COMPUTER OPERATIONS REQUEST #331a Date: November, 1981 Subject: Coding of EXPER and TENURE for respondents not in the labor force at the time of interview. To construct new codes for EXPER and TENURE for all 1975 respondents: Labor force experience and tenure in current job has been coded for all respondents who were in the labor force at the time of interview in 1975 (see Memo #043). EXPER is the proportion of time from June 1957 to the month of interview that the respondent was not known to be out of the civilian labor force, and TENURE is the time in months from the start of the current job to the month of interview. The purpose of this COR was to assign EXPER and TENURE codes to respondents who were not in the labor force at the time of interview. In the process of doing that, we checked the coding procedure used previously (described in Memo #043) and found and corrected some minor errors. WE ALSO FOUND ONE CODING DECISION WHICH LED TO INFLATED ESTIMATES OF THE EXPER VARIABLE FOR WOMEN WITH MORE THAN 4 CHILDREN. This was corrected in the new program. The EXPER. program assigns activity codes designating what the respondent did each month since she/he left high school. Each month is also assigned an intensity code which is an indicator of the proportion of time during that month that the respondent spent outside the labor force. Any respondent with missing values on any one variable used in the assignment of these codes are automatically given missing values on EXPER and TENURE. To reduce the number of cases with missing values, the original coding procedure checked missing value cases to see whether a reasonable assignment could be made for the variable in question. To assign codes for respondents not in the labor force the following was done: 1. To get a sense of the guidelines used in assigning values to some cases with missing data we went through the handwritten codes in the output folder for COR #301. Each case was checked by going back to the questionnaire and coding it independently. A comparison of these codes with the earlier coding decisions showed disagreement in 26 out of 49 cases. Most differences were small: in 13 cases we assigned a code where the previous coding did not; in the remaining 13 cases most of the discrepancies arose in the assignment of the duration of a time period - in only 5 of these cases were the differences greater than 5 weeks. Although differences were small and inconsequential we did change all missing value codes to a code that we agreed with. This was simply to assure that missing data were treated similarly for respondents in and outside the labor force. In addition to the hand coding of missing data, the original program summarily assigned codes to some variables with missing codes. For women this pertained to variables for hours of work in marriage and birth intervals. Missing values were replaced by 39 (hours per week) for the interval between marriage and first birth, 32 for the interval between the first and second birth, and 30 for subsequent intervals. For males the assignments made were identical to the original program. 2. The second step involved checking a listing of cases with missing values on the relevant variables for the sample of respondents who were not in the labor force at the time of interview. 112 cases out of 2079 were listed as having missing data on relevant variables, i.e. variables used by the EXPER program to assign codes for EXPER and TENURE. 17 of these 112 cases had missing values on a variable that turned out not to be used in the program - they were left alone. 15 cases contained numerous variables with missing values - they were left as they were. The remaining 80 cases had missing values on 99 variables all together. Of these, 43 were assigned new codes after the questionnaires had been checked. In all of these cases there was sufficient information in the questionnaire to change the code from missing data. Many codes were determined by estimating the midpoint between two events. To do this there had to be specific information on the two end points, and these end points could not be more than 2 years apart. All other coding decisions were based on specific information. 3. Coding experience and tenure for WOMEN NOT IN THE LABOR FORCE IN 1975 involved a number of changes in the program used in the original creation of EXPER and TENURE. All coding decisions regarding military service and education are unchanged. Memo #043 gives a detailed description of these decisions. For women not in the labor force in 1975, the computation of the upper time bound for the last marriage-birth interval was changed to take into consideration the month in which the woman reported to have left the labor force. In addition, the order in which each assignment procedure took place was changed. In the new version of the program, data for 1974 are processed before data based on the marriage-birth interval variables. This makes no difference in final outcome, but for some reason it made a difference internally in the program. The tenure code is only assigned to women who have held a job within 5 years of the interview - that is respondents for whom the variable CMJXCR is present._ The new program does all assignment and calculations the same way as the original program, with one exception.: FOR WOMEN WITH MORE THAN 4 CHILDREN, THE OLD PROGRAM IGNORED THE TIME INTERVAL BETWEEN THE THIRD AND THE NEXT TO LAST CHILD. This meant that this interval, which in some cases is quite long, most often was assigned the activity code 2 (In the labor force) and the intensity code 0, that is full time employment. These codes are not based on data for that interval, since they do not exist; they simply result frca the fact that all months at some earlier point in the Program are assigned the code 2. This code remains until it is overridden by other information. For women with more than 4 children no such information exists for that particular interval.This means that the EXPER variable provides an inflated estimate of labor force experience for women with more than 4 children. THIS DECISION RULE WAS REPLACED BY THE FOLLOWING: For all women with 5 or 6 children the activity code 2 months before and 2 raonths after the interval between the third and the next-to-last child are compared. If these two codes are the same, that code is assigned to the interval.- In thft overwhelming majority of cases the activity code was 6 and the intensity 1. out of 168 women with 5 children only 4 cases could not be assigned codes in this fashion. Out of 142 women with 6 children - 12 were not assigned codes. The assignment of EXPER for these 16 cases remains unchanged. For women with 7 or more children the assignment of codes for the interval between the third and the next to last child was done by hand. The experience and tenure codes were recalculated using the new intensity codes, and individual assignments were made for each case. Again, we found few instances where the decision rule was not straightforward to apply. Note that the WEXPER program does not correctly code months between the third and the next-to-last birth for women with 7 or more children. This change was made for all female respondents whether they were in the labor force at the time of interview or not. Comparison between the codes produced by the old program and the new WEXPER program shows very small differences for respondents who were in the labor force at the time of interview. The differences are easily explained by the correction of some coding errors. The exception is of course the code for women with more than 5 children. The old EXPER had a mean of .575 for women in the labor force with 3 or fewer children, the new EXPER mean is .580. For women with 5 children the mean drops from .534 to .433, for women with 6 or more children from .586 to .376. The new code for EXPER is a conservative estimate of the woman's labor force experience, while the old code was an extremely high estimate. Unfortunately we do not have the data to evaluate which of the two estimates are closest to the truth. OTHER USES OF THE EXPER PROGRAM: In its current version the WEXPER program assigns codes for experience and tenure in the last job and adds these codes onto the output pile. No record is made of the activity and intensity codes assigned each month since the respondent left high school. This information is potentially very valuable, and it should require little effort to obtain an output reccord which stores activity and intensity codes for each month. If this is done, it should be noted that the program in its current version does not deal adequately with the interval between third and next-to-last child for women with 7 or more children. A slight revision of the WEXPER program has for example produced records consisting of "spells" of time outside the labor force; beginning and ending month for such "spells" are recorded and in subsequent EXTRACT runs a file is created with the "spell" as the unit of analysis.