Natural experiments

While randomized controlled trials are increasingly prominent in the assessment of public policies, they cannot always be implemented, in particular when the experimentation protocol is too expensive, unacceptable from an ethical standpoint, or the reforms whose effects are being studied are already in place. In such cases, evaluation may include experiments that economists call “natural experiments”.

The use of exogenous “shocks”

A “natural experiment” describes a situation in which an exogenous “shock” affects the economic environment of the agents involved and permits the identification of causal relationships. Raising the compulsory school leaving age, gaps in the perceived amount of social support depending on the level of income, or the geographical targeting of certain policies are examples of natural experiments.

In the field of public policy, this approach can be applied when the beneficiaries of a scheme are selected on the basis of criteria that are difficult to manipulate, such as age, level of education or region of residence. The impact of the scheme can then be estimated by using the exogenous attribution criteria to identify comparable groups of individuals who are beneficiaries and non-beneficiaries.

The techniques used

Different econometric techniques have been developed for applying the methodological framework of natural experiments to the evaluation of public policies: instrumental variables, difference-within-differences, matching techniques and regression discontinuity design. The robustness of the techniques depends greatly on the quality of the data used for the evaluation, in particular the possibility of monitoring the population of beneficiaries and non-beneficiaries before and after the implementation of a reform.

Ce message est également disponible en : French