## wpiea2021103-print-pdf - Section 6 presents estimates of the economic costs of reversals and crises in terms of output

## Source details

**Canonical URL:** [wpiea2021103-print-pdf - Section 6 presents estimates of the economic costs of reversals and crises in terms of output](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021103-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021103-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021103-print-pdf.pdf.json)

---

### Definitions and data
- Credit reversal: binary indicator for episodes when credit contracts alongside expanding economic activity (intended to isolate supply-driven contractions in credit supply).
- Banking crisis (baseline): Laeven and Valencia (2017) definition requiring (i) significant signs of distress in the banking system (bank runs, losses, and/or bank liquidations), and (ii) significant banking policy intervention measures in response to large losses.
- Robustness crisis database: Reinhart-Rogoff (2009) definition (bank runs leading to closures/mergers/takeovers or large-scale government assistance of important financial institutions); Laeven and Valencia (2017) spans 160 countries, 1970‒2017; Reinhart-Rogoff (2009) covers 70 countries, 1800‒2010, extended to 2017 to match macro dataset.
- Data: yearly data for 179 countries, 1960‒2017, from IFS. Credit measured as domestic bank credit to the private sector (IFS line 22d). Credit converted to real terms using the GDP deflator.

### Methodology
- Credit cycles defined by CYGAP (credit-to-GDP gap), computed by detrending the ratio of credit-to-GDP with a two-sided HP filter with smoothing parameter λ=100; robustness checks use λ up to 1,600.
- Event-study approach drawing on Gourinchas et al. (2001); Mendoza and Terrones (2008, 2012).
- Three exercises:
  - Relate evolution of credit and GDP during positive CYGAP phases to subsequent cyclical dynamics (duration, pace, implied credit-to-GDP changes).
  - Study timing and likelihood of credit reversals and banking crises conditional on each other.
  - Estimate effects of credit reversals and banking crises on credit cycles and economic activity; compute output-forgone estimates.
- Event-study accumulation: accumulate GDP growth for up to six years after CYGAP turns negative and compare cycles with and without crises/reversals; starting point chosen where output gap differential is close to zero. Method assumes exogeneity of crises/reversals (if endogenous and inversely related to economic activity, estimates are upward biased).

### Key features of credit-to-GDP cycle dynamics
- Sample: 371 entire cycles (plus incomplete episodes); average entire cycle lasts about ten years.
- Average changes in credit-to-GDP:
  - Increase by about nine percentage points during positive gap episodes and drop by about two percentage points during negative gaps.
  - Extreme gaps average 15 percent (positive) and ‒18 percent (negative), wider in developing countries.
- Average credit growth remains positive in both positive and negative CYGAP phases; negative gaps reflect periods where GDP growth is relatively more dynamic than credit growth.
- Likelihood of observing a credit reversal: about 22 percent during the negative phase, almost two times higher than during positive gaps.

### Co-movement of credit growth and GDP growth
- Simple correlation between credit growth and GDP growth: 0.32 overall; 0.09 during episodes of negative credit growth.
- Directional co-movement:
  - Negative credit growth occurs in 23.4 percent of the years when economic activity expands.
  - Positive credit growth occurs in 47.9 percent of the years with contracting activity.
- Concordance index (Harding and Pagan, 2002):
  - Average concordance: 0.68 for developing countries and 0.73 for industrial countries.
  - Concordance driven almost exclusively by periods of positive co-movement.
  - Slight downward trend in concordance in industrial countries and a sharp drop after the 2008‒09 crisis.
- Conditional volatility:
  - Average credit growth is considerably more volatile than GDP growth, particularly in developing countries.
  - During positive CYGAP phases, average pace of credit growth is three to five times faster than GDP growth; during negative CYGAP episodes, single-year contractions of credit can be dramatic.

### Relationship between credit reversals, banking crises, and CYGAP cycles
- Timing and conditional probabilities:
  - Credit reversals are more likely to occur after banking crises.
    - Industrial countries: likelihood of a credit reversal conditional on a previous banking crisis rises from about 8 percent to 35 percent.
    - Developing countries: corresponding probabilities increase from 15 to 25 percent.
  - Banking crises tend to precede credit reversals; likelihood of banking crises increases in years leading to a credit reversal.
- Incidence within CYGAP phases:
  - Banking crises occur roughly once every eight years across country groups.
  - Credit reversals are more frequent, particularly in developing countries (about once every five years).
  - Banking crises are more likely when CYGAP is positive (driven by industrial countries); credit reversals are about two times more likely during negative CYGAPs.
- Evolution within cycles (ten-year window centered on CYGAP shift from positive to negative):
  - Developing countries: yearly probabilities of banking crises are rather flat throughout the window.
  - Industrial countries: likelihood of crises peaks early in the CYGAP cycle, four years before CYGAP turns negative.
  - Probability of credit reversals bottoms early in the positive phase and increases monotonically, peaking when CYGAP becomes negative.
- Cycle characteristics conditional on crises/reversals:
  - Cycles with banking crises: negative phase is on average 1.7 years longer than control group and features a differential drop in credit-to-GDP ratios of 5 percentage points.
  - Cycles with credit reversals: contraction in credit growth almost two-times higher than control; change in credit-to-GDP ratios associated with reversals is comparable in magnitude to crises.
- Magnitude interpretation (one-standard deviation frequency shock):
  - A 16-percentage point increase in frequency of banking crises (one standard deviation) leads to:
    - ‒0.5 percentage point change in credit-to-GDP ratios during negative CYGAP in developing countries and ‒0.9 in industrial countries (vis-à-vis control groups).
    - Increase in duration of negative phase by 0.328 years in developing countries and by 0.169 years in industrial countries.
  - Effects of credit reversals surpass those of banking crises across most variables in these comparisons.

### Economic costs of credit reversals and banking crises (Section 6 findings)
- Literature context: prior estimates of banking crisis costs typically range between 4‒15 percent of GDP.
- Event-study descriptive patterns:
  - Cycles with banking crises show wider output gap swings; developing countries close gap differentials more quickly.
  - For GDP growth, cycles with banking crises overperform control early but underperform substantially afterwards; differences remain protracted, especially in industrial countries.
  - GDP growth differentials for credit reversals are comparable in magnitude to those for banking crises and appear more persistent in developing countries.
- Cumulative GDP growth differentials (Table 9; Year 0 is the year when the credit-to-GDP gap turns negative):
  - Developing Countries (Banking Crises / Credit Reversals — Mean / Median by Year):
    - Year 0: 0.0000 / 0.0000 / 0.0000 / 0.0000
    - Year 1: -0.0189 / -0.0085 / -0.0018 / -0.0094
    - Year 2: -0.0224 / -0.0069 / -0.0143 / -0.0221
    - Year 3: -0.0194 / -0.0001 / -0.0241 / -0.0443
    - Year 4: -0.0018 / 0.0173 / -0.0332 / -0.0612
    - Year 5: -0.0176 / 0.0097 / -0.0590 / -0.0982
    - Year 6: -0.0503 / 0.0023 / -0.0547 / -0.1072
  - Industrial Countries (Banking Crises / Credit Reversals — Mean / Median by Year):
    - Year 0: 0.0000 / 0.0000 / 0.0000 / 0.0000
    - Year 1: -0.0167 / -0.0132 / -0.0181 / -0.0113
    - Year 2: -0.0395 / -0.0343 / -0.0293 / -0.0262
    - Year 3: -0.0650 / -0.0553 / -0.0392 / -0.0417
    - Year 4: -0.0776 / -0.0725 / -0.0639 / -0.0615
    - Year 5: -0.0897 / -0.0869 / -0.0876 / -0.0816
    - Year 6: -0.1010 / -0.1057 / -0.0984 / -0.0953
- Fixed-effect regressions of yearly GDP growth on event dummies (Table 10):
  - Dummy for Cycles with Banking Crises (DBC):
    - All Sample [1]: -0.013** [0.005]
    - All Sample [2]: -0.013* [0.007]
    - Developing [3]: -0.012 [0.010]
    - Industrial [4]: -0.013*** [0.004]
  - Dummy for Cycles with Credit Reversals (DCC):
    - All Sample [1]: -0.006*** [0.002]
    - All Sample [2]: -0.006** [0.002]
    - Developing [3]: -0.004 [0.003]
    - Industrial [4]: -0.011** [0.004]
  - Interaction DBC x DCC: estimates ~0.004–0.005 across columns.
  - Constant: 0.045*** / 0.045*** / 0.045*** / 0.044***.
  - Sample: Observations 7205 / 7205 / 5465 / 1740; Number of countries 177 / 177 / 139 / 38.
- Interpretation of regression results:
  - Banking crises associated with an estimated drop in GDP growth of 1.3 percent per year during entire credit-to-GDP episodes (ten-year cycles) → cumulative GDP differential about 14 percentage points (=1.013^10‒1). Using five-year average negative phase, implied cumulative drop around 7 percent (=1.013^10‒1).
  - Credit reversals have substantial economic costs, about one-half of those associated with banking crises per episode.
  - Relative frequencies: reversals occur about once every 7.6 years; crises about once every 5.1 years. Using relative frequencies, overall cost of reversals over time ≈ two-thirds of those associated with banking crises (0.69 = 0.006 / 0.013 × 7.6 / 5.1 as presented in text).
- Fixed-effect regressions of GDP growth on frequencies (Table 11):
  - Frequency of Banking Crisis in Cycle (PBC):
    - All Sample [1]: -0.021*** [0.008]
    - All Sample [2]: -0.021* [0.011]
    - Developing [3]: -0.021 [0.015]
    - Industrial [4]: -0.028*** [0.009]
  - Frequency of Credit Reversal in Cycle (PCC):
    - All Sample [1]: -0.019*** [0.005]
    - All Sample [2]: -0.019*** [0.005]
    - Developing [3]: -0.014** [0.006]
    - Industrial [4]: -0.033** [0.014]
  - Interaction PBC x PCC: estimates ~0.025–0.038 across columns.
  - Note: frequency is the proportion of years with banking crises or credit reversals within cycles.
  - Panel regression [2] result highlighted: a one percentage point increase in frequency of credit reversals leads to a drop of GDP growth of about 1.9 percentage point per year; estimated losses are larger for industrial countries.

### Policy implications and conclusions
- Credit reversals, defined stringently to capture supply-driven credit contractions during expanding activity, are frequent and economically costly.
- Relative costs:
  - Per episode, credit reversals impose about one-half of banking crisis costs.
  - Accounting for relative frequencies over time, reversals impose up to two-thirds of the overall cost of banking crises.
- Policy relevance:
  - Findings support macroprudential policies aimed at preventing supply-driven contractions in bank credit.
  - Evidence that banking crises peak early in positive CYGAP phases (especially in industrial countries) raises questions on CYGAP adequacy as the sole trigger for countercyclical capital buffers.
- Cross-country differences:
  - Impacts of reversals and crises on economic activity seem larger in industrial than in developing countries; developing countries often rebound faster after banking crises.
  - Possible channels include deeper financial markets in industrial countries, role of FX depreciation, and external demand in developing countries — left for future research.
- Caution: results are not causal due to possible endogeneity and omitted variables; if reversals/crises are more likely under sluggish growth, estimated costs are likely upward biased (conservative).

*Source: wpiea2021103-print-pdf - Section 6 presents estimates of the economic costs of reversals and crises in terms of output*

### Section 6 presents estimates of the economic costs of reversals and crises in terms of output

### wpiea2021103-print-pdf - Section 6 presents estimates of the economic costs of reversals and crises in terms of output

### Definitions of credit reversals and banking crises
- Credit reversal: an episode when credit contracts alongside expanding economic activity; intended to isolate supply-driven contractions in credit supply. Identified with a binary variable.
- Banking crisis (baseline): operational definition and dating of Laeven and Valencia (2017): (i) significant signs of distress in the banking system (bank runs, losses, and/or bank liquidations), and (ii) significant banking policy intervention measures in response to large losses.
- Robustness crisis database: Reinhart-Rogoff (2009) definition (bank runs leading to closures/mergers/takeovers or large-scale government assistance of important financial institutions). Laeven and Valencia (2017) spans 160 countries, 1970‒2017; Reinhart-Rogoff (2009) covers 70 countries, 1800‒2010, extended to 2017 to match macro dataset.

### Methodology and data
- Reference frame: credit cycles defined by CYGAP (credit-to-GDP gap), computed by detrending the ratio of credit-to-GDP with a two-sided HP filter with smoothing parameter λ=100. Robustness checks use λ up to 1,600.
- Event-study approach drawing on Gourinchas et al. (2001); Mendoza and Terrones (2008, 2012).
- Three exercises:
  - Relate evolution of credit and GDP during positive CYGAP phases to subsequent cyclical dynamics (duration, pace, implied credit-to-GDP changes).
  - Study timing and likelihood of credit reversals and banking crises conditional on each other.
  - Estimate effects of credit reversals and banking crises on credit cycles and economic activity; compute output-forgone estimates.
- Data: yearly data for 179 countries, 1960‒2017, from IFS. Credit measured as domestic bank credit to the private sector (IFS line 22d). Credit converted to real terms using the GDP deflator.

### Key features of credit-to-GDP cycle dynamics
- Sample: 371 entire cycles captured (plus incomplete episodes at sample ends); average entire cycle lasts about ten years, split between positive and negative gaps.
- Average deepening and changes:
  - Overall, credit-to-GDP ratios increase by about nine percentage points during positive gap episodes and drop by about two percentage points during negative gaps.
  - Extreme gaps average 15 percent (positive) and ‒18 percent (negative), wider in developing countries.
- Average credit growth remains positive in both positive and negative CYGAP phases; negative gaps reflect periods where GDP growth is relatively more dynamic than credit growth.
- Likelihood of observing a credit reversal:
  - About 22 percent during the negative phase, almost two times higher than during positive gaps.

### Co-movement of credit growth and GDP growth
- Simple correlation between credit growth and GDP growth: 0.32 overall; 0.09 during episodes of negative credit growth.
- Directional co-movement:
  - Negative credit growth occurs in 23.4 percent of the years when economic activity expands.
  - Positive credit growth occurs in 47.9 percent of the years with contracting activity (Table 2 figures reported in text).
- Concordance index (Harding and Pagan, 2002):
  - Average concordance: 0.68 for developing countries and 0.73 for industrial countries.
  - Concordance driven almost exclusively by periods of positive co-movement.
  - Slight downward trend in concordance in industrial countries and a sharp drop after the 2008‒09 crisis.
- Conditional volatility:
  - Average credit growth is considerably more volatile than GDP growth, particularly in developing countries.
  - During positive CYGAP phases, average pace of credit growth is three to five times faster than GDP growth; during negative CYGAP episodes, single-year contractions of credit can be dramatic.

### Relationship between credit reversals, banking crises, and CYGAP cycles
- Timing and conditional probabilities:
  - Credit reversals are more likely to occur after banking crises.
    - Industrial countries: likelihood of a credit reversal conditional on a previous banking crisis rises from about 8 percent to 35 percent.
    - Developing countries: corresponding probabilities increase from 15 to 25 percent.
  - Banking crises tend to precede credit reversals; likelihood of banking crises increases in years leading to a credit reversal.
- Incidence within CYGAP phases (Table 5 summary):
  - Banking crises occur roughly once every eight years across country groups.
  - Credit reversals are more frequent, particularly in developing countries (about once every five years).
  - Banking crises are more likely when CYGAP is positive (driven by industrial countries); credit reversals are about two times more likely during negative CYGAPs.
- Evolution within cycles (ten-year window centered on CYGAP shift from positive to negative):
  - Developing countries: yearly probabilities of banking crises are rather flat throughout the window.
  - Industrial countries: likelihood of crises peaks early in the CYGAP cycle, four years before CYGAP turns negative — suggesting CYGAP may provide insufficient lead time for countercyclical capital buffers.
  - Probability of credit reversals bottoms early in the positive phase and increases monotonically, peaking when CYGAP becomes negative.
- Cycle characteristics conditional on crises/reversals (Table 6 highlights):
  - Cycles with banking crises: negative phase is on average 1.7 years longer than control group and features a differential drop in credit-to-GDP ratios of 5 percentage points.
  - Cycles with credit reversals: qualitatively similar to crises but differences versus control group less prominent except for contraction in credit growth (almost two-times higher than control). Change in credit-to-GDP ratios associated with reversals is comparable in magnitude to crises.
  - Using within-cycle frequencies as intensity proxies: higher incidence of banking crises → longer and wider negative phases, substantially slower credit growth; reversals → deeper, faster, and wider drops in credit-to-GDP ratios, with contractions sometimes more severe than those associated with banking crises.
- Magnitude interpretation (one-standard deviation frequency shock; Table 7):
  - A 16-percentage point increase in frequency of banking crises (one standard deviation) leads to:
    - ‒0.5 percentage point change in credit-to-GDP ratios during negative CYGAP in developing countries and ‒0.9 in industrial countries (vis-à-vis control groups).
    - Increase in duration of negative phase by 0.328 years (about four months) in developing countries and by 0.169 years (about two months) in industrial countries.
  - Effects of credit reversals surpass those of banking crises across most variables in these comparisons.

### Economic costs of credit reversals and banking crises (Section 6 findings)
- Literature context: prior estimates of banking crisis costs typically range between 4‒15 percent of GDP, with wide dispersion across studies and methodologies.
- Event-study accumulation approach used in paper:
  - Accumulate GDP growth for up to six years after CYGAP turns negative and take difference between cycles with and without crises/reversals (starting point chosen where output gap differential is close to zero).
  - Method assumes exogeneity of crises/reversals; if likelihood is endogenous and inversely related to economic activity, cost estimates would be upward biased (conservative estimates).
- Empirical patterns (Figures 7 and 8 descriptive):
  - Cycles with banking crises show wider output gap swings; developing countries close gap differentials more quickly.
  - For GDP growth, cycles with banking crises overperform control early but underperform substantially afterwards; differences remain protracted, especially in industrial countries.
  - GDP growth differentials for credit reversals are comparable in magnitude to those for banking crises and appear more persistent in developing countries.
- Quantitative estimates (Table 9 summary and regression evidence):
  - Banking crises: large and protracted costs; in industrial countries estimated costs around 11 percentage points of foregone output seven years after start of negative CYGAP (median costs in developing countries much smaller than mean, indicating positive skew).
  - Credit reversals: protracted costs, comparable in magnitude to banking crises when estimated with same methodology.
  - Panel regression [1] results (cycles with vs. without crises/reversals; Table 10):
    - Banking crises associated with an estimated drop in GDP growth of 1.3 percent per year during entire credit-to-GDP episodes (ten-year cycles) → cumulative GDP differential about 14 percentage points (=1.013^10‒1). Using five-year average negative phase, implied cumulative drop around 7 percent (=1.013^10‒1).
    - Credit reversals have substantial economic costs, about one-half of those associated with banking crises.
    - Relative frequencies: reversals occur about once every 7.6 years; crises about once every 5.1 years. Using relative frequencies, overall cost of reversals over time ≈ two-thirds of those associated with banking crises (0.69 = 0.006 / 0.013 × 7.6 / 5.1 as presented in text).
  - Panel regression [2] results (sensitivity to one percentage point increase in frequency; Table 11):
    - A one percentage point increase in frequency of credit reversals leads to a drop of GDP growth of about 1.9 percentage point per year.
    - Estimated losses are larger for industrial countries.

### Policy implications and concluding points
- Credit reversals, defined stringently to capture supply-driven credit contractions during expanding activity, are frequent and economically costly.
- Relative costs: credit reversals impose substantial output losses — about one-half of banking crisis costs per episode, and up to two-thirds when accounting for relative frequencies over time — supporting macroprudential policies aimed at preventing supply-driven contractions in bank credit.
- CYGAP as a policy trigger: evidence that banking crises peak early in positive CYGAP phases (especially in industrial countries) raises questions on CYGAP adequacy as the sole reference metric for countercyclical capital buffers.
- Cross-country differences:
  - Impacts of reversals and crises on economic activity seem larger in industrial than in developing countries, and developing countries often rebound faster after banking crises.
  - Possible channels include deeper financial markets in industrial countries, role of FX depreciation, and external demand in developing countries — hypotheses left for future research.
- Caution: results are not causal estimates due to possible endogeneity and omitted variables; if reversals/crises are more likely under sluggish growth, estimated costs are likely upward biased (conservative).

*Source: wpiea2021103-print-pdf - Section 6 presents estimates of the economic costs of reversals and crises in terms of output*

### 2.5 yearsNot studiedStudy the effect of credit Reversals on the severity of

### wpiea2021103-print-pdf - 2.5 yearsNot studiedStudy the effect of credit Reversals on the severity of

### Costs of Banking Crises and Credit Reversals (Cumulative GDP Growth Differential, 1960‒2017)
- Table 9 reports cumulative GDP growth differentials between credit cycles with and without banking crises (or credit reversals). The starting point (Year 0) is the year when the credit-to-GDP gap turns negative.
- Developing Countries (Banking Crises / Credit Reversals — Mean / Median by Year):
  - Year 0: 0.0000 / 0.0000 / 0.0000 / 0.0000
  - Year 1: -0.0189 / -0.0085 / -0.0018 / -0.0094
  - Year 2: -0.0224 / -0.0069 / -0.0143 / -0.0221
  - Year 3: -0.0194 / -0.0001 / -0.0241 / -0.0443
  - Year 4: -0.0018 / 0.0173 / -0.0332 / -0.0612
  - Year 5: -0.0176 / 0.0097 / -0.0590 / -0.0982
  - Year 6: -0.0503 / 0.0023 / -0.0547 / -0.1072
- Industrial Countries (Banking Crises / Credit Reversals — Mean / Median by Year):
  - Year 0: 0.0000 / 0.0000 / 0.0000 / 0.0000
  - Year 1: -0.0167 / -0.0132 / -0.0181 / -0.0113
  - Year 2: -0.0395 / -0.0343 / -0.0293 / -0.0262
  - Year 3: -0.0650 / -0.0553 / -0.0392 / -0.0417
  - Year 4: -0.0776 / -0.0725 / -0.0639 / -0.0615
  - Year 5: -0.0897 / -0.0869 / -0.0876 / -0.0816
  - Year 6: -0.1010 / -0.1057 / -0.0984 / -0.0953

### Fixed-Effect Regressions of Yearly GDP Growth on Banking Crises and Credit Reversals (1960‒2017)
- Table 10: regressions using dummy variables for cycles with banking crises (DBC) and cycles with credit reversals (DCC).
- Coefficient estimates (with reported significance):
  - Dummy for Cycles with Banking Crises (DBC):
    - All Sample [1]: -0.013** [0.005]
    - All Sample [2]: -0.013* [0.007]
    - Developing [3]: -0.012 [0.010]
    - Industrial [4]: -0.013*** [0.004]
  - Dummy for Cycles with Credit Reversals (DCC):
    - All Sample [1]: -0.006*** [0.002]
    - All Sample [2]: -0.006** [0.002]
    - Developing [3]: -0.004 [0.003]
    - Industrial [4]: -0.011** [0.004]
  - Interaction DBC x DCC:
    - Estimates across columns: 0.004 / 0.004 / 0.005 / 0.004 (standard errors reported)
  - Constant:
    - 0.045*** / 0.045*** / 0.045*** / 0.044*** (standard errors reported)
- Sample and fit statistics:
  - Observations: 7205 / 7205 / 5465 / 1740
  - R-squared: 0.0040 / 0.004 / 0.002 / 0.027
  - Number of countries: 177 / 177 / 139 / 38
  - Sigma_u: 0.020 / 0.020 / 0.020 / 0.018
  - Sigma_e: 0.059 / 0.059 / 0.065 / 0.039
  - Rho: 0.103 / 0.103 / 0.091 / 0.171
  - Average obs per group: 40.7 / 40.7 / 39.3 / 45.8
  - Min obs per group: 2 / 2 / 15 / 22
  - Max obs per group: 57 / 57 / 57 / 57
- Notes:
  - Robust errors reported in columns where indicated.
  - Significance levels: *** p<0.01, ** p<0.05, * p<0.1

### Fixed-Effect Regressions of GDP Growth on the Frequency of Banking Crises and Credit Reversals (1960‒2017)
- Table 11: regressions using frequencies (proportion of years with events within cycles).
- Coefficient estimates (frequency measures):
  - Frequency of Banking Crisis in Cycle (PBC):
    - All Sample [1]: 1/ -0.021*** [0.008]
    - All Sample [2]: -0.021* [0.011]
    - Developing [3]: -0.021 [0.015]
    - Industrial [4]: -0.028*** [0.009]
  - Frequency of Credit Reversal in Cycle (PCC):
    - All Sample [1]: -0.019*** [0.005]
    - All Sample [2]: -0.019*** [0.005]
    - Developing [3]: -0.014** [0.006]
    - Industrial [4]: -0.033** [0.014]
  - Interaction PBC x PCC:
    - Estimates: 0.025 / 0.025 / 0.035 / 0.038 (standard errors reported)
  - Constant: 0.044*** / 0.044*** / 0.045*** / 0.041*** (standard errors reported)
- Sample and fit statistics mirror Table 10:
  - Observations: 7205 / 7205 / 5465 / 1740
  - R-squared: 0.0040 / 0.0040 / 0.002 / 0.023
  - Number of countries: 177 / 177 / 139 / 38
  - Sigma_u: 0.020 / 0.020 / 0.020 / 0.017
  - Sigma_e: 0.059 / 0.059 / 0.065 / 0.039
  - Rho: 0.103 / 0.103 / 0.091 / 0.162
  - Average obs per group: 40.7 / 40.7 / 39.3 / 45.8
  - Min obs per group: 2 / 2 / 15 / 22
  - Max obs per group: 57 / 57 / 57 / 57
- Note: 1/ Frequency is the proportion of years with banking crises or credit reversals within cycles.

### Credit and GDP Growth Concordance and Probabilities Around Events (Figures 1–6)
- Figure 1: plots yearly growth rates of GDP and credit (both in real terms using the GDP deflator) for 1960‒2017. Axes labels show ranges used for GDP Growth and Credit Growth (e.g., GDP Growth axis from -.2 to .2; Credit Growth axis from -.4 to .6).
- Figure 2: shows evolution of concordance between credit growth and GDP growth, 1960‒2017, with axes labeled 0 to 1 and time from 1960 to 2020 for both Industrial countries and Developing countries. Series include Positive Concordance and Total Concordance.
- Figure 3: presents yearly probabilities of credit reversals around the occurrence of banking crises (ten-year window centered at start of banking crises). Vertical axis labeled Probability from 0 to 50; horizontal axis Years -5 to 5. Series: Mean Prob. Credit Crunch; Prob. Credit Crunch; separate panels for Developing and Industrial. Note: Time window centered on banking crises.
- Figure 4: presents yearly probabilities of banking crises around the occurrence of credit reversals (ten-year window centered at start of credit reversal). Vertical axis Probability 0 to 40/50; horizontal Years -5 to 5. Series: Mean Prob. Banking Crisis; Prob. Banking Crisis; separate panels for Developing and Industrial. Note: Time window centered on credit reversals.
- Figures 5–6: show probabilities of banking crises (Figure 5) and probabilities of credit reversals (Figure 6) conditional on the Credit-to-GDP gap, 1960‒2017. Windows centered when the Credit-to-GDP gap goes from positive to negative. Plotted series include Mean Credit Growth, Mean Credit-GDP cygap, Credit Growth p10, Credit Growth p90, Prob. Banking Crisis / Prob. Credit Crunches. Axes and ranges include negative and positive values (e.g., -.2 to .6 on y-axis) and x-axis from -10 to 10 or -5 to 5 as labeled.

### GDP Dynamics Conditional on Events (Figures 7–8)
- Figures 7–8: depict GDP gaps and GDP growth dynamics centered when the Credit-to-GDP gap goes from positive to negative (window -5 to 5 years):
  - Figure 7: GDP Dynamics Conditional on Banking Crises, 1960‒2017.
    - Panels for GDP Gaps (Developing and Industrial) with ranges (e.g., -.001 to .003) and GDP Growth (Developing: .025 to .05; Industrial: .01 to .04). Series include Cycles without Crises and Cycles with Crises.
  - Figure 8: GDP Dynamics Conditional on Credit Reversals, 1960‒2017.
    - Panels for GDP Gaps (Developing and Industrial) with ranges (e.g., -.003 to .002) and GDP Growth (Developing: .03 to .055; Industrial: .01 to .05). Series include Cycles without Crunches and Cycles with Crunches.
- Note: Window centered when the Credit-to-GDP Gaps goes from positive to negative.

*Source: wpiea2021103-print-pdf - 2.5 yearsNot studiedStudy the effect of credit Reversals on the severity of (IMF, PDF chapter/section).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021103-print-pdf.pdf_