Some Skepticism About the “Separable Preferences” Approach to the Single Subject Rule: A Comment on Cooter & Gilbert

Response To:  Robert D. Cooter & Michael D. Gilbert, A Theory of Direct Democracy and the Single Subject Rule, 110 Colum. L. Rev. 843 (2010).

The "single subject" rule—mandating that voter initiatives embrace no more than one subject—has vexed courts and scholars alike for decades.  In the Columbia Law Review, Professors Robert D. Cooter and Michael D. Gilbert make a valiant attempt to bring rationality and consistency to judicial application of the rule.1  They propose a "separable preferences" decision criterion that they argue will improve the democratic process by reducing logrolls and depoliticizing adjudication.2

We are skeptical.  Though we agree that the decision criterion that Cooter and Gilbert propose could work in theory to limit the number of voter initiatives containing certain types of logrolls, we have two primary concerns.  First, their approach is premised on a normative hostility to logrolling that is not justified by existing political theory.  While it is true that some logrolls can lead to harmful outcomes, other logrolls can be socially beneficial, and the criterion that Cooter and Gilbert propose would eliminate both socially beneficial and harmful logrolls.  Second, we do not believe that the separable preferences criterion would lead to court consistency in application of the single subject rule.  Our own research finds that in states that allow courts to aggressively police single subject violations, judicial decisions in single subject cases are heavily influenced by the judge's political preferences rather than neutral, objective rules.  We suspect that courts in states that have shown restraint in application of the single subject rule would become much more aggressive if they adopted the separable preferences criterion, leading to greater politicization of adjudication.

I.  Cooter and Gilbert in a Nutshell

In a nutshell, Cooter and Gilbert view the initiative process as a useful supplement to the legislative process to deal with issues of legislative self-dealing and other agency problems.3  The initiative process—whereby a person or group drafts proposed legislation, secures enough signatures, and puts the proposed legislation up for a vote of the people—allows citizens to counteract legislative agency problems.  Initiatives can be a somewhat clumsy tool, however:  Initiatives cannot be changed once they go into circulation, and it is difficult for voters to negotiate and incorporate compromise policies via initiatives.  Cooter and Gilbert's main concern with initiatives is the possibility of logrolls or riders.4  Logrolls occur when two proposals, neither of which alone would gain majority support, are bundled together in an initiative for an up-or-down vote.  Riders are unpopular measures that are bundled with a popular measure and put together in an initiative for an up-or-down vote.  Initiatives combining multiple issues also may lack transparency and may be confusing to voters.  Cooter and Gilbert see the single subject rule as a way to solve problems arising from the bundling of issues into single initiatives.5

Implementation of the rule in practice has been difficult because the term "subject" is not self-defining.  Cooter and Gilbert propose that courts replace their various formulations of the single subject rule (such as the California requirement that initiative provisions be "reasonably germane"6 with one another), by having courts consider whether most voters have "separable preferences" over the various provisions of a bill.7  If a voter could decide whether one provision of the measure should pass without knowing if the other passes, then voter preferences are separable and the measure should be found to violate the single subject rule.  When voter preferences are not separable, there is no rational way to consider the individual issues in isolation, and it is proper for the voters to consider the issues as a package.  Cooter and Gilbert claim that this framework "clarifies rather than supplants existing [single subject] precedents [and, therefore, the separable preferences] theory could be applied immediately and without constitutional revision" in states applying the single subject rule to initiatives.8

II.  Normative Concerns with the Separable Preferences Criterion

Our first concern is that the normative argument for preventing bundling of separable issues is weak.  Cooter and Gilbert take the position that "[i]mplicit bargains [such as logrolling] in initiatives are likely to be socially harmful."9  We agree that multi-issue initiatives may be harmful, but it is entirely possible for them to be socially beneficial as well.  Consider the following example with three voters who are considering whether to go forward with two "projects," A and B.  The numbers represent each voter's utility from each outcome (for example, voter 1 receives utility 100 if project A is adopted and zero if project A is not adopted).10

In this case, voter 1 enjoys very high benefits from project A and is mildly hurt by project B; voter 2 enjoys very high benefits from project B and is mildly hurt by project A; and voter 3 is mildly hurt by both projects.  If we count the welfare of each person equally, the socially optimal choice is to approve both projects:  Project A produces a net gain of ninety-eight, as does project B.

If the projects are decided separately, both will fail:  Voters 2 and 3 will vote against project A, and voters 1 and 3 will vote against project B.  If the projects are bundled, then both will pass:  Voter 1 will support the package (the gain of 100 from A offsets the loss of one from B); voter 2 will support the package (the gain of 100 from B offsets the loss of one from A); and voter 3 will vote no.  Allowing the projects to be bundled brings about the socially optimal outcome.  Adoption of both projects is also majoritarian in the sense that two out of three voters prefer a world where A and B are adopted to the other option where neither is adopted.

In this example, voter preferences are separable in the Cooter-Gilbert sense because each voter knows his or her position on each issue, independent of the outcome on the other issue.  The separable preferences criterion would strike down an initiative that bundled A and B, and would require the issues to be voted on separately (or perhaps not at all if the proponents lacked the resources to fund separate initiatives).  In this situation, enforcement of a single subject rule using the separable preferences option would be socially harmful and countermajoritarian.11

Of course, there are also situations in which a logroll initiative can lead to a socially harmful outcome.  Cooter and Gilbert assert that the situation of harmful initiatives is "likely," while the situation of beneficial initiatives is unlikely, but they offer no evidence for this claim and it is not self-evident to us.12  Indeed, it seems plausible that groups with different interests may routinely package (or would routinely package, but for the single subject rule) rather disparate measures together in order to pool resources to qualify their measures for the ballot.  If logrolling is often welfare enhancing, the justification for the separable preferences criterion evaporates (and in fact, it would be socially harmful).

More broadly, Cooter and Gilbert argue that political bargaining is best done in legislatures, not via initiatives, and, in particular, that the bundling of issues is more likely to be done in a socially beneficial way in a legislature because negotiation and compromise is easier in that context.13  We agree that bargaining is easier in a legislature (although one should not ignore the fact that initiative proponents often form coalitions across diverse groups as well), but this only means that both good and bad logrolls are easier to put together in a legislature.  The question remains which is most likely to occur, and again we know of no evidence or compelling intuition in favor of legislatures.14  Moreover, the argument that voters should be prohibited from making laws because legislatures can do it better runs counter to the central premise of the initiative process.

We also note that Cooter and Gilbert do not fully explain why the separable preferences criterion would increase transparency and reduce voter confusion.  The point is not obvious:  Many complicated and confusing measures (about which voters have nonseparable preferences) will remain on the ballot, while other simple measures (about which voters have separable preferences), such as a measure to both close schools on Lincoln's birthday and build a library, would not be allowed.

III.  Practical Concerns with the Separable Preferences Criterion

Our second concern is with the practical implementation of the separable preferences criterion.  Imagine if states with the single subject rule adopted—either through state constitutional amendment or judicial rule—the separable preferences criterion for adjudicating "single subject" challenges.  Would this lead to neutral court adjudication of single subject disputes?  We suspect, to the contrary, that the principle would prove difficult to implement neutrally, and the attempt to enforce the single subject rule more aggressively would lead to less neutral decisions.

To illustrate, consider two proposed initiatives and ask yourself if either, or both, would violate the single-subject rule:15

Initiative A shifts responsibility for drawing state legislative and congressional districts from the state legislature to a redistricting commission.  The commission must draw single-member districts, changing current practice which allows multi-member districts for the state legislature.

Initiative B limits marriage to one man and one woman.  It also prevents localities from adopting "civil unions" for non-married couples that would give those in such unions any of the rights of married couples.

Under current single subject tests, it appears that jurisdictions applying a restrained test, such as California's "reasonably germane" test, would likely find no single subject violation.  In courts applying a more aggressive test, such as Florida's "oneness of purpose" test, the results are unclear,16 and may turn on how the judges view the merits of the initiatives.

Cooter and Gilbert would use neither a "reasonably germane" test nor a "oneness of purpose" test.  Instead, they would have judges determine whether voters have independent preferences on the various parts of the provision:  If a voter could decide how to vote on one aspect of the measure without knowing whether or not the other would pass, then the measure would violate the single subject rule.  In making this determination, judges are supposed to take evidence of voter opinion on separability—apparently based primarily on polling of voters in the jurisdiction—to determine if there is a single subject violation.  If a majority of voters see the parts of the initiative as contingent on each another, the package goes to the voters; if not, the initiative comes off the ballot.17

We believe polling data are unlikely to yield a clear-cut result in many cases.  It is well known that polls can be manipulated through wording choices.  A judge could well face competing evidence on polling about the separability of preferences.  If political operatives know that the single subject question will turn on polling, these operatives will have an incentive to try to influence public polling on the separability question.  For instance, imagine an advertisement that reads:  "Same sex marriage and civil unions.  Same issue.  Same problem.  Vote No."  And on issues about which the public is poorly educated, such as redistricting, polling results are likely to be unreliable.  Because polling on separability of preferences is likely to be all over the map, judges will end up with inconclusive evidence and may be forced to fall back on their intuitions and guesses about what voters believe.

The separable preferences criterion also calls for judges to be more aggressive in enforcing the single subject rule than they currently are in many states, such as California.  What would be the consequence of heightened judicial scrutiny in an environment where judges must rely on their own intuitions and educated guesses?  Dan Lowenstein argued that when judges are forced to make highly subjective decisions, it is hard for their reasoning not to be influenced by their belief systems, values, and ideologies.18  In a recent empirical study, we found strong support for this claim.19  We examined votes of state appellate court judges on single subject cases in five states during the period 1997-2006 (more than 150 cases and more than 700 individual votes).  We found that judges are more likely to uphold an initiative against a single subject challenge if their partisan affiliations suggest they would be sympathetic to the policy proposed by the initiative.  More important, we found that partisan affiliation was extremely important in states with aggressive enforcement of the single subject rule—the rate of upholding an initiative jumped from forty-two percent when a judge disagreed with the policy to eighty-three percent when he agreed—but not very important in states with restrained enforcement.  This evidence suggests that the separable preferences criterion, by increasing the aggressiveness of enforcement but placing judges in a position with significant discretion, could well lead to an increase in arbitrary or political decisions.

* * *

In sum, we share Cooter and Gilbert's concern for the subjective and political nature of current single subject adjudication in some states, but we are pessimistic that their proposed approach will improve the situation.  Their separable preferences criterion is designed to eliminate a large class of logrolls from initiatives, but even if it could be applied as intended, there is no compelling reason to believe that it will filter out initiatives that are harmful rather than helpful to social welfare.  Moreover, we are pessimistic that it could be applied as intended.  It strikes us as rather challenging to ascertain whether voters have separable preferences on any particular set of issues, and it seems likely that judges will have to introduce their own personal intuitions about the connectedness of issues in order to apply the rule in practice.  As we show in our own empirical work, the combination of heightened enforcement of the single subject rule and the need for judges to make personal assessments leads to adjudication that is heavily influenced by the political preferences of judges.  Rather than step up enforcement with such a new rule, we believe it is better for courts to simply take a restrained approach to adjudicating single subject questions, leaving the issue of the desirability of package legislation for the voters themselves to decide.

*William H. Hannon Distinguished Professor of Law, Loyola Law School, Los Angeles. 

**Charles F. Sexton Chair in American Enterprise at the Marshall School of Business, Gould School of Law, and Department of Political Science, University of Southern California.  We thank Inna Zazulevskaya for research assistance.

1 Robert D. Cooter & Michael D. Gilbert, A Theory of Direct Democracy and the Single Subject Rule, 110 Colum. L. Rev. 843 (2010).

2 Id. at 847–48.

3 Id. at 854–55.

4 Id. at 845–46.

5 Id. at 861–64.

6 Legislature v. Eu, 816 P.2d 1309, 1320 (Cal. 1991).

7 Cooter & Gilbert, supra note 1, at 868–70.

8 Id. at 871.

9 Id. at 856.

10 This example appears in John G. Matsusaka & Richard L. Hasen, Aggressive Enforcement of the Single Subject Rule, 9 Election L.J. (forthcoming October 2010), available at:  http://ssrn.com/abstract=1549824 (manuscript at 10–11, on file with the Columbia Law Review).

11 Cooter and Gilbert present an example that is formally equivalent to this one, and correctly observe that the two issues could only pass when bundled, but do not observe that failure of the bundle can be socially harmful and countermajoritarian.  Cooter & Gilbert, supra note 1, at 872–73.

The separable preference criterion can decrease voter utility in another way as well.  Imagine that a vast majority of voters have strong views in favor of a bundled measure allowing (or banning) both gay marriage and civil unions.  A vast majority of voters also prefer to have the ability to vote on bundled measures than to be denied the ability to vote on such measures.  But a judge removes the bundled gay marriage/civil union measure from the ballot because fifty-one percent of polled voters say they believe that the issues of gay marriage and civil unions are not linked.  By applying the separable preferences criterion, the courts would prevent voters from voting on an initiative package that commands widespread support.  Striking the initiative sacrifices social welfare for a technicality, using a mechanical rule not favored by voters.

12 Id. at 856.

13 Id. at 853–56.

14 It seems worth noting that most states have laws prohibiting logrolls in legislatures as well:  We might ask why the separable preferences criterion should not be applied to legislatures also.

15 The example comes from Richard L. Hasen, Ending Court Protection of Voters from the Initiative Process, 116 Yale L.J. Pocket Part 117, 118 (2006), at http://www.yalelawjournal.org/images/pdfs/71.pdf (on file with the Columbia Law Review).

16 As Hasen explains: 

In two opinions issued on the same day in March 2006, the Florida Supreme Court struck down Initiative A and upheld Initiative B against single-subject challenges.  As to the redistricting measure, the court ruled that federal redistricting and state redistricting are separate subjects, and both differ from the use of single-member districts.  As to the marriage proposal, the court held that both parts of the measure dealt with the subject of marriage.  It is not hard to imagine other courts reaching different conclusions.  Indeed, some have.  A California court upheld an election reform measure much more disparate than the Florida redistricting measure against a single-subject challenge.  A state court in Georgia struck down a measure very similar to Initiative B on grounds that same-sex marriage and civil unions are separate subjects (a decision later reversed by the Georgia Supreme Court).

Id.

17 Cooter & Gilbert, supra note 1, at 876–77.  Cooter and Gilbert also discuss the possibility of severing portions of the measure in the case of an initiative challenged on single subject grounds after it is passed.  Id. at 878–79.  We believe this could well lead to strategic actions on the part of initiative proponents to use sweeteners to get measures passed and then have those sweeteners removed by the courts.  This seems to defeat the majoritarian purposes that Cooter and Gilbert see for the single subject rule.  It also gives judges more discretion by allowing them to declare the "main purpose" of the initiative and then strike the provisions not in line with that main purpose.

18 Daniel Hay Lowenstein, California Initiatives and the Single Subject Rule, 30 UCLA L. Rev. 936, 967 (1983).

19 Matsusaka & Hasen, supra note 10, at 27–29.

Preferred Citation:  Richard L. Hasen & John G. Matsusaka, Some Skepticism About the "Separable Preferences" Approach to the Single Subject Rule:  A Comment on Cooter & Gilbert, 110 Colum. L. Rev. Sidebar 35 (2010), http://www.columbialawreview.org/sidebar/volume/110/35_Hasen.pdf.