“Does eating chocolate cause Nobel Prizes?” and other causal inferences

Studying human behavior in a lab poses myriad problems. Dr. Gordon Dahl developed a method that allows scientists to more accurately measure human behavior.

Dr. Gordon Dahl, a professor of economics at the University of California, San Diego, was the presenter at the 41st Annual Summer Institute of Applied Statistics at BYU. Dahl’s lecture series focused on natural experiments.

Dahl, an economist, uses statistics in his research on causality.

He began by discussing the maxim,“Correlation does not imply causation.”

He highlighted simple examples of this concept, such as “Does Internet Explorer use cause murder?” and “Does eating chocolate cause Nobel Prizes?” Just because two events occur at the same time does not mean one caused the other, even if one of these events occurred before the other.

One way scientists identify causation is through randomized experiments. In randomized experiments, the sample is split into two equal groups, and each group shares the same characteristics in order to get accurate results.

However, in the study of human behavior, it is often difficult or impossible to run experiments such as these. Because the experiment is run in an unnatural setting – the laboratory – and people who are in unnatural settings will not act or react normally, the results of the experiment can be inaccurate.

Dahl illustrated this difficulty using the example of trying to ascertain whether watching football can lead to family violence. It would be unethical and inaccurate to conduct a lab experiment to find the correlation and causality between football viewing and domestic violence. For one thing, inducing violence in the lab would be unethical. For another, domestic violence is unlikely to happen in front of a third party – such as observers in the lab.

On the other hand, researchers could use a natural experiment to study whether wins and losses in football games affect one’s emotions. Natural experiments are situations that provide naturally-occurring randomness. Football was chosen because there is a strong attachment to local NFL teams. When emotional responses are escalated – as may be the case with an unexpected loss in a football game – family violence is more likely to occur.

The natural experiment occurs as a result of an unexpected win or loss. Natural experiments essentially create themselves – they create a situation that provides randomness and allows researchers to simply observe and learn from them without having to intervene or set up the experiment to get the answers they want.

Dahl then introduced the theory of Gain-Loss Utility, which states that unexpected negative events have a larger impact than unexpected positive events.

He used the example of salary raises to explain gain-loss utility. If someone anticipates a 2 percent raise and the actual raise is 4 percent, then there is a positive shock to that person. However, if the anticipated raise is 4 percent and the actual raise is 2 percent, then there is a negative shock to that person. The theory is that the negative shock will have more of an impact than the positive shock.

This concept can be used to study responses to wins and losses in football games, as the same positive and negative shocks will occur with unexpected wins and unexpected losses.

With natural experiments, scientists can find causation and correlation between events without needing to do a lab or field experiment.

Dahl concluded that a lab experiment could not measure domestic violence, among other studies of human behavior, the way a natural experiment can. Natural experiments allow researchers to identify causality without requiring a lab experiment. Natural experiments can help economists and statisticians find answers to questions that may have seemed impossible to answer before.

By Jessilyn Gale Posted on