Causal Inference in Public Management

Session Submission Type: Panel

Abstract

In recent years, scholars in public administration have intensified their efforts to strengthen causal inference and have increasingly applied research methods drawn from other social sciences. A surge in experimental approaches has been widely documented (e.g., James et al. 2017) and identified as closely linked to the rise of Behavioral Public Administration (Battaglio et al. 2019; Bhanot and Linos 2020). However, on the one hand, a recent review of experimental public administration research shows that online experiments, particularly basic or factorial designs, have been especially popular, often at the expense of other types of experiments, such as field experiments and those exploring causal mediation (Belle and Belardinelli 2025). On the other hand, natural and quasi-experiments, such as difference-in-differences, regression discontinuity, and instrumental variables, have been largely overlooked in the public administration literature, with a few exceptions.

The goal of this panel is to discuss sound causal inference in public administration while promoting the adoption of less common experimental and quasi-experimental designs. It offers an opportunity to delve into methodological issues and emphasize causal identification strategies, while also highlighting the substantive contributions of the studies to the field. The papers complement and inform one another by (i) carefully addressing the internal validity of causal inference in public management research, and (ii) advancing the use of underutilized designs in this area.

Gregg Van Ryzin and Dahlia Remler highlight the limited use of quasi- and natural experiments in public management research. Using a new framework that outlines key dimensions of such studies, the paper assesses the literature for gaps and opportunities to employ these methods in generating causal evidence for theory and practice. Marylis Fantoni applies a natural experiment to examine trends in protective order (PO) denials in Indiana and the causal effect of judge gender on this process. The study leverages the random assignment of PO requests in half of the state’s courts. Findings show that from 2010 to 2013, female judges in those courts were 12% more likely to deny a PO than male judges. Nicola Belle combines randomized and quasi-experimental data from public healthcare workers to test a framework unifying biases in public decision-making as outcomes of selective attention and memory. This work demonstrates how belief instability and multimodality stem from cognitive mechanisms, thereby advancing theory in behavioral public administration. Elizabeth Linos, Jessica Lasky-Fink, and Heidi Wallace, are working in collaboration with a large US city to co-design a randomized controlled trial aimed at testing the effectiveness of behaviorally-informed outreach to connect low-income tenants with government housing stabilization resources, with the goal of preventing and mitigating evictions. This project offers a valuable case study in designing and conducting collaborative field experiments with government that answer critical public management questions while also responding to key policy challenges.

Policy Area

• Methods and Tools of Analysis

Secondary Policy Area

• Public and Nonprofit Management and Finance

Chair

• Carolyn J. Heinrich, Vanderbilt University

Discussant

• Paolo Belardinelli, Indiana University

Organizer

• Paolo Belardinelli, Indiana University

Individual Presentations

• Selective Attention and the Architecture of Belief: Instability and Multimodality in Public Decision-Making - Presenting Author: Nicola Belle, Sant'Anna School of Advanced Studies

• When a Protective Order is Denied: Causal Evidence on the Role of Judge Gender - Presenting Author: Marylis C Fantoni, Brigham Young University

• Connecting low-income renters to housing stabilization services: A field experiment - Presenting Author: Elizabeth Linos, Harvard University; Non-Presenting Co-Author: Jessica Lasky-Fink, Harvard University; Non-Presenting Co-Author: Heidi Wallace, Harvard University

• Quasi and Natural Experiments in Public Management Research - Presenting Author: Gregg G Van Ryzin, Rutgers University - Newark; Non-Presenting Co-Author: Dahlia Remler, Baruch College (City University of New York)