## wpiea2025029-print-pdf - 21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023).

## Source details

**Canonical URL:** [wpiea2025029-print-pdf - 21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023).](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025029-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025029-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025029-print-pdf.pdf.json)

---

### Program coverage and timing
- "21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023)."
- Each of the four programs had concluded by end-2021, while the repayment deadline to maintain the loan forgiveness portion of the CEBA ended in 2024.
- Note on sample: "Including incorporated and unincorporated businesses and partnerships."

### Methodology for aggregate productivity effects
- Aggregate labour productivity changes evaluated using the Griliches and Regev (1995) decomposition, adapted into zombie and non-zombie firm components.
- All items reported in real terms using a business sector GDP deflator.
- Caveat: "At the time of writing this paper, industry-level GDP deflators were only available up to 2020."

### Aggregate and historical dynamics of zombie firms
- Share of zombie firms since 2002 hovered around five to seven percent, peaked in 2011, and declined thereafter until 2019.
- Key prevalence figures:
  - 2011: 7.1 percent
  - 2019: 5.6 percent
  - 2022: 4.4 percent
- Between 2019 and 2022:
  - Zombie employment share fell from 5.4 to 4.5 percent.
  - Zombie capital share fell from 8.4 to 7.5 percent.
- Capital share remains elevated relative to historical levels.
- Alternative zombie definitions show similar dynamics; remaining analysis focuses on McGowan et al. (2018) definition.
- Conclusion: "the pandemic did not lead to a mass proliferation of zombie firms that might have inhibited the post-pandemic recovery."

### Industry-level changes (selected findings)
- Across industries, decline in zombie firm share between 2019 and 2022.
- Largest percentage point declines in prevalence:
  - Arts, entertainment and recreation: from 9.0 to 6.0 percent.
  - Retail trade: from 6.5 to 4.1 percent.
- Minimal decline example:
  - Health care and social assistance: declined by only 0.02 percentage points.
- Industries with increases in zombie employment share:
  - Educational services; arts, entertainment and recreation; transportation and warehousing.
- Industries with increases in zombie capital share:
  - Transportation and warehousing; arts, entertainment and recreation; health care and social assistance; accommodation and food services.
- Explanation for simultaneous decline in prevalence but rises in employment/capital shares:
  - i) surviving zombie firms growing faster than non-zombie firms;
  - ii) exit of smaller-than-average zombie firms and exit of larger-than-average non-zombie firms;
  - iii) replacement of existing zombie firms with larger-than-average non-zombie firms.
- Empirical finding: "the transition from non-zombie to zombie status is by far the most important factor since 2019."
- Transition of large non-zombie firms into zombie status explains the increase in industry employment and capital shares in 2022.

### Three-year transition probabilities and pandemic changes
- Historical average (2002–2019) three-year transitions (status in year t to year t+3) for a zombie firm:
  - Remain zombie: 24.5 percent
  - Become non-zombie: 46.7 percent
  - Exit: 28.8 percent
- Three-year transition period 2019–2022 for zombie firms:
  - Remain zombie: 19.3 percent
  - Zombie → exit: 34.4 percent
  - Zombie → non-zombie: 46.4 percent
- Changes in percentage points (2019–2022 vs historical):
  - Non-zombie: -1.0 percentage point remaining non-zombie; -1.7 percentage points transitioning to zombie; 2.7 percentage points higher exit rate.
  - Zombie: -0.3 percentage point transitioning to non-zombie; -5.2 percentage points in persistence as zombie; 5.5 percentage points higher exit rate.
- Interpretation: Over the pandemic, zombie firms had lower probability of recovery and higher probability of exit; non-zombie firms also experienced higher exit rates but to a smaller extent. "Taken together, the evidence shows that over the pandemic, cleansing effects were present."

### Decomposition of drivers of the declining zombie share
- Counterfactual applying pre-pandemic transitions to 2019 stock predicts zombie share of 6.0 percent in 2022 (observed 4.4 percent).
- Simulations (three offset scenarios) results:
  - Increasing zombie firm exit rate alone predicts zombie share of 5.9 percent (lower bound 5.7 percent; upper bound 6.0 percent).
  - Lower transition probability of non-zombie → zombie is the most prominent factor; bounds range from 4.5 to 4.6 percent.
  - Simulating higher exit rate of non-zombie firms shows only reducing non-zombie → zombie transition lowers share to 3.7 percent.
- Key conclusion: "the zombie firm share declined over the pandemic because of the exit of non-zombie firms that traditionally would have entered zombie status." Non-zombie firms that were impacted "skipped over zombification and directly exited the economy."

### Uptake of business support programs by zombie and non-zombie firms
- 58.5 percent of zombie firms in 2019 received pandemic-related support.
- 64.6 percent of non-zombie firms in 2019 received pandemic-related support.
- Higher uptake among non-zombie firms also found for the two largest programs, the CEWS and CEBA.
- Pre-pandemic zombie firms had a lesser likelihood of receiving support compared to the general population of firms.

### Empirical approach (multinomial logit models)
- Two multinomial logit models estimated on transitions into zombie, non-zombie, and exit:
  - Pandemic model: 2019 to 2022 transitions.
  - Pre-pandemic model: three-year transitions between 2002 to 2019 (last transition 2016 to 2019); excludes business supports and includes year dummies (base year = 2002).
- Pandemic model variables:
  - Dependent variable: Y indicating firm status at time t (reference = non-zombie).
  - Firm characteristics X: return on assets, debt-to-asset ratio, productivity, employment (measured at t−3).
  - Support dummy S: use of main business support programs (CEWS, CECRA, CERS, CEBA).
  - Demand shock D: percentage change in revenue of firm’s three-digit NAICS industry omitting the firm itself.
  - Industry dummies I at two-digit NAICS level.
- Pre-pandemic coefficients applied to 2019 population to generate counterfactual 2019–2022 transitions.

### Impact of firm characteristics (predicted transition probabilities, pre-pandemic model applied to 2019)
- Predicted probabilities for non-zombie firms:
  - Pre-pandemic characteristics: Non-zombie 73.7, Zombie 4.9, Exit 21.3 (Percent)
  - 2019 characteristics: Non-zombie 75.8, Zombie 4.2, Exit 20.0 (Percent)
  - Difference (pp): Non-zombie 2.3, Zombie -0.8, Exit -1.5 (Percentage points)
- Predicted probabilities for zombie firms:
  - Pre-pandemic characteristics: Non-zombie 46.5, Zombie 24.4, Exit 29.0 (Percent)
  - 2019 characteristics: Non-zombie 48.2, Zombie 23.7, Exit 28.1 (Percent)
  - Difference (pp): Non-zombie 1.8, Zombie -0.7, Exit -1.1 (Percentage points)
- Interpretation: 2019 firm characteristics increased probability of transitioning to non-zombie, decreased probability of transitioning to zombie, and decreased exit rates for both groups relative to pre-pandemic coefficients; thus changes in coefficients, the COVID-19 shock, or business supports drive observed pandemic transitions.

### Impact of business supports and coefficient shifts (pandemic model predictions)
- Pandemic model predicted probabilities for non-zombie firms (2019 characteristics):
  - (a) With observed supports: Non-zombie 72.8, Zombie 3.3, Exit 23.9 (Percent)
  - (b) Assuming no business supports: Non-zombie 62.5, Zombie 3.0, Exit 34.5 (Percent)
  - Differences (a) − (b): Non-zombie 10.3, Zombie 0.3, Exit -10.6 (Percentage points)
- Effects of changing pandemic coefficients to pre-pandemic values (non-zombie):
  - Pre-pandemic industry coefficient (h): Non-zombie 77.5, Zombie 3.0, Exit 19.5 (Percent); (a) − (h): -4.7, 0.3, 4.4 (Percentage points)
  - Pre-pandemic year coefficient (i): Non-zombie 77.2, Zombie 5.9, Exit 16.9 (Percent); (a) − (i): -4.4, -2.6, 7.0 (Percentage points)
- Interpretation for non-zombie firms:
  - Supports lowered exit rates by 10.6 percentage points, increased non-zombie persistence by 10.3 percentage points, and increased non-zombie → zombie probability by 0.3 percentage points.
  - Business supports mitigated exits mainly by increasing persistence rather than producing zombification.
  - Industry-specific effects increased non-zombie exit by 4.4 percentage points and decreased persistence by 4.7 percentage points.
  - Aggregate (year) effects increased non-zombie exit by 7.0 percentage points and decreased persistence by 4.4 percentage points, and lowered non-zombie → zombie transition.

- Pandemic model predicted probabilities for zombie firms (2019 characteristics):
  - (a) With observed supports: Non-zombie 46.4, Zombie 19.3, Exit 34.4 (Percent)
  - (b) Assuming no business supports: Non-zombie 35.0, Zombie 18.4, Exit 46.6 (Percent)
  - Differences (a) − (b): Non-zombie 11.4, Zombie 0.9, Exit -12.3 (Percentage points)
- Effects of changing pandemic coefficients to pre-pandemic values (zombie):
  - Pre-pandemic industry coefficient (h): Non-zombie 54.2, Zombie 19.0, Exit 26.7 (Percent); (a) − (h): -7.9, 0.3, 7.6 (Percentage points)
  - Pre-pandemic year coefficient (i): Non-zombie 48.6, Zombie 25.2, Exit 26.3 (Percent); (a) − (i): -2.2, -5.9, 8.1 (Percentage points)
- Interpretation for zombie firms:
  - Business supports helped zombie firms convert into non-zombie firms (11.4 percentage points higher) and lowered their exit by 12.3 percentage points, with a minor increase in zombie persistence of 0.9 percentage points.
  - The zombie → non-zombie transition was 10.1 percentage points lower due to worse economic conditions (sum of industry and aggregate effects), but business supports offset this by 11.4 percentage points.
  - Overall, pandemic increased probability of zombie firm exit, while business supports helped zombie firms recover into non-zombie status.
  - Evidence points to zombification declining over the pandemic; no evidence that the pandemic or uptake of supports propagated zombie firms.

### Firm-level productivity changes over the pandemic (2019 to 2022)
- Change in average firm labour productivity (real total revenue per employee) by transition:
  - Non-zombie → zombie: -31.0 percent
  - Zombie → zombie (persistent zombie): -19.8 percent
  - Non-zombie → non-zombie (persistent non-zombie): -6.3 percent
  - Zombie → non-zombie: +8.3 percent
- Comparison to pre-pandemic three-year transitions:
  - Zombie → non-zombie pre-pandemic improvement: +16.0 percent (vs +8.3 percent in pandemic)
  - Persistent non-zombie pre-pandemic: essentially no change (vs -6.3 percent in pandemic)
  - Zombie → zombie pre-pandemic decline: -5.8 percent
  - Non-zombie → zombie pre-pandemic decline: -17.5 percent
- Interpretation: Pandemic uniquely drove larger declines in firm productivity relative to historical three-year transitions; productivity increase for zombie → non-zombie supports objective that supports facilitated recoveries.

### Entry, exit, and productivity composition
- In 2019, zombie firms were on average 61 percent as productive as non-zombie firms.
- Exit timing by productivity:
  - Firms exiting in 2020 had the lowest 2019 productivity; exit cohorts in 2021 and 2022 had progressively higher 2019 productivity.
  - Indicates the least productive firms exited first during the pandemic.
  - Availability of supports did not prevent the weakest firms from exiting early.
- Entrants:
  - Entrants in 2019 and 2022 had lower productivity than non-zombie firms in the same years.
  - Productivity of entrants declined between 2019 and 2022.

### Decomposition of aggregate labour productivity changes (2019 to 2022)
- Framework splits contributions into within-firm, between-firm (reallocation), entry, and exit components, further split into non-zombie and zombie elements.
- Aggregate labour productivity in the business sector (real total revenue per employee) decreased by $9,299 between 2019 and 2022. This change resulted from large, offsetting movements across components.
- Key component contributions (dollars per worker):
  - Overall within-firm contribution (continuing firms): -$8,015
    - Continuing non-zombie firms: -$6,668
    - Continuing zombie firms: -$1,348
  - Contributions from firms that switched status:
    - Continuing firms switching from zombie to non-zombie: +$1,104
    - Continuing firms switching from non-zombie to zombie: -$1,167
  - Reallocation between continuing firms (between): +$1,009
    - Non-zombie contribution to reallocation: +$656
    - Zombie contribution to reallocation: +$353
  - Exit contribution: +$8,118
    - Exit of non-zombie firms: +$6,553
    - Exit of zombie firms: +$1,565
    - Exit of zombie firms contributed disproportionately relative to their share.
  - Entry contribution: -$10,411
    - Three-year entry rate in 2022: 23.5 percent (versus average 27.1 percent between 2009–2019)
  - Net entry (entry plus exit): -$2,293 (versus historical net entry contribution of +$2,661)
- The within-firm contribution from non-zombie firms shifted from +$3,989 (historical) to -$6,668 (2019–2022), driving much of the aggregate change.
- Zombie firms had relatively small overall contributions and did not materially change from historical levels despite disproportionate effects relative to their size.

### Impacts of pandemic business supports on productivity (by support status)
- Firms that received support: aggregate labour productivity increased by $6,295.
- Firms that did not receive support: aggregate labour productivity decreased by -$5,183.
- Support-associated increases were broad-based: within-firm, between-firm, and exit components.
  - Within: -$2,161 (Any support) versus -$5,855 (No support)
  - Between: +$1,753 (Any support) versus -$743 (No support)
  - Exit: +$6,703 (Any support) versus +$1,415 (No support)
- Interpretation:
  - Exiting firms that received support contributed $6,703, while exiting unsupported firms contributed $1,415 — indicating exiting support recipients had lower productivity than average, so creative destruction continued among supported firms.
  - Between-firm reallocation was positive for supported firms (+$1,753) and negative for unsupported firms (-$743).
  - Overall, pandemic-related supports are associated with higher aggregate labour productivity without compromising creative destruction.

### Comparative and contextual observations
- Reallocation over the pandemic remained positive but lower than historical levels for both non-zombie and zombie firms.
- Entrants in 2022 were uniquely unproductive compared to past entrants, driving the more negative entry contribution.
- Despite concerns, evidence indicates limited widespread zombification; supported firms made positive contributions in many dimensions.
- Both zombie and non-zombie firms displayed greater exit probabilities compared to historical trends; systemic industry and aggregate shocks were main contributors to higher exits (pandemic’s cleansing effects).

### Selected quantitative excerpts (Dollars per worker, 2019 to 2022)
- Total: 9,780 (Average of contributions across every four-year period between 2009 and 2019) versus -9,299 (2019 to 2022) and 6,295 (Any support) versus -15,593 (No support).
- Subtotal (excl. Entry): 6,295 (Any support) versus -5,183 (No support).
- Within: 3,102 (Average) versus -8,015 (2019 to 2022) and -2,161 (Any support) versus -5,855 (No support).
- Between: 4,018 (Average) versus 1,009 (2019 to 2022) and 1,753 (Any support) versus -743 (No support).
- Exit: 7,801 (Average) versus 8,118 (2019 to 2022) and 6,703 (Any support) versus 1,415 (No support).
- Entry: -5,140 (Average) versus -10,411 (2019 to 2022).
- Net entry: 2,661 (Average) versus -2,293 (2019 to 2022).

### Conclusion (paper’s synthesis)
- Pandemic policy responses were urgent and broad to keep businesses afloat and prevent failures, with a risk of zombification.
- Despite substantial support, there was no widespread increase in zombie firms; prevalence of zombie firms declined across industries.
- Business supports kept firms alive and operating as non-zombie firms and led some zombie firms to recover.
- Zombie firms had a limited contribution to aggregate labour productivity developments during the pandemic; supported firms made a positive contribution.

### Appendix highlights and selected multinomial logit coefficients
- Figures A1–A3: alternative zombie definitions and shares using adjusted McGowan et al. (2018) definitions.
- Table A1–A3: multinomial logit coefficients for firm status transitions (examples summarized in paper).
- Selected coefficients (Table A4 — status in year t, zombie firms in year t-3; base outcome = non-zombie; omitted industry: agriculture, forestry and fishing and hunting; omitted year: 2002; sample period: before 2019):
  - Outcome "Zombie in t" coefficients (examples):
    - Return on assets: -0.06 (Std. Error 0.01, p-value 0.00)
    - Debt-to-asset ratio: 0.04 (Std. Error 0.00, p-value 0.00)
    - Utilities: 1.06 (Std. Error 0.03, p-value 0.00)
    - Real estate and rental and leasing: 0.63 (Std. Error 0.02, p-value 0.00)
    - Arts, entertainment and recreation: 0.52 (Std. Error 0.02, p-value 0.00)
    - Constant: -1.05 (Std. Error 0.02, p-value 0.00)
  - Outcome "Exit in t" coefficients (examples):
    - Return on assets: -1.00 (Std. Error 0.01, p-value 0.00)
    - Utilities: 0.78 (Std. Error 0.03, p-value 0.00)
    - Retail trade: 0.43 (Std. Error 0.02, p-value 0.00)
    - Accomodation and food services: 0.27 (Std. Error 0.02, p-value 0.00)
    - Constant: -0.97 (Std. Error 0.02, p-value 0.00)
  - Year fixed effects for Zombie in t and Exit in t reported across 2003–2016 with varying coefficients and significance levels (see appendix tables for full listing).

*Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File (as presented in wpiea2025029-print-pdf).*

### 21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023).

### wpiea2025029-print-pdf - 21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023).

### Program coverage and timing
- "21.1 percent received the CERS, and 84.5 percent received the CEBA (Hoffarth and Lin, 2023)."
- Each of the four programs had concluded by end-2021, while the repayment deadline to maintain the loan forgiveness portion of the CEBA ended in 2024.
- Note on sample: "Including incorporated and unincorporated businesses and partnerships."

### Methodology for aggregate productivity effects
- Aggregate labour productivity changes are evaluated using the Griliches and Regev (1995) decomposition, which breaks down changes into contributions from entry, exit and survival of firms.
- The decomposition is adapted into zombie and non-zombie firm components to highlight each dynamic.
- All items are in real terms using a business sector GDP deflator.
- Caveat: "At the time of writing this paper, industry-level GDP deflators were only available up to 2020."

### Aggregate and historical dynamics of zombie firms
- Since 2002, the share of zombie firms in Canada hovered around five to seven percent, peaked in 2011, and declined after that till 2019.
- Between 2011, 2019 and 2022, the share of zombie firms fell from 7.1 to 5.6 to 4.4 percent, respectively.
- Between 2019 and 2022:
  - Zombie employment share fell from 5.4 to 4.5 percent.
  - Zombie capital share fell from 8.4 to 7.5 percent.
- The capital share remains elevated compared to historical levels.
- Alternative zombie definitions (Albuquerque and Iyer (2024); Storz et al. (2017)) show similar dynamics; authors focus on McGowan et al. (2018) definition for remaining analysis.
- Conclusion: "the pandemic did not lead to a mass proliferation of zombie firms that might have inhibited the post-pandemic recovery."

### Industry-level changes (selected findings)
- Across all industries, there is a decline in the share of zombie firms between 2019 and 2022.
- Largest percentage point declines:
  - Arts, entertainment and recreation: from 9.0 to 6.0 percent.
  - Retail trade: from 6.5 to 4.1 percent.
- Minimal decline example:
  - Health care and social assistance: declined by only 0.02 percentage points.
- Employment share increases for zombie firms occurred in:
  - Educational services; arts, entertainment and recreation; transportation and warehousing.
- Capital share increases occurred in:
  - Transportation and warehousing; arts, entertainment and recreation; health care and social assistance; accommodation and food services.
- Explanation for simultaneous decline in prevalence but rises in employment/capital shares:
  - i) surviving zombie firms growing faster than non-zombie firms;
  - ii) exit of smaller-than-average zombie firms and exit of larger-than-average non-zombie firms;
  - iii) replacement of existing zombie firms with larger-than-average non-zombie firms.
- Empirical finding: "the transition from non-zombie to zombie status is by far the most important factor since 2019."
- Transition of large non-zombie firms into zombie status explains the increase in industry employment and capital shares in 2022.

### Three-year transition probabilities and pandemic changes
- Average of all three-year transition rates between 2002 and 2019 (status in year t to year t+3):
  - For a zombie firm: 24.5 percent chance of remaining a zombie firm, 46.7 percent chance of becoming a non-zombie firm, and 28.8 percent chance of exiting.
- For the three-year transition period between 2019 and 2022:
  - Zombie firms were less likely to remain zombies: 19.3 percent.
  - Zombie to exit increased to 34.4 percent.
  - Zombie to non-zombie fell slightly to 46.4 percent.
- Changes in percentages (difference between historical and 2019–2022):
  - Non-zombie: -1.0 percentage point remaining non-zombie; -1.7 percentage points transitioning to zombie; 2.7 percentage points higher exit rate.
  - Zombie: -0.3 percentage point transitioning to non-zombie; -5.2 percentage points in persistence as zombie; 5.5 percentage points higher exit rate.
- Interpretation: Over the pandemic, zombie firms had a lower probability of recovering and a higher probability of exiting; non-zombie firms also had higher exit rates but to a smaller extent. "Taken together, the evidence shows that over the pandemic, cleansing effects were present."

### Decomposition of drivers of the declining zombie share
- Counterfactual: Applying pre-pandemic transitions to the 2019 stock of firms yields a predicted zombie firm share of 6.0 percent in 2022 (instead of the observed 4.4 percent).
- Simulations adjusting individual transition probabilities (with three offset scenarios: proportional offset; entire weight shifted to one other transition producing lower or upper bounds) yield:
  - Changing the zombie firm exit rate alone yields a predicted zombie share of 5.9 percent (lower bound 5.7 percent; upper bound 6.0 percent) — far from the actual 4.4 percent.
  - The most prominent factor accounting for the decline in the zombie firm share in 2022 is the lower transition probability of non-zombie to zombie status; bounds on the share in this case range from 4.5 to 4.6 percent.
  - When simulating the higher exit rate of non-zombie firms, only reducing the transition into zombie status lowers the share to 3.7 percent.
- Key conclusion: "the zombie firm share declined over the pandemic because of the exit of non-zombie firms that traditionally would have entered zombie status." In other words, "non-zombie firms that were impacted essentially skipped over zombification and directly exited the economy."

*Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File; IMF Working Paper excerpt.*

### 4.4 The economic conditions led more firms to exit, while the business supports

### 4.4 The economic conditions led more firms to exit, while the business supports helped firms to survive and operate as non-zombie firms

### Uptake of business support programs by zombie and non-zombie firms
- 58.5 percent of zombie firms in 2019 received pandemic-related support.
- 64.6 percent of non-zombie firms in 2019 received pandemic-related support.
- The higher share among non-zombie firms is also found for the two largest programs, the CEWS and CEBA.
- Pre-pandemic zombie firms had a lesser likelihood of receiving support compared to the general population of firms.

### Empirical approach (multinomial logit models)
- Two sets of multinomial logit models estimated on transition probabilities into zombie, non-zombie, and exit status:
  - Pandemic period: 2019 to 2022 transition (pandemic model).
  - Pre-pandemic period: three-year transitions between 2002 to 2019 (last transition between 2016 to 2019) (pre-pandemic model).
- Pandemic model (estimated in 2022 for firms in 2019) includes:
  - Dependent variable: 푌푌 indicating firm status at time t (reference = non-zombie).
  - Firm characteristics 푋: return on assets, debt-to-asset ratio, productivity, employment (measured at t−3).
  - Support dummy S: whether firm used any main business support programs (CEWS, CECRA, CERS, CEBA).
  - Demand shock D: percentage change in revenue of firm’s three-digit NAICS industry omitting the firm itself.
  - Industry dummies I at two-digit NAICS level.
- Pre-pandemic model excludes business supports and includes year dummies (base year = 2002) to pool pre-pandemic transitions.
- Coefficients from pre-pandemic model applied to 2019 firm population to predict 2019–2022 transitions (counterfactual).

### Impact of firm characteristics (predicted transition probabilities)
- Predicted transition probabilities from pre-pandemic model applied to 2019 population:
  - Non-zombie in pre-pandemic model:
    - Pre-pandemic characteristics: Non-zombie 73.7, Zombie 4.9, Exit 21.3 (Percent)
    - 2019 characteristics: Non-zombie 75.8, Zombie 4.2, Exit 20.0 (Percent)
    - Difference (pp): Non-zombie 2.3, Zombie -0.8, Exit -1.5 (Percentage points)
  - Zombie in pre-pandemic model:
    - Pre-pandemic characteristics: Non-zombie 46.5, Zombie 24.4, Exit 29.0 (Percent)
    - 2019 characteristics: Non-zombie 48.2, Zombie 23.7, Exit 28.1 (Percent)
    - Difference (pp): Non-zombie 1.8, Zombie -0.7, Exit -1.1 (Percentage points)
- Interpretation:
  - Firm characteristics in 2019 increase transition probability to non-zombie, decrease to zombie, and decrease exit rates for both zombie and non-zombie firms relative to pre-pandemic coefficients.
  - These firm characteristics do not explain the observed lower probability of non-zombie status in 2022 or the higher exit rates in 2022; thus changes in coefficients, the COVID-19 shock, or business supports drive pandemic transitions.

### Impact of business supports and coefficient shifts (predicted transition probabilities — pandemic model)
- Pandemic model predicted probabilities for 2019 characteristics (panel for non-zombie):
  - (a) 2019 characteristics: Non-zombie 72.8, Zombie 3.3, Exit 23.9 (Percent)
  - (b) 2019 characteristics assuming no business supports: Non-zombie 62.5, Zombie 3.0, Exit 34.5 (Percent)
  - Differences (a) − (b): Non-zombie 10.3, Zombie 0.3, Exit -10.6 (Percentage points)
  - Changing pandemic coefficients one at a time to pre-pandemic values:
    - Pre-pandemic coefficient on industry: (h) Non-zombie 77.5, Zombie 3.0, Exit 19.5 (Percent); difference (a) − (h): -4.7, 0.3, 4.4 (Percentage points)
    - Pre-pandemic coefficient on year: (i) Non-zombie 77.2, Zombie 5.9, Exit 16.9 (Percent); difference (a) − (i): -4.4, -2.6, 7.0 (Percentage points)
- Interpretation for non-zombie firms:
  - Presence of supports lowered exit rates by 10.6 percentage points, increased non-zombie persistence by 10.3 percentage points, and increased non-zombie → zombie probability by 0.3 percentage points.
  - Business supports mitigated pandemic impact on non-zombie exits mainly by increasing persistence rather than converting non-zombie firms into zombies.
  - Industry-specific effects increased non-zombie exit by 4.4 percentage points and decreased persistence by 4.7 percentage points.
  - Aggregate (year) effects increased non-zombie exit by 7.0 percentage points and decreased persistence by 4.4 percentage points, and lowered non-zombie → zombie transition.

- Pandemic model predicted probabilities for 2019 characteristics (panel for zombie):
  - (a) 2019 characteristics: Non-zombie 46.4, Zombie 19.3, Exit 34.4 (Percent)
  - (b) 2019 characteristics assuming no business supports: Non-zombie 35.0, Zombie 18.4, Exit 46.6 (Percent)
  - Differences (a) − (b): Non-zombie 11.4, Zombie 0.9, Exit -12.3 (Percentage points)
  - Changing pandemic coefficients one at a time to pre-pandemic values:
    - Pre-pandemic coefficient on industry: (h) Non-zombie 54.2, Zombie 19.0, Exit 26.7 (Percent); difference (a) − (h): -7.9, 0.3, 7.6 (Percentage points)
    - Pre-pandemic coefficient on year: (i) Non-zombie 48.6, Zombie 25.2, Exit 26.3 (Percent); difference (a) − (i): -2.2, -5.9, 8.1 (Percentage points)
- Interpretation for zombie firms:
  - Business supports helped zombie firms convert into non-zombie ones and lowered their exit (with a minor increase in zombie persistence of 0.9 percentage points).
  - The zombie → non-zombie transition was 10.1 percentage points lower due to worse economic conditions (sum of industry and aggregate effects), but business supports offset this by 11.4 percentage points.
  - Overall, pandemic increased probability of zombie firm exit, while business supports helped zombie firms recover into non-zombie status.
  - Evidence points to zombification declining over the pandemic; no evidence that the pandemic or uptake of supports propagated zombie firms.

### Firm-level productivity changes over the pandemic
- Change in average firm labour productivity (real total revenue per employee) by transition category (2019 to 2022):
  - Non-zombie → zombie: productivity declined by 31.0 percent.
  - Zombie → zombie (persistent zombie): productivity declined by 19.8 percent.
  - Non-zombie → non-zombie (persistent non-zombie): productivity declined by 6.3 percent.
  - Zombie → non-zombie: productivity increased by 8.3 percent.
- Comparison to pre-pandemic average three-year transitions (prior to 2019):
  - Zombie → non-zombie: pre-pandemic improvement 16.0 percent (compared to 8.3 percent in pandemic).
  - Persistent non-zombie: pre-pandemic essentially no change (compared to −6.3 percent in pandemic).
  - Zombie → zombie: pre-pandemic decline 5.8 percent.
  - Non-zombie → zombie: pre-pandemic decline 17.5 percent.
- Interpretation:
  - Pandemic period uniquely drove larger declines in firm productivity relative to historical three-year transitions.
  - The productivity increase for zombie → non-zombie during the pandemic supports the policy objective of business supports facilitating recoveries.

### Entry, exit, and productivity composition
- In 2019, zombie firms were on average 61 percent as productive as non-zombie firms.
- Exit timing by productivity:
  - Firms that exited in 2020 had the lowest 2019 productivity levels; exit cohorts in 2021 and 2022 had higher 2019 productivity levels progressively.
  - This indicates the least productive firms exited first during the pandemic.
  - Availability of business supports did not prevent the weakest firms from exiting early in the pandemic.
- Entrants:
  - Entrants in 2019 and 2022 had lower productivity than non-zombie firms in the same years.
  - Productivity of entrants declined between 2019 and 2022, consistent with the broad-based decline in productivity.

### Decomposition of aggregate labour productivity changes (Griliches and Regev (1995) framework, modified)
- Decomposition separates contributions into within-firm, between-firm, entry, and exit components, and further splits components into non-zombie and zombie elements.
- Definitions:
  - Within-firm component: weighted changes in firm productivity for continuing firms.
  - Between-firm component: impact of reallocation of labour between continuing firms with productivity above vs. below the aggregate average.
  - Entry contribution: entrants’ productivity relative to aggregate average.
  - Exit contribution: exiters’ productivity relative to aggregate average (split by exiting non-zombie and exiting zombie based on status in t−k).
- Aggregate labour productivity in the business sector (real total revenue per employee) decreased by $9299 between 2019 and

*Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File. — IMF Working Paper excerpt (4.4).*

### 2022. This change was the result of large changes in the various components that net each other.

### wpiea2025029-print-pdf - 2022. This change was the result of large changes in the various components that net each other.

### Decomposition of change in aggregate labour productivity (key findings)
- Overall change in productivity in 2019 to 2022 was the result of large, offsetting movements across components.
- Within-firm contribution (continuing firms) declined by -$8,015.
  - Continuing non-zombie firms: -$6,668.
  - Continuing zombie firms: -$1,348.
- Firms that switched status:
  - Continuing firms switching from zombie to non-zombie: $1,104 (positive contribution).
  - Continuing firms switching from non-zombie to zombie: -$1,167 (negative contribution).
- Reallocation between continuing firms contributed positively: $1,009.
  - Non-zombie firms contribution to reallocation: $656.
  - Zombie firms contribution to reallocation: $353.
- Exit of firms contributed positively to productivity: $8,118.
  - Exit of non-zombie firms: $6,553.
  - Exit of zombie firms: $1,565.
  - Exit of zombie firms contributed disproportionately relative to their share in the economy.
- Entry of firms made a negative contribution: -$10,411.
  - Three-year entry rate in 2022 was 23.5 percent, versus an average of 27.1 percent between 2009-2019.
- Net entry (entry plus exit) contributed negatively: -$2,293, in contrast to past periods where net entry contributed positively: $2,661.
- The within-firm contribution from non-zombie firms shifted from $3,989 (historical) to -$6,668 (2019–2022), driving much of the aggregate change.
- Zombie firms, while contributing disproportionately relative to their presence, had relatively small contributions overall and did not materially change from historical levels.

### Impacts of pandemic business supports on productivity
- Firms that received support experienced an increase in aggregate labour productivity of $6,295.
- Firms that did not receive support experienced a decline in aggregate labour productivity of -$5,183.
- Support-associated increases were broad-based across components: within-firm, between-firm, and exit components.
- Exception: within-firm contribution of firms transitioning between zombie to non-zombie status was higher for firms without support, but the magnitude was small.
- Exit contributions by support status:
  - Exiting firms that received support: $6,703.
  - Exiting firms that did not receive support: $1,415.
  - Interpretation: exiting support-receiving firms had lower productivity than the average, indicating creative destruction continued among supported firms.
- Between-firm component:
  - For supported firms: $1,753 (positive reallocation towards productive supported firms).
  - For unsupported firms: -$743 (misallocation among unsupported firms).
- Overall interpretation: pandemic-related supports are associated with higher aggregate labour productivity without compromising creative destruction.

### Comparative and contextual observations
- Reallocation over the pandemic remained positive but was lower than historical levels for both non-zombie and zombie firms.
- Entrants in 2022 were uniquely unproductive compared to past entrants, driving the more negative entry contribution.
- Despite concerns that supports might generate inefficiencies and zombification, the evidence indicates limited widespread zombification and, in many dimensions, improved aggregate productivity outcomes associated with supports.
- Both zombie and non-zombie firms displayed greater exit probabilities compared to historical trends; systemic industry and aggregate shocks were main contributors to higher exits (pandemic’s cleansing effects).

### Selected quantitative excerpts from Table 7 (Dollars per worker, 2019 to 2022)
- Total: 9,780 (Average of contributions across every four-year period between 2009 and 2019) versus -9,299 (2019 to 2022) and 6,295 (Any support) versus -15,593 (No support).
- Subtotal (excl. Entry): 6,295 (Any support) versus -5,183 (No support).
- Within: 3,102 (Average) versus -8,015 (2019 to 2022) and -2,161 (Any support) versus -5,855 (No support).
- Between: 4,018 (Average) versus 1,009 (2019 to 2022) and 1,753 (Any support) versus -743 (No support).
- Exit: 7,801 (Average) versus 8,118 (2019 to 2022) and 6,703 (Any support) versus 1,415 (No support).
- Entry: -5,140 (Average) versus -10,411 (2019 to 2022).
- Net entry: 2,661 (Average) versus -2,293 (2019 to 2022).

### Conclusion (paper’s synthesis)
- Pandemic policy responses were urgent and broad to keep businesses afloat and prevent failures, with a risk of zombification.
- Despite substantial support, there was no widespread increase in zombie firms; prevalence of zombie firms declined across industries.
- Business supports kept firms alive and operating as non-zombie firms and led some zombie firms to recover.
- Zombie firms had a limited contribution to aggregate labour productivity developments during the pandemic; supported firms made a positive contribution.

### Appendix highlights
- Figures A1–A3 present alternative zombie firm definitions and shares (percent of Firms, Employment, Capital) using adjusted McGowan et al. (2018) definitions (two-year consecutive, two-year exit, three non-consecutive years).
- Table A1–A3 present multinomial logit coefficients for firm status transitions (examples):
  - In Table A1 (Status in 2022, non-zombie firms in 2019): Return on assets coefficient for Zombie in 2022: -0.94 (Std. Error 0.01, p-value 0.00); Support coefficient for Exit in 2022: -0.91 (Std. Error 0.01, p-value 0.00).
  - In Table A2 (Status in 2022, zombie firms in 2019): Support coefficient for Zombie in 2022: -0.37 (Std. Error 0.03, p-value 0.00); Support coefficient for Exit in 2022: -1.08 (Std. Error 0.02, p-value 0.00).
  - In Table A3 (Status in year t, non-zombie firms in year t-3 before 2019): Return on assets coefficient for Zombie in t: -1.08 (Std. Error 0.00, p-value 0.00); selected year effects reported (e.g., Year 2003: -0.07, Std. Error 0.01, p-value 0.00 for Zombie in t).

*Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File (as presented in wpiea2025029-print-pdf).*

### 6.7 Table A4. Multinomial logit coefficients – Status in year t, zombie firms in year t-3

### 6.7 Table A4. Multinomial logit coefficients – Status in year t, zombie firms in year t-3

### Model specification
- Base outcome: non-zombie status.
- Omitted industry: agriculture, forestry and fishing and hunting.
- Omitted year: 2002.
- Sample period: before 2019.
- Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File.

### Coefficients for outcome "Zombie in t"
- Return on assets: Coefficient -0.06, Std. Error 0.01, p-value 0.00
- Debt-to-asset ratio: Coefficient 0.04, Std. Error 0.00, p-value 0.00
- Productivity: Coefficient 0.00, Std. Error 0.00, p-value 0.00
- Employment: Coefficient 0.00, Std. Error 0.00, p-value 0.00
- Demand: Coefficient 0.00, Std. Error 0.00, p-value 0.27
- Utilities: Coefficient 1.06, Std. Error 0.03, p-value 0.00
- Mining, quarrying, and oil and gas extraction: Coefficient 0.00, Std. Error 0.13, p-value 0.97
- Construction: Coefficient -0.08, Std. Error 0.02, p-value 0.00
- Manufacturing: Coefficient 0.26, Std. Error 0.02, p-value 0.00
- Wholesale trade: Coefficient 0.42, Std. Error 0.02, p-value 0.00
- Retail trade: Coefficient 0.42, Std. Error 0.02, p-value 0.00
- Transportation and warehousing: Coefficient 0.00, Std. Error 0.02, p-value 0.84
- Information and cutural industries: Coefficient 0.30, Std. Error 0.03, p-value 0.00
- Finance and insurance: Coefficient 0.44, Std. Error 0.02, p-value 0.00
- Real estate and rental and leasing: Coefficient 0.63, Std. Error 0.02, p-value 0.00
- Professional, scientific and technical services: Coefficient 0.29, Std. Error 0.02, p-value 0.00
- Administrative and support: Coefficient 0.17, Std. Error 0.02, p-value 0.00
- Educational services: Coefficient 0.13, Std. Error 0.03, p-value 0.00
- Health care and social assistance: Coefficient 0.06, Std. Error 0.02, p-value 0.01
- Arts, entertainment and recreation: Coefficient 0.52, Std. Error 0.02, p-value 0.00
- Accomodation and food services: Coefficient 0.40, Std. Error 0.02, p-value 0.00
- Other services: Coefficient 0.20, Std. Error 0.02, p-value 0.00
- Constant: Coefficient -1.05, Std. Error 0.02, p-value 0.00

Year fixed effects (Zombie in t)
- Year 2003: Coefficient -0.04, Std. Error 0.02, p-value 0.03
- Year 2004: Coefficient -0.03, Std. Error 0.02, p-value 0.19
- Year 2005: Coefficient 0.02, Std. Error 0.02, p-value 0.44
- Year 2006: Coefficient 0.12, Std. Error 0.02, p-value 0.00
- Year 2007: Coefficient 0.16, Std. Error 0.02, p-value 0.00
- Year 2008: Coefficient 0.14, Std. Error 0.02, p-value 0.00
- Year 2009: Coefficient 0.09, Std. Error 0.02, p-value 0.00
- Year 2010: Coefficient 0.10, Std. Error 0.02, p-value 0.00
- Year 2011: Coefficient 0.08, Std. Error 0.02, p-value 0.00
- Year 2012: Coefficient 0.07, Std. Error 0.02, p-value 0.00
- Year 2013: Coefficient 0.04, Std. Error 0.02, p-value 0.03
- Year 2014: Coefficient 0.02, Std. Error 0.02, p-value 0.37
- Year 2015: Coefficient 0.06, Std. Error 0.02, p-value 0.00
- Year 2016: Coefficient 0.11, Std. Error 0.02, p-value 0.00

### Coefficients for outcome "Exit in t"
- Return on assets: Coefficient -1.00, Std. Error 0.01, p-value 0.00
- Debt-to-asset ratio: Coefficient 0.00, Std. Error 0.00, p-value 0.04
- Productivity: Coefficient 0.00, Std. Error 0.00, p-value 0.00
- Employment: Coefficient 0.00, Std. Error 0.00, p-value 0.00
- Demand: Coefficient 0.00, Std. Error 0.00, p-value 0.48
- Utilities: Coefficient 0.78, Std. Error 0.03, p-value 0.00
- Mining, quarrying, and oil and gas extraction: Coefficient -0.15, Std. Error 0.12, p-value 0.23
- Construction: Coefficient 0.00, Std. Error 0.02, p-value 0.92
- Manufacturing: Coefficient 0.19, Std. Error 0.02, p-value 0.00
- Wholesale trade: Coefficient 0.35, Std. Error 0.02, p-value 0.00
- Retail trade: Coefficient 0.43, Std. Error 0.02, p-value 0.00
- Transportation and warehousing: Coefficient 0.07, Std. Error 0.02, p-value 0.00
- Information and cutural industries: Coefficient 0.14, Std. Error 0.03, p-value 0.00
- Finance and insurance: Coefficient 0.38, Std. Error 0.02, p-value 0.00
- Real estate and rental and leasing: Coefficient 0.20, Std. Error 0.02, p-value 0.00
- Professional, scientific and technical services: Coefficient 0.14, Std. Error 0.02, p-value 0.00
- Administrative and support: Coefficient 0.04, Std. Error 0.02, p-value 0.04
- Educational services: Coefficient 0.04, Std. Error 0.03, p-value 0.17
- Health care and social assistance: Coefficient -0.06, Std. Error 0.02, p-value 0.01
- Arts, entertainment and recreation: Coefficient 0.04, Std. Error 0.02, p-value 0.10
- Accomodation and food services: Coefficient 0.27, Std. Error 0.02, p-value 0.00
- Other services: Coefficient -0.02, Std. Error 0.02, p-value 0.22
- Constant: Coefficient -0.97, Std. Error 0.02, p-value 0.00

Year fixed effects (Exit in t)
- Year 2003: Coefficient 0.00, Std. Error 0.02, p-value 0.91
- Year 2004: Coefficient 0.07, Std. Error 0.02, p-value 0.00
- Year 2005: Coefficient 0.44, Std. Error 0.13, p-value 0.00
- Year 2006: Coefficient 0.17, Std. Error 0.02, p-value 0.00
- Year 2007: Coefficient 0.16, Std. Error 0.02, p-value 0.00
- Year 2008: Coefficient 0.10, Std. Error 0.02, p-value 0.00
- Year 2009: Coefficient 0.01, Std. Error 0.02, p-value 0.68
- Year 2010: Coefficient 0.02, Std. Error 0.02, p-value 0.21
- Year 2011: Coefficient 0.00, Std. Error 0.02, p-value 0.97
- Year 2012: Coefficient 0.01, Std. Error 0.02, p-value 0.69
- Year 2013: Coefficient 0.01, Std. Error 0.02, p-value 0.54
- Year 2014: Coefficient 0.20, Std. Error 0.02, p-value 0.00
- Year 2015: Coefficient 0.27, Std. Error 0.02, p-value 0.00
- Year 2016: Coefficient 0.32, Std. Error 0.02, p-value 0.00

*Source: Statistics Canada; authors' calculations using the National Accounts Longitudinal Microdata File.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025029-print-pdf.pdf_