← All projects

Tech Industry Layoff Analysis

"Private companies cut 32.4% of their workforce per layoff vs. 16.7% at public companies — a 2x gap, confirmed by a significance test across 2,723 layoff events"

Course: STAT 3080 (spring 2026)
Tools: R, tidyverse, ggplot2, knitr, R Markdown
Team: 3-person team, GitHub version control
Data: 2,723 layoff events from Layoffs.fyi (2020–2026), with company type derived from funding stage (Post-IPO = public, Seed and Series = private)
Methods: a one-sided Wilcoxon rank-sum test, chosen over a t-test because layoff percentages are bounded and heavily right-skewed; descriptive comparison of four industries (Finance, Healthcare, Retail, Consumer)
Headline finding: per layoff, private companies cut twice the share of their workforce that public companies do (mean 32.4% vs. 16.7%, median 20% vs. 10%); the test puts the typical shift at about 8 percentage points, and at least 7 with 95% confidence
Deliverable: two reproducible R Markdown reports

Why this question, and what surprised us

Going in, we expected some gap between how public and private companies handle layoffs — public companies answer to shareholders and analysts every quarter, private companies don't. What surprised us was the size of it: private companies cut twice the share of their workforce (32.4% vs. 16.7% on average), and the gap held up under a formal test, not just as a trend in a bar chart. Our explanation is funding: a startup whose next round falls through may have to cut deeply all at once, while a public company has access to capital markets and shareholders watching for measured cuts. The other surprise was Healthcare. Its public companies cut nearly as hard as its private ones, a gap of just 6.1 points against 15 to 26 in the other industries, which suggests pressure in digital health and biotech was strong enough to override that public-market restraint.

My contribution

I handled all of the coding for the project: cleaning the raw data and engineering features from it, the Wilcoxon rank-sum test, the ggplot2 charts, and the reproducible R Markdown reports.