Description of NOMINATE Data

(File names and formats follow these Descriptions)

13 July 2004

There are 4 kinds of NOMINATE Coordinates:

  1. The original D-NOMINATE Coordinates

    These 2-dimensional dynamic coordinates were originally estimated on the CYBER 205 supercomputer at Purdue University during 1988-89 and they are extensively analyzed in our book Congress: A Political-Economic History of Roll Call Voting, Oxford University Press, 1997). Each legislator's point is dynamic and is allowed to move as a linear function of time as measured by the Congress number (higher polynomials in time did not appreciably increase the fit). A legislator's point is constant within a Congress but "jumps" along a linear path between Congresses. A member had to serve at least 3 Congresses for the linear model to be estimated.

    Because of the "overlapping generations" nature of the estimation, scores in one Congress are directly comparable with scores in another Congress. However, as we caution in our book, cross-Congress comparisons should be conducted only between Congresses occurring during one of the stable 2-party periods of American History. Also, the D-NOMINATE scores cannot be compared across chambers.

    Please see Chapters 1 -3 of our book for a description of the spatial model underlying D-NOMINATE and Appendix A of our book for the technical details and Monte-Carlo tests of the D-NOMINATE procedure.

  2. W-NOMINATE Coordinates

    W-NOMINATE is a static (i.e., meant to be applied to only one Congress) version of D-NOMINATE, with a number of improvements being designed to increase the efficiency of the algorithm so that it can be run on a desktop personal computer. To this end W-NOMINATE differs from D-NOMINATE in two ways: It uses a slightly different deterministic utility function; and, because it is a static algorithm, it constrains the legislators and roll call midpoints to lie within an s-dimensional hypersphere of radius one (in contrast to the rather flexible constraint structure necessitated by the dynamic model).

    The original D-NOMINATE deterministic utility function was:

    uijy = b e[- (w2dijy2)/2]

    which is simply a normal distribution multiplied by a constant. Note that there is no index on w -- namely the weight is the same for all dimensions. Experimentally, we found that estimating both b and w was unnecesary so we set w2/2 = 1/8 (see Appendix A of our book for all the details).

    The W-NOMINATE deterministic utility function is:

    uijy = b e[- åk=1,s (wk2dijy2)/2]

    The weight on the first dimension, w1, is set equal to one and the other dimension weights are estimated (see Appendix A of our book for all the details).

    The W-NOMINATE coordinates are highly correlated with the D-NOMINATE coordinates for most Congresses (Pearson r's typically greater than .95 for both the 1st and 2nd dimensions). However, unlike the D-NOMINATE scores, W-NOMINATE scores are not directly comparable between Congresses.

    See Appendix A of our book for the technical details of the W-NOMINATE procedure. Jeff Lewis and I are computing Legislator Coordinate files with parametric bootstrapped standard errors for all Houses and Senates in American History. They can be found on the Parametric Bootstrap Page. For an explanation of how these are computed see:

  3. Common Space Coordinates

    The common space scores place the members of the House and Senate in the same space. This allows members to be compared across Chambers and across Congresses. The scores are derived from the W-NOMINATE scores utilizing a scaling technique explained in "Estimating a Basic Space From a Set of Issue Scales," American Journal of Political Science, 42 (July 1998), pp. 954-993.

    The W-NOMINATE coordinates of all members serving in both the House and Senate from 1937 - 2000 (Congresses 75 - 106) were used to estimate the common space coordinates. Each legislator is assigned a single coordinate throughout his or her career so that members who served in both chambers will have the same coordinate in both. The 1st dimension coordinates correlate with the corresponding DW-NOMINATE coordinates at .94 or higher for all Houses and Senates.

    The Presidents from Eisenhower through Clinton were scaled using the CQ Presidential Support Roll Calls in W-NOMINATE. See "Veto Power and Legislation: An Empirical Analysis of Executive and Legislative Bargaining from 1961-1986," Journal of Law, Economics, and Organization, 11, 1995, 282-312 (with Nolan M. McCarty), for an analysis of these W-NOMINATE scores. A President's W-NOMINATE scores were treated like any other member of the House and Senate for the purposes of fitting them into the common space.

    Note that the common space coordinates are adjusted so that the two dimensions are equally salient and lie within a unit circle. This assumption plus the assumption that a legislator serving in both Chambers of Congress has the same coordinate throughout his or her career means that these coordinates should be used with caution.

  4. DW-NOMINATE Coordinates

    DW-NOMINATE is a dynamic version of W-NOMINATE and is very similar to the original D-NOMINATE procedure. The only differences are that DW-NOMINATE is based on normally distributed errors rather than on logit errors and that each dimension has a distinct (salience) weight (the weight of the first dimension is always 1.0). DW-NOMINATE scores in one Congress are directly comparable with scores in another Congress. However, cross-Congress comparisons should be conducted only between Congresses occurring during one of the stable 2-party periods of American History. Also, the DW-NOMINATE scores cannot be compared across chambers. In addtion, because these coordinates were estimated using a weighted utility model, if distances are computed between legislators, this weighting must be taken into account! This was not true of the original D-NOMINATE coordinates.

    When Howard Rosenthal and I began our research in 1982, computer memory and speeds were a significant problem, and we took the logit approach because it was computationally more tractable. Because computer memory and speed are now minor impediments, we developed DW-NOMINATE because the normal distribution allows us to develop much more sophisticated models of correlated error (that work is now underway).

    We applied DW-NOMINATE to all roll call votes taken since the end of World War II as part of an on going study of post-War American politics. Our preliminary results are reported in Income Redistribution and the Realignment of American Politics (joint with Nolan McCarty, 1997, AEI Press).

    For all practical purposes, the coordinates estimated by DW-NOMINATE are identical to the original D-NOMINATE coordinates for the 1st to the 99th Congresses. For the 2-dimensional, dynamic model, the Pearson correlations between the two sets of coordinates for the House were .977 for the first dimension and .923 for the second dimension (n = 31,306). The corresponding correlations for the Senate were .974 and .901 respectively (n = 7619).

    Note that the average DW-NOMINATE coordinate for every legislator is constrained to lie within the unit hypersphere. However, some members may have large linear terms so that for some Congresses their coordinates can be greater than +1/-1. Also note that the weight on the 2nd dimension is about .3 for both the House and the Senate [see Income Redistribution and the Realignment of American Politics (joint with Nolan McCarty, 1997, AEI Press) for further details]. To reiterate the warning above, because these coordinates were estimated using a weighted utility model, if distances are computed between legislators, this weighting must be taken into account in any analyses using the distances.

File Names and Formats for NOMINATE Data

  1. KREH01.ORD to KREH99.ORD and KRES01.ORD to KRES99.ORD are D-NOMINATE coordinates. The format of the D-NOMINATE files is:
    
     1.  Congress Number
     2.  ICPSR ID Number:  5 digit code assigned by the ICPSR as 
                           corrected by Howard Rosenthal and myself.
     3.  ICPSR ID Number:  original 5 digit ICPSR code (this field is
                           usually blank).
     4.  State Code:  2 digit ICPSR State Code.
     5.  Congressional District Number (0 if Senate)
     6.  State Name
     7.  Party Code:  100 = Dem., 200 = Repub. (See PARTY3.DAT)
     8.  ICPSR Occupancy Code:  See any ICPSR Roll Call Voting Codebook
     9.  ICPSR Office Code:  See any ICPSR Roll Call Voting Codebook
    10.  Name
    11.  1st Dimension Coordinate
    12.  2nd Dimension Coordinate
    
    Note that if the 1st and 2nd dimension coordinates are exactly .000, then the Senator/Representative did not have enough roll call votes to be estimated in that Congress.

    To download these files go to the D-NOMINATE Scores Page.

  2. W-NOMINATE coordinates.

    
     1.  Congress Number
     2.  ICPSR ID Number:  5 digit code assigned by the ICPSR as 
                           corrected by Howard Rosenthal and myself.
     3.  State Code:  2 digit ICPSR State Code.
     4.  Congressional District Number (0 if Senate)
     5.  State Name
     6.  Party Code:  100 = Dem., 200 = Repub. (See PARTY3.DAT)
     7.  Name
     8-11. The cross-classications for the legislator.  The two middle numbers are the "errors".  
           For example, in the 106th House Callahan's cross classification numbers are:
    
            CALLAHAN     494  50  20 278  
    
            or in table form
    
                             ACTUAL
                           Yea     Nay
                     Yea   494      50
         PREDICTED
                     Nay    20     278
    
    12.  Geometric Mean Probability
    13.  1st Dimension Coordinate
    14.  2nd Dimension Coordinate
    15.  1st Dimension Standard Error
    16.  2nd Dimension Standard Error
    W-NOMINATE scores With Bootstrapped Standard Errors

    W-NOMINATE scores by year for 1939 - 1998

    W-NOMINATE Program

  3. BL75107.DAT are the Common Space coordinates -- W-NOMINATE coordinates for the 75th through the 107th Houses and Senates adjusted so that the the House and Senate members and the Presidents are in the same space. There is one record for each legislator in each Congress for this time period (n=17,871, 1937-2002). For an explanation of how this was done, see "Estimating a Basic Space From a Set of Issue Scales," American Journal of Political Science, 42 (July 1998), pp. 954-993.

    The format for BL75107.DAT is:
    
              1.  Congress Number 1 - 107
              2.  ICPSR ID Number:  5 digit code assigned by the ICPSR as 
                                    corrected by Howard Rosenthal and myself.
              3.  State Code:  2 digit ICPSR State Code.
              4.  Congress District Number (Senate and President = 0)
              5.  Name of State
              6.  Party Code:  100 = Dem., 200 = Repub. (See PARTY3.DAT)
              7.  Name
              8.  1st Dimension Coordinate
              9.  2nd Dimension Coordinate
             10.  Indicator For Chamber
                                0 = House
                                1 = Senate
                                2 = President (83 - 107)
    
    To download these files go to the Common Space Page.

  4. H761939.DAT to H1051998.DAT and S861959.DAT to S1051998.DAT are yearly W-NOMINATE coordinates.

    The format of these files is the same as the D-NOMINATE and W-NOMINATE files.

    W-NOMINATE scores by year for 1939 - 1998

  5. HL01107A1.DAT and SL01107A1.DAT are DW-NOMINATE legislator coordinates for the 1st to the 107th Congresses. HC01107A1.DAT and SC01107A1.DAT are the corresponding roll call coordinates.
    
    The format of the legislator files is:
    
     1.  Congress Number
     2.  ICPSR ID Number:  5 digit code assigned by the ICPSR as 
                           corrected by Howard Rosenthal and myself.
     3.  State Code:  2 digit ICPSR State Code.
     4.  Congressional District Number (0 if Senate)
     5.  State Name
     6.  Party Code:  100 = Dem., 200 = Repub. (See PARTY3.DAT)
     7.  Name
     8.  1st Dimension Coordinate
     9.  2nd Dimension Coordinate
    10.  1st Dimension Conditional Standard Error
    11.  2nd Dimension Conditional Standard Error
    12.  Log-Likelihood
    13.  Number of Votes
    14.  Number of Classification Errors
    15.  Geometric Mean Probability
    
    The format of the roll call files is:
    
     1.  Congress Number
     2   Roll Call Number
     3.  Spread on 1st Dimension    -- if the roll call was not scaled, there
     4.  Midpoint on 1st Dimension  -- are 0.000's in all four fields 
     5.  Spread on 2nd Dimension    --
     6   Midpoint on 2nd Dimension  --
    
    To download these files go to the DW-NOMINATE Page.

Note that the sign of the 2nd dimension D-NOMINATE coordinates for the Senate is the reverse of that for the House. We multiplied the Senate 2nd dimension coordinates by -1.0 when we used them in our published work so that they would be comparable to the House coordinates.

The state, congressional district, and party codes correspond to Ken Martis's The Historical Atlas of Political Parties in the United States Congress. Howard and I consider Martis's Atlas to be the best source of information about party affiliation of members of Congress so that our data base can be considered to be very "clean". Congresses 1 - 100 have been extensively checked by ourselves and other scholars. Congresses 101-107 have also been checked but cannot be considered to be completely "clean". If you find errors, please send E-Mail to Keith Poole kpoole@ucsd.edu so that corrections can be made.

The file PARTY3.DAT contains the dictionary for the party codes. Please cite Ken Martis's atlas as the original source if you use these codes in any way.

W-NOMINATE is written in FORTRAN and can be downloaded from this website.
[W-NOMINATE Computer Program]

Contact Keith Poole at KPoole@ucsd.edu if you need help with this data or the program.



Site Links

NOMINATE Data, Roll Call Data, and Software
Course Web Pages: UC San Diego (2004 - )
University of San Diego Law School (2005)
Course Web Pages: University of Houston (2000 - 2005)
Course Web Pages: Carnegie-Mellon University (1997 - 2000)
Spatial Models of Parliamentary Voting
Recent Working Papers
Analyses of Recent Politics
About This Website
K7MOA Log Books: 1960 - 2006
Bio of Keith T. Poole
Related Links