What is the importance of the Integrated Fertility Survey Series?
Though researchers from a broad range of disciplines have produced a large body of research on patterns in families and fertility, the ability to make comparisons over time -- a central task for understanding family change -- has been constrained by difficulties in using multiple datasets to perform time-series analyses. Such difficulties include changes in respondent universe, weighting procedures, imputation protocols, question wording, and variable availability across studies. This is especially true when attempting to include surveys from the earlier years. The IFSS project attempts to address these limitations by developing a data set that allows for comparisons across longer periods than were previously feasible. The primary aim of the IFSS is to establish a harmonized set of data and documentation across ten nationally representative surveys of fertility and family. It is expected that harmonization of common variables across multiple surveys will allow researchers, policymakers, students, and other constituencies to make comparisons across time.
What is harmonization? Why harmonize variables?
Harmonization is a process by which variables are made comparable across survey years. Harmonization schemes must be developed for each IFSS variable so that comparisons can be made across time. In general, harmonization in the IFSS project involves combining, into a single variable, information covering comparable substantive ground but from different files in the original GAF, NFS, and NSFG data sets.
Differences in question text, sample design, and respondent universes further complicate the harmonization process. Challenges posed by this part of harmonization include varying universes by survey (and survey year) and question text. For example, the IFSS data set will contain a variable for ever having used contraception. In the 1970 National Fertility Survey, the universe of respondents of whom the question is asked contains all respondents in the sample. However, in the 1965 National Fertility Survey, the universe is smaller; it contains only respondents who were not pregnant at the time of the survey interview. For other variables, differences in universe are not observed. For example, respondent's age is asked of all respondents in all ten component surveys; therefore, respondent's age poses no universe problems to those using the IFSS data set. Given the subtle variation in variables across surveys, the IFSS staff must exercise care in evaluating variable comparability.
To assist users of the IFSS data, IFSS staff are developing comprehensive variable documentation that will address important comparability notes across variables and studies. Especially serious, for example, are comparability problems that are not evident from the coding structure, including alterations in the survey question wording and changes in the variable universe. Such documentation will include notes about changes in respondent universe, question design, data collection instrument design, and other information necessary for informed use of IFSS data. The IFSS project staff will seek to make transparent all decisions made in the harmonization process so researchers can choose whether to use the harmonized version or to create one of their own.
How are variables selected to be harmonized? Are all variables harmonized across all data sets?
With consultation from the IFSS advisory panel -- composed of distinguished experts in fields as diverse as demography, economics, public health, survey methodology, and sociology -- variables selected for inclusion in the IFSS data set will be identified on the basis of their expected interest to social science researchers, graduate students, policymakers, and other constituencies. Not all variables in all surveys will be harmonized. In most cases, variables selected for inclusion in the IFSS data set will have comparable equivalents across three or more surveys.