This is an update to our original GiveCalc announcement from last year. TL;DR: GiveCalc is a free, open-source calculator that computes how charitable donations affect your taxes. It runs full tax simulations to account for interactions between donations, deductions, and credits. Supports US federal/state taxes and UK Gift Aid. What...
Testing Cross-Model Consistency in Economic and Policy Predictions Building on Pham & Cunningham (2024)'s findings on narrative-prompted predictions, I conducted a large-scale experiment comparing how different LLMs forecast economic and policy outcomes under different electoral scenarios. GPT-4o, GPT-4o-mini, and Grok all projected better outcomes under a Harris administration than a...
This is an update to our original GiveCalc announcement from last year.
TL;DR: GiveCalc is a free, open-source calculator that computes how charitable donations affect your taxes. It runs full tax simulations to account for interactions between donations, deductions, and credits. Supports US federal/state taxes and UK Gift Aid.
Multiplying donations by your marginal tax rate misses interactions that affect actual tax savings:
🇺🇸 US considerations:
🇬🇧 UK considerations:
Only if you itemize rather than taking the standard deduction!
For example, suppose you purchase a binary contract that pays $1 if an event occurs (and $0 if it doesn't) and you believe there's a 50% chance of the event. If you win, you receive $1 in gross winnings. With a 30% marginal tax rate and without itemizing, you'd pay tax on the full $1—leaving you with a net of $0.70. Given that you paid some amount for the contract, your net gain on a win is , while a loss means you lose the entire .
To break even, the expected value of the bet must be zero:
This simplifies to:
Thus, if you believe the event is 50% likely (and considering only taxation, not other factors like transaction fees or opportunity costs), you would only gain if you paid under $0.35 for the contract.
Building on Pham & Cunningham (2024)'s findings on narrative-prompted predictions, I conducted a large-scale experiment comparing how different LLMs forecast economic and policy outcomes under different electoral scenarios. GPT-4o, GPT-4o-mini, and Grok all projected better outcomes under a Harris administration than a Trump administration, and Grok had the largest gaps.
To elicit these forecasts, I prompted GPT-4o, GPT-4o-mini, and Grok to simulate the reporting of 2025 metrics under different administrative scenarios. I ran 500 trials per model/metric using narrative prompting techniques shown to improve LLM forecast accuracy. For example, I used this prompt for real GDP per capita:
... (read 274 more words →)Write a scene from January 2026 where a senior
Even trickier is the canonical obviously correct but deeply unpopular policy, the Carbon Tax, or its more accepted incomplete alternative the Gasoline Tax. Insanely, there are calls for a gas tax holiday or other cut, at exactly the time when we need to reduce consumption. That’s why the price is going up. A price is a signal wrapped up in an incentive.
Carbon taxes actually poll strongly: an average of +36 net favorability across the 8 polls I've seen since 2020.
This might not be the time to push a carbon tax, but as you say, avoiding gas tax holidays would be similarly good. I created a campaign to do this, starting in California. We're currently called Save the Gas Tax, but will probably rename to Stop Oil Subsidies soon.
Following in the UK's footsteps with household transfers would be a much better use of funds if we want to spend to reduce pressure at the pump.
I generally agree with this, but I'd like to check my optimism against some forecasting questions on US immigration policy with respect to the war. I don't see anything on Metaculus here.
Thanks for this post. To clarify, it sounds like your thesis is about duration of sleep, not quality of sleep? How's the evidence for sleep quality and life outcomes? (And what interventions improve it?)