Q: How does a US Presidential Primary work? What are these Delegate things?
A: A US Presidential primary is a series of elections that takes place in each US State and territory for each major party. Depending on the winner of each contest, delegates are given out to candidates which then vote to confirm that candidate at the Convention of that party. In 2020 both Democratic and Republican parties will have Presidential Primaries, but the incumbent President (Donald Trump) is expected to run mostly unopposed so there is no point in making a model for it. The Democratic Primary will be far more interesting, so therefore a model was made for it.
Q: How does the Democratic Primary work? Why is there an interactive map of State Congressional Districts?
A: In the Democratic Primary delegates are given out proportionally to each State and Congressional District. In accordance to Democratic National Committee (DNC) rules, only candidates that receive over 15% of the vote in each state or in each Congressional District get delegates for the Democratic National Convention (also DNC). Let’s examine a theoretical race between two candidates (Candidate A and B). Let’s say Candidate A receives 86% of the vote statewide to Candidate B’s 14%. In this case, Candidate A would receive 100% of the statewide delegates. Now let’s imagine that there are multiple Congressional Districts in that State, and that in one of the Congressional Districts Candidate A received 80% of the vote to Candidate B’s 20% of the vote. In that Congressional District, the Delegates would be split 80%-20% for Candidate B even though Candidate B did not win any statewide Delegates.
Q: What happens if no candidate receives 15% of the vote in a State/Congressional District? Is that incorporated in the Model?
A: If no Candidate receives 15% of the vote, the new cutoff to receive delegates becomes half of the highest percentage of the strongest Candidate. The Model has a built in Delegate Rounding Function specifically designed to address any delegate awarding issues, and will always give the correct amount of delegates in accordance to DNC rules.
Q: How many Delegates do you need to win the nomination? What happens if nobody has that amount?
A: 1885 Delegates are needed to win the nomination. If no candidate has that amount on the first ballot, then after the first ballot it becomes a contested convention, and then anything can happen (Chaos). There is a chart on the model page that shows exactly what the chances of “Chaos” are.
Q: Why does the Congressional District Map have two Congressional Districts in Montana? The last time I checked they only had one district in Congress. Also why do the Congressional Districts in Texas look weird?
A: The DNC lets each State determine how best to assign delegates. In the case of Montana the state created two Congressional Districts to distribute the delegates. In Texas, the State Democratic Party does not use the Congressional Districts, but instead uses the 31 State Senate Districts. Additionally, New Jersey pairs off each State House/Senate District to create “Delegate Districts” to distribute delegates, and the District of Columbia splits into two districts with 4 local wards in each district. All of this has been accounted for in the model, with the correct demographics for each region (with the exception of the DC wards, as they have roughly the same demographics as each other, both are given the same demographics in the model.
Q: Why is my favourite candidate not in the model? Why do you hate him/her?
A: The model only includes candidates that consistently poll over 3%. This is not to deliberately exclude certain candidates, but only to lower the run time of the simulations. The more candidates there are, the longer each simulation takes. Additionally, no candidate polling at 1% is going to win the nomination in any simulation. If you really want to see your preferred candidate in the model, go volunteer for them, to help them get over 3% in the polls. Once they consistently get 3% I’ll gladly let them in.
Q: So how does this model work?
A: The model uses State and National polling in order to estimate how many delegates each candidate would win if all of the Primaries occurred today. Once a Primary or Caucus occurs the results will be filled in and the model will become more accurate. This will allow the model to become more accurate once the Primaries begin, but scarifies some accuracy before the Primaries start.
Q: Most of these candidates have never been on the ballot in many of these states before. How does the Model calculate anything?
A: A Primary is much harder to simulate than a General Election, because in the General Election past results can show where Republicans or Democrats might perform better given new polling data. Also past primary data is unreliable because of changing coalitions. In order to build this model, a ground up method needed to be designed. The model starts will National Polls, averages Demographic data between them, and then calculates results for each State/Congressional District based on its Demographics (as obtained from the Census Bureau). Additionally, as new State polls are published, the model is able to adjust the results for that State and it’s Congressional Districts to better reflect those State Polls in accordance to the Demographic data from the National Polls. When there is no State Poll available, the model just uses National Polling Data and Demographics.
Q: Why is Amy Klobuchar winning Minnesota when there is no public poll there yet? How do you know what percentage of the vote she will receive there?
A: When no poll is available for a state, the model uses National Polls to calculate how each state/Congressional District will vote given it’s demographics. Additionally, incumbent advantages are given to home state candidates if there is no poll there. For example, Beto O’Rourke in Texas has been given the standard 15% incumbent advantage. Amy Klobuchar has been given a larger advantage because she does amazingly well in primaries in her home state. Once actual polling from those states is available, the incumbent advantage vanishes as the polling will account of that.
Q: How are you able to calculate who is winning in each County? Does it matter?
A: The County map is calculated the same way as each State or Congressional District. However, while each State and Congressional District are simulated thousands of times, the County map is calculated by using the average vote percentage for each Candidate in the model and then applying that vote through National polls, State Polls, and demographics for each County. This is mostly to cut down on the run time of the model, but also because it doesn’t matter how often a particular Candidate wins each County, as Counties do not give out Delegates.
Q: Why is Biden just winning almost every county? That seems wrong.
A: Given his large lead in the polls, the model predicts that Biden will pull out narrow to large wins in most States/Congressional Districts/Counties. Remember that there is a large amount of undecideds, and that several lower tier candidates are not in the model. Those votes are being redistributed by the model to the all of the candidates in the model, giving Biden a slightly larger lead than polls suggest. Once/if the race becomes closer, many areas will move toward the other candidates. Additionally, since the primary has barely begun, clear demographic trends have yet to emerge. Once stronger demographic trends emerge the interactive maps will begin to show it.
Q: OK, so you calculated a lot of demographics for each State/Congressional District/ County. However, this is a Democratic Primary. The Democratic Party does not look like the country as a whole. Did you account for that?
A: Yes. I used Pew research data, specifically their data from all interviews with callers from 2015. Their data shows demographic breakdowns for everybody that identified as a Democrat on the State level. That data was then compared to the actual State data that I calculated, and through proportions, I was able to calculate (roughly) the demographic breakdown of each Congressional District and County in each state using those proportions. When compared to 2016 Primary exit polls, that data matched up very well to the exit polls.