Development Economics after the Experimental Turn: A Critical Review of a Displaced Object of Inquiry

Abstract

Received: 11/08/2026, Accepted: 14/09/2026, Published: 18/09/2026

The experimental turn has been argued over as a quarrel about method. This article makes a different claim: what the turn produced was a substantial reweighting of the object of inquiry in development economics, programme-level causal effects becoming the dominant form in which development problems were rendered researchable. The review draws on some sixty texts published between 1995 and 2025, selected on three criteria: the founding and handbook statements of the experimental programme, the peer-reviewed methodological criticism it drew in general and in development journals, and the post-2015 work on aggregation, general equilibrium and scaling. It works from an explicit criterion of demarcation: an object has been displaced when the questions the apparatus cannot handle become markedly less central in the field's journals, not when they receive different answers there. Four shifts are documented. The unit of analysis moves from the trajectory of an economy to the average effect of a programme. The state moves from central object of explanation to implementing institution. The time horizon moves from structural transformation to the two or three years of a survey cycle. Causal credibility comes to rest more on assignment than on an explicit account of mechanisms. The magnitudes give the measure of the shift: the J-PAL network reports a cumulative portfolio rising from 103 randomised evaluations in 2003 to 2,566 in 2025 across 124 countries (J-PAL, 2026), and by 2015 empirical work, not experimental work specifically, accounted for more than 90 per cent of weighted publication pages in development. From this the article draws a proposition about the past decade: the corrections devised since 2015, Bayesian aggregation, general equilibrium estimation, the political economy of scaling, put back one by one the elements the apparatus had set aside, namely structure, aggregation and the state. Two results measure that return: the local transfer multiplier of 2.5 estimated on cash grants in Kenya, and the undetectable effect of contract teachers hired by the Kenyan administration against 0.18 standard deviations for the same contract delivered by an NGO. A bibliometric probe on 18,592 articles published in seven development journals between 1990 and 2025 puts figures on the reweighting: the share carrying an experimental vocabulary rises from 0.71 to 9.63 per cent of annual output between the 1990s and 2016-2025, reaching 18.4 per cent in the Journal of Development Economics, while the structural vocabulary follows a U-shaped path, halving during the 2000s and recovering afterwards everywhere except in that same journal. The epistemological consequence falls on the global South: the evidence available to policymakers there is composed of programme-level effects measured in the few countries cheapest to survey, while the questions on which their macroeconomic and industrial choices are decided remain, by construction of the method, the least researched