← Back to all news
arXiv Machine Learning By Anish Agarwal, Sukjin Han, Dwaipayan Saha, Vasilis Syrgkanis, Haeyeon Yoon

Synthetic Blips: Generalizing Synthetic Controls for Dynamic Treatment Effects

Read the original on arXiv Machine Learning →

The paper introduces "synthetic blips," a generalization of synthetic control methods for settings with dynamic treatment effects where units receive multiple sequential treatments based on an adaptive policy influenced by a latent, time‑varying confounding state. Under a low‑rank latent factor model, the authors develop an identification strategy for unit‑specific mean outcomes under any intervention sequence, using a backward induction process that expresses each treatment’s blip effect as a linear combination of other units’ blip effects, thereby avoiding combinatorial donor requirements. They provide estimation algorithms that yield consistent estimators and apply the method to Korean firm‑level panel data to estimate individualized dynamic treatment effects and optimal allocation rules for financial support to exporting firms.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

Related stories

arXiv AI
Sep 18

Risk-Set Transported Synthetic Control with Difference-in-Differences Adjustment under Staggered Treatment Adoption

The paper introduces Risk-Set Transported Synthetic Control with Difference-in-Differences Adjustment (RT‑SC‑DiD), a method for staggered treatment‑adoption studies that keeps the donor pool fixed by reallocating weights from exiting donors to similar surviving donors while applying a DiD baseline correction. It analyzes distortion from horizon‑by‑horizon re‑optimization, derives bounds on error propagation, and proposes diagnostics and a donor‑only placebo for tuning the transport penalty. Empirical simulations show that intermediate transport regularization reduces average RMSE compared to independent horizon‑specific estimation and strong anchoring, supporting the method’s bias‑variance trade‑off. "whyItMatters":"The method offers a principled way to stabilize synthetic‑control weights over time in staggered designs, potentially improving causal inference when donor support contracts as treatments roll out."

By Mojtaba Eslami
safety
More like this →
arXiv Statistics ML
3d ago

Causal Inference in Possibly Nonlinear Factor Models

The paper introduces a causal inference method for treatment effect models where confounders are measured noisily. It leverages many noisy proxies linked to latent confounders through an unknown, possibly nonlinear factor structure, using a local principal subspace approximation that combines K‑nearest‑neighbor matching and principal component analysis. The authors construct doubly‑robust estimators for various causal parameters, establish their large‑sample properties, and provide uniformly consistent estimators of the conditional average treatment effect, illustrated with an empirical study on political connections and stock returns and a Monte Carlo experiment.

By Yingjie Feng
More like this →