The topic of inequality has been trending globally for the past several years. Attention has focused especially on the very top of the income distribution in each country, which traditional measures of inequality, drawn from representative household surveys, struggle to capture accurately. In place of surveys, researchers have made use of tax data, which provide a more robust account of incomes of the richest segment of society. In a handful of countries, analysis reveals that the share of income accounted for by the top 1 percent has grown sharply. This presents a quandary. The more new information we uncover about top incomes, the less faith we have in traditional survey-based inequality measures, and the less knowledge we can claim to have about the distribution of income across an economy’s entire population.

Missing top incomes in household surveys is a long established problem in both developed and developing economies. It has two causes: the tendency for richer households to refuse to participate in surveys, and the high probability of missing high-income households through sampling error. A number of studies (here, here and here) in the United States have shown that survey response rates correlate negatively with incomes, and that more effort is required in contacting households with higher incomes. A study of Latin American economies found that in several countries, the income of the richest survey participant fell short of the local salary of a manager in a medium- to large-sized firm.

This has important implications for the reliable measurement of inequality within each country. When top incomes are missing from a survey, the right tail of the distribution is measured with error.¹ Contrary perhaps to intuition, the resulting inequality measures may bias upward or downward, as the error affects both the distribution and mean of income in the survey sample—see here for details. (It is worth noting that missing top incomes has a more benign effect on the measurement of poverty, which focuses on the bottom of the distribution—see here for details.)

Ideally, the results from survey and tax data would be reconciled to obtain accurate measures of inequality across the full population. Attempts to do this for the U.S. (here, here and here) indicate that levels and trends in the Gini coefficient may differ markedly from those reported in official data. However, such efforts are fiendishly difficult to carry out in practice and fail to deliver a singular answer, since the two sources employ different income definitions and different units of observation. Moreover, outside of rich economies, tax data rarely have sufficient population coverage to capture top incomes accurately, if the data exist at all.

Instead, various survey methods have been developed to mitigate the problem of the missing rich ex-ante or to attempt to correct for it ex-post. These include persistent attempts to contact households that do not respond, oversampling rich households, and the re-weighting of participating households using estimates of the relationship between response rates and income. These methods are undoubtedly useful but cannot fully resolve the problem.

A solution via national accounts

One more enterprising solution comes via another measurement problem: the tendency for household surveys to obtain lower estimates of national income and income growth than national accounts. In their seminal paper on the changing global income distribution, economists Christoph Lakner and Branko Milanovic argue that the two issues are closely related, and use the missing income between surveys and national accounts as a proxy for missing top incomes in surveys. In practice, this means adding the shortfall in income between surveys and national accounts to the income reported by the top 10 percent of the survey population.

This solution has obvious appeal. Unlike the reconciliation of tax and survey data, it can be applied to virtually all countries and implemented with relative ease. And there is certainly good reason to believe that the problems of missing top incomes and the discrepancy between national accounts and surveys are linked. However, attributing the entire discrepancy between national accounts and surveys to the missing rich is harder to justify, as there are other known factors that contribute to the discrepancy.²

First, just as surveys and tax records employ different definitions of income, so do surveys and national accounts employ different definitions of income or consumption. Two items that appear in national accounts but that surveys miss are the imputed value of financial services to households, and consumption by nonprofit institutions serving households. A study from India found that these two items could together explain 22 percent of the difference between survey and national account aggregates. An additional item that income surveys often overlook is the implicit rent for homeowners and occupiers. Another study from India estimated that this might account for half of the difference between surveys and national accounts. In the U.S., studies show that about half of the difference between survey and national account aggregates are due to differences in concepts and comprehensiveness. These definitional differences affect the measurement of welfare for a range of different households—not just those with high incomes.

Second, some participating households in surveys underreport their income or consumption. If underreporting is limited to the richest households, then the attribution of the entire discrepancy between national accounts and surveys to missing top incomes remains valid. Higher incomes are associated with underreporting to the extent that they entail more diverse consumption patterns and sources of income. For instance, in the U.S. underreporting is especially egregious for consumption of irregular purchases, which is more prevalent among rich households, and respondents who are frustrated with long questionnaires often lie in response to embedded screener questions in order to skip entire sections of a survey. However, underreporting has also been shown to be a significant problem among the poorest households in the U.S. And a range of other evidence points to underreporting that is not correlated with income. For instance, it is well established that the self-employed tend to underreport, and underestimate, their income. U.S. households significantly underreport their consumption of demerit goods, such as alcohol or gambling. Finally, survey design choices can contribute to underreporting: in the U.S., the diary method has been found to suffer from underreporting of consumption compared to an in-person questionnaire, despite being considered the gold standard of survey design in the developing world.

A more measured approach would be to attribute a portion of the discrepancy between national accounts and surveys to missing top incomes. This has some direct empirical basis: one study of India estimated that depending on the period under study, between 20 and 40 percent of the difference between national accounts and surveys could be explained by the missing rich.

An alternative set of Ginis

We take this approach in the exercise below. Our purpose is to explore how our understanding of inequality within countries changes if adjustments are made for missing top incomes. To do so, we generate an alternative set of over 1000 Gini coefficients covering the vast majority of the world’s economies spanning the past 25 years. In each case, we attribute half of the missing income between surveys and national accounts to missing top incomes.³ We assign the missing income to a new, elevated class that we append to the top of the distribution, rather than supplementing the income of those in the top 10 percent of the survey population. In other words, our methodology corrects for the exclusion of income from missing rich households but not for the unreported income from households included in the survey.

We determine the size of the new class and their incomes using a Pareto distribution whose shape we derive from the distribution of the top decile of the survey population and the discrepancy between survey and national accounts.⁴ This method ensures that our adjusted Ginis exceed those derived from surveys in all countries where income reported in national accounts exceeds that reported in surveys. For technical details, see the appendix and accompanying code.

On average, the appended new class accounts for 1 percent—the top 1 percent—of the population after our adjustment. In other words, our approach implies that the problem of missing top incomes in household surveys is typically restricted to a country’s richest 1 percent.

We interpret our revised Ginis as a plausible upper bound of the true value in each country and year. In 73 percent of cases, the share of income accounted for by the top 10 percent of the population after our adjustment remains under 50 percent, which is the maximum observed from tax data.

Figure 1 illustrates the current Gini coefficient for all countries, based on the most recent available survey results, alongside our alternative Ginis after adjusting for missing top incomes. On average, our adjustment pushes the Ginis up by 9 percentage points. This raises the average within-country Gini from 39 percent to 48 percent.

Figure 1: Ginis before and after adjustment for missing top incomes

Our alternative Ginis imply a large reshuffle in the rankings of countries by inequality. The ten countries in the world with the highest Ginis include five new entries: Nigeria, Mexico, Indonesia, Georgia, and Guatemala. Whereas the U.S. (0.41) and China (0.40) report very similar levels of inequality, when we adjust for missing top incomes the U.S. appears considerably more unequal than its rival (0.51 versus 0.44). Meanwhile, India has the dubious honor of leapfrogging both countries under our adjustment, as its Gini skyrockets from 0.36 to 0.56.

Our adjusted Ginis support the existing evidence that Latin American countries are, on average, the most unequal in the world, followed by countries in sub-Saharan Africa, although the gap between them has narrowed. Furthermore, it is unclear whether this different is robust given that Latin American countries use income (as opposed to consumption) as their survey welfare measure, which tends to generate higher measures of inequality. South Asia, once considered among the world’s most equitable regions, now ranks third with our adjusted Ginis, driven by the large discrepancy between surveys and national accounts in India. We find that the region with the lowest inequality is the Middle East and North Africa (0.37), contrary to the popular narrative of high inequality fueling the Arab Spring. However, we view this finding with particular caution given the woefully poor survey coverage in that region.⁵

The possibility that inequality is, on average, much higher than reported is cause for alarm. But is inequality within the average country worsening? This is a complicated question to answer, as there are many ways to average Gini coefficients across countries. A recent World Bank study provides a useful approach for us to follow. Given that Ginis are not available every year in each country, we group country Ginis around five-yearly benchmarks. We then generate four different trends depending on whether we use population weights to average country Ginis, and whether we use the same sample of countries in each benchmark year.

Figure 2 illustrates the results using our alternative Ginis. Under all four trends, inequality trended upward in the 1990s but since the turn of century, inequality has either stabilized or fallen. This mirrors the trends reported by the World Bank using official Gini coefficients. The stable or downward trend in inequality in the 2000s even after we adjust for missing top incomes is especially remarkable as the growing discrepancy between national accounts and surveys should force our alternative Ginis upward.

Figure 2: Average Gini across countries, adjusted for missing top incomes

How much do we really know?

The method we employ for adjusting Ginis for missing top incomes is one of many alternatives and is not beyond reproach. We are confident others could improve upon it. Its purpose is not to arrive at a definitive measure of inequality that includes top incomes but rather to highlight the fallibility of inequality narratives based on official survey data. We submit that we know far less about how inequality compares across countries than is commonly assumed. Ironically, perhaps the most robust stylized fact about inequality within countries—that it is not rising—is the least well known.

The use of tax data by inequality researchers has been a powerful force in expanding our understanding on the concentration of income among the top one percent in several countries. At the same time, it has exposed how little we know about the overall distribution of income.

       

The twin forces of technological change and globalization are reshaping the global economy in multiple and important ways. Nowhere are their effects more pronounced than in labor markets. Considerable attention is now being devoted to analyzing and anticipating changing patterns of employment and wages in advanced economies. Thus far, less focus has been given to understanding the implications for emerging economies.

The 2016 Brookings Blum Roundtable was convened to take on that agenda. How are the factors driving change in global labor markets playing out differently in developing economies? What are the jobs of the future and how will the terms of employment differ? What skills will those job demand and how will those skills be acquired? And finally, what are the implications of these changes for development prospects and for society?

This essay provides a brief account of the roundtable conversation. It is followed by six essays, authored by leading experts on this topic, that were commissioned to inform the roundtable discussion.

An era of change

The effects of technology—especially digital innovation—and globalization on labor markets are three-fold.

The first effect is disruption as jobs relocate to take advantage of lower costs, evolve to entail different tasks, or undergo wholesale change with the elimination of old jobs and the emergence of new ones. Disruption is a permanent feature of any dynamic economy, the upshot of markets responding to changing conditions. While the effects of disruption can be devastating for any particular individual or community, its effects for workers as a whole are positive if old jobs are replaced with new ones that are safer, less physically arduous, more stimulating, and provide greater autonomy. This has been the case for the median worker in rich economies throughout modern history up until recently, but less true for developing economies.[1] Today’s patterns of disruption are more discriminating, creating new groups of winners and losers, and can no longer be blithely assumed as a net positive. In addition, disruption is believed to be growing in intensity.

Technology has made capital goods cheaper and encouraged their substitution for workers. The result has been a shrinking share of national income accruing as wages as opposed to profits across rich and poor economies.

The second effect is a diminishing role of labor. Technology has made capital goods cheaper and encouraged their substitution for workers. The result has been a shrinking share of national income accruing as wages as opposed to profits across rich and poor economies. Some analysts view this phenomenon in combination with accelerating disruption as inevitably leading toward large-scale technological unemployment. This is an especially alarming prospect in the developing world where demographics are expected to increase the size of the global work age population by half a billion people by 2030. Another possible consequence is to reinforce the trend toward widening inequality.

The third effect is to decentralize economic activity away from corporations to which individuals provide their labor, toward the crowd in which workers participate as micro-entrepreneurs. This phenomenon is a facet of both the digital economy and globalization with the unbundling and contracting out of ancillary services from firms such as accounting and marketing. It is associated with changes in the terms of employment, both positive, such as increased flexibility, and negative, including the decline of unions and weakened workers’ bargaining power, the erosion of norms on pay equity, and reduced job security. These effects are most apparent in rich economies where the formal sector dominates.

While there is good evidence for each of the three effects, their speed and scale is surrounded with uncertainty. For instance, estimates on the share of jobs that are at risk of automation over the medium term vary from 9 to 47 percent for OECD economies; fewer estimates exist for the developing world meaning the possible range is arguably even greater. In another example, a year ago the World Bank reported evidence of the hollowing out of labor markets in the developing world, mirroring the pattern observed in industrialized economies. More recent analysis reveals no such pattern.[2]

Breaking free of a Western lens

One of the challenges in discussing the future of jobs in the developing world is that the jobs agenda and the semantics and metrics that go with it principally reflect a rich economy setting. Without recognition of this, such discussions can easily become divorced from reality.

In the West, being employed means generating an income; those that are not employed are assumed to be either idle or not in a position to take on a paid job. By contrast, individuals recorded as being employed in the developing world typically represent only a small portion of those that are economically active, with the majority instead engaged in low productivity activities in the informal sector. Thus, whereas raising employment in the rich world means moving people into paid employment, in the developing world it entails moving people into more productive—and likely better paid—lines of work.

This has important consequences. In industrialized economies the spread of automation implies the risk of redundancy for many workers. In developing economies, many workers are engaged in economic activities that are already some distance from the technology frontier—in other words, they could feasibly be done with greater technology and efficiency—and are paid accordingly. Automation needn’t imply the loss of that work, but rather the possibility of a further diminishing income. Thus, estimates of the share of jobs at risk of being eliminated in rich and poor economies have different consequences, though both are undoubtedly worrisome.

These differences can also result in labor market conditions in the developing world being mistakenly glorified. For instance, subsistence living in the developing world can be unhelpfully classified as entrepreneurship, implying a degree of choice and value that is clearly lacking. By the same token, subsistence living and informal work share some characteristics with the gig economy in rich economies, despite obvious and important differences.

Distinguishing a jobs agenda from a development agenda

The jobs agenda is increasingly becoming recognized as a priority within developing economies and the international development community.

Despite these challenges, the jobs agenda is increasingly becoming recognized as a priority within developing economies and the international development community. Demand for better jobs and anxiety about the sustainability of livelihoods consistently poll among the top issues in surveys of public opinion, including the vast My World exercise conducted to inform the contents of the Sustainable Development Goals.

At the same time, the development community has struggled to articulate how a jobs agenda differs from support for economic development more generally. This is understandable. Job opportunities in any economy are in part a function of a country’s level of income with poor countries constrained by low domestic demand and investment.

That’s why linking poor economies with consumers and investors in rich countries can be such a powerful force in expanding the range of opportunities for workers—and it is precisely this economic integration that underpins the dramatic economic convergence and poverty reduction in the developing world over the past quarter-century. New technologies have the promise to further expand these linkages, including platforms that couple workers with overseas firms, and, in the near future, instant translation services powered by artificial intelligence. The ability of poor economies to take advantage of these opportunities rests on equipping their workers with sufficient education and skills, as well as access to technology and digital infrastructure.

Even for countries that share the same level of income, the number and quality of jobs can vary significantly with the sectoral composition of their economies. For policymakers seeking to steer their economies toward more labor-intensive sectors, this can present a moving target. Whereas 20 years ago, manufacturing was lauded for its ability to absorb large numbers of relatively unskilled workers and to provide a stepping stone for economies seeking to develop capabilities and move into increasingly productive activities, today service-driven economies with a digital focus such as the Philippines and India may present the most promising and sustainable example to follow.

Just as low income in an economy constrains the number of good jobs that are available, it is also associated with various other adverse characteristics of labor markets including greater job insecurity, limited social protections and high information costs. A credible jobs agenda requires recognizing the constraint of low income and the extent to which it is binding for different labor issues. For instance, the biggest barrier to firms moving into the formal sector is arguably their low productivity and this cannot be divorced from a country’s level of development. By contrast, new technologies can be transformative in reducing information costs in even the poorest economies.

Skills for the future

Education has consistently been associated with high rates of return in both rich and poor economies, but the precise mechanism by which education generates those returns remains somewhat contested. Some view education principally in terms of the imparting of skills while others emphasize its ability to sort individuals by their innate capabilities. The recent emergence of evidence of dismal learning outcomes in schools across large parts of the developing world could be interpreted as swinging the debate more in favor of the latter explanation. That is problematic for a future where the skills demanded of most, if not all, workers are seen as increasingly complex and quickly evolving implying the need for lifelong learning.

Another area of contention is the skills that should be given emphasis in training tomorrow’s workers. If that determination is made narrowly on the basis of the pattern of job creation and elimination, it likely leads to an emphasis on STEM skills. If instead it is made on the basis of the increasing speed of disruption and the changing organization of economic activity toward micro-entrepreneurship, then this raises the importance of soft skills and business skills to help workers adapt to a changing marketplace and to make optimal choices. Workers themselves appear to see the wisdom in this assessment: the most popular course on the online education platform, Coursera, is on “learning to learn.”

Digital technology is ushering in broader changes in the education sector, by threatening to depose those institutions that deliver inadequate outcomes and replace them with ones better suited to the changing global economy. However, we remain a long way from building an ecosystem that can address workers’ needs in assessing capabilities, imparting skills, accrediting skills acquired, and guiding each individual’s trajectory through a path of lifelong learning. A particular institutional gap exists in establishing a credible and secure repository to record each individual’s portfolio of qualifications and capabilities.

As the institutions that dominate the education sector change, this inevitably leads to questions about the role of government in the sector. The enormous positive externalities associated with education would suggest an important role. However, this need not necessarily be as a service provider, but more in shaping policy and providing finance for education. The greatest externalities likely occur for early childhood interventions—an area where government’s role has historically been too limited.

The future of job matching platforms

Another area where digital technology is altering the institutions that shape labor markets is in the process of matching workers with vacancies. This is an exciting area of innovation and invention in the developing world. By linking workers and firms to the information they need, digital job matching platforms can both help markets clear more quickly and induce dynamic responses, including in the content of education programs and in the investment decisions of firms and workers. For example, when Apple adopted the new programming language, Swift, hundreds of freelancers on the digital job platform, Upwork, responded rapidly by training themselves to take advantage of new work opportunities.

A central objective for job matching platforms is to accelerate this feedback loop. That will likely require greater linkages with other actors including those involved in education provision and skill certification. Such partnerships hold enormous potential for transforming labor markets but have yet to be harnessed at scale.

A bolder challenge still is to assist those workers for which matching platforms consistently fail to find job opportunities. Solutions for these individuals may require partnering with NGOs or government, and more complex interventions such as supporting worker relocation.

The consequences of a weak labor movement

While changes in the nature of work associated with the changing global economy are most prominent in advanced economies, there are nevertheless some important patterns shaping the developing world. Perhaps the most prevalent is the absence of a well-defined labor movement in many developing economies and a consequently limited role for collective bargaining. This is notable given what a prominent role these played in securing political goals and improvements in quality of life in the advanced economies during their own process of economic development. Moreover, the effects of technology and globalization appear to be reinforcing this pattern. There is already some early evidence of gig work in the developing world resulting in workers’ rights being less effectively upheld.[3]

There are a number of creative efforts underway to mitigate the negative consequences of these circumstances. For instance, some job matching platforms identify one of their core services to workers as helping them to optimally price their work. However, any attempt to strengthen workers’ bargaining power in an atomized work environment is unlikely to achieve the kind of results won by organized labor. For this reason, some entities focus today on promoting alternative ownership structures to avoid pitting workers against each other and against capital owners.

For most workers in the developing world, a secure job remains only an aspiration. Should the pattern toward decentralized economic activity continue unabated, that possibility may diminish further, even as economies develop and workers enjoy greater earnings. That would fundamentally redefine our idea of what “middle-class” status represents.

Rethinking the role of government

The prospect of greater disruption, a diminished role for labor, and the decentralization of economic activity places a greater onus on the adequacy of safety nets. While safety nets are at different stages of maturity across different countries, those in the developing world have some advantages over those in rich nations in adapting to the needs of a changing global economy. These include programs being less fragmented and more harmonized, and avoiding the constraints that come when benefits are tied too closely to employment.

Compensation forms a central part of most public safety nets, which in its most comprehensive form means a universal basic income. The UBI concept has sparked excited debates in the West over the past year, so it is striking that the most extensive pilot, and one designed with rigorous evaluation in mind, is taking place in Kenya, and that the country arguably most poised to enact a UBI on a national scale is India. Other aspects of safety nets that require more innovative and effective solutions are those focused on retraining and relocating dislocated workers.

The redesign of social safety nets can be understood as one important component of a broader rebalancing required in the social contract between government and citizens, employers and employees, and the winners and losers of the changing global economy. The role of government is not just limited to helping those most vulnerable to change, but shaping change itself, whether by guiding the direction of technology through public investments in research and development, putting in place regulations that foster competitive markets and that are supportive of collective bargaining, and providing digital infrastructure for all.

Perhaps the most fundamental changes government can help usher in concern culture and norms that constrain worker opportunities in the modern global economy. This includes addressing discrimination in hiring and the workplace, removing the negative stigma associated with certain types of jobs, and altering how work is defined to more fully encapsulate all the productive activities by which people define themselves, such as bringing up children and civil participation.

Download the full report »

       

Introduction

The start of the 21st century has been characterized by both economic convergence and divergence.

Convergence occurs when poorer countries record faster growth in incomes than richer economies, thereby narrowing the gap in living standards between nations. This implies that a country’s rate of economic growth per capita is inversely correlated to its initial level of income. The blue line in Figure 1 shows evidence of this relationship across the world’s economies for the period 2000-2015.^[1] This represents a striking reversal of the pattern of the previous 200 years.^[2]

While evidence of convergence across the world’s economies on average is robust, some countries continue to be left behind, including many of the world’s poorest countries. The result is that, among the world’s low-income countries, living standards have diverged over the same period, with the poorest among them recording the most meager income growth on average. This is illustrated by the red line in Figure 1.

Figure 1: Convergence and divergence, 2000-2015

Note: Author’s calculations based on World Bank 2016. Country income groups refer to the World Bank classification in the year 2000. The negative correlation found among all economies is significant at the 1 percent level. The positive correlation found among all low-income countries is significant at the 5 percent level.

These two contrasting trends—convergence among the world’s economies, but divergence among its poorest—underpin some of the leading narratives in geopolitics and global development. The rapid rise of emerging economies has shaken up the world economic order; it has also spurred dramatic improvements in health, education, employment and access to energy, benefiting hundreds of millions of people. It has led to the emergence of a truly global middle class, representing nearly half of humanity.^[3] Meanwhile, some countries remain on the bottom rung of the development ladder, hardly budging from the levels of deprivation that prevailed in the poorest economies a generation ago. Included in this group are several fragile and conflict-affected states, whose spillover effects on others via an open and integrated global economy is seen as one of the world’s leading global risks.

The two trends are especially relevant to understanding global goals focused on the reduction of extreme poverty.

The first Millennium Development Goal sought to halve the rate of extreme poverty in the developing world between 1990 and 2015. That goal was achieved 7 years ahead of schedule—a result that is attributable to the improved growth performance of developing economies, led by China and India.^[4] On the latest count, the share of people living under $1.90 a day is down 74 percent since 1990, and the number of people below this threshold is down 1.2 billion, or 65 percent. The celebration of that goal serves as a tribute to the impact of convergence.

Despite that success, the record on global poverty reduction over the past quarter-century has not been unblemished. The number of people living in extreme poverty in fragile states has risen.^[5] The living standards of the very poorest people in the world—the “consumption floor”—has risen only meagerly.^[6]

The first Sustainable Development Goal (SDG)—to end poverty in all its forms everywhere— attempts to shift the focus to those who were left behind during the past 25 years by eradicating extreme poverty by 2030. Ending extreme poverty places demands on all countries where poverty exists today. (There are even some demands on countries that have already eliminated extreme poverty to stop poverty reemerging.) But the biggest challenge clearly rests with the world’s poorest countries that currently stand furthest from zero. In other words, to achieve the new goal, the pattern of divergence among the world’s poorest countries will almost certainly have to stop.

The World Bank is the organization responsible for monitoring global progress on poverty reduction. It has embraced the goal to end extreme poverty by enshrining it as one of two institutional goals. However, rather than adopting SDG 1 in a literal sense, it has instead set a target to lower the global poverty rate to 3 percent by 2030.

In explaining its figurative interpretation of the goal, the Bank makes an analogy between low levels of poverty and frictional unemployment:

 “It is…important to acknowledge that at any moment in time, some churning is likely to be taking place in which some people, possibly for reasons beyond their control, fall into poverty, even if only temporarily. It is difficult to imagine a world in which nobody at all is poor. For these reasons, it seems reasonable to view global poverty as having effectively ended even if some frictional poverty remains at a very low level. Hence, the global target is 3 percent or lower.”^[7]

This explanation is disingenuous on two levels.

First, whereas frictional unemployment is a necessary characteristic of functioning labor markets, there is nothing inevitable about people temporarily falling into poverty. Indeed, the Bank commits billions of dollars each year to support social protection programs in developing economies precisely so that households and communities that are vulnerable to shocks are able to withstand setbacks without falling into poverty.

Second, the attainment of a global poverty rate of three percent would almost certainly entail a much higher poverty rate in a minority of countries—counterbalanced by zero poverty in the majority of countries—where poverty could not reasonably be characterized as frictional or temporary. This means that the target could be met through ongoing convergence among the world’s economies while the very poorest countries continue to diverge and be left behind.

This is not to imply that the Bank’s 3 percent target lacks ambition. On the contrary, projections based on an extrapolation of recent country-level growth trajectories indicates a global poverty rate of around 5 percent in 2030.^[8] Meeting the Bank’s target would therefore require an improvement, on average, in the performance of developing economies—whether in terms of growth, its inclusivity, or both—relative to the already improved performance of developing economies since the start of the century. However, that would not preclude a minority of poor countries being excluded from these gains.

Thus, in the pursuit of global poverty reduction, a minority of countries risks being left behind both in terms of their levels of deprivation and in the way global progress is accounted for. This suggests that the “leave no-one behind” principle, originally intended to draw attention to marginalized groups such as lower castes or the disabled, might equally be applied to countries.

Identifying countries at risk of being left behind

Which countries are most at risk of being left behind in the fight against extreme poverty?

In an earlier publication we classified countries according to their poverty rate and their track record in poverty reduction over the preceding decade.^[9] Twenty-four countries combining high poverty rates and poor track records were identified as facing the greatest risk of failing to eliminate extreme poverty by 2030.

We adapt that approach here and make use of more recent data—several new household surveys and the latest International Comparison Program (ICP) round of international prices—to identify an up-to-date group of countries most at risk of being left behind.

We begin by taking the World Bank estimates of extreme poverty in each country in 2013—the most recent year for which a global poverty estimate is available—see Figure 2.^[10] We classify all countries with a poverty rate above 45 percent as being at risk. These countries would fail to reach a 3 percent rate of extreme poverty by 2030 even if they matched the speed of sustained poverty reduction achieved by the fastest performers on record: a reduction of 2.5 percentage points a year achieved by China and Vietnam in the 1990s and 2000s. It is striking that all these countries, other than Haiti, are located in sub-Saharan Africa.

Figure 2: Estimated poverty rates in 2013 ($1.90 PPP poverty line)

We add to this list those countries with a poverty rate above 20 percent that have recorded zero or meager progress in reducing poverty between 2002 and 2013, and between 2008 and 2013.^[11] We use both a shorter and longer timeframe to exclude countries that have achieved a recent turnaround in performance, such as Cote d’Ivoire.

Finally, we include countries that lack any household survey or valid international price data that we suspect would otherwise make the list. This generates a combined group of 30 countries that we judge as being most at risk of being left behind—see Table 1.

This approach is not beyond scrutiny. It relies on subjective judgment, artificial cut-offs and the continuation of past trends. Arguably its biggest weakness, however, is the weight it places on the reliability of poverty estimates. This could lead to errors of omission and commission.

First, the poverty estimates of countries that are included in our analysis in many cases rely excessively on extrapolation given the limited frequency of household surveys. This affects the accuracy both of countries’ poverty rates and their poverty reduction. Of the 22 countries we categorize as being most at risk of being left behind that have available poverty estimates, four have undertaken only one household survey in the 2000s. St. Lucia’s poverty estimates rely entirely on extrapolation from a survey in the 1990s. Of the 19 countries with poverty rates above 20 percent in 2013 that are not included in our group only on the basis of their satisfactory recent performance in poverty reduction, four countries have only a single household survey in the 2000s—see Table 2.

Second, even where countries have what appear to be a sufficient number of surveys, differences in survey design or implementation across surveys within the same country mean their results may not be strictly comparable. This is a particular problem for our assessment of each country’s track record in poverty reduction. A recent study by the World Bank assessed the comparability of household surveys according to their regional coverage, seasonality, reporting instruments, and recall period.^[12] Eleven of the 22 countries we categorize as most at risk of being left behind based on their poverty estimates do not have two comparable surveys since 1990.^[13] Of the 19 countries with poverty rates above 20 percent in 2013 that are not included in our group, a further seven lack two comparable surveys since 1990.

Table 1: Countries most at risk of being left behind

Table 2: Countries excluded from “left behind” group with poverty rate above 20 percent

There are several other problems with the reliability of poverty estimates that we do not elaborate on here, including differences in survey design and implementation between countries; the accuracy of domestic price data; and the accuracy of international price data.^[14]

Given our concerns with the reliability of poverty estimates, we also consider the performance of countries on non-monetary poverty measures. These measures can serve as a rough proxy for traditional poverty estimates and thus help in assessing their validity. They are also of intrinsic value given the multi-dimensional nature of poverty, which traditional monetary poverty measures don not capture.

A recent Brookings study assesses which countries are on or off track to meet some of the foremost Sustainable Development Goals for non-monetary aspects of poverty.^[15] As with our assessment of monetary poverty, the authors’ analysis for each goal is based on each country’s current distance from the target and their recent record of performance. They identify 37 countries that are off-track on each of four headline goals: child mortality (under-5 and neonatal), maternal mortality, access to drinking water, and access to sanitation.

Table 3 shows the status of the 30 countries we categorize as most at risk of being left behind on extreme poverty, on each of the four multi-dimensional goals. Twenty-one of the countries are off-track on all four goals. That lends some confidence to their inclusion in our group of at-risk countries. Five countries others are on-track on one goal. Another three countries, all of which have inadequate data for reliable poverty measurement, are on track on two goals.

Table 3: Trajectory of “left behind” countries on other goals (red = on-track; green = off-track)

Table 4 identifies the remaining 16 countries that are off-track on all four multi-dimensional goals but not included in our group of countries at risk of being left behind. Nine of these have reported monetary poverty rates above 20 percent and might reasonably be misclassified by our classification.

Table 4: Countries off-track on multidimensional goals that are not in “left behind” group

The preceding analysis highlights the difficulty of accurately identifying which countries are most of risk of being left behind on extreme poverty. Further analysis could be valuable in classifying countries with greater confidence.

What are the common characteristics of countries at risk of being left behind?

Identifying countries at risk of being left behind can serve a valuable purpose by drawing attention to them within the development community. A more ambitious objective is to understand what holds back the development of these countries, and to generate policy ideas that might raise their performance.

We begin here by simply identifying some common characteristics among the 30 countries we classify as being most at risk of being left behind—see Table 5.

Twenty-three of the 30 countries are in sub-Saharan Africa. The remaining seven are in the Caribbean (Haiti and St Lucia), the Middle East (Syria, Yemen), and South and East Asia (Afghanistan, North Korea, Myanmar).

Table 5: Common characteristics of “left behind” countries

Twenty-five of the 30 countries are relatively small, with populations below 30 million people. Each of these countries on its own is vulnerable to being overlooked by the World Bank in global poverty monitoring, given that the attainment of a 3 percent global poverty target is compatible with several small countries failing to make progress. (By contrast, a zero target would force attention on each and every country.) The Bank deserves credit for acknowledging this and stressing the moral imperative of reducing poverty in all countries.^[16] Collectively, however, the 30 countries can account for more than a third of people living under $1.90 a day in 2013, or 280 million people.

Eighteen of the 30 countries are considered fragile by the World Bank based on its more recent classification.^[17] This high share is consistent with prior analysis that has drawn attention to the rising share of global poverty that is located in fragile states.^[18] Nine countries are classified as resource rich by the IMF on the basis that more than 20 percent of their exports are nonrenewable commodities.^[19] Four of the 30 countries are both fragile and resource-rich.

An agenda for policy research

This note makes the case that, in the pursuit of global poverty reduction, a minority of countries risks being left behind both in terms of their levels of deprivation and in the way global progress is accounted for. Further research is needed to identify with greater confidence those countries at risk of being left behind, to understand the factors holding back their development, and to generate policy ideas that might ultimately help raise their performance.

       

“I think everyone in the room knows that this is a moment of extraordinary progress. Over the last 30 years, extreme poverty has been cut in half… When USAID was founded more than 50 years ago, donor governments were responsible for nearly three-quarters of all finance flowing to the developing world. Now we account for less than 10 percent.”

U.S. Agency for International Development Administrator Gayle Smith, March 2016¹

These remarks made by the head of the world’s largest bilateral aid agency capture two of the defining trends in global development of the past several years. Together, these trends imply a diminishing role and need for foreign aid.

In fragile states, however, neither of these trends is apparent. The stunning reduction in the number of people worldwide living under the extreme poverty line—from 1.8 billion in 1990 to 800 million in 2013—can be entirely attributed to stable countries. The total number of extremely poor people living in fragile states rose over the same period.² Similarly, international capital flows into fragile states look remarkably like the era that we are meant to have left behind. In 2012, the median fragile state still relied on aid for 50 percent of its foreign capital, the other half being made up of non-concessional loans, investment, and remittances. In other developing economies, aid represented only 10 percent.³

This is the crux of the argument behind calls to reallocate aid away from stable countries and toward fragile states: Unlike stable countries, fragile states continue to face urgent development needs and are at risk of being left behind. The U.K. government and the World Bank, the leading bilateral and multilateral donors in the global development community, have recently set targets to increase the share of their aid devoted to fragile states.

However, there remains a profound ambivalence about aid’s efficacy in these places. If development outcomes are driven by countries themselves rather than by outsiders, how much can aid realistically achieve in situations where partner governments are unable or unwilling to govern effectively? Even advocates of aid to fragile states are unlikely to argue that aid can dramatically improve how governments govern. Furthermore, aid-giving in fragile states seems especially vulnerable to causing unintended harm. One of the greatest criticisms of aid, forcefully made by Nobel Laureate Angus Deaton, is that it undermines the social contract that holds countries together by inserting donors into the relationship between states and citizens. In fragile states, a relationship that is already an Achilles heel risks being made weaker.⁴

These are valid concerns that cannot easily be argued away. Yet the evidence for aid’s impotence in fragile states is notably flimsy and has not stood the test of time. First, in the late 1990s and early 2000s a series of influential papers used cross-country growth regressions to show that the impact of aid was conditional on the quality of the receiving country’s policies and institutions—a damning result for fragile states where the quality of policies and institutions are lacking.⁵ But, in the following years, a much larger series of papers poked holes in the methods of this earlier research and discredited its results.⁶ Second, at the same time that these earlier findings were being taken seriously, the World Bank reported that the share of its projects in fragile states receiving unsatisfactory evaluations was double the share in stable countries. But when projects over the subsequent decade were evaluated, that difference vanished. The recent portfolio performance of the U.K. Department for International Development and the Global Fund in fragile states has been similarly favorable.⁷ As for aid’s capacity to unintentionally cause harm, some well-documented cases exist, as well as an emerging literature on aid’s incitement of conflict in certain conditions.⁸ But, to the extent that it can be measured, there is little evidence of any systematic negative effect of aid on recipient institutions.⁹

One upshot of these conflicting narratives is a focus on the operational aspects of aid giving. So long as donors are engaged in fragile states, they face a responsibility to ensure they are delivering aid as effectively as possible—a responsibility that is arguably more critical in fragile states, where aid’s usefulness is especially contested, than in other settings. Improving the operations of donors in fragile states offers a practical response to concerns regarding aid’s potential to do harm. It is these practices that are the focus of this paper.

Download the full report »

       

The election of Donald Trump demands a reevaluation of the future of globalization and our earlier optimism that the open global economic order will endure. It is time to consider the possibility that a single politician could reverse decades of global trends.

We published a short paper a month ago assessing the claim that globalization was on the verge of a retreat. Our conclusion was relatively sanguine. Based on an assessment of the movement of goods, money, and people across international borders, we found little evidence that globalization is already receding—see Figure 1. We also showed that the global economy is more integrated today than during the peak of the early 20th century, which we interpreted as a repudiation of the claim that globalization had reached unsustainable levels; we are skeptical that such levels exist in a literal sense. Looking to the future, we speculated that the years immediately ahead will be characterized either by stabilization in the level of globalization, or further integration but occurring at a more modest pace than in the past.

Figure 1: Globalization trends, 1870-2015

Authors’ calculations based on IMF 2015, Lane and Milesi-Ferretti 2013, Maddison 2001, the Maddison Project 2015, McKeown 2004, McKeown 2010, Riley 2009, U.N. 1999, U.N. 2015a, U.N. 2015b, UNCTAD 2015, U.S. Census Bureau 1975, World Bank 2015, World Bank 2016, and WTO 2016. Merchandise exports and foreign capital stock are expressed in market dollars as a share of global income expressed in international dollars, and will therefore differ with those cited elsewhere. Full description of sources and methodology are available in this report.

How, if at all, does the election of Donald Trump alter our views?

It is worth emphasizing that a resistance to globalization was arguably the foremost policy theme in Trump’s election campaign. In the speech announcing his presidential bid, Trump railed against the United States’ existing trade agreements, threatened to slap taxes on U.S. companies investing overseas, and pledged to build a wall to keep out migrants, whom he accused of being rapists. Trump’s plan for his first 100 days in office reaffirms the centrality of this theme, with a commitment to renegotiate or withdraw from NAFTA, abandon support for the Trans-Pacific Partnership (TPP), label China a currency manipulator, establish tariffs to discourage companies from off-shoring production and jobs, expel more than two million migrants, suspend immigration from terror-prone regions, and build the wall.

Let’s assume President-elect Trump succeeds in implementing this agenda. We see its effects on globalization playing out at three levels.

The first is the direct effect of the U.S. turning inward. The U.S. is the world’s largest economy, measured in market dollars, and its third most populated. A partial withdrawal from the global economy by the U.S. is therefore likely to register in measures of globalized stocks and flows, simply by virtue of the country’s size.

Indeed, America holds the largest share of global trade, foreign capital stocks, and migrants. Yet, relative to its size, America is not as globally integrated as many other countries. It is those areas where the U.S. is most globally intertwined, as measured by the vertical axes of Figure 2, that the direct impact of a Trump presidency on globalization could theoretically be greatest.

Figure 2: Global shares of trade, capital markets, and migration

Authors’ calculations based on Lane and Milesi-Ferretti 2014, U.N. 2015a, and World Bank 2016. Capital stock assets include estimates of external debt, foreign direct investment (FDI), and portfolio equity stocks. Full description of sources and methodology are available in this report.

For instance, despite the prominence of trade during the election campaign, the U.S. is a relatively closed economy. It accounts for only 11 percent of global trade volumes—far below its 24 percent share of global GDP. That said, this likely understates the country’s footprint in global trade given that the U.S. imports many final goods whose production occurs along international supply chains. By contrast, the U.S. plays a fuller role in global capital markets. Within those markets, the area that may be most directly vulnerable to Trump’s policies is outward foreign direct investment (FDI). The U.S. lays claim to 18 percent of global FDI assets. However, the U.S. is most globalized, relative to other countries, in terms of its openness towards migrants. The country is home to 19 percent of the world’s migrant stock, while accounting for only 4 percent of the world’s population. In fact, the U.S. is the top destination for migrants from 60 countries. The expulsion of large numbers of migrants and greater restrictions on the number of future entrants would directly alter this aspect of the global economy.

While the direct effects of the U.S. turning inward on global economic integration are important, they are still likely to be relatively small in terms of the three series presented in Figure 1. Much larger effects are possible in terms of the impact Trump’s policies could have by changing the behavior of other countries. This is the second level at which we see Trump’s effect on globalization unfolding.

We may see countries retaliate against U.S. protectionist policies. This is the basis for concerns that Trump could precipitate a trade war.

We may see countries retaliate against U.S. protectionist policies. This is the basis for concerns that Trump could precipitate a trade war. The threat has already been made explicit by the state-sponsored Chinese tabloid, Global Times, which proposed that China respond to aggressive trade policies by cancelling contracts with U.S. suppliers, imposing tariffs on U.S. imports, and limiting the number of Chinese students studying in American universities, and by French presidential candidate Nicolas Sarkozy who has suggested that the Europe Union impose a tax on U.S. products and limit the participation of foreign companies in EU public contracts if Trump withdraws from the Paris climate accord.

Countries and their leaders may instead imitate Trump’s policy agenda, whether in pursuit of similar electoral success or on the basis that his election gives his anti-globalization agenda legitimacy. In the past week, politicians from Italy, Hungary, Greece, and elsewhere have invoked Trump’s victory as justification for policies that reverse the pattern of globalization.

Alternatively, countries may repudiate global norms and institutions that underpin the globalized economy, if they feel that the U.S. is no longer committed to upholding the liberal economic order. This reflects the widely held belief that the stability of the existing economic order hinges on the example set by the U.S., as the longstanding global economic hegemon. Such an outcome foreshadows chaos, but with the era of U.S. global economic pre-eminence coming to a close, those norms and institutions, from World Trade Organization membership and rules, to the various U.N. conventions concerning the treatment of migrants and refugees, were always likely to be tested soon. Trump’s election may usher a more rushed transition as emerging markets that have benefited most from the open global economy come to its defense. There is already some evidence of this kind of realignment, as members of the TPP seek to patch together the deal without the U.S., and China makes the case for its alternative regional trade deals.

The third and final way in which we interpret the effect of Trump’s election on the future of globalization is the injection of a huge amount of uncertainty.

The breaking of globalization’s first wave a century ago is proof that the forces of global economic integration are neither irresistible nor irreversible.

The breaking of globalization’s first wave a century ago is proof that the forces of global economic integration are neither irresistible nor irreversible. Trump’s ascent to the White House adds to the evidence, representing the biggest shift in the U.S.’s orientation vis-à-vis the global economic system in the post-war period. This policy discontinuity is a source of uncertainty in and of itself. We have assumed that Trump will deliver on his anti-globalization agenda, yet for now it remains unclear whether he will pursue it in full and what constraints to its implementation will arise. Perhaps the most important risk concerns how he will respond to unanticipated events over the period of his presidency, through the prism of his anti-globalist perspective. This uncertainty alters the way we view the direct and indirect effects of Trump’s policies described earlier by reducing our confidence in them and expanding the range of possible outcomes.

Taking these considerations together, our view on globalization’s future has indeed changed. We are much less assured that the open global economic order will endure.

       

Cheng Li, senior fellow in Foreign Policy and director of the John L. Thornton China Center, talks about the rise of Chinese President Xi Jinping through the Chinese communist party leadership, which is the focus of his new book, “Chinese Politics in the Xi Jinping Era: Reassessing Collective Leadership.”

Also in this episode, Laurence Chandy, fellow in Global Economy and Development, examines how technology and globalization affect inequality.

Finally, Harsha Singh, executive director of the Brookings India Center, discusses his career, Brookings India, and current events in India.

Show Notes:

Hong Kong’s elections: Testing democratic reform in China

The end of collective leadership in China? Not really

Are technology and globalization destined to drive up inequality?

11 Global Debates

Brookings: @ Brookings Event Audio

Thanks to audio producer Gaston Reboredo and producer Vanessa Sauter, and also thanks for additional support from Eric Abalahin, Jessica Pavone, Nawal Atallah, Basseem Maleki, and Rebecca Viser.

Subscribe to Brookings podcasts here or on iTunes, send feedback email to BCP@Brookings.edu, and follow us and tweet us at @policypodcasts on Twitter.

BCP is part of the Brookings Podcast Network.

       

1.1 What’s the Issue?

Over the past several years, concerns that technology and globalization lead to ever greater inequality have reached fever pitch in the U.S. and beyond. To understand what’s behind this anxiety, three distinctions are useful.

First is to distinguish global inequality and its two components: inequality within countries and inequality between countries. Global inequality, as popularized by economist Branko Milanovic, looks at the distribution of income between all the world’s citizens irrespective of country borders. Inequality by this measure is exceptionally high. Over the past generation, between-country disparities fell, due to the fast growth of emerging economies, even while inequality within several countries has risen. The net effect has been a small reduction in recorded global inequality (Lakner & Milanovic, 2015). This pattern will continue if poor countries such as India continue to quickly converge on income levels prevailing in high-income countries and this convergence outweighs any widening of within-country distributions (Hellebrandt & Mauro, 2015). Yet that would not quiet grievances about inequality. On the contrary, the middle class in industrialized economies, one of the world’s most vocal and powerful constituencies, has seen global growth benefit high earners in their economies along with the expanding middle class in emerging economies, while their own incomes have stagnated. Their sense of being shortchanged is increasingly recognized as a source of political instability.

Since politics is organized principally around the nation state, it is the level and change in inequality within countries that is the most potent source of tension and debate.

This brings us to distinction two: inequality in developed versus developing economies. In the former, the trend is clear—nearly all developed economies have seen inequality rise over the past generation. In Anglophone countries, rising inequality has been especially pronounced at the top end of the distribution, with the top 1 percent of earners seeing their share of national income rise. In developing countries, on average, inequality rose in the 1990s but stabilized in the 2000s (Ravallion, 2014). In most developing economies where recent data exist, inequality is trending downward (World Bank, 2015). However, information about the top end of the distribution in developing economies is limited, given the absence of complete tax records.

In developing countries, on average, inequality rose in the 1990s but stabilized in the 2000s.

Distinction three is between inequality in market income and disposable income. Until now we have described the inequality of disposable income, net of the effects of government taxes and benefits, which serve to reduce the inequality of market outcomes. This redistributive effect tends to be greater in developed countries than in developing countries, where government is typically a smaller share of the economy. In most advanced economies, redistribution through taxes and benefits grew over the past generation, offsetting some but not all of the increase in market inequality. However, these effects have diminished on average since the late 1990s, due to policy choices such as the application of more stringent criteria to government benefits (OECD, 2011). Public policies can also shape the distribution of market income. For instance, weakened employment protection, such as rules regarding sick leave and severance pay, has contributed to widening inequality over this period.

1.2 What’s the Debate?

Debates over the causes of inequality are fraught, reflecting the multiple and complex channels through which technology and globalization are changing the global economy.

Arguably the most prominent effect of technology on inequality is through the increased premium it places on skills. Modern technology substitutes for many of the jobs and tasks traditionally performed by unskilled workers, while acting as a complement to skilled workers. In advanced countries, trade reinforces this effect by encouraging specialization in high-skill sectors in which those economies have a comparative advantage. The same logic should see income inequality narrow in developing economies that specialize in low-skill sectors. However, in practice, skilled workers in developing economies may take those jobs, so that distributions widen (Maskin, 2015).

By substituting for unskilled workers, technology has not only increased the premium on skills, but increased the role of capital in production. Historically the share of income that accrues to workers relative to capital owners was stable, but since the 1980s, it has declined across most countries and industries as technology has made capital goods ever cheaper (Karabarbounis & Neiman, 2013). This adds to inequality, as capital ownership is especially unequal and generates large investment income for many of the same individuals already earning high wages (Atkinson & Lakner, 2013).

Arguably the most prominent effect of technology on inequality is through the increased premium it places on skills.

Technology has often led to the creation of strongly monopolistic markets for new goods and services. This is especially apparent in the digital economy, where behemoths like Google and Apple dominate. Globalization has expanded the scale of these winner-takes-all markets, enabling vast salaries and profits to be shared among a narrow set of employees and shareholders.

At the same time, globalization and technology have served to lower market barriers and information costs. For instance, while digital platforms for taxis (Uber), retail (Amazon), and accommodation (AirBnB) are themselves quasi-monopolies, they have simultaneously lowered barriers to entry for self-employed drivers, sellers, and would-be hoteliers, creating highly contestable markets. This has redistributed rents and generated new income-earning opportunities for the unskilled.

Evidence is emerging of the hollowing out of labor markets in developing economies.

Finally, globalization has encouraged a race to the bottom on some regulations and redistributive policies, as the mobility of firms, investment, and skilled workers compels governments to match the conditions of their competitors so as to retain and attract business (Bertola & Lo Prete, 2008).

1.3 What to Watch Out For?

The effects of technology and globalization on inequality are neither inevitable nor entirely predictable. We identify three areas to watch closely:

Job automation. The past year has seen a rapid uptick in sales of robots, coinciding with breakthroughs in the capability of machines and artificial intelligence in increasingly complex, non-routine tasks such as driverless vehicles and semi-cognitive skills such as voice-recognition. This has led to growing anxiety over the prospect of widespread automation of jobs. Estimates on the share of jobs that are at risk of automation over the medium term vary from 9 to 47 percent for OECD economies (Frey & Osborne, 2013; Arntz et al, 2016). Equally uncertain are what, and how many, new jobs may emerge and the adjustment costs of moving lots of workers into new roles.

Prospects for developing economies. The replacement of workers by machines poses a threat to developing economies’ traditional comparative advantage in global markets—their surfeit of cheap labor. Evidence is emerging of the hollowing out of labor markets in developing economies, mirroring the pattern already observed in the west, and of premature deindustrialization as developing economies struggle to establish a manufacturing base, in stark contrast to the path taken by western economies and Asia’s tiger economies (World Bank, 2016; Rodrik, 2015). At the same time, the digital economy provides opportunities to link workers in poor economies with companies and customers in rich markets, thus offering a temporary reprieve from the risks associated with labor-saving technologies (Basu, 2016). It is unclear which of these two effects will win out in shaping developing economies’ fortunes in the near term. But the rate of their convergence with rich economy living standards is set to be a major determinant of global inequality trends.

Perceptions of inequality. Public anger over the inequitable effects of technology and globalization is cited as a cause of myriad social ills—from rising nationalism and identity politics, to disdain for institutions, and a fracturing of the rules-based international system. Whether that anger persists will depend less on any objective measure of inequality than on how inequality is perceived and managed (Nieheus, 2014). One important factor is the way global integration shifts the reference points people use to judge and compare their lives.

Policy has a vital role to play in promoting greater equality, both through redistribution, where taxes and benefits moderate the unequal distribution of market income into a more equitable distribution of disposable income, and “pre-distribution” where market forces and rules are engineered to improve the distribution of market income itself.

Given the alarming trends in inequality, and the tendency for political stalemate over changes in tax and benefits, attention is increasingly focused on policies that support pre-distribution. Some of the most creative ideas seek to reshape the forces of technology and globalization themselves. For instance, policies can be put in place to incentivize research and development on innovations that generate more jobs. Alternatively, governments can deploy public funds to acquire stakes in technological innovations and their commercialization so that the profits they generate can be shared with citizens rather than benefit only a narrow group of shareholders. With regard to globalization, multilateral efforts can eliminate tax inversions, whereby one corporation acquires another to re-domicile to a lower-tax jurisdiction.

More generally, there can be little doubt that focusing almost exclusively on average incomes and their growth has been a disservice to policymaking and to the economics profession. A growth strategy that doesn’t work for all members of an economy is incomplete and unsustainable, no matter how much redistribution there may be. The definition of economic success must therefore include the extent to which growth is inclusive. Inclusiveness cannot be an afterthought.

       

After two decades defined by growing integration, the global economy appears to be at an inflection point. This judgment has been prompted both by structural changes in the global economy, especially since the Great Recession, and political events over the past year illustrative of a backlash against past integration. Following one such event—the U.K.’s Brexit vote in June—The Economist magazine reported, in a funereal tone, that globalization now seems to be receding, inspiring comparisons with the rise and fall of globalization a century ago.¹

Globalization’s first wave, which lasted from 1870 to 1914, is viewed today as the embodiment of the liberal open economic paradigm. This period saw the spread of international trade, built on the exchange of Western manufactures for developing economies’ primary commodities along low-tariff corridors. The production of those commodities was financed mainly by the West and supported by a stable global exchange rate regime so that capital flowed freely from where it was plentiful to where it was scarce and could reap the highest returns. For instance, around half of all British savings were channeled abroad over this period, while half of Argentina’s entire capital stock was foreign owned by 1914. Migration was virtually unrestricted, with the exodus of Europeans across the Atlantic mirrored by even larger flows of laborers and merchants through North, South, and East Asia.

The retreat from globalization’s first wave was decisive and ruinous. The passage of the Smoot-Hawley Tariff act in the U.S. against the backdrop of the Great Depression set off a retaliatory backlash among the world’s major economies that crippled global trade. A moratorium on the repayment of war debts and reparations saw a drying up of private international lending. Many Western economies imposed controls on the exports of capital, while a number of developing economies defaulted on their liabilities. Economic hardship triggered anti-immigrant sentiment and governments responded by imposing drastic restrictions on newcomers for the first time.

The breaking of globalization’s first wave is proof that the forces of global economic integration are neither irresistible nor irreversible.

The breaking of globalization’s first wave is proof that the forces of global economic integration are neither irresistible nor irreversible. But can we be sure that we are at another turning point?

The essence of globalization is the movement of goods, money, and people across international borders. In this brief, we develop three series that capture this phenomenon over the past 150 years (Figure 1).

Those proclaiming globalization’s demise fall into two camps. Some argue that the process is already underway. Others argue that globalization has reached unsustainable levels as it did a century ago and so a backlash is inevitable.² Our three series allow us to verify whether globalization’s retreat has begun or not, and to gauge the degree of today’s global economic integration measured against the peak of the early 20th century.

Figure 1.

Authors’ calculations based on IMF 2015, Lane and Milesi-Ferretti 2013, Maddison 2001, the Maddison Project 2015, McKeown 2004, McKeown 2010, Riley 2009, U.N. 1999, U.N. 2015a, U.N. 2015b, UNCTAD 2015, U.S. Census Bureau 1975, World Bank 2015, World Bank 2016, and WTO 2016. Merchandise exports and foreign capital stock are expressed in market dollars as a share of global income expressed in international dollars, and will therefore differ with those cited elsewhere.

We find that, while the process of economic integration has slowed, there is only limited evidence so far of an absolute decline. In addition, today’s level of integration matches or exceeds the heights of globalization’s first wave. This could equally imply that globalization has reached unsustainable levels or that no such levels exist.

Generating any global data series over a century and a half is a complicated task. Few data series have complete coverage over space, incorporating all the world’s countries, which raises challenges about representativeness and scaling. Even fewer series have complete coverage over time, so incomplete ones have to be spliced together, which raises issues of comparability. Our results therefore come with numerous, often significant, caveats.

Two specific caveats should be highlighted from the outset—both of which are necessitated by the availability of data and the objective of maximizing comparability within each series. First, one of our three series (goods) is a measure of flows, whereas the other two (money and people) are measures of stocks. The latter provide moderately lagging indicators of globalization’s trajectory. Second, our series for goods and money are expressed in terms that differ from contemporary measures cited in economic reports and the media. Care is therefore required when interpreting the levels reported in our three series.

While these caveats impose limits on what can be reliably inferred from direct comparisons of the two waves of globalization, the process of assembling the three data series is itself illuminating. Each series can be understood as the aggregation of multiple underlying trends whose trajectories vary. Analyzing these trajectories helps explain the factors that are shaping globalization today and provides insights into its likely future.

In the following section, we look at the three series in turn. In each case, we begin by explaining how the series was generated and describing its caveats. We then assess the process of recent integration and the evidence that we are at an inflection point.

Goods

Our series for the international movement of goods is the most straightforward of the three to develop. Historical estimates of the dollar value of global merchandise exports for select years are provided by Angus Maddison. These are spliced together with annual estimates drawn from the World Trade Organization’s contemporary dataset of trade flows beginning in 1960, which covers the vast majority of the world’s countries.³ The series is weighted by global gross domestic product (GDP), expressed in international dollars.⁴

This last point warrants further elaboration. By presenting export values expressed in market dollars as a share of global income expressed in international dollars, we generate percentages that are out of step with those reported in standard analyses of trade. This unorthodox approach follows that of Maddison and ensures consistency and comparability across historical and contemporary estimates.

At the end of the first wave of globalization in 1913, merchandise exports peaked at 7.9 percent of global GDP.⁵ That peak was surpassed as early as 1970, when tariff reductions under the General Agreement on Tariffs and Trade were still at an early stage, the standardization of shipping containers was being established, and the rise of export manufactures from the developing world had yet to occur. Goods exports have since been propelled to far greater heights, reaching 19.7 percent of global income in 2008.

That share stood at 15.1 percent in 2015, having declined continuously for the previous four years. Underpinning this decline is a slowdown—not a reversal—in the growth of merchandise trade volumes, combined with low commodity prices and a stronger dollar.

The recent slump in trade growth has not gone unnoticed and has been extensively researched. Both cyclical and structural factors are at work.

The recent slump in trade growth has not gone unnoticed and has been extensively researched. Both cyclical and structural factors are at work. The latter include the exhaustion of gains from both the incorporation of previously-closed economies into global markets and the fragmentation of value chains across borders—commonly referred to as the second unbundling. When cyclical factors reverse, many analysts expect trade growth to resume and the merchandise exports’ share of GDP to stabilize.⁶

The potential for future trade growth in services, which are excluded from our long-term series, is further reason to doubt that trade’s role in the global economy is bound to diminish. While trade in services is considerably smaller than trade in goods,⁷ its value as a share of global GDP has doubled in the past 30 years and has proved more resilient during the recent trade slowdown, standing today at record levels (Figure 2).⁸

Concerns regarding the sustainability of global trade have been heightened by the inability of policymakers to advance multilateral and regional trade agreements. Some advocates of those agreements insist that closing new deals is crucial to unlocking trade growth, especially in services, where the largest barriers remain. In reality, modern trade agreements have become increasingly concerned with how the spoils of trade are distributed rather than with the volume of trade itself. Further trade integration will likely depend as much on technology as policy.

Figure 2.

Authors’ calculations based on the Maddison Project 2015, World Bank 2016, and WTO 2016. Merchandise and service exports are expressed in market dollars as a share of global income expressed in international dollars, and will therefore differ with those cited elsewhere.

A more serious concern for trade growth than the failure to agree new trade agreements is the trend towards creeping protectionism using non-tariff barriers. In the absence of improved global growth, that trend could worsen. This risk is real, but for now remains speculative. Although it is possible to envision circumstances under which trade’s share continues to decline, there is more reason to be optimistic as cyclical factors reverse and service exports rise.

Money

Our series for the international movement of money again starts with Maddison. He provides historical estimates of the value of the foreign capital stock in the developing world for a selection of years, based on an assessment of countries’ gross external liabilities.⁹ Contemporary annual estimates from 1990 onward are built at the country level from independent estimates of external debt, foreign direct investment (FDI), and portfolio equity stocks. These are drawn from a number of independent sources: external debt from the World Bank; FDI from United Nations Conference on Trade and Development; and portfolio equity from the “External Wealth of Nations” dataset developed by Philip Lane and Gian Maria Milesi-Ferretti, and supplemented in the most recent years from the IMF.¹⁰ In keeping with the approach for merchandise exports, the series is weighted by developing world GDP, expressed in international dollars. The levels of international capital reported will therefore differ with those cited elsewhere.

Four caveats are immediately worth noting. First, the group of countries that constitute the developing world is held constant across the series in accordance with Maddison’s historical estimates.¹¹ Thus, the series does not attempt to control for rising income levels and includes a small number of countries that are today classified as high-income. Sixteen percent of the group’s GDP is accounted for by high-income countries in 2014. Second, the developing world aggregates are assembled independently for each type of capital based on the available country coverage of each dataset. Of the three types of capital, external debt suffers the most from incomplete coverage.¹² Third, estimates of FDI and portfolio stocks do not accurately capture the divergent performance—and value—of these assets over time.¹³ Fourth, we exclude both remittances and foreign assistance in the form of grants—both important sources of foreign capital for the poorest economies—from our measure to ensure consistency with Maddison’s historical estimates.

Our focus on foreign capital in the developing world only is compelled by the availability of historical estimates, but has the advantage of concentrating our attention on the availability of money in areas where it is relatively scarce. The capital flowing from Western Europe to the New World and the colonies was crucial to the first wave of globalization, and similar dynamics are at play today.¹⁴

During the first wave of globalization, the developing world’s foreign capital stock peaked at 32.4 percent of GDP in 1914. The series is honing in on that peak today. Despite falling precipitously during the Great Recession, the foreign capital stock share has since rebounded, and at 30.1 percent, is at its highest level in a century.

Underlying the impressive recent growth of the foreign capital stock are changes in the composition of that stock. In Figure 3, we look beyond the total foreign capital stock share and trace the trajectory of the different types of capital separately.

FDI’s rise is a defining component of globalization’s second wave and is synonymous with the growing role of international finance beyond traditional areas such as railways and extractive industries into new sectors including commerce and industry. The value of FDI stock as a share of developing world income has risen almost continuously for the past 25 years, although the rate of growth has diminished since 2012, reflecting lower FDI inflows.¹⁵

The growth of foreign-held portfolio equity since 1990 is similarly pronounced but limited to more mature developing economies; the foreign-held portfolio stock as a share of low-income country GDP remains negligible at only 0.4 percent in 2014. In contrast to FDI, foreign-owned portfolio investment in the developing world was hit hard by the Great Recession and has yet to return to its pre-crisis peak.

While foreign capital in the form of FDI and portfolio equity have become an increasingly prominent part of many developing economies, the role of external debt has markedly diminished. This is especially the case for low-income countries where the debt stock as a share of GDP has fallen by two-thirds since 1990. Two factors lie behind this: debt relief, and the nurturing of domestic debt markets through the issuance of debt denominated in domestic currency and the encouragement of purchases by local financial institutions. While in a literal sense these factors reduce the extent of global economic integration and interdependency, they are rightly perceived as a mark of progress in reducing unnecessary exposure and risk.

Concerns about the de-globalization of foreign capital since the Great Recession have centered on a different phenomenon: the reduction in inter-bank lending.¹⁶ Its causes are believed to include regulatory changes, weakened bank balance sheets, and reduced access to wholesale funding. The fall has been largely limited to banks in advanced economies and so is not captured by our series. While this is certainly an example of international capital becoming less prominent, it should not overshadow the increasing role of foreign capital in other parts of the world.

Based on our review, the clearest evidence we have of global capital being in retreat is limited to examples of globalization’s excesses being curbed. Elsewhere, the data suggest that the pace of global integration has slowed but not reversed.

Figure 3.

Authors’ calculations based on IMF 2015, Lane and Milesi-Ferretti 2013, Maddison 2001, UNCTAD 2015, and World Bank 2015. Debt, FDI, and portfolio equity stock are expressed in market dollars as a share of global income expressed in international dollars, and will therefore differ with those cited elsewhere.

Migration

Our series for the international movement of people is the most problematic of the three to assemble.

Contemporary data on migration since 1960 are compiled by the U.N. and built from estimates of migrant stocks from national censuses. These estimates are notoriously weak, with censuses differing not just in terms of quality, but in their definition of a migrant. Additionally, estimates likely suffer from downward bias, as restrictions on immigration create incentives for migrants to avoid being counted.

Historical records of migration introduce a different set of problems. Flows—as opposed to stocks—of long-distance migration along certain routes are well documented from ports and customs statistics. Other routes, especially those overland, are not formally recorded. Accounts of short-distance migration are missing altogether for most of the world. Adam McKeown provides the most comprehensive and well-documented estimates of historical migration flows at a global level between 1850 and 1940. His estimates, which are provided for each five-year period, are based solely on known instances of international travel and include some long-distance travel within the same colonial jurisdiction which is not strictly cross-border.

To establish our series, we attempt to convert McKeown’s historical estimates of global migratory flows into stocks so that these can then be compared with contemporary U.N. estimates. This conversion adds together consecutive estimates of migrant inflows adjusting for the age at which migrants migrate, their life expectancy, and their return rate, which we obtain using the best available benchmarks taken from the period. Our benchmark for age at migration is the median age at entry of migrants to the U.S. each year, drawn from official U.S. records. Our benchmark for migrant’s life expectancy is U.S. life expectancy for the respective age, gender, and year of U.S. migrants, obtained again from U.S. records, and discounted by the ratio of U.S. life expectancy at birth to global life expectancy at birth.¹⁷ Our benchmark for the migratory return rate is drawn directly from McKeown who reports separate return rates for Chinese migrants, Indian migrants, and European migrants to the U.S.¹⁸

Our resulting series of the global migrant stock, which is marked by a gap between 1940 and 1960 when no records of global migration are available, provides an admittedly rough but illustrative account of the movement of people during the past 150 years. At the end of globalization’s first wave in 1914, the global migrant stock is estimated to have peaked at 2.5 percent of the global population, although it stayed close to this level into the 1930s before beginning its descent. The share began to climb again in the 1970s and surpassed its previous peak for the first time in 1988. It rose further in the 2000s and today stands at 3.3 percent. Given the possibility that our historical estimates of the global migrant stock are underestimated, a judicious assessment is that the share of global migrants today has only recently exceeded the levels reached a century ago.¹⁹

While contemporary migration at a global level may be comparable to that of a century ago in terms of proportional size, migratory patterns differ vastly across countries and regions. Migrants today come from an increasingly diverse set of origin countries, but over the last half century have concentrated in a shrinking pool of prime destination countries. Migration has therefore “globalized” from the perspective of destination countries but not from that of origin countries.²⁰

Two popular destinations for migrants—the U.S. and the U.K.—have extensive migration records that allow us to track the size of the migrant stock since the 19th century, and compare across globalization’s two waves, without resorting to statistical manipulations (Figure 4).

The U.S. migrant stock hovered at close to 14 percent of its population throughout globalization’s first wave before falling precipitously. That share has rebounded since 1970 and in 2015 stands a fraction short of its historical high at 13 percent. As a result, the U.S. is the top destination for migrants from some 60 sending countries and is home to one-fifth of the world’s migrants. By contrast, the U.K., which was an important source of migrants in globalization’s first wave, has only emerged as a significant destination for migrants in the post-war era. Its migrant share has risen especially sharply in the last decade and now stands at almost exactly the same level as the U.S.

Much commentary on global migration reinforces the notion that migration is growing unsustainably, driven by formidable emigration pressure from source countries, and that anti-migrant rhetoric will soon lead to anti-migrant policies.²¹ While our series shows that the share of the global population accounted for by migrants is at an all-time high, it does not corroborate this broader narrative.

Figure 4.

Based on preliminary estimates for 2015, growth of the global migratory stock as a share of the world population has slowed over the past five-year period, especially in the West. Moreover, while the global migrant stock is still rising, it would be wrong to assume this reflects ever increasing flows of migrants. Estimates of migratory flows derived from the U.N.’s migratory stock data suggest that both the absolute number and global share of migration flows has remained relatively stable in the past two decades.²² That means the increasing migrant stock must instead be explained by migrants’ characteristics: a lower age of migration, longer life expectancy for migrants, and/or lower return rates. As for the demand to migrate, modelling exercises up to 2025 forecast diminishing demand from Latin America and Asia, while emigration pressure from Africa will rise.²³

This gap between rhetoric and reality is in part because the most visible migrant flows are often not the largest. Refugees, for example, accounted for just 8 percent of the global migrant stock in 2015.²⁴

Of course, the likelihood of a clampdown on migration depends on perceptions of migratory trends—especially those at a national level—not their reality. If recent history is any guide, however, governments are more likely to strengthen their selectivity of migrants than impose absolute restrictions. Over the past 25 years, migration policies have become more discerning of migrants’ skills, class, and nationality while simultaneously allowing for a greater volume of migrants overall.²⁵ If the political backlash against migration ultimately focuses on restricting small but highly visible flows, then we should not expect global migration as a whole to decline.

Conclusion

Table 1 summarizes the results from our three series, comparing the degree of global integration across globalization’s two waves. On two of the three measures, the degree of globalization is continuing to rise based on the most recent available data, contrary to the claim that globalization is receding. And on two of our three measures, the global economy is more globalized today than during the peak of the early 20th century. Whether that implies globalization has reached unsustainable levels, or that no such levels exist, remains to be seen.

Table 1: Benchmarking globalization across two waves

Authors’ calculations based on IMF 2015, Lane and Milesi-Ferretti 2013, Maddison 2001, the Maddison Project 2015, McKeown 2004, McKeown 2010, Riley 2009, U.N. 1999, U.N. 2015a, U.N. 2015b, UNCTAD 2015, U.S. Census Bureau 1975, World Bank 2015, World Bank 2016, and WTO 2016. Merchandise exports and foreign capital stock are expressed in market dollars as a share of global income expressed in international dollars, and will therefore differ with those cited elsewhere.

As global economic integration continues, the composition of globalized stocks and flows evolves. The shifts in trade growth from goods to services, in foreign capital away from debt to FDI and portfolio equity, and in migration towards a more diverse set of origin countries concentrated in a shrinking pool of prime destination countries, represent defining features of the modern global economy.

While globalization is not yet in retreat, it seems likely that globalization is at an inflection point. The past two decades have been described as an era of hyper-globalization during which both the level and rate of global integration was judged as having intensified. Based on our analysis, it is reasonable to conclude that this period is over. The debate then is what will emerge next. Our speculative assessment is that the coming years will be characterized either by stabilization in the level of globalization, or further growth in the degree of integration but at a more modest pace than in the past.

       

“If we think about the progress of getting people out of extreme poverty, it is really impressive. But it is actually a much slower trend then what we have seen in the spread of digital technology. The speed with which mobile phone ownership has spread around the world, access to bank accounts, biometric identification cards, people getting online – these trends are happening even faster. We are seeing 100-300 million people each year getting access to a phone or biometric ID for the first time. These trends in getting people digitally connected. . .are progressing at such speed that they’re starting to reach some of the poorest people in the world. Digital technology is changing what it means to be poor because it’s bringing poor people out of the margins.”
– Laurence Chandy

“The role of governments will continue to be central in improving education. At the end of the day, they’re the only ones who have the duty and the mandate to care about the poorest of the poor. But they will more and more have to partner with organizations from the private sector, the philanthropic community and the non-profit community to try to reach the most marginalized kids for education.  Governments can set an enabling environment that lets these sets of actors bring their creativity, their new way of approaching intractable problems into a space where they’re given the ability to scale up their work.” – Rebecca Winthrop

In this week’s episode of “Intersections,” Laurence Chandy, fellow in the Global Economy and Development program, and Rebecca Winthrop, senior fellow and director of the Center for Universal Education, discuss progress toward meeting the Sustainable Development Goals of ending poverty and achieving education for all, and how digital technologies can be harnessed in that pursuit. Also, Winthrop and Chandy addressed the tools needed to reach the last 10%—those most marginalized.
Show Notes

What Works in Girls’ Education: Evidence for the World’s Best Investment
Millions learning: Scaling up quality education in developing countries
Connecting the poor is the best hope for ending poverty
Disrupting development with digital technologies

With thanks to audio engineer and producer Zack Kulzer, Carisa Nietsche, Sara Abdel-Rahim,  Brionne Smith, Eric Abalahin, Fred Dews and Richard Fawal.

Subscribe to the Intersections on iTunes, and send feedback email to intersections@brookings.edu.