PhD Courses in Denmark

Intuition and Interpretation in Causal Inference

Copenhagen Graduate School of Social Sciences

Department of Political Science

Dates and time: 22-23 April and 21-22 May 2026 from 9:00 to 16:00

Many graduate courses in causal inference equip students with powerful tools for drawing causal conclusions from observational data. These courses often emphasize either the formal, mathematical conditions under which causal identification is possible, or the practical implementation of methods in statistical software. While both approaches are essential, they can leave students with limited intuition for why particular designs work, what violations of identifying assumptions look like in practice, and how to meaningfully interpret the resulting estimates.

This PhD course focuses on building intuition for common causal inference designs. Rather than centering on proofs or step-by-step estimation procedures, the course emphasizes substantive understanding of identification strategies, realistic threats to identification, and careful interpretation of causal estimates. Particular attention is paid to the meaning of causal estimands – what exactly is being estimated, whom it applies to, and under what assumptions – and how these estimands relate to substantive research questions. And how does the estimand change under different specifications of the same general method?

The course assumes basic familiarity with causal inference concepts and designs. We will briefly revisit randomized controlled trials (RCTs) as a conceptual benchmark, before covering instrumental variables, regression discontinuity designs, and difference-in-differences. These designs are not covered as abstract techniques but are discussed and applied as research strategies embedded in substantive empirical contexts.

Through this approach, the course provides practical intuition for several important recent developments in applied econometrics. Topics include sensitivity analysis for unobserved confounding, staggered difference-in-differences designs under heterogeneous treatment effects, violation of parallel trends, profiling and interpretation of compliers in instrumental-variables settings, and power analysis for observational studies.

The course runs over four days in total. The first two days take place in April, and the final two days in May. Participants will have the option to write an essay on causal identification in their own dissertation project for additional ECTS credit.

Format
The course runs April 23-24, 2026, then there will be a one-month break, allowing students to work on their essays. Class meets again on May 21-22, 2026. The teaching format is highly interactive. Each causal inference design is introduced through an interactive lecture organized around a central case paper which is discussed throughout the session to illustrate identification logic, assumptions, and interpretation. This is followed by group discussion of an applied paper, where participants critically assess threats to identification, interpret the reported estimates, and discuss ways the analysis could be strengthened.

Optional essay
Participants will have the option to write a five-page essay about the topics discussed throughout the course, and how they relate to their own work. Completing this essay will result in additional ECTS points (3 points instead of 2). The course organizers will provide feedback and help think through the causal identification questions raised in all written essays. The deadline for submitting papers is Thursday 14 May 2026, at noon.

Course organisers and lecturers: 
Florian Hollenbach, Associate Professor, Copenhagen Business School
Benjamin Egerod, Assistant Professor, Copenhagen Business School

 

Preliminary program: Keywords and readings

Day 1:

  1. RCT (recap)
    1. Causality and identification
    2. Average treatment effects, heterogeneity and causal estimands
    3. Readings:
      Chapter 4 in Cunningham (2021)
      Case text: Blattman and Dercon (2018)
  2. Instrumental variables
    1. Identifying assumptions: Conditional independence, excludability, SUTVA.
    2. Compliers, always-takers, never-takers.
    3. Causal estimand: Local average treatment effect (LATE)
    4. Sensitivity analysis
    5. Readings:
      Chapter 7 in Cunningham (2021)
      Betz, Cook and Hollenbach (2019)
      Marbech and Hangartner (2020)
      Case text 1: Clingingsmith et al (2009)
      Case text 2: Baron and Gross (2025)
      Listen to this episode of the podcast series Probable Causation: Episode 64: Jason Baron on Foster Care Placement. Link: https://www.probablecausation.com/podcasts/episode-64-jason-baron. There is an extremely informative substantive discussion of compliance and how it matters for interpretation in the estimates (what we call the estimand).

Day 2:

  1. Regression Discontinuity
    1. Identification: Smoothness of potential outcomes vs. smoothness of compliers
    2. The LATE and external validity
    3. Effective sample size and power
    4. Robustness to specification choice
    5. Readings:
      Chapter 6 in Cunningham (2021)
      Marshall (2024)
      Case text 1: Eggers and Hainmueller (2009)
      Case text 2: Dell and Querubin (2018)

Day 3:

  1. Difference-in-differences
    1. Canonical design (brief recap): identification and causal estimand
    2. Staggered uptake and heterogeneity
    3. More on identification: What does the parallel trends assumption mean?
    4. Sensitivity analysis for violation of parallel trends
    5. Readings:
      Chapter 9 in Cunningham (2021)
      Egerod and Hollenbach (2024)
      Rambachan and Roth (2023)
      Ghanem, Sant’Anna and Wüttrich (2022)
      Case text 1: Egerod and Fouirnaies (2023)
      Case text 2: Dinas et al (2019)

Day 4:

  1. Statistical power in observational studies
    1. Bias and variance
    2. Error types
    3. Readings:
      Egerod and Hollenbach (2024)

 

Language: English

ECTS: 2 points for attendance including preparation + 1 extra point for participation with essay (optional).

Max. numbers of participants: 12

Course fee: The PhD School at the Faculty of Social Sciences participates in Denmark’s national network for PhD courses. This course is free of charge for PhD students enrolled at one of the participating PhD schools (PhD students enrolled at a Danish University, except Copenhagen Business School). Other PhD students will be charged a course fee of DKK 1,200 per ECTS for participation in the course (PhD students enrolled at Copenhagen Business School or a University outside Denmark).

Registration: Please register via the link in the box no later than 24 March 2026.

Further information: For more information about the PhD course, please contact the PhD Administration (phd@hrsc.ku.dk).

 

Literature

Baron, E. Jason, and Max Gross. "Is there a foster care-to-prison pipeline? Evidence from quasi-randomly assigned investigators." Review of Economics and Statistics (2025): 1-46.

Blattman, Christopher, and Stefan Dercon. "The impacts of industrial and entrepreneurial work on income and health: Experimental evidence from Ethiopia." American Economic Journal: Applied Economics 10, no. 3 (2018): 1-38.

Betz, Timm, Scott J. Cook, and Florian M. Hollenbach. "Spatial interdependence and instrumental variable models." Political science research and methods 8, no. 4 (2020): 646-661.

Clingingsmith, David, Asim Ijaz Khwaja, and Michael Kremer. "Estimating the impact of the Hajj: religion and tolerance in Islam's global gathering." The Quarterly Journal of Economics 124, no. 3 (2009): 1133-1170.

Cunningham, Scott. Causal inference: The mixtape. Yale university press, 2021.

Dell, Melissa, and Pablo Querubin. "Nation building through foreign intervention: Evidence from discontinuities in military strategies." The Quarterly Journal of Economics 133, no. 2 (2018): 701-764.

Dinas, E., Matakos, K., Xefteris, D., & Hangartner, D. (2019). Waking up the golden dawn: does exposure to the refugee crisis increase support for extreme-right parties? Political Analysis, 27(2), 244-254.

Egerod, Benjamin & Alexander Fouirnaies (2024): How Does Electoral Accountability Affect Legislator Behavior? Evidence from Life-Tenured Legislators in Denmark. Unpublished Manuscript.

Egerod, Benjamin CK, and Florian M. Hollenbach. "How many is enough? Sample size in staggered difference-in-differences designs." OSF Preprint (2024).

Eggers, Andrew C., and Jens Hainmueller. "MPs for sale? Returns to office in postwar British politics." American Political Science Review 103, no. 4 (2009): 513-533.

Ghanem, Dalia, Pedro HC Sant'Anna, and Kaspar Wüthrich. "Selection and parallel trends." arXiv preprint arXiv:2203.09001 (2022).

Marbach, Moritz, and Dominik Hangartner. "Profiling compliers and noncompliers for instrumental-variable analysis." Political Analysis 28, no. 3 (2020): 435-444.

Marshall, John. "Can close election regression discontinuity designs identify effects of winning politician characteristics?." American Journal of Political Science 68, no. 2 (2024): 494-510.

Rambachan, Ashesh, and Jonathan Roth. "A more credible approach to parallel trends." Review of Economic Studies 90, no. 5 (2023): 2555-2591.

Back