Wisconsin’s Differentiated Teacher Pay Improved Achievement and Recruitment

Caroline Hoxby argues that teacher compensation is fundamentally misaligned with the differences teachers make for students. Traditional pay systems generally reward years of service and credentials: teachers with the same experience and degree type receive the same salary, whether their measured value added falls near the top or bottom of the distribution.

That arrangement is difficult to reconcile with what economics of education has learned over roughly the past two decades: teachers differ substantially in their effects on student outcomes. The evidence is no longer confined to a small set of U.S. studies. Wherever researchers have sufficiently good education data, Hoxby says, they tend to find a similar result—large variation in teacher value added.

The important qualification is that conventional signals do not reveal that variation particularly well. A résumé, an esteemed teaching degree, or a particular college credential does not reliably identify an effective teacher. Some teachers without especially prestigious degrees are excellent, Hoxby says, while some with such degrees are not. The relevant question is not simply what qualifications a teacher brings to the classroom, but what changes for students after they have been taught by that person.

The policy decision is not only whether to measure teacher effects. It is whether districts are willing to use credible measures, alongside school-level judgment, to depart from salary schedules based principally on seniority and credentials.

The basic challenge in measuring a teacher’s contribution is that students do not arrive in classrooms with identical prior achievement, circumstances, or prospects. Caroline Hoxby frames the ideal method as a medical experiment. If students could be randomly assigned year after year to one teacher or another—her hypothetical teachers “Smith” and “Jones”—researchers could compare later outcomes and attribute systematic differences to the teachers.

Such randomized studies have been conducted, including a large randomized controlled trial in Los Angeles carried out by Mathematica, as well as trials funded by the Gates Foundation and by individual states. But assigning students experimentally to teachers on a continual basis is impractical as a general operating system for schools.

The alternative is to use observational data in a way that approximates experimental results. Researchers can estimate teacher effects by regressing students’ end-of-year achievement or other outcomes on a teacher-specific effect while controlling for students’ incoming achievement, grade level, testing year, and other observable factors. Incoming achievement matters because otherwise a teacher who receives students with lower expected achievement could be wrongly judged against a teacher assigned students with higher expected achievement.

According to Hoxby, these regressions produce extremely similar results to randomized controlled trials. That makes it possible to estimate teacher value added across many systems rather than only in occasional experiments.

Test scores are the most common outcome because they are often available, but they are not the conceptual boundary of value added. Hoxby lists a much broader set of measures: performance on Advanced Placement or International Baccalaureate tests; achievement in science projects, mathematics challenges, debate, writing, and arts competitions; attendance and disciplinary outcomes; college entry, persistence, and timely degree completion; and longer-run outcomes including employment, earnings, and family stability.

The further an outcome sits from a teacher’s classroom, Hoxby says, the harder it becomes to attribute an independent effect to that teacher: the student has accumulated exposure to more teachers along with other influences. She nevertheless describes the problem as statistically difficult rather than impossible.

Teacher effects are not uniform across outcomes or students. Caroline Hoxby explains that one teacher may be especially effective at raising measured academic achievement, while another may be especially effective at improving attendance and discipline. Those dimensions tend to be correlated, but they are not identical. Likewise, some teachers may be better suited to students who enter with lower achievement or cognitive difficulties, while others may be more effective with students of high cognitive ability.

That heterogeneity matters because a teacher’s impact compounds over time. Hoxby illustrates the point with a stylized comparison of students repeatedly assigned to low-, average-, and high-value-added teachers from first through 12th grade. Few students experience either extreme sequence, she acknowledges. But in her illustration, a student who repeatedly receives high-value-added teachers rises from a standardized score of about 4 to about 15, while a student repeatedly assigned low-value-added teachers rises from about 4 to about 7.

The effects are not confined to school performance. In the longer-run evidence Hoxby presents, teacher quality is associated with differences in college attendance, enrollment in higher-quality colleges, earnings at age 30, and earnings growth between ages 22 and 28.

She gives several other figures from those longer-run outcomes: about a 13% difference in attending college at age 20, an 18% difference in attending a high-quality college at that age, and substantially higher earnings growth between ages 22 and 28. Strong measures of short-term teacher value added, she argues, can forecast effects on these longer-term outcomes fairly well.

Formal statistical estimates are not the only way people recognize the differences. Principals or headmasters are often quite good at knowing a teacher’s value added even without being presented with the data, Hoxby says. Fellow teachers are also fairly good judges, as are students. That observation supports her practical premise that schools may have useful local knowledge about who is making a difference.

Caroline Hoxby uses a deliberately radical calculation to show how far prevailing salaries are from the economic value she attributes to strong teaching. Her starting assumptions are that each unit of teacher value added generates 9.2% higher earnings at age 30, that people work from age 25 through 65, and that each teacher instructs 20 students per year.

If teachers received the full additional lifetime earnings their value added generated, Hoxby calculates, the annual difference in pay between a teacher at the low end and one at the high end of her illustrative distribution would be about $1.6 million. She does not present that as a likely policy. Rather, it establishes the magnitude of the value being created under her assumptions.

Even a much smaller claim on that value would produce a large difference in compensation. Hoxby describes a 5% “equity stake” as analogous to teachers owning a small share in the gains generated for their students. On that basis, a teacher at the 90th percentile of value added would earn roughly $79,000 more per year than a teacher at the 10th percentile.

The analogy is a way to ask what compensation would look like if it reflected even a small share of the measured long-term value teachers create. Existing systems, by contrast, generally offer no salary distinction at all between a 90th-percentile and 10th-percentile teacher with the same tenure and credentials.

Hoxby calls this conventional arrangement “lock-step” pay, also sometimes “staircase” pay. Advancement occurs as teachers accumulate years of experience or additional credentials, not as their contribution to student outcomes is assessed to be higher. Moving away from it would require both a basis for distinguishing value added and discretion for districts to act on that distinction.

The practical question is whether schools can move away from lock-step compensation. Caroline Hoxby points to Wisconsin’s Act 10, which allowed districts to differentiate teacher pay from the conventional salary schedule.

Hoxby cites multiple studies of schools that chose to use differentiated pay. Her account addresses three distinct policy questions: whether the approach affects student outcomes, whether it changes the teacher labor market, and whether it can retain local political support.

The recruitment result is central to the economic argument. Compensation does not simply reward teachers already in place; it can affect where effective teachers choose to work, particularly earlier in their careers.

The Wisconsin example is narrower than the equity-stake calculation, but it supplies Hoxby’s practical case for differentiation. In her account, districts that used their latitude to differentiate compensation saw improved achievement, drew in effective teachers, and retained local support. She also cites charter schools and Teach for America as additional areas with relevant evidence, though she does not develop those examples here.