One: A Book Written on the Corpses of the Market
1934, the deepest year of the Great Depression. Graham and Dodd co-authored Security Analysis, a book so thick almost nobody finishes it.
What was the context? The Dow had just fallen 89% from its 1929 peak. Countless companies bankrupt, banks failed, investors jumped. The credibility of the entire U.S. financial industry was at its lowest point in history.
Graham was trying to rebuild one thing in this environment—to re-draw the line between 'investing' and 'speculation.'
The book opens with a near-foundational definition: investment is an operation that, upon thorough analysis, promises safety of principal and an adequate return. Operations not meeting these conditions are speculation.
90 years later, this definition is still something most people fail to achieve.
Every time I reread this book, that sentence stings. Because most so-called 'investing' has no thorough analysis, no safety of principal, and isn't seeking an adequate return—it's gambling. Acknowledging this is the starting point of any methodology.
Two: The First Timeless Yardstick—The Creditor's Perspective
The most distinctive feature of this book is its strong creditor's perspective—meaning, when you look at a company, first ask 'what can I get back in the worst case' rather than 'how much can I earn in the best case.'
Most retail investors enter through the stock market, naturally carrying a shareholder's perspective—looking at profit, growth, imagination. But Graham, writing in 1934, thought about something else: If this company goes bankrupt tomorrow, how much can I recover?
This view looks conservative, but it provides an extremely practical toolkit:
- Current assets vs. current liabilities—can the company survive a year?
- Total assets vs. total liabilities—after liquidation, how much is left for shareholders?
- Free cash flow vs. interest expense—can the company keep paying its debts?
These three ratios look like basic accounting. But spend ten minutes looking at the financials of some hot small-cap stocks in the Russell 2000, and you'll see that many are dangerously weak on these three ratios. During the 2023-2024 rate hiking cycle, roughly 40% of Russell 2000 constituents had a net interest coverage ratio below 1—meaning they don't earn enough in a year to cover interest payments.
Bull markets hide everything; bear markets reveal everything. That's why company stories always look better than financial statements at the top, while statements look better than stock prices at the bottom.
What Graham taught me is not how to value, but how to re-read a company through a creditor's eyes.
Three: The Second Timeless Yardstick—Normalized Earnings
This book repeatedly stresses one thing—don't look at one year's profit; look at the average profit of the past ten years.
Why? Because a single year's profit is noise; a ten-year average is signal.
Graham's specific prescription: use the average profit of the past 7 to 10 years, divided by the current stock price, to get a 'normalized P/E.' This number is much more honest than the current year's P/E.
Later, Shiller extended this into the famous CAPE (Cyclically Adjusted P/E). Shiller used it to gauge the overall U.S. stock market—and at the highs of 1999, 2007, and 2021, CAPE gave clear warning signals.
When I look at an individual company, I do a rough version: cumulative net profit over the past 10 years, divided by current market cap—a crude but stable yardstick. If this number is below 5%, I am very cautious, no matter how compelling the narrative.
Applied to U.S. stocks in 2026, this yardstick screens out several of the Mag 7—their current profits are high, but 10-year cumulative profit relative to market cap is still below 5%. That doesn't mean they are bad companies; it means current pricing embeds huge expectations of future profits, and whether those expectations materialize is the real risk.
Why is this method still effective? Because it fights one of the most stubborn human biases—extrapolating the recent best data into the future. Everyone gets carried away by 'this year was great.' Forcing yourself to look at ten years is the only way not to be fooled by a one-year story.
Four: The Third Timeless Yardstick—Margin of Safety as Attitude, Not Buffer
This book enshrined 'margin of safety' as the core of value investing.
But many people understand this term superficially—thinking margin of safety just means 'buying below estimated value.'
After reading it three times, I think what Graham really meant is something else: Margin of safety is a provision for your own cognitive errors, not for market volatility.
What's the difference?
If you think margin of safety is a 'buffer against market volatility,' your logic is 'I correctly estimated value, the market is temporarily mispricing it, so I wait.' That is confidence.
If you think margin of safety is a 'buffer against cognitive error,' your logic is 'my estimated value itself might be wrong. So I must leave room.' That is humility.
Confidence and humility lead to different order decisions.
The confident person sees an 80% discount and goes all in. The humble person sees an 80% discount but still controls position size—because he knows his estimated '100' might only be '60.'
Graham was the latter. When he wrote Security Analysis, he had just witnessed the 1929 crash—he saw countless confident analysts lose everything. He no longer believed in confidence; he chose to believe in structure.
Five: Where I Differ from Graham
On the fifth reading, I found myself with a few genuine disagreements with Graham.
First, his method is essentially useless for companies whose primary assets are intangible.
Graham's entire valuation framework is built on the balance sheet—tangible assets, accounts receivable, inventory, fixed assets. This method was extremely effective in the 1930s because the main assets of companies were those things.
But today, 8 of the top 10 U.S. companies by market cap have intangible primary assets—brand, user data, network effects, ecosystems, talent. These are almost invisible on Graham's balance sheet.
Applying Security Analysis to Apple, you'd get a very high P/B and conclude you shouldn't buy it. But Apple's real assets—the ecosystem of 2.2 billion active devices, the two-sided network effect of the App Store, the brand premium—were not priceable by Graham in his time.
In 2026, the true applicable scenarios for Graham's method are cyclical stocks, banks, insurance, traditional industrials, and energy—industries where assets are still mostly on the balance sheet.
Second, his 'diversification' assumption has been eaten by indexing.
Graham recommended defensive investors hold 10-30 stocks to diversify risk through quantity. This was entirely reasonable in the 1930s—there wasn't even a single index fund; true diversification required holding many names.
Today, buying one VTI (total U.S. stock market) diversifies you into 4,000 companies. Graham-style 'manual diversification' is meaningless compared to ETFs.
What really remains worth concentrating on is 'your unique insight relative to the index.' Security Analysis can't answer that—it was written in 1934, before indices even existed.
Third, he was almost ignorant of growth.
Graham's valuation method is essentially 'static valuation'—based on past earnings and existing assets. He barely considers future growth.
This method works in low-growth environments. But in U.S. stocks from 1990-2025, the real excess returns came almost entirely from high-growth companies—Microsoft's P/B was 30 in 1990, Google's P/E at IPO in 2004 was 80, Amazon was still a 'no-profit story stock' in 2010. Security Analysis would have said don't buy any of them. Yet the compounding from these three far exceeded everything Graham's method could find in the same period.
What I've learned is that Security Analysis is a 'don't lose' book, not a 'win big' book. It teaches you how not to be fooled, not to be fleeced, not to go bankrupt in a crash. It doesn't teach you how to generate 100x returns over 30 years. Those two goals require different tools.
Fourth, his accounting concepts have been partially obsoleted by modern U.S. GAAP.
When Graham wrote, GAAP was not yet formed, and management had enormous room to manipulate statements. So he emphasized 're-calculate it yourself'—adjust depreciation, revalue inventory, re-examine pension liabilities.
Today GAAP is much stricter, but the form of manipulation has changed—non-GAAP profit, adjusted EBITDA, whether stock-based compensation is deducted, whether AI capex is capitalized or expensed. The challenge today is not that numbers are fake, but that 'which adjusted number best reflects reality' is unclear.
Graham's attitude (question everything on the statement) remains, but what specifically to question has changed.
Six: Security Analysis vs. Buffett Letters—The Real Tension Between Graham and Buffett
Buffett has always called Graham his teacher. But look closely: their methods basically diverged after 1972.
Graham believed in 'quantitative value'—look for cheap cigar butts, pick them up for one last puff. His method worked in U.S. stocks from 1930-1960 because the market was full of undervalued cigar butts.
Under Munger's influence, Buffett moved toward 'quality value'—find great companies, buy at a reasonable price, hold forever. This method no longer relies on 'cheapness' but on 'certain compounding.'
This split has a specific date—1972, when Berkshire bought See's Candies. That was the first time Buffett clearly paid a not-cheap price for a 'brand moat.' Graham would never have bought such a company because its balance sheet didn't reflect See's true value. But Buffett bought it, and that company later generated cumulative cash flow for Berkshire more than 100 times the purchase price.
My own stance: In 2026 U.S. stocks, I use Buffett's 'quality value' as core and Graham's 'creditor perspective' as risk control. The two methods are not contradictory; the former decides what to buy, the latter decides what not to buy.
Seven: On Value Traps—What Graham Largely Missed
Security Analysis barely discusses one thing—value traps.
A value trap is a company that looks cheap (low P/E, low P/B, high dividend), but the cheapness comes from a systematically deteriorating business, not a temporary market mispricing. Graham's method would screen such companies out, but buying them leads to persistent losses.
Classic value traps over the past 20 years: Sears, JCPenney, Kodak, Yahoo, General Electric (2010s), IBM (2011-2018). At some point, each passed all of Graham's quantitative screens—cheap, had assets, had cash flow—but they all eventually lost.
The essence of a value trap: Graham assumes 'cheap' means 'the market is wrong,' but sometimes 'cheap' is 'the market's honest pricing of a dying company.'
I've added three screens for value traps that Security Analysis doesn't have:
- Is the industry still growing? — Graham's method almost always fails in shrinking industries.
- Has management's capital allocation been correct over the past 10 years? — Wrong buybacks and wrong acquisitions make a cheap company even cheaper.
- Is it threatened by technological substitution? — Graham couldn't foresee this in 1934, but today it must be checked.
Only with these three additions does Security Analysis become barely usable in 2026.
Eight: Why This Book Is Still Worth Reading 90 Years Later
Security Analysis is not an easy read. It's written in 1934's language, with 1934's cases and 1934's accounting rules. Most people put it down after 30 pages.
But its real value lies not in the specific techniques—those are obsolete—but in its attitude.
What it teaches you:
- First, admit that the market is unknowable;
- Then, admit that you yourself are unreliable;
- Under the twin premises of unknowability and unreliability, design a method that can survive.
This attitude has not changed in 90 years. And it will not change.
Buffett often says: I read this book at age 19, and it changed my life. He didn't mean the techniques changed; he meant the attitude changed—he went from a boy who wanted to win to an investor who didn't want to lose.
Nine: A Final Note
My advice on how to read this book: Don't start from the beginning. First pick three chapters—'Distinction Between Investment and Speculation,' 'Creditor Perspective,' and 'Margin of Safety.'
After those three chapters, you have the skeleton of value investing. The rest, use as a reference dictionary.
What Graham and Dodd left for posterity is not valuation formulas, but an attitude toward uncertainty.
This attitude is worth revisiting for every investor.
Because the market can be wrong ten thousand times, but you only need to be wrong once, and the game is over.
Graham, in 1934, had seen too many people eliminated by that one mistake. This book is his letter to 'the next generation that should not be eliminated'—and that sincerity can still be felt 90 years later.
But he wrote the skeleton, not the flesh. The flesh you have to grow yourself, in your own generation's market.
专注投资分析、市场洞察与资产配置。不追短期波动,只理解真正驱动长期回报的东西。
