## _wp16255

## Source details

**Canonical URL:** [_wp16255](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf.json)

---

### Strategic Corporate Layoffs — Abstract and Research Question
- Firms in the S&P 500 often announce layoffs within days of one another, despite the fact that the average S&P 500 constituent announces layoffs once every 5 years.
- Similar-sized privately-held firms do not display this high-frequency clustering of layoff announcements.
- The paper investigates whether clustering is driven by CEOs managing their reputation in financial markets and develops a theoretical framework to interpret empirical facts.

### Theoretical framework (Section 1)
- Managers delay layoffs during good economic states to avoid damaging the market’s perception of their ability.
- Model predictions:
  - Clustering in the timing of layoff announcements.
  - An amplification mechanism through which the cyclicality of firms’ layoff policies is intensified.

### Data and empirical strategy (Section 1)
- Two novel datasets: one of layoff announcements and one of actual mass layoffs (Mass Layoff Statistics).
- Empirical strategy: compare publicly-listed firms to observably similar matched private firms to estimate the impact of reputation-based incentives on the cyclicality of layoffs.

### Key quantitative findings (Section 1)
- Compared to observably similar matched private firms, public firms are 2.5-3.5 percentage points more likely to conduct mass layoffs in a recession month, indicating a doubling of layoff sensitivity to recessions.
- Firms that cluster layoff announcements at high frequencies are also the firms more likely to engage in mass layoffs during recessions.
- The average S&P 500 constituent announces layoffs once every 5 years (preserved phrasing from source).

### Robustness and exclusion of alternative channels (Section 1)
- Documented effects are not driven by:
  - Leverage differences.
  - Lifecycle differences.
  - The criteria used to match public and private firms.

### Interpretation and macroeconomic implications (Section 1)
- Reputation management by CEOs drives layoff policies at daily frequencies (clustering of announcements) and over the business cycle (amplified cyclicality).
- Reputation-driven actions can have significant macroeconomic consequences through more procyclical employment adjustments by public firms.

### Policy-relevant conclusions (Section 1)
- Incentives from market reputation influence firm-level layoff timing and cyclicality; policymakers and regulators should account for reputation-driven timing when assessing labor-market dynamics and macroeconomic impact of layoffs.
- Understanding reputation management helps interpret observed clustering of corporate layoffs and its contribution to employment volatility over the business cycle.

*Ruchir Agarwal and Julian Kolev, IMF Working Paper WP/16/255 (Section 1).*

### Sections 2–6: Extractability and content quality
- Sections 2, 3, 4, 5, and 6: supplied text is predominantly non-printable, garbled, and OCR-corrupted; coherent natural-language content, numeric projections, policy recommendations, and analytical findings are not recoverable from these sections as supplied.
- Repeated exact control-character-like sequences and fragmented tokens appear verbatim (examples preserved in the source).
- Implications for ingestion and indexing:
  - Treat these sections as unreadable/garbled and rely on other sections or canonical metadata for substantive information.
  - Exact corrupted character sequences can be used as detectors to flag OCR or encoding failures.
- Recommended remediation steps (as supplied):
  - Obtain a clean original of the PDF page or a higher-quality OCR output.
  - Run a robust OCR pipeline with manual verification; validate recovered text against other sections or the document’s table of contents before use.

*Source: _wp16255 - Sections 2–6 (supplied PDF content).*

### Empirical patterns, theoretical comparative statics, and event evidence (Section 7)
- Clustering detection and dynamics:
  - Figure 1: timelines and 30-day window clustering markers (within- and cross-industry).
  - Figure 2: Log p-values for sequence of tests on non-overlapping 60 (business) day windows; null = uniform distribution; horizontal red line = significance level of 0.05.
- Theoretical parameter-space mapping (Figure 3) and comparative statics (Figure 4):
  - Region right of solid lines: equilibrium = delay.
  - Region left of solid lines: equilibrium = terminate.
  - Between lines: mixed strategy.
  - When belief about aggregate state worsens, separating lines shift rightward, changing equilibrium-region boundaries.
- Event study evidence (Figure 5):
  - Response of layoff propensities to "Leader Layoffs" by largest 20 firms; dynamic regressions over 11-day event windows; controls include firm-level controls, annual/month/day fixed effects; standard errors clustered at firm level; point-wise 95% confidence intervals reported.
  - Panels compare public Fortune-500, Forbes-100 (private), same 1-digit NAICS industry, and responses to "News Media Events."

### Descriptive statistics: Public vs Private firms (Table 1 highlights)
- Revenue (in 2005 USD)
  - Public Mean 15143.298067.537064.76***; Private Mean 7162.076992.29169.78; Matched/Buyout sample means also reported (9218.318609.35608.97).
  - Median: Public 6035.374406.83; Private 5075.344903.13.
  - St. Dev: multiple precise values reported in table.
- Employees
  - Public Mean 46803.0136624.1510178.86***; Private Mean 29341.6629614.29-272.63; additional sample moments reported.
- Layoff Propensity
  - Public Mean 0.07210.05410.0180***; Private Mean 0.05840.05580.0027.
- Number Laid off
  - Public Mean 337.74357.88-20.14; Private Mean 281.59248.8232.77; medians and St. Devs provided.
- Share Laid Off
  - Public Mean 0.01340.0165-0.0031; Private Mean 0.01380.0169-0.0031.
- Sample sizes and observations reported per column in table (values preserved as printed).

### Public–Private differences over the business cycle (Table 2 key coefficients)
- Dependent variables: Layoff Indicator (mass layoff indicator in a month) and Share Laid Off.
- Selected coefficient estimates (preserved formatting from source):
  - Public Indicator: 0.00510.0020-0.00290.0029 -0.0064-0.0069***-0.0044-0.0060 (standard errors reported in table).
  - Recession Indicator: 0.0201**-0.0011 0.0240**0.0008 (standard errors reported).
  - Public x Recession interaction: 0.0247***0.0028*0.0259***0.0025 0.0320**0.0040*0.0307**0.0012 (standard errors reported).
- Sample moments and observations preserved exactly as printed; controls include industry fixed effects, quadratic time trend & seasonality, month fixed effects, firm fixed effects, matched-pair fixed effects, and interactions with log employees and log revenue.

### Matched-sample business-cycle totals and durations (Table 3 highlights)
- Dependent variables: total workers laid off over business cycle; median months since last layoff.
- Public Indicator in (1): 0.30390.0656* with (0.4251)(0.0363) reported.
- Sample moments: Mean 2.57457.76; Std. Dev 3.10776.69.
- Observations (matched sample): 236260.
- Specification windows: peak-to-peak (6 months prior to October 2000 to 6 months prior to July 2007) and trough-to-trough measures; matched-pair fixed effects used; standard errors clustered at firm level.

### Additional regression results (Table 4 excerpt)
- Public Indicator: -0.0079-0.0088 (standard errors (0.0261)(0.0272)).
- Public x Recession: 0.0604**0.0181 (standard errors (0.0258)(0.0176)).
- Mean: 0.06920.0142.
- Interpretation: Public x Recession positive and in some specifications statistically significant.

*Source: _wp16255 - Section 7 (PDF: _wp16255 - Section 7).*

### LBO targets and alternate explanations; heterogeneity and stock-market effects (Section 8)
- LBO-target sample (Table 5):
  - Unit: firm-months April 1995–December 2010; dependent variables: layoff indicator and share laid off.
  - Public × Recession coefficients (selected):
    - Column (1): 0.0478*** (0.0127)
    - Column (2): 0.0133 (0.0099)
    - Column (3): 0.0187* (0.0108)
    - Column (4): 0.0250** (0.0108)
  - Additional covariates: Leverage Ratio 0.0721*** (0.0137); Recession × Log Age -0.0341*** (0.0116).
  - Sample statistics and observations preserved as printed.
- Actual layoff propensity by public-firm splits (Table 6):
  - Public Indicator (selected):
    - Column (1): -0.0194** (0.0090)
    - Column (2): -0.0117*** (0.0041)
  - Public × Recession:
    - Column (1): 0.0312* (0.0185)
    - Column (2): 0.0115** (0.0048)
- Executive compensation and CEO tenure heterogeneity (Table 6 continued):
  - KV (Short CEO Tenure Indicator): 0.0032 (0.0050); 0.0025 (0.0062).
  - KV (Avg. Equity-Linked Compensation Share): 0.0072 (0.0071); 0.0164 (0.0091).
  - Key Variable × Recession: 0.0144* (0.0079); 0.0261** (0.0126); 0.0255** (0.0123); 0.0352** (0.0168).
  - Observations: 28746, 27882, 16818, 13410.
- Stock penalty of layoffs — Cumulative Abnormal Returns (Table 7):
  - Layoff by Top 20 Firm in Previous 3 days:
    - Column (1): 0.0049*** (.0017)
    - Column (2): 0.0052*** (.0019)
    - Column (3): 0.0042** (.0020)
    - Column (4): 0.0044** (.0021)
  - Baseline CAR constants: -0.0087*** (0.0013), -0.0081*** (0.0016).
  - Observations: 4796 (reported columns).
- Follower characteristics, analyst coverage, and compensation effects (Tables 8 and 9):
  - Avg. Equity-linked Compensation Share (Table 8, Column (1)): 0.0741*** (0.0186).
  - Total Compensation (Table 8, Column (1)): -0.0014*** (0.0004).
  - CEO Tenure = 0 - 4 years (Table 8, Column (1)): 0.0229*** (0.0075).
  - Analyst Coverage in Past 3 years (Table 9, Column (1)): -0.0254*** (0.0093).
  - Sample sizes and means/std. devs preserved as printed.
- Linking daily-frequency follower behavior to business-cycle outcomes (Table 10):
  - Past 5 year Follower Indicator: 0.0135*** (0.0035).
  - Past 5 year Counterfactual Follower Indicator: 0.0120*** (0.0036).
  - Recession × Past 5 year Follower Indicator: 0.0314*** (0.0085).
  - Observations: 70170 (per reported columns).

### Combining scan statistics and robustness (Section 8 appendices)
- Fisher's Method results (Table A1) for combining p-values across non-overlapping windows; selected printed rows preserved (examples include combinations with p-values 1.000, 0.736, 0.055, 0.0003, 0.041, 0.045, 0.011, 0.001, 0.0001, 0.007, 0.009).
- Robustness tests (Table A2) preserve Public Indicator and Public × Recession estimates as printed (e.g., Public Indicator -0.0194** (0.0090); Public × Recession 0.0312* (0.0185)).

*Source: _wp16255 - Section 8 (PDF: _wp16255 - Section 8).*

### Matched-sample robustness, follower characteristics, stock penalties, and CEO measures (Section 9)
- Matched-sample mass layoffs (4-digit NAICS matching)
  - Unit: firm-months April 1995–December 2010; Observations: 10,920,656.
  - Matching on size and 4-digit NAICS; matched-pair fixed effects; controls include previous year's log revenue/employees and interactions with recession; month fixed effects.
- Follower characteristics with firm fixed effects (Table A3)
  - Avg. Equity-linked Compensation Share: 0.1296*** (0.0356) in column (1).
  - Total Compensation: -0.0013*** (0.0004) in column (1).
  - CEO Tenure = 0 - 4 years: 0.0178 (0.0122) in column (1).
  - Observations and summary statistics preserved as printed.
- One-month cumulative abnormal returns (Table A4)
  - Firms that announce layoffs (excluding industry leaders), Column (1): Laid off within 1-week after Leader Layoff = 1: 0.0039 (0.0088); Constant 0.0040 (.0029); Observations 3,125.
  - Industry leader firms, Column (2): Industry leader with Follower Layoffs: 0.0229*** (.0068); Observations 1,882.
- Compassionate CEOs and layoff announcements over the business cycle (Table A5)
  - Years at Firm before CEO > 15 years: Years coefficient -0.1005* (0.0530); -0.1240** (0.0573) in specified columns.
  - Recession × Years at Firm before CEO > 5 years: -0.0160 (0.0372) and 0.0058 (0.0435) as printed.
  - Observations reported as 22951694 22951694 (preserved table footer formatting).
  - Mean layoff indicator: 0.1438; Std. Dev .3509.

*Source: _wp16255 - Section 9.*

### Section 1

### Strategic Corporate Layoffs

### Abstract and research question
- Firms in the S&P 500 often announce layoffs within days of one another, despite the fact that the average S&P 500 constituent announces layoffs once every 5 years.
- Similar-sized privately-held firms do not display this high-frequency clustering of layoff announcements.
- The paper investigates whether clustering is driven by CEOs managing their reputation in financial markets and develops a theoretical framework to interpret empirical facts.

### Theoretical framework
- Managers delay layoffs during good economic states to avoid damaging the market’s perception of their ability.
- The model predicts:
  - Clustering in the timing of layoff announcements.
  - An amplification mechanism through which the cyclicality of firms’ layoff policies is intensified.

### Data and empirical strategy
- Two novel datasets are used: one of layoff announcements and one of actual mass layoffs (Mass Layoff Statistics).
- The empirical strategy compares publicly-listed firms to observably similar matched private firms to estimate the impact of reputation-based incentives on the cyclicality of layoffs.

### Key quantitative findings
- Compared to observably similar matched private firms, public firms are 2.5-3.5 percentage points more likely to conduct mass layoffs in a recession month, indicating a doubling of layoff sensitivity to recessions.
- Firms that cluster layoff announcements at high frequencies are also the firms more likely to engage in mass layoffs during recessions.
- The average S&P 500 constituent announces layoffs once every 5 years (preserved phrasing from source).

### Robustness and exclusion of alternative channels
- The documented effects are not driven by:
  - Leverage differences.
  - Lifecycle differences.
  - The criteria used to match public and private firms.

### Interpretation and macroeconomic implications
- Reputation management by CEOs is an important driver of layoff policies both at daily frequencies (clustering of announcements) and over the business cycle (amplified cyclicality).
- These reputation-driven actions can have significant macroeconomic consequences through more procyclical employment adjustments by public firms.

### Policy-relevant conclusions
- Incentives stemming from market reputation influence firm-level layoff timing and cyclicality; policymakers and regulators should account for reputation-driven timing when assessing labor-market dynamics and the macroeconomic impact of layoffs.
- Understanding the role of reputation management can help interpret observed clustering of corporate layoffs and its contribution to employment volatility over the business cycle.

*Ruchir Agarwal and Julian Kolev, IMF Working Paper WP/16/255 (Section 1).*

### Section 2

### _wp16255 - Section 2

### Content characteristics
- The section's text in the supplied PDF is predominantly non-printable and garbled, composed of control-character-like sequences and fragmented tokens rather than coherent prose.
- Repeated exact sequences found verbatim in the section include:
  - " "
  - "	"
  - "#	"
  - "	  "
  - "	A"
- Visible isolated tokens and fragments preserved exactly as they appear:
  - "KA (66F- "
  - "J1L"
  - "I
	N"
  - "G 3"
  - "J"G'EK /\".&KE 2 IEKEL"
  - "5", "4 ", "0", "1", "0", "2"
- The section does not contain continuous sentences, numeric projections, policy recommendations, or clear analytical findings in machine-readable form.

### Implications for ingestion and indexing
- Direct extraction of substantive findings, projections, or policy recommendations is not possible from this section as supplied because coherent natural-language content is absent.
- Automated crawlers and LLM ingestion should treat this section as unreadable/garbled and rely on other sections or canonical metadata for substantive information.
- Exact character sequences above can be used as detectors to flag OCR or encoding failures for this PDF.

*Source: _wp16255 - Section 2 (supplied PDF content) *

### Section 3

### _wp16255 - Section 3

### Extractability summary
- The supplied Section 3 content is dominated by nonstandard control characters, fragmented tokens, and sequences that appear to be corrupted or the result of failed OCR.  
- No coherent paragraphs of substantive findings, projections, analysis, or policy recommendations could be reliably extracted from the supplied text unit.

### Observed content characteristics
- Repeated presence of control or non‑printable characters and isolated punctuation sequences (for example, frequent occurrences of symbols such as , , ).  
- Frequent short fragments of legible words (for example, "Section", "PDF", "B11", "01") interspersed with long runs of unintelligible glyphs.  
- Structural markers (line breaks, occasional capitalization) are present but do not form readable sentences.  
- No consistent numeric statements or intact sentences referencing economic indicators, percentages, dates, or dollar amounts were recoverable.

### Implications for processing and indexing
- Automated indexing, LLM ingestion, or AI crawling should treat this unit as non‑processable in its current form.  
- Downstream natural language tasks (summarization, topic extraction, entity recognition) will yield low‑quality or erroneous outputs if run on this content without remediation.  
- Reliance on this text for policy signals, statistics, or quotations is not advised.

### Recommended next steps
- Source recovery: obtain a clean original of the PDF page or a higher‑quality OCR output for Section 3.  
- If only this file is available, run a robust OCR pipeline with manual verification (human review of OCR zones) to reconstruct disrupted text segments.  
- Validate recovered text against other sections or the document’s table of contents to ensure logical placement and consistency before using for analysis or indexing.

*Source: _wp16255 - Section 3 — https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf*

### Section 4

### _wp16255 - Section 4

### Content quality and readability
- The provided text for Section 4 is largely unreadable and appears to contain extensive non-printable/control characters, fragmentary strings, and corrupted formatting.
- Examples of verbatim fragments from the supplied content:
  - "	      
 '	  
"
  - "6DB 	 "
  - "011 
 
"
  - "3)@ 
 /	   
 "
- The file contains many isolated alphanumeric strings and short sequences (for example "J\"G'EK %)3E 0 IEKEL") that do not form coherent sentences or identifiable analytical content.

### Findings, projections, and substantive content
- No coherent findings, quantitative projections, analytical narratives, or policy recommendations can be reliably extracted from the supplied text due to corruption of the content.
- There are scattered numeric and alphanumeric tokens present (e.g., "00", "01", "02", "3)@", "5", "0111", "21Q", "B11") but they are presented without contextual explanation; therefore no valid interpretation, preservation, or reassembly into substantive statistics or projections is possible from the supplied fragmentary sequences.

### Policy recommendations and actionable guidance
- The damaged state of the supplied section precludes identification of any explicit policy recommendations or actionable guidance within this content unit.
- Recovery actions (for the content pipeline/operators):
  - Re-ingest a clean copy of Section 4 from the canonical source.
  - Verify the PDF encoding and character set prior to automated extraction to avoid non-printable/control-character corruption.
  - If the canonical PDF is intact, replace the corrupted extract with a fresh text extraction run.

*Source PDF filename: _wp16255 - Section 4 — https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf*

### Section 5

### _wp16255 - Section 5

### Content notes
- The supplied text for this content unit contains extensive non‑printable characters, transcription artifacts, and garbled passages making the substantive findings, projections, analysis, and policy recommendations unreadable in their present form.
- The document as provided does not contain a coherent sequence of sentences or clear sections that can be reliably summarized for AI ingestion or search indexing.

### Text quality and extraction issues
- The section includes repeated control characters and fragmented tokens (for example sequences of symbols and isolated letter groups) that interrupt readable text flow.
- Several numeric fragments and short digit sequences appear in the raw stream but are not embedded in coherent statements and therefore cannot be interpreted as findings, projections, or policy parameters.

### Recommended next steps
- Supply a clean, machine‑readable version of Section 5 (for example a high‑quality PDF with embedded text or a plain‑text transcription) so that substantive content (findings, projections, analysis, and policy recommendations) can be extracted and summarized accurately.
- If only a scanned image version is available, provide a PDF that has been processed with OCR and reviewed for text recognition errors.

*Canonical source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf*

### Section 6

### _wp16255 - Section 6

### Content quality
- The supplied text for Section 6 is heavily OCR-corrupted and contains nonstandard characters, fragmented lines, repeated control characters, and uninterpretable segments.
- No coherent sentences, findings, projections, quantitative results, or policy recommendations can be reliably extracted from the supplied content.

### Extraction outcome
- Unable to identify substantive findings, projections, analysis, or policy recommendations from the supplied content due to pervasive corruption and illegibility.
- No numeric values, dates, or currency amounts could be confidently located or preserved from the text provided.

### Recommended next steps for ingestion (based solely on supplied content)
- Source text appears unreadable in this unit; obtain a clean copy of Section 6 (original PDF or higher-quality OCR) before attempting substantive extraction or summarization.
- Re-run OCR or text extraction on a higher-quality input to recover technical terminology, exact numeric values, and formal structure required for AEO/GEO/SEO overlays.

*Source: _wp16255 - Section 6 — https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf*

### Section 7

### _wp16255 - Section 7

### Empirical patterns in layoff announcements and clustering
- Figure 1: Timeline of layoff announcements by top 3 firms in the banking and automotive industry between 2001 and 2003. A dot represents a layoff announcement on a given day. Orange boxes of width 30 days mark periods with multiple announcements within 30 days; longer boxes indicate cross-industry clustering, shorter boxes indicate within-industry clustering.
- Figure 2: Reports Log p-values (multiplied by negative unity) of sequence of tests for each non-overlapping 60 (business) day window to identify clustering at 1, 5, 10 or 15 day horizons. Null hypothesis: layoff announcements are distributed uniformly over each 60 day window. Horizontal red line = significance level of 0.05; p-values exceeding the red line allow rejection at 5% level.

### Theoretical parameter space and comparative statics
- Figure 3: Parameter Space and Equilibrium Layoff Policy.
  - Region to the right of solid lines: equilibrium layoff policy = delay.
  - Region to the left of solid lines: equilibrium layoff policy = terminate.
  - Region between the two lines: equilibrium policy = mixed strategy (randomizes between terminate and delay).
- Figure 4: Comparative statics when belief about aggregate state worsens.
  - Dotted lines show mapping of parameter-space regions to equilibrium policies before comparative statics.
  - When belief about aggregate state worsens, separating lines shift rightward (represented by solid lines), changing the equilibrium-region boundaries.

### Event study evidence: leader layoffs and media events
- Figure 5: Response of layoff propensities to "Leader Layoffs" (layoff announcements by the largest 20 firms by prior-year revenue).
  - Coefficients from dynamic regressions predicting daily layoff propensities over 11-day event-time periods surrounding large-firm layoff announcements.
  - Controls: firm-level controls; fixed effects at annual, monthly, and day-of-week levels; standard errors clustered at the firm level; point-wise 95% confidence intervals reported.
  - Panels:
    - Panel A: response of layoff propensity for public Fortune-500 firms.
    - Panel B: response for Forbes-100 (private) firms.
    - Panel C: Panel A results restricted to firms in the same 1-digit NAICS industry as the leader firm.
    - Panel D: response of all public Fortune-500 firms to "News Media Events" (biggest negative news stories based on press coverage from 1970 to 2010).

### Descriptive statistics: Public vs Private firms (Table 1 highlights)
- Revenue (in 2005 USD)
  - Public Mean 15143.298067.537064.76***; Private Mean 7162.076992.29169.78; Matched/Buyout sample means also reported (9218.318609.35608.97).
  - Median: Public 6035.374406.83; Private 5075.344903.13; Matched/Buyout medians reported.
  - St. Dev: reported with multiple precise values (e.g., 30540.0510817.585122.214951.1414895.956995.246).
- Employees
  - Public Mean 46803.0136624.1510178.86***; Private Mean 29341.6629614.29-272.63; additional sample moments reported (e.g., Median 2158617649; St. Dev 98809.9057237.1320212.4426967.7868204.0444141.21).
- Layoff Propensity
  - Public Mean 0.07210.05410.0180***; Private Mean 0.05840.05580.0027; St. Dev reported (0.25870.2263 etc.).
- Number Laid off
  - Public Mean 337.74357.88-20.14; Private Mean 281.59248.8232.77; St. Devs and medians provided (Median 174200; 160150; 166191).
- Share Laid Off
  - Public Mean 0.01340.0165-0.0031; Private Mean 0.01380.0169-0.0031; Medians and St. Devs reported.
- Sample sizes
  - No. of Firms: values include 478135206744223 (table reports separate counts across columns).
  - No. of Observations: values include 7066510632, 1047010470, 55021803 (reported per sample column).
- Note: Table 1 reports means, medians, standard deviations, and pairwise differences in means with *** and ** denoting significance at the 1% and 5% levels. Samples cover 1995 to 2010 and include full samples, size-and-industry matched samples, and LBO-targeted samples.

### Public–Private differences over the business cycle (Table 2 key coefficients)
- Dependent variables: Layoff Indicator (mass layoff indicator in a month) and Share Laid Off (number laid off as share of prior year employees).
- Regression sets: four sets with two dependent variables each; specifications vary by fixed effects and time controls (industry 4-digit NAICS; quadratic time trend & seasonality; month fixed effects; firm fixed effects; matched-pair fixed effects).
- Selected coefficient estimates (preserve exact figures and standard errors where reported):
  - Public Indicator: 0.00510.0020-0.00290.0029 -0.0064-0.0069***-0.0044-0.0060 (standard errors in parentheses in table: (0.0085)(0.0013)(0.0127)(0.0017)(0.0090)(0.0025)(0.0063)(0.0037)).
  - Recession Indicator: 0.0201**-0.0011 0.0240**0.0008 (standard errors (0.0087)(0.0015)(0.0161)(0.0014)).
  - Public x Recession interaction: 0.0247***0.0028*0.0259***0.0025 0.0320**0.0040*0.0307**0.0012 (standard errors (0.0097)(0.0015)(0.0090)(0.0018)(0.0136)(0.0022)(0.0142)(0.0028)).
- Sample moments reported in Table 2:
  - Mean values: 0.06960.01370.06960.0137 0.05690.01550.05690.0155
  - Std. Dev values: 0.25440.07850.25440.0785 0.23180.02670.23180.0267
- Observations: 814145587814145587 and 205921167205921167 (reported across columns).
- Controls included across specifications: Industry Fixed Effect (4-digit NAICS), Quadratic Time Trend & seasonality controls, Month Fixed Effects, Firm Fixed Effects, Matched-Pair Fixed Effects, Log Employees and Recession Interaction, Log Revenue and Recession Interaction. Heteroskedasticity-consistent standard errors clustered at the firm level are reported under coefficient estimates. Significance markers ***, **, * denote 1%, 5%, and 10% levels.

### Matched-sample business-cycle totals and durations (Table 3 highlights)
- Dependent variables in matched-sample analysis: total workers laid off over business cycle; median months since last layoff.
- Coefficients:
  - Public Indicator in (1): 0.30390.0656* with (0.4251)(0.0363) as reported standard errors.
- Sample moments reported:
  - Mean: 2.57457.76
  - Std. Dev: 3.10776.69
- Observations (matched sample): 236260
- Notes: Specification (1) measures total workers laid off by a firm over a business cycle (peak-to-peak from 6 months prior to the 2001 recession peak [October 2000] to 6 months prior to the 2008 recession peak [July 2007]); each firm has one observation; matched-pair fixed effects control for pair-level unobservables. Specification (2) uses trough-to-trough period to study median distance between mass layoffs for each firm; also controls for matched-pair fixed effects. Standard errors clustered at firm level; significance markers used as above.

### Additional regression results (Table 4 excerpt)
- Reported coefficients:
  - Public Indicator: -0.0079-0.0088 (standard errors (0.0261)(0.0272))
  - Public x Recession: 0.0604**0.0181 (standard errors (0.0258)(0.0176))
  - Mean: 0.06920.0142
- Interpretation: Public x Recession interaction positive and in some specifications statistically significant, indicating public firms' layoff measures respond differently during recessions relative to private firms in the reported specifications.

*Source: _wp16255 - Section 7 (PDF: _wp16255 - Section 7).*

### Section 8

### _wp16255 - Section 8

### LBO Targets: Assessing Alternate Explanations for Public - Private Differences in Cyclicality of Mass Layoffs (Table 5)
- Unit of observation: firm-months between April 1995 and December 2010; dependent variable: layoff indicator and share laid off (mass layoff as per BLS Mass Layoff Statistics Program).
- Sample: firms that were targets of a leveraged buyout (LBO). Public firms = targets for whom LBO offer was withdrawn; private firms = targets for whom buyout offer was successful.
- Estimation details: month fixed effects included; controls for previous year's log revenue and its interaction with recession dummy, and previous year's log employees and its interaction with recession dummy (except when dependent variable is share laid off). Heteroskedasticity-consistent standard errors clustered at the firm level.

Key coefficient estimates (selected columns reported):
- Public Indicator:
  - Column (1): 0.0095 (0.0154)
  - Column (2): 0.0055 (0.0231)
  - Column (3): 0.0009 (0.0139)
  - Column (4): 0.0071 (0.0196)
- Public × Recession:
  - Column (1): 0.0478*** (0.0127)
  - Column (2): 0.0133 (0.0099)
  - Column (3): 0.0187* (0.0108)
  - Column (4): 0.0250** (0.0108)
- Additional covariates:
  - Leverage Ratio (in one specification): 0.0721*** (0.0137)
  - Log(1+Years since IPO): 0.0005 (0.0136)
  - Recession × Lev. Ratio: 0.0259 (0.0190)
  - Recession × Log Age: -0.0341*** (0.0116)
- Sample statistics (for displayed columns):
  - Mean: 0.0567, 0.0785, 0.0602, 0.0759, 0.0718
  - Std. Dev: 0.2314, 0.2689, 0.2379, 0.2639, 0.2582
- Fixed effects and controls varied across specifications: firm fixed effects in one column; industry fixed effects (4-digit NAICS) and month fixed effects present in many specifications.
- Observations reported across columns (examples): 4480547, 0194507, 8467466, 9621 (as printed).

### Actual Layoff Propensity over the Business Cycle (Public Firms), 1995-2010 (Table 6)
- Purpose: explore alternate explanations for public-private differences in layoff propensity; dependent variable = layoff indicator.
- Sample splits:
  - Column 1: restrict public firms to 'young' firms (time-since-IPO above median within calendar year).
  - Column 2: restrict public firms to 'old' firms.
  - Column 3: restrict to 'high leverage' public firms (leverage ratio above median).
  - Column 4: restrict to 'low leverage' public firms.
  - Column 5: public firms only; interaction between recession and both log years since IPO and leverage ratio.
- Controls: month fixed effects; log previous year's employees and its interaction with recession; log previous year's revenue and its interaction with recession. First four include industry fixed effects; last includes firm fixed effects.

Key coefficient estimates:
- Public Indicator (various samples):
  - Column (1): -0.0194** (0.0090)
  - Column (2): -0.0117*** (0.0041)
- Public × Recession:
  - Column (1): 0.0312* (0.0185)
  - Column (2): 0.0115** (0.0048)
- Sample statistics:
  - Mean: 0.0569, 0.0155 (as printed)
  - Std.: (not completed in printed excerpt)

### Executive Compensation and CEO Tenure: Industry Variation in Layoff Propensity (Table 6 continued / Table labels)
- Unit of observation: firm-months between April 1995 and December 2010; sample: Fortune 500 firms that ever announced a layoff (daily tracking described).
- Key variables:
  - KV = Short CEO Tenure Indicator (tenure between 0 and 4 years).
  - KV = Avg. Equity-Linked Compensation Share (Black-Scholes past five year average of value of stock-option grants as a share of average total compensation).
  - Recession indicator: months classified as recession months by NBER Business Cycle Dating Committee.
- Controls: month fixed effects, industry fixed effects, firm-level controls (Earnings-Price ratio, Book-to-Market Ratio, Market Capitalization, Leverage Ratio, Years since IPO) and their interactions with recession.

Key coefficient estimates (reported across specifications):
- KV (Short CEO Tenure Indicator):
  - 0.0032 (0.0050)
  - 0.0025 (0.0062)
- KV (Avg. Equity-Linked Compensation Share):
  - 0.0072 (0.0071)
  - 0.0164 (0.0091)
- Key Variable × Recession:
  - 0.0144* (0.0079)
  - 0.0261** (0.0126)
  - 0.0255** (0.0123)
  - 0.0352** (0.0168)
- Sample statistics:
  - Mean: 0.0787, 0.0790, 0.0812, 0.0831
  - Std. Dev: 0.2693, 0.2697, 0.2732, 0.2761
- Observations reported: 28746, 27882, 16818, 13410 (as printed)

### Stock Penalty of Layoffs: Cumulative Abnormal Returns (Table 7)
- Sample: daily observations from 1970 to 2010; dependent variable: Cumulative Abnormal Return (CAR) from day -2 to day +2 around layoff announcement.
- Key variable: indicator = 1 if firm announces a layoff within three days after a layoff announcement by any of the largest 20 firms (as measured by previous year's revenue); specifications include indicators for top-20 firm layoff within next 3 days (and other fixed effects).

Panel A key estimates:
- Layoff by Top 20 Firm in Previous 3 days =
  - Column (1): 0.0049*** (.0017)
  - Column (2): 0.0052*** (.0019)
  - Column (3): 0.0042** (.0020)
  - Column (4): 0.0044** (.0021)
- Layoff by Top 20 Firm in Next 3 days =
  - Reported as 1-0.0025 and -0.0027 (.0029) (.0029) (as printed; interpretation limited by excerpt)
- Constant (Baseline CAR):
  - -0.0087*** (0.0013)
  - -0.0081*** (0.0016)
- Fixed effects: year fixed effect, industry fixed effect in some specifications.
- Observations: 4796 in reported columns.

Panel B key estimates (cross-effects):
- CAR of Industry Leader that Laid off Within Last 3 Days when Follower Firms Lay off:
  - 0.0033*** (.0011)
  - Observations: 212
- CAR of Other Firms in Industry when Industry Leader Lay off:
  - -0.0003 (.0007)
  - Observations: 8286

### Follower Characteristics, 1970-2010 (Tables 8 and 9)
- Sample: Fortune 500 constituents excluding largest 20 by revenue; unit: firm-day observed 1970–2010.
- Dependent variables: follower indicator (firm announces layoff in a 5-day window following a large firm layoff, including day of large layoff) and counterfactual follower indicator (firm announces layoff in 5-day window prior to large layoff, excluding day of large layoff).
- Specifications: pairs estimated on (1) all firms, (2) all firms, (3)-(4) firms that laid off in given year, (5) firms that laid off within 5 days before/after large firm layoff. Year fixed effects and firm-level controls included; standard errors clustered at firm level.

Table 8 key coefficient estimates:
- Avg. Equity-linked Compensation Share:
  - Column (1): 0.0741*** (0.0186)
  - Column (2): 0.0186 (0.0165)
  - Column (3): 0.1850 (0.1546)
  - Column (4): -0.10860 (0.1415)
  - Column (5): 0.4689* (0.2833)
- Total Compensation:
  - Column (1): -0.0014*** (0.0004)
  - Column (2): 0.0004 (0.0011)
  - Column (3): -0.0043 (0.0029)
  - Column (4): 0.0110*** (0.0025)
  - Column (5): -0.0073* (0.0038)
- CEO Tenure = 0 - 4 years:
  - Column (1): 0.0229*** (0.0075)
  - Column (2): -0.0042 (0.0056)
  - Column (3): 0.1458** (0.0583)
  - Column (4): -0.0945* (0.0496)
  - Column (5): 0.2534*** (0.0928)
- Sample statistics (selected):
  - Means: 0.0292, 0.0176, 0.2764, 0.1645, 0.6238
  - Std. Devs: 0.1685, 0.1316, 0.4482, 0.3715, 0.4869
- Observations across columns: 21512, 21512, 28228, 22810, 1 (as printed — last entry likely truncated).

Table 9 key coefficient estimates (analyst coverage measure):
- Analyst Coverage in Past 3 years:
  - Column (1): -0.0254*** (0.0093)
  - Column (2): -0.0068 (0.0060)
  - Column (3): -0.2061** (0.0898)
  - Column (4): -0.0476 (0.0777)
  - Column (5): -0.1222 (0.1935)
- Sample statistics:
  - Means: 0.0361, 0.0246, 0.2464, 0.1836, 0.5940
  - Std. Devs: 0.1864, 0.1550, 0.4439, 0.3874, 0.4916
- Fixed effects: firm fixed effects in some specifications; year fixed effects and firm-level controls included.
- Observations reported: 71857, 18596, 49649, 64436 (as printed).

### Linking Daily-Frequency Layoff Behavior to Business Cycle Outcomes, 1995-2010 (Table 10)
- Unit: firm-months tracked April 1995–December 2010; dependent variable: layoff indicator.
- Key variables:
  - Past 5 year Follower Indicator (1 if firm announced a layoff in a 5-day window following a large firm layoff at least once in past five years).
  - Past 5 year Counterfactual Follower Indicator (1 if firm announced a layoff in a 5-day window prior to a large firm layoff at least once in past five years).
  - Recession × Past 5 year indicators included.
- Controls: month fixed effects; firm fixed effects; firm-level controls and their interactions with recession.

Key estimates:
- Past 5 year Follower Indicator:
  - 0.0135*** (0.0035)
- Past 5 year Counterfactual Follower Indicator:
  - 0.0120*** (0.0036)
- Recession × Past 5 year Follower Indicator:
  - 0.0314*** (0.0085)
- Recession × Past 5 year Counterfactual Follower Indicator:
  - 0.0041 (0.0095)
- Sample statistics:
  - Mean: 0.0722, 0.0722
  - Std. Dev: 0.2587, 0.2587
- Observations: 70170, 70170 (as printed)

### Fisher's Method for Combining Results from Independent Scan Statistics (Table A1)
- Purpose: combine p-values from independent scan statistics computed for non-overlapping intervals of 20 or 60 business-days between 1970 and 2010; null = uniform layoff propensity across windows.
- For each sub-window, reported: Window (days), Sub-window (days), # of months (number of individual tests), Test Statistic (Fisher's Method), Degrees of Freedom, p-value (Fisher's Method).
- Selected results (as printed):
  - 20 / 13 / 250 / 7.9965 / 0 / 1.000
  - 20 / 33 / 256 / 26.826 / 0.736
  - 20 / 53 / 270 / 8.6765 / 0.055
  - 20 / 103 / 2578 / 1.4165 / 0.0003
  - 60 / 11 / 0920 / 2.262 / 180 / 0.771
  - 60 / 31 / 0925 / 5.702 / 180 / 0.041
  - 60 / 51 / 0925 / 4.522 / 180 / 0.045
  - 60 / 101 / 0926 / 8.862 / 180 / 0.011
  - 60 / 151 / 0928 / 8.032 / 180 / 0.001
  - 20 / 125 / 2374 / 4.275 / 041 / 1.000
  - 20 / 325 / 2437 / 2.205 / 040 / 0.986
  - 20 / 525 / 2551 / 1.125 / 040 / 0.072
  - 20 / 102 / 5263 / 1.325 / 040 / 0.0001
  - 60 / 183 / 181 / 1.261 / 660 / 0.198
  - 60 / 383 / 192 / 2.811 / 660 / 0.076
  - 60 / 583 / 198 / 5.531 / 660 / 0.043
  - 60 / 108 / 3213 / 3.821 / 660 / 0.007
  - 60 / 158 / 3212 / 4.481 / 660 / 0.009
- Notes: top half covers all layoffs; bottom half restricts to layoffs in 3-digit NAICS industries. Degrees of freedom = 2 × number of individual tests. Reject null of no excess clustering if combined p-value < 0.05.

### Robustness Tests: Public–Private Comparison over the Business Cycle (Table A2)
- Selected coefficient estimates (as printed):
  - Public Indicator:
    - Column (1): -0.0194** (0.0090)
    - Column (2): -0.0117*** (0.0041)
  - Recession Indicator: (value not printed fully in excerpt)
  - Public × Recession:
    - Column (1): 0.0312* (0.0185)
    - Column (2): 0.0115** (0.0048)
- Sample means reported: 0.0569, 0.0155 (as printed)

*Source: _wp16255 - Section 8 (PDF: _wp16255 - Section 8).*

### Section 9

### _wp16255 - Section 9

### Robustness: Matched Sample Mass Layoffs (4-digit NAICS matching)
- Unit of observation: firm level tracked monthly between April 1995 and December 2010.
- Dependent variables:
  - Layoff indicator: equals 1 if the firm engages in a mass layoff in a given month (Mass Layoff Statistics Program definition).
  - Share laid off: number of workers laid off as a share of previous year's employees.
- Matching: matched public-private pairs based on size (revenue) and 4-digit NAICS industry; matched-pair fixed effects included.
- Controls included:
  - Previous year's log revenue and its interaction with the recession dummy.
  - Previous year's number of employees and its interaction with the recession dummy (except when dependent variable is share laid off).
  - Month fixed effects (so recession indicator is not separately identified).
  - Heteroskedasticity-consistent standard errors clustered at the firm level.
- Key sample statistic:
  - Observations: 10,920,656
- Coefficients / model markers quoted in table header (as presented):
  - Dev 0.23180.0267
  - Quadratic Time Trend & seasonality controls
  - Month Fixed Effects
  - Matched-Pair Fixed Effects
  - Log Employees and Recession Interaction
  - Log Revenue and Recession Interaction
- Notes: Significance notation ***, **, and * denote significance at the 1%, 5%, and 10% level (two-sided), respectively.

### Follower Characteristics, 1970-2010 (Table A3; robustness with firm fixed effects)
- Unit of observation: firm level observed every business day between 1970 and 2010.
- Sample: Fortune 500 constituents excluding the largest 20 firms by prior-year revenue.
- Dependent variables (indicator variables):
  - Follower indicator: equals 1 if in a given year the firm announces a layoff in a 5-day window following a large firm layoff (including the day of the large layoff).
  - Counterfactual follower indicator: equals 1 if in a given year the firm announces a layoff in a 5-day window prior to a large layoff (excluding the day of the large layoff).
- Specifications:
  - (1) and (2): entire sample (all Fortune 500 firms not in the top 20).
  - (3) and (4): only firms that laid off in a given year.
  - (5): firms that laid off within 5 days before/after a large firm layoff (two dependent variables perfectly negatively correlated; only one regression estimated).
  - All specifications include firm fixed effects, year fixed effects, and firm-level controls.
  - Heteroskedasticity-consistent standard errors clustered at the firm level.
- Reported coefficients and standard errors (columns (1)–(5)):
  - Avg. Equity-linked Compensation Share: 0.1296*** (0.0356), 0.0117 (0.0282), 1.2822*** (0.3119), -0.7038** (0.2738), 4.1088*** (0.4515)
  - Total Compensation: -0.0013*** (0.0004), 0.0009 (0.0014), 0.0006 (0.0047), 0.0105** (0.0046), -0.0343* (0.0179)
  - Ceo Tenure = 0 - 4 years: 0.0178 (0.0122), -0.0062 (0.0091), 0.3022* (0.1832), -0.1993 (0.1538), 0.9573*** (0.2286)
- Summary statistics shown:
  - Mean: 0.0292, 0.01763, 0.2764, 0.1645, 0.6238
  - Std. Dev: 0.1685, 0.1316, 0.4482, 0.3715, 0.4869
- Observations by column: 21512, 15122, 8228, 28101 (as presented); footer shows Observations 21512 15122 8228 28101 (table formatting preserved).
- Notes: Significance notation ***, **, and * denote significance at the 1%, 5%, and 10% level (two-sided), respectively.

### Stock Penalty of Layoffs: One-Month Cumulative Abnormal Returns (Table A4)
- Unit of observation: daily observations from 1970 to 2010.
- Dependent variable: Cumulative One Month Return of Firms that Lay off.
- Samples reported:
  - Firms that Announce Layoffs (Excluding Industry Leaders)
  - Industry Leader Firms
- Column (1): Firms that announce layoffs and are not industry leaders (based on prior-year revenue).
  - Laid off within 1-week after Leader Layoff = 1: 0.0039 (0.0088)
  - Laid off within 1-week before Leader Layoff = 1: -0.0168 (0.0113)
  - Constant (Baseline CAR): 0.0040 (.0029)
  - Observations: 3,125
- Column (2): Industry leader firms
  - Industry leader with Follower Layoffs: 0.0229*** (.0068)
  - Constant (Baseline CAR): 0.0023 (.0022)
  - Observations: 1,882
- Standard errors: Heteroskedasticity-consistent, clustered at the industry level.
- Notes: Significance notation ***, **, and * denote significance at the 1%, 5%, and 10% level (two-sided), respectively.

### Compassionate CEOs and Layoff Announcements over the Business Cycle, 1970-2010 (Table A5)
- Unit of observation: firm level tracked yearly between 1970 and 2010.
- Sample: contemporaneous constituents of Fortune 500 that ever announced a layoff.
- Dependent variable: Layoff Indicator (equals 1 when a firm announces a layoff in a given year).
- Compassionate CEO measures:
  - Indicator = 1 if tenure of given CEO at the firm before appointment > 5 years (specifications (1) and (2)).
  - Indicator = 1 if tenure of given CEO at the firm before appointment > 15 years (specifications (3) and (4)).
- Specification differences:
  - (1) and (3): all firms in sample.
  - (2) and (4): only firms with "home-grown" CEOs (no externally-appointed CEOs in the given year).
  - All specifications include year-fixed effects, firm-fixed effects, firm-level controls, and interaction of controls with recession indicator.
  - Firm-level controls: Earnings-Price ratio, Book-to-Market Ratio, Market Capitalization, Leverage Ratio, Years since IPO (from COMPUSTAT).
  - Year-fixed effects mean recession indicator is not separately identified.
  - Heteroskedasticity-consistent standard errors clustered at the firm level.
- Reported coefficients and standard errors:
  - Recession x Years at Firm before CEO > 5 years: -0.0160 (0.0372), 0.0058 (0.0435)
  - Years at Firm before CEO > 5 years: -0.0011 (0.0554), -0.0194 (0.0588)
  - Recession x Years at Firm before CEO > 15 years: -0.0129 (0.0413), 0.0048 (0.0437)
  - Years at Firm before CEO > 15 years: -0.1005* (0.0530), -0.1240** (0.0573)
- Summary statistics shown:
  - Mean: 0.1438 (all columns)
  - Standard Deviation: .3509 (all columns)
- Observations reported: 2,295,1694 2,295,1694 (presented as 22951694 22951694 in table footer formatting)
- Notes: Significance notation ***, **, and * denote significance at the 1%, 5%, and 10% level (two-sided), respectively.

*Source: _wp16255 - Section 9*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16255.pdf_