Jan Teichert-Kluge
Research
I develop double machine learning methods for causal inference when text and images are confounders, and I use these methods in empirical demand analysis, in particular to estimate price elasticities.
Selected papers
-
2026
Pragmatic DML with AI-Learned Representations
arXiv preprint
We study when AI-learned representations of text and images can be used as controls in causal analysis. With an imperfect representation, the target parameter is distorted by the product of two representation errors, and inference based on cross-fitted double machine learning is valid for the target defined by the representation.
-
2026
Valid Inference for Treatment Effects under Multimodal Confounding
Fifth Conference on Causal Learning and Reasoning (CLeaR 2026)
We develop a neural network architecture for the partially linear model in the double machine learning framework, with text and images as confounders, and we evaluate it on a semi-synthetic dataset with multimodal confounding.
Earlier version DoubleMLDeep: Estimation of Causal Effects with Multimodal Data (2024)
-
2024
Adventures in Demand Analysis Using AI
arXiv preprint · cemmap working paper
Using data on toy cars sold on Amazon, we represent each product by transformer-based embeddings that combine its text description and images with tabular covariates. With these embeddings, predictions of sales ranks and prices are more accurate and estimates of price elasticities are more credible.
In observational data on prices and sales, products with different prices also differ in attributes such as quality and branding, which are often recorded only in product descriptions and images. With Victor Chernozhukov, Martin Spindler and the DoubleML team, I work on neural network architectures for double machine learning and on the conditions under which AI-learned representations of text and images are valid controls in causal analysis.
Teaching
At the University of Hamburg Business School, I teach tutorials and seminars in statistics, statistical programming with Python, deep learning and causal inference, and I supervise seminar papers on large language models and foundation models. I also give workshops on causal machine learning outside the university, most recently the DoubleML trainings and a course at the summer academy of the Studienstiftung des deutschen Volkes.
Background
-
Current
Research stay at the Vector Institute
Toronto, Canada
-
2023–
Research associate and doctoral researcher, Statistics
University of Hamburg
-
2022–23
Data science, working student
Philips Medical Systems, Hamburg
-
2021–22
Business intelligence, working student
Siemens Gamesa, Hamburg
-
2020–23
M.Sc. Industrial Engineering
University of Hamburg and HAW Hamburg, thesis on causal machine learning with deep learning
-
2020
Hauke-Trinks-Award, 2nd place
Northern Institute of Technology Management
-
2016–20
B.Sc. Industrial Engineering
University of Hamburg and HAW Hamburg
Contact
jan.teichertkluge@uni-hamburg.de
University of Hamburg Business School, StatisticsMoorweidenstraße 18, Room 0025, 20148 Hamburg
+49 40 2395-21517