Some research  published in the Economist's Free Exchange column looks at recent antitrust behaviour in companies and finds that whether a company breaks the law has much more to do with simple economics than the morals of their CEO.
Instances of crime followed a clear cost:benefit pattern - if the odds of getting caught were too low or the punishment too light, breaking the law started too look far more attractive.
The column looked at cartels that operated between 1990 and 2005, and measured gains against potential penalties:
The first step is to measure the expected gain from crime which fines need to offset. In the study  by Messrs Connor and Helmers, the median amount that cartel members overcharged was just over 20% of revenue in affected markets.
Next, you need an assumption about the chances of being found out: a detection rate of one cartel in three would mean trustbusters were doing well. In this example, that would mean a fine of 60% of revenue is needed to offset an expected benefit of 20% of revenue — far higher than the fines in the study, which were between 1.4% and 4.9%.
It seems that for company collusion like this a 60 per cent revenue penalty is needed. This seems high, but in light of other recent corporate scandals the approach might be worth considering.