COR 315 - Addendum III                                     September 18, 1979

TO:  R. Williams
FROM:  R. Hauser

Purpose:  (A)  Reconstruction of best measures of 1957 Parental Income;
          (B)  the development of spline variables for IQ score and high school
               rank, (That is, we tried to find breaking points on the scales 
               of IQ and high school rank above and below which the effects of 
               those two variables would be linear;  the slopes need not be 
               the same above and below these breaking points.  SEE MEMO#49.);
          (C)    the new Social Security tape policy.

 

(A)    Background:  The current best measures of 1957 parental income (BMPIN1 
       and BMPIN2) were created using a combination of the variables PI5760 
       (1957 Wisconsin tax data) and YFML57 (1975 self-report on 1957 income). 
       When PI5760 had missing data, YFML57 was used.  BMPIN1 was then 
       truncated at $99,800, while BMPIN2 was truncated at $50,000.  It now 
       appears that the use of the self-report measure tends to give misleading
       results.  We therefore want to do the following:  

       (1)  Regress PI5760 on YFML57.  Also take logs of those variables and 
       regress them on each other.   Also run a breakdown of PI5760 by a 
       collapsed version of YFML57.  These runs will enable us to determine 
       whether a linear or nonlinear transform of YFML57is most appropriate.  

       (2)  After we have looked at the above results, create new best measures
       of 1957 parental income as follows:  
              
              (a)  When PI5760 is present, use its value.
              
              (b)  When PI5760 is missing but YFML57 is present, use a 
              regression estimate of PI5760 based on YFML57 from (1) as the 
              best measure.

              (c)  Truncate BMPIN1 at $99,800 and BMPIN2 at $15,000.

       (3)  See MEMO#49 for details on construction of BMPIN1 and BMPIN2.

       (4)  For SWL20 and SWL21 the output tapes should be the same as those 
       described in COR315.  For the Social Security tapes, the changes 
       described in COR315-Addenda I-III, should be made simultaneously.  The 
       relevant tapes are:  

            SWL20-SS Merge (6300 column version):
                  Current cobol master                     MACC#1943
                  Current STATJOB/FORTRAN master           MACC#1393
                  New cobol master                         MACC#1412
                  New STATJOB/FORTRAN master               MACC#1393

              SWL20-SS Merge (condensed, 3100 column version):
                  Current cobol master                     MACC#2002
                  Current STATJOB/FORTRAN master           MACC#3124
                  New cobol master                         MACC#2026
                  New STATJOB/FORTRAN master               MACC#3124

(B)    We also want to create spline variables for IQSCOR and HSRNRM (these 
       will not be put on the master tapes).  For each variable, create new 
       pairs of variables in which 90 and 100 are the cutting points.  For 
       example, SPLINE1 will equal 0 IQSCOR when IQSCOR < 90, and will equal 
       90 when IQSCOR >= 90.  SPLINE@ will equal 90 if IQSCOR <= 90, and will 
       equal IQSCOR if IQSCOR > 90.  SPLINE3 and SPLINE4 will use 100 as the 
       cutting point, and SPLINE5 through SPLINE8 will be constructed the same 
       way using HSRNRM instead of IQSCOR.

       After constructing the new variables, regress OCSX1, OCSXCR, and YRER74 
       on each pair of splines.  (For example, regress OCSX1 on SPLINE1 and 
       SPLINE2;  YRER74 on SPLINE7 and SPLINE8, etc.)

       Based on the above results, we will decide whether to use 90 or 100 as 
       our cutting point for future analysis.

(C)    IMPORTANT NOTE:  Up to this point, we have followed a policy of updating
both the large and condensed versions of the SS tapes.  This has proven to be 
extremely expensive and time consuming.  It is also unnecessary, since any 
information not on the condensed tape could always be retrieved from the 
original.  THEREFORE, FROM THIS POINT ON, WE WILL UPDATE ONLY THE CONDENSED 
VERSION OF THE SS TAPE.  MACC#1412 IS THE FINAL VERSION OF THE LARGE SS TAPE.