This article covers low confidence early forecasting of the 2016 Presidential Primaries and General Election.
Let me start by saying that everyone knows Hillary will have the Democratic nomination, unless she explicitly rejects it, which she probably won’t.
On the Republican side it’s much more interesting. There are 10+ candidates in consideration at this time. Using 3 sources (which in turn leverage multiple other sources) of polling and political futures price-implied probability, I am constructing my current low-confidence hypothetical primary election predictions below. Note that these are not my final or actual predictions in large part because these are not even the actual primary candidates:
- Romney (24%)
- Bush (11%)
- Christie (8%)
- Paul (8%)
- Carson (8%)
- Huckabee (6%)
- Ryan (6%)
- Walker (6%)
- Cruz (6%)
- Rubio (4.5%)
- Perry (3.5%)
- Jindal (3%)
- Kasich (2%)
- Santorum (2%)
- Fiorna (0%)
- Graham (0%)
- Pence (0%)
- Portman (0%)
The 3 sources used are:
- Real Clear Politics – 2016 Republican Presidential Nomination
- Wikipedia – Nationwide opinion polling for the Republican Party 2016 presidential primaries
- PredictIt – Who will win the 2016 Republican presidential nomination?
RCP is great because it aggregates many polls. Problems include the fact that 1) these polls are run on a straight average, which renders them insensitive to important fluctuations over time, and 2) these polls fluctuate in the people included in each poll. At the present time it is also missing the most recent national poll, conducted by The Economist, which Wikipedia records.
Wikipedia is useful because it documents a larger number of people running and it also includes the most recent poll.
PredictIt is great for ordinal likelihood of candidate win chance, but terrible for actual vote outcomes. This is because it operates on futures market principals. If one choice is clearly better, even if only marginally so, investments will pour into that choice at a rate disproportionate to the true outcome. Basically, people are not betting on a target outcome, they are betting on the expected winner and this distorts the ability to interpret a vote outcome from the market, but we can interpret an ordinal ranking of expected winner.
What I did is I basically took RCP values and slightly arbitrarily adjusted them based on two factors:
- An arbitrary additional weight toward more recent results.
- An adjustment to include additional candidates they did not consider, including a largely arbitrary selection of how much the added candidate would gain and how much the existing candidates would lose.
I would arbitrarily draw the relevance line at Rubio, who is ranked number 10 with an expected 4.5% vote. If we discount those after 10 we can probably bump Rubio’s expected vote up to 5% ish which is consistent with an arbitrary but reasonably drawn relevance line.
One really interesting thing I would note is that alot of people are talking these days about the fact that 6/10 Republicans want Mitt Romney to run. My results indicate that only half of those people would actually vote for him if he did run. Why would people want Romney to run if they aren’t going to vote for him?
I think we may be seeing strategic voter preference on an unprecedentedly large scale. I think many Republicans want Romney to run because he will radically diminish the votes obtained by Bush and Christie among others. The start difference between The Economist’s poll and the next most recent CNN poll is great evidence of this.
On the CNN poll Romney was not considered and Bush garnered 23%, but on The Economist’s poll Romney was considered and Bush dropped to 12%.
Many Republicans may be acting ironically because they will succeed in reducing votes for people such as Bush by advocating Romney jump in, but if Romney does jump in he seems to suddenly be the forerunner, which is not what these people will actually be desiring if their true concern is issue-based representation.