Back to Top

Fair progress? Economic mobility across generations around the world

by Ambar Narayan, Roy Van der Weide, Alexandru Cojocaru, Christoph Lakner, Silvia Redaelli, Daniel Gerszon Mahler, Rakesh Gupta N. Ramasubbaiah, and Stefan Thewissen

Published version (free)

Commentary by Jishnu Das

Should I stop kidding myself that learning leads to educational opportunity?

We want to make sure that children learn. We agree that schools aren’t delivering. But will policies to improve children’s test scores also enhance equity and opportunity in an increasingly unequal world?

I worry that they won’t even come close.

Think about it this way. Learning is a goal in itself; it is hard to participate with dignity in a globalised world without knowing how to read or write. But that’s not the only goal we are selling; we are also arguing that “learning is a ladder to opportunity” – what holds the poor back is the low quality of schools they now attend in increasingly higher numbers.

And in case you have any doubt that the poor are indeed held back, the World Bank’s recent report (2018) on education mobility makes for sober reading:

Mobility from the bottom half of the education ladder to the top quartile has fallen over time in developing economies, whereas persistence at the bottom has increased. In the median developing economy for the 1980s generation, less than 15 per cent born into the bottom half make it to the top quarter, while more than two-thirds stay in the bottom half […] These concerns are heightened for Africa and South Asia, where the prospects of children are still tied to the socioeconomic status of their parents more closely than in any other developing region, which suggests that relative mobility in these two regions will continue to be low in the near future.

In the U.S., test scores level the playing field …

One explanation for low mobility is that children from disadvantaged backgrounds are not learning in school and this prevents them from accessing the higher levels of education that children from more privileged backgrounds regularly reach. The lack of mobility in low-income countries tracks back to differences in test scores so that fixing schools will help build a fairer society.

This is indeed what we see in the U.S. Carneiro and Heckman (2003) showed that once you control for test scores, the difference in college attendance by socioeconomic status vanishes. Rather than the picture on the left, where academic ability has little to do with years of schooling, the U.S. looks like the picture on the right. Those from disadvantaged backgrounds still get fewer years of schooling, but these differences can be fully traced back to their academic performance. This powerful finding immediately puts school quality at the centre of all debates about social mobility; in the U.S., learning is indeed a (maybe the) ladder to opportunity.

Figure 1: Societies can have high returns to family background (left) or high returns to academic achievement (right)

…but in low-income countries, not so much

Similar data has been hard to come by in low-income countries because we need to know how test scores earlier in life lead to (more) schooling later, which requires long follow-up studies. But as these studies start to emerge, the picture looks very different. With Andres Yi Chang, I have been looking at new data from the Learning and Educational Achievement in Punjab Schools (LEAPS) project, which has now followed 5,000+ children from ages 7 to 9 through to their twenties. For around 700 of these children we have test scores from their primary schooling years.

Figure 2a and 2b plot child test-scores from 2003 to 2011. In 2003, the children were in Grade 3 and you can roughly think of each year as the subsequent grade, so that in 2006 most of these children are in Grade 6 (with the exception of those that are held back). The two lines are for children from “high” and “low” socioeconomic backgrounds, defined as an index of parental education and household wealth. Two basic facts emerge from this picture.

  • By the time they are in Grade 3, there is a difference in the test scores of high and low socioeconomic (SES) children. This gap corresponds roughly to how much children learn in a year, so children from low SES backgrounds are one year behind children from high SES backgrounds by the time they are in Grade 3. Through the primary years, there is a slight divergence, but this is neither statistically nor qualitatively significant.
  • By 2011, when (most of) these children are 17, large gaps of one standard deviation have emerged. As Figure 2b shows, this widening gap is entirely due to differential drop-outs. Among children who are still in school, the gaps remain the same as at the end of primary school. But because 61% of children from low SES backgrounds drop out of school compared to 35% from high SES backgrounds (we track and continue to test those who drop out), the average gaps in the cohort are much wider.

Figure 2a: Test score gaps have developed by Grade 3, remain steady over the primary schooling years, and then widen dramatically …

Figure 2b: … because children from low SES backgrounds drop out of school at higher rates than those from high SES backgrounds

Note: These graphs are based on longitudinal data collected as part of the LEAPS long-term follow-up study and are taken from ongoing work by Das and others on educational mobility and test scores in low income countries.

Figure 3 then plots the number of years of completed schooling in 2017 on the vertical axis (14 years after we first surveyed and tested these children), and the test scores of the children in 2006 on the horizontal axis. We show this relationship between test scores and completed years of schooling for children whose fathers have no education versus those whose fathers have more than middle-school education.

  • Although higher test scores are correlated with more years of schooling, family background is paramount. The highest performing children whose fathers have no education complete just as many years of schooling as the lowest performing children whose fathers have more than middle school education (9–10 years).
  • In regressions, every additional year of maternal and paternal education is associated with 0.42 more year of schooling, so that five years of maternal and paternal education implies two more years of schooling for a child. To get the same boost in years of schooling from higher test scores in third grade would require a two standard-deviation increase, equivalent to moving a child from the 50th to the 97th percentile of the test score distribution.

Figure 3: Children with higher test scores go on to complete more years of schooling … but family background remains critical.

Note: This figure is based on data from the LEAPS long-term longitudinal study.

Unfortunately, we don’t have a single intervention that has been shown to boost learning by even one standard deviation over the primary schooling years. In fact, based on these estimates, here is what we could honestly say to an illiterate parent wondering what our latest learning intervention will do for her child:

Our intervention boosts learning by 0.5 standard deviations, which will then increase your child’s years of schooling by 6 months. But even if this boosts your child to the 99th percentile of the test score distribution, she will still get less education than the children of parents with more education, even if they are at the 1st percentile of the test score distribution. But this is the best we have.

Fixing schools and improving learning will only go a small way towards addressing the deep equity issues we face today.

Yes, but, what if …

We can quibble. Maybe these results are specific to Pakistan. But work we are now doing with Abhijeet Singh finds remarkably similar patterns in the Young Lives countries of Peru, Ethiopia, Vietnam, and India. Maybe test scores have an independent effect on wages and later life outcomes regardless of their association with further educational attainment. This is again true for the U.S., where wages become more correlated with test scores and less correlated with educational attainment as workers gain labour market experience. But emerging evidence from China, Brazil, and Pakistan (admittedly, with imperfect studies) already show that wages are more strongly associated with years of schooling and labour market experience rather than test scores.

Should I stop kidding myself, or is there a way forward?

One possible takeaway from these results is that the systems are stacked against the poor. We can work at the margins but few of the things we are doing are going to fundamentally change their life circumstances.

But perhaps that is too pessimistic. Societies can, and do, change.

Consider, for example, the experience of the U.S., which in 1933 looked very much like low-income countries today with a high weight on family background and little weight on academic performance (Hendricks, Herrington and Schoellman 2017). Yet, by 1960 family had receded to the background with academic performance gaining importance.

By 1970, this trend strengthened further and for later cohorts, academic performance became the sole determinant of college attendance. The authors argue that much of this had to do with a surge in demand for college and the decreasing cost of college (in part due to subsidies for WW2 veterans through the G.I. Bill), which increased competition for high quality colleges, and allowed them to shift strategies from recruitment to selection.

Or, look more closely at India where Sam Asher has painstakingly computed educational mobility rates for India from 1950 to 1990. He shows that, even as overall educational mobility in India remained the same over this period, the mobility of scheduled castes (and to some extent, scheduled tribes) increased quite dramatically. One policy that could explain this improvement is reservations for these groups, both in higher education and in jobs.

Both examples show that in a (relatively) short time, societies can create structures that value academic achievement over family background.

But both examples also highlight that these require ongoing, systemic changes.

Ongoing because in the U.S., even as academic achievement became critical for college attendance, the overall importance of family background declined only slightly. Learning gained centre stage, but the correlation of family background and academic performance strengthened sufficiently to ensure that the children of the advantaged remained advantaged themselves. The battleground shifted, but the battle was not won.

Systemic because expanding these policies and implementing reservation policies across India has been a huge political battle with countless lives lost in the process. The rise in educational mobility for these groups has come only because scheduled caste and scheduled tribe populations mobilised and fought tooth and nail for a modicum of dignity in a country where most gains have accrued to the rich.

Ultimately, how much schooling children receive depends on how that schooling is valued in the labour market and how costly it is to obtain. In Pakistan, perhaps poor parents know that even if their very smart children continue in school, there will be no return in the labour market. Similarly, rich parents may already have a job lined up for their children – as long as they pass the critical high-stakes Grade 10 exam that opens up multiple opportunities in the public and private sectors. Or, perhaps the money is so tight that the poor cannot afford to send their children to school beyond a certain age.

At this point we don’t have the answers to these critical questions. But if we continue to believe that increasing learning in schools will be sufficient to enhance equity, we are kidding ourselves.

This note is based on ongoing work with Andres Yi Chang (World Bank) and Abhijeet Singh (Stockholm School of Economics). The data on Pakistan used for this project come from the LEAPS project and its long-term follow-up, funded by The World Bank’s SRP programme and the RISE programme. I thank Lee Crawfurd, Shanta Devarajan, Lant Pritchett and Zainab Qureshi for detailed comments on a previous draft. The findings, interpretations, and conclusions expressed in this paper do not necessarily represent the view of the World Bank, its executive directors, or the countries they represent. I declare no relevant or material financial interests related to the material presented in this note.

Correspondence to jdas1@worldbank.org.

References

Asher, S., Novosad, P., Rafkin, C., (2018), ‘Intergenerational Mobility in India: New Estimates from Administrative Data.’

Carneiro, P. and J. Heckman. (2003). ‘Human Capital Policy’, in Inequality in America: What Role for Human Capital Policies? ed. J. Heckman and A. Krueger (Cambridge, MA: MIT Press, 2003).

Hendricks, L., Herrington, C., and T. Schoellman. (2017). ‘The Changing Roles of Family Income and Academic Ability for US College Attendance.’ No. 1602.

Narayan, A., Van der Weide, R., Cojocaru, A., Lakner, C., Redaelli, S., Mahler, D.G., Ramasubbaiah, R., Gupta N. and S. Thewissen. (2018). ‘Fair Progress? Economic Mobility Across Generations Around the World.’ Equity and Development Series. © World Bank Group, Washington D.C.


To download this commentary in pdf form click here.

This commentary was originally published as part of the CfEE Annual Research Digest 2017-2018, in September 2018.

The volume is edited by Lee Crawfurd, Strategic Advisor with the Ministry of Education in Rwanda and the Tony Blair Institute for Global Change, and a CfEE Fellow.

To download a pdf of Lee’s introduction to the volume, visit this page.


Jishnu Das is a Lead Economist in the Development Research Group at The World Bank and a Senior Visiting Fellow at The Center for Policy Research, New Delhi. Jishnu’s work focuses on health and education in low and middle-income countries and he was member of the core team for the World Development Report on Gender and Development, 2011. He was the Flander’s Visiting Professor at McGill University in 2015, received the George Bereday Award from the Comparative and International Education Society, the Stockholm Challenge Award for the best ICT project in the public administration category in 2006, and the Research Academy award from the World Bank in 2017 and 2013. He has worked in India, Pakistan, Nepal, Paraguay,Kenya and Zambia.

Blog Category: 
About the author