Article

The Pivotal Politics of Temporary Legislation

I.     Introduction

Uncertainty is an unavoidable part of any democratic lawmaking process. Congress’s ability to change laws has broad implications. The behavioral effects of the law depend on how citizens and businesses expect the law to stand in the future. The long-run fiscal and macroeconomic implications of the law depend on its political stability. This political uncertainty is particularly salient with respect to temporary legislation—law that expires on a given date. Whether temporary legislation is renewed (and just as importantly, whether sophisticated citizens and businesses expect such legislation to be renewed) is key to understanding its behavioral effects and long-run policy implications. Before we can evaluate the desirability of any piece of temporary legislation, we must first understand the political uncertainty surrounding its renewal.

For example, consider the temporary capital gains rate cut in 2003.1 How did taxpayers respond to that cut? Was that response different because the rate cut was temporary? Did the temporary rate cut raise or lose revenue? Did revenue projections accurately or inaccurately reflect its cost? All of these inquiries are affected by uncertainty regarding the renewal of this provision. One cannot determine whether the temporary rate cut was desirable without first understanding the likeliness of its renewal. This Article fills a gap in existing literature by providing a theoretical and empirical framework for exploring that uncertainty.2

This framework relies on two key empirical findings in the political science literature. First, there are a number of key legislators (including the 60th senator,3 the President, and the median member of the House majority) whose preferences are particularly important in determining whether legislation will be enacted.4 Second, the tools of modern political science allow us to meaningfully compare legislators across Congresses.5 The intuition is that the renewal of temporary legislation can be explored by combining information regarding how legislators voted when the temporary legislation was originally enacted and how key legislators have changed in the interim. For those legislators who were not in Congress when the temporary legislation was originally enacted, we can predict how they will likely vote on renewal by reference to how their colleagues voted in the past.

This Article makes two primary contributions. First, it models the renewal process to yield the key insight that renewal does not depend on the temporary policy itself. Rather, renewal depends on the underlying permanent policy and its acceptability to certain key legislative actors.6 Thus, renewal can be studied by investigating how the preferences of these actors have changed between enactment and “sunset.”7 Second, the Article applies the model to several examples by empirically measuring these shifts in preferences. To take two concrete examples, this approach illuminates both the nonrenewal of the Assault Weapons Ban in 20048 and the renewal of the capital gains rate reduction in 2006.9

Although the primary goal of this Article is positive, there are some natural normative implications. Prior work (particularly in the context of revenue estimates) has implicitly treated all pieces of temporary legislation as if they have the same renewal-uncertainty. These estimates assume either that all temporary legislation will expire (the “current law” approach) or that all temporary legislation will be renewed (the “current policy” approach). However, various pieces of temporary legislation have (measurably) different renewal-uncertainty. This heterogeneity should influence how these provisions are evaluated—specifically, their revenue cost and their expected effect on taxpayer behavior. Incorporating renewal-uncertainty into these estimates could potentially improve their accuracy.

This Article proceeds in the following order: Part II presents the theoretical model. It describes the set of key legislators whose preferences are determinative of whether legislation can be enacted. It then argues that the renewal of temporary legislation depends on whether the underlying permanent policy, to which the law would default if the temporary legislation expired, is acceptable to these key legislators. Part III moves from theory to application. It presents an empirical approach to measure the uncertainty surrounding the renewal of temporary legislation and explores the normative implications. Part IV discusses whether renewal legislation will reliably reflect changing legislative preferences. The model suggests this will only be true in limited circumstances. The broader implication is skepticism toward the conventional wisdom that Congress should intentionally deploy temporary legislation as a responsive policy tool. Part V concludes.

II.     Modeling the Renewal of Temporary Legislation

Under the U.S. Constitution, legislation generally requires the approval of majorities in the House of Representatives and Senate, as well as the consent of the President.10 Given there are over 500 lawmakers with distinct preferences, figuring out whether legislation will be enacted may seem a Herculean task. However, the political science literature has established that we can focus on a few key legislators called “pivots.”11 Part II.A shows the preferences of these pivots are determinative with respect to the procedural veto gates in the U.S. legislative process. Indeed, legislation can be enacted if all of the pivots agree to change policy in the same direction. (Readers already familiar with pivot models may wish to skip to Part II.B.)

Part II.B considers the application of pivot models to temporary legislation. Temporary legislation involves both a temporary policy and an underlying permanent policy. This Article argues that the renewal of temporary legislation depends on whether the pivots agree on the undesirability of the underlying permanent policy. Counterintuitively, renewal does not depend on the temporary policy itself, but rather the underlying permanent policy and its acceptability to the pivots.12

A.     Modeling the Legislative Process with Pivots

Pivot models simplify the legislative process by focusing on a limited number of pivots. The preferences of these pivots determine whether the requisite legislative majorities (or supermajorities) support proposed legislation. This is best illustrated with an example. Consider the top marginal tax rate,13 and for the moment, focus on one part of the legislative process. Is there enough support in the Senate to pass a bill increasing the rate? Senate Rule 22 gives each senator the right to filibuster legislation subject to a three-fifths supermajority vote to end debate (“cloture”).14 A clumsy way to answer that question is to consider each senator in turn and count how many would vote on a proposed rate increase.

A more elegant approach first organizes the senators based on their ideal top rate. To simplify matters, consider a 12-member Senate with the following top-rate preferences.15

 

Table 1. Preferred Top Rates of Senators

Oh1 

To figure out whether a rate increase can pass the Senate, we can look simply at what Senator H wants. Why is that? Senator H is the last (or marginal) legislator whose agreement is required for any increase in the top rate. If Senator H supports the rate increase, Senators A to G will also support the increase because their preferred rates are even higher. Senators A to H have sufficient numbers to overcome a filibuster. On the other hand, if Senator H does not support the rate increase, then Senators H to L will manage to sustain a filibuster and prevent the bill from moving forward.

Whether a rate increase can overcome a filibuster in the Senate depends entirely on Senator H. Senator H is therefore called the “filibuster pivot.”16 Senator H is pivotal because of the ordering of legislator preferences, not because of any grant of power specific to her. On a different issue (where the order of senator preferences is different), the filibuster pivot may be a different legislator.

Returning to the original question, can a rate increase avoid filibuster in the Senate? We know that we have to focus on Senator H, the filibuster pivot. She will only support a rate increase if the current rate is too low (i.e., if the current rate is below 25%).

The basic approach of pivot models is to take each step of the legislative process and reduce it to the preferences of a particular legislator. The fact that bills must generally avoid filibuster is reduced to the preferences of the 60th senator. Since there are 100 senators, the 60th senator’s preference is pivotal in achieving the three-fifths supermajority necessary to close debate in the Senate. Similarly, other steps of the legislative process are reduced to the preferences of particular lawmakers. In order to enact legislation, both the House and the Senate must pass the bill.17 The President can either sign the bill or veto it.18 If the President vetoes the bill, the House and Senate can override the veto with a two-thirds supermajority vote.19

Whether a bill will clear any particular procedural hurdle can be determined by reference to the preferences of a single legislator. Will the bill pass the House? Consider the preferences of the median member of the House. Will the bill be vetoed? Look at the preferences of the President. If there is a veto, will the House and Senate be able to overcome that veto? Look at the preferences of the 34th senator and the 146th representative.20

 

Table 2. Steps in the Legislative Process and Corresponding Pivots

 Oh4 

The political science literature debates whether there are any additional pivots—whether there are any additional procedural veto gates. I discuss one here because it has significant support in the literature and because it more explicitly incorporates the influence of parties.

In their influential book on the role of the House majority, political scientists Gary Cox and Mathew McCubbins argue persuasively that there is an additional procedural step in the House.21 The House majority party exercises significantly more influence because bills are sometimes considered under more restrictive rules in the House than in the Senate. By controlling what gets onto the agenda, the majority party can avoid votes that “roll” the party—votes where a majority of the party is on the losing side. This may square with many readers’ intuition. How often does the House majority party get rolled because its members defect and vote with the minority party? Such a situation is extremely rare.22

But how does the majority party prevent rolls? Cox and McCubbins postulate that the power of the majority party is exercised through “negative agenda power.”23 Cox and McCubbins agree that under most circumstances, it is reasonable to doubt the ability of the majority party to force members to vote against their personal preferences.24 They argue, however, that parties can still exercise influence by controlling what gets onto the agenda.25 According to this theory, the majority party in the House exercises its power by controlling and disciplining a small number of legislators who have “special agenda-setting powers.”26 By controlling what gets on the agenda, the majority party ensures that only bills shifting policies towards the party median are enacted, even if rank-and-file legislators vote independently of the party.

Negative agenda power is easily incorporated into a pivot model. Pivot models reduce each veto gate to the preferences of a pivot. Here, that pivot is the median member of the House majority party.

This highlights a larger point. If different assumptions are made about the influence of parties (e.g., assuming that the Senate majority party can influence outcomes,27 or that minority parties influence pivots28) or the influence of committees,29 these assumptions could be incorporated by adding pivots. There is an ongoing argument in the political science literature about whether various pivots should be included.30 I will focus on the pivots listed in Table 2 and the majority-median pivot because this set of pivots has the strongest support in the political science literature. However, the discussion of temporary legislation that follows does not depend on a particular set of pivots and is generalizable.31

1.     When Is Legislative Action Possible?

Each pivot determines whether a hypothetical bill can pass a particular legislative hurdle. By combining all of the pivots and comparing those pivots to current law, one can determine whether legislation can be enacted. For example, assume the pivots have the following top-rate preferences.

 

Table 3. Preferred Top Rates of Pivots

Oh5

 

Whether legislation can be enacted depends on the status quo (the current tax rate) and its position relative to the pivots.

Example 1: Centrist Status Quo. In Example 1, assume the current tax rate is 35%. What would happen if a bill were proposed to increase the tax rate to 40%? The President would like that change because it would bring policy in line with his preferences. However, that bill would likely never reach his desk. Moving the tax rate from 35% to 40% would move policy away from two pivots: the House majority-median and the filibuster. If the House majority-median pivot opposes a bill, it means the House majority party will exercise negative agenda control so the bill never reaches a vote in the House. If the filibuster pivot opposes a bill, it means there are at least 41 senators who will filibuster. It is unlikely a bill increasing the tax rate could pass the House or Senate.

What would happen if a bill were proposed to decrease the tax rate to 30%? This bill could pass the House.32 However, the President would veto it. The veto-override pivots also oppose the bill. Any tax decrease would be vetoed by the President, and Congress would not override the veto.

In fact, there is no bill that can successfully displace the status quo. This is most easily demonstrated visually. Figure 1 maps the status quo and the pivots’ preferred rates. Any bill proposing to increase rates will make the filibuster and House majority-median pivots worse off. Those bills will not make it through Congress. Similarly, any bill that proposes to decrease rates will make the President and the veto-override pivots worse off. Those bills will be vetoed.


Figure 1. Pivots Relative to Status Quo of 35%

 Oh6

 

Thus, centrist status quos result in legislative inaction. Such status quos are inside what is called the “gridlock zone.”33 The gridlock zone is the set of policies between the preferences of the pivots.34 When the status quo is inside the gridlock zone (i.e., when current law is between the pivots), the model predicts no legislative action.35 In this example, the gridlock zone stretches from 25% to 40%. If the current rate falls in this range, legislation is unlikely. Any rate increase will not pass the House. The President will veto any rate decrease.

Example 2: Moderate Status Quo. On the other hand, when the status quo falls outside of the gridlock zone, new legislation can be enacted.36 In Example 2, assume the status quo is moderately conservative—the current rate is 20%. Legislative action is possible because all of the pivots want a rate increase. Figure 2 maps the pivots with the status quo of 20%. The gridlock zone is indicated by the hashed rectangle. As seen in Figure 2, the current rate falls outside of the gridlock zone, so legislation can be enacted.


Figure 2. Pivots Relative to Status Quo of 20%

Oh7

 

But what range of rates is possible? Could the rate be moved to the President’s preferred rate of 40%? Probably not. In determining how much the rate can be changed, the most important pivot is the House majority-median. Although that pivot would like to see an increase in the rate, it is unlikely that she would prefer 40% to the current rate of 20%.37 Since the current rate of 20% is already close to what that pivot wants, it is unlikely that a significant rate increase can pass through the House.

This provides a general rule of thumb: Where the status quo is relatively close to the gridlock zone, the pivot closest to the status quo determines how far policy can move.

Example 3: Extreme Status Quo. In contrast, if current law is far from the gridlock zone, the range of potential legislation increases. In Example 3, assume the status quo is extremely liberal—the current tax rate is 80%. As in Example 2, the status quo falls outside of the gridlock zone, and thus legislation can be enacted. However, in contrast to the prior example, there is a wide range of bills that can displace the status quo. For example, a 50% rate would have the requisite support of the House, Senate, and the President. A 40% rate or a 30% rate would also receive the requisite support.

What rate actually is enacted depends on who chooses the bill. If the President specifies the bill, then his preferred rate of 40% will likely prevail. If the House majority party specifies the bill, then a rate close to the median preference in that party (25%) will likely prevail.

More realistically, the President, the leadership in the House and Senate, and key committee members will all influence the outcome. The further the status quo falls outside the gridlock zone, and the more the pivots disagree, the wider the range of potential legislation.

As Examples 2 and 3 demonstrate, the extent to which policy can shift depends on how close the status quo is to the gridlock zone. The further the status quo is from the gridlock zone, the further policy can shift. The location of the status quo relative to the pivots is therefore critical to whether any legislation will pass, and what the content of that legislation will be.

2.     Shifting Pivots and Shifting Policies

Pivot models focus on the preferences of pivots relative to current law. Legislative action is possible when all of the pivots want to move policy in the same direction. This raises the question of how the status quo and pivots change. The status quo is determined by existing legislation but can shift as economic, demographic, and other factors change. Sometimes an exogenous shock will cause a discontinuous change to the status quo (e.g., the financial crisis). The status quo can also change incrementally as factors gradually shift (e.g., inflation).

Pivots move when elections replace legislators38 and when legislators’ preferences change.39 These preferences are sometimes influenced by new evidence regarding the effectiveness of current policy, information regarding constituency preferences, or the influence of lobbyists and special interests.40 Consider, for example, the rightward shift in preferences regarding top marginal rates over the past half-century.41 Politicians across the political spectrum began to prefer lower rates due to smaller deficits, increases in business lobbying, and mounting evidence of the efficiency costs of high tax rates.42

Shifting pivots provide a succinct explanation for the legislative “bump” often observed when a new President is inaugurated.43 At the beginning of many presidencies, there is a temporary increase in legislative activity that eventually tails off.44 Elections reveal information about constituency preferences (leading to shifts in legislator preferences) and also replace some legislators. This shifts the pivots. Certain policies that were within the gridlock zone prior to the election fall outside of the new gridlock zone after the election.45 New legislation replaces these status quos.

Once these policies are addressed, however, legislative action becomes much more difficult. This will be true even if there is a unified government (where one party holds both the presidential office and majorities in both houses of Congress).46 Absent some exogenous shock, new information, or significant change in the legislature, status quos will remain within the gridlock zone, and passing additional legislation will be difficult.47

However, not all status quos are sticky. Pivot models provide a lens for understanding the history of the alternative minimum tax (“AMT”), where the status quo has not been sticky due to inflation. The AMT was passed to ensure all taxpayers were paying a minimum amount of tax and not using deductions to reduce their tax liabilities to unreasonable levels.48 In calculating AMT liability, the taxpayer adds back certain deductible items to taxable income and then applies a reduced tax rate.49 If the AMT liability is higher than the taxpayer’s regular income tax liability, then the taxpayer pays the difference in addition to their regular income tax.50 Since the AMT was only intended to target high-income taxpayers, it included an exclusion that was intended to prevent the AMT from affecting moderate- and low-income taxpayers.51 However, the exclusion amount (as enacted in 1993) was not indexed for inflation.52

The exclusion was set at a level reflecting 1993 legislative preferences regarding the appropriate exclusion in real dollars. Assume for the sake of simplicity that legislative preferences regarding the AMT exclusion (in real dollars) did not change and the pivots therefore did not move. What happened? The status quo shifted each year due to inflation. The exclusion for an unmarried individual was set to $33,750 in 1993.53 In 1994, adjusted for inflation, that exclusion was $32,907.54 Table 4 presents the real-dollar decline in the exclusion due to inflation from 1993 to 2001.

 

Table 4. AMT Exclusion in 1993 Dollars from 1993–200155

Oh8 

Pivot models provide one explanation for why the AMT was not “patched” for the first eight years after it was enacted despite inflation averaging about 2.6% between 1993 and 2000. Inflation gradually moved the status quo. But, it was not until 2001 when the exclusion would have been only $27,537 (in 1993 dollars) that the status quo shifted far enough to no longer fall inside the gridlock zone.56 This shift is demonstrated in Figure 3.

Figure 3. AMT Exclusion Shifting Due to Inflation

 Oh9

 

In 2001, Congress passed the first temporary patch of the exclusion amount, moving the exclusion from $27,537 to $29,169 (in 1993 dollars).57 It is interesting to note how little the 2001 patch moved the exclusion. Recall that the model predicts that the closer the status quo is to the gridlock zone, the less policy can be shifted. This squares with the narrative that in 2001, the status quo had barely moved outside of the gridlock zone. The 2001 patch was scheduled to last from 2001 until 2004, but was similarly not indexed for inflation. For the next two years, inflation shifted the new status quo until 2003, when the exclusion would have been only $28,705 (in 1993 dollars). In 2003, Congress passed a second temporary patch of the exclusion amount, moving it to $31,609 (in 1993 dollars).58

Judging from this sequence of patches, one can estimate the low end of the gridlock zone is roughly $28,750 (in 1993 dollars). When the exclusion fell below this level, Congress consistently acted.59 It is not surprising that Congress continued to patch the AMT given how low the exclusion would otherwise have been. For example, if Congress had failed to patch the AMT in 2012, the exemption would have been only $21,241 (in 1993 dollars), far outside the gridlock zone.60

B.     Under What Conditions Will Temporary Legislation Be Renewed?

The pivot model literature predominantly focuses on permanent legislation and therefore focuses on the position of a single status quo relative to the gridlock zone.61 This Part extends these models to temporary legislation. It is possible to conceptualize temporary legislation as involving two distinct status quos. Prior to expiration, the relevant status quo is the temporary policy. However, when expiration is imminent, the relevant status quo is the underlying permanent policy. Thus, the decision to amend existing temporary legislation and the decision to renew expiring temporary legislation occur in the context of very different status quos.

1.     Renewal Depends on the Underlying Permanent Policy

Ultimately, it is the acceptability of the underlying permanent policy that determines whether temporary legislation is renewed. Whenever the AMT patch was due to expire, it was the underlying permanent policy—the unindexed 1993 exclusion—that determined the renewal of the patch. This is true even when the temporary legislation is never allowed to actually sunset (as in the AMT example). The key question in renewal is whether the underlying permanent policy is outside the gridlock zone when expiration is imminent.

The underlying permanent policy can take two different forms. Sometimes temporary legislation supersedes permanent legislation. For example, the first Bush tax cut temporarily reduced the income and estate tax rates.62 Upon expiration of the temporary legislation, those tax rates would have reset to the rates set by underlying permanent law.63 In other situations, temporary legislation writes on a blank canvas. Of course, this does not mean there is no underlying permanent policy. For example, the research and development (“R&D”) tax credit is a temporary tax provision that does not have underlying permanent legislation.64 If the temporary legislation expires, the credit disappears. The underlying permanent policy is an (implicit) zero percent tax credit for R&D.65

A few examples will illuminate how the underlying permanent policy influences the renewal process.

Example 4: Stable Pivots Lead to Renewal. Assume that the pivots have the tax rate preferences listed in Table 3 and that the current rate is 55%. The gridlock zone extends from 25% to 40%. Since the existing rate of 55% falls outside the gridlock zone, Congress can pass tax rate legislation. Assume, on December 31, 2012, the legislature passes and the President signs temporary legislation changing the rate to 35%. The temporary rate is scheduled to expire two years later on December 31, 2014.

For a moment, assume the pivots do not change. If the legislature considers any bill before sunset (e.g., six months later on June 30, 2013), no legislation could be enacted since the temporary rate falls within the gridlock zone. The pivots cannot agree on a change to the temporary rate. Whether the temporary rate can be changed prior to expiration depends on the temporary policy.

But what is the relevant status quo when Congress considers renewal? If the temporary legislation expires, then the rate will revert to 55%, i.e., the underlying permanent policy is a rate of 55%. Assuming the pivots have not changed, they will all agree the underlying permanent rate is too high. Since the underlying permanent policy falls outside of the gridlock zone, Congress and the President can renew the temporary legislation.

This example demonstrates a number of important characteristics of temporary legislation. First, the underlying permanent policy will initially fall outside the gridlock zone. Second, the temporary policy will initially be inside the gridlock zone, as shown in Figure 4.

 

Figure 4. Position of Temporary Policy and Underlying Permanent Policy Relative to Pivots and Gridlock Zone

 Oh10

 

More importantly, temporary legislation generally is renewed if the pivots do not change. If the pivots do not shift, the gridlock zone does not move. Since the underlying permanent policy fell outside the gridlock zone at enactment, that policy will also generally fall outside of the gridlock zone at expiration.66

2.     Expiration as Gridlock

Conversely, temporary legislation is not renewed when the pivots shift sufficiently such that the underlying permanent policy is inside the gridlock zone at sunset. This is demonstrated in the following example.

Example 5: Shifting Pivots and the Expiration of Temporary Legislation. Assume the existing permanent tax rate is 20%. Table 5 gives the rate preferences of the pivots in 2012 and 2014.

 

Table 5. Preferred Top Rates of Pivots in 2012 and 2014

Oh11

 

Since the existing rate of 20% falls outside the gridlock zone in 2012, it is possible to enact legislation increasing the tax rate. Assume, on December 31, 2012, the legislature passes and the President signs temporary legislation increasing the rate to 28%. The temporary rate is scheduled to expire two years later on December 31, 2014.

Two years later, when the temporary legislation is due to expire, the underlying permanent policy—the 20% rate—is no longer outside the gridlock zone. In the interim, one of the pivots has moved. Specifically, the median member of the House majority became more conservative. This could have occurred because Tea Party challengers displaced moderate Republican incumbents or because reelected Republicans shifted their tax rate preferences in a conservative direction. The result is that renewal legislation cannot be passed. The majority party in the House will block the enactment of any renewal legislation. The temporary legislation expires because the underlying permanent legislation is in the gridlock zone.

In Example 5, when the temporary legislation expires, the tax rate decreases to 20%. This is despite the fact that all the other pivots want a rate higher than 20%. In other words, when temporary legislation expires, it can shift policy away from what most legislators want. This can occur because renewal is driven by an underlying permanent policy that may have little connection to current legislative preferences. In some cases, the underlying permanent law is decades old. In every case, such law is not the most recent legislative pronouncement on that issue.67

Even if temporary legislation is renewed, the underlying permanent policy can influence the content of that legislation. If the underlying permanent policy is close to one of the pivots, that pivot will largely determine the content of renewal legislation. Remember that when the status quo is close to a pivot, that pivot will limit the extent to which legislation can move policy.68 Thus, the relative position of the underlying permanent policy to the gridlock zone influences the renewal of temporary legislation and the content of that legislation if it is renewed.

Part II.B has applied pivot models to the renewal of temporary legislation. The key intuitions are: (1) that renewal depends on the acceptability of the underlying permanent policy to the pivots; and (2) that the underlying permanent policy can affect the content of renewal legislation.

The next two Parts consider important objections to modeling renewal using pivot models. Part II.B.3 considers how the original motivation for enacting legislation as temporary affects the analysis of renewal. Part II.B.4 considers the assumptions of pivot models and qualifies some of our earlier conclusions.

3.     Renewal Can Also Depend on Why the Legislation Was Enacted as Temporary

Thus far, I have taken the existence of temporary legislation as given. But a natural question is why legislation is enacted as temporary in the first place.69 There are at least four different motivations for enacting legislation as temporary rather than permanent. These differing motivations can influence how we understand renewal.

In some cases, Congress intentionally employs temporary legislation as a responsive policy tool.70 As will be discussed further in Part IV.C, the hope is that temporary legislation can more quickly respond to changing circumstances and improved information.

In a second set of cases, the decision to enact policy using temporary legislation is simply a practical response to legislating in the shadow of increasing deficits, budgetary restrictions like PAYGO,71 and procedural restrictions on the use of reconciliation. In these circumstances, the enactment of temporary legislation is not an expression of legislator preferences on policy duration. For example, the Bush tax cuts were enacted as temporary in order to comply with the Byrd rule.72 Similarly, “tax extenders”73 are enacted on a temporary basis because the budgetary cost of making them permanent is considered too high.

A third type of temporary legislation responds to transitory circumstances. For example, the various stimulus provisions passed in response to the Great Recession were designed to accelerate spending and capital investments.74 Such acceleration depends on the provisions being temporary.

In a final set of cases, the decision to enact legislation as temporary rather than permanent may be strategic. Let’s return to Example 2 where the existing rate is 20%, and the President’s ideal rate is 40%. Would the President prefer a permanent 25% rate or a temporary 25% rate? It may depend on how he thinks the duration of that policy will affect future rates. For example, if the President thinks the gridlock zone will soon shift in a liberal direction, he may prefer the temporary 25% rate. The rate could then be revisited with the potential for a more liberal outcome. On the other hand, if he thinks the gridlock zone is going to shift in a conservative direction, he may prefer the permanent 25% rate. These strategic concerns are strongest when shifts in pivots are foreseeable.75

The likelihood of renewal can depend on why the legislation was temporary in the first place. The approach presented in this Article works particularly well for exploring the first and second categories, where temporary legislation is intentionally employed as a responsive tool or is motivated by some exogenous (usually budget-related) reason. In these situations, it is reasonable to ignore legislator preferences regarding duration. In the third set of cases, renewal can best be studied by looking at whether the transitory circumstances are still present.76 In the fourth set of cases, the model provides a useful structure for thinking about the strategic considerations of lawmakers, but provides less robust predictions regarding whether temporary legislation will be renewed.77

4.     One-Dimensional Preferences, Logrolling, and Strategic Disagreement

There is a further set of qualifications due to the assumptions underlying pivot models. First, pivot models are most helpful with respect to issues for which preferences can be arranged on one dimension. The examples have focused on tax rates, a policy with this feature. In order to identify the filibuster pivot, the House majority-median pivot, or the veto-override pivots, one must have the ability to arrange legislators in the order of their preferences.

At first blush, many policies seem to have this feature. How many immigrants should be allowed into the country in any given year? How many bullets should be allowed in a clip? How much should we subsidize renewable energy? But when we look closer at these policies, we see that preferences are sometimes much more complex. It is not just how many immigrants, but which ones. It is not just the size of the clip, but the caliber of the weapon. It is not just the size of the renewable energy subsidy, but which types of renewable energy should receive subsidies. Because policies are complex, there are many dimensions on which legislators may have differing preferences. Once legislator preferences can only be represented in multiple dimensions, the results of predictive models are not well-determined.78

On complex issues, legislator preferences may still approximately fall on a single dimension if one of two conditions holds. The first condition is if legislator preferences on the various dimensions are highly correlated. If that condition holds, then their preferences can be arranged on (i.e., collapse onto) a single dimension. The model will then reasonably approximate the renewal process. For example, if legislators who want a generous renewable energy tax credit also want the credit to apply to a wide range of renewable energy investments (and conversely, legislators who want a small credit want the credit to apply to a narrow range of investments), then a single dimension can meaningfully approximate legislator preferences. Similarly, if legislators who are in favor of restricting the number of bullets in a magazine are also in favor of regulating weapons of a certain caliber, then their preferences are potentially close enough to one-dimensional to employ the model.

The second condition is if legislators care significantly more about one dimension of policy than the others. If that condition holds, the model is a reasonable approximation of policy outcomes on the salient dimension of policy, but is silent on the less salient dimensions of policy. For example, if legislators care much more about the renewable energy tax credit percentage than the types of renewable investments that qualify, then the model is useful for thinking about outcomes regarding the tax credit percentage. However, it is less useful in studying policy outcomes regarding qualifying investments. Thus, the tax rate example is a particularly strong one because rates tend to be very salient to legislators.

Moreover, pivot models implicitly assume that Congress confronts each policy independently—that Congress decides what to do about tax rates and then separately decides what it will do about renewable energy. But what if legislators have different strengths of preferences and are willing to trade votes on one issue for votes on another, commonly referred to as “logrolling”?79 One can imagine situations where a status quo for one issue may fall within the gridlock zone, but the relevant pivot agrees to avoid exercising the veto gate in exchange for votes on some other issue of greater importance.

The possibility of logrolling cautions overzealous application of pivot models. It is quite possible, for example, that veto players are willing to crosstrade small deviations on certain issues. These concerns are more relevant with respect to less salient (smaller) policies. With respect to those policies, logrolling may make the borders of the gridlock zone a little fuzzy—one would expect logrolling to shrink the size of the gridlock zone, thereby making renewal more likely. On the other hand, with respect to salient issues like gun control or tax rates, it is much less likely that logrolling will meaningfully shrink the gridlock zone. Moreover, the increasing polarization between the parties and the homogenization within the parties suggest that logrolling currently is not occurring in the open “horse trade” fashion of the 1980s.80

Finally, pivot models do not incorporate strategic disagreement. These are situations where enough legislators prefer a bill to existing policy, but one party, group of legislators, or the President opposes legislative action for strategic purposes. If prevalent in the context of temporary legislation, strategic disagreement would undermine the conclusion that renewal occurs when the underlying permanent policy is outside of the gridlock zone. Strategic disagreement could result in nonrenewal even when legislator preferences suggest a high likelihood of renewal.

Authors have offered a number of rationales for such strategic behavior, and the explanations are grouped into two large categories. Politicians may choose not to compromise because of interparty competition (e.g., in order to maintain an issue with respect to which a party has an advantage) or intraparty competition (e.g., in order to avoid alienating key party constituencies or interest groups).81

There are reasons to think that strategic disagreement regarding temporary legislation will be limited. When renewing temporary legislation, politicians are compromising on an issue where they have previously compromised.82 Moreover, the temporary legislation itself often generates its own interest groups, making strategic disagreement regarding renewal even more unlikely. A good example is the tax extenders package. This group of temporary provisions has significant support from a variety of industries and interest groups. Even though there are occasionally political standoffs about renewing the tax extenders, these standoffs usually end with renewal (albeit retroactively at times).

However, strategic disagreement may in certain circumstances impede the renewal of temporary legislation. One would expect strategic disagreement to be most prevalent with respect to highly salient issues where interparty and intraparty competitive gains will be greatest.83 Perhaps a good example of this was the highly salient battle over extending the temporary Bush tax cuts in 2010 and 2012. But even these rate fights were eventually resolved—through a renewal of rates in 2010 and a partial expiration in 2012.84 Strategic disagreement is an important possibility to consider because it weakens the link between renewal legislation being possible and renewal legislation actually being enacted. However, the tax extenders and the tax rate examples suggest strategic disagreement and the attendant political posturing usually result in delayed renewal, rather than nonrenewal.

As discussed in this Part, pivot models have some very significant assumptions. However, if we focus on issues where preferences are approximated by a single dimension, these models are useful for studying renewal. Moreover, these theoretical models can form the basis of a more empirical exploration. The next Part moves from theory to application.

III.     Measuring the Likelihood of Renewal

The theoretical model of renewal predicts renewal of temporary legislation if the underlying permanent policy is outside of the gridlock zone at sunset. Since the underlying permanent policy was outside the gridlock zone at enactment (Figure 5(a)), that policy remains outside the gridlock zone at sunset so long as the pivots have not shifted significantly (Figure 5(b)).85 Conversely, nonrenewal results if the pivots have moved such that the underlying permanent policy is inside the gridlock zone (Figure 5(c)).

 

Figure 5. Pivots at Enactment of Temporary Legislation (a), Renewal (b), and Nonrenewal (c)

 Oh12

 

Renewal-uncertainty is therefore greatest when there are significant shifts in the pivots and when the underlying permanent policy is close to the gridlock zone at enactment. Part III.A presents a principled way to approach the heterogeneity of renewal-uncertainty across provisions and through time: measure the shift in pivots and the relative position of the underlying permanent policy. Part III.B then extends this methodology to other examples and considers the policy implications. It also explores several limitations of the approach.

A.     Unpacking Renewal-Uncertainty

In order to explore the renewal-uncertainty of a particular piece of temporary legislation, it is necessary to measure: (1) how far the gridlock zone has shifted since enactment; and (2) how close the underlying permanent policy was to the gridlock zone. The methodology described below provides useful proxies for both.

To keep the description accessible, I describe the methodology in the context of an example. I focus on the temporary reduction in the capital gains rate enacted in 2003.86 That piece of legislation temporarily reduced the capital gains rate from 20% to 15%. This reduction was initially scheduled to expire in 2008. In 2006, the rate reduction was extended until 2010.87 In 2010, the rate reduction was again extended until 2012.88 At the end of 2012, the temporary rate reduction expired, and the top capital gains rate returned to 20%.

1.     Measuring the Gridlock Zone

The gridlock zone shifts as the pivots change. In order to identify the pivots, the first step is to arrange legislators based on their policy preferences. The political science literature has developed sophisticated algorithms for arranging legislators based on their roll call voting. The most extensively used algorithm, developed by Keith Poole and Howard Rosenthal, is DW-NOMINATE, which uses roll call votes to estimate a score for each legislator from -1 to +1.89 Based on their voting history, DW-NOMINATE calculates that Senator Rand Paul (DW-NOMINATE = +1.00) is a very conservative Republican, Senator Olympia Snowe (DW-NOMINATE = +0.05) is a moderate Republican, Senator Ben Nelson (DW-NOMINATE = -0.01) is a moderate Democrat, and Senator Bernie Sanders (DW-NOMINATE = -0.64) is a liberal Democrat.90 The algorithm is a more empirically rigorous version of the legislative scorecards produced by Heritage Action or Americans for Democratic Action.91 It also has the advantage of allowing intertemporal comparisons of legislators across Congresses.92 The ordering of legislators produced by DW-NOMINATE is remarkably effective at explaining voting behavior—it correctly classifies about 85% of all roll call votes in Congress.93 The intuition is that the ordering of legislators based on their preferences is very similar across a broad range of issues.

I use these DW-NOMINATE scores as proxies for legislator and presidential preferences with respect to the capital gains tax rate. As shown in Table 6, these scores are very accurate in classifying the votes on legislation involving the capital gains rate.94 With respect to the 2003, 2006, and 2010 tax acts, the DW-NOMINATE algorithm correctly identifies 90.4% of the votes by senators and 91.3% of the votes by members of the House. It does a particularly good job of classifying the 2003 and 2006 votes (both of which were “cleaner” than the 2010 vote95). This suggests that DW-NOMINATE scores are a reasonable proxy for legislator preferences on this issue.

 

Table 6(a). House Roll Call Votes on Capital Gains Rate

Oh13 

Table 6(b). Senate Roll Call Votes on Capital Gains Rate

Oh14 

Using the DW-NOMINATE scores as proxies for the preferences of legislators, I calculate the position of the pivots. In 2003, for example, the DW-NOMINATE scores of the pivots were +0.72 (President), +0.41 (House majority-median pivot), -0.22 (filibuster pivot), +0.37 (House veto-override pivot), and +0.32 (Senate veto-override pivot). The gridlock zone therefore spanned from -0.22 to +0.41.96

Table 7 provides the DW-NOMINATE scores for the pivots that define the gridlock zone from 2003 to 2012. The “left-pivot” and “right-pivot” are, respectively, the liberal and conservative pivots that define the gridlock zone. Figure 6 graphs the gridlock zones.

 

Table 7. DW-NOMINATE Scores of Gridlock Zone

 Oh15


Figure 6. Gridlock Zone from 2003–2012

 Oh16

 

Figure 6 shows the gridlock zone expanded and shifted in a manner that nicely ties the model’s predictions to the history of the capital gains tax rate.

In 2006, the rate cut was extended.97 Figure 6 shows that between 2003 and 2006, there was almost no change in the pivots and a stable gridlock zone. The model predicts that renewal is likely where the gridlock zone is stable. The stability of the gridlock zone is unsurprising given the political context. In 2003, Republicans controlled the presidency and both chambers of Congress. The 2004 elections did nothing to disturb their hegemony. In that election, Republicans even gained seats in both chambers. Thus, when renewal legislation was considered in 2006, the legislative result was the same.

In 2012, the capital gains rate reduction expired.98 Figure 6 illustrates that between 2003 and 2012, the gridlock zone expanded substantially in both directions. Indeed, the leftmost pivot became more liberal, and the rightmost pivot became more conservative.99 It may be helpful to compare these pivots to current senators. In 2003, the leftmost pivot had slightly less liberal preferences than Senator Joe Lieberman, and the rightmost pivot had slightly less conservative preferences than Senator Mitch McConnell. In 2012, the leftmost pivot had roughly the same preferences as Senator Chuck Schumer, and the rightmost pivot had roughly the same preferences as Senator John Cornyn.100 The model predicts that where the gridlock zone expands, renewal-uncertainty is relatively higher. Thus, the expiration of the capital gains rate reduction in 2012 occurred at a time when the model predicts relatively higher renewal-uncertainty.

2.     Measuring the Proximity/‌Extremity of the Underlying Permanent Policy

The proximity or extremity of the underlying permanent policy is also important to the renewal of temporary legislation. If the underlying permanent policy is extreme—far from what any pivots want—renewal-uncertainty is small even if the pivots change. On the other hand, if the underlying permanent policy is proximate—close to the preferences of one or more pivots—renewal-uncertainty is larger.

If the mapping of the underlying permanent policy into the -1 to +1 DW-NOMINATE space is available, then it is easy to see whether the status quo is extreme relative to the pivots. For example, if we knew legislators with a DW-NOMINATE score of -0.25 preferred a 20% capital gains rate, then we could easily compare the underlying permanent policy to the gridlock zone. For example, -0.25 would fall outside of the gridlock zone in 2006 when the gridlock zone spanned from -0.20 to +0.42, but fall inside the gridlock zone in 2012 when the gridlock zone spanned from -0.35 to +0.49.

Unfortunately, mapping policies into the DW-NOMINATE space is difficult. DW-NOMINATE is very accurate at arranging legislators in order of their preferences and identifying the cutline (the score that separates “yes” and “no” votes for a particular roll call). But this information does not fully specify where the status quo falls in the DW-NOMINATE space. Both the status quo and the bill could be (equally) very far from the cutline, or both could be (equally) very close to the cutline. Either combination would result in the same predicted pattern of voting.

For example, assume that a particular roll call vote has a cutline at +0.20 with conservative legislators voting in favor. This means that DW-NOMINATE predicts that all the legislators with scores higher than +0.20 voted “yes” while all the legislators with scores lower than +0.20 voted “no.” This pattern of voting could be produced by a status quo located at +0.10 and a bill located at +0.30. It could also be produced by a status quo located at -0.50 and a bill located at +0.90.101 In fact, there are an infinite number of bill-status quo combinations that would result in the same cutline. Thus, figuring out where a status quo is located in the DW-NOMINATE space usually poses substantial difficulty (and generally requires more information than just the roll call vote).102

However, when dealing with temporary legislation, there is additional information that is not available with respect to permanent legislation. This key information is the votes enacting the temporary legislation. Intuitively, how legislators voted on the original temporary legislation provides information regarding how extreme the underlying permanent policy is. If the temporary legislation passed with significant margins in both the House and Senate, we can infer that the underlying permanent policy was extreme relative to the gridlock zone.

In addition, the DW-NOMINATE cutlines for these votes divide the legislators who prefer the underlying permanent policy from those who prefer the temporary policy. We can use the cutline to infer how legislators would vote if faced with that same choice between those two policies.103 The cutline allows us to predict how current legislators will likely vote on renewal by reference to how their colleagues voted in the past.104

Returning to the capital gains example, the cutline for the 2003 legislation was -0.07.105 Legislators with DW-NOMINATE scores higher than -0.07 are predicted to prefer the reduced 15% rate, and legislators with DW-NOMINATE scores lower than -0.07 are predicted to prefer the 20% rate. In other words, the underlying permanent policy (20%) maps into the DW-NOMINATE space somewhere to the left of -0.07, and the temporary policy (15%) maps into the DW-NOMINATE space somewhere to the right of -0.07.

Looking at the pivots in 2003, DW-NOMINATE predicts that the filibuster pivot (DW-NOMINATE score of -0.22) would have filibustered this bill. The 2003 legislation did not have the support of 60 senators and was only successfully enacted through reconciliation. When bills are passed through reconciliation, debate is limited. The 60th senator is no longer a pivot because reconciliation bills cannot be filibustered.106 Since only 51 senators are required to pass a bill, the median member of the Senate replaces the filibuster pivot when legislation is passed through reconciliation. Table 8 recalculates the pivots allowing for the possibility of legislation being passed through reconciliation, and Figure 7 graphs the cutlines and the shifting gridlock zone.

 

Table 8. DW-NOMINATE Scores of Gridlock Zone Allowing for Possibility of Reconciliation

 Oh17


Figure 7. Gridlock Zones and Cutline for Capital Gains Legislation

Oh18

 

In 2006, when the bill extending the capital gains rate cut was considered, the gridlock zone had shifted even further to the right. At that point, the gridlock zone stretched from +0.07 to +0.42. Recall the underlying permanent policy (20% capital gains rate) fell somewhere to the left of -0.07. Thus, in 2006, the underlying permanent policy was outside of the gridlock zone. The extension of the rate cut is unsurprising.

The cutline informs the likelihood of renewing the same 15% rate. According to the 2003 cutline, legislators more conservative than -0.07 preferred the 15% rate to the 20% rate. In 2006, all the pivots (other than the filibuster pivot) are predicted to prefer the 15% rate because their DW-NOMINATE scores are more conservative than -0.07.107 Thus, rate-extending legislation was possible to enact, but only through reconciliation. This is exactly what happened in 2006.108

Contrast this to 2012, when the temporary reduction in the capital gains rate was scheduled to expire. At this time, the gridlock zone stretched from - 0.35 to +0.49. Because of the shift in the pivots, the cutline falls inside of the gridlock zone. This means there is at least one pivot that prefers the underlying permanent policy to the temporary policy. In fact, there are two pivots to the left of the cutline—the median senator and the President. These pivots are predicted to prefer the 20% rate based on their DW-NOMINATE scores. Unsurprisingly, in 2012, the temporary capital gains rate reduction expired, and the capital gains rate reverted to 20%. The 20% rate could not pass the Senate and did not have the support of the President.109

The extension of the rate cut in 2010, however, shows the limits of this approach. The gridlock zone in 2010 stretched from -0.37 to -0.22. There were several pivots—the median senator, the President, and the House majority-median—that had DW-NOMINATE scores lower than -0.07. Their DW-NOMINATE scores suggest they would prefer the underlying permanent rate of 20% to the temporary rate of 15%. The model therefore predicts the temporary rate cut should have expired in 2010. But in 2010, these rate cuts were traded for extensions of stimulatory spending (namely extensions of the expanded earned income tax credit, the payroll tax cut, and emergency unemployment benefits) that liberals felt were necessary given the continued sluggishness of the U.S. economy.110

This relates back to the previous discussion of strategic disagreement and logrolling.111 Liberals held the necessary votes to prevent the capital gains tax rate cut from being renewed, but the cut was extended because tax rates were tied to other policy decisions. Thus, instead of thinking of the model as predicting renewal or nonrenewal, it is more accurate to use the model for forecasting lower or higher renewal-uncertainty.

B.     Normative Implications of the Methodology

Pieces of temporary legislation have heterogeneous levels of renewal-uncertainty. By comparing the cutline to the pivots at sunset, we can explore renewal-uncertainty. If the cutline falls outside the gridlock zone, there is low renewal-uncertainty. If the cutline falls inside the gridlock zone, renewal-uncertainty will be higher.112 The model of the renewal process allows a sorting of temporary provisions that have significant renewal-uncertainty from those provisions for which renewal is much more certain. The approach consists of: (1) measuring changes to the pivots; and (2) comparing those pivots to the cutlines for the votes enacting the temporary legislation. Such a sorting is useful in contexts where the permanence of temporary legislation is relevant.

For example, consider the scoring of temporary legislation for budgetary purposes. When a piece of temporary legislation is scheduled to expire before the end of a budgetary window, how should the revenue cost of that legislation be scored? Should the Congressional Budget Office and Joint Committee on Taxation use “current policy,” which assumes any temporary legislation will be renewed, or “current law,” which assumes any temporary legislation will expire?

Accounting for heterogeneous renewal-uncertainty recasts this debate and shows that each approach may inaccurately reflect the expected budgetary cost of certain provisions.113 Both the current policy and current law approaches implicitly assume all pieces of temporary legislation have the same renewal-uncertainty. Such assumption fails to acknowledge the heterogeneity of renewal-uncertainty. Some pieces of temporary legislation are functionally permanent while others are not. Explicitly accounting for that heterogeneity has the potential to improve the accuracy of revenue estimates for temporary legislation. Under such an approach, temporary legislation that enjoys wide bipartisan support (such as the tax extenders package) could be scored as effectively permanent while legislation with high renewal-uncertainty (like the capital gains rate cut) could be scored as temporary.

This heterogeneity also has implications for how the temporariness of legislation will affect the behavior of individuals and businesses. Should we expect temporary legislation to change taxpayer behavior relative to otherwise identical permanent legislation? For example, does the temporariness of the R&D tax credit affect the timing of investment in R&D? Did the temporariness of the capital gains rate cut affect the sales of capital assets?

For sophisticated actors, the answers to these questions may depend on renewal-uncertainty. Even without employing the methodology described in Part III.A, sophisticated parties may be able to roughly estimate the likelihood of renewal based on how often a provision has been renewed in the past and whether prior votes have been narrow or close to unanimous. Temporary policies fall on a spectrum from truly transient to effectively permanent. Where a piece of temporary legislation falls on that spectrum will influence the legislation’s effect on taxpayer behavior.114

Contrast the R&D tax credit and the capital gains rate cut. Sophisticated businesses may not meaningfully change their behavior in response to the temporariness of the R&D tax credit because the provision has broad bipartisan support and has been consistently renewed.115 This is not to say that businesses are indifferent to the credit’s temporariness.116 Whenever the credit is due to expire, businesses clamor for the credit to be made permanent.117 The question posed here is whether businesses actually change their behavior in response to the credit’s temporariness.118 It appears that most companies continue to plan their R&D expenditures under the assumption the credit will be renewed.119

On the other hand, sophisticated investors likely change their behavior in response to the temporariness of the capital gains rate cut.120 That cut was pushed through Congress at a time when Republicans had unified control of the government. There was much more uncertainty regarding its renewal in 2010 and 2012. Investors generally pay tax on the increase in the value of their assets when those assets are sold.121 By delaying the sale of assets, taxpayers can benefit from deferral. A taxpayer would rather pay a 15% tax a year from now rather than a 15% tax today because of the time value of money.122 However, when rates are expected to change, the benefit of deferral may be outweighed by the detriment of a higher rate in the future. A taxpayer may rather pay a 15% tax today than a 20% tax in the future.123 Thus, where there is substantial uncertainty concerning the renewal of a rate cut, there is some acceleration of capital asset sales.124

The political context of a piece of temporary legislation is important to understanding its renewal-uncertainty. That understanding can improve our grasp of the policy implications and behavioral effects of a particular piece of temporary legislation.

C.     Limitations and Extensions

Determining the renewal-uncertainty of different temporary provisions is potentially valuable. However, there are significant hurdles in more broadly applying the approach described in Part III.A. Take the example of the R&D tax credit.125

One hurdle is figuring out the specific gridlock zone for any given issue. Part III.A employed DW-NOMINATE scores as proxies for how the gridlock shifts and changes in size. This seems like a reasonable assumption for a salient dimension of policy like the capital gains rate. It is somewhat more questionable for less important policies like the R&D tax credit. How well do preferences on the R&D tax credit map into the DW-NOMINATE space? If the gridlock zone in 2012 extends from -0.35 to +0.49 in the DW-NOMINATE space, what is the gridlock zone in terms of preferences on the R&D tax credit?

There is also a significant issue in the second step of the analysis where the cutline is used to indicate legislator preferences as between the temporary policy and the underlying permanent policy. This becomes more difficult as the legislation becomes more complicated and the policy in question becomes relatively less important in the context of the overall bill. Temporary business provisions like the R&D tax credit have increasingly been enacted in the tax extenders package.126 Since these bills include dozens of temporary tax provisions, it is difficult to know what an extension means with respect to legislator preferences on any particular provision. How much should be read into the fact that any given temporary tax provision is included in the tax extender package? Is it safe to assume that a provision is only added to the tax extender package when there is broad political support for it?

Moreover, the extenders are sometimes tacked onto larger pieces of tax legislation. For example, the tax extenders were included in the Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of 2010.127 Legislator voting on that piece of legislation probably reflected legislator preferences on salient issues like marginal tax rates and the extension of unemployment benefits—making inferences on legislator attitudes towards individual temporary provisions ill-advised.128

It makes much more sense to consider the roll call votes when the extenders were voted on separately. For example, contrast the voting on the extenders in 1991 and 1993. In 1991, the extenders were considered separately, and the legislation was unanimously approved.129 In 1993, the extenders were considered as part of the Omnibus Reconciliation Act of 1993, where the vote barely passed—51–50 in the Senate and 218–216 in the House.130 The prior unanimous vote better indicates legislator preferences on the extenders package.131

Given these limitations, the procedure described in Part III.A should generally be reserved for very clean temporary legislation (e.g., when the tax extenders are considered separately) or for the most salient features of more complicated temporary legislation (e.g., changes in capital gains rates in the 2003 or 2006 tax acts).

Note that these limitations fundamentally come from comparing the underlying permanent policy to the gridlock zone within the DW-NOMINATE space. Another approach would make the same comparison within the policy space. Under this alternative approach, the challenge would be calculating the gridlock zone for each policy.132 In other words, it would require finding out what the pivots’ preferences are on each policy issue.133 Once the gridlock zone has been determined within the policy space, comparing the underlying permanent policy to that gridlock zone becomes an easy exercise.

The discussion has focused primarily on tax and spending policy, but it is interesting to consider extensions outside of that area. The model leans heavily on the assumption that legislator preferences on a given issue can be arranged along a single dimension. The model of the renewal process is a useful tool for studying temporary legislation in other policy areas where legislator preferences similarly map well onto one dimension.134 Moreover, if one wants to use the specific approach described in Part III.A, it is necessary for legislator preferences to map neatly into the DW-NOMINATE space. This is readily determined by considering how well DW-NOMINATE classifies the voting on the enacting legislation. Fortunately, DW-NOMINATE classifies a wide variety of votes extremely well. It correctly classifies over 84% of the roll call votes in the House and 82% of all the roll call votes in the Senate.135

Consider for example, the Assault Weapons Ban, which among other actions temporarily banned the manufacture and transfer of certain types of semiautomatic weapons and magazines.136 That piece of temporary legislation was enacted in 1994 with a scheduled expiration in 2004. DW-NOMINATE correctly classifies 93% of the votes in the Senate on the Assault Weapons Ban and 84% of the votes in the House. This suggests legislator preferences on this particular issue are relatively well described by DW-NOMINATE. When the temporary legislation was originally enacted, the rightmost pivot had a relatively centrist DW-NOMINATE score of +0.08 (roughly as conservative as Senator Snowe). The cutline for the enacting legislation was +0.30.137 Legislators with DW-NOMINATE lower than +0.30 were predicted to vote “yes,” and those with higher scores were predicted to vote “no.”

In 2004, when the temporary legislation was scheduled to expire, the rightmost pivot shifted from +0.08 to +0.41 (roughly as conservative as Senator McConnell). There was a significant expansion of the gridlock zone in the conservative direction. The cutline for the 1994 legislation is inside the gridlock zone in 2004. This suggests that at least one pivot opposed the renewal of the Assault Weapons Ban. Based on their DW-NOMINATE scores, both the President and the House majority-median pivot were predicted to oppose renewal. Thus, renewal-uncertainty was high in 2004—it is unsurprising that the Assault Weapons Ban was not renewed.138

 

Figure 8. Gridlock Zones and Cutline for Assault Weapons Ban

Oh19

 

The Voting Rights Act, on the other hand, demonstrates the limits of the approach. The Voting Rights Act was passed in 1965 in order to prohibit discrimination in voting.139 Several pieces of that legislation (most significantly the preclearance provisions in section 5) were scheduled to expire in 2007 and the act was amended in 2006.140 DW-NOMINATE (in one dimension) does a poor job of predicting the Senate and House votes on that piece of legislation—indicating that legislator preferences on this issue do not neatly map into the DW-NOMINATE space.141 The approach presented in Part III.A is therefore much less helpful in framing the renewal-uncertainty of this legislation.

Nevertheless, the political science literature has found that legislator preferences are increasingly well described in one dimension.142 Thus, with the aforementioned caveats, the application of this approach to recent legislation in tax and non-tax areas is promising.

IV.     How Does the Content of Temporary Legislation Change?

When temporary legislation is renewed, there is no guarantee that the content of that legislation will remain unchanged. In fact, renewal legislation often differs significantly from the temporary legislation that it replaces. In the extant literature, commentators have pointed to this content-uncertainty as both a potential positive and a potential negative. Some have argued that the content-uncertainty of temporary legislation is desirable because it may allow better adaptation of policy to changing information and conditions.143 Others have argued that content-uncertainty undermines the ability of taxpayers to plan their behavior.144 The tension between these two views highlights the lack of a theoretical or empirical understanding of content-uncertainty. This Part uses the model of the renewal process to begin to fill that void.

I identify two different sources of content-uncertainty: (1) changes in legislative preferences; and (2) the indeterminacy of legislative outcomes. The former source of content-uncertainty seems desirable—policy should respond to shifts in legislative preferences. However, the latter source of uncertainty undermines the extent to which the content of legislation will respond to preference shifts. I argue that this second source of uncertainty scales with the size of the gridlock zone—the greater the disagreement between the pivots on any given issue, the greater the uncertainty regarding the content of renewal legislation. This leads to the key insight that temporary legislation is a responsive policy tool when shifts in legislator preferences are significant relative to the size of the gridlock zone.

A.     The Effect of Shifting Pivots on the Content of Legislation

When legislative action is possible, there is generally a range of bills that can be enacted.145 The further the status quo is from the gridlock zone, the larger the space of potential bills.146 In his groundbreaking work on pivot models, Krehbiel goes beyond predicting when legislative action is possible.147 He relies on further assumptions to forecast the content of legislation. His two most important assumptions are: (1) that legislators’ preferences are symmetrical;148 and (2) that the median legislator selects the bill to be voted on.149

Table 9 reproduces Table 3 and adds the preferences of the median legislator in each chamber (and for the moment assumes that the median preference in the House and the Senate are the same). I revisit Examples 1–3 to demonstrate how these additional assumptions specify a particular legislative outcome.

 

Table 9. Preferred Top Rates of Pivots and Chamber Medians

Oh20 

In Example 1, the status quo is a 35% rate. Legislative action is impossible. The President will resist any rate decrease. The filibuster pivot and the House majority-median pivot will resist any rate increase. In this example, the additional assumptions are not relevant. In other words, these assumptions are not necessary to define the gridlock zone.

In Example 2, the existing rate is 20%. This rate falls outside of the gridlock zone, and thus, all of the pivots want to increase the rate. The first assumption—that legislators have symmetric preferences—determines the range of plausible bills. Recall that the pivot closest to the status quo determines how far policy can be shifted.150 Here, that pivot is the House majority-median. If that pivot has symmetric preferences, he or she will vote in favor of any rate between 21% and 29%. Any rate within that range is a smaller deviation from the pivot’s preferred rate of 25%.

The second assumption determines which rate in that space is enacted. Under that assumption, the median legislator selects the rate that is closest to her preferred rate. In Example 2, the median legislator will propose a rate of 29%,151 and that bill will be enacted. This assumption effectively assumes that bills are considered under open rules. The intuition is that under open rules, the median legislator can propose an amendment to shift any bill toward his or her preferred rate.152

In Example 3, the existing rate is 80%. All of the pivots agree this rate is too high. The status quo is so extreme that there is a large range of bills that could be enacted. In fact, that range includes the median legislator’s preferred rate of 35%. Under the second assumption, the median legislator can propose 35%, and the bill will be enacted. Under Krehbiel’s assumptions, if the current rate is sufficiently extreme, legislation will reflect the median preference of the legislature.

Turning to temporary legislation, the source of content-uncertainty under these assumptions is changing legislative preferences.153 If legislative preferences remain stable, the legislature will renew the same policy every time legislation is scheduled to expire. Consider Example 4. Temporary legislation setting the tax rate to 35% is enacted in 2012 when the gridlock zone spans from 25% to 40%. Since the pivots have not moved when the temporary legislation is scheduled to expire, the 35% rate is renewed.

However, when the pivots change, renewing legislation may differ from the temporary legislation it replaces. How the content changes depends on whether the underlying permanent policy is close to the gridlock zone. If the underlying permanent policy falls relatively close to the gridlock zone, the model predicts the closest pivot will influence the resulting legislation.154 Thus, the content of renewal legislation will generally reflect changes in the preferences of that pivot. In contrast, if the underlying permanent policy is far from the gridlock zone, Krehbiel’s assumptions suggest that renewal legislation will track the median preferences of Congress.

It may be desirable for current policy (whether temporary or permanent) to track the preferences of the median legislator rather than the preferences of a particular pivot. Majority rule is considered a fundamental keystone of democratic governing. If tracking the median legislator is the goal, then it is important that the underlying permanent policy is extreme relative to the pivots.155 In the usual case, where the underlying permanent policy is set by prior legislation, the importance of the underlying permanent policy seems merely descriptive. The content of renewal legislation is influenced by underlying permanent policy, but that policy is dictated by historic coincidence. However, Congress is not stuck with passively accepting underlying permanent policies. Congress could instead actively set backstops by enacting temporary legislation that defaults to some new permanent policy upon expiration.

The possibility of consciously picking a backstop casts the above discussion in a prescriptive light. When it is legislatively possible, how should Congress set that backstop?156 For example, imagine Congress wants to encourage the construction of electric cars but is unsure about what level of tax credit is appropriate. Imagine legislators are pretty sure that at least 20% is necessary but their best guess is 30%. Congress passes a temporary credit of 30% backstopped by a permanent 20% credit. Backstopping a temporary policy with a permanent policy seems intuitively appealing. There is a minimum subsidy for the construction of electric cars even if the temporary policy expires. The problem is this backstop may limit the extent to which future legislation can track changing legislative preferences. The backstop may be too close to the preferences of future pivots—thereby allowing legislative minorities to influence future policy away from majoritarian preferences.

An alternative approach is to intentionally make the backstop extreme. Setting extreme backstops is risky in an increasingly polarized political climate. The gap between the two parties is growing.157 Any backstop must fall very far from the gridlock zone in order to avoid the possibility of influencing the content of legislation in the future. It is not enough for the underlying permanent policy to be extreme relative to the current pivots. It must be extreme relative to the pivots when the temporary legislation is scheduled to sunset.

One might conclude the proactive manipulation of underlying permanent policies in this manner is therefore inappropriate.158 One can imagine a perverse situation where a backstop intended as extreme ends up being insufficiently so. An example is the recent sequester. In 2011, as part of the deal to avoid a debt ceiling crisis, Congress enacted the Budget Control Act of 2011.159 This legislation provided for sequestration (automatic mandatory and discretionary spending cuts) beginning in 2013 if Congress failed to produce a bill reducing the deficit by at least $1.2 trillion. The Budget Control Act established a backstop (the sequester) that most observers thought was sufficiently extreme to force legislators to come to the table and make the necessary budget cuts.160 Although Congress has reached temporary deals pushing some of the cuts off into the future,161 the sequester was insufficiently extreme and resulted in many cuts that may poorly reflect legislative preferences.162

B.     Legislative Indeterminacy Increases Content-Uncertainty

Under Krehbiel’s assumptions, the content of legislation changes as a result of shifting pivots. However, shifts in legislative preferences are not the only source of content-uncertainty. Incorporating more realistic assumptions into the model of the renewal process reveals that the indeterminacy of the legislative process is a further source of uncertainty. I consider each of these assumptions in turn.

The first assumption is legislators have symmetric preferences. Specifically, each legislator is indifferent to deviations from his or her ideal point in either direction. Relaxing this assumption does not affect the position of the gridlock zone and therefore does not affect the conclusion regarding the renewal of temporary legislation. Asymmetry in preferences, however, can strongly affect the content of any renewal legislation.163 Legislator preferences on certain issues are sometimes strongly asymmetric. For example, with respect to top marginal tax rates, one can imagine a Republican that prefers downward deviation from her ideal point to upward deviations. If legislator preferences are asymmetric, then pivots may exercise influence over the content of renewal legislation even where the underlying permanent policy is extreme. The space of enactable bills shrinks. This means that renewal legislation is even less likely to track median legislator preferences.

The second assumption is the median legislator selects the bill to be voted on. This is effectively an assumption that bills are considered under open rules, which is usually the case in the Senate. However, in the House, bills are often considered under closed or modified closed rules.164 When a bill is considered under a closed rule, no floor amendments are permitted. If a bill is considered under a modified closed rule, only designated amendments are considered.165 Closed rules are particularly important where the space of potential bills is large. When renewal legislation is considered under a closed rule, it seems more likely that the resulting legislation will reflect majority party or committee preferences rather than the floor median.

Bicameralism introduces further indeterminacy to the legislative process. Ultimately, both chambers must pass the same bill.166 Typically, after the House and Senate each passes a version of a bill, any differences are ironed out in the committee process.167 The final bill will fall somewhere between the House and Senate bills, but exactly where is indeterminate.

One response to the use of closed rules in the House and the underspecification of the conference committee process is to further specify the model. One might be tempted to replace the open rule assumption with a different assumption regarding whose preferences are reflected under closed rules: (1) the median member of the majority party; (2) the leadership of the majority party; (3) the median member of the relevant legislative committee; or (4) the leadership of the majority party in the relevant legislative committee. One could then make further assumptions regarding how differences between House and Senate bills are negotiated in the conference committee. These assumptions would lead to another set of precise predictions regarding the content of renewal legislation. However, these assumptions would be just as subject to challenge. What about bills that are considered under open rules in the House? Is the majority party, the substantive committee, or the rules committee most relevant in setting the agenda? What factors determine the relative negotiation power of the House and Senate in the conference committee process? What influence does the salience of a particular issue have on agenda-setting and the conference committee process? What role does the President play?

A better approach is to accept that the indeterminacy of the legislative process is a significant source of content-uncertainty. The effect of open vs. closed rules; the potential influence of parties, committees, or the President; and the conference committee negotiation make it unrealistic to precisely predict the content of legislation.168

The model can still provide more robust (albeit less precise) guidance regarding the content of renewal legislation. It is very likely that any renewal legislation will fall somewhere within the gridlock zone. To see why, consider what would happen if renewal legislation established a temporary policy outside the gridlock zone. In that case, further legislative action is possible. There would be sufficient legislative support for a subsequent bill that brought the temporary policy within the gridlock zone.169

This leads to the key insight that content-uncertainty scales with the size of the gridlock zone. This should square with one’s intuition. Where the House majority-median, the President, and the filibuster pivot are far apart on a particular issue, it is impossible to predict precisely what legislation will be enacted, but it does seem reasonable to assume that any legislation will establish a policy somewhere in between their preferences.

C.     When Is Temporary Legislation a Responsive Policy Tool?

Several commentators have argued that temporary legislation is a responsive policy tool suited for situations involving changing circumstances or imperfect information (such as legislative responses to emergencies or the use of new policy tools).170 The basic idea is an intuitively appealing one—temporary legislation can regularly be updated as conditions change and information improves.

The discussion in Part IV.A provides some support for the proactive use of temporary legislation. If the underlying permanent policy is sufficiently extreme, renewal legislation may track median preferences in the legislature. However, the discussion in Part IV.B casts serious doubt regarding how consistently or precisely renewing legislation will track changing preferences. Given a relatively stable set of pivots, there is little to gain from using temporary legislation. Changes in policy can reflect changes in the median preferences of Congress. But changes could also reflect special interest influence on legislators, the influence of pivots, the influence of committees and parties through bill selection and rules restricting amendments, or the trading that occurs in the conference committee process.171

It is somewhat more realistic to hope that temporary legislation can roughly reflect shifts in legislative preferences so long as those shifts are significant in relation to the gridlock zone. If the gridlock zone moves significantly, then any new policy that falls within the gridlock zone is likely a significant improvement over the old policy.172 Thus, when legislators consider proactively using temporary legislation in conditions of uncertainty, they should consider how narrow the gridlock zone is for that particular issue. How big is the gap between the preferences of the House majority, the President, and the senators that are likely relevant for filibuster purposes? Is the size of the gap driven more by lack of information or by fundamental partisan differences? If that gap is narrow or is driven by a lack of information, then proactively using temporary legislation may make sense. Under those circumstances, it seems likely that temporary legislation is meaningfully responsive. If on the other hand, the gap is significant or is driven by fundamental partisan differences, then intentionally using temporary legislation rather than permanent legislation does not seem advisable.173

V.     Conclusion

Political uncertainty is an unavoidable part of our legal system. This uncertainty is particularly important in the context of temporary legislation. This Article has provided a theoretical and empirical framework for exploring the renewal- and content-uncertainty of temporary legislation. I have shown that different pieces of temporary legislation have distinct renewal-uncertainty and that, in some cases, that renewal-uncertainty can be measured. This heterogeneity has important policy implications with respect to how the revenue cost of temporary legislation is estimated and how temporary legislation affects taxpayer behavior.

This highlights my broader approach to political uncertainty. The political uncertainty underlying any policy is itself an important dimension of that policy. The behavioral effects of the law depend on how citizens and businesses expect the law to change in the future. The long-run implications (e.g., fiscal, macroeconomic) of the law depend on its political stability. This is true for all laws–whether they take the form of judicial decisions, legislation, or regulation and whether they are nominally temporary or permanent.

  1. [1]. See infra notes 86–88 and accompanying text.
  2. [1]. See infra notes 86–88 and accompanying text.

  3. [2]. The extant literature focuses on the relative desirability of permanent and temporary legislation without engaging in this inquiry. For leading articles, see generally Jacob E. Gersen, Temporary Legislation, 74 U. Chi. L. Rev. 247 (2007); Rebecca M. Kysar, Lasting Legislation, 159 U. Pa. L. Rev. 1007 (2011) [hereinafter Kysar, Lasting Legislation]; Rebecca M. Kysar, The Sun Also Rises: The Political Economy of Sunset Provisions in the Tax Code, 40 Ga. L. Rev. 335 (2006) [hereinafter Kysar, The Sun Also Rises]; George K. Yin, Temporary-Effect Legislation, Political Accountability, and Fiscal Restraint, 84 N.Y.U. L. Rev. 174 (2009). The normative desirability of temporary legislation is also invoked in the debate over entrenching legislation (legislation that attempts to bind against subsequent legislative action). Compare Eric A. Posner & Adrian Vermeule, Legislative Entrenchment: A Reappraisal, 111 Yale L.J. 1665, 1676–77 (2002) (arguing that sunset clauses and entrenching clauses are “mirror image[s]”), with John O. McGinnis & Michael B. Rappaport, Essay, Symmetric Entrenchment: A Constitutional and Normative Theory, 89 Va. L. Rev. 385, 444 (2003) (“Sunset provisions raise none of the special problems... that make majoritarian entrenchment problematic.”). The debate over the normative desirability of temporary legislation goes back to the Founding. For example, in Federalist No. 26, Alexander Hamilton argued that appropriations for the armed forces should be temporary in order to oblige the legislature “to deliberate upon the propriety” of the level of funding. The Federalist No. 26, at 142 (Alexander Hamilton) (Am. Bar Ass’n ed., 2009).

  4. [3]. The preferences of the 60th senator are determinative of whether a bill can overcome filibuster. See infra text accompanying notes 15–19.

  5. [4]. This work was pioneered by three influential political scientists. See generally David W. Brady & Craig Volden, Revolving Gridlock: Politics and Policy from Jimmy Carter to George W. Bush (2d ed. 2006); Keith Krehbiel, Pivotal Politics: A Theory of U.S. Lawmaking (1998).

  6. [5]. This empirical work is most closely associated with two important political scientists. See generally Keith T. Poole & Howard Rosenthal, Ideology & Congress (2007).

  7. [6]. I use the term “renewal” broadly to include situations where temporary legislation changes significantly when renewed. I separately consider: (1) whether temporary legislation is renewed; and (2) how temporary legislation changes if it is renewed. Parts II and III focus on the former. Part IV focuses on the latter.

  8. [7]. “Sunset” is the date on which the temporary legislation is scheduled to expire. See Black’s Law Dictionary 1665 (10th ed. 2014) (defining “sunset law”); Posner & Vermeule, supra note 2, at 1676 (defining sunset clauses as “clauses that cause a statute to lapse, by operation of law, after a defined period”).

  9. [8]. See infra notes 136–38 and accompanying text; Figure 8.

  10. [9]. See infra notes 107–08 and accompanying text; Figure 7.

  11. [10]. See U.S. Const. art. I, § 7, cl. 2.

  12. [11]. Krehbiel, supra note 4, at 23 (“The focus of the modeling exercise is to discern which of n legislators or the president is pivotal in various lawmaking situations and why.”)

  13. [12]. The temporary policy may indirectly influence the renewal process if the temporary policy influences the preferences of the pivotal actors. See infra note 85.

  14. [13]. See Policy Basics: Marginal and Average Tax Rates, Center on Budget & Pol’y Priorities, http://www.cbpp.org/cms/?fa=view&id=3764 (last visited Dec. 27, 2014).

  15. [14]. S. Doc. No. 113-1, Rule XXII, at 20–22 (2013); see Brady & Volden, supra note 4, at 15–16. For now, we put aside the possibility of passing a bill through reconciliation. See infra note 31.

  16. [15]. In the political science literature, these preferences are called “ideal points.” See, e.g., Gary W. Cox & Mathew D. McCubbins, Setting the Agenda: Responsible Party Government in the U.S. House of Representatives 39 (2005); Joshua D. Clinton, Using Roll Call Estimates to Test Models of Politics, 15 Ann. Rev. Pol. Sci. 79, 83–84 (2012); Jeffrey B. Lewis & Keith T. Poole, Measuring Bias and Uncertainty in Ideal Point Estimates Via the Parametric Bootstrap, 12 Pol. Analysis 105, 105–06 (2004).

  17. [16]. For any given policy issue, there are two filibuster pivots, one for conservative changes in policy and one for liberal changes. In this example, the filibuster pivot for rate increases is Senator H, but the filibuster pivot for rate decreases is Senator E.

  18. [17]. U.S. Const. art. I, § 7, cl. 2.

  19. [18]. Id.

  20. [19]. Id.

  21. [20]. See Krehbiel, supra note 4, at 23–25 (considering the implications of the President’s position relative to the median voter). There are 100 senators, so to meet the two-thirds supermajority to override the President’s veto, the 34th senator’s preference becomes the important inquiry. Likewise, there are 435 members in the House, so to meet the two-thirds supermajority, the 146th representative’s preference becomes the important inquiry.

  22. [21]. See generally Cox & McCubbins, supra note 15.

  23. [22]. Id. at 91–94. This has been apparent in the recent immigration debate. House Speaker John Boehner has explicitly stated that he will not bring an immigration bill to a vote unless it has the support of a majority of House Republicans. Jamelle Bouie, Immigration Reform in the House?, Wash. Post (July 30, 2013), http://www.washingtonpost.com/blogs/plum-line/wp/2013/07/30/immigration-reform-in-the-house.

  24. [23]. Cox & McCubbins, supra note 15, at 37–49.

  25. [24]. Id. at 4–5. This skepticism is echoed through much of the political science literature. See, e.g., Brady & Volden, supra note 4, at 39, 47 n.50; Krehbiel, supra note 4, at 165–85 (questioning congressional majority party effects).

  26. [25]. Cox & McCubbins, supra note 15, at 37–49.

  27. [26]. Id. at 25 (internal quotation marks omitted).

  28. [27]. There is both a strong theoretical argument and strong empirical support for the proposition that the House majority party exercises negative agenda control. Sean Gailmard & Jeffery A. Jenkins, Negative Agenda Control in the Senate and House: Fingerprints of Majority Party Power, 69 J. Pol. 689, 689 (2007). It is less clear whether the majority party in the Senate also influences legislative outcomes. Id. at 689–90. Sean Gailmard and Jeffery Jenkins suggest that the majority party in the Senate may also exercise negative agenda control. Id. at 690. That article focuses on empirical evidence without giving a sound theoretical explanation for the exercise of such control. See generally id.

  29. [28]. At least one paper has suggested that the minority party may exercise influence over any pivotal legislators within the minority party to avoid minority party rolls. See Jesse Richman, Parties, Pivots, and Policy: The Status Quo Test, 105 Am. Pol. Sci. Rev. 151, 153–54 (2011) (measuring the gridlock zone from the majority party median to the minority party median). For example, the minority party in the Senate may influence the filibuster pivot if that pivot is a member of the minority party. The mechanism by which such influence is exercised over a pivot in the minority party is not clearly specified. Some potential mechanisms include: (1) threatened denials of access to party-controlled campaign funds; (2) threatened loss of committee roles; and (3) the possibility of future primary challengers.

  30. [29]. For example, if one thought that the tax committee chairman effectively had veto power over tax legislation, then one could incorporate the chairman as an additional pivot. Although tax committee chairs once wielded significant power, most observers would agree that they no longer exercise a veto over tax legislation.

  31. [30]. Compare Krehbiel, supra note 4, at 186­–224, with Cox & McCubbins, supra note 15, at 36–42, 171–97, 193–96.

  32. [31]. It is helpful to say a word about reconciliation since temporary tax legislation is sometimes passed through the reconciliation process. Reconciliation is a legislative process implemented by the Senate to accelerate consideration of a budget bill. See 2 U.S.C. § 641 (2012); Robert Keith & Bill Heniff Jr., Cong. Research Serv., RL33030, The Budget Reconciliation Process: House and Senate Procedures 2–3 (2005). It eliminates the filibuster and limits debate to 20 hours. The “Byrd rule” prevents the use of reconciliation for provisions that would increase the deficit outside the ten-year budgetary window. See 2 U.S.C. §644; Bill Heniff Jr., Cong. Research Serv., RL30862, The Budget Reconciliation Process: The Senate’s “Byrd Rule” 1 (2010); Yin, supra note 2, at 215–18. Reconciliation has been used recently to pass a number of significant tax bills. E.g., Tax Increase Prevention and Reconciliation Act of 2005, Pub. L. No. 109-222, 120 Stat. 345 (2006) (codified as amended in scattered sections of I.R.C.); Jobs and Growth Tax Relief Reconciliation Act of 2003, Pub. L. No. 108-27, 117 Stat. 752 (2003) (codified as amended in scattered sections of titles 26, 42 U.S.C.); Economic Growth and Tax Relief Reconciliation Act of 2001, Pub. L. No. 107-16, 115 Stat. 38 (2001) (codified as amended in titles 26, 29 U.S.C.); Taxpayer Relief Act of 1997, Pub. L. No. 105-34, 111 Stat. 788 (1997) (codified as amended in scattered sections of titles 1, 5, 19, 26, 29, 31, 42, 46 U.S.C.). In both the 2001 and the 2003 legislation, many of the deficit-increasing provisions were enacted as temporary (scheduled to sunset at the end of ten years) to comply with the Byrd rule. Jobs and Growth Tax Relief Reconciliation Act, §§ 107, 303, 117 Stat. at 755–56, 764; Economic Growth and Tax Relief Reconciliation Act, § 901(a), 115 Stat. at 150. When reconciliation is used, the filibuster pivot is no longer relevant because of the rule limiting debate. Instead, the median senator is a pivot.

  33. [32]. Whether the bill could pass the Senate depends on the position of the 41st senator. For any given policy, there are actually two different filibuster pivots—one for leftward changes in policy and one for rightward changes. I have suppressed one of the filibuster pivots in the table and figures for ease of explication.

  34. [33]. See Krehbiel, supra note 4, at 34–39; see also Brady & Volden, supra note 4, at 15–21. The political science literature in this area uses the term “gridlock” without the negative connotations popularly associated with the term. See Brady & Volden, supra note 4, at 3 (noting their discussion of gridlock “will not focus on the role of political parties, nor of special interests, nor of the media, and it does not rely heavily on presidential leadership”). Gridlock is the expected result whenever a status quo is relatively centrist given the supermajoritarian aspects of the U.S. legislative process. See, e.g., Krehbiel, supra note 4, at 38–39.

  35. [34]. Since the agreement of all the pivots is generally required for legislative action, the gridlock zone is defined by the most extreme pivots. Those pivots are in bold in Figure 1 (since legislation requires the support of the President or the veto-override pivots, it is the less extreme of those pivots that define the gridlock zone). Interior pivots can generally be ignored. In this example, the filibuster pivot will agree to any rate increase that the House majority-median pivot supports. The filibuster pivot will similarly agree to any rate decrease that the President supports. Thus, we can ignore the filibuster pivot in determining whether legislation can be enacted. The literature refers to such pivots as being “absorbed.” George Tsebelis, Veto Players: How Political Institutions Work 11–12, 26–30 (2002). The absorbed pivots are in gray in Figure 1.

  36. [35]. See Krehbiel, supra note 4, at 38.

  37. [36]. See id. at 35–38.

  38. [37]. For instance, if we assume that her rate preferences are symmetrical, then she would only be willing to raise rates up to 29%, which is closer to her ideal rate than the current 20% rate. See infra Part IV.A.

  39. [38]. Returning to the initial example in Part II.A and Table 1, consider what happens if incumbent Senator I loses an election to a liberal challenger whose ideal rate is 70%. The filibuster pivot is now Senator G with an ideal rate of 30%. Legislative turnover can result in a shift in the gridlock zone.

  40. [39]. Brady & Volden, supra note 4, at 26. I will often refer to the pivots as changing, shifting, or moving. By this, I mean that the preferences of the pivots are changing.

  41. [40]. Returning to the initial example in Part II.A and Table 1, even a small shift in preference for the filibuster pivot (Senator H) would move the pivots. However, a small change in preferences for non-pivotal legislators will have no effect on the pivots. In fact, it takes a shift large enough to make the legislator one of the pivotal legislators for the pivot to move. For example, Senator I’s preferred rate would have to shift above 25% before any of the pivots would move in the original example. At that point Senator I would replace Senator H as the filibuster pivot. In other words, a shift in legislator preferences or legislative turnover is necessary but not sufficient for pivots to shift. For example, if an extremely conservative legislator replaces a similarly conservative one, the pivots will likely not change.

  42. [41]. Jason S. Oh & Chris Tausanovitch, Policy as Probabilities: Estimating the Political Uncertainty of Tax Policy 16 (undated) (unpublished manuscript) (on file with author). See generally Tax Found., Federal Individual Income Tax Rates History (2013), available at http://tax
    foundation.org/sites/taxfoundation.org/files/docs/fed_rates_history_nominal_1913_2013_0.pdf.

  43. [42]. Oh & Tausanovitch, supra note 41, at 18.

  44. [43]. Brady & Volden, supra note 4, at 27; Krehbiel, supra note 4, at 46–47.

  45. [44]. Brady & Volden, supra note 4, at 27; Krehbiel, supra note 4, at 46–47.

  46. [45]. David Brady and Craig Volden refer to this as certain status quos being “released.” Brady & Volden, supra note 4, at 26.

  47. [46]. Consider the legislative performance of the 111th Congress when the Democrats held the presidency and sizable majorities in both the Senate and the House of Representatives. See Matt Bai, Democrat in Chief?, N.Y. Times (June 8, 2010), http://www.nytimes.com/2010/
    06/13/magazine/13midterms-t.html?pagewanted=all (discussing the inability of Democratic majorities in the House and Senate “to navigate the bulk of [Obama’s] ambitious agenda past a blockade of Republican vessels”); see also Brady & Volden, supra note 4, at 7, 20, 37, 164; Krehbiel, supra note 4, at 44–45 (showing that the 103rd Congress, despite Democratic control of the House, Senate, and White House, experienced gridlock).

  48. [47]. Krehbiel, supra note 4, at 45–47.

  49. [48]. Joint Comm. on Taxation, No. JCX-38-07, Present Law and Background Relating to the Individual Alternative Minimum Tax 5 (2007).

  50. [49]. Id. at 2–5; Leonard E. Burman et al., Brookings Inst., Options to Fix the AMT
    2–6 (2007).

  51. [50]. See Joint Comm. on Taxation, supra note 48, at 2–5; Burman, supra note 49, at 2.

  52. [51]. Tax Equity and Fiscal Responsibility Act of 1982, Pub. L. No. 97-248, § 201(a), 96 Stat. 321, 411–21 (1982) (codified as amended in scattered sections of I.R.C.); S. Rep. No. 97-494, vol. 1, at 108 (1982) (“The committee has amended the present minimum tax provisions applying to individuals with one overriding objective: no taxpayer with substantial economic income should be able to avoid all tax liability....” (emphasis added)); Burman, supra note 49, at 1.

  53. [52]. See Omnibus Budget Reconciliation Act of 1993, Pub. L. No. 103-66, § 13203(a), 107 Stat. 312, 461–62 (1993) (codified at I.R.C. § 55(b)(1) (1994)); Burman, supra note 49, at 6.

  54. [53]. Omnibus Budget Reconciliation Act § 13203(b)(2), 107 Stat. at 462 (codified at I.R.C. § 55(d)(1) (1994)).

  55. [54]. CPI Inflation Calculator, Bureau Lab. Stat., http://www.bls.gov/data/inflation_calculator.
    htm (last visited Dec. 27, 2014) (enter “33,750” into “$” field; change first year field to “1994;” choose “1993” in second year field; then click “Calculate”).

  56. [55]. Id. (enter “33,750” into the “$” field; change the first year field to 1994–2001; choose “1993” in second year field; then click “Calculate”).

  57. [56]. In addition to inflation, the status quo was further shifted in 2001 by the first round of Bush tax cuts. See Economic Growth and Tax Relief Reconciliation Act of 2001, Pub. L. No. 107-16, 115 Stat. 38 (codified as amended in titles 26, 29 U.S.C.). Due to the cuts in income tax rates, a significant number of taxpayers would have been affected by the AMT without an adjustment to the exclusion.

  58. [57]. Id. § 701(a)(2), 115 Stat. at 148 (codified at I.R.C. § 55(d)(1)(B) (2002)). In 2001 dollars, the exclusion was $35,750. See CPI Inflation Calculator, supra note 54.

  59. [58]. Jobs and Growth Tax Relief Reconciliation Act of 2003, Pub. L. No. 108-27,
    § 106(a)(2), 117 Stat. 752, 755 (codified at I.R.C. § 55(d)(1) (2003)). In 2003 dollars, the exclusion was $40,250. See CPI Inflation Calculator, supra note 54.

  60. [59]. I have assumed for purposes of this example that the pivots have not moved to demonstrate how a shifting status quo can lead to eventual legislative action. Focusing on the shifting of the status quo does an excellent job of explaining the patching of the AMT.

  61. [60]. This was until legislation was passed permanently indexing the AMT exclusion for inflation in 2013. See American Taxpayer Relief Act of 2012, Pub. L. No. 112-240, § 104(b)(1), 126 Stat. 2313, 2320–21 (2013) (codified at I.R.C. § 55(d)(4) (2013)); see also CPI Inflation Calculator, supra note 54.

  62. [61]. One exception is Keith Krehbiel’s description of appropriations. Since appropriations must be renewed each year, such bills are a specific example of temporary legislation. Krehbiel, supra note 4, at 203–05, 209–16.

  63. [62]. Economic Growth and Tax Relief Reconciliation Act, §§ 101 & 511, 115 Stat. at 41–44, 70–71 (codified as amended in scattered sections of I.R.C.).

  64. [63]. Other examples include the AMT patches, see supra Part II.A.2, and the temporary increase in bonus depreciation. See I.R.C. § 168(k) (2013).

  65. [64]. I.R.C. § 41; see infra notes 115–19 and accompanying text.

  66. [65]. Other examples of temporary tax provisions with no underlying permanent legislation include several renewable energy tax credits. See, e.g., I.R.C. § 45.

  67. [66]. It is possible that the underlying permanent policy has moved (e.g., due to inflation). See supra Part II.A.2.

  68. [67]. The most recent pronouncement is the temporary legislation itself. Permanent legislation, temporary legislation where the underlying permanent policy is extreme, and temporary legislation where the underlying permanent policy is moderate differ in their path dependence on prior legislation. If the most recent legislation is permanent, outcomes are determined by the pivots and the status quo established by that legislation (policyn+1 = f(policyn, pivotsn)). The n subscripts indicate successive legislative action. If the most recent legislation is temporary and the underlying permanent policy is extreme, there is no path dependence—all that matters are current legislative preferences (as in Example 4) (policyn+1 = f(pivotsn)). If the most recent legislation is temporary and the underlying permanent policy is moderate, that permanent policy influences current legislative outcomes (as in Example 5) (policyn+1 = f(policyn-i, pivotsn)).


         This adds an interesting caveat to the assertion that temporary legislation allows past legislatures to exercise undue influence on future legislatures. Some have argued that temporary legislation allows past legislatures to tie up legislative resources of future legislatures by forcing them to revisit certain issues. Gersen,
  69. [68]. See supra Part II.A.1, Example. 2.

  70. [69]. For a discussion of what political factors might influence the decision to enact legislation as temporary or permanent, see Gersen, supra note 2, at 279–86.

  71. [70]. Good examples of this first category of temporary legislation are the temporary provisions in the Patriot Act. See Uniting and Strengthening America by Providing Appropriate Tools Required to Intercept and Obstruct Terrorism Act of 2001(USA PATRIOT Act), Pub. L. No. 107-56, § 224, 115 Stat. 272, 295 (2001). See generally Charles Doyle, Cong. Research Serv., RL32186, USA Patriot Act Sunset: Provisions That Expire on December 31, 2005 (2004).

  72. [71]. The Budget Process: What is PAYGO?, Tax Pol’y Center, http://www.taxpolicycenter.
    org/briefing-book/background/budget-process/paygo.cfm (last visited Dec. 27, 2014).

  73. [72]. See supra note 31.

  74. [73]. The “tax extenders” are a group of temporary tax provisions that are enacted together on a temporary basis. See Molly F. Sherlock, Cong. Research Serv., R43124, Tax Provisions Expiring in 2013 (“Tax Extenders”) 2 (2013).

  75. [74]. See Council of Econ. Advisers, Economic Report of the President 51–55 (2010).

  76. [75]. This would be the case, for example, during periods of unified government. Most lawmakers will expect a significant future shift in pivots once the party out of power captures any one of the House, the Senate, or the White House. When there are not strong expectations regarding shifts in the gridlock zone, these strategic concerns will be less important.

  77. [76]. For example, the extension of unemployment benefits, the earned income tax credit, and the payroll tax cut were renewed in 2010 because the economy continued to be sluggish. Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of 2010, Pub. L. No. 111-312, §§103(c), 501–03, 601, 124 Stat. 3296, 3299, 3307–10 (2010) (codified as amended in scattered sections of I.R.C.).

  78. [77]. For example, one can imagine strategic disagreement leading to nonrenewal. See infra notes 81–84 and accompanying text.

  79. [78]. See generally Richard D. McKelvey, Intransitivities in Multidimensional Voting Models and Some Implications for Agenda Control, 12 J. Econ. Theory 472 (1976). There is theoretical work done on pivot models in multiple dimensions, but the results depend much more strongly on which pivots control the agenda. See generally Tsebelis, supra note 34.

  80. [79]. Black’s Law Dictionary 1084 (10th ed. 2014).

  81. [80]. See Michael Doran, Tax Legislation in the Contemporary U.S. Congress, 68 Tax L. Rev. (forthcoming 2015) (manuscript at 39–40), available at http://papers.ssrn.com/sol3/papers.
    cfm?abstract_id=2367760.

  82. [81]. John B. Gilmour, Strategic Disagreement: Stalemate in American Politics
    51–65 (1995).

  83. [82]. Id. at 61 (“Thus bargaining over the reauthorization of an existing program is less likely to generate avoidance tactics than legislation creating a new program.”).

  84. [83]. Id. at 25 (“In general, [strategies] of disagreement are most likely to be employed when politicians deal with issues that appeal to and excite large constituencies....”).

  85. [84]. In 2012, the top rate reverted to 39.6%, and the top capital gains rate reverted to 20%. The dividend rate stayed at the lower capital gains rate. American Taxpayer Relief Act of 2012, Pub. L. No. 112-240, § 102, 126 Stat. 2313, 2318 (2013) (codified as amended in scattered sections of I.R.C.); Jonathan Weisman, Senate Passes Legislation to Allow Taxes on Affluent to Rise, N.Y. Times (Jan. 1, 2013), http://www.nytimes.com/2013/01/02/us/politics/


    senate-tax-deal-fiscal-cliff.html?_r=0.
  86. [85]. Pivots may shift due to information about constituency preferences or policy effectiveness or because of elections. See supra Part II.A.2. It is also possible that pivots may be influenced by the temporary policy (reflecting, for example, a legislative preference for policy continuity). This would tend to shrink the gridlock zone, making renewal relatively more likely. Thus, the temporary policy may indirectly affect renewal if it influences the preferences of the pivots.

  87. [86]. Jobs and Growth Tax Relief Reconciliation Act of 2003, Pub. L. No. 108-27, §301(a)(2), 117 Stat. 752, 758 (codified as amended in scattered sections of I.R.C.).

  88. [87]. Tax Increase Prevention and Reconciliation Act of 2005, Pub. L. No. 109-222, § 102, 120 Stat. 345, 346 (2006).

  89. [88]. Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of 2010, Pub. L. No. 111-312, § 102, 124 Stat. 3296, 3298.

  90. [89]. Poole & Rosenthal, supra note 5, at 12–30. It is no exaggeration to say that Poole and Rosenthal’s work revolutionized the study of congressional voting.

  91. [90]. Data available at Royce Carroll et al., “Common Space” DW-NOMINATE Scores with Bootstrapped Standard Errors (Joint House and Senate Scaling), Voteview.com, http://www.voteview.
    com/dwnomjoint.asp (last visited Dec. 27, 2014) (scroll down and select “Legislator Estimates 1st to 112th Houses and Senates (Text File, 49,959 lines)” hyperlink).

  92. [91]. See Scorecard, Heritage Action for Am., http://heritageactionscorecard.com/ (last visited Dec. 27, 2014); Voting Records, Ams. for Democratic Action, http://www.adaction.org/
    pages/publications/voting-records.php (last visited Dec. 27, 2014). Those scorecards focus on a small number of votes. The DW-NOMINATE algorithm includes almost all roll call votes. It has been shown that DW-NOMINATE scores are highly correlated to the scores published by these organizations. Poole & Rosenthal, supra note 5, at 216–19.

  93. [92]. Poole & Rosenthal, supra note 5, at 96–104.

  94. [93]. DW-NOMINATE can be used to estimate higher dimensional mappings of legislator preferences. Notably, adding additional dimensions does not substantially increase the explanatory power of the mapping for recent Congresses. There have been times in the past where there was a meaningful second dimension to voting behavior (e.g., during the Civil Rights era). Id. at 57–66.

  95. [94]. Graphs based on author’s calculations using DW-NOMINATE data and congressional vote counts.

  96. [95]. In addition to the capital gains rate, the 2010 legislation involved other salient issues including extensions of the unemployment and the top marginal tax rate. See infra note 110 and accompanying text.

  97. [96]. Interior pivots are absorbed and therefore ignored. See Tsebelis, supra note 34, at
    26–30.

  98. [97]. Tax Increase Prevention and Reconciliation Act of 2005, Pub. L. No. 109-222, § 102, 120 Stat. 345, 346 (2006).

  99. [98]. Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of 2010, Pub. L. No. 111-312, § 102, 124 Stat. 3296, 3298.

  100. [99]. Considering the specific political context, this shift of the gridlock zone is unsurprising. In 2003, the Republicans held both the presidency and majorities in both chambers of Congress. In 2012, the Democrats held the presidency and a majority in the Senate.

  101. [100]. See Carroll et al., supra note 90.

  102. [101]. DW-NOMINATE provides estimates of where the status quo is located, but these estimates rely on questionable assumptions (including probabilistic voting) and are strongly dependent on the particular utility function assumed for legislators. See Poole & Rosenthal, supra note 5, at 28 (cautioning that status quo estimates are not generally reliable).

  103. [102]. In a working paper, Chris Tausanovitch and I estimate how policies map into the DW-NOMINATE space, which allows for a more direct comparison of the underlying permanent policy to the gridlock zone. Oh & Tausanovitch, supra note 41, at 5–19.

  104. [103]. Since preferences can shift, the most recent roll call vote provides the most relevant information where a piece of temporary legislation has been renewed several times.

  105. [104]. This is possible because DW-NOMINATE allows comparisons of legislators between Congresses. Poole & Rosenthal, supra note 5, at 96–104.

  106. [105]. This is the average cutline for the votes on the conference version. DW-NOMINATE calculates separate cutlines for each roll call vote and thus calculates separate cutlines for the House (-0.03) and the Senate (-0.10).

  107. [106]. See supra note 31.

  108. [107]. This approach implicitly assumes that legislators’ preferences regarding the capital gains rate were consistent between 2003 and 2006. This assumption is at least partially confirmed by the consistency of the cutline in 2003 and 2006. The average cutline for the 2006 legislation (-0.07) is virtually identical to the average cutline for the 2003 temporary legislation.

  109. [108]. See supra note 97 and accompanying text.

  110. [109]. See supra notes 98–100 and accompanying text.

  111. [110]. See David M. Herszenhorn & Sheryl Gay Stolberg, Obama Defends Tax Deal, But His Party Stays Hostile, N.Y. Times (Dec. 7, 2010), http://www.nytimes.com/2010/12/08/us/politics/
    08cong.html.

  112. [111]. See supra Part II.B.4.

  113. [112]. It still may be possible to pass a more modest piece of legislation. Consider two alternative assumptions regarding how the capital gains rates of 15% and 20% map into the DW-NOMINATE space. First, assume that the capital gains rate of 15% maps to a DW-NOMINATE score of +0.03 and the 20% capital gains rate maps to a DW-NOMINATE score of -0.17. The underlying permanent policy (20%) would fall within the gridlock zone in 2012, and the model would predict that no legislative action is possible. Assume alternatively that the capital gains rate of 15% maps to a DW-NOMINATE score of +0.23 and the 20% capital gains rate maps to a DW-NOMINATE score of -0.37. The underlying permanent policy (20%) would then fall outside the gridlock zone in 2012, and the model would predict that legislative action is possible. However, Congress would have to pass a more modest bill (a smaller reduction in the capital gains rate).

  114. [113]. It should be noted that there is a robust debate over the proper scoring of temporary legislation for budgetary purposes. See, e.g., Kysar, Lasting Legislation, supra note 2, at 1029–33; Kysar, The Sun Also Rises, supra note 2, at 359–62; Yin, supra note 2, at 187–208; Elizabeth Garrett, Comment, Accounting for the Federal Budget and Its Reform, 41 Harv. J. on Legis. 187, 197–98 (2004). Accuracy is just one of several relevant considerations. With respect to the larger debate, I remain agnostic because of the general theory of the second best. The normative (and theoretical) ideal would be a budgetary policy that fully accounts for the costs of all legislation indefinitely into the future with precision. Unfortunately, this is impossible, so there is a finite budget window. Analogizing to the general theory of the second best casts doubt on any “fix” that does not address the fundamental distortion of a limited budget window. Professor Kysar’s argument that temporary legislation does not take into account the full cost of legislation focuses on the failure of the budgetary process to properly account for costs of temporary legislation within the budget window. Kysar, Lasting Legislation, supra note 2, at 1038–39; Kysar, The Sun Also Rises, supra note 2, at 349, 360–62. Professor Yin’s argument instead focuses on how the budget process does not account for the costs of permanent legislation outside the budget window. Yin, supra note 2, at 208.

  115. [114]. Taxpayers may incorrectly perceive renewal-uncertainty and may make systematic errors in incorporating such probabilities into their decisionmaking. The various behavioral and psychological mechanisms underlying that misperception are beyond the scope of this Article. See generally Russell B. Korobkin & Thomas S. Ulen, Law and Behavioral Science: Removing the Rationality Assumption from Law and Economics, 88 Calif. L. Rev. 1051 (2000); Amos Tversky & Daniel Kahneman, Judgment Under Uncertainty: Heuristics and Biases, 185 Science (n.s.) 1124 (1974). I am making the more fundamental point that uncertainty and relative levels of uncertainty are important to understanding taxpayer behavior. Subsequent work could incorporate behavioral and psychological mechanisms into this analysis.

  116. [115]. The R&D tax credit was enacted in 1981 because Congress believed that a credit would lead companies to initiate or expand research and development programs. Section 41 of the Internal Revenue Code currently allows a taxpayer to take a tax credit equal to 20% of “qualified research expenses” for the taxable year over the “base amount.” I.R.C. § 41(b)–(c) (2013). The R&D tax credit was initially scheduled to sunset at the end of 1985, but has been repeatedly extended and was renewed for the 15th time in 2013. See American Taxpayer Relief Act of 2012, Pub. L. No. 112-240, § 301(a)(1), 126 Stat. 2313, 2326 (2013) (codified at I.R.C. § 41(h)(1)(B) (2013)); Gary Guenther, Cong. Research Serv., RL31181, Research Tax Credit: Current Law and Policy Issues for the 113th Congress 4, 18 (2013).

  117. [116]. For example, the temporariness of the R&D tax credit affects reported earnings. Emily Chasan, CFOs Warn Investors on Impact of Expired R&D Tax Credit, Wall St. J. (Jan. 22, 2014), http://
    blogs.wsj.com/cfo/2014/01/22/cfos-warn-investors-on-impact-of-expired-rd-tax-credit; Joe Harpaz, R&D Tax Credit Expiry Rears Its Head in Corporate Earnings Reports, Forbes (May 1, 2014), http://www.
    forbes.com/sites/joeharpaz/2014/05/01/rd-tax-credit-expiry-rears-its-head-in-corporate-earnings-reports/.

  118. [117]. See Chasan, supra note 116; Harpaz, supra note 116.

  119. [118]. There is evidence that companies have a strong elastic response to R&D tax credits generally. See, e.g., Nick Bloom, Rachel Griffith & John Van Reenen, Do R&D Tax Credits Work? Evidence from a Panel of Countries 1979–1997, 85 J. Pub. Econ. 1, 2 (2002).

  120. [119]. See Chasan, supra note 116; Harpaz, supra note 116.

  121. [120]. See Jane G. Gravelle, Cong. Research Serv., RS21014, Economic and Revenue Effects of Permanent and Temporary Capital Gains Tax Cuts 3 (2003), available at http://
    assets.opencrs.com/rpts/RS21014_20030129.pdf.

  122. [121]. More precisely, gains are taxed when “realized.” Realization generally includes the sale or exchange of an asset. I.R.C. § 1001(b) (2012); Gravelle, supra note 120, at 2.

  123. [122]. The taxpayer would be able to earn a return on the tax deferred. See Gravelle, supra note 120, at 3.

  124. [123]. The taxpayer’s preference would depend on their discount rate.

  125. [124]. See Gravelle, supra note 120, at 3.

  126. [125]. See supra notes 115–19 and accompanying text.

  127. [126]. A prominent reason for the increased use of the tax extenders is that the budgetary costs of these temporary provisions are limited (budgetary projections generally use the current law assumption). See The Tax Break-Down: Tax Extenders, Committee for Responsible Fed. Budget, http://crfb.org/blogs/tax-break-down-tax-extenders (last visited Dec. 27, 2014).

  128. [127]. Molly F. Sherlock, Cong. Research Serv., R42105, Tax Provisions Expiring in 2011 and “Tax Extenders” 12 (2011).

  129. [128]. See supra Part II.B.4 for a discussion of salience.

  130. [129]. See Tax Extension Act of 1991, Pub. L. No. 102-227, 105 Stat. 1686 (codified as amended in scattered sections of I.R.C.); H.R. 3909 (102nd): Tax Extension Act of 1991, GovTrack.us, http://
    www.govtrack.us/congress/votes/102-1991/h439 (last visited Dec. 27, 2014).

  131. [130]. Omnibus Budget Reconciliation Act of 1993, Pub. L. No. 103-66, 107 Stat. 312 (codified as amended in scattered sections of titles 2, 5, 7, 10, 12, 16, 19–22, 25–26, 28–31, 35–36, 38, 42–43, 47, 50 U.S.C.); H.R. 2264 (103rd): Omnibus Budget Reconciliation Act of 1993, GovTrack.us, http://www.
    govtrack.us/congress/votes/103-1993/s247 (last visited Dec. 27, 2014).

  132. [131]. But we still run into the aforementioned problem of inferring preferences with respect to individual temporary provisions. See supra note 126 and accompanying text.

  133. [132]. When the comparison is done in the DW-NOMINATE space, it is the underlying permanent policy that is difficult to locate precisely. When the comparison is done in the policy space, it is the gridlock zone that is difficult to measure precisely.

  134. [133]. This returns to the aforementioned challenge of mapping policy preferences onto the DW-NOMINATE space. One could, for example, estimate these preferences using public statements or the legislative record. See Oh & Tausanovitch, supra note 41, at 5–6.

  135. [134]. Legislator preferences can map onto one dimension, even when a given policy has multiple dimensions, so long as legislator preferences on those dimensions are correlated. See supra Part II.B.4.

  136. [135]. Poole & Rosenthal, supra note 5, at 34 (including all “scaled votes,” all votes where at least 2.5% of votes were on the minority side). These classification percentages are for DW-NOMINATE in a single dimension. Adding a second dimension to the algorithm adds very little explanatory power (two dimensions improves the classification by about 3% in both the House and Senate). Id.

  137. [136]. Violent Crime Control and Law Enforcement Act of 1994, Pub. L. No. 103-322,
    § 110102, 108 Stat. 1796, 1996–98 (codified at 18 U.S.C. § 922(v) (1994)) (expired 2004). There were proposals to renew the act. See, e.g., Assault Weapons Ban and Law Enforcement Protection Act of 2003, H.R. 2038, 108th Cong.

  138. [137]. This is the average cutline for the votes on the conference agreement. The House cutline was +0.52 and the Senate cutline was +0.08.

  139. [138]. When the prior cutline falls within the gridlock zone, this suggests that the temporary policy will not be renewed in the same form. However, it is still possible for the underlying permanent policy to be outside the gridlock zone and for less ambitious policy to be passed.

  140. [139]. Voting Rights Act of 1965, Pub. L. No. 89-110, 79 Stat. 437 (codified at 42 U.S.C. §§1973–1973aa-6 (2012)).

  141. [140]. Fannie Lou Hamer, Rosa Parks, and Coretta Scott King Voting Rights Act Reauthorization and Amendments Act of 2006, Pub. L. No. 109-246, 120 Stat. 577 (codified as amended in scattered sections of 42 U.S.C.).

  142. [141]. DW-NOMINATE does an excellent job of characterizing these votes if a second dimension is permitted. The voting behavior maps very neatly onto the North-South dimension rather than the liberal-conservative dimension. The North-South dimension of voting has become much less important recently. Since coalitions no longer form along this dimension, the roll call vote for the enacting legislation provides very little information regarding what would happen when this legislation was scheduled to expire. Poole & Rosenthal, supra note 5, at 54–59, 81–83.

  143. [142]. Id. at 55 (“The second dimension has withered away and all but disappeared.”).

  144. [143]. See infra Part IV.C.

  145. [144]. See Kysar, Lasting Legislation, supra note 2, at 1063–65. A report issued by Nina Olson, the current National Taxpayer Advocate (the head of the Office of the Taxpayer Advocate, an independent office within the Internal Revenue Service) lists the uncertainty caused by sunsets as one of the most serious problems encountered by taxpayers. 1 Nina E. Olson, Office of the Taxpayer Advocate, Internal Revenue Serv., National Taxpayer Advocate 2012 Annual Report to Congress 20 (2012), available at http://www.taxpayeradvocate.irs.gov/userfiles/
    file/Full-Report/Volume-1.pdf.

  146. [145]. See supra Part II.A.1, Examples 2–3.

  147. [146]. Id.

  148. [147]. See generally Krehbiel, supra note 4.

  149. [148]. Id. at 22.

  150. [149]. Id. at 23–25.

  151. [150]. See supra Part II.A.1, Example 2.

  152. [151]. The range of plausible rates is between 21% and 29%, and 29% is the closest rate to the median legislator’s preferred rate of 35%. See id.

  153. [152]. Krehbiel, supra note 4, at 24­–25 (“Though seemingly dictatorial, this one-player choice is more appropriately interpreted as a strategic simple-majoritarian action by the median voter on behalf of all voters with ideal points to one side of [the median legislator]. This is tantamount to assuming that the legislature decides under an open rule. That is, no restrictions are placed on amendments or on who can offer them.”).

  154. [153]. For now, I put aside the possibility that the underlying permanent policy could drift (as in the AMT exclusion example) in order to focus on the effect of shifting pivots. See supra Part II.A.2.

  155. [154]. See supra Part II.B.2, Example 5.

  156. [155]. Extreme does not mean just that the underlying permanent policy falls outside the gridlock zone but rather that pivots must prefer the median legislator’s preferred policy to the underlying permanent policy. It is not enough for the underlying policy to be extreme relative to the pivots at the enactment of temporary legislation. In order for renewing legislation to track the floor median, the underlying policy must be extreme relative to the pivots at the expiration of temporary legislation.

  157. [156]. This flexibility may not always be available. For example, if temporary legislation is passed through reconciliation, then the Byrd rule would make it difficult to enact a new permanent underlying policy if that shift in underlying permanent policy raises the deficit. See supra note 31.

  158. [157]. Poole and Rosenthal have found an increasing gap between the DW-NOMINATE scores of the Democrat and Republican parties. Poole & Rosenthal, supra note 5, at 104–10.

  159. [158]. There is also a question of whether it is possible to enact such legislation at all. It would require many legislators to vote for underlying permanent legislation that they find unpalatable. Intuitively, legislators would generally balk at supporting a backstop outside the gridlock zone even if that backstop is intended never to take effect.

  160. [159]. Budget Control Act of 2011, Pub. L. No. 112-25, 125 Stat. 240 (codified as amended in scattered sections of titles 2, 20, 31 U.S.C.).

  161. [160]. The sequester was intended to force both sides to come to the table and reach agreement. Megan S. Lynch, Cong. Research Serv., R42972, Sequestration as a Budget Enforcement Process: Frequently Asked Questions 1 (2013) (“Generally, sequesters have been used as an enforcement mechanism that would either discourage Congress from enacting legislation violating a specific budgetary goal or encourage Congress to enact legislation that would fulfill a specific budgetary goal.”).

  162. [161]. American Taxpayer Relief Act of 2012, Pub. L. No. 112-240, 126 Stat. 2313 (codified as amended in scattered sections of titles 2, 7, 10, 16, 26, 42, 45–46 U.S.C.).

  163. [162]. Bipartisan Budget Act of 2013, Pub. L. No. 113-67, §§ 111, 115–16, 127 Stat. 1165 (reducing the harshness of the sequestration cuts in 2014 and 2015).

  164. [163]. One can see the effect of asymmetric preferences on the content of legislation by turning back to Examples 1–2, supra Part II.A.1. Assume that the House majority-median pivot has asymmetric preferences. Specifically, the pivot’s preferred rate is still 25%, but she prefers downward deviations to upward deviations—that is, she is indifferent between 5% downward deviations and 1% upward deviations. Therefore, she is indifferent between 20% and 26%. The conclusion in Example 1 is unaffected. Her asymmetric preferences do not affect the position of the gridlock zone. However, the asymmetric preferences of the pivot do affect how far policy can move when the status quo falls outside the gridlock zone. Returning to Example 2, where the status quo was 20%, the pivot’s asymmetric preferences limit the extent to which the rate can be changed. Instead of shifting all the way to 29%, the rate can only move to 26%. It should be noted in passing that asymmetric preferences can lead to a situation where a legislator other than the originally identified pivot becomes the deciding vote in how far policy can be shifted.

  165. [164]. H.R. Rep. No. 111-714, at 22–23 (2011).

  166. [165]. Id.

  167. [166]. U.S. Const. art. I, § 7, cl. 2.

  168. [167]. Floyd M. Riddick & Alan S. Frumin, Riddick’s Senate Procedure: Precedents and Practice, S. Doc. No. 101-28, at 449 (1992).

  169. [168]. This is not to say that Krehbiel’s model does not guide our intuition. See generally Krehbiel, supra note 4. It is helpful to realize that the content of legislation can change in response to changing pivots. It is also helpful to realize that the farther the status quo is from the gridlock zone, the more flexibility there is in enacting policy. See supra Part II.A.1.

  170. [169]. This is a relatively robust claim that does not rely on symmetry of preferences or assumptions regarding open and closed rules. The only additional assumption required is that Congress would not enact legislation that could immediately be subject to further amendment. The exact probability distribution of those outcomes depends on many of the indeterminate and issue-specific factors discussed above. As a first approximation of that distribution, one could assume a uniform Bayesian prior distribution where each legislative result within the gridlock zone is roughly equally likely. One could then update such distribution to reflect specifics regarding the particular committee, party, and intercameral dynamics.

  171. [170]. Gersen, supra note 2, at 248 (“Because temporary legislation reduces background uncertainty and mitigates certain forms of cognitive bias, it is likely to provide far more advantages than drawbacks as a legislative response to newly recognized risks.”); Kysar, Lasting Legislation, supra note 2, at 1066–67 (arguing that “temporary legislation” may be necessary “[i]n crisis situations” and for “experimental” legislation).

  172. [171]. Jacob Gersen suggests, “If policy outcomes are entirely determined by the available information set, then a staged decision procedure is more likely to select the optimal policy than a single-stage enactment.” Gersen, supra note 2, at 267. He notes that there is no guarantee that new information will be used in the renewal process or that there will be deliberation. Id. at 275. There is a further caveat. Even if information is so incorporated, policy outcomes are also determined by the underlying permanent policy and the preferences of the pivots.

  173. [172]. I use “improvement” here to mean legislation that is more reflective of legislator preferences.

  174. [173]. But if temporary legislation should only proactively be used when there is the possibility of significant shifts in the gridlock zone, then how is it advantageous relative to permanent legislation? Remember permanent legislation can also be amended when there are significant shifts in the gridlock zone. Won’t both temporary and permanent legislation work in those circumstances? Perhaps the real benefit is in forcing future legislative bodies to revisit issues on particular timetables. When preferences shift drastically, both temporary legislation and permanent legislation may be amendable, but it is relatively more likely that temporary legislation will be timely amended. Some authors have argued that this phenomenon is unfortunate because it allows past Congresses to exercise control over the legislative agendas of future Congresses. See supra note 67. I am sympathetic to this concern. The proactive use of temporary legislation should only be reserved for situations where there is at least the possibility of significant shifts in legislative preferences.

*

Assistant Professor of Law, UCLA School of Law.



The author would like to thank Asli Bâli, Steven Bank, Samuel Bray, Ingrid Eagly, Dan Halperin, Allison Hoffman, Louis Kaplow, Sung Hui Kim, Hiroshi Motomura, Leigh Osofsky, Joanna Schwartz, Seana Shiffrin, Kirk Stark, Chris Tausanovitch, Kathleen DeLaney Thomas, Eric Zolt, and participants in seminars at Harvard University, University of Toronto, UCLA, and Loyola Law School (Los Angeles) for their helpful comments on earlier drafts. The author would also like to acknowledge the helpful assistance of Alexander Stevko and Kris Tayyeb.