## ch3annex

## Source details

**Canonical URL:** [ch3annex](https://www.imf.org/-/media/files/publications/gfsr/2025/october/english/ch3annex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2025/october/english/ch3annex.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2025/october/english/ch3annex.pdf.json)

---

### Data Description
- Sample: 14 emerging markets (EMs) — Brazil, China, Colombia, Hungary, India, Indonesia, Malaysia, Mexico, Peru, Poland, South Africa, Romania, Thailand, and Türkiye (note: the data sample after 2021m10 for Türkiye has been excluded).
- Period: 2012-2024.
- Investor categories compiled monthly: nonresidents, resident banks, resident nonbank financial institutions (NBFIs). Residual share (other resident groups such as households or government entities) averages 18%.
- Global shock measures:
  - CBOE VIX (implied volatility of US equities) as primary proxy for global investor risk aversion.
  - MOVE index used in robustness checks for fixed income market volatility.
  - Additional tested factors (not in main analysis): USD index for advanced economies, Global Financial Cycle Index (Miranda-Agrippino & Rey, 2020), Gilchrist-Zakrajšek credit spread, Excess Bond Premium, Risk-On/Risk-Off index (Chari, Dilts Stedman, & Lundblad, 2023).
- Market strain proxies:
  - Changes in spreads of zero-coupon local currency government bonds relative to US Treasuries (five-year maturity).
  - Bid-ask spreads of local currency bonds (five-year maturity).
- Macroeconomic controls included: 3-month benchmark rate, real GDP growth, monthly inflation, sovereign credit ratings, local currency government debt-to-GDP.
- Summary statistics:
  - Nonresident share (%) — Obs. 2110, Mean 22, Std. Dev. 12, Min 1, Max 58
  - Resident Bank share (%) — Obs. 2019, Mean 29, Std. Dev. 15, Min 2, Max 76
  - Resident NBFI share (%) — Obs. 1876, Mean 31, Std. Dev. 12, Min 7, Max 56
  - Other Resident Share (%) — Obs. 1876, Mean 18, Std. Dev. 10, Min 0, Max 46
  - ∆ 5y Yield Spread (bps) — Obs. 2132, Mean -1.42, Std. Dev. 36.75, Min -279.91, Max 439.35
  - ∆ 5y Bid-Ask Spread (bps) — Obs. 1976, Mean -0.02, Std. Dev. 1.65, Min -15.92, Max 18.58
  - VIX (%) — Obs. 156, Mean 18, Std. Dev. 6, Min 12 (at p10), Max 25 (at p90)
  - VIX >p75 — Obs. 40, Mean 25, Std. Dev. 7, Min 20 (at p10), Max 30 (at p90)
  - MOVE (%) — Obs. 156, Mean 76, Std. Dev. 25, Min 50 (at p10), Max 116 (at p90)
  - MOVE >p75 — Obs. 40, Mean 113, Std. Dev. 16, Min 92 (at p10), Max 132 (at p90)

### Empirical Design
- Baseline panel regression:
  - Dependent variable: monthly change in country i’s 5-year yield (or bid-ask) spread ∆y_{i,t}.
  - Main regressor: monthly change in VIX ∆VIX_t.
  - Expected sign: β1 > 0 (global shocks increase EM market stress).
- Controls included:
  - ∆% LCB/GDP (change in total local currency bond-to-GDP) — expected positive impact given market depth.
  - ∆ 3M benchmark rate — expected to raise local yields.
  - Lagged real GDP growth — may increase yields by raising expectations of future growth.
  - Lagged inflation — ambiguous expected effect.
  - ∆ sovereign rating — upgrades expected to lower yields.
- Fixed effects and inference:
  - Country fixed effects δ_i included.
  - Driscoll-Kraay standard errors used to allow cross-sectional and temporal dependence.
  - Time fixed effects excluded in baseline to avoid absorbing variation in global risk factors; robustness with year fixed effects tested.
- Investor interaction specifications:
  - Equation (2): adds lagged nonresident share (one-month lag) and interaction ∆VIX * nonresident_share_{i,t-1}.
  - Equation (3): adds lagged resident bank share and resident NBFI share (one-month lags) and their interactions with ∆VIX.
  - Interaction interpretation: negative interaction coefficient → stabilizing role (dampens impact of global shocks); positive → amplifying role.
  - Nonresident and resident shares estimated in separate regressions due to high correlation.
- Dynamics:
  - Panel local projection model (Jordà, 2005) estimated over next 12 months: ∆y_{i,t+h} regressed on contemporaneous ∆VIX_t, interaction ∆VIX_t * investor_share_{i,t-1}, controls, and country fixed effects.
- Reverse causality and lag sensitivity:
  - Reversed regression: changes in yields or bid-ask spreads do not significantly predict changes in VIX (Annex Table 3.1.8).
  - Replaced one-month lag of investor shares with three-month lag; results remain consistent (Annex Table 3.1.9).

### Regression Results — Main Associations
- Full sample (Annex Table 3.1.2, Panel A and B baseline):
  - A 1 percentage point (ppt) increase in the VIX is associated with:
    - 1.93 bps widening in 5y yield spreads (Baseline, Panel A).
    - 0.075 bps widening in 5y bid-ask spreads (Baseline, Panel B).
- Marginal impact of investor shares (per 1ppt increase in share; interaction coefficients):
  - Nonresident share:
    - Yield spreads: amplification of impact by 0.033 bps (interaction coefficient 0.033**).
    - Bid-ask spreads: amplification of impact by 0.001 bps (interaction coefficient 0.001**).
  - Resident bank share:
    - Yield spreads: attenuation of impact by -0.053 bps (interaction coefficient -0.053***).
    - Bid-ask spreads: attenuation of impact by -0.001 bps (interaction coefficient -0.001*).
- Selected control coefficients (Baseline, full sample):
  - ∆% LCB/GDP: -3.065** (Panel A baseline).
  - ∆ 3M rate: 31.583*** (Panel A baseline) and 0.700** (Panel B baseline).
  - Lagged inflation: -0.991** (Panel A baseline).
- Sample sizes:
  - Observations (Panel A baseline): 2,085; Number of groups: 14.
  - Observations (Panel B baseline): 1,929; Number of groups: 13.

### Regression Results — Stress Periods (VIX > 75th percentile)
- Stress subsample (Annex Table 3.1.3):
  - A 1ppt increase in VIX associated with:
    - 2.148*** bps widening in 5y yield spreads (Panel A baseline in stress periods).
    - 0.103*** bps widening in 5y bid-ask spreads (Panel B baseline in stress periods).
  - Interactions in stress:
    - Nonresident interaction: 0.027** (yield spreads) and 0.002*** (bid-ask spreads).
    - Resident bank interaction: -0.039** (yield spreads) and -0.001*** (bid-ask spreads).
  - Observations in stress subsample: 525 (Panel A baseline), Number of groups: 14.

### Regional and Subsample Heterogeneity — Selected Findings
- EM Asia (Annex Table 3.1.4):
  - Baseline ∆VIX coefficients: 1.478*** (yield spreads) and 0.060*** (bid-ask spreads).
  - Nonresident interaction: 0.061*** (yield spreads) and 0.003*** (bid-ask spreads).
  - Resident NBFI interactions: stabilizing effects in some specifications (e.g., ∆VIX * resident NBFI share: -0.078*** for yield spreads in Panel A).
  - Observations: 736 (Panel A baseline for EM Asia), Number of groups: 5.
- High government debt-to-GDP subsample (above median 47%) (Annex Table 3.1.5):
  - Baseline ∆VIX: 2.122*** (yield spreads) and 0.064*** (bid-ask spreads).
  - Nonresident interaction: 0.076*** (yield spreads).
  - Resident bank interaction: -0.130*** (yield spreads) and -0.006*** (bid-ask spreads).
  - Observations: 1,071 (Panel A baseline), Number of groups: 12.
- General heterogeneity conclusion:
  - Countries with weaker credit ratings and higher debt burdens are more vulnerable to global shocks when relying heavily on nonresident financing.
  - A larger resident investor base can provide greater stabilizing effect, potentially with costs (financial repression or increased sovereign-bank linkages).

### Robustness and Additional Tests
- Alternative volatility measure:
  - MOVE index used in robustness checks (Annex Table 3.1.7) — results broadly consistent with VIX-based findings.
- Time fixed effects:
  - Inclusion of year fixed effects in robustness weakens interaction coefficients somewhat but results remain similar to baseline.
- Reverse causality and lag sensitivity:
  - Reverse causality tests (Annex Table 3.1.8) show:
    - ∆5y yield_i,t coefficients on ∆VIX_t regressions: 0.033 (t-statistic 1.617) without controls and 0.032 (t-statistic 1.561) with controls.
    - ∆5y bid-ask_i,t coefficients: 0.591 (t-statistic 1.397) without controls and 0.601 (t-statistic 1.459) with controls.
    - No conventional significance reported, supporting interpretation that EM market stress does not contemporaneously drive VIX.
  - Quarterly lag specifications (Annex Table 3.1.9) using three-month lag of investor shares produce results consistent with one-month lag findings:
    - Panel A baseline ∆VIX_t: 1.930*** (t-statistic 3.935).
    - ∆VIX_t * NMShare_i,t−3: 0.030** (t-statistic 2.092).
    - ∆VIX_t * LMIRShare_i,t−3: -0.058*** (t-statistic -2.836).
    - Panel B baseline ∆VIX_t: 2.148*** (t-statistic 5.101).
    - ∆VIX_t * NMShare_i,t−3: 0.025** (t-statistic 2.156).
    - ∆VIX_t * LMIRShare_i,t−3: -0.047** (t-statistic -2.243).

### Additional Table Highlights (Selected coefficients from Online Annex Tables)
- Online Annex Table 3.1.6 (Full Sample, selected):
  - Panel A Baseline ∆VIX_t: 2.243*** (t-statistic 3.654); ∆VIX_t * NMShare_i,t−1: 0.064*** (t-statistic 3.406); ∆VIX_t * LMIRShare_i,t−1: -0.104*** (t-statistic -4.523).
  - Panel B Baseline ∆VIX_t: 0.083*** (t-statistic 3.328); ∆VIX_t * NMShare_i,t−1: 0.002*** (t-statistic 3.687); ∆VIX_t * LMIRShare_i,t−1: -0.002*** (t-statistic -2.850).
- Online Annex Table 3.1.7 (MOVE robustness, selected):
  - Panel A Baseline ∆MOVE_t: 0.454** (t-statistic 2.510); ∆MOVE_t * NMShare_i,t−1: 0.013** (t-statistic 2.458); ∆MOVE_t * LMIRShare_i,t−1: -0.016*** (t-statistic -2.632).
  - Panel B Baseline ∆MOVE_t: 0.021** (t-statistic 2.214); ∆MOVE_t * NMShare_i,t−1: 0.001*** (t-statistic 3.080); ∆MOVE_t * NLIRVShare_i,t−1: -0.001** (t-statistic -2.003).

### Country Classification of Local Currency Bond Markets — Methodology and Counts
- Sample context:
  - The chapter examines a sample of [55] EMDEs based on IMF classifications; elsewhere the chapter relies on a sample of 56 EMDEs with subsets used where data limitations apply.
- Classification categories and guiding criteria:
  - Major EMs (Emerging Markets) — Total: 12
    - Guiding criteria:
      - 1) Local currency marketable bonds-to-GDP >25 percent;
      - 2) 50 percent of outstanding bonds with size > $1 billion
    - Countries listed: Brazil, China, Colombia, Hungary, India, Indonesia, Malaysia, Mexico, Philippines, Poland, South Africa, Thailand
  - Other EMs (Emerging Markets) — Total: 7
    - Guiding criteria:
      - 1) Local currency marketable bonds-to-GDP >10 percent; and
      - 2) 20 percent of outstanding bonds with size > $1 billion
    - Countries listed: Argentina*, Chile, Dominican Republic, Peru, Romania, Saudi Arabia, Türkiye
  - Frontier Markets — Total: 37
    - Guiding criteria:
      - 1) JPM NEXGEM inclusion;
      - 2) LIC status with outstanding Eurobonds;
      - 3) Countries with * are those not meeting the other criteria but have:
        - a) Local currency marketable bonds-to-GDP >10 percent; and
        - b) 15 percent of outstanding bonds with size > $250 million
    - Countries listed: Algeria*, Angola, Armenia, Azerbaijan, Benin, Botswana, Cameroon, Congo, Rep, Costa Rica, Côte d'Ivoire, Ecuador, Egypt, El Salvador, Ethiopia, Gabon, Georgia, Ghana, Guatemala, Honduras, Jamaica, Jordan, Kazakhstan, Kenya, Mongolia, Morocco, Mozambique, Namibia, Nigeria, Pakistan, Paraguay, Senegal, Sri Lanka, Tunisia, Uganda*, Uzbekistan, Vietnam, Zambia
- Note:
  - The chapter notes use of subsets of countries where data limitations apply and that results should be interpreted with reference to the effective sample available for each analysis.

*Online Annex 3.1 Technical Annex—Measurement and Econometric Approach (Global Financial Stability Report Chapter 3 Annexes).*

### Section 1

### ch3annex - Section 1

### Data Description
- Sample: 14 emerging markets (EMs) — Brazil, China, Colombia, Hungary, India, Indonesia, Malaysia, Mexico, Peru, Poland, South Africa, Romania, Thailand, and Türkiye (note: the data sample after 2021m10 for Türkiye has been excluded).
- Period: 2012-2024.
- Investor categories compiled monthly: nonresidents, resident banks, resident nonbank financial institutions (NBFIs). Residual share (other resident groups such as households or government entities) averages 18%.
- Global shock measures:
  - CBOE VIX (implied volatility of US equities) as primary proxy for global investor risk aversion.
  - MOVE index used in robustness checks for fixed income market volatility.
  - Additional tested factors (not in main analysis): USD index for advanced economies, Global Financial Cycle Index (Miranda-Agrippino & Rey, 2020), Gilchrist-Zakrajšek credit spread, Excess Bond Premium, Risk-On/Risk-Off index (Chari, Dilts Stedman, & Lundblad, 2023).
- Market strain proxies:
  - Changes in spreads of zero-coupon local currency government bonds relative to US Treasuries (five-year maturity).
  - Bid-ask spreads of local currency bonds (five-year maturity).
- Macroeconomic controls included: 3-month benchmark rate, real GDP growth, monthly inflation, sovereign credit ratings, local currency government debt-to-GDP.

- Summary statistics (as reported):
  - Nonresident share (%) — Obs. 2110, Mean 22, Std. Dev. 12, Min 1, Max 58
  - Resident Bank share (%) — Obs. 2019, Mean 29, Std. Dev. 15, Min 2, Max 76
  - Resident NBFI share (%) — Obs. 1876, Mean 31, Std. Dev. 12, Min 7, Max 56
  - Other Resident Share (%) — Obs. 1876, Mean 18, Std. Dev. 10, Min 0, Max 46
  - ∆ 5y Yield Spread (bps) — Obs. 2132, Mean -1.42, Std. Dev. 36.75, Min -279.91, Max 439.35
  - ∆ 5y Bid-Ask Spread (bps) — Obs. 1976, Mean -0.02, Std. Dev. 1.65, Min -15.92, Max 18.58
  - VIX (%) — Obs. 156, Mean 18, Std. Dev. 6, Min 12 (at p10), Max 25 (at p90)
  - VIX >p75 — Obs. 40, Mean 25, Std. Dev. 7, Min 20 (at p10), Max 30 (at p90)
  - MOVE (%) — Obs. 156, Mean 76, Std. Dev. 25, Min 50 (at p10), Max 116 (at p90)
  - MOVE >p75 — Obs. 40, Mean 113, Std. Dev. 16, Min 92 (at p10), Max 132 (at p90)

### Empirical Design
- Baseline specification (panel regression) relates monthly change in country i’s 5-year yield (or bid-ask) spread ∆y_{i,t} to monthly change in VIX ∆VIX_t and controls:
  - Key expected sign: β1 > 0 (global shocks increase EM market stress).
- Control variables:
  - ∆% LCB/ GDP (change in total local currency bond-to-GDP) — expected positive impact given market depth.
  - ∆ 3M benchmark rate — expected to raise local yields.
  - Lagged real GDP growth — may increase yields by raising expectations of future growth.
  - Lagged inflation — ambiguous expected effect.
  - ∆ sovereign rating — upgrades expected to lower yields.
- Fixed effects and inference:
  - Country fixed effects δ_i included; Driscoll-Kraay standard errors used to allow cross-sectional and temporal dependence.
  - Time fixed effects excluded in baseline to avoid absorbing variation in global risk factors (VIX, MOVE); robustness with year fixed effects tested.
- Investor interaction specifications:
  - Equation (2): adds lagged nonresident share (one-month lag) and interaction ∆VIX * nonresident_share_{i,t-1}.
  - Equation (3): adds lagged resident bank share and resident NBFI share (one-month lags) and their interactions with ∆VIX.
  - Interaction interpretation: negative interaction coefficient → stabilizing role (dampens impact of global shocks); positive → amplifying role.
  - Nonresident and resident shares estimated in separate regressions due to high correlation.
- Dynamics:
  - Panel local projection model (Jordà, 2005) estimated over next 12 months to assess persistence: ∆y_{i,t+h} regressed on contemporaneous ∆VIX_t, interaction ∆VIX_t * investor_share_{i,t-1}, controls, and country fixed effects.
- Reverse causality and slow-moving shares checks:
  - Reversed regression: changes in yields or bid-ask spreads do not significantly predict changes in VIX (Annex Table 3.1.8).
  - Replaced one-month lag of investor shares with three-month lag; results remain consistent (Annex Table 3.1.9).

### Regression Results
- Main estimated associations (Annex Table 3.1.2, full sample):
  - A 1 percentage point (ppt) increase in the VIX is associated with:
    - 1.93 bps widening in 5y yield spreads (Baseline, Panel A).
    - 0.075 bps widening in 5y bid-ask spreads (Baseline, Panel B).
  - Marginal impact of investor shares (per 1ppt increase in share):
    - Nonresident share: amplification of impact by 0.033 bps for yield spreads, and 0.001 bps for bid-ask spreads (interaction coefficients: 0.033** and 0.001** respectively).
    - Resident bank share: attenuation of impact by -0.053 bps for yield spreads, and -0.001 bps for bid-ask spreads (interaction coefficients: -0.053*** and -0.001* respectively).
  - Selected control coefficient highlights (Baseline, full sample):
    - ∆% LCB/GDP: -3.065** (Panel A baseline), indicating a negative coefficient in baseline yield regression.
    - ∆ 3M rate: 31.583*** (Panel A baseline) and 0.700** (Panel B baseline).
    - Lagged inflation: -0.991** (Panel A baseline).
  - Sample sizes and groups:
    - Observations (Panel A baseline): 2,085; Number of groups: 14.
    - Observations (Panel B baseline): 1,929; Number of groups: 13.

- Stress periods (VIX > 75th percentile) results (Annex Table 3.1.3):
  - A 1ppt increase in VIX associated with:
    - 2.148*** bps widening in 5y yield spreads (Panel A baseline in stress periods).
    - 0.103*** bps widening in 5y bid-ask spreads (Panel B baseline in stress periods).
  - Nonresident interaction (stress): 0.027** (yield spreads) and 0.002*** (bid-ask spreads).
  - Resident bank interaction (stress): -0.039** (yield spreads) and -0.001*** (bid-ask spreads).
  - Observations in stress subsample: 525 (Panel A baseline), Number of groups: 14.

- Regional and subsample heterogeneity (selected findings):
  - EM Asia (Annex Table 3.1.4):
    - Baseline ∆VIX coefficients: 1.478*** (yield spreads) and 0.060*** (bid-ask spreads).
    - Nonresident interaction: 0.061*** (yield spreads) and 0.003*** (bid-ask spreads).
    - Resident NBFI interactions show stabilizing effects in some specifications (e.g., ∆VIX * resident NBFI share: -0.078*** for yield spreads in Panel A).
    - Observations: 736 (Panel A baseline for EM Asia), Number of groups: 5.
  - High government debt-to-GDP subsample (above median 47%) (Annex Table 3.1.5):
    - Baseline ∆VIX: 2.122*** (yield spreads) and 0.064*** (bid-ask spreads).
    - Nonresident interaction: 0.076*** (yield spreads).
    - Resident bank interaction: -0.130*** (yield spreads) and -0.006*** (bid-ask spreads).
    - Observations: 1,071 (Panel A baseline), Number of groups: 12.
  - General heterogeneity conclusion: Countries with weaker credit ratings and higher debt burdens are more vulnerable to global shocks when relying heavily on nonresident financing; a larger resident investor base can provide greater stabilizing effect, potentially with costs (financial repression or increased sovereign-bank linkages).

### Robustness and Additional Tests
- Alternative volatility measure:
  - MOVE index used in robustness checks (Annex Table 3.1.7) — results broadly consistent with VIX-based findings.
- Time fixed effects:
  - Inclusion of year fixed effects in robustness weakens interaction coefficients somewhat but results remain similar to baseline, implying the exclusion of time fixed effects does not substantially bias estimated associations from interactions.
- Reverse causality and lag sensitivity checks (as above) support interpretation that investor shares are slow moving and that EM market stress does not contemporaneously drive VIX.

*Source: Online Annex 3.1 Technical Annex—Measurement and Econometric Approach (Global Financial Stability Report Chapter 3 Annexes).*

### Section 2

### ch3annex - Section 2

### Impacts of VIX on Local Currency Bond Markets (LCBMs) — Full Sample (Online Annex Table 3.1.6)
- Panel A: 5y Yield Spread to UST (in bps)
  - ∆VIX_t: 2.243*** (t-statistic 3.654) — Baseline
  - ∆VIX_t: 0.958 (t-statistic 1.182) — NonResident
  - ∆VIX_t: 6.198*** (t-statistic 4.904) — Resident
  - NMShare_i,t−1: -0.093 (t-statistic -0.511)
  - ∆VIX_t * NMShare_i,t−1: 0.064*** (t-statistic 3.406)
  - LMIRShare_i,t−1: -0.187 (t-statistic -0.604)
  - ∆VIX_t * LMIRShare_i,t−1: -0.104*** (t-statistic -4.523)
  - NLIRVShare_i,t−1: -0.027 (t-statistic -0.088)
  - ∆VIX_t * NLIRVShare_i,t−1: -0.026 (t-statistic -0.921)
  - Controls: Yes
  - Observations: 1,211 (Baseline), 1,211 (NonResident), 1,099 (Resident)
  - Number of groups: 10

- Panel B: 5y Bid-Ask Spread (in bps)
  - ∆VIX_t: 0.083*** (t-statistic 3.328) — Baseline
  - ∆VIX_t: 0.037* (t-statistic 1.784) — NonResident
  - ∆VIX_t: 0.257*** (t-statistic 6.035) — Resident
  - NMShare_i,t−1: 0.007 (t-statistic 0.941)
  - ∆VIX_t * NMShare_i,t−1: 0.002*** (t-statistic 3.687)
  - LMIRShare_i,t−1: -0.023** (t-statistic -2.068)
  - ∆VIX_t * LMIRShare_i,t−1: -0.002*** (t-statistic -2.850)
  - NLIRVShare_i,t−1: -0.003 (t-statistic -0.252)
  - ∆VIX_t * NLIRVShare_i,t−1: -0.003*** (t-statistic -7.098)
  - Controls: Yes
  - Observations: 1,210 (Baseline), 1,210 (NonResident), 1,098 (Resident)
  - Number of groups: 10

- Notes:
  - t-statistics in parentheses.
  - Significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Impacts of MOVE and investor participation on LCBMs — Full Sample (Online Annex Table 3.1.7)
- Panel A: 5y Yield Spread to UST (in bps)
  - ∆MOVE_t: 0.454** (t-statistic 2.510) — Baseline
  - ∆MOVE_t: 0.201 (t-statistic 1.026) — NonResident
  - ∆MOVE_t: 1.133*** (t-statistic 2.865) — Resident
  - NMShare_i,t−1: -0.137 (t-statistic -1.018)
  - ∆MOVE_t * NMShare_i,t−1: 0.013** (t-statistic 2.458)
  - LMIRShare_i,t−1: 0.044 (t-statistic 0.213)
  - ∆MOVE_t * LMIRShare_i,t−1: -0.016*** (t-statistic -2.632)
  - NLIRVShare_i,t−1: -0.011 (t-statistic -0.066)
  - ∆MOVE_t * NLIRVShare_i,t−1: -0.006 (t-statistic -0.669)
  - Controls: Yes
  - Observations: 2,085 (Baseline), 2,056 (NonResident), 1,863 (Resident)
  - Number of groups: 14

- Panel B: 5y Bid-Ask Spread (in bps)
  - ∆MOVE_t: 0.021** (t-statistic 2.214) — Baseline
  - ∆MOVE_t: 0.005 (t-statistic 0.699) — NonResident
  - ∆MOVE_t: 0.040* (t-statistic 1.785) — Resident
  - NMShare_i,t−1: 0.008* (t-statistic 1.744)
  - ∆MOVE_t * NMShare_i,t−1: 0.001*** (t-statistic 3.080)
  - LMIRShare_i,t−1: -0.013** (t-statistic -2.070)
  - ∆MOVE_t * LMIRShare_i,t−1: -0.000 (t-statistic -0.194)
  - NLIRVShare_i,t−1: -0.005 (t-statistic -0.786)
  - ∆MOVE_t * NLIRVShare_i,t−1: -0.001** (t-statistic -2.003)
  - Controls: Yes
  - Observations: 1,929 (Baseline), 1,929 (NonResident), 1,817 (Resident)
  - Number of groups: 13

- Notes:
  - t-statistics in parentheses.
  - Significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Reverse Causality Test: Impacts of EM LCBMs on VIX (Online Annex Table 3.1.8)
- Dependent variable: ∆VIX_t (in percent)
  - ∆5y yield_i,t:
    - Column (1): 0.033 (t-statistic 1.617) — Controls: NO; Observations: 2,132; Number of groups: 14
    - Column (2): 0.032 (t-statistic 1.561) — Controls: Yes; Observations: 2,085; Number of groups: 14
  - ∆5y bid-ask_i,t:
    - Column (3): 0.591 (t-statistic 1.397) — Controls: NO; Observations: 1,976; Number of groups: 13
    - Column (4): 0.601 (t-statistic 1.459) — Controls: Yes; Observations: 1,929; Number of groups: 13
- Notes:
  - t-statistics in parentheses.
  - No significance markers reported at conventional levels here.

### Impacts of VIX and investor participation on LCBMs — Quarterly Lag (Online Annex Table 3.1.9)
- Panel A: 5y Yield Spread to UST (in bps)
  - ∆VIX_t: 1.930*** (t-statistic 3.935) — Baseline
  - ∆VIX_t: 1.268* (t-statistic 1.820) — NonResident
  - ∆VIX_t: 3.487*** (t-statistic 3.089) — Resident
  - NMShare_i,t−3: -0.088 (t-statistic -0.653)
  - ∆VIX_t * NMShare_i,t−3: 0.030** (t-statistic 2.092)
  - LMIRShare_i,t−3: 0.066 (t-statistic 0.372)
  - ∆VIX_t * LMIRShare_i,t−3: -0.058*** (t-statistic -2.836)
  - NLIRVShare_i,t−3: -0.001 (t-statistic -0.005)
  - ∆VIX_t * NLIRVShare_i,t−3: 0.002 (t-statistic 0.062)
  - Controls: Yes
  - Observations: 2,085 (Baseline), 2,028 (NonResident), 1,833 (Resident)
  - Number of groups: 14

- Panel B: 5y Bid-Ask Spread (in bps)
  - ∆VIX_t: 2.148*** (t-statistic 5.101) — Baseline
  - ∆VIX_t: 1.581*** (t-statistic 2.889) — NonResident
  - ∆VIX_t: 2.707** (t-statistic 2.187) — Resident
  - NMShare_i,t−3: 0.289 (t-statistic 0.730)
  - ∆VIX_t * NMShare_i,t−3: 0.025** (t-statistic 2.156)
  - LMIRShare_i,t−3: -0.543 (t-statistic -0.935)
  - ∆VIX_t * LMIRShare_i,t−3: -0.047** (t-statistic -2.243)
  - NLIRVShare_i,t−3: -0.053 (t-statistic -0.165)
  - ∆VIX_t * NLIRVShare_i,t−3: 0.021 (t-statistic 0.776)
  - Controls: Yes
  - Observations: 525 (Baseline), 523 (NonResident), 488 (Resident)
  - Number of groups: 14

- Notes:
  - t-statistics in parentheses.
  - Significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Country Classification of Local Currency Bond Markets — Methodology and Counts (Online Annex 3.2)
- Sample:
  - The chapter examines a sample of [55] EMDEs based on IMF classifications; elsewhere the chapter relies on a sample of 56 EMDEs with subsets used where data limitations apply.
- Classification categories and guiding criteria:
  - Major EMs (Emerging Markets)
    - Guiding criteria:
      - 1) Local currency marketable bonds-to-GDP >25 percent;
      - 2) 50 percent of outstanding bonds with size > $1 billion
    - Countries listed: Brazil, China, Colombia, Hungary, India, Indonesia, Malaysia, Mexico, Philippines, Poland, South Africa, Thailand
    - Total: 12
  - Other EMs (Emerging Markets)
    - Guiding criteria:
      - 1) Local currency marketable bonds-to-GDP >10 percent; and
      - 2) 20 percent of outstanding bonds with size > $1 billion
    - Countries listed: Argentina*, Chile, Dominican Republic, Peru, Romania, Saudi Arabia, Türkiye
    - Total: 7
  - Frontier Markets
    - Guiding criteria:
      - 1) JPM NEXGEM inclusion;
      - 2) LIC status with outstanding Eurobonds;
      - 3) Countries with * are those not meeting the other criteria but have:
        - a) Local currency marketable bonds-to-GDP >10 percent; and
        - b) 15 percent of outstanding bonds with size > $250 million
    - Countries listed (37 total): Algeria*, Angola, Armenia, Azerbaijan, Benin, Botswana, Cameroon, Congo, Rep, Costa Rica, Côte d'Ivoire, Ecuador, Egypt, El Salvador, Ethiopia, Gabon, Georgia, Ghana, Guatemala, Honduras, Jamaica, Jordan, Kazakhstan, Kenya, Mongolia, Morocco, Mozambique, Namibia, Nigeria, Pakistan, Paraguay, Senegal, Sri Lanka, Tunisia, Uganda*, Uzbekistan, Vietnam, Zambia
    - Total: 37

- Note:
  - The chapter notes use of subsets of countries where data limitations apply and that results should be interpreted with reference to the effective sample available for each analysis.

*Global Financial Stability Report Chapter 3 Annexes — IMF | October 2025*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2025/october/english/ch3annex.pdf_