Long story short: really really bad. But there's a optimistic remark coming: we're almost as good at forecasting that we are at counting. In other words, we are bad at both, and should potentially question more heavily the reliability of monthly employment figures.