If you live in a house like mine (I'll let you decide if that's fortunate or unfortunate), you have to put up with a lot of phone calls. Inevitably, one comes to notice that people markedly change their voices when talking on the phone. Obviously, we ourselves are not wont to notice this, but other people notice. Ask someone.
Anyway, can we explain this through economics? Like most things, yes. Assume that when we interact socially our body language is partially instinct: one's likely to be frightened on top of a roller coaster; but also partially controlled to impress: we're at lot less likely to scream at the top of said coaster if we're trying to impress someone.
When talking face-to-face, we have all sorts of weapons at our disposal to influence someone beyond our words: eye contact, body language, voice, physical positioning, etc, etc. Let's assume the law of diminishing marginal returns holds: you're likely to spread out your effort over these multiple avenues to optimize your chance to convey what you want. Thus, major changes - like a different voice - are unlikely, because they're "costly", in terms of social effort, compared to doing a few small things, like eye contact and a good handshake.
However, on the phone, we're reduced only to our voice. So, it's only logical that more effort is concentrated on the voice, even subconsciously, to produce the desired social result. Even if we're willing to "spend" less on phone interaction, some of the effort spend on other forms of body language gets shifted to the voice, resulting in a 'telephone voice'.