@rmetcalfe
Economist, Prof at Columbia University. Chief Economist: Centre for Net Zero (Octopus Energy Group). Co-editor: Journal of Public Economics. 1st gen, π΄σ §σ ’σ ·σ ¬σ ³σ Ώ https://www.rmetcalfe.net/
Applications are now open for the 2026 Price Theory Summer Camp!
Designed for PhD students outside UChicago, the program focuses on applying core economic tools to real problems.
ποΈ June 14β18
π Apply by March 1
π https://ow.ly/k6ON50Y8mQt
AI can cut peak electricity demand at scale. A large natural field experiment shows AI-managed EV charging cuts peak load 42 percent and shifts usage to cleaner hours, from @rmetcalfe.bsky.social, Andrew Schein, Cohen R. Simpson, and Yixin Sun www.nber.org/papers/w34709
Best workshop, best keynote and best time to visit London! Submit your paper/abstract. Deadline Feb 1.
Weβre very grateful to colleagues at the partner utility @octopus.energy for their work in delivery, and to many insightful commenters on earlier drafts.
www.nber.org/papers/w34709
Undated: www.rmetcalfe.net/_files/ugd/f...
Comments welcome!
We developed a theoretical model showing that low override rates make AI-managed charging welfare-superior to alternatives, including real-time pricing. The trial confirmed this prediction: empirical behavior matched theory, given the very low override rate we observed.
Generalizability: cross-market evidence suggests this algorithm scales. Similar plug-in behavior and override patterns in the UK, US, Germany, and Spain imply that AI-managed tariffs can deliver welfare gains internationally.
Economic & environmental benefits were substantial. Households saved Β£343/year (~18%); full adoption could flatten the load duration curve. COβ abatement cost was extremely low given; subsidizing the tariff could generate infinite MVPF.
Peak demand fell sharply. IV estimates show a 42% reduction in peak-period EV charging, with energy shifted to overnight hours. Total consumption stayed constant. In other words, weβre seeing temporal load shifting, not increased use.
Once enrolled, adherence to the algorithm was high. Daily overrides were rare (~1%), and only 2.3% of total EV energy was consumed outside AI control. This low overriding is important for assessing the welfare impact of the tariff.
We randomized an encouragement to adopt the tariff. Even a simple email increased enrolment in the AI tariff by 3.4pp; Β£50/month incentives roughly doubled this. Compatibility constraints mean these estimates are likely lower bounds.
The AI does a lot under the hood, optimizing over various energy market signals subject to user preferences and behaviors (some of which need to be predicted, such as plug-in and plug-out).
It combines this with a simple time-of-use set-up β any AI managed charging is at the low rate. The simplicity here is that the consumer does not have to optimize over these pricesβthe AI algorithm will do so on their behalf.
We examined the adoption of an AI-managed EV charging tariff (called IO Go) run by Octopus Energy, the largest energy retailer in the UK. Customers set their preference. The AI manages charging within those preferences (overrides β βbumpsβ β are allowed).
AI management of home energy use is something people have been talking about for decades. Hereβs Fred C Schweppe from MIT in 1980(!)
Mass EV adoption risks overwhelming electricity networks, increasing system costs and grid instability. You can see this in a graph of consumption profile of customers in our field trial before the trial started. EV customersβ consumption peaks during the system peak.
New @nber.org paper by myself, @the-scheining.bsky.social, Yixin Sun from @centrefornetzero.bsky.social, & Cohen Simpson. We study the largest field experiment on AI-managed EV charging, showing that it can reduce peak demand, lower costs, and cut emissions without disrupting household routines.
I attended the lecture by Hendren and Sprung - Keyser and it was excellent!
The AEA has posted eight "Recent Developments" lectures exploring highly topical issues in economics, presented by the best scholars in the field:
www.aeaweb.org/conference/w...
Well worth a watch!
Iβm really looking forward to this and hope youβll consider joining us!
Excited to announce that this year's Advances with Field Experiments conference will take place at the University of Chicago on September 17-18, 2026.
@johnlist.bsky.social and I will send out a call for abstracts early in the Spring.
bfi.uchicago.edu/events/event...
@katymilkman.bsky.social
Very excited to join one of the best environmental economics workshops out there.
π’ Call for Papers! The 9th Annual LSE/Imperial Workshop on Environmental Econ. will be held in London June 8-9
β
Submissions by Feb 1
π Full papers only (no abstracts)
π€ Keynote: Prof. @rmetcalfe.bsky.social
π more details here: www.lse.ac.uk/geography-an...
I look forward to catching up with folks at the Columbia Economics #ASSA2026 reception on Saturday evening in Marriott Downtown, Room 407.
www.aeaweb.org/conference/2...
Expanding on this. The rise of LLMs push the role of data scientists closer to systems engineers and away from coders. Systems engineering is about turning competing stakeholder demands into consensus criteria for judging a solution and coming up with ways to verify whether criteria have been met.
A large UK field experimentβ110,000 drivers, 60 percent of public chargersβfound that cutting prices by 40 percent increased charging by 117 percent. Consumers respond fast, large grid benefits, from Bernard, Hackett, @rmetcalfe.bsky.social, Panzone, and Schein www.nber.org/papers/w34600
New @nberpubs: "The Impact of Dynamic Prices on Electric Vehicle Public Charging Demand: Evidence from a Nationwide Natural Field Experiment" www.nber.org/papers/w34600
"Our findings suggest that dynamic pricing for public EV charging generated large consumer welfare gains."
Hannah's work shows that, when the minimum wage increases, schedule unpredictability goes up too. Perhaps not surprising (but very hard to show)--as the wage goes up, workers pay in other ways. She is on the market and has a number of very cool and important papers!
hannahfarkas.github.io
Super excited to see Hannah's JMP in the Economist this week! She uses detailed data to show how firms pass the risk of weather-driven shift changes on to workers in food & bev / retail, & that minimum wages actual *worsen* that effect!
#econsky #climsky #climate #jmp #jmc #susdev
Figure shows average hours worked per week by industry. It is 25.8 in leisure and hospitality, compared to an average of 34.7 for the whole private sector.
1. Schedule unpredictability means you don't work full time, but can't fill the other hours.
Avg hours worked per year in hospitality is 26, compared to 35-40 in most other industries.
But workers can't fill the spare hours with another job as they don't know when they'll be scheduled to work.
