It's Time to Ditch RCV, Embrace Approval Voting

It has been tried, and it has failed. Ranked-choice voting must go before it poisons the well for all electoral reform.

Introduction

Last year, I became heavily interested in electoral reform and voting systems. Like many I’ve talked to, my starting assumption is that “anything is better than the current Plurality system we have in the US”. In fact, that is a direct quote from my first blog post about Approval voting, which contains a number of statements and opinions I no longer stand by. However, after speaking to experts who have studied this deeply, after reading papers and reports, I have to admit I was wrong. Very wrong, actually. And I don’t mind admitting that. However, I feel it’s important to state up front that I had the apparently extremely common experience of thinking “Ranked-Choice Voting (RCV) is an improvement, but [insert favorite voting method] would just be better”. But then I started seeing the cracks in the system. The repeal efforts, the preemptive bans, the Condorcet failures, and the idea that RCV would actually improve our situation started to seem less and less plausible. Apparently, I was not alone there, by a long shot.

If you are still of the opinion that RCV is our best chance at replacing our current system, I would like to ask that you strongly consider the case I am about to present with an open mind. Because I was right there with you not too long ago, and I understand the appeal of RCV. But significant evidence changed my mind completely. This is not a matter of mathematics or philosophy. This is a case of pure practicality. In our common goal of successful and long-lasting electoral reform, I have found that of the main systems in conversation, RCV is not the answer.

The State of RCV

In 2024, ranked-choice voting (RCV) explicitly failed to pass in four U.S. states where it was on the ballot: Colorado, Nevada (despite having previously passed in 2022), Idaho, and Oregon. It was preemptively prohibited in Missouri, and a ballot measure in Arizona which could have implemented RCV also failed to pass. RCV only survived in Alaska by some 700 votes (160,230 to repeal, 160,973 to keep), making it a narrow 1:6 for RCV. This problem clearly isn’t isolated to red or blue states. Voters on both sides of the aisle are overwhelmingly rejecting RCV.

Despite over $60 million spent by advocacy groups that year alone, voters overwhelmingly rejected RCV. In 2025, six more states banned RCV : Wyoming, West Virginia, North Dakota, Arkansas, Kansas, and Iowa. In 2026, more states like Indiana are considering bans. Alaska is ready to try again to repeal it (after a narrow failure to do so in 2024). And Republicans are proposing the “Make Elections Great Again Act”, framed to ban RCV but would also ban all non-Plurality voting methods. That is, RCV is actively bringing all the other alternatives into the cross-fire. And as of January 2026, 18 states currently prohibit the use of RCV, while only 7 allow its use in some elections.

Update (March 2026): Indiana and Ohio have now become the 18th and 19th state to ban RCV, respectively this year.

It’s time to sit down and seriously reconsider if RCV is really our only hope to get any sort of electoral reform passed in the United States. Or, is the sunk-cost fallacy blinding us to the fact that RCV may be poisoning the well for all electoral reform?

This post is not for the voter unconvinced of the need for electoral reform . This is for those of us united in our desire to permanently replace our current choose-one voting system with something better. In my first post on voting systems, I claimed that anything is better than our current Plurality system, but I no longer believe that to be the case. After all of my research and investigation, I have come to one stark conclusion: RCV is one of the worst choices we could make for electoral reform. It fails in every conceivable way, and it is time to abandon it in favor of a better alternative, such as Approval voting.

This post is also not about convincing the electoral reformers who have already seen the flaws with RCV but are yet to be sold on Approval voting over their personal favorite method like STAR or a Condorcet method. However, I feel that the mistakes that have been made with RCV may soon potentially be repeated, and I firmly believe that Approval voting is the clear path forward to not repeat history. That said, I find the primary issue to be with the fact that RCV has sucked all of the air out of the room for all the better alternatives, and that is the main topic I wish to address.

So that we are all on the same page about what these systems are:

Ranked-choice voting (RCV) is a voting method where voters rank candidates in order of preference. Only one vote on the ballot is considered at a time, the rest of the rankings are ignored. If no candidate gets a majority, the candidate with the fewest votes is eliminated, regardless of whether they are ranked by a majority over other candidates. Voters who are currently voting for that eliminated candidate have their votes conditionally transferred to their next-preferred candidate still in the race. If no candidates remain on a voter’s ballot, their ballot is thrown out and ignored (exhausted). This repeats until one candidate has a majority of the remaining votes.

In comparison,

Approval voting is a voting method where each voter can vote for, or “approve” of, as many candidates as they like, and the candidate with the most votes wins. It uses the same ballot as Plurality, but you can put a check next to multiple names.

Approval asks the much simpler, arguably more important question: who do you consent to govern you? Something a ranked system simply cannot capture.

This post will tackle a number of general topics that are important in the consideration of a potential voting system to adopt (and why RCV does not stand up to scrutiny)

• Trust, Auditability, and Transparency
• No Broken Promises
• A Method to Last
• The Expressiveness Trap: Performance and Satisfaction

Trust, Auditability, and Transparency

Our Shared Goal: In a high-stakes election, particularly a close one, voters should have minimal room to doubt or question the legitimacy of the results. Mistakes and errors should be hard to make, but at least be easily spotted and corrected. Audits should be quick and easy to perform, else the results can be easily undermined and stalled.

RCV simply is not a method that can be easily audited. Computing the margin of victory in IRV elections is NP-hard (Blom et al., 2016) , meaning that determining whether an election outcome is robust enough to withstand auditing requires exponential computational complexity in the worst case. Combined with RCV’s lack of precinct summability, this makes proper auditing extremely difficult and in some cases practically impossible. Not to mention, slow to count and requiring centralized tabulation through a black box algorithm that has been shown to be vulnerable to errors that can change the outcome of an election.

This is not just theoretical, it has happened in a real election in Oakland . A mistakenly checked box resulted in counting errors in an RCV school board election that yielded the wrong winner, and it took four months and a lawsuit to resolve. The false winner had already voted on matters before the rightful winner was eventually put in to replace them. This is one of the major problems with a black box method like RCV. Results are slow and errors can propagate without being easy to spot. This creates a major vulnerability for RCV elections, as they can be easily undermined and stalled by errors or disputes. If this had happened in a federal senate or presidential election, it would be pandemonium. Voters won’t just shrug their shoulders in understanding. Imagine a Democrat were falsely projected to have won a senate election in South Carolina. Suppose that gives Democrats a narrow majority in the senate, and that Democrat goes to Washington and votes on high-stakes legislation like a Supreme Court confirmation, just to find out “oops, a box shouldn’t have been checked, that should have been a Republican”. That would be nothing short of a political crisis.

In contrast, Approval voting is fully compatible with existing voting infrastructure and machines, and is the easiest to understand and explain. It is no more difficult to audit or count than our current Plurality system. It is also precinct summable, making results extremely fast to count and know the winner. No system, except Plurality, can truly compare to the ease of counting and auditing of Approval voting.

Even other methods like STAR and Condorcet methods are very reasonable to count and audit. However, to fully avoid central tabulation, both STAR and Condorcet need precincts to count the number of ballots that place each candidate over each other candidate, which any precinct can sum locally and post on the door. STAR additionally needs score totals for each candidate, though this is a minor addition. Both methods are *categorically* faster than RCV, for which central tabulation is a necessity. For Approval, Condorcet, and STAR, you can likely know the winner the night of the election, like in our current Plurality system. RCV stands alone among the reforms in its absurd impracticality.

For many I have spoken to, this is already a disqualifying factor. In our era of distrust in elections as transparent and simple as our current system, an impossible to audit black box system is a nonstarter. Simplicity is a prerequisite for trust. However, it gets a lot worse for RCV.

No Broken Promises

Our Shared Goal: A voting method that delivers on its promises, and does not create new problems. If we oversell the limits of a voting method, we set ourselves up for failure when those limits are inevitably exposed and voters feel betrayed and lied to.

RCV promised to eliminate spoilers. It promised majority rule. It promised that a party could send multiple candidates to the general election without fear of vote-splitting. It broke every single one of those promises, and then some, in Burlington and Alaska. By brazenly overselling RCV’s benefits, we set ourselves up for failure when those promises were inevitably broken.

In the 2009 Burlington mayoral election, the Democrat, Montroll, was eliminated in the second round. The progressive Bob Kiss then won the final round against the Republican Kurt Wright. Republican voters were told they could vote for their true favorite, Wright, in blue Burlington without fear of vote-splitting. They were told that their votes would transfer if their first choice could not win. Instead, their viable second choice, Montroll, who was generally preferred by Republican voters over Kiss (28.8% of Wright first voters ranked Montroll second compared to 9.5% for Kiss), was eliminated first. The Republican Wright then went on to be defeated by the Progressive Kiss. If they had gritted their teeth and voted for Montroll first, they would have gotten a more preferred outcome. RCV failed to deliver on its guarantee that voters could safely vote for their true favorite without fear of vote-splitting. Their first choice could not win, but their vote did not transfer.

In fact, Montroll was preferred and ranked above all other candidates by a majority. That makes him a “Condorcet winner”: a candidate who would defeat every other candidate in a head-to-head. When you fail to elect such a candidate when they exist, you both 1. fail to honor majority rule, and 2. ignore the ranked data you made the voters go to the trouble of providing. This is a great way to make voters feel that majority rule was violated, and that the system swindled them out of a better choice due to the “jank” of the short-sighted elimination algorithm.

Burlington repealed RCV after this fiasco, but we didn’t learn. The same scenario played out in Alaska in 2022, where Republican voters were told they could safely vote for their true favorite, Sarah Palin, first and Nick Begich second without fear of vote-splitting . Republicans were, once again, told not to worry, because their votes would transfer if Palin could not win. Instead, the more viable second choice, Begich, was eliminated first, and the Democrat Mary Peltola went on to win over Palin. Republican voters were betrayed again. And given that this had already happened in Burlington, these promises become far more egregious. It’s no longer ignorance, but blatant lies to candidates and voters. And then we wonder why voters, particularly Republicans, are souring on RCV.

One may argue that Peltola would have won under Plurality too, so how can this be an RCV failure? The issue is that the only reason there were two Republicans in the race against the one Democrat is because Republicans were lied to that they could do so without splitting the vote. In 2024, they learned from this mistake and the other Republican dropped out, leading to Begich winning the later rematch. Everyone knows that Plurality has an issue with vote splitting, and so I don’t think Peltola would have won in 2022 if the Republicans had instead run only one Republican (particularly, Begich). By lying about the existence of the spoiler effect in RCV, that resulted in a trust-breaking failure mode that has led to repeal efforts in Alaska every election cycle since, and has fueled preemptive bans in other states.

The argument that first-choice preference matters more than the Condorcet criterion is a fallacy in two major ways. First, failing the Condorcet criterion is an objective failure of majority rule, which RCV uses as a central claim for its benefits. Second, that value judgment doesn’t land for the majority of voters who can see, blatantly in the ballot data, that they preferred the Condorcet winner over the actual RCV winner, and feel betrayed when that majority candidate was eliminated in favor of a minority preferred candidate.

We ideally want for a voting method to not be seen as “rigged” or “unfair” or “party-biased.” Otherwise, why on earth would the other party ever agree to implement it in the first place? And if you do manage to pass it, how can you ever expect that other party to not try to repeal it at the first opportunity? RCV has managed to terrify Republicans into preemptively banning it, and fueling repeal efforts to preserve the status quo. And they can convince their voters that RCV is “unfair” to them, and they would be able to do it without lying.

In contrast, Approval voting is extremely Condorcet-efficient, fully consistent with its ballots data, and would have very likely elected both Montroll in Burlington and Begich in Alaska. Palin-preferring Republicans would have been able to safely vote for their true favorite without fear of vote-splitting, and optionally support a viable second choice at the same time. By soaking up approvals from Republicans and Democrats who vote defensively against Palin, Begich was extremely likely to have won. Approval is the only system without major, trust-breaking failure modes. It always elects the Condorcet winner induced by the ballots. Like Plurality, no losing candidate can make a case they should have won based on the ballot data collected in Approval, but Approval also crucially eliminates the excuse of the spoiler effect. If the winner got more approvals than you, that means they checked the winner’s box, looked at yours, and said “no thanks”. STAR can fail to elect the Condorcet winner induced by the ballots, and Condorcet can have troubles with cycles. Approval voting gives the winner full legitimacy every time.

The fundamental trust differential is that with Plurality, everyone understands how it works, even if they don’t like the results. It’s bad, but doesn’t pretend not to be. It’s still honest and transparent, even when it delivers bad outcomes. Approval keeps that honesty and transparency while liberating the voter to vote for multiple candidates, fixing literally every problem Plurality voting has. Voters can verify that their votes counted exactly as cast. It doesn’t make promises like “over 50% of the vote” that RCV can’t deliver on either (see this election where the winner won with just 37.2% of the vote). All the while, Approval delivers better, more legitimate outcomes than RCV.

With RCV, voters can’t verify the results without centralized tabulation, and have to wait longer for results just to watch majorities lose to minorities through opaque eliminations. In an era of widespread electoral distrust, we cannot afford a system that makes verification harder and creates legitimate grievances about ‘stolen’ elections when the Condorcet winner, whom the voters clearly signified a preference over the actual winner, loses.

A Method to Last

Our Shared Goal: Widely implement a better voting method that can last for generations to come, and resist repeal efforts. If it keeps getting repealed, particularly due to objective failures of the system, that erodes public trust in electoral reform as a whole.

Sometimes, I think we get too caught up on getting something passed, rather than getting the right thing passed. What happens when you pass something so flawed that it creates its own problems that fuel repeal efforts which lead us back to square one? That is exactly what has happened with RCV. The fact that RCV has been implemented in many places in the US is not as impressive as it seems when you consider how many of those implementations have been repealed or are under threat of repeal.

Is being “battle-tested” really so important if it keeps failing the tests? Some point to Australia, which has used RCV for a century. But the US, in particular, has a serious issue with electoral trust and polarization that some reports (Atkinson, Foley, & Ganz, 2024) suggest RCV may actually exacerbate. That report also suggests that the Condorcet failures we have seen are not just unlucky flukes, but a sign of things to come: in the polarized environment of US states, Condorcet failures occurred in about 40% of all simulated elections, and would be far worse in the more pivotal and polarized states, including the swing states. RCV simply does not do well in polarized environments, compared to methods like Approval or Condorcet method which have strong evidence (Cox) for being more robust to diffusing polarization. The Australia comparison does not hold water when it has already failed multiple times in the US, and appears likely to continue to fail in the US.

Unlike STAR and Condorcet methods, Approval has actually been used in the US, albeit in only a few places. If the metric is “has been used in the US,” then Approval satisfies that metric, and also hasn’t been repealed by voters who were disappointed by the results, as RCV has. The 2025 repeal in North Dakota was a partisan move by the Republicans in the state legislature, not the voters, and it was not due to any objective failure of the system.

The fact that RCV has more money behind it than any other alternative method to get it implemented is not evidence that it is the best method, particularly when that coalition seems to have reached its limits and voters in states like Montana, South Dakota, Oregon, and Colorado have rejected it on ballot measures. This isn’t a red state versus blue state issue. Nor is it a Democrat versus Republican issue. There has been a bipartisan rejection of RCV. The idea that RCV is some proven winner because it has been implemented to often disastrous results, does not hold water if our goal is a permanent replacement of our current system.

The Expressiveness Trap

Our Shared Goal: We want a voting system that allows voters to give greater expression to their preferences than our current system so that the “best” candidate can win, and voters can feel more satisfied with the outcome.

The trap is the fallacy that “more expressive” necessarily means “better winners” or “more satisfaction”. This is not supported by data or theory.

Performance

RCV performs significantly worse than all commonly proposed non-Plurality methods in VSE simulations. In comparison to Plurality’s 72-86% VSE, RCV only achieves about 80-91% VSE. In contrast, Approval voting achieves about 89-95% VSE (Voter Satisfaction Efficiency), depending on the model used (see this VSE analysis ).

Compared to STAR’s 91-98% VSE, Condorcet’s 87%-99%, Approval voting is extremely comparable. Given that all three of these methods are such a significant improvement over Plurality, the benefits from going from the simplest solution Approval to the more complex STAR or Condorcet are somewhat negligible. But most stark is how much worse an even more complicated mess like RCV performs compared to these much simpler methods. Just because RCV has rankings, and uses a fancy algorithm unlike the “crudely simple” vote tallying of Approval, does not mean it is actually better at picking good candidates.

Satisfaction

Something that we, who are so interested in electoral reform and politics, often forget is that most people really do not find the idea of ranking 10 candidates to be particularly appealing. Especially for voters who do not have the time or means to be fully informed about all candidates.

Ranked ballots make every aspect more complicated (voting, counting, auditing), take longer (MIT found an extra minute per race with 5 candidates), create longer lines, and lower confidence. That MIT study also indicated that respondents reported significantly lower levels of overall satisfaction with the voting process in RCV and a marked decrease in confidence regarding the integrity of the election results. The data also showed a heightened perception among voters that the process was “slanted” against their specific political party.

We must keep in mind that the average voter is not a political science enthusiast who has spent hours researching candidates enough to give a full strict ranking of all candidates at a moment’s notice. The idea that some voters will be happy to be able to rank their candidates is not a universal truth that applies to all voters. A more complicated ballot has other consequences, such as logistical bottlenecks, longer lines, greater chance for ballot errors, and lower confidence in the system. These act as barriers to voting, and are indirect disenfranchisement. The psychological benefits of satisfaction with a more expressive but complicated ballot have to be weighed with such considerations.

In contrast, Approval voting is extremely simple and intuitive. Arguably more-so than our current Plurality system, because people naturally can like or want more than one thing. It is a major improvement in expressiveness without becoming too complicated. The claim that Approval is too strategic is common, but unsubstantiated. In a real election, there is no plausible reason to cast anything but a sincere vote. It’s always optimal to vote for your favorite candidate(s). And if you have an underlying ranking of the candidates in your head, then a best response strategy in a large electorate always involves drawing a line of acceptability somewhere in that ranking, and approving all candidates above that line. Voting sincerely based on your true feelings strictly and optimally helps increase the margin between all acceptable candidates against all the candidates you don’t find acceptable.

But for the strategic voters, certain intuitive strategies for drawing that line (Laslier ) paradoxically increase the majoritarianism and Condorcet-efficiency of Approval voting (potentially up to 100%), and only requires knowledge of the top two viable candidates. The criticism that Approval will devolve into strategic bullet voting all the way back to Plurality also does not hold up. In the real world elections in Fargo and St Louis, and more recently Utah, voters consistently approve multiple candidates (just check out approval.vote). Laslier proves it is probabilistically optimal to approve every candidate you prefer to whomever seems like the favorite to win.

But bullet voting is no silver “bullet” to Approval. Laslier also shows that bullet voting when your favorite is ahead is often an optimal strategy that paradoxically leads to more majoritarian outcomes. But when your favorites are not viable, it’s objectively optimal to vote for multiple candidates (ex. extending your “line of approval” to include a viable candidate). This is something we have already seen in St. Louis elections: high bullet vote rates for frontrunners, while over 80% of those who supported a non-frontrunner also approved someone else.

The claim that a ranked ballot is necessarily better because it is more expressive is not something that can be rightly claimed. Many voters may find the idea of ranking candidates to be overwhelming, and may prefer the simplicity of just bubbling names to cast a single vote for that name. I can understand the appeal of a more expressive ballot, but let’s not pretend that it is an objective improvement that necessarily leads to better outcomes or more satisfied voters. In fact, the data suggests the opposite.

It should be taken into account that a trust-breaking failure mode, such as a Condorcet failure (which means the expression that voters provided was not coherently honored), can significantly decrease voter satisfaction and trust in the system. Therefore, a robust and consistent method like Approval voting or Condorcet leads to greater satisfaction in the long run, even if they are less “expressive” on paper.

One argument against Approval voting is that it psychologically feels like a voter is hurting their favorite candidate by approving of a less preferred candidate. Whereas in RCV, it is claimed that ranking a candidate below your favorite cannot hurt them. Both claims should be addressed:

1. In Approval voting, you can approve of your favorite candidate and a less preferred candidate without hurting your favorite. By voting equally for both, you do not decrease the margin between them or hurt your favorite, and you strictly increase the margin for both of them over candidates you did not vote for. This makes one’s ballot far more defensive and safe. At the very worst, your ballot may fail to help distinguish between two candidates you both approved, or two you both didn’t. But a failure to help among equally-treated candidates is a significantly less severe failure mode than a ballot that can actually hurt a candidate you did vote for lose to one you did not vote for, which is the case with RCV.
2. In RCV, it is true that ranking a less preferred candidate below a more preferred candidate cannot hurt the more preferred candidate. But this is a misleading virtue, because it can result in getting an overall less preferred candidate elected, which is a much more significant issue. For example, in Alaska, ranking Palin first and Begich second strictly helped Palin, yes, but it resulted in neither Republican winning the race, because only Begich was actually able to beat Peltola. In contrast, your singular Approval ballot can only help a candidate you cast a vote for. It is impossible for your ballot to elect a candidate you did not vote for win over a candidate you did vote for.

The argument that Approval “feels” worse than a ranked system is fair, but must be weighed against all the other more pressing issues with RCV that come with this minor improvement in “feeling”. On the other hand, many find that the ability to support multiple candidates equally is liberating. Approval voting is not “settling for something worse”, but it is absolutely a compromise between our simple but flawed Plurality system and the more expressive but less practical systems also in the conversation (for those who can treat politics like a hobby). It lacks the bells and whistles of something like STAR, but manages to provide 95% of the better outcomes of those systems while maintaining the simplicity of Plurality voting, and fixing all of its problems. That simplicity also works to make it more consistent and robust. The things that make Approval better than other systems are quiet and subtle. It’s the stable, sturdy bridge that never collapses, even if it isn’t aesthetically pleasing. The optics of it aren’t “sexy”, but in an era of distrust, maybe boring but reliable is truly superior.

Unlike RCV, which collects more complex data but then ignores most of it, Approval voting collects a simpler signal: consent, but uses it faithfully, transparently, and honestly.

Conclusion

If you want a system to last, that delivers on its promises, that the public can trust, that is transparent and auditable, and that allows for greater expression without falling into the expressiveness trap, then Approval voting is the clear choice. RCV has failed in every single one of these categories, and it is time to ditch it before it poisons the well for all electoral reform.

The most persuasive arguments for RCV only really hold up if you ignore the real-world data and failures of RCV. If you come away from this post thinking “well, I’m not convinced about Approval voting, but Condorcet sounds pretty good”, then I’d even be satisfied with that. But I feel the most important considerations we have to make when dealing with a practical problem like “how to replace our current Plurality system with something better”, then there are a number of disqualifying factors for RCV, depending on what you value:

• If you value voter trust in the system, RCV has completely and utterly failed in Oakland, Burlington, and Alaska.
• If you want the system to permanently replace our current system, RCV has consistently failed to do this, due to its repeated voter-led repeal attempts.
• If you value getting the best outcomes, RCV is one of the worst choices for an alternative system. It consistently underperforms methods like Approval (despite Approval being simpler), Condorcet, and STAR.
• If you value logistics, like the expenses and complexity of running elections, RCV is absolutely the worst choice. Approval, Condorcet, and STAR are all far more practical, with Approval being the best by this metric.
• If you value expressiveness, the fact that RCV fails to be Condorcet consistent means that it does not respect the expression it actually collects. This leads to trust-breaking failure modes.
• If you value legitimacy, then RCV is one of the worst here. Again, when you do not elect the Condorcet winner expressed by the ballots then you fail to provide legitimacy to the results. STAR cannot guarantee this either, and even Condorcet struggles when a cycle occurs (despite their immense rarity, only occurring in small elections). Approval voting (and even Plurality voting to a far lesser extent) is the only system that can provide full legitimacy to the winner in every case (see this post).

I worry that a Condorcet failure or legitimacy crisis may occur under STAR, and we’re back to the same issues with RCV, but worse. If we get another chance after RCV, and we go with STAR–which is less tested, fails more criteria, and has far less literature than Approval voting–I don’t see us getting a third chance. I worry that a Condorcet cycle in a high stakes election may bring into question the integrity of whatever cycle-breaking mechanism is decided for Condorcet. We have seen that room for error and doubt can lead to disaster for RCV, and so I worry that any system that has a nonzero chance of a legitimacy crisis may be a nonstarter in the current political climate.

The reform conversation is split between convincing people that we need reform at all (or that reform is not a secret plot to undermine democracy), and squabbles between those who are convinced their solution is the only correct one. I hope that this post can help with the latter, and convince those who are already on board with reform that RCV is not the answer, and that Approval voting is the best path forward.

The only system that checks all of the boxes is Approval voting. It is the only system that makes sense in this era of distrust in our elections. It is the most reliable, robust, and practical solution available to us right now.

To the RCV advocates, and even the non-RCV advocates who have dismissed Approval, I ask that you consider these points carefully. Our goal to replace Plurality voting unites us, and I truly believe that the best chance we have is Approval voting.