Pandemic policy-making involves balancing reductions in coronavirus cases against other outcomes like mental health, education and the economy. People care about effects on different parts of society, even when adopting more equitable policies may mean saving fewer lives overall.
Policy decisions during a pandemic involve difficult trade-offs. Should younger people have to sacrifice their freedoms, and sometimes their livelihoods, to protect vulnerable elderly people from harm? Should schools close (as in the UK’s spring and winter lockdowns) or stay open (as in the autumn lockdown)? In the extreme, if NHS capacity limits are exceeded, medical professionals may face unprecedented choices about whom to make priorities for treatment and care. Decisions like these involve trade-offs between the wellbeing of different groups in society – and between different policy goals.
To inform these decisions, cost-benefit analysis requires estimates of the consequences of different potential policies. Metrics including the value of a statistical life (VSL), the value of a life year (VOLY) and the value of quality-adjusted life years (QALYs) put monetary values on health and safety consequences in different ways (as explained on the Economics Observatory by Coast and Sanghera, 2020).
The methods differ in their conceptual frameworks and in the empirical approaches for their estimation, yet all of them are blind to the distributional consequences of policies. Here, we examine how the distribution of safety should be considered in policy-making.
What is the role of altruism?
The VSL and VOLY metrics rest on a number of assumptions. Importantly, they assume that the appropriate perspective for the valuation of safety is self-interest, with people valuing reductions in their own risks but not those of others. The metrics also assume that the total amount of safety (that is, the total expected lives saved or life years gained) should be the policy objective for these measures.
Contrary to these assumptions, people care about others and value improvements to others’ wellbeing (Simon, 1993). For example, people give to charity and donate their time to good causes.
To develop a better understanding of the underlying preferences driving this behaviour, some economists conduct experiments that involve people making choices under controlled conditions. In some experiments, participants are given a sum of money and asked how they would like to split it between themselves and an anonymous stranger. Participants consistently give away some of their money to others, even when their actions are entirely anonymous and are therefore neither strategically useful nor helpful in improving the way they are viewed by others (Camerer, 2003).
In the context of the pandemic, people have shown remarkable willingness to incur personal costs by abiding by lockdown rules to protect others from harm as well as themselves. Recognising altruism, and ideally quantifying its extent, can help to predict how effective policies will be when they rely on people’s self-sacrifice.
Moreover, if altruism towards others is focused on safety, which implies that people’s wellbeing is enhanced specifically by knowing that others are being made safer, then estimates of VSL, VOLY and QALYs may be underestimated (Jones-Lee, 1992).
Whose safety matters most?
Covid-19 – and the methods used to combat it – affects groups in society differently. The elderly and those with pre-existing health problems are the most vulnerable to the virus and so they benefit the most from more restrictive policies. Younger people and those without health problems not only benefit less from these policies, but also tend to be harmed the most by them.
For example, children and their parents suffered when schools closed in the first and third lockdowns, and working age people in the hospitality sector face uncertainty and potential harm to their livelihoods as their industry is restricted by lockdowns and the tier system used to fight the subsequent waves.
When considering the distribution of scarce vaccination and testing capacity, we assign priority to NHS staff and other key workers. Our success in weathering the pandemic relies on their ability to work, so it is an efficient use of resources to protect them. In addition, individuals may prefer that society’s resources are directed towards those helping in the fight against the pandemic (see Persad et al, 2009 for evidence along these lines).
In this instance, efficiency and the preferences of society as a whole work in the same direction, and so making key workers the priority is uncontroversial. But what happens when this is not the case?
What do we learn from VSL, VOLY and QALYs?
Difficult choices must be made when public preferences are unclear – or when the option that best respects public preferences conflicts with the option that is most efficient in saving lives.
In aggregate, older people have fewer years of life remaining than younger people. Under the VOLY approach, policies that reduce the risk of dying for the young are more valuable than those that reduce the risk of dying for the elderly since the young have more years left to live. But under the VSL approach, all lives are valued equally, regardless of age.
Similarly, those who have pre-existing health conditions have fewer QALYs at stake. Under the QALY approach, policy-makers should give priority to saving the lives of those who are healthy, but the VOLY approach makes no such distinction. When different approaches fail to give a clear answer to these distributional dilemmas, where can policy-makers turn?
The answer may lie in measuring how people would prefer to trade-off different types of risks to different population groups using tailored experiments and surveys. The standard VSL, VOLY and QALY approaches all rest on the assumption that individuals’ preferences – or more specifically, their trade-offs between wealth and fatality risk – are the appropriate basis for policy-making. As such, it seems reasonable to suppose that their trade-offs between risks of different kinds and to different population groups is also a legitimate basis for policy-making. Currently though, this is not put into practice.
What tools does economics offer?
While the widely used tools of cost-benefit analysis ignore distributional concerns and measure the wealth-risk trade-offs from an explicitly self-interested perspective, the wider health and welfare economics toolkit has well-developed techniques for addressing these concerns empirically.
In the ‘person-trade-off’ approach (Nord, 1995), participants in research experiments are presented with scenarios describing groups of people at risk of harm. Those groups differ in some observable way, and the number of people in each group is also different. By eliciting the number of people in one group that is equivalent in harm to a different number in the other group, researchers are able to reveal the relative weight that the participants place on the health, safety or wellbeing of people in the two groups.
This method has been used in contexts such as age – where respondents are asked to prioritise saving more older people or fewer younger people – and health status – where they are asked to prioritise saving more people with a given health condition or fewer people with a different health condition. Analysing these priorities reveals the trade-offs that people are willing to make along these dimensions.
Person-trade-off studies are powerful and simple, but suffer from potential biases in the results as participants might give answers that they think the researchers want to hear or which show them in a positive light.
The discrete choice experiment approach is more subtle, masking the trade-off of interest by varying the groups in more ways than one. It is underpinned by an economic theory of decision-making called random utility theory (McFadden 1973). Statistical analysis of participants’ choices reveals their implicit rates of trade-off between different characteristics of the groups.
Seeking to understand how different characteristics affect the way that people around the world would prioritise the safety of different groups, researchers released an online platform called the ‘moral machine’ (Awad et al, 2020). The moral machine ‘game’, which is free to play and multi-lingual, has gained international attention, with 39.61 million decisions being made by people from over 200 countries.
In the game, participants see scenarios in which a self-driving car will hit a group of people unless a bystander diverts it, in which case the car’s occupants will die. The pedestrian and driver groups differ on observables including the number of people at risk, and their age and gender. They also differ in their social status, their fitness and their behaviour (for example, whether the pedestrians are crossing the road legally or illegally), and even whether they are humans or animals.
The results suggest a preference for sparing women over men, the young over the old, and the law-abiding over the unlawful, among other things. Differences between groups of countries are also apparent – for example, decisions made by people in the southern cluster of countries (predominantly Latin American countries and those characterised by French influence) demonstrate a much stronger preference for sparing women, whereas decisions made by those in the eastern cluster (predominantly representing Confucian and Islamic cultures) show a marked preference for sparing the law-abiding.
How should society balance the trade-offs between efficiency and equity in pandemic policy-making?
The VSL, VOLY and QALY approaches include only the value of an individual’s own risk reduction. By adding these values up across a population for cost-benefit analysis, it is implicitly assumed that the objective should be to maximise the sum of individual benefits for the population. But we have already questioned whether individual benefit is the right metric, arguing that safety-focused altruistic concern for others should also be considered. Next, we question whether the simple summation of the value of risk reductions is appropriate.
As well as the total reduction in risks to life, known as efficiency, any policy will also be characterised by the fairness of its allocation of risk reduction across different groups. The most efficient policy may not be the most equitable, and while standard approaches ignore such a possibility, this may not best respect the public’s preferences.
Evidence from my research with co-authors (Arroyos-Calvera et al, 2019) sheds light on this issue in the context of fatality risk reductions. We describe a city made up of two zones, identical except for the risk that they face from dying from exposure to waterborne bacteria.
The results of an experiment in which participants are presented with a simple choice reveal an overall preference for policies that save the most lives, but that people are willing to sacrifice some lives saved overall in order to reduce the risks in both zones of the city. Their implied preferences for saving the most lives and for equity outweigh the preference for policies in their own self-interest (although these are also preferred, all else equal).
In follow-up work, we replicated our earlier experiment during the first wave of the Covid-19 pandemic and our preliminary results point to remarkable stability of preferences. These new data give us confidence in suggesting that, in the context of pandemic policy-making, equitable policies that help a broad range of people may better respect the preferences of the public than opting for less equitable policies, even if these save the most lives.
Specifically, the odds that a participant would choose a policy that scored more highly on equity were 6.6% higher than the odds of choosing a comparable but less equitable policy. The odds of choosing a more efficient policy were 18.4% higher than the odds of choosing a less efficient but otherwise identical policy. This calls for a re-evaluation of the role of the VSL, VOLY and QALY approaches in guiding policy-making both during and after the pandemic, to seek a way to incorporate distributional concerns.
At what cost?
Choosing a policy with more evenly distributed benefits and costs, or one that aims to be more fair in its distribution of benefits across groups and over time, may come at a stark cost if it means saving fewer lives overall. Striking the appropriate balance between respecting a preference for equity versus saving the most lives may be especially difficult in the context of Covid-19 because the benefits of policies like lockdown are concentrated in one part of the population while the costs are concentrated in another.
This contrasts with applications like road safety, where risks and benefits are more evenly distributed across the whole population. Because of the concentrated nature of the risks and benefits, equity may be difficult to achieve, and it may result in a significant reduction in lives saved compared with a policy more focused on efficiency. Dedicated research using the tools we have described would be invaluable in helping to guide trade-offs in this context.
How can social welfare analysis help?
Social welfare analysis considers distributional issues. This approach recognises that government policies often have an impact on multiple aspects of people’s wellbeing (for example, their financial position as well as their safety); that policies often have different impacts on different groups, so that there are typically ‘winners’ and ‘losers’ from any given policy; and that there are often trade-offs between inequality and overall welfare.
The social welfare function is a mathematical representation of the way that society weights the wellbeing of different groups, and can guide policy-making beyond simple cost-benefit analysis. In this approach, the individual wellbeing of people in society is quantified to give a list of welfare scores for society overall. Different lists, representing the outcomes of different policy options, are then compared according to the social welfare function, which describes how society should prioritise different welfare distributions. In the social welfare function, weights may be placed on the distribution of safety as well as on total amount of safety that can be gained from a policy.
The appropriate social welfare function for practical application is a contested question. Experts debate the role that citizens’ views should play, and the practical application of the framework. But the theoretical underpinnings are well-developed, and work is in progress to determine the UK public’s preferences. The move from theory to practical application could prove an exciting development in the future.
Standard approaches such as the VSL, VOLY and QALY metrics focus on the total sum of benefits, but do not take account of the distribution of safety improvements. This is especially problematic in the context of the Covid-19 pandemic because of the clear distributional differences both in the burden of disease and mortality risk, and in the costs of mitigation.
Growing evidence from economics and related fields demonstrates that people are altruistic towards others, and hold preferences for equitable as well as efficient provision of safety. Yet forgoing the standard approach of cost-benefit analysis means accepting that in the pursuit of fairness, fewer lives may be saved. This is a tough balance to strike, and another difficult trade-off for society to make.
Where can I find out more?
- A scoping study on the valuation of risks to life and health: the monetary value of a life year (VOLY): Susan Chilton and colleagues explain how the technology exists now to generate a theoretically robust, evidence-based and updated valuation of risk to human life and health. They conclude that applying such values would lead to better and more informed policy decisions and would have major implications not only for efficiency of government spending but also for equity in population wellbeing.
- Principles for allocation of scarce medical interventions: In a study written for The Lancet in 2009, Govind Persad, Alan Wertheimer and Ezekiel Emanuel consider different rules that could guide decisions about how to allocate scarce medical resources such as organs and vaccines. They conclude that no simple allocation principle is sufficient to account for the complex moral considerations, and instead propose ‘multiprinciple allocation systems’, evaluating three existing systems and proposing a new one.
- The moral machine: Recent scientific studies in machine ethics have raised awareness about the how humans perceive machine intelligence making autonomous choices involving human life and limb, both in the media and public discourse. This website aims to take the discussion further, by providing a platform for: 1) building a crowd-sourced picture of human opinion on how machines should make decisions when faced with moral dilemmas; and 2) crowd-sourcing assembly and discussion of potential scenarios of moral consequence.