Predicting The Senate Vote on John Roberts

Royce Carroll, Jeff Lewis, Keith Poole, and Howard Rosenthal

29 September 2005

In 1987 two of us (Poole and Rosenthal, 1987, reproduced below) forecast the Senate confirmation vote on Judge Robert Bork based upon Senators' announced positions and an early one-dimensional version of the NOMINATE method of estimating congressional ideal points. We forecast that Judge Bork would fail to be confirmed by a vote of 41 for and 59 against. In the aggregate, this forecast was quite close to the actual vote of 42 for and 58 against and produced only 7 forecasting errors. Below we apply the same methodology to forecast the confirmation vote for Judge John Roberts.

We forecast the Roberts confirmation vote using the two-dimensional Poole and Rosenthal DW-NOMINATE ideological scores (technical details). The result is shown below. The predicted cutting line is based upon announced positions as of September 22. This cutting line (the best spatial division using the early announcements) predicted a vote of 69 for and 31 against. This forecast produced 11 voting "errors", that is, inconsistencies with their past voting behavior. Only one Senator, Bayh (D-IN), was predicted to vote "Yea" but announced "Nay". However, 10 Senators were predicted to vote "Nay" but announced "Yea" [9 Democrats and Jeffords (I-VT)]). Although the cutting line was forecast a week before the vote it remains the best line for forecasting based upon the spatial theory of voting, which predicts geometric consistency within the patterns of roll call votes from the same individuals. That is, even with the result known, this line cannot be repositioned to increase correct spatial classification.

To place this prediction in historical context, see below for discussions of the spatial predictability of other judicial confirmations. Although each of the votes produces deviations from spatial voting, the Roberts vote produced 10 of 11 errors on one side of the cutting line ("Nay"). This error pattern should be viewed in the context of the upcoming battle over the O'Connor seat. If Roberts was a one-shot vote so that strategic considerations were not present it is quite likely that most of the errors on the predicted Nay side of the line would not be there.

Note that drawing a cutting line parallel to the Roberts confirmation line but passing just to the left of Nelson (D-NE) and McCain (R-AZ) puts the five most moderate Republicans on the predicted "Nay" side of the line with a forecast of 50 - 50. This division is a likely line-up for a possible vote on the "Nuclear Option."

Our forecast cutting line is based upon announcements through 22 September 2005

Important Roll Calls on Judicial Nominations Prior to the Bush Administration

In the following sections, we place both the Roberts vote and our method of prediction in historical perspective, analyzing several of the more contentious judicial nomination votes over the last several decades.

The vote to invoke Cloture on the nomination of Abe Fortas for Chief Justice was recently in the news during the debate surrounding Senate rules changes to prevent judicial filibusters. President Lyndon Johnson nominated Fortas to replace Earl Warren as Chief Justice. The nomination became especially controversial when it was revealed that Fortas had accepted significant speaking fees, leading to the first ever filibuster of a Supreme Court nomination. On 1 October 1968 a Cloture motion failed 45 - 43 (two-thirds were required to invoke cloture). This roll call is a classic example of a "conservative coalition" vote, whereby Southern Democrats and most Republicans voted against Northern Democrats and liberal Republicans. Note that there are only 5 classification errors, all very close to the cutting line of the roll call.

Clement F. Haynesworth was nominated by President Richard Nixon in 1969. Facing conflict of interests charges and suspicions of being a segregationist, Haynesworth was rejected by the Senate by a vote of 45 - 55. This roll call is very similar to the Fortas roll call only with more classification errors -- 15. However, with exception of one Republican, the errors are again all close to the cutting line.

G. Harold Carswell was nominated by President Richard Nixon in 1970. Regarded as mediocre by many and having explicitly endorsed segregation in a 1948 political campaign, Carswell was rejected by the Senate by a vote of 45 - 51 . This nomination provides yet another example of a conservative coalition vote, with 13 classification errors all but two being very close to the cutting line.

Daniel A. Manion was nominated by President Ronald Reagan for the 7th Circuit Court of Appeals in 1986. Manion was intensely controversial because of a number of rulings as a state judge in Indiana. He was initially confirmed on 26 June 1986 by a vote of 48 - 46 after Robert Byrd switched from Nay to Yea so that he could make a motion to reconsider. Consequently, on 23 July 1986 there was a motion to reconsider the vote and Vice President George H. W. Bush voted to break a 49 - 49 tie. This vote is shown below. Note that the cutting line is nearly straight up and down reflecting the ongoing realignment of the South into the Republican Party. There are only 5 classification errors and they are all close to the cutting line.

William Rehnquist was nominated by President Ronald Reagan to replace Warren Burger as Chief Justice. Despite the reemergence of a controversy relating to Rehnquist's views on segregation in the early 1950s, the vote to confirm was 65 - 33. The cutting line is nearly up and down picking up some of the Southern Democrats. There are only 4 classification errors and they are all close to the cutting line.

Robert Bork was nominated by President Ronald Reagan in 1987. An intense and infamous confirmation battle resulted revolving around Bork's views (this case is discussed in detail below). Bork was rejected by the Senate on 23 October 1987 by a vote of 42 - 58. There are only 4 classification errors on the roll call and the roll call cutting line is nearly vertical, again reflecting the emerging preeminence of the party divide.

The Bork Nomination and the Enduring Liberal-Conservative Conflict

Keith T. Poole and Howard Rosenthal
Graduate School of Industrial Administration
Carnegie Mellon University
Working Paper # 14-87-88
October 1987

How members of the Senate responded to the Bork nomination can be understood nearly entirely as a matter of basic, long-run liberal-conservative orientations. Using NOMINATE, a one dimensional scaling algorithm for binary choice data (Poole and Rosenthal, 1985) we scaled the senators based on their 1985 voting records. NOMINATE results have been available to other researchers for some time and form the basis for the analysis in Macdonald and Rabinowitz (1987). The results match up perfectly with the outcome of the Senate committee hearings. The committee vote, in conjunction with the scaling, was used to predict that Judge Bork would fail to be confirmed by a 59-41 vote. The actual vote was 58-42.

The Committee

As the hearings opened, five senators, Kennedy, Eiden, Simon, Leahy, and Metzenbaum, declared their opposition to Bork; four, Byrd, Deconcini, Heflin, and Specter were undecided; and five, Simpson, Hatch, Humphrey, Grassley, and Thurmond supported the Reagan nominee. (See Table 1.) Table 1 shows that the scaled values of the five "against" were all to the left of the four undecided who were all to the left of the five Bork supporters. The first undecided senator to declare himself was Specter of Pennsylvania. Since Specter was to the right of Byrd, Deconcini, and Heflin, one could predict that the other three senators would also come out against Bork. They did, and the final committee vote was 9-5 against.

Fourteen of the fifteen senators on the committee were already serving in 1983. Our 1983 scaling shows, for these 14, exactly the same order as in 1985, demonstrating the great stability of the liberal-conservative dimension. Consequently, all the information about the potential voting alignments of the 14 most senior members was already "known" as early as 1983. The 15th member, Simon of Illinois, was generally known to be extremely liberal; he was a sure bet to have opposed Bork.

Predicting the Full Senate

In 1985, 51 senators were to the left of Specter and 41 to the right of Grassley. Consequently, 52 senators would be predicted to vote No in a final confirmation vote and 42 would be predicted to vote favorably to Judge Bork. If we split evenly the four senators (Heinz, Durenberger, Cohen, and Stafford) between Specter and Grassley, we would arrive at a 54-46 prediction. Interestingly, this could suggest the 1986 elections were not crucial to Bork's defeat. In other words, Specter was already to the right of the median of the 99th Senate. Of course, the 1986 elections may have been important for a variety of reasons. Specter and others, both inside and outside the Senate, may have interpreted the elections as a signal that the mood of the country had changed and that they would not find it politically damaging to differ with the White House. The committee chairmanship had changed hands. And, perhaps most importantly, the 1986 elections drove home the importance of black constituents to Southern Democrats.

What might have happened in the 99th Senate is speculation. What would actually happen when the 100th Senate came to a vote? Making a prediction is complicated by the fact that 14 new senators entered the freshman class of 1987. Two Democrats on Specter's left were replaced by Republicans. We predicted that these entrants would vote for confirmation. Two Democrats and one Republican on his left were replaced by Democrats; for these three senators, we predict an against vote. To Grassley's right, seven Democrats replaced Republicans. All these new Democrats were predicted to vote against confirmation. John McCain, a Republican who replaced Barry Goldwater, was predicted to vote for. The net effect of all the replacements is a gain of five votes against Bork's appointment. As a result, our prediction, based on the committee vote, was 59-41 against confirmation.

The Actual Vote

The actual vote was 58-42. Off by one vote in the aggregate, we misclassified seven individual senators. Most of the errors were offsetting. Favorable votes by Boren, Hollings, and Hatfield, all to Specter's left, were offset by negative votes from Chafee, Packwood, and Warner, all to Grassley's right. The group between Specter and Grassley split 3-1 rather than 2-2. Ex post classification errors are minimized by a "cutting line" between Specter and Heinz. In addition to the six offsetting errors, the cutting line prediction also results in Stafford being counted as an error.

True, predicting a strict party line vote would have done just as well. The party prediction would generate two more errors, Specter and Weicker, while correcting the error on Hatfield. So, in this case, a party model would make only one more error than a liberal-conservative model. But this comparison misses the interesting information in the liberal-conservative model. The liberal-conservative model implies that the errors will occur near the cutting line (Poole and Rosenthal, 1985). Disregarding replacements, we find that no errors occur among the 38 senators to the left of Boren nor among the 25 senators to the right of Warner. The seven "errors" are concentrated among the 23 senators in the middle of the spectrum.

The critical role of the middle is also shown in the sequence in which the senators announced their votes. On October 8, the New York Times reported that a large number of senators had made commitments the previous day. These were all either returnees from the 99th Congress concentrated near the middle of the dimension (Sasser, Chiles, Dixon, Exon, and Hatfield) or newly elected senators (Breaux, Conrad, Graham, Fowler, and Shelby) who could be expected to be cue takers rather than cue givers. On October 9, the Times published an exhaustive list of voting intentions as of the previous day. This establishes that, after October 7, only Proxmire, Stennis, Nunn, Heinz, D'Amato, Evans, and Warner were uncommitted among returnees. These late declarers were highly concentrated near the middle of the dimension. While reporting of commitments made prior to October 7 was less systematic, those senators who committed in October appear to be concentrated in the middle (See Table 1 for further details). Strong liberals and strong conservatives appear to have generally made their positions known in September or earlier.

Rehnquist and Manion

As a further check on the consistency of liberal-conservative positions as they apply to judicial confirmations, consider the Manion and Rehnquist votes in 1986. Confirmation of Daniel Manion as an Appeals Court judge was decided by Vice President Bush breaking a tie to confirm. The optimal cutting line on this vote is between Hollings and Heflin. There are only four classification errors (Zorinsky, Specter, Evans, and Kassebaum). A party model would have corrected the Zorinsky error but made errors on Long, Heflin, Weicker and Mathias. Again, the liberal-conservative model has errors that are contained in the middle of the spectrum.

A similar story applies to Rehnquist. Here, since the vote was not close, the errors are not in the middle of the spectrum but near the cutting line, as expected. The errors are Pell, Leahy, Bumpers, Johnston, and Exon. The cutting line is between Glenn and Ford. In this case, the party model does poorly. While the party model "corrects" the Johnston and boon errors, it makes additional errors an Ford, Bingaman, Chiles, Dixon, DeConcini, Bentsen, Proxmire, Stennis, Nunn, Hollings, Zorinsky, and Long.

Two Flashes from the Past

The last highly contested confirmation votes prior to the Reagan era involved the nominations of Haynesworth and Carswell to the Supreme Court by President Nixon. There were only 17 senators who served in 1985 who were also serving in this earlier period. Even if we look backwards and do not allow for changes in position over a 16 year period, we are still able to classify correctly 14 of the 17 senators. One of the errors, Byrd of West Virginia, clearly represents a rare case of spatial movement, a move likely connected to becoming majority leader. Again the errors are in the middle. indeed, the two remaining errors, Hatfield and Packwood, both of Oregon, were also errors on the Bork vote.

Summary and Discussion

What our analysis shows is that voting on judicial confirmations follows a persistent, temporally stable, ideological pattern. Contrary to conservative claims that the Bork case represented a fundamental change in the process, ideological voting is dejŠ vu. Once a nomination becomes controversial, the pattern of voting alignments will obey liberal-conservative consistency. What is not foreordained is the process that makes a nomination controversial. A year prior to Bork, a perhaps equally conservative jurist, Antonin Scalia, was confirmed 98-0. Similarly, while the Rehnquist nomination was debated, the cutting line fell far to the left of the Bork line. Those who face the practical concern of insuring a nominee's approval are probably wasting energy on rhetoric that argues that a nominee's political views should not influence the vote. Their attention would be better focused on getting a favorable location of the cutting line. Where the cutting line falls, as the Scalia-Bork contrast suggests, reflects the framing and timing of a nomination as well as more substantive concerns.


Macdonald, Stuart Elaine and George Rabinowitz, 1987. "The Dynamics of Structural Realignment", American Political Science Review, 81, 779-796.

Poole, Keith T. and Howard Rosenthal, 1985. "A Spatial Model for Legislative Roll Call Analysis." American Journal of Political Science, 29, 357-84.

Table 1.   Liberal-Conservative Positions and Confirmation Votes 1986-87
                         ----- Bork ----- 
                       Committee     Senate
          1985 Pos.  Early   Final  Final
                     Stance  Vote   Vote  Manion Rehnquist 
SARBANES     -0.917                 A <7     A     A      
HART, G      -0.907                (A <7)    A     A     Repl. by Wirth, Pred. Against
SIMON, P     -0.812     A     A     A <7     A     A      
INOUYE       -0.806                 A <7     A     A      
METZENBAUM   -0.795     A     A     A <7     A     A      
MELCHER      -0.789                 A <7     A     A      
HARKIN, T    -0.789                 A <7     A     A      
MATSUNAGA    -0.789                 A <7     A     A      
KENNEDY, E.  -0.764     A     A     A <7     A     A      
BURDICK      -0.748                 A <7     A     A      
LEVIN, C     -0.737                 A <7     A     A      
RIEGLE       -0.732                 A <7     A     A      
KERRY, J     -0.729                 A <7     A     A      
CRANSTON     -0.727                 A <7     A     A      
PELL         -0.703                 A <7     A     P      
EAGLETON     -0.700                (P <7)    A     A     Replaced by Bond, Pred. For
LAUTENBERG   -0.693                 A <7     A     A      
MOYNIHAN     -0.636                 A <7     A     A      
MITCHELL, G  -0.619                 A <7     A     A      
DODD, C JR   -0.613                 A <7     A     A      
BAUCUS, M    -0.610                 A <7     A     A      
ROCKEFELLER  -0.608                 A <7     A     A      
LEAHY        -0.596     A     A     A <7     A     P      
BIDEN J      -0.592     A     A     A <7     A     A      
GORE, A      -0.590                 A <7     A     A      
SASSER       -0.579                 A =7     A     A      
PRYOR, D     -0.572                 A <7     A     A      
BUMPERS      -0.545                 A <7     A     P      
BYRD         -0.527     U     A     A <7     A     A      
BRADLEY, W   -0.526                 A <7     A     A      
GLENN        -0.520                 A <7     A     A     
----------------------------------------------------------- Optimal Cut Line Rehnquist
FORD         -0.473                 A <7     A     P     
BINGAMAN     -0.471                 A <7     A     P      
CHILES L     -0.440                 A =7     A     P      
JOHNSTON J   -0.403                 A =1     A     A      
DIXON, A     -0.383                 A =7     A     P      
EXON, J      -0.380                 A =7     A     A      
DECONCINI    -0.365     U     A     A <7     A     P      
MATHIAS      -0.346                (A <7)    A     P     Repl. by Mikluski, pred. Ag.
BENTSEN L    -0.323                 A =1     A     P      
BOREN, D     -0.309                 P <7     A     P      
WEICKER L    -0.307                 A =6     A     P      
PROXMIRE     -0.304                 A >7     A     P      
STENNIS      -0.284                 A >7     A     P      
NUNN S       -0.278                 A >7     A     P      
HOLLINGS     -0.263                 P <7     A     P     
----------------------------------------------------------- Optimal Cut Line Manion
HEFLIN, H    -0.238     U     A     A <7     P     P     
HATFIELD     -0.178                 P =7     P     P     
ZORINSKY     -0.175                (P <7)    A     P     Repl. by Karnes, pred. For.
LONG, R.     -0.161                (A =7)    P     P     Repl. by Breaux, pred. Ag.
ANDREWS, M   -0.141                (A =7)    P     P     Repl. by Conrad, pred. Ag.
SPECTER, A   -0.140     U     A     A =1     A     P     
--------------------------------------------------------------- Ex Post Opt. Cut Line on Bork
HEINZ        -0.075                 P >7     P     P     After the committee vote,
DURENBERGER  -0.051                 P <7     P     P     these four senators were
COHEN, W     -0.048                 P <7     P     P     predicted to split 2-2,
STAFFORD R   -0.033                 A =6     P     P     but in fact split 3-1.
GRASSLEY, C  -0.032     P     P     P >7     P     P      
CHAFEE       -0.024                 A =5     P     P      
DAMATO, A     0.015                 P >7     P     P      
PACKWOOD      0.028                 A =4     P     P      
KASTEN, R     0.032                 P <7     P     P      
DANFORTH      0.103                 P <7     P     P      
PRESSLER, L   0.104                 P <7     P     P      
STEVENS       0.139                 P <7     P     P      
EVANS, D      0.158                 P <7     A     P      
KASSEBAUM     0.164                 P <7     A     P      
HAWKINS, P    0.164                (A =7)    P     P     Repl. by Graham, pred. Ag.
ABDNOR, J     0.167                (A <7)    P     P     Repl. by Daschle, pred. Ag.
GORTON, S     0.200                (A <7)    P     P     Repl. by Adams, pred, Ag.
WARNER, J     0.241                 A >7     P     P      
COCHRAN, T    0.270                 P <7     P     P      
RUDMAN, W     0.299                 P <7     P     P      
MURKOWSKI     0.304                 P <7     P     P      
NICKLES, D    0.315                 P <7     P     P      
TRIBLE, P     0.321                 P <7     P     P      
MCCONNELL     0.329                 P <7     P     P      
DOSCHWITZ     0.333                 P <7     P     P      
ROTH W        0.352                 P <7     P     P      
MATTINGLY     0.396                (A =7)    P     P     Repl. by Fowler, Pred. Ag.
LUGAR         0.403                 P <7     P     P      
WILSON        0.408                 P <7     P     P      
DOLE          0.425                 P <7     P     P      
DOMENICI P    0.434                 P <7     P     P      
QUAYLE, J D   0.481     P     P     P <7     P     P      
SIMPSON, A    0.482     P     P     P <7     P     P      
HATCH         0.556     P     P     P <7     P     P      
THURMOND      0.585                 P <7     P     P      
DENTON, J     0.597                (A =7)    P     P     Repl. by Shelby, Pred. Ag.
LAXALT        0.670                (A >7)    P     P     Repl. by Reid, Pred. Ag.
GARN          0.673                 P <7     P     P      
GOLDWATER     0.691                 P <7     P     P     Repl. by McCain, Pred. For.
ARMSTRONG     0.706                 P <7     P     P      
HUMPHREY, G   0.806     P     P     P <7     P     P      
GRAMM, P      0.807                 P <7     P     P      
HECHT         0.808                 P <7     P     P      
MCCLURE J     0.965                 P <7     P     P      
EAST, J       0.974                (A =2)    P     P     Repl. by Sanford, Pred. Ag.
HELMS J       1.000                 P <7     P     P      
WALLOP        1.000                 P <7     P     P      
SYMMS, S      1.000                 P <7     P     P      

Legend to Preceding Table:
Vote of replacements shown in parentheses 
>7 = Bork vote announced after 10/7 
=y =  Bork vote announced on 10/7
<7 = Bork vote announced before 10/7
Classification errors shown in boldface italics.
Sources: Bork; New York Times, Washington Post; Manion,
Rehnquist; Congressional Quarterly

Table 2.     1969-70 Confirmation Votes

           1985 Position  Haynesworth  Carswell
INOUYE          -0.806         A         A
KENNEDY,E.      -0.764         A         A    
CRANSTON        -0.727         A         A
PELL            -0.703         A         A    
EAGLETON        -0.700         A         A
BYRD            -0.527         P         P
MATHIAS         -0.346         A         A
PROXMIRE        -0.304         A         A 
------------------------------------------ Optimal Cutting Line
STENNIS         -0.284         P         P
HOLLINGS        -0.263         P         P
HATFIELD        -0.178         A         A 
LONG, R.        -0.161         P         P
PACKWOOD         0.028         A         A 
STEVENS          0.139         P         P
DOLE             0.425         P         P
THURMOND         0.585         P         P
GOLDWATER        0.691         P         P

Legend: A-Anti-confirmation; P-Pro-confirmation; 
Classification errors shown in boldface italics
Source: Congressional Quarterly