Explaining New Issuance Premiums in the US Corporate Bond Market (2016–2020)

The authors estimate the size of the new issuance premiums on the primary US corporate bond market from 2016 to 2020 to be about +1.5 bps. Linear regression of the NIP against new issue spread, market spread level and momentum, issue size, and tenor difference show only small predictive power. Therefore, the authors employ random forest models and use 17 explanatory features. The main determinants shown by the random forests are the number of syndicate banks, features capturing valuation uncertainty, the tenor of the bond, various market features, and most importantly the indicative premium at the initial pricing thoughts. The feature selection is largely confirmed by least absolute shrinkage and select operator (LASSO) regressions as a robustness check with a significantly lower model fit.