March 06, 2018

Beware, the more you trust data, the more you have to be absolutely sure about how to interpret it, and about what to do with it.

Sir, John Thornhill writes: “In his Alan Turing Institute lecture, MIT professor Sandy Pentland outlined the massive gains that could result from trusted data… the explosion of such information would give us the capability to understand our world in far more detail than ever before”, “Trustworthy data will transform the world” March 6.

Indeed, but that also leads to other bigger dangers, not only because we might trust that trusted data too much, but also because we might not know how to interpret or what to do with that trusted data.

Like for instance the regulators with their current risk weighted capital requirements for banks. These establish that the riskier an asset is perceived the larger the capital a bank has to hold against it. Does that make sense? Absolutely not!

It is not if the perceived risk is correct, meaning the ex ante risk perceived ends up being the real ex post risk, that poses any major danger for our banking system. It is if the risk perceived is incorrect, that the real big dangers arise. And, of course, the safer an asset is perceived, and the more bankers trust that perception to be right, the longer and the faster it can travel down the dangerous lane of wrong perceived risks.

What detonated the most the 2007 crisis? The securities backed with mortgages to the subprime sector rated AAA by “trustworthy” credit rating agencies, in fact so trusted that the Basel Committee, with Basel II, allowed banks to leverage 62.5 times their equity with such “safe” assets.