Errata for 1979-2000 Data Release

Errata for 1979-2000 Data Release

Important Information

The NLS Investigator contains the most recent release of each NLS cohort. Known problems with the 1979-2000 release of the NLSY79 are found below. Corrections have been made to items noted in the Errata of prior releases. For further questions, please contact NLS User Services.

Documentation Error

Subject: Breastfeeding, 1998-2000
The codebook for the 1998 and 2000 "months of age of child when breastfeeding ended" variables indicate the data have 1 implied decimal place. The data for these variables have 2 implied decimal places. The affected codebook pages are:

R6284100     ENTER # OF MOS OF CHILD? L1 1998
R6284200     ENTER # OF MOS OF CHILD? L2 1998
R6284300     ENTER # OF MOS OF CHILD? L3 1998
R6284400     ENTER # OF MOS OF CHILD? L4 1998
R6284500     ENTER # OF MOS OF CHILD? L5 1998
R6284600     ENTER # OF MOS OF CHILD? L6 1998
R6284700     ENTER # OF MOS OF CHILD? L7 1998
R6284800     ENTER # OF MOS OF CHILD? L9 1998

R6827100     ENTER # OF MOS OF CHILD? L1 2000
R6827200     ENTER # OF MOS OF CHILD? L2 2000
R6827300     ENTER # OF MOS OF CHILD? L3 2000
R6827400     ENTER # OF MOS OF CHILD? L4 2000
R6827500     ENTER # OF MOS OF CHILD? L5 2000
R6827600     ENTER # OF MOS OF CHILD? L6 2000
R6827700     ENTER # OF MOS OF CHILD? L7 2000
R6827800     ENTER # OF MOS OF CHILD? L10 2000

Data Errors

Subject: 2000 Relationship History Variables

A programming error in the creation of the Relationship History variables for the 1998 to 2000 period for number of spouses or partners reported and the relationship code of the current spouse/partner has led to some respondents being inaccurately coded as having no spouse or partner when one was present and vice versa. Preliminary analyses suggest that a similar problem may be present in 1998 as well for female respondents only. Evaluation of these data is underway, and corrected versions of any affected variable will be made available to users as soon as possible.

The variables known to be affected are:


Subject: Problems with Recipiency Event History Variables

Users should be aware of several problems that have been discovered in the calculation of the created Recipiency Event History variables (areas of interest RECIPIENT MONTH and RECIPIENT YEAR). These problems will also affect the Total Net Family Income and possibly the Poverty Status variables in previous survey years for some cases. Survey staff are currently reviewing all recipiency event history data, investigating the problems and making necessary corrections. Users should check this site for status updates concerning data corrections. CHRR anticipates that data corrections will be completed by March 2004. Any corrections completed by the time of the 2002 (round 20) data release will be included and documented on that release. The problems are described below.

  1. Dollar values for unemployment compensation: For calendar years 1978-2000, many of the dollar values for yearly and monthly unemployment compensation (UC) for both the respondent and spouse/partner are inaccurate. These dollar values were improperly edited to be substantially lower than they should actually be. From the 1993 interview until the 2000 interview, approximately 50% of those reporting UC receipt for themselves or their spouse/partners are affected. A much smaller proportion (around 10%) of UC recipients from the 1979-1992 interviews are affected. Until data corrections are available, users should edit the dollar values for cases that have an edit flag code (UNEMPR-EDIT-[YEAR] or UNEMPSP-EDIT-[YEAR]) of "3" or "5" and use the original value provided by the respondent to calculate correct monthly and yearly amounts.

    In addition, MONTHLY dollar values of UC are slightly off for data collected between the 1993 and 2000 interviews (inclusive). For data collected before 2002, MONTHLY values were calculated by multiplying the original WEEKLY values by "4" instead of "4.3". MONTHLY values may be corrected by dividing by "4" and remultiplying by "4.3".

  2. Edit flags for all recipiency programs: In general, for all survey years, cases which received a code "3" on the "[PROGRAM]-EDIT-[YEAR]" flags for variables other than UC receipt should be checked carefully. The dollar amounts reported by respondents in these cases were incorrectly edited (affecting both the resulting monthly and yearly dollar amounts). Those with a code "3" on the [PROGRAM]-EDIT-[YEAR] flags were edited under the incorrect assumption that a reported MONTHLY amount higher than an arbitrarily set maximum was actually a YEARLY amount. The values reported by respondents should be used, and researchers should make their own decisions about whether the value is "too high" or "too low" based on the benefit structure of the program in question.


  3. Seam problem for AFDC/TANF receipt: Monthly AFDC/TANF receipt information created from survey year 2000 (round 19) data contains a small seam problem. For data created from this year ONLY, some respondents were coded as "-4" in the interview month (or a month immediately before or after the interview month) when they should have received a value indicating the dollar amount of receipt. This problem affects only respondents who reported continuous receipt up until the interview month in that survey year, or who reported the interview month as their receipt stop month. This error results in deflated values for yearly AFDC/TANF dollar amounts and combined welfare dollar amounts. It also might cause users to overestimate the number of receipt spells for a given respondent, since it erroneously appears that respondents ended receipt and began another spell.


  4. Coding of non-receipt: In survey year 2000 (round 19), the conventions used for assigning "0" or "-4" on some of the recipiency created variables are not consistent with prior rounds. This is unlikely to affect many users since these codes both indicate non-receipt.


  5. Erroneous data for some June receipt: Respondents who reported during the 2000 interview that they had received program benefits during 1998 or earlier incorrectly may have been assigned a "-4" during the month of June. This problem is particularly prevalent for June 1998, because respondents interviewed in May 1998 or earlier reported June 1998 receipt in the 2000 interview. (For example, about 60 respondents out of over 200 who received AFDC have "-4" incorrectly assigned during June 1998.) The missing June data is far less prevalent for calendar years prior to 1998, but users should check receipt information for pre-1998 years carefully if it was reported in the 2000 (round 19) interview. Yearly receipt amounts are affected, with cases missing June data having a one-month undercount in the yearly total dollar receipt amount. An exception to this is a small number of respondents who did not know the dates for their receipt-their yearly totals may be correct.


  6. Recipiency stop month not counted: For data reported in survey year 2000, the stop month of receipt was not counted as a receipt month as it has been in recipiency event histories for other years. This leads to a truncated receipt spell and a slight downward bias in the yearly receipt totals.


  7. Total Net Family Income and Poverty Status variables: The problems described above may affect 1998 and 2000 Total Net Family Income values for some respondents. Those reporting UC or AFDC/TANF receipt in June 1998 and continuous AFDC/TANF receipt at the 2000 interview will be most likely to require adjustments to the Total Net Family Income values. In addition, the Poverty Status released for the 1998 and 2000 survey years may require adjustment for a subset of those cases.

    Last Modified Date: May 22, 2008 - 06:15 PM