Associated Press/Lee Jin-Man

Covid-19 and Global Income Inequality

 

angus deaton, winner of the Nobel Prize in Economic Sciences in 2015, teaches in the School of Public and International Affairs at Princeton University.

Published April 26, 2021

 

It’s widely known that the Covid-19 pandemic has damaged the lives and livelihoods of less-educated and lesswell- paid people more than those of the more educated and better paid — many of whom have the option to stay safely at home and continue to work. The saving grace: the resulting increase in domestic income inequality has been offset at least in part by large-scale government income support programs in many countries around the world. International inequality is another matter, and here there is a widespread belief that the pandemic has increased inequalities in income among countries. For example, Ian Goldin and Robert Muggah, writing for the World Economic Forum, say that “inequality is increasing both within and between countries.” The UN Development Program concludes that “the virus is ruthlessly exposing the gaps between the haves and the have nots, both within and between countries.”

Joe Stiglitz lays out the rationale: “The least developed economies have poorer health conditions, health systems that are less prepared to deal with the pandemic, and people living in conditions that make them more vulnerable to contagion, and they simply do not have the resources that advanced economies have to respond to the economic aftermath.”

This line of reasoning seems compelling. But it is good to check the data, which is what I do here. I demonstrate that global inequality — when defined as the dispersion of per capita income among countries, taking each country as a unit — has continued on its prepandemic downward trend, and has, if anything, fallen faster as a result of the pandemic. That is hardly the end of the story, however. Read on.

COVID-19 and Income

Figure 1 (below) shows a scatterplot across 169 countries of deaths per million against income per head in 2019. The areas of the circles are proportional to population — Brazil is much bigger than Ireland, for example. The circles are shown in blue for the OECD countries and in red for others. It turns out there is no statistical relationship between per capita income and Covid-19 deaths per million within the OECD, so the positive relationship shown in the scatterplot is dominated by the relationship between OECD and non-OECD countries, as well as by the relationship within the non-OECD countries.

Among the non-OECD countries, much depends on the statistical impact of India and China. Ignoring population size, the country by-country relationship in the non-OECD is close to that for all nations. Weighted by population size, the relationship holds only if China is excluded: China’s low death toll is an outlier, and its population is the largest in the world, so its inclusion annuls the relationship.

The positive relationship between death and income, as shown in the regression line in Figure 1, raises important issues, because it contradicts so many presuppositions. Ever since Samuel Preston’s famous 1975 paper, studies of global health and global income have universally found that higher-income countries have better health. They have better public and private health systems — both of which cost money — and usually have governments that are more effective at protecting their population’s health.

Such is the basis for Stiglitz’s argument, quoted above. And it is backed up by the comprehensive 2019 study of global health security by Johns Hopkins, the Nuclear Threat Initiative and the Economist Intelligence Unit. The Global Health Security (GHS) study published a set of global health indexes for 195 countries based on 140 questions that measure country capacity in six dimensions:

  • prevention of the emergence and release of pathogens
  • early detection and reporting for pandemics of potential international concern
  • rapid response and mitigation of the spread of a pandemic
  • sufficiency and robustness of the health system to treat the sick and protect health workers
  • commitments to improving national capacity, financing, and adherence to norms
  • risk environment and vulnerability to biological threats

These are presented separately and also aggregated into an overall index. In line with the “health is wealth” presupposition, the overall index correlates positively with per capita income (adjusted for purchasing power) over 166 countries. And much the same can be said with the subindexes listed in the bullet points.


Crispin Hughes/Panos Pictures/Redux

Yet, in spite of being designed to be more prepared for “high-consequence pandemic threats, such as a fast-spreading respiratory disease agent that could have a geographic scope, severity, or societal impact and could overwhelm national or international capacity to manage it,” and in spite of the evident care and thoroughness of the report, countries that did better on the indexes experienced more deaths per million from Covid-19 than those that did worse.

It seems that even distinguished and careful experts could not predict the international patterns of deaths in the pandemic, at least through the end of 2020. Nor is it clear that any country could have been adequately prepared for what happened. As nations learn lessons from the pandemic and try to better prepare for the future, they will presumably have to take measures, at least some of which are different from those proposed in the GHS report.

Set aside the possibility that the paradox is just the consequence of underreporting of deaths in low-income countries. There are alternatives. The low number in low-income countries has been linked by Pinelopi Goldberg and Tristan Reed to (the lack of) obesity, to the smaller fraction of the population over 70 and to the lower density of population in the largest urban centers.

Another alternative is to focus on demography. Patrick Heuveline and Michael Tzen provide age-adjusted mortality rates for each country by using country age-structures to predict what death rates would have been if the age-specific Covid-19 death rates had been the same as the U.S. The ratio of predicted deaths to actual deaths is then used to adjust each country’s crude mortality rate. This procedure scales up mortality rates for countries that are younger than the U.S. (Peru has the highest age- and sex-adjusted mortality rate) and scales down mortality rates for countries like Italy and Spain (which had the highest unadjusted rate) that are older than the U.S. (Peru has the highest age- and sex-adjusted mortality rate) and scales down mortality rates for countries like Italy and Spain (which had the highest unadjusted rate) that are older than the U.S.


Eduardo Parra/Europa Press via Associated Press

If Figure 1 were redrawn using the adjusted rates, the positive slope would remain, though the slope showing the relationship between death rates and income would be reduced from 0.99 to 0.47 — that is, the relationship would hold but would be less pronounced. In poor countries, many children suffer from ill health — diarrheal disease, respiratory infections, undernutrition — that could raise the risk of death conditional on infection, so they may not get as much benefit from a young age structure as would the U.S. On the other hand, poor countries are also warmer countries, where much activity takes place outside, and there are relatively few large, dense cities with elevators and mass transit to spread the virus.

It is also possible that Africa’s long-standing experience with infectious epidemics stood it in good stead during this one. People in countries with more-developed economies consume a higher fraction of income in the form of personal services, which makes infection easier. But such after-the-fact explanations are worth little without solid before-the-fact analysis — and again, the serious and thorough analysis in the GHS index report predicted just the opposite of what’s been happening in terms of Covid-19 deaths.

Perhaps the most surprising result in Figure 1 is the relatively high number of deaths in the highest-income countries. There has been much (well-deserved) condemnation of the Trump administration’s handling of the epidemic. But deaths per million in the U.S. are no worse than in several other rich countries.

Statements about the disproportion of deaths and population (e.g., that the U.S. has only 4 percent of the world’s population but 20 percent of the deaths, or that the U.S. has more than 30 times as many deaths as Pakistan) tell us a limited amount about how well or badly the pandemic was handled in the U.S. or elsewhere. Deaths in the U.S. are above the regression line in the figure. But, by that measure, the U.S. did about as well as Sweden, and better than Hungary, Spain, Poland, Portugal, Italy, the United Kingdom and France. (Belgium did worst of all, likely only because of its more comprehensive measure of Covid-19 deaths.)

 
In 1900, after a safe and effective vaccine had been available for more than a century, and in spite of already being the world’s richest country, the U.S. did worse than other rich countries in preventing smallpox deaths.
 

There’s another factor to consider here. In a 2015 book, Werner Troesken contends that Americans’ particular commitment to personal liberty includes a propensity to infectious disease. In 1900, after a safe and effective vaccine had been available for more than a century, and in spite of already being the world’s richest country, the U.S. did worse than other rich countries in preventing smallpox deaths. Troesken argues that this was “not despite being rich and free, but precisely because it was rich and free.”

Death and Growth

The second part of the story is the relationship between pandemic deaths and per capita GDP growth (positive or negative) in 2020. Here, I rely on forecast data, two sets of which are available: one from the IMF in October 2020, and one from the World Bank in early 2021. I use the earlier IMF numbers, but it should hardly matter. The World Bank numbers are close, and the cross-country correlations between the two sets of estimates are an excellent fit (0.945).

Figure 2 plots the IMF’s predicted growth rates from 2019 to 2020 against deaths per million. China, with few deaths, shows positive predicted growth, while the U.S., with many deaths, shows negative predicted growth. There are many cases that are not near the line, for reasons unrelated to Covid- 19. But, as would be expected, there are similarly sloped scatters of countries overall and both in and out of the OECD. The (population-weighted) regression — shown as the dotted line — has a slope of ‒0.015, implying that predicted GDP per capita growth decreases by one and a half percentage points for every unit increase in the logarithm of deaths per million. The slope of the population-weighted regression line is still negative (‒0.007).

I have also repeated these calculations using, not growth forecasts for 2020 in October 2020, but the revision to the 2019 to 2020 growth forecast between the 2019 and 2020 editions of the World Economic Outlook — the idea being to isolate the reduction in growth associated with the pandemic. The corresponding figure and regression results are similar to the originals, albeit with lower levels of predicted growth, so that, for example, all revisions to growth are negative.

It is perhaps not surprising that deaths from Covid-19 should bring economic destruction, nor that the relationship should be tighter than the relationship between deaths and income in 2019. But, once again, that this relationship should exist was not intuitively obvious before the pandemic. Indeed, in the early days of the pandemic, there was much discussion of the value of life and about a supposed trade-off between deaths and income — that lockdowns would save lives but destroy economies. Remember Dan Patrick, the lieutenant governor of Texas who, in March 2020, said that older Americans should be willing to sacrifice their lives to Covid-19 in order to keep the American economy growing?

But as previously noted by Martin Wolf, who looked at the advanced countries plus India and China, there is no evidence for the existence of any such trade-off. Instead, the route to growth lies through stopping deaths. It is not a matter of your money or your life, but your money and your life. This, by the way, should not necessarily be taken as an argument in favor of government-ordered lockdowns, because voluntary social distancing in the face of infection and death has also been important, and, according to Austan Goolsbee and Chad Syverson, perhaps more so.

Figure 3 plots the income changes from Figure 2 against the 2019 levels of income in Figure 1. It shows that richer countries had slower (or more negative) growth in 2020 than did poorer countries. The slope of the unweighted regression line in Figure 3 is ‒0.010, implying that every doubling of income shaves one percentage point off the predicted growth rate.

The population-weighted regression has a small slope of ‒0.003 because of the divergent experiences of India and China. China is growing because, in spite of its relatively high income, it has seen few deaths, while India, with more deaths per million than other countries at its (low) income level, shows a 10.2 percent decline in income. Each country is an outlier, but in opposite directions.

But if we ignore population size — treating, say, India and Japan as equivalent data points — the negative relationship between growth in 2020 and income in 2019 exists for the world as a whole, and within the non-OECD countries. Within the OECD, the better-off countries grew faster (or declined slower) in 2020, but the statistical relationship is not significant. That higher-income countries experience the largest decreases in income on average does not necessarily imply that there was a decrease in inequality in per capita incomes between countries.

Figure 3 plots the income changes from Figure 2 against the 2019 levels of income in Figure 1. It shows that richer countries had slower (or more negative) growth in 2020 than did poorer countries. The slope of the unweighted regression line in Figure 3 is ‒0.010, implying that every doubling of income shaves one percentage point off the predicted growth rate.

The population-weighted regression has a small slope of ‒0.003 because of the divergent experiences of India and China. China is growing because, in spite of its relatively high income, it has seen few deaths, while India, with more deaths per million than other countries at its (low) income level, shows a 10.2 percent decline in income. Each country is an outlier, but in opposite directions.

But if we ignore population size — treating, say, India and Japan as equivalent data points — the negative relationship between growth in 2020 and income in 2019 exists for the world as a whole, and within the non-OECD countries. Within the OECD, the better-off countries grew faster (or declined slower) in 2020, but the statistical relationship is not significant. That higher-income countries experience the largest decreases in income on average does not necessarily imply that there was a decrease in inequality in per capita incomes between countries.

Figure 4, closes the circle, showing estimates of between-country income inequality using a standard measure of inequality, the Gini coefficient, with and without country-population weights. (A Gini of 0 represents perfect equality; a Gini of 1 is “perfect” inequality.)

The top lines, marked “unweighted,” show the Gini coefficient of national per capita incomes, adjusted for purchasing power, with no account taken of relative population size — what Branko Milanovic calls the “Concept 1” measure of inequality. This measure has a slight upward trend (toward greater inequality) until its peak in 2000, and subsequently declined, aside from during the Great Recession (2008-2011). Note that it declined slightly faster from 2019 to 2020 than from 2018 to 2019.

The broken top line is also based on the IMF estimates of Concept 1 inequality. It differs from the solid line for two reasons. First (and not truly relevant here), it reflects the IMF’s change in base years for adjusting for purchasing power. Second (and very relevant here) the broken-line numbers for 2020 are predictions made in 2019 — not estimates after the fact. As such, it imagines a world in which the pandemic never happened, and thus (as in Figure 2) becomes my proxy for world inequality in 2020 without the pandemic.

Note that the Concept 1 broken line shows a small decline from 2019 to 2020, but less than the actual outcome. The difference between it and the solid line for the past few years only (circled in Figure 4) is the effect of the pandemic — implying that Concept 1 inequality declined a bit thanks to the pandemic.

The lower lines in Figure 4, marked “population-weighted,” reproduce the calculations, but with each country weighted by its population (Milanovic’s Concept 2). This measure of inequality has been falling for many years, largely because the world’s two largest countries, China and India, have grown rapidly. Indeed, growth has raised more than two billion people from near the bottom of the global income distribution to near its middle, where we can see them today in Figures 1 and 3.

This population-weighted (Concept 2) measure of global inequality grew (albeit slightly) between 2019 and 2020, in accord with the story that the pandemic has driven countries apart. Again, the “counterfactual” — my proxy for a world in which the pandemic never happened — is supplied by the broken line from the 2019 forecasts. There is no upturn in broken line. The gap between the Concept 2 lines thus grows, implying that the pandemic has increased Concept 2 inequality in 2020 over what it would have been without Covid-19.

The polar-opposite stories about changes in inequality told by the two pairs of lines comes from the counterbalancing effects of the two largest countries, India and China, which occupy very different positions in Figure 3. China’s 1.4 billion people experienced few deaths and continued to grow in per capita income, which took them closer to the richer countries of the world and decreased (population-weighted Concept 2) global inequality. India’s 1.4 billion people experienced many more deaths, as well as a large drop in income, which tended to increase (population-weighted) global inequality. When India, but not China, is excluded, the uptick in inequality vanishes. But when China alone is excluded, the uptick is even larger.

Ifs, Ands and Buts

For reasons that are only partially understood (and may include errors in measurement), poorer countries suffered fewer Covid-19 deaths per capita in 2020 than did richer countries. Moreover, each country’s loss in per capita income between 2019 and 2020 was positively related to its per capita Covid-19 death count — more income, more deaths.

These two facts together have meant that per capita incomes have, on average, fallen more in countries with higher per capita incomes in 2019. The 97 poorest countries lost an average of 5 percent of their 2019 per These two facts together have meant that per capita incomes have, on average, fallen more in countries with higher per capita incomes in 2019. The 97 poorest countries lost an average of 5 percent of their 2019 per capita GDP, while the richest 96 countries, with an average per capita income six-and-a-quarter times larger, lost an average of 10 percent. This need not have narrowed international income inequality, but in fact it did in terms of Concept 1 inequality: per capita incomes in equally weighted country comparisons are closer to one another now than in 2019.

The Small-Economy Conundrum

The presence or absence of small economies matters a lot for the unweighted measures of inequality. The two richest economies in the world in 2019, measured by per capita GDP in 2017, were Macau and Luxemburg, with populations of 670,000 and 614,000, respectively. After that, in positions three through eight were Singapore (5.7 million), Qatar (2.8 million), Ireland (4.9 million), Switzerland (8.5 million), Norway (5.4 million) and then the U.S. (328 million).

Look more closely at the impact of tiny Macau on our analysis. During the pandemic, Macau was predicted to lose just over half of its per capita GDP, not because of a large number of Covid-19 deaths, but because the gambling, entertainment and tourism on which it depends were hit so hard by the pandemic. This knocked Macau from first to ninth in the global per capita income rankings and had a large effect on unweighted global inequality. Indeed, the unweighted Gini coefficient actually rises from 2019 to 2020 if Macau is excluded from the calculation.

One way to deal with this irony would be to exclude economies like Macau — if, indeed, a gambling concession perched off the coast of China is a country at all. (And, remember, its international status is a Special Administrative Region of China, rather than a country.) But it is difficult to do this in a principled way. Should the cut-off be a population of a million or five million? Or we could simply ignore the unweighted measures and focus on the weighted measures. But, as we have seen, weighted measures have their own pitfalls.

Yet the smallness of the very richest countries is far from their worst problem in the context of measuring changes in inequality: per capita GDP is an exceptionally poor measure of material well-being. In 2019, the share of household consumption expenditure in GDP was just 25.4 percent in Macao, implying that little of Macau’s GDP supports consumer spending in Macau, and instead goes to some combination of foreigners, corporations and the wealthy. The figure for Luxemburg is 29.5 percent, for Qatar 24.5 percent, for Ireland 30.4 percent, all to be compared with 67.9 percent for the U.S.

Many mini-economies are tax havens, and much of their GDP is corporate profit, including profit accruing to non-citizens. So when we include these economies in global comparisons, we are not comparing like with like and are including much that is unrelated to the living standards of their citizens. Using consumption expenditures rather than GDP would offer a more realistic picture — though it would not solve the small-economy weighting problem. In any event, consumption numbers aren’t currently available for 2020. And while several arguments can be mounted for excluding Macau, there would be a good deal more discomfort if we were to exclude Singapore or Ireland.


Ahn Young-Joon/Associated Press

China had few deaths and experienced positive economic growth in 2020. Before the pandemic, China’s rapid growth had lifted more than a billion people up from the bottom of the global income distribution, and China dominates explanations for the reduction in global income inequality in recent decades when each country is weighted by its population. But this effect has been attenuating as China’s income has risen.

Today, out of the world’s population of 7.8 billion, 4.4 billion live in countries whose per capita income is lower than China’s, while only 2.0 billion live in countries whose per capita income is higher than China’s. During the pandemic, the Chinese economy grew, while most other economies shrank. While this had the effect of reducing populationweighted global inequality, the impact was not large enough to offset the inequalityincreasing effect of (much poorer) India’s loss of income. All told, then, populationweighted global inequality increased in 2020.

Contrary to preexisting trends, the pandemic reduced global unweighted inequality and increased global population-weighted inequality. That my findings are consequences of the pandemic is supported by comparing inequality measures using IMF income estimates pre- and post-pandemic.

It is important to be clear about what I am and am not claiming here. My results say nothing about whether the degree of suffering has been larger or smaller in poor countries. They are consistent with the pandemic increasing poverty around the world — in particular, with World Bank estimates that between 88 and 115 million people will be pushed into poverty. Even if all countries had the same decline in per capita income, the poorer countries would have had larger increases in poverty because they have many more people near the global poverty line. As it is, we know from World Bank research that, compared with richer countries, the suffering from the pandemic has hit poor countries more in terms of poverty, and less in terms of mortality.


Carol Smiljan/Nurpho via Associated Press

All of my results come from national accounts data, and there is a long history of national accounts data on consumption and income differing from consumption and income as recorded in the household survey data that are used for the assessment of poverty and within-country inequality. Beyond that, GDP per capita is often a poor indicator of material living standards, if only because GDP contains much — such as profits accruing to foreigners — that is not part of domestic consumption.

 
Even if all countries had the same decline in per capita income, the poorer countries would have had larger increases in poverty because they have many more people near the global poverty line.
 

Note, too, that my findings may reflect temporary circumstances. The pandemic is not done, there are more deaths to come, and they may fall more heavily on poorer countries. Indeed, given that the pandemic started along trade routes and affected urban before rural areas, it is plausible that current patterns will continue to change.

It is also possible that deaths are severely understated in poor countries, some of which do not have regular vital statistics systems that comprehensively report deaths even in normal times. My calculations use data up to the end of 2020, before vaccines had any chance to affect outcomes, and they say nothing about how the vaccines will be distributed among countries. It is entirely plausible that rich countries will recover more rapidly in 2021 and beyond, which will widen global inequality.

Conclusions and Reservations

Both concepts of inequality — unweighted on a country-by-country basis and populationweighted — raise uncomfortable issues. The unweighted measure, which is perhaps closest to the common notion of global income inequality, is sensitive to the inclusion or exclusion of small countries. The population-weighted measure does not have this drawback, because small countries get little weight. But changes in inequality across time will often critically depend on what happens with India and China — as is the case in the analysis above.

China, which is relatively well-off, did much better in the pandemic than rich countries, and much better than India, which is poorer than China, and which did much worse than the rich countries. Indeed, at risk of making you wonder why you plowed through the analysis above, given the facts in the previous sentence, not much more is added by looking at comprehensive measures of global income inequality.

main topic: Inequality
related topics: Public Health