There have been, conservatively, at least 20,000 COVID-19 deaths in NYC alone. NYC has a population of 8,300,000. 20,000 / 8,300,000 puts a hard lower bound on

There have been, conservatively, at least 20,000 COVID-19 deaths in NYC alone. NYC has a population of 8,300,000. 20,000 / 8,300,000 puts a hard lower bound on IFR of ~0.24%. That is, if we assume that the entire population of NYC has already had SARS-CoV-2, then the lowest mathematically possible IFR for NYC would be ~0.24%. Ioannidis' paper says 2.49%-4.16% of Santa Clara County has already been infected. Santa Clara County has a population of 1.9 million. That means we should expect (1900000 * 0.0249) * 0.0024 == 113 deaths at the lower bound, and (1900000 * 0.0416) * 0.0024 == 189 deaths at the upper bound. Santa Clara County has already recorded 134 deaths, so that number seem to fit. But it fits with the hard, categorical data from NYC only if we assume 100% of NYC to have already been infected. TL;DR: Unless you're prepared to believe that 1) most of NYC residents have already been infected, or 2) that IFR is significantly greater than 0.24% and the actual number of COVID-19 deaths in Santa Clara County is multiples of what's reported then it's clear that Ioannidis' paper is wrong. #2 seems far more plausible and may even be true, but the controversy is the implication of his paper on derived IFR. And if #2 is what's happening, the IFR is definitely greater than what people applying the paper (including, AFAIU, Ioannidis' himself) predict.