Can we recover a speech signal with error signal??

Let we have a speech signal which is transmitted over a channel. By omitting the error signal e(n), it is still possible to recover the speech signal to some
degree. Since the error signal is not transmitted, less data has to be transmitted, although at the cost of less intelligibility and quality of the reconstructed signal. What compression ratio has been achieved by omitting the error signal and only transmitting the model parameters?