Every class vote changes Frankenbanker's DNA. Not just the winning answer.
The class vote sets all four traits. These stay fixed for the rest of the experiment.
One slider raises detection and penalties, but cannot guarantee either perfect detection or zero fraud. Intense supervision also consumes resources.
One banker faces 65 hypothetical reporting decisions and 35 advice decisions. See whether your original regulatory setting deters fraud before management changes the incentives.
Raise the bonus or targets to encourage false reporting; obscure products to increase unsuitable advice. The banker and regulation stay fixed.
This is the visible enforcement record. It tells you how many fraudulent reports were caught, not how many were attempted.
Reveal the decisions the regulator could not fully observe. The banker's behaviour has already been determined by your earlier choices.
—
—
One banker, repeated choices. The class poll determines four traits. The same fictional person then faces 65 chances to falsify reporting and 35 separate advice decisions, not 100 employees.
Fraud rule: For each reporting decision, illustrative private gain (£000s) = 8 + 6 × greed level + 0.80 × performance bonus (£000s) + 14 × target-pressure level + an opportunity-specific variation of ±£25k. Fraud is attempted when that gain exceeds (perceived detection probability × personal loss if caught) + moral restraint. Moral restraint (£000s) = 17 − 4 × ruthlessness level.
Regulation: Intensity (0–100) jointly increases detection and the assumed personal loss if caught. Detection is reduced by cunning and opaque products; perceived detection is reduced by fearlessness. The highest modelled detection probability is below 100%. Reproducible fictional opportunities determine which attempts are caught. Maximum intensity still implies less than certain detection. The class selects the regulatory policy once; Slide 6 reveals how many fraud attempts were caught or missed under that policy.
Bank activity: “Legitimate services completed per 100 enquiries” = 86 + 14 × (1 − e^(−intensity/19)) − 0.17 × intensity − 0.0005 × intensity², rounded. It illustrates both confidence from safeguards and time/compliance frictions. This is not estimated sector output or a claim that heavier regulation necessarily harms growth.
Advice: Unsuitable recommendations are modelled separately from deliberate fraud and depend on targets, product opacity, character and oversight. Every numerical coefficient is a teaching assumption, not an empirical estimate, and the model omits many real-world mechanisms. You are testing logical sensitivity, not forecasting banking crime.