drdel
Tour Rookie
20,000 as a survey sample size is very large, much more than for the usual polls, and is statistically very likely to be accurate.
The survey really has no importance, but it's wrong to write it off as inaccurate.
But it can be skewed by, time of day(workers: not home),
Method: land lines not used by many,
Email: not used as much by elder people.
I could easily conduct a survey and by carefully choosing method, regions/post code, etc pretty much guarantee the result.