I held this workshop at the American University of Beirut (AUB) as a primer on causal inference for applied researchers. It’s designed to walk you through the core logic of identifying treatment effects when you can't rely on simple correlations.

The deck is a good starting point if you are just getting into evaluation and causal inference or if you are looking for a refresher on the standard toolbox of identification strategies. If you go through the slides and have questions about the DAGs or the specific assumptions behind the quasi-experimental designs, feel free to reach out.

What we covered:

• Why we can't observe the same unit in both the treated and untreated state at the same time, and how the counterfactual framework (Potential Outcomes) attempts to solve this.

• Selection Bias and Endogeneity: Identifying the specific threats to internal validity that usually crop up in observational data.

• A walkthrough of the big four quasi-experimental designs, including Randomized Controlled Trials (RCTs), Difference-in-Differences (DiD), Regression Discontinuity (RDD), and Instrumental Variables (IV).

Resources:

• Workshop Materials: [Download the Presentation (PPTX)]

I wrote this white paper to serve as a practical walkthrough for the power and sample size math we actually use in the field. It moves past the basic stuff and gets into the specific design complexities that usually trip up an evaluation design.

Feel free to use the derivations or the Stata commands for your own work, and definitely let me know if you have questions or catch any mistakes! 

What’s in the paper:

• Basic RCTs: The standard two-sample formulas for both continuous and binary outcomes.

• Multisite Designs: Specifically how to handle site-level heterogeneity and that inevitable allocation imbalance (S) that happens when sites don't split 50/50.

• Clustered Designs: Closed-form solutions for two-level and three-level nested structures.

I've mapped all the derivations to Stata implementation so you can run them directly.

Resources:

• Paper: [Download the White Paper (PDF)]

• Stata: You can pull my custom power commands directly from SSC:

  • ssc install cluster2

  • ssc install cluster3

I put this framework together because I wanted a one-stop shop for costing and economic evaluation that actually works across different sectors like education, health, and workforce development. The idea was to stop treating these as separate exercises and get back to core welfare-economic principles: identifying the causal impact and comparing it to the real resources used to get there.

I’ve also included a simplified Excel-based modeling tool for general use cases to handle the heavy lifting like discounting and sensitivity analysis. Feel free to grab both. If you find any bugs in the Excel logic or have thoughts on the intensive/extensive margin mapping, let me know.

What’s different in this approach:

• Most people are used to looking at fiscal budgets, but this framework includes the opportunity cost of participant time and social externalities to capture the true resource footprint.

• Rather than treating costing, cost-effectiveness (CEA), and benefit-cost analysis (BCA) as separate silos, this mapping derives all three from a single underlying causal model.

• It explicitly distinguishes between gains on the intensive margin (productivity or efficiency) and the extensive margin (increased participation or survival).

• I’ve integrated Monte Carlo simulation to show how results behave when your parameters vary within plausible ranges.

Note on Scope: This toolkit is designed specifically for intervention and policy-type treatments. It is distinct from typical health CEA that follows clinical effectiveness and the patient journey. I am currently working on a separate guide for health-specific CEA, so look out for that soon!

In-Depth Resources: If you're looking for a deeper theoretical lesson, you should check out the foundational work that informed this framework:

• General CBA/CEA: Boardman et al. (2018) and Jenkins et al. (2013) for principles.

• Education and Ingredients: Levin and McEwan (2001) and Dhaliwal et al. (2013) for the ingredients method and comparative logic.

Resources:

• White Paper: [Download the White Paper (PDF)]

• Analytical Tool: [Download the Excel Toolkit (XLSX)]