Across the social sciences, there is increasing interest in conducting randomized field experiments because of both their strong internal validity (from randomization) and external validity (by conducting the research in the real-world).

Field experiments have the added benefit of simplicity. At the end of the day, a well-executed randomized experiment ensures that experimental groups have the same potential outcomes, meaning that in most cases, all the analysis requires is a straightforward difference-in-means.

But before researchers can benefit from this simplicity, they are faced with a slew of decisions around the design and implementation of their projects. Examples of questions that researchers should consider include:

  • Am I getting the most statistical precision and power out of my existing design?

  • Does my design allow me to answer my research question or might there be confounding explanations?

  • What are the logistics of working with a partner organization? How do I work within their constraints while ensuring fidelity to the research design?

  • How can I set up a research procedure that is transparent and reproducible?

Before researchers spend the time and money on a field experiment, come meet with a D-Lab consultant! We are happy to work through your design and implementation plans.  

Author: 

Josh Kalla

I am a graduate student in the political science department. Much of my research involves partnering with political groups to conduct randomized experiments studying what influences the behavior and attitudes of voters and political decisionmakers. As a D-Lab consultant, I advise researchers on the design, implementation, and analysis of randomized field experiments.