During the boom, the computer always seemed to say “yes”. Automated credit scoring, where available data was crunched to spew out the likelihood of a consumer defaulting on a loan, meant only a few nuggets of information – a name, an address – were needed for a bank to price a loan and post an offer direct to someone’s doormat. Adherents claimed that this system was at least as effective as person-to-person banking. Still, a reliance on automatic credit scoring evidently failed to measure most of the risks that banks took on during the credit heyday.

One problem is that credit scores do not necessarily move with the cycle. During 2007 and 2008, analysis by VantageScore, a credit scoring provider, shows that the average score for the top third of the US population improved as newly cautious types paid down debt. The bottom third’s score deteriorated by about the same, with the middle ranks remaining stable. The trouble is that such scores represent a ranking of consumers, rather than overall default risk. Other factors, such as the development of riskier products (think Pick-a-Pay mortgages) and changing consumer behaviour, meant that the chances of default rose across the whole credit spectrum, even as rankings remained largely unchanged.