Viva Voce: Implications from the Disappearing Voice Vote, 1807-1990∗ Michael S. Lynch Assistant Professor Department of Political Science University of Kansas Lawrence, KS 66045 [email protected] Anthony J. Madonna Ph.D. Candidate Department of Political Science Washington University in St. Louis St. Louis, MO 63130 [email protected]

March 2008

∗

This manuscript was prepared for presentation at the 2008 meetings of the Midwest Political Science Association. The authors would like to Steven Smith, Matthew Gabel, and Yael Shomer for their helpful comments and suggestions in regards to this project. All remaining errors are our own.

Introduction Of all the relationships maintained by members of Congress, none are as important as the one he or she shares with the home constituency. It is that constituency that ultimately determines whether the member is returned to Congress. One of the most important factors conditioning this relationship is the legislators’ roll call voting behavior. A record that indicates support for the constituencies’ preferences can go a long way to lengthening a Congressional career. Conversely, if a voting record becomes erratic or inconsistent with pledges made to the district, that member may find himself quickly out of work. In short, roll call voting provides for transparency in the democratic process. It allocates tangible information to voters, allowing them to better exercise their electoral powers. Recognizing this, the founders provided for recorded voting in both houses of Congress in the Constitution at the behest of one-fifth of the chamber. As a result of this Constitutional provision, the roll call voting record has become an extremely powerful and popular tool for political scientists. Scholars have used the record to draw conclusions about how members base their roll call decisions. Later work has applied knowledge gleaned from the record to determine how political parties, rules of procedure, and member ideology (among others) have affected congressional development and outcomes. While this work has produced many important contributions to the study of legislative politics, few political scientists have actively studied the roll call voting process. Instead, these scholars implicitly assume that the process by which bills come to a roll call vote has remained constant. This assumption has two important implications. First, it has left a substantial theoretical gap in the political science literature regarding the utility of the roll call vote. Second, it has left studies of longitudinal data analysis of the U.S. Congress vulnerable to selection biases. In this article, we evaluate these two consequences by looking at the fluctuating usage of voting methods. Specifically, we compare the usage of recorded roll call votes to legislation passed viva voce, which are thus are not part of the roll call record. To address this question, 1

we track the proportion of landmark enactments that received recorded final passage votes across chambers and bill content from 1810 to 1990. We show that, unlike previous literature assumes, the level of transparency within the democratic process varies considerably across time, bill content and levels of contentiousness. Further, using alternative methods, we provide additional suggestive evidence confirming the recent finding of Clinton and Lapinski (n.d.) that the changing and contentious use of unrecorded voting implies methodological problems with using the roll call record. The fact that the process that generates the roll call record has changed over time means that measures of ideology derived from this record may be suspect. Finally, we discuss how our findings affect the literature on congressional development.

Recorded and Unrecorded Voting The United States Constitution states that “Each House may determine the rules of its proceedings. . . ”1 This clause largely embodies the framers attitudes towards chamber procedure. There are only a handful of references to legislative rules in the Constitution and there was very little debate over legislative rules during the Constitutional Convention (Binder 1997). However, one of the few congressional rules that the founders did specify was the requirement that “. . . the yeas and nays of the members of either House on any question shall, at the desire of one fifth of those present, be entered on the journal.”2 This roll call provision was debated in the Constitutional Convention on August 10, 1781. Gouverneur Morris of Pennsylvania, supported by Edmund Randolph of Virginia, proposed amending it to allow any one member to call for the yeas and nays. This was quickly countered by Roger Sherman of Rhode Island, Oliver Ellsworth of Connecticut, and 1

“The Constitution of the United States,” Article 1, Section 5, Clause 2.

2

“The Constitution of the United States,” Article 1, Section 5, Clause 3. The founders also saw it fit to mandate the yeas and nays be called on all votes to override presidential vetoes. See Article 1, Section 7, Clause 2 for that provision.

2

Nathaniel Ghorum of Massachusetts, who proposed eliminating the requirement altogether (Farrand 1966; Binder 1997). Morris professed concern about the ability of small states to reach the one-fifth threshold, while Ghorum worried about the practice of “. . . stuffing the Journals with [votes] on frivolous occasions,” and ”. . . misleading the people, who never know the reasons determining the votes (Farrand 1966, 255).” The apparent compromise was to keep the clause with the one-fifth threshold. While the founders were certainly concerned about allowing for roll call voting, the recorded vote was not set as the default mechanism for voting in either chamber. Primarily, there have been two additional methods of voting in the House and the Senate: voice and division votes (Tiefer 1989).3 A division vote can be requested by any member. Once requested, members rise if they take the affirmative on a question and they are then counted by the chair. This process is repeated for those in opposition. Division votes are not recorded and the chair’s count of the votes cannot be appealed.4 The default mechanism for voting in Congress is the voice vote. During a voice vote, the chair will put forward two questions: “all in favor say ‘Yea’,” and “all opposed say ‘Nay’.” Like a division vote, the counting is up to the chair and cannot be appealed. While members may make their opinions clearly known, voice votes produce no record of individual positions on a given bill.5 As discussed above, to receive a recorded roll call vote in either chamber a member needs a second of “. . . one-fifth of those present.”6 In the House of the Representatives, 3 An additional method, teller voting, is restricted to the House of Representatives and is used infrequently. While it is likely to yield more accurate vote totals than either voice or division votes, it is similar to these in that it also does not produce a record of how members cast their votes (Tiefer 1989, 352). 4

In the House, the chamber has the opportunity to request a recorded vote only after the voice or division vote. If this is not done, the result stands. In the Senate, the recorded vote can be requested at any time. 5

Because of the similarities between division votes and voice votes, the two types of votes are treated the same in this study. 6

It should be noted that the constitutional provision for recorded roll call voting does not apply to amendment voting within the Committee of the Whole in the House, which developed its own separate procedures. We focus almost exclusively on final passage votes in this study, as Roberts and Smith (2003) have already convincingly demonstrated the limitations inherent in aggregating roll call voting from both

3

once a sufficient second is voiced a roll call vote is taken. This voting has largely been done electronically since 1972. When the vote is called, members insert a personalized voting card into a station on the House floor and press either present, yea or nay. Members’ votes are then displayed on panels throughout the chamber. While the Speaker does have authority to extend votes, few last longer than the 15 minute requirement (Oleszek 2004). In the Senate, once the yeas and nays are ordered, the clerk begins to call the names of each senator alphabetically. The Senator then has, generally, 15 minutes to respond to his or her name (Tiefer 1989).7

Voice Votes in Political Science The roll call voting record has become an invaluable tool for observers of legislative behavior. Since the mid-20th Century, interest groups like the Americans for Democratic Action (ADA) have used select recorded votes to estimate the position of members of Congress on a liberal-conservative scale (Lynch n.d.). Seeking to reduce selection bias, political scientists have expanded this initial usage of the record to characterize members’ voting behavior using the entire roll call record (Clausen 1973; Kingdon 1973; 1977). Perhaps the most widely used of these efforts, NOMINATE, employs a scaling technique to detail legislators’ preferences onto an underlying policy space (Poole and Rosenthal 1990). Similar recent efforts have utilized Bayesian statistical methods to identify congressional member ideology using the complete roll call record (Clinton, Jackman and Rivers 2004). More recently, scholarship has sought to test non-voting theories of legislative institutions with roll call data. These institutions include legislative agendas (Campbell, Cox the House and the Committee of the Whole. 7

In addition to voting yea, nay or present, members in both bodies have the option of vote pairing. In a “live pair” a member who is present is paired with a member who is absent and neither cast votes. Generally, the member who is present announces that he has a pair with the absent member, and should the absent member have attended, he would vote yeah or nay. The practice is less common in the modern era, but in the 19th and early 20th Century, members would occasional form general pairs for as long as several months (Luce 1922).

4

and McCubbins 2002; Cox and McCubbins 2005; Gailmard and Jenkins 2007) intra-chamber rules (Krehbiel 1998; Wawro and Schickler 2004; 2006), and political parties (Snyder and Groseclose 2000; McCarty, Poole and Rosenthal 2001; Cox and Poole 2002). Literature using the roll call record ignores the process by which issues come to receive a recorded vote, as opposed to receiving an unrecorded one. Indeed, the two primary methods of dealing with voice votes are to treat them as all being unanimously-supported issues (Mayhew 1991) or to omit them from analysis altogether (Krehbiel 1998; Cox and McCubbins 2005).8 Making inferences from longitudinal data like the roll call record requires an understanding of the data-generating process by which the roll call record was created (Morton 1999). Ignoring the process by which some measures receive a recorded vote and others do not can be highly problematic, but only if this process has fluctuated throughout the history of Congress and these fluctuations are not accounted for. Specifically, ignoring voice votes will lead to suspect use of the roll call record if (1) the rate of issues receiving roll call votes, relative to those receiving unrecorded votes, has fluctuated over time and (2) the type of bills receiving a recorded vote has fluctuated over time. Criticism of the roll call record because of inconsistencies in the rate of issues that receive recorded votes is not new. Cox and McCubbins (2005) argue that negative agenda control is the primary cause of majority party dominance. Their theory is largely bolstered by analyses of partisan roll rates on final passage votes from 1877-1995.9 Responding to the ability of Cox and McCubbins to then compare these roll rates between Congresses, Smith (2007) points out that the ratio of public laws to final passage votes greatly fluctuates over 8

Commendably, Wawro and Schickler (2004; 2006) use both methods, although this is still not without its drawbacks. 9

Partisan roll rates work accordingly: when over half of a party votes for (against) a measure, but the measure still loses (wins), this qualifies as the party being rolled. For example, if bill A is taken up and being voted on, and the minority party, say the Republicans, oppose it unanimously - but the majority party, say the Democrats, favor it unanimously, then the Republicans will be rolled, due to the Democrats seat advantage. Like many analyses that extend into the late 18th and early 19th century Congresses, additional uncertainty in caused by the high level of abstention of paired voting in the errors.

5

time. He concludes that, “Only in the modern period are even a majority of enacted bills subject to a vote at final passage” (141). Furthering Smith’s point about public laws, Clinton and Lapinski (n.d.) trace public laws through the bill stage and enactment. They find that from 1889 to 1994 the percentage of enacted laws that receive a roll call vote has varied significantly from Congress to Congress. Studies in comparative politics has been particularly sensitive to how selection biases may be introduced into the roll call record if certain types of issues are more likely to receive a roll call vote than others. Using data from all votes cast in the European Parliament from July 1999 to June 2000, Carrubba, Gabel, Murrah, Clough, Montgomery and Schmbach (2006) show that the rate of issues receiving a roll call vote are inconsistent across issue areas, by an issue’s level of salience, and by the party that introduced an issue. Hug (2006) draws similar inferences in his study of all roll call votes cast in the Swiss lower house. Interestingly, both studies suggest such selection biases may not be a problem in the United States, where “almost all legislative votes are by roll call.” (Carrubba et al. 2006, 691). Consistent with these works, Clinton and Lapinski (n.d.) show that percent of public laws receiving a roll call vote in Congress is inconsistent across issue areas and levels of salience.10 In the next section, we examine the rate of voice votes, the type of issues receiving voice votes, and whether these have remained consistent throughout congressional history.

Examining the Choice of a Recorded Vote In order to assess the data-generating process that creates the roll call record, we analyze the percentage of votes on all landmark legislation that goes unrecorded. Specifically, we examine how the percentage of issues that receive voice votes has changed over time, consider the contentiousness of the issues that received voice votes, and examine the issue areas that are most likely to receive voice votes. 10

Somewhat similarly, VanDoren questions the utility of roll call voting data using data derived from all congressional decisions on policy proposals in the atomic energy and coal policy issue domains (1990).

6

To conduct this analysis we use a list of landmark enactments from 1807 to 1990 provided by Stathis (2003). We opt for this list primarily because his extended time series allows greater leverage on early Congresses.11 Restricting the analysis to landmark enactments has two distinct advantages. First, it ensures that the data is not biased by the large amount of ‘banal’ legislation the United States Congress produces (Cameron 2000; Clinton and Lapinski 2006). Second, these lists were created independently of the roll call voting record and allows researchers leverage in tracking the propensity of unrecorded votes.

0

.2

Proportion Voice Votes .4 .6 .8

1

Figure 1: Proportion of Unrecorded Lowess House Votes on Landmark Legislation, 1807-1990 smoother

20

40

60 Congress

80

100

bandwidth = .8

Note: Proportion of all landmark legislation that passed the House by voice vote from 1807 (10th Congress) to 1990 (101st Congress). The solid line is a loess smoothing line indicating the average percentage of landmark legislation passed via voce. The list of landmark enactments is provided by Stathis (2003).

Figure 1 reports the percentage of landmark legislation that passed the House via voice vote. This figure shows that in early Congresses the percent of landmark legislation that passed by voice vote was surprisingly high. Around the 10th Congress, approximately 40% of landmark legislation passed by voice vote. It dropped to around 30% just before Civil War and then climbed back to an average of 40% around the turn of the century (56th Congress). 11

This contrasts with other compilations of landmark enactments like Peterson (2001), Mayhew (1991) and Howell et al. (2000) which start from a much more recent date. The data from those lists reveal similar patterns as the Stathis list for the time periods they cover.

7

It then slowly fell to around the 10% rate that is common in today’s House. Figure 2 reports the percentage of landmark legislation that passed the Senate via voice vote. The percent of landmark legislation that received a voice vote around the 10th Congress (1808) was a little under 30%. This average rate slowly rose to above 40% around the 60th Congress (1908). It then slowly fell to around the 10% rate that is common in today’s Senate.

0

.2

Voice Vote Proportion .4 .6 .8

1

Figure 2: Proportion of Unrecorded Lowess Senate Votes on Landmark Legislation, 1807-1990 smoother

20

40

60 Congress

80

100

bandwidth = .8

Note: Proportion of all landmark legislation that passed the Senate by voice vote from 1807 (10th Congress) to 1990 (101st Congress). The solid line is a loess smoothing line indicating the average percentage of landmark legislation passed via voce. The list of landmark enactments is provided by Stathis (2003).

First, it is clear from both Figures 1 and 2 that the rate of legislation receiving a voice vote has not been constant over the history of either the House or the Senate. The rate of voice votes in individual Congresses ranges from 0% to 100% and the moving average ranges from above 40% to around 10%. This fluctuation is clear evidence that the data-generating process that creates the roll call record has not been consistent throughout the history of Congress. Instead, the likelihood that a bill would receive a voice vote instead of a recorded vote has oscillated greatly throughout the history of Congress. Because this data-generating process has been inconsistent, studies that use longitudinal roll call data without accounting for these changes risk introducing biases into their analyses. 8

Of course, if the issues that are receiving voice votes are all universally supported measures, then failing to account for the changing rates of bills that receive a recorded vote may be less problematic.12 But this is not the case. To account for the contentiousness of the bills receiving voice votes in this study, we rely on the fact that bills receiving a final passage vote in one body of Congress, must also receive a final passage vote in the other body. We examine all bills that receive a final passage voice vote in the House (Senate) and check to see if that bill received a recorded final passage vote in the Senate (House). If a bill receiving a voice vote in the House (Senate) is unanimously supported, we would expect that bill to receive a unanimous recorded vote in the other chamber.13 Table 1: Coalition Size on Landmark Legislation, House

Coalition Size Observations Senate Coalition Size Observations

10th–51st 52nd–91st 92nd–101st 77.17% 82.84% 88.73% 22 42 8 67.59% 86.10% 90.42% 37 98 27

Total 81.77% 72 82.59% 162

Note: House (Senate) coalition size indicates the percent of members from the Senate (House) that voted for a measure that received a voice vote in the House (Senate). Data from Wawro and Schickler (2004).

We find that bills receiving a voice vote in one chamber do not receive unanimous support in the other chamber. Instead bills receiving a voice vote in one chamber are often highly contentious in the other chamber (see Table 1). Of the 72 pieces of landmark legislation that received a voice vote in the House but received a recorded vote in the Senate, the average coalition size that supported the measure in the Senate was 81.77%. There are 3 instances (out of 72 observations) when the coalition size in the Senate was less than 60% of its members, indicating highly contentious measures. The results are similar for the Senate. Of the 162 pieces of landmark legislation that received a voice vote in the Senate 12 If only unimportant issues were receiving voice votes this too would make the failure to account for the changing rates of recorded voting seem less problematic. However, the data we use is made up entirely of landmark legislation. This indicates that there are fluctuating (and surprisingly high) rates of voice votes taking place even on the most important legislation. 13

This relies on the reasonable assumption that the ideological distributions of the House and Senate are similar and have been similar throughout the range of data we examine here.

9

but received a recorded vote in the House, the average House coalition size was 82.59%. In 18 instances (out of 162 observations) the House passed bills with coalitions of less than 60% of its members. Interestingly, the average coalition size has become bigger throughout the history of Congress. This indicates legislation receiving a voice vote in the past was more likely to be contentious then legislation receiving a voice vote in more recent Congresses. It also indicates the while treating all voice votes as though they are unanimous votes may be appropriate for analysis of recent Congresses, it is wholly inappropriate for analysis of earlier Congresses. This change in the contentiousness of bills receiving an unrecorded over time again provides evidence on longitudinal inconsistencies in process by which bills are chosen to receive a recorded vote. Another way to assess the contentiousness of measures receiving a voice vote is to consider the percentage of these measures that undergo amendment voting. We argue that bills receiving amendments are more likely to be contentious than are measures that pass in an unadulterated format. Of the legislation that passed the House via voice vote, an average of 24.60% had an amendment that received a recorded vote before final passage (see Table 2). The number of measures that underwent amendment voting was much higher (75.28%) before the introduction of Reed’s Rules in the 52nd Congress.14 Similarly, the Senate voted to amend 29.01% of the voice voted landmark legislation examined in this study. In earlier Congresses, 77.14% of such legislation received a recorded attempt to amend the measure. To assess whether certain types of issues are more likely to receive a roll call vote than others,we sort landmark legislation into six issues areas (government management, social welfare, agricultural assistance, international involvement, civil liberties, and miscellaneous) defined by Clausen (1973) and report the percentage of the legislation that received a voice vote. Table 3 reports the results for the House and Table 4 reports the results for the Senate. 14 We chose to split the data at the 52nd Congress to account for an era of increased partisanship that accompanied the introduction of Reed’s Rules in this Congress. We also split the data at the 92nd Congress to account for the introduction of electronic voting in the House. We argue that this should decrease the use of voice voting, because of the decreased cost in terms of time required to hold a recorded vote.

10

Table 2: Amendment Voting on Voice Voted Landmark Legislation 10th–51st 52nd–91st 92nd–101st House 75.28% 3.66% 6.90% (67/89) (7/191) (2/29) Senate 77.14% 12.00% 12.50% (81/105) (30/250) (6/48)

Total 24.60% (76/309) 29.03% (117/403)

Note: Percentage of landmark legislation that underwent a recorded attempt at amendment previous to final passage. Data from Wawro and Schickler (2004).

There is clear evidence that the rate of legislation receiving a voice vote is not consistent across issue areas. Highly salient issue areas like the legislation on civil liberties receive a very low percentage of voice votes (5.56% in the House and 18.06% in the Senate), while the far less salient category of government management received a much higher percentage of voice votes (36.29% in the House and 46.17% in the Senate).15 This strongly suggests that since some, arguably more salient, issues are more likely to receive roll call votes then other issue areas, selection biases are being introduced into the roll call record. These problems are exacerbated by the fact that the likelihoods of various areas receiving a recored vote has fluctuated over time. While 37.31% of government management legislation received a voice vote in the period from the 10th Congress to the 51st Congress, that rate had fallen to 11.63% from the 92nd Congress to the 101st Congress. In this section we find strong evidence that rate of legislation receiving a recorded vote has fluctuated greatly across congressional history. All of this evidence calls into question the consistency of the data-generating process that creates the roll call record and calls into question the validity of longitudinal use of the roll call record as a source of data. In the next section we employ multivariate analysis to check if these inconsistencies remain influential once other explanatory factors are considered.

15 The extremely low percentage of miscellaneous issues that receive a voice vote (0.00% for the House and 8.33% for the Senate) can be discounted due to the limited number of observations for this category.

11

Table 3: Percent of Voice Votes by Issue Area, House

Gov’t Management Social Welfare Agriculture Foreign Affairs/Defense Civil Liberties Miscellaneous Total

10th–51st 52nd–91st 92nd–101st 37.31% 42.53% 11.63% (72/193) (131/308) (10/86) 100.00% 21.51% 17.07% (3/3) (20/93) (14/82) 16.67% 19.23% 0.00% (1/6) (5/26) (0/2) 32.35% 34.74% 15.63% (11/34) (33/95) (5/32) 5.13% 9.09% 0.00% (2/39) (2/22) (0/11) 0.00% 0.00% 0.00% (0/3) (0/0) (0/2) 32.01% 35.11% 13.49% (89/278) (191/544) (29/215)

Total 36.29% (213/587) 20.79% (37/178) 17.65% (6/34) 30.43% (49/161) 5.56% (4/72) 0.00% (0/5) 29.80% (309/1037)

Note: Percent of landmark legislation receiving a voice vote sorted by issue area. Issue areas are defined by Clausen (1973). Issue area coding provided by voteview.com.

Table 4: Percent of Voice Votes by Issue Area, Senate

Gov’t Management Social Welfare Agriculture Foreign Affairs/Defense Civil Liberties Miscellaneous Total

10th–51st 52nd–91st 92nd–101st 43.01% 53.57% 26.74% (83/193) (165/308) (23/86) 66.67% 33.33% 23.17% (2/3) (31/93) (19/82) 66.67% 42.31% 0.00% (4/6) (11/26) (0/2) 21.15% 29.57% 15.79% (11/52) (34/115) (6/38) 12.82% 27.27% 18.18% (5/39) (6/22) (2/11) 0.00% 0.00% 50.00% (0/4) (0/6) (1/2) 35.35% 43.33% 23.08% (105/297) (247/570) (51/221)

Total 46.17% (271/587) 29.21% (52/178) 44.12% (15/34) 24.88% (51/205) 18.06% (13/72) 8.33% (1/12) 37.04% (403/1088)

Note: Percent of landmark legislation receiving a voice vote sorted by issue area. Issue areas are defined by Clausen (1973). Issue area coding provided by voteview.com.

12

Modeling the Choice to Record Understanding the factors that explain whether or not a bill will receive a recorded vote is the first step towards accounting for inconsistencies in the roll call record’s datagenerating process. In this section, we estimate a model to predict when votes will be recorded. We examine the impact of inter-chamber differences, issue areas, workload, and significant historical events, such as the introduction of Reed’s Rules and electronic voting. Differences in the operation of the House and Senate lead to differing expectations in the use of voice and recorded votes between the two chambers. Even before electronic voting, the introduction of Reed’s Rules in the 51st House of Representatives, and readoption of them in the 54th, the House provided majority parties strong agenda-control powers. This, along with amendment voting being concentrated in the Committee of the Whole, may provide the House with greater legislative stability, and thus, far lower use of unrecorded voting methods. While the Senate has never instituted electronic voting, the process of garnering onefifth present for a recorded vote is often considered pro-forma today (Smith 1987). However, this is not necessarily true throughout the chamber’s history. Tiefer notes that during the 1950s it was extremely difficult to for senators to obtain roll call votes without the support of either majority leader Lyndon Johnson (D-TX) or minority leader Everett Dirksen (R-IL) (1989, 532). Further, because the chamber often operates by unanimous consent, there are more strategic opportunities to utilizing voice voting. One example of this, not-surprisingly involving Lyndon Johnson, is detailed by Robert Mann (1996):

Typical was the day in 1955 when Johnson realized that Florida senator Spessard Holland, the chief opponent of legislation to increase the minimum wage, was absent from the Senate chamber. Late in the afternoon, Johnson looked around and said to [Minnesota Democrat Hubert] Humphrey, “I think we’ll pass that minimum wage bill now.” After a brief quorum call, Johnson engineered a voice vote. The bill passed. Minutes later, when Holland came storming onto the Senate floor, “jumping, screaming, hollering and pounding the desk,” Johnson said, “Well, Spessard, I had a little quorum call. If you fellows aren’t on the job 13

around here, I’ve got legislation to pass (p. 137).” The primary implication of these inter-chamber differences is that we would expect to see more frequent roll call voting in the House of Representatives. However, it is clear there may be reasons members would want some things on the record and other things off the record in both chambers. To account for the fact the members may prefer to be “on the record” for salient issues we control for issue areas, as discussed above. Specifically, we control for legislation on civil liberties because of the high probability that such measures will garner close scrutiny from constituents. Additionally, we control for measures focusing on international issues. Clinton and Lipinski find in an analysis of 3,500 “most significant” statutes that statutes dealing with international issues have an 11.5% higher probability of receiving a recorded vote (2007). We want to see if this results holds for landmark legislation. We also control for important historical events that are likely to influence the decision to hold a recorded vote. In 1973 the House instituted electronic voting. After electronic voting, the time that it took to physically hold a recorded vote decreased, which should increase the likelihood of a bills receiving a recorded vote. In a full model (Model 4), we control for this by including a dummy variable for all landmark legislation passed after the introduction of electronic voting and by including a House and electronic voting interactive variable to account for the fact that only the House has electronic voting. We also estimate a partial model for all Congresses after the introduction of electronic voting (92nd–101st, Model 3). Similarly, we control for the introduction of Reed’s Rules and the era of increased partisan control associated with the introduction of these rules by estimating a partial model Congresses before the introduction of Reed’s Rules (10th–51st, Model 1) and a partial model for the period between Reed’s Rules and the introduction of electronic voting (52nd–91st, Model 3). Additionally, we control for changing workloads throughout the history of Congress. The estimates from these models are presented in Table 5. 14

In the full model (Model 4), the House is more likely to use recorded voting than is the Senate. As expected, legislation on civil rights and foreign affairs are more likely to receive a recorded vote than other less contentious issues. Also, the introduction of electronic voting in the House made it more likely that votes in the House would be recorded. Congressional workload did not have an impact on the decision to hold a recorded vote. Table 5: Probit Estimates of Recorded Vote, 10th-101st Congresses Model 1 Model 2 Model 3 Model 4 10th–51st 52nd–91st 92nd–101st 10th–101st House 0.113 0.227∗ 0.388∗ 0.185∗ (0.075) (0.055) (0.074) (0.044) ∗ ∗ ∗ Civil Liberties 1.140 0.715 0.623 0.915∗ (0.256) (0.231) (0.207) (0.169) Foreign Affairs 0.389∗ 0.245∗ 0.190 0.288∗ (0.184) (0.123) (0.186) (0.090) ∗ ∗ Workload -0.141 0.226 1.072 0.061 (0.178) (0.075) (0.530) (0.049) Electronic Voting — — — 0.475∗ (0.134) Electronic*House — — — 0.196∗ (0.082) Constant 0.110 -0.091 -0.555 0.109∗ (0.128) (0.085) (0.613) (0.052) Observations 575 1114 436 2125 2 Pseudo R 0.053 0.030 0.041 0.050 2 Prob > χ 0.000 0.000 0.000 0.000

Note: Coefficients with robust standard errors clustered on Congress listed in parentheses, ∗indicates significance at the p = .05 level. The list of landmark enactments was provided by Stathis (2003). A vote is unrecorded if there was no applicable final passage vote on the bill or conference report.

In the partial models, civil liberties legislation was more likely to receive a recorded vote in all three time periods. Differences in the significant variables in the three time periods show further evidence of the changing roll call record data-generating over time. While foreign affairs legislation was more likely to receive a recorded vote in the two periods before electronic voting, it is not a significant factor after electronic voting. Similarly, the House was more likely to employ a recorded vote than the Senate after Reed’s Rules, but 15

not before.

Discussion and Conclusion The above analysis shows that the usage of voting mechanisms in Congress is, as the literature expects, largely consistent in the modern era. However, in historical Congresses, there appears to be substantial variation amongst the usage of recorded voting. It is clear that certain issue areas (civil liberties, foreign affairs) receive recorded votes at a greater proportion than other issue areas. It also clear that the percentage of landmark enactments receiving final passage votes fluctuates greatly in early Congresses. This calls into question studies that seek to compare roll call voting results between Congresses. Future work will seek to explore the implications of this on two levels. First, we will look to expand our work on democratic transparency by supplementing this with analyses looking at the fluctuating usage of abstention and vote pairing, as well as increased activity at the conference committee level. Additionally, we contrast the increase in roll call voting with an increased usage of omnibus legislating. Second, we believe these results have serious implications for future studies seeking to test theories using roll call voting data. One method for minimizing errors in this respect would be to parse the roll call record into various eras or issue subsets. Another would involve imbedding the process into models of roll call voting (Clinton 2006). Either approach would demonstrate the proper deference to the relatively disjointed process of what comes to a roll call vote.

16

Works Cited Binder, Sarah A. 1997. Minority Rights, Majority Rule: Partisanship and The Development of Congress. New York: Cambridge University Press. Cameron, Charles M. 2000. Veto Bargaining: Presidents and the Politics of Negative Power. Cambridge: Cambridge University Press. Campbell, Andrea C., Gary W. Cox, and Mathew D. McCubbins. 2002. “Agenda Power in the U.S. Senate, 1877-1986.” In David W. Brady and Mathew D. McCubbins, Eds., Party, Process, and Political Change in Congress: New Perspectives on the History of Congress. Stanford: Stanford University Press. Carruba, Clifford J., Matthew Gabel, Lacey Murrah, Ryan Clough, Elizabeth Montgomery, and Rebecca Schambach. 2006. “Off the Record: Unrecorded Legislative Votes, Selection Bias and Roll-Call Vote Analysis.” British Journal of Political Science. 36: 691-704. Clausen, Aage R. 1973. How Congressmen Decide: A Policy Focus. New York: St. Martens Press. Clinton, Joshua. 2007. “Lawmaking and Roll Calls.”. Journal of Politics. 69(2): 455-67. Clinton, Joshua, Simon Jackman, and Doug Rivers. 2004. “The Statistical Analysis of Roll Call Data.” American Political Science Review. 98(2):355-370. Clinton, Joshua and John Lapinski. 2006. “Measuring Legislative Accomplishment, 18771994.” American Journal of Political Science. 50 (1):232-49. Clinton, Joshua and John Lapinski. n.d. “The Relationship Between Laws and Roll Calls in the U.S. Congress, 1889-1994.” unpublished manuscript. Cox, Gary W. and Mathew D. McCubbins. 2005. Setting the Agenda: Responsible Party Government in the U.S. House of Representatives. Cambridge: Cambridge University Press. Cox, Gary W. and Keith T. Poole. 2002. “On Measuring Partisanship in Roll Call Voting: The U.S. House of Representatives, 1877-1999.” American Journal of Political Science. 46: 477-489. Farrand, Max. 1966. The Records of the Federal Convention of 1787, Volume II. New Haven, CT: Yale University Press. Gailmard, Sean and Jeffery A. Jenkins. “Negative Agenda Control in the Senate and House: Fingerprints of Majority Party Power.” Journal of Politics. 69(3): 689-700.

17

Hug, Simon. 2006. “Selection Effects in Roll Call Votes.” Center for Comparative and International Studies Working Paper. Kingdon, John W. 1973. Congressmen’s Voting Decisions. New York: John Wiley and Sons. Kingdon, John W. 1977. “Models of Legislative Voting.” Journal of Politics. 39(3): 563-595. Krehbiel, Keith. 1998. Pivotal Politics: A Theory of U.S. Lawmaking. Chicago: University of Chicago Press. Luce, Robert. 1922. Legislative Procedure. Boston: Houghton Mifflin. Lynch, Michael S. n.d. “Are They Asking the Right Questions: Assessing Interest Group Score Using Item Response Theory.” unpublished manuscript. Mann, Robert. 1996. The Walls of Jericho: Lyndon Johnson, Hubert Humphrey, Richard Russell, and the Struggle for Civil Rights. New York: Harcourt Brace and Co. Mayhew, David R. 1991. Divided We Govern: Party Control, Lawmaking and Investigating: 1946-1990. New Haven: Yale University Press. McCarty, Nolan, Keith Poole, and Howard Rosenthal. 2001. “The Hunt for Party Discipline” American Political Science Review. 95(3): 673-687. Morton, Rebecca. 1999. Methods and Models: A Guide to Empirical Analysis of Formal Models in Political Science. Cambridge: Cambridge University Press. Poole, Keith T. and Howard Rosenthal. 1997. Congress: A Political-Economic History of Roll Call Voting. New York: Oxford Press. Roberts, Jason M. and Steven S. Smith. 2003. “Procedural Contexts, Party Strategy, and Conditional Party Voting in the U.S. House of Representatives, 1971-2000.” American Journal of Political Science. 47(2): 305-317. Smith, Steven S. 1989. Call to Order: Floor Politics in the House and Senate. Washington, D.C.: The Brookings Institution. Smith, Steven S. 2007. Party Influence In Congress. New York: Cambridge University Press. Snyder, James M. and Tim Groseclose.“Estimating Party Influence in Congressional RollCall Voting.” American Journal of Political Science. 44(2): 193-211. Stathis, Stephen W. 2003. Landmark Legislation 1774-2002: Major U.S. Acts and Treaties. Washington, D.C.: CQ Press.

18

Tiefer, Charles. 1989. Congressional Practice and Procedure. New York City, NY: Greenwood Press. VanDoren, Peter M. 1990. “Can We Learn the Causes of Congressional Decisions from RollCall Data?” Legislative Studies Quarterly. 15(3): 311-340. Wawro, Gregory and Eric Schickler. 2004. “Where’s the Pivot? Obstruction and Lawmaking in the Pre-cloture Senate.” American Journal of Political Science. 48 (4): 758-774. Wawro, Gregory and Eric Schickler. 2006. Filibuster: Obstruction and Lawmaking in the U.S. Senate. Princeton University Press.

19

Appendix A: Unrecorded Votes on Landmark Enactments, 1807-1991 Year House 1807 4 1809 3 1811 15 1813 9 1815 10 1817 8 1819 6 1821 3 1823 2 1825 1 1827 1 1829 4 1831 3 1833 3 1835 6 1837 7 1839 1 1841 7 1843 5 1845 8 1847 3 1849 6 1851 4 1853 4 1855 3 1857 3 1859 6 1861 18 1863 10 1865 9 1867 8 1869 15 1871 11 1873 6 1875 3 1877 7 1879 1 1881 11

Unrecorded House Senate 0 4 2 3 5 15 4 10 4 11 6 10 3 6 2 3 0 2 0 1 0 1 1 4 0 4 3 3 3 6 1 7 0 1 1 8 1 6 4 9 1 5 0 7 2 4 1 7 0 3 0 4 1 6 6 18 4 10 0 9 0 9 3 15 4 12 0 6 2 4 0 7 0 1 4 12

20

Unrecorded Senate 1 0 5 4 5 6 2 2 0 0 0 1 0 1 2 2 0 1 1 3 1 1 3 4 1 0 2 5 4 1 0 5 6 1 2 4 0 6

Appendix A: Unrecorded Votes on Landmark Enactments, 1807-1991 (cont) Year House 1883 7 1885 11 1887 6 1889 20 1891 6 1893 10 1895 3 1897 13 1899 8 1901 14 1903 2 1905 13 1907 8 1909 14 1911 12 1913 13 1915 16 1917 18 1919 12 1921 20 1923 12 1925 10 1927 10 1929 9 1931 10 1933 33 1935 21 1937 18 1939 16 1941 23 1943 15 1945 17 1947 9 1949 12 1951 9 1953 13 1955 11 1957 13 1959 8

Unrecorded House Senate 4 7 5 11 4 6 8 20 3 6 7 10 3 3 5 14 1 9 5 15 2 3 9 14 4 8 8 14 8 12 6 13 6 16 2 18 1 12 6 23 6 12 3 11 7 11 5 10 0 10 14 33 9 21 10 18 7 16 7 23 6 15 8 18 2 10 4 13 3 10 5 13 3 15 5 14 2 9

21

Unrecorded Senate 3 7 5 8 4 7 3 5 4 10 2 10 7 7 5 7 8 3 4 11 5 5 6 4 5 15 11 11 6 8 7 9 3 6 6 6 5 7 1

Appendix A: Unrecorded Votes on Landmark Enactments, 1807-1991 (cont) Year House 1961 16 1963 13 1965 25 1967 15 1969 24 1971 18 1973 21 1975 20 1977 24 1979 21 1981 18 1983 28 1985 22 1987 23 1989 20 1991 22

Unrecorded House Senate 1 17 1 14 4 25 1 17 2 25 0 19 0 21 3 20 3 26 2 21 1 18 8 28 5 22 3 24 4 22 4 23

22

Unrecorded Senate 3 4 7 3 7 3 2 4 4 6 3 12 7 5 5 5