Center for Causal Inference

Applying methodological and statistical rigor to help draw proper inferences

At the heart of policy analysis is cause and effect: How can we determine or quantify the actual impact of an action or policy?

Without proper methods, recklessly inferring cause and effect from observed relationships can inform unhelpful policies at best, and dangerous ones at worst.

That’s where the Center for Causal Inference comes in. We innovate analytic approaches to yield estimates of causal relationships based on nonexperimental or observational data. And we recognize that even experimental data may require causal analysis.

Casual inference is involved in nearly every RAND research project, so it’s important that we get it right.

Our goals are to provide frameworks and estimation methods that reflect causality in the inferences we make, and to define what assumptions are necessary for those estimation methods to increase the certainty that we are drawing accurate conclusions.

Our Center develops, applies, and disseminates a toolbox of robust methodologies to ensure causal inferences are sound, and we offer training in those methods. We also provide seed funds to researchers to develop and improve upon causal inference methods—and in turn improve our ability to effectively analyze policy.

Our Focus

We apply methodological and statistical rigor to help our colleagues draw proper inferences. We encourage and support randomized studies, but when those aren’t feasible, we work with our colleagues to develop and use causal inference methods such as propensity scores, synthetic control, difference-in-difference, regression discontinuity, and instrumental variables, among others. We use these methods for estimations, and to determine the effectiveness of programs targeted at populations at risk or in need.

Learn More

Methods In Action

  • Event

    Third Annual Center for Causal Inference Symposium

    The third annual RAND Center for Causal Inference (CCI) Symposium, a free, virtual event, included three sessions, each featuring four speakers, and three breakout sessions.

  • Journal Article

    No Link Found Between Pets and Kids' Health

    Contrary to popular belief, having a dog or cat in the home does not improve the mental or physical health of children. Researchers used a weighted propensity score regression approach and double robust regression analyses to examine the association between living with a dog or cat and health outcomes, while accounting for confounding factors.

  • Project

    Toolkit for Weighting and Analysis of Nonequivalent Groups (TWANG)

    The Toolkit for Weighting and Analysis of Nonequivalent Groups, or TWANG, contains a set of functions to support causal modeling of observational data through the estimation and evaluation of propensity score weights. TWANG has been released as a package in R, and as downloadable macros for SAS users. The team has also released a series of tutorials intended to guide analysts using the toolkit for the first time.

  • Report

    Community Justice Center in San Francisco Is Associated with Lower Rearrest Rates

    Using a differences-in-differences design, researchers examined one-year rearrest rates among those arrested for eligible offenses within the four police districts that include a part of the CJC catchment area. Their findings support the hypothesis that the CJC reduces criminal recidivism and are robust to a number of sensitivity analyses.

  • Commentary

    Research Rigor Is Undermined by Translating Into Years of Learning

    The research field increasingly recognizes that we need metrics that everyone can understand. Translating the result into years of learning has become a popular approach, but this metric has major flaws. There are better options.

  • Journal Article

    Causal Inference Using Mixture Models: A Word of Caution

    Mixture models are useful for monitoring the behavior of data and for offering comparisons to supplemental data, especially in the presence of unobserved heterogeneity, but one should be highly cautious when drawing causal inferences as to which population each component of the fitted mixture model represents.

View more research

Student Spotlight

  • Applying Causal Inference to Workforce and Higher Ed Research

    Although she wasn't exposed to causal inference methods until she arrived at Pardee RAND, Hannah Acheson-Field (cohort '18) says she is grateful for her coursework and her work with the Center for Causal Inference, as they have opened her eyes to research and career possibilities.

Connect With Us


Want more information? Interested in working with us? We'd love to hear from you.