88462 INEQUALITY in FOCUS Ending Extreme Poverty and Promoting Shared Prosperity Could There Be Trade-o s Between These Two Goals? Kathleen Beegle, Pedro Olinto, Carlos Sobrado, Hiroki Uematsu, and Yeon Soo Kim, Raj Chetty: On Increasing I with Maximillian Ashwill Opportunities and n April 2013, the Joint Ministerial Committee of the Boards of Governors of the World Bank Group and the International Improving Income Mobility Monetary Fund endorsed two ambitious goals to galvanize What keeps people at the bottom of the income ladder from climbing up international development efforts by donors, governments and and out of poverty? the international development community, generally. These goals Raj Chetty, an award-winning are (1) to end extreme poverty (de ned as reducing the number economist and Harvard professor, of people earning $1.25 per day or less to no more than 3 percent uses an innovative blend of empirics globally) by 2030, and (2) to and theory, with help from big data, promote “shared prosperity” to produce work with important by boosting the incomes of implications for global inequality, the poorest 40 percent of the including showing which societal conditions allow people to population in every country. more easily climb the income ladder. This added focus on shared His work on the importance of U.S. teachers—and how the prosperity marks a major quality of teaching at all levels of education can substantially shift in the World Bank improve the outcomes and opportunities for students—has Group’s1 mission, which been cited by President Barack Obama, among others. In an interview with Inequality in Focus, Chetty discusses in the late 1940s was to these topics and others and the feasibility of adapting them to rebuild Europe, but evolved a broader, international context. | Page 7. in the 1970s toward poverty reduction. Ending extreme poverty has remained the sidered a necessary condition for poverty reduction. The growth primary focus of the World elasticity of poverty reduction—the percentage reduction in Bank Group for the past few poverty rates associated with a percentage change in mean (per decades, but the introduc- capita) income—is likely not high enough in many countries to tion of the goal to promote Photo by Dorte Verner achieve the goal of ending extreme poverty by 2030.2 Focusing shared prosperity introduces Could a focus on shared prosper- efforts to raise incomes for the lower part of the income distribu- ity turn global attention away a second primary focus. The inclusion of this tion, which is what shared prosperity does by focusing on the from the extremely poor? bottom 40 percent, can increase the growth elasticity of poverty second goal, to promote shared prosperity, has some important implications for global reduction. The implication, then, is that the shared prosperity development efforts. Notably, this new goal is a re nement of a goal is instrumental for reducing extreme poverty. However, to longer-standing, implicit focus on growth, which is widely con- our knowledge, it has never been stated clearly whether pro- moting shared prosperity is a secondary tool to reduce extreme poverty, or if the two goals are prioritized equally. In fact, the The authors would like to acknowledge the helpful peer review comments discussion around extreme poverty and shared prosperity—and of Francisco Ferreira, Peter Lanjouw and Martin Ravallion. The authors how these concepts should be interpreted and expected to inter- also thank Joao Pedro Wagner De Azevedo, Shaohua Chen, Jose Cuesta, act with one another—seems to be far from settled.3 Carolina Diaz-Bonnila, Leonardo Lucchetti, Rinku Murgai, David New- This raises some important questions. Could the shared prosperity house, and Minh Cong Nguyen for their suggestions on the draft and the goal compete with the extreme poverty goal over the allocation of poverty pro le calculations. development efforts and resources? Considering that the extremely Poverty Reduction and Equity Department :: www.worldbank.org/inequalityinfocus :: Volume 3, Number 1 :: May 2014 Figure 1 percent of income earners in each developing Distribution of the extreme poor, non-poor bottom 40 percent, and neither poor nor country. We rely on three sources of data. The bottom 40 percent based on data from 2010. The area of the figure represents the rst is the PovcalNet database,4 an online pov- total population of the developing world (5.9 billion people). Countries are listed along erty measurement tool maintained by the World the horizontal axis in descending order from the highest extreme poverty rate on the left to the lowest on the right. A vertical bar represents each individual developing Bank Group’s Development Research Group. country used in this sample, with the width of the bar corresponding to the country's The second are the World Development Indica- population, so, for example, India and China have the widest bars. tors,5 or the WDI, the World Bank Group’s primary collection of development indicators. The third is the International Income Distribu- tion Database or I2D2,6 a new dataset recently created by the World Bank Group. The I2D2 is a globally harmonized database that draws on more than 600 nationally representative house- hold surveys. By examining these datasets we can create pro les of ve relevant groups. These groups, visualized in gure 1, are: 1. The red group: The extremely poor (those who earn less than $1.25 per day). As gure 1 shows, in some countries (to the left) this population is much larger than 40 percent and in other countries (to the right) this population is much smaller than 40 percent. Before the addition of the shared prosperity goal, this was the World Bank Group’s main Source: PovcalNet and WDI. target population. poor are not distributed evenly across countries, how will the 2. The bottom 40 percenters: The bottom 40 percent of in- new mission alter the global development agenda? Within coun- come earners in each developing country, which can include tries, what are the implications? Could this weaken policy makers’ both extremely poor individuals and those who are not emphasis on ending extreme poverty? In other words, are there extremely poor, depending on the country (these people are trade-offs between boosting the incomes of the bottom 40 percent in located below the dotted line in gure 1). every developing country and ending extreme poverty globally? Our analysis suggests that there might indeed be trade-offs. Figure 2 To better understand this, we must rst answer the following two Composition of the extremely poor (red group), bottom 40 percent- ers, non-poor bottom 40 percent (blue group), and the new target questions: (1) How much importance will be placed on each goal? population (red group plus blue group) in the world in 2010. (2) Are the extremely poor and the bottom 40 percent of income earners different? If so, how? If the two populations are identical, and the bottom 40 percent are also extremely poor, then the rst question becomes irrelevant. In this case, boosting the incomes of the bottom 40 percent is equivalent to ghting poverty, and the two goals are well aligned. There is little or no trade-off. However, if the two populations are different, it then becomes crucial to specify how much weight will be placed on each of these two goals. For instance, will gains in shared prosperity be con- sidered a substitute for making progress toward ending extreme poverty? Or is promoting shared prosperity simply a secondary instrument to help reduce extreme poverty, which would remain the ultimate goal? These questions will need to be addressed if developing country governments and international donors are to ef ciently allocate resources aimed at achieving these goals. The rst question—how much importance will be placed on each goal?—is impossible to answer with data. In this article, therefore, we will take a look at the second question—are the two populations indeed different? We do this by comparing the characteristics of the world’s extremely poor to the bottom 40 Source: PovcalNet and WDI. 2 :: May 2014 :: Inequality in Focus Photo by Dorte Verner In UMICs (Upper Middle Income Countries without China), the introduction of the shared prosperity goal could shift the World Bank Group’s target population away from rural areas—and the agriculture sector—toward urban areas and likely the service and manufacturing sectors. 3. The blue group: These are individuals who are not extremely correspond to the World Bank Group's latest income classi ca- poor (earn more than $1.25 per day) but fall within the bot- tion from July 1, 2013.7 In this classi cation, countries were tom 40 percent. With the inclusion of shared prosperity this de ned as LIC if their per capita annual Gross National Income blue group represents people who were not previously a part (per capita GNI) in 2012 was below $1,035 per year. LMICs are of the main target population for the World Bank Group and countries in which the per capita GNI is between $1,036 and its partners, but will be now. $4,085. UMICs, as de ned here, are countries in which the per 4. The new target population: This group combines the ex- capita GNI is more than $4,086. tremely poor (the red group)—traditionally the World Bank Group’s main target group, with the bottom 40 percent in every Where in the world are the extremely poor (red group) developing country who are not poor (the blue group)—the and the bottom 40 percenters? World Bank Group’s additional target group. With the introduc- How will the introduction of the new shared prosperity goal shift tion of shared prosperity these two groups merge to become the the World Bank Group’s target population across country income World Bank Group’s new target population. classes? Figure 2 shows the distribution by country income class of 5. The green group: This group encompasses the top 60 percent the world’s (a) 1.2 billion extremely poor people (red group), (b) of income earners in each developing country who are not 2.4 billion bottom 40 percenters, (c) 1.3 billion bottom 40 percent- extremely poor. This does not include the individuals who be- ers who are not extremely poor (blue group), and (d) 2.5 billion long to the top 60 percent of the income distribution but who people of the “new target population” (red group plus blue group). are also extremely poor. These individuals are in red but above As can be seen, the introduction of the shared prosperity goal shifts the 40 percent dotted line, toward the left side of gure 1. the distribution of the target population away from LICs and India In this article we use the term “extremely poor” to refer toward UMICs and China. The weight of LMICs remains the same to those living with an income of less than $1.25 per day as at 22 percent. We can also see that while 62 percent of the world’s measured in 2005 Purchasing Power Parity (PPP) dollars. We extremely poor (red group) live in LICs and India, only 11 percent also separate developing countries into one of the following ve of the blue group live in those countries. On the other hand, while categories of country income classes: India, China, Low Income only 17 percent of the red group live in UMICs and China, 67 per- Countries (LICs), Lower Middle Income Countries without India cent of the blue group live in those countries. As a result, compared (LMICs), and Upper Middle Income Countries without China to when ending extreme poverty was the only goal, the new target (UMICs). The UMIC class also includes developing countries that population of 2.5 billion is distributed more or less evenly across have recently graduated to High Income status, such as Chile, the four top country income classes, each hosting approximately 20 Poland, Russia, and Uruguay. The LIC and LMIC income classes percent, with the LICs hosting only 16 percent. Inequality in Focus :: May 2014 :: 3 Figure 3 Figure 5 Share of agricultural Share workers of agricultural among workers amongworkers workers who are that are poor, poor, non-poor School-aged children School-aged out of children school out among of school children among thatare who children arepoor, poor, non-poor bottom bottom 40%40%, and neither and neither poor norpoor bottom nor40%, bottom 40%, income by country non-poor non-poor bottom 40%, and bottom neither 40% poor poor and neither nor bottom 40%, nor bottom by by 40%, by country income class. class country income class. country income class 80% 50% 46% 47% 68% 68% 43% 44% 70% 62% 45% 41% 40% 59% 58% 60% 56% 54% 53% 40% 35% 49% 50% 45% 35% 40% 40% 35% 27% 31% 30% 25% 30% 23% 25% 20% 12% 20% 10% 15% 11% 0% 8% Low income Lower middle income Upper middle income China India 10% 6% 5% Extremely poor Non-poor bottom 40% Neither poor nor bottom 40% 0% (red group) (blue group) (green group) Low income Lower middle income Upper middle income India Source: I2D2. Extremely poor Non-poor bottom 40% Neither poor nor bottom 40% (red group) (blue group) (green group) Figure 4 Source: I2D2. Share of children among the poor, non-poor bottom 40%, Share of children among the poor, non-poor bottom 40% and poor nor and neither neither bottom poor 40%,40%, nor bottom by country income by country class. income class comparatively older than the old target population. Figure 5 50% 43% shows the proportion of school-aged children (under 13) that 45% 40% 38% 35% 33% 35% 33% are not in school. It can be seen that there are more children out 35% 32% 30% 30% 24% 28% of school from the red group than there are from the blue group, 25% 21% 20% 17% 17% except in LICs. These results suggest that, at least in terms of 14% 15% 10% 9% enrollment, children’s education is a more pressing issue among 5% the extremely poor than it is for the blue group. Therefore, 0% Low income Lower middle Upper middle China India promoting shared prosperity may diminish the previous atten- income income tion received by children from the international development Extremely poor Non-poor bottom 40% Neither poor nor bottom 40% (red group) (blue group) (green group) community. Source: I2D2. Do the bottom 40 percent who are not extremely poor How important is agriculture for the extremely poor (blue group) have better access to public utilities than (red group) and the bottom 40 percenters who are not the extremely poor (red group)? extremely poor (blue group)? Figure 6 shows the difference in access to electricity between the Figure 3 shows the importance of agriculture for the red group, red group and the green/blue groups. In both LICs and LMICs, the the blue group, and the green group. As can be seen, agriculture red group has much less access to electricity than do the non-poor is an important sector for the majority of the red group and the (blue and green groups). For UMICs and China, the two popula- blue group in all country income classes, except for UMICs. tions have similar access. Thus, for the new target population, In these countries, less than half of the red group, and slightly access to electricity is less of an issue than it is for the extremely more than one-third of the blue group work in farming. More poor. This would likely hold true for other basic public utilities, important is the difference between the red and blue groups such as access to drinking water and proper sanitation. within UMICs. Compared with the red group, approximately 30 percent fewer of the blue group work in agriculture. Therefore, The political economy of shared prosperity: Could the in UMICs, the introduction of the shared prosperity goal shifts political willingness to end poverty wane? the target population away from rural areas—and the agriculture Looking again at gure 1, we see the distribution across countries sector—toward urban areas and likely the service and manufac- of 1.2 billion extremely poor people (red group), 1.3 billion non- turing sectors. poor people in the bottom 40 percent (blue group), and those 3.4 billion people who are neither extremely poor nor of the bottom 40 Does promoting shared prosperity turn the focus away percent (green group). We can see in the gure that the blue area is from children? about the same size as the red area. This means that the introduction It is hard to say with certainty. What can be said is that chil- of the shared prosperity goal will (approximately) double the World dren represent a smaller share of the additional target popula- Bank Group’s target population. Also, with anticipated population tion (blue group) than they did of the old target population growth and sustained poverty reduction in the developing world, (the red group). Figure 4 shows that, for every country income it can be expected that the blue group will continue to grow in class, there is a higher share of children under 13 in the red relation to the red group. Could this diminish the political will of group than in the blue group. Therefore, the new target popula- governments to end extreme poverty? Will advocacy for the poor be tion that is created by introducing the shared prosperity goal is weakened in countries where the poor are a small minority? 4 :: May 2014 :: Inequality in Focus Children represent a large share of the world's extremely poor population. Photo by Dorte Verner The political willingness of governments to end extreme poverty Figure 6 is unlikely to be impacted by the new goal in low-income/high- Access to electricity among population that is poor, non-poor poverty countries such as Nigeria and Bangladesh, where more bottom 40%, Access to neither and poor electricity nor among that isby bottom 40%, population country poor, income non-poor than 40 percent of the population is extremely poor. In those class. bottom 40% and neither poor nor bottom 40%, by country income class countries and others like them, promoting shared prosperity may 100% 88% 93% 100% 99% 99% 88% 86% 82% in fact refocus attention on the poorest segments of the popula- 80% 78% 74% 63% tion because the bottom 40 percenters are often poorer than the 60% 50% average poor person. In India, where approximately 33 percent of 40% 30% 22% the population lives in extreme poverty, the additional focus on the 20% 6% bottom 40 percenters might slightly dilute the focus on the poor, 0% Low income Lower middle income Upper middle income China India but probably not by much. Extremely poor Non-poor bottom 40% Neither poor nor bottom 40% The story is different for China and the UMICs such as Brazil, (red group) (blue group) (green group) Mexico, and Russia. In those countries, the introduction of the Source: I2D2. shared prosperity goal substantially expands the World Bank Group’s target population. Because the extremely poor are a very and the international development community names ending ex- small, and shrinking, minority in these countries, a refocusing of treme poverty as the higher priority, there might be a shift in focus the development dialogue away from ending poverty and toward away from LICs and India, where the majority of people in the red boosting the income of the bottom 40 percent may weaken the group currently live, toward UMICs and China, where the majority political willingness to end extreme poverty. Only 4 percent of of people in the blue group currently reside. the world’s extremely poor people reside in UMICs, but they still In terms of strategy within country income classes, we nd represent a large share of the population—approximately 45 mil- that in LICs and LMICs, the extremely poor (red group) and the lion people. Would a greater focus on shared prosperity mean that bottom 40 percenters are very much the same people. Therefore, those people can be more easily ignored? at least in the short run in those countries, the introduction of the Concluding remarks shared prosperity goal will not be likely to shift attention away For the rst time in more than 30 years, the World Bank Group from the reduction of extreme poverty. In these countries, what has enhanced its mission from ending extreme poverty to also is good for shared prosperity is also good for ending poverty and raising the incomes of the bottom 40 percent in every developing vice versa. country. Our analysis indicates that unless the World Bank Group In UMICs, however, the World Bank Group and the interna- Inequality in Focus :: May 2014 :: 5 tional community need to be cautious in their dialogues with 6 For this article and all I2D2-related gures we use data from client governments when they introduce the goal of shared a sample of countries from each income class, but not from all prosperity. If there is a sense that gains in the income of the countries. For more details about the I2D2 dataset, which is bottom 40 percent are as important as poverty reduction, then not yet public, refer to Olinto, Pedro, Kathleen Beegle, Carlos focusing on the extremely poor (and children) in these countries Sobrado, and Hiroki Uematsu (2013). “The State of the Poor: may become less important. The needs, location, and livelihoods Where Are The Poor, Where Is Extreme Poverty Harder to End, of the red group can be very different from those of the blue and What Is the Current Pro le of the World’s Poor?” World Bank group. As a consequence, by targeting the blue group in addition Group—Economic Premise 125 (2013): 1–8. The table below to solely the red group, the focus and energy spent on the red summarizes the coverage in terms of number of countries, total group might abate. population, and total poor population in the statistics presented In order to accelerate the rate of poverty reduction to a level in Figures 3 through 6 based on the I2D2. In each gure, we that ensures that the goal to end extreme poverty is achieved by selected the latest available survey for all available countries in 2030, the World Bank Group and international development the I2D2, with a cut-off year of 2000. For example, we use a total community may need to be explicit in how they prioritize the of 64 surveys in the I2D2 to calculate the share of workers in two goals. If it is made clear that gains in shared prosperity will agriculture in gure 3. These countries explain 83 percent of the be celebrated only if we are on track to end extreme poverty total population and 89 percent of the total poor population as of by 2030, then the addition to the World Bank Group’s target 2010. population of 1.3 billion people from the blue group, should not detract from the efforts to lift out of poverty the 1.2 billion Coverage* Figure 3 Figure 4 Figure 5 Figure 6 Total** people currently living on less than $1.25 per day. However, if Poor 1,032 1,061 897 1,030 1,165 an increase in the incomes of the bottom 40 percent in every population country is seen as a substitute for progress in reducing extreme in millions poverty, then the introduction of the second goal might work (% of total) 89% 91% 77% 88% 100% against the world’s poorest people. Population 4,653 4,838 3,390 4,507 5,574 in millions Notes 1 The World Bank Group consists of ve international orga- (% of 83% 87% 61% 81% 100% total) nizations, including: the International Bank for Reconstruction and Development (IBRD), the International Finance Corporation Number of 64 71 68 51 115 (IFC), the International Development Association (IDA), the countries International Centre for Settlement of Investment Disputes (IC- (% of 56% 62% 59% 44% 100% SID), and the Multilateral Investment Guarantee Agency (MIGA). total) 2 Yoshida, Nobuo, Hiroki Uematsu, and Carlos E. Sobrado Median 2006.5 2006 2007 2007 2009 (2014). Is Extreme Poverty Going to End? An Analytical Framework survey year to Evaluate Progress in Ending Extreme Poverty, Number 6740, The * Coverage is calculated assuming that each survey in I2D2 is nationally World Bank Group, Washington, DC. representative. 3 For example, in early discussions about the inclusion of the ** Based on countries included in 2010 poverty estimates in PovcalNet for shared prosperity goal, there was some agreement that the two which the survey year is not earlier than 2000. goals would be lexicographically ordered. This means that not only does extreme poverty have a much larger weight than share 7 The complete list of countries according to the latest World prosperity; but that no gains in shared prosperity should be al- Bank Group classi cations is available at: http://data.worldbank. lowed to compensate for a lack of progress in poverty reduction. org/about/country-classi cations/a-short-history. This implies that the marginal dollar of development resources is applied only to shared prosperity once it is assured that the goal The views expressed in this article are solely those of the authors and to reduce extreme poverty will be achieved. do not necessarily re ect the views or positions of the World Bank 4 The PovcalNet database can be accessed at http://iresearch. Group or peer reviewers. The authors are solely responsible for any worldbank.org/PovcalNet/index.htm. errors or omissions. 5 Population data used in gures 1 and 2 are obtained from WDI (http://data.worldbank.org/data-catalog/world-develop- ment-indicators) using wbopendata command in Stata (http:// data.worldbank.org/ developers/apps/wbopendata). Note that we use the latest population data available in WDI as of the time of data collection (March 1, 2014), which results in small discrep- ancies between our estimates and those in PovcalNet in terms of poor population and total population. 6 :: May 2014 :: Inequality in Focus Photo of Raj Chetty courtesy of the John D. & Catherine T. MacArthur Foundation. An Interview with Professor Raj Chetty Increasing Opportunities and Improving Income Mobility for the Poor R aj Chetty, the Bloomberg Professor of Economics at Harvard innovative blend of empirics and theory, and he has often tapped into large University, is producing work with very strong practical administrative data sets, or big data, for his work. and theoretical implications for global inequality, including Inequality in Focus: In your “Equality of Opportunity” research,1 the widely publicized Equality of Opportunity Project and you nd that upward income mobility (that is, the capability for those research on the long-term impacts of teachers that was cited in lower income classes to move up the economic ladder) varies sub- by U.S. President Barack Obama in his 2012 State of the Union address. stantially within the United States [see gure 1], and the areas with the Members of the Inequality in Focus staff recently interviewed Chetty, who greatest upward income mobility tend to have ve characteristics: less at 34 has already won two of the three most prestigious awards in econom- segregation, less income inequality, better schools, greater social capital, ics, including the MacArthur Fellowship, better known as the “genius grant,” and more two-parent families. Interestingly, you show, rst, that a larger in 2012, and the John Bates Clark Medal in 2013. Earlier, the Economist middle class has a greater positive in uence on upward mobility than magazine recognized him as one of the world’s eight brightest young econo- does having a lower concentration of wealth in the top 1 percent of mists. Chetty received his PhD from Harvard at 23, and at 29 became one incomes. Second, that racially segregated areas have lower economic of the youngest tenured professors in Harvard’s history. His research is an mobility for all races (whites as well as blacks or Hispanics) than non- Inequality in Focus :: May 2014 :: 7 Figure 1 This map shows the average percentile rank of children who grow up in below- median income families across areas of the United States (absolute upward mobility). Lighter colors represent areas where children from low-income families are more likely to move up in the income distribution. Source: www.equality-of-opportunity.org. segregated areas. Third, that areas with more two-parent-households analogy maybe explains why the size of the middle class matters. I can’t have higher upward mobility but having two parents is not the reason say we know for sure if that’s the explanation, but it would be consis- for this. Could you explain the causal mechanisms behind these three tent with the data. ndings? The second factor you discussed is segregation. There, I think, the Raj Chetty: As you know, it is hard to gure out exactly what is sociologists have produced many intuitive theories that would explain causing what because all of these characteristics are correlated with why segregation would be correlated with mobility. One example is many other omitted factors. But I can speculate to their causes based peer effects. If you’re around other high skilled, or high-achieving on some theories that have been discussed in prior literature. peers, you will also have role models, social networks, and connec- One is the idea that the size of the middle class, and the amount of tions that may improve your aspirations and help you see what is inequality, really affects people’s prospects for upward mobility. The feasible in your own life. When people are cut off from such peers, the visual analogy that people sometimes give is to think about climbing community evolves very differently. Another possible mechanism for the income ladder: If a lot of the rungs of the ladder are missing in the why segregation might matter is the funding of public goods; if we all middle of the distribution, and the space between the lowest rungs of live in an area where there are both high- and low-income people, the the bottom percentiles and the highest rungs of the top percentiles is low-income people are going to bene t from the better schools and very large, intuitively it is harder to climb up. This makes sense because the property tax payments of the high-income families, but if your there must be medium-skilled labor positions available for people community is very separated you might not bene t from those public with a high school education in order to give that family a prospect for goods. Also, when we say “segregation” we mean not just segregation upward mobility. However, if the only jobs available are those of high- by race but also by income. So places that are racially segregated also skilled engineers and low-skilled service workers there is much less tend to be more segregated by income, meaning the low- and high- scope for moving up. That type of explanation would also explain, as income people are not living in proximity to each other. This would you noted, why we nd a much weaker correlation between extreme explain why segregation tends to negatively affect both white and black upper tail inequality and upward mobility. To take an example from people, because regardless of race, poor people in segregated areas are California, San Jose and the Bay Area have a large middle class and a lot less likely to see the public good or peer-effect bene ts from a more of tech billionaires. In these cases, the existence of wealthy tech billion- integrated community. aires does not seem to be holding everyone else back from moving up Finally, a third comment on the two-parent-households. Theories in and there is still a high level of upward income mobility. So, the ladder sociology suggest that family structure or the strength of the commu- 8 :: May 2014 :: Inequality in Focus Figure 2 This figure shows that when a high value-added (top 5%) teacher enters a school, the end-of-school-year test scores in the grade he or she teaches rise immediately. High VA teacher School-grade-cohort mean test score arrives Year relative to entry of high value-added teacher Source: http://obs.rc.fas.harvard.edu/chetty/value_added.html. nity might matter. A key caveat here is that the fraction of two-parent that it is not valuable to improve the quality of schools for all, because households in an area matters even if your own parents are married, if we improve everyone’s outcomes, and bring some people above the which is to say that having more single-parent families in your neigh- poverty line, it is, at least in my view, a good thing even if inequality is borhood is correlated with poor outcomes for kids even if their own not reduced. parents are married. This suggests that it’s not the direct mechanism In terms of externalities, we have to remember that an important of your own parents being married that matters but something about part of kids doing well—especially the more af uent—comes back the community; the idea that more stable communities produce better to society as a bene t. This bene t can be in the form of tax dollars, im- outcomes for kids. This could also be consistent with other factors, for portant innovations, or job creation. We don’t have direct evidence that example, maybe places with more two-parent households tend to have precisely shows this, but it would go along with the idea that better better schools or lower teen birth rates. But again, this is hard for us to teachers improve kids’ outcomes through a variety of dimensions and disentangle. will have externalities that bene t both high- and low-income families. Inequality in Focus: Your research2 shows that students assigned to To sum, I do think that there is a strong argument for improving the higher-value-added teachers (a teacher’s value-added rating is de ned quality of education for all and providing public subsidies to increase as the average test-score gain for his or her students) are more likely to the value across the distribution. attend college, earn higher salaries, live in better neighborhoods, save Inequality in Focus: But some of the externalities you mention, more for retirement, and are less likely to have children as teenagers for example, innovation from high-income kids who receive a sub- [see gure 2]. You also show that replacing a teacher whose value- sidy, may not be that common. Isn’t it more likely that these more added rating is in the bottom 5 percent of the value-added distribution af uent kids will end up going to law school or medical school, in with one of average quality would generate cumulative earnings of which case wouldn’t these bene ts then be internalized? $80,000 per student, or more than $1.4 million for the average class- Raj Chetty: Bene ts are certainly partly internalized but I don’t room. But, your results also show that the impact from better teachers think they are entirely internalized. As I was saying, in the United is greater for richer kids than for poorer kids. If so, wouldn’t providing States, 30 to 40 percent of your income is going to taxes, so by free, high-quality education to all increase inequality, since by subsidiz- de nition that’s a scal externality on the government that students ing richer parents they can make complementary investments in their would not be internalizing. But also there is a question, and actual- kids’ education, via tutors, extracurricular activities and so on? Or, ly something we are doing some work on now: To what extent are do you think that there are enough positive externalities to justify the the returns to innovation captured by the individual as opposed public nancing of education for rich kids? to captured by other members of society? Certainly things like Raj Chetty: First of all, I agree to some extent with the points you research are bene ts that largely accrue to others. So you are right, make about the complementarity of public and private education. In for some things like law and medicine they would be internalized, levels, that’s exactly what we nd in the data; the high-income kids, in but this isn’t universally true. dollar terms, actually bene t more from better teachers so it does look Inequality in Focus: Currently, policy makers seem to be consistent with complementarity between inputs at home and inputs at emphasizing the importance of early-childhood education while school. So when you just improve the quality of schools for everyone, paying less attention to later education. This certainly seems to be it’s not necessarily true that you will reduce inequality. That is not to say the case at the World Bank. In your work, however, you show that Inequality in Focus :: May 2014 :: 9 the quality of teachers has a signi cant impact on all grades and macro issues that are going to be harder to study. Maybe by building not just on the youngest children. up from micro models we will have better macro models as well to Raj Chetty: Let me rst say that I think efforts to invest in early tackle those questions, but I still think that it’s going to be some time childhood education are valuable because I do think early childhood before we have a better empirical sense of those issues. education matters. Still, I strongly believe, after having seen more and Inequality in Focus: A lot of your work has very strong implications more evidence, that there are signi cant returns to improving the qual- for U.S. national policy but how applicable is it for other countries? ity of education much later—well beyond ages 5, 10 or 15— to even Raj Chetty: One of the approaches we try to take with these data when people are in college, and by no means should we be giving up sets is to extract lessons that are not only going to apply to the speci c on kids once they reach 5 or 8 years old. There is good evidence that context we are studying, but will have broader applicability. For you continue to improve kids’ outcomes in the long run if you improve instance, take the issue we were discussing about teachers mattering education at all age levels. This is extremely important when structur- not just at young ages, but at older ages. I intuitively think of that as a ing policy because it means that, while investments in early childhood pattern that is likely to generalize. If it’s true in the U.S., it is likely to be education are valuable, we should continue investing and trying to true in other developed countries, and perhaps in developing countries improve the quality of education at all points in time. as well. I think there is more low-hanging fruit in developing countries, Inequality in Focus: Much of your work uses large administrative meaning that policy solutions are perhaps much simpler and more data sets or big data. For example, in The Equality of Opportunity powerful than in the developed world and can have larger returns.For Project, you compile statistics from millions of anonymous earnings example, in the United States there is a complex debate on how to im- records. So far, such big data has been instrumental in providing clarity prove the quality of the teachers—if we should use value-added mea- to microeconomic questions, but because of limitations on what can surements of teaching or other methods, if we should train teachers, be tested it provides less clarity on macroeconomic questions. Do you and so on. In many developing countries, the issue is much simpler. think there is a role for big data, rst, in economics generally, and, sec- In India, for example, a problem is that teachers often do not show up ond, in providing clarity to long-standing macroeconomic questions, for work, so the policy response is not a sophisticated intervention to like how to overcome a recession or manage interest rates? improve the quality of teaching; it is simply getting teachers to show Raj Chetty: I think big data is going to play a fundamental role in up. We demonstrate, using data from the United States, that teachers economics because it allows us to approximate experiments, which is matter and it is likely that teachers matter everywhere. the hallmark of success in science. By having enormous data sets we While it is true that in poor countries the problems are bigger and are basically able to nd experiments in the data from which we can inequality is greater because you are farther away from the frontier in study, learn, and generalize. You already see this to a large extent in terms of growth, it is also true that even if you can’t close that huge gap, publications from the top economic journals today—70 or 80 percent a 1 percent improvement in India, in a sense is much more important of the empirical papers are using large administrative data sets. And the than a 1 percent improvement in the U.S., because there is so much era of survey-based research, at least in developed countries, is ending; more to be done. there are simply fewer and fewer people writing papers with those data I think, while recognizing the magnitude of the problem in devel- sets. In developing countries, survey data continue to be very impor- oping countries, it is important to recognize that making even some tant because those are the best data we have, but even there with the progress in these countries is incredibly valuable given from where you advent of technology, like mobile phones and information technology, I are starting. think we’ll start to have more large administrative data sources that will Inequality in Focus: In a recent presentation3 you gave at the World be useful. Bank, you questioned whether the United States can still be hailed The question of how well big data can answer macroeconomic as the “land of opportunity.” You show evidence to the contrary by questions is dif cult to answer because the types of methods that we comparing mobility between the U.S. and Denmark. But Denmark has are getting good at developing—causal experimental methods—lend a much narrower income distribution than the United States, which themselves well to cases where you can manipulate things at the micro makes it easier for someone in a low-income percentile to move to a level. Now, those micro level interventions can aggregate up to have big higher-income percentile. Wouldn’t this explain why you would likely macro impacts. For instance, changes in social safety nets, or changes see greater income mobility in Denmark? in unemployment bene t policies, do have important macroeconomic Raj Chetty: Yes, absolutely. Let me restate what you just said be- effects, but their nature is that they can be analyzed at the micro level. cause I agree with it: In Denmark it takes less to move across percen- Things like a recession, interest rates, or the level of in ation, are tiles of the income distribution in absolute dollars because, as you said, harder to manipulate at the micro level, which then makes them harder the income distribution is more compressed. So going from, say, the to analyze. So, in my view, it’s not a question of what important macro- 25th to 75th percentile is a smaller dollar movement in Denmark than economic questions big data can help answer, it is more a question of it is in the United States. This is precisely why I think the comparison what macro questions lend themselves to being studied with big data within the U.S., like what we focused on, is preferable to cross-country at a microeconomic level. My sense is that we are going to have within analyses that previous work has focused on. When we compare the next 10 to 15 years a much better understanding empirically and mobility across countries there are many different factors to take into not just theoretically of how these things work. Thereby we will be able account, which includes not only the income distribution, but also na- to build better macro models and have more agreement. But still, there tional levels of equality and the different institutions that exist in each are going to be other questions that are purely general-equilibrium country. So I think it is harder to learn across countries than it is from 10 :: May 2014 :: Inequality in Focus within the United States where we are able to hold more things xed. To give another anecdote, I have many cousins—who actually have If you wanted to more accurately compare the two countries, or any the same name as me; we were all named Nadarajan after my grand- others, you would need to look at global ranks. You could synthetically father—and it so happens that, while they are all living ful lling lives, combine Denmark and the United States’ income distribution and hold we had very different outcomes. One cousin is doing dif cult labor in the ranks xed. My sense is that you will nd less of a difference in mo- a temple in Singapore; others are working in manufacturing jobs. Very bility between the two countries when you do that, but I think you will different outcomes. still nd greater mobility within Denmark. I don’t know exactly how it I trace one of the reasons for this back one generation to our parents. would break down. It so happens that my mom was the one person in her family who Inequality in Focus: To conclude, we wanted to ask a more per- received a higher education and ended up going to medical school. sonal question. Your father was an economist and you accomplished Same thing with my dad. At that time in India, my grandparents did and contributed a lot in the eld of economics at a very young age, not have money to invest in the educations of all of their kids.  You so we wanted to know, rst of all, how your upbringing in uenced see the lasting impacts of those differences among the grandchildren your personal trajectory, and, secondly, did it in uence some of your and even the next generation. To me, this illustrates the importance of research ideas? opportunities. My cousins and I were quite similar to begin with. The Raj Chetty: One of the great things about this research is seeing how fact that we ended up in really different places further motivates me to it ts with my own introspection. Both of my parents are academics; understand why this happens and what we can do to give all children my dad is an economist, as you mention, and my mom is a physician who does research. I learned a lot from being around that environment an equal opportunity to succeed. and it also made me interested in research. So I think family played a This interview took place on March 7, 2014, between Professor Raj Chetty huge role for me, but so did my teachers. In high school, for instance, I and the Inequality in Focus editorial team: Pedro Olinto, Maximillian Ash- remember having a teacher who said to me “you are going to be a great will, Julie Barbet-Gros, and Fernanda Luchine. The interview was recorded economist someday” and I remember at that point really wanting to do and transcribed by Fernanda Luchine. The above dialogue is an interpreta- biomedical science. I wasn’t thinking about doing economics at all, and tion of the conversation written by Maximillian Ashwill and Pedro Olinto. I thought “oh well, there is no way I am going to do that.” Sometimes Notes it seems like your teachers know you better than you know yourself. Besides teachers, I think environment also plays a huge role and luck to 1 The Equality of Opportunity Project can be found online at some extent. I am very lucky to have had a great background and a lot www.equality-of-opportunity.org. of people investing in my coming up. 2 This research, “The Long-Term Impact of Teachers: Teacher Although I obviously had no sense of how test-score-based, value- Value-Added and Student Outcomes in Adulthood,” can be found added measures were going to capture things in my research or what online at http://obs.rc.fas.harvard.edu/chetty/value_added.html. patterns would emerge in the data, my background certainly contrib- 3 This presentation can be viewed online at http://live.world- uted to my interest in studying these questions. bank.org/improving-equality-opportunity. Inequality in Focus :: May 2014 :: 11 The Inequality in Focus series aims at informing the public debate on equity, inequality of opportunity, and socioeco- nomic mobility. It features articles written by World Bank Group staff, as well as researchers and policy makers from the broad development community. The views and interpretations in the articles are those of the authors and do not necessarily represent the views of the World Bank Group, its Executive Directors, or the countries they represent. The Inequality in Focus series is not copyrighted and may be reproduced with appropriate source attribution. Executive Editor: Pedro Olinto Managing Editor and Staff Writer: Maximillian Ashwill Copy Editing and Design: Mary Anne Mulligan Research Staff: Fernanda Luchine Ishihara and Julie Barbet-Gros Production: Anna Reva Web Designer: Maura K. Leary Sector Manager: Christina Malmberg Calvo Poverty  Reduction  and  Equity  Department   Poverty  Reduction  and  Economic  Management  Network  (PREM)