Colorado has managed to reduce teen birth rates by 39% and teen abortion rates by 35% in five years.* They did this with one simple program: providing free or low-cost, long-acting, reversible contraception (such as IUDs and implants) to low-income women.
It's hardly surprising that contraception is the most reliable way to reduce unintended pregnancies, since that's what contraception is for, after all. Those of you who have followed this will probably know that the cost of contraception is actually a huge barrier to entry for low-income women to use it; $40/month may not seem like a lot, but when your budget is already close to the line, it's a big difference. Reducing that cost therefore can have a big impact on use.
The "long-acting" part is one of the most interesting parts: to quote Dr. Stephanie Teal, head of the state Dept. of Public Health's family planning division, “With an implant or an IUD, if someone wants it out, we take it out, but once it’s in and they have to make an appointment to take it out, they really have to think, ‘OK, do I want a baby now, really?’ As opposed to the pill, you basically have to decide every single day, ‘Do I want to be not pregnant?’ And some days, you might want to be pregnant.” That's especially important for teens, who (as you may recall or know from your own kids) are not known for having the most reliable long-term thinking and planning skills. (For real reasons: the parts of the brain which control that don't fully mature until your mid-20's. The idea that your brain stops growing in childhood has turned out to be wrong, wrong, wrong.)
So what we have here is a simple, highly cost-effective method for reducing both unwanted pregnancies and abortions: it's based on reliable medicine, puts individual choice first (nobody is forced to have this!), makes the people getting it happier, and on the whole, saves the state a bundle.
This is what I call a good idea.
* Compared to a nationwide decline of 29% in teen birth rates. CDC (who did the study) reports the delta as statistically significant. Correlation is not causation, but cf the other studies referenced in the article, such as the Missouri one indicating the relative efficacy of long-term versus short-term birth control, and the drop in repeat pregnancies seen in the CAMP study. Not a smoking gun, but from an engineering perspective, this is what I would call "working well enough that I would definitely double down on the strategy."