Aaron Hamlin: How to Replace Our Broken Voting System

Elected offi­cials con­trol billions of dol­lars in spend­ing and make poli­cies that af­fect mil­lions of peo­ple. But we choose them through plu­ral­ity vot­ing, a sys­tem that en­courages ex­trem­ism and forces many vot­ers to choose be­tween their fa­vorite can­di­dates and those who ac­tu­ally stand a chance.

In this talk, Aaron Ham­lin of the Cen­ter for Elec­tion Science dis­cusses ap­proval vot­ing, a method that works much bet­ter with­out adding com­plex­ity, and how the Cen­ter plans to im­ple­ment it across the coun­try (they’ve already won a land­slide vic­tory in Fargo, North Dakota). Aaron also recom­mends this pod­cast.

Below is a tran­script of Aaron’s talk, which we’ve lightly ed­ited for clar­ity. You can also watch it on YouTube or read it on effec­tivealtru­ism.org.

The Talk

Elec­tions have vast con­se­quences for ev­ery­one, so it’s a real shame that we run them so poorly. Why is that?

Let me tell you a story about when I was in grad­u­ate school. This was in 2008 — dur­ing an elec­tion year. I went out to din­ner with my friends.


We were talk­ing about whom we were go­ing to vote for, and I no­ticed some­thing dis­con­cert­ing: All of my class­mates [planned to vote] for peo­ple whose ide­olo­gies and views didn’t al­ign with their own.

This dis­con­nect was re­ally alarm­ing to me. I walked away from that din­ner up­set.


I kept ask­ing “why” ques­tions: Why was it that my friends were vot­ing against their in­ter­ests? Why is it that we elect ter­rible peo­ple to office? Why is it that bet­ter peo­ple don’t run? Why are cer­tain can­di­dates marginal­ized?

Th­ese ques­tions led me to the vot­ing booth — and, more speci­fi­cally, the bal­lot box.


When you look at the bal­lot it­self within the vot­ing booth, you see di­rec­tions that say, “Vote for one.” They seem in­no­cent enough. But this has a huge im­pact. It’s called plu­ral­ity vot­ing, or first-past-the-post vot­ing. But for the pur­poses of this talk, I’m go­ing to call it choose-one vot­ing.

There are some symp­toms of the prob­lems caused by choose-one vot­ing.


We would like our vot­ing method to elect strong, con­sen­sus-style win­ners. In­stead, we see po­lariz­ing win­ners. We would like elec­tions to be more in­clu­sive, so that more can­di­dates run. In­stead, cer­tain can­di­dates don’t run for fear of not be­ing vi­able. We use proxy mea­sures for vi­a­bil­ity, like whether a cer­tain can­di­date has money or name recog­ni­tion, but those at­tributes aren’t nec­es­sar­ily good pre­dic­tors of whether some­one will do a good job in office. More in­clu­sivity would also mean that when peo­ple bring good ideas to the table, [we could mea­sure and see the amount of sup­port they gen­er­ate]. In­stead, we see ideas marginal­ized.

So why do we see these out­comes? The vot­ing method that we use en­courages us to be­tray our fa­vorite can­di­date.

Imag­ine there’s a can­di­date you like, but you view that can­di­date as un­vi­able. In­stead of choos­ing them, you choose some­one else whom you view as just okay among the fron­trun­ners. You don’t ac­tu­ally get to show sup­port for the can­di­date you like.

Also, since this vot­ing method only al­lows us to choose one can­di­date, if we have feel­ings about other can­di­dates, we don’t get to ex­press them.

Fi­nally there’s an is­sue with vote-split­ting; if there are mul­ti­ple can­di­dates who are similar, we can’t choose all of them, [which could re­sult in] a split vote be­tween them and an­other can­di­date who doesn’t have similar can­di­dates run­ning alongside them. Some peo­ple say, “We’ll just do a runoff,” but that doesn’t solve the is­sue.


We still miss out on that con­sen­sus can­di­date in the cen­ter, be­cause the vote gets di­vided be­tween the left and the right; the can­di­date in the mid­dle doesn’t make it to the runoff. This is true even with rank­ing meth­ods, such as ranked-choice vot­ing or in­stant runoff vot­ing, that at­tempt to simu­late se­quen­tial runoffs. That can­di­date in the mid­dle still gets knocked out.

So what hap­pens if this is un­treated — what kinds of con­se­quences are there when we use this ter­rible vot­ing method to elect peo­ple into office?


Peo­ple in office have a lot of re­spon­si­bil­ity, and one of their re­spon­si­bil­ities is to spend avast amount of money. Think about all of the causes that we would like to see pri­ori­tized with fund­ing. Wor­ld­wide gov­ern­ment spend­ing to­tals over $20 trillion. That’s not an easy num­ber to wrap our heads around, so imag­ine $100 bills in U.S. cur­rency stacked up to the height of skyscrap­ers sur­round­ing the Statue of Liberty.

Peo­ple who are elected do more than just spend money; they also con­trol the poli­cies that gov­ern our day-to-day lives — for ex­am­ple, crim­i­nal jus­tice re­form poli­cies that de­ter­mine whether we treat in­mates civilly and en­sure that they can move back into so­ciety, ver­sus poli­cies that keep them in soli­tary con­fine­ment. Con­sider men­tal health is­sues, de­ci­sions over send­ing peo­ple to war, bio-safety risks, and an­i­mal welfare. Do we al­low fac­tory farm­ing to con­tinue? Do we ad­dress en­vi­ron­men­tal is­sues or al­low CO2 lev­els to rise? Do we ad­dress AI safety or ig­nore it?

It’s im­por­tant to re­al­ize that these spend­ing and policy is­sues hap­pen on a wor­ld­wide scale. This map shows all of the places that use ei­ther the choose-one vot­ing method or a runoff type of sys­tem.


We see the same is­sues. We see (in green) places that use a rank­ing method to simu­late a runoff elec­tion — which, as I men­tioned ear­lier, can lead to a lot of the same is­sues. So as bad as this map looks, with all of red ar­eas us­ing that crappy vot­ing method [choose-one vot­ing], it’s ac­tu­ally worse be­cause rank­ing meth­ods aren’t that good ei­ther.

To be a lit­tle bit more op­ti­mistic: The map we were just look­ing at showed sin­gle-win­ner meth­ods for ex­ec­u­tive offices. Here is a map of places that use differ­ent types of pro­por­tional vot­ing meth­ods.


Pro­por­tional meth­ods are a big step up. They ad­dress a lot of is­sues, such as ger­ry­man­der­ing. But it’s also im­por­tant to get vot­ing for sin­gle-win­ner ex­ec­u­tive offices right, be­cause these are the peo­ple who sign in leg­is­la­tion — and they of­ten tend to be the most pow­er­ful lob­by­ists there are.

So what do we do about all of this? We have this ter­rible vot­ing method — what’s the al­ter­na­tive?


Ap­proval vot­ing is an­other vot­ing method that al­lows you to se­lect as many can­di­dates as you want. This is very easy to do. It works on the dumb­est of vot­ing ma­chines. You can hand-count it.

It also has some re­ally nice fea­tures. Imag­ine there’s a can­di­date you like, but they’re not per­ceived as very vi­able. Un­der ap­proval vot­ing, you can sup­port that can­di­date — and you can also sup­port one of the fron­trun­ners. You don’t have to worry about throw­ing your vote away. If there are mul­ti­ple can­di­dates that you like, you can sup­port mul­ti­ple can­di­dates. If you want to hedge your bets against a can­di­date that you don’t like, you can do that with an­other can­di­date — for ex­am­ple, a more mod­er­ate com­pro­mise can­di­date.

This vot­ing method also tends to elect more con­sen­sus win­ners. It pro­vides a much more ac­cu­rate re­flec­tion of sup­port for third par­ties and in­de­pen­dents. And it en­courages more can­di­dates to run who oth­er­wise wouldn’t, be­cause they don’t have to worry about the vi­a­bil­ity is­sue.

So how does this look in prac­tice? For in­stance, this is a 2007 French study.


We can see that un­der choose-one vot­ing, shown in red, there is a differ­ent win­ner com­pared to ap­proval vot­ing. In this French elec­tion, Sarkozy won un­der the choose-one method and un­der the runoff method. But Bay­rou, a more mod­er­ate, con­sen­sus-style can­di­date, would have won un­der ap­proval vot­ing.

In ad­di­tion to the win­ner chang­ing, we see that third par­ties and in­de­pen­dents do much bet­ter un­der ap­proval vot­ing; they re­ceive from five to 10 times as much sup­port. And it’s im­por­tant to note that this is some­thing we con­sis­tently see. You can’t marginal­ize can­di­dates or ideas in the same way when they have, for ex­am­ple, 20% sup­port ver­sus 1% sup­port.

You can see the same pat­tern in the 2009 Ger­man elec­tion.


The win­ner didn’t — and doesn’t always — change with the vot­ing method. But a vot­ing method has other jobs to do be­sides just se­lect­ing a win­ner. It’s im­por­tant to gauge sup­port for other can­di­dates. For in­stance, we can see can­di­dates re­ceiv­ing just a few per­centage points of sup­port un­der choose-one vot­ing. But un­der ap­proval vot­ing, when their ac­tual sup­port is shown, they re­ceive up to 30% sup­port.

Even if we were us­ing ap­proval vot­ing in this elec­tion and those can­di­dates [who re­ceived 30% sup­port] didn’t win, they would’ve got­ten their ideas heard. They may have even had cer­tain policy is­sues co-opted by the fron­trun­ner can­di­dates, which can be a big win for can­di­dates who just care about the policy is­sues.

It’s also im­por­tant to note that some can­di­dates didn’t change too much from the choose-one method ver­sus ap­proval vot­ing. And that’s be­cause ap­proval vot­ing doesn’t mag­i­cally make you a bet­ter can­di­date. You still have to be a good can­di­date, even un­der ap­proval vot­ing. It’ll cap­ture your sup­port — but your sup­port ac­tu­ally has to be there.

Here’s the 2016 U.S. elec­tion.


We see that the win­ner un­der our choose-one method and ap­proval vot­ing is the same; Clin­ton wins in both in­stances. Some of you may be think­ing, “I thought Trump won.” That’s true, but you have to re­mem­ber that in the U.S. we man­aged to take the worst vot­ing method there is [choose-one vot­ing] and make it even worse by nest­ing it un­der the Elec­toral Col­lege.

Also, look at the third-party can­di­dates. [Ap­proval vot­ing would have shown more sup­port for] John­son from the Liber­tar­ian party and Stein from the Green party. This is a fea­ture that we see time and again un­der ap­proval vot­ing — that more ac­cu­rate re­flec­tion of sup­port for third par­ties and in­de­pen­dents. Stein goes from 1% to 12%, and John­son goes from 3% to 21%. Th­ese are can­di­dates who were com­pletely ex­cluded from ev­ery sin­gle de­bate be­cause of the [seem­ingly] small amount of sup­port that they had. That doesn’t have to be the case with ap­proval vot­ing.

Also, in 2016, Gal­lup con­ducted a poll ask­ing peo­ple if they knew who in the world these [third-party] can­di­dates were. Two-thirds of peo­ple didn’t, be­cause these can­di­dates were get­ting marginal­ized in the me­dia. Imag­ine the pos­i­tive re­in­force­ment loop that would al­low them and other can­di­dates to re­ceive more at­ten­tion [un­der ap­proval vot­ing]. Re­mem­ber: Un­der our cur­rent vot­ing method, this can cause can­di­dates not to run. That was the case in 2016 when Bloomberg de­cided not to run. Surely there were other strong can­di­dates who also de­cided not to run, be­cause they ei­ther didn’t want to split the vote or didn’t per­ceive them­selves as be­ing vi­able.

So how does this much bet­ter al­ter­na­tive [ap­proval vot­ing] work in prac­tice? How do we change our cur­rent sys­tem?

I have an­other story for you. This is a story of Fargo, North Dakota.


North Dakota had an in­ter­est­ing elec­tion in 2015. It was a five-way race in which the win­ner had 22% of the vote. This was, frankly, em­bar­rass­ing for the City Com­mis­sion. So they cre­ated a task force. One of the task force mem­bers was par­tic­u­larly zeal­ous about this is­sue, had done their home­work about ap­proval vot­ing, and reached out to us.

We talked and he rec­og­nized that ap­proval vot­ing was go­ing to be a strong solu­tion for Fargo. He took it back to the task force. The task force got on board, pre­sented it to the com­mis­sion, and recom­mended ap­proval vot­ing for Fargo. And do you know what the com­mis­sion did af­ter the task force had recom­mended ap­proval vot­ing? It ig­nored the task force for an en­tire year! But that did not sit well with this par­tic­u­lar mem­ber of the task force. He gath­ered enough sig­na­tures to put ap­proval vot­ing on the bal­lot it­self, which had never been done be­fore.


We won. So Fargo, North Dakota be­came the first city ever in the U.S. to im­ple­ment ap­proval vot­ing. And we won by a lot: 63.5%. We won in ev­ery sin­gle precinct in the city of Fargo.

How do we repli­cate this?


There are cer­tain things that we look for:

* Lo­cal sup­port from the com­mu­nity. In the case of Fargo, we had a mem­ber from the com­mu­nity reach out to us who was also well-con­nected.
* Sin­gle-win­ner elec­tions. In Fargo, we fo­cused on the may­oral elec­tion.
* Cities with con­trol over their own vot­ing method. That was the case in Fargo be­cause North Dakota is a home rule state.
* Bal­lot ini­ti­a­tives. We have to use them.

You may be won­der­ing why was the Fargo com­mis­sion didn’t listen to the task force. One of the peo­ple on that com­mis­sion was the same per­son who won with 22% of the vote. And he was not very ex­cited to change the vot­ing method that al­lowed him to slide into office.

Right now we have this situ­a­tion where we are us­ing the worst vot­ing method there is to elect peo­ple to office, who then choose how we spend our im­mense amounts of tax­payer dol­lars and which poli­cies gov­ern not just our lives, but the lives of an­i­mals and other sen­tient be­ings who ex­pe­rience suffer­ing.

But this is a solv­able is­sue, as we showed in Fargo (which has a pop­u­la­tion of 120,000 peo­ple).


This is also scal­able, as we’ll show next year in the city of St. Louis, which has a pop­u­la­tion of over 300,000.


And this is repli­ca­ble, as we’ll show in other cities and states in the fu­ture — and not just in the United States, but across the globe.



It’s im­por­tant to re­mem­ber, too, that un­less we have the re­sources to push this for­ward, we will con­tinue to use this awful vot­ing method to elect peo­ple to office who make in­cred­ibly im­por­tant de­ci­sions about the spend­ing and poli­cies that con­trol not just the lives of peo­ple who ex­ist to­day, but the lives of many more in the fu­ture.

I’m not here to say that this is a silver bul­let. But ap­proval vot­ing is one of the most im­pact­ful in­ter­ven­tions that we have the power to achieve. Thank you.

Moder­a­tor: Thanks very much for that, Aaron. What is the biggest push­back you get [when pro­mot­ing ap­proval vot­ing]? Is it just in­er­tia? Sta­tus-quo in­cum­bents?

Aaron: We just cir­cum­vent that by not ask­ing them. We do a bal­lot ini­ti­a­tive in­stead. But in terms of bar­ri­ers, there’s not much push­back from mem­bers of the com­mu­nity. Fund­ing is the main is­sue that holds us back. Bal­lot ini­ti­a­tives are very ex­pen­sive to run.

Moder­a­tor: How can peo­ple act on that? Where would they donate?

Aaron: The Cen­ter for Elec­tion Science. Our web­site is elec­tion­science.org. Also, we com­mu­ni­cate about other part­ners that we work with and provide op­por­tu­ni­ties to donate to our part­ners di­rectly.

Moder­a­tor: Have you seen other ex­am­ples out­side of the U.S. of ap­proval vot­ing work­ing well?

Aaron: Right now we’re fo­cused on the U.S. Fargo is the only city in the en­tire world that has im­ple­mented ap­proval vot­ing, so we are real trailblaz­ers in this area.

Moder­a­tor: Could you ex­pand on why ranked-choice vot­ing is not a good solu­tion for the prob­lems that you’ve laid out?

Aaron: Ap­proval vot­ing is nice be­cause it has very low com­plex­ity costs. It’s very easy to im­ple­ment and it does a great job of elect­ing strong win­ners. It cap­tures sup­port even for peo­ple who don’t run. And it has a low bar­rier to en­try for peo­ple who are look­ing to run.

Ranked-choice vot­ing is much more com­pli­cated. Who­ever has more than half of the first-choice votes wins. If some­one doesn’t re­ceive more than half of the first-ranked votes, you look to the per­son with the fewest first-choice votes. You elimi­nate them and trans­fer the next-prefer­ence votes [of peo­ple who preferred the elimi­nated can­di­date] to the ap­pro­pri­ate can­di­date. Then you add up all the votes again, and if any­one has more than half of the first-choice prefer­ences, you have a win­ner. If not, you re­peat that pro­cess again. It’s com­pli­cated. With ap­proval vot­ing, you just choose as many as you want. And most of the time, you don’t have to worry about retrofitting the vot­ing ma­chines.

Also, with ranked-choice vot­ing, like I men­tioned be­fore, you must con­tend with the cen­ter-squeeze effect. You can have a very clear strong win­ner and not elect that win­ner. There are also sce­nar­ios with ranked-choice vot­ing in which you can hon­estly sup­port your fa­vorite can­di­date and end up with a worse out­come. And for third par­ties and in­de­pen­dents, the ranked-choice vot­ing al­gorithm doesn’t do a good job at show­ing all of the data con­cur­rently; it only con­sid­ers a por­tion of the data at any one point. That can re­sult in a poor pic­ture of the sup­port of third par­ties and in­de­pen­dents. With ap­proval vot­ing, in ad­di­tion to us­ing a sim­ple al­gorithm of just ad­di­tion, you can also see all the data at the same time. That means that you see all the sup­port for third par­ties and in­de­pen­dents.

Moder­a­tor: And is there any sort of rigor­ous study of how hard or easy vot­ers find it to un­der­stand ap­proval vot­ing com­pared to other meth­ods? In­tu­itively it’s very com­pel­ling, but have there been any stud­ies?

Aaron: There haven’t been any stud­ies ex­plic­itly com­par­ing un­der­stand­abil­ity. But in terms of ed­u­ca­tion cam­paigns and part­ners that we’ve spo­ken to, the sim­plic­ity of ap­proval vot­ing has a re­ally big im­pact com­pared to ranked-choice vot­ing.

There are differ­ent ways to mea­sure com­plex­ity. Per­haps one proxy mea­sure would be how long it takes to ex­plain ap­proval vot­ing ver­sus ranked-choice vot­ing. That could go a long way in study­ing [ease of un­der­stand­ing].

Moder­a­tor: What do you think of at­tempt­ing a statewide change of the choose-one method in a state such as Maine? Or do you feel city-level change is the best place to start?

Aaron: I think state-level change is im­por­tant and we will cer­tainly move up to that. Our strat­egy over­all has been to show a proof of con­cept, which we are do­ing in Fargo. Then, we will scale and repli­cate, as we’re do­ing in St. Louis.

The next tar­get will be even larger — a pop­u­la­tion of over half a mil­lion. And then we’ll be able to look at state-level change. We have to cre­ate a track record be­fore reach­ing the state level.

Moder­a­tor: What is your timeline for mak­ing that progress?

Aaron: Again, we’re largely fund­ing-con­strained. If the funds are there, we can move very quickly. At the end of 2017 we re­ceived a grant from the Open Philan­thropy Pro­ject. And within less than a year of re­ceiv­ing that grant, we made Fargo hap­pen. So we’re very quick and we’re very effi­cient with the funds.

Moder­a­tor: Great. And how big is the or­ga­ni­za­tion cur­rently?

Aaron: Our fourth hire will be start­ing soon.

Moder­a­tor: What do you think of the vi­a­bil­ity of the Na­tional Pop­u­lar Vote In­ter­state Com­pact?

Aaron: It’s a way of us­ing an in­ter­state com­pact or an agree­ment be­tween states [award their elec­toral col­lege votes to the can­di­date who wins the] na­tional pop­u­lar vote. I don’t think it’s worth cel­e­brat­ing too much. It’s [an im­prove­ment] — for in­stance, it would have re­sulted in elect­ing Clin­ton in the 2016 elec­tion. So it can be ma­te­rial in terms of the out­come.

But we have to re­mem­ber that even [with the in­ter­state com­pact] we’re still us­ing the worst vot­ing method.

One nice com­po­nent of ap­proval vot­ing is that it can in­ter­act with the in­ter­state com­pact, which means that it has fea­tures like precinct mo­bil­ity, which ranked-choice vot­ing does not have. Also, you can tab­u­late re­sults with dis­cor­dant states that aren’t us­ing ap­proval vot­ing; you can choose data from other states and add it to ap­proval-vot­ing data, but you can’t add rank­ing data and choose-one data at the same time.

Moder­a­tor: Great. Please join me in thank­ing Aaron.

No comments.