Page 768 - ISC PROCEEDINGS 21.4

commercial real estate demand, and municipal tax revenues, ultimately amplifying the
initial employment crisis through regional feedback loops.
3. Research methodology
3.1. Data source and collection
This study utilizes a comprehensive dataset compiled and maintained by Layoffs.fyi
(credited to Roger Lee), which tracks global technology layoffs from March 11, 2020—the
date when COVID-19 was declared a pandemic by the World Health Organization—
through April 21, 2025. The dataset aggregates publicly reported layoff events
systematically tracked through multiple authoritative business media sources, ensuring
comprehensive coverage and cross-validation of reported events. The primary data
sources include Bloomberg’s business and technology coverage, which provides detailed
reporting on public company restructuring announcements; San Francisco Business
Times' focused coverage of Bay Area technology companies; TechCrunch’s extensive
startup and venture-backed company reporting; and The New York Times' business
section coverage of major corporate developments. To ensure data integrity, the raw
dataset underwent a rigorous validation process. Incomplete entries lacking specific layoff
figures or verifiable company identities were excluded. Furthermore, large-scale layoff
events (exceeding 1,000 employees) were cross-referenced with official corporate
announcements, such as SEC 8-K filings or official press releases, to confirm the accuracy
of the reported scale and timing.
3.2. Analytical framework and computational tools
This analytical approach employs descriptive statistics and advanced Python-based
visualizations to ensure a transparent, reproducible, and insightful study. Utilizing a
robust data science ecosystem—including Pandas and NumPy for manipulation, and
Matplotlib, Seaborn, and Plotly for static and interactive graphics—all workflows were
documented in Jupyter notebooks to facilitate replicability. Given the unprecedented
convergence of a global pandemic recovery and the sudden emergence of generative AI,
predictive modeling faces significant limitations due to the lack of historical analogues.
Therefore, an exploratory descriptive approach was justified to map the structural
boundaries of this novel phenomenon before advancing to causal inferences.
3.3. Research limitations
This study acknowledges several methodological limitations. First, the Layoffs.fyi
database relies predominantly on publicly reported events, introducing media and
survivorship bias. “Quiet layoffs” or smaller reductions in early-stage startups often go
unreported, potentially underestimating the true scale of job losses. Second, there is a
geographic reporting bias; English-language media heavily covers the United States and
European markets, which may skew the geographic concentration findings. Finally, the
exact strategic rationale for each layoff event (e.g., AI pivot vs. pure financial distress)
cannot be definitively isolated for every company, necessitating reliance on aggregate
macroeconomic indicators.
4. Results and discussion
4.1. Industry-level analysis and sectoral vulnerabilities
Between March 2020 and April 2025, 808,091 employees were laid off across 2,863
companies. The distribution is heavily right-skewed, with a median of 65 layoffs per event
despite an average of 282, a pattern reflecting the power law distributions typical of firm
dynamics (Autor et al., 2003). While the early pandemic saw only 6% of total layoffs as
technology firms benefited from digital acceleration (Brynjolfsson et al., 2020), a massive

767