## 1. Decision-Based Analysis ## Source details **Canonical URL:** [1. Decision-Based Analysis](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024036-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024036-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024036-print-pdf.pdf.json) --- ### Definition and identification - Zombie banks: banks that are chronically insolvent (negative net worth as stated in banks’ financial statements) or gravely undercapitalized. - Decision-based zombie criterion: - An insolvent (undercapitalized) bank that has not undergone corrective “treatment” (recapitalization and/or restructuring by owners, mergers and acquisitions, or resolution measures, including intervention and liquidation) within a year of being initially reported as non-compliant is classified as a zombie. - Alternative (stricter) criterion: continued insolvency (undercapitalization) two years after initial recognition, regardless of evidence of interim treatment. - Rationale: accounting-, borrower-, or market-based identification infeasible due to cross-country data limitations; decision-based approach relies on observable corrective actions (or lack thereof). ### Data and sample coverage - Full universe of bank financial statements from Fitch and Orbis. - Cross-sectional annual dataset with 266 banks from 69 emerging and developing economies covering the period from 2001-18. - Inclusion criteria: - At least one year of technical insolvency (negative net worth—181 banks) or, alternatively, clear undercapitalization (266 banks). - Sample composition: - More than one third of the banks are from the Asia and Pacific region; remaining banks from Europe, Africa, and the Americas. - About 20 percent of the sample consists of state-owned banks. - 77 percent of all banks remained open for business two years after inadequate capitalization. - Treatment definition: recapitalization and/or restructuring by owners, mergers and acquisitions, or resolution measures (intervention and liquidation). ### Outcome-Based Analysis (complementary classification) - OBA classification: - Non-zombie if either (i) successfully recapitalized by the end of the second year after showing up as undercapitalized, or (ii) put in liquidation within that time frame. - In all other cases the bank is classified as a zombie. - Empirical prevalences: - Using decision-based classification, 34 percent of banks in our sample are classified as zombies. - Using outcome-based classification, 33 percent of banks in our sample are zombies. - Less than 7 percent of our sample received treatment one year following insufficient capitalization. - Two sample definitions: - Negative sample: banks with a negative equity-to-assets ratio (equity ratio). - Two percent sample: banks with equity-to-assets ratio below 2 percent. - Assumptions for the two percent sample: RWA density (RWA to total assets) of less than 50 percent. This translates into a capital adequacy ratio of no more than 4 percent. ### Variables, categories, and data collection - Total variables tested: 20 (five in each of four categories: bank-level, macroeconomic, structural, and crisis/program). - Bank-level indicators: equity-to-assets ratio (equity ratio), return on assets (ROA), bank size (contribution to total credit; size relative to GDP), dummy for state ownership. - Macroeconomic indicators: inflation, ratio of government debt to GDP, low-income country dummy (World Bank 2021 classification), growth rate of credit, ratio of credit to GDP. - Structural indicators: Worldwide Governance Indicators (government effectiveness, control of corruption, regulatory quality); IMF Financial Access Survey (number of bank branches and ATMs per 100,000 adults). - Crisis and program indicators: dummies for currency, banking, or debt crises in the year prior, year of, or year following undercapitalization (Laeven and Valencia 2020); IMF program participation dummy. - News searches to detect treatment begin with the bank name and, if necessary, the country name; translations applied as needed. ### Empirical strategy and econometrics - Control group: insolvent or severely undercapitalized banks that received corrective treatment. - Outcome variable: binary zombie classification (DBA: absence of treatment within 1 year primary; OBA: undercapitalization persist to 2 years or liquidation). - Regression framework: - Logit regressions with robust standard errors. - Bivariate regressions (one independent variable at a time) first, then parsimonious multivariate regressions. - Multivariate regressions take one variable from each of the four categories; reduced combinations used if no significant variable available for a block. - Samples: negative sample and two percent sample; regional disaggregation and country-level regressions also employed. ### Main regression findings — decision-based (negative sample) - Bivariate DBA results: - Significant predictors: equity ratio (bank-specific), low-income country status, crisis episodes. - Strongly negative equity ratios increase likelihood of remaining a zombie (deemed “too expensive” to fix). - Strong credit growth associated with a higher likelihood of being a zombie. - Higher inflation associated with a lower likelihood of becoming a zombie. - Banking or currency crisis occurrence associated with a lower likelihood of zombie bank status. - Governance indicators (government effectiveness, control of corruption, regulatory quality) not significant in global regressions. - Multivariate DBA results: - Specifications combining equity ratio with crisis dummies retain signs consistent with bivariate results. - Inflation and credit growth lose significance when included alongside crisis dummies. ### Main regression findings — outcome-based (negative sample) - Bivariate OBA results: - Larger banks are less likely to become zombies. - Equity ratio not statistically significant under OBA for the negative sample. - Banks in low-income countries more likely to be classified as zombies. - Higher government debt to GDP associated with a lower probability of zombie status. - Lower regulatory quality loosely linked to zombie status with an unexpected negative sign. - Banking crises and debt crises associated with lower probability of becoming a zombie. - Countries covered by an IMF program are less likely to develop zombie banks. - Multivariate OBA results: - Size and incidence of a debt crisis significantly reduce probability of zombie status when combined with other variables (size, government debt to GDP, regulatory quality, any-crisis dummy). ### Two percent sample results - Decision-Based (bivariate): - State-owned banks are more likely to be classified as zombies. - State ownership associated with lower profitability, higher risk, less capital, and higher NPLs. - Decision-Based (multivariate): - Including country dummies improves fit considerably; some variables (e.g., banking crisis) lose significance when country dummies are included. - Outcome-Based (multivariate): - Interactions tested between IMF program dummy and banking crisis dummy; included equity ratio, regulatory quality, and low-income dummy. - Banks in countries experiencing a banking crisis are less likely to be zombies, but this result holds only when country dummies are included. ### Regional disaggregation — main contrasts - Sample split into four regions (two percent sample): Africa, Asia and Pacific, Europe, Americas; multivariate regressions for each region. - Regional determinants (decision-based criterion): - Africa: - State ownership important driver. - Bank’s location in a low-income country increases likelihood of zombie status. - Asia and the Pacific: - ROA significant. - Inflation significant, suggesting relatively profitable banks may be allowed to persist as zombies. - Americas: - Bank size significant; smaller banks more likely to become zombies. - Lower control of corruption increases likelihood of zombie bank status. - Europe: - Higher government debt to GDP associated with a lower probability of zombie status. - Model fit: - Regional pseudo R2 between 0.1 and 0.48. - Global specifications without country dummies: pseudo R2 between 0.05 and 0.1. - Interpretation: many effects cancel out in global regressions but emerge regionally. ### Country-level regressions (share of zombie banks) - Dependent variable: share of zombie banks measured by ratio of zombie bank assets to system assets (two percent sample, DBA and OBA). - Bivariate OLS results: - DBA: inflation and banking crises lower probability of a high share of zombie assets. - OBA: higher shares of government debt to GDP associated with an increasing share of zombie assets (contrast with bank-level OBA results). - Most regressors not significant in isolation; multivariate analysis using zombie shares omitted for this reason. ### Interpretation, robustness, and limitations - Consistently significant across many specifications: equity ratio (bank-specific), low-income country status, crisis variables. - Credit growth associated with greater zombie incidence under DBA, pointing to macro-financial booms masking persistent impairment. - Counterintuitive associations: - Higher inflation and occurrence of banking/currency crises linked to lower zombie probability, consistent with treatment pressure or nominal effects on reported performance. - State ownership correlates with higher zombie probability in broader two percent sample. - IMF program participation associated with lower likelihood of zombie banks; significance of program dummy vanishes when combined or interacted with crisis variables. - Regression power relatively low; substantial influence of unobservable factors and omitted variable bias on stakeholders’ decisions to treat or tolerate insolvent banks. - Inclusion of country dummies approximately doubles regression fit, indicating important omitted drivers. - Structural/governance factors and financial inclusion indicators show little robust global significance; a few region-specific exceptions (regulatory quality, control of corruption). - Data limitations: - Inability to detect zombie banks with hidden/unrecognized losses. - Lack of cross-country data on guarantees and liquidity support. - Link between weak banks and weak borrowers recognized but not explored. ### Policy implications and recommendations - Strengthen regulatory framework and establish an effective resolution regime to reduce arbitrary decision-making and improve efficiency in treating undercapitalized banks. - Seek external support, including an IMF program, during crisis episodes to help manage failing banks that might otherwise become zombies. - Design policy taking into account region-specific drivers; global regressions may obscure heterogeneous regional determinants. - Reduce the random component of decision-making by installing proper recapitalization rules, resolution triggers, and clear processes to minimize moral hazard. ### Conclusions and avenues for further research - Contributions: - First comprehensive multi-country study on determinants of zombie bank status outside advanced economies. - Operationalization via objective undercapitalization and DBA/OBA identification. - Empirical testing across bank-specific, country, structural, and crisis-related variables. - Main conclusions: - Country-specific factors matter more than bank-specific or structural factors for the emergence of zombie banks. - Undercapitalized banks in low-income countries more likely to become zombies. - Banking or other crises associated with lower likelihood of undercapitalized banks remaining zombies (reverse relationship). - Some bank-specific evidence: state ownership and equity ratio matter. - Regional disaggregation reveals drivers (e.g., ROA, bank size) that are not significant globally. - Remaining unexplained variation: - Most variation empirically unexplained; many decision drivers unobserved and context-dependent. - Further research suggested: - Productivity implications of inadequate restructuring/resolution of zombie banks. - Tangible impacts of zombie banks on allocation of capital, hiding of losses, and need for state support. - Investigation of less apparent drivers of zombie bank status. *Source: IMF Working Paper — Determinants of Zombie Banks in Emerging Markets and Developing Economies (Decision-Based Analysis and Disaggregation by Region sections).* ### 1. Decision-Based Analysis ............................................................................................. ### 1. Decision-Based Analysis ### Definition and Identification of Zombie Banks - Zombie banks are defined as banks that are chronically insolvent (negative net worth as stated in banks’ financial statements) or gravely undercapitalized. - Decision-based zombie criterion: an insolvent (undercapitalized) bank that has not undergone corrective “treatment” (recapitalization and/or restructuring by owners, mergers and acquisitions, or resolution measures, including intervention and liquidation) within a year of being initially reported as non-compliant is classified as a zombie. - Alternative (stricter) criterion: continued insolvency (undercapitalization) two years after initial recognition, regardless of evidence of interim treatment. - Rationale: cross-country data limitations (inconsistent reporting of NPLs, loan loss reserves, and absence of market data for many banks) make accounting-, borrower-, or market-based identification infeasible; the decision-based approach relies on observable corrective actions (or lack thereof). ### Data and Sample Coverage - Sample construction: - Full universe of bank financial statements from Fitch and Orbis. - Cross-sectional annual dataset with 266 banks from 69 emerging and developing economies covering the period from 2001-18. - Inclusion criteria: at least one year of technical insolvency (negative net worth—181 banks) or, alternatively, clear undercapitalization (266 banks). - Sample composition: - More than one third of the banks are from the Asia and Pacific region; remaining banks from Europe, Africa, and the Americas. - About 20 percent of the sample consists of state-owned banks. - 77 percent of all banks remained open for business two years after inadequate capitalization. - Treatment definition: recapitalization and/or restructuring by owners, mergers and acquisitions, or resolution measures (intervention and liquidation). ### Empirical Strategy (Decision-Based) - Control group: insolvent or severely undercapitalized banks that received corrective treatment as defined above. - Outcome variable: classification as a zombie based on absence of treatment within the decision-based time windows (1 year primary; 2 years alternative). - Regression framework: multiple specifications reported (bivariate and multivariate regressions, samples including a “negative sample” and a “two percent sample”; regional disaggregation and country-level regressions are also employed). ### Key Findings (from decision-based and complementary analyses) - From the set of 20 candidate explanatory variables, only a few are consistently significant across most specifications: - Bank equity ratio (significant). - Low-income country status (significant). - Crisis episodes (significant). - Several macro and bank-specific variables are significant only for specific subsets of the analysis (i.e., not robustly across all specifications). - Structural indicators (e.g., governance variables or financial inclusion indicators) show hardly any evidence of significance in the global regressions. - Regional disaggregation uncovers additional significant drivers that are specific to individual regions; some variables that cancel out at the global level emerge as significant regionally. - Regression power is relatively low, suggesting substantial influence of unobservable factors and omitted variable bias on stakeholders’ decisions to treat or tolerate insolvent banks. - Empirical limitations acknowledged: inability to detect zombie banks with hidden/unrecognized losses and inability to measure non-observable drivers of supervisory or ownership decisions. ### Policy Implications (drawn from decision-based evidence) - Strengthen the regulatory framework and establish an effective resolution regime to reduce arbitrary decision-making and improve the efficiency of treating undercapitalized banks. - Seek external support, including an IMF program, during crisis episodes to help manage failing banks that might otherwise become zombies. - Recognize that policy design should account for region-specific drivers, as global regressions may obscure heterogeneous regional determinants. *Source: IMF Working Paper — Determinants of Zombie Banks in Emerging Markets and Developing Economies (Decision-Based Analysis section).* ### 1. Decision-Based Analysis ### 1. Decision-Based Analysis ### Decision-Based Analysis (DBA): definition and classification - A bank is coded as a zombie (dummy = 1) if there is no evidence of a treatment within the first year after becoming undercapitalized; coded as non-zombie (dummy = 0) if there is evidence of treatment within the first year. - Allowable treatment categories: start of liquidation, closure, declaration of bankruptcy, capital injection, nationalization or privatization, merger or acquisition (M&A), or restructuring to shore up profitability. - Using this classification, 34 percent of banks in our sample are classified as zombies. - The most common type of treatment was M&A, followed by recapitalization, restructuring, liquidation, nationalization, and closure. - Less than 7 percent of our sample received treatment one year following insufficient capitalization. ### Outcome-Based Analysis (OBA): complementary quantitative classification - A bank is classified as non-zombie if either: - (i) it is successfully recapitalized by the end of the second year after showing up as undercapitalized, or - (ii) it has been put in liquidation within that time frame. - In all other cases the bank is classified as a zombie. - Using this classification, 33 percent of banks in our sample are zombies. - Two samples are used: - Negative sample: banks with a negative equity-to-assets ratio (equity ratio) — treated as insolvent using total assets as the scaling factor (leverage ratio). - Two percent sample: banks with equity-to-assets ratio below 2 percent. - Assumptions for the two percent sample: - RWA density (RWA to total assets) of less than 50 percent. - This translates into a capital adequacy ratio of no more than 4 percent. ### Data collection note - News searches to detect treatment begin with the bank name and, if necessary, the country name; translations applied as needed. ### Key summary statistics and sample composition - 25 percent of our sample is classified as low income by the World Bank. - The analysis tests a total of 20 variables (five in each of four categories: bank-level, macroeconomic, structural, and crisis/program). ### Potential drivers of zombie status — categories and indicators - Bank-level indicators: - Equity-to-assets ratio (equity ratio), return on assets (ROA), bank size proxied by contribution to total credit, size relative to GDP, and dummy for state ownership. - Lack of cross-country data on guarantees and liquidity support prevents analysis of those factors. - Macroeconomic indicators: - Inflation, ratio of government debt to GDP, low-income country dummy (World Bank 2021 classification), growth rate of credit, ratio of credit to GDP. - Variables sourced from World Bank’s World Development Indicators and the IMF’s International Financial Statistics. - Structural indicators: - Worldwide Governance Indicators: government effectiveness, control of corruption, regulatory quality. - IMF Financial Access Survey: number of bank branches and ATMs per 100,000 adults. - Crisis and program indicators: - Dummy variables for occurrence of currency, banking, or debt crises in the year prior, year of, or year following a bank becoming undercapitalized (Laeven and Valencia 2020). - IMF program participation dummy to capture possible conditionality effects. ### Box 1: Bank restructuring and IMF program conditionality (selected country cases) - Monitoring of Funding Arrangements Database: 92 percent of programs initiated since 2006 include financial sector reform as a conditionality. - Kentikelenis et al. (2016): evaluation found 13,948 conditions associated with the financial sector, monetary policy, and central banking (1985–2014). - Selected country case summaries: - Nepal: 2003 Poverty Reduction and Growth Facility included restructuring troubled commercial and development banks and steps to increase loan recovery rates (IMF 2006). - Ukraine: December 2008 Stand-By Arrangement included commitments to bolster bank resolution framework and resolve a bank to restore confidence (IMF 2008a). - Pakistan: end-2008 Stand-By Arrangement included contingency plans for problem banks and steps to bolster bank resolution capacity (IMF 2008b). - Moldova: post-2015 special supervision of three large banks led to IMF program conditionality requiring enforcement actions, revamp of bank resolution framework, and restructuring (IMF 2016; IMF 2017). ### Econometric approach - Logit regressions with robust standard errors are run to evaluate determinants of zombie bank status for DBA and OBA. - Bivariate regressions (one independent variable at a time) are run first to test predictive power in isolation. - Parsimonious multivariate regressions are then crafted, incorporating several of the most significant variables identified in bivariate analysis. - Multivariate regressions take one variable from each of the four categories of potential drivers; reduced combinations used if no significant variable available for a block. - Both the negative sample and the two percent sample are analyzed with bivariate and multivariate regressions. ### Main regression findings — Decision-Based Analysis (negative sample) - Bivariate DBA regressions (negative sample): - The only significant bank-specific variable: the equity ratio. - Banks with strongly negative equity ratios may be deemed “too expensive” to fix and more likely to remain zombies. - Macroeconomic factors: - Strong credit growth is associated with a higher likelihood of being a zombie. - Higher inflation is associated with a lower likelihood of becoming a zombie. - Banks operating in low-income countries are more likely to be zombies. - Governance indicators (government effectiveness, control of corruption, regulatory quality) not found to be significant determinants. - Existence of a banking or currency crisis is associated with a lower likelihood of zombie bank status. - Multivariate DBA regressions (negative sample): - First specification combined equity ratio with banking crisis dummy; another combined equity ratio with a dummy for any crisis. - Signs of variables remain consistent with bivariate results. - Inflation and credit growth lose significance in multivariate settings when included alongside crisis dummies. ### Main regression findings — Outcome-Based Analysis (negative sample) - Bivariate OBA regressions: - Larger banks are less likely to become zombies. - The equity ratio is not statistically significant in OBA. - Banks in low-income countries are more likely to be classified as zombies. - Higher ratios of government debt to GDP are associated with a lower probability of zombie status. - Lower levels of regulatory quality loosely linked to zombie status with an unexpected negative sign. - Banking crises and debt crises associated with lower probability of becoming a zombie. - Countries covered by an IMF program are less likely to develop zombie banks. - Multivariate OBA regressions: - Combining size, government debt to GDP, debt crisis dummy, and regulatory quality indicator: size and incidence of a debt crisis significantly reduce probability of zombie status. - Similar results when combining size, government debt to GDP, regulatory quality, and any-crisis dummy. ### Two percent sample results (decision-based and outcome-based) - Decision-Based (two percent sample) bivariate regressions: - State-owned banks are more likely to be classified as zombies. - State ownership associated with inefficiencies: lower profitability, higher risk, less capital, and higher NPLs. - Decision-Based multivariate regressions (two percent sample): - Adding country dummies improves fit considerably; some dummies (e.g., banking crisis) lose significance when country dummies are included. - Outcome-Based multivariate regressions (two percent sample): - Interactions tested between IMF program dummy and banking crisis dummy; included equity ratio, regulatory quality, and low-income dummy. - Banks in countries experiencing a banking crisis are less likely to be zombies, but this result holds only when country dummies are included, highlighting substantial country heterogeneity. ### Overall empirical implications - Bank-specific weakness (notably very negative equity ratios) and country context (low-income status) increase the likelihood of persistent undercapitalization (zombie status) under DBA and OBA frameworks. - Rapid credit growth tends to be associated with greater zombie incidence (DBA), suggesting macro-financial booms can mask or precede persistent bank impairment. - Inflation and crisis occurrences exhibit counterintuitive associations (higher inflation and occurrence of banking/currency crises linked to lower zombie probability), consistent with treatment pressure or nominal effects on reported performance. - State ownership correlates with higher zombie probability in the broader two percent sample. - IMF program participation is associated with a lower likelihood of zombie banks, consistent with program conditionality promoting bank resolution. *IMF Working Paper: Determinants of Zombie Banks in Emerging Markets and Developing Economies — 1. Decision-Based Analysis* ### 1. Disaggregation by Region ### 1. Disaggregation by Region ### Regional regressions: key findings - Sample split: wider two percent sample divided into four regions: Africa, Asia and Pacific, Europe, and the Americas; multivariate regressions obtained for each region (see Table 13). - Determinants of zombie status (decision-based criterion) vary markedly by region: - Africa: - State ownership is an important driver of zombie status. - Bank’s location in a low-income country increases likelihood of zombie status. - Asia and the Pacific: - ROA (return on assets), a bank-specific variable, is significant. - Inflation is significant, suggesting relatively profitable banks are allowed to persist as zombie banks. - Americas: - Bank size is a significant factor; contrary to “too-big-to-fail”, it is the smaller banks that are not treated and become zombies. - Lower control of corruption increases likelihood of zombie bank status. - Europe: - Higher level of government debt relative to GDP is associated with a lower probability of zombie status (noted as somewhat counterintuitive). - Regional regressions produce a much tighter fit than aggregate regressions: - pseudo R2 of between 0.1 and 0.48 for regional specifications - compared to between 0.05 and 0.1 for global specifications without country dummies. - Interpretation: many factors in global regressions may be “cancelling out” effects present in specific geographical areas. ### Country-level regressions for share of zombie banks (DBA and OBA) - Approach: - Alternative regressions for the two percent sample using DBA and OBA. - Dependent variable: share of zombie banks measured by ratio of zombie bank assets to system assets. - Explanatory variables: each of 20 variables, using country average of each variable measured at the year a bank enters the sample. - Rationale: captures banking sector concentration; alternative to Herfindahl-Hirschman Index (which was found insignificant). - Bivariate OLS results with robust standard errors (Tables 14 and 15): - DBA results: - Inflation and banking crises lower the probability of a high share of zombie assets. - Consistent with bank-level results using a dummy variable for DBA as dependent variable. - OBA results: - Higher shares of government debt to GDP are associated with an increasing share of zombie assets in a given country. - This contrasts with bank-by-bank regressions for the OBA analysis where government debt to GDP was not significant. - Overall assessment: - Most regressors are not significant in isolation. - For this reason, multivariate analysis using zombie shares was omitted. ### Discussion of results: interpretation and robustness - Disaggregated regressions show stronger and more region-specific relationships than aggregate regressions. - Improved fit at regional level explained by significant regional results offsetting loose fits elsewhere. - At global level, mostly country-specific drivers are significant (e.g., low-income country status, banking crisis). - Bank-specific variables expected to be significant overall are tighter only at regional level: - Equity ratio and state ownership notable for Africa. - Return on assets stands out for Asia and the Pacific. - Bank size significant for the Americas. - Macroeconomic variables largely insignificant in determining zombie status: - Rationale: macro environment should impact probability of default but not necessarily the likelihood of being treated. - Correlation note: - Credit growth significance in countries with higher incidence of zombie banks may reflect level of economic development; credit growth and low-income dummy have a correlation of 0.3. - Concordance and divergence across DBA and OBA: - Consistently significant for both analyses: low-income status and crisis variables. - Equity ratio generally significant except in the OBA for the negative sample. - Decision to treat ailing banks differs from actual outcome: - Bank size does not appear to affect treatment decision, but smaller banks tend to end up as zombies regardless of prior treatment. - Credit growth and inflation are determinants in the treatment decision but not for eventual outcome. - Being under an IMF program does not influence treatment decision, but such programs lower incidence of zombie bank status after the fact; however, program dummy significance vanishes when combined or interacted with any crisis variables. - Crisis variables significant: during banking crises and/or other crisis episodes the zombie bank issue is addressed more extensively, often propelled by IMF programs requiring banking sector clean-up. - Omitted variables and unobserved drivers: - Inclusion of country dummies approximately doubles regression fit, indicating important drivers are not modelled. - Many decision-making factors are difficult to capture and may reflect decision-makers’ preferences to keep specific banks afloat. - The share of unobserved factors appears to dominate empirical results despite extensive regressors tested. - Structural and governance factors: - No strong evidence supporting structural factors’ involvement in zombie bank process. - Mild significance for regulatory quality variable in the two percent sample and control of corruption in the Americas regional regression. - Financial inclusion variables (e.g., number of bank branches or ATMs relative to population) tested initially do not show significant influence. - Lack of significance does not rule out unobserved bank-specific governance factors or access-to-finance considerations, as indicated by credit growth significance. - Data limitations: - Link between weak banks and weak borrowers recognized but not explored due to data limitations; left for future research. ### Conclusions and policy implications - Contributions of the paper: - First comprehensive multi-country study on determinants of zombie bank status outside advanced economies. - Operationalization: objective criterion of clear undercapitalization; identification of zombie banks by lack of treatment by owners and regulators (DBA) and continued undercapitalization regardless of prior treatment (OBA). - Empirical investigation testing variables across four types: bank-specific, country, structural, and crisis-related. - Main conclusions: - Country-specific factors matter more than bank-specific drivers or structural factors for emergence of zombie banks. - Undercapitalized banks in low-income countries are more likely to become zombies. - Countries experiencing banking or other crises are less likely to have undercapitalized banks end up as zombies (reverse relationship). - Some evidence for bank-specific factors: state ownership and equity ratio (size of capital gap matters). - Regional disaggregation reveals specific drivers in each region (e.g., ROA, bank size) that are not significant overall because they do not matter in other regions. - Remaining unexplained variation and recommended policy direction: - Most variation remains empirically unexplained; many factors behind treatment decisions are unobserved and difficult to model. - Decisions on treating or letting banks linger are complex and context-dependent, especially in less-developed countries with less well-defined regulatory recapitalization rules, resolution triggers, and processes. - Suggested policy priorities: - Reduce the random component of decision-making. - Install a proper regulatory framework. - Establish an effective resolution regime. - Both are essential to maximize efficiency in dealing with undercapitalized banks and minimize moral hazard. - Role of external support: - IMF programs in crisis situations can expedite needed bank restructuring and resolution. - Calls for further research: - Productivity implications of inadequate restructuring/resolution of zombie banks. - Tangible impacts of zombie banks: inefficient allocation of capital, hiding of losses, need for state support. - Investigation of less apparent drivers of zombie bank status. *Source: IMF Working Paper — Determinants of Zombie Banks in Emerging Markets and Developing Economies — Section 1. Disaggregation by Region* --- _Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024036-print-pdf.pdf_