ERIC EJ795876: Using Performance-Based Pay to Improve the Quality of Teachers

2007·0.12 MB·English

by  ERIC

05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 87
Using Performance-Based Pay
to Improve the Quality of Teachers
Victor Lavy
Summary
Tying teachers’ pay to their classroom performance should, says Victor Lavy, improve the cur-
rent educational system both by clarifying teaching goals and by attracting and retaining the
most productive teachers. But implementing pay for performance poses many practical chal-
lenges, because measuring individual teachers’ performance is difficult.
Lavy reviews evidence on individual and school-based incentive programs implemented in re-
cent years both in the United States and abroad. Lavy himself evaluated two carefully designed
programs in Israel and found significant gains in student and teacher performance. He observes
that research evidence suggests, although not conclusively, that pay-for-performance incentives
can improve teachers’ performance, although they can also lead to unintended and undesired
consequences, such as teachers’ directing their efforts exclusively to rewarded activities.
Lavy also offers general guidelines for designing effective programs. He emphasizes that the sys-
tem must measure true performance in a way that minimizes random variation as well as unde-
sired and unintended consequences. It must align performance with ultimate outcomes and
must be monitored closely to discourage gaming if not outright fraud in measured output. Goals
should be attainable. Incentives should balance individual rewards with school incentives, fos-
tering a cooperative culture but not at the expense of free riding. All teachers should be eligible
for the incentive offered, but only a subset of teachers should be rewarded in practice. If too
many teachers are rewarded, teachers may not need to exert much extra effort to benefit.
Many of the practical challenges faced by performance-related pay, Lavy says, can be ad-
dressed through careful design of the system. He emphasizes that setting up a performance-re-
lated pay system that works is not a one-time task. Even with the best preparation, initial imple-
mentation is likely to be problematic. But if the effort is seen as ongoing, it should be possible
to make progress gradually in developing incentives that motivate the desired teaching behav-
iors and that will be perceived by teachers as fair and accurate.
www.futureofchildren.org
Victor Lavy is professor of economics at the Hebrew University of Jerusalem and a research associate at the National Bureau of Economc
Research and the Centre for Economic Policy Research.
VOL. 17 / NO. 1 / SPRING 2007 87
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 88
Victor Lavy
I
nterest in improving public education to practicing teachers? How will incentives
is growing not only in the United affect the composition of applicants to
States but worldwide. One reason for teacher-training institutions and to teaching
the heightened public attention is the positions in the schools? What policy meas-
key role played by education in deter- ures can remedy existing distortions in teach-
mining both individual earnings and broader ers’ compensation? My intent is not to review
economic growth. Another is widespread dis- exhaustively what is known about perform-
satisfaction with the education sector’s per- ance-based pay in education, but rather to
formance of late: substantial increases in summarize selected key findings, highlight
spending on public schools have failed to some guidelines for designing effective
bring corresponding increases in student teacher incentives schemes, and identify
achievement.1 areas requiring additional evidence. My ob-
jectives are to present the theoretical bene-
The quest to improve public education has fits of performance-based pay as well as some
led policymakers and researchers to focus on of the practical obstacles to its effective im-
how to increase teachers’ effectiveness. One plementation, to review critically the empiri-
obvious means is compensation. According to cal evidence, and to draw policy conclusions.
many observers, the traditional basis for
teacher pay—years of service and educa- Teachers’ Compensation
tion—provides little incentive for excellence. Pay for performance is meant to solve the
To make teachers more effective, these crit- twofold problem of motivating high teacher
ics argue, pay should be tied to performance. performance while attracting and retaining
And some school districts, here and abroad, good teachers under conditions where their
are undertaking reforms to test those ideas. effort or ability is not readily measured or ob-
In November 2005, for example, Denver vot- served. In the teaching profession, earnings
ers approved a $25 million tax increase to are based primarily on input (that is, skills
fund a form of “merit pay” to reward elemen- and time worked), rather than on output.
tary and secondary school teachers along a Such a basis, critics say, is not “results-
variety of dimensions, including their own oriented.” Moving to an earnings structure
demonstrated knowledge and skills and stu- that ties pay—at least partially—to some per-
dent academic growth. Whether Denver’s formance indicators should thus improve the
new merit pay system will improve student current system. In theory, the idea makes
achievement remains uncertain; an earlier good sense. But implementing pay for per-
pilot study in Denver found mixed results.2 formance poses many practical challenges. In
the teaching profession effort and output are
In this article I examine academic and policy difficult to define and measure because the
analysis of performance-based reward pro- work is generally complex, unique, and often
grams for primary and secondary school results from team efforts, with any one
teachers. I stress, in particular, several ques- teacher’s effort difficult to disentangle from
tions. What are the pros and cons of imple- that of the others on the team. One key goal
menting teachers’ pay incentives in schools? of education is to give students the skills
What criteria are to be applied in designing needed to ensure a productive career and
optimal teacher incentives? How much is sustain their economic well-being. Yet, be-
performance affected by incentives offered cause it takes years for the adult earnings of a
88 THE FUTURE OF CHILDREN
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 89
Using Performance-Based Pay to Improve the Quality of Teachers
student to materialize, it is impossible to tie a that is equal for all winners, or it can vary and
teacher’s wages to his students’ earnings. Re- increase with the winner’s level of achieve-
cent studies do in fact suggest that students’ ment. The total amount of awards may be
test scores are strongly correlected with their predetermined (for example, only a certain
future earnings, but using test scores to number of teachers can win an award) or it
measure a teacher’s performance presents may be open. The performance criteria can
practical problems.3Peer or principal evalua- include outcomes for the teachers them-
tions are yet another way to measure per- selves, such as measures of absenteeism or
formance, given the drawbacks of testing and
the many teachers for whom testing would
not apply. But these approaches also have Pay based on performance
drawbacks.
usually involves some
Performance-Based Pay objective assessment of
Pay based on performance usually involves
schools’ or teachers’ efforts or
some objective assessment of schools’ or
teachers’ efforts or success or some measure success or some measure of
of their students’ performance. Performance-
their students’ performance.
based pay schemes have many variable fea-
tures. They can compensate teachers only for
their own performance or they can be struc-
tured as a team incentive program, with performance on a test. They can also include
group performance determining the total in- measures of the teacher’s students’ perform-
centive payment, which is then divided ance, such as attendance, grade retention,
among team members regardless of individ- dropout rates, or performance on tests.
ual performance. The group can include all These criteria are not mutually exclusive.
of the school’s teachers or a subgroup, such
as the teachers of a given grade or a specific The target set for determining award winners
subject. Performance-based pay schemes is critically important both for efficiency and
can, but need not, involve sanctions for for equity. For example, if schools are ranked
below-threshold performance. Although according to how many students attain a cer-
monetary rewards are the most common in- tain level of literacy as determined by an ex-
centive in performance-related pay, other in- amination, gains near the cutoff are most re-
centives can include reduced teaching load, warded. But if schools or teachers are
promotion, and public recognition of out- rewarded on the basis of average test scores
standing teachers. The reward can be just a or changes in those averages, then credit is
one-time event or it can be ongoing, leading given for gains at all parts of the achievement
to a permanent salary increase. It can be distribution, not just those close to the cutoff
based on a relative criterion (for example, the point. For example, students in the bottom 5
average test score gain of a teacher’s class rel- percent of the achievement distribution
ative to the classes of other teachers) or on an might be too far away from the literacy stan-
absolute criterion (such as the class average dard to pass the test after one or two years,
test score being higher than a predetermined but raising their test scores might be worth-
threshold). The reward may be a fixed sum while nonetheless. A hybrid measure could
VOL. 17 / NO. 1 / SPRING 2007 89
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 90
Victor Lavy
target both some cutoff as well as some aver- the basis of an agreed metric aligns incen-
age. It is also possible to design a scheme that tives directed at teachers or schools with
differentially weights improved test scores at those directed at students and potentially the
different points of the achievement distribu- entire society. If wages are based on student
tion, with the bottom weighted more than performance, for example, they provide
the top, the middle more than either ex- teachers or schools with powerful signals
treme, and so forth. Finally, an important, about what is valued and what is not. Absent
but potentially underappreciated, part of a such signals, even well-meaning teachers may
pay-for-performance plan is the identity of emphasize material that is obsolete or gener-
those who evaluate the teachers. The evalua- ally no longer valued by parents or the labor
tors can be external to the school or can be market. Similarly, if wages are based not only
peers, principals, or district supervisors. on the individual benefits of schooling to stu-
dents (social scientists call these “private re-
Despite the almost innumerable combina- turns”) but on the benefits to society as a
tions offered by these options, three proto- whole (“social returns”), teachers or schools
types of performance-based reward programs would take into account the social returns to
are most often implemented in education sys- education when making choices about their
tems and are commonly examined by re- work. A student, for example, may want to
searchers. The first model, merit pay, gener- drop out before completing high school be-
ally involves individual incentives based on cause he feels that the costs of staying in
student performance. The second, knowl- school outweigh his individual benefits. A
edge- and skill-basedcompensation, generally teacher considering only those individual
involves individual incentives based on benefits may not work as hard to discourage
teacher skills. Knowledge- and skill-based pay him from dropping out as would a teacher
differs from merit pay because it provides considering the costs and benefits to society
clear guidelines on what is being evaluated. as a whole.
The third model, school-based compensation,
generally involves schoolwide incentives, typi- Individual performance-based pay schemes
cally based on student performance. improve efficiency by helping correct distor-
tions in a teacher’s effort that might result
Potential Benefits of from gaps between her preferences and
Performance-Based Pay those of her students. For example, a teacher
Performance-based pay in education brings might fail to assign homework even though
with it many potential benefits but also many she knows its value for her students because
challenges. This section and the next present correcting and grading assignments involves
the main issues. more work for herself. Individual perform-
ance-based pay provides some incentive for
Productivity and Efficiency Considerations. the teacher to do the “right thing.”
In theory, pay based on output has two ad-
vantages over input-based pay in terms of ef- The second efficiency advantage of output-
ficiency (that is, producing “more” education based pay, mainly relevant for the merit pay
for the same cost). The first, most frequently model, involves sorting and selection. Assum-
noted efficiency advantage has to do with in- ing that the compensation system accurately
centives. Rewarding teachers or schools on identifies productivity, basing pay on per-
90 THE FUTURE OF CHILDREN
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 91
Using Performance-Based Pay to Improve the Quality of Teachers
formance will attract and retain the most pro- formance-based pay can thus make compen-
ductive teachers. Even if teachers are unable sation systems more equitable.
to alter their own behavior to enhance per-
formance, as measured, say, by students’ test Finally, performance-based pay may increase
scores, some people are still inherently better support for public education from politicians
than others at affecting test scores. Basing and members of the public who are con-
pay on output also tends to discourage teach- vinced that the reform will reverse the edu-
ers who cannot enhance their students’ per- cation sector’s poor reputation and perceived
formance from remaining in the profession. inefficient use of resources.8
A related point is that output-based pay will
create a market for teaching quality that will Potential Drawbacks to
help teachers move to schools where their Performance-Based Pay
talent is most highly valued. Equalization be- Despite its theoretical benefits, performance-
tween productivity and wages will result, with based pay offers many practical challenges.
poorly performing teachers receiving re-
duced wages and lower probabilities of pro- Measurement Problems. Performance meas-
motion, and more capable teachers com- urement poses two separate problems for per-
manding better options. Finally, if teachers formance-based pay. Incentive systems as-
are able to improve their classroom perform- sume that everyone can agree on goals; they
ance, linking compensation to performance also assume that it is possible to measure ac-
will provide all teachers incentives to im- curately progress toward these goals. Agree-
prove through professional development, ing on goals is particularly difficult in educa-
which will therefore induce still further pro- tion because competition between public
ductivity gains.4 schools is rare. In the private sector, market
mechanisms discipline firms into providing
Performance-related pay based on individual products that consumers value, but public
or schoolwide schemes could also improve schools lack market discipline. Schooling is
school productivity by inducing better gover- compulsory and public, and students are sim-
nance. For one thing, it requires school prin- ply assigned to attend their neighborhood
cipals to monitor closely the quality of their school. Parents and students who are unhappy
teachers’ work.5 It is also assumed to bring with what their schools offer generally have no
about more coherent and common teacher- alternative except to attend a private school or
management goals in addition to an im- move to another neighborhood or city—alter-
proved flow of information and feedback natives that are too costly for many.9
among all school agents. This result is as-
sumed to flow from a common interest in im- The other measurement issue represents the
proved outcomes.6 most common claim made against perform-
ance-based pay: evaluating progress toward
Other Potential Benefits. Critics of traditional the goal fairly and accurately is problematic.
pay schemes that reward experience and for- This is especially so when evaluation is based
mal qualifications instead of performance on proxies (as it often is), such as self-
argue that these schemes are unfair to highly reported effort and motivation.10 Identifying
motivated, effective, and efficient teachers precisely what one teacher contributes to a
whose extra efforts are not rewarded.7 Per- student’s performance and separating his
VOL. 17 / NO. 1 / SPRING 2007 91
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 92
Victor Lavy
contribution from those of other teachers, have unintended consequences. Teachers, for
the school, the principal, and the family is ex- example, may focus on the easiest way to in-
tremely difficult.11 Compounding the prob- crease the rewarded measure while ignoring
lem is the fact that students are often deliber- measures that schools and parents ultimately
ately assigned to specific teachers—that is, want to improve.13 Similarly, when one di-
the assignment of teachers to classes is not mension of output is easily measured but an-
random. A still further complication is how to other is not, teachers may dedicate their ef-
identify the contributions of previous teach- forts to maximizing the measurable at the
ers, who may have been superior or inferior. expense of the unmeasured dimension. Col-
lectively, such efforts could even begin to
constrict a school’s curriculum to measurable
Pay based on reading and subjects.14 A further risk is that because test
scores measure only certain skills, linking
math test scores, for example,
compensation to test scores might cause
might encourage teachers to teachers to sacrifice the nurturing of curiosity
and creative thinking to teaching the skills
favor those subjects at the
tested on standardized exams—a practice
expense of, say, music and known as teaching to the test.15Pay based on
reading and math test scores, for example,
art or values and civic
might encourage teachers to favor those sub-
responsibility. jects at the expense of, say, music and art or
values and civic responsibility.16 A teaching-
to-the-test mentality is thus assumed to sup-
Negative Effects on Motivation and Collegial- port the creation of a system where a narrow
ity. Another concern is that implementing in- curriculum necessarily restricts student
dividual-based incentives may create unfair achievement in domains not tested.
competition between teachers, especially in
the absence of transparent criteria, thus un- Unintended consequences may also arise if
dermining collaboration. Even if evaluation is teachers “game play” and develop responses
accurate and fair, teachers may still feel ag- that generate rewards contradicting the pro-
grieved if their competence is questioned. fession’s spirit.17 In other words, measuring
Evaluation may also create new hierarchies student output may stimulate teachers to par-
by giving administrators an additional source ticipate in inappropriate or deviant behavior,
of power over teachers and the curriculum. such as cheating. Using data from Chicago’s
Individual incentives could also undermine public schools, Brian Jacob and Steve Levitt
principal-teacher relationships because of the detected cheating in approximately 4 to 5
asymmetry in how each party views teacher percent of the classes in their sample.18They
evaluation: teachers use it to determine how also found that cheating responds swiftly to
they are performing and how they can im- changes in teacher incentives. After stan-
prove, while principals use it to measure dardized tests took on increased salience in
teachers’ contribution to the school.12 Chicago’s public schools in 1996, the preva-
lence of cheating rose sharply in low-
Unintended Consequences. Some analysts achieving classrooms, but not in classes with
caution that performance-based pay may average or above-average students. The
92 THE FUTURE OF CHILDREN
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 93
Using Performance-Based Pay to Improve the Quality of Teachers
prevalence of cheating also appeared to be atively risk-free input-based payment, per-
systematically lower where the costs of cheat- formance-based pay exposes employees to
ing were higher or the benefits of cheating earnings variability beyond their control. If
lower, as in classrooms where a large number teachers, like other workers, are risk averse,
of students’ test scores were excluded from inducing them to accept a risky compensa-
official calculations because they were bilin- tion package will entail higher average pay
gual. Other studies of unintended conse- overall.
quences include altering school lunch menus
during testing periods in an apparent attempt Teachers Are Motivated by Nonfinancial In-
to artificially increase student test scores and centives. A frequent criticism of performance-
manipulating who takes the test.19 based pay is that teachers, as professionals
relatively immune to motivation by pecuniary
Providing financial incentives to improve rewards, will not respond to financial incen-
performance may be counterproductive in tives. Monetary rewards could thus simply in-
other ways as well. First, it may demoralize flame resentment toward management and
teachers, resulting in reduced effort. In labo- decrease employee loyalty, both of which
ratory experiments, one study found that could reduce productivity. One study suggests
workers in high-powered incentive systems that nonmonetary rewards, such as additional
may become unmotivated and thus work less holidays, may be better motivators.25
than they would under a flat wage regime.20
Second, financial incentives may undermine Union and Teachers’ Opposition. Teacher
intrinsic motivation, that is, the sense of duty unions worldwide strongly oppose perform-
or satisfaction that motivates coming to ance-based pay.26Unions view wage differen-
work.21 This threat is particularly real for tiation on the basis of subject taught, as well
teachers, who, as a group, exhibit strong in- as any sort of subjective evaluation of teach-
trinsic motivation flowing from the value they ers, as threats to their collective bargaining
place on interacting with children and seeing strategies and therefore reject them outright.
them succeed.22 And union views weigh heavily: lobbying by
unions has often halted efforts to legislate
Another potential distortion is that teachers performance-based rewards.27 Union objec-
may focus disproportionate attention on tions appear to reflect opposition voiced by
those students who are most likely to im- teachers directly.28 Teachers see perform-
prove their test scores or to cross a desig- ance-based pay, supported by unfair evalua-
nated threshold.23 The highest- and lowest- tion, as a threat to their autonomy. Sanctions
performing students may consequently be against poorly performing schools, which are
neglected because they do not promise ade- included in some performance-based
quate returns on investments of teachers’ schemes, are another major source of union
quality time. and teacher opposition.29
Risks Posed to Teachers Could Increase Disappointing Experience with Past Merit
Costs. The risks posed to teachers by per- Pay Programs. The repeated failures of
formance-based pay could lead them to de- poorly designed and implemented merit pay
mand high compensation, which could in programs over the past two decades have un-
turn raise the cost of education.24 Unlike rel- dermined the credibility of new and better-
VOL. 17 / NO. 1 / SPRING 2007 93
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 94
Victor Lavy
designed initiatives. A key weakness in past quite good at identifying those teachers who
programs has been opaque goals, which make produce the largest and the smallest standard-
it hard for teachers to understand the pro- ized achievement gains in their schools, but
gram and undermine their support for it. far less able to distinguish between teachers
Opaque goals also make it difficult for admin- in the middle of the distribution.34 They also
istrators to explain why some staff members found that principals systematically discrimi-
receive a bonus and others do not. One study nate against male and untenured faculty.35
finds that even in established programs such
as those implemented in Kentucky and North A principal-based assessment system would
Carolina, many participants remain skeptical likely result in higher student achievement
that bonuses go to qualified teachers.30 than today’s input-based compensation sys-
tem. But Jacob and Lefgren cite an impor-
The High Cost of Performance-Based Pay tant limitation of their research. First, the
Schemes. Finally, implementing perform- principals whom they examined were not
ance-based pay is easier in small organiza- themselves evaluated explicitly on the basis
tions, such as private schools, than in large of their ability to identify effective teachers.
public school systems with sizable teaching Moving to a system where principals have
staffs. System size therefore impinges on the more authority and responsibility for moni-
observed high cost of performance-related toring teacher effectiveness might enhance a
pay, making the program infeasible. One principal’s capacity to identify the required
study argues that adequate evaluation of characteristics. But principals may be less
every teacher, expensive in itself, would re- willing to assess teachers honestly under such
quire considerable resources if performed a system, perhaps in response to social or po-
regularly.31 The time alone required to ad- litical pressures. Further, the inability of
minister a pay-for-performance system would principals to distinguish between teachers in
have severe budgetary implications.32 More- a broad middle range of quality suggests cau-
over, as a research study points out, improved tion in relying on principals for the finely
productivity in the private sector can gener- tuned performance determinations that
ate added income to help mitigate budget might be required under certain merit pay
problems, but enhanced productivity in pub- policies.
lic schools has no such effect.33
In response to the concern that merit pay
Overcoming Some of the Obstacles models may hamper collaboration, one could
Several of the many potential obstacles to im- structure the system to reward teacher coop-
plementing an effective performance-related eration, especially through group-based
pay system can be addressed. For example, pay.36 This strategy can foster both teacher
one solution to the measurement problem is interdependence and acknowledgement of
to compensate teachers on the basis of princi- that interdependence. That said, team-based
pal evaluations. Brian Jacob and Lars Lefgren incentive systems raise the risk of “free rid-
compared subjective principal assessments ing.”37 If each teacher’s share of the team re-
with measures of teacher effectiveness based ward is small relative to the cost of effort and
on gains in student achievement on standard- if effort is difficult to observe, every teacher
ized tests—measures often known as teacher in the team will have an incentive to shirk
value added—and found that principals are and free ride on the efforts of others.38 One
94 THE FUTURE OF CHILDREN
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 95
Using Performance-Based Pay to Improve the Quality of Teachers
way to avoid this problem is to encourage private school teachers viewed performance-
peer pressure and mutual monitoring within related pay more favorably than their public
the team.39 school counterparts.41 Teachers’ attitudes
thus appear relatively malleable and to de-
Fears that teachers would need to be highly pend on program design.
compensated for the increased risk in a per-
formance-based compensation system are Teachers’ and union objections were over-
probably overstated. If teachers are paid on come in some cases when specific interest
the basis of student performance, and if the groups and legislators supported perform-
number of students whom teachers teach
each year is high, the year-to-year variation in
average class test scores is likely to be small. Fears that teachers would
Furthermore, even under the most ambitious
need to be highly compen-
schemes only a fraction of teachers’ wages
would be tied to performance, thus making sated for the increased risk
compensation based on incentives only a
in a performance-based
marginal component of pay. It is thus unlikely
that earnings will fluctuate by more than a compensation system are
few percent annually around some basic
probably overstated.
trend.
The idea that teachers themselves—as re-
flected in the positions of their unions—op- ance-based pay. But political turnover makes
pose performance-based rewards may also be such support fragile, particularly in times of
overstated. One study found that most teach- economic recession, because the cost of
ers favor additional pay for additional duties performance-based pay is more visible than
per se and as a component of a career ladder are the benefits of improved student
where performance dictates the speed of ad- achievement.42
vancement.40 Unsurprisingly, performance-
based rewards are more popular when they Many of the practical challenges faced by
supplement, rather than replace, other forms performance-related pay, then, can be ad-
of salary. dressed through careful design of the system.
And despite the opposition of teachers
The same study found that the pay level in a unions to performance-based compensation,
school district appeared to have no influence it is not clear that the objections to such sys-
on teachers’ attitudes toward merit pay, al- tems come from the teachers themselves.
though teachers who were paid low salaries
and who belonged to ethnic minorities were Evidence on School-Based
more likely than others to support the pro- Performance Systems
gram. Attitudes toward merit pay were inde- In this section I review evidence on several
pendent of the number of students who were school-based incentive programs imple-
eligible for free lunches, suggesting that stu- mented in recent years both in the United
dents’ socioeconomic status did not affect States and in other countries. The programs
teachers’ views on merit pay. Interestingly, vary in their basic structure and details, with
VOL. 17 / NO. 1 / SPRING 2007 95
05 5564-7 lavy.qxp 1/15/2007 10:25 PM Page 96
Victor Lavy
some targeted at teams of teachers and oth- teristics to adjust to make the participating
ers at individuals.43 schools comparable to other schools in the
state. In addition, the test score gains in Dal-
Evidence from the United States las may have been part of a trend that started
Although school-based performance pay the- before the program was implemented.
oretically has many attractive features, re-
searchers have been able to find little causal The Dallas study also highlights some unin-
evidence that it is effective in U.S. programs. tended consequences. In an earlier study,
For example, three researchers studied Charles Clotfelter and Ladd had reported
school-based incentive pay systems in Ken- that in the Dallas program, schools of low so-
tucky, North Carolina (Charlotte-Mecklen- cioeconomic status rarely won awards.47 In
burg), and Maryland.44 They concluded that response, the state divided schools into five
in the Charlotte-Mecklenburg and Kentucky groups based on socioeconomic characteris-
programs, but not in the Maryland program, tics and rewarded the top performers in each
both teacher motivation and student out- group. But some of the lower-performing
comes improved. But because all three stud- schools in the upper socioeconomic bands
ies lacked a control group, they could not es- felt that they had been treated unfairly. Di-
tablish definitively that the program viding the schools into socioeconomic groups
itself—and not some other factor—was the also encouraged an undesired strategic re-
cause of the improvements.45 sponse from principals who realized that
their ability to gain an award was based on
Similarly Helen Ladd studied a school-based the socioeconomic category into which they
bonus program in Dallas.46 The program, were placed.
which began in the 1991–92 school year and
continued through 1995, ranked schools by Finally, two studies of a South Carolina per-
how well their students’ test scores compared formance-based program that included both
with state average scores, adjusting for stu- school-based and individual-based rewards
dents’ socioeconomic status. To avoid teach- found that student performance improved.48
ing to the test or other gaming behavior, the The studies, however, may overstate the in-
program relied on multiple measures of stu- centive effects because teachers could
dent outcomes, including two tests given choose whether to apply for an award. If, as
each year. Ladd compared gains in school- would be expected, only the most productive
level test scores in Dallas with gains in other teachers chose to apply, then part of the stu-
cities (adjusting for many school characteris- dent gains may be attributable not to the in-
tics, such as racial mix and relative depriva- centives but to the fact that participants were
tion) to evaluate the impact of this bonus better teachers in the first place.
scheme. She found that pass rates appeared
to increase more quickly in Dallas than in International Evidence
other cities. Effects were most positive for One of the stronger examples of a school-
Hispanics and whites and insignificant for based incentive program comes from Israel.
blacks. Although the study suggests that a In February 1995, Israel announced a com-
school-based program can be effective, it was petition for a monetary bonus for secondary
not conclusive. It had, for example, only a schools and teachers based on their students’
limited number of student and school charac- performance.49 The objectives were to re-
96 THE FUTURE OF CHILDREN