McNamara Fellows - 2022 Cohort Blogs

Shaping a Researcher from the Ground Up: The Robert S. McNamara Fellowship

Quite often, people reach out to me asking what economists do. Many of them expect an answer related to finance, business, politics, or even an emphasis on superior mathematical prowess. I do not pretend to answer that question, but it is worth noting that economics is not commonly associated with the realm of science. Surprising as it may seem, economists are, in fact, social scientists. This role can be challenging to comprehend, particularly in many developing countries where the research work in economics is not emphasized as a successful professional path. This misconception belongs to the prevailing paradigm in the development countries.

The Robert McNamara Fellowship is an effort to disrupt that paradigm. Matching potential researchers in economics from developing countries with researchers in the World Bank’s Development Impact Evaluation (DIME) and Development Research Group departments means to generate more opportunities for diversifying research groups and reducing the concentration of research activities.

For individuals who have not yet embarked on a doctoral program but possess a strong desire to explore and answer questions, the fellowship program offers a valuable opportunity to define their research preferences and sharpen the intuition required to begin a journey on economics research. For those currently enrolled in a doctoral program, the fellowship serves as a platform to strengthen their research agenda. Meanwhile, for those who are not specifically interested in pursuing a Ph.D. but still aspire to contribute to research and policy, the Robert McNamara Fellowship is an opportunity to gain insights into managing the research workflow and interact with government agencies, local communities, academia, and policymakers. By participating in high-quality research teams, fellows are provided with a pathway to shape their professional pursuits to make meaningful contributions to the science and policy formulation.

I worked on a research project within the Infrastructure and Climate Change unit at DIME. For a span of nine months, I actively participated in a transportation project aimed at evaluating the impact of subsidized transport fares on the most vulnerable segments of the population. The project takes advantage of a policy change in the methodology of subsidy targeting to assess its effects on the well-being of these vulnerable individuals, such as improved access to healthcare and job opportunities. But one step before that goal, the research team encountered a significant challenge: the low uptake of the subsidy. It immediately creates a puzzle around why people are not signing up for the subsidy. Therefore, as a team, we worked to construct a narrative that contextualized the baseline survey statistics and helped us gain a deeper understanding of the needs, perceptions, and challenges faced by the population. My research role was related to data analysis, survey design, and conducting pre-pilot surveys for the follow-up intervention. It was incredibly touching to be able to interact with some individuals, collecting their experiences, and discussed them within the team to then design an intervention aimed not only to determine the impact of such subsidies but also to assist the local government in designing more effective policies that address the specific needs of the most vulnerable population.

In addition to my project responsibilities, I have also had the opportunity to participate in various technical and knowledge trainings that focus on critical aspects of social research. These trainings cover essential topics such as research reproducibility and replicability, effective management of field research, and coding skills in R, Python, and STATA. Moreover, the opportunities extend further through a wide array of seminars, workshops, and conferences about relevant cross-cutting themes. Engaging in these events not only allowed me to explore different areas of economics that I am interested, but also provides a platform for connecting with researchers beyond the World Bank. These interactions have been invaluable, guiding, and inspiring aspiring researchers like me towards research-oriented careers.

The Robert McNamara Fellowship is a concise yet intensive program that serves as a guiding force for young researchers in pursuit of their professional aspirations. It not only provides an opportunity to engage in a cutting-edge research environment but also offers a platform to shape oneself as a researcher, even from the ground up. The research ethos at institutions like the World Bank inspires fellows to pursue a pathway where research goes beyond filling knowledge gaps and actively impacts the living conditions of peoples around the world.

David Concepcion Quiroz Barrios

Bridging Theory and Practice: Empirical Insights from General Equilibrium Models for Informed Policymaking

Economic activity exhibits significant disparities across spatial areas, even within the same city. These differences encompass a wide range of characteristics, including job and educational opportunities, crime rates, salary levels, and price variations. Understanding the underlying factors behind this uneven distribution of economic activity is crucial for governments to effectively allocate public resources and devise efficient policies that cater to the needs of their citizens. Quantitative General Equilibrium (QGE) models serve as valuable theoretical frameworks for investigating the interactions and consequences of economic activities within specific geographic regions. By employing these models, policymakers can gain insights into the dynamics of regional economies and make informed decisions that foster economic growth and enhance the well-being of their citizens.

In particular, the innovative QGE model developed by Ahlfeldt (2015) has made a significant impact in the field of urban economics by providing a framework to explore various topics through the lenses of agglomeration and congestion forces. Based on this model, researchers have successfully achieved to investigate a wide range of subjects, including trade (Allen & Arkolakis, 2014), labor informality (Zárate, 2022), transport infrastructure (Tsivanidis, 2017), and migration (Bryan & Morten, 2015).  

Overall, the QGE models have facilitated research in urban economics due to their adaptability to different contexts and analytical tractability. However, the widespread adoption of these tools among policymakers and operations professionals has been limited by two main constraints: the unfamiliarity with the model and the requirement of specific licensed software to estimate the model and conduct counterfactual exercises.

During my fellowship at DEC, I had the opportunity to participate in a project aimed at providing comprehensive training on QGE models to policymakers and local agents. The main objective was to emphasize the value of QGE models and enhance understanding and familiarity with these theoretical tools. As part of the project, we also developed an open-source package in R, which enables straightforward estimation of these models with minimal data requirements.

The primary use case for our R package was the evaluation of various public transport alternatives in the City of Cape Town. This process involved an iterative approach and relied heavily on the support of local agents from Cape Town. Initially, we collected and combined data from multiple sources, including census reports, house price data, and shapefiles of transport interventions, to meet the model's data requirements. We then calibrated the model using parameters from the literature, ensuring they accurately represented the context of Cape Town. Lastly, we tried to identify any weaknesses or abstractions in the model to properly frame the results of the model simulations.

By bridging the gap between research and policy practice, initiatives like these empower policymakers to make informed decisions that alleviate poverty and improve the well-being of their citizens. It is crucial to continue fostering collaborations, providing training, and developing user-friendly tools to facilitate the adoption of theoretical models and enhance their practical application in real-world scenarios.

Maria Del Mar

References

Ahlfeldt, G. M., Redding, S. J., Sturm, D. M., & Wolf, N. (2015). The Economics of Density: Evidence From the Berlin Wall. Econometrica, 83(6), 2127-2189. https://doi.org/10.3982/ECTA10876

Allen, T., & Arkolakis, C. (2014). Trade and the Topography of the Spatial Economy. The Quarterly Journal of Economics, 129(3), 1085-1140. https://doi.org/10.1093/qje/qju016

Bryan, G., & Morten, M. (2019). The Aggregate Productivity Effects of Internal Migration: Evidence from Indonesia. Journal of Political Economy, 127(5), 2229-2268. https://doi.org/10.1086/701810

Tsivanidis, N. (2022). Evaluating the Impact of Urban Transit Infrastructure: Evidence from Bogotá's TransMilenio. Retrieved from https://static1.squarespace.com/static/55bb98e2e4b0ba843f39599e/t/62882b12d140ce5b51ff6e24/1653091092837/TsivanidisTransMilenio_5.2022.pdf

Zárate Vasquez, R. D. (2022). Spatial Misallocation, Informality, and Transit Improvements: Evidence from Mexico City (English) (Policy Research working paper No. WPS 9990). Washington, D.C. World Bank Group. Retrieved from http://documents.worldbank.org/curated/en/099332303302232640/IDU04600b00f0f3e404c4f085520cd82c82a60a5

My master's degree in hand and driven by a strong motivation to “end poverty and promote shared prosperity”, I joined DIME (Development Impact and Monitoring Evaluation) last September as a McNamara fellow. I was motivated by the desire to contribute to researching complex questions related to unequal opportunities, gender disparities, education, and job opportunities across countries. These past few months have been an exceptional start to my career in development economics.

Being welcomed by the DIME community provided me with exposure to cutting-edge research. The DIME Analytics team has been instrumental in accelerating my progress by offering various courses and their expertise to enhance my technical skills. Being able to reach out to them for specific challenges allowed me to make tremendous progress.

DIME is a unit where it is possible to explore a broad range of topics and variety of contexts. I joined a project aiming to empower girls to fight gender-based violence in El Salvador through both targeting girls and their families. The project will allow us to understand the most effective target in decreasing gender-based violence and as such to inform potential policies. Additionally, I was able to support some projects in the Democratic Republic of Congo.

I consider one of the most enriching aspects of my fellowship to be the exposure to the early stages of various projects, which has significantly deepened my understanding of the junction of research, technical assistance, and impact in development economics.

This latter point can be summarized by Esther Duflo’s saying during the Development Impact Week:

“What matters enormously [in generating successful development impacts and knowledge] are the details of implementation that no one can think of in advance – and they are solved in partnership with our [government] partners” – Esther DUFLO

This became particularly evident to me during my participation in a mission in Tanzania. The Shule Bora program provides technical assistance to the government of Tanzania in reforming primary education through different areas: safe-schools, school management, inclusion and diversity. The objective of the visit was to gather different stakeholders to align the priority of each.

For researchers to think about the details of implementation in the “trial and adopt” endeavors is key. The specifics of the execution of a policy or project come to enrich the insights for design of research. For policymakers, actively engaging with researchers at early stages of design and implementation can provide them with a compass in their actions. The real-time feedbacks can allow to redefine actions and improve impact. Furthermore, such documentation can allow to scale actions and by extension, contribute to the larger goals of development.

For example, during the discussion on the safe school component of the Shule Bora project, the government members emphasize the need of more knowledge on reporting mechanisms of harassment in school. Consequently, our team took a direction in line with the need of policymakers.

In conclusion, the McNamara fellowship and my journey with DIME have provided me with valuable opportunities to explore complex development issues, enhance my technical skills, and gain practical experience in project implementation. The exposure to diverse projects and contexts has deepened my understanding of the intricacies of development economics. I am grateful for the supportive DIME community, and the chance to contribute to impactful research that can drive positive change and address pressing global challenges. Finally, along the way I made friends and colleagues who have been essential to reflect with and to learn from.

Muhsine Senart

Things I learned from reproducing 18 papers

After finishing my master's degree in applied development economics, I joined The World Bank’s Development Economics Vice Presidency Research Group as a Robert S. McNamara Fellow. During the first month, I got the opportunity to attend a training course on Reproducible Research Fundamentals where we were introduced to the importance of reproducibility and replicability. I learned how replicability matters for ensuring the credibility of research papers, building on the available knowledge, and using it to advance our understanding of the different research questions in development economics. It is worth mentioning that in this context the term ‘reproducibility refers to the ability to produce the same results as the research team by using the same dataset and code provided by them in their reproducibility packages.

As part of one of the projects I worked on during my fellowship, I reproduced 18 papers published in the American Economic Journal: Applied Economics, the American Economic Review, and the Journal of Development Economics. I re-ran regressions with around 750 outcomes and here is what I learned from this exercise:

1.       The coding style matters a lot in determining reproducibility. While 4 out of the 18 papers had only one do-file, the rest had multiple do-files ranging from 2 to 51 do-files. Papers with more than one do-file usually follow one of these approaches:

• Separate do-file for each stage (e.g., data cleaning, indicator construction, data analysis … etc.)
• Separate do-file for each phase of the research project (e.g., baseline, midline, endline, follow-up … etc.)
• Separate do-file for each figure and table in the final paper and the appendix.    
• Separate do-files for different tasks depending on the project
• Navigating many do-files can be very time-consuming, especially for someone that was not a member of the research team. What can make it easier/more efficient?
• Having a detailed README file that clearly specifies the purpose of each do-file as well as highlighting which do-file to use in case someone would like to replicate only one part of the paper
• Following clear and specific naming practices for the do-files and using informative names

2.       The content of the do-files is also of paramount importance. What makes it easier for non-team members to understand the code?

• Well-organized do-files divided into sections with clear titles.
• In the case of graphs and tables, use the same titles used in the final paper.
• Using informative variable names.
• Labelling the variables and using informative labels.
• Make sure the script is well-commented with the purpose of each section as well as the reasoning behind the code clearly stated.
• Using white spaces to facilitate the readability of the code.
• Reviewing code to avoid typos. It is crucial to run all the code from beginning to end before submitting the replication packages. This will help the team determine if there are any issues with the code and avoid any kind of typos.

3.       There is a clear trade-off between code efficiency (coding tasks are performed as simply as possible) and code readability (how easy it is for someone to understand the code).

a) The use of programs plays a role in ensuring higher code efficiency. However, sometimes sophisticated programs (such as ado programs) have lower readability for non-team members. Our study aimed at first running the results of these 18 papers and then exporting the outcomes to construct a dataset. While working with papers which wrote programs, the degree of transferability of the code (the extent to which it is easy for someone else to work on the same code) was lower.

b) The use of loops is a particularly useful way to avoid writing repetitive lines of code, which enhances code efficiency. But in some instances, loops were exceptionally long (spanning multiple lines of code) which reduces the readability of the code.

c) The use of locals and globals in Stata also contribute to enhancing code efficiency. However, do-files where there are multiple locals and/or globals with the same name can be confusing.

By looking at the scripts of 18 different research teams, I drew some lessons learned to improve my own coding style. First, it is important to work towards maintaining a good balance between code efficiency and code readability. Second, it is imperative to write extensive comments to increase readability of the code. Third, consciously consider what non-team members may not know about the project to ensure that the script will be as detailed and clear as possible. And finally, coding should not be considered a means to an end. In fact, coding is as important as all other stages of the research process in ensuring good research quality. You can also find DIME Research Reproducibility Standards here.

Coming soon: Along with David McKenzie and Jacobus Cilliers, we are currently studying the efficiency of using Post-Double Selection Least Absolute Shrinkage and Selection Operator (PDS LASSO) in selecting control variables in the case of field experiments. The paper aims at determining how much difference the use of machine learning tools makes in estimating coefficients, standard errors, and significance levels. In addition, we examine how many controls are selected by PDS LASSO and whether these controls predict the treatment variable, the outcome, or both. We hope to draw lessons for applied researchers designing and analyzing field experiments.

Nour Elashmawy