Integrated Fertility Survey Series

1
Integrated Fertility Survey Series

• Pam Smock
• Felicia LeClere
• Christopher Ward
• Lynette Hoelter
• Peter Granda

2
Overview of the Presentation

• Introduction
• Data files used in the IFSS data product
• Description of methods used to harmonized data
  items
• Proposed data releases

3
Introduction

• Why An Integrated Fertility Survey Series (IFSS)?

4
Reason 1

• Dramatic changes in family patterns (e.g.,
  marriage, divorce, childbearing, nonmarital
  childbearing, cohabitation) over past five
  decades

5
Median Age at Marriage of U.S. Women and Men,
1950-2004
(Source U.S. Bureau of the Census)
6
Total Fertility Rate, U.S. 1950-2003
7
Percent of Births Outside of Marriage, U.S.
8
Percent of First Marriages Preceded by
Cohabitation by Marriage Cohort, U.S.
Source Bumpass Lu (2000 Table 3) Bumpass
Sweet (1989 Table 2) Kennedy Bumpass (2008,
Table 4).
9

• Researchers, policy makers, and the public want
  to both document understand such changes
• IFSS will provide unique opportunity to examine
  family fertility change across 5 decades

10
Reason 2

• Researchers have produced an extremely large
  scientific literature with the 10 surveys we are
  harmonizing
• However most studies (76) use only 1 survey,
  missing chance to study change over time
• Strong impediments to using multiple surveys

11
Impediments

• Changes in target populations
• Changing weighting imputation procedures
• Changes in question wording, coding schemes,
  content
• Users faced with multiple codebooks files

12
Reason 3

• Goal of the Demographic and Behavioral Sciences
  Branch (DBSB) of National Institute of Child
  Health and Human Development (NICHD) is to
  broaden current user community by making data
  more accessible

13
Broaden user community

• The vast majority of research supported by the
  DBSB has focused on basic science, with academic
  scholars the primary audience. . Yet, many of the
  findings generated. . .could be useful to public
  policy-makers and the general public if the
  results could be made more accessible. . .

14
Sample Topics in IFSS

• Contraceptive use
• Pregnancy childbearing, including intentedness,
  timing, relationship status
• Marital fertility
• Nonmarital fertility
• Premarital intercourse
• Union (marriage, cohabitation) formation
  dissolution
• Fertility expectations
• Social class racial/ethnic differences

15
Sample Research Questions

• What factors (e.g., religiosity, education)
  affect whether a pregnant unmarried woman marries
  before childbirth? Have the influence of these
  factors changed over time? How does their
  influence vary by race and ethnicity?
• How have fertility expectations the number of
  children one expects to have - changed over the
  past several decades?
• How has contraceptive use among unmarried women
  changed over time? What characteristics predict
  the likelihood of using contraception? Have the
  influence of these factors altered over time?
• How have the odds of divorce changed over the
  past several decades and have the risk factors
  for divorce changed?

16
Data files in IFSS
17
DATA COLLECTIONS COMPRISING IFSS

• Growth of American Families 1955, 1960
• National Fertility Surveys 1965, 1970
• National Survey of Family Growth Cycles 1-6
  (1973, 1976, 1982, 1988, 1995, 2002)

18
Survey Samples

• NSF
• 1965 Currently married women, 18-54
• 1970 Ever married women, 18-44
• NSFG
• 1973, 1976, 1982 Ever married or single women
  with their own children living in the household,
  15-44
• 1988, 1995, 2002 All women 15-44 (some over
  samples by race)

• GAF
• 1955 Currently married White women, 18-39
• 1960 Currently married White women, 18-44
  Previously married White women, 23-44, who were
  married and living with their husbands in 1955
  Currently married non-White women, 18-39
  (oversampled)

19
Current Availability

• All of the original files are currently available
  through the IFSS website
• Have SAS, SPSS, STATA ready-to-go and syntax
  files.
• Have on-line analysis capabilities through SDA

20
Generating the Harmonized Files
21
Goal of Harmonization

• To create variables that are substantially
  consistent over time
• Allow researchers to perform effective trend
  analyses

22
Harmonization process

• Variable search and identification
• Translation table development
• Generation of data set

23
Variable list from SDA

• Establish map of variables from all ten studies
• Survey Documentation and Analysis (SDA) groups
  variables by substantive areas of interest (e.g.,
  sociodemographics, union history, childbearing)
• Additional, related variables located in other
  sections are added to list (e.g.,
  husbands/partners race located in union history)

24
Searching for concepts

• Social science variable database (SSVD) is
  primary means of searching
• Beginning with 2002 NSFG, conduct searches for
  every variable/concept in list
• Once finished, continue searching for variables
  1995 NSFG that do not appear in 2002
• Process continues with remaining NSFG data sets
  (e.g., variables in 1988 that do not appear in
  1988 or 2002)
• Continue search until 1973 NSFG

25
Identifying candidate variables

• Establishing a minimum number of studies for a
  variable/concept ensures that the variable is
  useful across time
• 5-study rule most variables/concepts must appear
  in at least half (5) of the studies
• 2/3-study rule variables judged to be highly
  salient to contemporary researchers may in appear
  in 3, or occasionally 2, studies

26
Criteria used to judge variable comparability

• Question text
• Location of question in questionnaire
  (response-order bias)
• Respondent universe
• Response categories
• Recoded/raw status
• Imputed status

27
Translation table development

• Depending on type, translation tables use
  variable metadata, including
• Variable names
• Variable values
• Value labels
• Each harmonized IFSS variable name contains an
  "IFSS_" prefix
• In most cases, IFSS variable names follow the
  variable name used in the study where the concept
  most recently appears
• IFSS variable value labels follow a similar
  convention

28
Structure of translation tables

• Excel spreadsheet
• Original variable names, values, and value labels
• Response categories from original studies are
  matched with IFSS variable categories across rows
• Priority flags used when multiple variables for
  one concept appear in a given study

29
(No Transcript)
30
Translation tables error-checking

• Structure of translation table is checked for
  accuracy (original values, value labels)
• Recodes (position of rows, logic) are checked
• Original study metadata checked for accuracy and
  proper inclusion

31
Translation tables comparability notes

• Due to nature of studies, imperfect comparability
  arises
• Where differences exist, comparability notes
  inform users
• Question text, respondent universe, and response
  categories differ most frequently

32
(No Transcript)
33
Generation of data

• After completion of error-checking, translation
  tables are processed through SAS
• SAS program generates harmonized IFSS variables
• SAS program also generates error-checking
  diagnostics

34
Error-checking of data

• Error-checking parameters include
• Correct recoding of response categories
• Accurate assignment of IFSS variable names and,
  variable labels, and value labels
• Correspondence between original study variable
  frequencies and harmonized IFSS variable
  frequencies
• Resolution of errors

35
Data release

• At the conclusion error-checking, data are
  processed through Hermes
• DDI and XML markup and standardized codebook are
  produced
• Final check and website beta
• Public release (data and updated website)

36
Data Release

• October 2009 Demographic Variables
• July 2010 Union Formation Variables
• January 2011 Fertility History Variables
• July 2011 Contraceptive Use Variables