On the SERS depolarization ratio