Tushar Kundu
@tushardonnay
Followers
598
Following
10K
Media
44
Statuses
949
Economics PhD student @columbia_econ. @swarthmore '17.
Alameda, CA
Joined December 2010
1/ In my JMP, I ask: Why do different kids specialize in different skills? In the same classroom you have: – the math whiz – the leader – the quiet empath Is that because they chose different skills to invest in… or because some skills are just harder for them to build? 🧵👇
1
29
134
Most young adults rely on family support as they enter adulthood, but youth who age out of foster care at 18 often lack that support. In her JMP, Kate Musen (@khmusen) finds that letting these youth stay in care until 21 substantially improves college and labor market outcomes.
1
13
33
In his JMP, Ricardo Pommer Muñoz shows that the difficulty inferring product quality can lead to under-demand of beneficial climate adaptive technologies. This friction also increases supplier incentives to provide low quality goods, making markets unsustainable.
1
8
22
Nikhil is smart, and cares deeply about using research to do good and shed light on the issues facing the most vulnerable. Some have said he is a “presence” and “will be missed” when he leaves Columbia. If you’re looking for someone like that in your workplace, hire Nikhil!
Time flies. I’m on the job market! Relatedly, if you’re at #APPAM, stop by and say hi at tomorrow’s evening reception, where I’ll have a poster on my JMP.
1
1
19
Unsustainable groundwater use in India threatens the livelihoods of millions and reflects a classic tragedy of the commons. In his JMP, @nikhilbasavappa shows externality severity may both justify an intervention—irrigation efficiency upgrades—and determine whether it backfires.
1
9
31
In his JMP, Tushar Kundu (@tushardonnay) studies why students specialize in different skills. In 5 Indian private schools, he surveys 3,404 parents to elicit perceptions of their children's academic and socioemotional skills, and their priorities over which skills to improve.
1
6
25
Currently in the process of writing hundreds of cover letters, and I’d be lying if I said it didn’t feel super wasteful. Never been a better time to rethink the hiring pipeline. LLMs wipe out the cover letter signal, but can they be used to design new types of signals?
What happens when online job applicants start using LLMs? It ain't good. 1. Pre-LLM, cover letter quality predicts your work quality, and a good cover gets you a job 2. LLMs wipe out the signal, and employer demand falls 3. Model suggests high ability workers lose the most 1/n
0
3
29
My full job market paper is posted here: https://t.co/L8CtP9PnxW
4
0
9
10/ If you made it this far, check out some of my other work! https://t.co/4Uz3gqLVLo I work on education and labor markets in India and beyond, designing and testing tools (including AI) to help schools, employers, and policymakers act on what people actually value.
tusharkundu.github.io
Economist
1
0
20
9/ Big picture: If you only look at test scores, you miss welfare. Families disagree on which skills matter most for their child, and teachers don’t automatically know that. But when teachers are shown what parents want, they adjust.
1
2
6
8/ What happens? In classes where teachers get the info: – Parents later report more improvement in the exact skill they cared most about – That skill drops in their priority list by the end (because it got better) So teachers do shift effort once they see what families want.
1
1
8
7/ So I test a fix. Some teachers get access to a web portal I built. They log in and see, for each child: – the child’s skill profile – which skills that child’s parent wants to improve Other teachers don’t get this info. Everything else in school stays the same.
1
1
5
6/ Do teachers already know what each family wants? I ask teachers: – for specific students in your class, what do you think their parents want most? I compare that to what parents actually said. Result: basically no correlation.
1
1
6
5/ My model says this points to learning costs. If costs drive specialization, parents should say “work on the weak skill next.” Arrows point toward the weak side. If benefits drive it, you don’t get that pattern. In the data, you get the cost pattern.
1
0
6
4/ What I expected: “Everyone says math.” And yes, lots of parents say math. But many say things like empathy, perseverance, leadership. Instead, the prevailing pattern: parents usually want to improve whatever their child is *weakest* at.
1
0
4
3/ I partner with 5 private schools in India to ask parents: How is your child doing in academic, social, and emotional skills? Which is most important to improve next? So for each child I see: – their skill profile – what skills they want next That data rarely exists.
1
0
3
2/ I highlight two forces: Benefits: Some skills might just be more valuable for a given child. Costs: Some skills are just harder for that child to improve. Policy usually asks “how much did test scores go up?” I ask: “which skills improved, and why?”
1
0
3
I can't close this chapter without giving credit where credit is due. I went back to the acknowledgments I wrote in my dissertation and wanted to share them here. It’s a bit long, so I’ll post it in parts. Thanks to my family:
1
1
22
I was today years old when I learned the difference between en dash and em dash
0
0
0
The contrast between Trump and Reagan couldn't be any sharper. Tear down this wall vs building walls, pro-trade vs most anti-trade president ever, immigration amnesty vs hunting immigrants, democracy vs love for authoritarians. And, yet, these clowns side with Trump
9
23
160