A Question about the MorningConsult Poll

[bluewater]

MorningConsult says their latest results are based on 16,188 interviews with registered voters collected over the last week.


My question is, do they call the same registered voters over and over?

Or are the interviewed registered voters selected randomly each week?

I ask because 538 only gives them a B- rating, despite MorningConsult interviewing more voters each week than most polls.

[crazytown]

1. Maybe they do call the same people each time

May 06-12 - Biden 39, Sanders 19
May 13-20 - Biden 39, Sanders 19
May 20-26 - Biden 38, Sanders 20
May 27-02 - Biden 38, Sanders 19
Jun 03-09 - Biden 37, Sanders 19
Jun 10-16 - Biden 38, Sanders 19
Jun 17-23 - Biden 38, Sanders 19

[DemocratSinceBirth]

2. Because it's an online poll and technically not a random sample.

I ask because 538 only gives them a B- rating, despite MorningConsult interviewing more voters each week than most polls.

And the rating isn't based on number of respondents. It's based on historical accuracy.

[bluewater]

3. It's an online poll?

I ask because they say their results are based on 16,188 interviews.

Thanks for clarifying how a 538 rating is assigned.

[DemocratSinceBirth]

4. It's an online poll designed to mimic a conventional random sample poll

To be distinguished from an online clickbait poll. But it is not a true random sample as everybody in the universe they are polling does not have the same theoretical chance of being included. The cost to contact 16,000 people by phone would be astronomical.

[bluewater]

5. Thanks!

I was baffled how they could poll 16,000 people every single week.

The unchanging nature of most of the weekly results also started to seem strange.

Thanks again for the info.

[pnwmom]

6. Right. But they could do a real probability poll , with a random selection,

of 1,000 people and get more accurate results.

This is just a cheap and dirty poll.