## wpiea2022194-print-pdf — Sections 1 and 4

## Source details

**Canonical URL:** [wpiea2022194-print-pdf — Sections 1 and 4](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022194-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022194-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022194-print-pdf.pdf.json)

---

### Context and motivation
- At the outset of the COVID-19 pandemic, prudential authorities focused on bank dividend policies because of the expected deterioration of bank asset quality and the threat to banking sector stability.
- Trade-off for banks: remunerate shareholders via dividends versus retaining earnings to absorb future losses or provide regulatory capital to support lending.
- Unlike in the global financial crisis, European regulators recommended that all banks should not make distributions to prevent outflow of regulatory capital before crisis-induced losses materialised.

### Regulatory framework and pandemic measures
- Basel Committee standards give supervisors tools to impose restrictions on payouts of dividends and share buybacks (BCBS, 2019, chapter RBC30).
- Macroprudential capital buffers include:
  - the capital conservation buffer, set to 2.5% of risk-weighted assets for all banks;
  - the countercyclical buffer applied to specific exposures;
  - firm-specific buffers for global and other systemically important institutions;
  - in some countries, systemic risk buffers.
- Use of macroprudential capital buffers triggers automatic restrictions (the ‘maximum distributable amount’ (MDA)) on distribution of profits and payment of coupons on Additional Tier 1 (AT1) instruments.
  - The MDA decreases in increments of 20 percentage points in lockstep with the amount of capital buffers which have been used.
  - A bank that has completely exhausted its capital buffers would not be allowed to pay out any dividends or coupons on AT1 instruments.
- Pandemic-era supervisory recommendations in the euro area:
  - ECB recommendation on 27 March 2020: banks should refrain from paying dividends and buying back shares at least until end-September 2020.
  - Recommendation extended in July 2020 until end-2020.
  - On 15 December 2020 the ECB asked banks to exercise extreme prudence in dividend and buyback decisions until 30 September 2021, expecting banks not to distribute more than 15% of their cumulative 2019 and 2020 profits and to limit payouts to less than 0.2% of total risk-weighted assets (ECB, 2020).
  - ESRB adopted a similar recommendation for all financial firms in the European Union (ESRB, 2020).
  - Similar actions taken by supervisors in Denmark, Sweden and the UK.
- State aid framework: EU state aid framework generally requires banks receiving state support to refrain from paying dividends; restrictions should be introduced as soon as a capital shortfall that would likely result in state intervention is identified (European Commission, 2013).

### Research objectives and empirical approach
- Objectives:
  - Investigate drivers of bank dividend policies in the euro area.
  - Assess counterfactual reaction of banks to the COVID-19 pandemic absent restrictions on dividend payouts.
  - Identify whether banks account for expected future economic conditions, economic uncertainty, or their own earnings’ prospects when deciding dividend payouts.
- Samples:
  - 69 listed banks observed over 2005–2019; data from SNL Financial.
  - About 1,400 banks (majority unlisted) observed over 2010–2019; data from Orbis BankFocus.
- Dependent variable: ratio of total dividend paid over total after-tax earnings.
- Exclusions: banks with negative after-tax earnings and negative capital ratios are excluded.
- Explanatory groups mapped to hypotheses:
  - Asymmetric information: ROA, ROE, opacity, growth opportunities.
  - Agency: log assets (size), market capitalisation, World Bank index of shareholder legal protection.
  - Business cycle: GDP growth.
- Econometric specifications:
  - Dynamic panel with lagged dividend payout ratio; extended to include forward-looking variables (model-based ROAP and market uncertainty indices).
  - Estimators: random effects (RE) and Tobit to account for truncation at zero.
  - Explanatory variables lagged by one year to mitigate endogeneity; time effects included.

### Key empirical findings (baseline results)
- Supported hypotheses and associations:
  - Asymmetric information: higher profitability associated with higher dividend payout.
  - Agency: larger banks pay higher dividends; superior investor protection associated with higher dividends.
  - Regulatory: banks with higher capital pay higher dividends.
  - Credit risk negatively associated with dividend payouts.
  - Past dividend payouts positively associated with future payouts (dividend smoothing).
- Economic significance (impact of one standard deviation change on dividend payout ratio, in percentage points):
  - Capital: ≈ 11 percentage points.
  - Size: ≈ 8 percentage points.
  - Investor protection: ≈ 10 percentage points.
  - Profitability and credit risk effects smaller than regulatory and agency effects.
- Selected coefficient estimates (coefficients with standard errors in parentheses; ***, **, * denote significance at 1%, 5%, 10%):
  - Lagged dividend payout ratio:
    - Listed banks RE: 0.598*** (0.045)
    - Listed banks Tobit: 0.603*** (0.040)
    - Unlisted banks RE: 0.431*** (0.011)
    - Unlisted banks Tobit: 0.215*** (0.012)
  - Capital:
    - Listed banks RE (col.1): 1.036*** (0.354)
    - Listed banks Tobit (col.2): 1.152 (0.291)
    - Unlisted banks RE (col.7): 0.413*** (0.067)
    - Unlisted banks Tobit (col.8): 0.384*** (0.068)
  - ROA:
    - Listed banks RE (col.1): 3.664 (2.745)
    - Listed banks Tobit (col.2): 5.006** (2.868)
    - Unlisted banks RE (col.7): 3.127*** (0.203)
    - Unlisted banks Tobit (col.8): 2.684*** (0.231)
  - Size (log assets):
    - Listed banks RE (col.1): 1.878*** (0.605)
    - Listed banks Tobit (col.2): 2.326*** (0.724)
    - Unlisted banks RE (col.7): 1.23*** (0.224)
    - Unlisted banks Tobit (col.8): 0.702*** (0.286)
  - Minority shareholder protection:
    - Listed banks RE (col.1): 2.721*** (0.676)
    - Listed banks Tobit (col.2): 3.402*** (0.817)
    - Unlisted banks RE (col.7): 5.862*** (0.622)
    - Unlisted banks Tobit (col.8): 13.29*** (0.862)
  - Credit risk (selected):
    - Listed banks Tobit (col.4): -2.583* (1.611)
    - Listed banks Tobit (col.5): -3.252** (1.622)
    - Unlisted banks RE/Tobit (cols.9–12): coefficients around -2.126*** (0.157) and -0.805*** (0.124) depending on specification.
- Model fit and samples:
  - R sq. reported as 0.54 (for listed banks columns reported).
  - Sample sizes vary by column (listed banks: N = 444, 439, 383; unlisted banks: N = 5,855 or 5,722).

### Role of MDA regime and structural break tests
- Structural break test for introduction of MDA in 2014 (dummy MDA = 1 from 2014; interacted with capital adequacy) finds no supporting evidence that the formal introduction of MDA altered the relationship between payouts and capital adequacy.
- Selected coefficients from Table 2 (listed banks; RE and Tobit) show MDA coefficients not statistically significant:
  - MDA: 0.612 (0.636) RE; 0.677 (0.537) Tobit.
  - Other coefficients remain consistent with baseline patterns.

### Forward-looking variables and uncertainty
- Market valuation and model-based forward-looking profitability:
  - Price-to-book ratio not statistically significant determinant of dividend payouts in main specifications.
  - Model-based forward-looking profitability proxy (ROAP) and market-based economic uncertainty indices (Citi indices) provide no strong evidence that banks systematically account for forward-looking variables when setting dividends; estimated coefficients not statistically significant in main specifications.
- Selected results from Table 3:
  - ROAP: -3.261 (4.547) and -3.259 (4.547) (not significant).
  - Economic forecast uncertainty index: 5.577 (3.476) (not significant at conventional levels).
  - Price-to-Book: -0.0122 (0.0217) (not significant).
  - Lagged dividend payout ratio remains highly significant: 0.687*** (0.0356), 0.695*** (0.0373), 0.693*** (0.0373) across columns.
- Caveat: lack of significance may reflect measurement error in forward-looking proxies.

### Robustness checks and stress-period analysis
- Alternative capital indicator (Equity/Total assets) yields qualitatively similar results but loses statistical significance in Tobit specifications.
- Alternative size indicator (market capitalisation) confirms relevance of size but may lose significance due to higher variability.
- ROE as alternative profitability indicator yields qualitatively similar results but ROE can lose statistical significance.
- Opacity proxy (residual-based forecast errors) shows no supportive evidence.
- Growth opportunities (loan book growth) support hypothesis that higher growth opportunities reduce expected dividend payout; selected coefficient: -0.260* (RE) and -0.229* (Tobit) with N = 377.
- Financial crises / stress episodes (Crisis dummy = 1 for 2008–2012):
  - Limited evidence of statistically significant drop in payouts during the stress period.
  - Interaction Crisis dummy * Capital positive and statistically significant: 1.286* (0.698) RE; 1.284** (0.625) Tobit, suggesting capital’s relationship with payouts differs during 2008–2012.
  - Crisis dummy coefficients: -4.30 (27.37) RE; -4.31** (2.17) Tobit (Tobit significant at 5%).
  - Table A.3.6 selected coefficients (RE and Tobit):
    - Lagged dividend payout ratio: 0.579*** (0.0446) RE; 0.647*** (0.0376) Tobit.
    - Size: 1.499* (0.902) RE; 1.708* (0.970) Tobit.
    - Minority shareholder protection: 2.415* (1.347) RE; 2.691** (1.140) Tobit.
    - Credit risk: -4.454 (3.573) RE; -4.947 (3.241) Tobit.
  - N = 383; regressions over 2005–2019 with time effects.

### Counterfactual simulations and policy implications
- Counterfactual simulations for a sample of 30 listed euro area banks using actual realizations up to Q3 2020 and econometric specifications (1)–(6):
  - Payout ratios would be adjusted downward by most banks, although the adjustment would be small, up to 4 percentage points for most banks, reflecting the drop in profits and a broadly steady level of capital.
  - Together with a projected moderate increase in payouts by two banks, this would lead to an almost unchanged total value of distributions of about EUR 19 billion for this group of banks.
- Supervisory recommendation effect:
  - ECB communicated that the recommendation contributed to maintaining €27 billion of regulatory capital within the banking sector, which might otherwise have been paid out.
- Interpretation: Without supervisory recommendations, some reduction in dividends paid out could have been expected in 2020, but it would likely have been limited in size; supervisory recommendations therefore played an important role in retaining capital during the most uncertain phase of the pandemic, increasing loss absorption capacity.
- Caveat: "the Covid-19 crisis led to economic disruptions that were unprecedented in Europe in peace-time, the relationships identified from the data may have no longer held. Its results should therefore be viewed with some caution as the magnitude of the reduction in dividends would likely be understated."

*Source: wpiea2022194-print-pdf — Sections 1 (Introduction) and 4 (Empirical Investigation).*

### 1. Introduction _____________________________________________________________________________________ 3

### 1. Introduction

### Context and motivation
- At the outset of the COVID-19 pandemic, prudential authorities focused on bank dividend policies because of the expected deterioration of bank asset quality and the threat to banking sector stability.
- Banks face a trade-off between paying dividends to remunerate shareholders and retaining earnings to absorb potential future losses or provide regulatory capital to support lending.
- Unlike in the global financial crisis, in the COVID-19 pandemic European regulators recommended that all banks should not make distributions in order to prevent an outflow of regulatory capital from the banking sector before the materialisation of crisis-induced losses.

### Research objectives and approach
- Investigate the drivers of bank dividend policies in the euro area and assess a counterfactual reaction of banks to the COVID-19 pandemic.
- Identify whether banks historically account for expected future economic conditions, economic uncertainty, or their own earnings’ prospects when deciding dividend payouts.
- Carry out counterfactual simulations to estimate how large euro area banks would have responded to the pandemic in the absence of restrictions on dividend payouts.
- Focus on two samples of European banks: listed banks and all banks (including non-listed banks).

### Key findings (from the paper)
- Banks paid out a larger share of profits as dividends when they were more profitable and better-capitalised.
- Higher dividend payouts are associated with larger banks and with superior shareholder protections, consistent with the agency cost hypothesis.
- Banks do not seem to systematically account for future expected economic developments, economic uncertainty, or their own earnings’ prospects when making dividend payout decisions.
- Counterfactual simulations suggest that some reduction in dividends paid out could have been expected in 2020, but it would have likely been limited in size.
- Implication: supervisory recommendations played an important role in retaining capital in the banking sector during the most uncertain phase of the pandemic, increasing loss absorption capacity.

### Organisation of the paper (overview)
- Section 2: literature review on dividend payout decisions by banks.
- Section 3: regulatory framework and summary of euro area restrictions introduced in 2020.
- Section 4: data and empirical methodology.
- Section 5: results and counterfactual simulations.
- Final section: key conclusions.

---

### Regulatory framework and measures during the COVID-19 pandemic
- Basel Committee standards grant supervisors supervisory tools including the possibility to impose restrictions on payout of dividend and share buybacks (BCBS, 2019, chapter RBC30).
- Post-global financial crisis reforms increased the capital levels at which regulation constrains payouts and introduced bank-specific automatic restrictions and a capital buffer framework.
- Macroprudential capital buffers include:
  - the capital conservation buffer, set to 2.5% of risk-weighted assets for all banks;
  - the countercyclical buffer applied to specific exposures;
  - firm-specific buffers for global and other systemically important institutions;
  - in some countries, systemic risk buffers.
- Use of macroprudential capital buffers triggers automatic restrictions (the ‘maximum distributable amount’ (MDA)) on distribution of profits and payment of coupons on Additional Tier 1 (AT1) instruments.
  - The MDA decreases in increments of 20 percentage points in lockstep with the amount of capital buffers which have been used.
  - A bank that has completely exhausted its capital buffers would not be allowed to pay out any dividends or coupons on AT1 instruments.

- During the COVID-19 pandemic, euro area regulators recommended tighter dividend restrictions than those implied by MDA rules, because MDA would only bind after losses materialised.
  - ECB recommendation on 27 March 2020: banks should refrain from paying dividends and buying back shares at least until end-September 2020.
  - ESRB adopted a similar recommendation for all financial firms in the European Union (ESRB, 2020).
  - Regulators extended and updated recommendations as uncertainty evolved:
    - Recommendation extended in July 2020 until end-2020.
    - On 15 December 2020 the ECB asked banks to exercise extreme prudence in dividend and buyback decisions until 30 September 2021, expecting banks not to distribute more than 15% of their cumulative 2019 and 2020 profits and to limit payouts to less than 0.2% of total risk-weighted assets (ECB, 2020).
  - Similar actions were taken by supervisors in Denmark, Sweden and the UK.

- State aid considerations:
  - The European Union state aid framework for banks generally requires banks receiving state support to refrain from paying dividends.
  - Such restrictions should be introduced as soon as a capital shortfall that would likely result in state intervention is identified (European Commission, 2013).

*Source: wpiea2022194-print-pdf - 1. Introduction*

### 4. Empirical Investigation

### 4. Empirical Investigation

### 4.1 Data
- Two complementary unbalanced panels of euro area banks:
  - Sample of 69 listed banks observed over 2005–2019; data from SNL Financial.
  - Sample of about 1,400 banks (majority unlisted) observed over 2010–2019; data from Orbis BankFocus.
- Dependent variable: ratio of total dividend paid over total after-tax earnings (following Baker and De Ridder (2018)).
- Exclusions to limit outliers: banks with negative after-tax earnings and negative capital ratios are excluded.
- Explanatory variable groups mapped to hypotheses:
  - Asymmetric information hypothesis: ROA, ROE, opacity, growth opportunities.
  - Agency hypothesis: log of assets (size), market capitalisation, World Bank index of shareholder legal protection (higher values = greater outsider minority protection).
  - Business cycle control: GDP growth from ECB Statistical Data Warehouse.
- Data definitions and sources described in Table A1 (in source).

### 4.2 Descriptive Analysis
- Dividend payouts:
  - Distribution truncated at zero; substantial fraction of banks choose not to pay dividends.
  - Payouts peaked immediately prior to the global financial crisis and declined thereafter.
  - High cross-sectional variance at any point in time, indicating strong bank-specific effects.
- Profitability and valuation:
  - Average bank profitability fell abruptly to a trough around 2012 and recovered only moderately thereafter.
  - Price-to-book ratio for traded bank stocks revalued downwards over the sample, reaching lowest point by 2019.
- Capitalisation and risk:
  - Regulatory capital ratios increased steadily between 2008 and 2013; further increased between 2015 and 2017 and stabilised at almost double pre-crisis levels.
  - Retained earnings contributed to an increase in total capital by about 8.3% over 2010–2012 for a sample of 31 large European banks, while average payout ratios declined from 38% to 26% (Cohen and Scatigna (2014), as cited).
  - Non-risk-weighted leverage ratios also increased, though less than risk-weighted ratios.
  - Credit risk receded slowly in the second half of the 2010s.

### 4.3 Econometric Methodology
- Baseline dynamic panel specification (variants estimated):
  - DP_it = α0 + α1 DP_it−1 + Σ α2,k AIHV_k,it−1 + Σ α3,k AHV_k,it−1 + Σ α4,k OV_k,it−1 + Σ ε_it
  - Extended specification adds forward-looking variables FLV_k,it−1: DP_it = α0 + α1 DP_it−1 + Σ α2,k AIHV_k,it−1 + Σ α3,k AHV_k,it−1 + Σ α4,k OV_k,it−1 + Σ α5,k FLV_k,it−1 + Σ ε_it
- Estimators:
  - Initial estimation with random effects (RE) to account for institutional features and slow-moving governance indicators.
  - Tobit model applied to account for dependent variable truncated at zero.
- Endogeneity mitigation: explanatory variables lagged by one year.
- Baseline analysis sequence: test asymmetric information and agency cost hypotheses, then include credit risk and GDP growth.
- Forward-looking analysis: include model-based forward-looking profitability proxy and market-based uncertainty indices.

### 4.4 Baseline Results
- Hypotheses supported:
  - Asymmetric information hypothesis: bank profitability positively associated with dividend payout; higher profit → higher dividends.
  - Agency hypothesis: larger banks pay higher dividends; banks in countries with superior investor protection pay higher dividends.
  - Regulatory hypothesis: banks with higher capital pay higher dividends.
- Additional findings:
  - Credit risk negatively associated with dividend payouts.
  - Past dividend payouts positively associated with future payouts (dividend smoothing).
- Economic significance (implicit elasticities; impact of one standard deviation change on dividend payout ratio, figures in percentage points):
  - Increase in capitalization by one standard deviation → increase in dividend payout ratio by about 11 percentage points.
  - Impact of size ≈ 8 percentage points.
  - Impact of investor protection ≈ 10 percentage points.
  - Profitability and credit risk effects smaller than regulatory and agency effects.
- Selected coefficients from Table 1 (listed banks and unlisted banks; coefficients with standard errors in parentheses; ***, **, * denote significance at 1%, 5%, 10%):
  - Lagged dividend payout ratio:
    - Listed banks RE: 0.598*** (0.045)
    - Listed banks Tobit: 0.603*** (0.040)
    - Unlisted banks RE: 0.431*** (0.011)
    - Unlisted banks Tobit: 0.215*** (0.012)
  - Capital:
    - Listed banks RE (col.1): 1.036*** (0.354)
    - Listed banks Tobit (col.2): 1.152 (0.291)
    - Unlisted banks RE (col.7): 0.413*** (0.067)
    - Unlisted banks Tobit (col.8): 0.384*** (0.068)
  - ROA:
    - Listed banks RE (col.1): 3.664 (2.745)
    - Listed banks Tobit (col.2): 5.006** (2.868)
    - Unlisted banks RE (col.7): 3.127*** (0.203)
    - Unlisted banks Tobit (col.8): 2.684*** (0.231)
  - Size (log assets):
    - Listed banks RE (col.1): 1.878*** (0.605)
    - Listed banks Tobit (col.2): 2.326*** (0.724)
    - Unlisted banks RE (col.7): 1.23*** (0.224)
    - Unlisted banks Tobit (col.8): 0.702*** (0.286)
  - Minority shareholder protection:
    - Listed banks RE (col.1): 2.721*** (0.676)
    - Listed banks Tobit (col.2): 3.402*** (0.817)
    - Unlisted banks RE (col.7): 5.862*** (0.622)
    - Unlisted banks Tobit (col.8): 13.29*** (0.862)
  - Credit risk (selected):
    - Listed banks Tobit (col.4): -2.583* (1.611)
    - Listed banks Tobit (col.5): -3.252** (1.622)
    - Unlisted banks RE/Tobit (cols.9–12): coefficients around -2.126*** (0.157) and -0.805*** (0.124) depending on specification.
- Model features and fit:
  - Time effects included in regressions.
  - R sq. reported as 0.54 (for listed banks columns reported).
  - Sample sizes:
    - Listed banks: N = 444, 439, 383 depending on column.
    - Unlisted banks: N = 5,855 or 5,722 depending on column.

### 4.5 Role of MDA Regime
- Structural break test for introduction of Maximum Distributable Amount (MDA) in 2014:
  - Dummy variable MDA = 1 starting from 2014; interacted with capital adequacy.
  - No supporting evidence for hypothesised structural break: introduction of formal MDA restrictions does not show a statistically significant change in the relationship between payouts and capital adequacy.
- Selected coefficients from Table 2 (listed banks; RE and Tobit):
  - Lagged dividend payout ratio: 0.585*** (0.0450) RE; 0.653*** (0.0378) Tobit.
  - Capital: 0.513 (0.573) RE; 0.307 (0.419) Tobit.
  - ROA: 1.848*** (0.661) RE; 2.133*** (0.663) Tobit.
  - Size: 5.812* (3.186) RE; 7.787** (3.172) Tobit.
  - Minority shareholder protection: 2.175*** (0.777) RE; 2.422*** (0.782) Tobit.
  - Credit risk: -2.779* (1.618) RE; -3.867* (2.057) Tobit.
  - MDA: 0.612 (0.636) RE; 0.677 (0.537) Tobit.
  - N = 383; time effects included.

### 4.6 Role of Forward-Looking Variables
- Price-to-book ratio (market valuation) tested as a determinant of dividend policy:
  - Result: banks do not seem to systematically account for their own market valuation when setting dividend payouts; estimated sign is as expected but not statistically significant in main specifications.
- Model-based forward-looking profitability proxy (ROAP) and market-based economic uncertainty indices tested:
  - ROAP constructed using bank-specific and macroeconomic fundamentals; proxy moved forward by one year.
  - Market-based uncertainty proxies: Citi indices of economic forecast uncertainty and consensus change.
  - Results in Table 3:
    - Lagged dividend payout ratio: 0.687*** (0.0356), 0.695*** (0.0373), 0.693*** (0.0373) across columns (1)–(3).
    - Capital: 0.849*** (0.303), 0.867*** (0.329), 0.865*** (0.329).
    - ROA: 5.587* (3.298) in column (1).
    - ROAP: -3.261 (4.547) and -3.259 (4.547) in columns (2)–(3).
    - Price-to-Book: -0.0122 (0.0217) in column (1) (not significant).
    - Economic consensus change index: 0.115 (0.0714) in column (2) (not significant at conventional levels).
    - Economic forecast uncertainty index: 5.577 (3.476) in column (3) (not significant at conventional levels).
    - Credit risk coefficients: -2.676 (1.874), -3.717* (2.120), -3.715* (2.120) across columns.
    - Minority shareholder protection: 3.305*** (0.751), 2.441*** (0.837), 2.438*** (0.837).
    - Size: 1.830*** (0.652), 0.449 (0.696), 0.448 (0.696).
    - N = 422 for column (1); N = 340 for columns (2) and (3); time effects included.
- Interpretation and caveats:
  - Neither model-based forward-looking profitability proxy nor market-based uncertainty indices provide strong evidence that banks account for forward-looking variables when setting dividend payouts.
  - Lack of significance may be driven by measurement error in forward-looking proxies.

*Source: IMF Working Paper — 4. Empirical Investigation (chapter/section from wpiea2022194-print-pdf).*

### 4.7 Robustness Checks

### 4.7 Robustness Checks

### Capital Adequacy
- Alternative indicator: capital over assets (Equity/Total assets) instead of capital over risk-weighted assets (Equity/RWA).
- Result: "does not qualitatively differ across Tobit specifications, although the capital over assets indicator loses statistical significance."
- Interpretation: likely reflects that "key prudential requirements are not set in terms of the capital over assets indicator, which only works as a backstop, but are binding in terms of capital over risk-weighted assets."
- Reference: Table A.3.1 shows the alternative-capital regressions.

### Size
- Alternative indicator: market-based measure defined as log of bank capitalization on the stock market (Market capitalization), compared to log(assets).
- Result: "The results do not materially change across the specifications, confirming the relevance of size for dividend payout policy, although the variable loses significance."
- Possible cause: "the higher degree of variability of this market-based measure."
- Reference: Table A.3.2 reports the alternative-size regressions.

### Profitability Indicators
- Alternative indicator: ROE tested instead of ROA.
- Result: "The results do not qualitatively differ across the board of various specifications, although the ROE indicator loses statistical significance."
- Reference: Table A.3.3 contains the ROE-based regressions.

### Opacity (Asymmetric Information)
- Alternative forward-looking variable: proxy for information opacity constructed from residuals of a fundamental-based model of profitability (drawing on Park (1999) and Crouzille et al (2004)); higher residuals = higher forecast errors = higher opacity.
- Result: "The results show no supportive evidence for this variable."
- Reference: Table A.3.4 presents regressions including the Opacity variable.

### Growth Opportunities
- Measure: loan book growth / annual growth rate of bank assets as proxy for growth opportunities.
- Hypothesis: "the higher the growth opportunity expressed in terms of loan book growth, the lower the expected dividend payout."
- Result: "The results confirm the relevance of this hypothesis."
- Reference: Table A.3.5 reports the Growth opportunity regressions.
- Exact estimates (selected): Growth opportunity coefficient reported as -0.260* (RE) and -0.229* (Tobit) with N 377 (Table A.3.5).

### Financial Crises / Stress Episodes
- Test: dummy for the period 2008–2012 constructed and interacted with explanatory variables to examine (i) change in dividend payouts during stress episodes and (ii) whether relationships changed during stress episodes.
- Result: "There is only limited evidence of a statistically significant drop in payouts during the stress period."
- Additional finding: "There is however some supportive evidence for the importance of capital during the stress episodes."
- Comparison: "This seems to be broadly consistent with the findings of Abreu and Gulamhussen (2013) who find the more prominent role of capitalization for bank dividend payouts during the Global Financial Crisis."
- Reference: Table A.3.6 (Annex 3) presents the stress-period interaction results.

### Additional robustness-summary points (from associated simulation and context)
- Counterfactual simulation context (linked to robustness of results): using econometric specifications (1) to (6) and actual realizations up to Q3 2020 for a sample of 30 listed euro area banks:
  - "payout ratios would be adjusted downward by most banks, although the adjustment would be small, up to 4 percentage points for most banks, reflecting the drop in profits and a broadly steady level of capital."
  - Outcome: "together with a projected moderate increase in payouts by two banks, this would lead to an almost unchanged total value of distributions of about EUR 19 billion for this group of banks."
- Regulatory note: the ECB communicated that the recommendation contributed to maintaining €27 billion of regulatory capital within the banking sector, which might have otherwise been paid out.
- Caveat stated in the source: "the Covid-19 crisis led to economic disruptions that were unprecedented in Europe in peace-time, the relationships identified from the data may have no longer held. Its results should therefore be viewed with some caution as the magnitude of the reduction in dividends would likely be understated."

*Source: IMF staff calculations and tables A.3.1–A.3.6 as presented in the provided document.*

### Section 4.

### Section 4

### Table A.3.6: Impact of Financial Crises — Main regression results
- Dep variable: (1) Dividend payout ratio (RE); (2) Tobit
- Coefficients (standard errors in parenthesis):
  - Lagged dividend payout ratio: (1) 0.579*** (0.0446); (2) 0.647*** (0.0376)
  - Capital: (1) 0.0602 (0.639); (2) -0.121 (0.499)
  - Size: (1) 1.499* (0.902); (2) 1.708* (0.970)
  - ROA: (1) 6.095 (4.921); (2) 6.559 (4.659)
  - Minority shareholder protection: (1) 2.415* (1.347); (2) 2.691** (1.140)
  - GDP growth: (1) -1.226 (1.746); (2) -1.184 (1.139)
  - Credit risk: (1) -4.454 (3.573); (2) -4.947 (3.241)
  - Crisis dummy: (1) -4.30 (27.37); (2) -4.31** (2.17)
  - Crisis dummy * Capital: (1) 1.286* (0.698); (2) 1.284** (0.625)
  - Crisis dummy * Size: (1) 0.688 (1.359); (2) 0.808 (1.333)
  - Crisis dummy * ROA: (1) 0.603 (7.590); (2) 3.023 (6.357)
  - Crisis dummy * Minority shareholder protection: (1) -0.608 (1.892); (2) -0.682 (1.577)
  - Crisis dummy * GDP growth: (1) 1.589 (2.198); (2) 1.143 (1.576)
  - Crisis dummy * Credit risk: (1) 4.188 (4.494); (2) 3.397 (4.225)

- Model features and sample:
  - Time effects: yes
  - N: 383
  - The regressions are panel regressions with random-effects and Tobit (Tobit only for listed banks).
  - The regressions are computed for an unbalanced panel over the period 2005–2019.
  - The baseline regression in Table 1 has been extended to include the “Crisis dummy” (and its interaction) which takes 1 in the period 2008–2012 and 0 elsewhere, as explained in Section 4.

- Statistical significance notation:
  - ***, ** and * denote statistical significance at the levels of, respectively, 1%, 5% and 10%; standard errors are reported in parenthesis.

### Substantive findings implied by the table
- Persistence in payout behavior:
  - Lagged dividend payout ratio is positive and highly significant in both specifications: 0.579*** (0.0446) and 0.647*** (0.0376).
- Role of capital during crisis periods:
  - The interaction Crisis dummy * Capital is positive and statistically significant in both specifications: 1.286* (0.698) and 1.284** (0.625), suggesting capital’s relationship with payouts differs during 2008–2012.
- Crisis-level effects:
  - The Crisis dummy coefficient is negative in both specifications: -4.30 (27.37) and -4.31** (2.17); only the Tobit specification reports statistical significance at 5% for the crisis dummy.
- Governance and payouts:
  - Minority shareholder protection is positive and statistically significant: 2.415* (1.347) and 2.691** (1.140), indicating stronger minority shareholder protection is associated with higher dividend payouts.
- Other coefficients (Size, ROA, GDP growth, Credit risk) show varying signs and generally lack consistent statistical significance across both specifications.

*Source: authors’ calculations (Table A.3.6; regressions over 2005–2019; Crisis dummy = 1 for 2008–2012).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022194-print-pdf.pdf_