You can do that with reasonable accuracy by testing a control group. Say for instance TTC drivers? You know how many were identified as infected before the testing, you test them all and see how many are actually infected. The trick is selecting an appropriate group for testing both in quantity and typical exposure to infection. I think TTC drivers would likely be slightly higher risk than the average person (and I'm not sure about union/ethical resistance to testing them all), but that was the best I could come up with now. You want a sample of thousands from a variety of areas (local hotspots and contact tracing may affect results if you choose a geographic group to sample).Some scientists think that the actual infection rate is ten times the reported in most cases, but how would you prove that?