I just completed a comprehensive statistical validation of this dataset using all 49,467 confirmed corporate deaths (1992-2025).
KEY FINDINGS: • 59.0% of delisted stocks lost >50% of their value before death (HIGHLY SIGNIFICANT, p<0.001) • Median price destruction: 66.6% • 33.2% lost >90% of value, 16.7% suffered >99% wipeouts
TEMPORAL TRENDS: • 1990s: 13,709 deaths, 60.8% median destruction • 2000s: 11,985 deaths, 71.7% median destruction • 2010s: 9,858 deaths, 73.7% median destruction ← PEAK SEVERITY • 2020s: 14,915 deaths, 66.7% median destruction ← MOST DEATHS
IMPLICATION: Corporate death is NOT sudden in price space—it's systematically foreshadowed by deep, sustained price deterioration. This validates the Three-Phase Mortality Model and has major implications for: - Risk management & distress prediction - Portfolio construction & drawdown dynamics - Early-warning systems for quantitative equity portfolios
The destruction severity has INCREASED over time (60.8% → 73.7%), suggesting market efficiency improvements make corporate decline more visible and systematic.
Full analysis: 6 cells of statistical validation + 4-panel visualization created in Google Colab, bypassing all Drive mount issues by downloading directly from GitHub.
I conducted a comprehensive literature review on corporate mortality patterns using Perplexity AI. The academic research strongly validates the Three-Phase Mortality Model:
Price Decline Before Failure: Multiple event-study papers document substantial negative abnormal returns in quarters/years leading to bankruptcy. Dawkins et al. report sharp "plunge" behavior around Chapter 11 filings. Japanese/TSE/NASDAQ studies show 1-year buy-and-hold returns before delisting of ~60% losses on average, with median losses of ~65% for main-board firms and ~84% for smaller markets.
Three-Phase Models: Several academic frameworks view failure as multi-phase rather than single-event. Recent JSF research categorizes firms as healthy→depressed→distressed, building three-phase models. Structural Merton-tradition models interpret phases as: asset value drift toward default, equity behaves like deep out-of-the-money option, intensifying volatility sensitivity.
Destruction Severity Across Eras: Literature confirms median pre-delisting losses in 50-70% range. Post-2000s studies show attenuation suggesting information flow evolution. The 2010s exhibiting MORE severe median destruction (73.7%) vs 1990s (~61%) aligns with findings that later-period failures show greater price destruction due to smaller, more distressed marginal firms.
Market Microstructure: Studies since 1990s document how decimalization, electronic trading, and high-frequency activity have altered distress manifestation. Flash Crash work highlights self-exciting dynamics and order flow. Delisting leads to tripled percentage spreads, doubled volatility, large depth drops even with high OTC volume.
KEY INSIGHT: The 59% rate of >50% value loss before delisting and 66.6% median destruction are highly consistent with documented one-year pre-delisting losses and event-window returns across multiple markets. This extends the literature by providing large-scale, multi-decade, all-cause delisting perspective.
Full Perplexity research: https://www.perplexity.ai/search/conduct-a-comprehensive-lit...
New_Person•1mo ago
I spent the last week analyzing a proprietary dataset of 49,315 delisted US stocks to understand "survivorship bias" from a microstructure perspective. Standard backtests usually ignore these companies, but I wanted to see what the order book looks like right before a firm goes to zero.
I built a pipeline to index 84GB of minute-level OHLCV data and cluster the failures using K-Means.
Key Finding: "Type III: The Zombie Churn." Stocks that have already lost 90%+ of their value, but volume explodes to 48x normal levels while price stays flat. It looks like a distinct signature of retail bag-holding vs. institutional exit.
The repo has the indexer script, the clustering logic, and the "Death Metrics" CSV for the top 1,000 failures (including Lehman Brothers and Enron).
Happy to answer questions about the parquet engineering or the metrics used!