Definitely want to see the dataset and estimates for that, but 20% for NYC fits my impressionistic sense and largely maps onto the seeming consensus that the IFR is somewhere in the .5-1% range.
That said, there's definitely the risk of sampling bias insofar as those who go out to grocery stores and big box stores are likely to have had more exposure opportunities in general. My elderly relatives have avoided all shopping for the past 5-6 weeks and relied on younger folks (like me) and grocery delivery services to get their essentials, and none of those (unexposed) folks would show up in their sample.
The other thing I want to note, as I've said elsewhere, is that I think we need to seriously question whether the denominator of the NYC population is anywhere near the census figure of 8.4 million people. Probably half the people I know whose permanent residence is NYC are upstate, on Long Island, or out of state, so my guess is that we're talking more like 6 million people actively living and working in the five boroughs right now. That doesn't change the antibody prevalence
right now, but it does mean "when we get back to normal" and those couple million of people return, there will be a substantial portion of the herd who never got immunity (assuming such immunity meaningfully exists).
I'd really like to see Google or Apple's cell phone GPS data for NYC. I have to imagine there's a brobdingnagian reduction in the number of cell tower pings as compared to this time last year, and not all of that is tourism and regional commuters.
---
Meanwhile, can anyone understand how Sweden's chief epidemiologist can substantiate his
claim of near-herd immunity in Stockholm? What are they doing differently that enables them to have a higher prevalence rate than NYC with a much, much lower IFR?