Our @voxdev.bsky.social blog about our recent EJ paper that develops a new method to analyze open-ended qualitative interviews with large samples without using LLMs (which we show can have extremely biased results with data like ours).
Why is intergroup hate rising across societies, all at once?
Our new paper, “The Cycle of Hate, and What We Can Do About It,” argues this it’s a self-reinforcing 10 reason cycle. We offer a diagnostic framework, and review the evidence on interventions. documents1.worldbank.org/curated/en/0...
We show in another paper in Sociological Methods and Research for data from non-Western contexts like ours LLM based qualitative analysis can be severely biased, and our bespoke method does much better. journals.sagepub.com/doi/10.1177/...
Many concepts—like aspirations, norms, and agency—are hard to capture with structured surveys. This work introduces a method that combines careful human coding of open-ended interviews with supervised NLP, allowing open-ended interviews to be analyzed at scale without losing interpretive depth.
New publication in The Economic Journal “Qualitative Analysis With Large-N: A New Method with An Application to Aspirations in Bangladesh.”https://academic.oup.com/ej/advance-article-abstract/doi/10.1093/ej/ueag005/8417166?redirectedFrom=fulltext Also available here vijayendrarao.org/wp-content/u...
Video of my farewell public lecture at the World Bank on ‘Policy for the People.’ Honored to have Deon Filmer chair, with insightful discussant comments from Nandini Krishnan & Robin Mearns. Grateful for all the warmth. youtu.be/_UzbwmOL80o
After 25 years at the World Bank’s Development Research Group, I’ll give my farewell talk — Policy for the People — on Nov 18 at 12:30 PM ET.
Honored to be joined by Nandini Krishnan, Robin Mearns & Deon Filmer.
🎥 worldbank.org/en/events/2025/11/18/policy-for-the-people
For sure. I plan to spend a lot of time
In Bangalore
This next phase is a continuation of what has always driven me — curiosity about how people create meaning and agency in their lives, and how social science can better honor, and practice, that complexity.
Grateful for the journey — and excited for what’s next.
I’ve been privileged to collaborate with extraordinary colleagues and partners in villages, governments, and research teams across the world — and to help build conversations between disciplines, and between people who don’t often speak to one another.
When I joined the Bank in 1999, I never imagined how much the institution — and I — would evolve. Working in the Bank’s research department has given me the freedom to explore how culture, politics, and economics intertwine in shaping people’s lives.
I will continuing my affiliations with CIFAR’s Boundaries, Membership & Belonging program and as a Non-Resident Fellow at the Center for Global Development.
I’ll be leaving the Bank effectively on December 15, 2025 (with my official departure on June 30, 2026) to make more time for independent research, writing, and teaching. From March to June 2026, I’ll be Distinguished Scholar-Practitioner in Residence at Northwestern University’s Buffet Institute.
Some news:
After 26 years in the World Bank’s Development Research Group, I have decided that it’s time for me to move on.
This year's #ABCDE2025, themed “Development in the Age of Populism,” sparked discussions on trade and geopolitics, poverty reduction, & more.
Following the conference, @eeshani.bsky.social shares what we learned—and didn’t—about populism & poverty in LMICs:
www.cgdev.org/blog/populis...
My cousin Rama Ranee changed my life—introducing me to yoga, meditation, and fieldwork in rural India.
For 35 years, she’s transformed a barren patch near Bangalore into a biodiverse, biodynamic sanctuary.
This stunning film by Good Food Movement captures her journey. youtu.be/W1XA-KOgKoA?...
Our study is the first predominantly qualitative paper accepted in The World Bank Economic Review.
It makes the case for listening carefully to what people say—especially when the numbers look good.
📄 Read it here: documents.worldbank.org/en/publicati...
Three key takeaways for development policy:
1️⃣ Growth is not enough.
2️⃣ We need to protect people from the market’s vagaries.
3️⃣ Reconnecting economic and social life is critical—especially amid rising populism.
Quantitative indicators—like GDP or consumption surveys—often miss this underside.
Open-ended conversations reveal the lived experiences and emotional undercurrents that data alone can’t capture.
This is what Hirsch meant by “social congestion”:
More money doesn’t buy better lives when education, housing, health, and leisure become rationed by class, and social networks fragment under economic strain.
Drawing on Karl Polanyi and Hirsch, we argue that economic growth can fracture the social fabric.
It prioritizes individual mobility over collective well-being—leading to anomie, congestion, and declining access to formerly free public goods.
Strikingly, the less developed western regions expressed less discontent.
There, people seemed more attuned to relative rather than absolute income—and less affected by what Fred Hirsch called “positional competition.”
We conducted 56 open-ended focus group interviews across Malaysia.
What we heard:
▪️ “Imbalance” between income & living costs
▪️ Reliance on multiple jobs
▪️ Rising debt
▪️ Stress & social disconnection
▪️ Ethnic polarization
Malaysia is often hailed as a success story of the “East Asian Miracle”:
✔️ High growth
✔️ Sharp poverty reduction
✔️ Falling inequality
But what do people themselves say about this progress?
Why do people feel worse even as their countries grow richer?
In an age of populism, we need to ask what policy misses when it focuses solely on growth.
A thread on our paper: Is There an Underside to Economic Growth? 👇
📄 documents.worldbank.org/en/publicati...
😂
Development folks (usually of a particular vintage) who are prone to making Denmark comparisons - note that you can fit six Denmarks in Delhi, five in Jakarta, four in Cairo….
Given that high quality codes are necessary in order to assess whether an LLM introduces bias, we argue that it may be preferable to train a bespoke model on a subset of transcripts coded by trained sociologists rather than use an LLM.
We find that using LLMs to annotate and code text can introduce bias that can lead to misleading inferences. By bias we mean that the errors that LLMs make in coding interview transcripts are not random with respect to the characteristics of the interview subjects.
We ask whether LLMs can help code and analyse large-N qualitative data from open-ended interviews, with an application to transcripts of interviews with Rohingya refugees and their Bengali hosts in Bangladesh.
