RCP Primary Calculator -- Data and Assumptions

X
Story Stream
recent articles

This is the technical manual for RCP’s new primary delegate estimator. I didn’t write this piece with a general audience in mind – I wrote it for academics, political junkies, data scientists, traditional journalists, my fellow data journalists, or anyone with an analytical bent who might want to drill down and see where I got my data and assumptions.

Data sources                                                                            

Rules for delegate allocation are far from standardized across the states. While the Republican National Committee has provided some overall rules (e.g. states aren’t allowed to be winner-take-all before a certain date), state parties have significant latitude in how they allocate their delegates.

I used three sources to get a good understanding of what the rules are in each state. I started with a summary of rules published by the RNC (it’s the same information that Time magazine’s Zeke Miller used here). That document gives the most important information – if a state is winner-take-all, if it’s proportional, what the thresholds are, when the contest will happen, etc.

But that fails to deal with other key issues such as rounding and the “backdoor winner-take-all” states (both described below). I used two sources to begin to fill in those details – Frontloading HQ and The Green Papers. My method was fairly simple – I used the information on FHQ and The Green Papers to fill in some gaps left by the RNC summary and then used a few assumptions to deal with the rest of the states.

Assumptions: Rounding

The first issue – rounding – is straightforward and only comes up when a candidate receives delegates proportional to his statewide vote share. If you multiply the percentage of support each candidate gets by the total number of delegates available and round to the nearest whole number, you can come up with situations where too few or too many delegates are counted. For instance, if four candidates received 40 percent, 40 percent, 13 percent and 7 percent of the vote in a state that has 27 available delegates and the totals were rounded to the nearest whole number, the candidates would get 11 delegates, 11 delegates, 4 delegates and 2 delegates respectively. That’s a total of 28 when the state has only 27 to give. So states developed different ways to deal with rounding that might cause delegates to be over- or under-allocated. Here are the three typical rules I found:

The first one gives or takes delegates to/from the candidates who are closest to the rounding threshold. So each candidate’s initial delegate allocation is his share of the vote multiplied by the total number of delegates available rounded to the nearest whole number. If there are too few delegates allocated then extra delegates are given to candidates whose delegate count was rounded down but were closest to the threshold (which is 0.5). So, if there is an unallocated delegate and three candidates had their totals rounded down from 8.3, 4.2 and 2.1 to 8, 4 and 2 respectively, then the 8.3 would get rounded up to nine. The system is reversed for over-allocated delegates – the candidate who had his delegate count rounded up but was closest to the rounding threshold loses a delegate.

The second rule is simpler. Every candidate has his delegate total rounded down, and one delegate is added to the first-place finisher, then one to the second-place finisher, etc. until all the delegates are allocated.

The third rule is maybe the most intuitive – each candidate’s delegate count is rounded to the nearest whole number. If too few delegates are allocated, the statewide winner gets an extra delegate. If there’s another unallocated delegate, the second-place finisher gets it, and then the third-place finisher, etc.

FHQ and The Green Papers provided information on which states use which rules, but as of now those sites don’t have all the information for every state. For states where we didn’t have those details, we defaulted to the third rule. It’s a somewhat arbitrary choice, but the third rule is simple, intuitive and fairly common.

Assumptions: Backdoor winner-take-all and when no candidate reaches threshold

The second issue is the “backdoor winner-take-all.” This phrase, coined by political scientist Josh Putnam, refers to a scenario in which only one candidate makes it over the threshold in a proportional state, thus winning all of the state’s delegates despite the fact that it’s proportional. It’s also possible that no candidate makes it over the threshold – and we need rules to deal with both of those cases.

If I didn’t find explicit rules on the winner-take-all backdoor, I assumed the state used the same rule that Texas uses. In Texas, if only one candidate achieves the threshold, then the delegates are allocated proportionally between the top two finishers. There are other ways to allocate delegates in this situation (e.g. allocate delegates proportionally without a floor when one or fewer candidates make it over the threshold), so the choice is somewhat arbitrary. But a number of states have eliminated the backdoor winner-take-all, so it’s good to assume that the backdoor is gone when no other information is given.

If no candidates reach the threshold in a proportional state, I assumed the delegates were apportioned proportionally as if there was no threshold. Again, this isn’t perfectly accurate but it’s a decent assumption that shouldn’t do much damage to the overall accuracy of the calculator.

Assumptions: Allocating congressional district and RNC delegates

The user will likely also notice that sometimes the delegate column auto-fills and sometimes it doesn’t. The explanation here is pretty simple – sometimes at-large and congressional delegates are allocated proportionally based on the statewide vote (rather than having at-large delegates allocated by statewide vote and congressional delegates allocated by votes in each district). In those cases, it makes more sense to take the math out of the user’s hands and deal with rounding and other issues within the code. If the congressional district delegates were allocated based on their district’s vote, then they show up in that column and the user can adjust those allocations in accordance with their best judgment. And if the congressional district delegates are all allocated to the statewide winner, they show up in the “district delegates” column, but there is no need to adjust that total.

In some states, RNC delegates are bound to the statewide winner (or bound through some rule related to the statewide vote rather than simply being unbound). I adjusted the data and code to reflect as many of those rules as I could find, but the rest are allocated in a “state” called RNC after all the normal states and territories have been passed through.

What to do about states that don’t quite fit

Finally there are some states that don’t quite fit into the molds of proportional or winner-take-all. For instance, in West Virginia, delegates are elected on the primary ballot and all specify who they intend to vote for at the convention. And in North Dakota, there is no preference poll and delegates remain unbound. I treat states and territories like these as proportional with no floor – that way users can allocate those delegates in an intuitive way without worrying too much about the minutiae of the rules in each state.

Future updates

As has been previously noted, there are a number of assumptions built into this calculator. Getting all the rules in one place at one time is a time-consuming and labor-intensive process. So I intend to update the calculator periodically as I obtain more information about the specific rounding and threshold rules in each state and territory.

David Byler is an elections analyst for RealClearPolitics. He can be reached at dbyler@realclearpolitics.com. Follow him on Twitter @davidbylerRCP.

Comment
Show commentsHide Comments
you might also like
This is the technical manual for RCP’s new primary delegate estimator. I didn’t write this piece with a general audience in...

Related Articles