Many Models Beat One

The Core Insight

Charlie Munger declined 1,500 shares of Belridge Oil at 115 dollars each, because paying for them meant raising 173,000 dollars. He had bought 300 shares at the same price the day before. Within two years Shell bought the company at about 3,700 dollars a share. The hesitation cost him about 5,400,000 dollars, and he wrote it down.

Poor Charlie's Almanack collects eleven talks given between 1986 and 2007. One argument runs under all of them. A person who carries the big ideas from every major discipline beats the specialist who carries one. The ideas hang on a latticework, and they get run as a checklist.

He puts the count at eighty or ninety models carrying about 90 percent of the freight, with a handful carrying most of that. He names the failure mode those models exist to prevent: the man-with-a-hammer tendency.

Most professionals believe depth wins. Master one discipline to the top of the field, then hire out the rest. Munger argues the reverse trade. The models that decide your outcome mostly sit outside your training, and a specialist reaches for the tool he owns.

The Framework

The latticework is a list you can write down, and he writes most of it down. The models he names come from a handful of disciplines, and none of them are advanced.

• Mathematics supplies compound interest, permutations and combinations, and the Gaussian distribution.
• Engineering supplies backup systems, redundancy, and the breakpoint where a loaded structure fails at once.
• Physics and chemistry supply critical mass and autocatalysis, where a reaction feeds itself.
• Biology supplies the Darwinian synthesis and the tragedy of the commons, borrowed by economics from a biologist.
• Accounting supplies double-entry bookkeeping, which reached the world through Venice.
• Psychology supplies the twenty-five standard causes of human misjudgment, the one list he built himself.

He runs the models as checklists, on the argument that no capable pilot skips the checklist because of talent.

Every hard call gets a two-track read. Track one asks what the interests are when considered rationally. Track two asks what the subconscious is doing to the person deciding, including you. The gap between the tracks is where the loss usually sits.

The circle of competence comes before both tracks. Every question goes into one of three baskets: in, out, and too hard. Technology lands in too hard by default. A boundary you cannot draw is not a competence, and he looks for one-foot fences with large rewards behind them.

Key Ideas

One Discipline Produces One Answer

Jacob Viner compared many academics to the truffle hound, bred for one narrow purpose and useless for anything else. Economists took up hard-form efficient market theory in part because it permits attractive mathematics. Whitehead called the wider condition the fatal unconnectedness of academic disciplines.

The sharper version of the error is overweighting what can be counted. Thomas Hunt Morgan banned the Friden calculator from the Caltech biology department. His reasoning was that a man panning gold on the Sacramento in 1849 reaches for visible nuggets, and spends nothing on placer mining.

If the training does not supply the other disciplines, supply them yourself.

The Incentive Explains the Behavior

Incentive-caused bias describes a decent person who drifts into bad conduct because the incentives point there, and then rationalizes it. Munger treats it as the most underrated force in human affairs.

The Federal Express system requires every package to move between planes at one central airport each night. The night shift never finished on time. Moral persuasion failed, along with everything else the company tried. Someone noticed the shift was paid by the hour, while the employer wanted the task finished fast. They switched to paying by the shift, and let everyone go home once the planes were loaded. The problem ended.

A Xerox commission plan pushed the older and inferior machine, until Joe Wilson came back from government and found it. A surgeon in Lincoln, Nebraska sent baskets of normal gall bladders to pathology for years. Westinghouse wrote hotel construction loans at 95 percent of value under accounting that showed high early income, and lost billions of dollars.

The design rule follows. Never reward what can be faked, and make the system hard to cheat. One friend ran a low-margin Texas plant where workers compensation fraud reached double-digit percentages of payroll. He moved the work to Utah, where the same expense runs 2 percent of payroll.

Nine Tendencies Carry Most of the Damage

Talk 11 lists twenty-five standard causes of human misjudgment. Nine of them carry the best evidence.

Reward and Punishment Superresponse sits first, and price fixing is the demonstration. It was common under modest fines, and it fell sharply once prominent executives went to federal prison. Inconsistency-Avoidance holds conclusions in place to conserve mental space. Franklin got a rival to lend him a book, and the rival trusted him afterward. The same move was the core of the Chinese handling of Korean War prisoners, and it beat torture.

Deprival-Superreaction weighs a loss heavier than an equal gain, including a reward barely missed. A man with 10,000,000 dollars in a brokerage account is irritated at losing 100 dollars of the 300 in his wallet. Slot machines earn about 25 percent more when the electronics raise the share of near misses. Social Proof works through imitation and through inaction, and the Kitty Genovese case is the second kind.

Authority-Misinfluence produced the order for an ear infection that ended with the word rear. The nurse put the drops in the patient's anus. Contrast-Misreaction sells a 1,000 dollar leather dashboard beside a 65,000 dollar car.

Excessive Self-Regard shows in lotteries, where play runs much heavier when the gambler picks his own number. His hiring rule falls out of it: underweigh the interview and overweigh the record. Denial covers the rest, and Alcoholics Anonymous reaching about a 50 percent cure rate also means a 50 percent failure rate. Reason-Respecting closes the set, because compliance rises when people are told why, and it rises even when the reason is empty.

Forces Compound Instead of Adding

Tendency twenty-five is the one he added himself. Several tendencies pointed at the same outcome do not add. They behave like critical mass, where nothing much happens below the threshold and the reaction runs away above it. He calls the result a lollapalooza, and he found it in none of the psychology textbooks he read.

The worked example is an aircraft evacuation test. Government rules require a realistic full-load evacuation before a new passenger plane can be sold. McDonnell Douglas ran one in a darkened hangar, with many old people, a cabin about about twenty feet above concrete, and flimsy rubber slides. The morning test produced about twenty serious injuries and failed the timing. The company repeated it that afternoon and produced about twenty more, including one permanent paralysis.

The checklist reading names six forces pointing one way. Reward superresponse, doubt-avoidance, authority-misinfluence, inconsistency-avoidance, and social proof, with nobody in the room objecting. Then deprival-superreaction, running the afternoon test the way a losing gambler places one more bet.

Milgram stood in the textbooks as a demonstration of authority alone, and the checklist finds at least six principles.

Inversion Is the Cheaper Search

Carl Jacobi told mathematicians to invert, and Munger took the instruction whole. Many hard problems yield only when they are addressed backward. A rustic in one of his stories wants to know the place of his death, so as to never go there.

Applied, it changes the question. To help India, ask how to hurt India, then avoid the answers. Darwin gets the credit for the working discipline: give priority attention to evidence that disconfirms a hypothesis you love.

The 1986 commencement address is the method executed in public. Asked to advise graduates, he gave prescriptions for a miserable life instead. Ingest chemicals to alter mood. Envy people. Nurse resentment. Be unreliable, which he says outweighs every virtue you have. Learn from your own experience rather than from other people. Go down at the first severe reverse and stay down.

His own guard against the reverse error is a standing rule. He is not entitled to an opinion until he can state the arguments against his position better than the people holding it.

Coca-Cola Was an Arithmetic Problem

A fifteen-minute exercise set in Atlanta in 1884 is the best worked example in the book. A rich eccentric named Glotz offers 2,000,000 dollars of 1884 money and keeps half the equity for his foundation. The other half goes to whoever plans that foundation to 1 trillion dollars in 2034, with the company at 2 trillion dollars.

By 2034 the world holds about 8 billion beverage consumers, each taking in about 64 ounces of water a day. That is eight servings of 8 ounces. Assume the new market and its imitators capture 25 percent of world water intake, and the company holds half of that. The company sells 2.92 trillion servings in 2034, and four cents a serving makes 117 billion dollars. Beverage purchasing power multiplies at least fortyfold over 150 years, so the 1884 price is one tenth of a cent.

The business is the creation and maintenance of conditioned reflexes. Operant conditioning supplies the reward from four sources: calories, flavor and aroma, a stimulant such as caffeine, and cooling. Pavlovian association attaches the drink to whatever people admire, which is why the name is exotic rather than descriptive. Universal availability protects the reflex, because a product never tried cannot build a rival habit. Spend more than 40 percent of the fountain price per serving on advertising. Make the bottlers subcontractors, rather than perpetual franchisees buying syrup at a frozen price.

Then he inverts and lists four things to avoid. A cloying aftertaste that stops consumption. Loss of half the trademarked name. The effects of envy. And any sudden large change in flavor, which is New Coke, and which deprival-superreaction explains.

In 1896 the real company had a net worth under 150,000 dollars and earnings near zero. It lost half its trademark and granted perpetual bottling franchises at fixed syrup prices anyway. It is worth about 125 billion dollars now, and 8 percent a year to 2034 reaches 2 trillion dollars. Munger asked his editor to warn readers that most people do not understand the talk, and he grades its delivery a failure.

Practical Applications

Price the incentive before you judge the behavior. Ask who is paid for what, and what the payment rewards. When the count is wrong, change the count, and the shift from hours to shifts at Federal Express is the pattern.

Fear advice most when it pays the adviser well. Learn enough of the adviser's trade to check the work, then replace a large part of what comes back.

Run the sort before the analysis. Put anything you cannot bound into the too hard basket. Then run both tracks in writing: the interests on a rational reading, and the tendencies working on you while you read.

Count the forces before sizing the bet. One tendency pointed at your preferred answer is normal. Four pointed the same way is the case for a very large bet or a hard stop.

Who This Is For

Operators making irreversible calls under bad information get the most from it. Anyone who hires, prices, or negotiates gets the psychology at a discount to firsthand tuition. Talk 11 and the Coca-Cola talk carry the load, and the rest is reinforcement.

The book is a compilation of talks he gave repeatedly, and not an argument written once. The editor states that the repetition is deliberate. The twenty-five tendencies overlap heavily and were never validated as a taxonomy. Doubt-avoidance, inconsistency-avoidance, and social proof cover much of the same ground. He assembled the list from memory over about fifty hours at eighty-one, having taken no course in psychology. His investing record shows that he made good bets, and it stands as no evidence that the list is correct.

Skip it for an investing method. The market commentary is dated, the foundation mechanics have aged, and the family reminiscences carry nothing. A reader who wants the machinery without the repetition can read three talks and stop.

The Decision

Take the decision you are most confident about this week and write the question backward. List every way to guarantee its failure, and name who is paid to make each one happen.

Then count. Go through the tendencies and mark each one pointing at the answer you already prefer. One mark is noise. Four marks pointing the same way is the case the checklist exists to catch.

Run it on the call you are surest about. Certainty is where the count runs highest.