Thread by @Peoples_Pundit

The public polling industry will never win back the public trust if outlets just smear others in order to cover for their own shortcomings.

People aren't stupid. They just see it as a narrative battle, and use polls to affirm what they want to believe. That's not their purpose.

This thread is long overdue. I was holding off on it until I convinced others, but that's just not happening. So, here it goes...

Like anything else, the best way for media pollsters to combat distrust is sunlight, meaning a new level of transparency.

For instance...

AAPOR's Transparency Initiative was a great place to start.

But...

Wouldn't people be more inclined to trust that a media poll was done honestly, even if it turns out not to be completely accurate, if providers were given, modeling was much more transparent? Not minimal 411.

I'm talking about a completely new standard, with various modes on each survey (B/C responses vary per bloc per mode), and a lot more daylight on weighting (for who/what/why) was provided?

Cost is not the issue. Live caller interviews are the most costly to run.

With minimal 411, I've seen unbelievable claims. You're not drawing a 10K list from a file & collecting 800 responses. No way.

So...

Eg. A random sample (targeted/not) was drawn from file X that was X in size. We made X number of calls to reach X people. X completed the survey.

People tend to believe/disbelieve polls based on party identification splits.

That is too simplistic. It's an education failure on behalf of media and pollsters. Properly weighting for region and education has been the problem, not party.

Consumers don't get that information.

Maybe at one time it wasn't necessary to provide granular level data. But a few things have changed.

1. Social media has increased interest not just in polling, but in the minutiae of polling.
2. At the same time, public trust collapsed.

It's time to change, too.

If I had to imagine the most transparent poll I could, it would look something like this...

Map points for each mixed-mode response, datatables etc. Region would be obvious and granular data would be available for each. Various results for each mode (live/IVR anon/online) AND...

AND... results could be filtered based on mode and turnout. Naturally, sponsors would be openly available, as would data providers for voter files, online panels, etc. If a native panel was used at all, recruitment and screening methods would be available, as well.

Lastly, it would admittedly be difficult but ideal for efforts to be bipartisan. Modeling would be agreed upon and detailed. For LVs, some pollsters ask respondents to rate from 0 to 10, some from Not at all - Certain to Vote. Most, including media polls, disclose nothing.

Gone are the days when Gallup (trusted) tracked a national electorate, and the states fell in line. It's getting less accurate, more convoluted, as interest and distrust grows. I just don't see how adding to the confusion without vastly increasing transparency serves voters.

To expand on what I mean for filtering based on mode, the samples would be large enough to provide a result (or best estimate) for each vote share for each mode.

A combined result would also be given for the entire mixed-mode. Point being, we could restore public trust...

Point being, we could restore public trust by not only being more accurate, but by educating and being more transparent. If consumers had more 411 at their fingertips, then it would be MUCH easier to explain WHY survey results were inaccurate IF they turn out to be. Obvious even.

Or, continue with the status quo, which is viewed as a narrative battle rather than a public service. But smearing each other while pretending the problem has been solved B/C nationals were fairly close while statewides were grossly inaccurate, doesn't seem all that wise to me.

The status quo means a continuation of narrative wars, and name-calling such as "they're Republican-leaning" and "they're Democratic-leaning" or "they're a fake pollster." It is divisive, and it is a disservice to the American people. They deserve a system they trust.

Thread Update:

Well, that thread aged well. @pewresearch finds (68%) believe made-up news and information has a big impact on their trust in government.

Naturally, they see media commissioned polls in the same light.

We need a Daylight Poll.
axios.com/americans-fake…