Share
Tax data suggest that while inequality between countries has fallen dramatically, inequality within countries has risen. Household surveys point to much lower within-country inequality, but may suffer from underreporting of income. This column uses a new method to estimate the degree to which the rich report a lower fraction of income than the poor do, and then compute the path of global within-country inequality for the last 40 years. The authors find that not only has overall global inequality fallen, but within-country inequality has stopped rising and has been falling for the better part of a decade, if not more.
Is global inequality falling? By now, there is a virtual consensus that the answer is yes, as major developing world countries – China, India, and many other countries in Asia, Eastern Europe, Latin America, and increasingly in Africa – have grown to attain middle-income or even upper-middle-income status, and enjoy standards of living previously associated with the developed world. However, there still remains debate as to why. One prominent account of the fall of global inequality (Chancel et al. 2022) is that while inequality between countries has fallen dramatically, inequality within countries has risen instead. It is well-documented that the US has experienced a sustained rise in top income shares, and it has been argued that many developing countries – including China and India – have followed in its stead. If present trends in within-country inequality continue, according to this account, global inequality may begin to rise again if the between-country convergence process slows.
This account relies on using tax data, or their imputations, to measure inequality within countries, both developing and developed. While tax data present a very high-quality view of developed country income distributions, they are available for only a very small fraction of people in developing countries. In contrast, if one uses unadjusted household surveys to measure inequality within countries (World Bank 2024), then within-country inequality peaked shortly after the global financial crisis and has been declining for the latter half of the 2010s. Within-country inequality is also much lower in the household surveys than it is in the tax data. On the other hand, household surveys almost certainly suffer from substantial underreporting of income, with the rich likely reporting much lower fractions of their income than do the poor both because of the additional difficulty in estimating their incomes and because of the concerns these data might not remain confidential.
In our research (Pinkovskiy et al. 2024), we try to reconcile the two sets of results by proposing a new method to estimate within-country inequality that adjusts household survey estimates by allowing the rich to report a lower fraction of their income on the household survey than the poor. We leverage the fact that most household surveys allow the computation of regional-level mean incomes, and that most countries compute regional GDP. It is intuitive that if the rich are reporting a lower fraction of their income in surveys than do the poor, but the regional GDPs accurately reflect the income earned in a given region, then inequality in the regional survey means should be lower than in the regional GDPs. Based on this insight, we can estimate for each survey its underreporting progressivity – the degree to which the rich report a lower fraction of income than do the poor – and use it to adjust each survey for misreporting. Aggregating across surveys in different countries and over time, we can compute the path of global within country inequality for the last 40 years. As many countries also have regional-level estimates of disposable income or consumption, we can also use them to adjust survey distributions to estimate the distributions of disposable income or consumption, which are more comparable income concepts to what is used by the World Bank to measure poverty.
We obtain four main sets of findings. First, we show that the poor also underreport their income on surveys, and they have been underreporting more and more over time. Whereas the bottom 50% of the global income distribution reported a higher fraction of their income in 1980 than did the top 10%, the reverse was true in 2019. Moreover, by 2019, the bottom 50% also accounted for a larger fraction of overall underreporting of disposable income than their survey income share. Much of this was driven by developing countries, and specifically the developing country poor, underreporting at higher and higher rates as they became richer. On average across surveys, underreporting progressivity experienced a slight decline between 1980 and 2019, evolving smoothly over time, suggesting that while developing countries increased their underreporting as a whole, the developing country poor increased their underreporting by more than the developing country rich (purple and pink lines in Figure 1). In contrast, the rates of average underreporting progressivity across surveys that would rationalise the estimates in the World Inequality Database must have experienced a dramatic kink after 2000, having been flat for the 1980s and 1990s but rising steeply afterwards (green line in Figure 1).
Figure 1 Underreporting elasticity over time, global average
Notes: An underreporting elasticity equal to 1 indicates that people underreport a constant fraction of their income, while value greater than 1 indicates that the rich underreport a larger fraction than do the poor. The purple line presents the population-weighted average underreporting elasticity of survey income relative to national accounts GDP, the pink line presents it relative to national accounts disposable income or consumption (HFCE) and the green line presents the population-weighted average implied underreporting elasticity of survey income that would rationalize discrepancies between survey reported income and estimates of income distributions in the WID.
Second, we replicate the findings of the earlier literature that global poverty and inequality have declined since the 1980s (Bhalla 2002, Sala-i-Martin 2002a, 2002b, 2006, Chen and Ravallion 2010, Chancel et al. 2022). Survey misreporting is very large, and could have been compatible with rising or flat as well as falling true global inequality after it was explicitly corrected for. We could have hypothetically found that the gains from growth in the past 40 years had been monopolised by the richest regions to a vastly greater extent in the national accounts than in the household surveys, leading us to overturn the consensus that global inequality has fallen, but even with the substantial changes we observe, the consensus remains. We find that once survey inequality from the World Bank PIP is corrected for underreporting, it increases substantially and is close in magnitude to the tax-based measures in Chancel et al. (2022). However, even after this correction, if national accounts disposable income is used to anchor each household survey, world inequality and poverty are lower and fall faster according to our estimates than they do in the World Bank PIP, which uses surveys alone. Intuitively, the poor also misreport their income on household surveys (Meyer and Sullivan 2015), and accounting for misreporting at all points on the income distribution does not fully compensate for the enormous difference between national accounts aggregates and household survey means.
Third, we find that not only has overall global inequality fallen, but within-country inequality has stopped rising and has been falling for the better part of a decade, if not more. Inequality within countries according to a range of measures, weighted to reflect countries’ populations, has declined since the mid-2000s to attain levels of the early to mid-1990s (purple and pink lines in Figure 2). Inequality has also fallen for several large countries such as China, India, and Indonesia. Therefore, the recent declines in inequality are driven by both falling across-country inequality and falling within-country inequality, rather than within-country inequality acting as a headwind to global inequality reduction. While our misreporting adjustment makes the within-country inequality reduction more salient, this result does not depend on our methodology, as the reduction in within-country inequality is present in the unadjusted household survey data as well, and data from the World Inequality Database (green line in Figure 2) are consistent with a recent stabilisation of within-country inequality.
Figure 2 Within-country average Gini
Note: Each line shows estimates of the within-country population-weighted average Gini over time. Brown line shows estimates over time based on surveys alone (World Bank PIP). Green line shows estimates based on the WID. Purple and pink lines show estimates based on survey distributions adjusted with estimates of underreporting elasticities relative to national accounts GDP and national accounts disposable income or consumption, respectively.
Finally, the world is doing much better than we thought not only at eliminating extreme poverty – reducing the $2.15-a-day headcount ratio – but at also reducing poverty at a higher poverty line. Using the World Bank’s preferred poverty lines of $3.65 a day and $6.85 a day (reflecting the medians of lower-middle-income and upper-middle-income country poverty lines), we find that poverty rates at these higher thresholds have declined to 30% and 50%, respectively, of their 1990 levels for the world distribution of disposable income or consumption (pink lines in Figure 3), 1 which is considerably lower than estimated by the World Bank using household surveys alone (brown lines in Figure 3). Hence, the world distribution of income is less defined by a large ‘precariat’ modestly above an extreme poverty line but liable to fall back into destitution following a global shock, but rather increasingly by a true ‘global middle class’ (Sala-i-Martin 2006) that is not poor even by upper-middle-income country standards.
Figure 3 Poverty ratios to 1990
Note: Each line shows estimates of the global poverty ratio to 1990 for the indicated poverty line (thus all lines are equal to 1 in 1990). Brown lines show estimates over time based on surveys alone (World Bank PIP). Pink lines show estimates based on survey distributions adjusted with estimates of underreporting elasticities relative to national accounts disposable income or consumption.
Our message is an optimistic one. Not only has inequality fallen dramatically across countries, as has been recognised now for decades, but it has also stopped rising and started declining on average within countries as well. Both within- and between-country inequality are helping global inequality decline. World poverty is also declining rapidly for a large set of reasonable poverty lines and faster than estimated by the World Bank PIP. On the eve of the pandemic in 2019, the world was rapidly becoming richer and more equal.
Source: cepr.org
