## onlineannex21 - 1. Variables in LSI 2. Alternative Aggregations of EM LSI

## Source details

**Canonical URL:** [onlineannex21 - 1. Variables in LSI 2. Alternative Aggregations of EM LSI](https://www.imf.org/-/media/files/publications/gfsr/2020/october/english/onlineannex21.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/gfsr/2020/october/english/onlineannex21.pdf.md)
- [Structured JSON version](/-/media/files/publications/gfsr/2020/october/english/onlineannex21.pdf.json)

---

### LSI variables and aggregation
- LSI components (category headings preserved):
  - Exchange Rate LSI
    - FX bid-ask spread
    - Realized volatility
    - Implied Volatility
    - Risk-reversal ratio
  - Local Bonds LSI
    - Bond bid-ask spread
    - Realized volatility
    - Bond asset swap
    - Term-premium estimate
    - Non-resident flows
    - Bond volume deviations
    - Convertibility basis
- Indicators and methodological notes:
  - LSI Cross-Correlation weights.
  - LSI Non-Real-Time estimation.
  - LSI (GFSR).

### Country event studies for Bond LSI (selected narrative findings)
- Hungary:
  - Purchases in May seem to have had a large impact.
  - MNB decided to step up its program in the summer as conditions deteriorated again.
- Poland:
  - Large purchases in April combined with a step up in issuance.
  - Purchases tapered off as conditions improved in June, but trading volumes remained low.
- Indonesia:
  - Purchases were frontloaded and combined with large issuance.
  - Conditions improved further following the burden sharing agreement in July.
- South Africa:
  - APP has been relatively limited in size and the market remained concerned with increased issuance.
- Thailand:
  - Saw a swift improvement in the LSI, despite halting its APP in early April.
- India:
  - Improvement remained limited until RBI announced simultaneous sell/buy curve operations.
- Data sources for event studies: Bloomberg Finance L.P.; Country Central Banks; and IMF Calculations.

### Measuring the drivers of FX surprises — specification and variables
- Empirical goal: shed light on drivers of exchange rate movements in emerging markets and the role of domestic policies and global factors.
- Regression specification (symbols preserved in source):
  - Dependent variable: 훥훥 퐶퐶퐶퐶 푟푟푟푟퐶퐶푛푛 퐶퐶푦푦
푐푐,푚푚 — negative of percentage difference between realized end-of-month spot and one-month forward (end of month D−1), normalized by the latter (represents percentage gain of domestic currency above the forward).
- Domestic policy variables (푫푓...):
  - FXI (FX intervention): valuation adjusted changes in the stock of reserves including spot and derivative markets; scaled by actual stock of reserves. Note: estimates do not adjust for FX bond sales/purchases in some cases (e.g., Mexico).
  - Domestic policy rate: end-of-month values from Bloomberg.
- Global factors (푮...):
  - Effective Federal Fund rate at end of each month.
  - VIX Index averaged over the month.
- Controls:
  - Four interactions between domestic policies and global factors are included.
  - 푋푋
푐푐,푚푚: macroeconomic control variables — surprise indices on economic activity and inflation sourced from Citi and Bloomberg.
  - Country fixed effects (μ
푐푐
) included; standard errors are robust.
- Samples and periods:
  - Two regressions:
    - COVID-19: January to May of 2020.
    - 2015 China sell-off: April 2015 to February 2016.
  - Sample: 14 emerging market economies — Argentina, Brazil, Chile, China, Colombia, India, Indonesia, Mexico, Malaysia, Philippines, Russia, Thailand, Turkey, and South Africa.
  - Robustness: removing China yields consistent results.

### Empirical results (Online Annex Table 2.1.1 — Effects of Domestic and Global Factors on FX Surprises)
- Coefficients with standard errors in parentheses:
  - FX Intervention:
    - 2020 COVID-19: -0.813 (0.991)
    - 2015 EM Sell-Off: 2.196*** (0.706)
  - Domestic Policy Rate:
    - 2020 COVID-19: -0.376 (0.309)
    - 2015 EM Sell-Off: 1.788** (0.813)
  - VIX:
    - 2020 COVID-19: -0.525*** (0.100)
    - 2015 EM Sell-Off: 7.928 (6.748)
  - Federal Fund Rate:
    - 2020 COVID-19: -9.659*** (1.634)
    - 2015 EM Sell-Off: 0.203 (0.125)
  - FX Intervention x VIX:
    - 2020 COVID-19: 0.012 (0.019)
    - 2015 EM Sell-Off: -0.165*** (0.048)
  - FX Intervention x Federal Fund Rate:
    - 2020 COVID-19: 0.206 (0.430)
    - 2015 EM Sell-Off: 4.095 (2.485)
  - Domestic Policy Rate x VIX:
    - 2020 COVID-19: 0.008* (0.004)
    - 2015 EM Sell-Off: -0.007 (0.012)
  - Domestic Policy Rate x Federal Fund Rate:
    - 2020 COVID-19: 0.160** (0.064)
    - 2015 EM Sell-Off: -2.219** (1.011)
- Additional table statistics:
  - Controls: Yes (both)
  - Country Fixed Effects: Yes (both)
  - R-Squared:
    - 2020 COVID-19: 0.635
    - 2015 EM Sell-Off: 0.244
  - Countries: 14 (both)
  - Observation:
    - 2020 COVID-19: 68
    - 2015 EM Sell-Off: 142
- Significance notation: *** p<0.01, ** p<0.05, * p<0.1
- Note on endogeneity: Some results might be overstated due to potential endogeneity issues.

### Impact of APP announcements on local-currency yields and currencies — method and identification
- Method: local projections (Jordà 2005) with extension by Teulings and Zubanov (2014) to capture dynamics and persistence of sovereign bond yield responses.
- Specification highlights (symbols preserved):
  - Dependent variable: 훥훥훥훥
푐푐,푡푡−1→푡푡+푝푝 — cumulative change in yield (percentage points) from t-1 to t+p for 10-year local-currency sovereign bonds (Bloomberg).
  - Key explanatory variables:
    - APP announcement dummy (퐴퐴퐴퐴퐴퐴): equals 1 on dates of APP announcements by EM central banks and 0 otherwise.
    - Global factor 퐺... : either (i) dummy for QE announcement by the Federal Reserve (assigned 1 on March 23) or (ii) the VIX index.
    - Percentage points decrease in policy rates (훥훥퐴퐴푓푓푝푝푖푖퐶퐶푦푦 푟푟퐺퐺푡푡퐶퐶
푐푐) from Bloomberg.
  - Four lags of all explanatory variables included; results robust to different lag specifications.
  - Country fixed effects (μ
푐푐
) absorb unobserved country-specific features.
  - Standard errors are robust (virtually identical if clustered at country level).
- Panel data:
  - Daily frequency from 13 emerging market economies.
  - Period: beginning of January 2020 to mid-May 2020.
  - Coefficient estimates reported for p = 0,...,6 (day 0 and 6 trading days after event) with one standard error confidence intervals.
- Currency impacts:
  - Same empirical set-up used for exchange rate cumulative changes; global factor in that analysis is QE announcement by the Federal Reserve.

### Dates for APP announcements (Online Annex Table 2.1.2 — sample of 13 EMs)
- Country — Date:
  - Chile — March 16, April 8
  - Colombia — March 23
  - Hungary — March 24, April 7, April 28
  - India — March 18, March 20, April 23
  - Indonesia — April 1
  - Korea — March 19, March 25, April 9
  - Mexico — April 21
  - Philippines — March 24, April 10
  - Poland — March 17, April 8
  - Romania — March 20
  - South Africa — March 25
  - Thailand — March 19, March 23, April 7
  - Turkey — March 31, April 17
- Note: sample expanded to include non-emerging markets (e.g., South Korea) and “operating twist” type announcements (e.g., Mexico, April 23 date for India). Results remain robust if individual countries are removed.
- Robustness checks and further steps:
  - Re-run tests dropping one EM at a time (12-EM panels) — results similar.
  - Address potential bias from closely spaced APP dates using Teulings and Zubanov (2014) extension to control for forward values of announcement dates.
  - Alternative combinations/identifications of APP dates produce similar results.

### Central bank policy responses and China case study
- Counting convention in Figure 2.1, panel 3: central bank policy options are counted only once (e.g., more than one rate cut counted as one action).
- Emerging market sample for Figure 2.1, panel 3 includes 50 central banks (list preserved in source).
- Online Annex Box 2.1 — China (key findings):
  - China did not experience the financial market stress seen in other EMs, but authorities faced challenges in maintaining supportive financial conditions.
  - PBOC cut short-term and one-year policy rates by 30 basis points.
  - Short-term interbank rates and one-year government bond yields fell as much as 180 and 100 basis points, respectively.
  - Repo-funded bond market purchases played an important role in policy transmission but were procyclical and amplified initial declines; later leverage reductions contributed to a bond sell-off.
  - Short-term interbank rate pass-through to bank funding costs was limited; deposit costs remained relatively sticky.
  - Yields on long-term government bonds fell by less than half as much as short-term bonds, steepening the yield curve to a five-year high.
  - Lower interest rates supported the economy but posed risks to bank profits and added to financial vulnerabilities.
- Figure note: Online Annex Figure 2.1.1 documents surge in repo borrowing amplifying initial declines and contributing to sell-off when leverage reduced; falling interbank rates provided relatively little pass-through to funding costs for key lenders such as banks and wealth management products.

### China: Daily Interbank Repo Trading — Market dynamics and interest-rate transmission
- Deposit alternatives saw surging inflows as interest rates fell, suggesting that banks might see funding cost pressures from further cuts to policy rates.
- Short-term rates had a limited impact on long-term yields, which remain linked to banking sector funding costs.
- Interest rates also led to a rise in asset management sector vulnerabilities.

### China: Key statistics and observed changes
- Net money market borrowing volumes by investment products surged 55 percent during the first half of 2020 to RMB 130 trillion.
- The PBOC has expanded its relending facilities to nearly RMB 2.2 trillion, targeted to micro and small businesses, the agricultural sector, and privately owned and manufacturing firms.
- China’s corporate-debt-to -GDP rose 10 percentage points in the first quarter.
- Household debt continued to rise, with rapid growth in housing-related debt and a rebound in retail stock market leverage.

### China: Policy actions taken
- Authorities lowered policy rates.
- The PBOC expanded relending facilities (low-cost funding for bank lending) to nearly RMB 2.2 trillion, targeted to vulnerable borrower segments.
- Authorities guided banks to increase lending and lower interest rates, particularly to micro and small businesses, the agricultural sector, and privately owned and manufacturing firms.

### China: Risks and vulnerabilities
- Expanded credit support measures may be adding to nonfinancial sector vulnerabilities (higher corporate and household leverage and debt servicing burdens).
- Asset management sector vulnerabilities increased as interest rates fell, with elevated leverage and interconnectedness to the broader financial system (net money market borrowing volumes at RMB 130 trillion).
- Bank funding cost pressures could arise if further policy rate cuts occur, given heavy flows into deposit alternatives and the continued linkage of long-term yields to banking sector funding costs.

### China: Policy priorities and recommendations
- Continue addressing interest rate transmission issues to increase the scope for traditional interest-rate-based monetary policy, easing debt servicing burdens and credit misallocation risks.
- Close remaining prudential regulatory gaps, particularly in the asset management sector.
- Improve market-based pricing of bank deposits.
- Accelerate bond market development by improving hedging mechanisms and diversifying the investor base.

*Sources: Bloomberg Finance L.P.; Country Central Banks; Citigroup; CEIC; People’s Bank of China; S&P Market Intelligence; IMF Staff Calculations; IMF staff.*

### 1. Variables in LSI 2. Alternative Aggregations of EM LSI

### onlineannex21 - 1. Variables in LSI 2. Alternative Aggregations of EM LSI

### LSI variables and aggregation
- LSI components shown (category headings preserved):
  - Exchange Rate LSI
    - FX bid-ask spread
    - Realized volatility
    - Implied Volatility
    - Risk-reversal ratio
  - Local Bonds LSI
    - Bond bid-ask spread
    - Realized volatility
    - Bond asset swap
    - Term-premium estimate
    - Non-resident flows
    - Bond volume deviations
    - Convertibility basis
- Indicators and methodological notes:
  - LSI Cross-Correlation weights.
  - LSI Non-Real-Time estimation.
  - LSI (GFSR).

### Country event studies for Bond LSI (selected narrative findings)
- Hungary:
  - Purchases in May seem to have had a large impact.
  - MNB decided to step up its program in the summer as conditions deteriorated again.
- Poland:
  - Large purchases in April combined with a step up in issuance.
  - Purchases tapered off as conditions improved in June, but trading volumes remained low.
- Indonesia:
  - Purchases were frontloaded and combined with large issuance.
  - Conditions improved further following the burden sharing agreement in July.
- South Africa:
  - APP has been relatively limited in size and the market remained concerned with increased issuance.
- Thailand:
  - Saw a swift improvement in the LSI, despite halting its APP in early April.
- India:
  - Improvement remained limited until RBI announced simultaneous sell/buy curve operations.
- Data sources for event studies: Bloomberg Finance L.P.; Country Central Banks; and IMF Calculations.

### Measuring the drivers of FX surprises — specification and variables
- Empirical goal: shed light on drivers of exchange rate movements in emerging markets and the role of domestic policies and global factors.
- Regression specification (symbols preserved in source):
  - Dependent variable: 훥훥 퐶퐶퐶퐶 푟푟푟푟퐶퐶푛푛 퐶퐶푦푦
푐푐,푚푚 — negative of percentage difference between realized end-of-month spot and one-month forward (end of month D−1), normalized by the latter (represents percentage gain of domestic currency above the forward).
- Domestic policy variables (푫푓...):
  - FXI (FX intervention): valuation adjusted changes in the stock of reserves including spot and derivative markets; scaled by actual stock of reserves. Note: estimates do not adjust for FX bond sales/purchases in some cases (e.g., Mexico).
  - Domestic policy rate: end-of-month values from Bloomberg.
- Global factors (푮...):
  - Effective Federal Fund rate at end of each month.
  - VIX Index averaged over the month.
- Controls:
  - Four interactions between domestic policies and global factors are included.
  - 푋푋
푐푐,푚푚: macroeconomic control variables — surprise indices on economic activity and inflation sourced from Citi and Bloomberg.
  - Country fixed effects (μ
푐푐
) included; standard errors are robust.
- Samples and periods:
  - Two regressions:
    - COVID-19: January to May of 2020.
    - 2015 China sell-off: April 2015 to February 2016.
  - Sample: 14 emerging market economies — Argentina, Brazil, Chile, China, Colombia, India, Indonesia, Mexico, Malaysia, Philippines, Russia, Thailand, Turkey, and South Africa.
  - Robustness: removing China yields consistent results.

### Empirical results (Online Annex Table 2.1.1 — Effects of Domestic and Global Factors on FX Surprises)
- Table entries (coefficients with standard errors in parentheses):
  - FX Intervention:
    - 2020 COVID-19: -0.813 (0.991)
    - 2015 EM Sell-Off: 2.196*** (0.706)
  - Domestic Policy Rate:
    - 2020 COVID-19: -0.376 (0.309)
    - 2015 EM Sell-Off: 1.788** (0.813)
  - VIX:
    - 2020 COVID-19: -0.525*** (0.100)
    - 2015 EM Sell-Off: 7.928 (6.748)
  - Federal Fund Rate:
    - 2020 COVID-19: -9.659*** (1.634)
    - 2015 EM Sell-Off: 0.203 (0.125)
  - FX Intervention x VIX:
    - 2020 COVID-19: 0.012 (0.019)
    - 2015 EM Sell-Off: -0.165*** (0.048)
  - FX Intervention x Federal Fund Rate:
    - 2020 COVID-19: 0.206 (0.430)
    - 2015 EM Sell-Off: 4.095 (2.485)
  - Domestic Policy Rate x VIX:
    - 2020 COVID-19: 0.008* (0.004)
    - 2015 EM Sell-Off: -0.007 (0.012)
  - Domestic Policy Rate x Federal Fund Rate:
    - 2020 COVID-19: 0.160** (0.064)
    - 2015 EM Sell-Off: -2.219** (1.011)
- Additional table statistics:
  - Controls: Yes (both)
  - Country Fixed Effects: Yes (both)
  - R-Squared:
    - 2020 COVID-19: 0.635
    - 2015 EM Sell-Off: 0.244
  - Countries: 14 (both)
  - Observation:
    - 2020 COVID-19: 68
    - 2015 EM Sell-Off: 142
- Significance notation: *** p<0.01, ** p<0.05, * p<0.1
- Note on endogeneity: Some results might be overstated due to potential endogeneity issues.

### Impact of APP announcements on local-currency yields and currencies — method and identification
- Method: local projections (Jordà 2005) with extension by Teulings and Zubanov (2014) to capture dynamics and persistence of sovereign bond yield responses.
- Specification highlights (symbols preserved):
  - Dependent variable: 훥훥훥훥
푐푐,푡푡−1→푡푡+푝푝 — cumulative change in yield (percentage points) from t-1 to t+p for 10-year local-currency sovereign bonds (Bloomberg).
  - Key explanatory variables:
    - APP announcement dummy (퐴퐴퐴퐴퐴퐴): equals 1 on dates of APP announcements by EM central banks and 0 otherwise.
    - Global factor 퐺... : either (i) dummy for QE announcement by the Federal Reserve (assigned 1 on March 23) or (ii) the VIX index.
    - Percentage points decrease in policy rates (훥훥퐴퐴푓푓푝푝푖푖퐶퐶푦푦 푟푟퐺퐺푡푡퐶퐶
푐푐) from Bloomberg.
  - Four lags of all explanatory variables included; results robust to different lag specifications.
  - Country fixed effects (μ
푐푐
) absorb unobserved country-specific features.
  - Standard errors are robust (virtually identical if clustered at country level).
- Panel data:
  - Daily frequency from 13 emerging market economies.
  - Period: beginning of January 2020 to mid-May 2020.
  - Coefficient estimates reported for p = 0,...,6 (day 0 and 6 trading days after event) with one standard error confidence intervals.
- Currency impacts:
  - Same empirical set-up used for exchange rate cumulative changes; global factor in that analysis is QE announcement by the Federal Reserve.

### Dates for APP announcements (Online Annex Table 2.1.2 — sample of 13 EMs)
- Country — Date (entries preserved):
  - Chile — March 16, April 8
  - Colombia — March 23
  - Hungary — March 24, April 7, April 28
  - India — March 18, March 20, April 23
  - Indonesia — April 1
  - Korea — March 19, March 25, April 9
  - Mexico — April 21
  - Philippines — March 24, April 10
  - Poland — March 17, April 8
  - Romania — March 20
  - South Africa — March 25
  - Thailand — March 19, March 23, April 7
  - Turkey — March 31, April 17
- Note: sample expanded to include non-emerging markets (e.g., South Korea) and “operating twist” type announcements (e.g., Mexico, April 23 date for India). Results remain robust if individual countries are removed.
- Robustness checks and further steps:
  - Re-run tests dropping one EM at a time (12-EM panels) — results similar.
  - Address potential bias from closely spaced APP dates using Teulings and Zubanov (2014) extension to control for forward values of announcement dates.
  - Alternative combinations/identifications of APP dates produce similar results.

### Central bank policy responses and China case study
- Counting convention in Figure 2.1, panel 3: central bank policy options are counted only once (e.g., more than one rate cut counted as one action).
- Emerging market sample for Figure 2.1, panel 3 includes 50 central banks (list preserved in source).
- Online Annex Box 2.1 — China (key findings):
  - China did not experience the financial market stress seen in other EMs, but authorities faced challenges in maintaining supportive financial conditions.
  - PBOC cut short-term and one-year policy rates by 30 basis points.
  - Short-term interbank rates and one-year government bond yields fell as much as 180 and 100 basis points, respectively.
  - Repo-funded bond market purchases played an important role in policy transmission but were procyclical and amplified initial declines; later leverage reductions contributed to a bond sell-off.
  - Short-term interbank rate pass-through to bank funding costs was limited; deposit costs remained relatively sticky.
  - Yields on long-term government bonds fell by less than half as much as short-term bonds, steepening the yield curve to a five-year high.
  - Lower interest rates supported the economy but posed risks to bank profits and added to financial vulnerabilities.
- Figure note: Online Annex Figure 2.1.1 documents surge in repo borrowing amplifying initial declines and contributing to sell-off when leverage reduced; falling interbank rates provided relatively little pass-through to funding costs for key lenders such as banks and wealth management products.

*Sources: Bloomberg Finance L.P.; Country Central Banks; Citigroup; IMF Staff Calculations; IMF staff.*

### 1. China: Daily Interbank Repo Trading

### 1. China: Daily Interbank Repo Trading

### Market dynamics and interest-rate transmission
- Deposit alternatives saw surging inflows as interest rates fell, suggesting that banks might see funding cost pressures from further cuts to policy rates.
- Short-term rates had a limited impact on long-term yields, which remain linked to banking sector funding costs.
- Interest rates also led to a rise in asset management sector vulnerabilities.

### Key statistics and observed changes
- Net money market borrowing volumes by investment products surged 55 percent during the first half of 2020 to RMB 130 trillion.
- The PBOC has expanded its relending facilities to nearly RMB 2.2 trillion, targeted to micro and small businesses, the agricultural sector, and privately owned and manufacturing firms.
- China’s corporate-debt-to -GDP rose 10 percentage points in the first quarter.
- Household debt continued to rise, with rapid growth in housing-related debt and a rebound in retail stock market leverage.

### Policy actions taken
- Authorities lowered policy rates.
- The PBOC expanded relending facilities (low-cost funding for bank lending) to nearly RMB 2.2 trillion, targeted to vulnerable borrower segments.
- Authorities guided banks to increase lending and lower interest rates, particularly to micro and small businesses, the agricultural sector, and privately owned and manufacturing firms.

### Risks and vulnerabilities
- Expanded credit support measures may be adding to nonfinancial sector vulnerabilities (higher corporate and household leverage and debt servicing burdens).
- Asset management sector vulnerabilities increased as interest rates fell, with elevated leverage and interconnectedness to the broader financial system (net money market borrowing volumes at RMB 130 trillion).
- Bank funding cost pressures could arise if further policy rate cuts occur, given heavy flows into deposit alternatives and the continued linkage of long-term yields to banking sector funding costs.

### Policy priorities and recommendations
- Continue addressing interest rate transmission issues to increase the scope for traditional interest-rate-based monetary policy, easing debt servicing burdens and credit misallocation risks.
- Close remaining prudential regulatory gaps, particularly in the asset management sector.
- Improve market-based pricing of bank deposits.
- Accelerate bond market development by improving hedging mechanisms and diversifying the investor base.

*Sources: Bloomberg Finance L.P.; CEIC; People’s Bank of China; S&P Market Intelligence; and IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/gfsr/2020/october/english/onlineannex21.pdf_