research methods in economics

How are econometric methods applied by researchers in development economics?

VoxDev Blog

Published 03.09.24
Paola Davila

Paola Davila

Deputy Managing Editor, VoxDev

Oliver Hanney

Oliver Hanney

Managing Editor, VoxDev

How has research featured on VoxDev used different econometric techniques? Here are some examples from recent development economics research, offering insights for students, teachers and academics.

It has been a busy year at VoxDev, and given the sheer volume of new articles we have published, we thought we would refresh our previous blog post highlighting VoxDev’s usefulness as an educational resource. This blog outlines some interesting recent applications of econometric methods that students might have been taught without much context. These articles and podcasts, and the underlying research papers, are perfect for adding to development economics courses and teaching materials in need of some interesting and applied development economics research. 

This post mainly focuses on articles published during 2024. These provide useful examples for students, either because they use a particular causal inference technique commonly taught in economics undergraduate or postgraduate degrees, or approach development issues in an interesting way. 

If you have a general interest in development economics, study economics, or teach a development economics course, our newsletter is the best way to keep up to date – you can sign up here.

Examples of Randomised Control Trials (RCTs) in developing countries 

Randomised Control Trials are often considered the gold standard for causal inference. As a website covering development economics, we have featured our fair share of RCTs – in fact, they are the most common method used by the research we feature. For a summary of the early stages of the randomised control trial method, and how its pioneers wound up winning a Nobel Prize, see this previous article - What does the 2019 Nobel mean for development economics? While incredibly useful for researchers, RCT’s are not a silver bullet - as explained here by Maitreesh Ghatak, here by Ted Miguel, and here by Paul Niehaus. 

Randomised Control Trials have offered valuable insights into the effectiveness of policies aiming to: 

The scale of the RCT’s that researchers are able to conduct is increasing: 

One criticism of experimental designs is that they may not capture long-run outcomes. More recently however, researchers have been able to return to earlier RCTs and measure their long-run impacts. For example, Tania Barham, Karen Macours and John A. Maluccio consider the effects of a conditional cash transfer programme ten years after its implementation.   

More examples of Randomised Control Trials.

What are ‘Natural Experiments’ in economics? How have they been used by research? 

Often researchers may not be able to run a RCT to research a policy question. Instead, they sometimes look for so-called ‘natural experiments’. What is a natural experiment? Natural experiments are studies that use shocks, policies or other events of interest that lead to some sort of 'random' variation in the exposure of individuals or groups. Often these natural experiments are historical ‘accidents’ that create differences between otherwise similar groups or individuals – this can help economists identify a good counterfactual. 

Researchers get creative in their search for these historical ‘accidents’. Examples of natural experiments on featured on VoxDev include: 

  • The dramatic expansion of visas for foreign nurses in the US, followed by a sudden reduction in visa availability. This enabled Paola Abarcar and Caroline Theoharides to explore the brain drain hypothesis in the Philippines. 
  • A bilateral treaty on migration between Malawi and South Africa was unexpectedly cancelled in 1974 following a deadly miner planer crash. Taryn Dinkelman, Grace Kumchulesi and Martine Mariotti use this to estimate the impact of international migration on Malawi. 
  • The rapid expansion of genetically modified seeds in Brazil allowed Mateus Dias, Rudi Rocha, and Rodrigo R. Soares to show that herbicides widely used in agriculture increase infant mortality. 

More examples of Natural Experiments.

Differences-in-differences (DIDs) in development economics 

Lots of establishing “causality” boils down to finding good comparisons, including a group that experience some sort of change (be it policy-driven or the result of a natural experiment), and a different group which did not and can be shown to be a good counterfactual. Difference-in-differences designs (there are many different types!) are a go-to method in this case in economics and social sciences.  

On VoxDev, the difference in difference approach has been used to show that: 

More examples of difference in difference models.

Regression Discontinuity Designs (RDDs): Cutoffs used by research in development economics 

Every economist loves a cutoff as it lets us compare people just above and just below, or areas just to the left and just to the right, to establish causal effects. The Regression Discontinuity Design revolves around this principle, allowing for cleaner comparisons and making it a popular choice for economists.

Regression Discontinuity analysis has been applied in various contexts on VoxDev. For instance:  

More examples of regression discontinuity designs.

Instrumental Variables (IVs): Examples of instruments used in development economics 

What is an instrumental variable in econometrics? A good instrument requires a strong first stage (i.e. it is strongly correlated with the explanatory variable of interest) and must satisfy the exclusion restriction (it must only affect the outcome of interest through its effect on the explanatory variable). 

Once an instrument has been found, this can be used in an instrumental variable estimation strategy to uncover causal effects in various policy-relevant contexts. Research featured on VoxDev has used an instrumental variable approach, with a wide variety of instruments, to explore a range of policy-relevant questions:

‘Bartik instruments’ have become increasingly common in development economics research. These instruments typically interact a ‘share’ (e.g. initial industry employment shares) with a ‘shift’ variable (e.g. a change in industry imports). David McKenzie’s World Bank blog post describes how these instruments are employed and this later post (Bryne, Kondylis and Loeser 2021) discusses the assumptions required to justify causal identification. Here are a couple of examples of how Bartik instruments have been used on VoxDev:  

More examples of the Instrumental Variable method.

Calibrated Models 

While we don’t feature purely theoretical research, many of our articles present results from calibrating models, i.e. plugging numbers from the real world into models.

These model-based approaches are an increasingly important methodology within development economics. While they are on the more advanced end of university economics courses, it is still interesting to see how these techniques can be used. Last year's podcast on the role of macro modelling is a good starting point. Francisco Buera and Joseph Kaboski outline how researchers are starting to combine the "micro" causal inference toolkit with "macro" modelling techniques to improve the relevance of research.

One area in which work using model calibration techniques is particularly important is considering how the world will adapt to climate change, as seen in Section IVA of our VoxDevLit on Climate Adaptation. Other examples on VoxDev include Allan Hsiao exploring how different trade policies for palm oil could reduce emissions, and Allan Hsiao, Jacob Moscona and Karthik Sastry showing that food policies in response to climate shocks often exacerbate global losses. 

Other articles have used similar calibrated model approaches to explore a wide range of issues including: 

More examples of model calibration in economics.

Heterogeneity: When average effects mask important differences 

This jargony word is loved by economists and we have become well accustomed with the synonyms. ‘Heterogeneity’ simply conveys that aggregate effects can mask differing impacts. For example, Derek Headey and Marie Ruel show that rising food prices are putting children in harm’s way, and estimate the heterogenous impacts of a 5% increase in real food prices on child wasting, across genders, geographies, and incomes.

Considering ethno-local norms is key when conducting research. For example, Tamara McGavock and Lindsey Novak unpack a key heterogeneity, based on local traditions, in the impact of droughts on a harmful traditional norm in Africa.

More examples of Heterogeneity.

Other VoxDev resources 

While we have mostly stuck to articles in this blog, we also have over 200 brilliant podcast episodes where researchers discuss findings from their research. Why not mix up your reading list with the inclusion of some of these episodes of VoxDevTalks?

Also, our growing library of VoxDevLits summarises the body of evidence on specific development topics. These are kept up to date, freely available, and focus on policy implications and accessibility. These living literature reviews also outline lots of ideas for where future research is needed – a potential starting point for ambitious and important dissertation projects. 

VoxDev and educating economists 

In many ways, VoxDev’s usefulness as an educational resource for students and educators is a spillover from our original mission, to make economic research accessible to policymakers who may not have had the technical training of an economist. This means that, as opposed to an academic publication, the audience for our economic content does not need to understand the methods or jargon of economists (in theory at least, there is always more we can do to make research more accessible).  

This makes our articles, podcasts, videos and VoxDevLits great gateways into development economics for students, many of whom will go to work in policy, who can read about the latest, high-quality applied research in an accessible manner with no restrictions or fees. 

Here are some great resources for those seeking a deeper understanding of the methods we have mentioned above: 

Editor’s note: If you think there is anything we should add to this post, or have any comments/suggestions, please email [email protected]. This blog will be a continually updated resource, like our VoxDevLits, and evolve over time as we feature new research using economic methods of the future. Also, if you are involved in teaching development and have used VoxDev as an educational resource, it is always great to hear about this – tracking this type of “impact” is always tricky and your stories help us understand our reach. 

RCT IV DID natural experiment RDD calibrated model heterogeneity

Share this voxdev blog: