Each component of a merit-pay plan creates incentives that will guide teacher behavior. The recommendations made here reflect research findings on merit-pay programs:[*]

  • Effective plans are noncompetitive for teachers. All teachers need to have the ability to earn performance awards if they meet performance criteria. Awarding only the top 10 percent of teachers, for instance, can create perverse incentives.

  • Effective plans judge teacher quality by student achievement growth. Our recommended plan measures student growth by evaluating the differences between student performance at the beginning and end of a single school year. Teachers are judged by their students’ learning progress, and they receive rewards whenever their students exceed the gains predicted statistically by the students’ prior rate of learning and other factors.[†]

    Such "value-added" statistical calculations of teacher effectiveness are already used in the Amphitheater Public Schools’ "Project EXCELL!" in Arizona.[‡] As a manual for the Arizona program notes in describing the fairness of the value-added method, "Value added data analysis helps to hold professionals accountable for the value added to student growth, without penalizing or rewarding for pre-existing differences of the students in the classroom/school."[35]

  • Effective plans allow a wide range of school personnel to participate.In our recommended plan, all instructional staff members, from teacher aides to principals, will earn rewards if students exceed projected growth targets. This inclusiveness promotes collaboration and a team approach to maximizing student learning.[§]

  • Effective plans offer significant awards. Our recommended plan offers bonuses for three different categories of professional personnel: "core" teachers, whose subjects are tested by the standardized exams and whose students are sufficient in number to allow meaningful measures of overall student gains; administrators, as well as "noncore" teachers, who either teach subjects untested by the standardized exams or who teach too few students to produce reliable aggregate measurements; and instructional support personnel, such as paraprofessionals and teacher aides, who help increase teachers’ gains in the classroom. There would be bonuses of up to $10,000 for core teachers, up to $7,000 for noncore teachers and administrators, and up to $3,000 for instructional support personnel.

  • Effective plans involve teachers in key decisions. Teachers must participate in the planning and design process for a plan to be effective. If they trust that their performance will be evaluated fairly, they are more likely to welcome the use of the complex statistical models that can better estimate their personal contributions to student learning.[**]

With these factors in mind, our recommended merit-pay bonus plan ties 80 percent of the educator’s award to value-added assessments of students’ achievement gains. For core teachers, 40 percent of the award would be tied to value-added calculations of their own contribution to their students’ achievement growth, and 40 percent would be tied to value-added calculations of the entire professional staff’s contribution to schoolwide achievement growth (a group measure that will provide core teachers with good reasons to collaborate with their colleagues). For noncore teachers and instructional support personnel, the entire 80 percent would be tied to value-added calculations of the staff’s contribution of schoolwide achievement growth. In each case, this heavy emphasis on student achievement would spur educators to focus their efforts on better classroom learning — the outcome that matters most.

The remaining 20 percent of educators’ awards would be based on supervisors’ evaluations. This component is meant as a safety valve to help correct any perceived flaws in the outcome of the value-added estimates. Although education research does suggest principals’ evaluations can be prone to certain subjective errors, the research also indicates that principals typically produce assessments that reflect a teacher’s ability to help students and satisfy parents.[36]

As noted earlier, our pilot plan offers different levels of maximum awards based on a teacher’s level of accountability. A higher possible award is available to core teachers, since their individual performance will face closer scrutiny and their students’ scores will directly impact the figure for schoolwide growth. In Graphic 1, the bonus plan is presented for the three categories of personnel.

Graphic 1: Payout Chart for Model Merit-Pay Bonus in Michigan

We recommend that student achievement growth be measured using the Northwest Evaluation Association evaluations, rather than the tests administered by the Michigan Educational Assessment Program. NWEA tests are used in the Arizona program mentioned earlier, and they are administered up to four times a year in math, reading, language usage and science. These computerized tests are "adaptive," meaning the questions get harder or easier based on a student’s previous answer until the test pinpoints the student’s exact achievement level.

By testing at both the beginning and the end of the school year, the tests identify a student’s starting point, his or her strengths and weaknesses, and his or her progress.[37] Reports on student progress are available within one to three days after each test administration.[38] The NWEA exams in Michigan are aligned with Michigan’s assessment system.[39]

Unlike the MEAP exams, which are given only in certain grades and only once in those grades, [40] NWEA testing can be used for students in grades K-10.[41] The company already has numerous customers in Michigan,[42] and the cost for a full year’s worth of testing is approximately $14 per student.

Obviously, different tests would be needed for 11th- and 12th-grade teachers. The Michigan Merit Exam (which is based in part on ACT scores) might be used in the 11th grade, with perhaps a second round of ACT scores or some other assessment used in the 12th. These and other details of our merit-pay program should be governed by the program administrators in consultation with a district advisory council that includes teacher representatives, district officials and a statistician skilled in testing and measurement. This advisory council can also recommend ongoing adjustments to the plan. Such committees exist in Denver, Chicago and many other plans.[43]

[*]Readers interested in "best practices" research should review the following: Solmon and Podgursky (2000); J.H. Barnett, "How Does Merit Pay Change Schools?" (2007). "A Review of the Research and Evaluation of the Impacts of the Achievement Challenge Pilot Project," unpublished doctoral dissertation, University of Arkansas; M.J. Podgursky and M.G. Springer (2006), "Teacher Performance Pay: A Review," available from the National Center on Performance Incentives at: www.performanceincentives.org; M.J. Holley, J.W. Ritter, and J.H. Barnett, (2007), "Merit Pay: Working Paper. Office for Education Policy," University of Arkansas, available at: www.uark.edu/ua/oep; H.G. Heneman, A. Milanowski, and S. Kimbrall (2007), "Teacher Performance Pay: Synthesis of Plans, Research, and Guidelines for Practice," Consortium for Policy Research in Education, available at: http://www.cpre.org/images/stories/cpre_pdfs/RB46.pdf; D. Goldhaber (2006), "Teacher Pay Reforms: The Political Implications of Recent Research," Center for American Progress, http://www.americanprogress.org/issues/2006/12/pdf/ teacher_pay_report.pdf; C.T. Clotfelter and H.F. Ladd (1996), "Recognizing and Rewarding Success in Public Schools," in Holding Schools Accountable: Performance-Based Reform in Education, Helen F. Ladd, editor, Brookings: Washington, D.C.

[†]This general principle still leaves details to be settled by those designing the program. For instance, teachers with multiple tested classes could be judged either on total average growth or on a weighted average of the classes’ results. These methods entail trade-offs, but choosing one through a collaborative process involving teachers, the district, foundation personnel and a statistical expert should not be prohibitively difficult.

[‡]For the Project EXCELL! manual, see http://www.amphi.com/ departments/teachlearn/projexcell/files/ 61CE0CCC74B94CDC85 C1A52CFC85497D.pdf.

[§]Some merit-pay plans include noninstructional personnel. We have chosen not to do so here.

[¶]A possible concern about this pay scale is that noncore teachers might feel frustrated by their inability to earn the core teachers’ larger bonuses. But noncore teachers will also face less scrutiny, since their students’ achievements in their subject area will not affect (and potentially reduce) their bonus. As the program progresses, the program’s administrators (advised by district officials and teachers’ representatives, as mentioned in the text) should monitor noncore teachers’ reactions to the bonus schedule. The program administrators may want to search for ways to increase the noncore teachers’ potential rewards by finding alternative tests of their students’ achievement. To avoid unfairness, these tests should be comparable to those faced by students of core teachers.

[**]The formula for compensating teachers involves a trade-off between transparency and complexity. A transparent model that uses straightforward arithmetic to calculate student achievement gains and the consequent teacher awards is more easily understood and verified, but is also more likely to overestimate or underestimate teachers’ contributions to student improvements. On the other hand, a "value-added" model like the one recommended above is better at isolating a teacher’s contribution from other factors affecting student achievement, but is harder to understand and verify. For a discussion of these trade-offs, see: "Roundtable Discussion on Value-Added Analysis of Student Achievement: A Summary of Findings" (The Working Group on Teacher Quality, 2007) http://www.talentedteachers.org/pubs/ value_added_roundtable_08.pdf (accessed June 28, 2008).