During Lent term, I have been one of the GTAs (i.e., Graduate Teaching Assistants) for MY457, or Causal Inference in Experimental and Observational Studies. As wonderfully inquisitive LSE students are, they have been asking me for further readings on the methods we have been discussing throughout the course. I thought that it might make sense to list some recurring suggestions I have made to them, if for nothing else, for a quick reference to anyone taking the course (or otherwise interested in the topics covered by it).
First and foremost, Paul Rosenbaum has just published a very well-written and surprisingly accessible book called Observation and Experiment: An Introduction to Causal Inference. Unlike his previous book, Design of Observational Studies, which can be sometimes heavy on equations and can feel a bit arcane for uninitiated readers, this new book is a real primer, which aims to get the readers invested in the new methods and new way of thinking without bothering them with some of the practical complexities. I have to confess, that so far I have only made it halfway through the book, but would be happy to recommend it to anyone just starting with causal inference.
Another very good introductory reading, on the more complex side, that I have been recommending is Imai et al. (2008). This article provides a very lucid and comprehensive picture on the advantages and disadvantages of pursuing designs permitting causal inference, by comparing the various causal inference methods to survey methods, and highlighting how these different approaches “trade errors” with each other.
Moving towards more concrete recommendations, the two articles I kept suggesting were Davies et al. (2017) (for matching) and Bind and Ruin (2017) (for instrumental variables). The reason these articles are so excellent is that they emphasise the considerations one has to make by leaving the observational world for making causal inference. At the same time, it is not always clear that the trade-offs are necessarily worth it, and these papers provide apposite examples and tests to help assess each case.
Finally, let me praise Skovron and Titiunik’s (2015) unpublished manuscript for providing a concise but well-rounded picture on RDDs, with great examples that highlight the utility of this approach.
Bind, M-A.C. & Ruin, D.B. (2017). Bridging Observational Studies and Randomized Experiments by Embedding the Former in the Latter. Statistical Methods in Medical Research https://doi.org/10.1177/0962280217740609
Davies, N. M., Thomas, K. H., Taylor, A. E., Taylor, G. M. J., Martin, R. M., Munafò, M. R., & Windmeijer, F. (2017). How to compare instrumental variable and conventional regression analyses using negative controls and bias plots. International Journal of Epidemiology, 46(6), 2067–2077.
Imai, K., King, G., & Stuart, E. A. (2008). Misunderstandings between experimentalists and observationalists about causal inference. Journal of the Royal Statistical Society, 171(2), 481–502.
Rosenbaum, P.R. (2009). Design of Observational Studies. Springer.
Rosenbaum, P.R. (2017). Observation and Experiment: An Introduction to Causal Inference. Harvard University Press.
Skovron, C. & Titiunik, R. (2015). A Practical Guide to Regression Discontinuity Designs in Political Science. Manuscript.