## wpiea2023125-print-pdf - Appendix A shows the industry breakdown, aggregated at the two-digit code: a large share of the

## Source details

**Canonical URL:** [wpiea2023125-print-pdf - Appendix A shows the industry breakdown, aggregated at the two-digit code: a large share of the](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023125-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023125-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023125-print-pdf.pdf.json)

---

### Data sources, cleaning, and sample construction
- Industry concentration:
  - Large share of firms concentrated in the industrial sector, consumer discretionary, information technology, and materials.
- Compustat (listed firms) cleaning steps:
  - Drop firms with missing assets, liabilities, debt, or the capital stock.
  - Drop firms with fewer than three years of data on the leverage ratio, the investment rate, the interest coverage ratio (ICR), and sales.
  - Restrict sample to firms whose total debt does not exceed 100 percent of their total assets.
  - Drop observations for countries with fewer than five firms for each quarter.
  - Winsorize key variables at the 2.5/97.5 percentiles at the country level.
  - Deflate nominal variables with the respective country CPI deflator.
- Compustat final sample:
  - Unbalanced panel of 42,760 nonfinancial firms over 2000q1-2021q4.
  - Resulting in 1,770,521 firm-quarter observations.
- Time coverage and trends:
  - Steady increase in the number of firms and countries since 2000.
  - On average, twice as many listed firms in AEs than in EMs, though the gap has been shrinking.

### Private firms (ORBIS) coverage and caveats
- Motivation:
  - ORBIS used to fill gap on private firms not covered by Compustat.
- Caveats and biases:
  - Covers fewer countries relative to Compustat; data on EMs is sparse, analysis falls mostly on AEs.
  - Some large economies are poorly represented (e.g. the US, UK, Japan, and Canada), biasing analysis toward European countries.
  - Important data lags and only annual data, limiting study of zombies to low-frequency analysis.
  - Cross-country differences in disclosure regulations introduce selection bias.
- ORBIS cleaning and sample:
  - Follow standard practice in the literature.
  - Select only private firms by keeping firms that are ‘Unlisted’ or ‘Delisted’.
  - Prioritize consolidated accounts to avoid firm‑year duplicates.
- ORBIS final sample:
  - Covers 43 countries (26 AEs and 17 EMs) in an unbalanced panel of 4,394,313 nonfinancial firms over 1997-2020.
  - Firms in EMs account for roughly 6 percent of the total sample, representing 269,667 firms.
- Industry distribution in ORBIS:
  - Almost half of the sample concentrated in manufacturing, and in wholesale and retail trade, followed by real estate and professional activities.
- Note on ICR computation:
  - For private firms ICR is computed as EBITDA over interest expenses (due to lack of EBIT).

### Country-level indicators and construction
- Banking sector health:
  - Main indicator: regulatory capital buffers = banks’ actual regulatory capital to risk-weighted assets minus minimum required risk-based regulatory capital ratio.
  - Minimum required ratio from the World Bank’s Bank Regulation and Supervision Survey (available for 55 countries from 2001 to 2019).
  - Banks’ actual regulatory capital to risk-weighted assets from IMF’s Financial Soundness Indicators.
  - Nonperforming loans to total gross loans (NPL) available from 2001 to 2021 used as an additional indicator.
- Macroprudential policies:
  - Use IMF’s iMaPP database covering 17 macroprudential policy measures for 182 jurisdictions over 1990-2020.
  - For each measure: tightening = 1, loosening = −1, no change = 0.
  - Indices transformed into cumulative sums from 1990 onward; proxy for overall stance is sum of all 17 indicators.
- Insolvency regime quality:
  - Use crisis preparedness indicator from Araujo et al. (2022), measuring five dimensions; time-invariant snapshot as of 2021; original covers 60 countries, matched with 43 in the firm-level dataset.
- Additional country-level controls:
  - Baseline: real GDP growth, three-month short-term interest rate.
  - Robustness: CPI inflation, real effective exchange rate (REER), current account balance as percentage of GDP, one-year government bond yield, Chinn-Ito index.

### Identification of zombie firms (definition and methodology)
- Conceptual:
  - Zombies broadly defined as firms that are risky, unproductive and unviable but avoid immediate default thanks to continued support.
- Operational (annual Compustat and ORBIS data; two-year persistence):
  - To be classified as a zombie for at least two consecutive years, a firm must satisfy:
    - (i) an ICR below one, and
    - (ii) a leverage ratio above the median firm in the same two-digit industry, and
    - (iii) negative real sales growth for two consecutive years.
  - ICR = EBIT/interest expenses for listed firms; EBITDA/interest expenses for private firms.
  - Leverage = total debt (short-term and long-term) divided by total assets; compared to industry median in same industry and country.
  - Two-year horizon for each indicator to avoid misclassification from cyclical fluctuations.
- Exit from zombie status:
  - A zombie exits only when for two consecutive years it records at least one of:
    - ICR above one, or
    - leverage ratio below the industry median, or
    - positive sales growth.
  - Classification on annual data; merged into quarterly dataset for higher-frequency analysis.
- Rationale vs alternatives:
  - Differs from subsidized-interest definitions; arguments given on bank relationships, interest capitalization/evergreening, low-rate environments, and data limitations on loan rates.
  - Differences from Favara et al. (2022): compare leverage to industry‑country median (not whole sample); require two-year exit conditions (Favara et al. use three); two-year horizon (results similar with three years).
  - Alternative rolling-averages approach yields practically same results.

### Empirical findings on zombie shares (listed vs private)
- Long-run trends:
  - Share of zombie firms for both listed and private firms has been steadily increasing over the last 20 years, especially after the GFC.
  - Low interest rates, ample liquidity, and search for yield may be related to the upward trend.
- Levels and changes:
  - Listed firms:
    - Zombie firms account for over 10 percent of all listed firms in the sample in 2021.
    - Zombie share was 6 percent in 2000.
  - Private firms:
    - Share rose from 1 percent in 1997 to 5.3 percent in 2020.
  - Upticks in zombie shares tend to occur after economic downturns (early‑2000s, after the GFC).
- Pandemic and recent dynamics:
  - Listed zombie share dipped from 2016 to 2019, then rose during the Covid-19 pandemic.
  - Share of private zombies ticked up slightly since 2018; steeper increase after the GFC.
- Weighting and employment:
  - Trends visible when weighting by total debt, total assets, or employment.
  - Zombie share levels similar when weighting by debt (and assets for private firms).
  - Zombie firms tend to account for a smaller share of employment, though employment data are sparse.
- Private vs listed comparison hypotheses:
  - Private firms show a somewhat lower share of zombies than listed firms.
  - Possible explanations: larger firms more likely to receive government subsidies; small private firms may be less able to get loans evergreened; political connections and state ownership (SOEs) as potential drivers.
  - Empirical note: zombification incidence among unlisted SOEs is very similar to other firms (Figure C.3 referenced).

### Exit rates and persistence of zombie status
- Exit-rate patterns:
  - Lower incidence of zombification among private firms largely explained by substantially higher exit rates of private firms compared to listed firms (applies to both zombies and nonzombies).
  - Conjecture: ‘Weak’ private firms prone to becoming zombies may be more likely to exit the market than weak listed firms.
  - Private zombie firms tend to exit the market at a faster pace than listed zombies.
  - Caveat: private-firm sample is less representative of the whole population than listed-firm sample.
- Measurement:
  - Exit-rate = number of firms that exit Compustat (ORBIS) in each quarter (year) divided by total number of firms.
- Persistence metrics:
  - Likelihood of a zombie firm remaining zombie: around 70-80 percent for listed firms, and around 65-70 percent for private firms.
  - Probability of a nonzombie firm becoming zombie: around 2 percent for listed firms, slightly higher compared to private firms.
  - Mean duration of zombie status: most recent available year suggests roughly four years.
  - Mean duration has steadily been increasing after the GFC.

### Cross-country and industry heterogeneity
- Cross-country:
  - Considerable heterogeneity; no discernible difference between AEs and EMs (exception: private firms before 2007, but private‑firm representativeness limited in EMs).
  - Smaller listed zombie shares in Austria, Finland, Spain, Germany, The Netherlands, and France; higher shares in Canada, Republic of Korea, and Australia (2000-21 sample).
  - Countries with higher listed zombie shares also tend to have higher private zombie shares.
  - Country characteristics associated with higher zombification: looser macroprudential stance, less-prepared insolvency frameworks, weaker GDP growth, and lower interest rates.
- Industry patterns:
  - Listed firms: higher zombie shares in real estate and energy; high shares in information technology, materials, and consumer discretionary; lower shares in utilities and consumer staples.
  - Private firms: highest zombie shares in entertainment, information and communication, professional activities, and real estate.
  - Overall: nontradable services and producers of durable goods tend to have higher zombification; utilities and consumer staples show lower zombification.

### Alternative zombie definitions and robustness checks
- Alternative definitions surveyed:
  - Storz et al. (2017): negative ROA for at least two consecutive years, negative net investments, and EBITDA/financial debt below 5 percent.
  - McGowan et al. (2018): ICR below one for three consecutive years, and at least ten years old.
  - Acharya et al. (2019): three-year median ICR implied rating of BB or lower, and interest expenses ratio lower than AAA-rated peers.
  - Altman et al. (2022): three-year moving average of ICR below one, and a Z-score or Z"-score below zero.
  - TFP alternative: replace negative sales growth with TFP below the median sample in each country (TFP estimated following Olley and Pakes (1996)); reduces sample.
- Comparison outcomes:
  - Baseline zombie shares stand between the lower end and the average of alternative measures.
  - Absolute number of zombie firms not very different across definitions, excluding Storz et al. (2017) and Acharya et al. (2019).
  - Acharya et al. (2019) is an outlier, indicating subsidized-interest definitions may capture firms with low ICRs but strong fundamentals.

### Balance sheet characteristics: empirical approach
- Regression specification:
  - Yi,c,n,t = β Zombiei,c,n,t + ζc,n,t + εi,c,n,t
  - Yi,c,n,t: balance sheet indicators for firm i in country c and industry n at time t.
  - Zombiei,c,n,t: binary indicator (1 if zombie, 0 otherwise).
  - ζc,n,t: country-industry-time fixed effects (quarterly for listed firms, yearly for private firms).
- Left-hand-side variables include (exact labels preserved):
  - Log K; ∆K; ∆Intan; ∆Emp; Liq.asset; Debt; ∆Debt; Int.rate; ∆Sales; TFP; Log Assets; ICR; ROA; PDAge; Age; Loan shr.

### Balance sheet characteristics: listed firms (Table 1 coefficients and key findings)
- Selected coefficients (Zombie effect; standard errors in parentheses):
  - Log K: -0.524 ∗∗∗ (0.011); Observations 1,764,996; Adjusted R2 0.254
  - ∆K: -2.604 ∗∗∗ (0.038); Observations 1,668,411; Adjusted R2 0.173
  - ∆Intan: -3.607 ∗∗∗ (0.051); Observations 1,315,336; Adjusted R2 0.144
  - ∆Emp: -9.762 ∗∗∗ (0.104); Observations 919,854; Adjusted R2 0.068
  - Liq.asset: -19.498 ∗∗∗ (0.125); Observations 1,759,840; Adjusted R2 0.140
  - Debt: 18.244 ∗∗∗ (0.091); Observations 1,768,959; Adjusted R2 0.165
  - ∆Debt: -2.519 ∗∗∗ (0.077); Observations 1,612,920; Adjusted R2 0.037
  - Int.rate: 0.033 (0.059); Observations 1,224,423; Adjusted R2 0.127
  - ∆Sales: -3.947 ∗∗∗ (0.052); Observations 1,330,104; Adjusted R2 0.140
  - TFP: -0.146 ∗∗∗ (0.004); Observations 743,705; Adjusted R2 0.007
  - Log Assets: -0.628 ∗∗∗ (0.009); Observations 1,768,959; Adjusted R2 0.250
  - ICR: -26.531 ∗∗∗ (0.736); Observations 1,172,042; Adjusted R2 0.083
  - ROA: -9.442 ∗∗∗ (0.083); Observations 1,314,380; Adjusted R2 0.142
  - PDAge: 0.829 ∗∗∗ (0.013); Observations 1,372,046; Adjusted R2 0.292
  - Age: -1.897 ∗∗∗ (0.099); Observations 1,382,276; Adjusted R2 0.276
  - Loan shr: 1.750 ∗∗∗ (0.134); Observations 1,767,958; Adjusted R2 0.149
- Key interpretations:
  - Listed zombies invest less: tangible capital stock is 57 percent smaller than nonzombies (Log K coefficient -0.524).
  - Tangible investment growth is on average 2.6 p.p. weaker (∆K -2.604).
  - Intangible investment growth roughly 3.7 p.p. lower (∆Intan -3.607).
  - Substantially smaller net liquid asset holdings (Liq.asset -19.498) and significantly more leveraged (Debt 18.244).
  - Less productive and profitable (TFP -0.146; ROA -9.442).
  - Tend to be smaller (Log Assets -0.628) and more dependent on bank loans (Loan shr 1.750 p.p. higher).
  - Slightly younger (Age -1.897); unconditional mean age for listed zombies is 33 years.
  - Implicit interest rate (Int.rate 0.033) not statistically significant at conventional levels.

### Balance sheet characteristics: private firms (Table 2 coefficients and key findings)
- Selected coefficients (Zombie effect; standard errors in parentheses):
  - Log K: -0.238 ∗∗∗ (0.015); Observations 35,525,890; Adjusted R2 0.247
  - ∆K: -12.146 ∗∗∗ (0.205); Observations 34,537,929; Adjusted R2 0.046
  - ∆Intan: -6.617 ∗∗∗ (0.225); Observations 17,898,246; Adjusted R2 0.036
  - ∆Emp: -9.896 ∗∗∗ (0.128); Observations 21,061,384; Adjusted R2 0.021
  - Liq.asset: -18.934 ∗∗∗ (0.422); Observations 36,735,014; Adjusted R2 0.092
  - Debt: 15.833 ∗∗∗ (0.412); Observations 36,735,193; Adjusted R2 0.202
  - ∆Debt: -9.747 ∗∗∗ (0.306); Observations 25,099,230; Adjusted R2 0.022
  - Int.rate: -7.350 ∗∗∗ (0.481); Observations 26,228,449; Adjusted R2 0.077
  - ∆Sales: -17.593 ∗∗∗ (0.397); Observations 35,382,471; Adjusted R2 0.047
  - TFP: -0.224 ∗∗∗ (0.005); Observations 19,527,399; Adjusted R2 0.010
  - Log Assets: -0.176 ∗∗∗ (0.016); Observations 36,735,193; Adjusted R2 0.255
  - ICR: -84.280 ∗∗∗ (2.241); Observations 32,829,521; Adjusted R2 0.099
  - ROA: -12.947 ∗∗∗ (0.145); Observations 35,476,335; Adjusted R2 0.080
  - Age: 2.974 ∗∗∗ (0.091); Observations 36,681,987; Adjusted R2 0.135
- Key interpretations:
  - Private zombies show much larger negative differences in investment growth (∆K -12.146; ∆Intan -6.617).
  - Much lower sales growth (∆Sales -17.593) and lower productivity (TFP -0.224).
  - More leveraged (Debt 15.833) and much lower ICR (ICR -84.280).
  - Appear to face a lower implicit interest rate burden than nonzombie counterparts (Int.rate -7.350).
  - Private zombies tend to be slightly older (Age 2.974); average age of private firms is 17 years.

### Synthesis and interpretation
- Core synthesis:
  - Zombies (by construction) have low ICR, high leverage, and weak sales growth; regressions show they also have substantially weaker balance sheets and are at higher risk of default.
- Listed vs private differences:
  - Listed zombies: somewhat younger (mean age 33), similar or slightly higher implicit interest rates compared with nonzombies, challenging identification via subsidized interest rates.
  - Private zombies: larger negative investment growth effects, lower implicit interest rates relative to private nonzombies, and higher leverage; private zombie firms exit faster, lowering observed incidence.
- Empirical alignment:
  - Results broadly align with McGowan et al. (2018), Banerjee and Hofmann (2022), and Favara et al. (2022), with differences on size likely reflecting definition and country coverage differences.

### Evolution of balance sheets of zombies (section 3.4): methodology and findings
- Methodology:
  - Follow Banerjee and Hofmann (2022) to track balance sheet indicators before and after listed-firm entry into zombie status.
  - Regression: Y_{i,c,n,t+h} = β_h Newzom_{i,c,n,t} + δ_h Oldzom_{i,c,n,t} + λ_h X_{i,c,n,t−20} + ζ_{c,n,t+h} + ε_{i,c,n,t+h}.
  - Newzom = firm becomes zombie at t=0; Oldzom = was zombie in t-1.
  - Coefficient β_h captures difference between new zombies and nonzombies from t-12 to t+12 quarters.
  - Country-industry-quarter fixed effects; standard errors clustered at country-industry level.
  - To mitigate survivorship bias, analysis focuses on firms alive in t+12.
- Pre-zombification dynamics:
  - Firms that become zombies show weaker balance sheets already three years before zombification.
  - Declines in ROA and ICR and lower Tobin’s q observed several quarters prior.
  - Balance sheet contractions begin two to three years before zombification: reductions in capital stock, total assets (including net liquid assets), and employment.
  - Debt-to-asset ratio trends upward, primarily reflecting asset shedding rather than increased borrowing.
  - 12‑month PD for new zombies starts increasing in the three years preceding zombification.
  - Implicit interest rate lower for firms before entering zombie status, suggesting easy financing conditions may contribute to zombification.
- Post-zombification dynamics:
  - Tentative and gradual recovery in financial performance after entry: some improvement in ROA, Tobin’s q, PD, and debt-to-assets, supported by a large fall in debt growth.
  - Despite some recovery, zombies’ financial performance remains well below nonzombies after three years.
  - Gaps in capital stock, total assets, and employment continue to widen over time.
  - Cost of debt repricing: zombies’ cost of debt goes above that of nonzombies after entry, indicating repricing of risk by lenders.
- Additional notes:
  - Implicit interest rate lower before zombification highlights a role for easy financing and credit misallocation.
  - Results on annual data for private firms similar to listed-firm findings.
  - 95 percent confidence bands reported with SEs clustered at country-industry level.

### Extensive margin (section 4.2): exit, entry, and congestion effects
- Measurement and methods:
  - Exit dummy = one in last quarter (or year in ORBIS) firm last reports; entry dummy = one in first quarter (or year) firm first reports.
  - Exclude major M&A by dropping observations when acquisitions larger than 5 percent of total assets.
  - Industry-level dynamics estimated with Jordà (2005) local projections; horizons h = 0,...,20 quarters (h = 0,...,5 years for private firms).
  - Industry-level entry/exit rates computed as number of exits (entries) in period within industry divided by total number of firms in that industry (denominator fixed at time t).
- Firm-level exit regressions (selected coefficients; exit dummy × 100):
  - Listed sample (Table 5):
    - Nonzombie: -0.218 ∗∗∗ (0.028); Nonzombie × zombie share: 0.004 ∗∗ (0.002); Observations: 1,536,524; Adjusted R2: 0.012
  - Private sample:
    - Nonzombie: -7.992 ∗∗∗ (0.226); Nonzombie × zombie share: 0.067 ∗∗∗ (0.024); Observations: 36,735,014; Adjusted R2: 0.097
- Interpretation:
  - Nonzombies less likely than zombies to exit the market.
  - Industries with higher zombie shares tend to drive healthy firms out of the market; effects quantitatively stronger for private firms.
- Industry-level local projections (one-standard-deviation increase in asset-weighted zombie share):
  - Listed firms: industries one-standard deviation above mean experience greater exit rates of healthy firms ~0.4 p.p. over medium term; entry rates tend to be lower.
  - Private firms: industries one-standard-deviation above mean experience greater exit rates of healthy firms roughly 10 p.p., with large uncertainty.
  - Congestion effects on entry and exit are larger since the GFC (post-2008Q1).
- Mechanism:
  - Congestion effects reflect impaired creative destruction: higher zombie shares depress market prices and keep wages high, raising productivity threshold for entrants; healthy firms face greater exit pressures and reduced credit access.
  - Echoes evidence from Japan's lost decade on reduced credit supply to competing nonzombies.

### Policy mitigation: banking-sector health, macroprudential stance, and insolvency preparedness
- Empirical test: triple-interaction augmentation of exit/investment equation with I_c,t (country-level indicator) to estimate β3 = marginal change in congestion effects for countries one-standard deviation above sample in I_c,t.
- I_c,t tested one at a time:
  - Banks’ NPLs; regulatory capital buffers; macroprudential stance (sum of iMaPP indicators).
- Main mitigation results (listed-firm, quarterly):
  - Stronger banks and tighter macroprudential policies mitigate congestion effects on nonzombies’ investment and debt growth.
  - Increasing banks’ capital buffers by one-standard deviation would mitigate congestion effects on nonzombies’ investment growth by 0.2 p.p. after five years (roughly 10 percent of estimated congestion effect from Figure 5).
  - Lower NPLs associated with smaller congestion effects.
  - Tighter macroprudential policies that constrain credit supply and improve bank capital reduce congestion effects; similar patterns across macroprudential sub-indices.
- Role of insolvency frameworks:
  - Using crisis preparedness indicator (Araujo et al. (2022)), mitigation from stronger bank capital or lower NPLs occurs only in countries with well-prepared insolvency frameworks (above median).
  - In less-prepared countries, healthy firms’ investment growth does not improve when regulatory capital buffers increase or NPLs fall.
  - Conclusion: improving bank health may be necessary but not sufficient; credible insolvency and restructuring frameworks are required to realize mitigation benefits.

### Policy implications and trade-offs
- Policies to mitigate zombification congestion effects:
  - Strengthen bank supervision and regulation to reduce evergreening incentives:
    - Bank recapitalizations.
    - Increased unannounced on-site inspections.
    - Credible stress-testing exercises.
  - Implement macroprudential measures that constrain credit supply to risky borrowers and improve bank capital.
  - Prepare and reform restructuring and bankruptcy laws to facilitate corporate restructuring and increase expected recovery values for lenders.
- Trade-offs:
  - Public support (credit guarantees, forbearance) can reduce short-term insolvencies and contain aggregate demand collapses during large shocks.
  - Untargeted policy support combined with lax macroprudential/supervisory measures and deficient insolvency frameworks may delay creative destruction, harming long-term productivity growth.

*Italic: Source — wpiea2023125-print-pdf*

### Appendix A shows the industry breakdown, aggregated at the two-digit code: a large share of the

### wpiea2023125-print-pdf - Appendix A shows the industry breakdown, aggregated at the two-digit code: a large share of the

### Data sources, cleaning, and sample construction
- Industry concentration: a large share of the firms are concentrated in the industrial sector, consumer discretionary, information technology, and materials.
- Compustat (listed firms) cleaning steps:
  - Drop firms with missing assets, liabilities, debt, or the capital stock.
  - Drop firms with fewer than three years of data on the leverage ratio, the investment rate, the interest coverage ratio (ICR), and sales.
  - Restrict sample to firms whose total debt does not exceed 100 percent of their total assets.
  - Drop observations for countries with fewer than five firms for each quarter.
  - Winsorize key variables at the 2.5/97.5 percentiles at the country level.
  - Deflate nominal variables with the respective country CPI deflator.
- Compustat final sample:
  - Unbalanced panel of 42,760 nonfinancial firms over 2000q1-2021q4.
  - Resulting in 1,770,521 firm-quarter observations.
- Time coverage and trends:
  - Figure A.1 shows a steady increase in the number of firms and countries since 2000.
  - On average, twice as many listed firms in AEs than in EMs, though the gap has been shrinking.

### Private firms (ORBIS) coverage and caveats
- Motivation: ORBIS used to fill gap on private firms not covered by Compustat.
- ORBIS caveats and biases:
  - Covers a smaller set of countries relative to Compustat; data on EMs is sparse, so analysis falls mostly on AEs.
  - Some large economies are poorly represented (e.g. the US, UK, Japan, and Canada), biasing analysis toward European countries.
  - Important data lags and only annual data, limiting study of zombies to low-frequency analysis.
  - Cross-country differences in disclosure regulations introduce selection bias: many private firms in some jurisdictions are not required to disclose balance sheet information.
- ORBIS cleaning and sample:
  - Follow standard practice in the literature (Kalemli-Özcan et al., 2015; Díez et al., 2021).
  - Select only private firms by keeping firms that are ‘Unlisted’ or ‘Delisted’.
  - Avoid firm-year duplicates by prioritizing consolidated accounts.
  - Appendix B lists detailed cleaning assumptions and data definitions.
- ORBIS final sample:
  - Covers 43 countries (26 AEs and 17 EMs) in an unbalanced panel of 4,394,313 nonfinancial firms over 1997-2020.
  - Firms in EMs account for roughly 6 percent of the total sample, representing 269,667 firms.
- Industry distribution in ORBIS:
  - Almost half of the sample concentrated in manufacturing, and in wholesale and retail trade, followed by real estate and professional activities.
- Note on ICR computation:
  - Due to lack of EBIT in ORBIS, for private firms ICR is computed as EBITDA over interest expenses.

### Country-level indicators and construction
- Banking sector health:
  - Main indicator: regulatory capital buffers = banks’ actual regulatory capital to risk-weighted assets minus minimum required risk-based regulatory capital ratio.
  - Minimum required ratio taken from the World Bank’s Bank Regulation and Supervision Survey (available for 55 countries from 2001 to 2019).
  - Banks’ actual regulatory capital to risk-weighted assets from the IMF’s Financial Soundness Indicators.
  - Nonperforming loans to total gross loans (NPL) available from 2001 to 2021 used as an additional indicator.
- Macroprudential policies:
  - Use IMF’s integrated Macroprudential Policy (iMaPP) database covering 17 macroprudential policy measures for 182 jurisdictions over 1990-2020.
  - For each measure: tightening = 1, loosening = −1, no change = 0.
  - Transform indices into cumulative terms by summing up dummy indices from 1990 onward to measure stringency.
  - Proxy for overall macroprudential stance is the sum of all 17 macroprudential indicators.
- Insolvency regime quality:
  - Use crisis preparedness indicator of insolvency systems from Araujo et al. (2022).
  - Measures five dimensions: out-of-court restructuring, hybrid restructuring, reorganization, liquidation, and institutional framework.
  - Indicator is time-invariant, giving a snapshot as of 2021.
  - Original indicator covers 60 countries; matched with 43 countries in the firm-level dataset.
- Additional country-level controls (used in baseline and robustness checks):
  - Baseline: real GDP growth, three-month short-term interest rate.
  - Robustness controls: CPI inflation, real effective exchange rate (REER), current account balance as percentage of GDP, one-year government bond yield, Chinn-Ito index.

### Identification of zombie firms (definition and methodology)
- Conceptual: zombies broadly defined as firms that are risky, unproductive and unviable but avoid immediate default thanks to continued support.
- Operational definition (annual Compustat and ORBIS data; two-year persistence):
  - To be classified as a zombie for at least two consecutive years, a firm must satisfy:
    - (i) an ICR below one, and
    - (ii) a leverage ratio above the median firm in the same two-digit industry,
    - and (iii) negative real sales growth for two consecutive years.
  - ICR computed as EBIT/interest expenses for listed firms; EBITDA/interest expenses for private firms (due to lack of EBIT).
  - Leverage = total debt (short-term and long-term) divided by total assets; compared to industry median in same industry and country.
  - Two-year horizon for each of the three indicators to avoid misclassification from cyclical fluctuations.
- Exit from zombie status (persistence requirement):
  - A zombie firm exits status only when for two consecutive years it records at least one of:
    - ICR above one, or
    - leverage ratio below the industry median, or
    - positive sales growth.
  - Classification performed on annual data; merged back into quarterly dataset for higher-frequency empirical analysis.
- Rationale vs alternative definitions:
  - Differs from subsidized interest rate definitions; several arguments given why subsidized-interest definitions may be problematic (banks’ relationships, interest capitalization/evergreening, low-rate environments, data limitations on loan rates).
  - Differences from Favara et al. (2022):
    - Compare leverage to median of same industry and country (not whole sample).
    - Require exit conditions for two consecutive years and allow exit if any one of the three conditions met.
    - Use two-year horizon (Favara et al. use three), with results qualitatively similar when taking three years.
  - Alternative rolling-averages approach yields practically same results.

### Empirical findings on zombie shares (listed vs private)
- Long-run trends:
  - Share of zombie firms for both listed and private firms has been steadily increasing over the last 20 years, especially after the GFC.
  - Low interest rates, ample liquidity, and search for yield may be related to the upward trend.
- Levels and changes:
  - Listed firms:
    - Zombie firms account for over 10 percent of all listed firms in the sample in 2021.
    - Zombie share was 6 percent in 2000.
  - Private firms:
    - Share rose from 1 percent in 1997 to 5.3 percent in 2020.
  - Upticks in zombie shares tend to occur after economic downturns (early-2000s, after the GFC).
- Pandemic and recent dynamics:
  - Listed zombie share dipped from 2016 to 2019, then rose during the Covid-19 pandemic, potentially reflecting unprecedented policy support and easy financing conditions.
  - Share of private zombies has ticked up slightly since 2018; steeper increase after the GFC suggests continued rise in 2021 and 2022 (speculative in text).
- Weighting and employment:
  - Trends visible when weighting zombie shares by total debt, total assets, or employment.
  - Zombie share levels are similar when weighting by debt (and assets for private firms).
  - Zombie firms tend to account for a smaller share of employment, though employment data are sparse.
- Private vs listed comparison and hypotheses:
  - Private firms show a somewhat lower share of zombies than listed firms, despite expectations that smaller private firms might be more vulnerable.
  - Country composition differences between Compustat and ORBIS do not explain the difference.
  - Possible explanations considered:
    - Larger firms may be more likely to become zombies because they are more likely to receive government subsidies.
    - Small private firms may find it more difficult to get banks to evergreen loans and thus default instead.
    - Political connections and state ownership (SOEs) as potential drivers; ORBIS allows identification of SOEs.
  - Empirical note: zombification incidence among unlisted SOEs is very similar to other firms (referenced Figure C.3).

*Source: wpiea2023125-print-pdf - Appendix A shows the industry breakdown, aggregated at the two-digit code: a large share of the*

### Appendix C).

### Appendix C)

### Exit rates and persistence of zombie status
- Lower incidence of zombification among private firms is most likely explained by substantially higher exit rates of private firms compared to listed firms (applies to both zombies and nonzombies).
- Conjecture:
  - ‘Weak’ private firms prone to becoming zombies may be more likely to exit the market than weak listed firms.
  - Private zombie firms tend to exit the market at a faster pace than listed zombies.
- Caveat: the sample of private firms is less representative of the whole population than the sample of listed firms.
- Exit-rate measurement: number of firms that exit the Compustat (ORBIS) dataset in each quarter (year) divided by the total number of firms.
- Persistence metrics:
  - Likelihood of a zombie firm remaining zombie: hovering around 70-80 percent for listed firms, and around 65-70 percent for private firms.
  - Probability of a nonzombie firm becoming zombie: around 2 percent for listed firms, slightly higher compared to private firms.
  - Mean duration of zombie status: most recent available year suggests roughly four years.
- Time patterns: mean duration of zombie status has steadily been increasing in the aftermath of the GFC.

### Cross-country and industry heterogeneity
- Zombie shares exhibit considerable cross-country heterogeneity, without a discernible difference between AEs and EMs (exception: private firms before 2007, but private-firm representativeness is limited in EMs).
- Country patterns for listed firms (2000-21 sample): smaller zombie shares in Austria, Finland, Spain, Germany, The Netherlands, and France; higher shares in Canada, Republic of Korea, and Australia.
- Country correlation: countries with a higher share of listed zombies also seem to have a higher share of private zombies.
- Country characteristics associated with higher zombification: looser macroprudential stance, less-prepared corporate insolvency frameworks, weaker GDP growth, and lower interest rates.
- Industry patterns:
  - Listed firms: higher zombie shares in real estate and in energy; high zombie shares in information technology, materials, and consumer discretionary; lower shares in utilities and consumer staples.
  - Private firms: highest zombie shares in entertainment, information and communication, professional activities, and real estate.
  - Overall: industries in nontradable services and those producing durable goods tend to be associated with higher levels of zombification; utilities and consumer staples show lower zombification.

### Alternative zombie definitions and robustness checks
- Storz et al. (2017): zombies = negative ROA for at least two consecutive years, negative net investments, and EBITDA/financial debt below 5 percent.
- McGowan et al. (2018): zombies = ICR below one for three consecutive years, and at least ten years old.
- Acharya et al. (2019): three-year median ICR implied rating of BB or lower, and interest expenses ratio lower than AAA-rated peers; ratings inferred from ICR when ratings unavailable.
- Altman et al. (2022): three-year moving average of ICR below one, and a Z-score or Z"-score below zero.
- Alternative using TFP: replace negative sales growth with TFP below the median sample in each country (TFP estimated following Olley and Pakes (1996)); reduces sample due to data constraints and estimation uncertainty.
- Comparison outcomes:
  - Baseline zombie shares stand between the lower end and the average of alternative measures.
  - Absolute number of zombie firms not very different across definitions, excluding Storz et al. (2017) and Acharya et al. (2019).
  - Acharya et al. (2019) is an outlier, indicating that defining zombies via subsidized interest rates can capture firms with low ICRs but strong fundamentals.

### Balance sheet characteristics: empirical approach
- Regression specification:
  - Yi,c,n,t = β Zombiei,c,n,t + ζc,n,t + εi,c,n,t
  - Yi,c,n,t: balance sheet indicators for firm i in country c and industry n at time t.
  - Zombiei,c,n,t: binary indicator (1 if zombie, 0 otherwise).
  - ζc,n,t: country-industry-time fixed effects (quarterly for listed firms, yearly for private firms).
- Left-hand-side variables include:
  - Log K (log of net capital stock)
  - ∆K (tangible investment growth: log percentage change in net capital stock)
  - ∆Intan (intangible investment growth: R&D costs + 30 percent of SG&A)
  - ∆Emp (employment growth: year-on-year log percentage change in total employees)
  - Liq.asset (liquid assets ratio: current assets net of current liabilities to total assets)
  - Debt (leverage ratio: total debt to total assets)
  - ∆Debt (change in total debt)
  - Int.rate (implicit interest rate: four-quarter rolling sum of interest expenses divided by total debt)
  - ∆Sales (sales growth: year-on-year change in net sales)
  - TFP (estimated total factor productivity)
  - Log Assets (firm size)
  - ICR (interest coverage ratio)
  - ROA (return on assets: four-quarter rolling sum of EBIT divided by total assets)
  - PDAge (probability of default in 12 months; modified Merton distance-to-default; not available for private firms)
  - Age (years since foundation)
  - Loan shr (share of bank loans in total debt; not available for private firms)

### Balance sheet characteristics: listed firms (Table 1 coefficients)
Panel A (listed firms)
- Log K: Zombie = -0.524 ∗∗∗ (standard error 0.011); Observations 1,764,996; Adjusted R2 0.254
- ∆K: Zombie = -2.604 ∗∗∗ (0.038); Observations 1,668,411; Adjusted R2 0.173
- ∆Intan: Zombie = -3.607 ∗∗∗ (0.051); Observations 1,315,336; Adjusted R2 0.144
- ∆Emp: Zombie = -9.762 ∗∗∗ (0.104); Observations 919,854; Adjusted R2 0.068
- Liq.asset: Zombie = -19.498 ∗∗∗ (0.125); Observations 1,759,840; Adjusted R2 0.140
- Debt: Zombie = 18.244 ∗∗∗ (0.091); Observations 1,768,959; Adjusted R2 0.165
- ∆Debt: Zombie = -2.519 ∗∗∗ (0.077); Observations 1,612,920; Adjusted R2 0.037
- Int.rate: Zombie = 0.033 (0.059); Observations 1,224,423; Adjusted R2 0.127

Panel B (listed firms)
- ∆Sales: Zombie = -3.947 ∗∗∗ (0.052); Observations 1,330,104; Adjusted R2 0.140
- TFP: Zombie = -0.146 ∗∗∗ (0.004); Observations 743,705; Adjusted R2 0.007
- Log Assets: Zombie = -0.628 ∗∗∗ (0.009); Observations 1,768,959; Adjusted R2 0.250
- ICR: Zombie = -26.531 ∗∗∗ (0.736); Observations 1,172,042; Adjusted R2 0.083
- ROA: Zombie = -9.442 ∗∗∗ (0.083); Observations 1,314,380; Adjusted R2 0.142
- PDAge: Zombie = 0.829 ∗∗∗ (0.013); Observations 1,372,046; Adjusted R2 0.292
- Age: Zombie = -1.897 ∗∗∗ (0.099); Observations 1,382,276; Adjusted R2 0.276
- Loan shr: Zombie = 1.750 ∗∗∗ (0.134); Observations 1,767,958; Adjusted R2 0.149

Notes: all regressions include country-industry-quarter fixed effects. Standard errors in parentheses clustered by country-industry-quarter. Asterisks denote statistical significance at the 10%, 5%, and 1% levels.

Key listed-firm findings:
- Listed zombie firms invest less than nonzombies: tangible capital stock is 57 percent smaller than nonzombies (Log K coefficient -0.524).
- Tangible investment growth is on average 2.6 p.p. weaker for zombie firms (∆K coefficient -2.604).
- Growth in intangibles is roughly 3.7 p.p. lower for zombies (∆Intan coefficient -3.607).
- Zombies have substantially smaller net liquid asset holdings (Liq.asset -19.498) and are significantly more leveraged (Debt 18.244).
- Zombies are less productive and profitable (TFP -0.146; ROA -9.442).
- Zombies tend to be smaller (Log Assets -0.628) and more dependent on bank loans (Loan shr 1.750 p.p. higher).
- Listed zombies tend to be slightly younger (Age -1.897), unconditional mean age for listed zombies is 33 years.
- Implicit interest rate result: Int.rate 0.033 (not statistically significant at conventional levels).

### Balance sheet characteristics: private firms (Table 2 coefficients)
Panel A (private firms)
- Log K: Zombie = -0.238 ∗∗∗ (0.015); Observations 35,525,890; Adjusted R2 0.247
- ∆K: Zombie = -12.146 ∗∗∗ (0.205); Observations 34,537,929; Adjusted R2 0.046
- ∆Intan: Zombie = -6.617 ∗∗∗ (0.225); Observations 17,898,246; Adjusted R2 0.036
- ∆Emp: Zombie = -9.896 ∗∗∗ (0.128); Observations 21,061,384; Adjusted R2 0.021
- Liq.asset: Zombie = -18.934 ∗∗∗ (0.422); Observations 36,735,014; Adjusted R2 0.092
- Debt: Zombie = 15.833 ∗∗∗ (0.412); Observations 36,735,193; Adjusted R2 0.202
- ∆Debt: Zombie = -9.747 ∗∗∗ (0.306); Observations 25,099,230; Adjusted R2 0.022

Panel B (private firms)
- Int.rate: Zombie = -7.350 ∗∗∗ (0.481); Observations 26,228,449; Adjusted R2 0.077
- ∆Sales: Zombie = -17.593 ∗∗∗ (0.397); Observations 35,382,471; Adjusted R2 0.047
- TFP: Zombie = -0.224 ∗∗∗ (0.005); Observations 19,527,399; Adjusted R2 0.010
- Log Assets: Zombie = -0.176 ∗∗∗ (0.016); Observations 36,735,193; Adjusted R2 0.255
- ICR: Zombie = -84.280 ∗∗∗ (2.241); Observations 32,829,521; Adjusted R2 0.099
- ROA: Zombie = -12.947 ∗∗∗ (0.145); Observations 35,476,335; Adjusted R2 0.080
- Age: Zombie = 2.974 ∗∗∗ (0.091); Observations 36,681,987; Adjusted R2 0.135

Notes: all regressions include country-industry-year fixed effects. Standard errors in parentheses clustered by country-industry-year. Asterisks denote statistical significance at the 10%, 5%, and 1% levels.

Key private-firm findings:
- Private zombies show larger negative differences in investment growth relative to private nonzombie peers (∆K -12.146; ∆Intan -6.617).
- Private zombies have substantially lower sales growth (∆Sales -17.593) and lower productivity (TFP -0.224).
- Private zombies are more leveraged (Debt 15.833) and have much lower ICR (ICR -84.280).
- Private zombie firms seem to experience a lower implicit interest rate burden than their nonzombie counterparts (Int.rate -7.350).
- Private zombies tend to be slightly older relative to private nonzombies (Age 2.974); average age of private firms is 17 years, making age comparisons with listed firms challenging.

### Synthesis and interpretation
- Zombies—by construction—exhibit low ICR, high leverage, and weak sales growth; regression results show they also have substantially weaker balance sheets across many dimensions and are at higher risk of default.
- Differences between listed and private zombies:
  - Listed zombies: somewhat younger (mean age 33), similar or slightly higher implicit interest rates compared with nonzombies, challenging the identification of zombies via subsidized interest rates.
  - Private zombies: larger negative investment growth effects, lower implicit interest rates relative to private nonzombies, and higher leverage; private zombie firms exit faster, contributing to lower observed zombification incidence among private firms.
- Empirical alignment: results broadly align with McGowan et al. (2018), Banerjee and Hofmann (2022), and Favara et al. (2022), with differences on the size dimension relative to McGowan et al. (2018) likely reflecting differences in definition and country coverage.

*Appendix C).*

### 3.4  Evolution of balance sheets of zombies

### 3.4  Evolution of balance sheets of zombies

### Methodology and empirical specification
- Follow Banerjee and Hofmann (2022) to track balance sheet indicators before and after a listed firm enters zombie status relative to healthy firms.
- Estimated regression:
  - Y_{i,c,n,t+h} = β_h Newzom_{i,c,n,t} + δ_h Oldzom_{i,c,n,t} + λ_h X_{i,c,n,t−20} + ζ_{c,n,t+h} + ε_{i,c,n,t+h}
  - Newzom dummy = 1 if a firm becomes zombie at t=0; Oldzom = 1 if firm was already classified as zombie in t-1.
  - Coefficient of interest β_h captures difference between new zombies and nonzombies in selected balance sheet variables from horizons t-12 to t+12 quarters.
  - Country-industry-quarter fixed effects (ζ_{c,n,t+h}) compare new zombies to healthy firms in the same country-industry-quarter.
  - Standard errors clustered at the country-industry level.
  - To mitigate survivorship bias from firms exiting the sample, analysis focuses only on firms still alive in t+12.

### Pre-zombification balance sheet dynamics (key findings)
- Firms that become zombies at t=0 exhibit weaker balance sheets already three years before zombification.
- Performance deterioration occurs several quarters before zombification, evidenced by:
  - Declines in ROA and ICR.
  - Lower growth opportunities measured by Tobin’s q.
- Balance sheet contractions begin two to three years before zombification:
  - Reductions in capital stock, total assets (including net liquid assets), and employment.
  - Debt-to-asset ratio trends upward, primarily reflecting asset shedding rather than increased borrowing.
- Market pricing of risk:
  - The 12‑month probability of default (PD) for new zombies starts to increase in the three years preceding zombification.
  - Implicit interest rate (proxy for cost of debt) is estimated to be lower for firms before entering zombie status, suggesting easy financing conditions may contribute to zombification.

### Post-zombification dynamics (key findings)
- Evidence of only tentative and very gradual recovery in financial performance after entering zombie status:
  - Some improvement in ROA, Tobin’s q, probability of default, and debt-to-assets relative to zombies, supported by a large fall in debt growth.
  - Despite some recovery, zombies’ financial performance remains well-below that of nonzombies after three years.
- Persistent and widening gaps:
  - The gap between zombies and nonzombies in capital stock, total assets, and employment continues to widen over time.
- Cost of debt repricing:
  - Zombies’ cost of debt goes above that of nonzombies after entry into zombie status, indicating repricing of risk by lenders.

### Additional notes from the analysis
- The implicit interest rate being lower before zombification highlights a role for easy financing conditions and potential credit misallocation in zombification.
- Results on annual data for private firms are similar to the listed-firm findings.
- Standard error inference: 95 percent confidence bands shown in the analysis (Figure 4) with standard errors clustered at the country-industry level.

*Italic: Source — wpiea2023125-print-pdf, section 3.4*

### 4.2  Extensive margin

### 4.2  Extensive margin

### Measurement and methods
- Exit and entry dummies:
  - Exit dummy = one in the last quarter (or year in ORBIS) in which the firm last reports its balance sheet; entry dummy = one in the first quarter (or year in ORBIS) in which the firm first reports its balance sheet.
  - Caveat: an exit in Compustat or ORBIS may not necessarily reflect bankruptcy or closure (listed firms: delisting; private firms: possibly going public).
  - Exclude most mergers and acquisitions by dropping observations when acquisitions are larger than 5 percent of total assets.
- Estimation strategies:
  - Augment Equation 4 with a post-GFC dummy (2008Q1 onwards) and its interaction with NZ_i,c,n,t × ZShare_c,n,t.
  - For industry-level dynamics, use Jordà (2005) local projections, running separate regressions for each horizon h = 0, 1, ..., 20 quarters (h = 0, 1, ..., 5 years for private firms).
  - Industry-level entry/exit rates computed as number of exits (entries) in each period within each industry divided by total number of firms in that industry; denominator fixed at time t to avoid spurious mechanical relationships.

### Extensive-margin findings: exit and entry dynamics
- Firm-level exit regressions (Equation 4 with exit dummy × 100):
  - Table 5 (congestion effects: probability of exit of healthy firms):
    - Listed sample:
      - Nonzombie coefficient: -0.218 ∗∗∗ (standard error 0.028)
      - Nonzombie × zombie share: 0.004 ∗∗ (standard error 0.002)
      - Controls included; Country×Industry×Time fixed effects included.
      - Observations: 1,536,524
      - Adjusted R2: 0.012
    - Private sample:
      - Nonzombie coefficient: -7.992 ∗∗∗ (standard error 0.226)
      - Nonzombie × zombie share: 0.067 ∗∗∗ (standard error 0.024)
      - Controls included; Country×Industry×Time fixed effects included.
      - Observations: 36,735,014
      - Adjusted R2: 0.097
  - Interpretation:
    - Nonzombies are less likely than zombie firms to exit the market (first-row results).
    - Industries with a higher share of zombie firms tend to drive healthy firms out of the market relative to industries with a lower share of zombie firms.
    - Effects are quantitatively stronger for private firms, which exhibit significantly higher exit rates than listed firms.
- Industry-level local-projection results (cumulative impulse responses to a one-standard-deviation increase in asset-weighted zombie share):
  - Listed firms:
    - Industries one-standard deviation above the zombie share mean experience greater exit rates of healthy firms in the order of 0.4 p.p. over the medium term.
    - Entry rates tend to be lower in industries with a greater share of zombie firms.
  - Private firms:
    - Industries one-standard deviation above the zombie share mean experience greater exit rates of healthy firms roughly 10 p.p., although these estimates are surrounded by large uncertainty.
  - The congestion effects on entry and exit are larger since the GFC (post-2008Q1).

### Mechanism and interpretation
- Congestion effects reflect "creative destruction" impairment:
  - Higher zombie shares depress market prices and keep wages high, raising the productivity threshold for new entrants.
  - Healthy firms face greater exit pressures and reduced access to credit.
- Reduced credit supply to nonzombies that compete with zombie firms echoes evidence from Japan's lost decade.

### Policy mitigation: banking-sector health, macroprudential stance, and insolvency preparedness
- Tested mitigants via triple-interaction augmentation of Equation (4):
  - Y_i,c,n,t = β1 NZ_i,c,n,t + β2 NZ_i,c,n,t × ZShare_c,n,t + β3 NZ_i,c,n,t × ZShare_c,n,t × I_c,t + λX_i,c,n,t + α_i + ζ_c,n,t + ε_i,c,n,t.
  - I_c,t = country-level indicators tested one at a time:
    - Banks’ non-performing loans (NPLs).
    - Regulatory capital buffers = actual regulatory capital to risk-weighted assets minus minimum required risk-based regulatory capital ratio.
    - Macroprudential stance: sum of 17 macroprudential indicators from IMF’s iMaPP database, transformed into cumulative terms (start of dataset 1990).
  - Main coefficient of interest β3 indicates marginal change in congestion effects for countries standing one-standard deviation above the sample in the indicator.
- Main mitigation results (listed-firm sample, quarterly frequency):
  - Stronger banks and tighter macroprudential policies mitigate congestion effects on nonzombies’ investment and debt growth.
  - Increasing banks’ capital buffers by one-standard deviation would mitigate the congestion effects on nonzombies’ investment growth by 0.2 p.p. after five years, which is roughly 10 percent of the estimated congestion effect from Figure 5.
  - Similar mitigation when exploring cross-country differences in banks’ NPLs (lower NPLs associated with smaller congestion effects).
  - Tighter macroprudential policies that focus on constraints on credit supply and improving bank capital reduce congestion effects; similar patterns across macroprudential sub-indices (demand, loan-targeted, supply-loans, supply-general, supply-capital, liquidity, SIFI, other).
- Role of insolvency frameworks:
  - Used crisis preparedness indicator from Araujo et al. (2022) to classify countries as well-prepared (above median) or less-prepared (below median) insolvency frameworks.
  - Strengthening bank capital buffers (or decreasing NPLs) mitigates congestion effects only in countries with well-prepared insolvency frameworks:
    - In countries with less-prepared frameworks, healthy firms’ investment growth does not improve when regulatory capital buffers increase or NPLs fall.
    - Differential effects show that improvements in bank health translate into stronger nonzombie outcomes primarily when insolvency regimes facilitate credible restructuring and recovery.
  - Conclusion: improving bank health may be necessary but not sufficient; credible, well-prepared insolvency and restructuring frameworks are required to realize mitigation benefits.

### Policy implications and trade-offs
- Policies that can help mitigate zombification congestion effects:
  - Strengthen bank supervision and regulation to reduce evergreening incentives:
    - Bank recapitalizations.
    - Increased number of unannounced on-site inspections.
    - Credible stress-testing exercises.
  - Implement macroprudential measures that constrain credit supply to risky borrowers and improve bank capital.
  - Prepare and reform restructuring and bankruptcy laws to facilitate corporate restructuring and to increase expected recovery values for lenders.
- Trade-offs for policymakers:
  - Public support (credit guarantees, forbearance) can reduce short-term insolvencies and contain aggregate demand collapses during large shocks.
  - Untargeted policy support combined with lax macroprudential/supervisory measures and deficient insolvency frameworks may delay necessary creative destruction, harming long-term productivity growth.

*Source: wpiea2023125-print-pdf - 4.2  Extensive margin*

### Appendix A: Sample selection and variable definitions – Com-

### Appendix A: Sample selection and variable definitions – Compustat

### Sample overview and cleaning adjustments
- Coverage: quarterly data on nonfinancial listed corporations for 63 countries, 32 EMs and 31 AEs, from S&P Compustat North America and Compustat Global.
- Final sample: an unbalanced panel of 42,760 nonfinancial firms over 2000q1-2021q4, resulting in 1,770,521 firm-quarter observations.
- Exclusions: financial firms (GICS codes ranging from 4010 to 4030).
- Adjustments and cleaning steps:
  - we convert non-USD to USD using: (i) end-of-period exchange rates for stock balance sheet data; and (ii) quarterly average exchange rates for income statement and cash flow data, and for financial market data
  - drop observations for missing assets and liabilities
  - replace negative values for assets and liabilities with zeros
  - drop observations if capital stock or total debt are missing
  - drop observations when acquisitions are larger than 5 percent of total assets to exclude potential mergers and acquisitions
  - drop firms with total debt larger than 100 percent of total assets
  - drop firms with fewer than three years of data on the leverage ratio, the capital stock, the ICR, and sales
  - drop observations for countries with fewer than five firms for each quarter
  - winsorize key variables at the 2.5/97.5 percentiles at the country level
  - drop countries with fewer than eight years of data on the leverage ratio, the capital stock, and the ICR from 2000 onwards
  - compute zombie firms for industries which have at least three firms per country for each quarter
  - drop countries with fewer than 12 years of data on zombie shares
  - we take four-quarter rolling sums of flow variables—EBIT, sales, interest expenses—before computing ratios when the denominator is a stock variable: e.g. ROA, and the implicit interest rate
  - deflate nominal variables with the respective country CPI deflator

### Variable definitions (Table A.1)
- Net capital stock — PPENTQ — Compustat
- R&D — XRDQ — Compustat
- SG&A — XSGAQ - XRDQ - RDIPQ — Compustat
- Intangible investment — R&D + 0.3×SG&A — Compustat
- Employment — EMP — Compustat
- Total debt (book value) — DLCQ + DLTTQ — Compustat
- Long-term debt (book value) — DLTTQ — Compustat
- Short-term debt (book value) — DLCQ — Compustat
- Total assets (book value) — ATQ — Compustat
- Current assets — ACTQ — Compustat
- Current liabilities — LCTQ — Compustat
- Net current assets — ACTQ - LCTQ — Compustat
- Cash + short-term investments — CHEQ — Compustat
- Net income — NIQ — Compustat
- Interest payments — XINTQ — Compustat
- Depreciation & amortization — DPQ — Compustat
- Stock prices — PRCQQ — Compustat
- EBITDA — SALEQ - COSGQ - XSGAQ — Compustat
- EBIT — SALEQ - COSGQ - XSGAQ - DPQ — Compustat
- Debt ratio — (DLCQ + DLTTQ) / ATQ — Compustat
- ICR — EBIT/XINTQ — Compustat
- ROA — EBIT/ATQ — Compustat
- Equity (book value) — SEQQ + TXDITCQ - Preferred stock — Compustat
- Tobin’s Q — (ATQ + PRCCQ×CSHOQ - Equity) / ATQ — Compustat
- Implicit interest rate — XINTQ / (DLCQ + DLTTQ) — Compustat
- Sales — SALEQ — Compustat
- Acquisitions — AQCY — Compustat
  - a. Transformation from year-to-date to quarterly.
- Loan share — (NP + DLTO)/(DLCQ + DLTTQ) — Compustat
  - b. Annual data interpolated to quarterly.
- Age — Foundation year — Capital IQ
- PD in 12 months — PRCQQNUS-CRI

### Industry composition (Table A.2)
- Energy101076,379   4.49.5
- Materials1510242,442  14.08.4
- Industrials2010-2030   403,778  23.26.5
- Consumer discretionary   2510-2550   327,083  18.87.9
- Consumer staples3010-3030   149,958   8.66.1
- Health care3510-3520   127,056   7.36.4
- Information technology    4510-4530   267,859  15.48.4
- Communication services   5010-5020   74,474   4.37.8
- Utilities551045,141   2.63.7
- Real estate601023,261   1.310.6

### Descriptive statistics: listed firms (Table A.3)
- log real capital stock1,757,775   3.60    2.490.29    3.69   6.75
- ∆ logreal capital stock1,664,330   0.55   11.55   -8.77   -0.38   10.39
- ∆ logintangible investment  1,313,860   3.15   15.34   -9.55   0.70   21.14
- log employment1,074,675   -0.04    2.05    -2.62   -0.07   2.61
- ∆ log employment922,093    1.39   17.18   -14.84   0.91   18.44
- Debt/assets1,761,732  23.65   18.76    1.43   20.71  49.79
- ∆ logreal debt1,608,888   -0.03   26.75   -22.62   -0.68   23.81
- Net current assets/assets   1,752,643  16.50   26.10   -10.86  15.81  48.27
- ∆ logreal sales1,328,797   2.07   13.09   -9.93   0.66   16.36
- TFP741,557    0.00    0.49    -0.38   0.01   0.41
- Log real assets1,761,732   5.21    2.102.66    5.12   7.94
- Implicit interest rate1,222,493  10.28   22.82    1.78    5.66   17.35
- ICR1,170,793  18.08  105.04   -7.09   2.97   46.34
- ROA1,312,600   2.47   21.59   -8.34   3.85   16.66
- PD in 12 months1,370,825   0.54    1.010.01    0.18   1.40
- Age1,381,173  38.25   33.61    8.00   27.00  86.00
- Loan share1,760,729  39.50   37.24    0.00   30.99  97.89
- Tobin’s q302,907    2.41   29.13    0.82    1.41   3.99

Sources: Compustat, Capital IQ, National University of Singapore’s Credit Research Initiative (CRI), and authors’ calculations.
Notes: The log is applied to nominal variables expressed in USD millions. All ratios and growth rates are expressed in percent.

---

### Appendix B: Sample selection and variable definitions – ORBIS

#### Sample overview and cleaning adjustments
- Coverage: nonfinancial private firms from Bureau van Dijk Orbis for 43 countries (29 AEs and 17 EMs).
- Final dataset: an unbalanced panel of 4,394,313 nonfinancial firms over 1997-2020, resulting into 36,774,234 firm-year observations.
- Exclusions: financial firms per NACE2 codes 64, 65, and 66; keep only private firms via ‘main_exchange’ equal to ‘Unlisted’ or ‘Delisted’.
- Consolidation rule: prioritize consolidated accounts; use unconsolidated only when firms consistently report unconsolidated accounts.
- Additional adjustments:
  - drop observations where total assets, sales, operating revenue turnover, capital expenditures, or total debt are missing
  - drop observations for negative assets, total fixed assets, and for sales
  - drop firms with total debt to assets above 100 percent
  - winsorize the data at the 2.5/97.5 percentiles at the country level
  - drop firms with fewer than three years of data
  - deflate nominal variables with the respective CPI deflator

#### Variable definitions (Table B.1)
- Net capital stock — tangible_fixed_assets
- Intangible investment — R&D + 0.3×SG&A
- Employment — number_of_employees
- Total debt (book value) — long_term_debt + loans
- Long-term debt (book value) — long_term_debt
- Short-term debt (book value) — loans
- Total assets (book value) — total_assets
- Current assets — current_assets
- Current liabilities — current_liabilities
- Net current assets — total_assets - current_liabilities
- Cash + short-term investments — cash_and_cash_equivalent
- Interest payments — interest_paid
- Depreciation & amortization — depreciation_and_amortization
- EBITDA — ebitda
- Debt ratio — Total debt / total_assets
- ICR — EBITDA/interest_paid
- ROA — EBITDA/total_assets
- Implicit interest rate — interest_paid / Total debt
- Sales — sales
- Asset turnover ratio — sales / total_assets
- Age — age
- State Ownership — state_shr

#### Industry composition (Table B.2)
- Agriculture01-03856,836   2.33.3
- Mining05-09119,173   0.33.7
- Manufacturing10-336,612,804  18.03.2
- Utilities35280,402   0.81.5
- Water supply36-39241,053   0.72.1
- Construction41-435,117,241  13.93.7
- Wholesale retail45-479,088,183  24.74.4
- Transportation49-531,658,854   4.52.7
- Accommodation & food55-561,904,805   5.24.2
- Information & comm.58-631,237,219   3.44.8
- Real estate683,327,063   9.04.6
- Professional activities69-752,851,198   7.84.7
- Administrative activities77-821,369,025   3.74.0
- Education85313,031   0.94.7
- Social work86-88826,733   2.22.7
- Entertainment90-93451,296   1.25.3
- Other services94-96512,431   1.44.3
- Miscellaneous84, 97-996,8870.05.5

#### Descriptive statistics: private firms (Table B.3)
- log real capital stock35,564,529  11.70   2.408.69   11.66   14.73
- ∆ logreal capital stock34,576,298   -0.56   49.22   -48.00   -6.93   55.02
- ∆ logintangible investment  17,908,235   -8.72   68.34   -72.49   -8.49   52.62
- log employment23,867,511   2.02    1.420.00    1.79    3.87
- ∆ logemployment21,093,232   1.18   27.38   -28.77   0.00   30.23
- Debt/assets36,774,234  20.56   23.71    0.00   11.50   57.66
- ∆ logreal debt25,127,763   -2.02   80.86   -76.69   -5.22   83.04
- Net current assets/assets   36,774,055  14.44   38.38   -27.18  14.96   59.32
- ∆ logreal sales35,421,435   3.59   51.39   -40.22   0.49   42.09
- TFP19,527,399   0.00    0.43    -0.36   0.01    0.38
- Log real assets36,774,234  13.73   1.78    11.63   13.59   16.01
- Implicit interest rate26,258,603  25.58  165.77   1.32    5.76   32.37
- ICR32,864,287  79.10  587.41   -3.39   6.38   114.46
- ROA35,514,808   8.90   15.96   -3.82   7.76   25.91
- Age36,720,924  16.60   13.41    4.00   13.00   33.00

Sources: ORBIS, and authors’ calculations.
Notes: The log is applied to nominal variables expressed in USD. All ratios and growth rates are expressed in percent.

*Appendix A and Appendix B: sample construction, variable definitions, industry composition, and descriptive statistics as reported in the source PDF.*

### Appendix C: Additional figures

### Appendix C: Additional figures (Summary)

### Prevalence of zombie firms: listed vs private; SOEs; duration
- Figure C.1: Shows share of zombie firms for listed and private firms, with measures:
  - Unweighted percentage share (blue bars).
  - Debt-weighted, Asset-weighted, Employment-weighted lines.
- Figure C.2 (Compustat sample): Compares percentage share of listed zombie firms in:
  - Full sample
  - Restricted sample (restricted to the country sample of ORBIS).
- Figure C.3: Compares private zombie shares for:
  - Non−SOEs (blue bars)
  - SOEs (khaki bars)
- Figure C.4: Mean duration of zombie status:
  - Reports mean number of years a firm remains zombie for Listed and Private firms.

### Cross-country and cross-industry patterns
- Figure C.5: Share of zombie firms by country group:
  - Advanced Economies (AE) and Emerging Markets (EM) for Listed and Private firms.
- Figure C.6: Average share of zombie firms by country: listed firms
  - Notes: Average zombie shares over the 2000-2021 period for a long country list.
- Figure C.7: Average share of zombie firms by country: private firms
  - Notes: Average zombie shares over the 1997-2020 period for a long country list.
- Figure C.8: Correlation between listed and private country zombie shares:
  - Each red circle = country-year pair; solid blue line = linear fitted values.

### Country panels and maps
- Figures C.9–C.12: Panels of country time series for listed and private zombie shares across many countries (1997–2022 and 1995–2021 horizons shown).
- Figure C.13: Map — Share of listed zombie firms in 2021:
  - Dark (light) blue = first (second) quartiles of country zombie shares in 2021.
  - Orange (red) = third (fourth) quartiles.
- Figure C.14: Map — Share of private zombie firms in 2020:
  - Same quartile color scheme as Figure C.13.

### Dispersion across countries and industries; industry-level averages
- Figures C.15 (listed) and C.16 (private): Dispersion in zombie shares across countries and industries:
  - Solid blue = median zombie share.
  - Light (dark) grey = 10th/90th (25th/75th) percentiles.
- Figures C.17 (listed) and C.18 (private): Average share of zombie firms by industry:
  - Listed: two-digit and four-digit GICS industry breakdowns, values on 0–10 (left) and 0–10 (right) scales shown.
  - Private: one-digit NACE2 industry level, values on 0–5 scale shown.

### Robustness to alternative definitions
- Figures C.19 (listed) and C.20 (private):
  - Left panels: Percentage share of zombie firms across definitions (labels: Base, Base: TFP, MAM (2018), AEEH (2019), ADW (2022), SKSW (2017) for listed; Base, Base: TFP, MAM (2018), AEEH (2019), SKSW (2017) for private).
  - Right panels: Total number of zombie firms in the sample (listed scale up to 8,000; private scale up to 300,000).

### Congestion effects — dynamics and heterogeneity
- Figures C.21–C.27: Impulse response analyses (cumulative over 20 quarters) of nonzombies’ exit rates, entry rates, investment growth, and debt growth to shocks in industry asset-weighted zombie share and to changes in bank health indicators or macroprudential measures. Key descriptive notes reproduced:
  - Figure C.21: Responses for Pre−GFC vs Post−GFC; 90 percent confidence bands; differences with 68/90 percent bands.
  - Figure C.22: Responses for countries with Loose iMaPP vs Tight iMaPP (iMaPP split at sample median); 90 percent bands and difference confidence intervals.
  - Figures C.23–C.27: Responses split by well-prepared vs less-prepared insolvency frameworks (Araujo et al. (2022) crisis preparedness indicator, split at median). Panels cover:
    - Bank regulatory capital buffers and NPLs on nonzombies’ debt (Figure C.23).
    - Tightening in macroprudential measures (iMaPP, Supply − loans, Supply − capital, Liquidity, Demand, Loan−targeted, Supply − general, SIFI, Other) on nonzombies’ investment (Figures C.24–C.25) and debt (Figures C.26–C.27).
  - For all panels: solid blue (red) lines = average responses for well- (less-)prepared insolvency frameworks with 90 percent confidence bands; right panels show 68/90 percent confidence bands for differences.

---

### Appendix D: Additional tables (Key regression coefficients and notes)

### Table D.1: Congestion effects on investment growth for alternative zombie definitions
- Dependent variable: log percentage change in real tangible capital stock (∆K).
- Columns and primary estimates (coefficients with standard errors in parentheses; significance stars as in source):
  1. Base: Nonzombie = 1.690*** (0.056)
     - Nonzombie×zombie share = -0.015*** (0.004)
     - Observations = 1,660,513
     - Adjusted R2 = 0.205
  2. Base: TFP: Nonzombie = 1.170*** (0.059)
     - Nonzombie×zombie share = -0.019*** (0.005)
     - Observations = 1,195,607
     - Adjusted R2 = 0.178
  3. MAM (2018): Nonzombie = 1.323*** (0.059)
     - Nonzombie×zombie share = -0.021*** (0.005)
     - Observations = 1,320,174
     - Adjusted R2 = 0.215
  4. AEEH (2019): Nonzombie = 0.183** (0.073)
     - Nonzombie×zombie share = -0.003 (0.002)
     - Observations = 1,644,995
     - Adjusted R2 = 0.205
  5. ADW (2022): Nonzombie = 0.126 (0.178)
     - Nonzombie×zombie share = -0.094** (0.039)
     - Observations = 1,631,468
     - Adjusted R2 = 0.205
  6. SKSW (2017): Nonzombie = 1.232*** (0.080)
     - Nonzombie×zombie share = -0.053*** (0.008)
     - Observations = 1,652,015
     - Adjusted R2 = 0.206
- Notes: Controls included; Firm FE; Country×Industry×Time FE; standard errors clustered by country-industry-time.

### Table D.2: Congestion effects from zombies and distressed nonzombies to nonzombies
- Dependent variables across columns: ∆K, ∆Intan, ∆Emp, ∆Sales, TFP, Int.rate, ∆Debt.
- Key coefficients:
  - Nonzombie:
    - ∆K = 1.679*** (0.061)
    - ∆Intan = 2.950*** (0.090)
    - ∆Emp = 6.888*** (0.147)
    - ∆Sales = 2.809*** (0.092)
    - TFP = 0.231*** (0.007)
    - Int.rate = -0.632*** (0.107)
    - ∆Debt = 3.669*** (0.132)
  - Nonzombie×zombie share:
    - ∆K = -0.015*** (0.004)
    - ∆Intan = -0.030*** (0.006)
    - ∆Emp = -0.024** (0.010)
    - ∆Sales = -0.032*** (0.006)
    - TFP = -0.001* (0.000)
    - Int.rate = 0.009 (0.010)
    - ∆Debt = -0.018** (0.008)
  - Nonzombie×distressed NZ share:
    - ∆K = 0.002 (0.006)
    - ∆Intan = -0.040*** (0.008)
    - ∆Emp = -0.041** (0.017)
    - ∆Sales = -0.011 (0.008)
    - TFP = 0.000 (0.001)
    - Int.rate = 0.002 (0.009)
    - ∆Debt = 0.005 (0.012)
- Observations vary by column (e.g., ∆K: 1,660,195; ∆Intan: 1,306,965; ∆Emp: 912,732; ∆Debt: 1,603,966).
- Adjusted R2 reported per column (e.g., ∆K = 0.205, ∆Debt = 0.044).

### Table D.3: Congestion effects from zombies to nonzombies: four-digit industry
- Same dependent variables as Table D.2, using four-digit industry fixed effects.
- Key coefficients (selected):
  - Nonzombie (∆K) = 1.676*** (0.056)
  - Nonzombie×zombie share (∆K) = -0.017*** (0.004)
  - Nonzombie×zombie share (∆Emp) = -0.032*** (0.010)
  - Observations: ∆K = 1,651,207; Adjusted R2 for ∆K = 0.209.
- Notes: Controls, Firm FE, Country×Industry×Time FE; clustered SEs.

### Table D.4: Congestion effects from zombies to nonzombies: pre-GFC vs post-GFC
- Interaction terms capture differential post-GFC effects:
  - Nonzombie (∆K) = 1.695*** (0.056)
  - Nonzombie×zombie share = 0.012 (0.008)
  - Nonzombie×zombie share×post-GFC = -0.033*** (0.008)
  - For ∆Intan: Nonzombie×zombie share×post-GFC = -0.062*** (0.013)
  - For Int.rate: Nonzombie×zombie share = 0.135*** (0.039); Nonzombie×zombie share×post-GFC = -0.152*** (0.037)
- Observations: e.g., ∆K = 1,660,513; Adjusted R2 for ∆K = 0.205.

### Table D.5: Congestion effects from zombies to nonzombies: AEs vs EMs
- Splits effects by EM indicator (interaction Nonzombie×zombie share×EM):
  - Nonzombie (∆K) = 1.687*** (0.056)
  - Nonzombie×zombie share (∆K) = -0.010* (0.006)
  - Nonzombie×zombie share×EM (∆K) = -0.009 (0.006)
  - For ∆Debt: Nonzombie×zombie share = -0.038*** (0.011); Nonzombie×zombie share×EM = 0.037*** (0.014)
- Observations: ∆K = 1,660,513; Adjusted R2 for ∆K = 0.205.

Notes (common to tables): Control variables not reported; firm fixed effects included; country×industry×time fixed effects included; standard errors clustered by country-industry-time; asterisks denote significance at the 10%, 5%, and 1% levels.

*Source: Appendix C and Appendix D figures and tables from the provided IMF content unit.*

### References

### References

### Studies on zombie lending and zombie firms
- Acharya, V. V., M. Crosignani, T. Eisert, and S. Steffen (2022). Zombie Lending: Theoretical, International, and Historical Perspectives. Annual Review of Financial Economics 14(1), 21–38.
- Caballero, R. J., T. Hoshi, and A. K. Kashyap (2008). Zombie Lending and Depressed Restructuring in Japan. American Economic Review 98(5), 1943–77.
- Banerjee, R. and B. Hofmann (2022). Corporate zombies: anatomy and life cycle. Economic Policy 37(112), 757–803.
- McGowan, M. A., D. Andrews, and V. Millot (2018). The walking dead? Zombie firms and productivity performance in OECD countries. Economic Policy 33(96), 685–736.
- Storz, M., M. Koetter, R. Setzer, and A. Westphal (2017). Do we want these two to tango? On zombie firms and stressed banks in Europe. Working Paper Series 2104, European Central Bank.
- Kulkarni, N., S. Ritadhi, S. Mukherjee, and K. Waldock (2021). Unearthing Zombies. Available at https://ssrn.com/abstract=3495660, SSRN.
- Favara, G., C. Minoiu, and A. Perez (2022). Zombie Lending to U.S. firms.  Available at https://ssrn.com/abstract=4065886, SSRN.
- Bonfim, D., G. Cerqueiro, H. Degryse, and S. Ongena (2023). On-site inspecting zombie lending. Management Science 69(5), 2547–3155.

### Bank interventions, bailouts, and policy responses
- Acharya, V. V., L. Borchert, M. Jager, and S. Steffen (2021). Kicking the Can Down the Road: Government Interventions in the European Banking Sector. The Review of Financial Studies 34(9), 4090–4131.
- Acharya, V. V., T. Eisert, C. Eufinger, and C. Hirsch (2019). Whatever It Takes: The Real Effects of Unconventional Monetary Policy. Review of Financial Studies 32(9), 3366–3411.
- Giannetti, M. and A. Simonov (2013). On the Real Effects of Bank Bailouts: Micro Evidence from Japan. American Economic Journal: Macroeconomics 5(1), 135–67.
- Faccio, M., R. W. Masulis, and J. J. McConnell (2006). Political connections and corporate bailouts. The Journal of Finance 61(6), 2597–2635.
- Becker, B. and V. Ivashina (2022). Weak Corporate Insolvency Rules: The Missing Driver of Zombie Lending. AEA Papers and Proceedings 112, 516–20.
- Araujo, J. D., J. Garrido, E. Kopp, R. Varghese, and W. Yao (2022). Policy Options for Supporting and Restructuring Firms Hit by the COVID-19 Crisis. Departmental Paper 2022/002, International Monetary Fund.
- Andrews, D. and F. Petroulakis (2019). Breaking the shackles: Zombie firms, weak banks and depressed restructuring in Europe. Working Paper Series 2240, European Central Bank.

### Credit allocation, financial constraints, and misallocation
- Schivardi, F., E. Sette, and G. Tabellini (2022). Credit Misallocation During the European Financial Crisis. The Economic Journal 132(641), 391–423.
- Bussolo, M., F. de Nicola, U. Panizza, and R. Varghese (2022). Politically connected firms and privileged access to credit: Evidence from central and eastern europe. European Journal of Political Economy 71, 102073.
- Peek, J. and E. S. Rosengren (2005). Unnatural Selection: Perverse Incentives and the Misallocation of Credit in Japan. American Economic Review 95(4), 1144–1166.
- Müller, K. and E. Verner (2023). Credit Allocation and Macroeconomic Fluctuations. Available at https://ssrn.com/abstract=3781981, SSRN.
- Caballero, R. J., T. Hoshi, and A. K. Kashyap (2008). Zombie Lending and Depressed Restructuring in Japan. American Economic Review 98(5), 1943–77.

### Intangible capital, investment, and firm value
- Corrado, C., C. Hulten, and D. Sichel (2009). Intangible capital and U.S. economic growth. Review of Income and Wealth 55(3), 661–685.
- Eisfeldt, A. L., E. Kim, and D. Papanikolaou (2022). Intangible Value. Critical Finance Review 11(2), 299–332.
- Peters, R. H. and L. A. Taylor (2017). Intangible capital and the investment-q relation. Journal of Financial Economics 123(2), 251–272.
- Ewens, M., R. H. Peters, and S. Wang (2019). Measuring Intangible Capital with Market Prices. NBER Working Papers 25960, National Bureau of Economic Research, Inc.
- Belo, F., V. Gala, J. Salomao, and M. A. Vitorino (2022). Decomposing Firm Value. Journal of Financial Economics 143(2), 619–639.
- Olley, G. S. and A. Pakes (1996). The Dynamics of Productivity in the Telecommunications Equipment Industry. Econometrica 64(6), 1263–1297.
- Goyal, V. K. and T. Yamada (2004). Asset Price Shocks, Financial Constraints, and Investment: Evidence from Japan. The Journal of Business 77(1), 175–200.

### Macroprudential policy, monetary policy, and financial development
- Akinci, O. and J. Olmstead-Rumsey (2018). How effective are macroprudential policies? an empirical investigation. Journal of Financial Intermediation 33, 33–57.
- Alam, Z., A. Alter, J. Eiseman, G. Gelos, H. Kang, M. Narita, E. Nier, and N. Wang (2019). Digging Deeper–Evidence on the Effects of Macroprudential Policies from a New Database. IMF Working Papers 2019/066, International Monetary Fund.
- Crouzet, N. (2021). Credit Disintermediation and Monetary Policy. IMF Economic Review 69(1), 23–89.
- Acharya, V. V., T. Eisert, C. Eufinger, and C. Hirsch (2019). Whatever It Takes: The Real Effects of Unconventional Monetary Policy. Review of Financial Studies 32(9), 3366–3411.
- Jordà, O. (2005). Estimation and Inference of Impulse Responses by Local Projections. The American Economic Review 95(1), 161–182.
- Chinn, M. D. and H. Ito (2006). What matters for financial development? Capital controls, institutions, and interactions. Journal of Development Economics 81(1), 163–192.

### Firm-level data, measurement, and methodological contributions
- Kalemli-Özcan, S., B. Sorensen, C. Villegas-Sanchez, V. Volosovych, and S. Yesiltas (2015). How to Construct Nationally Representative Firm Level Data from the Orbis Global Database: New Facts and Aggregate Implications. NBER Working Papers 21558, National Bureau of Economic Research, Inc.
- Olley, G. S. and A. Pakes (1996). The Dynamics of Productivity in the Telecommunications Equipment Industry. Econometrica 64(6), 1263–1297.
- Whited, T. M. and G. Wu (2006). Financial Constraints Risk. Review of Financial Studies 19(2), 531–559.
- Díez, F., R. Duval, J. Fan, J. Garrido, S. Kalemli-Özcan, C. Maggi, S. Martinez-Peria, and N. Pierri (2021). Insolvency Prospects Among Small-and-Medium-Sized Enterprises in Advanced Economies: Assessment and Policy Options. IMF Staff Discussion Notes 2021/002, International Monetary Fund.

*The Rise of the Walking Dead: Zombie Firms Around the World Working Paper No. WP/2023/125*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023125-print-pdf.pdf_