Randomistas
Page 11
Prisons are meant to serve four purposes: to rehabilitate inmates, to incapacitate them from harming the community, to make an example of them so as to deter potential wrongdoers, and to exact retribution on behalf of society. But the evidence – including from the California study – increasingly suggests that longer sentences may only serve one of these purposes: retribution. A five-year jail term costs society five times as much as a one-year sentence, but it’s doubtful it has five times the deterrent effect. If the aim is to make society safe, it’s better that we write our laws with thoughtful minds, not angry spleens.
*
If you’ve ever been the victim of a crime, you’ll know how tough it is to take a dispassionate view. But passion-driven policy often leads to laws like the ‘three strikes’ rule that saw Leandro Andrade sentenced to fifty years in jail when he stole $153 of videotapes from stores in Ontario, California.57 Andrade had previously been convicted of drug and burglary offences, so the videotapes were his third strike. He will be eighty-seven years old when released, and his incarceration will cost taxpayers over US$1 million.
When ‘Laura Norder’ shows up on the ballot paper, she wins a lot of votes. But as experimental criminologists have shown, our gut feel isn’t much of a guide to what actually works to improve community safety. Whether it’s prevention, policing, punishment or prison, the challenge isn’t just to carry out rigorous research, it’s also to do a better job of conveying the expert consensus to the public. Since 2011 the US Department of Justice has run the website CrimeSolutions.gov, which gives its highest rating to programs and practices backed by randomised experiments. Experimental criminologists such as Lawrence Sherman are slowly gaining the upper hand – but it’ll be a long time before we can finally put all those failed crime-fighting ideas behind bars.
7
VALUABLE EXPERIMENTS IN POOR COUNTRIES
Lariat Alhassan ran a business selling paint in Abjua, Nigeria.1 Business was passable, but she didn’t know how to expand. Alhassan sold the paint out of the boot of her car, travelling to meet her customers. To land bigger orders, she needed to have a showroom, but she couldn’t afford one without more sales.
In 2011, Alhassan heard an ad on the radio, inviting entrepreneurs to apply for a new competition. The advertisement told business owners: ‘It may be small today, but it won’t be after YouWiN! the youth enterprise and innovation competition.’ At first, Alhassan thought YouWiN! must have been a scam. It turned out to be a radical approach devised by the finance minister, Ngozi Okonjo-Iweala. As Okonjo-Iweala put it, ‘We had a large unemployment problem . . . especially with young graduates. And these are people who are ripe to be encouraged as entrepreneurs.’ So the Nigerian government ran a competition for entrepreneurial small businesses. Based on a short business plan, the government gave successful firms up to 10 million Naira (about US$64,000). The amount was equivalent to about a decade’s wages for the typical Nigerian.
Enter David McKenzie, a New Zealand–born development economist now working at the World Bank. To test the impact of YouWiN! grants on businesses, McKenzie persuaded the Nigerian government to randomise grants across some of the semi-finalists. The government announced 1200 winning businesses, with a majority chosen randomly by McKenzie. As the podcast Planet Money documented, one of the successful firms was Lariat Alhassan’s paint selling business. She used the money to hire staff, buy a delivery truck and rent a showroom.
Alhassan wasn’t the only winner who made good use of the money. Three years after the grants were distributed, McKenzie’s randomised trial compared the lucky winners with unlucky losers. He estimated that entrepreneurs who won a grant were more innovative, grew faster and took on more staff.2 Dividing the employment impact by the total cost of the program, each additional job cost the government $8500, making it far more effective than most similar programs. Reading the results, a prominent development economist asked, ‘Is this the most effective development program in history?’3
Perhaps no area of policy has seen such a rapid growth in randomised trials as development economics. In the 1990s there were fewer than thirty randomised experiments from developing countries published globally each year. In the 2010s there have been at least 250 randomised development studies published annually.4 Two-thirds of all impact evaluations in development economics now use randomisation.5 Randomised trials are proliferating across organisations such as the World Bank, the Bill & Melinda Gates Foundation, the UK Department for International Development and the US Agency for International Development. As Yale’s Dean Karlan put it, ‘There are hundreds and hundreds of randomized trials going on, and ten years ago that just wasn’t the case.’6
Part of the drive towards randomisation has been a recognition that there may not be a single answer to alleviating global poverty. There was a time when many development theorists thought the solution was simply for developing nations to be able to borrow on the international capital markets. Then there was a notion that the problem was overpopulation, and that providing free condoms would quickly raise living standards. After that came a push for debt forgiveness, on the principle that once countries were freed from their crushing loan repayments, they would grow quickly. Proposed by well-meaning idealists, none of these schemes lived up to their initial promises.7
The Greek poet Archilochus once wrote: ‘A fox knows many things, but a hedgehog one important thing.’8 In development, the randomistas are the foxes, and grand planners the hedgehogs. Where the planners have proposed a single reason why millions of people go to bed hungry each night, those who advocate randomised trials in development are searching for multiple solutions. Hedgehogs are confident they know what works. Foxes aren’t sure, but they’re sniffing everywhere.
Because the randomistas take an eclectic approach, tracking their progress is invariably messy. In this chapter I have sought to provide some order by grouping randomised trials in development into four groups: experiments that aim to create more business activity, experiments that seek to make government work better, experiments that try to improve health, and experiments whose goal is to raise educational standards.
*
In helping entrepreneurs build businesses, we have already seen from the YouWiN! experiment that access to cash can be crucial. If a smart youngster in Sweden has a brilliant idea for a start-up, she can probably borrow money from friends, family or a bank. An equally bright innovator in Somalia might not have access to any of these funding sources. But designing a financing system that works in a poor nation is no easy task. If you’ve ever lent money to someone who failed to repay the debt, you know the central challenge for a bank: how do you distinguish the conscientious from the crooked?
In 1976 Muhammad Yunus, a university professor in Bangladesh, lent US$27 of his own money to a group of poor village women who made bamboo furniture. As Yunus recounts, they repaid the loan and their businesses became more profitable. So he set up a bank to provide ‘microcredit’ – loans as small as a few dollars.
Fast-forward two decades, and microcredit was among the hottest trends in development. Yunus’s Grameen Bank had grown into a multi-billion-dollar organisation. The prime minister of Bangladesh argued that microcredit would ‘allow the world’s poorest people to free themselves from the bondage of poverty and deprivation to bloom to their fullest potentials’.9 Internationally, microcredit won supporters from the left for its focus on poor women and from the right for its emphasis on personal responsibility. A pair of supporters wrote: ‘If a social evangelist had a choice of picking one tool, one movement with the goal of emancipating the poorest women on earth, the microcredit phenomenon wins without serious competition.’10
As US president, Bill Clinton provided development assistance to microcredit programs and championed Muhammad Yunus for the Nobel Peace Prize.11 Awarding the prize to Yunus in 2006, the Nobel committee praised him for developing ‘micro-credit into an ever more important instrument in the struggle against poverty’. U2’s Bono wrote: �
��Give a man a fish, he’ll eat for a day. Give a woman microcredit, she, her husband, her children, and her extended family will eat for a lifetime.’12
Yet it turned out that the bold claims for microcredit were largely based on anecdotes and evaluations that failed to distinguish correlation from causation. By the 2000s, researchers had begun carrying out randomised trials of microcredit programs in Bosnia, Ethiopia, India, Mexico, Morocco and Mongolia. Summarising these six experiments, a team of leading development economists concluded that microcredit had no impact on raising household income, getting children to stay in school, or empowering women.13 Microcredit schemes did provide more financial freedom, and led people to invest more money in their businesses, but it didn’t make them more profitable.
Part of the reason that microcredit struggled to change lives is that it offered relatively small sums of money at high interest rates. Whereas the Nigerian entrepreneurs got to keep their YouWiN! prizes, microcredit borrowers had to pay back their loans with interest rates that sometimes exceeded 100 per cent. Microcreditors defended themselves by pointing out that they were cheaper than the knee-breaking loan sharks who came before them. But even the world’s best companies would struggle to pay 100 per cent interest rates and make a profit. As Dean Karlan puts it, there just aren’t that many golden-egg-laying geese around.14
As the gloss has come off microcredit, some economists believe that there may be larger benefits from helping people save than from helping them borrow. A survey of recent randomised trials finds that savings programs tend to boost income and wealth, increase spending on health care and education, and provide a buffer against unexpected shocks.15 Contrary to the past views of Bill Clinton and the Nobel Prize committee, the world’s poorest seem to have a greater need for bank accounts than credit cards.
‘The rich are different from you and me,’ wrote F. Scott Fitzgerald. ‘Yes,’ responded Ernest Hemingway, ‘they have more money.’16 In recent years, some researchers have explored an even simpler approach: giving cash to the poor. Working across Ethiopia, Ghana, Honduras, India, Pakistan and Peru, a research team tested the impact of providing the very poor with a package that included assets (typically livestock), income support, bank accounts and business training.17 A year after the end of the program, its effects could still be clearly seen. Compared with randomly selected households that did not get the resources, recipients were making more money from farming. They were less likely to be hungry or stressed, worked harder, and felt that they had higher status in their communities. Similar results have been found in a Ugandan program that gave cash to ultrapoor women. The impacts are not as transformative as in the Nigerian YouWiN! experiment, but appear promising nonetheless.
These results have led a group of US philanthropists to create a charity called GiveDirectly.18 As the name suggests, the organisation has one simple mission: to enable donors to give cash to the extreme poor. GiveDirectly is about to embark on a randomised experiment in Africa, in which it will give at least 6000 villagers a dollar each day, guaranteed for a decade.19 If successful, this could radically reshape how we think about aid.
For good reason, many programs to help businesses are focused on money. But we shouldn’t underrate the value of motivation. In a recent randomised trial, villagers in rural Ethiopia were shown four videos over the course of an hour.20 Each depicted people in poverty who managed to better themselves through hard work, goal-setting and careful decision-making. The villagers who watched the videos were extremely poor – with most making less than a dollar a day. They faced a raft of challenges, including illiteracy, corruption, sickness and bad roads.
One video told the story of Teyiba Abdella, beginning with glowing accounts from her neighbours.21 Abdella used to carry big loads even when she was pregnant, one neighbour says. But her life is getting better and better, another tells us. And then Abdella herself comes on the screen: a tall Ethiopian villager. We hear about how she used to carry 50 kilograms of flour on her back for three hours to walk to the market. Before she could afford a donkey, Abdella tells us, ‘I used to work like a donkey.’ Now, as well as flour, she sells eggs and chickens. The ten-minute clip finishes with her husband saying how proud he is of Abdella, a ‘role model to her fellow villagers’.
Could hearing stories like Abdella’s motivate someone to change their life? When the researchers followed up six months later, they found that just that one hour of watching motivational videos had made a lasting difference. Those who watched the videos had higher aspirations and greater savings rates, and were more likely to enrol their children in school. Summing up their findings, the research team admitted: ‘The extent and nature of their response has surprised us.’ Yet the result doesn’t seem to be a one-off. On the other side of the continent, a randomised trial in Togo found that personal initiative training for entrepreneurs boosted profits by almost one-third (a significantly larger impact than traditional business training).22
*
Poor countries are often poorly governed. The countries with the worst infrastructure are Congo, Chad and Yemen.23 The most corrupt nations are Somalia, South Sudan and North Korea.24 The most fragile states are the Central African Republic, Congo and South Sudan.25 Lacking money and well-trained public servants, it’s hardly surprising that governments in developing countries don’t operate with Swiss efficiency. But perhaps randomised trials can help these governments decide how to allocate their scarce resources.
In China, researchers have used randomised experiments to learn about how government responds. In an experiment across more than 2000 counties, a request for help in obtaining dibao, the basic social payment for those unable to find work, was posted on government web forums.26 In some counties, researchers simply posted a deferential request for help. In other counties, the request went further, adding either a threat of collective action (‘If you can’t help, we’ll try to figure out what we can do together about this situation’) or a threat of tattling (‘If this problem cannot be addressed, I’ll have to report it to upper-level government officials’). Both threats increased by about one-third the chances that local Chinese officials provided some kind of response.
Just as threats provoke better behaviour, bribes can lead to worse behaviour. A novel Indian experiment looked at what happened when a randomly selected group of young men were offered a cash incentive to obtain their driving licences quickly.27 To get the money, many in the study hired ‘agents’, who paid bribes to the testing authority. Shortly afterwards, the researchers gave everyone in the study a surprise driving test. Their focus was on how many people were incompetent behind the wheel, and yet had a licence. It turned out that those offered the cash incentive were 18 percentage points more likely to have their licence and yet be unable to drive.
My initial reaction to the licence study was to think that it was unethical, on the basis that the researchers were putting dodgy drivers onto the streets of Delhi. When I chatted about the findings with study author Sendhil Mullainathan, he acknowledged that I wasn’t the first to feel uneasy. But since everyone in the study got free driving lessons afterwards, he pointed out, ‘every single participant – treatment or control – was a better driver as a result of being in the experiment’. Multiple ethics boards approved Mullainathan’s research, because they understood the value of rigorously studying corruption in India.
Licensing corruption thrives because going down the honest route to get an Indian driving licence is a Kafkaesque nightmare. Without the help of an agent, the typical applicant makes three trips to the regional transport office and speaks with eight bureaucrats. If you don’t pay a bribe, getting an Indian driving licence typically takes more than five hours. When an experiment is the best way to shed light on a serious problem, we have to weigh the risk of doing a randomised trial against the risk of continued ignorance. I will return to these ethical issues in more detail in Chapter 11.
Other randomised trials in development work closely with governments. When Fabio
la Vázquez Saút, mayor of the Mexican city of Acayucan, found that council only had money to pave about half the streets, she saw an opportunity to avert some voter anger, and learn about the impacts of road paving. Rather than selecting the roads herself, she let researchers from Oxford and the University of Toronto randomly choose which streets to upgrade. The Acayucan experiment created something unusual – the world’s first study of the true impact of public road building on property values.28 Comparing homes on paved streets with those on unpaved streets, researchers found a considerable increase in value. Indeed, the increase in home values was as large as the cost of the street works. Households responded to their newfound wealth by buying home appliances and motor vehicles, contributing to a fall in material poverty.
In Kenya, economists worked with the national electricity utility to randomly give some households a discount on their connection fee.29 By varying the subsidy, the researchers were able to see how much households valued being connected to the grid. Unlike the Mexicans and their road experiment, Kenyan households seemed to place a much lower value on an electricity connection than it cost to provide it. Much as we might like to think about turning the lights on for households that now make do with kerosene lamps, it may be that scarce development dollars are better directed towards roads, schools or health clinics than to electrification in remote areas.
One researcher who has been at the vanguard of conducting experiments with governments in low-income countries is Karthik Muralidharan, an economics professor at the University of California, San Diego. Energetic, fast-talking and engaging, Muralidharan has persuaded several governments to conduct randomised trials at a scale that would only have been dreamt about a generation ago. When he heard that the Indonesian government planned to spend US$5 billion to double teacher pay, Muralidharan and his co-authors persuaded government officials – working with the World Bank – to agree to implement the policy early among a randomly selected group of a few thousand teachers.30 The study concluded that those teachers who had their pay doubled were happier, but there was no discernible impact on student outcomes.