eduzhai > Applied Sciences > Computer Science >

Natural Experiments

  • KanKan
  • (0) Download
  • 20210424
  • Save

... pages left unread,continue reading

Document pages: 40 pages

Abstract: The term natural experiment is used inconsistently. In one interpretation, itrefers to an experiment where a treatment is randomly assigned by someone otherthan the researcher. In another interpretation, it refers to a study in whichthere is no controlled random assignment, but treatment is assigned by someexternal factor in a way that loosely resembles a randomized experiment---oftendescribed as an "as if random " assignment. In yet another interpretation, itrefers to any non-randomized study that compares a treatment to a controlgroup, without any specific requirements on how the treatment is assigned. Iintroduce an alternative definition that seeks to clarify the integral featuresof natural experiments and at the same time distinguish them from randomizedcontrolled experiments. I define a natural experiment as a research study wherethe treatment assignment mechanism (i) is neither designed nor implemented bythe researcher, (ii) is unknown to the researcher, and (iii) is probabilisticby virtue of depending on an external factor. The main message of thisdefinition is that the difference between a randomized controlled experimentand a natural experiment is not a matter of degree, but of essence, and thusconceptualizing a natural experiment as a research design akin to a randomizedexperiment is neither rigorous nor a useful guide to empirical analysis. Usingmy alternative definition, I discuss how a natural experiment differs from atraditional observational study, and offer practical recommendations forresearchers who wish to use natural experiments to study causal effects.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...