Pages Menu

Categories Menu

Posted on Jan 28, 2016 in Bernie Sanders, Democratic Party, Hillary Clinton, primary, Public Opinion Polls | 3 comments

The Iowa Caucus: The Democrats–Updated: latest poll results and prediction as to who will win.

Iowa’s convoluted caucus system (and it’s a little bit different for Republicans than Democrats) will be held Monday. Today we are going to focus on the Democratic caucus. Later on this week, we’ll say something about the status of the Republican race.

As most of you know, Bernie Sanders had gained ground on Hillary Clinton in Iowa polls in the past two weeks. It always figured to get closer as the caucus neared, but about a week ago two polls (Quinnipiac and ARG) appeared that actually showed Sanders with small leads over Clinton–by 4 percent and 3 percent respectively. These polls were taken over varying periods ending January 21. Two others (NBC and Marist) that were taken more recently, covering rolling periods ending on January 26, and they put Hillary back in the lead by 5 percent and 3 percent. The partisan Democratic PPP poll, released January 29 gives Hillary an 8 point lead. Later that same day, Gravis, that uses one of the largest sized samples of the pollsters, had Hillary in front by 11 points. If all of the polls were equal it would appear that late momentum is favoring Clinton. But, as readers of this blog will remember, the various polls have different sized samples, they call-back, “no responses” (people who don’t answer their phone) different numbers of times, and weight their samples differently–as, for example, in the ratio of landlines to cell phones in their samples. The difference such seemingly small methodological variations can have on the results is apparent when data, upon further examination, reveal that Sanders has a sizable edge over Clinton among cell phone users. The polls also differ in their models of the likely caucus-goer. In Iowa registered Democrats, who usually only vote in the General Election, support Clinton over Sanders 52 percent to 41 percent. Will the Socialist label that Sanders has embraced encourage more Hillary supporters from this latter grouping to take the time necessary to attend a caucus precinct? How a poll answers this question in constructing their model of the likely voter could be a crucial difference. Iowa also allows voters to change their registration right up to and at the caucus. Which party’s will independent voters in Iowa be motivated to attend? When historical data are employed, and the sample is expanded to include those non-Democratic registered Iowans who have a history of voting in a Democratic caucus, Hillary’s lead shrinks a bit, 47 percent to 39 percent. Both are significant if not comfortably large leads. But what about the new voters? How many of them should be included in the poll samples? Historically, young voters turn out at the caucuses in much lower numbers than their elders. However, more were inspired by Obama–and they did turn out in greater numbers than usual for him–especially in his first race; Sanders has a huge lead among this group. The Monmouth poll weights the 18 to 34 year old cohort only 15 percent of their sample. By the same token, Hillary holds a significant lead among women. Historically more women turn up at the caucuses than men. Monmouth weights women 57 percent to 43 percent in their sample. Will those weightings match the actual turnout? These are just a few examples of how weighting of the samples according to assumptions about who are likely voters can radically affect the polls’ findings. Most such assumptions are data based–albeit historical data–with minor adjustments for assumed changes in the current race. As you can readily see, there is a certain amount of art to go along with the science in constructing poll samples. Even such a seemingly small thing as whether the voters attend the caucus alone or with someone else can affect  the outcome–Hillary has an advantage with those who expect to attend in the company of at least one other person. Primary election polls, where partisan sample respondents can’t simply rely on party labels to direct their vote, have always been less reliable than those for the general election. The very first primary election becomes even tougher. Both kinds of polls become more accurate the closer they are taken to the actual votes.

In Iowa, there is an additional monkey-wrench thrown in to confound prognosticators. Only about 20 percent of those eligible to vote actually come to the Democratic caucuses (and the votes are held at over 1600 different precincts). The method of voting at Democratic caucuses is equally unique–and challenging for pollsters trying to predict results. At each precinct, voters are asked to physically line-up in different parts of the room, according to the candidate they support. A lot of shmoozing goes on at Iowa Caucuses, and peer group pressures definitely come into play. Usually each grouping will select a few of their number to go over to the other groups and try and talk some of them to join their candidate’s group. This goes on for 30 minutes. Then votes are taken, tallied, and any candidate’s group who fail to reach the minimum threshold of 15 percent has to disband. If there were more candidates than just the three this time, bargaining could go on to allow unqualified groups to try and merge to allow one of them to qualify for a delegate. This time, in most precincts, Mark O’Malley’s groups won’t reach the threshold level. Which group will they join? And how many support him–polls vary from 3 percent to 6 percent. Caucus-goers are allowed an additional 30 minutes to haggle, argue, and plead with their neighbors to join their group. In a close race, the results could be determinative.

It’s not over yet, as precincts are then grouped, the results sent on to a higher level, and then yet to an even higher state level. The actual delegate selection follows a path as convoluted as the electoral college itself. By and large, however, the caucus night’s results determine the winner in Iowa. Meanwhile there is a systematic attempt once again, and very late in the in the Iowa race, to taint Hillary with emailgate. The trust problems that the polls show plagues Clinton, continues with Republican laser-focus attacks. This is much more of an issue with Republicans and Independents than with Democratic voters, but the refusal of the State Department to release some of the emails, subsequently labeled “top secret,” so close to the the caucus no doubt will be of concern to the Clinton camp. To what extent it will affect the voting isn’t known as the new revelation came before the most recent polls were taken. Many Democrats agree with Bernie Sanders when he said: “enough of those damned emails.” Even if it influences a small number of Democrats, however, this late in the campaign, and with little time to diffuse the implication of judgment error, it is possible that it could prove significant. I doubt that Bernie will try and exploit the issue. It is probably foolhardy to attempt to predict the result, especially with several days to go in the campaigning–and Iowa has fooled many a sophisticated prognosticator–but, I’ll chance it and say that Hillary will come out on top in Iowa. Bernie’s army of dedicated first-time voters may win the day, as Obama did defeating Hillary in Iowa in 2008. But I’m guessing that Hillary has the better organized ground game in Iowa and is more likely to get her supporters to the precincts across the state. That, and a discomfort with Sanders’ Democratic-Socialism, or at least, with the suggestion of higher taxes that his wildly ambitious program will entail, will motivate more voters in the 50 and older bracket (among whom Hillary leads by 54 percent to 34 percent) to make the effort to attend the caucus. Sanders’ best chance would be for an unusually large turnout. Monmouth’s model assumes 110,000 out of an eligible 150,000. As the turnout increases above their modeled number, polls indicate that Sanders’ support level rises. But there is some evidence that the all important late momentum is shifting Hillary’s way. However, at this writing a large snow storm is predicted for Iowa that is supposed to drop 6″ of snow. Will that skew not only the turnout size? Will it affect the usual age distribution of the non-attendees? Will elderly voters–who largely favor Hillary–be less inclined to risk life and limb by going outside in a snow storm? I do know that elderly Iowans take their voting responsibility very seriously–and the notion of “Iowa stubborn,” that was made famous in the Music Man musical, is legendary. Are the young Iowans, and others who don’t usually vote, and whose numbers favor Sanders, as determined as their elders? I don’t know of any data on the subject of which Iowans would brave a snowstorm to participate in their party’s caucus.

There is a great deal of excitement about this year’s Iowa caucuses, and who wins may be influenced by the size of a snowfall. How ironic after months of analyzing the polls. I am reminded of Shakespeare’s Midsummer Night’s Dream, where Puck observes: “Lord, what fools these mortals be.”

It may prove foolish using only polls taken before the new emailgate information, but I am sticking with my prediction of a Hillary victory.Late momentum, which is so important in primary elections, seems to be breaking her way. If the turnout is  normal, it could be an even bigger victory for her than most pundits expect.

Addendum Late Saturday, January 30

Late breaking news includes the following: The weatherman (or woman) is now hedging on the snow storm prediction. Now it is expected to arrive late Monday or on Tuesday. It all depends on the winds. Pollwise, the important Des Moines Register’s Iowa Poll was just released, and it has Hillary on top, but only by 3 points. Just as it appeared that the late deciding vote was breaking strongly for Hillary, based on the polls cited earlier, the Iowa Poll shows it a too close to call contest. That’s one way to look at their results. The other is that Hillary has gained three points from their last poll, while Sanders has gained only two. They do show, however, that Hillary’s supporters are firmer than Bernie’s. Turnout is still the crucial variable, as is whom the Register have identified as “likely voters.” Troubling for Hillary supporters is that the Iowa Poll was conducted over three days from January 26 to the 29th. They didn’t comment on whether there was any shift on the last day of polling, that in seems. The significant disparity between the two most recent polls, where Gravis had Hillary up by 11 points, and the Iowa Poll has her leading by only 3 points, is noteworthy. In both polls Clinton has improved her standing since they last reported, but the momentum may be less than it appeared. Iowa is infamocluded many hours after the new Hillary email information was made public. One might assume that if there were any major differences in responses on the 29th, as opposed to the earlier three days, that The Register would have pointed that out. I wouldn’t, however, bet on that assumption being correct, as logical as itus for late breaking shifts. Few expected Santorum to win the Iowa Republican race in 2012–not even Santorum staffers–not even this late in the game. Puck may be really chuckling right about now, not only at the poll disparity and the weather variability, but at anyone making a prediction at this juncture. I’ll give him more fodder for amusement by sticking to my prediction of a Hillary victory, but a big jump in expected turnout will likely leave me with egg on my face.

Addendum II Monday, February 1

Snow is starting to fall in parts of  Iowa and two new polls are just out with conflicting results. The small Emerson poll has Hillary leading by 8 points, but the much larger sampled Quinnipiac Poll shows the possible effects of the new email information, as it was completed just yesterday, January 31, and it shows Sanders now with the lead by 3 points. There is a very large gender gap and, as in 2008, Iowa men overwhelmingly are opposed to Hillary, while women support her by close to the same margin. Generally, in good weather, more women attend Iowa caucuses than men. The Quinnipiac Poll, like all of the earlier polls shows first time caucus-goers heavily for Sanders, while regular attendees break for Hillary. Turnout still appears to be the key to this election. It figures to be a nail-biter!

Check back for any late-breaking news.


  1. Based on the size of the crowds Sanders has attracted versus those of Hillary’s I expect a Sanders Victory.

    Also the Sanders Crowds Seem Much More Enthusiastic, My Money Is On Sanders For Those Reasons.

    • Frank, you are right about the big crowds Bernie draws. And his supporters are dedicated. But a big rally, whether for Bernie or Trump or whomever is not the same as a random sample of the larger voter poll. It is similar to Trump’s parading online polls, like Drudge, after each debate–a poll of Drudge users–who are overwhelmingly for Trump. When the real probability based polls of the larger Republican polity come out they always have numbers for Donald lower than Drudge indicated. In 2004, I was amazed at the big turnouts for Kerry—and he still lost the election. Gene McCarthy had GIGANTIC rallies, too in 1968. Yet he lost every primary, save the Oregon one, to Bobby Kennedy. McGovern, as well, in 1972–and he got slaughtered by Nixon in the election. We’ll want to see late polls to measure the effect, if any, that the new emailgate information has, but so far late momentum favors Hillary. Sanders’ only path seems to me to rest on producing huge caucus turnouts, even then he’ll have a steep uphill climb. Let’s revisit this after the election. And keep checking back for updates to this post as new poll results come in. I hope my analyses and methodological observations prove instructive to readers–ever the teacher.

      • See my Addendum posted late Saturday night about the just released new Des Moines Register’s Iowa Poll.

Post a Reply

Your email address will not be published. Required fields are marked *