These three activities have different QoS constraints: "voice" data is real time, lossy, and therefore has hardware assisted compression and routing; text messages and data are non-lossy, but have drastically different timing and retry constraints. So, even if they "traveled through the same towers and pipes" (they don't, a detail more expanded on by mdasen in his comment), it wouldn't make sense to charge the same for them as they have different cost requirements (although it might be somewhat more equal than it is now; in particular, text messages, which occupy valuable and limited control channel space, would likely become cheaper).