This is a slight derail, but to answer Vicxy on the previous page, and to allay any concerns around population surveys based on "small numbers" relative to the population, I have decided to do another post as this is important.
There is quite a bit of maths involved (for those who enjoy maths, this question and answers are relevant: math.stackexchange.com/questions/926478/how-does-accuracy-of-a-survey-depend-on-sample-size-and-population-size)
In particular, note the second answer which has less maths.
Generally speaking, the population size doesn't matter, the crucial factor is that the sample size is at least 1000. You get very diminishing returns for sample sizes larger than 1000, and good surveys are extremely expensive to conduct. I've worked on a number of surveys where I was doing population estimates, the survey was not face-to-face, and every additional person added to the sample size was a cost of around 10-20 pounds. This means that an extra 1000 people adds around 10,000 to 20,000 to the cost of collecting the data. This may not sound much, but when one is trying not to cut any corners with the survey (each question added can be another 500-2000 pounds, depending on the type of question as duration of answering the survey increases the costs), plus all the cognitive and field testing required where people are paid for their time and cognitive interviews require the survey people to sit with the people doing the cognitive interview and ask them questions about how they answer each question (and write a report with quotes, making recommendations for changing the original survey), you can see how costs start spiralling.
So organisations tend to go for the 1000 sample size. This may be adjusted upwards if we want small subpopulation groups to have accurate estimates made as well. A sample size of 1000 may increase to 1200 or 1500 so that the over-sampling of small groups does not have a negative effect on the overall population estimates ( we don't want to under-sample some groups to compensate for over-sampling others).
This is much more detail than I originally wanted to post. I am posting in solidarity with government departments who tend to use sample sizes around 1000 (because costs are always important for government departments as it is so difficult to get funding for surveys, as this requires additional funding over the normal departmental costs). Please trust the results of government surveys with this type of sample size.