"Business culture and dishonesty in the banking industry", Cohn et al 2014 https://pdf.yt/d/7aIznZHMLtx1cosW / https://www.dropbox.com/s/bqg59zs60m2v0uq/2014-cohn.pdf?dl=0 / http://libgen.org/scimag/get.php?doi=10.1038%2Fnature13977 (NYT)

Here we show that employees of a large, international bank behave, on average, honestly in a control condition. However, when their professional identity as bank employees is rendered salient, a significant proportion of them become dishonest. This effect is specific to bank employees because control experiments with employees from other industries and with students show that they do not become more dishonest when their professional identity or bank-related items are rendered salient. Our results thus suggest that the prevailing business culture in the banking industry weakens and undermines the honesty norm

...On average, the bank employees had 11.5 years of experience in the banking industry. Roughly half of them worked in a core business unit, that is, as private bankers, asset managers, traders or investment managers. The others came from one of the support units (for example, risk or human resources management). Subjects participated in a short online survey. After answering some filler questions about subjective wellbeing, subjects in the professional identity condition were asked seven questions about their professional background (for example, ‘‘At which bank are you presently employed?’’ or ‘‘What is your function at this bank?’’). Those in the control condition were asked seven questions that were unrelated to their profession (for example, ‘‘How many hours per week do you watch television on average?’’).

After the priming questions, all subjects anonymously performed a coin tossing task that has been shown to reliably measure dishonest behaviour in an unobtrusive way 18–20 and to predict rule violation outside the laboratory 17 . The rules required subjects to take any coin, toss it ten times, and report the outcomes online. For each coin toss they could win an amount equal to approximately US$20 (as opposed to $0) depending on whether they reported ‘heads’ or ‘tails’. Subjects knew in advance whether heads or tails would yield the monetary payoff for a specific coin toss. Moreover, subjects were informed that their earnings would only be paid out if they were higher than or equal to those of a randomly drawn subject from a pilot study. We introduced this element to mimic the competitive nature of the banking profession 9 . Given that the maximum payoff is approximately $200, subjects faced a considerable incentive to cheat by misreporting the outcomes of their coin tosses.

...We conducted a manipulation check in which subjects converted word fragments into meaningful words. For example, they could complete the word fragment ‘‘_ _ oker’’ with the bank-related word ‘‘broker’’ or an unrelated word such as ‘‘smoker’’. This allowed us to test whether the treatment increased professional identity salience. The frequency of bank-related words in the professional identity condition was increased by 40%, from 26% in the control to 36% (P 5 0.035, rank-sum test), indicating that our manipulation was successful.

Figure 1a shows the binomial distribution of earnings in the coin tossing task that would result if everyone behaved honestly, and the empirical distribution from the control condition. Both distributions closely overlap, suggesting that the control group behaved mostly honestly. On average, they reported successful coin flips in 51.6% of the cases, which is not significantly different from 50% (95% confidence interval: 48%, 56%). By contrast, the bank employees were substantially more dishonest in the professional identity condition (Fig. 1b). On average, they reported 58.2% successful coin flips, which is significantly above chance (95% confidence interval: 53%, 63%) and significantly higher than the success rate reported by the control group (P 5 0.033, rank-sum test). Figure 1 shows that the treatment effect appears to be driven by two factors: (1) a higher fraction of subjects claiming the maximum earnings; and (2) an increase in incomplete cheating (that is, reporting 6, 7 and 8 successful coin flips). Assuming that subjects did not cheat to their disadvantage, the rate of misreporting is 16% in the professional identity condition. Alternatively, we can compute the fraction of subjects who cheated, which is 26% (Supplementary Methods).

Regression analysis (Extended Data Table 1) shows that the treatment effect is robust even when we control for a large set of individual characteristics such as age, gender, education, income, and nationality (P 5 0.034, Wald test). When also controlling for business unit and experience in the banking industry, we find that employees in core business units were more dishonest than those in support units (P 5 0.008, Wald test). However, the treatment effect is not stronger in these units because the interaction between the professional identity condition and working in a core unit is not significant (P 5 0.960, Wald test).