Benjamin Graham's value investing framework centers on one idea: buy stocks for less than they are worth, with enough of a gap between price and value to protect against errors. He called this gap the margin of safety. In The Intelligent Investor, Graham described two approaches. The Defensive Investor applies strict mechanical filters that require almost no judgment. The Enterprising Investor digs into balance sheets to find stocks trading below their liquidation value, the so-called net-nets. Both approaches are quantitative screens that Claude can run across the entire US market.
This guide walks through both approaches step by step, starting with the simpler Defensive Investor screen and progressing to net-net analysis. For the screener overview and comparison table, see the Graham Value Screener.
Step 1: The Defensive Investor screen
Graham's Defensive Investor criteria from Chapter 14 of The Intelligent Investor are a mechanical checklist. If a stock passes all seven tests, it qualifies. No judgment calls, no industry knowledge, no forecasting. The criteria were designed to be applied by someone who does not want to spend time on security analysis.
The original seven criteria: (1) adequate size, (2) strong financial condition (current ratio above 2), (3) earnings stability (positive earnings in each of the last 10 years), (4) dividend record (uninterrupted dividends for 20 years), (5) earnings growth (at least 33% increase in EPS over 10 years), (6) moderate P/E (under 15), and (7) moderate price-to-assets (P/E x P/B under 22.5).
"Screen all US stocks for Graham's Defensive Investor criteria: P/E under 15, price-to-book under 1.5, product of P/E and P/B under 22.5, positive net income every year for the last 5 years, current ratio above 2. Market cap above $2 billion."
Claude checks each year's net income individually, not just a trailing average. If a company had a loss in any single year, it fails. The P/E x P/B product is the combined valuation ceiling that Graham used instead of applying the two ratios independently. Relaxing the earnings window from 10 years to 5 expands the universe while still requiring sustained profitability. Few screens on traditional platforms can check a product of two ratios or verify each year's earnings individually.
Step 2: Adding the Piotroski quality filter
A stock trading below book value could be cheap because it is improving or because it is deteriorating. The Piotroski F-Score separates the two groups. Piotroski's 2000 paper showed that buying high F-Score stocks among low price-to-book companies outperformed the market by about 7.5% annually. This is a natural overlay on any Graham screen.
"Find US stocks trading below book value with positive earnings, Piotroski F-Score of 7 or higher, and market cap above $500 million. Show the F-Score, P/B ratio, and net income."
This combines Graham's price discipline with Piotroski's quality signal. A low P/B ratio with a high F-Score means the stock is cheap and the business fundamentals are improving. For more on the F-Score, see the Piotroski F-Score Screener.
Step 3: Net-net screening (Enterprising Investor)
Graham's ultimate bargain: a company trading below its net current asset value. NCAV equals current assets minus total liabilities. If you can buy the company for less than NCAV, you are getting the fixed assets, the brand, and the ongoing business for free. Graham preferred stocks trading at two-thirds or less of NCAV.
"Net-net screen: US stocks where current assets minus total liabilities exceeds the market cap. Show NCAV, market cap, and the ratio. Filter to companies with positive operating cash flow to exclude cash burners."
The positive cash flow filter is important. Many net-nets are cheap because the business is declining and burning through its cash balance. Graham called these "melting ice cubes." Without the cash flow filter, a net-net screen returns companies that look cheap on the balance sheet but are losing value every quarter. Typically 5 to 15 stocks qualify at any given time, almost all micro-caps with thin trading volume.
Step 4: The Graham Number
The Graham Number estimates the maximum fair price for a stock based on earnings and book value: the square root of (22.5 x EPS x book value per share). It combines Graham's P/E ceiling of 15 and P/B ceiling of 1.5 into a single number. A stock trading below its Graham Number passes both valuation tests simultaneously.
"For each US stock with market cap above $1 billion, positive EPS, and positive book value, compute the Graham Number (square root of 22.5 times EPS times book value per share) and show stocks where the current price is at least 30% below the Graham Number."
The 30% discount to the Graham Number is the margin of safety. Graham did not buy at fair value; he bought well below it. Claude computes the Graham Number from EPS and book value per share and compares it to the current price in one pass. Stocks with negative EPS or negative book value are automatically excluded since the formula requires both to be positive.
Step 5: Historical backtesting
The final step: check whether these screens actually worked. Shibui has 20+ years of quarterly financials and daily prices, so you can run Graham screens at any point in history and measure subsequent returns.
"Find stocks that were trading below book value with positive earnings and Piotroski F-Score 7 or higher as of December 31, 2020. What was the average return 12 months later?"
Important caveats: survivorship bias applies. Companies that went bankrupt or were delisted are not in the dataset, so the backtest overstates real returns. The backtest also does not account for trading costs, bid-ask spreads (wide for micro-caps), or the difficulty of actually buying thin-volume net-nets at the quoted price. These are historical observations, not investment recommendations.
What Shibui cannot do
Shibui has balance sheet data, income statements, and valuation metrics. It does not have:
- Dividend history as a continuous record (dividends paid is a quarterly cash flow line item, but there is no "consecutive years of dividends" counter). Claude can check for dividend gaps by looking for quarters where dividends paid was zero.
- Adjusted book value (Graham recommended adjusting for intangibles and off-balance-sheet items). Tangible book value per share is available as a proxy.
- Qualitative factors: management quality, competitive position, industry dynamics. Graham emphasized that quantitative screens are the starting point, not the conclusion.
- Delisted companies. Historical backtests suffer from survivorship bias.
Frequently asked questions
What is the difference between Graham's Defensive and Enterprising Investor?
The Defensive Investor applies strict mechanical filters: adequate size, strong financials (current ratio above 2), earnings stability (positive earnings every year for 10 years), dividend record, and moderate valuation (P/E under 15, P/B under 1.5). The Enterprising Investor goes further, looking for net-net stocks trading below liquidation value and special situations. The Defensive approach requires less work; the Enterprising approach finds deeper bargains but demands more analysis.
How do you compute net current asset value?
Net current asset value (NCAV) equals current assets minus total liabilities. Graham considered a stock a bargain when it traded at two-thirds or less of NCAV, meaning you could theoretically buy the company, liquidate its current assets, pay off all debts, and still come out ahead. Shibui has current assets and total liabilities on quarterly balance sheets, so Claude computes NCAV across the entire market and compares it to each company's market cap.
Are Graham's criteria still relevant today?
The specific thresholds need updating. Graham wrote for a market where most companies paid dividends, capital-light business models did not exist, and a current ratio above 2 was standard. Modern screens often relax the dividend requirement, lower the current ratio threshold, and adjust the revenue floor for inflation. The underlying principle, buying below intrinsic value with a margin of safety, remains the foundation of value investing.
Can AI screen for net-net stocks?
Yes. On Shibui, you describe the screen in plain English and Claude computes NCAV for every stock, compares it to market cap, and filters by any additional criteria you specify, such as positive operating cash flow to exclude cash-burning companies. Traditional screeners like Finviz do not expose current assets or total liabilities as filtering fields, so net-net screening requires either a specialized tool or a database with balance sheet data. Graham recommended checking whether management was buying the stock. To overlay SEC insider buying data on a value screen, see the insider buying screener.