Freakonomics

What's inside (8 sections)

1. Incentives explain more than morals do
2. What teachers and sumo wrestlers have in common
3. Secrets lose power when everyone can read them
4. Crack gangs are a pyramid and houses are a signal
5. Abortion, crime, and the argument that would not die
6. What holds up and what bugs me
7. Key takeaways
8. FAQ

I picked up Freakonomics because I kept hearing it described as economics for people who hate economics. That sounded right. I did not want graphs about interest rates. I wanted stories. What I got was a book by University of Chicago economist Steven D. Levitt, written with journalist Stephen J. Dubner, that treats economics as a set of tools for asking odd questions and then chasing data until an answer shows up. The full title is Freakonomics: A Rogue Economist Explores the Hidden Side of Everything, first published in the U.S. in 2005 by William Morrow.

The core claim is simple and it runs through every chapter. Economics, at root, is the study of incentives. People respond to rewards and punishments, to praise and shame, to money and status, often in ways they would never admit out loud. If you can figure out what someone actually wants, and what information they have that you do not, you can often explain behavior that looks strange on the surface.

That sounds abstract. The book never stays abstract for long. One page you are in a Chicago classroom where test scores jumped in a way that made no sense. A few pages later you are in Tokyo watching sumo wrestlers. Then you are listening to a bagel seller in Washington, D.C., reading Ku Klux Klan newsletters, or sitting with a crack gang in Chicago public housing. Levitt loves data sets other economists ignored. Dubner loves turning those data sets into stories you can retell at dinner. Together they made a book that sold millions, annoyed plenty of experts, and still gets argued about.

I should say up front who does what here. Levitt supplied the research questions and the statistical work. Dubner supplied the narrative voice. The cover lists both men, and later editions even added an explanatory note about how they met when Dubner profiled Levitt for the New York Times Magazine in 2003. For this site I file the book under Levitt to match the reading list, but it really is a collaboration.

Incentives explain more than morals do

The introduction sets up the method. Levitt starts from the idea that conventional wisdom is often a story people repeat because it sounds right, not because anyone checked. Experts repeat it too. To get past it, you need to ask a different question, find the right data, and measure carefully.

His favorite line in this part is that morality describes how we wish the world worked, while economics describes how it actually works. He does not mean people are evil. He means people face pressures that push them away from what they say they believe.

He gives three kinds of incentives. Economic incentives are the obvious ones. You pay people to do something, or fine them if they do not. Social incentives are about approval and embarrassment. Nobody wants to look cheap or foolish in front of neighbors. Moral incentives are about conscience. People want to feel like good parents, good teachers, honest workers.

The tricky part is that incentives interact. Add a fine and you can accidentally erase guilt. Levitt tells the famous Israeli daycare story here, based on work by economists Uri Gneezy and Aldo Rustichini. A group of daycares had a problem with parents picking up kids late. Teachers had to stay late, unpaid, and felt resentful. The daycares introduced a small fine for late pickup, around three dollars. Late pickups went up, not down.

Why? Before the fine, being late triggered guilt. You were imposing on a tired teacher. After the fine, you were buying extra childcare. Three dollars felt cheap for an extra twenty minutes. When the daycares later removed the fine, lateness stayed high. The guilt did not come back. You had changed a moral trade into a market trade.

I think about that story a lot. It pops up whenever schools or companies try to fix behavior with a fee and then act surprised when people treat the fee as a price. Levitt returns to this pattern again and again. Cheating is rarely about a few bad apples. It is about a system that rewards a number so strongly that fudging the number starts to feel rational.

He also introduces information asymmetry in this opening stretch. Experts often know more than the rest of us, and they can use that gap. A real estate agent knows more about the housing market than a seller. A car dealer knows more than a buyer. A Klansman once knew rituals outsiders did not. Much of the book is about what happens when that gap shrinks. The internet did some of that work after 2005, which makes parts of the book feel dated and parts feel prophetic.

The introduction ends with a preview of the questions to come. Which is more dangerous, a gun or a swimming pool? What do teachers and sumo wrestlers have in common? Why do drug dealers live with their mothers? How much do parents really matter? Did legalized abortion cause crime to fall? Each one sounds like a bar bet. Each one turns into a lesson about measurement.

What teachers and sumo wrestlers have in common

This is Chapter 1 in the original, and it is still the best chapter to show how Levitt thinks. The title question sounds like a joke. The answer is cheating.

Start with Chicago. In the 1990s, Chicago Public Schools started holding teachers and principals accountable for test scores. Schools with low scores faced shutdown or reorganization. Teachers could lose jobs. That is a strong incentive to raise scores. Most teachers tried to teach better. A small fraction changed answers instead.

Levitt worked with Brian Jacob, then at Harvard, to find patterns that suggested fraud. They looked at Iowa Test of Basic Skills results for third through seventh graders. Their idea was straightforward. If a class learned a lot one year and then crashed the next year, with a different teacher, something might be off. Real learning tends to persist. Fake gains disappear.

They also looked at answer strings. In a normal classroom, kids miss different questions. Strong students miss hard ones. Weaker students miss easy ones too. If fifteen kids in a row get the first easy questions wrong and then all get the same hard questions right, that is strange. It looks like someone filled in bubbles afterward.

When Levitt and Jacob applied these filters, about 5 percent of Chicago classrooms each year showed suspicious patterns. They later sent retesters to a sample of flagged rooms. In most of the retested classes where cheating had been suspected, scores fell sharply. Kids had not learned the material. An adult had fixed the test.

Levitt is careful to say most teachers were honest. He also describes the pressure honestly. A teacher in a poor neighborhood, judged on one test, with a principal begging for better numbers, might see erasing a few wrong answers as helping kids keep their school open. That does not excuse it. It explains why moral appeals alone rarely stop cheating. Change the reward and the behavior changes.

Then he jumps to Japan. Sumo looks nothing like Chicago schools, but the incentive map is almost identical.

In professional sumo, rank matters enormously. Wrestlers fight fifteen bouts per tournament. A winning record, eight wins and seven losses, means promotion, higher pay, more respect, better treatment from the stable. A losing record, seven wins and eight losses, means demotion. The difference between 8-7 and 7-8 is huge. The difference between 10-5 and 9-6 is minor. So watch what happens on the final days when an 8-6 wrestler faces a 7-7 wrestler.

Levitt gathered data on thousands of matches. The 7-7 wrestler, desperate for that eighth win, won far more often than rankings or past performance predicted. That alone could be effort. Desperate people try harder. The tell came in the rematch. When the same two men met in the next tournament, and neither was on the bubble, the previous loser now won more than expected. It looked like a favor returned. You let me win when I needed it. I let you win later.

Critics said sumo wrestlers just fight harder when desperate. Levitt answered with the rematch pattern plus the fact that the effect was strongest where relationships were closest, among wrestlers from the same stable network or with a history of meeting. It was hard to explain with effort alone.

My favorite detail in this chapter is not sumo or schools. It is Paul Feldman, the bagel man. Feldman left a defense research job and started selling bagels to office workers on an honor system. He dropped off bagels and a cash box, then collected money later. He kept records for years. Payment rates fell when it was cold or during holidays, rose in smaller offices where people knew each other, and dipped among higher paid workers in a way that surprised him. It was a quiet, real world test of honesty without police or cameras. Most people paid. Enough people did not that Feldman could track the gap.

Put together, the chapter argues that cheating is predictable. Raise the stakes on a single metric, keep monitoring weak, and some people will game it. Lower the chance of being caught, increase the payoff, and more will join. Sumo had tradition and honor codes. Schools had mission statements. Neither stopped the math of incentives.

Secrets lose power when everyone can read them

Chapter 2 shifts from cheating to information. Levitt argues that experts keep power by controlling what clients cannot easily verify.

His main case is the Ku Klux Klan. In the 1940s, an activist named Stetson Kennedy infiltrated Klan-linked groups in Georgia and passed their secret rituals, passwords, and meeting details to the writers of the Superman radio show and other outlets. The show turned Klan secrets into jokes for kids. Levitt compares this to a business losing a patent. When outsiders learned the code words, the thrill of secret membership faded. Recruitment suffered. Fear remained, and Levitt does not minimize the violence, but the aura cracked.

I need to add a correction the authors themselves later added. After the first edition, historians and journalists pointed out that Kennedy embellished parts of his own story. In the revised and expanded edition, Levitt and Dubner admit they leaned too hard on a colorful source without checking enough. The broader point about information hurting groups that depend on secrecy still stands, but the hero story is messier than the first printing suggested. It is a good reminder that the book is strongest when it shows its data and weakest when it trusts a great anecdote.

The cleaner example, and the one that has aged better, is real estate.

Levitt looked at home sales data and found that agents kept their own houses on the market longer and sold them for higher prices than comparable client houses. The reason was commission structure. On a typical 6 percent commission split several ways, an extra ten thousand dollars on your sale price might mean only a few hundred dollars extra for the agent, while costing weeks of open houses and paperwork. For a client, ten thousand matters. For the agent, a fast sale at a slightly lower price often pays better per hour.

He then looked at listing language. Words like fantastic and charming and terrific appeared more often in listings that sold quickly at lower prices. Words describing concrete features, like granite, maple, or a specific location detail, appeared more often in houses that waited longer and sold higher. Correlation is not proof that the words caused the price, and Levitt knows it. His reading is that agents use vague praise when they want a fast sale and specific facts when they are willing to wait. The words leak what the agent really thinks.

The same logic hits other experts. Levitt talks about funeral directors, car salesmen, and stock promoters. The pattern repeats. When you cannot judge quality directly, you trust the expert. When data becomes public, through the internet or disclosure rules, the expert has to work harder. That chapter feels even more true now, after Zillow, Carfax, and rate review sites. Levitt did not predict those sites exactly, but he described the pressure that created them.

There is also a short, sharp section on campaign finance that still sparks arguments. Levitt argues that money does not buy elections as simply as people think. Good candidates attract both votes and donations, which makes spending look more powerful than it is. Cutting spending in half would not cut votes in half. I found this part less convincing than the real estate work because races differ so much, but it fits the theme. Follow the incentive to donate. Donors give to likely winners to buy access, not only to change outcomes.

Crack gangs are a pyramid and houses are a signal

Chapters 3 and 4 move from cheating and secrets to labor markets and parenting. They are the most story driven part of the book.

Chapter 3 asks why drug dealers still live with their moms. The phrase comes from Levitt working with sociologist Sudhir Venkatesh, a University of Chicago graduate student who walked into a Chicago housing project with survey forms and almost got himself hurt. Instead of leaving, he spent years with a gang that sold crack. He kept financial ledgers for a local leader Levitt calls J.T.

The ledgers were shocking. Street level sellers earned very little, often below minimum wage on an hourly basis, while facing high risk of arrest, injury, and death. J.T., the mid level manager, did well, with cars and watches and status. The top bosses did very well. The structure looked like a tournament or a pyramid. Many young men fought for a tiny chance at the top because legal jobs nearby paid poorly and offered little respect.

Levitt compares it to acting or pro sports. Thousands chase a dream. A few win big. Most earn little. The difference is that in the crack trade the losers risk prison or death. The chapter also tracks the crack era itself. Crack arrived in the 1980s, profits were high at first, then competition drove prices down. Violence rose as gangs fought over turf. By the 1990s, the market had matured and street wages fell further.

I liked how human Venkatesh made J.T. He was not a cartoon kingpin. He went to college for a time, worked in sales, and managed people with a mix of threats and pep talks. He kept books. He worried about morale. He even did community outreach when it suited him. That detail stuck with me more than the wage math. Criminal firms face the same management problems as legal firms, only with worse HR options.

Chapter 4 turns to real estate agents again, but from the seller side, and then to dating and baby names. Levitt asks whether agents really get you the best price. The data says they get themselves out fast. He also looks at online dating patterns and at names.

The names research, done with Roland Fryer, is fascinating and uncomfortable. Levitt tracks distinctively Black names versus distinctively white names and later, in follow up work, resumes with names like Lakisha and Jamal versus Emily and Greg. The resume studies found callbacks differed sharply. Employers favored white sounding names. That finding has been replicated and debated, and Levitt presents it as evidence that discrimination persists in hiring.

But he also argues that names themselves do not cause poverty or wealth once you control for background. A child named Winner and a child named Loser, both real cases he cites from New York records, did not succeed or fail because of letters on a birth certificate. Their names signaled the circumstances and resources of their parents. Names are a signal, like paint on a house. Changing the signal without changing the underlying facts rarely changes the outcome for long.

That sets up the parenting chapter, which many readers found either freeing or infuriating.

Levitt uses data from the Early Childhood Longitudinal Study, a federal project that tracked thousands of American children from kindergarten onward. He tests which family factors predict test scores. Being read to often correlates with higher scores, but he argues the reading itself may matter less than what it signals about the home. What predicts scores more strongly are factors like parental education, income, birth weight, and neighborhood stability. What matters less than people expect, in his regressions, are things like museum trips, spanking, watching a lot of TV, or whether a mother stayed home.

His punchline is blunt. Who you are as a parent matters more than what you do as a technician of parenting. Obsessive tactics matter less than the environment, habits, and resources you bring. Adopted children studies and birth outcomes point the same way. This does not mean parenting is irrelevant. It means the easiest things to market to anxious parents often have the smallest measured effect.

He opens the same chapter with the gun versus swimming pool question. In the U.S., a gun in the home is far more likely to be involved in an accidental shooting than pools are in drownings? Actually he flips the fear. Parents worry about guns, and they should take precautions, but for young children a backyard swimming pool carries a higher risk of death than a gun in the house, because exposure is constant and supervision lapses. The point is about risk perception. We fear vivid, rare events and underrate familiar ones. Experts and news coverage make this worse by highlighting dramatic cases instead of base rates.

Abortion, crime, and the argument that would not die

Chapter 5 is the famous one and the most contested. Levitt asks why violent crime in the U.S. fell so sharply in the 1990s after rising for decades.

He walks through popular explanations and rejects most of them. A strong economy? Crime fell during weak years too. More police and prisons? Those helped, but timing and scale do not fit the full drop. Gun control and new policing tactics in New York? Local effects were real, but crime fell in cities without those policies. The crack market calming down? That mattered, yet the decline started before crack faded and continued after.

Then he offers his answer with economist John Donohue. Legalized abortion after Roe v. Wade in 1973 reduced the number of children born into circumstances linked to higher risk of later criminal involvement. Fewer unwanted births, especially among teens and very poor households with limited support, meant fewer teenagers in high risk environments fifteen to twenty years later. Crime started falling around 1991, about eighteen years after 1973. States that legalized abortion earlier or had higher abortion rates saw earlier or larger drops. Arrest data by birth cohort pointed the same way.

Levitt stresses he is not making a moral claim. He is making a statistical claim about unwantedness and age structure. Children who grow up with more stable support, wanted and cared for, are less likely to commit violent crime. Anything that shifts that mix can affect crime rates years later.

This is where I have to slow down and be fair to critics, because this chapter drew heavy fire and some of it landed.

Economists Christopher Foote and Christopher Goetz found a coding error in the original paper version of the work. Levitt and Donohue acknowledged it and revised the analysis. Others, like Theodore Joyce, argued the link broke down when you looked more closely at specific cohorts and state differences, and that crack and policing deserved more weight. Later researchers pointed to lead paint removal, changes in policing, incarceration, and demographics as overlapping causes. Most scholars today treat legalized abortion as one possible factor among several, not the single hidden cause the chapter sometimes sounds like.

Levitt himself, in later interviews and in SuperFreakonomics, softened the tone a bit and admitted forecasting is hard. The 2005 chapter reads more confident than the evidence warranted. That confidence made for great sales and sharp debate, but it also taught a generation of readers the wrong lesson if they walked away thinking one variable explains crime.

I still think the chapter is worth reading, because the method question matters. How do you test a cause that acts with an eighteen year lag? How do you separate cohort effects from period effects? Even if you reject the abortion link entirely, the walk through bad explanations is useful. It shows how easy it is to credit a mayor, a police chief, or a new law for a trend that started before they acted.

The book closes with a short Epilogue about the path from research to storytelling. Levitt says he never set out to be a public figure. He liked riddles. Dubner liked making those riddles readable. They promise more questions rather than final answers.

What holds up and what bugs me

What holds up best is the habit of asking about incentives and measurement. After reading this book, I catch myself asking who gets paid for a number and whether anyone checks it. School testing scandals since 2005, from Atlanta to other districts, followed the exact script Levitt and Jacob described. The sumo pattern has been debated, but later studies of tournaments and judging still find favor trading where stakes spike. The real estate work reads stronger now that listings are public and researchers can test word patterns at scale.

The bagel man and the daycare fine also hold up as teaching tools. They are small, clean examples you can explain in two minutes. They changed how I think about honor systems at work. People are mostly honest when watched by peers they respect, and less honest when rules feel distant or unfair.

What bugs me falls into three buckets.

First, the book sometimes sells correlation as a reveal. Listing words correlate with price, but agents choose words and pricing strategy together. Names correlate with outcomes, but background drives both. Levitt usually notes this, then moves on fast to keep the story moving. A careful reader needs to supply the caution the prose skips.

Second, sourcing is uneven. The KKK material is the clearest miss, and to their credit the authors flagged it in the revised edition. Other stories lean on one data set or one field partner. Venkatesh did brave, long fieldwork, but it was one gang in one project at one time. Generalizing to all drug markets stretches it.

Third, the tone can sound smug about experts while acting expert. Levitt mocks criminologists, pundits, and realtors for bending facts, then presents bold claims with wide error bars. The crime chapter is the main case. The parenting chapter is the softer case. Telling tired parents that museum trips do not move test scores may be true on average, but averages hide a lot. A kid who loves art may still gain plenty from a museum, even if the test does not catch it.

Who should read it? Anyone curious about applied statistics who does not want a textbook. High school and college students get a lot from the first three chapters. Parents may find Chapter 4 freeing, though they should pair it with a child development book for balance. Activists and policymakers should read the crime chapter alongside its critics, not alone.

If you want the full picture, read the 2006 revised and expanded edition if you can find it. It adds New York Times Magazine columns and a note correcting the Klan section. The core chapters did not change much. The extra material shows how Levitt and Dubner kept hunting for new riddles after the book hit.

I finished Freakonomics faster than I expected for a book about regressions. That is Dubner doing his job. I also argued with it more than I expected. That is Levitt doing his. The best page in my copy is now the one where I wrote in the margin: who benefits if this number is high? It works for test scores, wrestling records, house prices, and crime stats. It even works for bestseller lists. A book that teaches you to ask that question, even when you doubt some answers, earns its spot.

FAQ

Who wrote Freakonomics?

Economist Steven D. Levitt supplied the research and journalist Stephen J. Dubner wrote it with him. The book is a true collaboration, filed here under Levitt to match the reading list.

What do teachers and sumo wrestlers have in common?

Both face sharp rewards for hitting one number, test scores or an eighth win. Levitt shows Chicago answer erasing and win trading in sumo follow the same incentive to cheat.

Is the abortion and crime theory accepted?

No, it remains contested. Later work found a coding error, timing problems, and rival causes like crack decline, policing, and lead removal. Most scholars now treat it as one possible factor, not the full cause.