Corporate scandals like Enron, Madoff, Wirecard, and FTX have left a durable impression that the finance industry attracts people willing to take enormous sums from strangers. Yet in laboratory experiments where participants can share or split money, most people behave surprisingly generously. How can both pictures be accurate?
A team of researchers set out to probe that contradiction by recreating, and then expanding, an experiment that was designed to capture the psychology of large-scale theft. Their results, published in the Journal of Economic Psychology, suggest that finance professionals are not more willing to rob strangers than university students, and that a common safeguard in the banking industry, the “four-eyes principle,” does not change the picture much either.
Rethinking Selfishness with the Big Robber Game
The original experiment that inspired this study, run by Carlos Alós-Ferrer and colleagues in 2022, introduced what they called the Big Robber Game. Rather than asking one player to share with one other person, as in a classical dictator game, it gave a single “robber” the option to take a chunk of the earnings of 16 anonymous “victims.” The authors found that more than half of their student participants grabbed the maximum amount on offer, which was half of the victims’ endowments. Their interpretation: when the harm is spread across many people, moral resistance weakens.
Armando Holzknecht and colleagues at the University of Innsbruck’s Department of Banking and Finance wanted to push this design further. They asked two questions. First, would people who work in finance, a group frequently associated in the public mind with corporate malfeasance, behave differently from students? Second, would requiring two people to agree on a decision, the so-called four-eyes principle widely used in banking to prevent fraud, encourage restraint?
How the Experiment Was Built
The researchers ran an online study with 874 participants in total. That included 186 finance professionals recruited from across the European Union through the CFA Society Austria and an online subject pool called BEFORE, which includes employees of major financial institutions in private banking, trading, investment banking, and portfolio management. Another 441 students at the University of Innsbruck took part in the main extension, and a separate group of 247 students was used to replicate the original 2022 experiment as closely as possible.
Every participant started with 25 points, later converted into euros. Those assigned the robber role chose what percentage of 16 anonymous victims’ endowments they wanted to take. Victims had no say. In the replication, robbers could pick 0%, 10%, 33%, or 50% (the same options as the original). In the extension, the choice set was expanded to go from 0% all the way to 100% in 10-point increments.
To test the four-eyes principle, each robber in the extension made two decisions: one alone, and one jointly with a randomly assigned partner, communicating via text chat until they agreed on a figure. To probe stereotypes, the researchers also asked victims what they expected student and finance-professional robbers to do. Robbers were similarly asked to predict the behavior of their peers.
Participants also completed the Equality Equivalence Test, a short task designed by economist Rudolf Kerschbamer that classifies people into social-preference categories such as “selfish,” “altruistic,” “inequality-averse,” or “maximin” (someone who prefers maximizing the worst-off outcome). A survey at the end collected demographic information.
One notable design choice: finance professionals earned three times as much per point as students, a common convention in experiments with industry participants to reflect their higher opportunity costs. The authors acknowledge that not disclosing this rate difference to participants constitutes a form of deception, and the journal’s editors added an unusual note explaining why they accepted a deception-based paper in this case.
The Replication Holds Up
The researchers first checked whether the original “selfishness to the masses” pattern reappeared. It did. Among the 119 students in the replication arm, 68.1% chose the maximum option of taking 50% of victims’ money, compared with 56.3% in the 2022 study. The overall distribution of choices was statistically indistinguishable from the original, and other secondary findings (such as the absence of gender differences) also matched.
Finance Professionals: Not the Villains of the Script
The headline comparison in the extension surprised the authors. When deciding individually, students took an average share that was significantly higher than what finance professionals took. Pooled across individual and paired decisions, students still robbed more than finance professionals. Both differences were statistically significant on conventional tests, but only barely, with p-values of 0.048 and 0.040.
Once the researchers adjusted for multiple hypothesis testing, these effects no longer cleared the 5% threshold. The two groups also differed in many background characteristics: finance professionals were older, more often male, and positioned themselves further to the political right than the students.
To see whether the raw differences survived those background factors, the team ran what is called a multiverse analysis, a method developed by Uri Simonsohn and colleagues. Rather than picking one model, they ran 960 different combinations of regression types, control variables, and outlier rules to see how robust the result was across reasonable analytical choices. Only 76 of those 960 specifications (about 8%) produced a significant difference between finance professionals and students. The authors interpret this as evidence that once observable characteristics are accounted for, finance professionals and students behave essentially the same in the Big Robber Game.
The Four-Eyes Principle Has No Detectable Effect
The second research question concerned whether paired decision-making changed behavior. Across all 960 specifications in the multiverse, none showed a significant difference between individual and paired decisions. Within each subject pool, solo and paired choices produced nearly the same average amounts taken. The interaction between subject pool and group size was also statistically flat.
The authors note this is worth registering given how much weight the four-eyes principle carries in real-world banking compliance. At least in this setting, requiring two people to agree did not meaningfully reduce how much they were willing to take from strangers.
A Stubborn Stereotype
The exploratory part of the study turned to beliefs. Student victims, when asked to predict how much robbers would take, expected finance professionals to grab significantly more than fellow students, both when deciding alone and in pairs. The actual data showed the opposite pattern, or at best no difference. The researchers interpret this as a case where a common stereotype about finance professionals is not borne out by behavior.
Both victims and robbers also expected the four-eyes principle to restrain finance professionals, even though it did not. Victims and robbers also tended to underestimate how much students would take, and how much professionals in pairs would take. Only the average taken by professionals deciding alone matched participants’ expectations reasonably well.
Finally, the classification of people into social-preference types predicted behavior in the robbing game, but mostly for students. Robbers categorized as altruistic, inequality-averse, “kick-down,” or maximin-oriented took significantly less than those categorized as selfish. Among finance professionals alone, no significant correlation appeared between these preference types and robbing behavior, though the authors note the professional sample may be too small to draw firm conclusions on that point.
Caveats and Takeaways
The authors are cautious about scaling their results up. The experiment used moderate monetary stakes in an anonymous online setting, very different from the multibillion-dollar scandals that motivated the question. The different conversion rates for professionals and students also complicate direct comparisons.
Still, a few observations stand out. Taking money from anonymous strangers when given the chance is not a behavior unique to, or even elevated in, the finance profession compared with a student population. Differences that do appear in raw comparisons can largely be explained by demographics, political orientation, and distributional preferences rather than by industry membership itself. And public perceptions of finance professionals as uniquely greedy may be, at least in settings like this one, out of step with how those professionals actually behave.
