According to the DTS paper, the argument for using DTS instead of spread duration itself is that relative spread volatility is more stable than absolute spread volatility. Then surely the same argument applies on yield as well. So why do people not use duration times yield?
Yield volatilities have not turned out to scale with yield. Otherwise yields would not have managed to become negative. The paper claims though that spread vol scales well with spread. Hence there are no similar arguments to apply .