Why do the percentages change so much with the smaller sampling size? Did they change the demographics of the samples?
Smaller sampling size = easier to control the sampling universe attributes...and easier to get the desired results.
The liberal at Reuters/Rooooteers don’t like the results they see in their own poll since it’s Trump who is always on top, so they manipulate the numbers polled. It’s easier to manipulate 346 (this poll) nationally than say around 550 to 600 that the usually poll. It’s not hard to play with who they poll and how many they want to poll. And it gives all the Libs, Fox News a narrative for them to talk about that Trump is losing support. Yippeeee.
If you don’t have 1,000 respondents in your sample, it’s not a reliable poll. And that counts if you narrow down your overall sample to a sub-group...that sub-group has to still have 1,000. That’s what I learned in statistics class.
Like someone said above, it is a subgroup of the overall sample.
I'll drill down even farther for you to explain the Reuters graphic that I posted. The usual question they, as pollsters, determine "likely voters" - have you voted in the last election, or have you voted in the last presidential primary or caucus? Or another variation thereof of the question. If the persons polled says "yes" they get categorized as likely voters.
And yeah, the bigger the sample, the smaller the Margin of Error (MOE).