There's a broad literature on this exact question. Of course, we do our own analysis within the department. Personally, I'm not in the camp of a “this thing pays for itself” type of conclusion. Depending on how you calibrate your models and expect feed-through into the economy, there is some literature out there that suggests that, dollar for dollar, it would pay for itself.
I think the bottom line is that, in terms of things you could be doing to generate either labour force participation or increased productivity, this is one of those things that's at the very top of the list in terms of return on investment from a government perspective. With either conclusion, there's a real payback here.