## Quantitative Easing and Credit Rating Agencies

## Source details

**Canonical URL:** [Quantitative Easing and Credit Rating Agencies](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022113-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022113-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022113-print-pdf.pdf.json)

---

### Eligibility criteria for CSPP purchases
- Securities must be eligible as collateral for Eurosystem credit operations.
- Securities must be denominated in euro.
- Securities must have a minimum first-best credit assessment of at least credit quality step 3 (rating of BBB- on S&P’s scale or equivalent) obtained from an external credit assessment institution. The four credit rating agencies recognized by the Eurosystem are Standard & Poor’s, Moody’s, FitchRatings and DBRS.
- Securities must have a minimum remaining maturity of six months and a maximum remaining maturity of 30 years at the time of purchase.
- Securities must have a yield to maturity above the deposit facility rate (DFR) at the time of purchase.
- The issuer must:
  - be a corporation established in the euro area, defined as the location of incorporation of the issuer,
  - not be a credit institution,
  - not have any parent undertaking which is a credit institution,
  - not be an asset management vehicle or a national asset management and divestment fund established to support financial sector restructuring and/or resolution.
- Conditional note: Corporate debt instruments issued by corporations incorporated in the euro area whose ultimate parent is not based in the euro area are also eligible for purchase under the CSPP, provided they fulfill all the other eligibility criteria.

### Additional restrictions and risk controls applied by the Eurosystem
- An issue share limit of 70% per international securities identification number (ISIN) on the basis of the outstanding amount.
- A limit at issuer group level.
- Extensive risk monitoring and due diligence activities are performed regularly prior to purchases.
- The Eurosystem Credit Assessment Framework (ECAF) incorporates comprehensive ex-post performance monitoring activities to mitigate credit risk associated with explicit reliance on CRAs.

### CSPP holdings and purchase patterns (key statistics)
- By June 2021, the ECB’s cumulative CSPP holdings amounted to EUR 282 billion.
- Around 78% of CSPP bond purchases were in the secondary market.
- During June 2016 to December 2017, the ECB purchased more than EUR 130 billion eligible corporate debt from euro area firms.
- Net CSPP purchases accounted for less than 10% of the APP purchases in 2016 and 2017, and reached larger shares from 2018 onwards as the monthly pace of net purchases of government bonds was reduced.

### Theory: The credit rating channel (mechanism and implications)
- Framework centers on conflicts of interest from CRAs’ “issuer pays” business model.
- Two key incentive problems:
  - CRAs have heterogeneous beliefs and issuers can “rating shop” for favorable ratings.
  - CRAs have incentives to inflate ratings to increase the probability of being selected to rate deals.
- Conjecture: Common knowledge of the ESCF’s design (notably the first-best rating rule), combined with oligopolistic competition and conflicts of interest among CRAs, may facilitate rating upgrades for bonds initially below but close to the CSPP eligibility frontier — the “credit rating channel.”
- Prediction: Marginal rating upgrades are more likely for borderline bonds around the eligibility frontier because reputational costs for such upgrades may be smaller than the gains from becoming CSPP-eligible.

### Market structure and CRA market shares
- The credit rating industry in the EU is highly concentrated (ESMA, 2017):
  - The “Big Three” control approximately 93% of the ratings business.
  - Moody’s and S&P together control 77.5% of the European market.
  - Fitch controls a further 15,65%.
  - DBRS has less than 2% of the total market share.

### Data and sample construction
- Dataset frequency: monthly (beginning of the month) from January 2015 to December 2017.
- Main data source: Bloomberg (security-level information including ISIN, outstanding amount, currency, security type, issue date, maturity date, eligibility for Eurosystem credit operations, yield-to-maturity, bid-ask spreads, credit ratings).
- Data cleaning steps:
  - Drop bonds with no rating over the full period (focus on credit rating vector).
  - Exclude on a monthly basis all bonds that do not comply with CSPP requirements except the minimum first-best credit assessment of at least credit quality step 3.
  - Exclude bonds with first-best rating of at least BBB- over the entire period but not eligible for Eurosystem credit operations.
- Final identification: 1750 publicly traded corporate bonds across 16 euro area countries.
- Pre-announcement distribution (beginning of March 2016):
  - Out of 1750 corporate bonds, around 65% were potentially eligible for the corporate QE and 35% were below the “minimum first-best credit assessment of at least credit quality step 3”.
- “Frozen list” approach: bonds available in March 2016 are frozen to track rating changes before and after the CSPP announcement on the same sample.
- The share of securities in the frozen list with at least a BBB- was roughly stable at around 63% from January 2015 to March 2016; after the corporate QE announcement a gradual increase in the share of potentially CSPP-eligible bonds is observed.
- Control variables: unemployment rates, industrial production, risk-free short-term rates, GDP and interest rate forecasts, PMI indices, stock market indices, Citigroup economic surprise indices, sovereign long-term interest rates, VSTOXX; firm-level variables from Orbis Europe and firms’ annual reports.

### Identification strategy and main empirical specification
- Key identifying rule: rule #3 — minimum first-best credit assessment of at least credit quality step 3.
- Dependent variable: QEeligible_{i,j,k,t} = 1 if CSPP rules #1 to #6 are satisfied; 0 otherwise.
- Deterministic eligibility indicator: QEeligible i,t = 1 if max_r R_{r,t} ≥ C_ECB where C_ECB is the “BBB-” cut-off and r ∈ {Standard & Poor’s, Moody’s, Fitch, DBRS}.
- Baseline linear probability model:
  - QEeligible i,j,k,t = Fˆ α + β1 1(t>March2016) + γ1 Λ_{i,j,k,t} + FEs + ε_{i,j,k,t}
  - 1(t>March2016) equals 1 after the CSPP announcement and 0 before.
  - Λ includes time-varying macro, country, industry, firm, and bond controls; fixed effects include remaining maturity, country, bond type, industry.
- Rating vectors taken at first day of each month; controls contemporaneous when taken at end of previous month; financial variables use average of previous month.
- Remaining maturity fixed effects use March 2016 reference with six buckets: 0-1 year, 1-3 years, 3-5 years, 5-8 years, 8-10 years, more than 10 years.
- Baseline standard errors clustered at the bond level; two-way clustering and probit specifications in robustness checks.

### Baseline estimation results and interpretation
- Key coefficient β1 (post-CSPP indicator effect):
  - Column (1), no controls: probability of becoming eligible increased by around 2.9 percentage points after the announcement.
  - Columns (2)-(4), with macro and volatility controls: β1 increased to around 3.5 percentage points and remained strongly significant.
  - With time-invariant unobserved characteristics and double-clustering, β1 oscillates between 3.0 and 4.5 percentage points, remaining statistically significant at the 1 percent level.
- The estimate of the constant suggests around 59% of the euro area corporate bonds are CSPP-eligible before March 2016.
- Interpretation: non-eligible corporate bonds are more likely to jump into CSPP-eligible territory after March 2016; magnitude stability supports causal interpretation via the credit rating channel.

### Temporal dynamics and placebo evidence
- Placebo regressions simulate treatment in every month from January 2015 to December 2017:
  - β_τ fluctuates around zero before March 2016; increase in probability of being CSPP-eligible becomes gradually visible after March 2016 without fading out.
- A strong and statistically significant temporal jump in probability occurred a few months after the announcement.

### Cross-sectional evidence around the eligibility frontier
- Bonds just below the frontier (BB+ in March 2016) much more likely to jump into eligibility than bonds with worse first-best ratings.
- Around 17% of bonds with an initial first-best rating BB+ were eligible as of December 2017.
- Timing: increase begins a few months after Eurosystem purchases start.
- Rating activity (∆Rating): bonds rated BB+ experienced statistically significant positive rating activity after March 2016; bonds rated BB or BBB- did not show significant changes.
- Matches to purchases:
  - Estimated that 14% of bonds in sample with first-best BB+ in March 2016 were purchased as a result of rating upgrades in the post-CSPP period.
  - Benchmark: 45% of the bonds rated at least BBB- in March 2016 in the sample were purchased under the corporate QE.

### Robustness checks and heterogeneity analyses (selected findings)
- Non-euro denominated assets as control group:
  - Expanded sample: 2180 bonds, 22 currencies; around 50% denominated in euro.
  - Difference-in-differences: post-CSPP dummy positive and significant overall; interaction (euro × post-CSPP) is positive and significant while post-CSPP alone is not — increase driven by euro-denominated securities.
  - For BB+ bonds, Euro effect positive and significant a few months after purchases begin; for BB and lower buckets Euro effect ≈ 0 or not significant.
- Non-recognized CRAs as control group:
  - Sample: 284 bonds rated by Egan Jones, Scope Ratings, AM Best; using first-best among these yields lower probability of jumping into eligibility after March 2016 (contrasts baseline).
  - Conclusion: credit rating channel driven by the four recognized CRAs and the first-best rating rule.
- Two-period difference-in-differences (pre vs post):
  - Balanced sample (~1200 bonds): bonds initially close but below frontier have around 16% higher probability of jumping into eligibility post-CSPP versus worse-rated bonds.
  - Rating upgrades: only bonds initially BB+ had significantly higher probability of experiencing one upgrade post-CSPP.
- Controlling for CRAs competition:
  - Competition measured by number of publicly available ratings in March 2016 (0–4); around 50% of bonds rated by two CRAs; among 875 bonds with two ratings, 77% by Moody’s and S&P.
  - Being rated by more CRAs associated with higher credit rating; interaction shows effect weakens somewhat after CSPP.
  - No single CRA consistently responsible for upgrades; baseline results driven by bonds with two and three ratings.
- Rating shopping margins:
  - Bonds with no change in number of ratings: post-CSPP coefficient similar to baseline.
  - Bonds with at least one change in number of ratings: post-CSPP coefficient stronger than baseline, suggesting rating shopping may have played a role.
- Step-by-step migration:
  - Main findings driven by marginal (one-step) credit rating changes (from BB+ to BBB-).
- Cross-country analysis:
  - Countries classified into vulnerable (Ireland, Greece, Spain, Italy, Cyprus, Portugal, Slovenia) and less vulnerable.
  - Shift from non-eligible to eligible observed for both groups; weak evidence of differential impact across country groups — adjustments broad-based.
- Bonds with no rating in March 2016 (19 bonds):
  - CSPP spurred demand for new credit ratings; majority obtained one that allowed them to become eligible.
  - Upward adjustments for unrated bonds were more rapid than for bonds initially below but close to the frontier.

### Additional methodological notes and reported statistics
- Main frozen-list regressions: 1,750 bonds (54,507 bond-month observations).
- Unfrozen-list regressions: 2,909 bonds (65,435 observations).
- Main reported coefficient examples:
  - QEeligible 1(t ≥ March2016): 0.031*** to 0.036*** across pooled specifications (standard errors typically 0.006–0.008).
  - Probit specification: 0.029*** to 0.036*** (standard errors 0.006–0.008).
  - Size-weighted (log amount issued): 0.570*** to 0.871*** (standard errors 0.130–0.159 in earlier columns).
- Significance notation: *** p < 0.01, ** p < 0.05, * p < 0.1.
- Standard errors: clustered at bond level; some specifications two-way clustered at bond and month levels.
- Notable covariate coefficients in sample specifications (examples preserved as reported):
  - Citigroup ESIt: 0.371*** (0.051) in early specification, declining to 0.171*** (0.041).
  - ΔUnemployment_{j,t}: -0.179*** (0.040) in initial specification.
  - ΔStockMarket_{j,t}: 0.986*** (0.341) in initial specification.
  - ΔVIX_t: negative and sometimes significant (examples: -0.054** (0.025); -0.068*** (0.026)); positive in small subsample (0.113** (0.054)).

### Extensions and alternative specifications
- Probit and logit specifications: marginal effects numerically close to linear probability model.
- Weighting by bond size: eligible amount issued increased significantly after CSPP.
- Unfrozen list results: probability of becoming CSPP-eligible increased significantly after March 2016 with estimates ranging from 3% to 7% across specifications.
- New issuance analysis: share of bonds issued after announcement ending up with one upgrade more pronounced for bonds rated BBB-; effect sharper when considering amount issued.

### Conclusion — financial and macro readings
- Financial view:
  - The discrete eligibility criteria of the Eurosystem corporate QE had a significant and asymmetric impact on CRAs’ rating activity.
  - Rating upgrades concentrated among bonds initially below, but close to, the eligibility frontier consistent with incentives created by the policy design.
- Macro view:
  - Given the small size of the euro area corporate bond market and the magnitude of estimates, localized rating adjustments induced by the corporate QE design are unlikely to have had adverse macroeconomic or financial stability implications.
  - Evidence suggests more recent “credit rating easing” policies during Covid-19 helped lower borrowing costs and avoided a downward spiral of downgrades and defaults.
  - Recommendation: consequences of relying explicitly on CRAs should be better acknowledged by central banks when designing monetary policy.

*Source: Excerpt from wpiea2022113-print-pdf — Quantitative Easing and Credit Rating Agencies (Working Paper No. WP/2022/113)*

### 1.  they are eligible as collateral for Eurosystem credit operations;

### 1.  they are eligible as collateral for Eurosystem credit operations;

### Eligibility criteria for CSPP purchases
- Securities must be eligible as collateral for Eurosystem credit operations.
- Securities must be denominated in euro.
- Securities must have a minimum first-best credit assessment of at least credit quality step 3 (rating of BBB- on S&P’s scale or equivalent) obtained from an external credit assessment institution. The four credit rating agencies recognized by the Eurosystem are Standard & Poor’s, Moody’s, FitchRatings and DBRS.
- Securities must have a minimum remaining maturity of six months and a maximum remaining maturity of 30 years at the time of purchase.
- Securities must have a yield to maturity above the deposit facility rate (DFR) at the time of purchase.
- The issuer must:
  - be a corporation established in the euro area, defined as the location of incorporation of the issuer,
  - not be a credit institution,
  - not have any parent undertaking which is a credit institution,
  - not be an asset management vehicle or a national asset management and divestment fund established to support financial sector restructuring and/or resolution.
- Conditional note: Corporate debt instruments issued by corporations incorporated in the euro area whose ultimate parent is not based in the euro area are also eligible for purchase under the CSPP, provided they fulfill all the other eligibility criteria.

### Additional restrictions and risk controls applied by the Eurosystem
- An issue share limit of 70% per international securities identification number (ISIN) on the basis of the outstanding amount.
- A limit at issuer group level.
- Extensive risk monitoring and due diligence activities are performed regularly prior to purchases.
- The Eurosystem Credit Assessment Framework (ECAF) incorporates comprehensive ex-post performance monitoring activities to mitigate credit risk associated with explicit reliance on CRAs.

### CSPP holdings and purchase patterns (key statistics)
- By June 2021, the ECB’s cumulative CSPP holdings amounted to EUR 282 billion.
- Around 78% of CSPP bond purchases were in the secondary market (reference: Figure 2A in the Appendix).
- During June 2016 to December 2017, the ECB purchased more than EUR 130 billion eligible corporate debt from euro area firms.
- Net CSPP purchases accounted for less than 10% of the APP purchases in 2016 and 2017, and reached larger shares from 2018 onwards as the monthly pace of net purchases of government bonds was reduced.

### Theory: The credit rating channel (mechanism and implications)
- The framework centers on conflicts of interest from CRAs’ “issuer pays” business model (White, 2010; Bolton et al., 2012; Efing and Hau, 2015; Baghai and Becker, 2018).
- Two key incentive problems:
  - CRAs have heterogeneous beliefs and issuers can “rating shop” for favorable ratings.
  - CRAs have incentives to inflate ratings to increase the probability of being selected to rate deals.
- Reputational concerns may not suffice to ensure truthful reporting (Mathis et al., 2009).
- Conjecture: Common knowledge of the ESCF’s design (notably the first-best rating rule), combined with oligopolistic competition and conflicts of interest among CRAs, may facilitate rating upgrades for bonds initially below but close to the CSPP eligibility frontier — termed the “credit rating channel.”
- Prediction: If bonds around the eligibility frontier have similar risk profiles and similar macroeconomic responses, reputational costs for marginal rating upgrades may be smaller than the gains from becoming CSPP-eligible, increasing upgrade likelihood for borderline bonds.

### Market structure and CRA market shares (as reported)
- The credit rating industry in the EU is highly concentrated (ESMA, 2017):
  - The “Big Three” control approximately 93% of the ratings business.
  - Moody’s and S&P together control 77.5% of the European market.
  - Fitch controls a further 15,65%.
  - DBRS has less than 2% of the total market share.

### Data and summary statistics (sample construction and descriptive facts)
- Dataset frequency: monthly (beginning of the month) from January 2015 to December 2017.
- Main data source: Bloomberg (security-level information including ISIN, outstanding amount, currency, security type, issue date, maturity date, eligibility for Eurosystem credit operations, yield-to-maturity, bid-ask spreads, credit ratings).
- Data cleaning steps:
  - Drop bonds with no rating over the full period (focus on credit rating vector).
  - Exclude on a monthly basis all bonds that do not comply with CSPP requirements except the minimum first-best credit assessment of at least credit quality step 3.
  - Exclude bonds with first-best rating of at least BBB- over the entire period but not eligible for Eurosystem credit operations (Eurosystem restricts accepted collateral to simple and transparent debt instruments).
- Final identification: 1750 publicly traded corporate bonds across 16 euro area countries.
- Pre-announcement distribution (beginning of March 2016):
  - Out of 1750 corporate bonds, around 65% were potentially eligible for the corporate QE and 35% were below the “minimum first-best credit assessment of at least credit quality step 3”.
- “Frozen list” approach:
  - Bonds available in March 2016 are frozen to track rating changes before and after the CSPP announcement on the same sample.
  - The share of securities in the frozen list with at least a BBB- was roughly stable at around 63% from January 2015 to March 2016; after the corporate QE announcement a gradual increase in the share of potentially CSPP-eligible bonds is observed.
- Distributional evidence:
  - January 2015 distribution is bimodal and similar to March 2016 (Kolmogorov-Smirnov test for equality of distribution functions cannot be rejected at the 1% level).
  - By December 2017, the distribution appears shifted rightward toward CSPP-eligibility and looks more unimodal and “closer” to a normal distribution, suggesting a gradual increase of CSPP-eligible securities from March 2016 onwards.
- Control variables collected at euro area and country level: unemployment rates, industrial production, risk-free short-term rates, GDP and interest rate forecasts, PMI indices, stock market indices, Citigroup economic surprise indices, sovereign long-term interest rates, Euro Stoxx 50 Volatility index (VSTOXX).
- Issuer-level variables matched from Orbis Europe and firms’ annual reports: total assets, leverage ratios, operating revenues, number of employees, other balance sheet and income statement figures.
- The list of bonds actually purchased under the CSPP is used and is available on the ECB website on a weekly basis.

### Empirical strategy and main results (identification logic)
- Unit of observation: bond i, country j, industry k, month t.
- Main dependent variable defined as QEeligible_{i,j,k,t} = 1 if CSPP rules #1 to #6 are satisfied; 0 otherwise.
- Identification exploits the legal restriction that the ECB can only purchase corporate bonds “with at least a BBB-” from at least one of the four recognized CRAs.
- Central claim: CSPP non-eligible bonds located below, but close to, the CSPP eligibility threshold are more likely to experience rating upgrades via the credit rating channel.
- Methodology compares rating changes of borderline non-eligible bonds (treatment group) to those of bonds farther from the eligibility frontier (control group), arguing that borderline high-yield bonds are more likely to jump into CSPP-eligibility after the announcement.
- Challenges addressed:
  - Monetary policy announcements respond to current and anticipated aggregate shocks, and CSPP itself may improve macroeconomic conditions (reducing financing costs, stimulating issuance, strengthening pass-through), complicating causal attribution.
  - Identification assumption: bonds located around the CSPP-eligibility frontier should be affected similarly by macroeconomic shocks, allowing isolation of the CSPP design’s effect on CRAs’ behavior.

*Source: Excerpt from wpiea2022113-print-pdf - 1.  they are eligible as collateral for Eurosystem credit operations;*

### Section II.A, where the most important rule we rely on for identification is the rule #3 (i.e. minimum

### Section II.A — Identification and empirical strategy for CSPP eligibility

### Identification rule and dependent variable
- The most important rule for CSPP identification is rule #3 (i.e. minimum first-best credit assessment of at least credit quality step 3).
- CSPP-eligibility follows a deterministic rule expressed as an indicator:
  - QEeligible i,t “ 1 if max r tR r u ≥ C ECB , where 1t.u is the indicator variable, max r tR r u represents the first best rating, R r is rating assigned by the rating agency r (r = Standard & Poor’s, Moody’s, Fitch and DBRS) converted in numerical values and C ECB is the “BBB-” cut-off.
- The dependent variable in the baseline panel regression is the probability for bond i to be eligible for the ECB’s corporate QE at month t (QEeligible i,j,k,t ).
- Main specification (linear probability model baseline):
  - QEeligible i,j,k,t = Fˆ α + β1 1t>March2016u + γ1 Λ i,j,k,t + FEs + ε i,j,k,t
  - 1t>March2016u is an indicator equal to 1 after the announcement of the CSPP and 0 before.
  - Λ i,j,k,t includes time-varying controls: (i) euro area forward-looking macroeconomic and financial variables (e.g. expected GDP growth, expected slope of the yield curve), (ii) global euro area risk indexes (e.g. VIX, Citigroup economic surprise index), (iii) country-level characteristics (e.g. unemployment rate, PMI indices, stock market indices), (iv) industry-level characteristics (e.g. stock market indices), (iv) firm-level characteristics (leverage ratio), and (v) bond-level characteristics (bid-ask spread, return, amount issued).
  - FEs are fixed effects (remaining maturity, country, bond type, industry).
- Rating vectors are taken at the first day of each month; controls are considered contemporaneous when taken at the end of the previous month; for financial variables the average of the previous month is used.
- Remaining maturity fixed effects use March 2016 as reference and six buckets: 0-1 year, 1-3 years, 3-5 years, 5-8 years, 8-10 years, more than 10 years.
- Baseline standard errors are clustered at the bond level; alternative clustering and probit specifications are considered in robustness checks.

### Baseline estimation results and interpretation
- Key coefficient of interest: β1 (post-CSPP indicator effect).
- Column (1), no controls: probability of becoming eligible increased by around 2.9 percentage points after the announcement of the corporate QE.
- Columns (2)-(4), controlling for expected euro area macroeconomic outlook and financial volatility: β1 increased to around 3.5 percentage points and remained strongly significant.
- When controlling for time-invariant unobserved characteristics (columns (5)-(8)) and using double-clustering (column (9)), β1 oscillates between 3.0 and 4.5 percentage points, remaining statistically significant at the 1 percent level.
- The estimate of the constant suggests that around 59% of the euro area corporate bonds are CSPP-eligible before March 2016.
- Interpretation: non-eligible corporate bonds are more likely to jump into CSPP-eligible territory after March 2016; magnitude stability after adding controls supports a causal interpretation of the credit rating channel.

### Placebo test and temporal dynamics
- Placebo regression simulated treatment in every month from January 2015 to December 2017 using:
  - QEeligible i,j,k,τ = Fˆ α + Στ≠March2016 βτ 1t=τu + FEs + ε i,j,k,τ
- Figure 6 (described): βτ fluctuates around zero before CSPP announcement; increase in probability of being CSPP-eligible becomes gradually visible after March 2016 without fading out.
- A strong and statistically significant temporal jump in probability occurred a few months after the announcement.

### Cross-sectional variation in CRAs response — non-linear effects below the eligibility frontier
- Hypothesis: bonds just below the eligibility frontier in March 2016 are more likely to jump into eligibility after the programme compared to bonds farther away.
- Regression for each first-best rating bucket ω below the frontier (D to BB+), months April 2016–December 2017:
  - QEeligible i,j,k,τ = Fˆ ρ ω 1t max r tR r u March2016 = ω u + ε i,j,k,τ  for ω ∈ Ω , τ ∈ T
  - ρ ω represents fraction of bonds not eligible in March 2016 that became eligible in a specific month afterwards.
- Key quantitative findings:
  - Bonds with a first-best rating of BB+ in March 2016 are much more likely to jump into eligibility than bonds with worse first-best ratings.
  - Around 17% of bonds with an initial first-best rating BB+ were eligible as of December 2017.
  - Timing: the increase in this share begins a few months after the start of purchases by the Eurosystem.
- Interpretation: evidence consistent with sluggishness of credit ratings; CSPP had a significant impact on eligibility of bonds initially slightly below the cut-off.

### Cross-sectional variation around the eligibility frontier — rating activity (∆Rating)
- To test differential rating activity for bonds just below (BB+) vs just above (BBB-), estimate:
  - ∆Rating i,j,k,t = α + β1 1t>March2016u + γ1 Λ i,j,k,t + FEs + ε i,j,k,t , for ω ∈ Ω F (from BB to BBB-)
  - Dependent variable ∆Rating is change in first-best credit rating from month t to t+1.
- Results (Table 2):
  - Bonds below but close to the frontier (BB+) experienced a statistically significant positive rating activity after March 2016 (more upgrades than downgrades).
  - Bonds rated BB or BBB- did not show significant changes in rating activity post-CSPP.
- Matching purchases data:
  - Estimated that 14% of bonds in sample with first-best BB+ in March 2016 were purchased as a result of rating upgrades in the post-CSPP period.
  - As a benchmark, 45% of the bonds rated at least BBB- in March 2016 in the sample were purchased under the corporate QE.

### Incorporating additional controls
- Alternative specifications include country and other euro area time-varying controls (unemployment rates, stock market indices, Citigroup economic surprise index) and bond-level information (lagged bid-ask spread, lagged return, amount issued).
- Table 3 and Table 8A: coefficients for additional controls display expected signs and are statistically significant in most cases.
- Post-CSPP time dummy coefficient remains statistically significant across specifications and oscillates around 3.0 percentage points, consistent with baseline.
- Conclusion: adding controls has little effect on the post-CSPP coefficient, lending further support to the credit rating channel prediction.

### Further evidence supporting identification assumption
- Identification relies on assumption that, absent CSPP, bonds slightly below frontier (bond J) and slightly above (bond I) respond similarly to business cycle fluctuations after controlling for bond type, maturity, industry, country and pricing-risk characteristics.
- Firm-level evidence around frontier (BB+ vs BBB- in March 2016):
  - Kernel density of log total assets in 2016 for BB+ (32 firms) and BBB- (27 firms) are very close; Kolmogorov-Smirnov test for equality of distribution functions cannot be rejected at the 1% level.
  - Distributions for total assets, leverage ratio, number of employees, operating revenues, net income, profit margins and return on equity in 2015 and 2016 are similar across groups.
- Interpretation: results are not driven by systematic differences in observable firm-level characteristics.

### Robustness checks — overview and selected results
A. Non-euro denominated assets as a control group
- New sample: 2180 bonds covering 22 currencies; around 50% denominated in euro.
- Difference-in-differences with euro denomination as treatment:
  - Column (1) Table 4, no controls: post-CSPP dummy positive and significant; probability of jumping into rating eligibility increases by around 1.3 percentage points after March 2016 (regardless of currency).
  - Columns (2)-(8): interaction term (euro × post-CSPP) is positive and significant while post-CSPP coefficient is no longer different from zero — implying increase is driven by euro-denominated securities.
- Extended Eq.4 including Euro i,j dummy:
  - Rating max r tR r u ≥ C ECB = Fˆ ρ ω 1t max r tR r u March2016 = ω u + ν ω Euro i,j + ε i,j,k,τ
  - ν ω measures additional euro-denomination effect for each rating bucket.
- For bonds initially BB+, ν ω positive and significant a few months after purchases begin; for BB bonds ν ω not significant; for lower buckets ν ω ≈ 0.
- Conclusion: upward rating adjustments post-CSPP are concentrated on euro-denominated bonds close to frontier.

B. Non-recognized CRAs as a control group
- Restrict sample to ratings by non-recognized CRAs (Egan Jones, Scope Ratings, AM Best): 284 bonds have at least one rating from these CRAs in Jan 2015–Dec 2017.
- Repeating baseline regression using first-best rating among these CRAs and defining eligibility as if these were recognized:
  - Table 5: probability of jumping into eligibility territory is lower after March 2016 (contrasts with baseline).
  - Restricting to bonds rated in March 2016 (55 bonds) yields same conclusion (Table 9A).
- Interpretation: credit rating channel is driven by the four recognized CRAs and the first-best rating rule.

C. Two-period difference-in-differences (pre vs post)
- Balanced sample: pre defined as March 2016 eligibility, post as December 2017 eligibility (around 1200 bonds still in sample; for bonds leaving sample use latest available month).
- First exercise (eligibility movement): interaction term positive and significant only for bonds initially close but below frontier; these bonds have around 16% higher probability of jumping into eligibility post-CSPP versus worse-rated bonds.
- Second exercise (rating upgrades): dependent variable = dummy for experiencing only one rating upgrade. Only bonds initially first-best rated BB+ had significantly higher probability of experiencing one upgrade in post-CSPP versus other buckets.
- Results not valid for more than one upgrade, confirming marginal upgrade importance.

D. Controlling for CRAs competition
- Competition measured as number of publicly available ratings in March 2016 (discrete 0–4); around 50% of bonds are rated by two CRAs; among 875 bonds with two ratings, 77% by Moody’s and S&P.
- Distinguish high competition (≥ sample median 2) vs low competition (< 2).
- Table 10A: being rated by more CRAs associated with higher credit rating; interaction term shows effect weakens somewhat after CSPP.
- Subsample analysis dividing bonds into four groups by which CRAs rated them: baseline results appear driven by bonds with two and three ratings as of March 2016; no single CRA consistently responsible for upgrades.

E. Rating shopping — extensive vs intensive margin
- Bonds with no changes in number of publicly available ratings (Jan 2015–Dec 2017): post-CSPP coefficient positive and similar to baseline (Table 12A).
- Bonds with at least one change in number of available ratings: post-CSPP coefficient stronger than baseline (Table 6), suggesting rating shopping may have played a role.

F. Step-by-step rating migration
- Compare bonds with only one change in first-best rating vs those with more than one change:
  - Table 7 and Table 13A: main findings driven by marginal (one-step) credit rating changes (from BB+ to BBB-), consistent with theory.

G. Cross-country analysis
- Classify countries into vulnerable (Ireland, Greece, Spain, Italy, Cyprus, Portugal, Slovenia) and less vulnerable (others) per Altavilla et al., 2016.
- Kernel distributions show shift of density from non-eligible to eligible territory post-CSPP for both groups.
- Table 14A: weak evidence of different impact across country groups — adjustments are broad-based.

H. Bonds with no rating in March 2016
- Sample includes 19 bonds with no rating in March 2016 but with a rating during the period; out of these, 12 bonds obtained a rating after March 2016 (analysis continued beyond excerpt).

*Source: wpiea2022113-print-pdf — Section II.A*

### 2016.  In this sub-section, we investigate whether the rating they got after the CSPP announcement

### wpiea2022113-print-pdf - 2016.  In this sub-section, we investigate whether the rating they got after the CSPP announcement

### Impact on new credit ratings and eligibility
- We estimate Eq.4 for the 19 bonds with no publicly available rating in March 2016.
- The estimates of the key coefficient (plotted in Figure 16A in the Appendix) suggest that the CSPP has spurred the demand for new credit ratings.
- The vast majority of bonds with no ratings in March 2016 obtained eventually one that allowed them to become eligible for the corporate QE.
- Upward rating adjustments for these unrated bonds were more rapid than the one observed in the subsample of corporate bonds initially located below, but close to, the CSPP eligibility frontier.

### I. Probit model
- Table 15A in the Appendix reproduces the baseline specification using a probit model (F = Φ).
- For tractability, marginal effects are reported.
- The key coefficient is numerically very close to the one in the linear probability model, suggesting robustness to the selection of the functional form.
- The results are also robust to a logit model.

### J. Weighting by bond size
- Baseline specification treats bonds equally; concern that findings might be driven by small-sized bonds.
- We include bond-size information by weighting the dependent variable by the log of the amount issued for each bond.
- Results reported in Table 16A show that the eligible amount issued increased significantly after the announcement of the CSPP.

### K. Unfrozen list of bonds
- Baseline analysis uses the “frozen list” (securities with at least one credit rating as of March 2016).
- Results for the “unfrozen list” (reported in Table 17A in the Appendix) confirm that the probability of becoming CSPP-eligible increased significantly after March 2016.
- The estimates of this increase range from 3% to 7% and are statistically significant across all specifications.
- The larger impact is likely linked to the increase of bond issuance activity in the post-CSPP period, as documented by Abidi and Miquel-Flores (2018), Arce et al., (2018) and Grosse-Rueschkamp et al., (2018).
- Focusing on new issuances (securities not in the frozen list): the share of bonds issued after the announcement and ending up getting one rating upgrade with respect to the rating of bonds of the same issuer at March 2016 was more pronounced for bonds rated BBB- (Figure 17A in the Appendix).
- This finding is even sharper when considering the amount issued instead of the number of bonds and is fully in line with the prediction of the credit rating channel.

### VI. Conclusion — Financial and macro readings
- Financial view:
  - The discrete nature of the eligibility criteria of the Eurosystem corporate QE had a significant and asymmetric impact on CRAs rating activity.
  - After the launch of the policy, rating upgrades were mostly noticeable for bonds initially located below, but close to, the eligibility frontier.
  - This effect is concentrated where CRAs’ and firms’ incentives are expected to be more sensitive to the policy design.
- Macro view:
  - Given the small size of the euro area corporate bond market and the magnitude of our estimates, the localized rating adjustments induced by the design of the corporate QE are unlikely to have had adverse macroeconomic or financial stability implications.
  - Evidence suggests that the more recent "credit rating easing" policies (e.g. "frozen ratings" in the ECB’s collateral easing package as well as the inclusion of high-yield securities in the Fed’s QE programs) provided during Covid-19 helped lower the cost of borrowing and avoided the downward spiral triggered by an avalanche of downgrades and defaults.
  - The consequences of relying explicitly on CRAs must be better acknowledged by central banks when designing monetary policy.

*Source: wpiea2022113-print-pdf - 2016.  In this sub-section, we investigate whether the rating they got after the CSPP announcement*

### References

### References

### Literature on quantitative easing, asset purchases, and monetary policy
- Abidi, N. and Miquel-Flores, I. (2018),Who benefits from the corporate QE? A regression disconti-nuity design approach, Working Paper Series No 2145, European Central Bank
- Altavilla, C., Carboni, G. and Motto, R. (2016),Asset purchase programmes and financial markets: Lessons from the euro area,Working Paper Series No 1864, European Central Bank
- Arce, O., Gimeno, R. and Mayordomo, S. (2018),Making room for the needy:  The credit-reallocation effects of the ECB’s corporate QE, Working Papers No 1743, Banco de Espa ̃na
- Bartocci,  A.,  Burlon,  L.,  Notarpietro,  A.  and  Pisani,  M.  (2017),Macroeconomic  effects  of  non-standard monetary policy measures in the Euro Area:  The role of corporate bond purchases, Working Paper Series No 1136, Banca d’Italia
- Bauer, M. D. and Rudebusch, G. D. (2014), The signaling channel for Federal Reserve bond pur-chases,International Journal of Central Banking, 10(3), 233-290
- Christensen, J. H. E. and Rudebusch, G. D. (2013), The response of interest rates to US and UK quantitative easing,The Economic Journal,122(564), F385-F414
- Di Maggio, M., Kermani, A. and Palmer, C. (2016),How quantitative easing Works:  Evidence on the Refinancing Channel, NBER Working Paper No 22638
- Gagnon,  J.,  Raskin,  M.,  Remache,  J.,  Sack,  B.  (2011),  The  financial  market  effect  of  Federal Reserve’s Large-Scale Asset Purchases,International Journal of Central Banking, 14(3), 3-43
- Joyce,  M.  A.  S.,  Lasaosa,  A.,  Stevens,  I.,  and  Tong,  M.  (2011),  The  financial  market  impact  of quantitative easing in the United Kingdom,International  Journal  of  Central  Banking, 7(3), 113-161
- Krishnamurthy, A. and Vissing-Jorgensen, A. (2011),The effects of quantitative easing on interest rates:  Channels and implications for policy, NBER Working Paper No 17555
- Kuttner,  K.  N.  (2018),  Has  quantitative  easing  earned  a  place  in  the  Fed’s  toolbox?,Journal  of Economic Perspectives, 32(4), 121-146
- Dell’Ariccia, G., Rabanal, P., and Sandri, D. (2018), Unconventional monetary policies in the Euro Area, Japan, and the UK,Journal of Economic Perspective, 32(4), 147-172
- Borio,  C.  and  Zabai,  A.  (2016),Unconventional  monetary  policies:  a  re-appraisal,  BIS  Working Papers No 570, Bank for International Settlements
- Greenlaw, D., Hamilton, J. D., Harris, E. S. and West, K. D. (2018),A skeptical view of the impact of the Fed’s balance sheet, NBER Working Paper No 24687

### Studies on corporate sector purchase programmes, bond markets, and disintermediation
- De  Santis,  R.,  Geis,  A.,  Juskaite,  A.  and  Vaz  Cruz,  L.  (2018),The  impact  of  the  corporate  sec-tor  purchase  programme  on  corporate  bond  markets  and  the  financing  of  euro  area  non-financial corporations,Economic Bulletin, May 2018, European Central Bank
- ECB  (2017),The  ECB’s  corporate  sector  purchase  programme:   Its  implementation  and  impact, Economic Bulletin, Issue 4/2017, European Central Bank
- Ertan, A., Kleymenova, A., and Tuijn, M. (2018),The  real  effects  of  financial  disintermediation: Evidence  from  the  ECB  Corporate  Sector  Purchase  Program,  Chicago Booth Research Paper No 18-06
- De Santis,  R. and Holm-Hadulla,  F. (2017),Flow  effects of central bank asset purchases on euro area sovereign bond yields: evidence from a natural experiment,Working Paper Series No 2052, European Central Bank
- Cahn, C., Duquerroy, A., and Mullins, W. (2017),Unconventional monetary policy and bank lending relationships, Working Paper Series No 659, Banque de France
- Montagna, M. and Pegoraro, S. (2017),The transmission channels of quantitative easing:  Evidence from the cross-section of bond prices and issuance, Unpublished Manuscript
- Di Maggio, M., Kermani, A. and Palmer, C. (2016),How quantitative easing Works:  Evidence on the Refinancing Channel, NBER Working Paper No 22638

### Credit rating agencies, rating behavior, rating shopping, and conflicts of interest
- Altman,  E.  I.,  Kao,  D.  L.  (1992),  The  implications  of  corporate  bond  ratings  drift,Financial Analysts Journal, 48(3), 64-75
- Altman, E. I., Rijken, H. A. (2004), How rating agencies achieve rating stability,Journal of Banking & Finance, 28(11), 2679-2714
- Altman,  E.  I.,  Rijken,  H.  A.  (2006),  A  point-in-time  perspective  on  through-the-cycle  ratings, Financial Analysts Journal, 62(1), 54-70
- Cantor,  R.  (2004),  An  introduction  to  recent  research  on  credit  ratings,Journal  of  Banking  & Finance, 28(11), 2565-2573
- Cheng, M., Neamtiu, M. (2009), An empirical analysis of changes in credit rating properties:  Time-liness, accuracy and volatility,Journal of Accounting and Economics, 47(1-2), 108-130
- Baghai, R. P. and Becker, B. (2018), Non-rating revenue and conflicts of interest,Journal  of  Financial Economics, 127(1), 94-112
- Efing, M. and Hau, H. (2015), Structured debt ratings:  Evidence on conflicts of interest,Journal of Financial Economics, 116(1), 46-60
- Mathis,  J.,  McAndrews,  J.  and  Rochet,  J.  (2009),  Rating  the  raters:   Are  reputation  concerns powerful enough to discipline rating agencies?,Journal of Monetary Economics, 56(5), 657-674
- Griffin, J. M., Nickerson, J., and Tang, D. Y. (2013), Rating shopping or catering?  An examination of the response to competitive pressure for CDO credit ratings,The  Review  of  Financial  Studies, 26(9), 2270-2310
- He, J., Qian, J. and Strahan, P. E. (2015), Does the market understand rating shopping?  Predicting MBS losses with initial yields,The Review of Financial Studies, 29(2), 457-485
- Farhi, E., Lerner J., and Tirole J. (2013), Fear of rejection?  Tiered certification and transparency, The RAND Journal of Economics, 44(4), 610-631
- Skreta,  V.  and  Veldkamp,  L.  (2009),  Rating  shopping  and  asset  complexity:  A  theory  of  rating inflation,Journal of Monetary Economics, 56(5), 678-695
- Sangiorgi, F., and Spatt, C. (2013), Opacity, credit rating shopping, and bias.Management Science, 63, 4016-4036
- Sangiorgi, F., and Spatt, C. (2017), The economics of credit rating agencies,Foundations and Trends in Finance, 34(12), 1-116
- Sean Chu, C. (2014),Empirical  analysis  of  credit  ratings  inflation  as  a  game  of  incomplete  infor-mation, Working Paper
- S&P (2018),2017 Annual global corporate default study and rating transitions, S&P Global
- ESMA (2017),Report on CRA market share calculation, European Securities and Markets Authority
- Garcia Alcubilla, R. and Ruiz Del Pozo, J. (2012),Credit rating agencies on the watch list:  Analysis of European regulation, Oxford:  Oxford University Press
- White, L. J. (2010), Markets:  The credit rating agencies,The  Journal  of  Economic  Perspectives, 24(2), 211-226
- Partnoy,  F. (2010),  Historical perspectives on the financial crisis:  Ivar Kreuger, the credit-rating agencies, and two theories about the function, and dysfunction, of markets,Yale Journal on Regulation, 26(2), 431-443

### Central bank collateral, risk management, and operational frameworks
- Bindseil,  U.  and  Papadia,  F. (2009),Risk  management  and  market  impact  of  central  bank  credit operations, in:  Bindseil, U., Gonzalez, F. and Tabakis, E.Risk Management for central banks and other public institutions, Cambridge University Press
- Bindseil, U., Gonzalez, F. and Tabakis, E. (2009),Risk Management for Central Banks and Other Public Investors, Cambridge University Press
- Bindseil, U., Corsi, M., Sahel, B. and Visser, A. (2017),The  Eurosystem  collateral  framework  ex-plained, Occasional Paper Series No 189 European Central Bank
- Coppens, F., Gonzalez, F. and Winkler, G. (2009),The performance of credit rating systems in the assessment of collateral used in Eurosystem monetary policy operations, Working Paper Series No 65, European Central Bank
- Gonzalez, F. and Molitor, P. (2009),Risk mitigation measures and credit risk assessment in central bank policy operations, in:  Bindseil, U., Gonzalez, F. and Tabakis, E.Risk Management for central banks and other public institutions, Cambridge University Press
- ECB (2015),The  financial  risk  management  of  the  Eurosystem’s  monetary  policy  operations, European Central Bank
- Nyborg, K. (2016),Collateral frameworks:  The open secret of central banks(1st Ed.), Cambridge: Cambridge University Press
- Hartmann, P., Smets, F., (2018),The first 20 years of the European Central Bank:  Monetary policy, Working Paper Series No 2219, European Central Bank

### Empirical analyses, data, and methodological contributions
- Alp, A. (2013), Structural shifts in credit rating standards,Journal of Finance, 68(6), 2435-2470
- Lando, D., Skodeberg, T. M. (2002), Analyzing rating transitions and rating drift with continuous observations,Journal of Banking & Finance, 26(2-3), 423-444
- L ̈offler, G. (2004), An anatomy of rating through the cycle,Journal of Banking & Finance, 28(3), 695-720
- L ̈offler,  G.  (2005),  Avoiding  the  rating  bounce:   Why  rating  agencies  are  slow  to  react  to  new information,Journal of Economic Behavior & Organization, 56(3), 365-381
- Matthies,  A.  B.  (2013),Empirical  research  on  corporate  credit-ratings:  A  literature  review,  SFB 649 Discussion Paper 2013-003
- Coppens, F., Gonzalez, F. and Winkler, G. (2009),The performance of credit rating systems in the assessment of collateral used in Eurosystem monetary policy operations, Working Paper Series No 65, European Central Bank
- Bongaerts, D., Cremers, K. M. and Goetzmann, W. N. (2012), Tiebreaker:  Certification and multiple credit ratings,The Journal of Finance, 67(1), 113-152
- Keys, B. J., Mukherjee, T., Seru, A., and Vig, V. (2010), Did securitization lead to lax screening? Evidence from subprime loans,The Quarterly Journal of Economics, 125(1), 307-362

### Appendix note and authorship
- Appendix for “Quantitative Easing and Credit Rating Agencies” — Nordine Abidi, Matteo Falagiarda, Ixart Miquel-Flores. Contains additional findings and robustness checks, presents results in the order they appear in the main text. Author affiliations and e-mail addresses provided in the appendix. The views expressed are those of the authors and do not necessarily reflect those of the European Central Bank, the Eurosystem or the International Monetary Fund.

*Source: wpiea2022113-print-pdf - References*

### 2017.  The dependent variable is a dummy indicating the eligibility for the corporate QE. Standard errors are clustered 

### wpiea2022113-print-pdf - 2017.  The dependent variable is a dummy indicating the eligibility for the corporate QE. Standard errors are clustered

### Main empirical findings on QE eligibility
- Baseline bond-level regressions (1750 frozen-list bonds, January 2015 - December 2017) show a positive and statistically significant main effect for the post-March 2016 indicator on the QE eligibility dummy:
  - QEeligible 1(t ≥ March2016) coefficient: 0.031*** (columns (1)-(4) and pooled specifications report 0.031*** to 0.036***; standard errors typically 0.006–0.008).
  - Probit specification: QEeligible 1(t ≥ March2016) = 0.029*** to 0.036*** across models (standard errors 0.006–0.008).
  - Weighting by log amount issued: QEeligible 1(t ≥ March2016) = 0.570*** to 0.871*** across specifications (standard errors 0.130–0.159 in earlier columns; higher point estimates than unweighted regressions).
- Robustness and fixed effects:
  - Maturity fixed effects, country fixed effects, bond type fixed effects, and industry fixed effects are included progressively; R-squared increases substantially when adding these (examples: R-squared from 0.001 to 0.479 in main frozen-list tables).
  - Standard errors are clustered at the bond level; in column (9) of multiple tables standard errors are two-way clustered at the bond and month level.
- Additional covariates with notable coefficients (frozen-list and subsamples):
  - Citigroup ESIt: 0.371*** (0.051) in an early specification, declining to 0.171*** (0.041) in later specs.
  - ΔUnemploymentj,t: -0.179*** (0.040) in initial specifications, effects attenuate when richer controls added.
  - ΔStockMarketj,t: 0.986*** (0.341) in initial specifications, with variability across models.
  - pE r GDPs t,next year and pE r Slopest,next year enter positively in several subsample regressions (e.g., for bonds with no change in number of CRAs: pE r GDP next year = 0.044*** (0.008); pE r Slope next year = 0.035*** (0.006)).
  - ΔVIXt displays negative and sometimes significant coefficients in many weighted and subsample models (examples: -0.054** (0.025); -0.068*** (0.026)) and positive in a small subsample of bonds with multiple rating-bucket changes (0.113** (0.054)).

### CRAs, ratings dynamics, and competition
- CRAs competition (1750 frozen-list):
  - Competitioni,March2016 coefficient: positive and significant (e.g., 0.169*** (0.025) in column (2)).
  - Interaction Competitioni,March2016 * 1(t ≥ March2016) is negative and sometimes significant (example: -0.023* (0.012) to -0.028** (0.013)), indicating differential post-March2016 effects by competition measure.
- Pivotal CRA role and number of ratings:
  - Results by number of ratings in March 2016:
    - For 1-rating bonds: QEeligible 1(t ≥ March2016) = 0.032* (0.017) in column (1) and 0.009 (0.021) in column (2).
    - For 2-rating bonds: QEeligible ≈ 0.040*** (0.009) to 0.037*** (0.007).
    - For 3-rating and 4-rating bonds coefficients vary (e.g., 0.022** (0.010) for 3 ratings; 0.055 (0.056) for 4 ratings).
  - Interactions with pivotal CRA indicators produce mixed effects; one interaction shows a negative and significant effect: 1(t ≥ March2016) * pS&P & Fitch1 = -0.115*** (0.031) in a specification (column corresponds to 2-ratings group).
- Non-recognized CRAs hypothetical exercise (55 bonds in frozen list also rated by non-recognized CRAs):
  - QEeligible 1(t ≥ March2016) coefficients are negative in these restricted regressions (example: -0.092* (0.047) to -0.096** (0.048)), suggesting hypothetical ineligibility if non-recognized CRAs were recognized by the Eurosystem.
  - Observations: 1,716; R-squared moves from 0.026 to 0.490 across nested specifications.

### Vulnerability, frozen/unfrozen lists, and subsample dynamics
- Vulnerable versus less vulnerable countries (full frozen-list regressions):
  - Main post-March2016 effect: 0.024*** (0.007) in simple specification, rising to 0.040*** (0.007) in richer specifications.
  - Interaction Vulnerablej * 1(t ≥ March2016) shows positive coefficients in some specifications (example: 0.035* (0.018), 0.049*** (0.019)), indicating larger post-March2016 changes for vulnerable-country bonds in some specifications.
- Unfrozen list (2,909 bonds):
  - QEeligible 1(t ≥ March2016) = 0.031*** to 0.038*** across columns (standard errors 0.007–0.008); weighted and richer-fixed-effect specifications show similar patterns.
  - pE r GDP next year and pE r Slope next year positive and significant in many unfrozen-list specifications (examples: pE r GDP next year = 0.043*** (0.008) in column (2); pE r Slope next year = 0.040*** (0.006)).
- Subsamples by rating-change history:
  - No change in number of CRAs (1,405 bonds): QEeligible 1(t ≥ March2016) = 0.031*** to 0.039***; pE r GDP next year and pE r Slope next year positive and significant; ΔVIXt negative and significant in several columns (e.g., -0.054** (0.025)).
  - More than one single rating-bucket change (220 bonds): QEeligible coefficients generally small and not significant in many columns (example: 0.016 (0.012) to 0.018 (0.015)); ΔVIXt positive and significant in some columns (0.113** (0.054); 0.159*** (0.061)).

### Distributional evidence and auxiliary figures
- Firm- and bond-level distributions (selected figure notes):
  - Figure 5A: Kernel density of log total assets for firms with BB+ vs BBB- rated bonds in March 2016 (2016 data). Kolmogorov-Smirnov test: null hypothesis for equality of distribution functions cannot be rejected at the 1% level.
  - Figures 6A–12A: Present median and 25–75 percentiles for total assets, leverage ratio (total liabilities/total assets), number of employees, operating revenues, net income, profit margins, and ROE for BB+ vs BBB- issuer groups in 2015 and 2016.
  - Figures 14A–15A: Dynamic kernel-density distributions of bonds by first-best rating for January 2015, March 2016, and December 2017 for less vulnerable and formerly vulnerable countries.
  - Figure 13A: Number of ratings per bond in March 2016 (19 bonds rated during sample but not in March 2016).
  - Figure 16A: Coefficient from Eq.4 for 19 bonds with no publicly available ratings in March 2016 (they had at least one rating during the sample period and were excluded from the frozen list).
  - Figure 17A: Share of bonds issued after March 2016 ending up getting one upgrade relative to the rating of bonds of the same issuer in March 2016.

### Methodological and robustness notes
- Sample coverage: Main frozen-list regressions use 1,750 bonds (54,507 bond-month observations reported across many tables); unfrozen-list regressions use 2,909 bonds (65,435 observations reported).
- Standard errors: clustered at the bond level across most specifications; column (9) in many tables implements two-way clustering at bond and month levels.
- Significance notation used throughout: *** p < 0.01, ** p < 0.05, * p < 0.1.
- Regression variants: OLS linear probability models, Probit models, size-weighted regressions (log amount issued), subsamples by rating-change patterns, by number of ratings, and exercises replacing recognized CRAs with non-recognized CRAs.

*Source: Quantitative Easing and Credit Rating Agencies — Working Paper No. WP/2022/113 (figures and tables and accompanying notes as provided in the content unit).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022113-print-pdf.pdf_