Ballot Order Effect for Minor Party Candidates
My interest in the Prohibition Party is well-documented and it actually started a few years ago because of a weird outlier I noticed while going through 2016 election data. 3rd parties had a great year between Libertarian Gary Johnson and Green Jill Stein but in Arkansas county, Arkansas, someone else came in third place. It was a random Independent I had never heard of called James Hedges, who turned out to be the Prohibition Party candidate. And Arkansas county wasn’t the only one with weird results for minor parties which seemed to vary wildly from county to county.
That’s actually because Arkansas has an obscure election rule outlined in this news clip where each county has a random ballot order that is decided by chance with campaigns drawing lots. There has been a lot of different research on ballot order but they all seem to focus on the choice in low info elections or placement in primaries, or even propositions. There has been some for primetime general elections but even those focus on the major party effects, finding an impact though it was muted. All of these are fascinating and probably of more help to the major players, which is why they’ve received attention.
As an obsessive into the most niche and useless of issues, I wanted to instead look at the impact of ballot order on the most minor of candidates, those running for third parties and Independents, with no resources and only a ballot line (or the same name as a famous singer). This analysis chooses to focus on Arkansas because while other states like Texas have similar laws, Arkansas has relaxed enough standards to allow more candidates to make the ballot and in 2020, there were a total of 13 people on the ballot. What I’ve found is likely to be of more help to those parties when deciding where to do outreach but I’ve also found several instances where the behavior of the protest (or almost protest) voter is revealed and how people make their decision when they step into the voting booth. Different counties have different ways to decide the order, with games of random chance like getting campaign representatives to draw a ping pong ball with a number on it. Given this random and repeated nature, could this have an impact on how many votes someone receives by sheer coincidence and voting behavior. The more I dug in the more I found that at the margins of protest voting, it is not just how voters think that impacts the results, but the technology that voters are presented with.
Process for Finding the Order and The Lucky Draw:
Finding the actual ballot order was way more difficult than I thought it would be. Some states just have it online but Arkansas decided to make everything way more difficult. I was able to find some screenshots in local news reports but for a lot of them I had to use VoterView. With this site, have to put in a first name, a last name, and a date of birth for a registered voter to look up their absentee status and at the bottom of most pages, a pdf with their sample ballot. To find a registered voter, I would google a generic name in google with the county name and then click on one of those sites where you can see someone’s basic information but need to pay for more. I would then plug in those names into the VoterView. For other counties, I called up the county clerks and they were very helpful in tracking down and reading out or emailing me 2020 sample ballots from their county. After this most recent update, I managed to find the ballot order for all 13 candidates in 75 of the 76 total counties.
For the sake of helping others also look at possible effects, I have published a Google Sheet with the order and the performance for every candidate right here:
https://docs.google.com/spreadsheets/d/1CLZguyNcukuMUiJaO3bC6RBYmbLMau6I0kC1QiIfqBg/edit?usp=sharing.
I’ve visualized below how many times a candidate was on the ballot in a certain order for all the counties I had data for and they are ordered by their statewide performance starting with Jorgenson at the top.
These really do seem random and the distributions are not alike, though a lot of people near the top had more time in the #10 spot which we will get to later. There was also no correlation or pattern between the average ballot order spot and performance. If I wanted to look for an effect, I would have to dig deeper.
Page Effect Factor:
The first plot was a simple one, taking the order for each candidate and then making a box and whisker plot for their raw share of the vote. This chart and all others are available and interactive at this link, and you can use them to filter by candidate if you wish:
https://public.tableau.com/profile/noah7751#!/vizhome/ArkansasBallotOrder/BallotOrder
This chart is immediately illuminating. My initial thoughts going into this would be that third party candidates would do better when they started in first place and then gradually decline as they went further down the list. This is not the only trend that is happening but there are instead a few things that other factors that stand out. First place does seem to give a boost but it’s not as large as I would have anticipated, even with a surge over 2nd place. Additionally, while there is downward trend after 1st place, it is a smooth decrease and there is a large surge starting at the end beginning with 10th place. There is also a huge overperformance when the candidate is in last place on the ballot, maybe from cycling through everyone and then just marking the final option.
Along with the jump in the last four options, there are some other curious outliers. 5th place is higher than 4 and 6 and overall, 6 is the nadir of third parties. While it took effort to track down and set up, it appears that the real difference is that Arkansas also uses different voting machines by counties, as shown in the map below using data collected from the Arkansas Secretary of State: https://www.sos.arkansas.gov/elections/voter-information/arkansass-voting-machines/arkansass-voting-machines-ivotronic
The different methods for in person voting machine were:
Retain Existing ELECTronic DRE — This means they used very old and outdated equipment, probably an optical scanner but I couldn’t track it down and they’re probably all different.
Solution A — Model 100 Optical Scan Precinct — The voter fills out a paper ballot, listing all the presidential candidates in one column, and then inserts it into the machine that will read that ballot.
Solution B — iVotronic Touch Screen — This method is an electronic voting machine where the candidates are displayed on a touch screen that the person selects from. In sample online ballots I found from counties using the iVotronic, they were presented on four separate pages, which I will talk about later. An example is shown below.
Solution C — Model 650 Central Optical Scan Tabulator — While this was also named as a scanner, online sample ballots were presented with all the candidates on two pages and news reports from these counties showed people voting online, not only with paper ballots.
Below is a chart showing the same ballot order effect as the previous plot but now it’s broken up by voting machine. The side is no longer the raw total vote but rather the relative share a third-party candidate received when they were in that line on the ballot against how well they did statewide so we can adjust for overall candidate performance.
Solution A, the optical scanner where people used a paper ballot, actually followed my initial hypothesis. When presented with every option on the same page third parties did better when they were in the top 2 and saw an increase from being last place but the differences were not as severe as the other two. Solution B and C saw wilder swings with higher first place results and larger advantages for being near last. The next two sections will dive into the nuances of the options displayed on the electronic voting machines but here it’s clear to see that the voting machines cause different results than only using paper ballots.
The Two Page Effect:
When counties with Solution C had voting machines, there were two pages. There’s also a boost in the 10th spot so there may actually be some with four pages but I could not find evidence of that. The electronic display shown in the sample ballots was arranged below with 6 candidates on the first page and the remaining seven on the second one.
Since people read left to right, you will see an improvement for third parties in the 3rd and 5th spot as well. Since we’ve already seen the results overall in the plot below, I wanted to check out a different effect: Where Biden and Trump were. My hypothesis was that people were selecting a minor candidate when on a page without seeing Trump or Biden. This would be compounded when the major party candidates were on the first page and then the voter toggled to the second page and didn’t want to click the back button. The plot below shows the median relative performance for a third party by the ballot line divided up by Trump and Biden placement. The orange line is when Trump and Biden were both on the first page, purple is when one is on each, and the blue line is when both major party candidates are on the second page.
The results here are stark. When both major party candidates are on the first page, third parties in the last few slots get twice as much or more of the vote in that county as they do statewide. This same large advantage for the last few ballot lines does not appear when Trump or Biden are also on the page. When both major party candidates are on the second page, there is a real advantage to being in the first slot that the other configurations don’t have.
Additionally, there is another pattern that seems to show up here. It’s a bit difficult to see at first but there does at least appear to be a bias towards odd-numbered positions. Those are on the left side of the screen and what people read first. The chart below confirms this pattern but with a bit of a twist.
When a major candidate is on the same page as the 3rd party candidate, in orange, there is no real difference between the left and right side position. But when its a page without a major party candidate and all minor party candidates there, there is a 40% difference where candidates on the left did a lot better.
Page and position effect is already strong when a voter would have to hit the forward or back button one time but what happens when the candidates are listed on four different pages?
The Four Page Effect:
This is the scenario presented in the most counties, where people would vote using Solution B — the iVotronic Touch Screen. The display below shows that this lists three candidates at a time, except for the last page which had four. If you wanted to scroll through all of the pages, you would have to hit the forward button three times, and once on the last page, would hit the back button three times.
If it’s anything like the last one, we would expect that people would not want to toggle back to the beginning after going to the last page. The first plot below shows the median candidate performance for every possible combination placement of Trump and Biden. For example, the orange line is 1st & 1st, which means both major party candidates are on the first page out of the four.
This chart is absolutely nuts. In the instances that both major party candidates are presented on the first page, minor party candidates on the last [page received a whopping 700% increase in their vote compared to statewide. There were also last place bonuses when Trump and Biden were on the 1st and 2nd or 2nd and 3rd pages. A cleaned up chart below plots the performance based on if there is a major party candidate on the last page (yellow) against whether nobody major is on the last page (blue).
The differences here are exactly what you would expect. When Trump or Biden or both are on the last page, there is not much of an overperformance for being on the last page, especially compared to when neither of them are present. However, if neither major party candidate is on the last page, candidates in positions 10 through 13 saw a very large boost, as opposed to nothing for when they were. There is, however, a larger first page bonus from these. People’s votes were heavily influence by how easy it was to find a major party candidate and more chose to throw their vote away in a safe state rather than hit a button three time and navigate an interface with obstacles.
When a major party candidate is on the first page, there is no bonus for being in first place, as opposed to a 30% bonus. Additionally, when neither Trump nor Biden or on the first page, we see a consistent decline as the ballot order goes down to 4th. Being on the last page now makes no difference.
Lastly, I looked at position effect. We saw on the 2 pager that left and right made a difference and when it’s four pages, it seems that the bonus effect is when you’re on the top of a page. If a major candidate is on your page, there is no real effect on position. But when you only have other third partes, we see a significant difference not found for other positions.
The Candidates
The cup runneth over for choices available to Arkansas voters last cycle. Besides Biden and Trump, there were a lot of other candidates and for the most part, they did about the same overall, with the exception of the Libertarian candidate. The screenshot below from uselectionatlas.org has the final results.
Below are the options in order of performance along with a graphic showing where they did better, the order on the ballot, and the relative performance in that county based on where on the ballot they were located. The darker the color on the map on the left, the higher up on the ballot the candidate was located.
Jorgenson/Cohen: The Libertarian ticket this cycle was more left-leaning then in the past, with an emphasis on BLM messaging and anti-police. Jorgenson was the 1988 VP candidate for the first time the party was on all 50 states. Her VP this time was originally on the ticket at the convention with meme candidate Vermin Supreme and had an outrageous platform (Personal note: I disagree with the linked article) that was moderated for the general.
You’ll notice that ballot order doesn’t have as much of an effect here. As an established named party, Jorgenson did better in the suburban center and Northwest corner where Republicans have been moving towards Democrats but stopping over for third parties in the last few cycles. She faltered much more down in the rural and African American heavy areas which don’t usually vote third party. This is interesting in a wider context of protest voting but isn’t as instructive to see a ballot order effect.
West/Tidball: On the ballot in a few states largely due to ratfucking efforts by GOP consultants, the famous rapper is a mishmash of ideas and held only one unhinged event during the cycle. Following a pro-life turn, his vice-presidential pick was biblical life coach Michelle Tidball.
Though he was a better known name which allowed him to place 4th overall, a lot of his strength came from culturally conservative Black voters in the East but he also benefitted when he was listed last on the ballot or in 10th place.
Hawkins/Walker: The Green Party wasn’t on every state, largely due to the pandemic, but it remains America’s 4th most serious party. After a heavy-handed primary that saw several splinters near the end, Hawkins chose Walker and, with neither one of them are officially GP members, steered the party in an explicitly Eco socialist direction.
Though the Green party has a history in Arkansas of occasionally being the only opposition party in the past, Hawkins’ had a disappointing performance. Here his performance very closely follows a lot of the same effects we’ve seen: A boost from 1st, 7th, 10th, and 13th and very poor performance when it’s not in a prime position.
Collins/Joe Parker: No, this is not the singer who singlehandedly made Tarzan a memorable film for everyone my age. Phil Collins is the name of the Prohibition Party nominee, and his personal belief system is that on the religious right, a shift from last cycles liberal views held by Hedges. Collins was a town council member in Libertyville, IL and ran for Mayor of Las Vegas as a Republican recently. Billy Joe Parker is a Georgia man who hates both alcohol and socialism referring to them as “human captivity” and this pairing was actually the third iteration of the Prohibition ticket after the last two in line to be nominees dropped out, which will matter later.
The famous name effect is super strong here, with a giant boost in the 10th spot and a massive drop to 11th, plus a strong first and last place effect. If I ever run for office, I’m also changing my name to someone famous from the boomer’s era, maybe like Jimmy Stewart, Independent.
Pierce/Ballard: Brock Pierce was a child actor from The Mighty Ducks who struck it rich in cryptocurrency and is now an entrepreneur. A lot of times you get these random people with a lot of cash and advisors willing to take that cash and they get on a few ballots or see some “polls” and never end up running. Pierce was on the ballot in the easier dozen plus and a write-in in about a dozen more, though he did do a lot of media hits. Ballard, a direct descendent of Aaron Burr, is another tech entrepreneur and made BLM a focus of her campaign.
Pierce was actually on the ballot last in a few counties and they really stand out as places he did better. It must be the ballot order because his platform was scattered but his overperformance is so scattered across the state in exactly those locations.
Blankenship/Mohr: In a minor party field filled with easily meme-able candidates, Blankenship provided some stiff competition. The ex-coal company CEO allegedly responsible for a mine explosion who turned his anti-establishment rhetoric into an anti-Mitch McConnell Senate run in West Virginia barely spent any cash and no attention on this race and it’s unclear why he ran in the first place. As an outsider, he caused a lot of state parties to pick other slates and he wasn’t as effective as past nominees in shutting down other right wing minor parties like the American Solidarity Party from choosing other candidates. William Mohr is an ex-Republican who joined the Constitution Party’s Michigan branch called the US Taxpayer’s Party in 2005 and has since run in several cycles for local and state house races. They both have moustaches.
Blankenship is a bit more well known but looking at the y-axis, he really did better in the first slot than almost any other candidate. His conservative creds and Constitution Party ties probably helped him in conservative rural areas but performance in the Southwest and Northeast corners are still strongly correlated with ballot order.
Carroll/Patel: An evangelical party first and foremost, the American Solidarity Party puts stopping abortion first and foremost. They are also anti-COVID measures, but the party struggles with a strong wing that cozies a bit up to misogyny and racism but this cycle saw two “establishment” party members come on board between Brian Carroll and former Chair Amar Patel.
Not much to say here, this is a weird pattern but a big drop in 6th place. It looks like Montgomery county was the huge overperformance where Myers were Carroll was last but other counties where he was last was an early voting African-American heavy county like Pulaski so the numbers appear more muddled.
Gammon/Collins: C L Gammon was the VP of the original Prohibition Party ticket with Bill Bayes. Bayes, after being the 2016 VP nominee, was ousted by the party committee for neo-Confederate views. Gammon, a historical author and collector like several of the members of the party, had health issues so he had to drop out though he was stuck on the ballot in Arkansas. His VP pick, Phil Collins, is talked about above as the new ticket was also admitted.
This just really follows the pattern we see overall with positives in the last four and a bonus on the 1st, 10th, and 13th. A model minor party candidate for my results.
Myers/Lusk: Myers is a frequent Alaska candidate but used to be a member of the Constitution Party and Alaskan Independence Party, having run in 2016 for VP of the former with no luck. He considers himself a Reagan Republican. Having split off, Life and Liberty is a brand-new right-wing party focused on ending abortion but also GMOs and the environment. Tiara Lusk is a pro-marijuana Idahoan and went from the Republican Party in 2018 to the Constitution Party as a delegate up to the 2020 convention where she left and joined this ticket. Arkansas is the only state that they received votes in.
As we get to the really weak overperformers, they start to look the same where the patterns repeat. Here the weakness of the 6th position stands out.
La Riva/Freeman: The Party of Socialism and Liberation is a communist party established in 2004 that has a party organization and has nominated La Riva as their nominee in every cycle it has existed with the exception of 2012, though she was a stand-in in some states. She is a perennial candidate and was the VP nominee in the Worker’s World Party for decades before. Her VP nominee was officially Leonard Peltier, a Native American civil rights activist imprisoned since 1977 for the murder of FBI agents during a shooting. He dropped out to due health reasons so Sunil Freeman is acting as a placeholder. With more ballot access, their vote share stayed the same since 2016 and a slightly larger number of votes.
La Riva got the last page more than any other candidate and it really shows up here good performance correlated to that in the South of the state with the outlier counties being darker on the left and not as big overperformances.
De la Fuente/Richardson: A man singularly bent on running for office as many times in as many states as humanly possible, Roque De La Fuente and now his sons constantly file for offices across the nation at all levels with all different parties. De La Fuente is independently wealthy and uses the attention from running to promote a message of equality and the need to fix ballot access laws.
Conclusion:
This experiment leads to the conclusion that ballot order plays a huge role in the performance of third parties, even when it’s just a protest vote. Additionally, the interaction with voting machine technology can create outsized impacts based on how much effort it takes someone to vote. For fairness’ sake, we should probably be presenting everyone with every option on one screen, so as not to add additional effort to finding a preferred candidate. For third parties in Arkansas moving forward, this is a good guide on where to concentrate limited resources down to the county, boosting name recognition when they see you first or last on the ballot or page. While ballot line order has been studied before, not a lot of attention has focused on how the choices are presented to voters and little focus on minor parties until now and I hope this spurs more states and counties to consider how people approach voting.
