https://sukjinhan.com Sukjin Han Fri, 28 Nov 2025 16:50:21 +0000 https://sukjinhan.com en https://sukjinhan.com/mining_causality Mon, 16 Sep 2024 05:43:00 +0000 Sukjin Han https://sukjinhan.com/mining_causality Mining Causality: AI-Assisted Search for Instrumental Variables︎︎︎ arxiv︎︎︎&nbsp;&nbsp;slides︎︎︎ Aug 2025 Revise &amp; ResubmitJournal of Econometrics Finding IVs is a heuristic and creative process, and justifying exclusion restrictions is largely rhetorical. We propose using large language models (LLMs) to systematically search for new IVs through narratives and counterfactual reasoning. *This project is featured in the causal inference course at Stanford︎︎︎ and the&nbsp;Lindau Nobel Laureate Meetings 2025︎︎︎. https://sukjinhan.com/LGR Tue, 03 Oct 2023 17:48:02 +0000 Sukjin Han https://sukjinhan.com/LGR Estimating Causal Effects of Discrete and Continuous Treatments with Binary Instruments︎︎︎ with Victor Chernozhukov, Iván Fernández-Val, Kaspar Wüthrich arxiv︎︎︎&nbsp; slides︎︎︎ Dec 2024 Revise &amp; ResubmitEconometrica We identify average and quantile treatment effects for binary, ordered and continuous treatments with only binary IV under local copula invariance. The resulting semiparametric estimation procedures are very easy to implement. https://sukjinhan.com/partial_winners Tue, 21 Sep 2021 14:56:53 +0000 Sukjin Han https://sukjinhan.com/partial_winners Inference for Interval-Identified Parameters Selected from an Estimated Set︎︎︎ with Adam McCloskey arxiv︎︎︎ Mar 2025 Revise &amp; ResubmitQuantitative Economics We develop new inference tools for interval-identified welfare at a policy chosen from an estimated set (e.g., an estimated identified set). https://sukjinhan.com/font Tue, 21 Sep 2021 14:08:14 +0000 Sukjin Han https://sukjinhan.com/font Shapes as Product Differentiation︎︎︎ with Eric Schulman, Kristen Grauman,&nbsp;Santhosh Ramakrishnan arxiv︎︎︎ &nbsp;slides︎︎︎ Nov 2022 Revise &amp; Resubmit RAND Journal of Economics Many differentiated products have key attributes that are high-dimensional (e.g., design, text). We consider one of the simplest design products, fonts, and quantify their shapes by constructing neural network embeddings. Using the embeddings and data from the world's largest online market place for fonts, we study the causal effect of a merger on the merging firm's creative decisions of product differentiation. *This project is featured in a typography magazine︎︎︎ and included in the MIT graduate machine learning course︎︎︎. <img width="3152" height="1689" width_o="3152" height_o="1689" data-src="https://freight.cargo.site/t/original/i/05371e7389269b09e39334ef63f6a2389ec4d752ae2068a929f6ff55640a4088/tsne_sample400_perp70_exg30_edited.png" data-mid="119531749" border="0" src="https://freight.cargo.site/w/1000/i/05371e7389269b09e39334ef63f6a2389ec4d752ae2068a929f6ff55640a4088/tsne_sample400_perp70_exg30_edited.png" /> <img width="1216" height="603" width_o="1216" height_o="603" data-src="https://freight.cargo.site/t/original/i/785d78957a2398185b59aec583a0ab914f42f76271e94561b4c1dd95c77938f6/Supply_side_Mean_Distances_Time_Plot_02252020.png" data-mid="119531780" border="0" src="https://freight.cargo.site/w/1000/i/785d78957a2398185b59aec583a0ab914f42f76271e94561b4c1dd95c77938f6/Supply_side_Mean_Distances_Time_Plot_02252020.png" /> https://sukjinhan.com/copyright Sat, 24 Feb 2024 18:08:04 +0000 Sukjin Han https://sukjinhan.com/copyright Copyright and Competition: Estimating Supply and Demand with Unstructured Data︎︎︎ with Kyungho Lee arxiv︎︎︎&nbsp; ssrn︎︎︎&nbsp; slides︎︎︎ Sep 2025 Extended abstract at EC’25 To understand the role of copyright policy in the presence of generative AI in markets for products with visual attributes, we estimate a structural model of supply (e.g., product positioning) and demand (e.g., tastes for visual attributes) using image data. Visual similarity, calculated using neural network embeddings, serves as a crucial metric for the analysis.*Media: KBS Radio 1︎︎︎ https://sukjinhan.com/LLM_eval Fri, 28 Nov 2025 16:50:21 +0000 Sukjin Han https://sukjinhan.com/LLM_eval Can LLMs Predict Human Behavior? A Measure of their Pretrained Knowledges︎︎︎ with Wayne Gao, Annie Liang arxiv︎︎︎ Jan 2026 new draft coming soon We measure the predictive capability of an LLM by the size of the training sample required for a conventional ML or economic model to match its predictive performance. We derive the inference theory associated with this measure. https://sukjinhan.com/synthetic_blip Sat, 09 Aug 2025 21:13:31 +0000 Sukjin Han https://sukjinhan.com/synthetic_blip Synthetic Blips: Generalizing Synthetic Controls for Dynamic Treatment Effects︎︎︎ with Anish Agarwal, Dwip Saha, Vasilis Syrgkanis, Haeyeon Yoon arxiv︎︎︎ Nov 2025 new draft coming soon By generalizing synthetic controls, we identify and estimate dynamic individual treatment effects for policy evaluation and individualized recommendation. The procedure is applied to export financial support using unique Korean firm-level data. https://sukjinhan.com/BCE Tue, 21 Sep 2021 15:51:58 +0000 Sukjin Han https://sukjinhan.com/BCE Testing Information Ordering for Strategic Agents︎︎︎ with Hiro Kaido, Lorenzo Magnolfi arxiv︎︎︎ &nbsp;slides︎︎︎ Jan 2026 We develop a test of information ordering to examine if the true information structure is at least as informative as a proposed baseline. We utilize the notion of Bayes Correlated Equilibrium (BCE). <img width="2430" height="1822" width_o="2430" height_o="1822" data-src="https://freight.cargo.site/t/original/i/87d2888ca2044a3c1f371f01200b29b1a0989589dde7935da6ac7d0955c1fba5/BCE_fig1_small.jpeg" data-mid="228798583" border="0" src="https://freight.cargo.site/w/1000/i/87d2888ca2044a3c1f371f01200b29b1a0989589dde7935da6ac7d0955c1fba5/BCE_fig1_small.jpeg" /> <img width="2430" height="1822" width_o="2430" height_o="1822" data-src="https://freight.cargo.site/t/original/i/5156c8ddf9174f5caf84ab3c12c5d98e10dc1b1e9e5ce4127bdb55aceec37e35/BCE_fig2_small.jpeg" data-mid="228798584" border="0" src="https://freight.cargo.site/w/1000/i/5156c8ddf9174f5caf84ab3c12c5d98e10dc1b1e9e5ce4127bdb55aceec37e35/BCE_fig2_small.jpeg" /> https://sukjinhan.com/Set-CF Thu, 18 Jan 2024 10:41:11 +0000 Sukjin Han https://sukjinhan.com/Set-CF Set-Valued Control Functions︎︎︎ with Hiro Kaido arxiv︎︎︎ &nbsp;slides︎︎︎ Feb 2025 new draft coming soon The classic control function approach requires invertibility or point-identified controls, which limits its applicability (e.g., discrete treatments, controls as intervals). We allow the control function to be set-valued and derive sharp bounds on structural parameters. https://sukjinhan.com/multi_IVs Tue, 08 Mar 2022 16:40:59 +0000 Sukjin Han https://sukjinhan.com/multi_IVs On Quantile Treatment Effects, Rank Similarity, and the Variation of IVs︎︎︎ with Haiqing Xuarxiv︎︎︎&nbsp; slides︎︎︎ Aug 2023 We provide a novel interpretation for rank similarity, which motivates us to relax it in a way that is useful to bound the QTEs using multi-valued IVs. The bounds are easy to calculate in practice and are shown to be informative.