(CNN)The Iowa caucuses are arguably the most difficult assignment for any pollster. That is because relatively few people attend them, and people can, and do, change their minds in the caucus room.
How proposed Democratic caucus rules make polling even harder, and how we crafted our best shot
Any poll done ahead of that faces the challenge of finding likely caucusgoers and gathering enough data to assess the "what ifs?" of caucus night. I've said many times that it should not be possible to be accurate in polling caucuses. All we can do is take our best shot.
For the past several caucuses, that's worked out OK for the Iowa Poll.
Next year, a lot will change. The Iowa Democratic Party is finalizing rules that will add a new universe of participants to the mix. Six virtual caucuses will be held in an effort to make the first-in-the-nation contest more accessible to those who might not be able to schedule several hours on a February Monday evening to take part in person.
These virtual caucuses, as they are currently envisioned, will proceed much like the in-person meetings, except via phone or computer. There will be a message from the party chair, statements from candidates' campaigns, and perhaps some other party business. Then participants will be asked their candidate preference, and that will be done five times to help the party work out which candidates are and are not viable. So, it is still a commitment of time.
All of this doesn't change much of what a pollster does. We can easily expand our definition of a likely caucus participant to include those who intend to engage via a virtual caucus.
But, here's the twist. No matter how many Iowans choose to participate virtually, under the proposed rules, they will account for 10% of the delegate equivalents that will be reported out on caucus night. Those who participate in person will account for 90% of the delegate equivalents.
Another twist is that the party will report out numbers never made public in previous Democratic caucuses: raw counts of both the initial preferences and the final preferences recorded on caucus night -- both for the virtual caucuses and for the in-person meetings.
Today's reporting in The Des Moines Register and on CNN shows our best shot at producing three estimates for the horse race. You'll see the results of combining both in-person and virtual caucus participants, weighted so that the virtual group is 10% of the total.
We found that virtual caucus participants were 28% of the people who passed our screen as a likely caucus participant. But we made them look like 10%, to reflect what the party will do on caucus night. This is our estimate for the delegate equivalents, which are likely to be the first numbers reported out.
In addition, you'll also see the preferences of the in-person caucusgoers and the virtual caucus participants reported separately on the methodology document we release online. We cannot do anything about the final preferences -- we never could, and we cannot now.
Because of these changes, we start afresh with this poll of two groups -- those who intend to caucus in person and those who intend to caucus virtually, with no comparisons to past polls that were not conducted this way. The results show many similarities, but some major differences, too.
Those who say they are likely to choose the virtual option are younger, more moderate, and more likely to be currently registered as "no party," which of course they must change to participate in the Democratic Party caucuses. They are also less committed to caucusing than those who say they intend to show up on caucus night.
If the goal was to bring new blood into the caucuses, this approach appears to be doing just that. A majority of virtual participants in our poll say this will be the first time they've ever participated in any caucus. By comparison, just 17% of in-person participants say it will be their first time.
It's easy to think about 2008, when a majority indicated in the caucus night entrance poll that they were attending their first caucus. That group put Barack Obama far ahead of Hillary Clinton and John Edwards, so he carried the evening.
It is hard to imagine an influx into the virtual caucuses so great as to affect the outcome of how the delegates are distributed. No matter how many pick that option, they will make up 10% of the final delegate count.
There will be plenty to talk about on caucus night, for sure. Many will look at initial and final preferences for the virtual caucus and in-person participants and likely argue about how many candidates can claim a top-three ticket out of Iowa, historically a prerequisite for winning their party's nomination.
Our aim in designing this poll was to best capture the opinions of likely in-person and virtual caucus participants today, then apply the processes that will be used on caucus night. We offer this detailed explanation in the interest of full transparency. We launched this particular poll with the expressed understanding it would be an experiment -- to see how our best-laid plans would work with real data. And, it is our best shot.
J. Ann Selzer, who has a doctorate in communication theory and research from the University of Iowa, is president of Selzer & Co. Her firm is conducting Iowa Polls on the 2020 caucuses for the Des Moines Register, CNN and Mediacom. She has conducted Iowa Polls for the Register as a staff pollster or contractor for 27 years.