Search Type Title Author/Investigator Search Advanced Search

Table of Contents | Previous | Next

The NLSY97 Data, Its Strengths and Limitations

Background of the NLS Program and Structure of the NLSY97 Cohort

The National Longitudinal Survey of Youth is part of a broader collection of surveys sponsored by the Bureau of Labor Statistics within the Department of Labor, and collectively known as the National Longitudinal Surveys, or NLS. Their primary function is to gather information at multiple points in time on the labor market experiences of groups of men and women. The data are longitudinal, meaning that once a cohort is sampled the selected individuals are then re-interviewed (either annually or biennially) over several years. The NLS includes a total of eight cohorts of men and women. Four of these were selected in the 1960s and those surveys have since ended; the four more recent cohorts continue to be interviewed. Thus, for several panels of men, women and children, the data capture a significant segment of the individuals’ lifetime, while also reflecting detailed changes from one year to the next. The table below summarizes these cohorts as described in the official NLS Handbook of 2005, including definition of the cohort by age and birth year, original sample size, and the span of years covered by the interviews.

Cohorts of the NLS

Survey Group	Age cohort	Birth year cohort	Original sample	Initial year/ latest year	Number of surveys	Number at last interview	Status
Older men	45-59	4/1/06-3/31/21	5,020	1996 / 1990	13	12,092	Ended
Mature women	30-44	4/1/22-3/31/37	5,083	1967 / 2003	21	2,237	Ended
Young men	14-24	4/1/41-3/31/52	5,225	1966 / 1981	12	3.398	Ended
Young women	14-24	1943-1953	5,159	1968 / 2003	22	2,859	Ended
NLSY79	14-21	1957-1964	212,686	1979 / 2004	21	37,724	Continuing
NLSY79 children	birth-14	-	4-	1986 / 2004	10	33,190	Continuing
NLSY79 young adults	515 and older	-	4-	1994 / 2004	6	34,238	Continuing
NLSY97	12-16	1980-1984	8.984	1997 / 2004	8	67,756	Continuing

1 Interviews in 1990 also were conducted with 2,206 widows or other family members of deceased respondents. 2 After dropping the military (in 1985) and economically disadvantaged nonblack/non-Hispanic oversamples (in 1991), the sample contains 9,964 respondents eligible for interview. 3 The latest sample size available is from the 2002 survey. 4 The size of the NLSY79 child sample depends on the number of children born to female NLSY79 respondents, attrition over time, and the gradual aging of the children into the young adult sample. The size of the young adult sample depends on the number of children who reach age 15 in each survey year. Information about the number interviewed in each survey is available in chapter 4. 5 In 1998 only, the young adults eligible for interview were limited to those ages 15 to 20. 6 The latest sample size available is from round 7.

The most recent cohort, the NLSY97, is the focal point of this paper. Its purpose as stated by BLS in the NLS Handbook is to “document the transition from school to work and from adolescence to adulthood.” As such, it is representative at the national level of individuals born between 1980 and 1984, i.e., between the ages of 12 and 16 in 1997. The sample includes 8,984 members, including an over-sample of just over 2,000 additional blacks and Hispanics. The progression from Round 1 to Round 8 reflects a high level of sample retention, just over 86 percent, which is typical for surveys in the NLS program.

It is important to note that not all these respondents represent unique households. The NLSY97 interviewed all adolescents of qualifying age within a household. As a result, just over 5,000 individuals were from single respondent households and just under 4,000 were from multiple sibling households. The inclusion of siblings as respondents represents a unique opportunity to study variability of outcomes within a household. BLS recommends caution when generalizing from studies of siblings within the NLSY97, however, because these households are not representative of all sibling households.

In addition to the annual youth questionnaire, the NLSY97 contains special data supplements in various rounds. These include a detailed interview of a youth’s resident parent in round 1, a household income update completed by a residential parent in rounds 1 through 5, the collection of high school transcripts in 2000 and 2004, surveys of secondary schools located within the primary sampling units in 1996 and 2000, and test scores for the Armed Services Vocational Aptitude Battery (ASVAB) in 1997. In addition, the data contain screener information for all members of the household in round 1 and roster information for household members and immediate family outside the household in all rounds.

Topical Content

Although the study of labor market outcomes is a primary goal of the NLSY97 survey, it also contains a rich set of covariate outcomes for the respondents--pertaining to schooling, training, income, assets and program participation, family formation and family background, expectations, attitudes, behaviors and time use, health, environmental variables and event history variables. For a full description of the array of variables within each of these domains see the NLSY97 Users’ Guide. 2 Here we briefly highlight the content of particular interest to the marriage and family formation agenda.

A recent ACF review of data sources available to conduct research on marriage and the family concluded that the NLSY data are well-suited for measuring a variety of factors related to family formation, including economic, cultural and demographic influences. 3 Among several valuable data sets reviewed, it is shown to be the only survey that combines rich measures of relationship quality with detailed relationship histories (including cohabitations), extensive employment and human capital measures, and additional well-being outcomes with annual interviewing and a nationally representative sampling frame.

Specific details on NLSY97 content are presented below within the four marriage-related topics presented to the panel for discussion:

Pathways to cohabitation and marriage:

• Detailed annual questions on cohabiting and marital partners have been assembled into event history variables that track relationship stability and transitions.
• Demographic variables on the other biological parent of the respondent’s child, even if they have never cohabited or married.
• Variables are included that measure expectations of marriage within the coming year.
• Marital history of youths’ parents captures intergenerational influences.

Relationship between family formation/marital status and employment:

• Detailed employment data on all jobs held by the respondent include weekly employment status, as well as benefit eligibility and average weekly hours and earnings.
• Job descriptions include measures of industry, occupation, employer size, circumstances of hire, level of satisfaction and union status.
• Employment gaps are described in terms of timing, reasons and job search efforts.
• Other covariates possibly related to selection into marriage are included, such as education, training, and aptitude scores (ASVAB).
• Related outcomes, such as child-care use by type of arrangement and welfare receipt, are also covered in detail.
• The event history structure allows an analyst to capture timing of transitions more precisely, to better enable estimation of temporal relationships.

Marriage outcomes:

• Extensive measures of relationship quality between youths’ parents (or parent figures), both as reported by the youth and as reported by the responding parent, include supportiveness, degree of conflict and conflict resolution. Youth-reported measures are collected over multiple years.
• As youth transitions into young adulthood, there are similar measures of relationship quality between him/her and his/her partner are included as well.
• The data support detailed breakouts by marital status of parental interactions with youth--including measures of family routines, parental engagement, decision making and limit setting—and contain information on contact with nonresident parents.

Child wellbeing in different family structures:

• Youth well-being is measured across multiple domains, including educational attainment, employment, obesity, risk-taking behaviors and emotional well-being, measured at multiple points in time. Parents also provided more detailed information on youth health status during round 1.
• The analyst is able to observe variation in youth wellbeing across multiple family structures--including married step, married bio, cohabiting step, cohabiting bio, adoptive and single-parent households.
• Parents’ retrospective information allows researchers to measure impact on children of total time spent during childhood in married versus nonmarried households, as opposed to capturing only a point-in-time relationship.
• Future data collection will offer the potential to observe well-being of the youths’ children and to observe changes over the life-course.

Data Access and On-Line Resources

There are currently three methods for obtaining the NLSY97 data. (1) The user may search and extract variables from the NLSY97 using the on-line extraction tool known as Web Investigator, available through links on the BLS website. This tool contains documentation on each variable, including number of observations and missing values, as well as links to the NLSY97 user guide. (2) One can order a CD that will contain all cohorts of the NLS as zipped files, along with extraction software developed by the Center for Human Resource Research (CHRR) at Ohio State University. (3) One can download a particular cohort, which is accompanied by a copy of the extraction software to install on the user’s hard drive. Additional details on data access can be found in Appendix A of this paper. User guides and user handbooks are also available for download from the BLS website. 4

Current Limitations of the NLSY97

In addition to its many recognized strengths, the NLSY97 has several limitations in its utility for marriage and family research. These limitations, which are summarized below, are less a sign of weakness in the data than a function of the extensive capacity of the data, plus the fact that the survey was originally developed to study labor market issues. Some limitations particularly relevant to the data’s utility in the topical areas discussed above include the following:

• Richness of data. The sheer volume of measures within the NLSY97 data makes it a powerful tool for studying marriage-related outcomes, but also lends great difficulty to identifying particular measures of interest. For example, an on-line search of the NLSY97 variable names for the word ‘cohabitation’ yields nearly 600 variables.
• Ambiguity across generations. The content on family formation outcomes for multiple generations is a powerful benefit of the NLSY97. But variable names often inadequately distinguish between them (e.g., it can be difficult to tell whether a variable is referring to the youth’s father or the father of the youth’s child).
• Lack of specificity in Users’ Guide. Given the broad array of outcomes covered in the NLSY97 Users’ Guide, it can only touch on some topics such as family formation in general terms. While it describes the types of outcomes covered, there is little detail on specific variables and how to use them.
• Skip patterns that are complex to disentangle. Because the questionnaire is extensively tailored to respondent circumstances, the same construct may be collected from different respondents across a number of different variables. As a result, numerous variables have relatively small numbers of valid responses. When a user discovers a variable with a large proportion of missing data, uncovering the implication for one’s analysis can be quite time consuming. As there is no consistent information on the universe of eligible respondents for each question, identifying why certain observations are missing requires one to trace back through skip patterns across several preceding variables.
• Key measures that are hard to calculate. Several measures fundamental to the family formation agenda can be calculated based on the existing data. This requires extensive effort as there are no corresponding created variables. A primary example of this is whether the youth’s current cohabiting partner is the father of the youth’s child.
• Lack of interview questions on other key outcomes. Key outcomes not addressed include measures of in-kind contributions by the nonresident parent of Respondent’s child, welfare receipt by Respondent’s mother beyond 1997, non-cash assistance Respondent received from family if not living at home (other than child care, gifts and allowance), and wellbeing measures for child of Respondent.
• Lack of control totals on some key subpopulations. As noted above, several key subpopulations are difficult to isolate due to the lack of a single identifying variable. The lack of control totals for these groups further limits researchers’ ability to have confidence in their estimates.
• Limited information on parenting of young children. By definition, the respondents to this survey were already teens when they were first interviewed. Thus, we lack information on how their earlier childhoods were affected by the type of parenting they received. At this point we also lack information on parenting behaviors of the youth themselves because many are just now entering adulthood.
• Inability to support sub-national estimates. With initiatives at the state and local level to promote healthy marriage, contextual measures of family formation outcomes at the sub-national level would be valuable. However, sample sizes within the NLSY surveys are only sufficient to represent national-level populations. This is also true of most other surveys containing the same level of detail, e.g., the National Survey of Family Growth and the Survey of Income and Program Participation.

---

 

Table of Contents | Previous | Next