Tim Harford The Undercover Economist

Articles published in June, 2015

The psychology of saving

‘There is one dramatic success for behavioural economics — the way it has shaped pensions’

“THERE ARE IDIOTS. Look around.” So began a famous economics paper by Larry Summers — a lauded academic before he became US Treasury secretary. It is perhaps the most concise expression of behavioural economics, the branch of economics that tries to take psychology seriously.

Behavioural economics is appealing not only because it is realistic but also because it is vastly more charming than the traditional variety. Championed by economists such as Richard Thaler (co-author of Nudge and author of a new book, Misbehaving ) and psychologists such as Nobel laureate Daniel Kahneman (author of Thinking, Fast and Slow), it has triumphed in the “smart thinking” section of the bookshop and exerted increasing influence in academia.

It can be hard to turn psychological insights into rigorous academic models, and even harder to turn them into good policy. But there is at least one dramatic success for behavioural economics — the way that it has shaped pensions. At a recent Financial Times event, Professor Thaler rightly celebrated this as the field’s greatest triumph.

Other than Thaler’s own evangelism, the reason for this success is twofold. First, when it comes to pensions there is a large gap between what we do and what we should do. Second, bridging that gap is fairly simple: we need to encourage people to save more, and in most cases those savings should flow into simple, low-cost equity tracker funds. The only comparable example that springs to mind is smoking: many smokers are making themselves unhappy and would be better off if they could find a way to stop. And as a classic research paper in behavioural economics concludes, taxes on cigarettes seem to make smokers and potential smokers happier by prompting them to quit, or never start.

Given the problem — people need to be nudged into saving more — the biggest pension policy breakthrough has been automatic enrolment, a cornerstone of modern UK pension policy and widely used in the US too. A typical defined contribution pension invites people to pay money into a pension pot, often enjoying tax advantages and matching contributions from an employer. Yet people procrastinate: money seems tight, retirement is a long way off, and who wants to fill in forms? Automatic enrolment reverses the default, deducting pension contributions from our payslips unless we take active steps to opt out. The process respects our autonomy — you can opt out if you wish — but makes it easy to do what we probably should be doing anyway.

A clever supplement to this approach is “Save More Tomorrow”, a scheme in the US whereby people make an advance commitment to redirect part of their pay rises into the pension. At a 50/50 ratio, for example, a 2 per cent pay rise becomes a 1 per cent pay rise and a 1 percentage point increase in pension contributions. It doesn’t take long for a 3 per cent contribution (which is inadequate but typical in the US) to become something more sensible, such as 10 or 15 per cent. Thaler has been a driving force behind both ideas.

Of course, these tactics do not work for everyone. I once spoke at a book festival in Australia and found that a slice of my modest fee had been automatically invested in an Australian pension for me. This benefited nobody except some administrators and the postman who delivered the letters from Australia, detailing the evaporation of my tiny pension pot.

A more serious difficulty is choosing the right level of default contribution. A default that is too aggressive — automatically deducting 25 per cent of salary — jolts most people into opting out. A default that is too low, such as 3 per cent of salary, could conceivably be worse than the old opt-in default of zero. Many people who might have taken an active choice to save 6 or 7 per cent rather than nothing end up settling instead for the default. As mentioned, 3 per cent is a common level for automatic enrolment in the US, for no good reason other than historical accident. Yet it is dangerously low.

There is also a painful conflict of interest at the heart of any corporate pension plan. From the perspective of classical economics, companies will offer generous pensions if they want to attract capable staff. It is expensive to subsidise a pension but staff value the subsidy, making it worthwhile.

From the more realistic perspective of behavioural economics, a tension emerges. A benevolent planner, armed with behavioural insights, would nudge people into a passive pension investment with almost no conscious thought. But a corporate human resources director would want to remind employees how generously their pensions are being subsidised. That means frequent reminders and ample opportunity to admire the pension pot — even if such admiration leads to anxiety about uncertain returns, or expensively trading shares within the pension.

There are approaches that might keep both the behavioural economist and the HR director happy. For example, a pension pot that is expressed in terms of daily or weekly income in retirement, adorned with photographs of cruise ships, seems more appealing than an abstract and rather meaningless lump sum.

We cannot blame behavioural economics for this tension but it is real. As automatic enrolment becomes the norm, it will be important to keep an eye on how corporations respond.

Written for and first published at ft.com.

Free email updates

(You can unsubscribe at any time)

Teamwork gives us added personbyte

‘Complex products require elaborate networks of teamwork, and only a few places manage the trick’

Is economic life getting more complicated? In some ways, no. It’s much easier to use a computer than it used to be, or to make an international phone call, or to buy avocados. But, in many ways, complexity is on the rise.

This is true for products. When once we used to buy simple chunks of matter, such as copper, tin or wheat, now we buy smart watches, movie downloads and ready meals — these are things whose structure is a vital part of their value. You can melt and cool copper, or scatter and then rebuild a pile of wheat, and no great harm will come to either. Put your phone into a blender and you’ll find it has changed in ways that matter a great deal.

Scientific ideas are also becoming more complex. Benjamin F Jones, an economist, has used large databases of academic articles and patents to show that researchers and inventors are getting older and have narrower specialisations. This seems to be because there is much more science to be known, and scientists must devote more time to mastering what is already known before they can contribute original research.

César Hidalgo, a physicist at Massachusetts Institute of Technology and author of Why Information Grows, coins the word “personbyte” to describe the amount of knowledge that one person can reasonably know. The personbyte isn’t getting any smaller but — relative to the knowledge that needs to be mustered to produce a modern scientific paper, or a computer, or a car — the personbyte looks ever more inadequate.

The way to escape the constraint of the personbyte is to work in larger teams, and this is exactly what Jones finds in academic and patent databases: research teams are bigger than they were 40 years ago. This is a natural consequence of the fact that a personbyte isn’t big enough to process the knowledge required for modern science or engineering. One person cannot hold all the necessary know-how in her head, so she must work together with others.

How easy is this collaboration? It depends. Some knowledge is easily copied. Once someone has invented the wheel, anyone else can simply copy the idea and no more elaborate teamwork is needed than that. Other knowledge can be embodied in a product and widely dispersed. I don’t know how to make a laptop computer but I know how to use one. In a sense, I am standing on the shoulders of Ada Lovelace, Alan Turing and Bill Gates. But I do not need to spend any time in meetings with these people, which is just as well. Better yet, my laptop is built from simpler standalone modules, such as the central processing unit, the keyboard, the operating system and the hard drive. Individuals — or teams, or firms — can work on these modules even if no organisation has mastered the skills necessary to build every part of the laptop from scratch. The modular nature of the computer makes it straightforward to use earlier knowledge.

But some knowledge requires far more challenging collaborations. This knowledge is tacit, hard or perhaps impossible to describe. It may be easy to send data around the world but data may not be enough. Knowledge may be weightless in principle but, as César Hidalgo points out, we find it easier to move heavy copper from mines in Chile to factories in Korea than to move manufacturing know-how from Korea to Chile.

Hidalgo argues persuasively that networks of people and companies with such tacit knowledge are an essential part of a modern economy. They form essential capabilities: how to make a plasma display, or champagne, or a financial derivative. Simpler products require simpler networks of collaboration, and can be produced almost anywhere. More complex products require elaborate networks of teamwork, and only a few places manage the trick.

More than 20 years ago, economist Michael Kremer published “The O-ring Theory of Economic Development”. His title refers to a simple seal whose failure destroyed the Challenger space shuttle and killed seven astronauts. Kremer wanted us to think about weak links. A string quartet is not much better than its worst player. A gourmet meal could be ruined by a clumsy chef, a faulty oven, a rude waiter, a decaying ingredient, or a rat scurrying across the dining room. These are O-ring products.

The logic of an O-ring world is that the most skilled workers end up collaborating with each other, using the best equipment. Chef Heston Blumenthal does not work at Burger King. It makes more sense for Joshua Bell to play a Stradivarius and for a street busker to play a worm-eaten fiddle than for the two musicians to swap.

Inequality soars in O-ring worlds because the more complex a product or service, the greater the value of someone who can avoid errors. And a weak link somewhere in your economy can spread like a cancer. Why should a young person in Nigeria study hard if her efforts will be dissipated by electrical blackouts, criminal gangs or corrupt officials?

The economic world is unlikely to become simpler. But we may rise to the challenge better if we think about both the social and institutional support that helps make complex collaborations possible — and the simple modular engineering that makes complex collaborations unnecessary.

Written for and first published at ft.com.

The truth about our norm core

‘Social pressure matters but it is not the only thing that matters. Facts can trump groupthink’

While not quite as infamous as Philip Zimbardo’s Stanford prison simulation, or Stanley Milgram’s obedience research, Solomon Asch’s conformity experiments remain among the most celebrated in psychology. In 1951, Asch’s research showed that our judgments about simple factual matters can be swayed by what people around us say. The finding echoed down the decades. Milgram found in 1961 that people were willing to administer apparently dangerous electric shocks when ordered to do so by an experimenter. In 1971, Zimbardo set up an imitation prison in a Stanford University basement with subjects given the role of guards and prisoners, then observed as the guards humiliated the prisoners.

Between them, the three academic psychologists taught us that in order to fit in with others, we are willing to do almost anything. That, at least, is what we are told. The truth, as so often, is more interesting.

Asch gave his subjects the following task: identify which of three different lines, A, B or C, was the same length as a “standard” line. The task was easy in its own right but there was a twist. Each individual was in a group of seven to nine people, and everyone else in the group was a confederate of Asch’s. For 12 out of 18 questions they had been told to choose, unanimously, a specific incorrect answer. Would the experimental subject respond by fitting in with the group or by contradicting them? Many of us know the answer: we are swayed by group pressure. Offered a choice between speaking the truth and saying something socially convenient, we opt for social convenience every time.

But wait — “every time”? In popular accounts of Asch’s work, conformity tends to be taken for granted. I often describe his research myself in speeches as an example of how easily groupthink can set in and silence dissent. And this is what students of psychology are themselves told by their own textbooks. A survey of these textbooks by three psychologists, Ronald Friend, Yvonne Rafferty and Dana Bramel, found that the texts typically emphasised Asch’s findings of conformity. That was in 1990 but when Friend recently updated his work, he found that today’s textbooks stressed conformity more than ever.

This is odd, because the experiments found something more subtle. It is true that most experimental subjects were somewhat swayed by the group. Fewer than a quarter of experimental subjects resolutely chose the correct line every time. (In a control group, unaffected by social pressure, errors were rare.) However, the experiment found that total conformity was scarcer than total independence. Only six out of 123 subjects conformed on all 12 occasions. More than half of the experimental subjects defied the group and gave the correct answer at least nine times out of 12. A conformity effect certainly existed but it was partial.

This surprised me, and it may surprise others who have read popular accounts of the so-called conformity studies. I doubt that it surprised Asch. Conformity was already a well-established finding by 1951, and his experiments were designed to contrast with earlier research on social norms. This previous research showed that people conformed to social pressure in situations where there was no clear correct answer — for instance, when asked to identify which of two ungrammatical sentences was the most ungrammatical. But Asch wanted to know if peer pressure would also wield influence when the crowd was unambiguously wrong. His research provided an answer: social pressure is persuasive but, for most people, the facts are more persuasive still.

Myths about famous experiments have always grown in the telling. It seems most unlikely that Archimedes ran naked through the streets of Syracuse yelling “Eureka!”, and an apple probably did not strike Newton’s head. But there seems to be something particularly attractive about these famous psychology experiments that paint us all as sheep — even when the experiments may have been flawed, impossible to replicate or (as with Asch’s work) have simply found something much more subtle than the myth would have us believe.

The psychologist Christian Jarrett comments, “the resistance to tyranny shown by many participants in Zimbardo’s prison study has largely been ignored, and so, too, has the disobedience shown by many participants in Milgram’s seminal work.”

Zimbardo’s Stanford prison experiment was shocking stuff, and raised serious questions about research ethics. But we should also ask questions about what Zimbardo really found. By his own admission he gave a strong steer to the guards, and cast himself as their ally in a quest to dehumanise the prisoners. “We’re going to take away their individuality in various ways,” he told them. Other psychologists have suggested that this was more a test of obedience to Zimbardo than a demonstration that sadism blooms given the opportunity.

Few textbook accounts of the study mention Zimbardo’s attempt to influence the guards; nor do they point out that two-thirds of the guards refrained from sadism.

Social pressure matters but it is not the only thing that matters. Solomon Asch showed that facts can trump groupthink. It would be ironic if our own biased recollections of his finding proved him wrong.

Written for and first published at ft.com.

Down with mathiness!

‘In the recent UK election campaign, a diet of numbers was stuffed into voters like feed into French ducks’

The American Economic Review isn’t usually the place for trash talk but a brief new article by Paul Romer is the closest academic economics is likely to come to a spat between boxers at a pre-fight weigh-in. Romer, a professor at New York University, is worried that a new trend in economics — “mathiness” — is polluting the discipline. And he names names — including Robert Lucas and Edward Prescott, both Nobel laureates, and inequality guru Thomas Piketty.

In a follow-up comment, “Protecting the Norms of Science in Economics”, Romer says: “I point to specific papers that deserve careful scrutiny because I think they provide objective, verifiable evidence that the authors are not committed to the norms of science.”

Romer adds that if his suspicions are confirmed, such people should be ostracised — suggesting that Nobel Prize winners should be ejected from academic discussion because of their intellectual bad faith. This is strong stuff.

Romer, though, has rarely stuck to the academic script. In the late 1980s he developed a new approach to thinking about economic growth that mathematically modelled the development and spread of ideas, an achievement that many regard as worthy of the Nobel memorial prize in economics. But Romer then drifted away from academia, first founding an online learning platform called Aplia, and then campaigning for a radical development idea, “charter cities”.

Does economics have a mathiness problem? Many casual observers would say, “of course”. Economics has a reputation for producing rigorous nonsense.

But Romer’s attack is much more focused. He doesn’t mean that economics uses too much mathematics but that some economic theorists are pushing an ideological agenda and using fancy mathematics to disguise their intentions. They can redefine familiar words to mean unfamiliar things. They can make unrealistic assumptions. They can take hypothetical conclusions and suggest they have practical significance. And they can do all these things with little fear of detection, behind a smokescreen of equations. If Romer is right, some economics papers are Orwellian Newspeak dressed up as calculus.

In his short essay “Politics and the English Language”, Orwell argued that there was a “special connection between politics and the debasement of language”. While some people wish to communicate clearly, the political writer prefers a rhetorical fog. And the fog can spread. Writers who should know better imitate sloppy writing habits. Readers become jaded and stop hoping that anyone will tell them the truth.

Romer fears a similar rot at the heart of economics. As some academics hide nonsense amid the maths, others will conclude that there is little reward in taking any of the mathematics seriously. It is hard work, after all, to understand a formal economic model. If the model turns out to be more of a party trick than a good-faith effort to clarify thought, then why bother?

Romer focuses his criticism on a small corner of academic economics, and professional economists differ over whether his targets truly deserve such scorn. Regardless, I am convinced that the malaise Romer and Orwell describe is infecting the way we use statistics in politics and public life.

There being more statistics around than ever, it has never been easier to make a statistical claim in service of a political argument.

In the recent election campaign in the UK, a diet of numbers was stuffed into voters like feed into French ducks. A fact-checking industry sprang up to scrutinise these numbers — I was part of it — but the truth is that most of the numbers were not false, unhelpful. Instead of simply verifying or debunking the latest number, fact checkers found themselves spending much effort attempting to purify muddied waters.

This is infuriating — for the public, for the fact checkers, and for the scientists and statisticians who take such pains to gather evidence. Imagine their dismay when the politicians seize that evidence and hold it up for protection like a human shield. Good statistics matter; without them it is almost impossible to understand the modern world. Yet when statistics are dragged into political arguments, it is not the reputation of politics that suffers but the reputation of statistics. The endgame isn’t pretty: it becomes too much trouble to check statistical claims, and so they are by default assumed to be empty, misleading or false.

Just as the antidote to Newspeak isn’t to stop using language, the antidote to mathiness isn’t to stop using mathematics. It is to use better maths. Orwell wanted language to be short, simple, active and direct. Romer wants economists to use maths with “clarity, precision and rigour”. Statistical claims should be robust, match everyday language as much as possible, and be transparent about methods.

Some critics believe that economics should conduct itself in plain English at all times. This is, I think, unreasonable. Mathematics offers precision that English cannot. But it also offers a cloak for the muddle-headed and the unscrupulous. There is a profound difference between good maths and bad maths, between careful statistics and junk statistics. Alas, on the surface, the good and the bad can look very much the same.

Written for and first published at ft.com.

Mind the fair trade gap

‘If fair trade does deliver higher incomes for farmers, it may prove too successful for its own good’

In 2001, the world price of coffee sank to its lowest ebb for decades, threatening dreadful hardship for the often-poor farmers who grow the sainted berry. It was also around that time that fair trade coffee seemed to come of age, with a common certification mark launched in 2002, and the product becoming a familiar sight in supermarkets and coffee chains.

The premise of fair trade is that the disparity between poor coffee farmers and prosperous drinkers presents both a problem and an opportunity. The problem is that farmers often live a precarious existence: geographically isolated and growing a crop with a volatile price. The opportunity is that many western consumers care about the earnings and conditions of the people who grow their coffee, and have some money to spare if only it might reach those people.

Unlike a taxi driver or a waiter, you can’t just tip the guy who grew your coffee. The fair trade answer to the conundrum is a labelling scheme: an inspector verifies that all is well on the farm, with good conditions and a higher price paid for coffee; this information is conveyed to consumers by way of a recognisable trademark, the most famous of which is the Fairtrade logo. It’s an appealing idea — a voluntary scheme that helps people who want to help people. (Or rather, several voluntary schemes: there is more than one fair trade label, alongside diverse certification schemes such as Organic or Rainforest Alliance.) Who wouldn’t want a better deal for farmers who are poor and work hard? But there are problems with the idea too.

The most obvious problem is that this labelling scheme costs money. Flocert, a certification body set up by the Fairtrade Labelling Organization, charges farmer co-operatives €538 merely to apply for certification, plus an initial audit fee of €1,466 even for a small co-op. Cynics might suspect bureaucratic bloat but the costs may well be real. It cannot be cheap to check pay and conditions in some remote Peruvian coffee plantation. But every euro spent on certification is a euro that the farmer cannot spend on his family. And larger co-operatives from richer, better-connected countries are more likely to find it worthwhile to pay for certification. For this reason, economist and fair trade critic Ndongo Sylla says that fair trade benefits “the rich”. That seems too strong; but it is certainly a challenge for the fair trade model to reach the poorest.

A second problem is that fair trade certification cannot guarantee fair trade sales. If coffee importers want to put the Fairtrade mark on their coffee, they must find a Fairtrade certified producer and pay the Fairtrade price, which reflects both a modest premium and a guaranteed minimum price. But importers are not obliged to buy fair trade coffee and may avoid it when it gets too expensive, exactly when the premium is most needed. A study by Christopher Bacon found that during the price slump of 2000 and 2001, Nicaraguan coffee farmers were earning twice as much per pound when selling fair trade coffee as when selling the uncertified stuff. But much of their coffee could not find a buyer at such rates and was sold at market rates instead; as a result, the average price premium, while substantial, was much lower at around a third.

Another study, by Tina Beuchelt and Manfred Zeller, found the fair trade certified farmers in Nicaragua started at a similar income level to conventional farmers and, if anything, slipped backwards. A recent survey by Raluca Dragusanu, Nathan Nunn, and Daniele Giovannucci was more upbeat but still found the evidence in favour of fair trade “mixed and incomplete”.

A final irony is that if fair trade does deliver higher incomes for farmers, it may prove too successful for its own good. If coffee farmers are able to sell more coffee at a premium price, more people will want to become coffee farmers. One possible result is that the market price for uncertified coffee falls and, on balance, coffee farmers are no better off.

As the development economist Paul Collier once wrote, fair trade certified farmers “get charity as long as they stay producing the crops that have locked them into poverty”. It is a telling point. For all the good I may wish the people who make my coffee, a globalised tip jar makes a precarious foundation for their future prosperity.

Written for and first published at ft.com.