Research · September 2026

Which sectors fail, which fall, and which compound

We followed 4,618 companies listed on the NYSE, Nasdaq and NYSE American that were worth at least $300 million in July of any year from 2015 to 2021, including 1,429 that have since been delisted, and counted how often each sector lost half its value, went bankrupt, or doubled.

The sectors where a single stock most often lost half its value within five years were biotech (49.5%), advertising (44%), healthcare software (42%), oilfield services (39%), internet services (36%) and pharma (34%). Across all stocks the figure was 18%.
Chip-equipment makers, homebuilders, wholesale distributors, banks and utilities rarely lost half (1–6%).
Bankruptcy is a different list. Among all 6,700 companies filing annual reports with the SEC in 2013–2015, oilfield services (2.6% a year) and printing and publishing (2.5%) failed most often, followed by healthcare services, oil and gas producers and telecom (1.6–1.9%). Banks (0.15%) were the only large group reliably below the 0.62% average.
Two common beliefs held up only in part. Software stocks lost half somewhat more often than the market (23% against 18%) but also doubled a little more often. Vehicle makers collapsed mostly among electric-vehicle companies that listed around 2018–2021, but the average auto stock has returned only 2.4% a year since 2000.

How often a stock loses half in five years

For each July from 2015 to 2021 we took every listed company with a market value of at least $300 million and measured its share-price change over the next five years. If the company was delisted, we used its last price. Where no price history survives, we used the change in market value, which also reflects new share issuance (about a quarter of cases). Entries with corrupted historical market values were removed (see "How we checked").

Share of companies whose stock was down 50% or more five years later. Black line: 95% interval (bootstrap by company). Dashed line: all stocks. Sectors with fewer than 25 companies are not shown. 4,618 companies, 21,767 company-years.

The sectors at the top share one of three features. Biotech companies often depend on one or two drug trials, and a failure leaves them worth little more than their cash. Energy and telecom companies borrow heavily against long-lived assets, so a drop in oil prices or revenue can leave them unable to pay their debts. In internet and healthcare-software companies, the losses concentrate in the companies that entered in 2019–2021, near the market's valuation peak. At the bottom are businesses that are regulated (banks, utilities) or that sold into a long investment boom (chip equipment, homebuilders). The period includes the AI data-center build-out, which likely helped chip equipment.

Risk and reward: which sectors fall and which double

The same five-year windows, seen from both sides. Each point is a sector: how often its stocks lost half their value (left to right) against how often they doubled (bottom to top). Top left is the best combination of the period, bottom right the worst, top right a lottery with many failures and many big winners.

Each point is a sector; its number is its place in the first chart (by share of companies that lost half). Horizontal: share of companies down 50% or more after five years. Vertical: share that doubled. Dashed lines: all stocks (18.3% and 20.3%). Sectors with fewer than 25 companies are not shown.

Key to the numbers in the chart above: sector, share that lost half and share that doubled within five years.

How often companies go bankrupt

Companies file a Form 8-K with Item 1.03 when they enter bankruptcy or receivership. We matched those filings to every company that filed a 10-K in the prior year and counted only the first filing of each case.

Bankruptcy filings per company per year, 2013–2015, all SEC registrants filing a 10-K (about 6,700 a year, any size). 125 bankruptcies, counting a company's first Item 1.03 filing only. Sectors with fewer than 150 company-years are not shown. The window includes the 2014–2015 oil price collapse, so energy is higher than in a calm period. Intervals are wide: oilfield services, oil and gas producers, printing and publishing, healthcare services and telecom sit reliably above the average and banks reliably below it; most other sectors cannot be told apart from the average.

Sector returns since 1970

Five years is one market cycle. For a longer view we used the industry portfolios in the Kenneth R. French Data Library at Dartmouth, which cover U.S. common stocks and include dividends. "Typical stock" weights every company equally, which is closer to the experience of owning an average company in the industry. "Index" weights by market value.

Since 1970, tobacco, defense, aircraft and semiconductors compounded fastest. Software has swung harder than most industries: since 2000 it lost more than 30% in four calendar years (2000, 2002, 2008, 2022), and it fell 77% from 2000 to 2002. Among the industries shown, autos have the widest gap since 2000 between the index and the typical stock: a few large automakers carried the index while the average company grew 2.4% a year.

Software and autos

Software stocks lost half their value in 23% of cases, somewhat more often than the market's 18%, and doubled in 24% of cases against 20% for the market. Their bankruptcy rate was 0.4% a year against a 0.6% average, although with only four software bankruptcies in the window that difference is not statistically reliable. The sector's risk shows up as sharp drawdowns across the whole group rather than as individual failures.

Automakers are a small group: 27 listed companies worth $300 million or more, including foreign makers with U.S. listings. The numbers are noisy (95% interval 17–46%). 29% of five-year windows ended with the stock down by half, above the market's 18%. Almost all of those were electric-vehicle companies that listed around 2018–2021. Among established manufacturers, Ford (2015–2020, share price before dividends), Tata Motors (2015–2020) and Stellantis (2021–2026) each lost half once; GM, Toyota and Honda did not in any window we measured.

What this means for selling put options

A put seller loses when a stock falls below the strike before expiry, usually within days or weeks. Option prices already scale with each stock's volatility, so the useful question is how often a stock falls further than its own volatility suggests. On daily prices from 2012 to 2026 we took every rolling five-day window and divided the price change by the stock's volatility over the previous 60 days. A normal distribution would put 2.28% of windows below two standard deviations. Companies with too few report dates on file, mostly foreign filers, are left out, because their report weeks cannot be separated.

The biggest factor is the company's own earnings report. In weeks with a report, stocks fell more than two standard deviations 9.7% of the time; in weeks without one, 2.6%. The stock's own history comes second: the fifth of stocks with the fattest tails in 2012–2018 went on to 3.1% of such weeks in 2019–2026, the thinnest fifth to 2.3%. A recent 20% drawdown (2.7% against 2.6%) or a share price under $20 (2.4% against 2.8%) made little difference.

Share of rolling five-day windows without an earnings report in which the sector's stocks fell more than two standard deviations of their own recent volatility. Daily prices, 2012–2026; all stocks 2.61% (3.20% including report weeks). Sectors with fewer than 15 companies are not shown. The interval resamples companies and does not capture that sell-offs arrive on the same dates, so the real uncertainty is wider.

Without report weeks, the fattest tails belong to sectors hit by shared shocks in rates, credit and commodities: REITs, oilfield services, asset managers, oil and gas producers, airlines and gas utilities. The thinnest include communications equipment, water utilities, aluminum, food and apparel makers. REITs are near the top of this chart even though they are usually seen as calm businesses, so a quiet sector is not automatically a safer put. We did not test how option prices compare across sectors, so this chart describes the stocks, not the premium a seller would receive.

How we checked the numbers

The main figures were rebuilt from the raw files with separately written code. The industry returns and the bankruptcy counts matched.
That check caught two errors, which we corrected before publishing. Repeat bankruptcy filings for the same company (plan confirmation, emergence) had been counted as new cases, which put the average rate at 0.72% a year instead of 0.62%. Some historical market values were corrupted by share-count mismatches after reverse splits, which let very small companies through the $300 million filter. We now drop an entry when the market value jumps implausibly, when our own price times the share count in the company's SEC filings puts it below $300 million while the recorded value is more than three times higher, when the implied share count shrinks more than fivefold, or when price and market value tell opposite stories: 1,163 of 22,930 entries.
We ran the five-year study on a second, independently built sample: our own daily price archive plus the SEC delisting catalog, with sectors from SEC industry codes and January cohorts 2012–2021. On the 2,689 companies both samples contain, the results agree (15% lost half in each). The second sample's lower overall level (15% against 18%) comes mainly from 1,929 companies it lacks, 21% of which were later delisted. Both samples put biotech and oilfield services near the top and banks, utilities and chip equipment near the bottom. Oil and gas producers and automakers moved a lot between samples, so their positions are less certain.

Limitations

Five-year results cover July 2015 to July 2026, one cycle. Price returns exclude dividends, which understates high-yield sectors such as REITs and utilities by several points a year and banks by a few.
A few small companies whose price history was restated after many reverse splits cannot be verified by any of the checks and may still carry inflated early market values; an independent review put the effect at one to two points for automakers and near zero for the market as a whole.
For delisted companies we use the last recorded price or market value, which understates sudden collapses. When a ticker passes to a different company after a bankruptcy, the old company's collapse can be hidden. Both effects make losses somewhat understated.
Bankruptcy rates come from 2013–2015 only.
Industry codes are self-reported and sometimes out of date. Chip-equipment and materials makers are a named list (27 met the size filter) because standard industry classifications do not separate them from chipmakers.

Sources: SEC EDGAR, Kenneth R. French Data Library, The Compound Family. This article describes historical statistics. It is not investment advice or a recommendation to buy or sell any security.