Sebastian Roing

Out of the box and into this world: behavioural economics and international development

Studies show that people often make bad decisions. Annoyingly bad decisions, for annoyingly petty reasons. Like that time you procrastinated getting back to a friend, for no obvious reason, and it turned out you missed out on a nice dinner, an evening at the pub or some such. In fact, you probably didn’t need studies to inform you that you sometimes make bad decisions. You knew that perfectly well already.

But there is an incredible value to knowing people, in general, are not rational decision-makers. And there is an even higher value to knowing what specific biases influence their decisions and what lapses in judgement people are prone to commit. This value consists in being able to design policy interventions that 

, flagship publication of the World Bank. In this blog article I will outline the thinking behind the WDR and highlight a few promising opportunities for low-cost interventions.

 looks at how small changes in the costs of applying to college affected application rates for students from low-income households. The small changes in costs originated from score reports of standardized tests, such as the ACT, that have to be sent by the test administrator directly to American colleges. In this case, the number of score reports that were sent free of charge increased from 3 to 4. Sending additional reports incurred a small cost of $6. What happened was that the mean number of score reports sent increased from almost exactly 3 to almost exactly 4. The effect was the same across low-income and high-income households.

Why is this strange? Well, it suggests that the students didn’t weigh the costs of sending an additional score report against the potential benefits, they simply went with the default option. Unthinking acceptance of default options seems to be pervasive in human societies, and in some cases it may have huge consequences. The students from low-income households in the example above were foregoing, on average, an additional $10,200 of lifetime earnings by not sending an additional $6 score report.

Automatic acceptance of default options is only one of many cognitive biases that people are subject to. Others include anchoring (an excessive focus on a single piece of evidence, usually the first one presented, when making a decision) and confirmation bias (the tendency to search for, interpret and focus on information in a way as to confirm your preconceived ideas).

These biases, recurring systematically and pervasively, are not in accordance with the underlying assumption of rational, utility-maximizing agents of economic models. 

 do what they would really want to do. They fall in the traps of automatic thinking and pre-defined mental models.

This implies that a heavy usage of economic models with rational agents can lead to a situation in which we ignore 

incredible opportunities to improve people’s lives through interventions that recognize their cognitive biases.

 What’s more, this is particularly relevant for development policy, as 

poor countries lack many of the institutions that make good decisions default options in the West

Figure 1: Good luck maximizing utility, baby.

The World Development Report 2015 draws heavily on the work of behavioural economists Daniel Kahneman and Amos Tversky, and for an overview of their research I would really recommend 

Three principles of human decision-making

In the World Development Report 2015, the World Bank sets up a framework where human decision-making is characterized by three principles.

 Lots of decisions we just make automatically, without much effort or consideration. Accepting the default option is one example of automatic thinking. Hence, 

minor changes to the decision-making environment can have disproportionate effects on outcomes and aggregate well-being.

 Such changes may relate to the timing of the decision or the design of the documents relevant for the decision.

One example comes from Mexico City, where people from low-income households were invited to pick the best $800, one-year loan from a randomized list of loan products representative of the local market. A mere 39% could identify the lowest-cost product from the brochures distributed by the banks themselves. However, 68% could do so with the help of a consumer-friendly leaflet designed by the Consumer Financial Credit Bureau. Making it easier for people to select what they want improves aggregate well-being.

Another example comes from India, where a study by Mani et. Al. (2013)

 showed how financial distress can deplete mental resources needed for important decisions. In the part of India studied, 99% of sugar cane farmers were indebted just before harvest, as opposed to 13% just after the harvest when farmers had received most of their income for the season. The financial distress had a considerable impact on farmers’ mental resources. In a series of cognitive tests, farmers performed consistently worse before harvest relative to after harvest, controlling for other factors such as sleep, work hours, nutrition etc. The mental depletion corresponded, on average, to the lack of a full night’s sleep.

With this in mind, it is perhaps less surprising that another study (Duflo, Kremer & Robinson, 2011)

 found that farmers on average make significantly higher-return investments in fertilizer straight after harvest relative to later in the season. Policy makers should therefore take real care in the timing of important decisions for the poor. Information about better seeds and crops should be distributed 

 harvest, not before, and farmers should be able to decide on theirs children’s schooling 

 Humans are social animals and have innate preferences for cooperation and altruism. We are strongly affected by the social norms in our communities. This is why prevalent corruption can largely be considered a social phenomenon – if you think everyone else will accept a bribe in your position, there is little stopping you from doing so yourself. Opportunities to change social norms exist at large.

An illustrative example of this is included in 

 by Abhijit V. Banerjee and Esther Duflo. Here we are told of school funds in Uganda, and that only 13% of the intended funds ever reached the right schools in the late 90s. After this had been publicized, there was an uproar, and the Ministry of Finance started providing the main national newspapers with statistics on how much money had been sent to local schools. The hypothesis was that the money got stuck in the pockets of local officials. 5 years after, more than 80% of the school funds reached their intended destinations.

 We use a variety of mental models when we think, passed down to us by history and culture. They are not necessarily social norms in the sense that they apply pressure on us to act in a specific way, but they are rather ways of thinking that are easily accessible. Thinking that people are rational is in some sense a mental model, particularly common amongst economists.

An example of how changes to institutions can help change mental models comes from the West Bengal in India. Here, affirmative action led to some regions having women leaders for the first time ever. Only seven years after the leaders had been elected, aspirations for teenage women were much higher. And perhaps more importantly, parents’ aspirations for their teenage daughters were higher and men’s bias in evaluating women in leadership positions was reduced (Beaman et Al, 2012)

Low-cost interventions that change behavior systematically look really promising from a development perspective. Particularly since many of the results in the World Development Report and those mentioned throughout this text have been asserted using robust statistical techniques, primarily randomized control trials. 

Improving decision-making environments and making good choices default options can and should improve well-being in the future. Finding such opportunities should be a clear focus for development professionals.

As a last note, we should always be aware of our own mental models and cognitive biases when thinking about development and policy interventions. And we should make sure that personal liberty is not unjustly compromised in the process of development. 

, by recognizing cognitive biases, shouldn’t be considered an infringement on personal liberty. In the words of John Stuart Mill:

[It] is a proper office of public authority to guard against accidents. If either a public officer or anyone else saw a person attempting to cross a bridge which has been ascertained to be unsafe, and there were no time to warn him of his danger, they might seize him and turn him back, without any real infringement of his liberty; for liberty consists in doing what one desires, and he does not desire to fall into the river. (Mill, 1859)

What values do we need to keep in mind about the DALY? Part I

Should we value the life of a 25 year-old more than that of a 65 year-old simply because they have different social roles? Should we discount future health benefits because it is uncertain whether they will arise or not? Should we think differently about the individual burden of a disease when a whole community is affected instead of just one individual?

This post gives a brief history of the health statistic known as the DALY – Disability Adjusted Life Year. The DALY has come to influence many decisions in global health over the last 20 years. I raise a few problems with the statistic,and emphasise its benefits as a statistic that is comparable across causes.

The DALY is an attempt to merge data on morbidity and mortality into one statistic, and make it comparable across different diseases and causes. It assigns a numerical value between 0 and 1 to health outcomes, where death, the worst possible case, is assigned the value 1. Thus, it allows for simple addition of different health outcomes. This enables agencies such as the WHO to calculate the disease burden of a particular region, accounting for all the different diseases prevalent in that region.

Moreover, the DALY can also be used to guide resource allocation, by allocating resources to the intervention that saves the most DALYs per dollar. So how do we construct the DALY and put a numerical value on health outcomes? Well, at face value it’s a simple formula:

**DALY = YLL (Years of life lost) + YLD (Years of life lost due to disability)**

Suppose a young adult of 25 years is living in a remote area where there is little access to health care. Tragically, this individual has contracted AIDS, and receives no antiretroviral treatment. Nonetheless, the 25-year-old survives for a year, but then dies of an infection.

This scenario can be converted into DALYs. First, we need to account for the year lived with AIDS without receiving antiretroviral treatment. The 

 on this health outcome is 0.547. These weights are numbers between 0 and 1 and they signify how bad a particular health outcome is compared to death. 0 represents a state of full health, whereas 1 represents death. The figure below lists a few common causes and their disability weights. The full list of disability weights is available 

To get the YLD for the 25-year-old we therefore multiply the number of years lived with AIDS by the disability weights (1 * 0.547 = 0.547). The other component, the YLL, is simply the years of life lost relative to the agreed standard life expectancy (87.8 – 26 = 61.8). The DALYs lost and the total burden of the disease for the 25-year-old is 62.347.

**DALY = YLL (standard life expectancy – age at death) + YLD (years lived with disability * disability weight)**

But wait. Aren’t there at a hundred ethical problems with this method? Well, there are a few. I will describe the four issues I think are most important in this post. Two of these issues have been resolved in the latest update of the Global Burden of Disease study from 2010.

History of the DALY and two (resolved) issues

The original DALY was developed at the Harvard School of Public Health in the late 1980’s. Other measures, such as YLL (Years of Life Lost) and QALY (Quality Adjusted Life Years) were already popular, but they offered no standardised way of measuring morbidity and mortality.

The DALY was subsequently adopted by the Global Burden of Disease 1990 study of the WHO (GBD 1990); a huge project which estimated the total, global disease burden by using DALY methods. The GBD series was later updated in 2004 and most recently in 2010. Over these 25 years, the DALY statistic changed significantly following a lively debate about, chiefly, two ethical issues inherent in the DALY; age-weighting and time discounting.

Reasonably, we might ask ourselves – Is it worse if a 25-year-old suffers from AIDS than if a 75-year-old suffers from AIDS? What about infants? Is the death of an infant as bad as the death of a 25-year-old? The original DALY envisaged the weighting of age according to social roles played by people of different ages. C. J. Murray explains the technical basis of the DALY in his 1994 article 

Quantifying the burden of disease: the technical basis for disability-adjusted life years

. The weighting of age in the original DALY followed the function in the figure below.

The weighting peaks at around 25 years, and decreases as age increases. Furthermore, the weighting starts at 0 at birth and increases quickly until the age of 25. This implies that the death of an infant, at birth, is weighted as 0 DALYs (nothing). So the death of an infant is definitely not regarded as bad as the death of a 25-year-old. And the weighting of age is linear. This implies that, in terms of DALYs, one 25-year-old suffering from AIDS for one year is approximately as bad as two 75-year-olds suffering from AIDS for one year.

Murray did not justify age-weighting by some differential intrinsic valuation of life. Rather, age-weighting was an attempt to capture different social roles at different ages. The young and the elderly depend on young adults for care, emotional and financial support. Sickness and death among this group of individuals is therefore particularly serious.

But if we open up for the possibility of weighting according to social roles, shouldn’t we also weight individuals across professions and income? What about doctors and nurses? Since people depend on them for care shouldn’t they also be given a higher weight? It turns out this kind of weighting is quite problematic. By weighting people according to their age we open up many other issues that would have to be considered before we could justify the application of age-weights. Furthermore, numerous critics voiced concerns about the universalism of a human life. The practise of age-weighting was consequently 

A related issue is time discounting, a constantly relevant problem in cost-benefit analysis. Toby Ord and Robert Wiblin have written extensively on this issue in our section on 

. Discounting is the practise of applying higher weights to benefits that arise in the present relative to the future. For example, most people would prefer getting a piece of chocolate today than tomorrow (see, for example, 

 experiment on present-bias). And I wouldn’t exchange $100 this year for $100 next year, chiefly because I could take the $100 today and invest them such that my return would give me more than $100 next year. We usually model this phenomenon using geometric discounting.

However, the practise of discounting is not as intuitive when it comes to health benefits. Is it better to cure an individual of tuberculosis today rather than next year? Not really. But could this cured individual contribute to society in such a way that we are able to cure more people of tuberculosis next year? Well, maybe, but Ord and Wiblin suspect that most benefits 

Time discounting was originally used in the DALY statistic for technical reasons. Suppose we let DALY estimates guide our resource allocation. What would be the benefits of investing in a research program that produces a vaccine for HIV? If the time horizon is infinite, then the benefits would be infinite, because we would prevent an infinite number of people from contracting HIV. Consequently, if we don’t discount benefits in the future, all resources should be invested in vaccination programs.

What is the problem with this reasoning? As pointed out by Ord and Wiblin, it is not appropriate to assume that no one would ever think of a vaccine against HIV if it weren’t done by one research program in particular. It is not the appropriate counterfactual. It is better to construct an explicit model that assumes that someone else would think of a vaccine in x years. Furthermore, discounting the DALY creates a lot of intergenerational inequality that is quite arbitrary. Originally, the DALY used a discount rate of 3%. This gives a weight of approximately 0.22 to DALYs that arise in 50 years. Is it fair to consider 5 deaths in 50 years as serious as 1 death today?

The most serious implication of age-weighting and time discounting, as noted by 

 was that no one could get rid of these practises without re-doing the entire analysis. Once a disease burden had been calculated for a particular setting it was more or less impossible to exclude time discounting and age-weighting without going back to the original data. This was a serious obstacle to transparency and transferability of results in the global health debate. I join the crowd of positive supporters of the WHO decision to ignore time discounting and age-weighting in the latest Global Burden of Disease study.

In the next post in this series, I will look at the DALY and its role as a measure of disease burden that is transferable across different conditions and illnesses. I will also discuss the future of the DALY, its limitations as an indicator of resource allocation, and an additional method that could include community effects into the DALY.

Sebastian read Economics at King's College, Cambridge. He enjoys kitesurfing and books in his spare time.

Sebastian Roing

About

Articles

Out of the box and into this world: behavioural economics and international development

What values do we need to keep in mind about the DALY? Part I