Last week the second largest bank in France, Societe Generale, announced that it had uncovered a 4.9 billion euro ($7.14 billion) of fraudulent trades, allegedly committed by a 31-year-old trader named Jerome Kerviel.

Before we decide which parties are to blame, let me tell you about some experiments we recently conducted on cheating with MIT and Harvard students.

We gave a large group of students a sheet of paper with 20 simple math problems but only five minutes to solve these problems. A third of the students submitted their sheets and got paid 50 cents per correct answer. Another third were asked to tear up their worksheets, stuff the scraps into their pockets, and simply tell the experimenter their score in exchange for payment–making it possible for them to cheat. The final third were also told to tear up their worksheets and simply tell the experimenter how many questions they had answered correctly. But this time, the experimenter wouldn’t be giving them cash. Rather, she would give them a token for each question they claimed to have solved. The students would then walk 12 feet across the room to another experimenter, who would exchange each token for 50 cents.

What is the point of all of this? We had the intuition that people could easily take a pencil from work home without thinking of themselves as dishonest, but that they could not take 10¢ from a petty-cash box and feel good about themselves. In essence we wanted to find out if the insertion of a token into the transaction–a piece of valueless, nonmonetary currency–would affect the students’ honesty? Would the token make the students less honest in tallying their answers?

What were the results? The participants in the first group (who had no way to cheat) solved an average of 3.5 questions correctly (they were our control group). The participants in the second group, who tore up their worksheets, claimed to have correctly solved an average of 6.2 questions. Since we can assume that these students did not become smarter merely by tearing up their worksheets, we can attribute the 2.7 additional questions they claimed to have solved to cheating. But in terms of brazen dishonesty, the participants in the third group took the cake. They were no smarter than the previous two groups, but they claimed to have solved an average of 9.4 problems–5.9 more than the control group and 3.2 more than the group that merely ripped up the worksheets. This means that when given a chance to cheat under ordinary circumstances, the students cheated, on average, by 2.7 questions. But when they were given the same chance to cheat with nonmonetary currency, their cheating increased to 5.9–more than doubling in magnitude. What a difference there is in cheating for money versus cheating for something that is a step away from cash!

So what’s going on here and what lesson can we learn about it for Societe Generale? As it turns out, it is much easier for us to be dishonest when we are one step removed from cash. This is why we are more comfortable taking office supplies home than cash; Why it is relatively easy for executives to cheat by backdating their stock options; and this I suspect is why Jerome Kerviel was able to erase $7.14 billion for Societe Generale. After all, he was dealing with stock derivatives that are multiple steps removed from cash. This might seem a very pessimistic perspective on human nature, but if we accept that when we deal with more abstract and nonmonetary currency our morality is less able to guard us against dishonesty, we might be able to learn some lessons from this disaster and reduce the likelihood of waking up one day to another Societe Generale.