Natural experiment
A natural experiment describes an empirical investigation method in which the test subjects are divided into an experimental group and a control group based on natural events that cannot be controlled by the researcher . The researcher only registers as an observer what happens without his intervention. Natural experiments are quasi-experiments , as they are controlled (there is a control group), but not randomized in the strict sense (the group is not assigned by the researcher at random) and can therefore be differentiated from randomized controlled studies . Ideally, however, the natural allocation of subjects is similar to randomization. Furthermore, natural experiments are field studies , since they take place in the field and not in the laboratory under standardized conditions, and observational studies , since the independent variable is not actively manipulated by the researcher.
Natural experiments are particularly useful when there is a clearly defined exposure for a certain subpopulation that does not affect another subpopulation, for example the introduction of a new school system in a region ( experimental group ), but not in the socio-demographically very much similar neighboring region ( control group ). They are used when the realization of a randomized controlled study is not possible for logistical or ethical reasons.
The derivation of causal conclusions from natural experiments is possible because there is an experimental group and a control group, but is subject to restrictions due to the possibly imperfect randomization.
history
One of the most famous early natural experiments took place in 1854 at the time of a cholera epidemic in London district of Soho . Doctor John Snow used city maps of London, which showed new infections and deaths, to identify a public water pump as the source of the cholera outbreak. Snow was able to show a strong association between the use of the water from the public well and the number of new infections and deaths. His further research revealed that the water company, which was supplying water to the areas with high rates of reinfection, was getting its water from a lower part of the Thames that was contaminated by sewage. Areas in which the new infection rates were lower, on the other hand, obtained water from another water company, which took water from the Thames further upstream that had not come into contact with sewage. Since the supply of the city districts by the different water companies was largely random, Snow described his study as an "experiment ... on the highest scale". Since Snow was of course unable to control which people drank which water, this type of investigation is viewed as a natural experiment.
Examples from recent history
Health consequences of ionizing radiation
Nuclear weapons tests release ionizing radiation that can damage the cells of living things. Before 1963, i.e. before the ban on nuclear weapons tests , the effects of these rays on human cells could be investigated. The effects of ionizing radiation on people born before 1963 can still be studied today. After the Chernobyl reactor accident , changes in the frequency of occurrence of various diseases, such as cataracts , were recorded in the affected areas and compared with the frequencies immediately before the disaster.
Economic decisions in quiz programs
Quiz programs can be seen as natural experiments to investigate economic behavior, for example decisions in risky situations or cooperative behavior. Even if the situation is an artificial one, quiz shows can be said to be a natural experiment, since the researcher neither brought about the situation nor intervened in it.
TV shows and suicide rates
The broadcast of the television program Death of a Schoolboy by ZDF in 1980 and 81 was used by scientists to check whether the representation of a rail suicide on television influences the actual rate of rail suicide (see Werther Effect ). As a comparison, the suicide rates in the phases before and after the broadcast of the program were used. There was a clear increase in rail suicides during and immediately after the broadcast compared to the control period.
Individual evidence
- ↑ J. DiNardo: Natural experiments and quasi-natural experiments . In: Steven N. Durlauf, Lawrence E. Blume (eds.): The New Palgrave Dictionary of Economics , Second ed .. Edition, Palgrave Macmillan, 2008, doi : 10.1057 / 9780230226203.1162 .
- ^ Snow, J. (1855). On the Mode of Communication of Cholera (2nd ed.). London: Churchill. Excerpted in MacMahon, B. & Pugh, TF (1970). Epidemiology . Boston: Little Brown.
- ↑ Freedman, DA Statistical Models: Theory and Practice (Cambridge University Press) [1] , Chapter 1.3, pp. 6-9.
- ^ Post, Van den Assem, Baltussen, Thaler: Deal or No Deal? Decision Making under Risk in a Large-Payoff Game Show . In: American Economic Review . 98, No. 1, 2008, pp. 38-71. doi : 10.1257 / aer.98.1.38 .
- ^ Van den Assem, van Dolder, Thaler: Split or Steal? Cooperative Behavior When the Stakes Are Large . In: Management Science . 58, No. 1, 2012. doi : 10.1287 / mnsc.1110.1413 .
- ↑ Death according to the pattern. In: spiegel.de . Spiegel , November 3, 1986, accessed June 2, 2016.