It is more logical to send a higher voltage through the cable, and then use a local voltage regulator on the other side.
I have a homebrew automation network at home, which sends 24V power and RS485 data over CAT-5 cables, and each node has it's own SMPS. CAT-5 is limited to about 1A, with linear regulators (each node 10mA to 100mA) the maximum would be around 20 nodes on a cable, but with the SMPS, a higher input voltage results in less current though the cable for the same power, and I guess I could add 100 nodes before I have to add some auxiliary power supply in another location.
Voltage drop over the GND wire is also a concern, although not a problem when done correctly. With CAT-5 and 1A, the voltage drop on a long cable can be as much as 2V per conductor (they are not trans atlantic). That would mean that data received on a node at the far end of the cable, can be 2V below it's local GND level if the data is sent from a node near to the power supply. GND levels change along the cable, but data does not, because current is low. That was one of the reasons for choosing RS485 drivers. These are officially specified to work with data levels between -7V and +12V. (which is 7V above the "normal" 5V power rail)