or loan and bond prices exist, a weighted average is taken using the ratio of public and publicly guaranteed commercial bank debt to public and publicly guaranteed bonds reported in the GDF database.
  - When both Brady and bond prices are reported, issue-amount information from Bloomberg, bradynet.com, and World Debt Tables is used as weights.
- On combining datasources:
  - In Brady vs non‑Brady bond aggregation across datasources, amounts outstanding or amount issued are used as weights.
  - In loan and bond aggregation across datasources, the ratio of public and publicly guaranteed bank and bond debt stocks from the GDF database is used.

### J.P. Morgan EMBI data specifics (instrument inclusion and pricing)
- EMBI Global inclusion requirements:
  - available daily prices;
  - a minimum of $500 million outstanding;
  - at least 2 ½ years of remaining maturity to be first included (and at least one year to remain in the index);
  - must be able to settle internationally.
- EMBI Global instruments include: Brady bonds, emerging markets loans, Eurobonds and local market debt instruments issued by sovereign and quasi‑sovereign entities; all denominated in US dollars.
- EMBI Plus instruments: more stringent requirements; typically a subset of EMBI Global.
- For each instrument the authors use: face amount outstanding (FACE OS) and Current Face Price Bid or Close Bid (BID).
- Bid prices are used for consistency with Financial Flows and the Developing Countries, which mostly reports bid prices.

### Bloomberg data handling (1988–2001)
- Bloomberg data on international government bonds (GOVT TK) were collected for all available issues and sorted by Bloomberg market issue.
- Brady bonds and other restructured instruments are identified via Bloomberg category "restructured debt"; remaining issues are treated as "non restructured debt" (new issue bonds).
- Because Bloomberg does not keep a historical record of amount outstanding for each bond, information on amount issued and issue currency is used to compute weights based on issue volume in U.S. dollars. These are merged with Bloomberg historical bid prices to compute weighted annual prices.

### Other data considerations
- Salomon Brothers and NMB data provide one loan price per country; no weighting required.
- FFDC reports secondary market prices for a single instrument per country; used only in rare cases because it chooses a specific issue when restructurings occur.
- On rare occasions, combining a Brady bond, non‑Brady bond, or loan price from different datasources is necessary; weighting rules above are applied.

### Representativeness of price data (summary statements and causes of deviation)
- Representativeness is calculated as total amount issued or amount outstanding used for price aggregation divided by the corresponding debt stock from the GDF, in percent.
- For early years when no issue amounts are available and country-specific representative loan prices were used, the tables show the ratio of GDF public and publicly guaranteed bank loans to the stock of privately held debt or PPG privately held debt, respectively.
- The authors note: "The results are generally reassuring, with most representativeness rates in the 50 to 100 percent range."
- Causes of representativeness rates above 100 percent:
  - using "amount issued" data rather than "amount outstanding" in weighting.
- Causes of very low representativeness rates:
  - use of a price source that provides data on only one benchmark issue (e.g., LDC Debt Report for early bond issues).
- Limitations in JP Morgan data:
  - does not include Euro‑denominated Eurobonds and bonds with less than $500 million outstanding.
  - Proposed remedies: obtain Euro‑EMBIG data and incorporate instruments underlying JP Morgan’s LEI index for smaller issues (Latin American countries).

### Appendix III — Computation of net transfers on alternative investment (method and example)
- Objective: compute IRRs on an alternative investment assuming net disbursements that went into emerging debt markets had instead been used to purchase an alternative asset A.
- Case: alternative asset is a zero coupon U.S. treasury bond of maturity τ ≥ 1. Denote 0 the year first purchased; last purchase year is T. Bonds are held to maturity.
- Net transfers from creditor perspective defined as:
  - A_t = A_{t-τ} + d^A_t − int_{t-τ}^A , 0 ≥ t ≥ +Tτ  (*)  (equation as presented in source)
- Variables:
  - d^A_t are gross disbursements at time t.
  - i_t is the yield to maturity of the bond in period t.
- Constructing flows from net debt disbursements nd_t and debt stock D_t:
  - Initial "disbursement" going to emerging markets is p_0 D_0, and authors set d^A_0 = p_0 D_0, i.e. "000 Dpd_A =". (notation as in source.)
  - Net investments in the alternative asset thereafter are set equal to net disbursements to emerging markets.
  - Using the fact that repayments today equal disbursements 1≥τ periods ago, the relation used is:
    - d^A_t ≡ d_t − d_{t−τ} = D_{t} − D_{t−1} − ... (notation and recursion as given in source)
- Recursive example for τ = 3 (as given in source):
  - 000 D p d^A =  
  - 011 D D d^A − =  
  - 122 D D d^A − =  
  - d^A_0 23 = D + D d_3  (notation follows source presentation)
  - ...  
  - d^A_T = D_T − D_{T−τ} + ...  (recursive construction as in the source)

*Source: Appendix Tables and notes from Appendix II and Appendix III (as provided).*

### APPENDIX III

### APPENDIX III

### Alternative asset: return-indexed net transfers
- Now suppose that the “alternative asset” is an asset class for which a return index in available (such as the U.S. stock market). This index enables us to compute annual returns which reflect capital gains as well as any regular payments. We thus compute net transfers for this alternative index assuming it is held for one year. Let tρ denote the annual return at time t, i.e. the percentage difference between the return index at the end of year t and the end of year t-1. Net transfers in year t are:

  A
  tt
  A
  t
  A
  t
  ddnt
  1
  )1(
  −
  ++−=−ρ , 0≥≥tT, where

- 000
  Dpd
  A
  =,

- A
  ttt
  A
  t
  dDDd
  τ−−
  +−=
  1
  for 0>>tT, and

- 0=
  A
  t
  d for Tt≥

### APPENDIX IV

### Country-level results on volatility: Appendix Table 7 (Returns, Volatility, and Betas)
- Appendix Table 7 reports country-level returns, volatility, and betas (Standard Errors in Parentheses). Examples from the table (as presented):
  - Argentina8.816.419.8198728.2-0.560.290.132.030.690.10
    - (0.58)   (0.55)   (0.23)   (1.11)   (0.69)   (0.07)
    - MSCI returns...17.124.21987111.13.042.711.747.406.96...
    - weighted...314. 583.9(2.19)   (2.04)   (0.75)   (4.46)   (1.92)...
  - Brazil8.211.114.3198762.1-0.68-1.490.32-0.10-0.75-0.23
    - (1.31)   (1.15)   (0.50)   (2.79)   (1.57)   (0.30)
    - MSCI returns...27.321.2198760.92.320.861.180.313.17...
    - weighted...104. 760.8(1.10)   (1.18)   (0.35)   (2.73)   (1.23)...
  - Colombia8.19.36.3199210.50.360.030.190.140.520.06
    - (0.31)   (0.26)   (0.08)   (0.56)   (0.46)   (0.14)
    - MSCI returns......-6.8199230.40.500.110.30-0.761.40...
    - weighted......-16.1(0.98)   (0.76)   (0.31)   (1.61)   (1.34)...
  - Chile10.214.323.4198716.7..................
    - MSCI returns...14.620.9198737.10.550.240.730.911.46...
    - weighted...45.634.5(0.78)   (0.73)   (0.21)   (1.64)   (0.84)...
  - Mexico9.714.415.0198715.1-0.050.130.151.230.470.17
    - (0.32)   (0.30)   (0.12)   (0.57)   (0.36)   (0.07)
    - MSCI returns...13.325.1198750.31.010.871.041.751.81...
    - weighted...53. 473.5(1.03)   (0.96)   (0.27)   (2.19)   (1.16)...
  - Venezuela9.112.412.4199237.10.540.450.311.322.010.37
    - (0.86)   (0.65)   (0.27)   (1.36)   (1.01)   (0.16)
    - MSCI returns......4.0199237.0-0.200.030.27-1.421.37...
    - weighted......13. 8(1.65)   (1.26)   (0.54)   (2.66)   (2.35)...
  - Indonesia10.47.18.3198913.9-0.10-0.300.20-0.55-0.100.13
    - (0.31)   (0.27)   (0.11)   (0.63)   (0.37)   (0.07)
    - MSCI returns...-17.1-15.2198955.41.51-0.031.07-1.60-0.03...
    - weighted...1.62.3(1.15)   (1.15)   (0.37)   (2.53)   (1.47)...
  - Korea, Rep.9.4...6.0199317.5..................
    - MSCI returns...-4.2-0.4199371.73.240.870.73-2.45-2.35...
    - weighted......21.8(2.10)   (1.83)   (1.02)   (3.95)   (3.71)...
  - Malaysia9.98.88.919885.6-0.04-0.120.06-0.210.030.04
    - (0.12)   (0.10)   (0.04)   (0.24)   (0.14)   (0.03)
    - MSCI returns...1.0-4.7198852.11.53-0.011.15-0.780.26...
    - weighted...13. 421.2(1.05)   (1.04)   (0.25)   (2.34)   (1.34)...
  - Philippines6.624.318.5198729.80.900.760.611.921.170.47
    - (0.58)   (0.55)   (0.17)   (1.21)   (0.68)   (0.10)
    - MSCI stock re...2.94.2198751.91.940.671.122.282.36...
    - weighted...62.337.0(0.94)   (1.01)   (0.26)   (2.22)   (1.12)...
  - Thailand10.87.58.1198910.70.07-0.030.060.270.080.11
    - (0.24)   (0.22)   (0.10)   (0.49)   (0.28)   (0.06)
    - MSCI returns...-9.8-11.5198948.21.600.141.050.461.27...
    - weighted...1.3-2.6(0.96)   (1.00)   (0.28)   (2.25)   (1.21)...
  - Turkey9.37.87.8199012.50.590.480.211.020.810.03
    - (0.28)   (0.23)   (0.09)   (0.47)   (0.22)   (0.04)
    - MSCI returns...6.66.61990109.75.161.192.04-2.720.00...
    - weighted...30. 030.0(2.47)   (2.46)   (0.76)   (5.10)   (3.17)...

- Sources for Appendix Table 7: Morgan Stanley, and authors' calculations. Notes: 1/ Gross US dollar returns. Debt returns refer to public and publicly guaranteed debt, private creditors, and average of "direct" and "indirect" approaches. 2/ Based on annual returns.

### Returns over one-year holding periods: Appendix Table 8
- Appendix Table 8 reports one-year holding period returns for countries and asset classes across 1986-87 through 1999-2000. Representative series and selected annual values (as presented):
  - All countries 2/ ... 20.1 14.9 45.9 -16.4 19.7 21.8 5.8 -4.3 18.6 7.2
  - All Brady countries 2/ -20.5 -5.0 15.2 67.4 19.2 17.8 62.8 -23.8 21.5 30.7 9.1 -5.8 18.7 9.9
  - All non-Brady countries 2/ ... 22.1 9.6 15.4 1.8 16.6 5.0 -2.2 -0.6 18.2 1.0
  - Latin America countries 2/ -18.0 -2.8 10.2 70.9 17.2 16.9 61.3 -23.9 21.0 28.5 9.1 -6.4 18.5 10.4
  - Argentina series (1986-87 to 1999-2000): -43.2 -17.9 50.4 72.5 64.1 38.4 10.9 -25.9 27.7 23.7 11.5 4.3 7.1 4.8
    - MSCI stock returns: ... 102.6 81.2 -6.1 405.0 -38.5 58.4 -23.6 12.9 20.3 24.6 -24.3 34.3 -25.1
  - Brazil series (1986-87 to 1999-2000): -22.8 8.2 -19.2 146.1 -34.0 12.3 181.3 -28.0 17.3 30.9 9.1 -17.6 36.8 11.4
    - MSCI stock returns: ... 103.4 38.7 -61.6 172.2 6.7 77.7 65.7 -19.2 42.5 27.3 -39.6 67.2 -11.4
  - Colombia series: -9.1 0.9 23.6 18.8 31.4 10.5 19.9 13.6 21.2 2.0 -3.8 -5.1 16.0 -2.9
    - MSCI stock returns: ... 31.3 21.3 -25.5 11.1 41.8 -42.2 -13.7 -38.9
  - Mexico series (selected): 3.4 1.8 23.5 33.2 34.8 14.2 32.6 -22.2 19.9 23.6 13.7 1.0 15.1 17.2
    - MSCI stock returns: ... 72.0 89.2 62.7 126.0 25.0 49.4 -40.6 -20.4 18.7 53.9 -33.5 80.1 -20.5
  - Venezuela series: -11.1 -16.3 13.9 73.7 35.0 -8.8 34.3 -29.2 39.4 56.4 11.2 -18.8 24.5 11.0
    - MSCI stock returns: ... 20.9 -14.5 -22.7 131.2 28.3 -49.2 8.7 4.5
  - Memorandum: Chile series (selected): 10.8 13.3 37.5 55.4 38.4 8.6 8.7 13.9 ... 11.2
    - MSCI stock returns: ... 40.2 64.0 31.1 116.1 21.9 36.6 44.8 -2.7 -13.5 5.5 -28.5 39.0 -15.1
  - Asian countries 2/ (aggregate): ... 11.3 19.2 10.7 17.3 6.6 7.1 7.2 -6.9 2.9 20.0 5.3
  - Indonesia series (selected): ... 10.4 15.4 10.5 5.5 9.7 11.6 9.6 6.0 -9.4 -23.2 35.4 12.9
    - MSCI stock returns: ... 258.1 81.9 6.4 -44.5 -0.2 105.7 -25.9 9.2 27.5 -74.1 -31.5 93.5 -61.9
  - Malaysia series (selected): ... 6.0 13.8 15.2 15.8 11.7 9.1 -0.6 4.0 0.0 11.7 14.6 5.6
    - MSCI stock returns: ... 27.3 57.0 -7.0 5.8 18.6 109.5 -20.1 4.7 25.5 -68.0 -30.8 114.3 -16.0
  - Philippines series (selected): -19.6 -7.7 82.2 -7.4 65.2 24.1 65.7 -5.1 16.3 25.2 -4.8 11.1 18.2 -6.4
    - MSCI stock returns: ... 43.1 65.0 -45.6 85.5 39.0 122.3 -7.9 -11.5 17.7 -62.6 13.5 3.3 -45.0
  - Thailand series (selected): ... 12.4 21.6 7.3 10.9 2.9 6.0 0.7 -16.4 27.1 9.4 8.8
    - MSCI stock returns: ... 47.3 114.4 -27.3 22.7 35.4 103.6 -9.0 -3.8 -36.6 -73.4 11.6 47.2 -56.3
  - Memorandum: Korea, Rep. (selected): ... -10.0 26.0 -0.2 -27.2 18.0 20.4 7.1
    - MSCI stock returns: ... 96.6 1.7 -27.3 -15.5 1.6 31.4 23.7 -3.3 -38.1 -66.7 141.1 92.4 -49.6
  - Turkey (selected): ... 30.3 4.5 13.0 -5.9 28.0 3.0 2.0 -1.0 16.8 -6.6
    - MSCI stock returns: ... -59.3 547.2 -5.3 -18.5 -44.6 220.5 -50.5 -2.8 36.9 118.1 -52.5 252.4 -45.6

- Market and benchmark series (selected):
  - MSCI U.S. Index 3.9 15.9 31.4 -2.1 31.3 7.4 10.1 2.0 38.2 24.1 34.1 30.7 22.4 -12.5
  - MSCI AC World Stock Return ... 24.0 17.6 -16.5 19.9 -4.2 24.9 5.0 19.5 13.2 15.0 22.0 26.8 -13.9
  - MSCI Emerging Markets Stock ... 40.4 65.0 -10.6 59.9 11.4 74.8 -7.3 -5.2 6.0 -11.6 -25.3 66.4 -30.6
  - 10-Year Treasury Benchmark -3.22 6.46 16.39 6.72 17.25 6.44 11.97 -7.90 23.72 0.08 11.20 13.05 -8.35 14.46
  - ML US Treasury Master 1.98 7.06 14.30 8.65 15.20 7.21 10.63 -3.35 18.45 2.61 9.62 10.03 -2.38 13.37
  - ML US Corporate Master 1.84 9.74 14.07 7.43 18.24 9.12 12.43 -3.36 21.53 3.43 10.39 8.72 -1.89 9.13
  - ML US High Yield Master 4.47 13.36 2.31 -4.36 39.17 17.44 16.69 -1.03 20.46 11.27 13.27 2.95 2.51 -5.12

- Sources for Appendix Table 8: Morgan Stanley, Merrill Lynch, and Authors' calculations.
- Notes: 2/ Refers only to countries shown in Tables 9 and 10, excluding memorandum countries. 1/ Gross returns. Except where otherwise noted, returns refer to public and publicly guaranteed debt, private creditors, average of "direct" and "indirect" approaches.

*Source: _wp0413 - APPENDIX III and APPENDIX IV (tables and notes) as provided in the source content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2004/_wp0413.pdf_