Storable votes: Can we solve gridlock and yet protect the minority?

by Andrew Gelman on June 7, 2012 · 21 comments

in Legislative Politics,Political Theory

This is a guest post from Alessandra Casella, an economics professor at Columbia University, and Sébastien Turban, an economics Ph.D. candidate there. Alessandra Casella’s book on Storable Votes is available here. Sébastien Turban joined the project when the manuscript was already written but contributed substantially to its final shape. 

The Senate has become less efficient in passing significant legislation. The public is upset at the gridlock in Congress and wants politicians to cooperate.  Political commentators believe that significant responsibility for such malfunctioning rests with the Senate filibuster.  Sarah Binder, Greg Koger, Greg Wawro and Eric Shickler on The Monkey Cage and in their respective books, Barbara Sinclair and Ezra Klein in The Washington Post, have all discussed the desirability of possible reforms in detail over the last years.

It seems to us that the debate on the filibuster points to a more general question on the appropriate protection of the minority.  The filibuster exists to protect the minority party, and indeed taming the tyranny of the majority is essential to the legitimacy of democracy.  But the filibuster does so very inefficiently: it amounts to a super-majority rule, with its built-in bias towards the status-quo.

The problem is that sanctioning simple majority rule, treating all votes and voters identically, and protecting the minority seem mutually incompatible goals.  Using storable votes, this need not be the case.  Suppose that a group of voters were asked to pass or fail a number of independent proposals. Each proposal is decided by majority vote, and each voter is granted the same number of votes.  However, subject to the overall budget of votes, each voter is allowed to decide how many votes to cast on any individual proposal.  With ten proposals and ten votes, an individual voter could cast one vote on each proposal, several votes on some of the proposal and none on the others, or all ten votes on one proposal.  The voter will cast more votes on the proposals he feels more strongly about, increasing his probability of winning when it matters.  Storable votes (1) equip every voter with the same number of votes; (2) allow voters to cast more votes on decisions that matter to them most, and (3) empower the minority to prevail occasionally.

The central idea is the possibility of shifting one’s own votes from one contest to another, of storing votes not spent on decisions that are low priorities for use on decisions that matter more.  Voters’ behavior thus reflects truthfully the relative intensities of their preferences over issues.  By cumulating votes on one or few of the proposals, a cohesive minority can win when its preferences are most intense.  And because the majority generally holds more votes, it is in a position to overrule the minority if it cares to do so: the minority can win only those issues over which its strength of preferences is high and the majority’s preference intensity is weak.  Note again that all voters are treated equally: all individuals are granted the same number of votes and all votes are equal.

The mechanism has been tested  by Alessandra Casella and her co-authors (Shuky Ehrenberg, Andrew Gelman, Thomas Palfrey, Raymond Riezman, Jie Shen) in laboratory experiments, and in one natural setting. The results confirmed the intuition: the minority wins occasionally, but only when the costs to the majority are small. The conclusion continues to hold when one party is granted some limited agenda power.

Storable votes resemble cumulative voting, a semi-proportional electoral system used in local jurisdictions and in corporations to facilitate the election of minority candidates.  Storable votes, however, are a voting system for decision-making, as opposed to an electoral system, and in this they differ not only from cumulative voting but also from most other well-known voting rules: for example, Borda rule, Approval Voting, Range Voting, or Single Transferable Vote.  In an electoral system, all candidates are compared, and a fixed number of winners is elected.  With storable votes, each proposal is compared only to its own rejection, and all could in theory pass, or all could fail.  Nor do storable votes resemble vote-trading: voters can shift their individual votes across different proposals but are not allowed to trade votes interpersonally.  In fact, the system discourages informal exchanges of favors: if not cast, a vote can be used over a different decision, and because it preserves its value, spending it to satisfy others’ requests is costly, even in exchange for promises of future favors.

In principle, storable votes could be implemented easily: they can have large effects, but are a reasonably minor deviation from simple majority voting.  In the case of the filibuster, it is tempting to think, for example, that storable votes could help to address the increasing delays in confirming nominations.  Consider the following procedure.  At fixed intervals of time, the Senate is presented with a list of nominees, each nominated for a specific position, and each senator is granted a number of votes equal to the number of names on the list.  A vote is then called on each nominee. A senator can cast as many of his total votes as he wishes in his preferred direction (for or against the nominee), distributing the number of votes at the senator’s disposal only over the specified set of nominees.  The nominee is confirmed if a simple majority of votes cast on his name are in favor. New votes are assigned whenever a new list is voted upon.

Much of the scheme’s possible success would depend on the details of its rules:  How frequently will the lists be presented?  Should all nominations be treated equally? We do not want to argue here that storable votes are ready for immediate implementation. Rather, we want to suggest a potential new direction for solving gridlock in polarized environments.

P.S. from Andrew Gelman: Many commenters below suggest potential problems with storable votes. These are all worth looking into, but let’s not forget that our current system is a storable votes system with 1 fixed vote on each issue and 0 storable votes. Casella would keep the fixed vote; the only change is to increase the number of storable votes.

{ 21 comments }

Thomas Leeper June 7, 2012 at 12:55 pm

Doesn’t this system theoretically allow multiple contradictory proposals to be simultaneously implemented? In other words, couldn’t two minority groups force through opposite policies by storing all of their votes for a single proposal (e.g., one minority could put all of their votes behind completely banning abortion and another minority could put all of their votes into making it legal in all circumstances), leaving the decision-making body with two adopted but contradictory policies? Or, at least open the possibility of leaving the legislature with a constant cycle between radically different policy alternatives?

Jim June 7, 2012 at 1:08 pm

This is intriguing, but one can see how the majority leadership would use it to their advantage. The legislative items grouped together for votes would most definitely be done to try to embarrass the minority and vote (not vote) for politically unfavorable (favorable) items, or pull votes towards or against specific items.

Interest groups, as well, would complicate the system. They would likely change their yearly scorecards to reflect not just whether or not a senator voted for or against a specific proposal, but how many votes that senator allocated towards it. These considerations would seriously complicate the system, and probably bias voting even more towards parochial or group interests. As it stands, whenever senators or MCs have to take numerous votes in a row, it tends to empower leaders and interest groups as they are trying to quickly assess the political situation they are voting under for a variety of issues in a short period of time.

Additionally, a simple reform to allow storable voting in the Senate wouldn’t solve the filibuster problem. A simple reform like that would still allow single senators to block votes since the Senate operates largely under unanimous consent. It also would raise new issues — could senators place a hold on just one of the items in a group vote, or would they have to hold all of them?

It is a very interesting idea, but I would like to see the authors incorporate some of the political and procedural realities of the Senate for us to understand how storable voting might actually function in the modern U.S. Senate — or any other legislative chamber, for that matter.

RobC June 7, 2012 at 3:07 pm

It may well be that “the public is upset at the gridlock in Congress,” but the citation the authors provide for that proposition doesn’t speak to that issue. That citation, to polls of Congressional job approval, shows that the public doesn’t approve much of Congressional job performance but doesn’t show the reasons for the disapproval. The authors’ implicit assumption seems to be that what the public wants is more legislation and accordingly roadblocks to such additional legislation are the cause of their disapproval of Congress. For many people that may be the reason for their disapproval, but other reasons might be members’ grandstanding, political opportunism or perceived disconnection with voters, or indeed the public’s dissatisfaction with the legislation that is passed (e.g., the Affordable Care Act).

To be sure, this is only a blog post, not a scholarly paper, and it doesn’t lend itself to extensive footnoting. That said, I hope it’s not out of line to point out when the hyperlink doesn’t provide authority for the proposition to which it’s attached.

Andrew Gelman June 7, 2012 at 3:24 pm

RobC:

I haven’t looked at these citations but you might very well be correct on this. Casella is an economist and her interest in storable votes is more generally as an approach to aggregating preferences of varying intensity. Concerns about Congress might motivate trying out new voting rules but the particular issues of gridlock are not crucial to the larger ideas of Casella and her collaborators.

I agree with you that it would be an important research project in itself to connect some of these general ideas of voting systems to specific concerns about the U.S. Congress, connecting the economics to the political science.

Sebastien Turban (@PtitSeb) June 7, 2012 at 5:55 pm

RobC,

Sebastien here. You’re right, this was a poor choice of link. In looking for an answer to your comment, I found it far harder than I thought to justify that the public is upset because of gridlock (Sorry). It is not clear, as you suggested.

Here is what I can say quickly (as a disclaimer, I am not a public opinion or Congressional expert…):
- The adjectives you get when asking one-word impressions of Congress are “Dysfunctional”, “Corrupt”, “Selfish”, “Inept”, “Confused”, “Incompetent”, “Ineffective”, “Lazy”
- Most people think that the Congress has accomplished less than previous ones, and think that the members are the problem.
- Most Americans want the two parties to cooperate, and would like the parties to compromise.
- Most Americans think that the delays in Congress are due to the fact that members want to score political points, as opposed to serious differences on issues. In the same poll, they want both Congress and the President to take the lead on the national agenda, on determining the federal budget, on going to war. Incivility in Congress is seen as a major problem.
- Most Americans are “angry” and “want to throw” elected officials in Washington “all out”, are “frustrated because all they do is fight”, and “wish they would compromise”

Thanks for the comment!

Jim June 8, 2012 at 1:33 pm

Worth considering is that the public may say it dislikes things about Congress, but may be uninformed, or be misdirecting their anger.

Take, for example, that most Americans say they want the parties to cooperate and that they want compromise. To a degree, many Americans think this because they do not realize the degree of conflict over the major issues within society itself. They believe most Americans are in agreement about what the government should do on most things and see politicians as the reason for conflict either because they are too extreme, disconnected from the public, corrupt, or something else.

I suggest you and your colleagues read the books ‘Congress As Public Enemy’ and ‘Stealth Democracy’ by John Hibbing and Elizabeth Theiss-Morse for more on the bases of public perceptions of Congress.

alessandra and sebastien June 7, 2012 at 6:43 pm

Thanks for pointing out the wrong citation – our mistake. As Andrew points out, our main point is that if there is a will to beat the gridlock then different voting systems, and storable votes in particular, may be one possible direction.

Thomas Leeper: If both proposals are on the agenda, the chances of this happening seem slim – the first minority would need to think about keeping enough votes to beat the second proposal. More generally, the comment is important because it stresses the role of the agenda: voters should either know precisely which proposals will be voted upon, or be well-informed about the exact procedure through which new proposals are called to a vote and be able to take that into account when deciding how many votes to cast. Not by chance, we tailor our proposal to nominations with a fixed list of names, each attached to a different vacancy.

Jim: We agree: before any concrete talk of application, much more detail is needed on Senate procedures. On increased potential for manipulation: well, maybe or maybe not. Embarrassing the other side may be counterproductive: for example putting on agenda proposals whose only purpose is to deplete the other side’s votes will not work unless the majority is also willing to spend some of its votes on the same proposals. The question of whether to make public the number of votes cast, in addition to the direction of one’s vote, is also an interesting question.

Thomas Leeper June 8, 2012 at 1:57 pm

Thanks for your answer. I guess I wonder about the agenda-setting process then. Is it set by majority rule or by a single agenda-setter? Because that would prevent this voting strategy from ever translating into policy wins for a minority bloc because they would never get on the agenda, right?

But if the agenda itself is set by storable votes, then questions emerge about whether votes can only be used within one of two stages {agenda-setting, voting} or whether votes can be stored from an agenda-setting phase and used later in a voting stage. I think it gets murky what would actually happen (outcome-wise) once you introduce agenda-setting procedures. My apologies, as I haven’t read the book, but just wondering how viable this really is under realistic conditions.

alessandra and sebastien June 9, 2012 at 4:29 pm

We are not proposing to change how the list of nominees is arrived at. But, given the list, storable votes increase the minority’s ability to vote a candidate down – and thus presumably at least in part its influence on the names that are presented. I think we all agree that the institutional details need to be studied further.

Jon M June 7, 2012 at 7:27 pm

Even in the case of a public and fixed set of names the agenda setting problem is not solved. The whole proposal leads to far too many incentives for strategy in agenda setting. For instance, in the nominees case, the president has every incentive to nominate as unacceptable as possible candidates to draw opposition votes. If the agenda setter knows that she can’t pass all items she should try and make the items that won’t pass as aggravating as possible to the other side. This seems like a perverse incentive and one that will end badly when somebody miscalculates.

alessandra and sebastien June 8, 2012 at 10:31 am

Putting a name on the list just to use the other side’s votes is a costly proposition. If the majority does not intend to spend votes on that name, the minority can win (vote the candidate down) with very few votes. If instead the majority intends to spend votes on that name, then it does so at the cost of other candidate it may feel more strongly about.
In the experiments we ran, we let a chair choose either the order of the proposals or which subset of proposals to call, sequentially, out of a known set. We did not see any evidence of manipulation, mostly because the agenda setter gave priority to the proposals (s)he felt strongly about, as opposed to focusing on the preferences of the other voters and trying to distort the voting. That, by the way, was also close to the chair’ optimal theoretical policy.

Jon M June 8, 2012 at 1:16 pm

There seems to be two alternative problematic situations depending on whether votes are allowed to be changed or added to.
Under the circumstances where votes are secret until the voting is finished the agenda setter can use up all the votes of his opposition by using the following strategy:
Announce in advance that they will propose several utterly unacceptable (to the minority) candidates (n). Announce that the agenda setters allies will put “(total opposition votes/ n)” votes on one of these candidates at random.
Provided the payoffs to the opposition of the unacceptable candidates are negative enough (Ariana Huffington and Michael Moore as federal judges), and the commitment credible, the opposition rationally have to split all their votes on blocking the unacceptable candidates. The agenda setters party can then use their remaining votes to elect all their preferred candidates at leisure.
This strategy is slightly less effective if the results of each vote are announced but would still greatly increase the power of the agenda setter’s faction. Politics would descend into people sending signals that this vote was going to be the clash of the century and then not wasting any votes on it.
One way round this would be to allow voters to assign additional votes in response to the tactics of others. But then the system collapses into majority voting as the majority just assign extra votes to any proposal that the minority are about to win. Seen from this side the minority successes are really just majority failures to predict the voting behaviour of the minority.

alessandra and sebastien June 9, 2012 at 4:22 pm

Jon,
These are interesting points. The scenario you describe depends on both the majority and the minority succeeding in maintaining strict discipline under secret voting. And even then, neither group is acting optimally: the majority should not spend any votes if the minority is committed to using enough to always win on the “problematic” candidates; but then of course, neither should the minority, etc.
If you are interested in agenda problems, take a look at chapter 4 of the book.
Your comment below about collusion is also a very good question. In the experiments we ran, the minority was able to maintain collusion; the majority (for whom collusion matters less) mush less so.

Jon M June 10, 2012 at 6:25 am

On the first point, you might be right if the agenda setter is in the majority but the random voting on unacceptable candidates also works for a minority with agenda setting power.

For instance let’s say the Republicans win a small majority in the senate and Obama retains the presidency. Obama nominates 3 candidates who the Republicans cannot stand the thought of but the Democrats are slightly positive or indifferent to. He then announces the random voting strategy for these three which will use up all of the Republicans votes to guarantee none are elected.

All the remaining Democrat votes can be used to elect Obama’s entire roster of nominees. The outcome is pareto efficient. Neither side can change their strategy without being worse off: if the Republicans don’t play along, they end up with the worst candidate imaginable elected and the Democrats are getting nearly all their nominees passed whilst only having a minority of seats.

Since the voting is set up so that *any* Republican defections will lead to disaster, party discipline should be easy to maintain. And the democrats have a huge incentive to do this as they get to act like a majority with only a minority of seats.

Jon M June 10, 2012 at 6:54 am

Actually possibly not pareto efficient but definitely an equilibrium.

Jon M June 8, 2012 at 8:59 am

Also, there is ample room for vote trading. It doesn’t make sense to swap votes but it makes a lot of sense to swap non-votes. If two voters on opposite sides collude to not vote on a party line issue, they can free ride on the rest of their party in order to save votes for voting on pork barrel projects.

Tom Sciortino June 8, 2012 at 6:02 pm

It’s interesting what the vote totals would reveal about the relative importance of the issues.

Grant June 10, 2012 at 8:11 pm

Politicians already do have a system like this, their time. Every Senator and Representative has a limited amount of time to spend at D.C. voting and they are required to spend a good deal of time doing politicking and money-raising at home. Therefore they can’t all be there for every vote on every bill, but must budget their time and appear only for votes that will win them favors or votes that they actually care about.
Besides that, what if they should run out of votes just before the news breaks that the administration has been funneling weapons and money to far-right soldiers in South America and they want to pass a law stopping it?

anon June 11, 2012 at 8:36 am

Several potential issues with a storable vote system as applied to nominations have been identified by posters here. But this seems just the tip of the iceberg. Examination of Senate practice will reveal that applying the concept to legislation would be significantly more complicated. At some level, nominations seem to be comparable items of Senate business (with obvious variations in salience among them). But how does this work in practice when the Senate is facing legislation that ranges from narrow-in-scope to broad-in-scope? For instance, recall that germaneness is typically not required for amendments. So one bill, for example, could have in it a major transportation reauthorization, abortion policy, bankruptcy reform, and innumerable unrelated policy proposals (not to mention the same [worse?] problem commonly presented by omnibus appropriations bills). Doesn’t a storable vote system approach create inevitable distortions in such cases (and become so complicated as to make ex ante prediction of effects quite difficult)? The effects would apply to the process by which these various unrelated policy pieces get put into the same bill in the first place (a process which may happen in the early agenda-setting stages, or alternatively, one that may occur after the Senate has already decided to move to a bill), as well as to the process by which they are decided in the end. To be sure, this situation already has interesting and complicated effects on Senators’ behavior under current practices, but I think it would have to be specifically considered if one were to evaluate the effects of a storable vote approach.

Jon M July 3, 2012 at 11:31 am
Jack Condorcet February 12, 2013 at 1:48 pm

One idea regarding agenda setting I had while reading this post:
• Create a set schedule where events A, B, and C occur on a cyclical basis. For example, A happens every Monday at 6pm, B every Wednesday at 6pm, and C every Friday at 6pm.
• A. Each member of the chamber may make a proposal.
• B. Members vote using a Proportion Representation system on which proposals from event A are featured in event C. There is a set number of proposals that are chosen to advance each time.
• C. Members vote on proposals chosen in event B using storable votes.

I’m not saying this solves all the problems surrounding strategic agenda setting, but by allowing the minority to set some of the agenda, it may give them the ability to counter-strategize. More analysis is needed here.

This proposal may also introduce other issues, such as: What if a proposed bill is too long or too complex to be read and analyzed in the amount of time scheduled? What provision should be made for crisis legislation? Again, more analysis is needed.

Comments on this entry are closed.

Previous post:

Next post: