Limited Obstruction: Recent Examples

During the 111th Congress (2009–10), Democrats were triply blessed: their party’s leader, Barack Obama, occupied the White House; they held a commanding 256-178 majority in the House of Representatives; and, most unusually, the size of their Senate caucus at times reached the coveted filibuster-proof 60-40 majority.Footnote ¹ Liberals leveraged their party’s numerical dominance by enacting Barack Obama’s economic stimulus package, funding health insurance for poor and middle-class children, setting aside public lands for wildlife preservation, strengthening protections against pay discrimination, reforming the American financial sector, and passing the Patient Protection and Affordable Care Act. Some of this legislation passed the Senate relatively swiftly and quietly over mild Republican objections, but other bills were subjected to blistering public attacks and concomitant aggressive use of dilatory tactics. Most notably, the Senate had to stay in session for over a month to pass President Obama’s crown jewel, the Affordable Care Act, because filibuster-determined Republicans forced the Democrats to invoke cloture at nearly every opportunity. Even though Democrats had the votes to invoke cloture, delay persisted due to layover requirements and to Republicans’ insistence on consuming the full 30 hours of postcloture debate.Footnote ² While the Affordable Care Act was the most salient demonstration of limited obstruction in the 111th Congress, the Republicans engaged in smaller scale obstruction on several other bills.

The Republicans scored major victories in the 2010 elections, seizing control of the House and substantially reducing the Democratic majority in the Senate. In this political context, expert observers predicted that the ensuing lame-duck session would be uneventful, as Republicans would have an incentive to stall until their newly elected majority in the House was seated (Austin Reference Austin2011). To the surprise of these observers, the lame-duck session proved to be productive. Most notably, the Senate ratified the New Strategic Arms Reduction Treaty (New START) and repealed Don’t Ask, Don’t Tell (DADT). Even though the majority of Republicans voted against these measures, these bills passed without strong resistance or delay. Over the previous two years, Republicans had shown the willingness and capacity to drag floor debates out for many weeks, as they did with the Affordable Care Act and the Dodd-Frank Act. Why didn’t the Republicans generalize this strategy by using obstruction still more aggressively to run out the clock and prevent the Democrats from passing at least one of these two bills?

Puzzles

In addition to the immediate question of why the Republicans did not run out the clock, the 111th Congress in general and its lame-duck session in particular raise two broader puzzles for which existing literature has no satisfactory solution.

1. Variation in the degree of obstruction. The passage of the Affordable Care Act and the Dodd-Frank Act each consumed about a month of floor time. Other controversial items that were opposed by the majority of Republicans, such as a bill to expand the ratification of New START and the Small Business Jobs and Credit Act of 2010, encountered relatively little obstruction and took only a few days each. Early in the session, the Republicans had filibustered several noncontroversial bills that eventually passed unanimously or near unanimously. These bills collectively consumed several weeks of floor time (Mann and Ornstein Reference Mann and Ornstein2016). Why—especially in a unified government with a filibuster-proof Senate—is there so much variation in the degree of obstruction encountered by different bills that is not explained simply by the extent of disagreement in the legislature, the intensity of the preferences of individual legislators, or, in a word, polarization?

2. Unwillingness to adopt streamlining reform. Most observers of the contemporary Senate agree that passage of most legislation requires 60 votes de facto. Even so, the modern filibuster procedure, as laid out by Senate Rule XXII and its associated precedents, is much more complex and time-consuming than a hypothetical streamlined process of simply bringing a bill to the floor, debating it for a limited time, and finally subjecting it to a 3/5 super-majority vote.Footnote ³ Given that the rules of the Senate are endogenously determined under Article 1 Section 6 of the Constitution, why doesn’t the Senate drastically reduce its collective costs of delay by eliminating the holds/filibuster/cloture rules and precedents and replacing them with a simple streamlined 60-vote requirement for final passage?

Literature: Some Hints

Scholarly interest in obstruction in representative assemblies goes back at least as far as Rutherford (Reference Rutherford1914), who found that dilatory tactics were part and parcel of plenary action in the English House of Commons. In a similar vein he made frequent references to strategies of obstruction in the U.S. Senate, Canadian House of Commons, and both the French and Italian Chambers of Deputies. His description of “transient obstruction” is essentially synonymous with our conception of limited obstruction. Bücker (Reference Bücker1989, 243–4) extends Rutherford’s list to include 33 parliaments worldwide and finds forms of obstruction in nearly all of them. Döring’s (Reference Döring and Doring1995) extensive comparative essay titled, “Time as a Scarce Resource: Government Control of the Agenda,” affirms with greater specificity that the main ingredients of our analytic approach are of interest well beyond the U.S. Senate. Similarly, there is no shortage of excellent works on the filibuster, the earliest of which, from 1940, is Burdette’s dated yet still keenly insightful book, Filibustering in the Senate.

Leaping forward in time, Cox and McCubbins (Reference Cox and McCubbins2007), Cox and McCubbins (Reference Cox, McCubbins, Edwards, George and Schickler2011), and Johnson and Lewis (Reference Johnson and Lewis2017) come closest to our checklist of theoretical desiderata when they, like Döring (Reference Döring and Doring1995), emphasize the joint consideration of scarce floor time and agenda setting. However, in contrast with our framework, they do not consider the role of obstruction or any other form of resistance to their agenda setter, and they take variation in the time required to pass bills as exogenous.

Other studies that have some bearing on variation in the degree of obstruction encountered by bills (Puzzle 1) can be grouped according to four focal points in the broader study of Senate behavior: end-of-session effects, party polarization, preference intensity, and signaling.

First, a number of studies contend that delay tactics are most potent near the end of the session (Binder, Lawrence, and Smith Reference Binder, Lawrence and Smith2002; Wawro and Schickler Reference Wawro and Schickler2004; Oppenheimer Reference Oppenheimer1985).Footnote ⁴ Perhaps the time until the end of the session is related to the amount of obstruction a bill encounters. However, it is unclear whether this observation holds in equilibrium. An agenda-setting majority should anticipate that obstruction is more potent near the end of the session and factor this into which bills are proposed and when. It is difficult to discern precisely what this research implies about the degree of obstruction that a given bill encounters.

A second possibility is that obstruction is driven by partisan conflict. Binder, Lawrence, and Smith (Reference Binder, Lawrence and Smith2002) find a strong positive relationship between the frequency of filibusters and the strength of the majority party. Woon and Anderson (Reference Woon and Anderson2012) likewise find a positive relationship between polarization and the delay in the passage of appropriations bills. Koger (Reference Koger, Dodd and Oppenheimer2013) carries this line of reasoning to its logical conclusion: an obstinate minority can delay bills to reduce the majority’s total productivity over the course of the session, and so the minority has an incentive to drag its feet. However, most conceptions of partisan polarization in legislatures suggest that it changes relatively slowly; indeed, polarization is typically measured at the level of an entire session. So, while polarization may explain variation in the time needed to pass bills across sessions, it offers little leverage in explaining variation within a given session.

Third, perhaps obstruction is related to the intensity of preferences of individual legislators. Wawro and Schickler (Reference Wawro and Schickler2006) present a model of the filibuster in the Senate before the adoption of cloture (Rule XXII). In their model, obstruction is a mechanism for a minority to communicate the intensity of its preferences to the majority. In turn, the majority is responsive to intense preferences out of respect for a variety of Senate norms. Broadly consistent with this hypothesis, Black, Madonna, and Owens (Reference Black, Madonna and Owens2011) find that Senators are more likely to issue blue slips (a kind of hold on judicial nominations) when they are ideologically distant from—hence, in intense opposition to—the president. But, if so, why did Republican senators significantly slow down several bills that ultimately passed unanimously but allow the ratification of the New START, which most Republicans opposed, essentially unimpeded?

Fourth, as in Patty (Reference Patty2016), obstruction might offer legislators a costly opportunity to signal their grit or ideological purity to constituents outside of the legislature. We agree that legislators often obstruct in order to send signals to their constituents.Footnote ⁵ However, although Patty’s signaling model provides useful results as to the types of legislators who will be inclined toward obstruction, it is not clear which bills will be obstructed if the legislator is given a choice among many bills. Extending Patty’s model slightly, legislators should choose the bill that is least costly to obstruct. But as we have seen, if preference intensity determines the cost of obstruction, then the models predictions are not consistent with the data.

Collectively, these hypotheses and claims beg the question of why the agenda-setting majority tolerates rules that allow individual Senators to slow the pace of legislation. This is our second puzzle, and its possible solutions can be divided into two categories.

One possibility is that obstructers restrain themselves out of respect for norms and/or out of fear that the rules will be changed. This is the position taken by Overby and Bell (Reference Overby and Bell2004) as well as Wawro and Schickler (Reference Wawro and Schickler2006). It is also considered but rejected by Wolfe (Reference Wolfe2004) in a study of the Japanese Diet. Furthermore, Wawro and Schickler are admittedly and understandably reluctant to apply this logic to the polarized and partisan milieu of the modern Senate. The norms that formed the foundation of their analysis seem to have eroded. Indeed, Sinclair (Reference Sinclair, Dodd and Oppenheimer2013, 22), citing an interview with a Senate staffer, dismisses the role of norms of civility and reciprocity in sustaining unanimous consent agreements in the modern Senate.

The second possible explanation is that a sufficiently large coalition of the Senate prefers the policies that are chosen when limited obstruction is permitted to those that would prevail in its absence (Binder and Smith Reference Binder and Smith1997). Koger (Reference Koger2006) finds that the minority party and those far from their party medians on both sides—that is, those with preferences farthest from the presumptive agenda setter—are consistently less likely to support filibuster reform. We agree with this assessment, but it leaves an important question unanswered. If a requisite majority or supermajority coalition can agree regularly to enact an agenda of desirable legislation, and if scarce time prevents that supermajority from passing all of the bills it would like to pass, why is it unwilling or unable to change the rules so that it can pass legislation more quickly?

Diagnosis

Although existing research on obstruction in the form of dilatory tactics provides a broad base of empirical knowledge on which to build, it is less directly helpful in terms of theoretical guidance. Specifically, we still lack a sufficiently explicit theory that explains why some bills encounter more obstruction than others, and why self-governing organizations such as the Senate exhibit strong reluctance to adopt reforms to streamline their processes.

Our initial diagnosis and subsequent operating premise is that too little effort has been devoted to understanding the strategic position of senators on the losing side of policy-making. When considered at all, proposal-opposing minority factions or parties are portrayed either as potted plants that passively accede to the will of the majority, or as bulls in a china shop that wreak havoc on the majority’s agenda. No model has imbued these presumed losers with procedural rights and strategic sophistication comparable to that attributed to the canonical, ever-influential agenda setter. We uniquely consider the possibility that obstructers, like the agenda setter, try to influence which bills ultimately pass and which bills are effectively obstructed.

Players and Payoffs

• • B is the countable set of all possible bills that the agenda setter, s, can pass if time permits.

• • $u^{s} : B\rightarrow \mathbb {R}$ is a function that maps each bill, b ∈ B, to the utility that the agenda setter, s, accrues if b passes.

• • $u^{o} : B\rightarrow \mathbb {R}$ is a function that maps each bill, b ∈ B, to the utility that the obstructer, o, accrues if b passes.

Strategies

• • A⊆B is the agenda—a set whose members are the bills that pass during the session. This subset is chosen by s.

• • t : B → {1, 2} is the amount of obstruction o engages in if s puts bill b on the agenda. These values are chosen by the obstructer, o.

Game Form

1. 1. The obstructer, o, announces a delay schedule t : B → {1, 2}, which specifies whether each bill will require one or two periods to pass.

2. 2. The agenda setter, s, chooses an agenda A⊆B subject to the constraint ∑ _{b ∈ A} t(b) ≤ 2.

3. 3. Payoffs are awarded. The setter s gets utility U ^s (A; t) = ∑ _{b ∈ A} u ^s (b), while the obstructer o gets utility U ^o (A; t) = ∑ _{b ∈ A} u ^o (b).

We offer three interpretations of the parameters in the model as they apply to the U.S. Senate.

First, although the agenda setter, s, is referred to as a unitary actor both for purposes of empirical motivation and analytic convenience, the game is structured so that s implicitly represents a filibuster-proof coalition that collectively possesses the powers ascribed to the model’s agenda setter. An obvious example of such a coalition is the union of (a) most majority party members with (b) the cloture-invoking requisite number of relatively moderate minority party members. Such a coalition might form, for instance, when, in accordance with Senate Rule XXII, a 3/5 supermajority votes to invoke cloture (i.e., to close debate).

Second, the obstructer, o, can be interpreted as either a lone-wolf obstructer or as the floor manager of a bill-opposing coalition. Like the agenda setter(s), the obstructer(s), too, have readily available party interpretations. It bears emphasis, however, that the model readily accommodates situations of bipartisanship and/or nonpartisanship, too. We will later show how to model o as a coalition of obstructers with different preferences.

Third, B consists of the set of all bills that are passable, meaning that, given sufficient finite time not greater than the duration of the full legislative session, the setter s and coalition she represents can invoke cloture or circumvent any other dilatory tactics and pass the bill. Although a modicum of mind-stretching is required, B can be thought of as all forms of all bills, including potential substitute bills, logrolls, sweeteners, and omnibuses. To keep things tractable, we do not explicitly model bundling and amendment strategies on the floor but instead conceive of these as all possible forms and versions of legislation that are included in B along with all conventional bills as introduced.Footnote ⁸

Finally, no assumption is needed on the structure of the payoffs, u ^s and u ^o . This is an asset because it allows for flexibly diverse interpretations of several different institutional settings. For instance, this is not a spatial model nor is it necessarily a model of parties in legislatures. As such, although the model is amenable to partisan and spatial interpretations, concepts such as party polarization, party discipline, party-based coalition composition, and so on, are not presumed in model or in its interpretation. Similarly, due to the minimal structure on the payoffs, the model can implicitly incorporate the actions of other institutions. The payoff is accrued when the Senate brings the bill to a final vote. Thus, the utility can be interpreted as a lottery over whether the bill passes the House, is vetoed by the president, is implemented faithfully by the bureaucracy, survives judicial scrutiny, and is well received by the electorate. It can also capture the returns from passing the bill in the current session over waiting and trying to pass it in some future session.

Equilibrium for T = 2

To illuminate the core logic of limited-obstructive behavior as an equilibrium phenomenon, first consider the setter’s problem of constructing an optimal agenda conditional on the obstructer’s choice of t. Suppose there are at least two bills that the agenda setter can pass in one period each.Footnote ⁹ The agenda setter contemplates two possibly viable strategies. She can pass two bills, each of which consumes only one period. Or, she can pass one bill that consumes either one or two periods. Therefore, in describing the agenda setter’s best response to an obstructer’s schedule t, three bills are critical: the setter’s two most-preferred bills that can pass in one period each and her single most-preferred bill among all bills.

Formally, let $\bar{b}_1^s$ be the highest utility bill (for s) that can be passed in one period, let $\bar{\bar{b}}_1^s$ be the second highest utility bill that can be passed in one period. Next, let $\bar{b}^s$ be bill that gives the highest utility to the agenda setter out of all bills in B.Footnote ¹⁰ Note that it may be that $\bar{b}_1^s = \bar{b}^s$ . Parsing the notation, the subscript shows how many periods the bill takes to pass, and the superscript shows whose utility is under consideration. The absence of a subscript in $\bar{b}^s$ indicates that the bill may take either one or two periods to pass. One bar denotes the most-preferred bill that the agenda setter can pass in the subscript-denoted number of periods. Two bars means that it is the second most-preferred bill in the specified set.

If the setter prefers passing both $\bar{b}_1^s$ and $\bar{\bar{b}}_1^s$ to passing only $\bar{b}^s$ , then her optimal agenda is to pass the pair of bills that consume one period each. Otherwise, her optimal agenda is to pass only $\bar{b}^s$ , whether that takes one or two periods. Formally:

Now consider the obstructer’s problem of choosing the optimal delay strategy t*. If he obstructs every bill, then the setter spends the entire session of this simple two-period game passing her most-preferred bill. Accordingly, the challenge for the obstructer is to find a pair of individual bills that both he and the agenda setter prefer to spending the entire session passing only the agenda setter’s most-preferred bill.

To derive the equilibrium strategies, it is helpful to define distinct pairs of bills b′ and b″ that, as a pair, are preferred by both players to the setter’s most-preferred of all bills $\bar{b}^s$ (i.e., the highest utility bill for the agenda setter).Footnote ¹¹ Formally,

$\mathcal {C}$ is the set of pairs of bills for which (a) the setter is willing to give up her most-preferred alternative, and (b) the obstructer is willing to facilitate as a compromise by assigning t = 1 to the bills in the pair. For the sake of brevity and mnemonic convenience, we refer to $\mathcal {C}$ as the compromise set.Footnote ¹²

If $\mathcal {C}$ is empty, then the obstructer’s optimal strategy is to obstruct every bill and to anticipate that the setter responds with $a^*(B,t)= \lbrace \bar{b}^s\rbrace$ , her most-preferred bill.Footnote ¹³ If $\mathcal {C}$ is nonempty, however, then the obstructer should choose the pair of bills within the compromise set, $\mathcal {C}$ , that gives him the greatest utility; proclaim a willingness to accept those bills without delay (i.e., assign both of them values of t(b) = 1); and state his intention to obstruct all other bills in B (by setting t(b) = 2). This strategy ensures that the setter opts for inclusion of the obstructer’s favorite pair of bills from $\mathcal {C}$ —namely, $\lbrace \bar{b}_{\mathcal {C}}^o, \bar{\bar{b}}_{\mathcal {C}}^o\rbrace$ —and aborts the setter’s pursuit of her favorite bill, $\bar{b}^s$ . This reasoning is summarized in the following proposition.

Proof. The proof straightforwardly parallels the backward-induction verbal derivation that precedes Proposition 1.

Proposition 1 illustrates that the outcome hinges on $\mathcal {C}$ , the set of compromise agendas. If it is nonempty, then the equilibrium agenda has two bills; otherwise, the equilibrium agenda has only one bill. Each player’s procedural rights and preferences determine the composition of $\mathcal {C}$ . The agenda setter’s right to pass any bill from B implies that she always has the option to pass her most-preferred bill, $\bar{b}^s$ , even if doing so requires two periods. This makes the agenda setter’s most-preferred bill, $\bar{b}^s$ , the benchmark against which all potential compromise agendas must be compared. Every compromise agenda must yield a higher utility to the agenda setter than $\bar{b}^s$ alone, because otherwise she passes $\bar{b}^s$ instead of putting a cumulatively less-liked pair of bills on the agenda. Similarly, every compromise agenda must yield a higher utility to the obstructer than $\bar{b}^s$ alone, because otherwise he would not allow the bills in the compromise agenda to pass in one period each. Finally, because the obstructer selects which bills are allowed to pass in one period, he ultimately chooses which compromise agenda from $\mathcal {C}$ is available to the agenda setter. $\Box$

The Lame-Duck Session of the 111th Congress

Earlier, we highlighted seemingly aberrant behavior in the lame-duck session of the 111th Congress: Senate Republicans allowed the Democrats to repeal DADT and ratify the New START relatively quickly even though they opposed both bills. In the language of the model, they allowed these bills to pass in one period each even though they had the power to make passing either bill take both of the remaining periods in the session. Why?

The model clarifies that there is an important piece of information missing from the exposition of the lame-duck session: B, the set of bills that the Senate Democrats could have passed given sufficient time. One bill in particular stands out as particularly important: the Democracy Is Strengthened by Casting Light On Spending in Elections (DISCLOSE) Act. This campaign finance reform bill was vehemently opposed by Republicans. The Democrats had attempted to bring this bill to the Senate floor earlier in the session. The motion to invoke cloture failed (by only one vote), but so too had the first three attempts to bring the Dodd-Frank Act to the floor. The Democrats had demonstrated their ability to overcome initial setbacks before, and the Republicans had reason to fear that a determined Democratic caucus would be able to pass the DISCLOSE Act over their objections.

Table 1 provides a numerical example that illustrates how the model can explain the perplexing outcome of this lame-duck session. The scenario is one with three bills: the DISCLOSE Act (x), the DADT Repeal (y), and the New START Ratification (z), which the two players rank in opposite order with a modicum of lumpiness in their utility assessments (top panel). Such cardinality can be interpreted as reflecting intensity of preference. All possible two-period agendas are shown in the bottom panel. Each cell entry is the sum of the utilities for the two items on the agenda.

TABLE 1.

Example of Equilibrium in the Two-Period Game*

*The upper-left-most cell entry, u ^s (x) = 6, represents $\bar{b}^s$ , which plays a key role in the algorithms. The bold-faced column {y, z} describes the equilibrium agenda and payoffs. This agenda maximizes the obstructer’s utility u ^o within the three-element set of agendas $A \in \mathcal {C}$ (which in this example happens to be a singleton).

In the example, the basis for comparison is the setter’s most-preferred singleton $\bar{b}^s$ , which is bill x (the DISCLOSE Act). Next we identify the set of compromise agendas, namely the set of 2-tuple agendas that both players prefer to the singleton {x}. The first two of these agendas do not meet this condition; the obstructer prefers passing only x to {x, y} or {x, z}. The third agenda, {y, z} (both the DADT Repeal and the New START ratification), does satisfy the condition, providing a utility of 8 to the agenda setter and –4 to the obstructer. Although the agenda setter is unable to pass her most-preferred bill, she prefers to pass both y and z to only x. Even though he dislikes both of these bills in an absolute sense, the obstructer allows y and z to pass in only one period each, because by doing so he is able to prevent the passage of his most-dreaded bill x.

To review, in equilibrium, the obstructer sets t*(x) = 2 and t*(y) = t*(z) = 1. The agenda setter responds by passing the agenda {y, z}. Substantively, the Senate Republicans would have subjected the DISCLOSE Act to maximum permissible obstruction had it been brought to the floor, but they instead agreed to refrain from subjecting either of the other two bills to excessive delay. While the Senate Democrats would have preferred to pass the DISCLOSE Act to either of the other two bills, they preferred those two bills combined to the DISCLOSE Act on its own. Accordingly, they used the lame-duck session to repeal DADT and to ratify the New START.

The preceding analysis is not intended to be either an empirical test of the model or a comprehensive account of the lame-duck session. Rather, it is intended only to illustrate that the model can help explain otherwise anomalous behavior by drawing attention to abandoned bills that, given sufficient time, could have been enacted. The key feature in our model that is missing from other theories of obstruction is its taking into account not only the pro–con preferences of the players but also their relative intensities of preferences within and across all bills. Such is the essence of optimal limited obstruction.

The example also illustrates another unique and subtle insight of the model: limited obstruction gives the obstructer substantial influence over the putatively setter-monopolized agenda. In the absence of limited obstruction, the agenda setter is able to pass any pair or singleton she likes. But with limited obstruction, the obstructer takes most of the two-bill agendas off the table by imposing delay on most bills. While the agenda setter indeed has the right ultimately to choose the agenda, rationality dictates that she choose from a limited menu that is determined by the obstructer’s execution of optimal delay. The agenda setter’s so-called monopoly agenda setting right is, in fact, dramatically circumscribed by limited obstruction.

Implications

The model shows that equilibrium behavior is driven by the composition of B—those bills that the agenda setter could pass given sufficient time (up to T units). Several implications with real-world empirical referents can be derived directly by examining substantively important properties of B. The first implication addresses the relationship between legislative productivity and the availability of bipartisan bills. A byproduct is revelation of a distinctive form of agenda power that can emerge under limited obstruction.

Proof. $u^s(b) + u^s(\bar{b}^s) > u^s(\bar{b}^s)$ and $u^o(b) + u^o(\bar{b}^s) > u^o(\bar{b^s})$ , so $\mathcal {C} \ne \emptyset$ . By Proposition 1, $\mathcal {C} \ne \emptyset \Rightarrow \exists \; \lbrace b^{\prime }, b^{\prime \prime }\rbrace$ such that t*(b′) = t*(b″) = 1 and $u^s(b^{\prime }) +u^s(b^{\prime \prime }) \ge u^s(\bar{b}^s)$ , so |a*(B, t*)| = 2. $\Box$

To facilitate interpretation of Implication 1, define a bipartisan bill as one from whose passage both players receive positive utility, and define productivity as the number of bills that pass during the session. Under our assumptions, the legislature is productive if it passes two bills in one period each; it is unproductive if one bill consumes both periods. Implication 1 says that the existence of a bipartisan bill is sufficient for high legislative productivity.Footnote ¹⁵

The logic may not be transparent immediately. The obstructer sifts through bills b ∈ B hoping to find a bill other than $\bar{b}^s$ (the setter’s most preferred bill) that both players prefer to no bill at all. Should she find such a bill, legislative productivity is guaranteed in the simple case of T = 2. As an illustration, consider the worst case for the obstructer. Suppose there is only one qualifying b > 0 for both players and that its payoff to the obstructer is only negligibly positive u ^o (b) = ε. In sharp contrast, the obstructer’s payoff under the setter’s most preferred bill $u^o(\bar{b}^s)$ is large and negative. Even in this scenario, the obstructer is better off assigning t = 1 to these bills than delaying either or both of them. The reason stems from the interplay of finite time and limited obstruction. The obstructer can slow but not kill $\bar{b}^s$ by assigning t = 2. However, given that this bill will pass regardless of the obstructer’s strategy, the obstructer may as well expedite the disdained bill’s consideration to make room on the agenda for the compromise bill that gives the obstructer a small but positive payoff. Of course, other compromise agendas may exist that are better for the obstructer than this purposely pessimistic example, and a rational obstructer would steer the setter into selecting one of these. But the more immediate point is that as long as at least one compromise agenda exists, the legislature will be productive. In summary, these conditions for legislative productivity under limited obstruction seem fairly weak.

The second implication begins to untangle the relationship between polarization and gridlock. Define polarization as the extent to which players’ preferences for or against bills are qualitatively opposite one-another. That is, if the setter is in favor of passing a bill (u ^s (b) > 0), then the obstructer is against it (u ^o (b) < 0), and vice versa. Suppose the obstructer opposes every bill in B, while the setter favors every such bill. In other words, the body is highly polarized.Footnote ¹⁶ Conventional wisdom seems strongly to suggest obstruction (t = 2) and/or gridlock occurs in the presence of such high polarization. In contrast, our model shows why this is not necessarily the case.

Proof. See the counterexample described in Table 1. $\Box$

Current research makes strong, confident, rarely contested claims about a strong positive relationship between polarization (of parties and preferences) and gridlock (low or no legislative productivity). Implication 2 cautions against making such sweeping generalizations. In effect, it shows that, even in an extreme case in which the obstructer (alternatively, a sizable minority coalition) dislikes any and all possible bills that the agenda setter (backed by a majority coalition) would like to pass, the obstructer does not necessarily use gridlock-enhancing obstruction.

With a stronger notion of polarization based on preference intensity, however, we can obtain an implication of the model that comports better with conventional wisdom. Suppose players’ preferences were mirror images of one another bill-by-bill without exception, so u ^o (b) = −u ^s (b)∀b ∈ B. Clearly, this is an extreme conception of zero-sum polarization; the obstructer abhors each bill precisely as much the agenda setter adores it. The important analytic effect of the condition is that it renders crucial compromise set $\mathcal {C}$ empty. As such, no productive negotiation, reciprocity, or compromise between players is possible. The resulting equilibrium agenda is a singleton that takes two periods to pass, hence the expected result: polarization (of an extreme, perfectly counterbalancing form) leads to gridlock.

Proof. For any b′ and b″, $u^o(b^{\prime }) + u^o(b^{\prime \prime }) \ge u^o(\bar{b}^s) \Rightarrow u^s(b^{\prime }) + u^s(b^{\prime \prime }) \le u^s(\bar{b}^s)$ . Thus, $\mathcal {C} = \emptyset$ , which by Proposition 1 implies that t*(b) = 2 for all b ∈ B. $\Box$

Implications 2 and 3 illustrate that relative intensities of preferences are crucial elements of the model. Our final implication elaborates under more realistic conditions. A key, more specific feature for determining whether gridlock or legislative productivity prevails under specified preference conditions is the abhorrence level of the obstructer with respect to the setter’s most-preferred bill, $\bar{b}^s$ . Implication 4 states that legislative productivity is weakly increasing in the intensity of the obstructer’s disdain of the agenda setter’s most-preferred bill $\bar{b}^s$ . In other words, the more baneful $\bar{b}^s$ is to the obstructer, the greater is his desire to find a pair of bills from the compromise set that can be passed without delay. Formally:

Proof. Define $\mathcal {C}(B)$ as the compromise agenda set associated with B, and define $\mathcal {C}(\tilde{B})$ analogously. Consider b′, b″ ∈ B such that $\lbrace b^{\prime }, b^{\prime \prime }\rbrace \in \mathcal {C}(B)$ and $\bar{b}^s \not\in \lbrace b^{\prime },b^{\prime \prime }\rbrace$ . $u^s(b^{\prime }) + u^s(b^{\prime \prime }) \ge u^s(\bar{b}^s) = u^s(\tilde{b}^s)$ and $u^o(b^{\prime }) + u^o(b^{\prime \prime }) > u^o(\bar{b}^s) > u^o(\tilde{b}^s)$ , so $\lbrace b^{\prime }, b^{\prime \prime }\rbrace \in \mathcal {C}(B) \Rightarrow \lbrace b^{\prime }, b^{\prime \prime }\rbrace \in \mathcal {C}(\tilde{B})$ . Now suppose $b^{\prime \prime } = \bar{b}^s$ . Then trivially $u^s(b^{\prime }) + u^s(\bar{b}^s) \ge u^s(\bar{b}^s) \Rightarrow u^s(b^{\prime }) + u^s(\tilde{b}^s) \ge u^s(\tilde{b}^s)$ and $u^o(b^{\prime }) + u^o(\bar{b^s}) > u^o(\bar{b}^s) \Rightarrow u^o(b^{\prime }) + u^o(\tilde{b}^s) > u^o(\tilde{b}^s)$ , so $\lbrace b^{\prime }, \bar{b}^s\rbrace \in \mathcal {C}(B) \Rightarrow \lbrace b^{\prime }, \tilde{b}^s\rbrace \in \mathcal {C}(\tilde{B})$ . These conditions exhaust all elements of $\mathcal {C}(B)$ , so for every element in $\mathcal {C}(B)$ , there is a corresponding element in $\mathcal {C}(\tilde{B})$ . From Proposition 1, it is clear that |a*(B, t*)| = 2 if and only if $\mathcal {C} \ne \emptyset$ , which combined with the previous statement implies $|a^\ast (B,t^\ast )| \le |a^\ast (\tilde{B}, t^\ast )|$ . $\Box$

Informally, this comparative static characterizes the change in obstruction due to changes in the intensity of the obstructer’s distaste for the agenda setter’s most-preferred bill or top priority. The more the obstructer dislikes this bill, the larger is the set of compromise agendas. Since legislative productivity (here meaning passing two bills rather than just one) is high if and only if the set of compromise agendas is nonempty, this implies that legislative productivity is weakly increasing in how intensely the obstructer dislikes the agenda setter’s most-preferred bill.

Multiple Obstructers

The game readily extends to instances with n obstructers.

1. 1. Obstructer o ₁ announces a delay schedule t _i (b) : B → {1, 2}.

2. 2. Repeat Step 1 for i = 2, . . ., n.

3. 3. The setter, s, chooses agenda A⊆B subject to the constraint ∑ _{b ∈ A} max _i t _i (b) ≤ 2.

4. 4. Payoffs are awarded. The setter, s, gets utility U ^s (A; t) = ∑ _{b ∈ A} u ^s (b), while obstructer o _i gets utility U ^o_i (A; t) = ∑ _{b ∈ A} u ^o_i (b).

This game is the same as the original game except that a bill requires two periods to pass if any of the obstructers decides to slow down the passage of that bill. Setting t(b) = max _i t _i (b), the agenda setter’s best response function is the same as before.

The final obstructer, o _n , faces a strategic situation similar to the obstructer’s in the original game, except that he operates under the constraint that $t(b) \ge {\max \atop i = 1, \dots , n - 1} t_i(b)$ . Qualitatively, he behaves just as the obstructer behaves in the original game if some subset of bills automatically required two periods to pass.

Formally,

Intuitively, some bills cannot be used to construct compromise agendas because other obstructers have already committed to obstruct them. If any agendas that both o _n and s prefer to $\bar{b}^s$ in two periods remain by the time o _n takes his turn, then he obstructs all bills except for the two in his most-preferred compromise agenda. Otherwise, he obstructs every bill. The preceding obstructers, of course, must anticipate o _n ’s behavior. Collusion or cooperation is not presumed among obstructers. They must obstruct bills that are unacceptable to them, but they must beware of the possibility that o _n does not tolerate a compromise agenda that leaves him worse off than passing only $\bar{b}^s$ .

Iterating this logic yields the subgame perfect equilibrium strategies. If there are any compromise agendas that are acceptable to o _i and all subsequent obstructers, then o _i obstructs all bills except her most-preferred agenda in this set. Otherwise, he obstructs every bill. The outcome can be inferred from the strategy of o ₁. If there is at least one compromise agenda that the agenda setter and every obstructer prefers to the $\bar{b}^s$ in two periods, then the first obstructer’s most-preferred agenda in this set is the realized agenda. Otherwise, the agenda is $\lbrace \bar{b}^s\rbrace$ .

Formally, o ₁’s strategy (as well as the equilibrium schedule of delay) can be written as follows:

Thus, the multiple obstructer equilibrium is almost identical to the original equilibrium. The only difference is that compromise agendas must be acceptable to every obstructer. Implications analogous to those presented above therefore hold in the multiple obstructer case as well.

Additionally, the extension to multiple obstructers clarifies the relationship between the size of the compromise agenda set (and, by extension, the prospects for passing two bills) and the number of obstructers.

The proof is trivial from the equilibrium of the multiple obstructer game. Perhaps counterintuitively, this suggests that the prospects for passing two bills are better when the minority party acts like a unified actor.

Our model, therefore, begins to solve our first of two puzzles: Why is there so much variation in the degree of obstruction encountered by different bills? Obstructers deliberately and predictably create this variation to induce the agenda setter to select an agenda that is more to their mutual liking.Footnote ¹⁸

Streamlining Reforms

This leaves the second puzzle: Given that the Senate can change its own rules, why does a coalition that is large enough to pass legislation over a filibuster allow obstructers the right to delay the passage of legislation? Why doesn’t the Senate replace Rule XXII with a simple 60-vote requirement for passage of Senate bills? In the terms of the model, such a reform would remove the first step of the game and instead set t(b) = 1 for every bill.

Analysis of the possibility of reform is uninteresting for a two player game, so we consider a slightly modified game with three players. The players are a liberal agenda setter s, a moderate m, and a conservative obstructer o. Player s requires m’s support in order to pass bills (equivalently, all bills in B satisfy the condition u ^s (b) > 0 and u ^m (b) > 0). Because this requirement can be cast as a constraint to the space of parameters, it does not affect the equilibrium strategies or the game form. Players m and o have the ability to obstruct. Finally, s needs the support of either o or m to change the rules, so the support threshold for changing the rules is the same as that for passing bills. Their votes to streamline the rules by prohibiting limited obstruction are given by r _m and r _o , where one indicates a vote in favor the reform and zero indicates a vote against. We assume that s votes for the reform because it always makes him weakly better off.

Formally,

1. 1. o announces r _o ∈ {1, 0}

2. 2. m announces r _m ∈ {1, 0}

3. 3. If r _o + r _m = 0, then o announces t _o :B → {1, 2}

4. 4. If r _o + r _m = 0, then m announces t _m :B → {1, 2}

5. 5. If r _o + r _m = 0, then s chooses A⊆B subject to ∑ _{b ∈ A} max{t _o (b), t _m (b)} ≤ 2. Otherwise, s chooses A⊆B subject to |A| ≤ 2.

To make the example concrete, suppose there are only three bills—x, y, and z—and the three aforementioned players whose preferences are given in Table 2.Footnote ¹⁹

TABLE 2.

Payoffs in a Three-Player Obstruction Game

Note that the setter s and the moderate m have similar preferences. Their most-preferred bill is x and they both assign positive utility to all bills. The more extreme obstructer, o, on the other hand, dislikes x and y and likes only z. The equilibrium outcome of the subgame in which obstruction is permitted is {x, z}, because this is the only agenda that all three players prefer to x alone (i.e. {x, z} is the only pair in $\mathcal {C}$ ). The equilibrium outcome of the subgame in which limited obstruction is forbidden is {x, y}, the pair that is most appealing to s. Because m and o both prefer {x, z} to {x, y}, neither votes to change the rules and limited obstruction is the equilibrium institution.

This example falls short of a full-fledged theory of endogenous institutional choice, but it offers a helpful lead as to how the filibuster could survive internal reform attempts, which it has many times since the Senate’s adoption of Rule XXII in 1917. Moreover, the configuration of preferences and the resulting agenda in the example are intuitively plausible. The bill most-preferred by both the agenda setter and the moderate passes even though obstruction is permitted. However, limited obstruction essentially eliminates from the agenda the setter’s second-highest priority (which the obstructer dislikes and which the moderate finds only slightly appealing) and replaces it with a bill that appeals to all players. In this example, limited obstruction is aggregate welfare enhancing, and the only net beneficiary of the streamlined procedure is the agenda setter.

In our repeatedly useful case of the 111th Congress, for example, a liberal agenda setter might enact universal healthcare and create a cap-and-trade system in a no-obstruction Senate. Because of obstruction, the liberal agenda setter must settle for enacting universal healthcare and cutting taxes on the middle class. This is to the advantage of conservatives and moderates, both of whom prefer the tax cuts to cap-and-trade. Therefore, the conservative and moderate band together to prevent any changes of the rules. Critically, this does not require the moderate liberals to actively obstruct the agenda of their own party. Recall that there is an equilibrium in which the first obstructer (whom we suppose is the conservative) is the one who effectively sets t, and in doing so, he anticipates the preferences of the other obstructers. The conservatives can (and happily will) do all of the actual obstructing, secure in their knowledge that the moderates support them if the extreme liberal agenda setter attempts to eliminate the filibuster.

The conditions necessary to sustain the endogenous institution of limited obstruction can be stated precisely. For expositional simplicity, assume there are at least two bills that provide the agenda setter with positive utility. Let $\mathcal {C}_o = \lbrace \lbrace b^{\prime }, b^{\prime \prime }\rbrace \;:\; u^i(b^{\prime }) + u^i(b^{\prime \prime }) > u^i(\bar{b}^s) \text{ for $i$ = $o$, $m$, $s$} \rbrace$ , the set of compromise agendas from the perspective of the first obstructer. Let $\lbrace \bar{b}_{\mathcal {C}_o}^o, \bar{\bar{b}}_{\mathcal {C}_o}^o\rbrace = {{\rm argmax} \atop \lbrace b^{\prime }, b^{\prime \prime }\rbrace \in \mathcal {C}_o }\; u^o(b^{\prime }) + u^o(b^{\prime \prime })$ , o’s most-preferred pair in the compromise set.

Proof. Because either m or o can change the rules by voting in favor of the rule change, limited obstruction survives in the subgame perfect Nash equilibrium if and only if both m and o prefer the subgame where the rules are not changed. From the analysis of the multiple obstructer game above, (1) and (2) describe the conditions under which both m and o prefer the subgame where the rules are not changed to the subgame where the rules are changed. $\Box$

The proposition states that limited obstruction as an endogenous institutional feature survives if it induces an agenda that both m and o prefer to the agenda that would result from giving s unfettered agenda control (the agenda setter’s two most-preferred bills).

Finitely Long Agenda

The model can also be extended to cases where the number of periods in the session, T, takes an arbitrary but known value. The details of this extension are complicated because the players’ equilibrium strategies must be characterized algorithmically rather than in closed form, but the logic of their strategies are straightforward extensions of their strategies in the two period case. We assemble and explicate the specifics in the Online Appendix and, therefore, provide only a brief sketch of them here.

The agenda setter iteratively applies the logic of her two-period strategy; she decides whether she prefers to pass the two best one-period bills or the single best two-period bill. If she prefers the pair, she adds the better member of the pair to the agenda; if she prefers the singleton, she adds the singleton to the agenda. She then repeats this procedure until the time required to pass her concatenated agenda equals the session length.

The obstructer, in contrast, optimizes by constructing a generalized version of the compromise set. He considers all agendas that are politically feasible (i.e., that are the agenda setter’s best response for some choice of t) and chooses t to induce the agenda setter to pass his most-preferred strategy in this compromise agenda set. The key to identifying whether an agenda is politically feasible is that the agenda setter must prefer the two worst bills (from the perspective of s) that pass in one period to the best two-period bill that does not make it onto the agenda.

This more general game clarifies an important point about obstruction. Dislodging the agenda setter’s most-preferred bill is difficult. Doing so requires large number of attractive pairs, because the agenda setter must prefer least attractive of these pairs to her most-preferred bill. Limited obstruction typically does not prevent the agenda setter from passing her most-preferred bills. It more commonly replaces her secondary priorities with compromise pairs that are attractive to the agenda setter but not so distasteful to the obstructer.