* This paper is based upon the author’s dissertation research, under the direction of Dr. Donald F. Kettl

Performance-based contracting (PBC) is becoming increasingly attractive to state human service agencies. PBC incorporates performance measures and makes contract payment, extension, and renewal (fully or partially) contingent on successfully achieving them. Under PBC, public managers only specify desired results in advance, leaving contractors considerable discretion to prescribe service delivery method and use funds. These efforts are changing the landscape for public funding of social programs. Traditionally, human service contracts are run on a fee-for-service (FFS) basis, in which contractors are closely monitored in delivering the services prescribed by public managers and reimbursed by unit of service incurred, regardless of service outcome. In contrast, PBC, making contractors accountable for their final outputs, is expected to encourage higher level of performance. 

Despite the promising benefits, to date there has been little systematic study on how PBC affects service performance. Actually, human service is not always considered compatible with performance measure. Also, PBC creates high risks for both governments and contractors. PBC brings high fiscal risks for contractors’ serving difficult clients and less control for government to oversee and monitor the appropriateness and quality of services, leading to “creaming” and “gaming” behaviors.

In this study, using a quasi-experimental approach, we evaluates the effectiveness of PBC, by comparing the performance of state vocational rehabilitation programs in Indiana (started using PBC in 2006) with those in Michigan (still using FFS) over a period from 2003 to 2009. Specific indicators include average time to placement, rate of placement, length of job retention, employment wage rate, and cost per placement outcome. Because of the nonrandom assignment, the selection bias may be substantial. Thus, propensity score matching is used as a “correction strategy” that corrects for the confounding effects of variables other than the intervention. Particularly, individual samples in two states are matched based on demographic factors, work disincentive, and type of vocational rehabilitation services received. The data are obtained from the RSA 911 database of the Rehabilitation Service Administration, which reports records pertain to all individuals whose case records were closed in a given fiscal year.

The finding shows a mixed result. Generally, PBC performs better in placement rate, time to placement, and cost per placement. However, the differences in the areas that are not targeted by PBC, including wage and job retention, are not significant.