Four S&P 500 companies pass Graham's defensive test
Under Benjamin Graham's original 1973 criteria, 4 of 503 companies qualify. All four are financials. Sixty percent of the index fails both halves of his balance-sheet test.
Benjamin Graham published a checklist for the defensive investor in The Intelligent Investor — criteria a company had to meet before someone unwilling to spend their life analysing businesses should own it. Adequate size. A strong balance sheet. A decade of uninterrupted profits. Growth. And strict ceilings on what you pay.
I run that checklist against every company in the S&P 500, nightly, from data filed with the SEC.
As of 22 August 2026, four companies pass.
| Ticker | Company | Sector |
|---|---|---|
| ACGL | Arch Capital Group | Financials |
| EG | Everest Group | Financials |
| FISV | Fiserv | Financials |
| RF | Regions Financial | Financials |
Three reinsurers or payments processors and a regional bank. Graham excluded banks and insurers from his working-capital tests, so the bar for financials is five of six testable criteria rather than six of six — but the pass count still has to clear a fixed threshold. It does not shrink when filed data is missing.
Regions passes on a marginal P/E: 14.4× against Graham's 15× ceiling on three-year average earnings. A price rise of roughly four percent would drop RF to Enterprising without changing a single filing.
Why so few pass
Graham's defensive standard is deliberately harsh. Under classic 1973 rules, the twenty-year dividend record cannot be tested on the filing history we store, so every company is scored on six fixed slots — not a smaller denominator when balance-sheet or earnings fields are absent from our extract.
That matters. Several names that previously appeared to pass were clearing a lower bar because missing inputs silently dropped whole criteria from both the numerator and the denominator. Homebuilders that looked defensive on five of five evaluable criteria were not failing Graham's balance-sheet test; the test was never run. Once the denominator is held constant, they no longer qualify.
The honest reading is not that the index got worse overnight. The scoring now matches the rule the site already described in its methodology: six of seven criteria testable under classic rules, with a fixed pass threshold.
What actually fails
Here is how often each criterion rejects a company, across the 482 with enough filing history to score:
| Criterion | Companies failing | Rate |
|---|---|---|
| 1. Adequate size | 0 | 0.0% |
| 2. Strong financial condition | 343 | 71.2% |
| 3. Positive earnings every year for a decade | 142 | 29.5% |
| 4. Uninterrupted dividends for 20 years | not evaluated | — |
| 5. Earnings growth of at least a third | 131 | 27.2% |
| 6. P/E no higher than 15 | 408 | 84.6% |
| 7. P/E × P/B no higher than 22.5 | 397 | 82.4% |
The price ceilings reject the most, which is unsurprising in an expensive market. But criterion 2 is where the interesting result is.
Graham's second criterion has two parts: a current ratio of at least two, and long-term debt not exceeding net working capital. Taken separately, 299 companies fail the current ratio test and 330 fail the debt test.
The overlap is almost total:
| Pattern | Companies |
|---|---|
| Fail the current ratio test only | 13 |
| Fail the debt test only | 44 |
| Fail both | 286 |
Two hundred and eighty-six companies — 60% of everything scoreable — fail both halves of Graham's balance-sheet test. Only thirteen fail one without the other.
That isn't two problems. It's one. Companies that run thin on working capital are the same companies carrying debt beyond what Graham thought prudent, and they are the substantial majority of the largest businesses in the United States.
That's not a scandal. Working capital management has changed enormously — just-in-time inventory, revolving credit and commercial paper mean a modern company can operate safely on a balance sheet that would have alarmed Graham. But it does mean the failures aren't a story about valuation alone, and a cheaper market wouldn't fix them.
Almost nobody fails on one thing
| Criteria failed | Companies |
|---|---|
| 0 | 8 |
| 1 | 29 |
| 2 | 107 |
| 3 | 205 |
| 4 | 102 |
| 5 | 31 |
Twenty-nine companies are within a single criterion of passing. Four hundred and forty-five fail on two or more, and the most common outcome by some distance is failing three.
This is the part that surprised me most. I expected a large group of good businesses failing narrowly on price — companies that would qualify after a correction. That group is 29 names.
Which has an uncomfortable implication for anyone waiting for a drop, including me: it's the premise this whole site is built on. A 20% fall in prices would fix criteria 6 and 7. It would do nothing whatsoever about the 343 companies failing Graham's balance-sheet test, the 142 without a decade of unbroken profits, or the 131 without his growth.
Cheaper prices produce more Graham candidates. They don't produce nearly as many as a loose denominator once implied.
Where everything lands
| Classification | Count |
|---|---|
| Defensive | 4 |
| Enterprising | 90 |
| Net-net | 0 |
| Speculative | 388 |
| Insufficient data | 21 |
Graham used "speculative" as a technical term, not an insult — it means the price being asked isn't supported by the criteria he set out. Three hundred and eighty-eight companies, roughly four-fifths of the scoreable index, sit there today.
Zero net-nnets, across all 503. The strategy that made Graham's name — buying a company for less than its liquidation value — has no candidates whatsoever among the largest 500 American companies.
What this analysis cannot do
The twenty-year dividend criterion is not evaluated. Graham required an uninterrupted dividend record going back two decades. The filing data available covers at most ten years, so this criterion is marked "not evaluated" for every company rather than silently passed. Every stock page shows how many years are actually on file. Seven of Graham's criteria exist; six are currently testable under classic rules, with a constant denominator that does not shrink when other fields are missing.
These are Graham's 1973 thresholds, not adapted ones. Some are arguably obsolete. The dividend requirement would disqualify a company for a single cut during 2008 or 2020, punishing a decision many would now call prudent. A modernised ruleset is available on the site as a labelled alternative, and it produces a different answer.
None of this predicts anything. Companies failing this test have been excellent investments for years and may continue to be. Companies passing it can still lose money. The checklist screens for cheapness and financial strength; it has nothing to say about whether a business will still be good in a decade.
Method
Fundamentals come from company filings with the SEC via EDGAR. Prices update daily from a market data provider. The checklist runs nightly against all 503 index constituents, and every company's individual result — including which criteria it passes, fails, or cannot be evaluated on — is published at grahamwise.com. It's free, there's no signup, and there's nothing to buy.
Twenty-one companies are unscored and excluded from the percentages above: ten with genuinely short filing history on file, and eleven where our pipeline cannot yet extract enough EPS and revenue from SEC filings that do exist. See the next section.
The count changes nightly as prices move.
Why the scoreable base moved
Until August 2026, our parser missed revenue or EPS on dozens of companies whose filings were fine. The scoreable base rose from 450 to 482 across five stages of fixes — not because Graham's rules changed, but because we learned to read what was already in the filings.
Retailers and utilities often tag the face income-statement line with RevenueFromContractWithCustomerIncludingAssessedTax — the Including name, even when the figure is already net of sales tax. We walked the tag list in the wrong order and left revenue empty for names like TJX and Duke Energy.
Banks and broker-dealers do not file a line called "Revenue." They report net interest income and noninterest income, or a single tag RevenuesNetOfInterestExpense when they publish total net revenues. Gross interest income is not the analogue of industrial sales — it includes interest paid to depositors. On Wells Fargo's 2025 10-K, gross interest plus noninterest income is about $123.5 billion; reported total revenue is $83.7 billion. Using the obvious interest tags would overstate a bank's "sales" by nearly half.
EPS gaps turned out to be sibling tags, not missing data. Airbnb, Monster Beverage, Kimco, and Regency Centers file diluted EPS as IncomeLossFromContinuingOperationsPerDilutedShare rather than the standard EarningsPerShareDiluted name.
REITs after ASC 842 stopped filing Revenues and report rental income as OperatingLeaseLeaseIncome. Equity Residential was the clearest case.
One revenue figure was wrong, not absent. Camden Property Trust had revenue on file — but at $13 million, roughly 1% of its face property revenues of $1.57 billion. The parser had locked onto an ASC 606 non-lease slice instead of the lease-income line that matches the 10-K total. That was the only company failing adequate size on a bogus number. Criterion 1 now rejects zero companies.
The genuinely novel finding of the whole exercise: every gap we closed was a sibling tag — a different GAAP name for the same face line, or a component tag sitting next to the total in the filing. The data was there. We were reading the wrong label.
What remains unscored
Only twenty-one companies sit outside the tables above — down from fifty-three before any of this work.
Ten have genuinely short filing history (fewer than five fiscal years with both EPS and revenue on file). BlackRock (BLK) is the best-known example: four years on file, not a parser failure.
Eleven are extraction gaps — SEC filings exist, but we still cannot read enough EPS and/or revenue to run the checklist honestly. The largest names:
- Berkshire Hathaway (BRK-B) — revenue on file, but no annual diluted EPS tag in the XBRL we read.
- Visa (V) — neither EPS nor share-count tags in companyfacts.
- Synchrony Financial (SYF) — deliberately left unscored. Synchrony does not report a labeled total-revenue line; retailer-share arrangements sit between net interest income and other income. We show nothing rather than invent a total.
The remaining extraction-gap names are mostly partnerships and specialty filers with non-standard tag layouts. Each stock page states the reason plainly.
Data as of 22 August 2026. Educational analysis, not investment advice.