2-Day In-Person Workshop

Partial Identification in Practice: Bounding Causal Effects

Dr. Hannah Wich

Stephen F. Austin State University

May 7, 2025, Wednesday. 2:00 – 5:00PM

AND

May 8, 2025, Thursday. 9:00AM – 12:00PM & 2:00 – 5:00PM

RSVP  HERE

Join us IN PERSON on the Texas A&M University Campus, College Station, TX in Teague 326 

 Lunch will be provided for in-person attendees.

Workshop Abstract:

Causal estimates help quantify the impact of one variable on another, making them invaluable in social science research for evaluating interventions and exposures. However, identifying causal effects is inherently challenging since we can only observe one possible outcome for any given intervention—leaving the counterfactual unobserved. While randomized controlled trials (RCTs) are the gold standard for establishing causality, they are often impractical, expensive, or unethical, making them rare in many applied settings. As a result, researchers frequently rely on observational data, which requires strong assumptions to infer causality. These assumptions—such as no unobserved confounding or no measurement error—are often unrealistic, limiting the credibility of traditional causal inference methods. This has fueled interest in alternative approaches that relax these assumptions while still providing meaningful insights from nonexperimental data.

Partial identification offers a powerful alternative by shifting the focus from point estimates—often reliant on strong, and sometimes unrealistic assumptions—to credible bounds on causal effects. Rather than point identifying a parameter of interest, partial identification starts with minimal, highly plausible assumptions and derives set-valued estimates. As assumptions are gradually strengthened, these bounds become more informative, by shrinking the size of the set estimates.

This workshop introduces partial identification techniques with an emphasis on program evaluation, particularly in settings affected by selection bias and nonrandom measurement error. Through applied illustrations, we will demonstrate how to use partial identification to derive informative bounds on key parameters of interest. Participants will gain practical insights into how these methods can support robust conclusions under credible and relatively weak assumptions.

Key topics covered in this workshop

• Introduction to Partial Identification – Understanding the concept of set-valued estimates, how they differ from point estimates, and their role in causal inference when strong assumptions are not feasible.
• Addressing Selection Bias and Measurement Error – Exploring how partial identification techniques help mitigate selection bias and nonrandom measurement error in program evaluation.
• Applied Examples – Demonstrating partial identification methods through real-world case studies that illustrate how to derive informative bounds on parameters of interest in the presence of imperfect data and limited identifying assumptions.

BIO

Dr. Hannah Wich is. an Assistant Professor in the Department of Economics and Finance at Stephen F. Austin State University. Originally from Germany, she earned her bachelor’s degree from the University of South Dakota and her Ph.D. in Economics from Iowa State University in 2023

Dr.Wich’s research focuses on consumer food choice decisions and strategies to alleviate food insecurity, with interests spanning public economics, health economics, and applied microeconomics. She utilizes retailer data to examine food purchasing behavior, particularly the role of SNAP participation and bulk purchasing decisions. Additionally, she explores the intersection of mental health, food insecurity, and SNAP, contributing to a deeper understanding of how food assistance programs impact well-being. Her methodological interests include using partial identification techniques to obtain causal estimates, allowing for more robust policy-relevant insights in settings where traditional causal inference methods face limitations.