Directorates
·	Director General's Office
·	Data Development
·	Strategic Evaluations
·	Partnership and Continuous Evaluations
Publications
·	Evaluation Reports
·	Evaluation Briefs
·	Labour Market Development Agreement Series
·	Lessons Learned Series
·	Evaluation Tool Kit Series
·	Evaluation Forum Newsletter
·	From Theory To Practice

A critical step in a summative evaluation is to select the best measures for assessing outcomes. "An irrelevant or unreliable measure can completely undermine the worth of an impact assessment by producing misleading estimates" (Rossi and Freeman, 1993, p. 234).

Outcome measures relate to the impact the program is supposed to have, so suitable outcome measures focus on the program's objectives. In general, HRDC training programs have all or a subset of the following goals: higher educational achievement; improved transition to the labour market with enhanced employability and earnings; reduced dependency on passive income support; and improved work attitudes. These outcomes are easily quantified and would be treated as dependent variables in the econometric models (since it makes sense to assess the program in terms of its intended effects). Thus, good outcome (post-program) variables for training programs include education level, employment status (i.e., working or not), time spent working or in school, annual earnings, months spent on social assistance, weeks spent on EI, and attitudes towards work, education and passive assistance.

It is important to note that pre-program measures of the outcome are highly desirable for any quasi-experimental evaluation. Very occasionally, the program's management information system will have adequate pre-program data for participants and non-participants. For example, HRDC has complete and accurate data on EI use that are very useful for impact evaluations. HRDC's Service and Outcome Measurement System (SOMS) also has good longitudinal data on employment status, and time spent in school or work. Unfortunately, reliable pre-program data on social assistance use and attitudes are very rare. Nevertheless, this information should be gathered if possible to conduct an impact evaluation. Some pre-program information can be gathered during a follow-up survey, although problems of accurate recall and lost records render the data imprecise. Other pre-program information, especially attitudes, is impossible to reconstruct after the program.

Collecting pre-program data on the sample through a "baseline survey" is the best recourse. The fundamental purpose of a baseline survey is to establish the pre-program characteristics of participants and non-participants in support of a future summative evaluation. A good baseline survey would feature questions that aim to establish what the person was doing in terms of the outcome variables (work, school, earnings, social assistance, EI, and so on) before the program. It should explore pre-program events each year for at least two years before the program began: for example, the number of months on social assistance during 1995, 1996 and 1997. A good section on attitudes, and a section on demographics (especially if the administrative system is unreliable or non-existent) should also be included in the questionnaire. Finally, the baseline survey should ask for at least two contacts (family or friends) who can help locate the individual for follow-up purposes, because the target group for training programs tends to be very mobile.