Optimal Classification Page

Optimal Classification Scores

Updated 18 February 2010

The files below contain Optimal Classification two dimensional scores for the 1^st to the 111^th Congresses (1789 - 2009). These coordinates were computed using OC. The algorithm is explained in "Non-Parametric Unfolding of Binary Choice Data." Political Analysis, 8:211-237, 2000, and Non-Parametric Unfolding of Binary Choice Data (1998 APSA Paper) (PDF file). The coordinates for the 1947 - 2002 (80^th through 107^th Congresses) are analyzed in Changing Minds? Not in Congress (PDF file).

The files below contain OC coordinates for all members of the House, all members of the Senate, many Presidents [for Presidents prior to Eisenhower these are based on roll calls corresponding to Presidential requests. These roll calls were compiled by an NSF project headed by Elaine Swift ( Study No. 3371, Database of Congressional Historical Statistics, 1789-1989). Many of these scores are based upon a small number of roll calls so use them with caution!], and all roll calls with at least 2.5% in the minority (44,070 roll calls in the House and 44,095 roll calls in the Senate). The scaling was done on both chambers simultaneously using the 635 members who served in both the House and Senate as "glue" to define a common metric. See the discussion in Changing Minds? Not in Congress for the advantages and disadvantages of this assumption.

The overall fit of the scaling was 88.00 percent in two dimensions -- 13,814,903 of 15,698,103 total choices were correctly classified for an APRE of .6416. There were 10,543 unique members of the House, 1,858 unique members of the Senate, and 635 legislators served in both chambers. Hence, there were 10,543 + 1,858 - 635 = 11,766 unique legislators who served in Congress in American history. Known major political party switchers (e.g., Strom Thurmond) are in the data twice. (For lists of party switchers, see Congressional Party Defection in American History).

The unique members file looks like this:


         1 3698 NORTH D  200 AANDAHL,        110      10  0.909  0.363 -0.511
         2 40 4 VIRGINI  100 ABBITT W       2600     340  0.869  0.096  0.992
                            etc etc etc
                            etc etc etc
     99909 99 0 USA      100 CLINTON         963      95  0.901 -0.386 -0.154
     99910 99 0 USA      200 BUSH            119       5  0.958  0.396 -0.076

The first column is the ICPSR ID number (as corrected by myself and Howard Rosenthal -- see Corrected ICPSR Member ID Numbers Congress 1 - 107 for a discussion and listing of all members of Congress). The second column is the ICPSR State Code (99 if President), the third is the congressional district (0 if Senate or President), name of state, party code (see the party codes page for a complete listing of the party codes), name of member, total choices, errors, proportion correct ([total - errors]/total), and the last two columns are the two dimensional coordinates.

All Unique Members of the House and Senate Estimates 1^st to 111^th (Text File, 11,766 lines)

All Unique Members of the House and Senate Estimates 1^st to 111^th (Excel File, 11,766 lines)

All Unique Members of the House and Senate Estimates 1^st to 111^th (Stata 8 File, 11,766 lines)

All Unique Members of the House and Senate Estimates 1^st to 111^th (Stata 7 File, 11,766 lines)

The separate legislator coordinate files by Congress look like this:


   1  9062  198 CONNECT 5000 STURGES      0.567  0.350
   1  9706  198 CONNECT 5000 WADSWORTH    0.697  0.013
                  etc etc etc
                  etc etc etc
 107 14657 25 9 WISCONS  200 SENSENBRENN  0.594 -0.804
 107 29584 68 1 WYOMING  200 CUBIN        0.571  0.280

The first column is the Congress number, the second is the ICPSR ID number, the third is the ICPSR State Code (99 if President), the fourth is the congressional district number (0 in the Senate file), the fifth is the state name, the sixth is the party code, name of member, and the two dimensional coordinates. Note that these coordinates are the same as those in the unique members file above.

Legislator Estimates 1^st to 111^th Houses (Text File, 36,627 lines)

Legislator Estimates 1^st to 111^th Houses (Excel File, 36,627 lines)

Legislator Estimates 1^st to 111^th Houses (Stata 8 File, 36,627 lines)

Legislator Estimates 1^st to 111^th Houses (Stata 7 File, 36,627 lines)

Legislator Estimates 1^st to 111^th Senates (Text File, 8,851 lines)

Legislator Estimates 1^st to 111^th Senates (Excel File, 8,851 lines)

Legislator Estimates 1^st to 111^th Senates (Stata 8 File, 8,851 lines)

Legislator Estimates 1^st to 111^th Senates (Stata 7 File, 8,851 lines)

The separate roll call coordinate files by Congress look like this:


   1    1      1  41   8  49   1 6 1 -0.710 -0.966  0.260
   1    2      2  36   1  37   1 1 6  0.675 -0.560  0.829
                 etc etc etc
                 etc etc etc
 107  988  44774 366  19 385  17 6 1 -0.649  1.000 -0.027
 107  990  44776 244 116 360  27 6 1 -0.306  0.998 -0.061

The first column is the Congress number, the second column is the roll call number (note that unscaled roll calls are omitted!), the third is an overall counter for the chamber, the fourth is the number of Yeas, the fifth the number of Nays, the sixth the total number of votes cast, the seventh the number of classification errors, the eighth and ninth numbers indicate where the Yeas and Nays fell relative to the projected midpoint on the normal vector (if the first number is a "6" that means that Nay was predicted below the cutpoint, if "1" Yea), the tenth number is the projected midpoint on the normal vector, and the last two numbers are the normal vector for the roll call.

Roll Call Estimates 1^st to 111^th Houses (Text File, 44,070 lines)

Roll Call Estimates 1^st to 111^th Houses (Excel File, 44,070 lines)

Roll Call Estimates 1^st to 111^th Houses (Stata 8 File, 44,070 lines)

Roll Call Estimates 1^st to 111^th Houses (Stata 7 File, 44,070 lines)

Roll Call Estimates 1^st to 111^th Senates (Text File, 44,095 lines)

Roll Call Estimates 1^st to 111^th Senates (Excel File, 44,095 lines)

Roll Call Estimates 1^st to 111^th Senates (Stata 8 File, 44,095 lines)

Roll Call Estimates 1^st to 111^th Senates (Stata 7 File, 44,095 lines)