Poor economics and its missing mechanisms: The case for causal mediation

• Published date: 01 November 2024
• Author: Sunil Mitra Kumar,Ragupathy Venkatachalam
• Date: 01 November 2024
• DOI: http://doi.org/10.1111/rode.13029

SPECIAL ISSUE ARTICLE

Poor economics and its missing mechanisms:

The case for causal mediation

Sunil Mitra Kumar

1

| Ragupathy Venkatachalam

2

1

King's India Institute and Department of

International Development, King's

College London, London, UK

2

Institute of Management Studies,

Goldsmiths, University of London,

London, UK

Correspondence

Sunil Mitra Kumar, King's India Institute

and Department of International

Development, King's College London,

London, UK.

Email: sunil.kumar@kcl.ac.uk

Abstract

A key aim of studying development is to understand the

factors that shape socioeconomic progress and explain

inequalities. In empirical work, the predominant focus

has been on posing these questions in the language of

causal inference: how one or more variables effect an

outcome of interest, with the estimation of Average

Treatment Effects (ATE) becoming prioritised as the key

objective. The ‘credibility revolution’and the emphasis

on randomised controlled trials in research on develop-

ment has cemented this dominance, because randomisa-

tion is well-suited to estimating the ATE. This paper

argues that this dual dominance—ATE as main ques-

tion of interest, and experiment as preferred method—is

narrow and restrictive. We propose causal mediation

frameworks as an alternative, which are routinely used

in disciplines including epidemiology, psychology, soci-

ology and political science where causal mechanisms

are an equally important focus. We introduce key con-

cepts and definitions of path-specific effects, and discuss

identification and estimation approaches. We illustrate

applications for development and demonstrate how

causal mediation brings the focus back to contextual

knowledge, combining this with empirical rigour.

KEYWORDS

causal mechanisms, causal mediation, development economics

Received: 9 July 2022 Revised: 25 June 2023 Accepted: 27 June 2023

DOI: 10.1111/rode.13029

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distrib ution and

reproduction in any medium, provided the original work is properly cited.

© 2023 The Authors. Review of Development Economics published by John Wiley & Sons Ltd.

2014 Rev Dev Econ. 2024;28:2014–2033.

wileyonlinelibrary.com/journal/rode

1|INTRODUCTION

Amongst the key aims of the study of development across disciplines is to understand the fac-

tors that shape socioeconomic progress and explain underlying, often persistent variations

across groups and countries. Such factors can operate and be defined at macro level

(e.g., nations, institutions, structural constraints) and micro level (e.g., individuals and house-

holds). In practice, interest in this larger query is often broken down into smaller sub-questions

at different levels to make theoretical and empirical investigations tractable. These involve

focusing on isolating a smaller subset of variables, developing a theoretical framework to

explain observed empirical patterns, potential determinants, and where relevant offer policy

insights. In development economics, the predominant focus has been on posing these questions

in the language of causal inference: how one or more variables effect an outcome of interest.

This interest in causal inference has shaped the field in distinct ways, with the estimation of

Average Treatment Effects (ATE) becoming prioritised as the key objective, be it using experi-

mental or observational data. For example, the effects of a job-training programme for the

unemployed on future employment levels, or of a remedial teaching intervention on learning

scores amongst school pupils and so on.

Experimental methods in particular have enjoyed disproportionate attention in the past two

decades and there is a sizable literature on their advantages (e.g., Banerjee & Duflo, 2011;

Haynes et al., 2012) and key limitations, both methodological and ethical (e.g., Deaton &

Cartwright, 2018; Teele, 2014). The ATE focuses attention on the causal effect of a single vari-

able on a single outcome. Randomisation directly enables this form of causal inference, while

observational data can also be used to infer these effects using appropriate statistical

adjustment.

However in our view, this dual dominance—of the ATE as main question of interest, and

experiment as preferred method—represents a narrow and restrictive form of causal inference.

The limitation is twofold. First, a narrow focus on what works has consequently led to neg-

lecting causal mechanisms or why it works. The processes through which various factors

together shape developmental outcomes are seldom the object of inquiry, or at best indirectly

so. This is in contrast to disciplines like epidemiology, psychology, sociology and political sci-

ence where causal mechanisms are an important focus. Second, a failure to fully exploit the

power of causal inference frameworks to glean insights into such mechanisms from observa-

tional data has further narrowed the scope of scholarship.

In contrast, understanding development more holistically necessitates moving beyond mea-

suring ATEs alone and the related disproportionate focus on experimental methods. Account-

ing for causal mechanisms is a crucial element of such understanding, and thereby the myriad

ways in which context, structure and processes interact to shape developmental outcomes. In

order to understand mechanisms, we need to identify and disentangle the pathways through

which causal effects are manifested. This involves studying intermediate variables—media-

tors—and drawing upon contextual background knowledge to specify the causal structures

involved. This also requires bringing the emphasis back to practitioner expertise and contextual

or domain knowledge, which otherwise have a tendency to be relatively deprioritised.

In this paper, we illustrate this argument and offer a concrete methodological solution for

investigating causal mechanisms through the toolbox of causal mediation methods. Mediation

methods are common in a variety of disciplines including psychology, epidemiology, political

science and sociology, so forth, but their adoption has been very limited within economics, and

even less so within the study of development. For example, while Deaton (2010a,2010b) has

KUMAR and VENKATACHALAM 2015