My main take-away from this is that political philosophy must not spend much time with actual data if this is the level of numeric discourse. The models they use for very complicated decisions like the one described seem just crashingly unsubtle from this.