Why does range voting have to be averaged and not tallied?

[Marcia]

I thought I understood it. But I guess I don't.  You score all of the candidates and then you average? Why aren't you adding them?

Thanks

[Jeremy Macaluso]

You can do either one. The average and the sum are only different by a factor of the number of people voting as long as everyone votes for all the candidates. The difference is mostly how you treat write-in candidates, but the system is still called range voting either way.

[Warren D Smith]

The summing and averaging methods are equivalent (up to dividing
by #voters) IF all voters score all candidates. But if some voters
leave some candidates unscored, then the two methods differ
(note: the voters who leave Nixon unscored, do not affect Nixon's average).

It is probably a bad idea to treat non-scores as the same thing as zero scores
(which is what summing does), since that causes a large artificial bias against
lesser-known candidates. They already are victimized by the fact a lot
of human voters tend to downgrade candidates they know little about,
includign giving
them zero. Further victimizing them by telling the voters who TRY TO ADMIT
their ignorance about Nixon by not scoring him: "sorry, I refuse to accept
your ignorance and am going to insist that you really hate Nixon" is too much.

--
Warren D. Smith
http://RangeVoting.org <-- add your endorsement (by clicking
"endorse" as 1st step)

[Ted Stern]

Averaging vs. Summing does cause range voting to fail some voting criteria, doesn't it?

--
You received this message because you are subscribed to the Google Groups "The Center for Election Science" group.
To unsubscribe from this group and stop receiving emails from it, send an email to electionscience+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Andy Jennings]

If I remember correctly, I have never seen a "quorum rule" besides summing which satisfies "participation".  Participation says a new voter showing up and voting A=8, B=7 should never switch the election from A winning to B winning.

https://groups.google.com/d/msg/electionscience/SpLDc1Z0hzE/v7dSAizLV78J

[Warren D Smith]

yes, Jennings is correct that averaging causes range voting
to fail "participation" (although passes it with summing)
because your ballot can then exert unequal "pulls" on various candidates,
and will "pull harder" when you score a little-scored candidate.

But keep in mind, averaging and summing are exactly the same thing if
all voters score every candidate.

[David Hollander]

On Monday, June 19, 2017 at 3:55:00 PM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

It is probably a bad idea to treat non-scores as the same thing as zero scores
(which is what summing does), since that causes a large artificial bias against
lesser-known candidates.

This is not the case in zero-centered range, ex [-1,+1]. This is only the case in zero-plus range, ex [0, +1].

 

They already are victimized by  the fact a lot
of human voters tend to downgrade candidates they know little about,
includign giving
them zero.  Further victimizing them by telling the voters who TRY TO ADMIT
their ignorance about Nixon by not scoring him: "sorry, I refuse to accept
your ignorance and am going to insist that you really hate Nixon" is too much.

Giving a candidate the score in the middle of the range when summing, such as a 0 in a {-2,-1,0,+1,+2} range, is not equivalent to asserting that they hate the candidate, as voters who hate a candidate will presumably give that candidate a minimum score rather than a middle score. If the ballot instructions establish that all candidates receive a zero-score in the middle of the range and ignorance \ no-preference is assumed by default, then the only thing voters admit to is which candidates they support or oppose.

[Warren D Smith]

yah, I was implicitly assuming nonnegative ranges; as DH said
zero-centered ranges are another matter.

But zero-centered ranges are problematic for other reasons,
namely negative signs confuse too many people too easily.

[David Hollander]

I think including a number line at the top of the ballot that displays a symbol of negative emotion (thumbs down, frowning face) next to negative numbers and a symbol of positive emotion (thumbs up, smiling face) next to positive numbers would be sufficient to clear up confusion. Adding an explict '+' mark next to positive scores to balance the '-' mark next to negative scores may also help. For small number ranges such as range-3 and range-5, it is possible to include a unique emoticon adjacent to each score, as done on the Vote de Valeur website. For larger number ranges greater than 5, it may be sufficient to use two emoticons, and display the negative emoticon to the left of the negative endpoint and the positive emoticon to the right of the positive endpoint.

I hope to include a paper ballot generator and mock election simulator on the range5.org website in the future so that I can get user feedback on this. Vote de Valeur may have already obtained some data on voter satisfaction with their voting method. I suspect the pros outweigh the cons in regards to the zero centered scale. Voter dissatisfaction generated by mapping preferences to a range including negative numbers may only be a short term cost borne of unfamiliarity which is paid for up front during the first time they use the election method. If the only source of voter dissatisfaction is initial unfamiliarity, then there may be no recurring cost to pay in future elections.

[Phil Uhrich]

While I agree that the chances of a dark horse candidate running away with it is small, that isn't the same as impossible. All it takes is one bad experience to sour voters to a new system. For instance, I have mentioned before that switching to a system that is clone independant creates the condition where there is no reason not to run, so every remotely plausible candidate has every incentive to run to get their name out there for future races. This definitely happened here in Minneapolis as soon as we had an open Mayoral contest with IRV. We had something like 23 candidates; this year we already have about 10 vs an incumbent and a month left to file. Even total political junkies like me were frustrated and didn't get a chance to look into all of them. I could easily see a high profile race getting something like 50 candidates once they realize the ramifications of clone independance. I don't see any voter sitting there for 10 min filling out 0's for 50 people.

IMO the ideal solution would be changing the process for qualifying for the ballot. Saying that there will be no more than 10 candidates allowed on the ballot per race and you have a filing deadline, if 10 or less file they will all be on the ballot. If more than 10 do all candidates will be notified that they now have X days (30?) to collect signatures and the top 10 signature getters get on. Signatures collected before the window are not valid; this might make the public less likely to sign for candidates too, and it would require more retail politicking which is always a plus.

I could also see voters getting soured over a quorum throwing an election between two relatively well known candidates.
Pre Quorum:
Sally has an average of 4.5 from 1000 voters
Joe wins with an average of 4.51 from 950 voters

Quorum of 50 switches the winner.

I prefer a version of you "half of rule d" http://rangevoting.org/WhyHalf.html
With just that, a candidate just needs to convince Hitler to run so everyone zero's him and the threshold is 50%.

Either that method with the caveat that if the result is greater than 33%, 33% becomes the threshold.
Or
Calculate the average of the the 2 or 3 candidates that were most voted on and use the score they received to calculate a weighted average on votes cast on them and then make that the threshold.

[David Hollander]

On Tuesday, June 27, 2017 at 9:25:05 AM UTC-5, Phil Uhrich wrote:

I don't see any voter sitting there for 10 min filling out 0's for 50 people.

If paper ballot instructions establish that candidates will automatically receive a default score in the middle of the range if no preference is indicated, then this would be unnecessary. No preference would be assumed by default. Voters would only have to make marks next to candidates they support or oppose and could leave the rest blank.

Pre Quorum:
Sally has an average of 4.5 from 1000 voters
Joe wins with an average of 4.51 from 950 voters

Quorum of 50 switches the winner.

Assuming 1000 total voters, 100% of voters holding an opinion of Sally, and a 0->5 range, then Sally would receive 4.5 and Joe would receive 4.41 if missing preferences resulted in a default score of 2.5 in the center of the range. If the vote was held using a 1->5 range with a middle score of 3, then Sally would receive 4.5 and Joe would receive 4.43. With a zero-centered, default 0, range 5 election counted using a simple sum of scores, Sally receives +1500 and Joe receives +1435. Using a simple sum of scores and a default score in the middle of the range seems like it would not only prevent voter frustration by Sally supporters in the scenario you mention, but also make life simpler for the election counters. Individuals responsible for tabulating the results would no longer have to perform any division operations and could determine the winner using only integer arithmetic.

[Warren D Smith]

On 6/27/17, David Hollander <dhl...@gmail.com> wrote:
> On Tuesday, June 27, 2017 at 9:25:05 AM UTC-5, Phil Uhrich wrote:
>>
>> I don't see any voter sitting there for 10 min filling out 0's for 50
>> people.

--good. With averaging, they will then not affect those candidates
they are too ignorant or
lazy to score. If there really were 50, then I think few voters would know
enough about all 50 to score them all; and if most of them
then due to either laziness or honest confession
of ignorance in fact did NOT score them, then that'd be excellent and
a situation
greatly to be desired.

Re the suggestion of both-signed scores:
as I said negative signs worry me, but
an indication that it can work is provided by Venice,
http://www.rangevoting.org/VenHist.html
which employed {-1, 0, +1} range voting for centuries,
and they did not have frowny face icons to help them, either.

[David Hollander]

A voter score in a [0,1] range with no default value could perhaps be intuitively modelled as a random sample of the fractional probability that a candidate will increase social welfare once taking office. Increasing the number of voters increases our confidence in the final estimate obtained. However, asking voters to self-assess the degree of confidence they have in their individually generated score, and exclude themselves from contributing to that candidate's average if they have an arbitrarily low level of confidence in it, will not necessarily increase our confidence in the accuracy of the final estimate. This is because voters who self-assess a low level of confidence in their individual estimate of a candidate's probability of increasing social welfare may actually be more knowledgable and taking more variables into consideration than voters who self-assess a high level of confidence in their individual estimate of a candidate's probability of increasing social welfare. That is, voters who are biased to exclude themselves from the average (if given the opportunity) may actually be more knowledgable than voters who are biased to include themselves in the average. If there was instead a default score which voters must conciously decide to marginally adjust from, scores may be more likely to encode judgements of both confidence and probability rather than only probability. If confidence is already a criteria factored into decision making when encoding scores, then I suspect it is unnecessary to add a second channel of information to account for it.

[David Hollander]

Perhaps a more productive way to analyze this is as follows:

Consider a range 2+X ballot where voters indicate whether an option is good, bad, or unknown.

Let a certain good (CG) be an option which all voters indicate is good, no voters indicate is bad, and no voters indicates is unknown.
Let an uncertain good (UG) be an option which 1/2 voters indicate is good, no voters indicate is bad, and 1/2 voters indicate is unknown.
Let a certain bad (CB) be an option which no voters indicate is good, all voters indicate is bad, and no voters indicate is unknown.
Let an uncertain bad (UB) be an option which 1/2 voters indicate is bad, no voters indicate is good, and 1/2 voters indicate is unknown.
Let a certain neutral (CN) be an option which 1/2 voters indicate is bad, 1/2 voters indicate is good, and no voters indicate is unknown.
Let an uncertain neutral (UN) be an option which 1/3 voters indicate is bad, 1/3 voters indicate is good, and 1/3 voters indicate is unknown.
Let a pure unknown (U) be an option which no voters indicate is bad, no voters indicate is good, and all voters indicate is unknown.

The 'highest average' method produces the following group preferences:

CB = UB < CN = UN < CG = UG
U = CG, U = UG, U = CB, U = UB, U = CN, U = UN

The 'highest sum, default bad' method produces the following group preferences:

U = CB = UB < UN < CN = UG < CG

The 'highest sum, default neutral' method produces the following group preferences:

CB < UB < UN = CN = U < UG < CG

The 'highest sum, default neutral' method seems to produce a stricter ordering of preferences. Increasing the chances of unknown options winning relative to bad options but decreasing the chances of unknown options winning relative to good options might affect group stability in iterative games.

[William Waugh]

I recommend that voters lodge full opposition to any candidates they don't bother to research. The ballot forms should offer an "all others" entry to be given the minimum score, to make it easy to take my advice.

[Warren D Smith]

--I recommend voters who know nothing about candidate X
1. say so
2. if they do, then those voters do not affect that candidate

--I do not see why it is that you think Adolf Hitler should,
as a matter of principle, be given an artificial huge advantage
versus comparatively non-famous candidates like Ralph Nader.

--Also, since I have not bothered to research "William Waugh"
and nor have most other people, I take it you are fine
with the idea that we should automatically
assume you are the worst human being on the planet
(and certainly a worse person than anybody who
has a lot of money to spend on ultra-annoying TV ads),
until such time as we "research" you "enough"?

[Phil Uhrich]

Even if you have the nil votes count as the median score if your brand new voting system comes on line and attracts 50 people to run in a race you could elect Jesus and still have a city full of people who disliked the process.

[David Hollander]

My proposal was to use a default score in the center of the range of possible preferences, rather than a 'median' score in center of distribution of actual preferences, so that votes could be counted in a single pass.

While I believe this would do a great job at minimizing the 'marginal cost' to voters of adding 'one more' candidate to the ballot, I believe that you are correct in assuming that if the marginal cost to voters of adding an additional candidate to the ballot exceeds 0, then at some arbitrarily high number of candidates voter dissatisfaction will exist.

Your proposal 'N number of candidates with greatest number of signatures in X days' (ex N=10, X=30) seems like it would work on the surface. However the hard-cap of N is problematic. The reason is because it may encourage political parties to start developing strategies to bump the N+1 candidate from the ballot.

For instance, suppose we had a race where only two candidates initially wanted to run, Bob and Alice. If Bob thinks he can gather more signatures than Alice, then he might get X-1 of his friends to also run, and have petitioners gather signatures for his friends at the same time they gather signatures for him. If his petitioners gather more signatures than Alice's petitioners, then Alice might get bumped from the ballot entirely, despite her potentially being a major party opponent to Bob. So the existence of a hard cap on the number of candidates could actually create preverse incentives which encourages more people to run than otherwise would. A small window on number of days may also be problematic if it gives the official in charge of validating signatures the potential to 'loose' or 'misplace' the signature roll submitted by members of an opposition party until after the deadline as elapsed.

I believe an alternate ballot access rule currently used in many states is 'all candidates with signatures >= K% of ballots cast during the previous election for office sought' (ex K=2%). While there is not a hard-cap, ballot access is still constrained by a monotonic difficulty function in order to deter spam. The signature requirement raises after elections with high turnout, and lowers after elections with low turnout.

[Phil Uhrich]

"For instance, suppose we had a race where only two candidates initially wanted to run."
No problem, they both automatically get access. Only once you have over 10 people file an intent to run (probably with a filing fee) does anyone need to get signatures.

No repeat signatures. Anyone found to have signed for two candidates gets their name stricken from both.

I would agree with you if there was going to only be 2 or 3 slots, but any party that had the organization to get 10 candidates with more signatures than any other candidate was going to win that election anyways. The benefit to doing it this way would be that candidates for less sexy positions (like state auditor) would have a much easier time getting on the ballot and no matter how much hype a race got (like GA-6 special election) you still only have 10 candidates. It incentivises the community to get involved too because there is always the possibility that the other candidates could out sign you. Also, I imagine that candidates would want to demonstrate their viability by posting high numbers here. It would be an avenue for less well known candidates to demonstrate to media organizations that they have an enthusiastic base of support and are worthy of more coverage. There will always be trouble with who is in charge of validating signatures, but that is why we have courts.

[David Hollander]

There are a very large number of problems with your proposal to prevent repeat signatures.

It would create a much higher administrative overhead for petitioners, as they would have to take extra steps to ensure the signer had not already signed another petition, or would not sign another petition later. An independent candidate could believe they had collected sufficient signatures, only to find out that some of their petitioners were fraudulently told by other petitioners that they could sign multiple petitions. At the last minute they could become completely disenfranchised by the sitting party in power, when a new roll of signatures 'turns up' that happens to have a large number of names from their original petitions on it, causing a large number of their signatures to be invalidated at the last minute before they would be allowed to collect new ones.

It would create a very large administrative overhead for the election authority, as they would be essentially be forced to run a large scale, single-choice plurality voting election and create a database of names to prevent double voting. This database would have to be transparent in order to ensure that those overseeing elections, who may be highly partisan appointees, are not unfairly excluding ballot access signatures for opposition candidates. The construction of such a database would completely violate the principle of the secret ballot. It would tie voter names to support of specific candidates and parties. If this single-choice signature proposal does not disenfranchise independent candidates during the signature validation process, and allows an independent candidate to make it on a ballot, then the list of voters who signed that candidates petition may be leaked to a partisan political consultant, in order to obtain a list of voters to target for disenfranchisement and intimidation.

When a voter is instead allowed to sign a ballot access petition for multiple candidates, although their signature would not be secret, they still have complete deniability over whether they actually support the views, positions, or policies of the candidate whose ballot access petition they signed. All that they would be indicating in the multi-signature scenario is that they believe the candidate is seeking to participate in the electoral process in good faith, and should be allowed to appear on the ballot. If a candidate is not allowed to appear on the ballot, then they will likely be excluded from public debates. An undecided voter may be interested in three candidates, and wish to see all three participate in a debate against each other, so that all three have a better opportunity to explain their views on a complex set of issues of interest to them. Even if the voter currently dislikes all three candidates, they may believe that all should be allowed to participate in debates and appear on the ballot if they believe that it will generate better public discussion. With a single-choice signature system, we are reintroducing all of the problems associated with plurality voting that prevent such debates from occuring.

Courts and litigation are not an alternative to having an objective and fair list of rules likely to head off the problem in the first place. If an election authority decides to drop a candidate from the ballot right before an election, they may not be able to go through the legal process to get on the ballot in time, and the uncertainty over whether they will be allowed to appear on the ballot could be considered a form of voter intimidation.

If a ballot signature gathering procedure exists, then repeat signatures should be allowed, there should be multiple filing periods or a rolling submission system for signature validations, and campaigns should be allowed to know the precise number of signatures they must obtain ahead of time so that they can allocate resources efficiently.

[Phil Uhrich]

All of those considerations are already in place in most states for current signature validation procedures. It would require new legislation to implement voting reform and a streamlining and adjustment to best practices could be used. There should be ample time for signature verification and legal challenges before the election. Simply informing perspective signers that they must be a registered voter and are only allowed to sign a petition for one candidate per office would eliminate much of the confusion. If privacy about political affiliation is a concern mandate that signature sheets remain confidential after verifying that the signator is a registered voter. Any increases in overhead would pale in comparison to the cost of running a primary.

https://ballotpedia.org/Valid_signature

[David Hollander]

Preventing multiple votes for one candidate per voter is reasonable and already in place. Preventing one vote for multiple candidates per voter is unreasonable and not already in place. My issue is solely with the later and not the former. The later is a move away from approval voting and towards single choice plurality voting. If voters are only allowed to sign the petition for one candidate, then undecided voters will be discouraged from signing ballot petitions for unknown candidates which they would like to see participate in debates and hear more from. This is because they will be incentivized to treat their signature as an artificially scarce resource and save it to sign strategically, which increases the cost for lesser known candidates to obtain signatures in comparison to well known candidates.

The current ballot signature process in many states is already very corrupt and heavily biased towards invalidation rather than validation. Adding more opportunities for invalidation will encourage more corruption, not less corruption. Major parties are already granted special priviledges for automatic ballot access which prevent them from having to collect signatures, and signature collection is a very large campaign expense in both dollars and volunteer hours for minor parties and independent candidates. Minor parties and independent candidates have to spend a large proportion have their campaign resources on signature collection, and routinely have large amounts of their signatures invalidated. These campaigns are already regularly excluded from the ballot via collusion by state authorities, and have to come up with the legal funds to fight protracted court battles for multiple months, where they have to file Freedom of Information Act requests in order to get the information necessary to prove their case. Adding a legal mandate is not sufficient to protect privacy if the fox is guarding the hen house and there is no empirically reliable method for detecting and measuring whether the crime has occurred. Adding new criminal penalties would not fix all of the other problems with the single-choice signature proposal.

I don't see any advantages to one candidate per voter rule over multiple candidates per voter. If you wish to limit the number of candidates on the ballot in a general election more heavily to reduce the hypothetical damage caused by the hypothetical 50 candidate scenario, one can simply increment the variable percentage in the K% rule, or use independent variables other than voter turnout in the previous election, such as the number of candidates on the ballot in the previous election. One could possibly use a weighted combination of multiple independent statistics. The important part is that any proof-of-work difficulty function to prevent ballot spam should apply equally to all candidates and parties, that it should not impose hard caps, that the precise level of difficulty for obtaining ballot access should be known by the start of campaign season, and that the level of work required to obtain ballot access should not fluctuate or alter once campaigning has commenced.

I don't see a direct connection between this public polling for parties which hold primaries. The last point seems like a false dillema.

[Phil Uhrich]

Fine tell each voter they can sign for no more than 5 candidates per office to prevent a single party gaming the process.

It really shouldn't be any different validating vs invalidating, and even if it was and it cost the state a bunch of money to set up a new procedure to validate signatures it would still be much less expensive than the cost of having party primaries which the state currently pays for.

There are already wolves guarding the hen house and you can buy voter rolls and voting history for most states.
http://www.sos.state.mn.us/election-administration-campaigns/data-maps/registered-voter-list-requests/
https://www.elections.ny.gov/NYSBOE/FOIL/2016SubjectMatterList.pdf
Strict privacy laws should work fine.

Requiring a fixed amount of signatures for every person for every race will end up with no one running for dog catcher in middle of nowhere's ville. A signature requirement only for highly contested offices would work much better.

[David Hollander]

On Thursday, July 6, 2017 at 7:23:02 PM UTC-5, Phil Uhrich wrote:

Fine tell each voter they can sign for no more than 5 candidates per office to prevent a single party gaming the process.

Why specifically do you think allowing voters to sign petitions for multiple candidates will allow a single party to game the process? The way parties currently game the process is by making it harder for independent candidates to reach signature quotas, which is what your proposal also does.
 

It really shouldn't be any different validating vs invalidating

If a state official can stall ballot access for third party and independent candidates, or setup a procedure for validating signatures which is likely to invalidate them rather than validate them, then they can keep third party and independent candidates off the ballot. If they can keep opposition candidates off the ballot, their party is more likely to stay in power, and they get rewarded by their party for loyalty and helping to win the election. With the single-candidate per signer rules, there are more opportunities for state officials to invalidate signatures than with multi-candidate per signer rules.  With both rules they can invalidate signatures by showing that a signer double signed a petion for one candidate or is not a state resident. However with single-candidate per signer rules, they can also invalidate signatures by claiming that a signer signed for multiple candidates. This would create additional opportunities for election fraud where one candidates signatures previously determined to be valid can beinvalidated without even having to modifying their list of signatures, but by inserting a new list of signatures for a different candidate at a later date, that puts some of the signers for the orignal candidate over their individual limit. This allows anyone overseeing the signature validation process to retroactively subtract signatures previously determined to be valid from a candidates total if they wish for them to be dropped from the ballot.
 

There are already wolves guarding the hen house and you can buy voter rolls and voting history for most states.

It's unnecessary for states to determine individual party affiliations if they only provide open primary services. Additionally, voters are free to vote for independent candidates and parties which do not utilize public primaries, and records about party registration for primaries do not necessarily indicate who a voter supported on election day, so there is still deniability. With multi-candidate signatures for independent candidates, voters still have very strong deniability as to whether they actually will support that candidate in election day. With single-candidate signatures for independent candidates who do not participate in primaries, voters have substantially less deniability. Also, the fact that many states are needlessly doing a bad job does not mean that we should require to do a bad job by law, especially when the benefits we are purchasing with this cost are questionable and have not been clearly described.

Strict privacy laws should work fine.

No, they won't. If an elected official like the Secretary of State is the one responsible for violating the law, they are not going to get caught. They just ask an office worker to hand off the data to someone in their office during lunch, and a consultant for their political party has all that data. The concern I raised was not the public obtaining the information. The concern I raised was the possibility of the political party whose members hired the people who are responsible for the validating signatures obtaining the information. The only solution I see without secret ballots is making sure that the signature data is a poor statistical representation of who the signer is likely to vote for on election day, which is accomplished by encouraging petition signers to sign many petitions for multiple candidates, which is the opposite of what you are proposing. Additionally, privacy laws do not change the fact that limiting signatures creates a bias against independent and minor party candidates, as discussed previously.

Requiring a fixed amount of signatures for every person for every race will end up with no one running for dog catcher in middle of nowhere's ville.  A signature requirement only for highly contested offices would work much better.

Yes, signature requirements (if they exist) should be per office. A signature requirement for state governor should certainly not be the same as for municipal dogcatcher. Signatures are only needed for when a race is expected to have such an overwhelming number of candidates that it will generate voter dissatisfication, such as in your initial 50+ candidate hypothetical. I would lean towards as low of signature requirements as possible. My position concerning fixed number of signatures was that candidates should have to gather the exact same number as their opponents, and that the number of signatures they need to gather to be guaranteed ballot access should be known at the beginning of the race. When you introduce a 'hard cap' on the number of candidates allowed to appear on the ballot during the general election, candidates will not know until the middle of the race whether or not they will be required to gather signatures, and they will never know the precise threshold of signatures they must gather to be guaranteed access. They won't know whether they got bumped from the ballot until it is too late. So, if you absolutely have to impose a 'cap' on the number of candidates which appear on the ballot for a specific office, what I am advocating for is a 'soft cap' rather than a 'hard cap'. With a 'soft cap' there is only a probablistic limit on the number of candidates which will appear on the ballot for a specific office, based on historical data about past races for the office in question.