To send a $10 \, KHz$ bandwidth of speech (Non Filtered speech), Nyquist criteria says that sampling duration should be $50 \, \mu s$. While for a $4 \, KHz$ bandwidth of filtered speech, sampling duration should be $125 \, \mu s$ (Much larger than $50 \, \mu s$.).
Now an excerpt taken from my textbook:
"In general, the wider the $BW$ of a signal, the more "expense" it will be to transmit the signal. So it is much "cheaper" to send a filtered speech over a communication link."
What are the factors related to the words "expense" and "cheaper" mentioned here? What does author mean by using these two words?