Write-Down-Only Cocos

As previously deduced, the conversion risk premium on cocos can become steep when the recovery rate [ρ] is zero and the annual probability of conversion [π] is as high as 4%. Line 20a of Table 4 above shows, for instance, an expected value of that premium [R – r *] of 4.46% for ρ = 0 and π = 4%. This internally consistent inference from a pricing model must still be confronted with market data where the conversion risk premium is represented by Bloomberg estimates of G-spread and π is implied. It turns out that the average of the values of π in the last column of Table 7 that satisfy the basic equation with ρ = 0 and r * given by the yield curve on US Treasuries for the respective G-spread is 3.96% unweighted and 3.80% when weighted by the US dollar size of each of the 10 issues. Hence the value of π implied by actual data for the cocos issues of the 10 banks featured in Table 7 is roughly the same as the 4% value set for π in the model solution described at the beginning of this paragraph. For this reason, the average of the G-spreads in Table 7 is also expected to be close to the conversion risk premium of 4.46% deduced from the basic equation as described before. Even though this premium is not likely to be the only contributor to the G-spread, the actual average G-spread was lower than the average spread deduced from the model for the conversion risk premium alone. It was 4.18% unweighted and 4.01% weighted by size of issue compared with 4.46%. Hence there is rough—but meaningful—correspondence between actual data and model-based inferences.