Removing a birth removes a full lifetime of emissions (all the cars that person would have bought) as well as any future children of that kid; over the long term, the extra emissions from adding a person grow exponentially, while adding a larger car (for a kid that would have existed anyway) is a one-time fixed cost. So, as t->infinity, fewer kids is going to win from a co2 perspective.
So the question is just where the crossover point is (if any; the 8000 fewer kids might already simply offset any larger cars bought within the first year - i don't know). I suppose that requires estimating the number of larger cars bought for third children per year, and comparing to the yearly emissions for people as they age.