## Sign-Restriction VAR Analysis on Bond Issuance

## Source details

**Canonical URL:** [Sign-Restriction VAR Analysis on Bond Issuance](https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025012.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025012.pdf.md)
- [Structured JSON version](/-/media/files/publications/selected-issues-papers/2025/english/sipea2025012.pdf.json)

---

### Introduction and context
- Pension fund assets declined from approximately 80 percent of GDP to less than 60 percent of GDP due to three rounds of pension fund withdrawals during the pandemic.
- Public debt-to-GDP ratio rose to about 40 percent of GDP.
- Banco Central de Chile (BCCh) introduced Facility of Credit Conditional on Lending Increase (FCIC) measures in 2020–2021, accepting commercial bank credits as collateral, corresponding to about 8 percent of banks’ total liabilities.
- FCIC and other exceptional liquidity measures (including a cash purchase and forward sale program — CC-VP — and temporary repo facilities) were unwound in April and July 2024.
- The reduction of about 20 percent of GDP in pension fund assets is comparable to pension contributions for about 9 years.

### Changes in the structure of local financial markets
- Shift toward foreign funding:
  - Non-financial firms relied more on foreign investors and less on banks, insurers, and pension funds for issuance of debt securities and loan funding during the pandemic.
  - Government debt holders shifted from pension funds to banks and foreign investors; this shift persisted until at least September 2024.
  - Issuance shifted to offshore markets particularly for sovereign and corporate bonds amid lower demand for local bonds.
- Bank funding and deposit composition:
  - Banks became more dependent on retail deposits and less on pension funds and money market fund deposits.
  - The share of government bonds in banks’ total financial assets doubled from about 3 to 6 percent since the pandemic.
  - Banks’ bond funding amounts remained stable and did not increase in line with bank asset growth, partly due to reduced funding from pension funds.
  - Loans from the BCCh increased during the pandemic under FCIC but returned to pre-pandemic levels after the FCIC was fully unwound in July 2024.
- Currency and dollar funding:
  - Financial and non-financial sectors have relatively low foreign-currency denominated debt, concentrated in larger firms (including state-owned firms and mining and electricity, gas, and water sectors).
  - Dollar funding costs in US$-dominated bond markets were broadly stable; onshore dollar spreads were also stable.
  - Non-financial corporations’ currency mismatch is generally limited due to exporters’ natural hedges and use of derivatives; banks’ currency mismatch is restricted by regulations on foreign currency funding gaps.

### Financial depth: developments and indicators
- Pre-pandemic standing:
  - Chile’s financial institutions depth (based on IMF Financial Development Index components) was comparable to the average of advanced economies and much higher than regional peers (index available until 2021).
  - Financial markets depth sat between the average levels of advanced economies and regional peers.
- Post-withdrawal changes:
  - Financial institutions depth declined significantly in 2021, mainly reflecting the decrease in pension fund assets to GDP; it had not rebounded by 2023.
  - Other components of financial institutions depth also declined and had not fully recovered by 2023.
  - Financial markets depth declined in 2021 and indicators remained broadly unchanged through 2023.
- Bond and equity market specifics:
  - Stock of local bonds to GDP declined in 2021 and remained below pre-pandemic levels.
  - Issuance of bank and corporate bonds returned to pre-pandemic trend levels, but no pent-up issuance observed.
  - Average bond maturity moderately declined; average interest rates increased.
  - Secondary bond market turnover has not returned to pre-pandemic levels; corporate bond spreads remain higher than pre-pandemic.
  - Volatility (standard deviation) of corporate and bank bond spreads has been above the 2018 average since 2019; sovereign spreads show similar, though more modest, patterns.
  - Equity market: new issuance stagnated, market capitalization declined; turnover, PER, and PBR are lower than pre-pandemic; small-cap liquidity deterioration more significant.

### Presence in FX markets and shock absorption
- Pension funds’ presence in gross currency derivative positions and turnover declined; foreign investors’ presence in spot and derivative markets increased.
- Exchange rate volatility remains elevated relative to the 2018 average.
- The role of pension funds as a natural offset (shock absorber) to non-resident capital outflows has weakened because the cushioning function is proportional to pension fund transaction volume and asset size.

### Empirical assessment: sensitivity of local financial variables to global risk
- Model specification (local projection, daily data):
  - Estimated model: y_{t+h} − y_{t−1} = α_{t+h} + β_{t+h} GFS_t + γ_{t+h} X_{t−1} + ε_{t+h}, for h = 0,...,10.
  - Dependent variables: sovereign bond spreads, corporate bond spreads, equity prices (log-level), exchange rates, onshore spreads, and a local stress index for sovereign bond and exchange rate markets.
  - GFS_t is the global financial stress index developed by the Office of Financial Research (OFR) in the U.S.
  - Controls X_{t−1} include the one-period lag of the dependent variable (y_{t−1} − y_{t−2}) and lag of the global financial stress index (GFS_{t−1}).
  - The coefficient β_{t+h} represents the estimated cumulative response of the dependent variable to a one standard deviation increase in global financial stress from t−1 to t.
- Sample and estimation details:
  - Daily data sample from January 2000 to July 2024 (or the longest available period within that timeframe).
  - Models estimated separately for periods before September 2019 and after October 2019 (separation based on start of 2019 social unrest).
  - Standard errors corrected for heteroskedasticity and autocorrelation (HAC estimators).
  - Responses are presented with 90 percent confidence intervals for a one standard deviation increase in global financial stress.
- Interpretation benchmarks:
  - Historical bottom-peak movements of the OFR global financial stress index: about 3, 5, 1, and 3 standard deviations for the subprime mortgage crisis, the global financial crisis (GFC), the Euro debt crisis, and the pandemic, respectively (from January 2000 to July 2024).
  - The estimated responses capture both the direct impact of global financial stress and endogenous responses of other local financial variables.

### Increased sensitivity of local financial variables (main findings)
- Comparative responses (post-social unrest / post-pandemic versus pre-pandemic):
  - Sovereign bond spreads: increased; difference statistically significant at the 10 percent level.
  - Corporate bond spreads: tended to increase.
  - Equity prices: responses increased; difference statistically significant at the 10 percent level.
  - Exchange rates: responses increased compared to pre-pandemic periods.
  - Onshore spreads: results are mixed.
- Local financial market stress index: more strongly heightened in response to a one standard deviation increase in the global financial stress index.
- Robustness: results broadly robust to different separation periods, pre-pandemic sample periods, and number of lags for control variables (Annex III).

### Sign-Restriction VAR analysis on bond issuance (Box 1)
- Objective: historical decomposition of deviation from log-linear trend of 12-month moving averages of monthly total issuance of corporate and bank bonds (percent) in Chilean capital market.
- Model specification:
  - Variables: bond issuance (UF-denominated focus), GDP growth (IMACEC y/y percent), CPI inflation (y/y percent), monetary policy rate (percent).
  - Four lags.
  - Sample: January 2001 to June 2024.
- Identified shocks:
  - Demand shock: positive effects on GDP, inflation, policy rate, and bond issuance.
  - Supply shock: positive effects on GDP and bond issuance; negative effect on inflation.
  - Monetary policy (MP) shock: negative effects on GDP, inflation, bond issuance; positive effect on policy rates.
- Key findings:
  - Pension fund withdrawals had a negative impact on bond issuance during the pandemic, with no sign of pent-up demand after that.
  - “Macro-financial shocks” explained most of the decline during the GFC but could not account for the decrease in issuance after the pandemic.
  - Since the pandemic, “other shock” has not contributed to bond issuance in either direction.
  - Focus on UF-denominated issuance because it accounts for approximately 85 percent of total issuance in the sample.
  - Replacing monetary policy rates with long-term (10-year) interest rates in UF terms leaves contributions of “macro-financial shocks” and “other shock” broadly unchanged.

### Risk assessment and policy recommendations
- Structural changes and consequences:
  - Pension fund withdrawals, higher public debt, and BCCh liquidity injections during the pandemic reshaped Chilean financial network.
  - Reduced demand for local bonds led large non-financial corporations and the government to rely more on offshore markets and foreign investors for issuance.
  - Retail deposit inflows from pension fund withdrawals and BCCh liquidity narrowed banks’ wholesale funding channels (MMF deposits and bank bonds).
  - Banks’ exposure to sovereign bonds increased; sovereign-bank nexus remains lower than regional peers and part of increase may be temporary due to collateral switching for FCIC unwinding (ended July 2024).
  - Dilution of local market depth following pension fund withdrawals has made the market more sensitive to external financial stress.
  - Estimated increased responsiveness of local financial variables to global financial stress (estimate based on entire period since late 2019; does not account for within-period variation).
- Policy recommendations:
  - Restore pension fund size and ability to invest in relatively illiquid assets to rebuild market depth and resilience.
  - Avoid further pension fund withdrawals.
  - Implement proposed increase in pension contribution rate from 10 to 16 percent to accelerate rebuilding of market sizes.
  - Smooth implementation of Financial Market Resilience Law to:
    - Develop interbank repo market.
    - Enhance BCCh’s crisis response capabilities.
    - Strengthen mutual fund liquidity management framework.
    - Facilitate internationalization of the Chilean peso to diversify counterparties and deepen CLP role in cross-border transactions, reducing dependence on foreign currency funding.
  - Evaluate and support measures that enhance secondary market liquidity and broaden the investor base (including initiatives to integrate regional exchanges to attract global investors).
  - BCCh measures in implementation:
    - Self-securitization scheme to increase supply of collateral instruments.
    - Primary market maker system for sovereign bonds in implementation phase.
    - Basel III capital and liquidity requirements being implemented, including counter-cyclical capital buffers: currently 0.5 percent of risk-weighted assets; plans to gradually increase to a 1 percent neutral level.
  - Regulatory and market measures noted:
    - Recent limits imposed on frequent transfers between different types of pension funds to reduce precautionary liquidity.
    - Fintech Law allows providing new financial products such as Mini-bonds.
    - Financial Market Resilience Law features (BCCh preparing regulations, 2024H2 FSR) include enabling BCCh to provide liquidity to non-bank financial entities and exceptionally offer repos to non-banks in systemic financial stress; enhances CMF’s powers to impose liquidity buffer requirements on mutual funds; simplifies procedure for obtaining a Tax Identification Number for CLP accounts held by non-residents.

### Annex I — Chilean pension funds: key facts (highlights)
- Size and trends:
  - Pension fund assets to GDP: reached around 80 percent until 2019; dropped to around 60 percent after withdrawals.
  - 80 percent pre-2019 is significantly higher than the median of regional and OECD peers (approximately 20 percent).
  - Current ~60 percent level comparable to the 75th percentile of OECD peers.
- Investment profile and holdings:
  - Primary investments in private domestic bonds; increasing foreign investments.
  - Share of bonds in pension fund assets similar to regional and OECD medians; share of private bonds in total bond investments higher than medians.
  - Share of foreign assets increased steadily; level comparable to OECD median and significantly higher than regional peers.
  - Before withdrawals, pension funds held around half of sovereign and bank bonds and about 20 percent of corporate bonds; these shares declined after withdrawals.
- Fund types and behavior:
  - Five fund types (A–E) since 2002 reform, with A riskiest and E safest.
  - Frequent shifts among fund types by investors; during crises shifts to D and E; in 2022 shifts from D and E to A and B.
  - Recent regulatory limits on frequent fund-type shifts introduced.
- Duration strategy:
  - After withdrawals, type E funds lost prominence.
  - Pension funds have actively used interest rate derivatives to increase duration: long positions in long-term interest rates (receiving fixed) and short positions in short-term rates (paying variable).

### Annex II — Network analysis: funding structure shifts (highlights)
- Debt securities network:
  - Liquidity provision from pension funds to government, non-financial corporations, and banks decreased.
  - Liquidity provision from overseas to these sectors increased.
  - Liquidity provision from other investment funds to banks and from insurers to non-financial firms narrowed.
  - BCCh liquidity provision to banks increased during pandemic via bank bond purchase program (started March 2020) and bank bonds reinvestment program (started January 2021).
- Deposits network:
  - Households now primary providers of liquidity to banks.
  - Money market funds and pension funds became smaller liquidity providers.
  - Banks’ deposits in BCCh rose during pandemic (reflecting FCIC and bank bond purchase programs) and decreased as FCIC was unwound in April and July 2024.
  - Government deposits to banks have gradually increased since the pandemic.
- Mutual funds and loans networks:
  - Pension funds’ presence in overseas markets decreased.
  - Liquidity provision from households and firms to mutual funds (MMFs and other) increased.
  - Increasing provision of liquidity from overseas to non-financial firms.
  - BIS locational banking statistics: loans from the United States, Spain, and Hong Kong SAR increased from Q2 2019 to Q2 2024.

### Annex III — Robustness of sensitivity analysis (summary)
- Separation period sensitivity:
  - Separating samples at end of January produces results almost unchanged from main text.
  - Separating at end of May 2020 (first round of pension fund withdrawals) yields broadly robust results, especially for exchange rates and equity prices; difference for sovereign and corporate bond spreads becomes less striking.
  - Shifting separation to end-May 2020 reduces sample size by approximately 25 percent for corporate bonds and 20 percent for other variables.
  - Note: during-and-post-pandemic sample periods are about four years for corporate bonds and five years for other variables.
- Pre-pandemic sample period sensitivity:
  - Starting pre-pandemic sample from 2014: results remain broadly unchanged or become sharper.
  - Corporate bond spreads response becomes significantly less sensitive in the earlier pre-pandemic sample, widening differences with during-and-post-pandemic period.
  - Exchange rates responses become less persistent in pre-pandemic period, widening differences five days after stress date.
- Lag structure and controls:
  - Results robust when extending lagged controls beyond a single lag (e.g., adding y_{t−2}−y_{t−3}, y_{t−3}−y_{t−4}, ..., GFS_{t−2}, GFS_{t−3}, ...).

*Source: sipea2025012.*

### 1. Sign-Restriction VAR Analysis on Bond Issuance _____________________________________ 15

### 1. Sign-Restriction VAR Analysis on Bond Issuance

### Introduction and Context
- Pension fund assets declined from approximately 80 percent of GDP to less than 60 percent of GDP due to three rounds of pension fund withdrawals during the pandemic.
- The public debt-to-GDP ratio rose to about 40 percent of GDP.
- The Banco Central de Chile (BCCh) introduced Facility of Credit Conditional on Lending Increase (FCIC) measures in 2020–2021, accepting commercial bank credits as collateral, corresponding to about 8 percent of banks’ total liabilities.
- The FCIC and other exceptional liquidity measures (including a cash purchase and forward sale program — CC-VP — and temporary repo facilities) were unwound in April and July 2024.
- The reduction of about 20 percent of GDP in pension fund assets is comparable to pension contributions for about 9 years.

### Changes in the Structure of Local Financial Markets
- Shift toward foreign funding:
  - Non-financial firms relied more on foreign investors and less on banks, insurers, and pension funds for issuance of debt securities and loan funding during the pandemic.
  - Government debt holders shifted from pension funds to banks and foreign investors; this shift persisted until at least September 2024.
  - Issuance shifted to offshore markets particularly for sovereign and corporate bonds amid lower demand for local bonds.
- Bank funding and deposit composition:
  - Banks became more dependent on retail deposits and less on pension funds and money market fund deposits.
  - The share of government bonds in banks’ total financial assets doubled from about 3 to 6 percent since the pandemic.
  - Banks’ bond funding amounts remained stable and did not increase in line with bank asset growth, partly due to reduced funding from pension funds.
  - Loans from the BCCh increased during the pandemic under FCIC but returned to pre-pandemic levels after the FCIC was fully unwound in July 2024.
- Currency and dollar funding:
  - Financial and non-financial sectors have relatively low foreign-currency denominated debt, concentrated in larger firms (including state-owned firms and mining and electricity, gas, and water sectors).
  - Dollar funding costs in US$-dominated bond markets were broadly stable; onshore dollar spreads were also stable.
  - Non-financial corporations’ currency mismatch is generally limited due to exporters’ natural hedges and use of derivatives; banks’ currency mismatch is restricted by regulations on foreign currency funding gaps.

### Financial Depth: Developments and Indicators
- Pre-pandemic standing:
  - Chile’s financial institutions depth (based on IMF Financial Development Index components) was comparable to the average of advanced economies and much higher than regional peers (index available until 2021).
  - Financial markets depth sat between the average levels of advanced economies and regional peers.
- Post-withdrawal changes:
  - Financial institutions depth declined significantly in 2021, mainly reflecting the decrease in pension fund assets to GDP; it had not rebounded by 2023.
  - Other components of financial institutions depth also declined and had not fully recovered by 2023.
  - Financial markets depth declined in 2021 and indicators remained broadly unchanged through 2023.
- Bond and equity market specifics:
  - Stock of local bonds to GDP declined in 2021 and remained below pre-pandemic levels.
  - Issuance of bank and corporate bonds returned to pre-pandemic trend levels, but no pent-up issuance observed.
  - Average bond maturity moderately declined; average interest rates increased.
  - Secondary bond market turnover has not returned to pre-pandemic levels; corporate bond spreads remain higher than pre-pandemic.
  - Volatility (standard deviation) of corporate and bank bond spreads has been above the 2018 average since 2019; sovereign spreads show similar, though more modest, patterns.
  - Equity market: new issuance stagnated, market capitalization declined; turnover, PER, and PBR are lower than pre-pandemic; small-cap liquidity deterioration more significant.

### Presence in FX Markets and Shock Absorption
- Pension funds’ presence in gross currency derivative positions and turnover declined; foreign investors’ presence in spot and derivative markets increased.
- Exchange rate volatility remains elevated relative to the 2018 average.
- The role of pension funds as a natural offset (shock absorber) to non-resident capital outflows has weakened because the cushioning function is proportional to pension fund transaction volume and asset size.

### Empirical Assessment: Sensitivity of Local Financial Variables to Global Risk
- Model specification (local projection, daily data):
  - Estimated model: y_{t+h} − y_{t−1} = α_{t+h} + β_{t+h} GFS_t + γ_{t+h} X_{t−1} + ε_{t+h}, for h = 0,...,10.
  - Dependent variables: sovereign bond spreads, corporate bond spreads, equity prices (log-level), exchange rates, onshore spreads, and a local stress index for sovereign bond and exchange rate markets.
  - GFS_t is the global financial stress index developed by the Office of Financial Research (OFR) in the U.S.
  - Controls X_{t−1} include the one-period lag of the dependent variable (y_{t−1} − y_{t−2}) and lag of the global financial stress index (GFS_{t−1}).
  - The coefficient β_{t+h} represents the estimated cumulative response of the dependent variable to a one standard deviation increase in global financial stress from t−1 to t.
- Sample and estimation details:
  - Daily data sample from January 2000 to July 2024 (or the longest available period within that timeframe).
  - Models estimated separately for periods before September 2019 and after October 2019 (separation based on start of 2019 social unrest).
  - Standard errors corrected for heteroskedasticity and autocorrelation (HAC estimators).
  - Responses are presented with 90 percent confidence intervals for a one standard deviation increase in global financial stress.
- Interpretation benchmarks:
  - Historical bottom-peak movements of the OFR global financial stress index: about 3, 5, 1, and 3 standard deviations for the subprime mortgage crisis, the global financial crisis (GFC), the Euro debt crisis, and the pandemic, respectively (from January 2000 to July 2024).
  - The estimated responses capture both the direct impact of global financial stress and endogenous responses of other local financial variables.

### Implications and Policy Considerations
- Market resilience and sensitivity:
  - Shallower financial depth and reduced pension fund size have likely increased market volatility and sensitivity to external shocks.
  - Increased reliance on foreign investors for sovereign and corporate issuance raises exposure to external stress on a stock basis, though current redemption schedules and limited foreign-currency debt mitigate immediate systemic risk.
- Policy priorities noted:
  - Restoring pension funds and strengthening market resilience and crisis response capabilities are essential for future financial stability.
  - Continued monitoring of the sovereign-bank nexus is warranted given the doubling of government bond holdings in banks from about 3 to 6 percent of total financial assets.
  - Evaluate and support measures that enhance secondary market liquidity and broaden the investor base (including initiatives to integrate regional exchanges to attract global investors).

*Source: IMF staff analysis (Chapter: "Post-Pandemic Changes to Chile’s Financial Markets"; data and text as in the provided content unit).*

### 15.      Local financial variables appear to have become more sensitive to global financial

### 15.      Local financial variables appear to have become more sensitive to global financial stress.

### Increased sensitivity of local financial variables
- Responses compared to (pre-social unrest and) pre-pandemic periods:
  - Sovereign bond spreads: increased; difference statistically significant at the 10 percent level.
  - Corporate bond spreads: tended to increase.
  - Equity prices: responses increased; difference statistically significant at the 10 percent level.
  - Exchange rates: responses increased compared to pre-pandemic periods.
  - Onshore spreads: results are mixed.
- Local financial market stress index: more strongly heightened in response to a one standard deviation increase in the global financial stress index.
- Robustness: results broadly robust to different separation periods, sample periods for the pre-pandemic periods, and the number of lags for control variables (Annex III).

### Authorities’ measures to restore depth and resilience
- Pension fund reform (proposed):
  - Increase in pension contribution rate from 10 to 16 percent.
  - Recent limits imposed on frequent transfers between different types of funds to reduce precautionary liquidity.
- Fintech Law:
  - Allows providing new financial products such as Mini-bonds.
- Financial Market Resilience Law (implementation underway):
  - Aims to increase depth of the interbank repo market by providing legal certainty to repo participants (BCCh preparing regulations, 2024H2 FSR).
  - Enables the BCCh to provide liquidity to non-bank financial entities, including systemic credit unions and financial market infrastructures.
  - Establishes framework for BCCh to exceptionally offer repos to non-banks in systemic financial stress.
  - Enhances CMF’s regulatory powers to impose liquidity buffer requirements on mutual funds.
  - Intends to facilitate internationalization of the Chilean peso by simplifying procedure for obtaining a Tax Identification Number for CLP accounts held by non-residents.
- Additional measures:
  - BCCh implementing a self-securitization scheme to increase supply of collateral instruments.
  - Primary market maker system for sovereign bonds in implementation phase to enhance secondary market liquidity.
  - Basel III capital and liquidity requirements being implemented, including counter-cyclical capital buffers:
    - Currently 0.5 percent of risk-weighted assets.
    - Plans to gradually increase to a 1 percent neutral level.

### Sign-Restriction VAR analysis on bond issuance (Box 1)
- Objective: historical decomposition of deviation from log-linear trend of 12-month moving averages of monthly total issuance of corporate and bank bonds (percent) in Chilean capital market.
- Model specification:
  - Variables: bond issuance (UF-denominated focus), GDP growth (IMACEC y/y percent), CPI inflation (y/y percent), monetary policy rate (percent).
  - Four lags.
  - Sample: January 2001 to June 2024.
- Identified shocks:
  - Demand shock: positive effects on GDP, inflation, policy rate, and bond issuance.
  - Supply shock: positive effects on GDP and bond issuance; negative effect on inflation.
  - Monetary policy (MP) shock: negative effects on GDP, inflation, bond issuance; positive effect on policy rates.
- Key findings:
  - Pension fund withdrawals had a negative impact on bond issuance during the pandemic, with no sign of pent-up demand after that.
  - “Macro-financial shocks” explained most of the decline during the GFC but could not account for the decrease in issuance after the pandemic.
  - Since the pandemic, “other shock” has not contributed to bond issuance in either direction.
  - Focus on UF-denominated issuance because it accounts for approximately 85 percent of total issuance in the sample.
  - Replacing monetary policy rates with long-term (10-year) interest rates in UF terms leaves contributions of “macro-financial shocks” and “other shock” broadly unchanged.

### Risk assessment and policy recommendations
- Structural changes in funding and exposure:
  - Pension fund withdrawals, higher public debt, and BCCh liquidity injections during the pandemic reshaped Chilean financial network.
  - Reduced demand for local bonds led large non-financial corporations and the government to rely more on offshore markets and foreign investors for issuance.
  - Retail deposit inflows from pension fund withdrawals and BCCh liquidity narrowed banks’ wholesale funding channels (MMF deposits and bank bonds).
  - Banks’ exposure to sovereign bonds increased; sovereign-bank nexus remains lower than regional peers and part of increase may be temporary due to collateral switching for FCIC unwinding (ended July 2024).
- Consequence:
  - Dilution of local market depth following pension fund withdrawals has made the market more sensitive to external financial stress.
  - Estimated increased responsiveness of local financial variables to global financial stress (estimate based on entire period since late 2019; does not account for within-period variation).
- Policy recommendations:
  - Restore pension fund size and ability to invest in relatively illiquid assets to rebuild market depth and resilience.
  - Avoid further pension fund withdrawals.
  - Implement proposed increase in pension contribution rate to accelerate rebuilding of market sizes.
  - Smooth implementation of Financial Market Resilience Law to:
    - Develop interbank repo market.
    - Enhance BCCh’s crisis response capabilities.
    - Strengthen mutual fund liquidity management framework.
    - Facilitate internationalization of the Chilean peso to diversify counterparties and deepen CLP role in cross-border transactions, reducing dependence on foreign currency funding.

### Annex I — Chilean pension funds: key facts
- Size and trends:
  - Pension fund assets to GDP: reached around 80 percent until 2019; dropped to around 60 percent after withdrawals.
  - 80 percent pre-2019 is significantly higher than the median of regional and OECD peers (approximately 20 percent).
  - Current ~60 percent level comparable to the 75th percentile of OECD peers.
- Investment profile:
  - Primary investments in private domestic bonds; increasing foreign investments.
  - Share of bonds in pension fund assets similar to regional and OECD medians.
  - Share of private bonds in total bond investments higher than medians.
  - Share of foreign assets increased steadily; level comparable to OECD median and significantly higher than regional peers.
- Market holdings:
  - Before withdrawals, pension funds held around half of sovereign and bank bonds and about 20 percent of corporate bonds; these shares declined after withdrawals.
  - Significant holdings in mutual funds; limited share in direct equity investments.
- Fund types and investor behavior:
  - Five fund types (A–E) since 2002 reform, with varying exposure to variable-income assets (A riskiest, E safest).
  - Frequent shifts among fund types by investors; during crises shifts to safer funds (D and E); in 2022 shifts from D and E to A and B.
  - Shifts driven by many individual investors often guided by financial advisory firms; led funds to maintain ample liquidity, hampering investment in illiquid assets.
  - Recent regulatory limits on frequent fund-type shifts introduced.
- Investment limits and rules:
  - Pension funds can invest up to 80 percent abroad, with limits ranging from 5 to 80 percent depending on fund profile.
  - Allowed to invest in alternative assets since 2016; investment limits have been increased.
- Duration strategy:
  - After withdrawals, type E funds lost prominence.
  - Pension funds have actively used interest rate derivatives to increase duration: long positions in long-term interest rates (receiving fixed) and short positions in short-term rates (paying variable).

### Annex II — Network analysis: funding structure shifts
- Debt securities network:
  - Liquidity provision from pension funds to government, non-financial corporations, and banks decreased.
  - Liquidity provision from overseas to these sectors increased.
  - Liquidity provision from other investment funds to banks and from insurers to non-financial firms narrowed.
  - BCCh liquidity provision to banks increased during pandemic via bank bond purchase program (started March 2020) and bank bonds reinvestment program (started January 2021).
- Deposits network:
  - Households now primary providers of liquidity to banks.
  - Money market funds and pension funds became smaller liquidity providers.
  - Banks’ deposits in BCCh rose during pandemic (reflecting FCIC and bank bond purchase programs) and decreased as FCIC was unwound in April and July 2024.
  - Government deposits to banks have gradually increased since the pandemic.
- Mutual funds network:
  - Pension funds’ presence in overseas markets decreased.
  - Liquidity provision from households and firms to mutual funds (MMFs and other) increased.
  - Liquidity provision from other mutual funds to overseas markets moderately increased.
- Loans network:
  - Increasing provision of liquidity from overseas to non-financial firms.
  - BCCh loans to banks increased during FCIC period but returned to pre-pandemic levels after unwinding in April and July 2024.
  - BIS locational banking statistics: loans from the United States, Spain, and Hong Kong SAR increased from Q2 2019 to Q2 2024.

*Source: sipea2025012.*

### Annex III. Robustness of the Analysis on Sensitivity of Local

### Annex III. Robustness of the Analysis on Sensitivity of Local Financial Variables to Global Risk

### Overview
- This Annex examines robustness of the local projection analysis regarding the sensitivity of local financial variables to global financial stress presented in the main text.
- Robustness is explored from three perspectives:
  - Separation period between pre-pandemic and during-and-post-pandemic periods.
  - Estimation period for the pre-pandemic period (focus on more recent samples).
  - Adequacy and lag structure of control variables.

### Separation of sample periods
- If samples are separated at the end of January (i.e., exactly before and after the pandemic), results remain almost unchanged from those in the main text.
- If samples are separated at the end of May 2020 (the first round of pension fund withdrawals):
  - Results remain broadly robust, especially regarding exchange rates and equity prices.
  - The difference between the pre-pandemic and the during-and-post-pandemic periods becomes less striking for sovereign and corporate bond spreads.
  - Shifting the separation period to the end of May 2020 reduces the sample size by approximately 25 percent for corporate bonds and 20 percent for the other variables, respectively.
- Note on sample length: sample periods for the during-and-post-pandemic periods are only about four years long for corporate bonds and five years long for other variables.

### Sample periods for pre-pandemic periods
- If the pre-pandemic sample period starts from 2014, estimation results:
  - Remain broadly unchanged or even become sharper.
  - Show that the response of corporate bond spreads becomes significantly less sensitive in the pre-pandemic period, producing a sizable difference in sensitivity between the pre-pandemic and the during-and-post-pandemic periods.
  - Show that the response of exchange rates becomes less persistent in the pre-pandemic period, resulting in a wider difference in sensitivity five days after the stress date.
  - Are consistent with a widening difference in the local stress index for exchange rates.

### Control variables and lag structure
- Estimation results remain quite robust when extending the number of lagged control variables from one period to multiple periods.
- Example of extended controls added to X_{t−1}: (y_{t−2}−y_{t−3}, y_{t−3}−y_{t−4}, ..., GFS_{t−2}, GFS_{t−3}, ...).

### Robustness summary
- The results are broadly robust to:
  - Different separation periods between pre-pandemic and during-and-post-pandemic samples.
  - Different sample periods for the pre-pandemic estimation (including starting from 2014).
  - Adding multiple lagged control variables beyond a single lag.

*Source: Annex III. Robustness of the Analysis on Sensitivity of Local Financial Variables to Global Risk*

---


_Source: https://www.imf.org/-/media/files/publications/selected-issues-papers/2025/english/sipea2025012.pdf_