Endogenous Selection Bias: The Problem of Conditioning on a Collider Variable

2014/06/06 by Felix Elwert, Christopher Winship · 2 voices · 53 citations
Decision Sciences · Mathematics · #Advanced Causal Inference Techniques #Game Theory and Applications #Statistical Methods and Inference

paper · pdf · doi:10.1146/annurev-soc-071913-043455

openalex publication_date 2014/06/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/11

Abstract

Endogenous selection bias is a central problem for causal inference. Recognizing the problem, however, can be difficult in practice. This article introduces a purely graphical way of characterizing endogenous selection bias and of understanding its consequences (Hernán et al. 2004). We use causal graphs (direct acyclic graphs, or DAGs) to highlight that endogenous selection bias stems from conditioning (e.g., controlling, stratifying, or selecting) on a so-called collider variable, i.e., a variable that is itself caused by two other variables, one that is (or is associated with) the treatment and another that is (or is associated with) the outcome. Endogenous selection bias can result from direct conditioning on the outcome variable, a post-outcome variable, a post-treatment variable, and even a pre-treatment variable. We highlight the difference between endogenous selection bias, common-cause confounding, and overcontrol bias and discuss numerous examples from social stratification, cultural sociology, social network analysis, political sociology, social demography, and the sociology of education.

Citations

• Association, Causation, And Marginal Structural Models
• The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations.
• Estimating causal effects of treatments in randomized and nonrandomized studies.
• Sample Selection Bias as a Specification Error
• Homophily and Contagion Are Generically Confounded in Observational Social Network Studies
• Graphical Causal Models
• The Consquences of Adjustment for a Concomitant Variable That Has Been Affected by the Treatment
• Direct and Indirect Effects
• Causal Diagrams for Epidemiologic Research

Cited by

• Men and Women’s Employment Status and Union (In)stability: Does Contextual Gender Equality Matter?
• Has the Credibility of the Social Sciences Been Credibly Destroyed? Reanalyzing the “Many Analysts, One Data Set” Project
• Through the lens of causal inference: Decisions and pitfalls of covariate selection
• The long-term effects of childhood residential mobility on social capital
• PIPCFR: Pseudo-outcome Imputation with Post-treatment Variables for Individual Treatment Effect Estimation
• What Is Your Estimand? Defining the Target Quantity Connects Statistical Evidence to Theory
• Parenthood and Gendered Mental Health: The Role of Paid Work and Housework Time
• The Insufficiency of Statistics for Detecting Racial Discrimination by Police
• The Psychological Science Accelerator: Advancing Psychology Through a Distributed Collaborative Network
• A caution on sibling comparisons in studying effects of the rearing environment
• How Do Internet-Related Characteristics Affect Whether Members of a German Mixed-Mode Panel Switch from the Mail to the Web Mode?
• Misery Needs Company: Contextualizing the Geographic and Temporal Link between Unemployment and Suicide
• Regression-with-residuals Estimation of Marginal Effects: A Method of Adjusting for Treatment-induced Confounders that may also be Moderators
• A Generalized Back-Door Criterion for Linear Regression
• Cross-ethnic integration through participation in leisure organisations? Evidence from two-wave panel data
• Do grandparents really matter? The effect of regular grandparental childcare on the second-birth transition
• Addressing the joint occurrence of self-selection and simultaneity biases in the estimation of program effects based on cross-sectional observational surveys: case study of travel behavior effects in carsharing
• Thinking Clearly About Correlations and Causation: Graphical Causal Models for Observational Data
• Unemployment among single mothers and adolescent children's mental health
• Counterfactual Slopes and Their Applications in Social Stratification
• Social origins and educational attainment: The unique contributions of parental education, class, and financial resources over time
• The intergenerational effect of educational expansion: evidence from a natural experiment in Spain
• To Evaluate the Age–Happiness Relationship, Look Beyond Statistical Significance
• The Future Strikes Back: Using Future Treatments to Detect and Reduce Hidden Bias
• The code of cohesion: adolescent network centrality, offending, and the downside of school cohesion
• Causal Inferences from Digital Behavioral Data
• Causal Machine Learning: A Deductive–Inductive Framework for Sociological Research
• The very early antecedents of NEET: understanding the role of birthweight
• A college on every cape: gender equality, gender segregation and local college openings
• Time-Varying Confounding Bias in Observational Geoscience with Application to Induced Seismicity
• Aggressive policing and undermined legitimacy: assessing the impact of police stops at gunpoint on perceptions of police in São Paulo, Brazil
• Visual Commonsense R-CNN
• Treatment Effect Learning Under Sequential Randomization
• In sole or joint names? The role of employment and marriage biographies for married women’s asset ownership in later life
• Causal inference on human behaviour
• Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
• Ethnic and social class discrimination in education: Experimental evidence from Germany
• Recent Developments in Causal Inference and Machine Learning
• Characterization and Learning of Causal Graphs with Latent Confounders and Post-treatment Selection from Interventional Data
• How Conditioning on Posttreatment Variables Can Ruin Your Experiment and What to Do about It
• We, Not Just I, Must Trust Doctors: Network Medical Trust and Health Care Utilization
• Timing of citizenship acquisition and immigrants’ children educational outcomes: a family fixed-effects approach
• An Assessment and Extension of the Mechanism-Based Approach to the Identification of Age-Period-Cohort Models
• Misery Needs Company: Contextualizing the Geographic and Temporal Link Between Unemployment and Suicide
• Why Did the Gender Gap in Adolescent Life Satisfaction Grow? Evaluating Methodological and Demographic Explanations
• Common effect models: Positive or negative manifold?
• You and whose economy? Group‐based retrospection in economic voting
• When Perception Trumps Reality
• Early-Life Exposures and Social Stratification
• Spillover effects of gestational age on sibling’s literacy
• Identifying the role of high school in educational inequality: A causal mediation approach
• Do Moral Agency for Aggressive Behavior and Low Peer Substance Use Protect Against Non-Substance Use Delinquency? A Cross-Lagged Analysis of Youth at High Risk for Delinquency
• On the role of surrogates in the efficient estimation of treatment effects with limited outcome data

Discussions

• More for a sociology audience, but is clear about the issues: doi.org/10.1146/annu... [bsky, 10 points, 1 comments]
• There are some great papers on collider bias. Mine provides examples of collider bias in economic history/historical demography, but this paper by Elwert and Winship is more general: www.annualreviews [bsky, 3 points, 1 comments]

Related

• Simple graphical rules for assessing selection bias in general-population and selected-sample treatment effects
• Toward a Clearer Definition of Selection Bias When Estimating Causal Effects
• Foundations of Structural Causal Models with Latent Selection
• On the Graphical Rules for Recovering the Average Treatment Effect Under Selection Bias
• On a Class of Bias-Amplifying Variables that Endanger Effect Estimates