What is the economic incentive to cheat? How does an experiment capture exogenous deviations?

Question

In my undergraduate career an economic experiment was conducted on my class:

• one class was the control
• two classes were the experimental (I was in an experimental class)

The basic premise was a modified prisoner's dilemma whereas two individuals had to make a decision regarding whether or not to 'cheat'. As this was a two stage game, the first person can choose whether or not they will cheat the second. The second can choose to cooperate or cheat. Each individual are in different classes to prevent collusion.

The monetary value (USD) was 1 for both individuals choosing to cheat, 5 if one cheated the other, and 3 if both cooperated.

Nonetheless, the experiment itself went well with no hiccups. But I approached the PI's of the experiment with the following:

• This would a be a poor measure of individuals within a population willingness to deviate from established rules as you've established what the options to cheat are, issues that are prevalent within society are often based upon individuals looking for overlooked loopholes, obscure mechanisms, or outright illegal means to obtain an advantage over their competitors. As a result, the results in this experiment are predetermined and not representative of a random and at times chaotic economic structure.
• If you wanted to explore an alternative measurement of behavioral economics, look for individuals who are unique in their attempts to cheat. Who go beyond the rules that you place in front of them and seek alternative means to maximize gains.
• An example: Given that the experiment happened sequentially, a student who was in a previous class holds the details of the experiment as leverage in return for a greater monetary reward, threatening to disclose the details (effectively blackmail). The individual-level analysis concludes that the student realizes that there is a greater 'pot' of money that can be had (if the total class sizes is 100, then there is a budget of at least $500 to cover the situation where everyone cooperated) and the options at hand: to play according to the experiment rules (1, 3, or 5 payoff) or not to and blackmail the experimenters for a greater payoff (>5 but <500). From the experimenter's perspective, a slightly higher expense to payoff the rogue student and in the process, preserve the integrity of the experiment (and the research value of the expended monies of conforming students).

Of course, I am in full knowledge that the the deviation itself in the real world would constitute a crime and liable to criminal prosecution then and now. But this does not deter crime given the prevalence of well... crimes committed for monetary gain.

Although the reply was along the lines of "We hope you don't tell the other classes about the experiment" I retorted that my goal wasn't monetary gain, but wanted to learn more about experimental methodology and parameters.

The basic premise remains unanswered. When an economic experiment is designed to have predetermined outcomes of ethical and unethical decisions, how does it address the extreme levels of deviation that is present within historical record? Enron, Bernard Madoff, Arthur Anderson, to name a few.

I am sure I am missing a couple of literary research pertaining to the experiment and would be open to any suggestions.

Answer 1

A couple of comments to your question below:

When an economic experiment is designed to have predetermined outcomes of ethical and unethical decisions, how does it address the extreme levels of deviation that is present within historical record?

1) External validity --- whether results obtained in the lab can be generalized to settings outside the lab --- has long been one of the main criticisms of experimental economics, and practitioners in this field are well aware of this. In light of this concern, one must therefore exercise caution in interpreting experimental results. Lab results should be extrapolated to an external environment only if the settings of the experiment sufficiently resemble those in that environment. (To some extent, the same can be said about generalizations of results obtained using field data.)

In the experiment that you participated, decisions were labeled as "cooperate" and "cheat". However, the labels could very well be "A" and "B", or "Left" and "Right", or "Aspiring" and "Cautious". The point is that the labeling of the actions is arbitrary, and you shouldn't (over)interpret the results just because the experimenter happen to choose a non-neutral frame.

2) Judging by your very limited description of the experiment --- e.g. you didn't say what the control treatment was like, or whether the subject moving at stage 2 can observe what was chosen in stage 1 --- I suspect that the experiment was more for a pedagogical purpose than a research one. Generally speaking, experiments for research purposes should avoid recruiting subjects from the experimenter's own class, since there are strong outside incentives that may affect/dominate the salience of the payoff from within the experiment. Moreover, experiments should generally adopt neutral framing, since non-neutral frames tend to bias the results in a certain direction.

Therefore, I think you are over-interpreting the results in an experiment that is not very rigorously conducted.