June 10 2016

How incentives could eliminate cheating

Ismael-Rodriguez-Lara MiddlesexSenior Lecturer in Labour Economics Dr Ismael Rodriguez-Lara explores the reasons why people cheat and proposes a radical solution that could put an end to dishonest financial behaviour.

In December 2008 the world learned that Bernie Madoff had perpetuated the largest fraud in US history, losing his investors more than $18 billion. However while the scale of Madoff’s fraud was unprecedented, as an example of human cheating it is almost depressingly ordinary.

In the years since Madoff’s fraud was uncovered we have seen further high profile examples, such as the LIBOR scandal and the rogue trader who cost UBS more than $1 billion. These are merely well known examples of much broader trend – in 2015 financial fraud in the UK increased by 25 per cent.

Why do people cheat?

Such cheating comes with great cost. The European Commission estimates that economic losses from tax evasion within the eurozone alone are as much as €1 trillion per year. The non-economic costs of cheating, such as the loss of reputation or trust, are also significant (if somewhat harder to quantify). Given the high price of dishonesty, one might consider why people seem to cheat so often?

Over the last decade experimental research has begun to provide insight into the factors that influence cheating. One significant factor is the presence of incentives and research has shown that team incentives, performance-based bonuses and random bonuses all seem to promote cheating behaviour. However incentives are not necessarily corrupting and if well designed they can create better outcomes.

In 2015 we ran an experiment to test whether the design of incentives would affect patterns of cheating. We recruited 210 participants and split them into three groups. The participants in all groups were paid to complete a simple task – roll a ten-sided die which had an equal chance of landing on any number from one to ten – and then report their score. No one could see them doing this so only they would know their real score. The differences between the groups were when and how much the participants were paid.

The control group

Group A - cheating

The participants in Group A were the ‘control group’. They were paid a fixed amount of €2.50 to roll a die and then report their score. As we expected, the participants reported a roughly even distribution of scores from one to ten. This suggests that there is no cheating when there is no associated financial gain.

The gain group

Group B - cheatingThe participants in Group B were the ‘gain group’. Before the experiment they were told that they would be paid based on their score – participants reporting a score of ten would be paid €5 while those reporting a score of one would be paid nothing at all. Perhaps unsurprisingly the participants in this group reported a suspiciously high number of scores of eight, nine and ten.

The loss group

Group C cheatingThe participants in Group C were the ‘loss group’ and they were also told they would be paid based on their score. However, instead of being paid after reporting their score they were given the maximum payment of €5 before the experiment. When we collated the results from this group we found that they had reported a roughly even distribution of scores from one to ten and this was very unexpected.

Loss averse

Humans are ‘loss averse’. We do not like losing what we already have and find losses twice as painful as we find gains pleasurable. On this basis we had expected the participants in the ‘loss group’ to cheat even more than the participants in the ‘gain group’. However based on the results they do not appear to have cheated at all. This result is both surprising and exciting as it suggests that upfront payments may reduce cheating behaviour. What is not yet clear is exactly why this would happen.

One explanation is that the upfront payments increase the psychological cost of cheating. The theory of ‘self-concept maintenance’ suggests that when contemplating cheating, a person will weigh up their opportunity to make a financial gain with their desire to maintain a positive self-image an honest individual.

What we can say with confidence is that the costs of cheating are enormous.

This explains why although there was clear evidence of cheating among participants in our ‘gain group’ in general they did not maximise the benefits of cheating (by reporting a score of ten). Instead many cheated by reporting scores of eight or nine – sacrificing some of the potential financial gain from cheating in order to maintain the perception that they are ‘mostly’ honest individuals. It is possible that participants who receive an upfront payment perceive that cheating will come at greater cost to their positive self-image than those who only receive payment after completing a task.

An alternative but related explanation is that upfront payments draw more attention to cheating behaviour and reduce it as a result. Other experiments have shown that people are less likely to cheat if they are asked to recall and then write down the Ten Commandments before they complete a task (this holds true even if people cannot accurately recall any of the Ten Commandments). Drawing attention to a moral standard causes people to pay more attention to their own behaviour and reduces the incidence of cheating. It is possible that upfront payments also draw more attention to the act of cheating and so decrease this behaviour.

Support our research

What we can say with confidence is that the costs of cheating are enormous and amount to trillions of pounds every year. We know from previous research that incentives can drive this behaviour and yet our results suggest that carefully designed incentives may also play a role in preventing it. If we can understand how to structure incentives which prevent cheating in our experiments, we may eventually be able to use these insights to prevent real world corruption and fraud.

We have just launched a crowdfunding campaign to support the next phase of our research. Please consider donating to help us discover how incentives might be used to prevent, cheating, corruption and fraud.

Share this post

 

Tags: , , , , , , , , , , , ,

Leave a Reply

Latest tweets from Middlesex Minds

Follow MDXminds

@

Related blogs