Why In a phase modulator the frequency deviation produced rises in proportion to the modulating audio frequency, so a PM transmitter automatically emphasizes the highs by about 6 dB per octave (the same curve as the 75 microsecond pre-emphasis network used ahead of true FM modulators). A receiver detector recovers that boosted audio, so a de-emphasis network with the complementary falling response is needed to restore a flat, natural-sounding baseband. That is what lets one receiver design work with both frequency-modulated and phase-modulated transmitters.
Watch out Rolling off the highs does cut high-frequency hiss and some impulse noise energy, which makes the noise-reduction answer tempting, but that is a side benefit; the design reason the network exists is to undo the transmitter's pre-emphasis so PM and FM sound alike.
Pre-emphasis at the transmitter (or built into PM), de-emphasis at the receiver: the two curves must cancel.