101. Costs and benefits of natural experiments in political science
- Author:
- Elena Labzina
- Publication Date:
- 12-2011
- Content Type:
- Journal Article
- Journal:
- Central European University Political Science Journal
- Institution:
- Central European University
- Abstract:
- Natural experiment is a research design that is widely employed in modern Political Science. Yet, in the existing literature the features of the concept are ambiguous. The major aim of this article is to refine the related theory and argue natural experiments provide valid estimates in terms of causal inference. First, the author summarizes briefly and develops the related theory: the 'as-if' treatment randomization assumption is redefined with the introduction of the expected exchangeability treatment assumption, which enables their classification more as natural experiments. Second, based on the renewed theory, the author proposes an algorithm for the assessment of assumed natural experiments. Third, the algorithm is applied to five illustrative cases of recent natural experiments from Political Science. As a result, it is found that only two of them may be considered capable of providing valid inference. The major empirical finding is these two valid natural experiments are 'hidden' experiments, e.g. the individual(s) who were unaware that they assigned treatment had performed the randomization. This leads to the conclusion that the mysterious “nature” in nature experiments is human beings.