The RAPIDS data provide insight into many characteristics of apprentices, apprenticeships, and sponsoring programs. Most importantly, given they include the universe of apprentices and programs, counts of each provide a broad measure of entrants at any time or the total number of apprentices or programs over time. As such, these data can be used for program reporting, for monitoring overall enrollment trends, and for showing a range of worker and program characteristics. For example, data on the variety of sources of related instruction (e.g., community college, correspondence course) reveal several ways programs interface with the community when training apprentices.
Of special interest to this project is the use of the data by the RTI/Urban team to help identify programs for site visits and for thinking about what is possible for a large-scale evaluation. Specifically, the list of programs by region and counts of currently enrolled apprentices by major occupational group can be used for helping select sites to visit in the current project. After identifying a subset of possible programs to visit, the RAPIDS data could be used to narrow the subset according to other characteristics in the data. For example, decisions about choices of site visits will be informed by data on programs that have several apprentices in the advanced or specialty occupations or programs that have both a large proportion of cancellations and a large number of currently enrolled apprentices.
Because the RAPIDS data are usually entered by sponsors and not through a centralized data entry function, much of the RAPIDS data lack certain characteristics for use in quantitative analysis for a future evaluation. First, there are large amounts of missing data in many data fields. Second, timeliness of data entry or updates may vary across employers, which would be important for data that is particularly time dependent (e.g., wages, enrollment status). Third, some problems encountered by employers interviewed for this report may prevent employers generally from entering completely reliable data.
Fourth, while some data probably have little or no measurement error in terms of data entry accuracy because they change little over time or are observed without error (e.g., demographic characteristics), other data had obvious problems due to either data entry error or variability in coding the same data fields by different programs. For example, according to DOL staff, the starting wage could be entered as either the actual beginning wage of an apprentice or the wage entered on the apprentice agreement, which may or may not be the actual wage. As a result, our preliminary analysis of the wage data showed that the pre-apprentice wage may be larger than the starting wage, and the ending wage may be less than the starting wage. Because of these issues, for quantitative data that may have a narrow range or are expected to change but only by small amounts over time (e.g., wages), a future evaluation contractor would probably need to collect such data in a centralized fashion and employ a range of data entry quality control measures to reduce potential measurement error.
Fifth, the data lack certain important measures of the effects of apprenticeship over time. For example, because RAPIDS includes no data on whether an apprentice is still employed at the same or another employer, one cannot estimate placement and retention rates. Such data could potentially be acquired from unemployment insurance data records. In addition, the system lacks data on employer costs and benefits because the OA does not have the authority to collect it. Finally, the RAPIDS system does not include data on one of largest future sources of apprenticeships, the expected large-scale implementation of apprenticeships in Washington State. Given the size and scope of that future project (and potential other large projects in various states), any future evaluation will have to obtain data that goes beyond the current RAPIDS database.
The RAPIDS data are useful for monitoring enrollment trends and understanding various characteristics of apprentices and programs. But, additional data sources designed explicitly for any future evaluation effort will be needed to support quantitative analysis, especially key outcomes such as turnover and program costs. Further, at their current scale, most LTC RAP programs are too small and too dispersed geographically to support detailed analyses and separate evaluations. Any future evaluation will depend on having a large enough number of apprentices participating in apprenticeships over time to provide a large enough sample in each occupation for evaluation. As of December 2009, only 1,444 apprentices were currently participating in training (registered), with only approximately 300-600 apprentices in each major occupation, except for HHAs, of which there were only approximately 100 currently registered apprentices.