The utility of a quasi-experimental evaluation design for estimating benefits derived from a program to match job-seekers to openings listed by employers was examined in three studies of Job Service referrals and placements in Washington and Oregon. Data examined included a mail survey of a sample of 587 Washington residents referred to job openings in the first half of 1998 and records of 328,815 spells of unemployment experienced by unemployment insurance (UI) claimants in 1987-1995. The effectiveness of job service placement services and referral services were estimated using program cost-benefit ratios, duration of joblessness, success in obtaining interviews, and successful job placement for job-seekers with strong and spotty work records. An expert panel examining the methodology concluded that: 1) survey results were imperfect because of the possibility that some job-seekers who failed to secure interviews were screened out by employers; 2) the sample was small and may not have been representative; 3) improving the survey and using telephone follow-up should overcome the problems; and 4) the design holds promise for obtaining credible estimates of the value of being placed that also can be used as benchmarks in assess the bias in non-experimental estimators. (Contains 11 tables.) (MN/CG)