The data throughput for a channel is defined by two things:
- Signal bandwidth
- Signal-to-noise ratio
This is expressed in the Shannon Information Theorem. That is:
$$C = W *log_2(1 + SNR)$$
Where W represents the bandwidth and C is the capacity of the channel.
Mind, this is the information capacity for data that has infinite entropy, that is, the data is truly random and thus isn't compressible by any means.
More about the Shannon Information Theorem here: http://www.inf.fu-berlin.de/lehre/WS01/19548-U/shannon.html
And, C. E. Shannon's actual paper: http://people.math.harvard.edu/~ctm/home/text/others/shannon/entropy/entropy.pdf
If we employ compression, the throughput can be increased substantially. This can range from about 2:1 for a simple scheme like G.711 mu-law to much higher using transform+entropy based codecs like MP3 and others.
A comparison of compression formats here: http://www.sericyb.com.au/audio.html