How economic data can often be both 'worse' and 'good'