Over the last six months, the COVID-19 (coronavirus) pandemic has generated massive losses in well-being around the world – through greater mortality, widespread suffering, higher unemployment and rising poverty. Attempts to quantify these welfare costs have so far struggled with the fact that mortality losses and economic losses are typically expressed in different units: the former in numbers of deaths and the latter in monetary units. To combine them, economists have often resorted to various ways of calculating the “value of a statistical life” in monetary units. Unfortunately, all such methods have proven deeply contentious in the public debate, essentially because most people seem uncomfortable with putting a price on human life, and policymakers (rightly) care a great deal about what most people think…
Yet, being able to express the health and economic costs of the pandemic in a common unit is important, and not only because it would enable us to quantify its total welfare cost. It matters even more because some of the policy responses to the crisis face a trade-off between lives and livelihoods. Of course, there are some policies which imply no such trade-offs: developing a vaccine, for example, would help both save lives and restore livelihoods. But important policy responses to a viral epidemic, such as lockdowns, social distancing and various forms of containment do, quite clearly, imply some trade-offs: stricter lockdowns save lives, but lead to increased bankruptcies, unemployment and poverty. Indeed, some have wondered whether these policy responses have been appropriate. As Peter Singer and Michael Plant note, without a common metric for valuing the economic and life costs of the crisis, making the right policy choices is much harder.
In a recent paper, we propose a simple approach to estimating the welfare costs of the crisis, using life years as a common unit. Our approach covers mortality and poverty costs only, for 150 countries. For each country, we estimate how many years of life have been lost through COVID-19 deaths (“lost years”) as of early June. We also estimate how many additional years are spent in poverty due to the pandemic and the subsequent individual and public policy responses (“poverty years”).
Take for instance Belgium, the country with the highest per-capita mortality resulting from COVID-19. We find that the number of poverty years in Belgium is 3.5 times larger than the number of lost years. Here is how we compute this ratio: We estimate the number of lost years as the number of deaths (9,600) multiplied by the average residual life-expectancy of those who died in Belgium (9.5 years), i.e. about 91,000 lost years. In order to estimate the number of poverty years, we start from the forecast GDP shock, which is a reduction of 8.5% for 2020, according to the Belgian National Bank. Assuming – conservatively – that all incomes are affected in the same proportion (Lakner et al., 2020), we can compute the additional number of people whose incomes fall below the national income poverty threshold, which is 320,000 individuals. Assuming – again, conservatively – that these individuals stay poor for one year only, we obtain 320,000 poverty years. We conclude that the number of poverty years is 3.5 times larger than the number of lost years.
Which source of welfare loss is dominant as of early June: mortality or poverty? The answer depends on the value of a key normative parameter, α, which captures how many poverty years are as bad as one lost year. This parameter embodies the key normative trade-off between lives and livelihoods. One way to form an opinion on α is to ask yourself the following question: how many years of your remaining life would you be willing to spend in poverty in order to increase your lifespan by one year? Importantly, we remain agnostic about the value of α and let the reader form their own opinion. We simply present the empirical ratios of the number of poverty years to the number of lost years. If the reader thinks that 5 poverty years are as bad as one lost year, then mortality is a more important source of welfare losses in Belgium than poverty, since we find an empirical ratio of 3.5.
We follow essentially the same approach for all countries in our sample although, for countries that provide less detailed data than Belgium, we need to make some additional assumptions (e.g. on the age distribution of COVID-19 deaths) in order to estimate their empirical ratios. Three main findings arise:
1.The 406,000 COVID-19 deaths confirmed worldwide by June 9 have generated an estimated 4.3 million lost years. Using the World Bank’s latest growth forecasts for 2020 and the extreme international poverty line of $1.90 per person per day (and assuming no changes in inequality) we estimate that the crisis has led to 68.2 million additional poverty years, yielding an empirical ratio of 16.
Of course, the $1.90 line may be deemed too extreme for all but the poorest countries, so we repeat our calculations using the World Bank’s income-class specific poverty lines: $1.90 for low-income countries; $3.20 for lower middle-income countries; $5.50 for upper middle-income countries; and $21.70 for high-income countries. Using those somewhat more generous – but still frugal – thresholds, the number of poverty years in the world (235 million) is 54 times the number of lost years.
2. Poverty costs relative to mortality costs decline sharply with GDP per capita.; 620 for lower middle-income countries; 45 for upper middle-income countries and 9.5 for high-income countries. These are striking differences. Figure 1, which uses log scales on both axes, plots the empirical ratios for the 150 countries in our sample. Despite using progressively higher poverty lines as countries grow richer, the figure shows a very pronounced negative gradient.
Figure 1: The poverty years/lost years ratios for 150 countries
These numbers suggest very high poverty costs from the pandemic and associated responses, both public and private. For most developing countries, one would have to hold unrealistically high values of α (“I’d rather spend 600 years in absolute poverty if that would add one year to my current life expectancy”) to judge that the mortality effects dominate the poverty effects as sources of welfare losses. (Despite the fact that our approach treats the value of human life exactly the same way across all countries.)
But that is not to say that a laissez-faire, “no-intervention” response to the pandemic would have been superior. To answer that question, one would need to observe rich and poor countries at the same stage of the pandemic, which they clearly are not currently at. And one would have to simulate a counterfactual scenario where governments had not responded, and the epidemic evolved until countries (hypothetically) reached herd immunity.
3. We constructed such a counterfactual scenario using the Banerjee et al. (2020) assumption that herd immunity is reached at an 80% infection rate and found that such a no-intervention policy would have led to higher welfare losses, particularly in richer countries. For high income-countries we estimate that the number of lost years under this scenario would have been five times larger than the sum of lost years and poverty years estimated as of early June (based on the $21.70 poverty line). This implies that the welfare consequences of no-intervention in these countries would have been at least 5 times worse than the consequences measured as of early June. (Five times is based on α =1. The ratio obviously rises with larger values of α.)
Per-capita GDP levels matter here too. At equal infection rates, developed countries suffer more lost years than developing countries, largely because they have older populations and higher residual life expectancies (at all ages). This means that the hypothetical welfare losses under no-intervention in poorer countries are closer to those observed as of early June. But they are larger nonetheless: 4 times larger in upper middle-income countries; 1.1 times larger in lower middle-income countries and 1.5 times larger in low-income countries. And these ratios entirely disregard the poverty years that would have been generated under no-intervention. We can therefore be fairly certain that
Although there are various limitations to our approach, we feel that the analysis usefully documents the importance of accounting for increased poverty in any assessment of the welfare losses associated with the pandemic – particularly in poor countries. Furthermore, although “no-intervention” would be a bad policy choice in almost all settings, our results also suggest that