Doing Good Better: Effective Altruism and a Radical New Way to Make a Difference

by William MacAskill

The law of diminishing returns also explains why, in general, it makes less sense to donate to disaster relief than it does to donate to the best charities that fight poverty. Every day people die from easily preventable diseases like AIDS, malaria, or tuberculosis. This is a disaster far beyond that of Haiti, or Tōhoku, or Sichuan. Every day over 18,000 children – more than the number of people who perished in Tōhoku – die from preventable causes.

Every year cancer kills 8.2 million people and is responsible for 7.6% of all deaths and ill health worldwide (measured in terms of QALYs lost). $217 billion per year is spent on cancer treatment. Malaria is responsible for 3.3% of QALYs lost worldwide. In terms of its health impacts, cancer is about twice as bad as malaria, so if medical spending were in proportion to the scale of the problem, we would expect malaria treatment to receive about $100 billion per year. In reality only $1.6 billion per year is spent on malaria treatment: about sixty times less than we would expect.

Again, we see The 100x Multiplier at work. We’re about one hundred times richer than the poorest billion people in the world, and we can do several hundred times more to help them than we can to help others in the rich countries we live in.

Key question 4: What would have happened otherwise?

Instead, we should look to a much more unlikely hero: Viktor Zhdanov, a Ukrainian virologist who died in 1987. At the time of this writing, he has a mere four-paragraph Wikipedia page, and there are only a few grainy black-and-white photos of him available online. I’m not aware of any major accolades for his work.

Of course I’m not a hero. The good I do is not a matter of the direct benefits I cause. Rather, it is the difference I make.

Looking at what would have happened otherwise is a fundamental piece of scientific reasoning, referred to as ‘assessing the counterfactual’. But the mistake of neglecting the counterfactual is rife within the world of altruism, and this mistake can have terrible consequences.

The goal was to make him reconsider his life choices before he ended up in jail for good. Brandon is the central character in an episode of Beyond Scared Straight.

Arnold Shapiro.

Nine high-quality studies have been done on the programme, assessing the progress of one thousand juveniles overall. The Cochrane Collaboration, a non-profit institute that rigorously assesses the evidence behind health and social programmes, looked at these studies and found that two of them had no significant effect, while the remaining seven showed increased rates of criminality among juveniles. The authors of the review estimated that the Scared Straight programmes that had been studied increased the odds of offending by about 60%.

The researchers concluded that, because Scared Straight was increasing rates of crime, with associated penitentiary costs and costs to the local community, every $1 spent on Scared Straight cost society $203.

I suspect the apparent effectiveness of Scared Straight can be explained by a phenomenon called regression to the mean.

But that only explains why Scared Straight can appear to be effective when it really does nothing. Why, then, does the programme increase rates of criminality? No one knows for sure, but one hypothesis is that the inmates – who play up how tough they are for surviving life in prison – act as role models rather than deterrents for the delinquents.

Earning to give means exactly what it sounds like: rather than trying to maximise the good you do directly via your job, you instead try to increase your earnings so you can donate more, improving people’s lives through your giving rather than your day-to-day work.

Most people don’t consider this option when choosing a career that ‘makes a difference’. But time and money are usually interchangeable – money can pay for people’s time, and your time can be used to earn money – so there’s no reason to assume the best careers are only those that benefit people directly through the work itself. If we’re serious about doing good, earning to give is a path we should consider.

There is plenty more to be said about doing good through your career, and the whole of Chapter 9 will be devoted to it. But before we can address this question properly, and before we can see why earning to give is merely one path, and not always the most effective career choice, we need to look at the fifth and final key question of effective altruism.

Key question 5: What are the chances of success, and how good would success be?

So far in this book we have focused on measurable, concrete ways of helping others. The real world, unfortunately, is not always so simple. Often we don’t know whether our actions will be successful, and, given the difficulty of knowing what would have happened otherwise, we will usually never know whether our actions really make a difference.

Within economics and decision theory the standard way to do this is to look at an action’s expected value.

To calculate the expected monetary value of each bet, you look at all the possible outcomes of that bet. For each outcome, you take the monetary gain or loss and multiply it by the probability of the outcome. In this case, there are two possible outcomes, Heads and Tails. Each has a 50% chance of occurring. The expected monetary value of taking the bet is therefore (50% × +$2) + (50% × −$1) = $0.50. The expected monetary value of refusing the bet is $0. Taking the bet has the higher expected value, so you should take the bet.

Maximising expected value is generally regarded as the best strategy for making decisions when you know the value and the probability of each option. It’s the strategy used by economists, statisticians, poker players, risk management experts and pretty much anyone else who regularly needs to deal with uncertain outcomes.

That’s rational because money has diminishing returns: if you’re like most people, the improvements you’d make to your life by spending the first $1,000 of your $100,001 lottery winnings will bring you more joy than the improvements you make to your life by spending the last $1,000. The same isn’t true, however, when we’re thinking about philanthropy.

To see how useful the idea of expected value is, let’s consider a morbid but important application: assessing the risk of death from different activities. Smoking, riding a motorbike, scuba diving, taking ecstasy, eating peanut butter: all these things increase your risk of death. How much should you worry about each of them? Public health experts use the concept of a ‘micromort’ to compare the risks, where one micromort equals a one-in-a-million chance of dying, equivalent to thirty minutes of expected life lost if you’re aged twenty, or fifteen minutes of expected life lost if you’re aged fifty.

By comparing different activities in terms of micromorts, we can easily assess their relative dangers. The results can be surprising.

When thinking about risk from transport, you can think directly in terms of minutes of life lost per hour of travel. Each time you travel, you face a slight risk of getting into a fatal accident, but the chance of getting into a fatal accident varies dramatically depending on the mode of transport. For example, the risk of a fatal car crash while driving for an hour is about 1 in 10 million (so 0.1 micromorts). For a twenty-year-old, that’s a 1 in 10 million chance of losing sixty years. The expected life lost from driving for one hour is therefore three minutes. Looking at expected minutes lost shows just how great a discrepancy there is between risks from different sorts of transport. Whereas an hour on a train costs you only twenty expected seconds of life, an hour on a motorbike costs you an expected three hours and forty-five minutes.

The Fukushima safety engineers were trying to prevent harm with their safety assessment, and they failed by ignoring an important but low-probability event. In just the same way, when trying to do good, we need to be sensitive both to the likelihood of success and to the value of that success. This means that low-probability, high-payoff activities can take priority over sure bets of more modest impact. It also shows that people are often confused when they say that ‘one person can’t make a difference’. Voting provides a vivid illustration.

Given the concept of expected value, however, Levitt’s reasoning is too quick. We can’t just say that the chance of affecting the outcome by voting is so small as to be negligible. We need to work out how large the benefit would be if we did indeed affect the outcome.

That’s the sense in which voting is like donating thousands of dollars to (developed-world) charities.

We used the idea of expected value to show why voting for the better party is often a high (expected) impact altruistic activity.

On many issues, I find that people hold the following two views: If many people did this thing, then change would happen. But any individual person doesn’t make a difference. Holding that combination of views is usually a mistake when we consider expected value.

This isn’t just a theoretical argument. Economists have studied this issue and worked out how, on average, a consumer affects total sales by declining to buy one unit of a product. They estimate that, on average, if you give up one egg, total production ultimately falls by 0.91 eggs; if you give up one gallon of milk, total production falls by 0.56 gallons. Other products are somewhere in between: if you give up one pound of beef, beef production falls by 0.68lb; a pound of pork: 0.74lb; a pound of chicken: 0.76lb.

The chance of you being the person who makes the difference is very small, but if you do make the difference, it will be very large indeed.

Again, this isn’t just a speculative model. Professors of political science at Harvard and Stockholm Universities analysed Tea Party rallies held on Tax Day, 15 April 2009.

They found that policy was significantly influenced by those rallies that attracted more people, and that the larger the rally, the greater the degree to which those protestors’ representatives in Congress voted conservatively.

For our purposes, the most important use of expected value reasoning is in comparing concrete, measurable ways of doing good with more speculative but potentially higher-payoff strategies. One example of this concerns how to compare different careers. Whereas earning to give in order to donate to charities like the Deworm the World Initiative provide a reliable way of doing good, others, like politics, are much less certain. How should we compare these?

‘I expect to fail,’ Laura Brown told me while putting down her cup of coffee. We were sitting at the Grand Café in Oxford, the oldest coffee house in the UK, and I was there to discuss Laura’s career plans. A second-year student of Philosophy, Politics, and Economics (PPE), Laura had recently read a newspaper article discussing the chances of a PPE student being elected to Parliament. Intrigued, she traced the original source, finding work by my organisation 80,000 Hours, and studied the calculations carefully. The research persuaded her to pursue a career in politics. ‘I’m most likely going to fail to become a high-flying politician. But I could do so much good if I did succeed that I think it’s worth taking the chance,’ Laura told me.

The most naïve estimate of the odds of becoming an MP would be to work out how many people alive in the UK today will ever serve as an MP (which we estimated at about 3,100, five of whom would become prime minister), and divide by the total population of the UK (64 million), giving a 1 in 20,000 chance of becoming an MP, and a 1 in 12 million chance of becoming prime minister.

However, most people don’t try to enter politics, and the representation of politicians in the UK is highly skewed towards people of certain backgrounds. In particular, graduates of Oxford are dramatically over-represented within UK politics, with graduates of PPE especially over-represented. For example, both David Cameron and Ed Miliband earned a degree in PPE from Oxford. Of the 650 members of Parliament, over a hundred studied at Oxford (despite there being only 3,000 graduates every year), of which thirty-five studied PPE (despite there being only 200 PPE graduates each year). Of Cabinet ministers (those MPs who hold executive power) 32% studied PPE at Oxford, and of the thirteen prime ministers since 1945, nine studied at Oxford and three of those studied PPE.

We worked out that a graduate in PPE from Oxford who chooses to enter politics had a historical chance of 1 in 30 of becoming an MP and 1 in 3,000 of becoming prime minister. Laura’s background gave her remarkably good odds of entering Parliament, but, it was nonetheless still far more likely that she would fail, so entering politics as a means of doing good was still a high-risk stratagem.

The conclusion that earning to give isn’t always the highest-impact career path isn’t a conclusion that’s unique to Oxford PPE graduates. In Chapter 9 I’ll discuss several other career paths, such as research and entrepreneurship, that have a low probability of success but a very high upside when success occurs and that seem highly competitive with earning to give.

In 1790 Benjamin Franklin, for example, wrote a letter petitioning US Congress to end slavery.

However, this debate is strange for another reason: even if scientists had not already shown that man-made climate change is happening, the mere fact that man-made climate change might be happening is enough to warrant action.

Our predicament with climate change is no different. If climate change is happening and we don’t take action, millions of lives will be lost and the world economy will lose trillions of dollars. If climate change isn’t happening and we do take action, the costs are much lower. We would have wasted some amount of resources developing low-carbon technology and slowed economic progress a bit, but it wouldn’t, literally, be the end of the world.

The second way in which expected value is relevant when figuring out how to respond to the threat of climate change is that it shows why individuals have reason to mitigate climate change just as much as governments do. Over your lifetime, your individual greenhouse gas contribution will only increase the temperature of the planet by about half a billionth of a degree Celsius. That, you might think, is such a small difference as to be negligible, so you shouldn’t bother trying to reduce your personal emissions.

Finally, and most importantly, expected value is important when assessing just how bad climate change is, and how much of a change we should make. When I first looked into economic assessments of the damages of climate change, I was surprised to find that economists tended to assess climate change as being not all that bad. Most estimate that climate change will cost only around 2% of global GDP. This, of course, is huge when measured in the trillions of dollars lost. But it’s not that large when compared to somewhat typical rates of slow economic growth. Economic growth per person over the past decade has been about 2% per year; so a loss of 2% of global GDP due to climate change is therefore equivalent to going one year without economic growth. The thought that climate change would do the equivalent of putting us back one year economically isn’t all that scary – 2013 didn’t seem that much worse than 2014.

However, this standard economic analysis fails to faithfully use expected value reasoning. The standard analysis looks only at the effects from the most likely scenario: a 2–4ºC rise in temperature. It doesn’t consider what the consequences would be if our best-guess estimates are wrong. This is especially important because the climate is an incredibly complex system that is difficult to predict so we can’t be sure that our estimates are correct. When climate scientists make estimates about temperature rise, they have to acknowledge that there is a small but significant risk of a temperature increase that’s much greater than 2–4ºC. The Intergovernmental Panel on Climate Change gives more than 5% probability to temperature rises greater than 6ºC, and even acknowledges a small risk of catastrophic climate change, of 10ºC or more. To be clear, I’m not saying that this is at all likely, in fact, it’s very unlikely.

But it is possible, and if it were to happen, the consequences would be disastrous, potentially resulting in civilisational collapse.

In cases where people seem to neglect the risks of worst-case outcomes, helping to prevent these outcomes might be a particularly effective altruistic activity. This is what the Skoll Global Threats Fund focuses on, trying to reduce the chances of global catastrophes arising from climate change, pandemics, and the proliferation of nuclear weapons. The charity evaluator GiveWell is currently investigating these sorts of activities in an attempt to work out how effective donations in these areas can be.

In this book so far, I’ve introduced the key questions to help you think like an effective altruist: How many people benefit, and by how much? Is this the most effective thing you can do? Is this area neglected? What would have happened otherwise? What are the chances of success, and how good would success be? Now it’s time to examine how we apply those questions to the real world and put effective altruism into action. That is the subject of Part II.

Using what is called the M-Pesa system, cell phones are used as makeshift bank accounts, thereby enabling an easy transfer of money from foreign bank accounts to the poor. GiveDirectly uses satellite images to find households with thatched roofs (a strong indicator of poverty, compared to iron roofs) and then contacts those households to discuss the programme. If the household is willing, GiveDirectly transfers them a lump sum of $1,000, which is equal to a little more than one year’s total income for that household. In their words: Recipients use transfers for whatever is most important to them; we never tell them what to do. An independent evaluation of our work in Kenya by Innovations for Poverty Action found that recipients use transfers for a wide variety of purposes that on average generate large income gains. Common uses range from buying food to investing in tangible assets such as housing and livestock to investing in children’s education. Of these three charities, which do you think can do the most good with your $100 donation? In this chapter I’ll provide a framework that will help you determine the answer.