The National Survey of Family Growth

Transcription

1 The National Survey of Family Growth Anjani Chandra, Ph.D. National Center for Health Statistics Using Secondary Data for Analysis of Marriage and Family 2009 ICPSR Summer Program Workshop Ann Arbor, Michigan 1

2 Overview of presentation 9:00 9:30: 9:30 10:30: 10:30 10:45: 10:45 12:00: 12:00 12:15: 12:15 12:30: Overview of the NSFG Contents of NSFG Break Accessing & using the NSFG Update on the Continuous NSFG Wrap-up 2

3 3 Overview of the National Survey of Family Growth

4 Contacting the NSFG team at NCHS: NSFG staff: Mail: Joyce Abma, Anjani Chandra, Casey Copen, Jo Jones, Gladys Martinez, Bill Mosher (Team Leader), Stephanie Ventura (Branch Chief) NSFG Team, NCHS, 3311 Toledo Rd, Rm 7318, Hyattsville, MD Voice mail: Web site: 4

5 a 5

6 On the NSFG web site, you will find: Most everything you need to do research with NSFG Cycles 1-6: Data files (with program statements for selected software packages) User s Guides & Codebooks (incl. interactive codebooks) Questionnaires Key Statistics from the NSFG Cycle 6 Lists of 570+ publications from NSFG Cycles 1-6 A shortcut to subscribe to the NSFG Listserv New information about NSFG and announcements 6 Short description of NSFG (including information on continuous interviewing)

7 Purposes of the NSFG 7 Responding to Sec 306 of the PHS Act: NCHS shall collect statistics on family formation, growth, & dissolution. Explaining birth rates: contraception, sexual activity, infertility Measuring risk of HIV and sexually transmitted diseases: # of sexual partners, condom use, drug use, same-sex contact Collecting data on: Families: Marriage, divorce, cohabitation Roles of fathers in raising children Attitudes about marriage, parenthood, and gender roles

8 Social factors Race/ethnicity Religion Labor force participation Education Income Access to health care Family background Community environment (economic, social, etc) Intermediate variables Intercourse variables: Timing of first intercourse Percent of women who ever had intercourse Time spent in marriage (separation, divorce, cohabitation) Frequency of intercourse Conception variables: Contraceptive use Sterilization Infertility Fertility 8 Pregnancy outcome (gestational) variables: Miscarriage and stillbirth Induced abortion Live births

9 NSFG history in brief Cycle Year Scope Ever-Married Women Ever-Married Women All Women # of Interviews Over-Samples Average Length Incentive 9,797 Black Women 60 Minutes No 8,611 Black Women 60 Minutes No 7,969 Black Women Teens 60 Minutes No Women ,450 Black Women 70 Minutes No Women ,847 Black Women Hispanic Women 100 Minutes $ Women Men (First time) 12,571 W = 7,643 M = 4,928 Blacks Hispanics Ages W= 85 min M= 60 min $ (contin -uous) Women Men Goal: 5,000 per year Blacks Hispanics Ages W=80 min M=60 min $40 9

10 Agencies supporting the NSFG National Center for Health Statistics Office of Population Affairs Eunice Kennedy Shriver National Institute for Child Health & Human Development (NIH) OASPE Ass t Secretary for Planning & Evaluation ACF s Children s Bureau CDC s Division of HIV/AIDS Prevention CDC s Division of Reproductive Health CDC s Division of STD Prevention 10

11 Cycle 6 NSFG: Key survey features Contractor: Univ Michigan s Institute for Social Research (ISR) Multi-stage probability sample in 121 areas Blacks, Hispanics, year olds over-sampled Independent samples of males & females in U.S. household population (not couple-based) 12,571 total (7,643 female and 4,928 male) In-person interviews using laptop computers Average length of 60 minutes for males & 80 minutes for females CAPI (interviewer-administered) for majority of questions ACASI (self-administered) averaged 20 minutes 1,794 (about 14% of total) interviews were conducted in Spanish 11 Informed consent procedures Signed consent for adults (18-44) Signed assent & signed parental consent for minors (15-17)

12 Cycle 6 Sample size by subgroup FEMALE MALE TOTAL TOTAL 7,643 4,928 12, years 1,150 1,121 2, years 6,493 3,807 10,300 Hispanic 1,589 1,123 2,712 Black 1, ,460 White/other 4,524 2,875 7,399 Response rate: Total 79% Female=80% Male=78% 12

13 NSFG - sample weights Adjust for: Oversampling Screener and Interview non-response Adjusted to Census Bureau population estimates (age, race, sex) It s getting easier to use the weights: SAS, STATA, SPSS all have software that will use weights and compute correct sampling errors. Examples are on the NSFG web site. 13 So use the weights and design variables to make valid national estimates.

14 Who uses NSFG data? Researchers traditional academic research in the social sciences and (increasingly) public health, using the public use data files --MA theses, PhD dissertations, journal articles, book chapters Translators: Nat l Campaign to Prevent Teen & Unwanted Pregnancy, Child Trends, Guttmacher Institute: --short reports for policy makers, program field staff, funders, etc. Browsers: Reporters, Congressional staff, students, teachers who want a quick statistic 14

15 The NSFG data have been used for: Over 570 published journal articles and NCHS reports: see HP 2010 and 2020 Objectives on Family Planning, HIV, & STD. Statistical Abstract of the United States (tables 91-95) National Campaign to Prevent Teen & Unwanted Pregnancy Briefings for government officials at all levels. 15

16 Selected research articles using NSFG Bratter JL, King RB. But Will It Last? : Marital Instability among Interracial and Same-Race Couples. Family Relations 57(April): Lehrer EL. Age at marriage and marital instability: revisiting the Becker Landes Michael hypothesis. Journal of Population Economics 21: Hayford SR, Morgan SP. The Quality of Retrospective Data on Cohabitation. Demography 45 (1): Hynes K, Joyner K, Peters HE, DeLeone F. The transition to early fatherhood: National estimates based on multiple surveys. Demographic Research 18(12):

17 Selected research articles using NSFG Lamb KA. I Want to Be Just Like Their Real Dad : Factors Associated with Stepfather Adoption. Journal of Family Issues 28(9): Manlove J, Logan C, Ikramullah E, Holcombe E. Factors associated with multiple-partner fertility among fathers. Journal of Marriage and Family 70(2): Lichter DT, Graefe DR. Men and Marriage Promotion: Who Marries Unwed Mothers? Social Service Review 81(3): Teachman J. Complex life course patterns and the risk of divorce in second marriages. Journal of Marriage and Family 70 (2):

18 18 Selected findings generated from NSFG data

19 Percentage of men and women who ever had a biological child, by age: US, Percent Women Age Men Martinez GM, Chandra A, Abma JC, Jones, J, Mosher, WD. Fertility, Contraception, and Fatherhood: Data on Men and Women from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 26. May Hyattsville, MD: National Center for Health Statistics. Chandra A, Martinez GM, Mosher, WD Abma JC, Jones, J. Fertility, Family Planning, and Reproductive Health of U.S. Women: Data from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 25. Dec Hyattsville, MD: National Center for Health Statistics

20 Percentage of men years of age who have ever cohabited or are currently cohabiting, by education: United States, 2002 Not HS grad or GED HS diploma or GED Some college Bachelor's degree or higher Ever cohabited Cohabiting experience Currently cohabiting Martinez GM, Chandra A, Abma JC, Jones, J, Mosher, WD. Fertility, Contraception, and Fatherhood: Data on Men and Women from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 26. May Hyattsville, MD: National Center for Health Statistics.

21 Probability of first marriage for men and women years of age, by age : US, 2002: Men Women Age 21 Martinez GM, Chandra A, Abma JC, Jones, J, Mosher, WD. Fertility, Contraception, and Fatherhood: Data on Men and Women from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 26. May Hyattsville, MD: National Center for Health Statistics. Chandra A, Martinez GM, Mosher, WD Abma JC, Jones, J. Fertility, Family Planning, and Reproductive Health of U.S. Women: Data from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 25. Dec Hyattsville, MD: National Center for Health Statistics.

22 Percentage of ever married women who had their first intercourse 5 or more years before marriage, by year of first marriage: US, Before Year of First Marriage 22 SOURCE: Chandra A, Martinez GM, Mosher, WD Abma JC, Jones, J. Fertility, Family Planning, and Reproductive Health of U.S. Women: Data from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 25. Dec Hyattsville, MD: National Center for Health Statistics.

23 Percentage of men and women years of age who agree or strongly agree with the statement, It is all right for unmarried 18 year olds to have sexual relations if they have strong affection for each other, by importance of religion: US, Men Women Very important Somewhat important Not important 23 Martinez GM, Chandra A, Abma JC, Jones, J, Mosher, WD. Fertility, Contraception, and Fatherhood: Data on Men and Women from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 26. May Hyattsville, MD: National Center for Health Statistics.

24 24 Contents of the NSFG

25 Female questionnaire (2002 and current) A. Age, race, education, family & parental background B. Pregnancy history, care of nonbio kids, & adoption C. Marital & cohabitation history; first sexual partner; sex education; sexual partner history for last 12 months D. Sterilization operations; fecundity impairments E. Contraceptive history; all methods ever used, first method used, used in last 3 years, partner specific use, consistency of condom use; wantedness of births (old & new measures) 25

26 Female questionnaire (2002 and current) cont d F. Use of medical services for birth control & for general health G. Desires & intentions to have children in the future H. Infertility services; reproductive health conditions; HIV testing I. Health insurance; religion; work; child care; attitudes on sex, parenthood & marriage J. Audio Computer-Assisted Self-Interview (ACASI) 26

27 Male questionnaire (2002 and current) A. Age, race, education, family & parental background; numbers of wives & cohabiting partners B. Ever had sex; sex education; ever had a child; number of partners in lifetime and last 12 months; vasectomy; infertility; dates of 3 most recent partners in last 12 months or last sex ever C. Current wife or cohabiting partner: characteristics; children with her; contraceptive use D. 3 most recent partners in last 12 months; children with them; contraceptive use; first sexual partner ever E. Former wives and first premarital cohabiting partner; children with them; contraceptive use 27

28 Male questionnaire (2002 and current) cont d F. Other births and pregnancies fathered: characteristics of their mothers G. Activities with biological and adopted children respondent lives with, and does not live with H. Desires and intentions to have children in the future I. Health insurance; health conditions; use of family planning, infertility, and other health care services; HIV testing J. Religion; military service; work; attitudes on sex, marriage, parenthood K. Audio Computer-Assisted Self-Interview 28

29 Types of data collected in Audio CASI Height & weight Pregnancies had or fathered Alcohol, smoking (females only), and drug use in last 12 months Vaginal, oral, and anal sex with opposite-sex partners Same-sex sexual experience (greater detail for males) Non-voluntary sex (18-44 only) HIV/STI risk behaviors Sexual orientation and attraction STI history 29 Income and public assistance received

30 Percentage of sexually experienced women years of age whose first intercourse was not voluntary, by age at first intercourse: US, Under and over Age at first intercourse 30 SOURCE: Chandra A, Martinez GM, Mosher, WD Abma JC, Jones, J. Fertility, Family Planning, and Reproductive Health of U.S. Women: Data from the 2002 National Survey of Family Growth. Vital & Health Statistics Series 23, Number 25. Dec Hyattsville, MD: National Center for Health Statistics.

31 Collecting data on sex in CAPI & ACASI In CAPI (interviewer administered): We collect data only on heterosexual vaginal intercourse. Sex Babies In ACASI (self-administered), we ask about: Opposite sex and same sex partners All types of sexual activities (vaginal, oral & anal) Sex risks for STDs and HIV 31

32 Percentage of males and females years of age who have had each type of sexual contact: US, Males Females Any opp-sex contact Vaginal intercourse Oral sex with opp-sex Anal sex with opp-sex Same-sex contact 32 Source: Mosher WD, Chandra A, Jones J. Sexual Behavior and Selected Health Measures: US, Advance Data Report 362. Sept 15, 2005.

33 Sexual experience of females years of age, by age: US, vaginal intercourse at all oral sex but no vaginal 33 Source: Mosher WD, Chandra A, Jones J. Sexual Behavior and Selected Health Measures: US, Advance Data Report 362. Sept 15, 2005.

34 34 Accessing NSFG data & documentation (including questionnaires)

35 35 NSFG Questionnaires

36 NSFG Questionnaires for Cycle 6: 2 Formats CAPI Reference Questionnaire (CRQ) format Detailed specifications for the interview CAPI-Lite format Abridged version of question wording variants Shorter description of skip patterns 36

37 CAPI Reference Questionnaire (CRQ) format What is included in the CRQ Flow checks Specify precise routing through interview Provide details about Blaise-computed variables Edit checks Data consistency checks Indication of survey aids Show cards, Help screens, Life History Calendar references 37

38 Female CRQ Section C Marital & Relationship History Titles of Subsections 38

39 Flow Checks in CRQ Computed variable definition Routing 39

40 CAPI-Lite Version Shorter skip pattern 40

41 Available data files from NSFG Public-use files for Cycles 1 thru 6 are available free on CD- ROM. Cycle 6 PUF data also downloadable from webpage. ACASI data files from Cycles 5 & 6 are available free on request but require a signed User Agreement. Contextual data files for Cycles 5 and 6 are available through the Research Data Center at NCHS. Interviewer observation data file is also available through the RDC 41

42 NSFG Cycle 6 (2002) public-use files Female respondent file one record for each of the 7643 women Male respondent file one record for each of the 4928 men Pregnancy file one record for each of 13,593 pregnancies Information on completed and current pregnancies 42

43 Public-use data files from Cycle 6 (2002) NSFG Available free on CD-ROM 43

44 44 Cycle 6 Public-Use Data & Documentation

45 Download Data & Program Statements from Webpage/FTP site Program Statements ASCII data files 45

46 46 Cycle 6: Other Data Files

47 ACASI files for Cycle 6 (2002) The data and documentation from the self-administered questionnaire are available on request and without charge. Provide a brief summary, on your organization s letterhead, of how you propose to use the data. (If you are a student, a faculty advisor should also provide a letter of support for the proposed research.) Submit a signed User Agreement Available on the NSFG website All users must sign the agreement 47 A CD-ROM will be sent to you. Omitted items files for Cycles 4 and 5 are also available

48 Cycle 6 (2002) Contextual data files (in RDC) Use remotely or in the NCHS Research Data Center (RDC) application required (available on website), charges apply Community characteristics are shown for State, county, census tract, block group 2 points in time (2000, 2002) Linkage to over 1000 contextual variables Codebook and variable list are available on the website State of residence only available through RDC, but enables linkage to user-provided state-based indicators

49 Finding based on NSFG contextual data Percent of women years of age who are currently married by race and Hispanic origin and community male unemployment rate in 1990: United States,

50 a 50

51 Key Statistics from the NSFG (from A to Z) Link to cited reports 51

52 52 Key Statistics from the NSFG (from A to Z)

53 53 Cycle 6 User s Guide & Codebooks

54 54 Codebook entry for a variable, taken from bookmarked pdf of full codebook

55 Cycle 6 Webdoc & User s Guide Interactive codebook & user s guide in pdf 55

56 56 Search engine in webdoc

57 Webdoc Example location/origin Description or question text Links to recode specifications 57

58 User s Guide (pdf format) Main Text File indexes Recode specs 58

59 Main Text of User s Guide Includes information on: Organization and use of data files Questionnaires and codebooks Data preparation, including methods to prevent inadvertent disclosure of respondent identities Item-specific user notes Missing data Imputation procedures Date codes (century month calculation) 59

60 Appendix 1: File Indexes Contains listings of all variables included on the public-use male, female, & pregnancy files Variable name Column locations Description (aka, short labels ) Includes question number if raw/asked item Type of variable: raw, computed, recode 60

61 Types of variables provided on the data files Raw : most variables. Usually not edited. Not imputed. May have some missing data. (e.g. AB-1 MARSTAT) Blaise-computed : defined within the instrument during interview primarily to help with routing, etc. Some are included in the data set. (e.g. fmarit) 61 Recode : Best to use recodes if available. Specs provided in User s Guide. All missing cases imputed (expected value generated by multiple regression). Inconsistencies may exist between raw variable and recodes because recodes are edited and imputed. Imputation flags indicate whether & how imputation was done (e.g., RMARITAL recode and RMARITAL_I imputation flag)

62 Appendix 2: Recode Specs Contains list of recodes for each of the data files listed by section Contains recodes specs (instructions for creating recodes) as well as some notes on imputation constraints 62

63 63 Using NSFG Data FAQ s

64 Frequently Asked Questions from NSFG Users 1) Why is it so important to use NSFG s recodes? 2) What is the difference between the female respondent file and pregnancy file? 3) Do I need to use the weights and design variables? 4) How do I combine male & female data from Cycle 6? 5) How do I pool data across NSFG cycles? NOTE: A lot of these answers are in the User s Guide or on the NSFG webpage. 64

65 1) Why is it so important to use NSFG s recodes? Constructed variable on which missing values are imputed Imputation flags included on file to indicate whether & how imputation was done Allow comparable, reproducible measurement of key variables, which is critical for valid national estimates Some recodes are very simple, and some are quite complex RCURPREG simple (straightforward transfer from a single raw variable) CONSTAT complex (draws upon many variables) Full specifications are provided in User s Guide Appendix 2 NSFG data cleaning and editing focuses on recodes. (We don t edit most of the other variables on the file. As noted in the documentation, inconsistencies will exist.) 65

66 2) What is the difference between the female respondent file and pregnancy file? Respondent File 1 record per female respondent Selected arrays of pregnancybased variables included to minimize need for merging (see page 8 of User s Guide for list) Cycle 6 has 7,643 records with a record length of 4,927 Pregnancy File 1 record per pregnancy Selected respondent-based variables included to minimize need for merging (see pages 8-9 of User s Guide for list) Only contains records for respondents who have reported any pregnancies to interviewer Cycle 6 has 13,593 records (lrecl=447) representing pregnancies reported by 5,033 ever-pregnant respondents 66 Merging of respondent and pregnancy-based data can be done in either direction, and relies on CASEID and PREGORDR variables. See pages of User s Guide for SAS syntax, adaptable for other software.

67 3) Do I need to use the weights and design variables? YES, if you want to make valid national estimates Ignoring the weights will skew your point estimates because each respondent represents thousands of people. (Also, Cycle 6 weights vary considerably more than prior cycles.) Ignoring the multistage probability design features (in particular, the oversampling and nonresponse adjustments) will skew your variance estimates and inferences. Design variable Cycle 6 (2002) Stratum variable SEST values 1-84 Cluster/Panel Variable Final poststratified, fully adjusted case weight values 1-2 SECU_R --female resp file SECU_P --female preg file SECU --male file FINALWGT 67 The next several slides will describe variance estimation for Cycle 6, but much of it applies to other cycles as well.

68 3a: Variance estimation examples on the NSFG webpage 68

69 3b. Variance Estimation Examples on the NSFG webpage (cont d) We provide full syntax and output (in pdf files) using 4 major software packages: SAS 9.1 SUDAAN STATA 8.0 WesVar 4.1 (for Windows) Users can then modify to suit their own computer setup and analysis project. 69

70 3c. Variance Estimation Examples on the NSFG webpage (cont d) 9 separate examples to illustrate common estimation tasks: Percentages & percentage distributions Total counts or numbers Mean numbers Pooling data for males and females Use of subsetting statements (e.g., subpopn) Linear regression (continuous DV) Logistic regression (binary DV) 70

71 71 Variance Estimation Examples

72 3d. Variance Estimation Examples on the NSFG webpage (cont d) All examples illustrate comparability of estimates across all packages: Point estimates are equivalent Standard errors differ slightly Design effects vary a bit more 72

73 4) How do I combine male & female data from Cycle 6? Make sure you are using truly comparable variables Consult recode specs and other documentation This point also important when pooling data across Cycles See example under Variance Estimation Examples on webpage: Define new variables for stratum and cluster separately for males and females (finalwgt same varname for M&F) Append male and female case records (in SAS, use SET statement) Run analyses using the new variables you ve defined 73

74 5) How do I pool data across NSFG cycles? (female data) Design variable Cycle 5 (1995) Cycle 6 (2002) Stratum variable COL_STR SEST Cluster/Panel Variable Final post-stratified, fully adjusted case weight PANEL POST_WT SECU_R female resp SECU_P female preg FINALWGT Cycle 5 NSFG: stratvar=col_str; panelvar=panel; weightvar=post_wt; survey=1995; Cycle 6 NSFG: stratvar=sest *100; panelvar=secu_r*100; weightvar=finalwgt; survey=2002; 74

75 Cycles 1-5 of the NSFG Public use files from all previous cycles are available free on CD-ROM and can be downloaded from the website Codebooks, and documentation, are on the NSFG web site Omitted Items files (Cycle 4 & 5) are, upon request (similar process as for Cycle 6 ACASI files) Contextual files for Cycle 5 data available through RDC 3 points in time (1990, 1993, 1995) 4 levels (state, county, census tract, block group) Charges apply 75

76 76 Cycles 1-5 Data & Documentation

77 77 Cycle 5 tools on webpage

78 78 The Continuous NSFG

79 Cycle 7 not really cycles anymore! Continuous interviewing for NSFG has begun. Interviewing began June 2006 Why did we move from periodic to continuous interviewing? More timely and cost effective ability to supply data on: Populations at risk for HIV/STD and unintended pregnancy Emerging public health and fertility concerns: HPV vaccine New contraceptive methods on the market Emergency contraception (ever, last 12 months, provider) Ability to quickly make corrections to the instrument (if needed). Year 2 revisions were incorporated into field instrument starting June Cost efficiencies derive from stability of instrument over time and smaller staff/management requirements.

80 Cycle 6 (2002) versus Continuous NSFG (2006-present) Cycle 6 Continuous NSFG Instrument: content comparable to Cycle 6 Interviewing period: Mar 02 Mar 03 Design: multistage probability sample Staff size: 250 temporary/part-time interviewers working nationally in 121 PSUs over the fieldwork period Sample size:12,571 Data release: 20 months after the last interviews in March 2003 Interviewing period: June 06- Design: multistage probability sample Staff size: 40 full-time interviewers working nationally in 33 PSUs in any given quarter Sample size: 5000 a year Data release (goal): months after the last interviews in Dec 2008 files released every 2-3 years thereafter 80

81 How is the NSFG done under continuous interviewing? Survey setting: face-to-face survey (CAPI). Screen to locate year olds (60% of households). one person is selected per household. Female interviewers, trained for 1 week. $40 Incentive for the interview NEW IN CYCLE 7: Continuous interviewing organized in 4 field periods of 12 weeks each phase 1 of 10 weeks phase 2 of 2 weeks ( double sample ) 81

82 Upcoming changes related to marriage, cohabitation, & family data New questions: Marriage/cohabitation attitudes & expectations (M&F) Mo/Yr of Bachelor s degree (M&F) Females: Children with each husband or cohabiting partner Males: More detail on nonbiological children New recodes: Describing men s marriages beyond the 1st Non-marital and marital cohabitations Household structure 82

83 Sample sizes in the 2002 & NSFG data sets * Change Total 12,571 13, Male 4,928 6, Female 7,643 7, Teen 2,271 2, Black 2,460 2,451-9 Hispanic 2,712 2, * numbers subject to small changes.

84 Schedule for data release from continuous interviewing Late 2009: First dataset released, around 13,500 interviews conducted in Late 2011: Second dataset released, around 20,000 interviews conducted in Release dataset every 2-3 years beginning in 2009 much better than every 6-7 years! 84

85 85 Wrap Up

86 Advantages of using NSFG data Nationally representative of men and women years of age Don t ignore the complex sample design: SAS, SPSS, STATA can help with weights and variance estimates now. Detailed information on life events Marriage Cohabitation Fertility Household information Fathering (men) Attitudes toward marriage/family 86

87 Advantages of using NSFG data cont d Recodes use them whenever possible NSFG webpage and interactive tools provide almost everything you need to know Experienced staff willing to assist you nsfg@cdc.gov Phone:

88 88 Questions?