It’s a statistical argument based on the observed distribution of asteroids. It’s not precisely true, and could even be widely incorrect, but they think it’s likely to be in the correct neighborhood.
Also, after a certain size things become significantly easier to detect with your existing instruments, so there is some confidence from that.
Related to this (though maybe not directly applicable), in certain scenarios there is a difference between what can be reliably detected vs tracked. For example, in the context of debris orbiting the earth, our radars can take LEO census measurements down to 3-4mm, even though we can only reliably track things on the ~10cm size range.
We can therefore semi-confidently say we know how many 1cm objects there are out there, even if we can’t predict the trajectory of any single one. (Similar to atoms, really)
Also, after a certain size things become significantly easier to detect with your existing instruments, so there is some confidence from that.