## wp1823

## Source details

**Canonical URL:** [wp1823](https://www.imf.org/-/media/files/publications/wp/2018/wp1823.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1823.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1823.pdf.json)

---

### Data and sample construction
- Panel of 38 countries with sufficient bond issuance and BIS credit data.
- Dealogic covers bond issuance from 1980-2016; after filters the sample is 41 countries, and BIS coverage yields 38 countries with sufficient issuance.
- Sample consists of about 110,000 bonds for issuers operating in these 38 countries.
- Exclusions: financial issuers, money market instruments, floating-rate bonds, and bonds where yield to maturity cannot be calculated.
- Country entry years range from 1980 (France and the US) to 2002 (China and India); coverage begins by 1985 for 16 countries; number of years per country ranges from 37 to 15.
- Four main country-year quantities:
  - IMF WEO series on real GDP growth.
  - Credit to GDP ratio (ratio of BIS series on credit obtained by private non-financial borrowers to IMF WEO GDP series, both nominal, and in dollars), reported in percentage points.
  - High-yield (HY) share of issuance: proceeds from high yield issues as a share of proceeds from all issues, in percentage points.
  - Price-based measures (credit spreads): defined as the difference between the 90th percentile yield and the 10th percentile yield within country year, divided by the mean of this difference by country over the full sample.
- Data treatments:
  - HY share and spread winsorized at the 5th and 95th percentile, and imputed with lags when missing.
  - Growth in the credit to GDP ratio measured as the log change over five years, multiplied by 100.
  - Credit booms: country years where five year growth in credit/GDP is above the 75th percentile of this quantity in the previous ten years (rolling historical window).
  - Five year change in HY share and five year change in spread used to study evolution of lending standards and spreads.
- Notes:
  - Dealogic provides a dummy for investment grade issues; results similar if HY share constructed based on rating.
  - Construction differences vs Greenwood & Hanson (2013): assign issues to countries based on issuers’ countries of operations, include government bonds, use level rather than log of HY share, and define the share based on funds raised rather than principal amounts to be repaid.
  - BIS reports total credit obtained by domestic resident borrowers; BIS recently added data for a 44th country, Colombia.

### HY share as a measure of lending standards: stylized facts
- Two stylized facts:
  - Lending standards in bond markets broadly move in line with survey measures of bank lending standards, when both available.
  - Lending standards in bond markets are procyclical: they tend to loosen during economic booms and tighten during busts.
- Survey coverage:
  - Of the 38 countries in the sample, loan officer surveys are available for 14.
  - US SLOOS conducted since 1990; most other country series begin in early 2000s.
- Empirical relationships:
  - HY share and bank loan officer surveys move in line where both available; correlations sign depends on survey scaling.
  - Procyclicality tests: for horizons k = 1 to 5, HY share rises during periods of rising growth and falls during periods of falling growth.
  - Regression evidence (Table 4):
    - Panel A (country fixed effects only): coefficients on 1_{G_{t,t−k}>0} for k=1..5: 3.50, 1.37, 1.60, 1.60, 1.22 (t-stats: 4.78, 2.98, 3.05, 3.06, 2.84).
    - Panel B (country and year fixed effects): coefficients for k=1..5: 1.40, 0.10, 0.56, 0.82, 1.07 (t-stats: 1.76, 0.13, 0.80, 1.25, 2.12).
- Robustness on issuer age:
  - Matched issuers to Worldscope in US, Canada, India, Japan, and Korea.
  - High yield issuers are not substantially younger: average differences in age around two years; small differences in fraction of firms older than five years.
- Interpretation:
  - HY share is a reasonable cross-country market-finance measure of lending standards; it moves with survey measures when both available but reflects only a small portion of credit.
  - Dynamics consistent with Minsky (1986) quotation on history of success diminishing margins of safety.

### Main results: lending standards separate good booms from bad booms
- Summary:
  - Episodes of high credit growth tend to be followed by lower output growth.
  - Unconditionally, no clear relationship between evolution of lending standards and subsequent growth.
- Conditional on credit booms (defined using the 75th percentile of five year credit/GDP growth over the previous ten years):
  - Substantial variation in subsequent three year real GDP growth persists even after conditioning on booms.
  - Panel B of Table 5 key observations:
    - The worst booms are not simply the biggest in terms of credit growth; after conditioning on credit booms there is no clear relationship between magnitude of credit growth and subsequent GDP growth.
    - Evolution of lending standards over booms helps flag problematic booms: booms followed by the worst subsequent growth are accompanied by rising HY shares.
    - ΔHY_{t,t−5} monotonically falls across columns as subsequent growth rises in Panel B (worse subsequent growth associated with larger increases in HY share during the boom).
- Quantitative local-projection evidence (section 4.1):
  - Baseline regressions follow Jordà (2005) local projection approach.
  - Key regressors: ∆HY_{t,t−5} (scaled to unit variance), Credit boom dummy, Credit boom × ∆HY_{t,t−5}; controls include five year growth in credit/GDP and two lags of real GDP growth; horizons h ∈ [1,5].
  - Main finding: interaction coefficient δ negative and statistically significant out to four years; peak effect three years ahead.
  - Quantitative effect: Given a credit boom, a one standard deviation increase in the HY share over the course of the boom is followed by cumulative growth lower by nearly 1.5 percentage points three years later.
    - Note: one standard deviation move in the average change in the HY share is around 4 percentage points per year.
  - Non-parametric: within credit booms, the probability that subsequent three year real GDP growth is within the lowest quintile is three times higher given a “bad HY indicator” (highest quintile average change in HY share) relative to a “good HY indicator” (lowest quintile).
- Spreads and supply vs. composition:
  - Spreads defined as difference between 90th and 10th percentile yields within country-year, scaled by country mean of this difference.
  - Spreads generally compress in the buildup to credit booms, consistent with an outward shift in credit supply.
  - Spreads rise when HY share rises, particularly during credit booms, but for booms with deteriorating lending standards spreads do not rise substantially; they compress less but are, on balance, flat.
  - Regressions (Table 7) show coefficient on credit boom dummy and interaction term balance out, implying credit booms with a one standard deviation increase in HY share have flat, not rising, credit spreads.
  - Interpretation: joint movement suggests supply shocks play a significant role; spreads not rising enough implies supply shocks at least as important as borrower composition shocks.

### Additional results and robustness (section 4.2)
- Global lending standards:
  - Regressions include interaction of credit boom dummy with average five year change in global HY share (∆HY^G_{t,t−5}, scaled).
  - Credit boom × ∆HY^G_{t,t−5} coefficients (1yr–5yr): -0.61, -1.15, -2.27, -2.65, -2.63 (t-stats: -2.32, -2.04, -2.21, -1.89, -1.54).
  - Global evolution supports view that shifting credit supply is key.
- HY share reversals:
  - A one standard deviation increase in HY share over previous five years is followed by a one-third standard deviation reduction in HY share over subsequent five years (Table A.2).
  - Coefficients on ∆HY_{t,t−5} for subsequent 1–5 years: -0.30, -0.33, -0.35, -0.30, -0.37 (t-stats: -6.09, -4.25, -3.99, -3.52, -3.74).
- Definition of credit boom:
  - Baseline: 75th percentile of five year growth in credit/GDP over previous ten years; full-sample alternative threshold close to 22 percent.
  - Results similar using full-sample threshold (Table 9): Credit boom × ∆HY_{t,t−5} coefficients (1yr–5yr): -0.51, -1.11, -1.53, -1.21, -0.61 (t-stats: -2.86, -3.41, -2.51, -2.10, -0.72).
  - Stricter definitions (e.g., Dell’Ariccia et al. (2012)) yield much weaker results; looser versions closer to baseline.
  - Using BIS credit/GDP gap as proxy for high credit growth yields robust results; Table 10: interaction coefficients (1yr–5yr): -0.54, -1.01, -1.28, -0.78, -0.38 (t-stats: -3.74, -2.59, -2.06, -1.43, -0.60). Effects with credit gap are two to three years ahead rather than three to four years.
- Advanced economies vs emerging markets:
  - Sample: 25 advanced economies and 13 emerging markets.
  - Table 11: results stronger than baseline for advanced economies in statistical significance and magnitude.
  - Advanced economy Credit boom × ∆HY_{t,t−5} (1yr–5yr): -0.49, -1.16, -1.92, -1.87, -1.44 (t-stats reported).
  - Emerging markets coefficients have same signs but smaller magnitudes and not statistically significant.
- Household credit and house prices:
  - Household credit booms (defined using five year growth in household credit to GDP above 75th percentile of full sample) that are also broad credit booms are followed by substantially lower growth over subsequent five years (Table 12: Credit boom × 1_{HH Credit boom} coefficients (1yr–5yr): -0.89, -2.26, -3.33, -3.72, -3.68).
  - HY share informative during household credit booms: HH Credit boom × ∆HY_{t,t−5} coefficients (1yr–5yr): -0.82, -1.38, -1.19, -1.10, -1.38 (Table 13).
  - Baseline results robust to controlling for five year growth in house price indices relative to GDP (Appendix Table A.7); Credit boom × ∆HY coefficients (1yr–5yr): -0.53, -1.11, -1.63, -1.47, -0.77.
- Further robustness exercises where baseline holds:
  1. Additional credit quantity controls (Credit/GDP level, Bond Proceeds/GDP, Credit/GDP gap).
  2. Shorter horizons (three- and four-year definitions of ∆HY) yield similar patterns.
  3. Winsorization of dependent variable at 5th and 95th percentiles retains results.
  4. Alternative HY share definitions (scaled by country mean, defined by number of issues, excluding new issuers) yield qualitatively similar results; results weaker when HY defined by number of issues.
  5. Post-1995 sample results similar.
  6. Using subsequent real GDP per capita as dependent variable yields similar results, confirming economic rather than demographic effects.
- HY share and spreads together:
  - Table A.16: including both ∆HY_{t,t−5} and S_t − S_{t−5} shows both related to subsequent output growth; Credit boom × ∆HY coefficients (1yr–5yr): -0.40, -0.91, -1.36, -1.29, -0.88; Credit boom × S_t − S_{t−5} coefficients (1yr–5yr): -0.46, -0.68, -0.67, -0.48, -0.62.

### Information in prices (section 4.3)
- Credit spreads in primary bond markets:
  - Five-year change in spread during credit booms (S_t − S_{t−5}) associated with lower subsequent growth when interacted with credit boom.
  - Table 14: Credit boom × (S_t − S_{t−5}) coefficients (1yr–5yr): -0.51, -0.78, -0.80, -0.59, -0.67 (t-stats: -2.27, -2.46, -2.11, -1.46, -1.21).
- Heterogeneity:
  - Spreads-based results stronger in advanced economies (Table A.15 Panel A) with Credit boom × S_t − S_{t−5} (1yr–5yr): -0.55, -0.98, -1.15, -1.01, -1.36.
  - For emerging markets Panel B suggests rising credit spreads are unconditionally followed by lower growth, but interaction coefficients not significant.
- Complementarity:
  - HY share and spreads together provide complementary information for contemporaneous evaluation of credit booms.

### Key numeric highlights from main tables
- Table 6 (baseline local projections):
  - ∆HY_{t,t−5} coefficients (1yr–5yr): 0.14, 0.33, 0.37, 0.21, 0.25 (t-stats: 1.32, 1.48, 1.42, 0.89, 0.90).
  - Credit boom coefficients (1yr–5yr): -0.35, -1.00, -1.47, -1.54, -1.67.
  - Credit boom × ∆HY_{t,t−5} (1yr–5yr): -0.45, -1.00, -1.45, -1.34, -0.93 (t-stats: -2.56, -3.21, -2.36, -2.39, -1.09).
  - Mean subsequent cumulative real GDP growth: 1yr 2.40, 2yr 4.81, 3yr 7.16, 4yr 9.50, 5yr 11.76.
- Table 5 (Panel B, conditional on credit boom) quintiles:
  - Real GDP growth t,t+3 by quintile Q1..Q5: -6.1, 2.5, 6.2, 9.5, 17.3.
  - ∆Credit/GDP t,t−5 by quintile Q1..Q5: 38.2, 33.2, 31.4, 33.3, 39.5.
  - ∆HY t,t−5 by quintile Q1..Q5: 1.4, -0.8, -0.9, -1.2, -2.4.
- Table 7 (five-year change in spreads regressions):
  - ∆HY_{t,t−5}: 0.14 (2.91)
  - Credit boom × ∆HY_{t,t−5}: 0.26 (2.78)
  - Mean of depvar: -0.06.
- Table A.2 (HY share reversal): a one standard deviation increase in past HY share followed by ~one-third standard deviation reduction in subsequent five years (coefficients: -0.30 to -0.37 across 1–5 years).

### Policy implications and interpretation
- Practical construction:
  - Real-time measures of lending standards can be constructed from currently available data (market issuance, HY share).
- Monitoring and early warning:
  - Lending standards help separate good booms from bad booms: credit booms accompanied by deteriorating lending standards are followed by worse growth.
  - Policy makers should pay particular attention to credit booms with deteriorating lending standards.
- Causality interpretation:
  - Timing (peak adverse effect several years after booms) and procyclicality of HY share make reverse causality less likely.
  - Two interpretations: HY share as a useful indicator formed from other information, or HY share as part of causal mechanisms; procyclicality aligns more naturally with behavioral/extrapolative narratives (Minsky, Kindleberger, Bordalo et al. 2016) but results remain policy-relevant under either view.

*Italicized source attribution: Content based solely on the supplied excerpt from wp1823.*

### 1980.  Krishnamurthy & Muir (2016) use secondary market prices, including from historical sources that allow earlier

### wp1823 - 1980.  Krishnamurthy & Muir (2016) use secondary market prices, including from historical sources that allow earlier

### Data and sample construction
- Panel of 38 countries with sufficient bond issuance and BIS credit data.
- Dealogic covers bond issuance from 1980-2016 for bonds issued by firms operating in 81 countries; after filters the sample is 41 countries, and BIS coverage yields 38 countries with sufficient issuance.
- Sample consists of about 110,000 bonds for issuers operating in these 38 countries.
- Exclusions: financial issuers, money market instruments, floating-rate bonds, and bonds where yield to maturity cannot be calculated.
- Country entry years range from 1980 (France and the US) to 2002 (China and India); coverage begins by 1985 for 16 countries; number of years per country ranges from 37 to 15.
- Four main country-year quantities:
  - IMF WEO series on real GDP growth.
  - Credit to GDP ratio (ratio of BIS series on credit obtained by private non-financial borrowers to IMF WEO GDP series, both nominal, and in dollars), reported in percentage points.
  - High-yield (HY) share of issuance: proceeds from high yield issues as a share of proceeds from all issues, in percentage points.
  - Price-based measures (credit spreads): defined as the difference between the 90th percentile yield and the 10th percentile yield within country year, divided by the mean of this difference by country over the full sample.
- Data treatments:
  - HY share and spread are winsorized at the 5th and 95th percentile, and imputed with lags when missing.
  - Growth in the credit to GDP ratio is measured as the log change over five years, multiplied by 100.
  - Credit booms: country years where five year growth in credit/GDP is above the 75th percentile of this quantity in the previous ten years (rolling historical window).
  - Five year change in HY share and five year change in spread are used to study evolution of lending standards and spreads.
- Notes:
  - Dealogic provides a dummy for investment grade issues; results similar if HY share is constructed based on rating instead of the dummy.
  - Construction builds on Greenwood & Hanson (2013) HY share for the US; differences: assign issues to countries based on issuers’ countries of operations, include government bonds, use level rather than log of HY share, and define the share based on funds raised rather than principal amounts to be repaid.
  - BIS reports total credit obtained by domestic resident borrowers; BIS recently added data for a 44th country, Colombia.
  - Syndicated loans noted as potential but incomplete international datasets.

### HY share as a measure of lending standards: stylized facts
- Two documented stylized facts:
  - Lending standards in bond markets broadly move in line with survey measures of bank lending standards, when both are available.
  - Lending standards in bond markets are procyclical: they tend to loosen during economic booms and tighten during busts.
- Survey coverage:
  - Of the 38 countries in the sample, loan officer surveys are available for 14.
  - US SLOOS has been conducted since 1990; most other country series begin in early 2000s.
- Empirical relationships:
  - When available together, HY share and bank loan officer surveys move in line: for countries where survey indices rise when standards tighten, survey measures are negatively correlated with the HY share; for countries that scale indices oppositely, correlations are positive.
  - Procyclicality tests:
    - For each horizon k (k = 1 to 5), sample is split by whether growth was rising or falling on average from t−k to t.
    - The average change in the HY share from t−k to t is computed conditional on rising vs falling growth.
    - Findings: HY share rises during periods of rising growth and falls during periods of falling growth, for horizons from 1 year to 5 years.
    - One-year horizon: increases in the growth rate coincide with increases in the HY share, and vice versa.
  - Regression evidence (Table 4):
    - Dependent variable: average change in the HY share from t−k to t.
    - Main independent variable: dummy for whether growth was rising from t−k to t.
    - Controls include five year growth in credit/GDP and two lags of GDP growth; standard errors clustered at the country level.
    - Panel A (country fixed effects only): strong procyclicality at all horizons.
    - Panel B (country and year fixed effects): smaller coefficients, statistical significance only at a horizon of five years, indicating some synchronization across countries.
- Robustness on issuer age:
  - Matched issuers to Worldscope in US, Canada, India, Japan, and Korea.
  - High yield issuers are not substantially younger: average differences in age around two years; small differences in fraction of firms older than five years.
- Interpretation:
  - HY share is a reasonable cross-country market-finance measure of lending standards.
  - HY share moves with bank survey measures when both available, but directly reflects only a small portion of credit.
  - Dynamics consistent with Minsky (1986) quote: “Current views about financing ... reflect the past and, in particular, the recent past. A history of success will tend to diminish the margin of safety that business and bankers require ... [while] a history of failure will do the opposite.”

### Main results: lending standards separate good booms from bad booms
- Summary context:
  - Episodes of high credit growth tend to be followed by lower output growth, consistent with the credit boom literature.
  - Unconditionally, no clear relationship between evolution of lending standards and subsequent growth.
- Conditional on credit booms:
  - Credit booms defined using the 75th percentile of five year credit/GDP growth over the previous ten years (rolling window).
  - Even after conditioning on credit booms, substantial variation in subsequent three year real GDP growth persists.
  - Key observations from Panel B of Table 5:
    - The worst booms are not simply the biggest in terms of credit growth; after conditioning on credit booms there is no clear relationship between magnitude of credit growth and subsequent GDP growth.
    - Evolution of lending standards over booms helps flag problematic booms: booms followed by the worst subsequent growth are accompanied by rising HY shares.
    - The change in the HY share over the previous five years, ΔHY_t,t−5, monotonically falls across columns of Panel B as subsequent growth rises (i.e., worse subsequent growth associated with larger increases in HY share during the boom).
- Additional findings summarized in Section 4.1:
  - Within credit booms, spreads do rise for booms with deteriorating lending standards, but do not seem to rise enough, suggesting a role for credit supply constraints or insufficiencies in price-based correction.
- Section 4.2 (overview of further results):
  - Provides further evidence on credit supply.
  - Discusses sensitivity to the definition of credit booms.
  - Examines relative strength of results for advanced economies and household credit booms.
- Section 4.3:
  - Investigates information contained in spreads.

### Policy implications
- Real-time measures of lending standards can be constructed from currently available data (market issuance, HY share).
- Lending standards help separate good booms from bad booms: credit booms accompanied by deteriorating lending standards are followed by worse growth.
- Policy makers should pay particular attention to credit booms with deteriorating lending standards.

*Italicized source attribution: Content based solely on the supplied excerpt from wp1823.*

### 4.1    The HY share and output growth

### 4.1    The HY share and output growth

### Research question and empirical approach
- Objective: Assess whether the evolution of lending standards (measured by the HY share) during credit booms helps separate “good” credit booms from “bad” ones and forecasts subsequent real GDP growth.
- Empirical method: Jordà (2005) local projection approach following Krishnamurthy & Muir (2016).
- Baseline regression (summary):
  - Dependent variable: cumulative subsequent real GDP growth at horizon h (ln(y_{i,t+h} / y_{i,t})).
  - Key regressors: ∆HY_{t,t−5} (scaled to have unit variance), Credit boom dummy, Credit boom × ∆HY_{t,t−5}.
  - Controls: year fixed effects, country fixed effects, five year growth in credit/GDP, two lags of real GDP growth.
  - Standard errors clustered at the country level.
  - Horizons considered: h ∈ [1,5].

### Main findings on output
- Booms with deteriorating lending standards (rising HY share) are followed by lower subsequent growth.
- The interaction coefficient δ (effect of a rising HY share over the course of a boom) is negative and statistically significant out to four years; the peak effect is three years ahead.
- Quantitative effect:
  - Given a credit boom, a one standard deviation increase in the HY share over the course of the boom is followed by cumulative growth lower by nearly 1.5 percentage points three years later.
  - Note: A one standard deviation move in the average change in the HY share is around 4 percentage points per year.
- Non-parametric evidence:
  - Within the sample of credit booms, the probability that subsequent three year real GDP growth is within the lowest quintile is three times higher given a “bad HY indicator” (highest quintile average change in HY share) relative to a “good HY indicator” (lowest quintile average change in HY share).
  - Much of the predictive information lies in the tails of the distribution of changes in the HY share during booms.

### Interpretation and causality considerations
- Timing: The peak adverse effect occurs several years after the boom, which, together with the procyclicality of the HY share, makes reverse causality (output declines causing HY rises) less likely.
- Two interpretations:
  - Lending standards are useful indicators of future downturns even if they reflect market participants’ expectations formed from other information.
  - It is not natural to interpret rising HY share as lenders anticipating downturns, because the HY share is procyclical (lenders are unlikely to extend lower-quality credit if they expect a downturn).
- The procyclicality of the HY share aligns more naturally with behavioral/extrapolative narratives (Minsky, Kindleberger, Bordalo et al. 2016) than with classical amplification-only models; however, results are policy-relevant under either view.

### Credit spreads and supply vs. composition
- Definition: Credit spreads = difference between the 90th and 10th percentile yields within country-year, scaled by the country mean of this difference.
- Empirical pattern:
  - Spreads generally compress in the buildup to credit booms, consistent with an outward shift in credit supply.
  - The relationship between lending standards and credit spreads has the expected sign: spreads rise when the HY share rises, particularly during credit booms.
  - However, for booms with deteriorating lending standards, spreads do not rise substantially; they compress less but are, on balance, flat.
  - In regressions (Table 7), the coefficient on the credit boom dummy and the interaction term balance each other out, implying credit booms with a one standard deviation increase in HY share have flat, not rising, credit spreads.
- Interpretation:
  - The joint movement of lending standards and credit spreads suggests supply shocks play a significant role during credit booms.
  - The HY share can move because lenders evaluate credit risk differently (supply) or because the composition of borrowers changes (demand/composition).
  - Because spreads do not rise enough for booms with deteriorating lending standards, supply shocks seem at least as important as borrower composition shocks.

### Practical implications and external validity
- Simplicity and scope:
  - The empirical approach uses only country and year fixed effects, historical credit growth, the evolution of lending standards (HY share), and lagged growth; nonetheless it produces meaningful predictive results.
  - The HY share can be constructed internationally and provides time series visibility for a reasonably broad cross section of countries.
- Policy relevance:
  - The evolution of lending standards in bond markets appears to be a reasonable measure of broader lending standards and helps distinguish good from bad credit booms.
  - Whether policymakers should attempt to tighten lending standards depends on interpretation (indicator vs. causal determinant), but the HY share is clearly relevant for monitoring and early warning.

*Source: IMF Working Paper — section 4.1, "The HY share and output growth".*

### 4.2    Additional results and robustness

### 4.2    Additional results and robustness

### 4.2.1    Further evidence on the role of credit supply
- Global lending standards at the time of a credit boom in a given country are informative about subsequent GDP growth: credit booms that coincide with deteriorating lending standards globally are followed by lower GDP growth in subsequent years.
- Regressions include an interaction of the credit boom dummy with the average five year change in the global HY share (weighted by proceeds).
- The evolution of global lending standards supports the view that shifting credit supply is the key factor (demand or borrower composition in any one country unlikely to dominate global lending standards).
- HY share reversals:
  - Increases in the HY share over the previous five years are generally followed by a reduction in the HY share over the subsequent five years (see Appendix Table A.2).
  - Dependent variable: average change in the HY share over the subsequent 1-5 years; change on both sides is scaled to have unit variance.
  - Magnitude: a one standard deviation increase in the HY share over the previous five years is followed by a one-third standard deviation reduction in the HY share over the subsequent five years.
  - The overall size of the reversal grows over time.
- Consistency with other findings: López-Salido et al. (2017) find that increases in the HY share in the US are followed by a reduction in firms’ reliance on debt relative to equity, suggesting credit supply subsequently shifts inwards.

### 4.2.2    Definition of credit boom
- Baseline definition:
  - Credit booms defined backward looking using the 75th percentile of five year growth in credit/GDP over the previous ten years as the threshold.
  - Evolution of lending standards based on HY shares of issuance only up until yeart.
- Alternative full-sample definition:
  - Results similar if using the 75th percentile of five year growth in credit/GDP over the full sample.
  - Full sample threshold for five year growth in credit/GDP is close to 22 percent.
  - Figure 10 (impulse response) and Table 9 show qualitatively and quantitatively similar results when using the full-sample threshold.
- Stricter definitions:
  - Krishnamurthy & Muir (2016) use the 92nd percentile in their work.
  - Dell’Ariccia, Igan, Laeven & Tong (2012) methodology identifies only 39 country years as credit booms in this sample, versus 173 in the baseline definition.
  - Results using the Dell’Ariccia et al. (2012) definition are much weaker (Panel A of Appendix Table A.4).
  - A looser version of the Dell’Ariccia et al. methodology yields results closer to baseline (Panel B of Appendix Table A.3 and Appendix Table A.4).
  - Panel B of Table 5: conditional on a boom defined at the 75th percentile, booms followed by lower growth are not systematically bigger.
- US experience:
  - According to baseline methodology, the US has no credit booms over the sample period.
  - Five year growth in credit/GDP in the run up to the Great Recession in the US was high but consistently below the 75th percentile of previous international experience.
  - The Dell’Ariccia et al. (2012) methodology and its looser version do not flag any US sample years as a credit boom.
  - Two alternative definitions that flag some US history produce weaker results concentrated in the short term (75th percentile within country’s history; median five year growth in credit/GDP over the previous ten years—see Appendix Tables A.5 and A.6).
- Credit gap alternative:
  - Using the credit/GDP gap from the BIS as a proxy for high credit growth yields robust results.
  - Table 10: HY share informative conditional on a positive credit gap.
  - Effects with credit gap are only two to three years ahead, rather than three to four years ahead in baseline.

### 4.2.3    Advanced economies and emerging markets
- Sample composition: 25 advanced economies and 13 emerging markets (emerging markets typically with shorter coverage).
- Table 11: baseline methodology results separately for advanced and emerging markets.
  - Results are stronger than baseline for advanced economies in statistical significance and magnitude.
  - For emerging markets, coefficients have same signs but smaller magnitudes and are not statistically significantly different from zero.
- Possible explanations for the split:
  - Sample skew toward advanced economies: emerging market sample may be too small and statistical power too low.
  - Measure of lending standards is based on market finance and may be less relevant for bank-dependent emerging markets.
  - Market finance may be noisier in emerging markets due to global capital flow cycles, making the signal less useful; Section 4.2.1 tests this and finds consistent but statistically insignificant results.
- Sovereign downgrade note:
  - As government bond issuance is included in the HY share, concern that sovereign downgrades drive results is addressed: there are no episodes in the sample where advanced economy sovereign debt is downgraded from investment grade to high yield during a credit boom.

### 4.2.4    Household credit and house prices
- Household credit booms:
  - BIS provides household credit data with shorter time series coverage.
  - Household credit booms defined where five year growth in household credit to GDP is above the 75th percentile of the full sample of this variable.
  - Table 12: three dummies partition credit booms into (i) booms that are both household and broad credit booms, (ii) booms of one kind but not the other.
  - Episodes that are booms of both kinds are followed by substantially lower growth over the subsequent five years.
  - Note: In regressions without credit boom dummies, growth of household credit/GDP is statistically significantly related to subsequent growth, while growth of credit/GDP is not.
  - Two thirds of household credit booms are broad credit booms, and vice versa.
- HY share informative during household credit booms:
  - Table 13: baseline specification substituting household credit booms for broad credit booms shows HY share interaction coefficient negative and statistically significant out to two years, with magnitudes similar to baseline.
  - Suggests lending standards evolve consistently across asset classes; HY share moves in line with survey measures of bank lending standards (Section 3).
  - Addresses concern that rising HY share merely identifies household credit booms.
- House prices:
  - Baseline results robust to controlling for evolution of house prices.
  - Appendix Table A.7 includes five year growth in house price indices relative to GDP and an interaction of this growth with the credit boom dummy as controls.
  - Results similar to baseline (Table 6).
  - Data on house prices from the OECD and the Global Property Guide.

### 4.2.5    Further robustness exercises
- Six further robustness modifications where baseline results hold:
  1. Additional controls based on credit quantities:
     - Appendix Table A.8: Panel A adds level of credit/GDP and bond market proceeds/GDP as controls; Panel B adds the credit to GDP gap as a control.
  2. Shorter horizons:
     - Panels A and B of Appendix Table A.9 use horizons of three or four years (both credit boom dummy and average changes in HY share defined over these shorter horizons).
  3. Winsorization of dependent variable:
     - Panels A, B, and C of Appendix Table A.10 show results robust to winsorizing subsequent cumulative growth at the 5th and 95th percentile in the full sample, by country, or by year, respectively.
  4. Alternative HY share definitions:
     - HY share scaled by country mean before looking at five year changes; HY share defined by number of issues; excluding new issuers.
     - Results qualitatively similar but weaker magnitudes with alternative definitions.
     - Weaker results when HY share defined by number of issues suggests investor asset allocation information in baseline proceeds-based HY share is important and that increases do not primarily reflect lending to large numbers of new borrowers.
     - Appendix Tables A.11 and A.12 show these specifications.
  5. Post-1995 sample:
     - Results similar for the post-1995 sample (Appendix Table A.13).
     - Figure 4: constructed US HY share (adjusted for coverage) is similar to Greenwood & Hanson (2013) US HY share after 1995.
  6. Real GDP per capita:
     - Results similar when using subsequent growth in real GDP per capita rather than real GDP, confirming effects concern economic rather than demographic component of growth (Appendix Table A.14).

### 4.3    Information in prices
- Credit spreads in primary bond markets:
  - Spreads defined as difference between the 90th and 10th percentile yields within a country year, divided by the country level mean of this difference.
  - Use change in this spread from yeart−5 tot, including both level and interaction with credit booms.
  - Table 14: credit booms accompanied by rising spreads are followed by lower growth over subsequent years, with statistical significance out to three years.
  - Some specifications suggest spiking credit spreads after a credit boom might act as a trigger for worse growth in subsequent years.
- Advances vs emerging markets:
  - Appendix Table A.15: results based on credit spreads stronger in advanced economies (Panel A) with larger magnitudes and stronger significance at longer horizons; interaction coefficients not significant for emerging markets (Panel B).
  - Panel B suggests rising credit spreads are unconditionally followed by lower growth in emerging markets.
- HY share and spreads together:
  - Appendix Table A.16: regressions including evolution of both HY share and credit spreads show both are related to subsequent output growth in a horse race.
  - Coefficients similar to regressions including only one variable at a time, though results somewhat weaker for spreads.
  - Suggests both indicators might be useful in going beyond quantities for contemporaneous evaluation of credit booms.

### Conclusion (from section 5)
- Primary debt capital markets provide a measure of lending standards: the high-yield (HY) share of bond issuance constructed from bond-level issuance data across countries.
- Key empirical findings:
  - Lending standards measured by the HY share are procyclical and move in line with survey measures of bank lending standards.
  - Credit booms with a rising HY share are followed by lower output growth over the subsequent three to four years.
  - Outward shifts in credit supply are likely to play a role: for credit booms with deteriorating lending standards, credit spreads do not seem to rise substantially.
- Policy implication:
  - Policy makers should pay particular attention to credit booms accompanied by deteriorating lending standards in bond markets.
- Interpretation:
  - Procyclicality of lending standards suggests lenders expect booms to continue, consistent with behavioral narratives of credit build ups and crises (Minsky 1977, Minsky 1986, Kindleberger 1978).
  - Dynamics of bad credit booms are difficult to explain solely by views of credit-driven crises based only on financial frictions.

*Source: wp1823 - 4.2    Additional results and robustness*

### References

### wp1823 - References

### Major thematic areas covered by the references
- Credit cycles, leverage, and financial instability
  - Fostel, A. & Geanakoplos, J. (2008), ‘Leverage cycles and the anxious economy’, The American Economic Review98(4), 1211–1244.
  - Fostel, A. & Geanakoplos, J. (2014), ‘Endogenous collateral constraints and the leverage cycle’, Annu. Rev. Econ.6(1), 771–799.
  - Kiyotaki, N. & Moore, J. (1997), ‘Credit cycles’, Journal of Political Economy105(2), 211–248.
  - Minsky, H. (1977), ‘The financial instability hypothesis: an interpretation of keynes and an alternative to standard theory’, Challenge20(1), 20–27.
  - Minsky, H. (1986), Stabilizing an Unstable Economy, Yale University Press.
  - Bordo, M. D. & Haubrich, J. G. (2010), ‘Credit crises, money and contractions: An historical view’, Journal of Monetary Economics57(1), 1–18.
  - Kindleberger, C. P. (1978), Manias, Panics, and Crashes: A History of Financial Crises, Basic Books.
  - Reinhart, C. M. & Rogoff, K. S. (2009), This time is different: Eight centuries of financial folly, Princeton University Press.

- Mortgage markets, loan originations, and defaults
  - Adelino, M., Schoar, A. & Severino, F. (2015), Loan originations and defaults in the mortgage crisis: Further evidence, Technical report, National Bureau of Economic Research.
  - Adelino, M., Schoar, A. & Severino, F. (2016), ‘Loan originations and defaults in the mortgage crisis: The role of the middle class’, The Review of Financial Studies29(7), 1635–1670.
  - Foote, C. L., Loewenstein, L. & Willen, P. S. (2016), Cross-sectional patterns of mortgage debt during the housing boom: Evidence and implications, Technical report, National Bureau of Economic Research.
  - Mian, A. & Sufi, A. (2009), ‘The consequences of mortgage credit expansion: Evidence from the us mortgage default crisis’, The Quarterly Journal of Economics124(4), 1449–1496.
  - Mian, A. & Sufi, A. (2017), ‘Fraudulent income overstatement on mortgage applications during the credit expansion of 2002 to 2005’, The Review of Financial Studies30(6), 1832–1864.
  - FCIC (2011), The financial crisis inquiry report: Final report of the national commission on the causes of the financial and economic crisis in the united states, Technical report, The Financial Crisis Inquiry Comission.

- Banking, lending standards, and macrofinancial stability
  - Bassett, W. F., Chosak, M. B., Driscoll, J. C. & Zakrajˇsek, E. (2014), ‘Changes in bank lending standards and the macroeconomy’, Journal of Monetary Economics62, 23–40.
  - Dell’Ariccia, G., Igan, D., Laeven, L. & Tong, H. (2012), Policies for macrofinancial stability: Dealing with credit booms and busts, Staff Discussion Note 12/06, International Monetary Fund.
  - Dell’Ariccia, G., Igan, D. & Laeven, L. U. (2012), ‘Credit booms and lending standards: Evidence from the subprime mortgage market’, Journal of Money, Credit and Banking44(2-3), 367–384.
  - Stein, J. C. (2012), ‘Monetary policy as financial stability regulation’, The Quarterly Journal of Economics127(1), 57–95.
  - Bernanke, B. & Gertler, M. (1989), ‘Agency costs, net worth, and business fluctuations’, The American Economic Review79(1), 14–31.

- Asset prices, credit spreads, and macroeconomic linkages
  - Gilchrist, S. & Zakrajˇsek, E. (2012), ‘Credit spreads and business cycle fluctuations’, The American Economic Review102(4), 1692–1720.
  - Greenwood, R. & Hanson, S. G. (2013), ‘Issuer quality and corporate bond returns’, The Review of Financial Studies26(6), 1483–1525.
  - Stock, J. H. & Watson, M. W. (2003), ‘Forecasting output and inflation: The role of asset prices’, Journal of Economic Literature41(3), 788–829.

- Empirical methodologies and historical macrofinancial analysis
  - Jordà, Ò. (2005), ‘Estimation and inference of impulse responses by local projections’, The American Economic Review95(1), 161–182.
  - Jordà, Ò., Schularick, M. & Taylor, A. M. (2011), ‘Financial crises, credit booms, and external imbalances: 140 years of lessons’, IMF Economic Review59(2), 340–378.
  - Jordà, Ò., Schularick, M. & Taylor, A. M. (2013), ‘When credit bites back’, Journal of Money, Credit and Banking45(s2), 3–28.
  - Jordà, Ò., Schularick, M. & Taylor, A. M. (2017), ‘Macrofinancial history and the new business cycle facts’, NBER Macroeconomics Annual31(1), 213–263.
  - Schularick, M. & Taylor, A. M. (2012), ‘Credit booms gone bust: monetary policy, leverage cycles, and financial crises, 1870–2008’, The American Economic Review102(2), 1029–1061.
  - Bordalo, P., Gennaioli, N. & Shleifer, A. (2016), Diagnostic expectations and credit cycles, Technical report, National Bureau of Economic Research.

- Financial conditions, sentiment, and vulnerable growth
  - Hatzius, J., Hooper, P., Mishkin, F. S., Schoenholtz, K. L. & Watson, M. W. (2010), Financial conditions indexes: A fresh look after the financial crisis, Technical report, National Bureau of Economic Research.
  - Adrian, T., Boyarchenko, N. & Giannone, D. (2016), Vulnerable growth, Staff Reports 794, Federal Reserve Bank of New York.
  - López-Salido, D., Stein, J. C. & Zakrajˇsek, E. (2017), ‘Credit-market sentiment and the business cycle’, The Quarterly Journal of Economics132(3), 1373–1426.
  - Krishnamurthy, A. & Muir, T. (2016), ‘How credit cycles across a financial crisis’.

- IMF work and global financial stability
  - IMF (2017), Household debt and financial stability, Global Financial Stability Review Chapter 2, October 2017, International Monetary Fund.

### Notable dataset and figure references included in the source
- Figures referenced in the source:
  - Figure 1: Summary of sample coverage — summarizes sample coverage showing the 38 countries covered by the decade in which coverage starts.
  - Figure 2: Real GDP growth for selected countries — shows GDP growth, in percentage points, from 1980, or when the country enters the sample.

*References list as provided in wp1823 - References.*

### 2016.  Panel A shows a selection of advanced economies, while Panel B shows a selection of emerging markets.  Refer

### wp1823 - 2016

### High-yield (HY) share of bond issuance: patterns and comparisons
- Figure 3: HY share of bond issuance, in percentage points, from 1980 (or country entry) to 2016, presented separately for selected advanced economies (Panel A) and selected emerging markets (Panel B).
- Figure 4: Three versions of the HY share are shown: the corporate-only US HY share, the global HY share, and Greenwood and Hanson’s US HY share (exponentiated).
- Figure A.2: Country-level average HY share (non-financial corporate issuance only) plotted against the standard deviation of five-year growth in credit/GDP; separate panels for advanced economies and emerging markets.

### Credit conditions, lending standards, and HY share
- Figure 5: US corporate investment-grade share is plotted alongside US SLOOS (Senior Loan Officer Opinion Survey on Bank Lending Practices); SLOOS rises when lending standards tighten. The investment grade share is plotted so both measures increase when lending standards tighten.
- Table 3: Correlations between bank loan officer surveys and HY share (annual level).
  - Panel A (measures increase when standards tighten): example correlations include Austria -0.18, Canada -0.41, France -0.34, Germany -0.40, Netherlands -0.35, Portugal -0.56, Spain -0.28, United States -0.28.
  - Panel B (measures increase when standards ease): Japan 0.83, Poland 0.63, United Kingdom 0.25.

### HY share dynamics and GDP growth: average effects
- Table 4: Average change in HY share by whether growth is rising (dummy 1_{G_{t,t−k}>0}), regressions over k∈[1,5].
  - Panel A (country fixed effects only): coefficients on 1_{G_{t,t−k}>0}:
    - k=1: 3.50 (t-stat 4.78)
    - k=2: 1.37 (t-stat 2.98)
    - k=3: 1.60 (t-stat 3.05)
    - k=4: 1.60 (t-stat 3.06)
    - k=5: 1.22 (t-stat 2.84)
    - Mean of depvar across horizons: -0.15, -0.06, -0.01, -0.04, -0.06
  - Panel B (country and year fixed effects): coefficients on 1_{G_{t,t−k}>0}:
    - k=1: 1.40 (t-stat 1.76)
    - k=2: 0.10 (t-stat 0.13)
    - k=3: 0.56 (t-stat 0.80)
    - k=4: 0.82 (t-stat 1.25)
    - k=5: 1.07 (t-stat 2.12)

### HY share, credit booms, and subsequent GDP growth (interaction evidence)
- Table 6: Regressions of subsequent cumulative real GDP growth (horizons 1 to 5 years) on ∆HY_{t,t−5} (scaled to unit variance), Credit boom dummy, and Credit boom × ∆HY_{t,t−5}. Controls: 5-year credit/GDP growth and two lags of real GDP growth.
  - Coefficients on ∆HY_{t,t−5}:
    - 1yr: 0.14 (1.32)
    - 2yr: 0.33 (1.48)
    - 3yr: 0.37 (1.42)
    - 4yr: 0.21 (0.89)
    - 5yr: 0.25 (0.90)
  - Coefficients on Credit boom:
    - 1yr: -0.35 (-0.95)
    - 2yr: -1.00 (-1.35)
    - 3yr: -1.47 (-1.55)
    - 4yr: -1.54 (-1.66)
    - 5yr: -1.67 (-1.60)
  - Coefficients on Credit boom × ∆HY_{t,t−5}:
    - 1yr: -0.45 (-2.56)
    - 2yr: -1.00 (-3.21)
    - 3yr: -1.45 (-2.36)
    - 4yr: -1.34 (-2.39)
    - 5yr: -0.93 (-1.09)
  - Mean of dependent variable (subsequent cumulative real GDP growth): 1yr 2.40, 2yr 4.81, 3yr 7.16, 4yr 9.50, 5yr 11.76.

- Table 10: Using a positive credit/GDP gap dummy instead of boom:
  - Coefficients on ∆HY_{t,t−5}:
    - 1yr: 0.27 (2.50)
    - 2yr: 0.56 (1.78)
    - 3yr: 0.62 (1.65)
    - 4yr: 0.16 (0.46)
    - 5yr: 0.13 (0.33)
  - Credit/GDP gap>0:
    - 1yr: -0.66 (-2.28)
    - 2yr: -1.13 (-2.27)
    - 3yr: -1.09 (-1.53)
    - 4yr: -0.82 (-0.87)
    - 5yr: -1.15 (-1.18)
  - Interaction (Credit/GDP gap>0 × ∆HY_{t,t−5}):
    - 1yr: -0.54 (-3.74)
    - 2yr: -1.01 (-2.59)
    - 3yr: -1.28 (-2.06)
    - 4yr: -0.78 (-1.43)
    - 5yr: -0.38 (-0.60)

- Table 9: Results using full-sample definition of credit boom:
  - ∆HY_{t,t−5} coefficients:
    - 1yr: 0.16 (1.61)
    - 2yr: 0.39 (1.70)
    - 3yr: 0.43 (1.64)
    - 4yr: 0.18 (0.72)
    - 5yr: 0.17 (0.59)
  - Credit boom:
    - 1yr: -0.32 (-0.90)
    - 2yr: -0.90 (-1.26)
    - 3yr: -1.36 (-1.46)
    - 4yr: -1.29 (-1.29)
    - 5yr: -1.56 (-1.50)
  - Credit boom × ∆HY_{t,t−5}:
    - 1yr: -0.51 (-2.86)
    - 2yr: -1.11 (-3.41)
    - 3yr: -1.53 (-2.51)
    - 4yr: -1.21 (-2.10)
    - 5yr: -0.61 (-0.72)

### Global HY share and cross-country heterogeneity
- Table 8: Regressions using the average five-year change in the global HY share (∆HY^G_{t,t−5}, scaled to unit variance), interactions with Credit boom and Emerging Markets (EM) dummy, and triple interaction.
  - Credit boom × ∆HY^G_{t,t−5}:
    - 1yr: -0.61 (-2.32)
    - 2yr: -1.15 (-2.04)
    - 3yr: -2.27 (-2.21)
    - 4yr: -2.65 (-1.89)
    - 5yr: -2.63 (-1.54)
  - EM × ∆HY^G_{t,t−5}:
    - 1yr: -0.51 (-1.34)
    - 2yr: -0.78 (-1.39)
    - 3yr: -0.15 (-0.17)
    - 4yr: 0.24 (0.21)
    - 5yr: 0.40 (0.30)
  - EM × Credit boom × ∆HY^G_{t,t−5}:
    - 1yr: -0.53 (-0.73)
    - 2yr: -1.38 (-1.01)
    - 3yr: -2.66 (-1.68)
    - 4yr: -1.90 (-1.11)
    - 5yr: -2.41 (-1.17)

### Heterogeneity: advanced economies vs emerging markets
- Table 11: Separate estimates for advanced economies (Panel A) and emerging markets (Panel B); independent variables include ∆HY_{t,t−5}, Credit boom, and Credit boom × ∆HY_{t,t−5}.
  - Panel A (Advanced economies), ∆HY_{t,t−5}:
    - 1yr: 0.23 (2.32)
    - 2yr: 0.48 (1.71)
    - 3yr: 0.52 (1.72)
    - 4yr: 0.31 (1.50)
    - 5yr: 0.37 (1.49)
    - Credit boom × ∆HY_{t,t−5}:
      - 1yr: -0.49 (-2.80)
      - 2yr: -1.16 (-3.93)
      - 3yr: -1.92 (-2.77)
      - 4yr: -1.87 (-3.32)
      - 5yr: -1.44 (-1.44)
  - Panel B (Emerging markets), ∆HY_{t,t−5}:
    - 1yr: 0.19 (0.52)
    - 2yr: 0.56 (1.05)
    - 3yr: 0.96 (1.36)
    - 4yr: 0.87 (1.11)
    - 5yr: 1.09 (1.16)
    - Credit boom × ∆HY_{t,t−5}:
      - 1yr: -0.41 (-0.77)
      - 2yr: -0.98 (-1.23)
      - 3yr: -0.77 (-0.77)
      - 4yr: 0.27 (0.25)
      - 5yr: 1.44 (1.20)

### HY share changes and future growth quantiles
- Table 5: Summary statistics by quintiles of future cumulative 3-year real GDP growth.
  - Panel A (Unconditional):
    - Real GDP growth t,t+3 by quintile Q1..Q5: -1.0, 4.8, 7.8, 10.5, 17.4
    - ∆Credit/GDP t,t−5 by quintile Q1..Q5: 17.6, 12.4, 9.7, 9.3, 5.6
    - ∆HY t,t−5 by quintile Q1..Q5: -0.2, -0.6, -0.7, -0.0, -1.1
  - Panel B (Conditional on credit boom):
    - Real GDP growth t,t+3 by quintile Q1..Q5: -6.1, 2.5, 6.2, 9.5, 17.3
    - ∆Credit/GDP t,t−5 by quintile Q1..Q5: 38.2, 33.2, 31.4, 33.3, 39.5
    - ∆HY t,t−5 by quintile Q1..Q5: 1.4, -0.8, -0.9, -1.2, -2.4

### Spreads, HY share, and macro outcomes
- Table 7: Five-year change in spreads (S_t − S_{t−5}) regressions.
  - Credit boom coefficient:
    - Specification 1: -0.25 (-2.34)
    - Specification 2: -0.24 (-2.28)
  - ∆HY_{t,t−5}: 0.14 (2.91)
  - Credit boom × ∆HY_{t,t−5}: 0.26 (2.78)
  - No Credit boom × ∆HY_{t,t−5}: 0.09 (1.65)
  - Mean of depvar: -0.06

- Table 14: Subsequent cumulative real GDP growth and 5-year change in spread during credit boom.
  - S_t − S_{t−5} coefficients on subsequent growth:
    - 1yr: 0.13 (1.62)
    - 2yr: 0.19 (1.46)
    - 3yr: 0.09 (0.50)
    - 4yr: 0.03 (0.12)
    - 5yr: 0.15 (0.58)
  - Credit boom × (S_t − S_{t−5}):
    - 1yr: -0.51 (-2.27)
    - 2yr: -0.78 (-2.46)
    - 3yr: -0.80 (-2.11)
    - 4yr: -0.59 (-1.46)
    - 5yr: -0.67 (-1.21)

### Credit boom types and household credit
- Table 12: Subsequent cumulative real GDP growth by type of credit boom (household vs baseline).
  - Credit boom × 1_{HH Credit boom}:
    - 1yr: -0.89 (-1.91)
    - 2yr: -2.26 (-2.44)
    - 3yr: -3.33 (-2.65)
    - 4yr: -3.72 (-2.65)
    - 5yr: -3.68 (-2.30)
- Table 13: During household credit booms, interaction of HH Credit boom × ∆HY_{t,t−5}:
  - ∆HY_{t,t−5}: coefficients 1yr..5yr = 0.15 (1.53), 0.38 (1.65), 0.21 (0.93), 0.06 (0.19), 0.42 (1.34)
  - HH Credit boom: -0.81 (-2.09), -1.96 (-2.69), -2.51 (-2.56), -2.90 (-2.52), -2.92 (-2.18)
  - HH Credit boom × ∆HY_{t,t−5}: -0.82 (-2.52), -1.38 (-2.13), -1.19 (-1.62), -1.10 (-1.15), -1.38 (-1.08)

### Dynamics and mean-reversion of HY share
- Table A.2: Reversal of HY share regressions (dependent variable: subsequent average change in HY share over horizons 1–5 years).
  - Coefficients on ∆HY_{t,t−5} (scaled, both sides):
    - 1yr: -0.30 (-6.09)
    - 2yr: -0.33 (-4.25)
    - 3yr: -0.35 (-3.99)
    - 4yr: -0.30 (-3.52)
    - 5yr: -0.37 (-3.74)
  - Credit boom × ∆HY_{t,t−5} coefficients small and generally statistically weak across horizons.

### Sample and data coverage
- Table 1: Sample consists of 110k bond issues from Dealogic, associated with issuers' countries of operation; excludes bonds with maturity <1 year, financial issuers, and countries with no bond issuance for 5 years. For each country the sample starts with the first year with at least 10 bonds and at least one non-financial corporate bond issued; final restriction: countries must have at least 10 years with at least 10 bonds and at least one non-financial corporate bond issued.
- Representative country-level sample statistics (as reported in Table 1 excerpt):
  - Argentina: First year 1993, N 23, Missing 11, Issuers 98, Bonds in sample 184, Mean Share/GDP 132.45, Boom years 0, Advanced 0
  - United States: First year 1980, N 37, Missing 1, Issuers 2,941, Bonds in sample 32,971, Mean Share/GDP 6.10, Boom years 1, Advanced 1
  - China: First year 2002, N 15, Missing 5, Issuers 225, Bonds in sample 12,715, Mean Share/GDP 5.06, Boom years 0, Advanced 0
  - United Kingdom: First year 1985, N 32, Missing 1, Issuers 598, Bonds in sample 3,994, Mean Share/GDP 5.61, Boom years 2, Advanced 1
  - Japan: First year 1984, N 33, Missing 1, Issuers 734, Bonds in sample 10,596, Mean Share/GDP 10.51, Boom years 10, Advanced 1

*Italic: Source: wp1823 - 2016 (figures and tables excerpt).*

### 1.5 times its standard deviation and the annual growth rate of the credit-to-GDP ratio exceeds 10 percent; or (ii) the

### wp1823 - 1.5 times its standard deviation and the annual growth rate of the credit-to-GDP ratio exceeds 10 percent; or (ii) the

### Credit-boom definitions and sample overlap
- Baseline SDN credit-boom definition (as used in Panel A figures/tables):
  - A boom occurs if either: (i) the deviation from trend is greater than 1.5 times its standard deviation and the annual growth rate of the credit-to-GDP ratio exceeds 10 percent; or (ii) the annual growth rate of the credit-to-GDP ratio exceeds 20 percent.
  - Trend estimated using a cubic regression over the past 10 years.
- Less stringent SDN definition (Panel B):
  - A boom occurs if either: (i) the deviation from trend is greater than 1.5 times its standard deviation and the annual growth rate of the credit-to-GDP ratio exceeds 5 percent; or (ii) the annual growth rate of the credit-to-GDP ratio exceeds 10 percent.
- Overlap counts shown in Panels A and B (preserve table cell values as presented):
  - Panel A counts: 0 5 6 6 8 5 7 4; 11 4 23 11 73; Total 70 83 97 47
  - Panel B counts: 0 5 2 4 5 0 5 7 4; 18 3 90 173; Total 60 71 40 74 7
  - (These counts are taken verbatim from the panels as displayed.)

### Main regression specification and interpretation
- Dependent variable: subsequent cumulative real GDP growth over horizons from 1 year to 5 years.
- Key independent variables:
  - ∆HY t,t−5: average change in high-yield (HY) share over the previous 5 years (scaled to have unit variance unless otherwise noted).
  - Credit boom dummy (variously defined: SDN, less stringent SDN, country-level, median-based).
  - Interaction term: Credit boom × ∆HY t,t−5.
- Common controls: 5-year credit/GDP growth and two lags of real GDP growth. Some specifications include additional controls (see robustness section).
- Fixed effects: Country FE = Y in all reported regressions; Year FE = Y in all reported regressions.
- Clustering: Standard errors are clustered (by country or by country-year variations noted in specific tables).

### Key empirical findings (selected coefficient magnitudes and statistical notes)
- Baseline (Panel A, SDN-defined boom):
  - ∆HY t,t−5 coefficients (1yr–5yr): 0.02, 0.08, -0.02, -0.18, 0.09 (t-stats: (0.16), (0.44), (-0.11), (-0.70), (0.30)).
  - Credit boom coefficients (1yr–5yr): 0.27, -0.20, -0.66, -1.37, -1.60 (t-stats: (0.43), (-0.17), (-0.39), (-0.78), (-0.89)).
  - Credit boom × ∆HY (1yr–5yr): -0.37, -0.82, -0.63, -0.85, -1.88 (t-stats: (-1.05), (-1.58), (-1.24), (-0.91), (-1.38)).
  - R2 (1yr–5yr): 0.56, 0.59, 0.63, 0.69, 0.74.
  - Country years (1yr–5yr): 747, 709, 671, 633, 595.
  - Countries: 38 across all horizons.
- Less stringent SDN (Panel B):
  - ∆HY t,t−5 (1yr–5yr): 0.07, 0.20, 0.09, -0.08, 0.25 (t-stats: (0.69), (1.16), (0.57), (-0.28), (0.95)).
  - Credit boom (1yr–5yr): -0.05, 0.08, 0.16, 0.20, 0.24 (t-stats: (-0.12), (0.12), (0.16), (0.18), (0.26)).
  - Credit boom × ∆HY (1yr–5yr): -0.40, -0.93, -0.82, -0.91, -1.77 (t-stats: (-1.92), (-2.14), (-1.25), (-0.91), (-1.49)).
  - R2 (1yr–5yr): 0.56, 0.60, 0.63, 0.69, 0.74.
  - Country years (1yr–5yr): 747, 709, 671, 633, 595; Countries: 38.
- Country-level boom definition (Table A.5):
  - ∆HY t,t−5 (1yr–5yr): 0.13, 0.29, 0.11, -0.23, 0.01 (t-stats: (1.07), (1.20), (0.58), (-0.79), (0.04)).
  - Credit boom (1yr–5yr): -0.14, -0.99, -1.51, -1.52, -1.45 (t-stats: (-0.51), (-1.73), (-1.88), (-1.89), (-1.81)).
  - Credit boom × ∆HY (1yr–5yr): -0.53, -1.01, -0.70, -0.00, -0.23 (t-stats: (-2.27), (-1.88), (-1.33), (-0.01), (-0.29)).
  - R2 (1yr–5yr): 0.56, 0.60, 0.63, 0.69, 0.74.
  - Country years (1yr–5yr): 747, 709, 671, 633, 595; Countries: 38.

### Robustness checks and additional specifications
- Boom defined at median (Table A.6):
  - Credit boom coefficients (1yr–5yr): -0.64, -1.11, -1.33, -1.85, -2.36 (t-stats: (-2.11), (-1.66), (-1.44), (-2.03), (-2.65)).
  - Credit boom × ∆HY (1yr–5yr): -0.35, -0.40, -0.49, -0.40, -0.00 (t-stats: (-2.24), (-1.48), (-1.37), (-0.77), (-0.00)).
- Controlling for house prices (Table A.7):
  - ∆HY t,t−5 (1yr–5yr): 0.16, 0.35, 0.40, 0.14, 0.08 (t-stats: (2.08), (1.74), (1.56), (0.67), (0.32)).
  - Credit boom × ∆HY (1yr–5yr): -0.53, -1.11, -1.63, -1.47, -0.77 (t-stats: (-3.06), (-3.38), (-2.41), (-3.28), (-1.33)).
  - R2 (1yr–5yr): 0.64, 0.68, 0.71, 0.74, 0.78.
  - Country years (1yr–5yr): 647, 611, 575, 539, 504.
- Additional credit quantity controls (Table A.8):
  - Panel A (controls include Credit/GDP level and Bond Proceeds/GDP):
    - ∆HY t,t−5 (1yr–5yr): 0.17, 0.40, 0.47, 0.31, 0.38 (t-stats: (1.56), (1.79), (1.76), (1.24), (1.24)).
    - Credit/GDP coefficient (1yr–5yr): -0.68, -1.47, -2.28, -2.75, -3.11 (t-stats: (-2.14), (-2.09), (-2.25), (-2.12), (-1.90)).
    - Bond Proceeds/GDP (1yr–5yr): 0.26, 0.49, 0.40, 0.04, 0.52 (t-stats: (1.76), (1.52), (1.17), (0.10), (1.09)).
    - Credit boom × ∆HY (1yr–5yr): -0.41, -0.92, -1.40, -1.35, -0.74 (t-stats: (-2.36), (-2.93), (-2.15), (-2.09), (-0.91)).
  - Panel B (controls include Credit/GDP gap):
    - Credit/GDP gap (1yr–5yr): -0.65, -1.35, -1.62, -1.60, -1.48 (t-stats: (-1.68), (-1.69), (-1.68), (-1.84), (-2.15)).
    - Credit boom × ∆HY (1yr–5yr): -0.34, -0.83, -1.41, -1.41, -1.03 (t-stats: (-1.58), (-2.74), (-2.59), (-2.60), (-1.19)).
- Shorter horizons (Table A.9):
  - Four-year horizon (∆HY t,t−4): coefficients (1yr–5yr): 0.06, 0.25, 0.38, 0.15, 0.22; Credit boom × ∆HY (1yr–5yr): -0.45, -0.96, -1.45, -1.78, -2.07 (t-stats: (-2.09), (-2.84), (-3.14), (-2.88), (-2.64)).
  - Three-year horizon (∆HY t,t−3): coefficients (1yr–5yr): 0.06, 0.23, 0.36, 0.25, 0.23; Credit boom × ∆HY (1yr–5yr): -0.45, -0.93, -1.19, -1.13, -1.92 (t-stats: (-1.99), (-2.63), (-2.87), (-2.09), (-2.88)).
- Winsorized dependent variable (Table A.10):
  - Panel A (winsorized over full sample): ∆HY t,t−5 (1yr–5yr): 0.12, 0.19, 0.25, 0.20, 0.28; Credit boom × ∆HY (1yr–5yr): -0.41, -0.81, -1.05, -1.13, -0.67 (t-stats: (-2.72), (-3.47), (-2.93), (-2.81), (-1.39)).
  - Panels B and C (winsorized by country and by year) show similar patterns with coefficient magnitudes and t-statistics reported in the tables.
- Alternative scalings and definitions of HY share:
  - Scaled by country mean (Table A.11): ∆HY t,t−5 (1yr–5yr): 0.12, 0.25, 0.26, 0.14, 0.19; Credit boom × ∆HY (1yr–5yr): -0.29, -0.73, -1.16, -1.15, -0.67 (t-stats: (-1.37), (-2.12), (-1.71), (-1.86), (-0.98)).
  - HY share defined by number of issuers (Table A.12 Panel A): ∆HY t,t−5 (1yr–5yr): 0.15, 0.19, 0.26, 0.25, 0.14; Credit boom × ∆HY (1yr–5yr): -0.16, -0.29, -0.42, -0.49, -0.52 (t-stats: (-3.40), (-4.06), (-2.35), (-2.14), (-2.22)).
  - HY share excluding new issuers (Table A.12 Panel B): ∆HY t,t−5 (1yr–5yr): 0.20, 0.49, 0.51, 0.31, 0.29; Credit boom × ∆HY (1yr–5yr): -0.11, -0.26, -0.34, -0.23, -0.12 (t-stats: (-2.16), (-4.07), (-3.33), (-1.25), (-0.39)).
- Post-1995 sample (Table A.13):
  - ∆HY t,t−5 (1yr–5yr): 0.12, 0.33, 0.34, 0.15, 0.20; Credit boom × ∆HY (1yr–5yr): -0.48, -1.06, -1.52, -1.38, -0.99 (t-stats: (-2.52), (-3.30), (-2.37), (-2.29), (-1.00)).
- Per capita growth dependent variable (Table A.14):
  - ∆HY t,t−5 (1yr–5yr): 0.12, 0.31, 0.32, 0.04, -0.01 (t-stats: (1.17), (1.38), (1.27), (0.20), (-0.03)).
  - Credit boom × ∆HY (1yr–5yr): -0.43, -0.96, -1.37, -1.14, -0.54 (t-stats: (-2.49), (-2.83), (-2.19), (-2.16), (-0.60)).
- Advanced vs emerging markets: spreads interaction (Table A.15)
  - Advanced economies (Panel A): S t − S t−5 (1yr–5yr): 0.21, 0.29, 0.19, 0.16, 0.23 (t-stats: (2.22), (1.90), (1.04), (0.68), (0.86)); Credit boom × S t − S t−5 (1yr–5yr): -0.55, -0.98, -1.15, -1.01, -1.36 (t-stats: (-2.30), (-2.42), (-2.66), (-2.26), (-2.35)).
  - Emerging markets (Panel B): S t − S t−5 (1yr–5yr): -0.38, -0.65, -0.96, -0.96, -0.53 (t-stats: (-1.03), (-1.33), (-2.00), (-1.74), (-0.81)); Credit boom × S t − S t−5 (1yr–5yr): 0.52, 1.00, 1.32, 1.46, 1.50 (t-stats: (0.85), (1.26), (1.51), (1.48), (1.56)).
- HY share and spreads together (Table A.16):
  - ∆HY t,t−5 (1yr–5yr): 0.13, 0.33, 0.38, 0.22, 0.25 (t-stats: (1.27), (1.45), (1.43), (0.91), (0.86)).
  - S t − S t−5 (1yr–5yr): 0.13, 0.19, 0.09, 0.03, 0.14 (t-stats: (1.65), (1.45), (0.53), (0.13), (0.53)).
  - Credit boom × ∆HY (1yr–5yr): -0.40, -0.91, -1.36, -1.29, -0.88 (t-stats: (-2.29), (-3.15), (-2.27), (-2.33), (-1.07)).
  - Credit boom × S t − S t−5 (1yr–5yr): -0.46, -0.68, -0.67, -0.48, -0.62 (t-stats: (-1.99), (-2.11), (-1.70), (-1.09), (-1.13)).
  - R2 (1yr–5yr): 0.57, 0.61, 0.64, 0.69, 0.74.

### Consistent patterns across specifications
- Positive baseline association between increases in HY share (∆HY) and subsequent cumulative real GDP growth in several specifications (positive ∆HY coefficients).
- Negative and often statistically significant interaction terms (Credit boom × ∆HY) across many specifications, indicating that the growth-enhancing effect of an increase in the HY share is weaker or reversed during credit booms.
- Credit boom dummy alone often has negative coefficients at longer horizons (e.g., -1.37 to -2.36 across various tables and definitions), indicating booms are associated with lower subsequent cumulative growth over multiple horizons in several specifications.
- Robustness to alternative definitions (median-based, country-level, post-1995), alternative HY constructions (number of issuers, excluding new issuers), inclusion of house price controls, credit quantity controls, shorter horizon specifications, winsorization, and joint inclusion of HY share and spreads.

*Source: wp1823 (text and tables as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1823.pdf_