This section of the report presents the methodology for the Formative Evaluation of the Provincial Benefits and Provincial Measures under the Canada/Manitoba LMDA. For a more detailed description of the methodology, see the Technical Report under separate cover. The final design of the evaluation consisted of 28 key informant interviews, 12 focus groups, a survey of 1,400 program participants (1,100 PBPM participants and 150 participants in each of the PBT and POT programs)16 and a survey of 500 non-participants. The following is a description of the methodology used for each component of the data collection process. 2.1 Key Informant InterviewsA series of 28 key informant interviews was conducted for the formative evaluation. The list of potential key informants was prepared by the Joint Evaluation and Accountability Committee (JEAC). The interview respondents included a variety of program managers and policy makers who collectively were in a position to comment on the design, implementation and delivery of the PBPMs that were assessed in this evaluation. The key informants included the following:
2.2 Focus GroupsA total of 12 focus group discussions were conducted. Focus groups had an average of eight to 10 participants each and were conducted in each of the three service regions of the province: Winnipeg; Southern Manitoba (Brandon); and Northern Manitoba (Thompson). There were five types of participants: clients; staff; third-party delivery agents; employers; and community partners (including employee groups, employer groups, sectoral associations and others). The distribution of the 12 focus groups is presented in Table 2.1. For recruiting clients, program participants were selected from the sample for the survey of participants (representing a variety of PBPMs). For employers, community partners and delivery agents, the focus group recruitment relied on lists provided by the JEAC.
2.3 Survey of Participants and Non-ParticipantsSurvey of Participants: PopulationThe final participant data set included only those participants who started a PBPM intervention on or after November 27, 1997 and completed/terminated before August 15, 1999. The length of time since completing or terminating the intervention was a key variable for stratifying the sample in the final analyses as the length of time since completion is related to employability outcomes. Survey of Participants: Data SamplingThe data files were originally developed to include participants who participated in LMDA employment programs and services at any time between November 27, 1997 and August 15, 1999. As mentioned, these files were aggregated, yielding a single data file containing information for 19,252 participant cases, with the individual client as the unit of analysis. This was not equal to the sum of all the cases from the administration and data files because clients that had taken part in more than one intervention could appear in more than one file. Following the removal of all cases without valid phone numbers, start and end dates for EI benefits, and start and end dates for most recent interventions, the final data file consisted of 14,784 individuals. Three primary sampling variables were taken into account when developing the survey sample frame. These variables were program type (EAS17, Wage Subsidies, Self-Employment, Employment Partnerships, SLG/Enhanced Fee Payer, Purchase of Training, Apprenticeship and Project-Based Training), region (Winnipeg, Southern Manitoba and Northern Manitoba), and claimant status (active EI claimant, near reachback, mid reachback and far reachback).18 Table 2.2 presents the sampling plan for the participant survey (in the column labelled "quota"), as well as the number of interviews that were actually completed (in bold) with participants who differed according to these three variables. The targets for individual cells were adjusted based on the distribution of participants by geographical area, program and claim status, thus attempting to ensure that a minimal sample size (e.g., 40 cases) was available to permit segmented analysis based on any two of the variables simultaneously. Given the size of the final sample (i.e., intended to be 1,400 interviews), segmenting the sample on the basis of all three variables simultaneously for the purpose of analysis would require that the sample be divided into 96 cells, reflecting the product of the number of categories in each of the sampling variables (i.e., eight programs multiplied by three regions and four claim status categories). For some cells, however, the available sample was not sufficient to support a quota of 40 interviews for the analysis of the data by two of the sampling variables (i.e., Employment Partnerships in Northern Manitoba). Thus for some programs, additional interviews were completed in other regions that had sufficient sample (i.e., Winnipeg) and, if no additional sample was available within the same region, attempts were made to complete additional interviews within the same program and claimant status category. A total of 1,393 interviews were completed. The distribution of sample drawn to obtain the desired number of survey completions is presented with response rate information in Table 2.2. Survey of Participants: Response RateThe response rates and refusal rates for participants in each type of program are presented in Table 2.3. The response rate is the proportion of cases from the functional sample who responded to the survey, while the refusal rate represents the proportion of cases from the functional sample19 who declined to participate in the survey.
The response rate for the survey was lower than expected, ranging from 31.7 per cent among SLG/Enhanced Fee Payer participants to 55.2 per cent for Purchase of Training respondents, with an overall response rate of 39.2 per cent. The overall refusal rate was also higher than expected (13.3 per cent) and ranged from 9.1 per cent for Apprentices to 20.7 per cent for Wage Subsidies participants. The sampling error for the survey is ±2.5 per cent. That is, the overall survey results are accurate within ±2.5 percentage points, 19 times out of 20. The margin of error is higher for the sub-group analyses. Survey of Participants: Weighting of DataThe grouping of weighting variables was dependent on the type of analysis. For the overall analyses, the participant data were weighted according to program type, region, claimant status and age. For analyses by program type, the data were weighted by region, claimant status and age. For analyses by claimant status (i.e., active EI claimant versus reachback clients), the data were weighted by program type, region and age. Comparison Group Survey: Sample Frame CreationThe comparison group case file was drawn from a file of EI claims that were active at some time between 1994 and 1998. The administrative files included T1, Status Vector, ROE, NESS, and BNOP files. This dataset of comparison group members was then linked to the administrative data files, yielding a single file of 204,531 active EI claimants and reachbacks from which to draw the comparison group sample. The comparison group active EI claimant and reachback samples were developed separately. To sample comparison group active EI claimants, the comparison group data file was matched to the participant data file based on the time periods for which members of the comparison group were receiving EI. To accomplish this, three time periods were defined according to observed values for program end dates in the population of active EI claimant participants. The comparison group active EI claimant sample was drawn in the same proportions as were observed for active EI claimants in each of the three time period cohorts in the participant population. To develop a comparison group sample for reachbacks, the proportion of different reachbacks (i.e., near, mid and far reachbacks) in each of the three time cohorts was determined for the participant population. Comparison group reachbacks (i.e., they were not on claim during any of the three theoretical reference dates) were then randomly assigned to one of the three time cohorts. Reachback status (i.e., near, mid, far) was determined based on the length of time since the end-date of their most recent EI claim period and the theoretical reference date (i.e., mid-point of the time cohort). Near, mid and far reachback comparison group cases were then sampled such that they reflected the proportion of these cases in the three time period cohorts that were observed for the participant population. To correct for the fact that a comparison group case could possibly fall into any of the three time cohorts, each time period cohort was sampled separately. The final comparison group sample frame consisted of 2,633 cases in three mutually exclusive time period cohorts from a population of 204,531. Comparison Group Survey: Response RateThe response rate for the survey is presented in Table 2.4. The response rate is the proportion of cases from the functional sample who responded to the survey. Conversely, the refusal rate represents the proportion of cases from the functional sample who declined to participate in the survey. The response rate for the survey was 34 per cent20 and the refusal rate was 26.4 per cent. The sampling error is ±4.4 per cent. That is, the overall survey results are accurate within ±4.4 percentage points, 19 times out of 20. Comparison group survey results presented by claimant status (i.e., active EI claimants versus reachback) were weighted by region and gender.
|