MEASURING THE EFFECT OF NAPSTER ON RECORDED MUSIC SALES: DIFFERENCE‐IN‐DIFFERENCES ESTIMATES UNDER COMPOSITIONAL CHANGES