26.2 Robustness Checks
Robustness checks are essential to demonstrate that findings are not driven by model specification choices.
Recommended robustness checks (Goldfarb, Tucker, and Wang 2022):
- Alternative Control Sets
- Show results with and without controls.
- Examine how the estimate of interest changes.
- Use Rosenbaum bounds for formal sensitivity analysis (Altonji, Elder, and Taber 2005).
- In marketing applications, see (Manchanda, Packard, and Pattabhiramaiah 2015) and (Shin, Sudhir, and Yoon 2012).
- Different Functional Forms
- Check whether the results hold under different model specifications (e.g., linear vs. non-linear models).
- Varying Time Windows
- In longitudinal settings, test different time frames to ensure robustness.
- Alternative Dependent Variables
- Use related outcomes or different measures of the dependent variable.
- Varying Control Group Size
- Compare results using matched vs. unmatched samples to assess sensitivity to sample selection.
- Placebo Tests
- Conduct placebo tests to ensure the effect is not spurious.
- The appropriate placebo test depends on the specific quasi-experimental method used (examples provided in later sections).
References
Altonji, Joseph G, Todd E Elder, and Christopher R Taber. 2005. “Selection on Observed and Unobserved Variables: Assessing the Effectiveness of Catholic Schools.” Journal of Political Economy 113 (1): 151–84.
Goldfarb, Avi, Catherine Tucker, and Yanwen Wang. 2022. “Conducting Research in Marketing with Quasi-Experiments.” Journal of Marketing 86 (3): 1–20.
Shin, Jiwoong, K Sudhir, and Dae-Hee Yoon. 2012. “When to ‘Fire’ Customers: Customer Cost-Based Pricing.” Management Science 58 (5): 932–47.