Another thing to consider is a level of participation when it comes to PPM. David could probably speak to this more.
Considering the majority of KIXI listeners probably fall under the senior citizen category, I think it's safe to assume that folks in assisted living or senior living situations, are probably less likely to be approached by Nielsen, let alone understand what participating involves and sign up. That, and getting back to my point #3; the Seattle area, particularly East Side, suffered a lot of senior deaths from COVID. That assisted living place in Kirkland was one of the first in the nation with a huge number of deaths. This sounds weird, but if your sample on a good month is a .7 share, it would be easy to imagine your share would drop noticeably if even 10% of your audience died, literally.
This is a complex question.
The PPM study is not a random probability sample like the diary survey is. It is a panel, recruited specifically to mirror the total population.
If 65 and over are 20% of the population is over 65, then 20% of the panel is of that age group. The subsets are also proportional by gender and income and county of residence and other stratification variables.
On the other hand, the panel is not proportional by occupation or retirement status or residence in a hospital or such. If a panelist does not participate for a few days, they are contacted. If they don't respond, they are dropped. But they are quickly replaced with another person from the same set of qualifications; Nielsen is constantly recruiting and they try to "immediately" replace drop-out. In fact, normally about 6% to 8% of the panel drops each month. Some have reached the end of the max of two years. Some are dropped because a member of the dwelling unit fails to comply. And other just get tired or don't like it any more. So they are recruiting all the time.
If, in this case, portions of the population die in significant numbers, that will not affect the ratings as Nielsen bases their estimates on projection of the panel into the universe as determined by annual Census Bureau updates as reported via private demographers. So if a group is reduced, you won't see the effects until Fall of next year. The panel will continue to be based on last year's population estimates and won't be changed until this year's data is released sometime next year. There is always about a year's worth of delays.
That said, it's never perfect. it's an approximation. And that is all we expect.
Generally a panel is a family or household unit, not a single person; however, persons who do live alone are recruited but there is not a family size quota. But the object is to have the right number of people in each age and gender and ethnicity, not specific occupations.
And the data used to recruit is less likely to have extended care patients listed as that, like a college dorm, is transitory. The survey uses broad strokes and your questions are very specific. Your assumption, though, is correct. There are people who won't be approached and there are some that won't accept. There are transients and homeless persons or those who change residence often who can´t be reached. And I am pretty sure that Mr Bezos or Mr Gates and most of their close friends don't join the panel, either.
Ratings are a sample, not a census. While in elections, a tiny fraction of a percent can separate winning and losing, in radio ad buys wobbles of a whole share point are common and ad buyers know that they are getting, as I said, an approximation... a metric... a way of tagging prices to delivery.