4 Likes, Wants, and Needs

THE INCREASINGLY OPEN EXPRESSION of unvarnished animus toward people of different race, religion, ethnicity, and even gender has become the staple of populist leaders throughout the world. From the United States to Hungary, from Italy to India, leaders who offer little more than racism and/or bigotry as their policy platform are becoming a defining feature of the political landscape, a ground force that shapes elections and policies. In the United States, in 2016, the degree to which a person deeply identified as white was one of the strongest predictors of support for Donald Trump among Republicans, much more than, for example, economic anxiety.1

The vicious vocabulary our leaders employ daily legitimizes the public expression of views some people probably had already, but were rarely spoken or acted upon. In one instance of everyday racism, a white woman in a supermarket in the US called the police on a black woman whom she suspected, on the basis of a phone conversation she was overhearing, was trying to sell food stamps—and in the process rather tellingly exclaimed, “We are going to build this wall.” On the face of it, the comment made no sense: the accused was an American citizen who belonged on the same side of the hypothetical wall as her white critic.

But of course we all know what she meant. She was expressing her preference for a society free of people different from her, with President Trump’s metonymic wall separating the races. This is why the wall has become such a flashpoint in American politics, an image of what one side dreams of and the other fears.

Preferences, at one level, are what they are. Economists make a sharp distinction between preferences and beliefs. Preferences reflect whether we prefer cake or cookies, the beach or the mountains, brown people or white. Not when we are ignorant of the merits of each and may therefore be swayed by information, but when we know everything we could possibly need to know. People can have wrong beliefs but they cannot have wrong preferences—the lady in the supermarket can insist she is under no obligation to make sense. Yet it is worth trying to understand why people have such views before we sink deeper into the morass of racism, especially because it is impossible to think about the policy choices we will confront in this book without getting a handle on what these preferences represent and where they come from. When we discuss the limits of economic growth, the pain of inequality, or the costs and benefits of protecting the environment, there is no way to avoid dealing with the distinction between what individuals need and what they want, and how society at large should value those desires.

Unfortunately, traditional economics is ill-equipped to help us here. The attitude in mainstream economics has been very much one of tolerance of people’s views and opinions; we may not share them, but who are we to pass judgment? We can shout out the facts to make sure people have the right information, but only they can decide what they like. Moreover, there is often a hope the market will take care of the problem of bigotry. People who happen to have petty, narrow-minded preferences should not survive in the marketplace, since being tolerant is good business practice. Take, for example, a baker who does not want to bake cakes for same-sex weddings. He will lose sales from all same-sex weddings, which will go to other bakers. They will make money; he will not.

Except it does not always work that way. Bakers who don’t want to bake for same-sex weddings don’t go bankrupt, in part because they win support from like-minded people. Bigotry can be good business, at least for some, and it seems to be good politics as well. As a result, economics in recent years has had to reckon with preferences, and we have gained some useful insights about how we might be able to get out of this mess.

In 1977, in a famous piece titled “De Gustibus Non Est Disputandum” (usually translated as “There Is No Accounting for Tastes”), Gary Becker and George Stigler, Nobel Prize winners and founders of the Chicago school of economics, made an influential case for why economists should avoid getting entangled in trying to understand what lies behind preferences.2

Preferences are part of who we are, Becker and Stigler argued. If, after we go over all the information we have, two of us still disagree on whether vanilla is better than chocolate or polar bears are worth saving, the presumption ought to be this is something intrinsic to who we each are. Not a whim or a mistake or a response to social pressures, but a considered judgment reflecting what we value. While they recognized that this is surely not always true, they argued that it is still the best place to start when we set out to understand why people do what they do.

We have some sympathy for the idea that people’s choices are coherent, in the sense of being thought out rather than a collection of random acts of whimsy. It is both patronizing and wrong-headed, in our view, to assume people must have screwed up just because we might have behaved differently. And yet society routinely overrules people’s choices, especially if they are poor, supposedly for their own good, for instance when we give them food or food stamps rather than cash. We justify this on the grounds we know better what they really need. To partially combat this attitude—only partially, because we don’t deny there are many misjudgments in the world—in our book Poor Economics we took some pain to argue that the choices of the poor often make more sense than we give them credit for.3 For example, we told the story of a man in Morocco. After he made a compelling case that he and his family really did not have enough to eat, he showed us his largish television with a satellite connection. We might have suspected the television was just an impulse purchase he had subsequently regretted. But that was not at all what he said. “Television is more important than food,” he told us. His insistence made us ask how this could make sense, and once we went down that road it was not that hard to see what was behind this preference. There was not much to do in the village, and given he was not planning to emigrate, it was not clear that better nutrition would buy him much more than a fuller stomach; he was already strong enough to do the little work that was available. What the television delivered was relief from the endemic problem of boredom, in these remote villages where there was often not even a tea stall to relieve the monotony of daily life.

The Moroccan did very much insist his preference made sense. Now that he had the television, any more money, he told us several times, would go to buying more food. This is entirely consistent with his view that televisions serve a greater need than food. But it flies in the face of most people’s instincts and many of the standard formulations in economics. Given that he bought a television when there was not enough food in the house, the presumption would be that any extra money in his hands would go even faster down some drain, since he evidently was the sort of person given to irrational impulses. This is at the base of the case against giving money to poor people. And yet, a number of recent studies from across the world, published after we made the case in Poor Economics that he knew what he was doing, have found that when randomly selected very poor people get some extra cash from government programs, they do spend a very large fraction of this extra money on food.4 Maybe after they buy that TV, exactly as the Moroccan man had promised.

So we learned something by being willing to suspend our disbelief and trust that people know what they want. Becker and Stigler however want us to go a step further—to assume preferences are stable, in the sense they are not influenced by whatever is going on around us. Neither schools, nor the exhortations of parents or preachers, nor the stuff we read on billboards or on our many screens, in this view, change our true preferences. This rules out conforming to social norms and being influenced by one’s peers, like getting a tattoo because everyone else has one, wearing a headscarf because it is expected of one, buying a flashy car because the neighbors have one, and so on.

Becker and Stigler were too good as social scientists not to realize this was not always the case. But they believed it was more useful to ponder why a particular seemingly irrational choice might actually make sense, rather than to close our minds to its potential logic and attribute it to some form of collective hysteria. This view was enormously influential; many, perhaps most, economists bought into this agenda of sticking to what have come to be known as standard preferences, meaning preferences that are coherent and stable. For instance, many years ago Abhijit was living in Manhattan and teaching at Princeton, and thus often found himself on a train. He noticed people often formed lines at specific places on the platform to wait for their train, but as often as not, the front of the line would be nowhere near a door to the train. It was a fad.

A natural conclusion might have been that people just went with the flow because they preferred doing the same as everyone else. This would have violated the idea that preferences are stable, because their preference for one place on the platform over another depended on how many people were there. To explain why people join fads without simply assuming they happen to like behaving like everyone else, Abhijit constructed the following argument. Suppose people suspect that others know something (perhaps the train door will open at a particular spot). They would then join the crowd (perhaps at the cost of ignoring their own information that the train is likely to stop somewhere else). But that would make the crowd bigger, and so the next person coming along would see an even bigger crowd and be even more likely to think this conveyed useful information. They might also join the crowd, for the same reason. In other words, what looks like conformity could be the outcome of rational decision making by many individuals with no interest in conforming, but who believe others might have better information than they. He called it a “simple model of herd behavior.”5

The fact that each individual decision is rational does not make the outcome desirable. Herd behavior generates informational cascades: the information on which the first people base their decision will have an outsized influence on what all the others believe. A recent experiment nicely demonstrates the power of random first moves to generate cascades.6 Researchers worked with a website that aggregates advice on restaurants and other services. Users post comments, and other users add an up- or down-vote. In their experiment, the website randomly chose a small fraction of comments and gave them one artificial up-vote as soon as they were posted. They also randomly chose another small batch to get a down-vote. The positive up-vote significantly increased the probability that the next user also gave an up-vote, by 32 percent. After five months, the comments that had received one single artificial up-vote at the beginning were much more likely to get a top grade than those that got a single down-vote. The influence of that original nudge persisted and grew, despite the fact that the posts had been viewed a million times.

Fads, therefore, are not necessarily inconsistent with the paradigm of standard preferences. Even when our preferences do not directly depend on what other people do, the behavior of others can convey a signal that alters our beliefs and our behavior. In the absence of a strong reason to believe otherwise, I might infer from other people’s actions that a tattoo does look good, that drinking banana juice will make me slim, and that this harmless-looking Mexican man is really a rapist at heart.

But how can we explain that people will sometimes do things they know not to be in their immediate self-interest (for example, getting a tattoo they find ugly or lynching a Muslim man at the risk of being arrested) just because their friends do it?

It turns out that just as fads can be rationalized by standard preferences, so can sticking to social norms. The basic idea is that those who violate the norm will be punished by the rest of the community. And so will those who fail to punish violators, and those who fail to punish those who fail to punish those who fail to punish, and so on. One of the great achievements of the field of game theory is the folk theorem, a formal demonstration that this argument can be made in a logically coherent way and can therefore be a candidate for explaining why norms are so powerful.7

Elinor Ostrom, the first (and so far the only) woman to receive a Nobel Prize in economics, spent her career demonstrating instances of this logic. Many of her examples were drawn from small communities—cheese makers in Switzerland, forest users in Nepal, or fishermen on the Maine coast or in Sri Lanka8—who live by a norm about how community members were supposed to behave that everyone stuck to.

In the Alps, for example, Swiss cheese producers had for centuries relied on common ownership of a pasture for cattle grazing. If there had been no communal understanding, this could have led to disaster. The land might have been overgrazed to barrenness since it belonged to no one and everyone had a reason to want to feed their own cows more, potentially at the expense of the others. However, there was a set of clear rules for what cattle owners could and could not do on the common pasture, and those rules were followed because violators were excluded from future grazing rights. Given that, Ostrom argued, collective ownership was actually better for everyone than private property. Dividing the land into small parcels, each owned by a separate person, increases risk, since there is always the possibility of some disease hitting the grass in any given small area.

This kind of logic also explains why, in many developing countries, a part of the land (for example, the forest abutting the village) is always held as common property. As long as the common land is used sparingly, it provides a resource of last resort for those villagers whose own economic plans have hit some headwinds; foraging in the forest or selling grass cut from the common land helps them survive. The intrusion of private property into these settings, generally inspired by economists who don’t understand the logic of the context (and love private ownership), has often been a disaster.9

It also suggests a selfish reason for why people in villages often seem to help each other out; it is probably partly in anticipation of receiving similar help when they need it.10 The punishment sustaining the norm is that those who refuse to help will themselves be excluded from the community’s help in the future.

Systems of mutual help are vulnerable to collapse if some community members have opportunities outside. Then the risk of being excluded is not that terrifying anymore, making it tempting to default on obligations. Anticipating this, community members may be more reluctant to help, further increasing the temptation to default. The whole system of mutual support may unravel totally, leaving everyone worse off. The community is therefore very alert to, and protective against, behavior that seems to threaten the communal norms.

Economists have generally emphasized the positive role communities play.11 But the fact that norms can be self-enforcing does not necessarily make them good. The discipline they impose could be directed toward some reactionary, violent, or destructive cause. A now classic paper showed that both racial discrimination and India’s notorious caste system can be sustained by the same logic, even if no one actually cares about race or caste.12

Suppose no one actually gives a damn about caste, but anyone who crosses caste lines in sex or marriage is charged with miscegenation and treated like an outcast, meaning nobody will marry into their family and no one will befriend or associate with them. And finally suppose that anyone who defies this norm and marries an outcast also becomes an outcast. Then as long as people are sufficiently forward-looking, and that they do want to get married, this will be enough to stop everyone from breaking the rule, however arbitrary everyone feels it is. Of course, this could shift if enough people start defying the norm. But there is no guarantee it will ever happen.

This is very much the story at the heart of Samskara, a wonderful 1970 Indian film directed by Pattabhi Rama Reddy, in which a Brahmin (and hence a member of the so-called highest caste) becomes “polluted” by sleeping with a low-caste prostitute. When he suddenly dies, no other Brahmin is willing to cremate him for fear of being polluted by contact with him. His body is left to rot in public.The norm becomes a perversion of the community’s rules precisely because the community is stuck in enforcing its own standards.

This tension between the community that binds and the community that bullies is of course age old and universal. And it translates into the tension between a state that protects the individual and a state that undermines the community, which is at the heart of a battle ongoing in countries as diverse as Pakistan and the United States. The fight is partly against the bureaucratization and impersonality that come with state interventions, and partly to preserve the right of the community to pursue its own goals. Even if these goals include, as they often do, discriminating against people with different ethnicity or sexual preferences, as well as the enforcement of religious diktats over those of the state (teaching creation science, for example).

In the Indian national movement, Gandhi famously represented the view that the new Indian nation should be based on decentralized self-reliant villages, havens of peace and fellow-feeling. “The future of India lies in its villages,” he wrote. His most remarkable opponent in the movement was Dr. B. R. Ambedkar, the man who would eventually draft the Indian constitution. Born into the very lowest caste, not allowed to enter the classroom in the local school, he was so brilliant that he nevertheless ended up with two PhDs and a law degree. He famously described the Indian village as “a sink of localism, a den of ignorance, narrow-mindedness and communalism.”13 For him, the law, the state as its enforcer, and the constitution from which it derived its force, were the best guarantors of the rights of the underprivileged against the tyranny of the locally powerful against the community.

The history of independent India has been a reasonable success in terms of integrating the castes. For example, the wage gap between the traditionally disadvantaged castes (SC/STs) and others dropped from 35 percent in 1983 to 29 percent in 2004.14 This does not look so spectacular, but is more than the improvement in the wage gap between blacks and whites in the United States over a similar time period. In part this is the result of the affirmative action policies Ambedkar put into place, which gave historically discriminated groups privileged access to educational institutions, government jobs, and the various legislatures. Economic transformation also helped. Urbanization, by making people more anonymous and less dependent on their village networks, has permitted greater mixing of the castes. New jobs lowered the importance of the caste network in finding employment opportunities and increased the incentives for young people from lower castes to get educated. In part the village community was also perhaps less bad than Ambedkar had feared. Villages have proven capable of collective action that transcends caste lines, for example, when they embraced universal primary education and free school meals for all children, regardless of caste.

This is not to say the problem of caste has been solved. At the local level, caste prejudice is alive and well. A study of 565 villages in eleven Indian states found that despite legal bans, some form of untouchability continued to be practiced in almost 80 percent of the villages. In almost half of the villages, Dalits (members of the lower castes) could not sell milk. In about a third of them, they could not sell any goods on the local market, they had to use separate utensils in restaurants, and access to water for irrigating their fields was restricted.15 Furthermore, while traditional forms of discrimination are weakening, upper castes also react with violence to the perceived threat of the economic progress of lower castes. In March 2018, a young Dalit man in the state of Gujarat was killed for owning and riding a horse, something apparently only upper castes are allowed to do.

To complicate matters, a newer pattern of conflict is emerging; caste groups now see each other as closer to equals but also as potential rivals for power and resources.16 In politics, there is increasing caste polarization in voting; an increasing fraction of the votes of the upper castes go to the Bharatiya Janata Party (BJP), the one party not committed to affirmative action.17 Other parties have emerged to cater specifically to different caste groups. This polarization has consequences. In Uttar Pradesh, India’s most populous state, the complexion of politics changed drastically between 1980 and 1996. Areas dominated by lower castes voted more and more for the two parties identified with the low castes, whereas the areas dominated by upper castes continued to vote for the parties traditionally associated with them. During the same period, corruption exploded. An increasing number of politicians had a case opened against them, some even fighting (and winning) reelection campaigns from jail. Abhijit and Rohini Pande found there was a connection: corruption increased the most in areas where either the upper castes or the lower castes were a large majority.18 In those areas, as a result of caste-based voting, the candidate of the dominant caste was all but assured to win, even when he was extremely corrupt and the opponent was not. Nothing like that happened in areas where the population was balanced.

At the same time, the importance given to caste loyalty also allows the community to exercise control over its members, often in clear violation of the law of the land. For example, the caste panchayats (essentially local caste associations) have actively resisted the state’s writ on sex and marriage in the name of tradition. In a grotesque incident in the state of Chhattisgarh, a fourteen-year-old girl who had been raped by a sixty-five-year-old man was advised by the local caste panchayat not to go to the police about it. When she persisted, she was thrashed by some of the elders of the community, both male and female. A strong community can oppress its weakest members (the Dalits yesterday, today the young woman), and the state is largely powerless to stop it, in part because a majority of community members find it in their interest to uphold community control. As long as they conform, the caste collective offers members access to a web of support and comfort in time of need, and while its brutal underside might bother them from time to time, it takes a brave man or woman to take on the entire community.

This 2008 headline from the satirical newspaper the Onion captures just how remarkable Barack Obama’s presidential candidacy was for the United States. The play on words highlighted the contrast between the stereotype of a black man as a freeloader (begging for small change) and Obama as an inspirational leader (asking for cultural change). It is easy to forget there were fewer than forty-five years between the Freedom March and the election of the first African American president. Much has changed in race relations in the years since the civil rights movement, a lot for the better. This made it possible for the country to elect Obama, just as the president and the prime minister of India in 2019 were from the erstwhile backward castes, something equally unthinkable forty-five years ago.

On the other hand, while the African American population today is much better educated than it was in 1965, the income gap between white and black men with similar education has been growing and is now as much as 30 percent, more than that between the scheduled castes and the other castes in India.20 Black Americans have substantially lower rates of upward mobility and higher rates of downward mobility than whites.21 This clearly is related to the much discussed large gap in incarceration rates between black males and everyone else,22 but it is also related to a persistent segregation in neighborhoods and schools.

Despite the fact that white males seem to have no reason to feel economically threatened by African Americans, there is evidence of rising (or at least more open) articulation of anti-black sentiments in recent years. According to the FBI, the number of hate crimes rose by 17 percent in 2017. It was the third consecutive year they increased. They started rising in 2015, after a long period when they had been flat or declining. Three out of five hate crimes targeted a person’s ethnicity.23 Nine candidates who were self-described white supremacists or had close ties to white supremacists ran for office in the congressional elections in 2018.24

The dominant story in the United States since the 2016 elections, however, is not the mistrust of African Americans, but the open rage against immigrants, which goes well beyond purely economic resentment.

Immigrants do not only “take” “our” jobs; they are “criminals and rapists” who threaten the very survival of whites. Interestingly, within the US the fewer immigrants there are living in a state, the less liked they are. Nearly half of residents in states with almost no immigrants—like Wyoming, Alabama, West Virginia, Kentucky, and Arkansas—believe immigrants represent a threat to American culture and values.25

This suggests the worry has more to do with identity than with economic anxiety. The logic seems to be more that in the absence of much contact, it is easy to imagine that the unseen group is fundamentally different.

This phenomenon predated 2016, but Trump’s election made it that much easier to talk about openly. In a clever experiment highlighting this, researchers recruited online respondents in eight deep-red states: Alabama, Arkansas, Idaho, Nebraska, Oklahoma, Mississippi, West Virginia, and Wyoming.26 Just before the 2016 election, they offered respondents a financial incentive to give money to an anti-immigration charity. Specifically, they asked respondents to authorize them to make a $1 donation to the organization on their behalf, and offered to pay them an extra fifty cents if they agreed. For some people the choice was purely private. For some randomly selected others the offer was presented in a way that implied a small chance they would be personally called by a member of the research team to verify their decision—so at least one person would observe their decision and discuss it with them. Before the election, people in this second group were less likely to agree to donate than people who could do it purely privately (34 percent versus 54 percent). But when the same experiment was conducted right after the election, that difference entirely disappeared! The victory of someone who expressed overt anti-immigration views had freed respondents up to openly give money to an anti-immigrant group.

It is perhaps reassuring that previous waves of migrants to the United States experienced similar rejection before they were ultimately accepted. Benjamin Franklin hated the Germans: “Those who come hither are generally of the most ignorant Stupid Sort of their own Nation.… Not being used to Liberty, they know not how to make a modest use of it.” Jefferson thought the Germans were unable to integrate. “As to other foreigners it is thought better to discourage their settling together in large masses,” he wrote, “wherein, as in our German settlements, they preserve for a long time their own languages, habits, and principles of government.”27 America tried to limit Chinese immigration as early as the nineteenth century and it was eventually banned. In 1924, quotas were introduced with the aim of limiting immigration of Eastern and Southern Europeans (Italians and Greeks).28

And yet each wave of immigrants eventually was accepted and assimilated. The first names they chose for their children, the occupations they ended up in, the way they voted, and what they bought and ate converged with those of the local population. In turns, the locals adopted the once-foreign first names and foods: Rocky is an American hero and pizza is one of the five basic food groups.

The same phenomenon happened in France. French people rejected the Italians. Then they rejected the Poles. Then they rejected the Spaniards and the Portuguese. Every wave of migrants eventually became integrated, but in each case at first the French believed it was “different this time.” By 2016, it was the Muslims’ turn to be rejected.

Where do these preferences and attitudes come from? Why do we seem to look for a new enemy even as we become reconciled to the previous one?

There are potentially some simple economic explanations for bigoted behavior toward other groups, very much in the spirit of Becker and Stigler’s standard model. Intimidation sometimes serves an economic purpose. Between 1950 and 2000, Hindu Muslim riots in India were much more likely to occur in a particular city in a particular year if the Muslim community happened to be relatively well off. And they were less likely to occur if the Hindu community happened to be doing well.29 This is consistent with detailed accounts of some of the large riots, where Muslim businesses were specifically targeted in the midst of what may have seemed like random violence. Violence is often a convenient camouflage for theft.

It is also true that sometimes individuals feel the need to express intolerance and prejudice (including sentiments they do not actually share) in order to signal fealty to their group. For example, during the Indonesian economic crisis, membership in Koran reading groups increased. The display of intense religiosity was a sign of loyalty to earn a place in a mutual assistance circle.30 In other contexts, sometimes people keep quiet about racism (and/or sexism), or even echo what they hear because they do not want to lose their jobs or valuable social connections.

And, finally, there is what economists call statistical discrimination. We met an Uber driver in Paris who was very enthusiastic about his job. He said that in the old (pre-Uber) days, if a North African man like him was seen driving a nice car, everyone assumed he was either a drug dealer or had stolen the car. Most people believed, correctly, that most normal North Africans tended to be relatively poor and therefore unlikely to be able to afford a new car, and on the basis of that statistical association their presumption was that the individual North African driver of a nice car was a criminal. Now they assume he is an Uber driver, which is clear progress.

Statistical discrimination explains why the police in the United States justify stopping black drivers more often. And how the Hindu majoritarian government of the state of Uttar Pradesh recently explained why so many of the people “accidentally” killed by the state police (in what are called “encounter deaths”) are Muslim. There are more blacks and Muslims among criminals. In other words, what looks like naked racism does not have to be that; it can be the result of targeting some characteristic (drug dealing, criminality) that happens to be correlated with race or religion. So statistical discrimination, rather than old-fashioned prejudice—what economists call taste-based discrimination—may be the cause. The end result is the same if you are black or Muslim, though.

A recent study on the impact of “ban the box” (BTB) policies on the rate of unemployment of young black men provides a compelling demonstration of statistical discrimination. BTB policies restrict employers from using application forms where there is a box that needs to be checked if you have a criminal conviction. Twenty-three states have adopted these policies in the hope of raising employment among young black men, who are much more likely to have a conviction than others and whose unemployment rate is double the national average.31

To test the effect of these policies, two researchers sent fifteen thousand fictitious online job applications to employers in New Jersey and New York City, just before and right after the states of New York and New Jersey implemented the BTB policy.32 They manipulated the perception of race by using typically white or typically African American first names on the résumés. Whenever a job posting required indicating whether or not the applicant had a prior felony conviction, they also randomized whether he or she had one.

They found, as many others before them, clear discrimination against blacks in general: white “applicants” received about 23 percent more callbacks than black applicants with the same résumé. Unsurprisingly, among employers who asked about criminal convictions before the ban, there was a very large effect of having a felony conviction: applicants without a felony conviction were 62 percent more likely to be called back than those with a conviction but an otherwise identical résumé, an effect similar for whites and blacks.

The most surprising finding, however, was that the BTB policy substantially increased racial disparities in callbacks. White applicants to BTB-affected employers received 7 percent more callbacks than similar black applicants before BTB. After BTB, this gap grew to 43 percent. The reason was that without the actual information about convictions, the employers assumed all black applicants were more likely to have a conviction. In other words, the BTB policy led employers to rely on race to predict criminality, which is of course statistical discrimination.

That people are using statistical logic does not, of course, mean they are always drawing the right inferences from it. In one study, researchers asked Ashkenazi Jews (European or American Jews and their descendants) in Israel to play a trust game with Eastern Jews (Asian and African immigrants and their descendants). The trust game is one of the mainstays of experimental economics. It is played by two people, one of whom, the sender, is given a certain amount of money and asked to share some part of it with the other person, the receiver. The amount could be zero and is entirely left to the sender’s discretion. However, they are both told that if the sender shares any of it, that shared amount will be tripled and given to the receiver, who then has full control over the money. The receiver has the option of sharing some of his gains with his benefactor but can opt not to do so. The point of this game is to infer what the sender thinks about the receiver; the less selfish the sender believes the receiver to be, the more he should share.

The trust game has been played thousands of times in laboratory settings. Typically, the sender shares half or more of the original amount and gets back more than was sent. Senders are trusting and receivers are trustworthy. This is also what the researchers found when the two players were both Ashkenazi. But things fell apart when the receiver was an Eastern Jew. In that case, the sender shared about half of what would have been sent to an Ashkenazi. As a result, both senders and receivers got less.

It could be that this happens because the Eastern receivers are not trusted to return the gift. Or it could be because they are disliked, and Ashkenazi senders are willing to hurt themselves just to hurt Eastern receivers as well. But when players were asked to just voluntarily give some of their money to a partner with no expectation any of it would come back, they gave about as much to Eastern partners as they did to the Ashkenazi; the source of the different behaviors in the trust game seems to be suspicion rather than animosity.

Interestingly, the suspicion extends to Eastern senders in the trust game. They were no more trusting of their co-ethnics than the others. There seems to be a stereotype of Eastern Jews that everyone has bought into. But the twist is that the stereotype is entirely unfair. There is absolutely no evidence the Eastern players in this game act in a less trustworthy way; their pattern of returning the money is exactly the same as that of the Ashkenazis. The participants in the experiment thought they behaved rationally, but they were being led astray by imaginary suspicions.

The ubiquity of self-discrimination, or discrimination against one’s own group, was powerfully brought to light by a well-known experiment by the American psychologist Claude Steele, which demonstrated the power of what he called a “stereotype threat.” In his original experiment, he found that black students performed comparably with white students when told a test they were taking was “a laboratory problem-solving task.”33 However, black students scored much lower than whites when test takers were told the test was meant to measure their intellectual ability.

Minorities aren’t the only ones vulnerable to stereotype threat. Female college students performed better on a hard mathematics test when it included at the beginning the statement “You may have heard that women typically do less well at math tests than men, but this is not true for this particular test.”34 Conversely, white male math and engineering majors who received high scores on the math portion of the SAT (a group of people quite confident about their mathematical abilities) did worse on a math test when told the experiment was intended to investigate “why Asians appear to outperform other students on tests of math ability.”35 These types of experiments have been repeated many times in different contexts to test different types of self-discriminatory prejudice.

Self-discrimination is often self-reinforcing; people perform differently when they are reminded of their group identity, which makes them doubt themselves even more. The same goes for discrimination against other groups. In a now infamous (once famous) experiment in psychology from the 1960s, teachers were tricked into believing one group of their students (a fifth of the class) was gifted and therefore expected to develop much faster than the rest in terms of IQ. In reality, this group was randomly selected and roughly identical to the rest.36 The students for whom teachers had higher expectations gained twelve IQ points over the course of the year, while the rest gained only eight. The original experiment was criticized for a variety of reasons, including the morality of such interventions, but numerous other experiments have shown the power of self-fulfilling prophecies.

In France, a study of young cashiers in a French grocery store chain, a sizable share of whom were minorities of North African and Sub-Saharan African origin, showed that biased supervisors invested less in the workers they managed.37 The cashiers worked with different supervisors on different days and had virtually no control over their schedule. The study showed that assignment to a supervisor more or less biased against a minority affected the performance of minority and nonminority workers differently. On days when they were scheduled to work with biased supervisors, minority cashiers were more likely to be absent. When they did come to work, they spent less time working; they also scanned items more slowly and took longer to serve the next customer. Such effects were completely absent for nonminority workers. The reason for the lower performance of minority workers when assigned to a biased manager seemed to be not so much overt hostility (minority workers did not report disliking working with biased supervisors more, or that biased supervisors disliked them) as less-effective management. Minority workers reported, for example, that biased supervisors were less likely to come over to their cashier stations and encourage them to perform better.

Discrimination against women in leadership positions often carries the same flavor of self-fulfilling prophecy. In villages in Malawi, male or female farmers were randomly selected to learn a new technology and teach it to other farmers.38 Women retained more information from the training, and those who were trained by them and listened to them did in fact learn more. But most farmers did not listen. They assumed women were less able, and therefore paid less attention to them. Along the same lines, when women in Bangladesh were trained to become line managers, they were just as good as men based on an objective assessment of their leadership and technical qualities, but they were perceived as less good by their line workers. And, presumably as a result, the performance of their lines also suffered, perversely confirming the prejudice that they were worse managers.39 What started as an unjustified preference against women resulted in women actually doing worse through no fault of their own, and this reinforced their inferior status.

What is strange about these self-fulfilling prophecies is just how predictable they are. It is always a traditionally disadvantaged person who ends up as the victim of a biased, but self-fulfilling prediction; you never hear about white males being systematically underestimated in anything except sports. The bias stems from a stereotype rooted in the social context.

A study of African American and white Princeton undergraduates shows how deep this runs.40 The students, who had no prior experience of golf, were asked to perform a series of golf exercises of increasing difficulty. In a first experiment, half of them were asked to indicate their race in a questionnaire before they played (the standard way to “prime” race; that is, to bring group identity to the top of the mind), and half were not. All students were then presented the golf exercises as a test of “general sports performance.” When race was not primed, white and black students performed very similarly. But once race was made salient, the fact that golf is a “white” sport (this was before Tiger Woods) made the African Americans worsen their performance and the white students improve theirs, creating a large gap between the two.

In a second experiment, researchers did not prime race, but instead the students were randomly assigned to one of two treatments.41 In both groups, the instructions said the exercises would become increasingly challenging. In one group, the instructions said the test was designed to measure personal factors correlated with natural athletic ability. Natural athletic ability was defined as “one’s natural ability to perform complex tasks that require hand-eye coordination, such as shooting, throwing, or hitting a ball or other moving objects.” In the other, the same test was presented as measuring “sports intelligence,” or “personal factors correlated with the ability to think strategically during an athletic performance.” In the “natural ability” condition, the African Americans did much better than the whites. In the “sports intelligence” condition, the whites did much better than the African Americans. Everyone, including the blacks themselves, had bought into the stereotype of the African American natural athlete and the white natural strategic player. And this was at Princeton…

It is hard to square this evidence with the Becker-Stigler construct of coherent and stable preferences. It seems clear that the way the groups thought about themselves (and others) was a product of these largely ephemeral social constructs of “sports intelligence” and “natural ability” and their supposed connection to race.

Becker and Stigler want us to stay away from the social context behind preferences, but the social context keeps creeping back in. We have preferences not only about what to eat or where to live, but also about who we should spend time with.

We avoid people we are suspicious of, move to neighborhoods where there are more of us. In turn, this segregation affects life chances and breeds inequality. When a neighborhood is mostly poor and black it also gets fewer resources, and all of this has lasting influences on the lives of the children who grow up there. When black people moved to white towns in the north between 1915 and 1970, during what is known as the Great Migration, whites moved away, often leaving behind worse schools, declining infrastructure, and fewer job opportunities.42

These neighborhoods became poorer and more derelict, more crime prone, and less and less conducive to economic success. The chance for a black kid to move from the bottom quintile to the top quintile of the distribution of income is much lower in neighborhoods abandoned by the whites during the Great Migration than in others.43 There are obviously many factors at play, but one of them is that people consciously and unconsciously end up playing by the rules of their neighborhoods. Violence becomes the norm in a neighborhood where it is expected, just as taking five courses when four are required is the norm for MIT undergraduate students.

In a clever experiment illustrating the power of these norms, a group of mostly Hispanic high school students in Los Angeles were offered the option to sign up for a free SAT prep.44 Some students, chosen at random, were told their choice would remain a secret, while some of them were led to believe their choice might become public. In non-honors classes, the latter group of students were less likely to sign up for the course (61 percent versus 72 percent), presumably because they did not want their friends to find out they had academic aspirations.

It is true that the folk theorem could explain what is going on here. Perhaps it is true that students would be dropped by all their friends if they found out the students were nerds, and anyone who talked to them would also be excommunicated. But it is not accidental that this norm has taken hold with Hispanic students, where there is a history of resenting the norms of white culture, sometimes with very good reason; these Hispanic boys and girls, it seems, were worried about “acting white.” That worry has deep roots in their history. We never hear of Asian kids in the United States who have adopted a habit of avoiding their friends who work too hard. In the Becker-Stigler world, since the norms are norms only because people have submitted to them, there is no reason why Hispanic students would not sometimes turn out to be hard working and the Asians the slackers. It is history and the social context that seem to be guiding us toward one norm rather than the other.

To investigate the way the social context influences us, researchers at the University of Zurich recruited a group of bankers as experimental subjects and asked them to flip a coin ten times and report online the outcomes they got.46 They were told that if they had more than a threshold number of heads (or tails) they would get twenty Swiss francs (about $20) for each extra head (or tail) they reported. Nobody checked whether or not they reported accurately, which created a very strong incentive to cheat.

The key comparison was between those who, before the experiment began, were asked about their favorite leisure activity, highlighting their role as a “regular” person, and those questions about their role as a banker, effectively highlighting their banker identity. Those made to think of themselves as bankers reported many more heads, so many more that it could not have been pure chance. The estimated cheating rate went from 3 percent for those thinking of themselves as regular people to 16 percent for those thinking of themselves as bankers.

This was not because the bankers were better at figuring out how to do well in the game; everybody in the game was a banker, and what got highlighted about them (banker or not) was chosen at random. But being reminded of their profession seems to have brought out a different moral self, one more willing to cheat.

In other words, people seemed to act as if they had multiple personalities, each with different preferences. The context picks the personality that gets to decide in a particular situation. In the Swiss experiment, the context was whether or not the person saw himself as a banker, but in life it is often the people we are with, the schools we went to, what we do for work or for play, the clubs we belong to, and the clubs we would like to belong to that form us and shape our preferences. We economists, in our fealty to standard preferences, have tried very hard to keep all of that out, but it is increasingly obvious this is a hopeless quest.

Once we begin to acknowledge that our beliefs and even what we take to be our deep preferences are determined by context, many things fall into place. One important insight comes from the Nobel Prize–winner Jean Tirole’s work with Roland Bénabou on motivated beliefs.47 They argue that a big step toward understanding beliefs is not taking them too literally. Our beliefs about ourselves are shaped in part by our emotional needs; we feel terrible when we disappoint ourselves. The emotional value we put on beliefs about ourselves also leads us to distort our beliefs about others; for example, since we want to shield ourselves from our own prejudices, we couch them in the language of objective truth (“I have nothing against North African cashiers, but they would not respond to my encouragement anyway, so I don’t bother”).

We don’t like changing our minds because we don’t like to admit we were wrong to begin with. This is why Abhijit insists it is always the software’s fault. We avoid information that would force us to confront our moral ambiguities; we skip over news about the treatment of migrant children in detention centers to avoid thinking about the fact we have supported a government that treats children in this way.

It is easy to see how we may get trapped by these strategies. We don’t like to think of ourselves as racists; hence, if we have negative thoughts about others, it is tempting to rationalize our behavior by blaming them. The more we can persuade ourselves migrants are to blame for bringing their children with them, the less we worry about the children in their little cages. Instead, we look for evidence that we are right; we overweight every piece of news, however thin, that supports our original position, ignoring the rest.

Over time, the instinctive defensive reaction we started from is replaced by a carefully constructed set of seemingly robust arguments. At that point, we start feeling that any disagreement with our views, given how “solid” the arguments are, has to be either an insinuation of moral failure on our part or a questioning of our intelligence. That’s when it can get violent.

Recognizing these patterns has a number of important implications. First, obviously, accusing people of racism or calling them the “deplorables,” as Hillary Clinton famously did, is a terrible idea. It assaults people’s moral sense of themselves and puts their backs up. They immediately stop listening. Conversely, one can see why calling egregious racists “fine people,” and emphasizing there are bad people “on both sides,” as President Trump did, is clearly an effective strategy (however morally reprehensible) to gain popularity, since it makes those who make these remarks feel better about themselves.

It also explains why facts or fact-checking don’t seem to make much of a dent on people’s views, at least in the short run, as we observed in chapter 2, in the context of migration. It remains possible that in the longer run, when the initial “How dare you challenge my beliefs?” reaction fades, people will adjust their views. We should not stop telling the truth, but it is more useful to express it in a nonjudgmental way.

Since most of us like to think we are decent people, forcing someone to affirm their own values before exercising a judgment involving others might reduce prejudice. Psychologists these days encourage parents to tell their children not that they should be nice, but that they are nice, and all they have to do is to behave in conformity with their natural kindness. That applies to all of us.

This strategy is more likely to work when self-esteem is not already battered. Part of the problem low-income whites face in areas where anti-immigrant and anti-black sentiments are the strongest is that in some observable ways their lives come very close to their own caricature of how those despised “others” live. In 1997, William Julius Wilson wrote in the context of what was happening in the black community that “the consequences of high neighborhood joblessness are more devastating than those of high neighborhood poverty… Many of today’s problems in the inner-city ghettos—crime, family dissolution, welfare, low levels of social organization, and so on—are fundamentally a consequence of the disappearance of work.”48

Twenty years later, J. D. Vance wrote in Hillbilly Elegy: “Wilson’s book spoke to me. I wanted to write him a letter and tell him that he had described my home perfectly. That it resonated so personally is odd, however, because he wasn’t writing about the hillbilly transplants from Appalachia—he was writing about black people in the inner cities.”49

That Wilson’s description of the social problems in black neighborhoods applies so well to white communities in the Rust Belt now literally adds insult to injury. Since the perception of their own worth is tied to a sense of superiority with respect to blacks and migrants, the convergence in social circumstances exacerbates the poor white American’s sense of crisis.

There are two ways to proceed to restore the sense of self. One is denial (for instance: “We can afford to be resolutely anti-abortion since none of the girls in our community ever get pregnant”). The other is increasing the distance between us and them by turning the other into a caricature. For a white person who has to be on disability because it is the only way to get welfare, it is not sufficient anymore to say a black or Latino single mother must be a welfare queen; that was a Reagan-era insult. Now that white people have to be on welfare as well, the insult has to be ratcheted up; she must be a gang member.

This underscores why we need social policies to reach beyond economic survival and try to restore the dignity of those whose occupations are threatened by technological progress, trade, and other disruptions. The policies must effectively counterbalance the loss of self-confidence; old-fashioned government handouts are not going to work by themselves. What is needed is a wholesale rethinking of the social policy apparatus, the subject of chapter 9.

We know that people will go to great lengths to avoid evidence that would force them to revise their opinions on what they consider to be their core value system (including their opinion about other races or immigrants), because it is so related to their views of themselves. Unfortunately, it does not follow that people are particularly thoughtful about forming those initial opinions.

In one of the most famous experiments in the field of behavioral economics, Daniel Kahneman and Richard Thaler chose college students randomly to receive a mug or a pen. Immediately following the gifts, they offered to buy them back from the newly endowed mug and pen owners. At the same time, they also offered those who did not get a mug or a pen the opportunity to buy what they did not get. Strikingly, the price at which the newly endowed sellers were willing to part with their mugs or pens was often two to three times greater than what those who did not have the pen or mug wanted to pay for them.51 Since who ended up with a mug or a pen was entirely random, there was absolutely no reason why the arbitrary act of being chosen to get one of them would create such a divergence in valuations. The difference in the bids must have been because those who ended up with a mug started liking their mug more, while those who got a pen did the same with the pen, suggesting there is relatively little intrinsic or deep about how people value things like mugs and pens.

An even more dramatic form of arbitrariness was revealed by another experiment. Students were asked to bid on trackballs, wine bottles, and books. Before bidding, they were asked to write down the last two digits of their social security number with a dollar sign in front of it and imagine it was a possible price for the product they had. They obviously knew their social security number had nothing to do with the price of a wine bottle, but nevertheless they were influenced by the “price” they had written down. Students with social security numbers ending in the number eighty or larger bid between 200 percent and 350 percent more for the same good than those whose social security numbers ended in a number less than twenty. In most other ways, they still behaved according to the standard model: for example, they were less willing to buy as the price went up and were most likely to buy cheaper items. But they seemed to have no idea how much these products were worth to them in absolute terms.52

But of course mugs and pens are not immigrants and Muslims. Are we really seriously implying this arbitrariness applies to preferences on these much more serious issues as well? We are indeed.

Something similar shows up in social preferences, what economists call preferences that concern other human beings. In 1954, Muzafer Sherif and Carolyn Wood Sherif carried out an experiment in which twenty-two eleven- and twelve-year-old boys were invited to a summer camp in Robbers Cave, Oklahoma.53 The boys were randomly divided into two groups. Each group spent some time living in a different location of Robbers Cave, so that the groups were initially unaware of each other’s existence. Then at some point the two groups were introduced to each other and made to compete, for example, at tug-of-war. This created animosity between the groups, leading to name-calling and attempts to vandalize the other group’s possessions. In the final days, the researchers artificially induced a water shortage, making it useful for the two groups to work together. After some initial hesitation, they did so and mostly forgot their animus.

Some version of this experiment has been repeated many times, and the basic insight has proven very robust. Interestingly, the fact that arbitrary labels heavily influence our loyalties is true even without the bonding experience the initial isolation provided. Just giving a different name to a randomly chosen group of participants got in-group members to favor their own over the others. This was as true of adults as of eleven-year-olds.

Both parts of the Robbers Cave experiment are important: the fact that it is easy to divide as well as the fact that it is possible to reunite. That it is easy to divide is a strong reason to be extremely frightened by the xenophobes and the cynical manipulators of xenophobia who rule so many countries today. The damage they do is not permanent, but unless it is carefully undone it can leave a terrible scar on a nation. In Rwanda, the Belgian colonialists created the myth of the superior Tutsis and the inferior Hutus out of a more or less homogenous population as a way of securing allies in the process of governing. In the immediate post-colonial period, the Tutsis embraced their purported superiority, much to the resentment of the Hutus, and this became a crucial contributory factor to the horrific genocide of 1994.54

At the same time, the fact that preferences are not necessarily internally consistent makes attaching ad hominem labels—such as “racist,” other “ists,” or for that matter “deplorables”—to other human beings suspect, because many people are both racist and not, and their expressions of prejudice are often expressions of pain or frustration. Those who voted for Obama and then Trump may be confused about what each candidate stood for, but to dismiss them as racists after they voted for Trump is both unfair and unhelpful.

Since our preferences are strongly influenced by whom we associate with, social divides are particularly costly because there is very little mixing across these divides; people tend to associate with others like themselves. In US schools, black teenagers mostly associate with blacks, and whites with whites.55 This is what sociologists call homophily. For obvious reasons, this is especially true of those from the largest social group in the school. Those who are a part of a small minority have no choice but to have relatively more friends outside their group.56

This does not have to be evidence of intense prejudice. That students in the biggest group do not reach out to outsiders can easily be explained by the fact it is easy for them to meet others like them, and therefore as long as they have a mild preference for their own group, they have no reason to reach out beyond it.

The source of the mild preference does not have to be a negative view of anyone else; it could just be that it is easier to be with people who speak the same language, who share the same gestures and the same sense of humor, who watch the same TV shows and enjoy the same music, or who make the same unstated assumptions about what is appropriate or not. Abhijit, who is from India, is always struck by how easy it is for him to talk to people from Pakistan, notwithstanding the past seventy years of animus between India and that country. The sense of what is funny or what is private (hint: South Asians are nosy), what creates intimacy and what distracts from it, is something, he says, instinctive in all of us South Asians, something partition did not manage to destroy.

The downside to this very natural behavioral pattern becomes evident when we meet people from other groups. We hold back; we walk on eggshells, rationing our human warmth because we worry we might be misunderstood. Or we blunder forward, giving offense when none was intended. Either way, something important gets lost, with the result that we are less likely to communicate smoothly with people from other groups.

This is partly why people mostly marry people like themselves. A little over fifty years after the landmark decision Loving v. Virginia, which in 1967 struck down prohibition of interracial marriage in the United States, only about one in six American newly married couples was biracial.57 In India, 74 percent of families say they believe marriages should be made within castes. Our research suggests this is in part because the men in each caste are looking for women who are the equivalents of their sisters (in other words, the familiar) and likewise for women, and the best place to find such a match is naturally within the group they belong to.58

Such behavior leads to accidental and probably largely unconscious segregation. We may not realize that if all of us choose to hang out with friends like us, we end up forming entirely separate islands of similar people. This feeds into the intensification of apparently bizarre preferences and/or extreme political views. One obvious downside of sticking to our own is we don’t get exposed to other points of view. As a result, differing opinions can persist, even on points of fact such as whether vaccines cause autism or where Barack Obama was born, and even more obviously on matters of taste. We earlier observed that people might rationally choose to suppress their own opinions and join the herd, but of course not being exposed to any opinions outside the herd only makes things worse. We end up with multiple closed groups with contrasting opinions and very little capacity for communicating respectfully with each other. Cass Sunstein, a law professor at Harvard and a member of the Obama administration, describes these as “echo chambers,” where like-minded people whip themselves into a frenzy by listening only to each other.59

One result of this is extreme polarization on what should be more or less objective facts; for example, 41 percent of Americans believe human activity causes global warming, but the same number either say warming is due to a natural cycle (21 percent) or say there is no warming at all (20 percent). According to the Pew Research Center,60 public opinion about global warming is deeply segmented along political lines: “Democrats are far more likely than Republicans to say there is solid evidence that temperatures are rising (by a margin of 81% to 58%), and that human activity is the root cause (by 54% to 24%).” That does not mean Democrats are necessarily more pro-science. The scientific consensus, for example, is that GMO foods are not harmful to health, but a strong majority of Democrats think they are and are in favor of labeling them.61

Another result of constantly talking to the same people is that the members of a group tend to have shared opinions on most issues. Eclectic political positions become increasingly untenable in the face of a resolute herd, even one that is resolutely wrong. In fact, Democrats and Republicans do not even speak the same language anymore.62 Matthew Gentkzow and Jesse Shapiro, two economists who are leading scholars of the media, write about members of the US House of Representatives: “Democrats talk about ‘estate taxes,’ ‘undocumented workers’ and ‘tax breaks for the wealthy,’ while Republicans refer to ‘death taxes,’ ‘illegal aliens,’ and ‘tax reform.’ The 2010 Affordable Care Act was ‘comprehensive health reform’ to Democrats and a ‘Washington takeover of health care’ to Republicans.” It is now possible to predict the political affiliation of a congressman simply by listening to the words they use. Unsurprisingly, partisanship (defined as the ease with which an observer can infer a congressperson’s party from a single sentence) has exploded in recent decades. Between 1873 to the early 1990s, it did not change, increasing from 54 percent to just 55 percent during this period. But it increased sharply after 1990; by the 110th session of Congress (2007–2009) it was 83 percent.

This convergence of opinions and vocabulary is precisely why access to Facebook data was so useful to Cambridge Analytica and to political campaigns in the UK and the US. Since most Massachusetts Democrats, for example, have more or less the same views across a wide range of questions and use the same words, it takes just some snippets of our opinions to predict our politics, how we should be targeted, and what types of stories we are likely to like or dislike. And, of course, once real people embrace this cardboard cutout predictability, it becomes that much easier to invent characters, create fake profiles, and inject them into an online conversation.63

This insularity also creates an opportunity for skilled political entrepreneurs to present themselves very differently to very different people. During the run-up to the 2014 election for prime minister that he won in a landslide, Narendra Modi in India managed to be at many rallies at the same time by using full-scale, three-dimensional holograms that many voters took to be real. He also managed to be at more than one place in ideological terms. To the generation of globally connected ambitious young urban Indians, he was the embodiment of political modernization (emphasizing innovation, venture capital, and a slick pro-business attitude, and so on); the new entrants into the expanding middle class saw him as the one most likely to uphold their vision of nationalism rooted in Hindu tradition; for the economically threatened upper castes, he was the rampart against the (largely imagined) growing influence of Muslims and lower castes. If members of these groups had met together and each had been asked to describe “their” Modi, their answers would probably have been largely unrecognizable to the others. But the networks in which these three groups operated were sufficiently separate that there was no need for internal consistency.

The sharp segmentation of the electorate goes much deeper than just policy disagreements. Americans of different political hues have started to positively hate each other. In 1960, roughly 5 percent of Republicans and Democrats reported they would “[feel] ‘displeased’ if their son or daughter married outside their political party.” By 2010, nearly 50 percent of Republicans and over 30 percent of Democrats “felt somewhat or very unhappy at the prospect of inter-party marriage.” In 1960, 33 percent of Democrats and Republicans thought an average member from their own party was intelligent, compared to 27 percent who had the same view about someone from the other side. In 2008, those numbers were 62 percent and 14 percent!64

What explains this polarization? One of the most important changes since the early 1990s, when partisanship started its sharp increase, is the expansion of the internet and the explosion of social media. As of January 2019, Facebook had 2.27 billion monthly active users globally, while Twitter had 326 million.65 In September 2014, more than 58 percent of the US adult population and 71 percent of the US online population used Facebook.66 (That does not include us, so everything we have to say about these networks is second hand.)

Originally, virtual social networks were billed as the new public place, the new way to connect, and hence something that should have reduced homophily. In principle, they provided an opportunity to connect with distant people with whom we shared some specific interest, say in Bollywood movies, Bach cantatas, or raising babies. These people might not have been like us in other ways, offering us a more eclectic choice of friends than what would result from mere physical proximity. They would have had almost nothing to do with each other, so to the extent we would get to exchange views about things other than the precise topic that brought us together, we would all be exposed to a variety of opinions. Indeed, on Facebook, 99.91 percent of the two billion people on it belong to the “giant component,” meaning that almost everyone is everyone else’s friend of a friend of a friend.67 There are only about 4.7 “degrees of separation” (the number of “nodes” you have to cross) between any two people in the giant component. This implies that in principle we could easily be exposed to pretty much everyone’s views as they travel through the social network.

Nevertheless, virtual social networks have mostly failed to integrate their users on divisive issues. A study of 2.2 million politically engaged users on Twitter (defined as those who followed at least one account associated with a candidate for the US House during the 2012 election period) in the United States finds that while there are roughly ninety million network links among these users, 84 percent of the followers of conservative users are other conservatives, and 69 percent of the followers of liberal users are liberal.68

Facebook and Twitter function as echo chambers. Democrats pass on information produced by Democratic candidates, and Republicans do the same for Republicans. Eighty-six percent of first retweets of tweets by Democratic candidates come from liberal voters. The corresponding number for Republicans is an amazing 98 percent. Taking into account retweets, liberals get 92 percent of their messages from liberal sources, and conservatives get 93 percent of their messages from conservative sources. Strikingly, this is not just true of political tweets; for these politically engaged people, the exposure is just as skewed for nonpolitical tweets. Apparently, even to chat about fly fishing on Twitter, people prefer to connect with a fellow liberal or conservative. The virtual community that social networks have created is at best a fragmented public space.

But is there anything specific about social media that causes this polarization? The political strategies to divide the population and plant fake news were invented long before Facebook. Newspapers have always been highly partisan, and political mud-slinging was the bread and butter of the print media in colonial America, and continued into the early days of the American Republic (in the musical Hamilton, it is the threat of scurrilous press coverage that forces Hamilton to own up to his affair). The “Republican noise machine” was perfected on cable TV and talk radio in the 1990s, as David Brock powerfully documents in his book bearing that title.69

An even more powerful demonstration of just how destructive old media can be comes from the Rwandan genocide. Before and during the genocide, Radio Télévision Libre des Mille Collines (RTLM) called for the extermination of the Tutsis, whom they called “cockroaches,” justified it as self-defense, and talked about the supposed atrocities committed by the Rwandan Patriotic Front (or RPF, the Tutsi militia). Villages that RTLM reached experienced significantly more killings than villages it did not due to the mountains blocking radio wave transmissions. Altogether, RTLM propaganda is estimated to be responsible for 10 percent of the violence, or about fifty thousand Tutsi deaths.70

Gentzkow and Shapiro computed an “isolation index” for the year 2009 (which in some ways feels like a lifetime ago, but the internet was already quite vibrant) both for online and offline news. This was defined as the difference in the share of news items with a conservative slant a conservative was exposed to and the share of news items with a conservative slant a liberal was exposed to. What they found seemed to suggest polarization was happening offline just as much as it was online. The average conservative’s exposure to conservative views online was 60.6 percent of their total news consumption, the equivalent of a person who gets all their news from usatoday.com. The average liberal’s exposure to conservative positions was 53.1 percent, at the same level as cnn.com. The isolation index for the internet (the difference between the two) was therefore just 7.5 percentage points, a little bit higher than the isolation index for broadcast news and cable television news, but lower than that for national newspapers. And it was much lower than the segregation of in-person contact. It was already true in 2009 that conservatives had mostly conservative friends, and the opposite for liberals. The isolation index is low because, in their data, both conservative and liberal users visited mostly “centrist” sites, and those most likely to visit extremist sites (like Breitbart) also visited many others, including those with opposite perspectives.71

While it is true that polarization has increased among online users, it has also increased in other spheres of life. Indeed, while polarization has increased in all demographic groups since 1996, it increased the most among those sixty-five or older, who are the least likely to be on the internet, and increased the least among the youngest people (those aged eighteen to thirty-nine).72 Polarization has also increased in traditional news media. A textual analysis of the content of cable news showed that since 2004 the language used by Fox News has become increasingly slanted to the right, while MSNBC has moved to the left.73 The audiences have also diverged. Until 2008, Fox News had a stable share of about 60 percent Republicans among its viewers. This share increased to 70 percent between 2008 and 2012. Over the years, Fox News became increasingly conservative, which attracted more conservative voters, who in turn pushed it to be even more conservative. This has begun to affect voting patterns. We know this because in some counties in the United States Fox News shows up at a less accessible part of the dial, for purely accidental reasons, and therefore people are less likely to tune in to it.74 In those counties people are also less prone to vote for conservatives.

So what is it that changed? In Congress, according to Gentzkow and Shapiro, the turning point seems to have been 1994, the year of Newt Gingrich’s takeover of the Republican Party and his “contract with America.”75 This was also the first year political consultants played a key role in designing and testing messages, which is something that as social scientists interested in the design and testing of innovations, including in messaging, we find rather disturbing.

Even if political polarization did not wait for the internet, it is hard to be entirely sanguine about the effects of the virtual social networks and the internet on our policy preferences, and the ways they are expressed. For one, we don’t really know the counterfactual; what would the world be like without these innovations? Comparing those with and without access to the internet, like the young and the old, does not answer this question, for many obvious and less obvious reasons. In particular, the internet is often the place where rumors get manufactured and circulated before they make their way to Fox News, where older people get to hear them. Perhaps younger people are less moved by these rumors because they know the internet is full of errors and exaggerations and can correct for them, whereas older people, used to trusting the booming authority of television anchors, are more gullible.

There are other concerns as well. The first is that the circulation of news on social media is killing the production of reliable news and analysis. Producing fake news is of course very cheap and very rewarding economically since, unconstrained by reality, it is easy to serve to your readership exactly what they want to read. But if you don’t want to make things up, you can also just copy it from elsewhere. A study found that 55 percent of the content diffused by news sites and media in France is almost entirely cut and pasted, but the source is only mentioned in less than 5 percent of the cases.76 If a piece of news produced by a team of journalists is immediately cut and pasted onto many other sites, how does the original source get rewarded for its production? It is no surprise that the number of journalists in the United States has plummeted in the last few years, going from nearly 57,000 in 2007 to almost 33,000 in 2015.77 There are both fewer journalists in total and fewer journalists per newspaper. The economic model that sustained journalism as a location for “public space” (and correct information) is collapsing. Without access to proper facts, it is easier to indulge in nonsense.

The second concern is that the internet allows for endless repetition. The problem with echo chambers is not just that we are only exposed to ideas we like; we are also exposed to them again and again and again, endlessly, throughout the day. The fake users used to “boost” stories on Facebook plus the real persons paid to “like” content accentuate the natural tendency for some messages to be repeated and acquire a life of their own. The endless repetition whips people into a frenzy (much like the way political demonstrations use repeated chants), making it harder for them to stop and check the stories.

And even if the truth eventually gets out, the many repetitions of a falsehood can raise the salience of a divisive issue and harden extremist views. We remember only the endless talk about Mexicans (who we never trusted in any case) and not the fact that first-generation immigrants, legal or otherwise, are actually less likely to be criminals than native-born Americans.78 This of course creates a very strong reason to flood the markets with alternative facts. A hundred and fifteen pro-Trump fake news stories that circulated before the 2016 presidential election were viewed thirty million times (pro-Clinton fake news stories also existed, but they were viewed only eight million times).79

The third is that the crabbed language of internet communication (which Twitter takes to an extreme) encourages directness and abbreviation, contributing to the erosion of the norms of civic discourse. As a result, Twitter has turned into a lab for trying out the latest nasty pitch. Political entrepreneurs are happy to plant their wildest claims on Twitter and watch them play out, with an eye to whether they have gone too far. If it seems to be working, at least among the targeted group (as measured by retweets or likes, for example), they add it to the pack of potential strategies for the future.

Fourth, there is automatic customization. In 2001, when Sunstein was writing about echo chambers, he was worried about the opportunity users have to choose the news they consume. Increasingly, there is no need to choose. Sophisticated algorithms use machine-learning prediction techniques to try to figure out what we might like based on who we are, what we have searched before, etc. The objective, quite explicitly, is to get people what they like so they spend more time on it.

Facebook came under pressure for the algorithm it used to push stories to users, and in 2018 it promised to reprioritize its feeds, putting posts from friends and family ahead of media content. But you do not need to be on Facebook for this to happen. On Esther’s Google home page on July 2, 2018, there was one article from the Atlantic, “The Trade Deficit Is China’s Problem”; Paul Krugman’s latest op-ed in the New York Times; one article from the New York Times on millennial socialists; one article on the soccer World Cup; one article from the Boston Globe on Lawrence Bacow, the new president of Harvard; one article on Simone Veil’s burial; a Huffington Post article on Senator Susan Collins’s view on the choice of the latest Supreme Court justice; and the unavoidable article about the Pixel Watch. There were only two stories she was not obviously interested in: one about a criminal escaping a French prison by helicopter (which turned out to be lots of fun) and a piece on Fox News about Busy Philipps fighting with Delta Air Lines for rebooking her and her child on different flights. That last piece was her entire exposure to right-wing media for the day. Such customization is ubiquitous. Even the National Public Radio app (“NPR One” to the cognoscenti) calls itself “Pandora for Public Radio,” referring to the app that gives you the music you like based on what you have listened to in the past. Within the echo chamber for liberal ideas that is NPR, an algorithm will filter for the user exactly what the user likely will want to hear.

This matters because when users actively choose what they are reading, they are at least conscious of what they are doing. They may prefer to read articles from familiar sources, but be sophisticated enough to acknowledge their own biases reflected in those sources. An unusual experiment in South Korea demonstrated that this kind of sophistication is very real. From February to November 2016, two young Koreans created an app offering users access to curated articles from the press on topical issues and regularly asked them their opinions on the articles and on the issues themselves. At first, all users received a randomly selected article about each issue. After a number of rounds, some randomly selected users got to choose the news sources from which they received their articles, while others continued to receive randomly selected articles. The experiment yielded three important results. First, users did respond to what they read: they updated their views in the direction of what was being presented to them. Second, as expected, those given the option chose articles generally aligned with their partisan preferences. Third, however, at the end of the experiment, those who got to choose their articles had updated their preferences more than those who did not, and they had generally updated toward the center! This is the opposite of the echo chamber effect. On balance, the option to choose slanted material made users less partisan. The reason is they understood exactly how biased the source they chose was, and partly undid the bias, while being receptive to the information; whereas with randomly assigned stories, users did not recognize the bias and therefore remained skeptical about the content, not changing their opinion much.80

It would be very interesting to replicate this experiment in the United States. The effect may also depend on how politically engaged the reader is. It is not entirely clear that many internet readers in the US make a conscious effort to correct the bias in what they read. But this study suggests a key problem of seamless customization: its very seamlessness. Correcting slant requires an understanding of what the source’s slant is. When we always read news from the same source, we are familiar with it. But when an algorithm is serving us articles from all over the internet, some of which comes from known sources and some from more unfamiliar corners, and some of which may be entirely fake, we won’t know how to read those signals. Moreover, because we have not made the choice ourselves, we may not even remember to make the correction.

As we lose the ability to listen to each other, democracy becomes less meaningful and closer to a census of the various tribes, who each vote based more on tribal loyalties than on a judicious balancing of priorities. The biggest coalition of tribes wins, even if its candidate is a known child molester, or worse. The winner does not need to deliver economic or social benefits even to his own supporters as long as the supporters worry enough about the possibility of takeover by the other side; knowing that, he or she will do their best to stoke those fears. In the worst case, the winner can then use the power gained in this way to take control of the media to shut down any alternative voice, so there is no more competition to worry about. Prime Minister Orbán has successfully done this in Hungary, and many others are not far behind.

Moreover, there is an expanding circle of violence—against blacks, women, and Jews in the United States, against Muslims and lower castes in India, and against immigrants in Europe—that is probably not uncorrelated with the unabashed expressions of vituperation the current polarized climate permits, including by heads of state. The murderous mobs in India and Brazil, and the recent shooters and pipe-bomb senders in the United States or New Zealand seem to all emerge from those vortices of paranoid thinking, where the same falsehoods bounce back and forth. It has not yet reached proportions of a civil war or a genocide, but history suggests that it could.

As we have already seen, our reaction to the other is closely tied to our self-confidence. Only a social policy founded on respect for the dignity of the individual can help make the average citizen more open to ideas of toleration.

There are also possible interventions at the group level. Racism, anti-immigrant views, and the lack of communication across party lines originate, for many people, with an initial lack of contact. Gordon Allport, a professor of psychology at Harvard, formulated what he called the contact hypothesis in 1954.81 This is the idea that under appropriate conditions, interpersonal contact is one of the most effective ways to reduce prejudice. By spending time with others, we learn to understand and appreciate them, and as a result of this new appreciation and understanding, prejudice should diminish.

The contact hypothesis has been intensively studied. A recent review identifies twenty-seven randomized controlled trials (RCTs) investigating Allport’s idea. Overall, these studies find that contact reduces prejudice, although the review calls attention to the importance of the nature of the contact.82

If this is right, schools and universities are obviously key. They bring together young people from different backgrounds in a single location, at an age when everyone is much more plastic. In one large US university, where roommates were assigned at random, a study found that white students who happened to end up with African American roommates were significantly more likely to endorse affirmative action, and that white students assigned roommates from any minority group were more likely to continue to interact socially with members of other ethnic groups after their first year, when they had full freedom in choosing whom to associate with.83

This process of socialization could start even earlier. A policy change in Delhi demonstrated the power of bringing together young children from very different backgrounds. Starting in 2007, elite private schools in Delhi were required to offer places to poor students. In an ingenious study on the impact of this policy change, randomly chosen children were given the responsibility to select teammates for a relay race.84 Some of them attended schools that had already admitted poor children, and some attended schools that had not done so yet. And, within schools, some children were in study groups with poorer children (based on the first letter of their first name), and some were not. To help them decide who they wanted to partner with in the race, they were all given a chance to observe everyone else run a test race. There was, however, a catch. They had to agree to have a playdate with whomever they picked for their team. The study found that those students from affluent families who had not been exposed to poor students in their school avoided picking them, even when they were better runners, to avoid having to spend time with them. But those who’d had some exposure to children from less-well-off families in their schools, thanks to the new policy, were much more likely to pick the best runner even if the child was from a poor family, because the prospect of a playdate was no longer all that daunting. And those who were in a study group with poor children were even more likely to invite poor child to run and play with them. Familiarity performed its magic.

One implication of this evidence is that diversity in the student body of educational institutions is valuable in and of itself, because it durably affects preferences. Affirmative action was originally envisioned in the United States partly as compensation for historical injustice, and partly as a way to level the playing field between the whites, who had the advantage of many generations of advanced education, and the rest. But it goes much beyond that. What the twenty-seven RCTs on the effect of contact on tolerance imply is that this mixing is one of the most powerful instruments we have for making society more tolerant and more inclusive. The problem is that affirmative action itself is now a polarizing idea.

In the spring of 2018, New York City struggled with the redesign of the admission system for its elite public schools, which is currently based on an exam and lets in very few Latinos and African Americans. At the same time, Asian Americans were suing Harvard for discrimination, on the grounds that, in order to achieve its diversity goals, Harvard artificially limits the number of Asian American students it admits. Additionally, the Trump administration has been urging schools to stop taking race into consideration in their admission decisions. The US Supreme Court has so far resisted pressure to forbid any discrimination based on race, but it is not clear how long it will hold out.

In India, the debate is framed in terms of the actual quotas in educational institutions and government jobs for the castes historically discriminated against. These quotas are much resented by the upper castes, resulting in frequent protests and lawsuits challenging the validity of the law, especially on the grounds that a disproportionate share of the reserved slots end up with the more privileged among the lower castes, who perhaps need them less. (They are poetically referred to as the “creamy layer.”) The Indian court system has been sympathetic to this complaint, and has made eligibility for the quotas subject to an income qualification: you have to be poor enough to qualify. At the same time, other social groups have been lobbying to be included in the quotas, which would serve to dilute them. As a result, the system of reservations is almost incessantly being fought over somewhere or the other in the country, with not infrequent outbreaks of violence.

The idea of “merit” plays a key role in this debate. At the heart of the argument is the idea that test scores provide an objective measure of merit, a measure of how suitable the candidate is for the job or a place in the university, and therefore affirmative action discriminates against “meritorious” candidates, as they are called in India. Given everything we have seen in this chapter, that seems like a very unlikely proposition. Self-discrimination undermines confidence and test performance. A history of being underestimated, patronized, ignored, or despised by teachers and supervisors because you happen to be from a particular group will make it harder to achieve. Moreover, as we both know, growing up in a household where books are everywhere and dinner table conversation often centers around fine points of mathematics or philosophy, whether or not you always enjoy it, becomes a distinct advantage when you are writing your college essays. A lower-caste candidate who performed as well as Abhijit in the high school leaving exam had to jump through more hoops to get there and is for that reason likely to be more talented.

The fuzziness of the notion of merit was the bone of contention between two first-rate empirical economists, David Card and Peter Arcidiacono, who were retained by the two sides in the Students for Fair Admission v. Harvard trial. On the plaintiffs’ side, Arcidiacono argued that Asians must be discriminated against because admitted Asians have higher grades and higher test scores than any other group. In other words, given the same test scores, an Asian student is less likely to be admitted to Harvard than a white student (or an African American).

On the Harvard side, Card had a number of objections to Arcidiacono’s analysis, including the point that the objective of diversity in parental background and intended major was legitimate. But the most striking difference came from their interpretations of the “personality rating,” meant to capture the candidate’s leadership qualities and integrity. Asian students systematically have higher academic and extracurricular ratings, but lower personality ratings, and once we account for that they are no less likely to be admitted than white students.

For Card, this proves there is no discrimination. Arcidiacono contends that the personality rating is exactly the way Harvard discriminates against Asians. In the debate, a rather ironic parallel with history did not go unnoticed. In the 1920s, then Harvard president Abbott Lawrence Lowell attempted to introduce quotas to limit the admission of Jews to the university. This failed, but he put in place the system of “holistic” admission, meaning a system that values personal characteristics beyond grades, which was used to limit the number of Jews. Students for Fair Admission wants to make the case this is happening again.

The debate illustrates the essential treacherousness of the idea of merit, and the very notion of what constitutes quality. On the one hand, “personal qualities” may reflect (perhaps unwittingly) a form of belonging to a club, with secret handshakes not taught in the average public school. The personality rating may indeed be a not so subtle way to keep a certain type of student out (whether or not they are Asian) and ensure the smooth intergenerational transmission of elite status. On the other hand, the fact that among applicants African Americans systematically have higher personality ratings than whites or Asians may well reflect what we mentioned earlier: since admissions at Harvard require a sterling academic record, a child from a disadvantaged background must have quite unusual personal skills to be even considered, especially since the child might have had to survive worse schools and perhaps a more challenging home environment.

There is no evident solution to this problem. As a flagship producer of the next generation of leaders, Harvard clearly needs to find a place for students from all social groups, and a massive overrepresentation of any particular social group relative to its weight in the population is both perhaps undesirable in a democracy and likely to lead to political problems. But we need a more transparent social conversation about the design of affirmative action. The current implementation of affirmative action policies, which dances around the concept of race instead of directly confronting it, is probably not anywhere close to ideal. The Harvard challenge is both inevitable and perhaps desirable in that it makes society confront its own inconsistencies.

From the perspective of the narrow objective of affecting preferences by increasing contact between social groups, the growing resentment of affirmative action poses a problem. Allport’s original hypothesis was that contact would reduce prejudice, but only if some conditions were satisfied. In particular, he held that reduced prejudice would result when the contact happened in a setting where there was equal status between the groups in the situation, common goals, intergroup cooperation, and the support of authorities, law, or custom. Extremely contentious integration is unlikely to produce these conditions. For example, if high school students feel they are competing for slots in college and, worse, if they have the impression this competition might be tilted against them, they may come to resent the other group even more.

That this is a very real concern is demonstrated by a clever recent study.85 In the state of Uttar Pradeshin in India, a researcher ran an eight-month-long cricket league involving 800 players, all young men, chosen randomly out of 1,261. In the league, about a third of the players were assigned to homogeneous-caste teams; the others were in mixed-caste teams. Like others, the study found many positive effects of collaborative contact. Compared to those who played on single-caste teams, the young men who played on mixed teams were more likely to be friends with people from other castes after the experiment, and not just those from their teams. When they had a chance to select their teams, they selected better teams for future matches, since they made their choices based on talent, not caste.

But who they played against mattered. Those in teams randomly assigned to play against other-caste teams were less likely to make friends with people from other castes than those who played only against their own caste, or even those who never got to play anyone. Competition undermined contact.

These somewhat less optimistic results make the important point that contact may not be enough to produce tolerance; it may be necessary to have shared goals. Both in 1998 and in 2018, the victory of France’s team at the soccer World Cup had exactly this effect on the nation as a whole. In particular, the fact that some of the team’s champions grew up and learned their skills in the suburbs of Paris notorious for their dilapidated housing projects and their car-burning riots did create a sense of goodwill and shared purpose. In that moment, everyone could see that the kids from the 93 (as one disadvantaged district in the north of Paris is known) were not all lazy bums who skipped school and committed petty crimes. Behind France’s winning black-blanc-beur (“black-white-Arab”) team was the effort and the discipline of tens of thousands of young kids working hard to make it.

Since there are obvious limits on integration through universities, mixed neighborhoods offer a useful alternative. The problem is that mixed neighborhoods have a proclivity toward being unstable, as Tomas Schelling, who won the Nobel Prize in economics, demonstrated.86 Suppose homeowners are happy to live in mixed neighborhoods, but not in neighborhoods mostly dominated by another group. Then they must live in fear of the day when, by chance perhaps, a few of their own move out and are replaced by the other. The neighborhood becomes a little less attractive to people like them, and now they all start worrying that if a few more leave, let’s say because they are also having the same thoughts, or because they are less tolerant, they will be forced to leave as well. The tension of whether and when that may happen can become unbearable, so anyone who can get out leaves. This is what Schelling called the tipping point.

David Card studied the increase in segregation that happened in the United States in the 1970s, 1980s, and 1990s, and it does indeed look like there is this tipping point property.87 If the fraction of blacks in the neighborhood was less than some number, it remained stable; if it became higher, there was a large outflow of the white population in subsequent years. Chicago, for example, had a particularly low tipping point. If the black population in a neighborhood was less than 5 percent in 1970, it remained at that level afterward, but if it was any more than that the fraction of whites soon plummeted. On average across US cities, Card and his colleagues found tipping points ranging between 12 percent and 15 percent.

The way to prevent the segregation implied by the tipping point logic is to build public housing targeting low-income residents and disperse that housing throughout the city, so there are no “pure” neighborhoods available. In a fancy neighborhood in Paris, where we spent a year, the building next to us was a housing project. The children all attended the same neighborhood school and played in the same park. At that age, they clearly inhabited the same universe. It may not be possible to be as bold as Singapore, where strict quotas ensure some amount of mixing between ethnic groups in every block of residential housing, but it seems possible to reserve a certain fraction of public housing in every neighborhood.

The challenge of implementing such a policy is mostly political. It seems easy enough to imagine how to do it well if the political will is there: spread the public housing around, give everyone a lottery number, have a public lottery every time new housing becomes available, and make it easy to check that the winners get the housing. The difficulty is that public housing in fancy neighborhoods is very tempting for local politicians to use as patronage, but with enough political will it can probably be overcome.88

Nevertheless, in the near future, while most poor people still live in low-income neighborhoods, shared schools are another way to integrate the population. For this to happen, children will need to be moved. Busing a large number of children to foster school diversity, as it was done in Boston at some point, is however unpopular, in part for the very good reason that young children do not enjoy being bused. The best idea may be to allow children from designated low-income neighborhoods to attend schools outside their neighborhoods. The METCO program in the United States, which organized the busing of minority children to majority schools, was shown to be beneficial to the minority children without any harm to the test scores of majority children. The latter, who would have mostly spent their lives in largely white enclaves, ended up being exposed to a much more diverse population, which as we have seen durably affects worldviews and preferences.89

The sum total of all our proposals might seem modest in the face of what feels like a tsunami of prejudice. But that would be to miss the main point of this chapter, which is that such preferences are as much part of the symptoms of the malaise as its cause, perhaps more. Prejudice is often a defensive reaction to the many things we feel are going wrong in the world, our economic travails, and a sense that we are no longer respected or valued.

This has four important implications. First, and most obviously, the expression of contempt for those who express racist sentiments, fraternize with racists, or vote for them (“deplorables”) serves only to reinforce those sentiments, founded in the suspicion the world no longer respects us. Second, prejudice is not an absolute preference; even so-called racist voters care about other things. North India in the 1990s and early 2000s saw a period of mostly caste-based polarization. However, by 2005 this had run its course. The lower castes who had aligned themselves with explicitly caste-based parties (as against the less transparently caste-based BJP, Prime Minister Modi’s party) had begun to question whether they were getting enough from their parties. Mayawati, the leader of one of those parties, decided to rebrand herself as the leader of all poor people, including poor upper castes, and won the 2007 Uttar Pradesh state elections on that basis. She went for broad inclusivity, not narrow sectarianism.

More recently, in the United States we are struck by the curious history of the once much hated Affordable Care Act, or Obamacare. As the signal policy initiative of the much despised black Kenyan Muslim Barack Obama, it was something that many Republican governors refused to have anything to do with, and many refused federal subsidies to expand Medicaid, a key mechanism to extend health care coverage under the Affordable Care Act. Yet by the 2018 midterm election initiatives to expand Medicaid were on the ballot in the deep-red states of Utah, Nebraska, and Idaho. They were approved in all three. Kansas and Wisconsin also elected new Democratic governors who vowed to expand Medicaid where their Republican predecessors had not. This is not because people in these places became Democrats; they still voted for Republican congressmen and senators, often with very conservative views. But on this issue many seem to have decided to ignore the warnings of the Republican establishment and go with their own understanding of what was going to be good for them. Economics trumped Trump.

This is related to our third point. The fact that voters put a premium on race or ethnicity or religion, or even the articulation of racist views, does not have to mean they feel very passionately about them. Voters do realize political leaders choose to play the ethnic or race card when convenient. Part of the reason they still vote for those politicians is they are deeply cynical about the political system, having convinced themselves all politicians are more or less alike. Given that, they might as well vote for the guy who looks or sounds like them. In other words, ethnic or bigotted voting is often just an expression of indifference. But that means it is surprisingly easy to make them change their minds by highlighting what is at stake in an election. In 2007, in Uttar Pradesh, an Indian state famous for its caste-based politics, Abhijit and his colleagues managed to make 10 percent of voters move their vote away from their own caste-party using only a combination of songs, a puppet show, and some street theater—all carrying the simple message “Vote on development issues, not on caste.”90

Which leads us to our final and perhaps most important point. The most effective way to combat prejudice may not be to directly engage with people’s views, natural as that might seem. Instead, it may be to convince citizens it is worth their while to engage with other policy issues. That leaders who promise them a great deal and even make grand gestures toward it may not actually deliver much more than those gestures, in part because doing anything more is not easy. In other words, we need to reestablish the credibility of the public conversation about policy, and prove that it is not just a way to use big words to justify doing very little. And of course we need to try to do what it will take to assuage the anger and deprivation so many feel, while acknowledging it will be neither easy nor quick.

This, as we explained in chapter 1, is the journey we started in this book. We started with the issues where the most is known and understood: immigration and trade. Even there, there is a strong tendency for economists to pronounce on these issues with categorical answers (“immigration is good,” “free trade is better”) without accompanying detailed explanations and necessary caveats, which massively undermines credibility. We now turn to issues that are much more contentious, even among economists: the future of growth, the causes of inequality, the challenge of climate change.

We will attempt to do the same exercise of demystification for these topics, while recognizing that what we have to say will occasionally be based on more abstract arguments than the ones we have made so far, and somewhat less well grounded in evidence. These issues are nonetheless so central to our view of the future (and the present) that there is no way to talk about how to do better economic policy without embracing them.

In all of this the role of preferences is crucial. It is obviously impossible to talk about growth and inequality and the environment without thinking of needs and wants, and therefore preferences. We have seen that wants may not be needs—people seem to value bottles of wine based on their own social security number rather than the pleasure of drinking—and needs may not be wants—is a television a need or a want? These will of course be central concerns in the coming chapters, implicit and sometimes explicit in the arguments we make and the view of the world we project.