MEMO #26                                

The section on Military Service has been cleaned and the
variables constructed with the following observations:

A) Question 16a has a non-mutually exclusive set of responses. 
   The individuals in the [RESERVES/NATIONAL GUARD] category were
   really enlisted people.  They enlisted in the reserves or
   National Guard.

B) Questions 16c and 16i refer to dates of ACTIVE duty; however,
   respondents did not answer this question consistently.  It's
   validity is questionable.

C) Question 16j refers to rank only and we got branch of service
   as a secondary piece of information.  We should have obtained  
   it properly with the Reserves/National Guard options indicated 
   in Q 16a.  Also, we failed to get this piece of information 
   from people on active duty at the time of interview.

D) Questions 16k and 16m should not have used the word
   SEPARATIONS and should have been couched in the mental      
   framework of starting and ending dates of the broken periods   
   of service rather than the starting and ending dates of the    
   separations.

E) IMPLICATIONS OF THE ABOVE PROBLEMS

   1) Because of B), the ranks respondents gave were often the 
      highest ranks that they achieved in the Reserves, not the 
      highest ranks obtained while they were on active duty.  The 
      rank they gave was generally consistent with the dates that 
      they answered for 16c and 16i.  If these dates included 
      reserve time, then they gave their "reserve" rank.

   2) Also, largely but perhaps not entirely because of B), there 
      are conflicts between the periods when individuals were on 
      "active" duty and the dates which they completed high 
      school, first attended college, or started their first job. 

      While it is not impossible to start a job or attend college
      while on active duty, it appears that the dates of "active" 
      duty were a frequent source of error.

   3) Because of A), B), and C), we really don't know whether the 
      individual was a "reserve" in the short program (6 months), 
      a "reserve" after an active duty of longer duration, or in 
      fact whether a person was a "reserve" at all. That is--it 
      would be possible for a person to have been in the reserves 
      and not show up in either 16a or 16j.

F) SPECIAL CHANGES MADE TO THE DATA

   1) The only changes made to the data were those where there 
      were obvious coding errors found during the wild punch, 
      extreme value, consistency and contingency checks with the 
      exception of the changes stated in 2), 3), and 5) below.

   2) When starting date of military service was prior to 
      graduation from high school and the respondent indicated 
      that he graduated in 1957 with the rest of his cohort or 
      had completed 12 years of high school prior to entry or 
      both, then the year of start of service was changed to 1957 
      and the month was left as it was.

   3) People who were currently on active duty in 1975 had their 
      rank and branch of service codes lifted from their current 
      occupation section of the questionnaire.  Officers of 
      unspecified rank were coded 35 and enlisted men of
      unspecified rank were coded 30.  Army privates not further 
      specified were coded 01 if they were in for 6 months or 
      less or were in the reserves. They were coded 02 otherwise. 
      There were 8 officers, 29 enlisted, and about 50 privates 
      not further specified respectively.

   4) Conflicts between the time of active duty and the start of 
      college did not cause any dates to be altered.  It was 
      determined that there was no uniform way to determine which 
      cases should be changed.  There were about 60 cases in 
      which this conflict occurred.  Individuals working with the 
      data should be made aware that the dates of active duty 
      which we obtained from the respondents are suspect, 
      likewise the duration of active duty variable which we 
      obtain by subtracting the starting date from the ending 
      date is flawed.

   5) Conflicts between the time of active duty and start of 
      first job were generally resolved by changing the ending 
      date of active service to conform to the date the 
      respondent started their first job.  Exceptions to this 
      general treatment occurred when it was obvious that the 
      date of first job was wrong, or when no appropriate date 
      change could be made. In this latter case, military service 
      dates were changed to (DK=997).  All together there were 62 
      cases which had dates of military service changed.  These 
      changes were made only on the analysis tape--not the raw 
      data tape.

   6) For individuals currently on active duty, the duration of 
      military service variable was created by subtracting the 
      century month in which they began their active duty from 
      the century month for June 1975 (906).

G) CLEANING AND RECODING MILITARY SERVICE VARIABLES IN THE 1977 
   SIBLING STUDY
  
        The military section in 1977 was greatly shortened     
   from that used in 1975.  The questions involving training      
   programs while respondent was in the armed forces, military    
   rank, branch of service and separations from service were      
   omitted in 1977. For individuals currently on active duty,     
   the duration of military service variable (XMILDUR) was      
   created by subtracting the century month in which they began   
   their active duty (XCMMLBG) from the century month for      
   February 1977 (926). Date of interview could have been used    
   instead of the February 1977 (926), but in 1975 a fixed      
   value was used instead of the interview date, and we wanted    
   to be consistent. The only changes to the data were those      
   where there were obvious coding errors found during wild      
   code, consistency and contingency checks.