Look at the patterns in our database. The celebrity who killed a pedestrian and settled for ₩500 million: suspended. The chairman who beat employees over years and settled with each one: short sentence relative to count. The public figure whose stalking case was 'voluntarily withdrawn' after a confidential payment: fine + suspension. Settlement does work in the data — and it works disproportionately for defendants who can write large checks.
The doctrinal architecture is hard to fault on its surface. The Criminal Act, Article 51, instructs judges to consider 'circumstances after the offense.' Restitution is exactly that. Victims often genuinely benefit from receiving meaningful payments. The state benefits from reducing prison populations and case backlogs. So the system, in its way, encourages these resolutions.
The asymmetry shows up at the edges. When the same statute treats two defendants very differently, what differs is rarely the harm and almost always the defendant's ability to pay. The 22-year-old job-seeker who stole ten ramen packs cannot 'settle' the harm to the convenience store; the harm is too small and the store doesn't want money. The chaebol heir whose case turns on tens of billions can write a meaningful settlement check. Both defendants face the same statute, both technically have the same mitigators available — but one of them has the structural ability to *use* them.
This isn't about whether settlements should exist. It is about whether 'sincere efforts at restitution' means the same thing in two pockets of identical depth, or whether it means something different in pockets of different depth. The data, repeatedly, points to the latter.