Yesterday (April 29, 2025) I was at a conference of scholars who work mainly on the USA (the first time in my life I have been to a two day event devoted exclusively the the USA). I met someone new and she asked “What is it you do?” and I said “I am a development economist” and she said “What is that?” and I said “I do the economics of poor countries like, say, India and Indonesia” and she said “So, like micro-finance?”
I did not know whether to laugh or cry. So I laughed. I said: “No, nothing like micro-finance, nothing at all. I am the type of development economist who worries whether the financial system in India (banks, stock markets, corporate bonds) can support the level and allocation of financial services (including the financing of private sector investment) that India’s economy needs to reach the level of worker productivity of the USA or France or Japan.”
But I was crying inside. How did we get here? How did the noble vision of development economics as the study of, and if possible, support to, the four-fold transformation of national development get reduced to the evaluation of discrete ‘interventions’ (which don’t even pass a smell test as important to national development)? How did development economics become understood in the global North as the field that generates evidence about how to do more effective charity work?
Sometimes when you get very lost the only way back is to trace your path back to the first wrong turn and start over from there. The root of the problem is that, stunningly, a large part of the development field acquiesced in a measure of development progress, poverty measured at low-bar poverty lines–that excludes the progress of most of the the world.
That was the result of three steps.
One, the claim that “development is about poverty reduction” became widely accepted. So far, so okay, but that statement generates consensus only by constructive ambiguity and its implications depend entirely on what is meant by “poverty.”
Two, it became commonplace to measure “poverty” by a class of measures (Foster, Greer, Thorbecke 1984) that have the property that above some threshold of income, called a “poverty line”, the incremental contribution of income to poverty reduction is exactly zero.
Whoa, careful here. Already this second step implied these poverty measures were no longer compatible with standard economics of welfare measurement. Unlike the silly caricature that economists believed GDP per capita (an average) was itself a measure of wellbeing, most economists believed in distribution sensitive measures of wellbeing because improvements of the less well-off should count for more in assessing progress than the same progress of the better off. But the idea that gains to wellbeing above some threshold of income should count for exactly zero in an aggregate measure of wellbeing even though people themselves rank this income as valuable is explicitly paternalistic. This violates the (strong) Pareto principle (that if one person was better off and no person worse off then an overall ranking should increase), a widely accepted axiom of an adequate measure of aggregate wellbeing.
There was a debate with poverty measures about the tradeoff between a measure of wellbeing that satisfied the (strong) Pareto Principle for social welfare measures and what Amartya Sen proposed as the “focus axiom” (which was, properly speaking, never an “axiom”) that a poverty measure should be invariant to increases in income (or redistribution among) “the rich.”
This opened up a gap between mainstream economics and “poverty” economics. Mainstream economics was always perfectly open to social welfare functions that counted gains to poorer people as more important than gains to richer people but did not accept the paternalism of counting gains to some people at zero. Squire and van der Tak 1976 proposed that the World Bank’s cost-benefit analysis of projects should incorporate methods that gave more importance to the gains of poorer people, a suggestion which was not at all controversial or rejected in principle.
But “poverty economics” adopted poverty measures which were explicitly paternalistic. The practical severity of the differences between standard distribution sensitive money -metric measures of wellbeing (such as the Atkinson index) and poverty measures (such as the FGT class) depended on where one drew the poverty line. As Angus Deaton has said in the past: “I have no problem with poverty measures as long as the poverty line is infinity.” I would guess among most development economists would have little problem with a “focus axiom” that was invariant to the income of “the rich” if “the rich” meant those with income over 100 dollars a day. Or even perhaps if “the rich” were defined as those with consumption over 50 dollars a day. But this is not what happened.
The third step is where it all went wrong.
The World Bank’s annual flagship report, the World Development Report was titled Poverty. It was felt (and this is first person report oy my conversations with the people responsible at the time) that the report needed a headline number for “How many people in the world are poor?” The choice was made to report the number of poor in the world based on a poverty line of (roughly) dollar-a-day in income/consumption per person per day (adjusted for PPP). This poverty was chosen because it was the lowest plausible global poverty line. The only rationale for this particular, very low, poverty line was that it was the poverty line of the poorest countries in the world (Ravallion, Datt and van de Walle 1991)–with the claims that there was a lower asymptote in the relationship between national poverty lines and GDP per capita).
It was obvious (to me at the time, and I expressed these objections at the time) that using the dollar-a-day line makes the problems with FGT poverty measures as bad as they could possibly be.
With a low poverty line the discrepancy between a standard distribution sensitive measure of social welfare and the poverty measure becomes very stark. Any reasonable social welfare function is going to treat the gains of those just below dollar-a-day as very important–much more important than the gains to “the rich” at, say, 100 dollars a day. But any reasonable social welfare function is also is going to treat gains to those just above dollar-a-day as very important, and much more important than the gains to those at 100 dollars a day. And any reasonable social welfare function is going to treat gains to those just above the poverty line and those just below the poverty line about the same.
FGT poverty measures with a dollar-a-day poverty line are completely unreasonable measures of social welfare. The gains to those just above dollar-a-day are treated as worth zero in reducing poverty. This has two implications. One, the ratio of the poverty gains from income gains of those just below and just above the poverty line is infinitely large (a positive number divided by zero). Two, the gains to those just above dollar-a-day and those at 100 dollars a day are treated as exactly alike–both count for zero.
Think of the “focus axiom” that a poverty measure should be invariant to changes in the income of “the rich.” But a low-bar poverty line like dollar-a-day does not divide the distribution into “the poor” and “the rich,” it divides people into “the extremely poor” and the “not-extremely-poor” where the “non-extremely-poor” category includes both very poor people (those just above the dollar-a-day poverty line) and very rich people. So the focus axiom with a dollar-a-day poverty line implies the measure of poverty should be indifferent if we took some money from someone at P$1.25 pppd consumption and gave it to Bill Gates. As a proposal for how to measure human wellbeing in a country this is not just wrong, it is completely nuts.
The combination of (i) “development is about poverty reduction”, (ii) poverty measurement with poverty lines, and (iii) the adoption of a extremely low poverty line as the primary (or exclusive) poverty line has radical implications.
And yet that last step never had any justification sufficient to merit the radical implications. That is, one can accept dollar-a-day as one poverty line among many, a measure of one type of poverty, called “extreme poverty”, but there was never any reason given why the lowest plausible poverty line–and only proposed and rationalized as such–should be treated as “the” international poverty line or become the special or even most important measure of global poverty.
The dollar-a-day line was never widely accepted among countries in the Global South. After all, since the line was chosen because it was (roughly) the average national poverty line of the poorest countries this necessarily implies most countries had, for their own measurement of national poverty, a higher, sometimes much higher, poverty line.
Nevertheless, powerful political forces in the Global North favored this penurious standard and hence its use spread.
When the Millennium Development Goals (MDGs) were adopted in 2000, Goal 1 was “Eradicate extreme poverty and hunger” and the only target for poverty was Target 1A: Halve, between 1990 and 2015, the proportion of people living on less than $1.25 a day. (which was the inflation updated dollar-a-day standard). So in ten years progress on poverty measured at the dollar-a-day poverty line had gone from being plucked pretty much out of thin air and adopted because it was such a low threshold to become a central goal of development.
Fast forward another 11 years and in 2011 the book Poor Economics (Banerjee and Duflo 2011) is published and becomes a wildly popular and influential book. They proposed that many problems of poverty could be improved through a development economics that relied more on the use of rigorous evidence about the impact of specific, targeted, interventions, with priority given to the use of randomized controlled trials (RCTs). They argued that this “small” approach could nevertheless lead to big impact.
Agnes Labrousse (2020) in a brilliant (but widely ignored) essay explains the puzzling success of this book as due to its “rhetorical superiority.” She argues it is extraordinarily weak as scholarship but a brilliant exemplar of Aristotle’s three principles of ethos, pathos, and logos.
She also points out that the tell a narrative about poverty that ignores nearly all of development economics. “They leave out of the field, without specifying this—it is in this way that the framework is manipulative—fundamental questions of development economics.”
She provides a list of words never mentioned (or mentioned only once):
These off-camera elements are, in fact, out-of-discourse. The organization of production and business, innovation dynamics, meso-economic and territorial questions, local and international financial and commodity flows, macroeconomic dynamics and politics, the environment and inequalities are largely absent. As such, there are no instances in the body of the text for inequal* and unequal, Gini coefficient, income/wage disparity/ies, justice, ethics*, dependency, terms of trade, import, comparative advantage, commodity/ies, stabilization, specialization, international relation*, industrial revolution, capitalism, market economy, modernization, westernization, globalization, tariff*, reserves, foreign investment, capital flow*/flight*, brain drain, volatility, instability, speculation/tive, deregulation, Dutch disease, monetary policy, fiscal/budgetary policy, redistribution*, protection/ist*, lost decade, (Post)Washington consensus, IMF, structural adjustment, foreign debt, foreign investment, fair/free trade, regional development, value chain, production network, corporate governance/interests, innovation fund, technology gap, patent, license, intellectual property, agrarian reform, land grabbing, deforestation, commons/common pool, natural resources, climate change, greenhouse (gas), bio diversity, public good. Industrial policy appears only once, which is the same for domination, dynamics (familial), inequity (intrafamilial), trade (the idea of trade credit), remittance, diversification (of risks), pollution (“pollution inspectors”), externalities (“treatment externalities”), global warming, carbon emission, liberalization (“early years of Chinese liberalization”), privatization (“privatization voucher” for school fees) or recession. Energy is used only in the psychological sense (3 uses); the same is true for 5 out of 7 instances of depression. The results were similar for structure and macro (cf. Section 8.5). This is revelatory of the fundamental difficulty of RCTs in tackling historical dynamics (including microeconomic dynamics), and meso and macro questions. These issues are not amenable to RCTs.”.
So roll forward to 2019. Banerjee and Duflo and Kremer won the Nobel Prize. The scientific background statement says: “This year’s Prize in Economic Sciences rewards the experimental approach that has transformed development economics, a field that studies the causes of global poverty and how best to combat it. In just two decades, the pioneering work by this year’s Laureates has turned development economics ― the field that studies what causes global poverty and how best to combat it ― into a blossoming, largely experimental field.”
This statement reveals the centrality of the complete embrace of low-bar poverty lines and how that re-defines what development economics is about.
If by “global poverty” one means the very wide variety of poverty measures that countries use, or the higher poverty lines used by the World Bank, or more ambitious measures of global poverty enshrined in the SDGs, or global poverty lines such that those not in “poverty” are “prosperous”, then it is true this is broadly what the field of development economics studies. But then it is not true that this field and these questions have been particularly helped by experiments.
Take Pakistan. At the World Bank dollar-a-day (now P$2.15) line the headcount poverty rate is 4.93 percent. But at P$6.85, a poverty line the World Bank also routinely reports–which is still a low poverty line relative to developed countries or what we argue in our paper are defensible upper-bound development poverty lines–the headcount poverty rate is 84.53 percent of the population. One can imagine that targeted programs informed by the results of RCTs might help reduce dollar a day poverty in Pakistan. But it is not at all plausible to think that the level of P$6.85 poverty line poverty does not require broad based improvements in the overall level of productivity of the economy. Programs for “the poor” are not relevant to 84.53 percent of the population.
When the statement says this “transformed development economics” one might think this means there was a field of development economics that addressed an array of questions and the advent of experimental methods improved that field. But that is not what happened. What happened was that there was a very, very narrow subset of questions within development economics which were, perhaps, in principle, amenable to experimental method and a large set of questions important to countries becoming productive and prosperous that these methods were not amenable to. What the randomista movement did was shift the attention within and among a group of academic researchers to the set of questions that were amenable to their method and just, as in Poor Economics ignore the other questions.
The point here is that the question of whether or not RCTs are important contribution to development economics hinges on the question of the poverty line. If one accepts that the only, or primary, or most important, meaning of “global poverty” is a measure of poverty that uses a very low poverty line then this debate can be entertained. But if one rejects that dollar-a-day represents the only, primary, or most important definition of global poverty then claims for the importance of RCTs to reducing global poverty are just obviously false.
Take Niger. There was a recent paper, published in the prestigious magazine Nature doing an RCT about adding various design elements to a cash transfer program, including a “psychosocial” component. Does a study like this contribute significantly to reducing global poverty? At the World Bank P$6.85 poverty line headcount poverty in Niger is 96.5 percent. So the answer is obviously no. Does this paper show there are interventions which can cost-effectively add very very small amounts to the incomes of very very poor people in Niger? Yes. Was that an important question in development economics? That is only even conceivably true if one takes the low-bar poverty lines as being much, much more central to development than development country governments or people themselves do (that is, no one in Niger thinks being just above a dollar-a-day is discontinuously important).
All of this is background to why I think my new paper with Martina Viarengo titled “Raise the Bar” (and my previous paper “dollar-a-day” poverty was a mileage marker, not the destination) are interesting and relevant. In our new paper we start from the fact that the poverty goal in the 2015 Sustainable Development Goals is radically different from the MDG of 2000. Instead of the narrow goal of “End extreme poverty and hunger” the SDG is “End poverty in all its forms everywhere.” This is an explicit rejection of the notion that “extreme poverty” (dollar-a-day poverty line) is the primary definition of global poverty.
This embrace of an array of poverty lines raises the question: “If dollar-a-day is the lowest plausible line for global poverty, what is the highest global poverty line for a development goal?” The paper analytically and empirical justifies a global upper-bound poverty line of P$21.5, ten times the dollar-a-day standard.
With “global poverty” defined as the fraction of people living below P$21.5 (and with more weight on improvements on those with lower incomes within that) I am then happy to define development economics as: a field that studies the causes of global poverty and how best to combat it. With that idea of global poverty we need a better development economics that can answer the important and pressing questions political leaders and policy makers in developing countries face, very few of which are amenable to improved techniques for project evaluation. And, to circle back to the opening anecdote, in that development economics, microfinance is, at most, a tiny concern.