6
Safety in Numbers
Many credit markets behave as if they are inferring from
only a few dozen years of relevant observations and frac-
tional evidence of default. Their beliefs are modeled most
plausibly as highly dispersed distributions, where the
consensus might be wildly wrong. The uncertainty calls
for much larger contingent reserves on top-rated credits
than standard regulatory calculations recommend.
Modern bond markets are highly sensitive to perceived default risk.
Credit spreads are normally mea sured in basis points (bps), which are
one- hundredth of percentage points. While default might not trigger com-
plete loss, the residual or salvage value of a defaulted claim is oft en less
than half the nominal value. Ideally the market wants a 1 or 2 bps accuracy
in its estimates of default risk per annum.
Th ere is no way to achieve that degree of accuracy for the debtors we
care most about. Knowledge is always rooted in directly relevant observa-
tion, and we don’t have nearly enough. So analysts branch into indirectly
relevant or possibly relevant observation and draw on their imaginations.
Th ey are lucky to obtain one- in- one- hundred part accuracy.
Th nancial risk analysis. Yet it receives is problem is fundamental to fi
scant attention. To redress the balance, let’s start with a real- life example.
Real-Life Ignorance
Th e coun- e Soviet bloc imploded around 1990, give or take a few years. Th
tries that emerged from the wreckage were a motley crew. Some had strong