This is a daily update to my model projecting the results of the 2016 Democratic Presidential Primary. For details and links to previous posts, scroll to the bottom.
Click here for a complete projection for all Congressional Districts with current national polls (37% Sanders — 50% Clinton). [warning, large image]
Click here for a complete projection for all Congressional Districts if the race narrows to a 45%-45% national tie. [warning, large image]
The significant changes since yesterday are:
-
Updated NH polls, of course. This changes the model’s prediction for NH results to 57.1% for Sanders to 42.9% for Clinton. It now projects 56.5-43.5 in NH-01 and 57.7-42.3 in NH-02. I expect that the difference between the two CDs may be a bit larger than that. However, while NH-02 is pretty sure to be better for Sanders than NH-01, NH-02 does include some significant, heavily populated but geographically small areas that one would expect to be better for Clinton (around Nashua and Salem) — and maybe the model is picking up on that.
-
56.25% is the crucial delegate cutoff for both Congressional Districts and also for the statewide delegates. If a candidate gets more than that, they split the delegates 5-3. If it is closer than that, the delegates split 4-4. The handy chart below shows how delegates are allocated for any State or Congressional district that has fewer than 50 delegates. To see how to use it, look at the row labeled 8 (for a CD with 8 delegates available). Then look at the column labeled 5, and you will see that 56.3% of the vote is required to get 5 delegates (for a 5-3 split) in a CD with 8 total delegates. For 6 delegates, look over one more column to the right, and you will see that you need 68.8% for a 6-2 split.
-
NH-01 has most of the Boston exurb areas in the “southern tier,” as well as Manchester and its suburbs like Bedford and Goffstown. It also includes Derry and Londonderry, all large towns that were important to Clinton’s win in 2008. But it also has at least some towns that will probably be strong for Sanders, like the University of New Hampshire (Durham), Exeter, and towns on the Seacoast like Portsmouth, and some more rural areas along the Maine border up to the areas around Conway and Laconia. There is a possibility that Sanders could do better in Manchester than did Obama, due to his greater appeal to working class whites.
-
NH-02 includes Concord, the Upper Valley, Keene, and Hanover (Dartmouth University), where Sanders is likely to be strong and where Obama did well in 2008. But it also has Nashua, which Clinton won in 2008. Furthermore, it also does has a few large and important towns like Salem and Windham, Pelham, and Hudson in the “southern tier” next to MA, which helped carry Clinton to victory in 2008.
-
If the race were a 45-45 dead heat nationally, the model would expect Sanders to win by about 63.9-36.1 (63.3-36.7 in NH-01 and 64.4-35.6 in NH-02). If Sanders gets a win on that sort of scale, it would suggest that he may already have gotten a big bump nationally that hasn't been generally picked up since there haven't been any quality live phone polls other than Quinnipiac. But it won't be enough to get any more delegates unless he could get more than 68.8%, which is pretty unlikely.
- On the other hand, the purely Demographic/National Poll portion of the model is more skeptical of Sanders — it ignores the State polls and predicts Sanders 55.4 - Clinton 44.6 (54.9-45.1 in NH-01 and 56.0-44.0 in NH-02) on the basis of demographics and current national poll averages. If the result were to be like that, Sanders would miss the cutoffs for 5-3 delegate splits and NH’s delegates would be shared evenly between Clinton and Sanders.
-
A New York Poll (Siena)! — There is a New York poll out from Siena, showing Clinton up 55-34. That might not seem to be all that significant, since New York doesn't vote any time soon. But this is significant for the model, because it is the first polling indication we have of how New York might vote. I had been assuming for the national polls/demographics portion of the model that Clinton would receive a large home state advantage in New York, but I wasn't sure how true that would end up being. After all, Sanders grew up in New York, so it is sort of his home state too. Although Clinton is well ahead in this poll, it is actually a very good result for Sanders. The reason is that the model previously expected that Sanders should do substantially worse in New York than his national average, and Clinton should do substantially better than her national average, in large part because of the assumed home state advantage for Clinton. Although the poll does have Clinton doing a bit better than her national average and Sanders doing a bit worse, it is not by as much as the model expected. So it does not appear, from this poll, that Clinton is getting a very large home state advantage in New York. Accordingly, I reduced the assumed home state advantage in the national poll/demographic sub model by half. This, as well as the effect of the poll itself, is enough for Sanders to gain about 3 points. I’ll revisit this if and when we get more New York polls that might provide more clarity as to what sort of home state advantage (if any) Clinton seems to be getting in New York.
-
Two Michigan Polls — There are two new Michigan polls from Mitchell and IMP/Target Insyght, which have Clinton up 57-28 and 62-30. As a result of these strong performances, Clinton picks up about 2 points as compared to what the model previously thought was the state of the race in Michigan.
-
A low quality Florida Poll — There is a poll from Florida Southern College that has the race in Florida at 43% Clinton - 26% Sanders. Although the poll is low quality (from an unknown pollster), it is the only poll since November, sot he model does give it some weight. This has a ridiculously high level of undecideds (31%). Both Clinton and Sanders are lower than where the model previously expected they should be (30.3% for Sanders and 59.1% for Clinton). But Sanders is only a bit lower, while Clinton is substantially lower. The model interprets this by increasing the undecideds, which helps Sanders by about 2 points as compared to yesterday (with undecideds assumed to split evenly).
-
A North Carolina poll from High Point University has Clinton up 55-29. This strengthens her slightly in NC.
Projected vote in a 45-45 national popular vote dead heat:
Here's what the model projects the state of the race would be if Sanders gets enough of a bounce nationally out of Iowa and New Hampshire to pull even with Clinton:
National Vote by Demographics
I had meant to include this yesterday, but forgot. The projections for the national vote by different demographics have shifted a bit since the last time I posted these about a week ago. That is mainly as a result of taking into account the Iowa Caucus exit polls (actual voting!). These are derived from averages from 25 national polls since November (50% weight), and also from Iowa exit polls (50% weight). Methodology is explained in previous posts (see links at the bottom).
These charts show the estimated vote by demographic subgroups for current levels of national poll averages as well as for a hypothetical 45-45 tied national race.
This is a continuing part of an ongoing series using polling data, past exit poll data, census data, and other data sources to analyze the 2016 Democratic Primary.
Previous posts are:
- Poll Meta-Analysis: The Bernie and Hillary 2016 Coalitions, and how they compare to 2008 Obama/HRC
- Poll Data Analysis: The Current State of the Democratic Primary
- How the delegate math shakes out for Bernie and Hillary down to the Congressional District level
- Bernie Sanders Did Much Better With Non-Whites In Iowa Than You Think
- Democratic Primary Model, Feb 8 Update (Pre-NH)
For more detail on how delegates are allocated across different states, check out this excellent resource from Torilahure.