Alleviating poverty with experimental research: The 2019 Nobel laureates

Article

Published 01.11.19
Photo credit:
Dominic Chavez/World Bank

The Nobel nomination’s emphasis on the practical applications of Banerjee’s, Duflo’s and Kremer’s methods represents a monumental and welcome change

‘‘Most of the people in the world are poor, so if we knew the economics of being poor, we would know much of the economics that really matters.” 
(Schultz 1980)

Theodore Schultz and Arthur Lewis were awarded the Nobel “for their pioneering research into economic development research with particular consideration of the problems of developing countries” (Royal Swedish Academy of Sciences 1979). It was 1979 and despite this global recognition, development economics was on its way out of the mainstream into a backwater where it remained for decades, like an odd cousin who is too embarrassing to be seen with in polite society. There were exceptions, of course, but it took mavericks like Chris Udry to be able to make fieldwork respectable by combining it with state-of-the-art theory.

Development economics had no PhD courses, no group at the NBER or CEPR, and hardly any publications in top journals until the early 2000s. What this year’s Nobel laureates did was to build the infrastructure to make fieldwork widely accessible and the methods to make the analysis credible. What they did, and what they were awarded for, is to put development economics back on centre stage. 

The prize to Abhijit Banerjee, Esther Duflo, and Michael Kremer is unusual in many ways. Passionate about statistical trivia, the economist on the street will quickly point out that the three are very young, that Esther Duflo is only the second woman to win it, and that she is the youngest ever economics laureate. 

This is indeed unusual – but it is largely irrelevant. What is unusual and relevant is that the nomination explicitly mentions that the winners lead a group effort: “The Laureates’ research findings – and those of the researchers following in their footsteps” (Royal Swedish Academy of Sciences 2019).

What is even more unusual and extremely relevant is that the nomination emphasises the practical applications of their methods, which “have dramatically improved our ability to fight poverty in practice”. This is a monumental change, and one that the profession should welcome for the obvious reason that making the world a better place is a desirable goal. 

To be clear, each of them could have easily won the prize the ‘usual’ way – that is, by doing research of the highest quality, which has had lasting influence both in theoretical and applied economics. The economist (still) on the street might notice that of the top three cited papers for each of the three laureates, only two are randomised controlled trials (RCTs). 

While their non-RCT work is not explicitly mentioned by the Academy, it is worth revisiting because it illustrates the laureates’ vision and their common interest in understanding the persistence of poverty and the huge differences in living standards across countries.

Abhijit Banerjee’s “Occupational choice and the process of development” (joint with Andy Newman) and Michael Kremer’s “The O-ring theory of economic development”, both published in 1993, build bridges between market failures due to asymmetric information at the micro level and aggregate output and growth at the macro level, thereby laying the foundations for modern growth theory.

They formalise how individual occupational choice decisions map onto aggregate employment and output through the creation of firms. Both papers model plausible mechanisms that create poverty traps, and can therefore explain why societies starting out in very similar place can end up in very different equilibria. 

The mechanism studied in Banerjee and Newman (1993) links inequality to credit market imperfections, which determine whether individuals engage in wage work, small entrepreneurship or manage to hire others and start a firm. An economy that starts poor and equal will remain so because nobody will ever be able to start a firm, thus forcing everyone into subsistence entrepreneurship.

By contrast, an equally poor economy with enough inequality to allow someone to start a firm will end up in a higher-income equilibrium. Importantly, the occupational structure determines inequality, creating vicious or virtuous circles. The paper has had, and continues to have, a deep influence on how economists think about growth, in particular how market imperfections link inequality with growth. 

Kremer’s O-ring theory (1993) studies growth through the lens of organisational economics, thereby connecting what goes on inside a single firm with aggregate economic performance. The paper challenges the view of labour as a homogeneous factor of production and explicitly models the complementarities between workers within different talents doing different tasks within the same firm.

The key assumption is that the value produced by a given worker in a given task depends on the quality of the output produced by workers responsible for other tasks. This generates assortative matching and implies that small differences in skill levels will result in huge differences in productivity and income. These are amplified by the fact that with imperfect information, individuals will under-invest in education, implying that small differences in education policy will produce even larger differences in income. 

Importantly, the paper opened up the possibility that misallocation, in this case through the mismatch of workers, can explain cross-country differences. Today, the misallocation of both capital and labour is seen as key (Restuccia and Rogerson 2017).

Market failures call for government intervention, but whether such interventions can be effective in practice is an open question. Esther Duflo’s first paper answered it by providing evidence on the effect of government investments in schools on educational achievements and earnings.

Duflo (2001) exploits a rapid and very large school construction effort in Indonesia, which varies in intensity across geographical areas. She combines this variation with the observation that only children who were sufficiently young when the schools were built could have possibly benefitted from them. This allows her to estimate the effect of school construction on enrolment, exploiting ‘differences-in-differences’ between young and old cohorts and between high programme intensity and low programme intensity areas. 

Having established that building schools increases educational attainment and having ruled out every plausible, and even some implausible, endogeneity concerns, she uses school construction as an instrument for education in a Mincerian wage regression, providing one of the most reliable estimates of the returns to education. Duflo (2001) makes substantive contributions to the empirical literature on education and elevates the standards for empirical evidence to a much higher level. For both reasons, the paper is hugely influential.

The need for policy to fix market failures generates the need for tools to evaluate policy. RCTs were developed to achieve this in a systematic way. They and other experimental methods had been widely used in the natural sciences and to a lesser extent in economics well before Abhijit Banerjee, Esther Duflo, and Michael Kremer began their work. Their contribution was to make them accessible to a large number of researchers, creating a research ‘firm’ which, like those in Banerjee and Newman (1993) and Kremer (1993), combines the talent of many to produce more than the sum of its individual components.

The basic principle of RCTs is very simple: random assignment into ‘treatment’ and ‘control’ breaks the link between selection into treatment and outcomes, therefore creating a ‘clean’ counterfactual. This is easier said than done as, in practice, subjects might refuse treatment or drop out at a later stage, or treatment might spill over to control subjects. These, along with other limitations, are acknowledged and met with solutions (Duflo et al. 2007).

The rapid growth in the use of trials in development economics has also created a lively literature on the theory (Chassang et al. 2012, Narita 2019), econometrics (Young 2016, 2018), and philosophy of RCTs (Cartwright 2011, 2012). Deaton and Cartwright (2017) present a lucid and comprehensive overview of the issues, both those that can be solved using appropriate techniques and those that cannot and instead require other methods.

Further issues arise because organisations that are willing to put their policies through serious evaluation are not themselves randomly selected, and because the interventions that researchers are allowed to evaluate are positively selected, as no organisation will allow researchers to evaluate a policy that they know to be detrimental (Bandiera et al. 2011).

There is no doubt that RCTs are not a silver bullet, that theory is essential and that the set of questions that RCTs can answer is only a subset of the questions we ought to answer. Nonetheless, RCTs have made careful identification salient, and have set a benchmark for other empirical work. Most importantly, they have dragged many economists out of their offices and into the real world to collaborate with firms, NGOs, and governments.

Perhaps some of the results that they generate are only relevant in their specific setting and, most probably, many of them got their standard errors wrong. But it is difficult to believe that diverting the efforts of talented and highly skilled individuals towards finding solutions to real problems is welfare-reducing. The first order impact on the allocation of talent will come later, from a different place.

Bringing academics to the field has exposed a large number of locals to research ideas and, most importantly, to the idea of research. Among the many field managers, surveyors and research assistants who greeted the prize announcement with great enthusiasm, there will be some who will see an option they did not know of. Many will have the talent and some will have the inclination and the opportunity to take it. My guess and hope is that before too long, a column like this will be discussing their research. 

Editor's note: This column also appeared on VoxEU.org.

References

Bandiera, O, I Barankay and I Rasul (2011), ‘Field experiments with firms’, Journal of Economic Perspectives 25(3): 63-82.

Banerjee, AV (1992), ‘A simple model of herd behaviour’, Quarterly Journal of Economics 107(3): 797-817.

Banerjee, AV, and E Duflo (2011), Poor economics: A radical rethinking of the way to fight global poverty, Penguin. 

Banerjee, AV, E Duflo and M Kremer (2016), ‘The influence of randomized controlled trials on development economics research and on development policy’, in The State of Economics, The State of the World, conference at the World Bank.

Banerjee, AV, and AF Newman (1993), ‘Occupational choice and the process of development’, Journal of Political Economy 101(2): 274-98.

Bertrand, M, E Duflo and S Mullainathan (2004), ‘How much should we trust differences-in-differences estimates?’, Quarterly Journal of Economics 119(1): 249-75.

Cameron, DB, A Mishra and AN Brown (2016), ‘The growth of impact evaluation for international development: how much have we learned?’, Journal of Development Effectiveness 8(1): 1-21.

Cartwright, N (2011), ‘A philosopher's view of the long road from RCTs to effectiveness’, The Lancet 377(9775): 1400-01.

Cartwright, N (2012), ‘Presidential address: Will this policy work for you? Predicting effectiveness better: How philosophy helps’, Philosophy of Science 79(5): 973-89.

Chassang, S, PI Miquel and E Snowberg (2012), ‘Selective trials: A principal-agent approach to randomized controlled experiments’, American Economic Review 102(4): 1279-1309.

Deaton, AS, and N Cartwright (2018), ‘Understanding and misunderstanding randomized controlled trials’, Social Science and Medicine 210: 2-21.

Duflo, E (2001), ‘Schooling and labor market consequences of school construction in Indonesia: Evidence from an unusual policy experiment’, American Economic Review, 91(4): 795-813.

Duflo, E, R Glennerster and M Kremer (2007), ‘Using randomization in development economics research: A toolkit’, Handbook of Development Economics 4: 3895-3962.

Kremer, M (1993a), ‘The O-ring theory of economic development’, Quarterly Journal of Economics 108(3): 551-75. 

Kremer, M (1993b), ‘Population growth and technological change: One million BC to 1990’, Quarterly Journal of Economics 108(3): 681-716.

Miguel, E, and M Kremer (2004), ‘Worms: Identifying impacts on education and health in the presence of treatment externalities’, Econometrica 72(1): 159-217.

Narita, Y (2019), ‘Experiment-as-market: Incorporating welfare into randomized controlled trials’, Available at SSRN 3094905.

Restuccia, D, and R Rogerson (2017), ‘The causes and costs of misallocation’, Journal of Economic Perspectives31(3), 151-74.

Royal Swedish Academy of Sciences (1979), Press release: The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 1979.

Royal Swedish Academy of Sciences (2019), Press release: The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2019.

Schultz, TW (1980), ‘Nobel lecture: The economics of being poor’, Journal of Political Economy 88(4), 639-51.

Udry, C (1994), ‘Risk and insurance in a rural credit market: An empirical investigation in Northern Nigeria’, Review of Economic Studies 61(3): 495-526.

Young, A (2016), ‘Improved, nearly exact, statistical inference with robust and clustered covariance matrices using effective degrees of freedom corrections’, manuscript, London School of Economics.

Young, A (2018), ‘Channeling Fisher: Randomization tests and the statistical insignificance of seemingly significant experimental results’, Quarterly Journal of Economics 134(2): 557-98.

Endnotes

[1] For completeness, Abhijit Banerjee is only the second Indian (after Amartya Sen) and Michael Kremer only the second ‘Michael’ (after Spence).

[2] The top three most cited works (according to Google Scholar or Web of Science) are Banerjee (1992), Banerjee and Newman (1993) and Banerjee and Duflo (2012) for Abhijit Banerjee; Bertrand et al. (2004), Banerjee and Duflo (2012) and Duflo (2001) for Esther Duflo; and Kremer (1993a) Kremer (1993b) and Miguel and Kremer (2004) for Michael Kremer.