Course Introduction
Propensity Score Matching (PSM) is a widely used statistical technique for estimating programme effects when participants are not randomly assigned to treatment and comparison groups. In development programmes, public policy initiatives, social protection interventions, education projects, health services, and livelihood programmes, evaluators often need to compare outcomes between beneficiaries and non-beneficiaries using observational data. This five-day training course provides a structured and practical foundation for understanding how propensity score matching can improve the comparability of programme groups and strengthen the credibility of impact assessment findings.
The course introduces the counterfactual framework, potential outcomes, selection bias, confounding, and the assumptions that underpin causal inference using observational data. Participants will explore why differences in outcomes between programme participants and non-participants cannot automatically be interpreted as programme effects. Through guided discussions and applied examples, they will learn how observed characteristics influence programme participation and how statistical matching can help construct a more comparable counterfactual group when randomization is unavailable or impractical.
Participants will develop practical skills in specifying treatment and comparison groups, selecting appropriate pre-intervention covariates, estimating propensity scores, and assessing the quality of matched samples. The course examines commonly used estimation approaches, including logistic regression and other suitable classification models, while emphasizing that a sophisticated prediction model does not necessarily produce a credible causal estimate. Particular attention is given to the distinction between predicting participation accurately and achieving covariate balance between treated and comparison observations.
A major component of the programme focuses on matching algorithms and implementation decisions that affect evaluation results. Participants will examine nearest-neighbour matching, caliper restrictions, radius matching, kernel matching, and alternative approaches to constructing comparable samples. They will learn how to assess common support, identify unmatched observations, evaluate standardized mean differences, and interpret balance diagnostics before and after matching. Practical exercises will demonstrate how inappropriate matching specifications, poor overlap, and unsuitable covariate choices can undermine the validity of estimated programme effects.
The training also addresses the interpretation, reporting, and limitations of propensity score matching in real-world programme evaluations. Participants will examine average treatment effects on the treated, the importance of defining the target estimand, and the implications of unobserved confounding, missing data, measurement error, and model misspecification. The course introduces sensitivity analysis and robustness checks to help determine whether findings are stable across alternative specifications. It also explains why propensity score matching cannot eliminate bias arising from unobserved factors simply by balancing the characteristics included in the analysis.
By the end of the five-day programme, participants will be equipped to design, implement, review, and communicate propensity score matching studies with greater methodological confidence. They will practice developing a transparent analytical workflow, documenting methodological decisions, presenting balance diagnostics, and translating statistical results into evidence that programme managers, policymakers, donors, and evaluation commissioners can use. The course is suitable for professionals seeking to improve the rigour, reproducibility, and practical usefulness of programme comparison studies across diverse development and public service contexts.
Duration
5 days
Who Should Attend
-
Monitoring and evaluation specialists responsible for assessing programme effectiveness and results.
-
Impact evaluation researchers conducting nonexperimental programme comparison studies.
-
Economists undertaking applied microeconometric analysis and policy evaluation.
-
Development practitioners evaluating donor-funded projects and social interventions.
-
Government planning and policy officers assessing public programme performance.
-
Research analysts working with beneficiary and non-beneficiary datasets.
-
Statisticians and quantitative research professionals seeking applied causal inference skills.
-
Data analysts working with household surveys and administrative programme records.
-
Social protection specialists evaluating cash transfers and targeted assistance programmes.
-
Public health researchers comparing service users and non-users or intervention groups.
-
Education researchers assessing the outcomes of school-based programmes and learning interventions.
-
NGO and civil society professionals responsible for evidence generation and programme accountability.
-
Donor agency staff reviewing evaluation methodologies and causal impact claims.
-
Academic researchers and postgraduate students conducting applied evaluation research.
-
Consultants and evaluation team leaders designing credible comparison strategies for programme assessments.
Course Objectives
-
Explain the counterfactual framework, potential outcomes, selection bias, and the role of propensity score matching in observational programme evaluation.
-
Identify appropriate treatment and comparison groups and define the target population, treatment indicator, outcomes, and estimand for a matching study.
-
Select theoretically justified pre-treatment covariates that influence programme participation and outcomes while avoiding inappropriate post-treatment variables.
-
Estimate propensity scores using suitable statistical models and interpret how model specification affects the distribution of scores across groups.
-
Apply nearest-neighbour, caliper, radius, and kernel matching techniques while considering their assumptions, practical advantages, and limitations.
-
Assess common support, overlap, and covariate balance using graphical and numerical diagnostics before accepting a matched sample for analysis.
-
Estimate and interpret treatment effects for matched observations, particularly the average treatment effect on the treated, with appropriate statistical caution.
-
Identify methodological threats including unobserved confounding, missing data, poor overlap, model misspecification, and sensitivity to matching choices.
-
Conduct robustness checks and sensitivity analyses to evaluate the stability and credibility of programme comparison results.
-
Prepare transparent analytical reports that document matching procedures, diagnostic evidence, uncertainty, limitations, and implications for programme decisions.
Comprehensive Course Outline
Module 1: Foundations of Programme Comparison and Causal Inference
-
Understanding the difference between descriptive comparisons, associations, and causal estimates in programme evaluation.
-
Exploring potential outcomes, counterfactual reasoning, treatment assignment, and the fundamental problem of causal inference.
-
Identifying selection bias, confounding, and baseline differences between programme participants and comparison groups.
-
Comparing randomized evaluations, quasi-experimental designs, regression adjustment, and propensity score matching as analytical strategies.
Module 2: Study Design, Data Requirements, and Estimand Definition
-
Defining the research question, intervention exposure, eligible population, comparison group, and outcomes of interest.
-
Distinguishing the average treatment effect, average treatment effect on the treated, and other population-specific estimands.
-
Reviewing beneficiary registers, household surveys, administrative records, and linked datasets for matching suitability.
-
Establishing data quality procedures for missing observations, inconsistent records, measurement errors, and treatment classification.
Module 3: Covariate Selection and Propensity Score Estimation
-
Selecting pre-intervention covariates based on programme participation mechanisms, subject knowledge, and causal reasoning.
-
Specifying logistic regression models for estimating participation probabilities and examining alternative model formulations.
-
Understanding nonlinear relationships, interaction terms, categorical predictors, and the implications of model misspecification.
-
Avoiding post-treatment variables, inappropriate adjustment, and covariate selection based solely on statistical significance.
Module 4: Matching Algorithms and Implementation Strategies
-
Applying nearest-neighbour matching with replacement and without replacement while evaluating practical trade-offs.
-
Using caliper and radius matching to limit matches to observations with sufficiently similar propensity scores.
-
Exploring kernel matching and alternative weighting approaches for constructing comparison samples from eligible observations.
-
Comparing matching specifications using sample retention, match quality, balance diagnostics, and suitability for the research question.
Module 5: Common Support, Overlap, and Balance Diagnostics
-
Visualizing propensity score distributions for treatment and comparison groups before and after matching.
-
Identifying common support restrictions, extreme propensity scores, and observations without credible comparison counterparts.
-
Calculating standardized mean differences and examining variance ratios and other covariate balance indicators.
-
Interpreting balance tables and diagnostic plots to determine whether the matched sample is suitable for effect estimation.
Module 6: Estimating and Interpreting Programme Effects
-
Estimating average treatment effects on the treated using matched observations and clearly defined analytical populations.
-
Calculating outcome differences and interpreting effect magnitudes in relation to programme objectives and baseline conditions.
-
Understanding uncertainty estimation, standard errors, confidence intervals, and the implications of dependence introduced by matching.
-
Communicating statistical significance, practical importance, and causal interpretation without overstating the evidence.
Module 7: Practical Implementation Using Statistical Software
-
Preparing analytical datasets, coding treatment indicators, defining covariates, and documenting reproducible data-processing steps.
-
Implementing propensity score estimation and matching workflows using suitable tools in R, Stata, or Python.
-
Generating balance tables, propensity score plots, matched-sample summaries, and outcome comparison outputs.
-
Organizing scripts, recording software settings, and maintaining an auditable trail of analytical decisions and results.
Module 8: Bias, Limitations, and Threats to Validity
-
Examining unobserved confounding and explaining why matching on observed characteristics cannot guarantee unbiased causal estimates.
-
Assessing the effects of missing data, measurement error, treatment misclassification, and inappropriate covariate definitions.
-
Understanding how poor overlap, small comparison groups, and discarded observations can limit generalizability.
-
Evaluating the circumstances in which alternative methods may be more appropriate than propensity score matching alone.
Module 9: Robustness Checks and Sensitivity Analysis
-
Comparing alternative matching algorithms, caliper widths, replacement settings, and covariate specifications.
-
Testing the stability of estimated effects across reasonable analytical choices and alternative matched-sample definitions.
-
Exploring sensitivity analysis for potential hidden bias arising from unmeasured confounding.
-
Establishing transparent criteria for documenting robustness findings and distinguishing stable results from specification-sensitive estimates.
Module 10: Reporting, Interpretation, and Applied Evaluation Project
-
Developing a complete propensity score matching analysis plan aligned with the programme evaluation question and available evidence.
-
Preparing methodological documentation covering data sources, covariates, estimation procedures, matching rules, and diagnostic decisions.
-
Presenting treatment effect estimates alongside balance assessments, uncertainty measures, sensitivity findings, and study limitations.
-
Completing an applied programme comparison exercise and translating findings into responsible recommendations for policy and programme improvement.
Training Approach
This course will be delivered by our skilled trainers who have vast knowledge and experience as expert professionals in the fields. The course is taught in English and through a mix of theory, practical activities, group discussion and case studies. Course manuals and additional training materials will be provided to the participants upon completion of the training.
Tailor-Made Course
This course can also be tailor-made to meet organization requirement. For further inquiries, please contact us on: Email: training@upskilldevelopment.com Tel: +254 721 331 808
Training Venue
The training will be held at our Upskill Training Centre. We also offer training for a group (at a discount of 10% to 50%) at requested location all over the world. The Onsite course fee covers the course tuition, training materials, two break refreshments, buffet lunch, airport transfers, Upskill gift package, and guided tour.
Visa application, travel expenses, dinners, accommodation, insurance, and other personal expenses are catered by the participant
Certification
Participants will be issued with Upskill certificate upon completion of this course.
Airport Pickup and Accommodation
Airport pickup and accommodation is arranged upon request. For booking contact our Training Coordinator through Email: training@upskilldevelopment.com, +254 721 331 808
Terms of Payment:
Unless otherwise agreed between the two parties’ payment of the course fee should be done 3 working days before commencement of the training so as to enable us to prepare better.








