## 2.1    A Statistical Indicator of Euro Area Financial Fragmentation

## Source details

**Canonical URL:** [2.1    A Statistical Indicator of Euro Area Financial Fragmentation](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf.json)

---

### Overview
- Financial fragmentation defined as breakdown of financial markets into fragments or segments (geographically, by product type, or participant type).
- Focus: the euro area and the link between financial fragmentation and the stability/fragility of sovereign and corporate bond markets.
- Key hypothesis: bond markets are more fragile in the higher-fragmentation regime, reflected by a relatively larger responsiveness of euro area bond market risk premiums to a normalized global risk shock.
- Notable quotation and date in source: "History shows that fragmentation comes not only with high economic cost to growth and jobs, but also a potentially high political one, with loss of public confidence in European institutions" — Financial Times, June 24 2022.
- Clarification: financial fragmentation is not equivalent to geopolitical or geo-economic fragmentation.

### Construction of the fragmentation indicator
- Core idea: use divergences in market dynamics—assets in an integrated market respond homogeneously to common shocks; divergences indicate fragmentation.
- Method:
  - PCA of weekly sovereign bond yield changes for five largest euro area economies: France (FR), Germany (DE), Italy (IT), Spain (ES), Netherlands (NL).
  - PC1 captures co-movements; PC2 captures divergence (core vs periphery).
  - Fragmentation indicator = variance of yield movements explained by PC2 computed over a twelve-month moving window.
- Transformations:
  - Compute first differences for weekly yield-data (∆y). Let ∆y be the T×n matrix of T = 1286 weekly yield-changes for n = 5 jurisdictions.
  - Focus on 5-year maturity (also consider 10-year).
- Interpretation:
  - PC1: components all same sign; PC1 explains around 68 percent of variance for 5-year sovereign yields over full sample.
  - PC2: splits signs (Italy and Spain vs France, Netherlands, Germany); PC2 explains around 20% of variance for sovereign bonds over full sample and around 5% for corporate bond OAS over full sample.
  - During higher-fragmentation weeks PC2 accounts for 31.0% of variance; during lower-fragmentation weeks PC2 explains 11.4% of variance.

### Data and three-step methodology
- Three steps:
  1. Construct fragmentation indicator via PCA on yield changes.
  2. Construct identified global risk shocks via sign-restricted BVAR following Brandt et al. (2021).
  3. Assess fragility via non-linear panel local projections (Jordà, 2023), distinguishing higher and lower fragmentation regimes.
- Data specifics and frequencies:
  - Fragmentation indicator dataset: weekly sovereign and corporate bond yields for FR, DE, IT, ES, NL; sample Jan. 2001 - Sep. 2024 (1286 data points).
  - Identified shocks: daily data for BVAR (Jan. 2008-Sept. 2024, 20012 data points).
  - Local projections: daily sovereign CDS spreads at the 5-year maturity and daily option-adjusted corporate spreads at index-level (Jan. 2008-Sept. 2024, 20012 data points).
- Variables used:
  - Step 1: sovereign yields (maturities: 2y, 5y, 10y) for DE, ES, FR, IT, NL; IG corporate yields (index weighted) for same countries.
  - Step 2 (BVAR): 10y euro OIS rate, 10y US Treasury yield, EURO STOXX index returns, US S&P500 index returns, EUR/USD.
  - Step 3: sovereign CDS spreads (5y) for DE, ES, FR, IT, NL; IG corporate OAS (index weighted) for same countries.
  - Other inputs: VStoxx for euro area volatility; Citi Economic Surprise Index (CESI) for macro conditions.
- Additional country-level macro variables used in heuristic analyses: unemployment, government debt-to-GDP ratios, government deficits, GDP growth (ECB repository); Citi Surprise Index sourced from Bloomberg.

### Identified shocks and empirical strategy
- Global risk shocks identified using sign-restricted BVAR (Brandt et al., 2021) on financial market variables; BVAR also yields EA monetary policy shocks (secondary).
- Identifying variables in BVAR: 10-year euro OIS rate, 10-year US Treasury yield, EURO STOXX and S&P500 returns, EUR/USD.
- Local projections: non-linear panel local projections with a sigmoid smooth transition function G(z t ;θ(κ,c)) = 1 / (1 + exp(−κ(z t − c))); κ set to 10 in main analysis; c is median of fragmentation indicator.
- Control set X t = [EA policy t, EA macro t, US policy t, US macro t] are structural shocks from the BVAR.

### Main empirical findings — sovereign and corporate fragility
- Core finding: euro area financial fragmentation is associated with heightened fragility in bond markets.
- Sovereign CDS (selected quantified effects):
  - Higher fragmentation, global risk shock at h(1): one standard deviation increase in global risk factor → 1.4 percent rise in sovereign CDS premiums (reported coefficient: 1.427, significance ∗∗∗).
  - Fixed effects (FE) panel contemporaneous impact during higher fragmentation estimated at 1.5 percent; LP panel model ≈ 1.4 percent.
  - Statistical significance of regime difference for Global Risk Shock h(1): h(1)+ = 1.427, h(1)− = −0.200, ∆Coeff. = 1.626, SE Diff = 0.319, T-stat = 5.101, P-value = 0.000.
  - Impact largely contemporaneous (absorbed at h(1)); no further significant increases at h(2)−h(3).
- Corporate OAS (selected quantified effects):
  - Higher fragmentation, Global Risk Shock h(1): coefficient 0.551 ∗∗∗ (≈ 0.6 percent increase in OAS for one standard deviation shock).
  - Responses at h(2) and h(3) in higher fragmentation: 0.283 ∗∗∗ and 0.171 ∗∗∗ respectively; cumulative effect by h(3) ≈ 1 percent.
  - Regime difference for Global Risk Shock h(1): h(1)+ = 0.551, h(1)− = 0.114, ∆Coeff. = 0.436, SE Diff = 0.135, T-stat = 3.238, P-value = 0.001.
- Core vs periphery:
  - Under higher fragmentation, contemporaneous sovereign ∆CDS responses: Core = 1.13 percent; Periphery = 1.46 percent.
  - Both core and periphery sovereign CDS widen to global risk shocks; corporate OAS responsiveness appears larger for firms domiciled in core countries than periphery in some specifications.
- Validation via panel regressions:
  - Fixed effects panel: during high-fragmentation episodes, global risk shocks add 0.6 to the change in credit spread (Global Shock + interaction aggregate: 0.615 ∗∗∗ (0.081) reported).
  - Selected coefficient estimates from Table 9 (Model (1) and Model (6) examples preserved):
    - Global Shock: 0.998 ∗∗∗ (0.061) in Model (1); 0.601 ∗∗∗ (0.035) in Model (6).
    - EA Macro: 1.422 ∗∗∗ (0.162) in Model (1); 0.845 ∗∗∗ (0.067) in Model (6).
    - US Macro: 0.204 ∗∗∗ (0.043) in Model (1); 0.112 ∗∗ (0.032) in Model (6).
    - EA Policy: 0.854 ∗∗∗ (0.133) in Model (1); 0.591 ∗∗∗ (0.080) in Model (6).
    - US Policy: 0.065 ∗ (0.025) in Model (1); 0.160 ∗∗ (0.040) in Model (6).

### Local projections specification and identification considerations
- Dependent variable: log-difference of credit risk premium ×100: log(CDS i,t / CDS i,t−1) ×100 ≈ ∆CDS i,t / CDS i,t−1 ×100; analogous for OAS.
- Local projections equation (non-linear with smooth transition G):
  - ∆CDS i,t+h−1 = β0,h + β1,h S t + γh G(z t ;θ) S t + sum_{k=0}^{H} δk,h X t−k + α i + u i,t+h−1.
- Identification and endogeneity concerns addressed:
  1. Reverse causality: inclusion of euro area-specific shocks from the BVAR; Granger causality tests performed (Appendix E).
  2. Omitted variable bias: include EA and US macro and policy shocks from BVAR as controls.
  3. Common confounders: contrast higher vs lower fragmentation regimes and control for regional shocks.
- Serial correlation and diagnostics:
  - ACF diagnostics and Wooldridge test found no significant serial correlation in ∆CDS t in panel setup.
  - Transition smoothness κ set to 10 in main analysis; robustness checks use κ ∈ [1,10].

### Robustness, sensitivity, and uncertainty combination
- PCA robustness:
  - PC1/PC2 decomposition consistent across instruments, maturities, sample periods, and country selections.
  - Bootstrap: 10,000 resamples; PC2 composition (loadings) highly stable; bootstrap-derived indicator trajectories broadly similar.
- Moving-window sensitivity:
  - 12-month moving window chosen as pragmatic balance; 3M yields more volatile series; 24M produces smoother, right-shifted series.
- Kalman filter alternatives:
  - Univariate Kalman and Kalman with Italian-German spread as exogenous input produce smooth dynamics with marginal adjustments.
- Combined estimation uncertainty:
  - Procedure: 500 BVAR identification matrix draws; for each draw re-estimate LP and simulate 20 values uniformly within LP confidence interval to approximate conditional IRF distribution at h(1).
  - Distributional findings: high fragmentation IRF distribution centered well above one; low fragmentation distribution concentrated around zero with lower dispersion.
  - Statistical tests: Kolmogorov–Smirnov D = 0.997, p<0.001; Wilcoxon rank-sum p<0.001 — distributions differ significantly across regimes.
- Sensitivity to κ:
  - κ ∈ [1,10] yields robust conclusions: larger and more persistent impulse responses in higher fragmentation regime across κ values.
- Autocorrelation:
  - Autocorrelation at lag 1 modest and decays quickly; Wooldridge test F = 1.26e5, p-value = 0.9972 — no evidence of first-order serial correlation in panel residuals.
  - Including ∆CDSt−1 in LP specifications yields insignificant coefficient and does not alter main conclusions.

### Heuristics, liquidity, redenomination risk, and fundamental drivers
- Fragmentation vs CDS premiums:
  - Fragmentation indicator captures divergence in yield dynamics independent of levels; examples: 2014–2018 indicator elevated while CDS premiums subdued; GFC credit premiums rose broadly while fragmentation indicator stayed low.
- Re-denomination risk (Appendix G):
  - Market-based measure: spread between euro and US dollar denominated CDS contracts (quanto credit default spread).
  - Quanto-CDS spreads peaked during European sovereign debt crisis and policy-uncertainty episodes (e.g., Italian elections March 2018).
  - Quanto-CDS spreads positively correlated with fragmentation indicator; steeper/convex relationship for Italy and Spain.
  - Caveat: fragmentation can be elevated without redenomination risk and vice versa.
- Liquidity:
  - Liquidity measure from Bloomberg: deviations from a smooth spline fit; higher values → lower liquidity.
  - In lower fragmentation regime, German and Italian liquidity more closely aligned; in higher fragmentation, periphery liquidity evaporates faster.
  - Stress proxies used: VIX (US equity implied volatility) and SMOVE (swaption implied interest rate volatility); both show faster deterioration of Italian liquidity in higher fragmentation regime.
  - Authors opted for VIX (S&P 500 based) over VSTOXX to reduce reverse causality concerns.
- Fundamental drivers and probit regression (correlates of high fragmentation episodes):
  - Random-effects panel probit; McFadden’s pseudo-R2 = 0.291. Significance levels: ***p <0.001, **p <0.01, *p <0.05.
  - Coefficient estimates:
    - Unemployment (%) — Estimate: 0.557; Std. Error: 0.048; t-value: 11.662; p-value: 0.000
    - Government debt-to-GDP (%) — Estimate: 0.019; Std. Error: 0.003; t-value: 6.822; p-value: 0.000
    - Government Deficit (%) — Estimate: 0.172; Std. Error: 0.025; t-value: 6.967; p-value: 0.000
    - Citi Surprise Index — Estimate: 0.003; Std. Error: 0.001; t-value: 4.696; p-value: 0.000
    - Economic Policy Uncertainty — Estimate: 0.002; Std. Error: 0.001; t-value: 2.235; p-value: 0.025
  - Key insight: estimated coefficients unstable across specifications; substantial portion of variation remains unexplained (PseudoR2 = 0.291). Fragmentation not fully explained by fundamentals alone.

### Structural breaks, Granger causality, and regime identification
- Structural breaks:
  - Chow test with dummy equal to one when fragmentation exceeds median: p-value < 2.2e-16; first significant structural break dated 07/05/2010 when fragmentation first exceeds its median.
  - Bai-Perron procedure identifies breakpoints aligning with higher-fragmentation periods.
- Granger causality:
  - Tests for lags 1–5 show bidirectional Granger causality for most countries; global risk shocks and sovereign CDS often influence each other.
  - Italy exception: global shocks primarily Granger-cause Italian CDS (Italy tests for CDS→Global Shock not significant; Global Shock→CDS significant).
- Use as regime identifier:
  - Fragmentation indicator can be used to define higher vs lower fragmentation regimes (median split) and is supported by statistical evidence of structural breaks.

### Policy implications and conclusions
- Policies aimed at reducing financial fragmentation within the euro area could enhance bond market stability.
- The fragmentation–fragility link suggests harmonization and market-integration efforts may foster resilience across sovereign and corporate bond markets.
- Paper treats fragmentation as a latent state variable proxied by divergences in dynamics and stops short of causal attribution; both structural and behavioral drivers acknowledged.

*Source: wpiea2025194-source-pdf - 2.1 A Statistical Indicator of Euro Area Financial Fragmentation (excerpt). Canonical source URL: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf*

### 2.1    A Statistical Indicator of Euro Area Financial Fragmentation  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  

### 2.1    A Statistical Indicator of Euro Area Financial Fragmentation

### Overview
- Financial fragmentation: breakdown of financial markets into fragments or segments (geographically, by product type, or participant type).
- Context: paper focuses on the euro area and the link between financial fragmentation and the stability/fragility of sovereign and corporate bond markets.
- Key hypothesis: bond markets are more fragile in the higher-fragmentation regime, reflected by a relatively larger responsiveness of euro area bond market risk premiums to a normalized global risk shock.
- Notable quotation and date in source: "History shows that fragmentation comes not only with high economic cost to growth and jobs, but also a potentially high political one, with loss of public confidence in European institutions" — Financial Times, June 24 2022.
- Clarification: financial fragmentation is not equivalent to geopolitical or geo-economic fragmentation.

### Construction of the fragmentation indicator
- Core idea: use divergences in market dynamics—assets in an integrated market respond homogeneously to common shocks; divergences indicate fragmentation.
- Method:
  - Principal component analysis (PCA) of weekly sovereign bond yield changes for the five largest euro area economies: France (FR), Germany (DE), Italy (IT), Spain (ES), Netherlands (NL).
  - First principal component (PC1) captures co-movements in sovereign yields across the euro area.
  - Second principal component (PC2) captures a divergence pattern in which yield changes move in opposite directions across two country-blocks (the “core” and “periphery” groupings).
  - Fragmentation indicator = variance of yield movements explained by PC2 computed over a twelve-month moving window.
- Historical behavior:
  - The fragmentation indicator peaks during the European sovereign debt crisis and declines afterwards.
  - The indicator correlates with other proposed measures, such as the quanto-CDS spreads.
- Nomenclature: the paper uses "core" and "periphery" jurisdiction labels to refer to the two country groupings.

### Heuristics and interpretation
- Rationale for focusing on dynamics (first differences) rather than levels:
  - Differences in credit risk premium levels or yield levels can reflect fundamentals; therefore divergences in dynamics are used to proxy latent fragmentation.
- Relationship to existing literature:
  - De Vette and Mosk (2022): the indicator was first introduced as the Dynamics Divergence Indicator (DDI).
  - Fontana and Scheicher (2016), Fabozzi et al. (2016), Candelon et al. (2022) find similar PCA patterns (homogeneous PC1, two-block PC2) but differ in data frequency, variables (CDS vs yields), or use of levels vs dynamics.
  - Kakes and van den End (2023) offer an alternative measure based on higher moments of the sovereign-spread distribution not explained by fundamentals.
- Limits: fragmentation treated as an unobserved (latent) state variable proxied by divergences in dynamics; structural fundamentals do not fully explain the fragmentation indicator (large unexplained variance).

### Data and methodology (three-step approach)
- Three steps:
  1. Construct financial fragmentation indicator via PCA on yield changes.
  2. Construct identified global risk shocks via sign-restricted Bayesian Vector Auto-regression (BVAR) following Brandt et al. (2021).
  3. Assess fragility of bond markets via non-linear panel local projections (Jordà, 2023), distinguishing higher and lower fragmentation regimes using the fragmentation indicator.
- Data specifics:
  - Fragmentation indicator dataset: weekly sovereign and corporate bond yields for FR, DE, IT, ES, NL; sample Jan. 2001 - Sep. 2024 (1286 data points).
  - Identified shocks: daily data used for BVAR (Jan. 2008-Sept. 2024, 20012 data points).
  - Local projections: daily sovereign CDS spreads at the 5-year maturity and daily option-adjusted corporate spreads at index-level (Jan. 2008-Sept. 2024, 20012 data points).
- Variables used (as summarized in the methodology table):
  - Step 1 variables: sovereign yields (maturities: 2y, 5y, 10y) for DE, ES, FR, IT, NL; IG corporate yields (index weighted) for DE, ES, FR, IT, NL.
  - Step 2 variables (BVAR): 10y euro OIS rate, 10y US Treasury yield, EURO STOXX index returns, US S&P500 index returns, EUR/USD.
  - Step 3 variables: sovereign CDS spreads (maturities: 5y) for DE, ES, FR, IT, NL; IG corporate OAS (index weighted) for DE, ES, FR, IT, NL.
  - Other inputs: euro area volatility measured via the VStoxx; macroeconomic conditions proxied by the Citi Economic Surprise Index (CESI).
- Frequency and ranges:
  - Step 1: Weekly, Jan. 2001 - Sep. 2024 (1286 data points).
  - Step 2: Daily, Jan. 2008 - Sept. 2024 (20012 data points).
  - Step 3: Daily, Jan. 2008 - Sept. 2024 (20012 data points).
- Additional country-level macro variables used in heuristic analyses (Section 2.1.2 and 4.3): unemployment, government debt-to-GDP ratios, government deficits, GDP growth (sourced from ECB repository); Citi Surprise Index sourced from Bloomberg.

### Identified shocks and empirical strategy
- Global risk shocks: identified using sign-restricted BVAR following Brandt et al. (2021) on financial market variables.
- BVAR also yields identified euro area monetary policy shocks (secondary interest).
- Local projections: non-linear panel local projections method used to estimate dynamic responses of sovereign and corporate credit spreads to identified shocks, conditional on fragmentation regime (higher vs lower fragmentation).

### Main empirical findings (summary)
- Core finding: euro area financial fragmentation can be associated with heightened fragility in bond markets.
- Specifics:
  - Global risk shocks induce a larger response in sovereign credit risk premiums in the higher-fragmentation regime.
  - Corporate bond spreads (option-adjusted spreads) also react more strongly to global risk shocks in the higher-fragmentation regime.
  - The elevated sensitivity appears across both country-blocks (core and periphery), but is most pronounced in the country-block of more vulnerable jurisdictions.
- Robustness and extensions:
  - Section 4 tests robustness of the fragmentation indicator and establishes confidence intervals that combine uncertainty from multiple estimation steps.
  - Section 4.3 presents a probit regression exploring correlates of fragmentation; variables like debt-to-GDP ratios, fiscal deficits, and unemployment correlate with the fragmentation measure, but a large share of variance remains unexplained and variables like debt-to-GDP have economically insignificant coefficients.
  - Paper does not attempt to fully explain causes of fragmentation; structural and behavioral drivers are acknowledged.

### Policy implications
- Policies aimed at reducing financial fragmentation within the euro area could enhance bond market stability.
- The fragmentation–fragility link suggests harmonization and market-integration efforts may foster resilience across both sovereign and corporate bond markets.

### Paper structure (as provided)
- Section 2: Methodology — construction of PCA-based fragmentation indicator; construction of identified financial market shocks; implementation of local projections; heuristics in Section 2.1.2.
- Section 3: Results — local projections response functions for sovereign CDS premiums and corporate option-adjusted spreads to identified global risk shocks.
- Section 4: Robustness — tests for the fragmentation indicator and combined uncertainty; validation via panel regression.
- Section 5: Conclusions and Discussion.
- Annexes A–H cover definitions, PCA supporting statistics, BVAR shock construction, panel regression, Granger causality, sensitivity to smoothness parameter κ, structural breaks in sovereign fragmentation indicator, and autocorrelation in sovereign CDS spreads.

*Source: IMF working paper chapter 2.1, "A Statistical Indicator of Euro Area Financial Fragmentation" (excerpt). *

### 2.1    A Statistical Indicator of Euro Area Financial Fragmentation

### 2.1    A Statistical Indicator of Euro Area Financial Fragmentation

### Construction of the Fragmentation Indicator
- Sample and data:
  - Sovereign yields for five largest euro area economies: France (FR), Germany (DE), Italy (IT), Spain (ES) and the Netherlands (NL).
  - Weekly data from January 2001 to September 2024.
  - Primary focus on benchmark yields for the 5-year maturity; also consider 10-year benchmark yields.
  - Also apply methodology to weighted option-adjusted spreads (OAS) for investment grade corporate bond indices for the same countries.
- Transformations and method:
  - Compute first differences for weekly yield-data (∆y) to focus on yield dynamics rather than yield levels.
  - Let ∆y be the T×n matrix of T = 1286 weekly yield-changes for the n = 5 jurisdictions.
  - Perform principal component analysis (PCA) on ∆y and focus on the first two principal components, PC1 and PC2.
- Interpretation of principal components:
  - PC1: Co-movements
    - Components of PC1 each have the same sign, reflecting co-movements of yields.
    - Magnitude of individual components varies; larger magnitude for Italian and Spanish yield-changes implies these sovereign yields tend to be more volatile.
  - PC2: Core-periphery divergences
    - Components of PC2 split into two groups by sign: Italy and Spain (periphery) versus France, the Netherlands and Germany (core).
    - PC2 reflects yield-dynamics where core and periphery yields move in opposite directions.
- Fragmentation indicator definition:
  - The indicator is the share of the variance explained by the second principal component, computed over a twelve-month moving window.
  - Rationale: PC2 captures divergent yield-moves and thus asymmetric pricing/fragmentation.

### Key Empirical Findings and Statistics
- Variance explained (full sample, based on weekly yield changes):
  - For 5-year sovereign yields: around 68 percent of the variance is explained by the first principal component.
  - For corporate bond OAS: the first factor accounts for 85% of the variance over the full sample.
  - The second factor explains around 20% of the variance for sovereign bonds and around 5% for corporate bond OAS over the full sample.
- Time-varying behavior (12-month moving window):
  - Before the Global Financial Crisis (GFC) and European Sovereign Debt crisis: the measure shows lower readings, accounting for around 10% of the variance.
  - During the euro area sovereign debt crisis: increases to almost 50% for sovereign yields and 30% for corporate bond OAS.
  - Post-crisis: both sovereign and corporate fragmentation indicators declined gradually; in recent years the bulk of euro area bond yield dynamics can be explained by the common factor (PC1).
- Table 2 — Summary statistics of main variables (sample: daily data for five euro area countries from 2008 to September 2024; CDS and OAS measured in basis points; shocks identified using BVAR model of Brandt et al. (2021); pooled at country level for CDS, OAS and Equity market returns):
  - Sovereign CDS: Mean 0.03, St. Dev. 4.25, Min −67.56, Max 79.45, N 20,012
  - Corporate OAS: Mean 0.01, St. Dev. 1.71, Min −34.12, Max 30.51, N 20,012
  - Equity returns: Mean 0.003, St. Dev. 1.46, Min −18.54, Max 13.48, N 20,012
  - VSTOXX: Mean 23.41, St. Dev. 9.19, Min 10.68, Max 87.51, N 20,012
  - Citi Surprise Index: Mean 1.10, St. Dev. 67.22, Min −304.60, Max 212.40, N 20,012
  - EA Policy Shock: Mean −0.004, St. Dev. 0.83, Min −4.86, Max 5.72, N 20,012
  - EA Macro Shock: Mean 0.02, St. Dev. 0.86, Min −4.71, Max 5.77, N 20,012
  - Global Risk Shock: Mean 0.01, St. Dev. 0.93, Min −7.44, Max 5.37, N 20,012
  - US Macro Shock: Mean −0.02, St. Dev. 0.85, Min −5.31, Max 5.40, N 20,012
  - US Policy Shock: Mean −0.01, St. Dev. 0.84, Min −7.15, Max 4.72, N 20,012

### Robustness, Sensitivities, and Design Choices
- Sensitivity checks and choices:
  - Maturity choice: experiments with 10-year yields show close correspondence with 5-year based measures.
  - Aggregation/averaging windows: shorter windows yield a more volatile measure; longer windows smooth variations. Twelve-month window chosen as a pragmatic balance between stability and responsiveness.
  - Country sample: limited to five largest euro area economies for liquidity and stability of PCA; broader samples (e.g., including Greece, Portugal, Austria, or Ireland) could add granularity but introduce data comparability and estimation stability challenges (missing values, illiquidity, absence of benchmark 5-year yields).
  - Asset choice: sovereign yields chosen because sovereign yields and their divergence during the European Sovereign Debt Crisis are central to euro area financial fragmentation and sovereign bonds trade in relatively liquid markets.
- Cross-asset patterns:
  - Similar PC1/PC2 patterns emerged for option-adjusted spreads for corporate bonds (Figures referenced in source).

### Heuristics, Interpretation, and Relation to Other Measures
- Fragmentation indicator versus sovereign CDS premiums:
  - The indicator captures divergence in yield dynamics independent of yield or spread levels; CDS measures capture market perceptions of credit risk.
  - Example behaviors:
    - 2014–2018: fragmentation indicator remains elevated while CDS premiums are relatively subdued — yield dynamics diverge for reasons other than credit risk levels.
    - Global Financial Crisis: credit premiums increased across the board while the fragmentation indicator remained relatively low, reflecting a common shock affecting all jurisdictions.
  - Conclusion: the fragmentation indicator is not simply a one-for-one reflection of sovereign CDS premiums and captures distinct dimensions of market behavior.
- Fragmentation and investor behavior:
  - Weekly frequency analysis emphasizes the role of external/global shocks affecting investor sentiment and risk perception as key drivers of variance in yield and risk premium dynamics.
  - During higher fragmentation, "flight to safety" dynamics (e.g., investor flows out of periphery into core) play a key role; evidence suggests outflows were driven largely by investment funds while investors domiciled in periphery jurisdictions absorbed sales (Longaric et al. (2025) as cited).
  - The extent to which variations in risk sentiment translate into PC2-like dynamics (as opposed to PC1-like dynamics) is what the indicator captures.
- Limits and caveats:
  - The indicator is an empirical construct relying on discrete choices: moving window length, moving average, jurisdiction selection, maturity selection, and choice of sovereign yields versus spreads or CDS premiums.
  - Results are descriptive and highlight correlations; causal relationships are not investigated here.

### Structural Breaks and Regime Identification
- Structural break testing:
  - Applied Chow break test using a dummy equal to one when fragmentation exceeds its median to assess structural breaks.
  - Chow test result: p-value < 2.2e-16, indicating a significant structural break on 07/05/2010 when the fragmentation indicator is above its median for the first time in the sample.
  - Additional analysis using the Bai-Perron procedure supports these findings; identified breakpoints align closely with periods classified as higher fragmentation regimes.
- Use as regime identifier:
  - The fragmentation indicator can be used to identify regimes in which sovereign debt markets are more fragmented based on statistical evidence of structural breaks.

*Source: wpiea2025194-source-pdf - 2.1    A Statistical Indicator of Euro Area Financial Fragmentation*

### Appendix G).

### Appendix G)

### Fragmentation and Re-denomination Risk
- Concept:
  - Re-denomination risk is defined as the risk that assets would be re-denominated into post-break-up currencies in the event of a break-up of the monetary union.
  - The paper constructs a market-based measure of re-denomination risk as the spread between euro and US dollar denominated CDS contracts (the quanto credit default spread), following De Santis (2015, 2019).
  - A higher quanto-CDS spread signals an additional premium for the euro-denominated contract on the same underlying asset, reflecting the risk and cost of re-denomination.
- Empirical observations:
  - Historically, the quanto CDS spread peaked during the European sovereign debt crisis and during periods of policy uncertainty (example given: the Italian elections of March 2018).
  - Ex-ante expectation: the financial fragmentation indicator should positively correlate with re-denomination risk for all jurisdictions, particularly for periphery countries such as Italy and Spain.
  - During systemic episodes (e.g., the euro area sovereign debt crisis), fragmentation and re-denomination risk tend to move in tandem.
- Statistical/visual evidence:
  - Quanto-CDS spreads are positively correlated with the fragmentation indicator; this relationship is reflected by upward-sloping curves in Figure 6.
  - The steepness of the curves is higher for periphery jurisdictions — Italy and Spain — and shows some positive convexity.
- Interpretive caveats:
  - Fragmentation can be elevated without redenomination risk (e.g., due to uneven liquidity, market depth, investor base, or regulatory differences).
  - Redenomination risk can increase without broader fragmentation if concerns are concentrated in a single jurisdiction (e.g., investor doubts about one member state’s political commitment to the euro).

### Fragmentation and Liquidity
- Liquidity measure:
  - Analysis uses a liquidity measure from Bloomberg L.P., which reflects sovereign bond yields’ deviations from a smooth spline fit. Under this measure, higher values point to lower liquidity.
  - Rationale: deviations from a smooth yield-curve fit point to arbitrage opportunities; in a very liquid market, traders remove such arbitrage opportunities quickly.
- Regimes:
  - Higher- and lower-fragmentation regimes are defined by splitting the sample at the fiftieth percentile of the fragmentation indicator.
- Empirical findings:
  - Figure 7(a): In the lower fragmentation regime, German sovereign bond liquidity is more closely aligned with Italian sovereign bond liquidity than in the higher fragmentation regime.
    - Implication: in the higher fragmentation regime, liquidity may evaporate much more quickly in “periphery countries” compared to “core jurisdictions.”
  - Figures 7(b) and 7(c): For the Italian sovereign market, using stress proxies:
    - VIX US Equity Implied Volatility Index (US VIX implied volatility index) as one stress proxy.
    - Swaption Implied Interest Rate Volatility Index (SMOVE) as another stress proxy.
    - For both stress measures, a faster deterioration of liquidity conditions is observed in the higher fragmentation regime.
- Notes on stress proxies:
  - The VIX Index is based on S&P 500 Index (SPX) option prices.
  - SMOVE (Swaption Merrill Option Volatility Estimate) is a yield-curve weighted index of the normalized swaption-implied volatility, reflecting implied-volatilities for 2-year, 5-year, 10-year, and 30-year interest rate swaps.
  - The authors note the option to consider VSTOXX for European stocks; they opted for VIX to reduce reverse causality concerns.

### Fundamental Drivers of Fragmentation
- Stance and approach:
  - The paper does not attempt to fully explain causes of euro area financial fragmentation.
  - It treats financial fragmentation as a latent state variable proxied through dynamics of the euro area financial system.
- Potential drivers (discussed, not fully explanatory):
  - Structural factors: differences in policies and regulations.
  - Other drivers: concerns around debt sustainability, sovereign credit risk, euro area break-up risk.
  - Amplifiers: self-reinforcing dynamics and behavioral biases that may act non-linearly.
- Probit regression (panel) results (Table 3) — correlates of high fragmentation episodes:
  - Regression type: random-effects panel probit estimating probability of high financial fragmentation episodes; dependent variable is a dummy equal to one when higher fragmentation equals one. Panel includes DE, NL, FR, IT and ES. Model includes an intercept. Standard errors are robust. McFadden’s pseudo-R2 = 0.291. Significance levels: ***p <0.001, **p <0.01, *p <0.05.
  - Coefficient estimates and statistics:
    - Unemployment (%) — Estimate: 0.557; Std. Error: 0.048; t-value: 11.662; p-value: 0.000
    - Government debt-to-GDP (%) — Estimate: 0.019; Std. Error: 0.003; t-value: 6.822; p-value: 0.000
    - Government Deficit (%) — Estimate: 0.172; Std. Error: 0.025; t-value: 6.967; p-value: 0.000
    - Citi Surprise Index — Estimate: 0.003; Std. Error: 0.001; t-value: 4.696; p-value: 0.000
    - Economic Policy Uncertainty — Estimate: 0.002; Std. Error: 0.001; t-value: 2.235; p-value: 0.025
- Key insights and limitations:
  - Estimated coefficients are not robust to changes in model specification, indicating some instability.
  - Even with a comprehensive set of explanatory variables, a substantial portion of variation in the fragmentation indicator remains unexplained (PseudoR2 = 0.291).
  - Conclusion: financial fragmentation in the euro area cannot be fully explained by fundamental variables alone and may be shaped by nonlinear dynamics, feedback loops, and behavioral mechanisms beyond traditional macro indicators.

### Construction of Identified Market Shocks
- Purpose:
  - Assess fragility of bond markets by analyzing responses of sovereign CDS premiums and corporate OAS to identified shocks.
- Methodology summary:
  - Follows Brandt et al. (2021): asset price movements in a sign-restricted Bayesian vector auto-regression (BVAR) are used to derive the extent to which euro area rates and US yields, equity prices, and the EUR/USD exchange rate are jointly driven by monetary policy, macro-news, and global risk factors.
  - The identified factors correlate with the VIX Index (global risk factor), and euro area and US Citi surprise indices (EA macro and US macro factor).
  - Data used: daily data for the 10-year euro OIS rate, the EURO STOXX and S&P 500 equity index returns, the EUR/USD exchange rate, and the 10-year US Treasury yield for the sample period.
- Relevance:
  - The global risk shock is most relevant for the paper’s analysis.
  - Previous literature cited: sovereign CDS premiums are largely driven by a global risk factor (Srivastava et al., 2016; Longstaff et al., 2011; Corzo et al., 2020).
  - Increased sensitivity of CDS pricing to the global risk factor may imply markets are more prone to flight-to-safety or disorderly dynamics, impairing financial stability (De Santis (2014)).
- Identification sign restrictions (as used by Brandt et al. (2021); reproduced in the paper):
  - Table of sign restrictions by region/shock type (rows) and variables (columns). Summary of signs:
    - Favorable Global risk: EA long-term yields +++++ ; EA equity prices -++ ; US equity prices -++ ; USD/EUR exchange rate ++--+ ; EA-US long-term spread ++---
    - Restrictive monetary policy (EA and US) and Favorable macro news entries are specified in the table; the table is used to identify shocks via sign restrictions.
- Implementation:
  - Methodology implemented using the BEAR-toolbox of Dieppe et al. (2016).
  - Further details and visualizations are indicated as available in Appendix C.

*Source: Appendix G), wpiea2025194-source-pdf - Appendix G).*

### 2.3    Assessing Market Fragility with a Local Projections Method

### 2.3    Assessing Market Fragility with a Local Projections Method

### Methodology: dependent variable and model specification
- Dependent variable:
  - Log-difference of the credit risk premium (×100): log(CDS i,t / CDS i,t−1) ×100 ≈ ∆CDS i,t / CDS i,t−1 ×100.  
  - For corporate bonds, an analogous construct based on option-adjusted spreads (OAS) is used.
- Rationale:
  - Using the log-difference absorbs any linear relationship between the level of the credit risk premium (CDS t−1) and its response (∆CDS t), thereby controlling for factors (e.g., debt/GDP ratio, primary and overall deficits, economic growth) that affect responsiveness and are reflected in CDS levels.
  - Implicit assumption: nonlinear contributions from credit fundamentals to responsiveness are small or negligible.
- Dynamic panel local projections model (non-linear with smooth transition function):
  - Estimated equation:
    - ∆CDS i,t+h−1 = β0,h + β1,h S t + γh G(z t ;θ) S t + sum_{k=0}^{H} δk,h X t−k + α i + u i,t+h−1.
  - Notation and components:
    - i indexes countries, t represents time, h represents local projections horizon.
    - S t is the shock variable (e.g., global risk) at time t.
    - X t = [EA policy t, EA macro t, US policy t, US macro t] are exogenous controls (structural shocks from a sign-restricted BVAR following Brandt et al. (2021)).
    - G(z t ;θ) is a sigmoid smooth transition function: G(z t ;θ(κ,c)) = 1 / (1 + exp(−κ(z t − c))), with θ = (κ,c).
    - κ governs smoothness of the transition; κ is set to 10 in the main analysis and robustness checks use κ between 1 and 10.
    - c reflects the 50th percentile (median) of the fragmentation indicator; G allows a continuous change in shock impact across fragmentation regimes.
  - Nonlinear impulse responses computed following Adämmer (2019), so responses depend on shock size/direction and degree of fragmentation.

### Identification and endogeneity considerations
- Main identification: shocks S t are identified and orthogonal but not strictly exogenous; three endogeneity concerns discussed:
  1. Reverse causality: a move in euro area CDS premiums could become globally systemic and drive global risk shocks.
     - Mitigation: inclusion of euro area-specific shocks (orthogonal to the global risk shock) from the BVAR; Granger causality tests assessed whether peripheral sovereign CDS cause global risk shocks (details in appendix E).
  2. Omitted variable bias: important control factors may be missing.
     - Mitigation: specification includes EA and US macroeconomic and monetary policy shocks from the BVAR (see section C) as controls.
  3. Common confounders/new information affecting both global risk shocks and fundamentals:
     - Focus is on differences between higher and lower fragmentation regimes (elasticity of CDS response to global shocks across regimes).
     - Augmenting local projections with US and EU macro and policy shocks helps isolate regional influences and reduces conflation with domestic conditions.

### Controls and diagnostics
- Exogenous controls: EA policy, EA macro, US policy, US macro (structural shocks derived via sign-restricted BVAR).
- Serial correlation tests:
  - ACF diagnostics and a Wooldridge test found no significant evidence of serial correlation in ∆CDS t in the panel setup.
  - Annex D includes a model with ∆CDS t included; the coefficient is not significant. Annex H provides further autocorrelation test details.
- Transition parameter choices:
  - κ set to 10 in main analysis (relatively sharp transition); results robust to κ between 1 and 10.

### Results — general setup and interpretation
- Standardized global risk shock:
  - Introduced at first period h(1).
  - Magnitude normalized to one standard-deviation; magnitude is consistent across analyses.
  - Dependent variable is log-difference of credit risk premium (∆ CDS) ×100 — approximately the percentage change.
- Local projections reported at horizons h = 1−3 in tables and plotted through h(1)−h(15) in impulse response figures.
- Fragmentation regimes:
  - “+” indicates higher fragmentation (above median of fragmentation indicator).
  - “-” indicates lower fragmentation (below median).

### Results — Sovereign credit risk (sovereign CDS)
- Main finding:
  - Higher fragmentation regime is associated with increased sensitivity of euro area sovereign CDS premiums to global risk shocks.
- Key quantified effects:
  - Higher fragmentation, global risk shock at h(1): a one standard deviation increase in the global risk factor leads to a 1.4 percent rise in sovereign CDS premiums (reported coefficient: 1.427, significance ∗∗∗).
  - Fixed effects (FE) panel model contemporaneous impact during higher fragmentation estimated at 1.5 percent; LP panel model ≈ 1.4 percent.
- Dynamics:
  - Impact appears fully absorbed at h(1); no further significant increases at h(2)−h(3); initial impact is fairly persistent.
- Control factor effects (higher fragmentation regime):
  - Euro area macro-news factor: positive and significant (e.g., coefficient 1.977 ∗∗∗ at h(1) in Table 5).
  - Euro area monetary policy factor: positive and significant (e.g., coefficient 1.033 ∗∗∗ at h(1) in Table 5).
  - US macro-news factor: positive and significant with some lag (e.g., 0.347 ∗∗ at h(1) and 0.549 ∗∗∗ at h(2) in Table 5).
- Lower fragmentation regime:
  - Responsiveness to a standardized global risk shock is significantly lower; immediate reaction at h(1) is no longer observed.
  - Sovereign CDS premiums less sensitive to macroeconomic news factors during lower fragmentation.
  - Monetary policy tightening shocks become statistically insignificant in low-fragmentation regime.
- Statistical significance of regime differences (Table 7):
  - Global Risk Shock h(1): h(1)+ = 1.427, h(1)− = −0.200, ∆Coeff. = 1.626, SE Diff = 0.319, T-stat = 5.101, P-value = 0.000.
  - Global Risk Shock h(2): h(2)+ = 0.014, h(2)− = 0.801, ∆Coeff. = −0.788, SE Diff = 0.352, T-stat = −2.235, P-value = 0.025.
  - Global Risk Shock h(3): h(3)+ = −0.207, h(3)− = 0.430, ∆Coeff. = −0.637, SE Diff = 0.353, T-stat = −1.807, P-value = 0.071.

### Results — Corporate credit risk (investment grade OAS)
- Main finding:
  - Fragmentation increases sensitivity of corporate credit risk premiums to global risk shocks; higher fragmentation regime shows significantly higher sensitivity.
- Key quantified effects (Table 6):
  - Global Risk Shock h(1), higher fragmentation: coefficient 0.551 ∗∗∗ (i.e., OAS increases by around 0.6 percent for a one standard deviation global risk shock).
  - Responses at h(2) and h(3) in higher fragmentation: 0.283 ∗∗∗ and 0.171 ∗∗∗ respectively; cumulative effect by h(3) ≈ 1 percent.
- Dynamics and interpretation:
  - Corporate OAS responses are not limited to first period; significant responses persist at h(2)−h(3), possibly due to lower market liquidity and slower price adjustment in corporate debt markets.
- Control factor effects (higher fragmentation regime):
  - Euro area macro and monetary policy factors significantly increase corporate risk premiums (e.g., euro area macro +0.654 ∗∗∗ at h(1); euro area pol +0.419 ∗∗∗ at h(1) in Table 6).
  - US macro and US policy show positive and sometimes significant effects (e.g., US macro +0.268 ∗∗∗ at h(1); US policy 0.173 ∗∗∗ at h(2)).
- Statistical significance of regime differences for corporate OAS (Table 8):
  - Global Risk Shock h(1): h(1)+ = 0.551, h(1)− = 0.114, ∆Coeff. = 0.436, SE Diff = 0.135, T-stat = 3.238, P-value = 0.001.
  - Global Risk Shock h(2): h(2)+ = 0.283, h(2)− = −0.149, ∆Coeff. = 0.432, SE Diff = 0.122, T-stat = 3.543, P-value = 0.000.
  - Global Risk Shock h(3): h(3)+ = 0.171, h(3)− = −0.002, ∆Coeff. = 0.173, SE Diff = 0.112, T-stat = 1.543, P-value = 0.123.

### Robustness and consistency with literature
- Robustness:
  - Results are robust to alternative κ values between 1 and 10 (gradual to sharp transitions).
  - Local projection results are consistent with fixed effects panel model results reported in Annex D.
- Consistency with literature:
  - Findings align qualitatively with existing literature identifying global risk factor as a key determinant of sovereign CDS premiums (Longstaff et al., 2011; Pan and Singleton, 2008; Augustin, 2018).
  - Corporate market findings relate to prior work on corporate credit dynamics (Zaghini, 2016; Bedendo and Colla, 2015).

*Source: IMF — 2.3 Assessing Market Fragility with a Local Projections Method (wpiea2025194-source-pdf).*

### 3.3    Core and periphery credit risk

### 3.3    Core and periphery credit risk

### Sample split and methodology
- Sample split into "core" and "periphery" jurisdictions:
  - Core countries: France, Germany and The Netherlands.
  - Periphery countries: Italy and Spain.
- Analysis re-estimates the non-linear panel local projections for each sub-sample to assess whether main-sample results are driven by periphery jurisdictions.

### Sovereign CDS: contemporaneous responses and comparisons
- Under higher fragmentation regime:
  - Global risk shock impact on sovereign ∆CDS is positive and significant for both groups.
  - Core countries: 1.13 percent.
  - Periphery countries: 1.46 percent.
- Under lower fragmentation regime:
  - Difference between core and periphery in response to a global risk shock is 0.12 percent.
- Additional observations:
  - Sovereign CDS spreads widen for both core and periphery in response to a global risk shock, even for "safe" core assets.
  - Liquidity premiums can dominate during flight-to-safety episodes, partially explaining widening CDS spreads for core sovereigns despite safe-asset demand dynamics.
  - Widening credit risk premiums for both groups suggests spillovers among euro area countries.

### Corporate investment-grade OAS: region and fragmentation effects
- Global risk shock has a less pronounced impact on corporate investment grade ∆OAS at h(1) in general.
- Under higher fragmentation, responsiveness of corporate ∆OAS is larger, although estimates feature wider confidence intervals.
- Impact appears more pronounced for bonds issued by firms domiciled in core countries than for issuers domiciled in periphery countries.

### Robustness and uncertainty in indicator construction
- Fragmentation indicator is PCA-based on weekly 5-year yield changes and treated as a latent state variable.
- Bootstrap analysis:
  - 10,000 samples drawn with replacement from 1,287 weekly observations were used.
  - The second principal component consistently captures core–periphery divergences.
  - Bootstrap-derived indicator trajectories are broadly similar; confidence bands wider during lower indicator values but overall uncertainty appears limited for regime distinction.
- Aggregation window sensitivity:
  - Moving-average windows compared include 3 months, 6M, 12M, and 24M.
  - Shorter windows (e.g., 3 months) yield a more volatile, responsive indicator; longer windows produce smoother, right-shifted series.
  - Authors adopt a 12-month moving-average as baseline for its balance of responsiveness, stability, transparency, and replicability.
- Kalman filter alternatives:
  - Two specifications examined: univariate Kalman filter (no exogenous inputs) and Kalman filter with Italian-German yield spread as exogenous variable.
  - Both produce smooth dynamics and avoid right-shift bias; including the yield spread leads to only marginal adjustments.
  - Concerns noted about assuming indicator evolves independently (univariate Kalman) and about selection and functional form of exogenous variables.

### Combined estimation uncertainty (BVAR identification + LP estimation)
- Procedure to combine uncertainty:
  - 500 BVAR identification matrix draws were used.
  - For each BVAR draw, the LP model was re-estimated and 20 values were simulated uniformly within the reported LP confidence interval to approximate the full conditional distribution of the IRF at h(1).
- Distributional findings:
  - High fragmentation: IRF distribution centered well above one; global risk shocks trigger a significantly larger increase in sovereign CDS spreads.
  - Low fragmentation: IRF distribution concentrated around zero with lower dispersion.
- Statistical tests:
  - Kolmogorov–Smirnov test: D = 0.997, p<0.001, strongly rejects equality of distributions across regimes.
  - Wilcoxon rank-sum test: distributions differ significantly, rejecting equal medians (p <0.001).

### Validation via panel regressions
- Fixed effects panel regressions (as robustness check) examine impact of BVAR-identified shocks on sovereign ∆CDS.
- Specifications include interaction terms with a fragmentation dummy, macro and policy controls, lagged dependent variable, sovereign debt crisis dummy, and time-fixed effects.
- Key panel regression result:
  - During high-fragmentation episodes, global risk shocks exert a significantly stronger marginal effect, contributing an additional 0.6 percentage points to the change in CDS spreads.
- Conclusion from validation: panel regression evidence supports the local projections findings and underscores the amplifying role of financial fragmentation.

### Summary of substantive findings
- Both sovereign and corporate credit risk premiums are more reactive to global risk shocks during episodes of elevated euro area financial fragmentation.
- This amplified responsiveness holds for both core and periphery jurisdictions, implying fragmentation affects market stability across the single monetary union rather than being confined to the periphery.
- Corporate option-adjusted spreads show larger and more persistent responses under elevated fragmentation, with a more pronounced effect for firms domiciled in core countries.

*Source: 3.3 Core and periphery credit risk (excerpt from wpiea2025194-source-pdf).*

### References

### References

### Methodology and econometrics
- Adämmer, Philipp, “lpirfs:  An R package to estimate impulse response functions by local projections,” The R Journal (2019), 2019,11(2), 421–438.
- Bai, Jushan and Pierre Perron, “Computation and analysis of multiple structural change models,” Journal of applied econometrics, 2003,18(1), 1–22.
- Chow, Gregory C, “Tests of equality between sets of coefficients in two linear regressions,” Econometrica: Journal of the Econometric Society, 1960, pp. 591–605.
- Dieppe, Alistair, Romain Legrand, and Björn Van Roye, “The BEAR toolbox,” 2016.

### Sovereign spreads, CDS, and credit risk in the Euro area
- Attinasi, Maria-Grazia, Cristina Checherita, and Christiane Nickel, “What explains the surge in euro area sovereign spreads during the financial crisis of 2007–09?,” Public Finance and Management, 2010,10(4), 595–646.
- Augustin, Patrick, “The term structure of CDS spreads and sovereign credit risk,” Journal of Monetary Economics, 2018,96, 53–76.
- Bampinas, Georgios, Theodore Panagiotidis, and Panagiotis N Politsidis, “Sovereign bond and CDS market contagion:  A story from the Eurozone crisis,” Journal of International Money and Finance, 2023, 137, 102902.
- Beber, Alessandro, Michael W Brandt, and Kenneth A Kavajecz, “Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market,” The Review of Financial Studies, 2009,22(3), 925–.
- Bedendo, Mascia and Paolo Colla, “Sovereign and corporate credit risk: Evidence from the Eurozone,” Journal of corporate Finance, 2015,33, 34–52.
- Blommestein, Hans, Sylvester Eijffinger, and Zongxin Qian, “Regime-dependent determinants of Euro area sovereign CDS spreads,” Journal of Financial Stability, 2016,22, 10–21.
- Drago, Danilo and Raffaele Gallo, “The impact and the spillover effect of a sovereign rating announcement on the euro area CDS market,” Journal of International Money and Finance, 2016,67, 264–286.
- Fabozzi, Frank J, Rosella Giacometti, and Naoshi Tsuchida, “Factor decomposition of the Eurozone sovereign CDS spreads,” Journal of International Money and Finance, 2016,65, 1–23.
- Costantini, Mauro, Matteo Fragetta, and Giovanni Melina, “Determinants of sovereign bond yield spreads in the EMU: An optimal currency area perspective,” European Economic Review, 2014,70, 337–349.

### Financial fragmentation, spillovers, and global risk
- Auerbach, Alan J and Yuriy Gorodnichenko, “Output spillovers from fiscal policy,” American Economic Review, 2013,103(3), 141–146.
- Brandt, Lennart, Arthur Saint Guilhem, Maximilian Schröder, and Ine Van Robays, “What drives euro area financial market developments?  The role of US spillovers and global risk,” 2021.
- Candelon, Bertrand, Angelo Luisi, and Francesco Roccazzella, “Fragmentation in the European Monetary Union:  Is it really over?,” Journal of International Money and Finance, 2022,122, 102545.
- Catalan, Mr. Mario, Mr. Salih Fendoglu, and Tomohiro Tsuruga, “A Gravity Model of Geopolitics and Financial Fragmentation,” IMF Working Papers 2024/196, International Monetary Fund September 2024.
- Corzo, Teresa, Laura Lazcano, Javier Márquez, Laura Gismera, and Sara Lumbreras, “A common risk factor in global credit and equity markets: An exploratory analysis of the subprime and the sovereign-debt crises,” Heliyon, 2020,6(6).
- Fernánde z-Villaverde, Jesús, Tomohide Mineyama, and Dongho Song, “Are We Fragmented Yet? Measuring Geopolitical Fragmentation and Its Causal Effect,” NBER Working Papers 32638, National Bureau of Economic Research, Inc June 2024.
- Bampinas, Georgios, Theodore Panagiotidis, and Panagiotis N Politsidis, “Sovereign bond and CDS market contagion:  A story from the Eurozone crisis,” Journal of International Money and Finance, 2023, 137, 102902.

### Market structure, liquidity, and official reports
- Beber, Alessandro, Michael W Brandt, and Kenneth A Kavajecz, “Flight-to-quality or flight-to-liquidity? Evidence from the euro-area bond market,” The Review of Financial Studies, 2009,22(3), 925–.
- Financial Stability Board, “FSB Report on Market Fragmentation,” Report, Financial Stability Board June

*Source: wpiea2025194-source-pdf - References*

### 2019.  Available at FSB website.

### Euro Area Financial Fragmentation and Bond Market Stability — Working Paper No. WP/2025/194

### Definitions of Financial Fragmentation
- Conceptual definition used: follows the Financial Stability Board (FSB) market-based angle — breakdown of markets into fragments either geographically or by product type or participant; fragmentation in securities markets: trading or clearing of economically similar assets across multiple venues; fragmentation in banking: pools of capital and liquidity segregated within local markets and unable to move freely across jurisdictions (see Financial Stability Board (2019)).
- Loose operational definition used in the paper: financial fragmentation is a situation wherein financial markets are segmented, causing a breakdown of market efficiency across segments and disparities in access to capital, interest rates, and investment opportunities.
- Drivers of fragmentation cited: geopolitical relationships, differences in legal frameworks and law enforcement’s, demographic and geographic factors, varying degrees of market development, and differences in investor base.
- Comparative definitions noted:
  - OECD emphasizes heterogeneous economic systems, policies, rules, laws and industry practices that block business efficiency (see OECD (2016)).
  - IMF discusses factors affecting cross-border investment, international payment systems, and asset prices with focus on geopolitical tensions (see IMF (2023)); Catalan et al. (2024) define fragmentation as a policy-driven weakening of financial links between countries.

### Supporting Statistics for Principal Component Analysis (PC2)
- PCA decision rules discussed:
  - Kaiser criterion: retain components with eigenvalues greater than one; in terms of variance explained: 1/n = 20% (n = 5 countries).
  - Parallel analysis: compare observed eigenvalues to those from random data; component significant if eigenvalue exceeds the 95th percentile of simulated distribution (Horn (1965)).
- Role and significance of PC2:
  - PC2 explains approximately 20% of the total variance in yield changes over the full sample.
  - During periods of higher fragmentation (weeks in which the fragmentation indicator is above its median), PC2 accounts for 31.0% of the variance in yield changes.
  - During periods of lower fragmentation, PC2 explains 11.4% of the variance.
- Statistical robustness:
  - PC2 is borderline significant under the Kaiser criterion and inconclusive under parallel analysis (falls below 95% threshold but within 95% confidence interval of the median of random draws).
  - Bootstrap analysis shows the composition (loadings) of PC2 is highly stable across resampled datasets (Figure 11 referenced).
  - The PC1/PC2 decomposition (PC1: common co-movements; PC2: core–periphery divergence) is consistent with prior literature and robust across instruments, maturities, sample periods, and country selections.

### Construction of Shocks (BVAR)
- Methodology: identified shocks constructed following Brandt et al. (2021) using financial market data and sign restrictions to identify monetary policy shocks, macro fundamental shocks in both the US and euro area, and global risk shocks.
- Empirical behavior:
  - BVAR model identifies spikes corresponding to major events (e.g., Global Financial Crisis, start of Quantitative Easing, COVID-19 pandemic).
  - Global risk factor correlates with equity market implied volatility (VIX and VSTOXX) and the euro zone CISS index (Hollo et al. (2012)), with the added benefit of being symmetrical (measuring both adverse and benign risk sentiment episodes).

### Financial Fragmentation and Bond Market Stability: Panel Regression Findings
- Panel setup: fixed effects panel model relating changes in sovereign CDS premiums (∆CDSit) to identified BVAR shocks and interaction terms for high-fragmentation regimes (high frag dummy equals 1 when fragmentation indicator above median); a sovereign debt crisis dummy equals one between May 2010–January 1, 2014.
- Main empirical specification (equation form reported): includes global shock, euro area policy, euro area macro, US macro, US policy, crisis dummy, high frag dummy, interactions (high frag × shocks), country fixed effects, and time-fixed effects in some models.
- Key findings:
  - In lower fragmentation regimes, global risk shocks and euro area macro shocks have a significant (p<0.05) positive relationship with sovereign CDS premiums.
  - Interaction terms of BVAR shocks with the high-fragmentation regime show a positive and significant additional impact — shocks have exacerbated effects during higher fragmentation.
  - During higher fragmentation regimes, a global risk shock adds 0.6 to the change in credit spread.
  - Adding time-fixed effects does not materially change coefficients and significance, suggesting robustness.
- Selected coefficient estimates from Table 9 (standard errors in parentheses; significance stars preserved):
  - Global Shock: 0.998 ∗∗∗ (0.061) in Model (1); 0.601 ∗∗∗ (0.035) in Model (6).
  - EA Macro: 1.422 ∗∗∗ (0.162) in Model (1); 0.845 ∗∗∗ (0.067) in Model (6).
  - US Macro: 0.204 ∗∗∗ (0.043) in Model (1); 0.112 ∗∗ (0.032) in Model (6).
  - EA Policy: 0.854 ∗∗∗ (0.133) in Model (1); 0.591 ∗∗∗ (0.080) in Model (6).
  - US Policy: 0.065 ∗ (0.025) in Model (1); 0.160 ∗∗ (0.040) in Model (6).
  - Crisis Dummy: −0.064 ∗∗∗ (0.013) in Model (5).
  - Global Shock + interaction term aggregate: 0.615 ∗∗∗ (0.081) reported.

### Granger Causality Results
- Approach: Granger causality tests for lags 1 to 5 to examine directional relations between sovereign CDS premiums and global risk shocks for IT, ES, FR, DE and NL; also tests whether CDS Granger-causes fragmentation index or vice versa.
- Overall result: bidirectional Granger causality for most countries — CDS premiums and global shocks often influence each other or react simultaneously to common risk events; directionality not clearly one-way in most cases.
- Italy exception: global shocks primarily Granger-cause Italian CDS (Italy shows non-significant tests for CDS→Global Shock and significant tests for Global Shock→CDS), reducing likelihood that global risk shocks originate in Italian CDS markets.
- Selected entries from Table 10 (CDS → Global Shock; F statistic and P value excerpts):
  - France lags 1–5: F statistics 11.08, 5.36, 3.68, 2.92, 2.36 with P values 0.00***, 0.00***, 0.01***, 0.02***, 0.04*** respectively.
  - Italy lags 1–5: F statistics 0.93, 0.58, 0.39, 0.32, 0.30 with P values 0.34, 0.56, 0.76, 0.86, 0.92 (not significant).
- Selected entries from Table 11 (Global Shock → CDS; F statistic and P value excerpts):
  - Italy lags: 1: F = 5.78, P = 0.02***; 2: F = 4.21, P = 0.01***; 3: F = 2.78, P = 0.04***; 4: F = 2.10, P = 0.08**; 5: F = 1.90, P = 0.09**.

### Sensitivity to the Smoothness Parameter κ (Local Projections with STF)
- Baseline choice: κ = 10 (following Auerbach and Gorodnichenko (2013)), implying a relatively sharp transition between fragmentation regimes.
- Robustness exercise: κ varied between 1 and 10; local projection models re-estimated at both extremes.
- Result: main conclusions robust across κ ∈ [1,10]:
  - Impulse responses of sovereign CDS spreads and corporate OAS to a standardized global risk shock are larger and more persistent in the higher fragmentation regime for both κ = 1 and κ = 10.
  - Differential responses remain statistically and economically significant across specifications and do not significantly vary across κ values.

### Structural Breaks in Sovereign Fragmentation Indicator
- Method: Bai-Perron multiple breakpoint test (Bai and Perron (2003)) applied to endogenously identify structural breaks in the mean of the fragmentation indicator; Chow test applied at the first identified high-fragmentation point.
- Findings:
  - Estimated breakpoints align closely with periods where the fragmentation indicator exceeds its historical median (grey shaded high-fragmentation periods).
  - Chow test at first high-fragmentation occurrence strongly rejects the null of parameter stability, indicating a statistically significant shift in the level of sovereign fragmentation.
  - Alignment between Bai-Perron breakpoints and high-fragmentation periods supports the indicator’s ability to capture meaningful regime shifts.

### Autocorrelation in Sovereign CDS Spreads
- Empirical patterns:
  - Autocorrelation at lag 0 equals one by construction; autocorrelation at lag 1 is modest and decays quickly, with most subsequent lags within 95% confidence bounds (ACFs for first ten lags shown in Figure 21).
  - Fixed effects panel regression of CDS on its first lag: Wooldridge test gives F = 1.26e5, p-value = 0.9972 — provides no evidence of first-order serial correlation in panel residuals.
  - Local projection specifications including ∆CDSt−1 show coefficients for ∆CDSt−1 are not significant; adding lagged CDS changes does not alter main conclusions.
- Interpretation: sovereign CDS spread changes exhibit limited serial dependence at daily frequency after controlling for fixed effects, consistent with rapid and efficient price adjustment.

*Italic source attribution: Source: wpiea2025194-source-pdf - 2019. Available at FSB website. Canonical source URL: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025194-source-pdf.pdf_