Newsletter

November 1995 - Number 2

User Questions

This section of the FLS Newsletter incorporates questions of general interest raised by data users. The section deals largely with questions that clarify existing variables and relationships between variables, respondents, and subfiles in the publicly released databases. Topics covered relate to questions not addressed in public-release documentation.

How do I identify the 1,096 Children sample respondents in MFLS-2? If I take those people in the MF21ROST file for the Panel and Children sample who have values of 11-40 or 61-90 for the variable MFLS1, I get many more than 1,096.
The MF21ROST.PC file for the Panel and Children sample contains all Panel and Children sample household members. Any children of the Panel woman who were in the 1976 MFLS-1 household and now appear in an MFLS-2 household will have a value in the variable called MFLS1 (the individual identifier in the MFLS-1 data). The 1,096 Children sample respondents are a subset of the adult children of the Panel women who appear across the MFLS-2 households. To isolate the specific 1,096 Children sample respondents, you need to take the MFLS1 identifier variable from MF21ROST.PC and merge it onto the MF22SUM.PC and MF23SUM.PC files using CASE, SPLIT and PERSON. Those MF22 and MF23 respondents (the people in the MF22SUM.PC and MF23SUM.PC files) with a value of MFLS1 that is between 11 and 40 or between 61 and 90 are the Children sample respondents. This method yields 1,095 people. There is one MF23 respondent in the Children sample who was an adopted child and has an MFLS1 code of 51. This gives a total of 1,096.
There appear to be some discrepancies in the MFLS-2 data between the variables HHTYPE and CASE for the household, and between HHTYPE and household composition. New sample households with CASE < 7000 (New only or New and Senior) have HHTYPE=8 (Senior only). How can that be? Panel only households (HHTYPE=4) have adult children of the Panel woman in the same MFLS-2 household as the Panel woman. Why aren't such households HHTYPE=6 (Panel and Children)?
The variable HHTYPE is based on what type of respondents are in the household and not on the original sample into which the household fell. For example, if a household selected as part of the New and Senior sample (i.e., households where both a woman aged 18 to 49 (New respondent) and a person aged 50 and over (Senior respondent) could be selected for interview) had only the Senior respondent actually complete an interview, the HHTYPE would be 8 (only a Senior respondent) and not HHTYPE=9 (New and Senior respondents). Likewise, in the Panel and Children data, if a household has both a Panel woman and adult children of that Panel woman, but the selected adult child does not complete an interview, the household will be classified as HHTYPE=4 (Panel respondent only).
Some individuals have seemingly inconsistent responses between the MF22/MF23 work histories and the MF25INC data on income earning activities in the past year. Also some individuals in MF25INC appear to have implausible totals for number of weeks worked in the past year across all activities mentioned. What does this mean? What should I do?
The responses in MF25INC are given by one person, the head of the household in most cases, and this person provides information about the income earning activities of everyone in the household. The MF25INC respondent is listed in the MF25SUM file. In some cases the MF25INC respondent may not really know how many weeks the individual has worked in the past year or how many hours per week were worked. The MF25INC respondent may believe that the individual basically worked full time (or part time) and gives responses consistent with that belief. Discrepancies may also exist when the respondent is the same person in both the work history and MF25INC. Some such discrepancies may result from the fact that the work history covers a type of work whereas MF25INC covers actual activities. If someone had related activities in the past year (i.e., the same type of work), characteristics of all those related activities may not show up in the work history. Also, the respondent may be giving a job characteristic as opposed to how long he/she actually has worked at a job. Generally, analysts have to determine how they wish to handle such inconsistencies in relation to their research. Sometimes crosschecking all data in the MFLS related to work for the individual in question may provide information to make reasonable guesses as to what inconsistent values should be. As with all survey data, there are always responses that don't make sense--people are people and none of us is perfect.