Matrix Completion Methods for Causal Panel Data Models

DOI10.1080/01621459.2021.1891924MaRDI QIDQ5881958zbMATH OpenOpenAlexWikidataFDO

Authors Susan Athey, Mohsen Bayati, Nikolay Doudchenko, Guido W. Imbens, Khashayar Khosravi

Publication date 14 March 2023

Published in Journal of the American Statistical Association (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1710.10251

zbMATH Keywords

synthetic controlscausalitylow-rank matrix estimationunconfoundednessinteractive fixed effects

Mathematics Subject Classification ID

Matrix completion problems (15A83)

Abstract: In this paper we study methods for estimating causal effects in settings with panel data, where some units are exposed to a treatment during some periods and the goal is estimating counterfactual (untreated) outcomes for the treated unit/period combinations. We propose a class of matrix completion estimators that uses the observed elements of the matrix of control outcomes corresponding to untreated unit/periods to impute the "missing" elements of the control outcome matrix, corresponding to treated units/periods. This leads to a matrix that well-approximates the original (incomplete) matrix, but has lower complexity according to the nuclear norm for matrices. We generalize results from the matrix completion literature by allowing the patterns of missing data to have a time series dependency structure that is common in social science applications. We present novel insights concerning the connections between the matrix completion literature, the literature on interactive fixed effects models and the literatures on program evaluation under unconfoundedness and synthetic control methods. We show that all these estimators can be viewed as focusing on the same objective function. They differ solely in the way they deal with identification, in some cases solely through regularization (our proposed nuclear norm matrix completion estimator) and in other cases primarily through imposing hard restrictions (the unconfoundedness and synthetic control approaches). The proposed method outperforms unconfoundedness-based or synthetic control estimators in simulations based on real data.

Recommendations

• Matrix Completion, Counterfactuals, and Factor Analysis of Missing Data
• Bayesian matrix completion approach to causal inference with panel data
• Robust synthetic control
• Collective matrix completion
• Matrix completion with covariate information

Cites work

• scientific article; zbMATH DE number 3136275 (Why is no real title available?)
• scientific article; zbMATH DE number 3938400 (Why is no real title available?)
• scientific article; zbMATH DE number 845714 (Why is no real title available?)
• A simpler approach to matrix completion
• A unified framework for high-dimensional analysis of M-estimators with decomposable regularizers
• About the constants in Talagrand's concentration inequalities for empirical processes.
• Causal inference for statistics, social, and biomedical sciences. An introduction
• Determining the Number of Factors in Approximate Factor Models
• Dynamic linear panel regression models with interactive fixed effects
• Efficient Estimation of Average Treatment Effects Using the Estimated Propensity Score
• Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure
• Estimation of (near) low-rank matrices with noise and high-dimensional scaling
• Estimation of high-dimensional low-rank matrices
• Exact matrix completion via convex optimization
• Inferential Theory for Factor Models of Large Dimensions
• Linear regression for panel with unknown number of factors as interactive fixed effects
• Longitudinal data analysis using generalized linear models
• Matched Sampling for Causal Effects
• Matrix Completion From a Few Entries
• Matrix completion and low-rank SVD via fast alternating least squares
• Matrix completion from noisy entries
• Noisy low-rank matrix completion with general sampling distribution
• Nuclear-norm penalization and optimal rates for noisy low-rank matrix completion
• Panel data models with interactive fixed effects
• Recovering Low-Rank Matrices From Few Coefficients in Any Basis
• Restricted strong convexity and weighted matrix completion: optimal bounds with noise
• Robust synthetic control
• Spectral regularization algorithms for learning large incomplete matrices
• Statistical inference for average treatment effects estimated by synthetic control methods
• The Power of Convex Relaxation: Near-Optimal Matrix Completion
• The central role of the propensity score in observational studies for causal effects
• User-friendly tail bounds for sums of random matrices

Cited in

(34)

• Inference for large dimensional factor models under general missing data patterns
• Dynamic treatment effect estimation with interactive fixed effects and short panels
• Color image inpainting via robust pure quaternion matrix completion: error bound and weighted loss
• Direct and spillover effects of a new tramway line on the commercial vitality of peripheral streets: a synthetic-control approach
• Indian Buffet process factor model for counterfactual analysis
• Statistical Inference For Noisy Matrix Completion Incorporating Auxiliary Information
• Bayesian matrix completion approach to causal inference with panel data
• Revisiting event-study designs: robust and efficient estimation
• Individualized dynamic latent factor model for multi-resolutional data with application to mobile health
• High-dimensional latent panel quantile regression with an application to asset pricing
• A Penalized Synthetic Control Estimator for Disaggregated Data
• Imputation of Counterfactual Outcomes when the Errors are Predictable
• A new hybrid regularization scheme for removing salt and pepper noise
• Robust density power divergence estimates for panel data models
• Inference for Low-Rank Models Without Estimating the Rank
• Inference for low-rank models
• Synthetic learner: model-free inference on treatments over time
• Confidence intervals of treatment effects in panel data models with interactive fixed effects
• Inference for low-rank completion without sample splitting with application to treatment effect estimation
• Causal models for longitudinal and panel data: a survey
• The policy is always greener: impact heterogeneity of Covid-19 vaccination lotteries in the US.
• Counterfactual inference in sequential experiments
• Matrix Completion, Counterfactuals, and Factor Analysis of Missing Data
• Assessing the causal effect of binary interventions from observational panel data with few treated units
• Treatment effects in interactive fixed effects models with a small number of time periods
• Doubly robust proximal synthetic controls
• Estimating the effects of a California gun control program with multitask Gaussian processes
• Nuclear norm regularized quantile regression with interactive fixed effects
• Noisy matrix completion for longitudinal data with subject- and time-specific covariates
• On the Properties of the Synthetic Control Estimator with Many Periods and Many Controls
• Nonrandom exposure to exogenous shocks
• Synthetic control as online linear regression
• Same root different leaves: time series and cross-sectional methods in panel data
• Perspectives on optimizing transport systems with supply-dependent demand

Describes a project that uses

Uses Software

• softImpute

This page was built for publication: Matrix Completion Methods for Causal Panel Data Models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5881958)