Behavioral Economics Books: A Critical Reading Guide

Most people who start a new job do not fill out the retirement-plan paperwork right away. At one large US company, putting it off used to mean you were not in the 401(k). Then the company switched to automatic enrollment: new hires were signed up unless they said no. Brigitte Madrian and Dennis Shea studied what happened next. Among employees with three to fifteen months at the company, participation rose from about 37% to about 86%. The people had not changed. What happened when they did nothing had.

That is a big result if your goal is getting people into the plan. It says less if your goal is helping them retire with more money. Most of the new members stayed with the company’s preset choices: 3% of pay, invested in a money-market fund. The researchers checked the company’s stock purchase plan and found little sign that people simply moved money out of it. But they could not see anyone’s other savings or debts, so they could not say whether workers ended up better off at retirement.

Keep that gap in mind as you read the books below. Most of them lean on results like this one. For each result, ask two questions: what actually changed, and does that change tell you about the outcome you care about? The guide covers 13 books, from Thinking, Fast and Slow and Nudge to Misbehaving and Noise; the table below matches each to the question it helps you answer.

Choose a book for your question

What you want to understand	Start with	What to watch for as you read
When to trust a quick judgment and when to think it through	Thinking, Fast and Slow	Whether a fast answer rests on practice with feedback or only feels convincing
How the way choices are presented affects decisions	Nudge	What happens if a person does nothing, what it takes to choose otherwise, and who benefits
How to do something you keep meaning to do	How to Change	What is actually stopping the person, and whether the suggested help addresses it
How other people persuade us	Influence	Whether a tactic changes a decision, why it might work, and whether saying yes is good for the person who says it
Why behavioral economics challenged older models of choice	Misbehaving	Whether a puzzling observation leads to a tested prediction or a useful intervention
How a shortage can affect attention as well as choices	Scarcity	Which problems come from having too little and which come from how people respond to it
How the field’s famous ideas and researchers became influential	The Undoing Project	Where the story of a discovery ends and the evidence for it begins
Why people given the same case can reach different judgments	Noise	Whether making decisions more consistent would also make them more accurate
Why a promising result may fade when a program grows	The Voltage Effect	Who took part in the first test, who would join later, and what the larger program would cost
How to read risk claims without being misled by a big percentage	Risk Savvy	How many people are affected, not only the relative change
How cues and surroundings shape repeated actions	Good Habits, Bad Habits	Whether making it easier to start also helps someone keep going
Why an investment plan that looks sound on paper can fail the person following it	The Psychology of Money	Whether the person could stick with the plan through a loss or a change in circumstances
Whether a product’s design explains why people come back to it	Hooked	The user’s actual reason for coming back, not just a diagram of the product

These books come from more than one field. Misbehaving and The Undoing Project tell the history of behavioral economics. Nudge is about how choices are designed. Thinking, Fast and Slow draws heavily on cognitive psychology. The last three are related books on habits, money and product design. Reading across them can help you ask better questions, but a lesson about one kind of decision does not automatically explain another. For an account of the field itself, see what behavioral economics is.

These are short reading notes, not 13 full book reviews. Five of the titles (How to Change, Influence, Good Habits, Bad Habits, The Psychology of Money and Hooked) also have full reviews on this site. The notes on the other eight rest on each book’s introduction, an authorized excerpt or the publisher’s description, plus the studies discussed here. They do not assess every chapter of those books.

Thinking, Fast and Slow

Would you buy shares in a company because you liked its cars? In the introduction to Thinking, Fast and Slow, Daniel Kahneman describes the chief investment officer of a large financial firm who had recently bought Ford stock. Asked why, the officer talked about how much he had liked the Ford cars he saw at an auto show. Liking the cars and knowing whether the shares are worth their price are different questions. Kahneman does not say how the investment turned out. That is not his point. The story shows how a quick, satisfying answer can take the place of the answer a decision actually needs.

Two questions, two kinds of thought

The officer’s good impression of the cars came easily. Judging whether Ford stock was underpriced would have taken information about the share price and the company’s prospects. Kahneman uses two characters to describe the difference. System 1 produces quick thoughts. System 2 handles the slower work that takes concentration. They are a way of describing mental activity, not two separate decision makers in the brain. Psychologists study the distinction in dual-process theories.

So the trouble in the Ford story was not simply that the officer answered quickly. He answered an easier question than the one in front of him.

Fast thinking can also draw on real expertise. Kahneman describes a skilled chess player who recognizes a promising move almost at once. The officer and the chess player both answered fast. One answered the wrong question. The other recognized a pattern learned through play. How can you tell which kind of fast answer you are getting?

When has a quick judgment earned your trust?

Kahneman and Gary Klein took up that question in a 2009 paper on intuitive expertise. They name two conditions that can support accurate intuition. The task has to have patterns that people can learn, and the person needs enough chances to learn those patterns through experience and feedback. If the task changes unpredictably, or nobody checks how earlier judgments turned out, years of making judgments may teach a person little.

Try that test on a designer who has watched many customers use signup forms. She looks at a new form and spots, at once, a label like one that customers have repeatedly misread. Then she predicts that changing the label will increase purchases. She may feel equally sure of both judgments. Watching people use forms has taught her what a confusing label looks like, but that kind of watching does not show what label changes do to sales. Unless she has also seen purchase results after similar changes, her experience backs the first judgment and not the second.

Speed, then, tells you little on its own. What matters is whether the person has had the chance to learn to answer this particular question well.

How to check a quick answer

Ask three questions. First, which question does the quick answer actually address? The Ford officer answered one about cars when he needed one about shares. Second, what experience and feedback does the person have with decisions like this one? The designer has learned a lot about confusing labels and much less about sales. Third, what facts does the decision still need?

Slowing down helps when you use the time to ask the better question or check those facts. Time alone cannot supply information you do not have.

Nudge

Someone always decides what doing nothing means

Go back to the company at the top of this page. Before the change, a new hire who did nothing stayed out of the 401(k). After it, a new hire who did nothing was in. Either way, the company had decided what would happen to people who took no action. There is no way to avoid making that decision. Richard Thaler and Cass Sunstein use choices like this one in the preface and introduction to Nudge: The Final Edition (2021).

They call the way options are set up choice architecture. It includes which option is selected in advance (the default) and what someone has to do to choose something else. A nudge, in their sense, changes that setup. It leaves every option available, does not substantially change economic incentives such as prices or cash rewards, and is easy and cheap to avoid.

What automatic enrollment changed

Under opt-in enrollment, you have to sign up to join. Under automatic enrollment, you are in unless you decline, and you can change the preset contribution rate and investment. Nobody has to become more motivated to save first. The change works on what happens when people do nothing.

Madrian and Shea’s study at that employer (2001) found much higher participation after automatic enrollment, and most new members kept both preset choices: how much of their pay went in and where it was invested. The researchers compared groups of employees hired before and after the policy change; they did not randomly assign individual employees to the two systems.

That result matters. Employees joined an actual retirement plan, and money went into their accounts. If your aim is to increase participation, changing the enrollment rule can help you get there.

What does enrollment tell us?

It tells us who is in the plan. It does not tell us how much someone saves overall, or what they own once their debts are subtracted. Money going into one account does not, by itself, tell us what happened to other accounts or debts. Nor does automatic enrollment show that anyone learned to budget or began making a deliberate choice to save, again and again. Each of those claims needs its own evidence. The behavioral economics examples follow this question into later research on contributions and borrowing.

That gap is the point of my criticism of defaults presented as behavior change. It matters because it shapes what people try next. A system can enroll you in an exercise class. It cannot do the exercise for you. If regular exercise is the goal, you have to measure whether people actually exercise.

The default itself also needs scrutiny. If employees keep the default settings, those settings decide how much they contribute and how the money is invested. At this employer, that meant 3% of pay in a money-market fund. Higher participation does not show that every part of that arrangement serves each employee well.

The authors limit their own claims. Their introduction does not present defaults as the answer to every problem, and it discusses settings where experience and feedback help people choose for themselves. Those qualifications make their argument clearer, not weaker.

Read Nudge with a real decision in mind, such as joining a retirement plan. Sketch what happens if someone does nothing, what it takes to choose something else, and what result the designer wants. Then ask whether that result is good for the person choosing. One more caution: a default that helps in one setting does not show that every other technique called a nudge will help.

How to Change

Find the obstacle first

Suppose someone misses a flu-shot clinic. If they forgot about it, a reminder might help. If the clinic closes before their shift ends, no reminder will solve the problem. Katy Milkman’s How to Change (2021) focuses on obstacles such as forgetting, putting things off, giving in to immediate temptations, sticking with what you already do, and a lack of confidence. It also covers getting started, social influence, and how change can last. Before using one of its techniques, I would check for a simpler obstacle outside the person’s control, such as the clinic’s hours.

The obstacle decides what is worth trying. The two studies below test one of the book’s techniques, making a specific plan, and ask a concrete question: when people were asked to plan, did more of them actually follow through?

How a date-and-time plan raised workplace flu shots

In a workplace vaccination experiment by Milkman and colleagues (2011), 3,272 employees of a Midwestern utility were mailed information about free flu-shot clinics at work. Some mailings also had a space to write down the date they planned to go; others asked for both the date and the time. Writing in the space did not book an appointment. It made a private plan.

Here is the share of employees who got vaccinated at a workplace clinic:

• Mailing with no planning prompt: 33.1%.
• Mailing that asked for a date only: 35.6%, a difference from the no-prompt group that the researchers could not distinguish from chance.
• Mailing that asked for a date and time: 37.1%.

The gap between the date-and-time group and the no-prompt group was 4 percentage points, or roughly four additional vaccinations per hundred eligible employees. After accounting for measured differences such as employees’ age and work location, the researchers estimated the increase at 4.2 points.

All three groups were sent the same information about the free clinics. The date-and-time mailing added only a space to write down when they would go, and the employees who got that space were more likely to get vaccinated. The outcome was an actual vaccination, not a stated intention. That is evidence that the prompt helped in this setting.

Notice what the prompt had to work with. The clinic was at work and the shot was free; the prompt created neither. If you already offer something that convenient, a planning prompt may be worth testing. If you do not, this experiment does not tell you that a planning box will make up for it.

When gym members made plans

Planning gave a different result in a separate randomized study of 877 gym members by Carrera and colleagues (2018). Members asked to plan the days and times of their visits did not go to the gym more often over the following two weeks, according to the gym’s check-in records. The researchers also confirmed that people took part in the planning exercise, so the result cannot be explained away by saying nobody made a plan.

This was not a repeat of the vaccination trial. Getting one shot at a clinic at work is a different task from going to the gym again and again, and the two studies were designed differently. They do not tell us exactly why the results differed. They do show why “planning works” is too broad a conclusion. Before recommending a plan, ask whether the person needs to show up once or keep coming back week after week.

Does the book help you choose the right technique?

My full review of How to Change separates two questions. Did a particular technique help people in a study? And can the book help you choose the technique most likely to help you? The vaccination experiment is evidence of the first kind. The second needs a different test: one of how the book matches advice to people.

Picture two people who both say they are forgetful. One misses a class because it slips their mind. The other remembers but cannot leave work on time. A reminder could help the first and do little for the second. To test the book’s matching, researchers would first need to find out what is actually stopping each person. Then they would compare the book’s recommendation with other reasonable ways to help. A collection of successful examples, sorted into chapters, does not show that the book can make that match.

Use the book to find possible approaches. Then check that the obstacle it names is the one actually stopping you.

Read the chapter on lasting change

Chapter 8, “Changing for Good,” describes programs that increased gym attendance while they were running. Some of the increase lasted after the programs ended, but the gains were smaller than the researchers had hoped. Milkman uses the experience to argue for continued support.

That lesson should change how you read the earlier chapters. If reminders help you keep going to a class, do not assume you will keep going once the reminders stop. Continued access and continued effort may matter too. Count that support and work as part of the approach. Needing ongoing support does not make an approach a failure; what matters is whether the benefit is worth the ongoing cost.

Start with the chapter that describes your problem, read the chapter on making change last, and check the evidence appendix before relying on any particular study. The review identifies which techniques have stronger evidence and which claims need correction.

Influence

Could a small favor change what someone buys? In Influence, Robert Cialdini discusses an experiment that tested this with a soft drink and some raffle tickets. It is a good case for three separate questions: did the favor change what people bought, why might it have done so, and did the purchase serve the buyer?

What the soft-drink experiment showed

In Dennis Regan’s experiment (1971), some students were given a soft drink by the man who later asked them to buy raffle tickets. Others got a drink from the experimenter, or no drink at all. Regan analyzed results from 77 male students. Those who got a drink from the man doing the asking bought more tickets than either of the other groups.

That is more than an anecdote about a clever salesperson. Regan counted tickets people actually bought, and he compared the requester’s gift with two alternatives: a gift from someone else and no gift. That design supports a conclusion about this experiment. Whether a favor would have a similar effect on other people or in another setting needs more evidence.

Cialdini’s seven principles

Cialdini treats the requester’s gift as an example of reciprocation. His 2021 New and Expanded edition organizes its account of persuasion around seven principles:

• Reciprocation: feeling obliged to return a favor.
• Liking: being more willing to agree with someone you like.
• Social proof: looking at what other people do to decide what you should do.
• Authority: taking direction from someone you see as knowledgeable or in charge.
• Scarcity: finding something more desirable because it is hard to get or might become unavailable.
• Commitment and consistency: wanting your choices to agree with what you have already said or done.
• Unity: responding to someone as “one of us,” with a sense of shared identity.

These principles give you possible explanations to examine, not proof of why a particular person agreed. Nor do they tell you whether agreeing was a good decision for that person.

Why did they buy, and did buying help them?

Back to Regan’s students. Why did a drink from the requester lead them to buy more tickets? They may have felt obliged to return the favor. But the purchases alone cannot tell us whether obligation was the reason, or how much of the effect it explains.

Did buying the tickets help the buyer? A tactic can increase sales without leading the customer to a better decision. The book’s defensive advice needs the same kind of test. If it suggests a way to resist an unwanted request, we need to know whether people who use that advice actually get better at saying no.

Some of that advice has been tested. In three experiments by Brad Sagarin, Cialdini and colleagues (2002), participants learned to tell a legitimate authority, someone with relevant expertise, from an illegitimate one. Afterwards, ads that relied on illegitimate authority seemed less persuasive to them, and ads with legitimate authority seemed more persuasive. The lesson worked best when participants also saw that they themselves could be fooled. The studies measured how people judged ads, not whether they later avoided purchases they regretted.

So Regan’s experiment supports a conclusion about purchases, but on its own it cannot tell us why people bought or whether buying helped them. My full review of Influence examines the evidence for particular persuasion techniques and the wider claims made about them.

You can put any of the seven principles through the same questions. Take authority. A recommendation from someone with relevant expertise can be useful because they may know something you do not. The same recommendation from someone whose impressive title has nothing to do with the question deserves less weight. Whether you should listen depends on what the person knows, not on the word “authority.”

Now try scarcity. Suppose a seller says a product will be unavailable after tonight. If the deadline is real, you have learned when the product will stop being available; you have not learned whether you need it. If the deadline is invented, the seller has misled you. Cialdini himself warns against false claims of scarcity, popularity or expertise. You can check whether a claim is true, and whether acting on it serves you, without assuming that every attempt at persuasion is dishonest.

Read Influence with a real request in mind. Write down what you are being asked to do and what makes you want to say yes. Then ask whether saying yes serves what you care about. Use the evidence appendix to look into a particular experiment or the book’s advice on resisting persuasion. The book can help you notice what makes a request appealing; you still need to check the evidence for each explanation.

Misbehaving: Where did behavioral economics come from?

Suppose you bought a bottle of wine years ago for about $5. You could sell it today for $100. You would not pay more than $30 for a bottle, yet now and then you open this one and drink it. Richard Thaler tells this story in his Nobel lecture about the development of behavioral economics. If you could really sell the wine for $100, drinking it means giving up that $100. So why does selling a bottle you own feel so different from buying an identical one? Thaler uses the story to illustrate the endowment effect: valuing something more once it is yours.

Misbehaving is the book to read if you want to follow that question into the history of the field. Thaler describes the cases he collected in which people behaved differently from the simplified decision makers in economic models, and the resistance he met when he tried to make those observations part of economics. The publisher’s description points to questions about buying, selling, saving and markets. This is the story of a research program, not a catalogue of tricks to use on customers.

The wine story also shows how to read the book without accepting every interpretation in advance. The price paid years ago should not decide whether selling today is worthwhile. But the fact that someone drinks an old bottle does not, by itself, tell us why. They might value that particular bottle, dislike the hassle of selling it, or simply not think of a possible sale as a cost.

So Thaler’s examples can show you where a model leaves something out. To decide what belongs in its place, look for a prediction tested against actual choices. And before using an insight to change a policy or product, ask whether it improved the result you care about, not just whether the original story sounded familiar.

Scarcity: What happens when there is too little to go around?

You have enough money to cover this month’s rent or the car repair that gets you to work, but not both. Which bill gets your attention first? A payment reminder might help if you forgot a due date. It cannot pay a bill you cannot afford. Keep that difference in mind as you read Sendhil Mullainathan and Eldar Shafir’s Scarcity. By scarcity, they mean having too little time or money for what you need. (In Influence, Cialdini uses the same word for the appeal of a product that is hard to get. That is a different problem.)

The opening of Scarcity sets Sendhil’s growing list of overdue work beside Shawn’s mounting debts and late fees. Both keep making reasonable plans to get ahead. Both find that the next urgent demand eats up time or money they had already committed. The authors propose that having less than you feel you need does more than restrict your options. It also pulls your attention toward the immediate shortage. Sometimes that helps you focus on it, while making other tasks harder to handle.

Read the book if you want to understand that proposed link between shortage and attention, rather than treating every missed payment as a failure of character.

The idea has evidence behind it. In one study by the authors and colleagues, the same sugarcane farmers did better on cognitive tasks after the harvest than before it, when money was tighter. Testing the same people twice is more informative than simply comparing rich people with poor people, who differ in many other ways. It still does not show that a reminder, a budgeting lesson or any other particular program will help people facing a shortage.

How far the idea reaches is still an open question. A test in the United States did not find the same pattern. Carvalho, Meier and Wang (2016) randomly assigned low-income households to answer an online survey either before or after payday. Those surveyed before payday had less money in cash and in their bank accounts, and in choices about money they behaved as if they were more impatient. But the researchers found no before-and-after difference in performance on cognitive tasks, in risk-taking or in the quality of decisions.

The two studies involve different people and different shortages: a once-a-year harvest income in one, the days before a regular paycheck in the other. One result does not cancel the other. Together, they show that the harvest finding cannot simply be assumed to hold wherever money runs short.

The wider evidence is mixed. A later review found scarcity theory coherent but not fully consistent with the available evidence. A replication audit found mixed results for brief scarcity prompts used in experiments. The audit’s authors corrected errors in 2022, including removing one reported replication. In 2023, the original researchers challenged the audit’s choice of studies and methods. In their reply, the audit’s authors accepted that three of the included studies had measured scarcity rather than created it, but defended their broad selection and the value of independent replications.

Keep two distinctions in view. A brief prompt in an experiment is not the same as living with poverty. And the mind may respond to a shortage, but the shortage itself remains real.

The Undoing Project: When a scout and a model disagree

In 2007, a statistical model used by the Houston Rockets liked a basketball player named Marc Gasol. The team’s scouts saw an unflattering photograph, gave him a nickname and cooled on him. The Rockets passed, and Gasol went on to become an All-Star. In the opening of The Undoing Project, Michael Lewis uses the episode to show the problem facing the Rockets executive Daryl Morey: a vivid impression can overwhelm a prediction built on far more observations.

If the story ended there, it would be easy to declare the model the winner. Lewis does not let us. The model also strongly favored Joey Dorsey, who disappointed, and dismissed DeAndre Jordan, who became a star. In Jordan’s case, a scout had information the model lacked. So the useful question for Morey was not which side was right. He needed to learn when the model’s data helped, when a scout saw something the data missed, and whether either source improved predictions across many players.

That opening leads into Lewis’s story of Daniel Kahneman and Amos Tversky: their collaboration, their questions about judgment under uncertainty, and the influence those questions had far beyond psychology. Read the book if you want to understand the people and the history behind many claims you will meet elsewhere in behavioral economics. The opening alone shows Lewis’s skill at explaining why a question mattered to the people who faced it.

Keep the Rockets question with you as you read. The story gives you reasons to doubt both unexamined intuition and blind faith in a model, but it does not compare the two across every kind of decision. In the same way, a compelling account of Kahneman and Tversky’s partnership cannot establish the size, reliability or practical value of each effect linked to their work. For that, read the original studies and later tests, and ask what each one actually measured. A history like this can explain why researchers asked their questions. It should make you more curious about the evidence, not less.

Noise: A Flaw in Human Judgment

Imagine giving the same insurance application to two underwriters. One quotes a much higher price than the other. Even before you know which price is right, you have a problem: what the customer pays depends on who opened the file. Daniel Kahneman, Olivier Sibony, and Cass Sunstein call this kind of unwanted variation in judgment noise. In an excerpt from Noise, they make a point that is easy to miss when we focus only on bias: you can measure how far judgments spread without knowing the correct answer.

One practical idea from the book is the noise audit. Give several people the same realistic cases, have them decide independently, and compare their answers. In an interview about the book, Kahneman and Sibony explain why ordinary work hides the problem: a case usually goes to one loan officer, interviewer or judge, so colleagues rarely learn how differently they would have handled it. An audit brings those differences into view. If a hiring team gives the same application very different ratings, it can ask which facts each person noticed and which criteria each thought mattered.

I would be careful about what an audit proves. A team can agree because everyone follows the same poor rule. An insurer can quote consistently bad prices. Interviewers can agree on a trait that has little to do with doing the job well. Less spread is valuable when people in the same role should treat comparable cases alike. It does not, by itself, show that the decisions are accurate or fair. To learn that, you also have to ask what happened after each decision and who paid for a mistake.

Read Noise if you make repeated, comparable judgments or run an organization that does. Ask first whether the variation is real and unwanted. Then keep the harder question open: if everyone followed the same process, would they make better decisions for the people affected?

The Voltage Effect: How to Make Good Ideas Great and Great Ideas Scale

John List helped build a preschool research project in Chicago Heights and hoped other school districts could use what it learned. Should they? The introduction to The Voltage Effect turns that hope into a question about scale. Can an idea that helps a small group keep helping when more people use it, in places with different staff, budgets and participants?

One part of the Chicago Heights project was a Parent Academy, and List coauthored a randomized study of it. Parents were offered classes and on-site childcare. They could also earn money for attending sessions, for completing activities with their children, and for their children’s performance on assessments. So the study tested the whole package, not the payments alone.

The researchers reported higher scores on vocabulary and early-learning tests, and on measures of attention and self-control, for Hispanic children and a small group of White children. They found no clear benefit for Black children. The White estimate rests on only 15 children with measured outcomes, and the study could not establish why the results differed by group. Nor did this one-year test show what would happen if another district took the program to scale.

List’s five checks give a school district a better starting point:

• Was the first result a false alarm?
• Were the first participants like the people the new program will serve?
• Did success depend on something that cannot be reproduced at scale, such as exceptional staff?
• Will a wider rollout have side effects that work against the program?
• Will it cost too much to sustain at a larger size?

The introduction is useful because it makes those questions hard to skip. No checklist can answer them for a particular program, though. Before expanding one, test it with the people and resources it will actually have, and measure the result you care about instead of assuming an encouraging pilot will carry over.

Risk Savvy: How to Make Good Decisions

Suppose a campaign raises the share of people who complete an application from 10% to 11%. You could call that a 10% improvement. It is also one extra completion per hundred people. Both descriptions are correct, but they invite different reactions. Which one would you want to see before approving the campaign?

Gerd Gigerenzer’s work pushes back on the idea that every disappointing decision reveals an irrational mind. Sometimes the problem lies in how the risk was explained. In a paper on helping doctors and patients make sense of health statistics, Gigerenzer and his colleagues argue for showing absolute risks and counts, not only relative percentages. A clearer description can give someone a better chance to understand the choice in front of them.

Risk Savvy (2014) takes that argument to general readers. Its publisher describes a book about how people misjudge risk, how often professionals misunderstand statistics, and how simple rules can support decisions about health and money.

Clearer numbers do not make a program worth running, though. In the application example, you would still need to know whether the campaign caused the extra completion, what that completion led to, and what else the money could buy. Read Risk Savvy when a big-sounding percentage is doing more work in an argument than the actual result.

Good Habits, Bad Habits

Why does a sensible plan disappear the moment your workday begins? Imagine that you mean to spend ten minutes deciding what matters most today. You sit down, open your inbox, and answer the first message. By the time you remember the plan, the morning is gone. Another lecture about discipline probably won’t change what happens when you sit down tomorrow.

Wendy Wood’s Good Habits, Bad Habits (2019) points you to that moment and the surroundings that shape it. Here, that could mean putting a blank planning sheet beside the keyboard and leaving your email closed until you have filled it in. Now the plan is the first thing you see, and opening your email takes a deliberate step. A familiar cue can help you start; making the competing action less convenient can also help. That is more useful than assuming every repeated failure reveals weak character. Wood does not dismiss goals, though. You still have to decide which action is worth repeating.

Now suppose it works and you begin planning every morning. Choosing between two projects still takes thought. Starting can become familiar while the work itself stays difficult. That distinction also helps with a famous number from Wood’s research. In two diary studies, Wood and her colleagues called an action habitual when a person reported doing it almost every day in the same place. In one of the studies, those actions made up an average of 43% of each participant’s sampled activities. The diaries also asked what people were thinking about, but the 43% figure alone does not show that nearly half of all action occurred without thought.

Read Wood if you want to work on the cues, friction and rewards around an action you have already decided is worth doing. Then ask whether those changes help you start, help you keep going, or merely make an unsuitable activity easier to repeat. My full review of Good Habits, Bad Habits follows the larger claims about habit, effort and lasting change through the evidence.

The Psychology of Money

Two people can understand the same investment chart and still feel differently about taking a risk. One began earning during a long market rise. The other watched a parent lose a job and then saw their own savings fall during a crash. If their plans differ, which one has misunderstood the numbers? Neither has to have misunderstood anything.

Morgan Housel’s The Psychology of Money (2020) asks a better question: what can each person live with, financially and emotionally, when the plan stops looking good? The chapters on staying wealthy, choosing a reasonable plan and leaving room for error are at their best when they put that question front and center. A plan that requires you to stay invested through a loss has to keep working when you need the money, lose your income, or find the losses harder to bear than you expected.

Part of Housel’s account has research behind it. Malmendier and Nagel (2011) found that people who had lived through lower stock-market returns reported less willingness to take financial risks and were less likely to own stocks, even after the researchers accounted for age, survey year and measured household differences. That is evidence of a link between what people lived through and how they invest. It does not show that any one person’s caution was caused by a past crash, or that experience matters more than income, education and opportunity. Nor do two memorable life stories settle which route to wealth works for most people.

Read this book if financial advice has ever sounded mathematically elegant but impossible to follow. Use it to ask whether a plan fits your actual needs and constraints. Its stories are good for seeing a problem; check the evidence before treating one as a general rule. My full review of The Psychology of Money examines those stories, the book’s numbers, and its claims about wealth and happiness in more detail.

Hooked

Suppose you check a transit app every morning. Opening it has become part of leaving home. You have saved your usual route, and the arrival time changes from day to day. A product team could map that onto a four-step loop: a prompt, an action, a changing result and an investment. But the app is also simply useful. It tells you when the bus will arrive. Does the loop explain your return visits any better than the app’s usefulness does?

Nir Eyal’s Hooked (2014) calls those four steps trigger, action, variable reward and investment. The sequence is easy to remember, and it can help a team notice where a product is hard to use or gives people no reason to return. Eyal presents it as a practical tool for product builders. That is a good way to read it: as a set of design questions, with the user’s problem at the center.

The trouble starts when the diagram becomes the explanation for every return visit. You check the bus time because you need the information; its unpredictability is not necessarily a special reward. Saving a route may improve the app, but that does not show the saving made your checking automatic.

A later six-week randomized study of a health app gives a narrower test of one ingredient. Its 61 participants were split into three groups that received fixed or variable reward arrangements. The study did not find a clear engagement advantage for variable rewards. It did not test the whole Hook sequence or a transit app, and a sample that small cannot settle the broad design question.

I am named in the book’s acknowledgments and have criticized its model before; the full review records those ties. Read Hooked if you build products and want ideas for experiments. First decide what a useful result would be for the person using the product. Then test a proposed change against a credible alternative, measuring that result rather than only return visits. My full review of Hooked examines the complete model, its individual studies, and what its success stories do and do not establish.

Successes, failures and open questions

The books above rest on very different kinds of evidence. Thaler’s wine story and Lewis’s Rockets draft show why people began asking some of the field’s questions; neither tells us whether a behavior-change program works. The vaccination trial tests a small change to a mailing. List’s Parent Academy trial tests a larger package of classes, child care and incentives in one district. Each teaches something worth knowing, and each has a different limit. A critical reader explains a success as carefully as a failure, and keeps an unfavorable result separate from a question nobody has answered yet.

Take the vaccination experiment. It is favorable evidence for adding a planning prompt to a clinic-information mailing in that setting. It did not test whether the same prompt can build a lasting exercise routine. Accept the finding it supplies and leave the second question open. Dismissing planning altogether would throw away this evidence. Calling it a general method for changing people would add a claim the experiment did not test.

How much should you expect these techniques to help in real government programs, compared with published academic trials? DellaVigna and Linos (2022) compared trials run by two government teams that test nudges with a separate set of trials published by academic researchers. The average improvements were positive in both sets, but much smaller in the government trials: 1.4 percentage points, against 8.7 in the academic trials. These were different trials, not the same interventions followed from small studies into large programs. The comparison also excluded defaults, so it does not tell us whether the retirement-enrollment example above works.

Both facts matter. The interventions helped on average, and the average improvement in the government trials was much smaller. That gives you a reason to test specific, inexpensive changes, and less reason to expect large improvements wherever you use them. Whether a small gain is worth pursuing depends on its cost, how many people it reaches and what else the money could do. A real improvement can be too small to justify the expense. A small one can be well worth it if it helps enough people at little cost.

My strongest objection is to the leap from a selection of interesting results to a promise of reliable behavior change across settings. That promise needs its own evidence. A field’s reputation, an author’s confidence and a memorable example cannot supply it.

When you read a claim about a technique, ask which of these four situations you are looking at:

• The study supports a result under the conditions tested. Give it credit, but be clear about what changed and what it was compared with.
• A later study repeats the test and does not support the original finding. Ask how much confidence the original still deserves, and whether differences in how the studies were done might matter.
• The same intervention works differently in another setting. Find out what changed before claiming that it works everywhere or nowhere.
• The study never tested the larger claim. Leave that question open. Not having an answer is different from having evidence that the claim is false.

Sorting claims this way lets you be demanding without deciding the answer in advance. It also lets you give a positive result its due without claiming more than the study showed.

For research papers rather than books, the behavioral economics reading list collects replication failures and critical views of the field’s findings.

Before you act on a claim

Pick one claim from one of these books. What example made it seem plausible? If the book cites a study, what did that study actually measure? What would you need to know before relying on the claim? For a history book, the claim may simply be about why someone began asking a question; that kind of book does not test every answer.

If the book recommends a technique, try finishing this sentence: “This might help someone do ___, because ___, provided that ___.” To fill the blanks, you have to name an action, a possible reason it might help, and the conditions it needs.

For the vaccination study, that might be: “A date-and-time plan might help someone attend a free workplace clinic, because it makes the intended visit more specific, provided they have a realistic opportunity to attend.” Notice that the middle blank is a proposed explanation, not a finding. The experiment showed an increase in vaccinations. It did not establish that making the visit more specific was the reason for that increase.

Now change the conditions. Suppose the clinic is a two-hour journey away and opens only during the person’s shift. A more specific plan removes neither obstacle. The workplace result still stands, but you would have reason to deal with travel and work hours before asking anyone to make a more detailed plan.

Then compare the tactic with other ways of reaching the goal. If someone wants more physical activity, reminders for an inconvenient gym class are only one option. An evening class, swimming or a walk with a friend may suit them better. This is the question behind choosing a behavior that fits: which action is worth supporting in the first place? That comparison needs testing too. If people switch to a better-fitting activity, do they actually stay active?

The practical-effect calculator can help you compare two percentages that measure the same thing. The evidence exercises let you practice separating what a study found from what it leaves unanswered. The full reviews give detailed assessments of individual books.

References