Towards these ends, recent efforts have focused on validating the consistency of reporting over time for each individual in the study as well as between mothers and children. Initial work checked the reporting patterns of mothers against that of their children, and compared the patterns with reports of the children's status early on in the study. For example, if a mother reported that the study child had been given up for adoption, then there should not be any interviews for the study child at later times. In cases where interviews were discovered, the actual interviews were pulled and checked for mis-reporting of the child's status or interview notes that a child other than the study-child had been interviewed. Such consistency checks were made for reports of the mother's or child's death at any point in the study. To clarify all possible situations, a review of reasons for dropping out was completed for the mothers and is still in progress for the children. At this point, the current status of each mother or reason for dropping out (e.g. the woman died, was too ill to answer the survey, was in jail, moved, etc.) has been determined for more than 80% of the original sample. Such validations of the data are time-consuming as they require sophisticated programming for comparing the electronic data sets as well as chart reviews of the original interviews to on-cover data entry errors (which are rare) and to identify potential cases mis-reporting.

To produce consistent life course trajectories for the different dimensions described above, it is necessary to first check for consistent reporting of various events over time and then to identify the potential components of each dimension at each time period. The type of work required at this stage can be described using education as an example. Consistency of reporting is simply examined by how the woman reports her highest grade attained at each point and then correcting, where possible, for obvious errors (either data entry or mis-reporting). If a woman reports having reached the 10th grade at Time 1, the 11th grade at Time 2, and the 10th grade at Times 3 and 4, then it is clear that the report at Time 2 was incorrect. Whenever such inconsistencies are encountered, the respondent's entire educational history is examined from the electronic data and compared to the actual reporting on the interview sheets. Both sources of information are used to determine the most reasonable correction to be made to the data.

Identifying potential components of the different dimensions at each time period requires a comparison of all questions asked over the entire period of the study. The first phase of this task was accomplished by the creation of a spread sheet, organized by the different life dimensions represented in the various surveys (e.g. education, marriage, work, health, contraceptive behavior, etc.) and detailing the specific variable names and question numbers for the "like" concepts over the 7 surveys. The second phase of this task, which was only recently initiated, involves an examination of the correlations among similar items within as well as over surveys, re-coding of questions to increase similarity across surveys and the exploration of potential scales for creating a single dimension from varying questions at the different times.

At a more macro level, the Baltimore data has been restructured into a relational data base which makes the relationships between the mothers, their children, and their mothers (the mothers' mothers) more readily identifiable. The new data structure will greatly enhance the ease of performing multiple analyses on the Baltimore data. It will also make the data more "user friendly" for collaborators who are unfamiliar with the data.

-Kathy Foley