1. T he NLSY79 Child Assessments NLS Summer Research Workshop Paula Baker CHRR at The Ohio State University July 2007

2. Aim & scope of the Child assessments Primary Aim - To assess key aspects of physical, cognitive, social and emotional development What assessments are used? What do they measure? How were they collected? What scores are available?

3. Types & modes of assessments Two types of assessments: 1. Established scales 2. Other indicators of child wellbeing – some scaled, some individual items Three modes of data collection: 1. Interviewer-administrated to child (in Child Supplement) 2. Mother report (in Mother Supplement) 3. Interviewer observation

4. The assessment domains: What do they measure? Cognitive – memory; math & reading achievement, word reception & recognition Socioemotional – developmental milestones; temperament; perceived self-competence Behavioral – prosocial; behavior problems; peer conflicts; risk taking Physical Health & Development – hgt & wgt measurements; self-reports

5. Cognitive: Interviewer-administered Memory - Wechsler Digit Span Achievement/Ability: PIAT Math PIAT Reading Vocabulary recognition/retention: PPVT-R

6. Cognitive assessments for little kids Short-term memory: McCarthy Verbal Memory Memory for Location Receptive vocabulary: Body Parts

7. Socioemotional Assessments Motor & Social Development (MSD) – from NHIS Temperament (What My Child is Like) – from Rothbart, Campos, and Kagan Self-Perception Profile for Children (SPPC) – Harter scales

8. Other socioemotional measures Child moods: Child 10-14 in Child Self-Administered Supplement Risk aversion scales: Child 10-14 report in CSAS Shyness: Interviewer-report in Child Supplement Psych referral: Mother-report

9. Behavioral, Mother Report Behavior Problems Index (BPI) Measures incidence and severity of child behavior problems Global Measure & subscales: 1. antisocial 2. anxious or depressed 3. hyperactive 4. headstrong 5. dependent 6. peer conflicts Trichotomous vs Dichotomous

10. The HOME-SF Mother Report Mother Report emotional and verbal responsiveness maternal acceptance of and involvement with her child organization of the environment presence of materials for learning variety of stimulation

11. The HOME Interviewer Observations Interviewer observes physical environment & parent-child interactions Interviewer Observations – not available if … No interaction observed No visit to the home (interview by phone) Child too young for assessment

12. The HOME-SF scales Three Scales: Overall (conservative imputation for missing) Emotional Support Cognitive Stimulation

13. Pattern of Assessment Administration

14. Assessment Scores What types of scores are available? Summary scores for each assessment See tables & appendices in users guide How are they computed? Procedures in users guide Conservative imputation for missing data: HOME, BPI, SPPC Not all assessments are normed: No normed scores for Temperament or SPPC Raw & standardized scores for Digit Span (no pctile)

15. Norms – Std vs %ile Types of scores: Standardized vs Percentile vs Raw Standard Score equal-interval measure one of the most common metrics can be compared across tests can be arithmetically manipulated Percentile Score ordinal scale more intuitive caution against using to manipulate scores by differencing or taking an average to represent growth or change

16. Testing conditions For each child interview: Overall Interviewer evaluation of testing conditions Type & amount of interference Child rapport w/ interviewer Attitude, motivation Physical limitations Perseverance, cooperation Language of administration For each assessment: Interruptions or premature termination Child’s energy level Others present & their impact

17. Valid assessment scores for interviewed children

18. Child Age Number of Interviews 12345678Total All Ethnicities <1 Year 11 1 Year 27 2 Years 24 16 40 3 Years 13 62 75 4 Years 12 3051 93 5 Years 5 2097 122 6 Years 10 1573 52 150 7 Years 17 1230146 205 8 Years 9 152395 73 215 9 Years 11 202138141 231 10 Years 12 181734113 92 286 11 Years 13 152321 56203 331 12 Years 10 172116 48151119 382 13 Years 12 111815 33 94223 406 14 Years 17 6 24 33 53152135 437 Total2032573914414975934941353011 Hispanic Total 50 71100 95109106 81 21633 Black Total 58 56 80104127158108 29720 Non-Black, Non-Hispanic Total 95130211242261329305 851658 Repeat Interviews: Children under age 15 as of 2004

19. Repeat assessments: PIATs & PPVT

20. New in the 2006 NLSY79 Child Mother Supplement embedded in the main Youth interview Child Self-Administered Supplement (CSAS) integrated into the Child Supplement (CS) Interviewers assess children age 4-14 – no younger ages Mothers part of “Early Bird” field effort – some children assessed later in survey year

21. Child Care, Family Structure, and Child Outcomes: Possibilities for Analysis Using the Children of the NLSY79 David Blau, OSU NLS User Workshop July 17, 2007

22. Introduction The NLSY79 and Children of the NLSY79 (CNLSY79) data are the best available resource for analyzing the determinants and consequences of child development – Sample size; overall, and by race/ethnicity – Depth and breadth of survey content – Longitudinal coverage – Intergenerational coverage – Sibling and cousin coverage

23. Intro - 2 There are some limitations: – The children are not a representative sample from a well-defined cohort – Child care data are somewhat limited – Not much information on time use or purchased goods: it is not a time use or expenditure survey – Less information on children born to young mothers – Not much information on absent fathers – Changes in survey design and content over time

24. Intro - 3 Outline of presentation: – Brief conceptual framework – Examples of data organization and analysis from 3 papers using CNLSY79: – The Effect of Child Care Characteristics on Child Development, David Blau, Journal of Human Resources, 34 (Autumn, 1999), pp. 786-822. – Child Care Choices and Children’s Cognitive Achievement: The Case of Single Mothers, Raquel Bernal and Michael Keane, working paper, 2006 (http://www.faculty.econ.northwestern.edu/faculty/bernal/Bernal_Keane_IV_Single_Mothers.pdf)http://www.faculty.econ.northwestern.edu/faculty/bernal/Bernal_Keane_IV_Single_Mothers.pdf – A Demographic Analysis of the Family Structure Experiences of Children in the United States, David Blau and Wilbert van der Klaauw, working paper, 2007, http://www.unc.edu/depts/econ/profs/blau/demog%20paper%20version%203.pdf

25. Conceptual Framework A “child outcome production function:” S ijt =  1 T ijt +  2 C ijt +  3 G ijt +  4 X ijt + µ j +  it + ε ijt, where: – S ijt is an outcome for child i of mother j at age t (e.g., a cognitive test score) – T ijt is a measure of cumulative maternal time input since birth – C ijt is a measure of cumulative child care input since birth – G ijt is a measure of cumulative goods inputs since birth – X ijt is a set of controls for the child’s endowment and environment – µ and δ are unobserved components of the endowment – ε ijt is a transitory shock and/or measurement error

26. Interpretation Distinction between a production function and a demand equation: only proximate determinants belong in a production function Coefficients represent the “technology” of child development; causal effects, other things equal Each input could be a vector: quantity and quality; effects could vary by age; non-linear; interactions

27. Typical Empirical Implementation Little information on maternal time input; assume it equals all maternal time not spent working for pay Little information on goods inputs: – Use the “HOME” score as a proxy (Blau) – Substitute income for goods (Bernal & Keane) Other variables (X): typically mother’s education, AFQT score, age, race, ethnicity; child’s age, birth weight.

28. Child care data Monthly record of child care used from birth through age 36 months (1986, 88, 92, 94+ surveys): num. of arrangements, type, location Child care in the 4 weeks prior to the survey: – type, location, payment arrangement (82-86, 88) – group size, number and training of caregivers (85, 86, 88) [“quality”]

29. Does child care “quality” matter for child cognitive and behavioral development? (Blau, JHR 1999) Group size, num. of staff, and staff training are not very good measures of child care quality; but are often used as proxies Data are a snapshot of the 4 weeks prior to interview date, and were measured in only three survey years, in the 1980s – cannot construct complete histories of quality Reported by the mother, not recorded by observers This paper uses data through 1992

30. Specification of child care variables Average quality from ages 0-2 (infant-toddler) Average quality from ages 3-5 (preschool) Other child care variables also averaged within these two age groups: – Mode, hours per week, months per year, num. of arrangements, payment

32. Outcomes Peabody Picture Vocabulary Test (PPVT), ages 3+, some repeat assessments Peabody Individual Achievement Tests (PIAT) in mathematics and reading comprehension; ages 5+, repeated Behavior Problems Index (BPI); ages 4+, repeated All normed to national samples with mean 100 and SD 15 Measured in even-number survey years, beginning in 1986

33. Some results

34. Interpretation Consider a group of ten children cared for by one provider. Adding a second adult leaves GS unchanged and increases the SCR by 0.1 (from 0.1 to 0.2). Splitting the group in half and providing a teacher for each of the two smaller groups causes group size to fall by five and SCR to rise by 0.1. Based on the results in Table 4 the estimated impacts of these two hypothetical experiments are: BPIPIAT-MPIAT-RPPVT Add a second adult: -.00.31.50*.27 Split the group in half: -.40.00 -1.35*.02 Adding a second care giver is predicted to increase the reading score by 0.50 (3.5 percent of a SD) and to have smaller and statistically insignificant effects on the other outcomes. Reducing GS from ten to five while also raising SCR by 0.1 is predicted to reduce the Reading score by 1.35. These are relatively small effects in view of the large changes in GS and SCR considered in these experiments, and in the latter case the net impact is negative rather than positive.

35. Does child care quantity matter for child cognitive and behavioral development? (Bernal & Keane, 2006) Much simpler specification of child care: cumulative time spent in child care, birth through age t Cumulative income as proxy for cumulative goods input Single mothers only Estimation by Instrumental variables, using state welfare waivers and welfare reform as the main source of variation in child care use: exploits geocode data Outcome is a combination of PPVT and PIAT math and reading test scores; includes test dummies; ages 3-6

36. Age profiles of maternal employment and child care use

37. Some results

38. Interpretation Baseline specification is column (2) in Table 8. This model implies that each additional quarter of full-time day care reduces a child’s test score by approximately 0.70%. Thus, a year of full-time child care is associated with a reduction of about 2.8% in child test scores. This corresponds to approximately.0282/.1861 = 0.15 standard deviations of the score distribution. Viewed another way, a 2.8% test score reduction at age 6 would translate into about a.053 to.070 year reduction in completed schooling. Interpretation: this is the effect of child care time relative to the effect of mother’s time, plus the effect of any change in goods inputs that the mother may choose as a result of using day care

39. Family structure and child outcomes (Blau and van der Klaauw, 2007, and in progress) Effect of growing up in a single-parent family Effect of family structure disruption Effect of presence of a step-parent Effect of living in a “blended family” Effect of cohabitation versus marriage And many other interesting questions

40. Data needed for family structure analysis Marital history Cohabitation history Pregnancy and birth history Household roster and relationships Child residence history Identity of men in relation to children

41. NLSY79 Family Structure Data In 1979, the survey collected information on the beginning and ending dates (to the nearest month) of up to two marriages. In subsequent waves, information has been collected on up to three changes in marital status that occurred since the previous interview: marriage, separation, re-uniting after a separation, divorce, death of a spouse, and re-marriage Cohabitation: household roster with relationship codes, including “partner” and “other non-relative.” Beginning in 1990, retrospective report of the beginning date of the cohabitation, if in progress at the interview date. Also, if married at the interview date: whether cohabited before the marriage began, and begin date. Complete redesign of the cohabitation questions in 2002. Beginning and ending date of cohabitations that did not turn into marriages. Cohabitations that lasted less than three months are ignored.

42. Identifying fathers Beginning with the 1984 interview, the mother is asked for every biological child present in her household whether the biological father of the child is present. Thus, when a woman lives with a man before or during the conception or birth of a child, identifying fathers is straightforward. Cases in which a woman conceives and bears a child while single are more difficult: can identify father only if she subsequently moves in with a man. If she never enters a union with a man following the birth of a child, we cannot identify the child’s father. If she moves in with a man and the union ends before the 1984 interview, then we cannot determine the identity of the child’s father. If a man moves in and out between interviews, we cannot determine the father of the child.

43. Analysis to date Mainly descriptive, demographic Also some analysis of policy and labor market determinants of family structure Some descriptive statistics and illustrative results from hazard models

47. Conclusion There are many interesting and useful questions about child care, family, structure, and child outcomes that can be addressed with the CNLSY79 It is not the ideal data set for all such questions But it is probably the first place to look for answers to such questions The most important feature of the survey: the children are being followed into adulthood => eventually, long term consequences of childhood experiences can be studied