Read on Twitter
I have some experience with this (of course).
I once ran a 60 station CATI market research phone room.
One customer study alone paid all the bills - a new car company called #LEXUS.
We did a daily collection of data for a continuous survey, 7 days/week. https://twitter.com/unseen1_unseen/status/1312555558818525184
I once ran a 60 station CATI market research phone room.
One customer study alone paid all the bills - a new car company called #LEXUS.
We did a daily collection of data for a continuous survey, 7 days/week. https://twitter.com/unseen1_unseen/status/1312555558818525184
We called each time zone using RDD (random digit dialing). I'll never forget one survey I did where the respondent wanted to know how I got her number.
Me: its random.
Her: but how'd you get MY number from random ones?
Me: we made it up.
Her: no, really, how?
Me: ok....
Me: its random.
Her: but how'd you get MY number from random ones?
Me: we made it up.
Her: no, really, how?
Me: ok....
Me:...your neighbor told us.
Her: i knew it! That batch!
Anyway, we had random numbers for every area code. We had a quota of surveys to get, by hour, by time zone, by sex.
The hardest to get are/were New York City folk.
We called to get people's attitudes regarding cars.
Her: i knew it! That batch!
Anyway, we had random numbers for every area code. We had a quota of surveys to get, by hour, by time zone, by sex.
The hardest to get are/were New York City folk.
We called to get people's attitudes regarding cars.
We were checking if advertising being run impacted people's attitudes.
Hourly quotas were important bc they cross referenced the hour with their TV advertising schedule.
Believe me the same thing is being done w political market research.
Hourly quotas were important bc they cross referenced the hour with their TV advertising schedule.
Believe me the same thing is being done w political market research.
People who have been doing this a while know where the biased results will come from. Folks collecting surveys will over-sample biased area codes at will.
These surveys are being used to gauge efficacy of advertising as well as measure opinion of trump.
These surveys are being used to gauge efficacy of advertising as well as measure opinion of trump.
"Opinions" cannot be "measured."
One really frustrating part of phone surveys is getting people to measure how much they like something (eg "on a scale from 1 to 10 w 10 being best, how would you rate the color of your car?").
Respondent: idk, a 4 or 5
Me: is it 4 or is it 5?
One really frustrating part of phone surveys is getting people to measure how much they like something (eg "on a scale from 1 to 10 w 10 being best, how would you rate the color of your car?").
Respondent: idk, a 4 or 5
Me: is it 4 or is it 5?
Respondent: I guess 4.5
Me: I have no 4.5 choice, so is it 4 or is it 5?
And so on.
I later got an MBA where I learned why there was no 4.5. I would explain it now but this thread is already too long.
Anyway, point is surveys mean zero when trying to describe population.
Me: I have no 4.5 choice, so is it 4 or is it 5?
And so on.
I later got an MBA where I learned why there was no 4.5. I would explain it now but this thread is already too long.
Anyway, point is surveys mean zero when trying to describe population.
And nationwide surveys that oversample liberal area codes mean even less than zero when forecasting an electoral college vote.
Trust no polls except the election results. Even those may be suspect in some parts of the world!
Trust no polls except the election results. Even those may be suspect in some parts of the world!
