---
title: Machine Learning for Causal Inference: Is a Nonlinear First Stage Really Forbidden in 2SLS?
url: https://www.ml-quant.com/papers/ssrn/4772060/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: SSRN 4772060
source_url: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4772060
featured: 2024-03-27
citations: 0
topic: Econometrics & Forecasting
---


# Machine Learning for Causal Inference: Is a Nonlinear First Stage Really Forbidden in 2SLS?

The paper shows that the bias in the two-stage least squares estimator can be split into an observable and unobservable bias, without needing to specify the first stage's functional form or validate the instrumental variable.

- Source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4772060
- Identifier: SSRN 4772060
- Released: 2024-03-25
- First featured: Quant Letter No. 42 (2024-03-27): https://www.ml-quant.com/issues/2024-03-27/
- Citations (Semantic Scholar): 0
- Published in: not yet
- Topic: Econometrics & Forecasting

## Related

- [Robust agents learn causal world models](https://www.ml-quant.com/papers/arxiv/2402.10877/): The research suggests that intelligent agents must learn an approximate causal model to generalize to new domains, impacting fields like transfer learning and causal inference.
- [Expert-aided causal discovery of ancestral graphs](https://www.ml-quant.com/papers/arxiv/2309.12032/): A novel method for causal inference uses ancestral graph sampling and expert feedback to refine causal discovery, providing uncertainty estimates and accounting for unobserved confounders.
- [Exploring the heterogeneous impacts of Indonesia’s conditional cash transfer scheme (PKH) on maternal health care utilisation using instrumental causal forests](https://www.ml-quant.com/papers/arxiv/2501.12803/): The research uses machine learning to study the effects of Indonesia's conditional cash transfer scheme on maternal health care, finding significant variations based on supply-side factors and poverty indicators.
- [The impact of extracurricular education on socioeconomic mobility in Japan: an application of causal machine learning](https://www.ml-quant.com/papers/arxiv/2506.07421/): A machine learning study found that private tutoring in Japan can have positive socioeconomic impacts, but these are undermined by economic disparities among households.
- [Double/Debiased ML in Stata](https://www.ml-quant.com/papers/repec/tsj-stataj-v-24-y-2024-i-1-p-3-45/): The article presents a package named ddml for double/debiased machine learning in Stata, supporting estimators of causal parameters for five econometric models and is compatible with various supervised machine learning programs.
- [Epistemic Limits of Empirical Finance: Causal Reductionism](https://www.ml-quant.com/papers/ssrn/4646664/): Causal Reductionism: The research criticizes the use of one-way causation in capital market studies, suggesting that current quantitative finance tools may only be suitable for after-the-fact causal inference.