iiinova — where your digital self keeps evolving

Reposted by Will Matcham

Alex Usher @alexusherhesa.bsky.social · 28/08/2026

MIT put out a report on AI Use in Teaching, Learning, and Research Training. It's a) mostly pretty sensible and b) not really about tech. Good read.

aiandeducation.mit.edu

Report – AI and Education

076

Will Matcham @willmatcham.bsky.social · 28/08/2025

Made it to sunny Valencia and excited to attend @earie-news.bsky.social 2025 today and to present my research on credit card lenders tomorrow!

171

Will Matcham @willmatcham.bsky.social · 26/08/2025

Prof Mark Schankerman and I are hiring two full-time research assistants to work with us at LSE on research projects relating to productivity, innovation, and intellectual property rights. Apply here: jobs.lse.ac.uk/Vacancies/I/... Deadline 30th September 11:59 UK

jobs.lse.ac.uk

Research Assisant: ERC Research Grant on Incentives and Dynamics of Innovation

Research Assisant: ERC Research Grant on Incentives and Dynamics of Innovation, , <p style="text-align: center;"><em><span>LSE is committed to building a diverse, equitable and truly inclusive univers...

031

Will Matcham @willmatcham.bsky.social · 01/07/2025

Almost two years at @royalholloway.bsky.social now, and still nowhere near over how gorgeous this campus is.

130

Will Matcham @willmatcham.bsky.social · 21/03/2025

Before I submit any writing, I find and remove all instances of "It is worth mentioning that" (if it's not worth mentioning, why am I writing it??) and "It is important to note that" (convince the reader it's important through the substance, not empty words..). Pure academic throat clearing, gone.

010

Will Matcham @willmatcham.bsky.social · 10/03/2025

I am really looking forward to presenting and submitting this paper going forward, knowing that I have addressed this frequent comment. I can finally be at peace with the paper, and now just need to find a home for it. Wish me luck! (9/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

All other findings in the paper go through, including my favourite figure from the paper, included here. It shows that lenders use interest rate tailoring in the counterfactual to price discriminate, rather than control default risk (which is done separately through tailored credit limits). (8/10)

plot showing two features. First, under EU regulation, credit card interest rates are flat along customers' interest rate elasticity, for both high- and low-risk types. Second, when regulation is relaxed, lenders give higher interest rates for inelastic borrowers, and do this in the same way for high- and low-risk types.

110

Will Matcham @willmatcham.bsky.social · 10/03/2025

My estimated costs to lenders of individualizing interest rates are substantial, are higher at lenders known to have more sophisticated customers and correlate positively with individuals’ incomes. So, this change was not just to patch up my machinery, but generated new insights (7/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

Then, the counterfactual could be operationalized by turning off the regulatory constraint and setting these costs to zero. Now I am even more confident that the model and counterfactuals are internally consistent. (6/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

I finally cracked how to model lenders’ interest rate and credit limit choices subject to the regulation. The key trick was to include linear costs to lenders of individualizing interest rates, which I could estimate off the few cases where they did choose to deviate from the rate advertised. (5/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

While I sympathized with these comments, it was never obvious to me how I could give lenders the potential to individualize interest rates and credit limits in the baseline, subject to the regulation. That all changed at the start of this year, however.. (4/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

Prior to this revision, I was often confronted with the comment that my baseline model did not explain interest rate choices and was therefore not appropriate for a counterfactual in which I take away the regulation limiting lenders in individualizing interest rates. (3/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

In the paper I use statement-level data to show that UK lenders tailor credit limits, but NOT interest rates, to customers’ risk. They are inhibited by the costs of individualizing interest rates that come from EU-wide regulations requiring them to advertise an interest rate for each product (2/10)

100

Will Matcham @willmatcham.bsky.social · 10/03/2025

I have just posted a noteworthy revision of my paper “Risk-Based Borrowing Limits in Credit Card Markets.” The main novelty of the paper and its findings are the same. Why is the revision of interest, then? Read on to find out: (1/10)

230

Will Matcham @willmatcham.bsky.social · 11/12/2024

Cute detail in @jeffgortmaker.com's JMP: in a paper about open source software, pyBLP is used for the demand estimation, which is Jeff's own open source software! It's a great paper - you should take a look at it.

020