Dylan T. Moore

@dylantmoore.bsky.social

Tax economist (optimal taxation, bunching methods, behavio(u)ral PF, political economy). Canadian expat. Assistant Professor at the University of Hawaiʻi Economic Research Organization (UHERO) & University of Hawaiʻi at Mānoa Department of Economics.

My tests of Deepseek so far indicate consistently weaker performance than o1-pro & also regular o1 on basic applied theory tasks in econ. Often o1-mini also outperforms it. More interestingly, Gemini's latest "thinking" model outperforms Deepseek on my tasks more often than not.

Bild

In economics, editors, referees, and authors often behave as if a published paper should reflect some kind of authoritative consensus. As a result, valuable debate happens in secret, and the resulting paper is an opaque compromise with anonymous co-authors called referees. 1/

Bild

Let me say though, when I asked it to present the derivation of the standard optimal nonlinear income tax formula using the perturbation approach (formally, using Gateaux derivatives, etc) it failed, ignoring the behavioral response effects of tax reforms. Claude 3.5 Sonnet nailed it.

Dylan T. Moore@dylantmoore.bsky.social · 2y ago

If you do applied theory in econ, and you haven't tried using ChatGPT o1 yet, give it a go. It is much better than earlier models at doing math reliably. Deriving and interpreting FOCs. Doing simple comparative statics. Developing model extensions. It is good enough to be very useful!

If you do applied theory in econ, and you haven't tried using ChatGPT o1 yet, give it a go. It is much better than earlier models at doing math reliably. Deriving and interpreting FOCs. Doing simple comparative statics. Developing model extensions. It is good enough to be very useful!