Because people are more than just the skills they have that happen to apply to their current job function. And because management requires a skill set that -- like many others -- can be attained by someone who is willing to work to attain it.

I think of it in terms of leverage. If I knew for sure that my becoming a manager would let me be a force multiplier for my team, that they would all be enough better to more than compensate for losing me as an individual contributor, I would consider making the switch. Having been a developer myself, I would have insight into what gets in their way, and I could use my managerial powers For Good™ to get those things out of their way. At least that would be my intent. I've had excellent managers who had been good developers who chose this path.

Having said all that, one of my first managers early in my career was a high-functioning developer who was moved to a leadership role because that was the default expectation. He was a terrible manager; he played favorites and treated his responsibility as authority to be wielded against those he didn't like. I was fortunate that he liked me, but he stifled the early careers of some of my friends who were at least as good at the job as I was. So there is something to be said for not having developer-to-manager as a default expectation.

Recent research suggests no positive correlation between individual performance and manager performance in the same line of work. If a company insists on moving ICs to managers, even though this is irrational, the best thing they can do is retrain their lowest performing ICs into people management roles.