Fundamental Base Rates
Do strong companies stay strong? Measured on all US companies, not on the showcase examples.
What this is: Across this platform we test popular claims by measuring them as base rates — how often something holds when you count every case, not just the good examples. This page applies the same method to two classics of stock analysis: "high-margin companies stay on top" and "growth companies keep growing" — measured on 9,930 US companies from their mandatory SEC filings, as first reported.
The result in one sentence: High margins mostly persist (79.6 % are still in the top quarter a year later, against 25.8 % by chance); high revenue growth persists far less often (44.1 %) — and is often just volatility.
No prices, no returns, no stock picks — the numbers say whether a metric persists, never what the stock does next.
Operating income ÷ revenue, per fiscal year from the company's own 10-K as first reported.
Top and bottom
| Condition | p | q | Δ pp [95 % CI] | n |
|---|---|---|---|---|
| Top, 1 year | 79.6 % | 25.8 % | +53.7 [+52.6; +54.8] | 6,039 |
| Top, 3 years in a row | 88.3 % | 27.4 % | +60.9 [+59.2; +62.6] | 1,690 |
| Bottom, 1 year | 82.4 % | 23.0 % | +59.4 [+58.4; +60.6] | 5,513 |
| Bottom, 3 years in a row | 89.1 % | 19.5 % | +69.6 [+67.9; +71.2] | 1,387 |
| REIT, top, 1 year | 82.2 % | 25.5 % | +56.7 [+50.8; +62.2] | 213 |
| REIT, top, 3 years in a row | 88.7 % | 25.4 % | +63.3 [+54.7; +72.9] | 62 |
p = share in the same quartile next year · q = share in that quartile over all windows of equal length in the same period · non-overlapping windows · CI from a company-cluster bootstrap. REITs run separately because depreciation distorts their margins.
Does it scale with the level?
Share in the top quartile next year, by tenth in year t (1 = lowest values). Spearman 0.988.
Rising monotonically — the shape a real effect has.
Before and after the rate shock
| Top | Outcome year ≤ 2021 | ≥ 2022 |
|---|---|---|
| 1 year | Δ +53.4 [+52.2; +54.7] · n 4,269 | Δ +54.4 [+52.8; +56.1] · n 1,770 |
| 3 years in a row | Δ +60.8 [+58.7; +62.7] · n 1,133 | Δ +61.2 [+58.5; +63.6] · n 557 |
No shift larger than the width of the confidence intervals is visible. The difference itself was not bootstrapped separately, so "unchanged" is not measured.
By sector
| Sector (SIC) | 1 year: p · q · n | 3 years in a row: p · q · n |
|---|---|---|
| Chemicals & pharma (28) | 73.5 % · 18.9 % · 600 | 86.8 % · 20.8 % · 151 |
| Industrial machinery & computers (35) | 80.8 % · 29.0 % · 317 | 95.2 % · 31.8 % · 84 |
| Electronic equipment (36) | 72.4 % · 21.2 % · 358 | 92.4 % · 23.1 % · 92 |
| Transportation equipment (37) | 76.8 % · 22.1 % · 138 | 88.9 % · 22.6 % · 36 |
| Instruments (38) | 83.3 % · 23.0 % · 413 | 90.6 % · 23.6 % · 127 |
| Communications (48) | 80.9 % · 43.5 % · 298 | 82.1 % · 46.9 % · 95 |
| Utilities (49) | 92.5 % · 66.4 % · 596 | 96.1 % · 68.3 % · 205 |
| Software & business services (73) | 84.1 % · 22.7 % · 734 | 90.2 % · 23.4 % · 215 |
| All other sectors | 76.9 % · 25.4 % · 2,585 | 84.7 % · 26.9 % · 685 |
A sector gets its own row from 50 companies with at least 10 consecutive years; the rest runs as "all other". "small" = fewer than 30 conditional cases — read as an anecdote. A high q (e.g. utilities on margin) means the metric is structurally high there, so Δ comes out small.
Survivorship and method
- 54.1 % of companies no longer file a 10-K after FY2023 (acquisition, delisting, bankruptcy, going private). EDGAR keeps them, and so does the sample. The conservative number counts each of them as "did not hold"; the truth lies between both numbers, because some drop-outs are acquisitions of successful companies.
- Every number comes from one regime window: fiscal years 2011–2025, which contain exactly one recession (NBER, Feb–Apr 2020, two months).
- Source: SEC EDGAR companyfacts (companyfacts-2026-09-26.zip), periods from 2011-06-15, tag rules tag-rules-v1: period from the value's end date, own 10-K as first reported, declared tag priority per metric. 776 banks excluded, 482 REITs separate.
- Quartiles per fiscal year over all companies with a value. CI from 1,000 bootstrap draws over companies. Before any reading, a synthetic control had to show that the method finds built-in persistence and creates none without it. Measured 2026-09-26.
Machine-readable: Operating margin · Return on assets · Revenue growth (Knowledge Objects, also via API and MCP).
Fundamental Base Rates — Do High Margins and Fast Growth Persist?
"Quality persists" and "growth persists" are two of the most repeated claims in stock analysis, and almost never published as a base rate. This page measures them on 9,930 US companies from their SEC filings as first reported: how often a company in the top quartile of its year stays there, against how often any company lands there in the same period.
- Operating margin, top quartile, 1 year
- 79.5 % stay vs 25.8 % base rate (Δ +53.7 pp, n 6,039)
- Return on assets, top quartile, 1 year
- 74.8 % stay vs 25.9 % (Δ +48.9 pp, n 6,324)
- Revenue growth, top quartile, 1 year
- 44.1 % stay vs 23.2 % (Δ +20.9 pp) — U-shaped, partly volatility
- Survivorship
- 53.8 % of companies stop filing a 10-K after FY2023; they stay in the sample
- Regime
- Fiscal years 2011–2025, one recession (Feb–Apr 2020)
- Not
- A stock pick, a forecast, or anything about prices and returns
Method: quartiles are set per fiscal year over all companies with a value, so the threshold moves with the market. Windows do not overlap. Every rate stands next to the unconditional rate of the same period and a 95 % confidence interval from a company-cluster bootstrap. Banks (776) are excluded because they report no comparable revenue; REITs are a separate group. Before any number was read, a synthetic control had to show that the method finds persistence where it was built in and none where it was not.
What it shows: margins and returns on assets are highly persistent — after three years in the top quartile, 88.3 % (margin) and 83.7 % (return on assets) are still there a year later. Revenue growth persists far less, and extreme growth in either direction tends to follow extreme growth. Counting every company that stopped filing as a failure lowers every rate by 6 to 17 points; the truth lies between the two numbers, because some of those companies were acquired for doing well.