Date post: | 26-Nov-2023 |
Category: |
Documents |
Upload: | michiganstate |
View: | 0 times |
Download: | 0 times |
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
Economics of Education Review 000 (2015) 1–16
Contents lists available at ScienceDirect
Economics of Education Review
journal homepage: www.elsevier.com/locate/econedurev
Review
The impact of teachers’ unions on educational outcomes: What
we know and what we need to learn
Joshua M. Cowen a,∗, Katharine O. Strunk b,1
a College of Education, Michigan State University, 116-F Erickson Hall, East Lansing, MI 48824, United Statesb University of Southern California, United States
a r t i c l e i n f o
Article history:
Received 6 February 2015
Revised 17 February 2015
Accepted 22 February 2015
Available online xxx
JEL classification:
I20
J5
J51
J58
Keywords:
Economics of education
Teacher unionization
a b s t r a c t
In this paper we consider more than three decades of research on teachers’ unions in the
United States. We focus on unions’ role as potential rent-seekers in the K-12 educational
landscape, and specifically how teachers’ unions impact district and student outcomes. We
review important methodological improvements in the identification of union impacts and
the measurement of contract restrictiveness that characterize a number of recent studies. We
generally find that the preponderance of empirical evidence suggests that teacher unionization
and union strength are associated with increases in district expenditures and teacher salaries,
particularly salaries for experienced teachers. The evidence for union-related differences in
student outcomes is mixed, but suggestive of insignificant or modestly negative union effects.
Taken together, these patterns are consistent with a rent-seeking hypothesis. We conclude by
discussing other important union activities, most notably in the political arena, and by noting
that recent changes in state laws pertaining to teachers and teacher unions may provide
context for new directions in scholarship.
© 2015 Elsevier Ltd. All rights reserved.
1. Introduction
Teachers’ unions, perhaps more than other public sector
unions, have remained controversial since their early incep-
tion in the 1850s. Although teachers’ associations were origi-
nally intended to advocate and provide support for members,
they have evolved into highly active and influential players
in local, state and national contexts. Teachers make up 26.6%
of state- and local-public sector workers and are unionized at
a higher rate than many other public sector workers (Sanes &
Schmitt, 2014). The two major teachers’ unions, the National
Education Association (NEA) and the American Federation of
Teachers (AFT), have grown from just 700,000 members in
1957 to a combined membership of over 4 million educators
∗ Corresponding author. Tel.: +1 517 355 2215.
E-mail addresses: [email protected] (J.M. Cowen),
[email protected] (K.O. Strunk).1
Tel.: +1 213 740 2190.
http://dx.doi.org/10.1016/j.econedurev.2015.02.006
0272-7757/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of t
and what we need to learn, Economics of Education Review (2015)
and education support providers. In fact, the NEA today is
the largest labor union in the United States. The two teach-
ers unions spend more than any other public sector union on
federal lobbying activities.2
It is not just their size and political power that make teach-
ers’ unions important for focused study. Teachers’ unions
play an active role in setting school district policy through
their role as collective bargaining agent for teachers in the
45 states (plus the District of Columbia) that require or per-
mit teachers’ unions to collectively bargain with district ad-
ministrators. The resulting collective bargaining agreements
(CBAs, or contracts) regulate nearly every aspect of teachers’
work and school operations, which has led one scholar to note
that union contracts are the most important policy document
governing school district operations (Hill, 2006). CBAs regu-
late education policy regarding teacher salaries and benefits,
2 Data retrieved February 2014 from the Center for Responsive Politics,
www.opensecrets.org.
eachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
2 J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
teacher assignment and transfers, teacher evaluations, class
size, grievance procedures, leaves, association rights, student
placement, instruction and curriculum, layoffs, preparation
periods and non-instructional duties and more (Goldhaber,
2013; Moe, 2009; Strunk, 2012, 2014). Although teachers’
unions cannot bargain on behalf of public school teachers
in the five states in which collective bargaining is explicitly
prohibited they may still have a voice in district-level policy-
setting (Fowles & Cowen, 2014).3 The two major teachers’
unions have local affiliates in every state, and often work
as professional associations in non-union locations, helping
teachers to understand their rights, advocating and lobbying
for teachers’ interests and taking action in local and state-
wide elections.
Although teachers’ unions are central actors in the pro-
vision and governance of public education, the literature on
teachers’ unions remains relatively thin, with myriad unan-
swered questions. However, the research base on teachers’
unions has improved dramatically over the past two decades.
In particular, recent studies have brought about two method-
ological innovations in the study of the impact of teachers’
unions on education outcomes, in both the identification of
plausibly causal union impacts and in the measurement of
union strength. As unions continue to occupy center stage in
policy debates over public education—for example in elec-
toral contests in states like Wisconsin and California, and in
battles over regulations for which unions have advocated and
fought such as tenure and due process—it is important to re-
turn to the literature to determine what is and is not known
about the impacts of teachers unions on education outcomes.
In what follows, we first review the current context in
which teachers’ unions are operating and discuss theoretical
motivations for union activity. We then review the literature
that has traditionally informed research and policy about
teachers’ unions, focusing on the methodological limitations
in earlier work and more recent innovations in both the
identification of union impacts and the measurement of
bargaining strength. Next, we review the extant literature
on the relationship between unions and student and district
outcomes. We close by discussing how these trends may be
extended and enhanced by studies exploiting new changes
to union-related policies nationwide, arguing ultimately
for an increased focus on teachers’ unions in economic and
policy analytic research.
2. The current context for teachers’ unions and collective
bargaining in the United States
As noted above, teachers’ unions have grown in both size
and political stature over the past half-century. In spite of,
or perhaps in part due to unions’ presence in the political
arena, many recent reform efforts intended to improve the
quality of teaching in American public schools have targeted
teachers’ collective bargaining rights as well as areas of legis-
lation that safeguard teacher protections fought for by their
3 In 15 states bargaining is permissible but not required. In these states
whether teachers negotiate collectively is determined at local district levels;
whether bargaining includes wages, working hours, conditions such as class
size, dismissal policies and teacher evaluation varies considerably between
these states (National Council on Teacher Quality, 2014a).
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of
and what we need to learn, Economics of Education Review (2015)
unions. States have begun to weaken teacher job security
and change seniority provisions that have historically guided
teacher assignment and transfer. Traditional tenure protec-
tions are also being diminished across the country: 11 states
now make teaching effectiveness (rather than experience)
the preponderant criterion for attaining tenure and nine more
states are including student performance among those cri-
teria. Teacher ineffectiveness is now grounds for dismissal
in 20 states (National Council on Teacher Quality, 2014b).
Where legislative action has failed to materialize, reformers
are turning to the judiciary. In California, for example, a group
of students backed by school reformers has sued the state in
an attempt to remove seniority protections during layoffs and
limit teachers’ due process and tenure protections (Medina,
2014). In June 2014, the California Superior Court ruled that
these “challenged statues impose a real and appreciable im-
pact on students’ fundamental right to equality of education
and that they impose a disproportionate burden on poor and
minority students,” essentially violating the equal protection
clause of the California Constitution (Vergara v. State of Cal-
ifornia, 2014, p. 8). The ruling is now being appealed by the
state and the two teachers’ unions, who asked to be named
as co-defendants in the case. A similar lawsuit has been filed
in New York (Wright v. New York, 2014) and others are being
discussed in states around the country.
More fundamentally, policymakers across the country are
acting to restrict unions’ fiscal and membership resources.
Twenty-six states now limit or prohibit teachers’ unions’ abil-
ity to collect membership dues from teachers who do not ex-
pressly consent to paying such agency fees (National Council
on Teacher Quality, 2014b), and multiple court cases are sur-
facing that aim to reduce teachers’ and other public sector
unions’ abilities to collect membership dues from employees
who do not choose to join them. The U.S. Supreme Court re-
cently ruled in Harris v. Quinn (Case No. 11-681) that, in some
cases, public employee labor unions cannot require employ-
ees to pay membership dues (Estlund & Forbath, 2014). In Cal-
ifornia, Freidrichs vs. California Teachers’ Association (US Dis-
trict Case No. SACV 13-676-JLS) is challenging the California
Teachers’ Association’s right to require teachers to pay union
dues. The U.S. 9th Circuit Court of Appeals has ruled that the
case should be passed on to the U.S. Supreme Court, where
many believe the more conservative judiciary may overturn
the precedent-setting Abood vs. Detroit Board of Education
(Supreme Court Case 75-1153) 1977 decision that originally
legalized unions’ rights to collect “fair share” (Fensterwald,
2014).
At the federal policy level, large initiatives such as waivers
to the Elementary and Secondary Education Act, Race to the
Top and the Teacher Incentive Fund have incorporated re-
quirements that directly counter long-held union-supported
protections. Elements of these three federal programs require
the creation of large data systems that link teachers to their
students, teacher evaluation systems that expressly calcu-
late teacher performance in part based upon their students’
achievement, and systems that tie teacher compensation to
their classroom performance rather than solely to experience
and educational credentials. In short, teachers’ unions, along
with other public employees, are facing clear challenges not
only to their closely held policy priorities but to their exis-
tence itself (Freeman & Han, 2012).
teachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16 3
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
4 There is a third view of unions’ role in education policy that takes the per-
spective that teachers’ unions act as professional organizations. Advocates
of this point of view conceive union motivation to be based in the desire to
raise the professionalism of teaching by ensuring better working conditions
for teachers, providing teachers with individual supports to enhance their
practice, and helping teachers to become “knowledge workers” (e.g., Bascia,
1994, 2000; Kerchner, 2003, 2004; Kerchner & Koppich, 1993, 1999, 2000;
Kerchner, Koppich, & Weeres, 1997, 1998).
The immediate result of these changes is that unions
are operating across a changing and increasingly diverse
set of policy and political conditions (Winkler, Scull, &
Zeehandelaar, 2012). Yet still little is known about the im-
pact of teachers’ unions on district and student outcomes,
and therefore about the implications, if any, of current leg-
islative, judicial and policy changes. This paper is intended to
clarify what is, and is not, understood about the impacts of
teachers’ unions on these outcomes, and to highlight fruitful
avenues for future research.
3. Theoretical perspectives on union motivation
Before an extended discussion of the innovations in the
literature on teachers’ unions, it is important to contem-
plate why teachers’ unions theoretically might impact stu-
dent and district outcomes. From one perspective, that most
closely associated with the fields of labor and public eco-
nomics, unions impact outcomes as actors in the market
for educational production. From this standpoint, unions are
rent-seekers (Hoxby, 1996), looking to gain from their in-
volvement in public education through increases in salaries
and enhanced working conditions, exhibited through pro-
visions such as class size (smaller classes are easier for
teachers to manage, but also provide greater membership
for unions), longer planning periods, shorter work days and
school years, and the like. The rent-seeking model of teach-
ers’ unions suggests that teachers’ unions extract rent from
school districts by negotiating increases in their own salaries
and working conditions. Because unions are also able to ne-
gotiate strong job protections, districts may hire and retain
more teachers, and specifically more lower-quality teach-
ers, than is optimal under the same levels of salary ex-
penditures (Eberts & Stone, 1984). The predicted result of
such conditions is a scenario in which unions achieve these
rents while providing no commensurate gains to—perhaps
even harming—district productivity as measured by student
outcomes.
Political science has also contributed a theoretical orienta-
tion germane to understanding how teachers’ unions impact
outcomes in educational policy. The work of Terry Moe in par-
ticular has stressed the role of teacher unions as organized
interest groups with stakes in the policymaking process. In
Moe’s framework (e.g. Moe, 2011, forthcoming) union activ-
ity is not only geared toward maximizing salaries and protect-
ing teacher jobs, but also toward the opposition of any policy
change that weakens the union position vis-à-vis those and
other priorities. Moe’s framework is also clarifying in that
it includes the notion of unions as actors within the public
sector, rather than simply one side of a bilateral negotiation
over resources. Through this lens, which is in some ways an
extension of the rent-seeking view taken by economists, ar-
guments such as NEA President Lily Eskelsen-Garcia’s that
unions serve the public interest by fighting for public schools
and school children and even other social causes (e.g. Bryant,
2014) are predictable responses to any efforts to curb their
power. Like other special interests groups jockeying for po-
sition, unions’ activity functions not to promote the public
good per se, but to align the definition of that public good to
their own preferences. As a result, unions’ support of fixed
salaries, lifetime tenure, and seniority-based staffing, as well
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of t
and what we need to learn, Economics of Education Review (2015)
as their opposition to charter schools, school accountability
programs and teacher evaluation, become matters of fairness
and equity, due process, the value of public schools and the
limits of standardized testing.4
Although we focus on the rent-seeking perspective in
this paper, we do discuss (albeit briefly) the notion of
unions as politically active interests groups, which leads
to very different—though not necessarily incongruent—
organizational points of emphasis for any comprehensive
discussion of union effects. Critics of collective bargaining
itself may (or may not) also fault union electoral or lob-
bying activity—and supporters may likewise defend these
tendencies—but they occur at different levels of government,
are carried out by different individuals and affiliate groups,
with different sets of resources and, however the two may
ultimately align, toward a different set of outcomes and ob-
jectives.
4. Empirical challenges and new developments in studies
of union impacts
Over the past three decades a number of studies have as-
sessed the proposition that teachers’ unions are rent-seeking
by attempting to estimate the impact of unions on achieve-
ment and other outcomes. In large part, this research has done
so by modeling the relationship between a simple measure
of unionization (e.g., whether the state or district is union-
ized or the proportion of unionized teachers in a district) and
the outcome of interest. The most recent complete review
of the research on the impact of teachers’ unions on student
achievement, published in 2006, made clear that the body of
research up to that date rested on “shaky empirical ground”
(Goldhaber, 2006, p. 157). Goldhaber’s book chapter reviewed
extant literature through the year 2000, and noted that only
one of the economic papers on teachers’ union impacts at-
tempted to correct for potentially endogenous relationships
between unionization status and any number of district or
student-level outcomes.
Table 1 (below) outlines all of the papers that have at-
tempted to address the impact of teachers’ unions on district
and student outcomes, listing them chronologically and sum-
marizing their main methodological strategies. These stud-
ies can be split into two distinct areas: those that study the
impact of teachers’ unions on district expenditures, most of
which focus on union impact on teachers wages, and those
that examine the impact of teachers’ unions on productiv-
ity as measured by student outcomes (achievement tests
and graduation rates). The earliest of these suffer from two
main impediments to the accurate prediction of the impact
of teachers’ unions on student achievement: 1.) they utilize
econometric techniques that fail to correct for bias due to en-
dogeneity or other issues with specification; and 2.) they use
simple indicators of union status that may no longer reflect
eachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
4J.M
.Co
wen
,K.O
.Strun
k/E
con
om
icso
fE
du
catio
nR
eview
00
0(2
01
5)
1–
16
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]
Table 1
Summary of union/CBA studies relating to district and student outcomes.
Author Year Data Method Results
Chambers 1977 Two cross-section samples of school districts
from a sample of 39 elementary districts
and 50 unified districts in California,
combined with data provided by the
California Teachers Association.
OLS: average teacher salaries and school
expenditures predicted by school and
district characteristics, including controls for
non-bargaining districts located near
bargaining districts (regional bargaining
effects).
Bargaining districts spend between 5.7 and 12.2% more on teacher
salaries; 3.5–12.2%; more on administrative salaries; regional
bargaining effects of 5.3–7.9% on district spending overall.
Gallagher 1979 1974 expenditure data from 133 bargaining
(65) and non-bargaining (68) small/med
Illinois districts, stratified by district
wealth.
OLS: spending on district bargaining status and
controls. N � 40–50/regression (stratified by
wealth).
Bargaining districts spend more on teacher salaries than do
non-bargaining districts. Medium- and high-wealth bargaining
districts spend more overall (8–13%) and on non-teacher salary
expenditures than do non-bargaining med/high wealth districts.
Eberts 1983 1976–1977 CBA data from New York State
paired with district-level NY DoE data on
finances and student and teacher
characteristics.
Examines 53 individual regulations and 5
areas in CBAs.
Generates 2 measures of contract
strength/restrictiveness: index of #
provisions and Guttman scaling
technique-based measure.
Descriptive statistics and content analysis of
CBAs.
OLS: spending on CBA measures and
controls.
CBA strength and the existence of specific provisions (especially
regulating RIF and dismissal) are associated with district resource
allocation. Magnitudes are relatively small (e.g., having specific RIF
provisions is associated with increase of about 0.5–1 cent per pupil
spending on instruction, 0.75 of a cent increase per pupil in
spending on salaries).
Eberts and Stone,
chap. 2
1984 1972–1973 and 1976–1977 CBA data from
New York State. Focus on RIF protection as
a representative provision. Matched to
district financial and student
characteristics data.
Logits: regress whether districts have, lost or
gained the RIF provision between 1972 and
1976 on budget measure (expenditure level
or change in spending on instruction,
benefits or services) and district controls.
Economic events such as budget changes, enrollment changes or
school closures do not impact whether CBAs contain RIF provisions.
However, decreases in district total operating expenditures are
associated with the loss of RIF CBA protections and increases in
expenditures are associated with gaining a RIF protection.
Eberts and Stone,
chap. 4
1984 1972–1973 and 1976–1977 CBA data from
New York State. Focused on 80 CBA
provisions. Measure bargaining strength
through class size and RIF seniority
provisions as well as change of total
budget going to personnel.
Matched to random sample of elementary
and secondary classroom teacher data and
to census data.
Level and first-differenced OLS wage
regression: regressed change in ln(wage) on
a set of first-differenced covariates and an
indicator of becoming a union member
between the two time periods for each
teacher.
Greater bargaining strength is associated with stronger contracts in
terms of employment security provisions.
The estimated premium for a union joiner is 8% greater in 1977
versus 1973, and is 4% in 74 and 12% in 77.
Eberts and Stone,
chap. 5
1984 1972–1973 and 1976–1977 CBA data from
New York State. Code for 80 provisions
within CBAs. Also generate a measure of
CBA restrictiveness using a Guttman
scaling technique. Also examine 18
individual items and an index made up of
summing those items.
Matched to random sample of elementary
and secondary classroom teacher data and
to census data.
OLS: regressed expenditure areas (all per
pupil: instruction, administration, benefits,
other, salaries, teacher, average salary) on
individual CBA items and a set of controls.
Alternative regressions use an index variable
as measure of CB strength.
Stronger CBAs are associated with increases in spending overall and
in subareas: having greater than average # of items is associated
with increases in funds to instruction, benefits and salaries, and
decreased spending on "other." CBA Arbitration and grievance
protections are not associated with non-wage budget items.
However, protections against RIFs and dismissals are associated
with greater overall spending and greater expenditures on teacher
salaries.
Eberts and Stone,
chap. 6
1984 1972–1973 and 1976–1977 CBA data from
New York State. Matched to district
financial and student characteristics data.
OLS: regress achievement on district
covariates in both union and non-union
districts. Then sum the coefficients of the
parameters earlier research indicates are
associated with CB.
Achievement is higher in unionized districts (�4%). Unions have
strongest benefit for students who are near the average pretest
score – but are less beneficial for students at the tails.
(continued on next page)
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
J.M.C
ow
en,K
.O.Stru
nk
/Eco
no
mics
of
Ed
uca
tion
Rev
iew0
00
(20
15
)1
–1
65
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]
Table 1 (continued)
Author Year Data Method Results
Woodbury 1985 1977 Census of Government data for
national sample of 4851 districts.
OLS and 2SLS: regress student–teacher ratios
and teacher salaries on indicators of state
legality of bargaining class size and district
CBA class size regulations and controls.
Districts that bargain have higher student–teacher ratios (2–4%). This
results from tradeoffs with salaries (smaller class sizes and lower
salaries). Districts in states that must bargain over class size have a
student–teacher ratio that is 6.3% lower and teacher pay that is
15.7% lower.
Eberts and Stone 1986 Sustaining Effects Study (200 elementary
schools); High School and Beyond (893
high schools).
OLS predictions of district expenditures as a
function of district demographics and
bargaining status; Heckman corrections
included as sensitivity analysis.
Unionized districts have 7–15% higher costs; similar impacts when
union effect is allowed to vary by level of district productivity
measured by student math achievement.
Eberts and Stone 1987 Sustaining Effects Study including
14,000 + students in 328 schools.
OLS predictions of student achievement
differences as a function of school, student
and teacher demographics.
Positive union effect for average students; negative effect for higher
and lower achieving students, estimated separately by union and
non-union district status; post-estimation predictions used to
calculate productivity differentials; at the sample mean
union-nonunion differential = 0.67; similar but opposite direction
in favor of nonunion schools for children very high above or low
below mean; average union effect is positive at 0.33 across sample.
Kurth 1987 Author compiled state-level data on SAT
average scores; state educational
expenditures; teacher labor market
attributes 1960–1980.
OLS predictions of % change in state average
math/verbal SAT from 1972 to 1982 as a
function of state demographic changes and %
of teachers under CBA and meet-and-confer.
State-level SAT scores 0.7–0.8 points lower as % of teachers covered
by CBA increases; 0.8–0.11 points lower as % of teachers under
meet-and-confer increases.
Easton 1988 1969–1982 sample of all 55 large unified
Oregon districts paired with financial data.
OLS of district expenditures, class sizes and
teacher salary levels on set of covariates, run
on data from before and after legalized CB.
First-difference regression to predict change
in salary levels between two years post CB.
2SLS as specification check.
The onset of CB did not change relationships between ability or
willingness to pay and spending outcomes.
Kleiner and Petree 1988 Author compiled state-level data from 1972
to 1982, including measures of union
participation rates; changes to bargaining
laws; state averages on college entrance
exams and graduation rates from NEA, U.S.
Census and Educational Testing Service
sources.
OLS and state fixed effects models of the
impact of union membership and coverage
rates on college entrance exam and
graduation rate averages.
Increase in proportion of union members from 0 to 1 increases
SAT/ACT scores between 6 and 8%; elasticity of test scores to
unionization is 0.5–0.7; insignificant relationships between union
coverage and outcomes; inconsistent impacts on graduation rate.
Register and Grimes 1991 National Assessment of Economic Education
Data on 2360 students in 1987
supplemented with author-collected data
on CBA status of each district.
OLS predictions of SAT and ACT scores as a
function of student and school
characteristics, with Heckman-corrected
estimates to adjust for selection bias in
sitting for college exam.
Students in CBA districts have average 46.3 SAT (or SAT-equivalent)
points (4.7%) higher than non-CBA districts.
Argys and Rees 1995 National Educational Longitudinal Survey
(1988); tenth grade standardized testing
data.
OLS controlling for 8th grade test scores. Students in unionized schools have an average of 1.1% higher
achievement scores in mathematics; negative impacts of
approximately 0.5% for students at high and low ends of
achievement distribution; distributional effects mitigated by
inclusions of class characteristics in production function.
Duplantis, Chandler,
and Geske
1995 Survey data of 82 school districts with more
than 10,000 students nationwide.
Instruments include questions on
certification elections; election results;
CBA negotiations and superintendent
perceptions of union growth.
OLS predictions of log wages, log employment
and log district expenditures as a function of
CBA presence and union endorsements of
board candidates.
9.5% increase in teacher salaries and 15.6% increase in expenditures
in districts with bargaining agreements; 4% increase in
expenditures associated with union endorsements; no union
impacts on employment.
(continued on next page)
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
6J.M
.Co
wen
,K.O
.Strun
k/E
con
om
icso
fE
du
catio
nR
eview
00
0(2
01
5)
1–
16
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]Table 1 (continued)
Author Year Data Method Results
Zwerling and
Thompson
1995 1984 Administrator Teacher Survey of NELS;
High School and Beyond.
OLS prediction of teacher salaries as a function
of teacher professional and demographic
characteristics, school working conditions,
local school enrollment trends and teacher
union density by state.
Unionized districts spend on average 5.5% more on teacher salaries;
no differences after adjusting for statewide union density
(spillover); both direct union and spillover effects concentrated at
higher experience levels, no difference for entry-level salaries.
Hoxby 1996 Author compiled district-level data
(1960–1992) from NEA, US Census of
Governments, Census of Population and
Housing sources.
OLS, difference-in-differences; instrumental
variables predictions of per pupil
expenditures, teacher salary levels,
student–teacher ratio and drop-out rates as
functions of district demographics and
district adoption of bargaining agreements;
diff-in-diff and IV used to correct for
endogenous district characteristics.
Unions increase per pupil expenditures by 2.9–12.3%; teacher salaries
by 1.6–5.0%; reduce student–teacher ratios by 1.1–1.7 students;
worsen drop-out rate by 1.8 points, while drop-out rates less
responsive to input changes in union schools.
Nelson and Rosen 1996 Author compiled state-level data on 1995
SAT scores; 1994 NAEP scores
supplemented with state-level data from
NCES Schools and Staffing Survey and ETS.
Descriptive statistics; OLS predictions of
state-averaged SAT and NAEP scores as a
function of state testing-taking
demographics.
CBA states have regression-adjusted SAT scores of average 36 points
higher than non-CBA states; regression-adjusted NAEP scores in
CBA states are 9 points higher.
Zigarelli 1996 Collective Bargaining Law Data Set (1984)
combined with High School and Beyond
(HSB, 1982) and the Administrator and
Teacher Survey (1984) from HSB.
OLS predictions of natural log of teacher salary
and average student time per day spent in
class as a function of state law indicators
associated with unionization (right to strike,
arbitration) controlling for teacher and
school characteristics.
Right-to-strike laws associated with 11.5% higher teacher salaries;
37 min of class time fewer per day; de facto right to strike increases
salaries by 5.7% and reduces class hours by 44 min; arbitration
availability increases salaries by 3.6% and reduced class hours by 70
fewer minutes.
Milkman 1997 High school and beyond 1979–1980 and
1982 follow-up; include 2684 minority
student surveys and scores.
OLS estimates of math scores as a function of
prior math, student, teacher and school
characteristics including; estimated
separately by union non-union status.
Union/nonunion productivity differential is 0.24 for minority
students compared to 0.39 for all students; -0.14 for minority
students in majority schools, 0.31 for minority students in minority
schools.
Carini, Powell, and
Steelman
2000 Author compiled state-level data based on
Schools and Staffing Survey (1993–1994)
1993 SAT scores; 1994 ACT scores and
state demographics from US Census.
OLS predictions of SAT and ACT scores as a
function of state demographics and
unionization rate calculated from SASS.
1 point change in unionization rate associated with SAT increases
between 0.27 and 0.51 in preferred models. Similar results when
ACT averages are the dependent variables.
Goldhaber et al. 2008 1999–2000 Schools and Staffing Survey. LPM predictions of the adoption of formal and
informal merit pay systems as a function of
district performance, costs, and community
characteristics; tested against instrumental
variables models based on state partisan
control of legislatures.
Districts with non bargaining agreement 6% more likely to adopt
formal merit pay plans; more than 8% more likely in right-to-work
states.
Lovenheim 2009 Author compiled district-level data from
case laws (LexisNexis); union certification
election results from Public Employment
Relations Board in IA, IN, MN; Census of
Governments.
State by year and district fixed effects models
of effect of union certification on teacher
salaries, employment levels,
student–teacher ratios, district expenditures
and drop-out rates.
Few consistent, significant union increase on district resources; no
long-term union impact on drop-out rates (1.8% increase in
short-run, insignificant after 7.7 years).
Moe 2009 1989-9 CBA data from 371 randomly
selected California districts. Measure of
CBA restrictiveness/strength: uses factor
analysis incorporating 41 CBA items into
15 factors.
Matched to district achievement and
demographic data.
School-level OLS: regressed school
achievement growth over five years on
measure of CBA restrictiveness
(district-level), base year achievement, base
year demographics and changes in
demographics. Clusters SEs to the district
level. Separate regressions for elementary
and secondary schools.
OLS: regressed contract strength on district
covariates.
More restrictive CBAs are associated with smaller achievement
growth over time (impact sizes of -0.24 and -0.32 in elementary
and secondary districts, respectively). This is driven by the
relationship between CBA strength and achievement growth in
large districts (impact sizes of -0.44 and -0.57). In large districts,
the negative relationship between contract restrictiveness and
achievement growth is exacerbated in districts with greater
proportions of minority students. In addition, larger districts have
more restrictive CBAs.
(continued on next page)
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
J.M.C
ow
en,K
.O.Stru
nk
/Eco
no
mics
of
Ed
uca
tion
Rev
iew0
00
(20
15
)1
–1
67
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]Table 1 (continued)
Author Year Data Method Results
Cowen 2009 1999–2000 Schools and Staffing Survey,
focus on 1066 districts in 14 states with
substantial variation in district
non-bargaining/bargaining status;
supplemented with expenditure data from
School District Finance Surveys.
District-level OLS prediction of log
expenditures on salaries/benefits as a
function of district covariates supplemented
with state-level data on laws affecting
bargaining; state fixed-effects; interactions
between bargaining status and
student–teacher ratio.
CBA-related increases in compensation (salaries or benefits) of
approximately 20%; bargaining-related differences greatest at low
levels of students/teacher ratios and for district with high levels of
locally sources revenue.
Strunk and Grissom 2010 2005–2006 CBA data from 113 California
school districts with 4+ schools. Generate
a measure of overall CBA restrictiveness
using 639 items in a Partial Independence
Item Response (PIIR) Model; generate PIIR
restrictiveness measures for transfer and
vacancies, evaluation, association rights
and class size sub-areas.
Matched measure to survey data from
school board members re. union power
and direct political mechanisms through
which unions exert influence (fundraising
assistance, volunteer mobilization and
endorsements) as well as other special
interest group power and influence.
Matched to district demographic and
financial variables.
OLS: regress CBA restrictiveness (overall,
subarea, provision) on survey measures of
union power/influence, relative to business
group power/influence and district controls.
2SLS regression of CBA restrictiveness on
measures of teacher salary, instrumented
with revenue limit assigned to district.
More powerful unions negotiate more restrictive CBAs than do less
powerful/influential unions (e.g., a 1 SD increase in CR is associated
with a 0.14, 0.20 and 0.33 SD increase in "other working
conditions," "transfer and vacancy" and "evaluation" subareas).
Rather than tradeoffs occurring between contract restrictiveness
and teacher salaries, the two are positively and significantly
related.
More powerful unions negotiate CBAs with more restrictive
individual provisions.
Boards with more educators on them have more restrictive CBAs,
boards with more Republicans have less restrictive CBAs.
Strunk and Reardon 2010 2005–2006 CBA data from 466 California
districts with 4+ schools (83%). Utilizes
639 objectively chosen items from CBAs.
Uses a Partial Item Independence Response
(PIIR) model to generate a measure of
contract strength.
PIIR measure of union strength has a number of advantages over
other measures of contract strength/unionization. It is an objective
and statistically sound approach and generates transparent
measure along a probabilistically-based interval scale. Places
contracts according to their specific level of contract restrictiveness
and provides standard errors of measurement and measure
reliability. Allows for measurement of restrictiveness of provisions
within CBAs.
Rose and Sonstelie 2010 1999–2000 data from 771 California districts
that provided CA Department of Education
with their salary schedules (J90 dataset).
Matched to California achievement and
demographic data.
OLS: regress bargaining outcomes (base salary,
experience premium and teacher–pupil) and
achievement on a measure of union power
(# of eligible voters in the school district,
students/eligible voter or % homeowners in
district) on district-level controls and
measures of budget constraint. Cluster SEs to
the region. 2SLS robustness checks.
Union power (district size) is positively associated with bargaining
outcomes (teachers’ salaries – both base and experience premium
– and pupil–teacher ratio). The 10 largest districts in the sample
would have a base salary about 7% higher and an experience
premium about 11% higher than the average district. The 10 largest
districts would have teacher–pupil ratios about 6% lower than the
average sized CA district. Achievement is negatively related to
union power (district size).
Strunk 2011 2005–2006 CBA data from 465 California
districts with 4+ schools (82%). Generates
a measure of overall CBA restrictiveness
using 639 items in a Partial Independence
Item Response (PIIR) Model.
Matched with district financial,
achievement and demographic data.
OLS: regress expenditures (overall and
categories) on measure of contract
restrictiveness and covariates; regress
achievement and one-year growth in
achievement on contract restrictiveness and
covariates.
CBA restrictiveness is positively associated with overall expenditures
(1 SD stronger CBA spends 13% more). Spending driven by
increased expenditures on administrator salaries (1 SD stronger
CBA spends 0.15%), classified personnel benefits and
instruction-related functions. CBA restrictiveness is negatively
associated with expenditures on textbooks and other instructional
materials (1 SD stronger CBA spends 0.20% less on textbooks and
supplies) and on schoolboard related activities.
CBA restrictiveness is negatively associated with same-year
achievement score, but not with growth in achievement (1 SD
stronger CBA is associated with a �4.5 point decrease in
achievement).
(continued on next page)
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
8J.M
.Co
wen
,K.O
.Strun
k/E
con
om
icso
fE
du
catio
nR
eview
00
0(2
01
5)
1–
16
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]
Table 1 (continued)
Author Year Data Method Results
Strunk and McEachin 2011 2005–2006 CBA data from 465 California
districts with 4+ schools (82%). Generate a
measure of overall CBA restrictiveness
using 639 items in a Partial Independence
Item Response (PIIR) Model.
Matched with district achievement and
demographic data from 2005–2006
through 2008–2009.
OLS and Logits with year FEs: regress
achievement on CBA restrictiveness and %
minority, % low income, % proficient and
district covariates. Iterations of models
include interactions between CBA
restrictiveness and indicators of
hard-to-staff districts.
HLM and HGLM for models using school and
district level data.
CBA restrictiveness is negatively associated district aggregate student
achievement (district math and ELA proficiency levels and
graduation rates; positively associated with likelihood of failing
NCLB (PI) and continuing to fail NCLB (higher levels of PI)). A 1 SD
increase in CBA restrictiveness is associated with a 7.7% increase in
the probability that a district will be in PI and a 4.5% increase in the
likelihood that district will progress to higher levels of PI, and with
a 1.3% decrease in grad rates. Results are amplified in hard-to-staff
districts.
At the school level, there is no relationship between CBA
restrictiveness and schools’ likelihood of failing NCLB, nor between
CR and school-level proficiency or graduation rates. However,
negative relationships between hard-to-staff school characteristics
and school achievement measures are amplified in districts with
more restrictive contracts.
West and Mykerezi 2011 Schools and Staffing Survey and TR3
(National Council on Teacher Quality).
OLS estimates of effect of presence of
bargaining agreement on a set of dependent
variables related to teacher pay, controlling
for district characteristics; IV estimates
based on exogenous variation in timing of
state laws affecting unionization and public
sector union membership rates.
Presence of a CBA decreases the probability of merit pay adoption by
20 percentage points; increase in average salaries by 3.9%; similar
returns to experience (1.1–1.2%) for early and later career teachers.
Winters 2011 1999–2000 Schools and Staffing Survey. GMM estimator accounting for spatial
correlation in salaries between districts:
salaries in nearby districts are instrumented
with distance-weighted averages of
explanatory variables in adjacent and other
local districts.
Relative to states with no collective bargaining membership, states
with all teachers as CBA members have average salaries 18% higher
for experienced teachers; states with greatest membership density
have 28% higher salaries for experienced salaries than those with
lowest density of membership.
Strunk 2012 2005–2006 CBA data from 465 California
districts with 4+ schools (82%). Generates
a measure of overall CBA restrictiveness
using 639 items in a Partial Independence
Item Response (PIIR) Model. Also focuses
on 95 representative provisions in CBA
areas: compensation, evaluations, leaves,
transfer/vacancy, class size, non-teaching
duties, and school calendar/year. Matched
with district demographic data.
Descriptive analysis examining 95 provisions
by provision district % of FRPL, % minority
and urban location.
OLS: regressed existence of provision (LPM)
and overall CBA restrictiveness (OLS) on
district demographic variables.
Although CBA regulations largely constrain principal and district
autonomy, there is some flexibility in CBAs and CBA provisions
serve to enhance teachers’ professional working conditions.
Hard-to-staff districts have particularly restrictive CBA provisions,
but also have some of most flexibility-enhancing provisions.
Districts with greater % minority, controlling for other covariates,
have more restrictive CBAs, as do urban and larger districts.
(continued on next page)
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
J.M.C
ow
en,K
.O.Stru
nk
/Eco
no
mics
of
Ed
uca
tion
Rev
iew0
00
(20
15
)1
–1
69
ART
ICLE
INPRESS
JID:ECO
ED
U[m
3Gdc;M
arch23,
2015;17:12]
Table 1 (continued)
Author Year Data Method Results
Brunner and Squires 2013 2007–2008 Schools and Staffing Survey data
of nationally representative sample of
districts. Matched with demographic data
from the 2005–2009 American
Community Survey and NCES data.
OLS with labor market and state FEs: regressed
bargaining outcomes (base salaries, 2
experience premium measures and
teacher–pupil ratio) on the # of eligible
voters in the district (size, or union power),
the interaction between size and a state
indicator of CB status, and vector of controls.
In states that mandate CB, union strength (district size) is positively
associated with bargaining outcomes (higher beginning salaries
and salary returns to experience, lower teacher–pupil ratios). In
states in which CB is not permitted, there is a positive relationship
between union power (size) and certain bargaining outcomes
(higher beginning teacher salaries and lower teacher–pupil ratios),
but there is a negative relationship between district size and
returns to experience for teachers. Results suggest that more
powerful unions bargain for more generous returns to teacher
seniority at the expense of staffing ratios and base salaries. In states
that mandate CB, moving from a district with 1 SD below the mean
to one SD above the mean in district size would increase the base
salary of teachers by 4.5% and the salary premium paid to
experienced teachers by 2.4% for teachers with a BA and 3.6% for
teachers with an MA.
Goldhaber,
D-Entremont, Fang,
Lavery, and
Theobald
2013 Uses 2010–2011 CBA CBA data from all 270
districts in Washington state. Generate a
measure of overall CBA restrictiveness
using 633 items in a Partial Independence
Item Response (PIIR) Model, as well as
measures of subarea restrictiveness
(Accessibility; association rights;
evaluation; grievance;
layoffs/benefits/leaves; hiring and
transfers; workload; a cherry picked
subset of dramatic provisions).
Matched to district demographic data.
Replicates Strunk and Reardon (2010) PIIR
method in WA state to assess overall CBA
restrictiveness and the restrictiveness of
subsets of CBA items. Uses correlations and
t-tests to assess relationships between areas
of CBAs and with district characteristics.
WA state CBAs that contain many provision in one area of the CBA
tend to also contain many provisions that regulate other areas of
policy. Highly restrictive CBAs are larger, have more Asian/PI, black,
bilingual students, and get greater % of funding from local funds
but less from state.
Lott and Kenny 2013 721 nationally drawn districts with more
than 10,000 students, drawn from
www.schooldatadirect.com;
supplemented with IRS and NCES data on
school/district characteristics plus union
dues and expenditures per student.
OLS predictions of 4th/8th grade math and
reading proficiency percentages as a
function of school/district demographics and
union dues/expenditures plus IV corrections
using right-to-work laws as instruments.
1 s.d. rise in union dues/teacher decreases math and reading
proficiency percentages by 3.7 points; 1 s.d. rise increase in union
expenditures/student decrease proficiency by 3.2–3.4 points for
math and reading, respectively. IV estimates of dues and spending
relationships comparable at 2.2–3.7 decreases per 1 s.d. increase of
union resources.
Strunk 2014 2008–2009 CBA data from 506 California
districts with 4+ schools (90%).
Descriptive statistics. CBA regulations largely constrain principal and district autonomy to
respond to education reforms although there is some flexibility in
CBAs to enable reform.
Goldhaber, Lavery,
and Theobald
2014 Uses 2010–2011 CBA CBA data from all 270
districts in Washington state. Generate a
measure of overall CBA restrictiveness
using 633 items in a Partial Independence
Item Response (PIIR) Model, as well as
measures of subarea restrictiveness
matched to district demographic and
geographic data.
Replicates Strunk and Reardon (2010) PIIR
method in WA state and uses measures of
CBA restrictiveness in a spatial econometric
approach to explore impact of bargaining
outcomes in nearby districts on an
individual’s district’s contract.
OLS and ML estimation of spatial lag model,
see OLS is likely biased and use ML
estimates. Also use 2SLS in case of OVB using
proximity as instruments.
Spatial relationships play an important role in determining
bargaining outcomes as measured by CBA restrictiveness, although
shared institutional bargaining structures seems to be what really
matters (Education Service Districts, or ESDs) and Uniserv councils.
The influence of geographic distance found in previous studies of
teacher wages may actually reflect influence of bargaining
structures. Spatial relationships are consistent over all subareas of
CBAs, except transfer and hiring provisions (Uniserv drives spatial
relationship) and grievance provisions (ESD drives relationship).
Ple
ase
citeth
isa
rticlea
s:J.M
.C
ow
en
,K
.O.
Stru
nk
,T
he
imp
act
of
tea
che
rs’u
nio
ns
on
ed
uca
tion
al
ou
tcom
es:
Wh
at
we
kn
ow
an
dw
ha
tw
en
ee
dto
lea
rn,E
con
om
icso
fE
du
catio
nR
ev
iew
(20
15
),http
://dx
.do
i.org
/10
.10
16
/j.eco
ne
du
rev
.20
15
.02
.00
6
10 J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
the context in which unions operate. In what follows, we first
outline the empirical challenges faced by much of the earlier
set of research addressing union impacts. We then discuss
what is known in light of these challenges about the impacts
of unions on district spending, and especially on teachers’
wages, and on student achievement.
4.1. Identifying union impacts
Most of the studies that addressed union impact through
the early 2000s did so by estimating models that did not ac-
count for endogeneity or other sources of bias. In particular,
the possibility that the presence of a local union was en-
dogenous to variation in the outcome of interest was largely
(but not wholly) unexplored in this early work. Until 1996,
the dominant empirical studies of union impacts were those
conducted by Eberts and Stone (e.g. 1984,1986, 1987) which
essentially adjusted mean differences in district or student
outcomes via Ordinary Least Squares regressions, accounting
for a number of demographic and institutional characteris-
tics. Later studies (e.g. Argys & Rees, 1995; Kleiner & Petree,
1988; Kurth, 1987) employed a similar approach. Only two
studies during this period moved beyond a basic OLS frame-
work: Eberts and Stone (1984)’s estimate of a teacher-level
hedonic wage model, which incorporated teacher fixed ef-
fects that also captured time-invariant district characteris-
tics, and Woodbury’s (1985) two-stage least squares (2SLS)
models incorporating variation in community demographics
as a source of exogenous variation in resources. Neither study,
however, was explicitly concerned with the unobserved loca-
tional differences that may have predicted both unionization
status and the outcomes of interest.
These limitations reflect the field at the time, and remain
important contributions to the base of knowledge concern-
ing the relationship between unionization and district out-
comes and productivity. However, while this early research
assesses the difference in outcomes in unionized versus non-
unionized districts, these studies’ ability to interpret these
differences as causal union impacts is limited. In particular, it
is entirely plausible that districts choose to unionize and/or
greater proportions of teachers join a union when district ad-
ministrators spend money in specific ways, or when student
achievement is particularly high or low. As such, the findings
of this early research should be considered descriptive rather
than causal.
Hoxby’s (1996) work was the first to pay explicit atten-
tion to the possibility that unobserved determinants of dis-
tricts’ unionization status may also determine resource al-
location and district productivity. As we describe in greater
detail below, Hoxby leveraged a combination of difference-
in-differences and instrumental variables approaches (pre-
viewed to some extent a decade earlier in the Eberts and
Stone (1984) and Woodbury (1985) studies) as an empiri-
cal solution to this theoretical problem. A number of other
scholars (e.g. Lott & Kenny, 2013; Lovenheim, 2009; West &
Mykerezi, 2011) have since followed suit, while other recent
work (Brunner & Squires, 2013; Winters, 2011) has exploited
variation in local or regional labor markets to identify union
effects on wages. We discuss these approaches in our survey
of the evidence below.
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of
and what we need to learn, Economics of Education Review (2015)
4.2. Measuring contract strength
The discussion above highlights how research about
the impacts of teachers’ unions has improved in the
identification strategies since the earliest studies undertaken
in the 1970s. However, even the more recent studies largely
consider unionization in a binary sense: districts are either
unionized or they are not; teachers either are members of
their union or not. Studies that attempted to measure the
impacts of unions on student achievement and other impor-
tant education outcomes during this time period were able to
measure union impacts as a function of whether a district did
or did not have a collective bargaining agreement, a registered
teachers’ union, or in some cases the proportion of teachers
who were unionized in a district. For instance, the Eberts and
Stone studies measured unionization as a simple dichoto-
mous indicator of union versus non-union school districts in
New York state in the 1970s. Similarly, using national survey
datasets from the National Center on Educational Statistics
(NCES), other work by Eberts and Stone (1986, 1987) mea-
sured unionization as a function of district’s collective bar-
gaining status. Similar measures of unionization were em-
ployed by nearly all researchers tackling questions of union
impacts around this time (e.g., Argys & Rees, 1995; Milkman,
1997; Nelson & Rosen, 1996; Register & Grimes, 1991). These
dichotomous indicators of union status fit the context at the
time, when unions were just entering the scene.
At about the same time, researchers with sufficient data
began to expand on union measures, focusing on the propor-
tion of teachers at the state and/or local level at three levels of
union activity: operating under a collective bargaining agree-
ment; under a meet and confer agreement (by which unions
can meet and consult with district administrators, but results
of these conferences are not binding); and without CBA pro-
tections (Kleiner & Petree, 1988; Kurth, 1987). Hoxby (1996),
used a stricter definition of unionization than much of the
previous work at that time, measuring unionization as dis-
tricts with at least 50% union membership, that reported that
they collectively bargained and that had in place a collec-
tively bargained contract. In addition, she instrumented for
unionization using the timing of the passage of laws explic-
itly enabling collective bargaining and those allowing teacher
agency shops and union shops.
Many of the more recent studies about unions have fo-
cused on the methodological challenges associated with at-
tempting to measure union strength, focusing on the activity
that gives unions the most power at the local level: the col-
lective bargaining agreement (CBA) itself. These studies have
worked to assess contract “restrictiveness” (variously called
“strength” or “determinativeness”). Four sets of researchers
have attempted to use statistically-based approaches to com-
pile multiple CBA items into a single measure of contract
restrictiveness. Eberts (1983) and Eberts and Stone (1984)
were the first to do so, using a Guttman scaling technique
to generate a measure of contract strength based on the dif-
ficulty of negotiating each of a set of 18 items within New
York union contracts in the late 1970s. Moe (2009) gener-
ated a measure of California school districts’ contract re-
strictiveness, taken from the 1998–1999 school year, using
a factor analytic approach that incorporates a selection of 41
contract items into 15 factors. Strunk and Reardon (2010)
teachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16 11
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
expand on these techniques using a larger set of 639 pro-
visions from California contracts in place during the 2005–
2006 school year, selected objectively rather than due to any
particular rationale regarding their likelihood of impacting
district and school operations. Strunk and Reardon’s (2010)
measure uses a Partial Independence Item Response (PIIR)
Model to model the entire contract as a function of a contract-
specific latent level of restrictiveness. This method has since
been replicated in Washington State by Goldhaber and col-
leagues, who have confirmed its internal validity as a mea-
sure of contract restrictiveness or strength (Goldhaber, Lav-
ery, Theobald, D’Entremont, & Fang, 2013). In later work,
Strunk and Grissom (2010) also confirmed the external va-
lidity of Strunk and Reardon’s (2010) PIIR-based measure
of contract restrictiveness in a study that shows that con-
tract strength is associated with school board-reported union
strength.
4.3. Summary of developments
As discussed in the preceding paragraphs, a number of re-
searchers have begun to address either the identification or
measurement problems highlighted in the preceding discus-
sion. Given persistent challenges concerning the availability
of data, few papers have been able to address both the iden-
tification and measurement difficulties. Because of the rela-
tively nascent stage of research on the latter, it is no surprise
that the majority of studies that examine relationships be-
tween contract determinativeness and relevant predictors or
outcomes lack strong causal inference. As more scholars con-
tinue to delve into issues of teachers’ unions and contracts
and build longitudinal and cross-state datasets, ideally more
work will emerge that can identify cause-and-effect relation-
ships while recognizing considerable variation in bargaining
agreements themselves. As we note below, many of the policy
changes associated with teaching and teacher unionization
may indeed provide empirical leverage to this end. The po-
tential for these future developments notwithstanding, the
current state of research is far better able to inform policy
discussions and future research about union impacts in ways
that it was not in Goldhaber’s 2006 review. In the section
below we discuss the current evidence for what we know
about the impact of teachers unions on expenditures and
achievement.
5. Unions and district outcomes: what does the evidence
say?
5.1. Expenditures and wages
The preponderance of the evidence available over the
years of union scholarship generally indicates that average
teacher salaries and payments for fringe benefits may be
higher in unionized districts, with estimates of union salary
advantages at least as high as 5% (Baugh & Stone, 1982;
Brunner & Squires, 2013; Cowen, 2009; Duplantis, Chandler,
& Geske, 1995; Eberts, 1983; Eberts & Stone, 1984, 1986;
Freeman, 1986; Hoxby, 1996; Kaspar, 1970; Lott & Kenny,
2013; Winters, 2011; Zigarelli, 1996), with some studies
showing considerably larger differences. At the very least,
unionization may raise overall levels of district spending by
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of t
and what we need to learn, Economics of Education Review (2015)
2–15%, depending on the study, even if salaries themselves
are not notably increased (Easton, 1988; Kleiner & Petree,
1988; Lovenheim, 2009). Variations in these figures may be
partly explained by variations in the distribution of bargain-
ing districts within a given state (Zwerling & Thomason, 1995)
or region (Chambers, 1977), or by the overall budget flexibil-
ity of the district (Gallagher, 1979).
Although many of the early studies in the field provide
preliminary evidence about the existence of a relationship,
more recent work more precisely and accurately identifies
the magnitude and significance of these relationships. As
discussed above, although earlier studies have raised the
question of endogenous relationships between unionization
and outcomes of interest, the effort to mitigate such con-
cerns began in earnest with Hoxby’s (1996) study of union
rent-seeking. Hoxby exploits changes to state laws affect-
ing unionization between 1960 and 1992 in both difference-
in-differences and instrumental variable estimates of union
impacts on district expenditures, finding union effects on
salaries between 1 and 5%. However, employing a strat-
egy based on within-state variation in the timing of union
certification election in a number of Midwestern districts,
Lovenheim (2009) found few overall differences in overall
district spending—especially in the long run—and no impact
on teacher pay associated with historical adoptions of bar-
gaining agreements over time. More recently, Brunner and
Squires (2013) exploit within-state and within-local labor
market variation in district size (a proxy for union strength),
finding as much as a 4.5% increase in teacher salaries due
to union strength and a 3.6% higher salary premium paid
to experienced teachers with a masters degree. This is sim-
ilarly the case in California: Rose and Sonstelie (2010) em-
ploy a sample of 771 districts in the 1999–2000 school year
and find that the 10 largest districts in the sample have base
teacher salaries that are approximately 7% higher and expe-
rience premiums about 11% higher than the average district.
In addition, these districts have teacher–pupil ratios that are
about 6% smaller than the average-sized California school
district.
Regardless of absolute level of pay, unionization does ap-
pear to promote compensation systems that reflect rent-
seeking behaviors. Arguing that teacher salaries are at least
partly a function of neighboring district salaries, Winters
(2011) employs a generalized method of moments (GMM)
estimator in which salaries in neighboring districts are instru-
mented with distance-weighted averages of explanatory de-
mographics and staffing characteristics from neighboring dis-
tricts. Depending on the measure of union activity, Winters
finds little evidence of union-related differences in salaries
for beginning teachers, but gains to salaries of experienced
teachers—between 18 and 28%. This can be read as evidence
of rent-seeking given strong evidence in the field that gains
in experience—at least for very senior teachers—is not as-
sociated with improvements in teacher performance (e.g.
Chingos & Peterson, 2011; Clotfelter, Ladd, & Vigdor, 2006;
Rivkin, Hanushek, & Kain, 2005; Rockoff, 2004). Along the
same lines, Grissom and Strunk (2012) find that unionized
school districts favor salaries that reward veteran teachers
above novices, even when such salary schemes do not appear
to produce increases in student achievement outcomes. Pro-
viding further evidence that teachers’ unions promote com-
eachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
12 J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
pensation mechanisms that reward characteristics that are
not known to be associated with student performance, West
and Mykerezi (2011) use OLS models alongside robustness
checks based on Hoxby’s instrumental variables approach to
show that unionization leads to compensation on the basis of
credentialing and experience, but reduces the probability of
pay-for-performance adoption by as much as 20 percentage
points. Goldhaber, DeArmond, Player, and Choi (2008) sim-
ilarly find unionization reduces the likelihood that districts
employ pay-for-performance compensation structures. All of
these salary structures promoted in unionized districts may
limit district capacity to recruit teachers via higher salaries for
higher ability (Figlio, 2002). Indeed, Hoxby and Leigh (2004)
argue that, due to a union-related compression of the salary
scale in which pay is not determined by job performance
but rather by experience and education credits, highly qual-
ified teachers face high opportunity costs to entering the
profession.
Only three studies have directly examined how contract
strength—at least as defined by specific provisions or the
determinativeness of district CBAs—are associated with dis-
trict spending and resource allocation (Eberts, 1983; Eberts
& Stone, 1984; Strunk, 2011). The most recent work from
California shows a one standard deviation increase in CBA
strength is associated with a 13% increase in total expendi-
tures (Strunk, 2011). This finding is remarkably consistent
with earlier work that shows that unionized school districts
have 8–15% higher operating expenditures (Eberts & Stone,
1986; Hoxby, 1996).
In addition, all three studies show that contract strength
or restrictiveness is associated with the ways districts spend
money. Eberts (1983) and Eberts and Stone (1984) use data
from New York State and show that contract strength and
the existence of specific provisions—especially those regu-
lating reductions in force and dismissal procedures—are as-
sociated with the ways in which districts allocate resources.
Specifically, they spend more money on instruction, bene-
fits, salaries, and average salary but less on other areas of the
budget that they speculate may not be preferred by teachers.
Conversely, Strunk (2011), using data from California nearly
three decades later, shows that districts with more restrictive
CBAs have greater overall expenditures, but that this spend-
ing does not seem to be driven by the increased allocation
of resources to teachers’ salaries or benefits. Rather, the in-
crease in total spending is partly driven by increased funds
for administrators’ salaries, classified personnel benefits and
instruction-related functions that include training, support
and evaluation of classroom teachers. In particular, Strunk
(2011) shows that districts with contracts that are one stan-
dard deviation more restrictive than the mean California CBA
spend more on administrator salaries, effectively giving these
administrators a 3.6% raise, or an increase of $3824 in aver-
age administrator salaries. Although districts with stronger
CBAs spend more on instruction-related services, such that
a one standard deviation increase in contract strength is as-
sociated with a 0.41% increase in spending on such services
(equivalent to $38 per student in the 2005–2006 school year),
districts with stronger contracts spend less on textbooks and
other instructional materials. Districts with CBAs that are one
standard deviation more restrictive than the mean spend ap-
proximately $54,000 less than the median district on school
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of
and what we need to learn, Economics of Education Review (2015)
board-directed activities, which include a range of activities
including negotiations with labor unions.
Altogether, the research that explores the relationship
between unions or contracts and fiscal outcomes suggests
that unionized districts and districts with more restrictive
contracts spend more money, and spend it differently. The
majority of studies find that unionized districts have higher
spending and specifically higher spending on teachers’
salaries, and especially salaries for veteran teachers. In addi-
tion, research that examines the restrictiveness of contracts
in unionized districts finds that, once unionized, districts with
more powerful unions spend more overall, although differ-
ent studies point to different drivers of costs. Interestingly,
the most recent research in this space indicates that con-
tract strength may no longer be associated with increased
teacher salaries, as much as other elements of district spend-
ing (Strunk, 2011).
5.2. Student outcomes
Using a variety of datasets and methodological ap-
proaches, scholars have attempted to discern the relationship
between unions and student outcomes (e.g. Carini, Powell, &
Steelman, 2000; Eberts & Stone, 1984, 1986, 1987; Hoxby,
1996; Kurth, 1987; Kleiner & Petree, 1988; Lovenheim, 2009;
Moe, 2009; Nelson & Rosen, 1996; Register & Grimes, 1991;
Strunk, 2011; Strunk & McEachin, 2011). Similar to the work
described above that explores the impact of unions on ex-
penditures, research on the relationship between unions and
student outcomes has grown more attentive to the poten-
tial for endogenous relationships between these two sets
of variables. One early view of the evidence suggests that
unionized school districts perform better on average while
students at the tails of the performance distribution have
lower test scores and higher dropout rates (Eberts & Stone,
1984, 1987). These studies estimate that achievement scores
are as much as 4% higher in unionized districts on average,
though students at both tails of the achievement distribu-
tion may not benefit from this average positive impact. In a
later study, Argys and Rees (1995) also find evidence of such
an inverted-U shaped relationship to achievement, but find
that larger class sizes for such students in such schools may
explain some of this difference. Three studies with the most
rigorous methodologies paint different pictures. In particular,
the Hoxby (1996) study discussed above shows that union-
ized, and “stronger” unionized districts have higher dropout
rates by nearly 2 percentage points, indicating that students
at the lower tail of the performance distribution are harmed
by teacher unionization. Similarly, Lott and Kenny (2013) ex-
ploit differences in state law—in their case, whether states
had Right-to-Work requirements—as an exogenous source
of variation in unionization efforts (operationalized as union
dues per student and union expenditures per student), es-
timating between a 2.2–3.7 point drop in student math or
reading proficiency rates. On the other hand, Lovenheim’s
(2009) study based on differences in the timing of district
union certification elections found that adopting union con-
tracts had no significant impact on student outcomes. When
taken in conjunction with the research reviewed above that
shows that unionized districts spend more, likely due to the
provision of higher salaries and better working conditions for
teachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16 13
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
teachers, it seems that the rent-seeking hypothesis may be
justified.
Recently, a small number of studies have explored the
relationship between the determinativeness of CBAs in con-
straining administrator autonomy and student outcomes—all
in California (Moe, 2009; Strunk, 2011; Strunk & McEachin,
2011). Strunk (2011) and Strunk and McEachin (2011) find
that more restrictive contracts are negatively associated with
both an aggregate district-level measure of student achieve-
ment (the Academic Performance Index, API) and graduation
rates. Specifically, a one standard deviation increase in CBA
restrictiveness is associated with approximately a 6.5% de-
crease in API and with a 1.3% decrease in district graduation
rates. They also show that contract strength is positively as-
sociated with the likelihood that schools and districts are in
Program Improvement and at higher levels of Program Im-
provement under the No Child Left Behind Act, such that a
district with a contract that is 1 SD more restrictive has a
7.7% greater likelihood of failing to achieve NCLB targets and
being placed in Program Improvement. In addition, Strunk
and McEachin (2011) show that the negative relationship be-
tween contract strength and student achievement is ampli-
fied in schools and districts that have higher proportions of
minority, low-income and low-achieving students. However,
while Strunk (2011) does not find evidence of a relation-
ship between CBA strength and two-year growth in student
achievement (API), Moe (2009) finds that stronger contracts
are associated with smaller API growth over the ensuing five
years. His results indicate that a one standard deviation in-
crease in contract strength is associated with 0.24 and 0.32
of a standard deviation decreases in five-year API growth for
elementary and secondary schools, respectively. In addition,
Moe (2009) finds that the relationship between CBA strength
and achievement growth varies greatly with district size. In
large districts, contract strength is significantly and substan-
tially associated with achievement growth, whereas in small
districts this relationship is close to zero. The discrepancy
in results between Strunk (2011) and Moe (2009) may oc-
cur because the two authors’ studies differ in the sample of
districts and CBAs included, the methods they use to assess
the relationship between contract strength and achievement
growth, and the measures of contract restrictiveness them-
selves. Given these differences in methods, measures and
sample, and the focus on contracts in only one state, more
research is clearly needed on this topic.
In sum, the research that explores the relationship be-
tween unions or contracts and achievement outcomes sug-
gests that unions may be rent-seeking, although the degree
to which this is the case is unclear. Hoxby (1996) and oth-
ers find that unionized districts see lower student achieve-
ment and higher dropout rates, but it is as yet unclear the
impacts of stronger unions on achievement. Meanwhile, the
studies cited above suggest that there is a negative relation-
ship between contract restrictiveness and average student
outcomes (e.g., Moe, 2009; Strunk, 2011; Strunk & McEachin,
2011). It remains to be seen whether CBA strength drives
lower achievement or if stronger CBAs simply exist in lower-
achieving districts. The current state of research does not
allow us to draw certain conclusions about the direction-
ality of CBAs’ impact on resource distribution or student
achievement.
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of t
and what we need to learn, Economics of Education Review (2015)
6. Union influence on state and federal policy
Although our focus in this discussion has been on studies
that either directly or indirectly consider models of unions as
rent-seekers within school districts, we acknowledge an en-
tirely separate body of work—largely from political science—
stressing the role unions may play in setting state and, in-
creasingly, federal educational policy. These policies in turn
may shape many of the district outcomes prioritized in the
rent-seeking framework. Indeed the political influence of
teachers’ unions, as one of the leading scholars on the topic
has argued, “may be even more consequential than the power
they wield in collective bargaining” (Moe, 2011, p. 275). If
identification and measurement pose continued challenges
to studies of union impacts on district resource allocation or
productivity, the literature on political influence remains for
the most part even further behind on this score. Descriptive
evidence may be instructive, however, particularly in a pol-
icymaking context in which district expenditures, especially
on salaries, and district performance have increasingly drawn
the attention of decision-makers at higher jurisdictional
levels.
Three general themes are apparent in this line of work.
The first stresses, as noted in the introduction, teachers’
unions’ enormous political and lobbying expenditures on
political candidates and union-friendly causes, which have
been among the highest contributors to state and federal
election candidates (Moe, 2011; Winkler et al., 2012). The
NEA and AFT combined spent more than $59 million on fed-
eral elections between 1989 and 2010—more than any other
organizational contributor—and 95% of these contributions
went to Democratic efforts (Moe, 2011). Other evidence in-
dicates that unions similarly influence elections to state leg-
islatures and governorships, as well as ballot initiatives di-
rectly pertaining to reform agendas (Hess & Leal, 2005; Lott &
Kenny, 2013; Moe, 2011), with $24 million spent on statewide
candidates and $37 million on ballot initiatives in 2008
alone.
The second theme has centered on teachers’ unions’ role
in setting policy, and particularly in opposing education re-
forms (e.g. Hartney & Flavin, 2011; Moe, 2011). The rationale
for teachers’ unions’ championing the status quo and block-
ing reforms stems from their relative position of advantage
in the current operation of school districts and their inherent
motivation to protect their own job interests (Moe, 2001a,
2003a, 2003b, 2006a, 2006b, 2011, 2014). The evidence on
unions’ role in shaping No Child Left Behind, while limited,
follows this pattern. The law itself says little about collective
bargaining, per se, except for a short provision that stipu-
lates that the new provisions cannot override locally agreed
contracts. Although the national unions—and in particular
the NEA—took credit in communication with its members
for vigorously defending that provision, some observers have
suggested that union influence was actually overstated, and
that the focus on explicit provisions pertaining to bargain-
ing shifted union efforts away from the more substantive
changes regarding teacher quality (Manna, 2006). During the
law’s implementation stage, both the NEA and AFT lobbied—
apparently successfully—to prevent an unfriendly regulatory
interpretation of the bargaining provision by the U.S. Depart-
ment of Education (Manna, 2006).
eachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
14 J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
The final point of emphasis in studies of union politi-
cal power is located within school districts and may pro-
vide some explanation, albeit non-systematic, for the way
unions achieve their rent-seeking objectives outlined above.
At the local level, where bargaining occurs, unions can
influence negotiations not simply by stressing particular
contract provisions, but by determining at least in part the
priorities of the school district itself. As Moe (2005, 2006b,
2006c) and Hess and Leal (2005) have shown, unions are ac-
tive players in the election of local school board officers who
oversee them. Strunk and Grissom (2010) lay out two likely
venues for teachers’ unions to influence school board elec-
tions: through campaign activity in favor of supported candi-
dates, such as providing union endorsements and working on
behalf of favored candidates to fundraise and mobilize voters,
and through mobilizing union members themselves to vote
in board elections.
Extant research suggests that unions are effective at both
sets of activities (Hess & Leal, 2005; Moe, 2006a). Even subse-
quent to election, the policy preferences of union-endorsed
candidates are better aligned (by a factor of 0.52 standard de-
viations on survey responses) with union preferences (Moe,
2005). Once school boards are seated, the evidence suggests
that those with more members traditionally sympathetic
to teachers (such as former educators themselves) negoti-
ate contracts that limit administrative flexibility, while more
flexible contracts are negotiated by boards with members
less aligned with union interests (such as Republican mem-
bers and those with ties to the business community; Strunk
and Grissom 2010).
7. Current policy debates: a path for future research
Public debates over teachers’ unions tends to divide the
question in normative terms: are teachers’ unions “good” or
“bad” for American education? The literature reviewed above
suggests that the answer to this question is not as simple as
it once was perceived to be. Teachers’ unions, though fac-
ing new challenges to their priorities, continue to be forceful
actors in local, state and national education policy. Whether
unions are rent-seeking, and the results of any rent-seeking
behavior, may be less important today than are the results
of the activities unions undertake and the policies they en-
dorse and/or block. The intent of this article is to update our
knowledge about the impacts of teachers unions themselves
and the ways that we might start to examine the charac-
teristics of unions and their actions that are more germane
to today’s policy context. In so doing, we hope to shift the
discussion away from the normative question about unions
towards more substantive questions about union strength
and actions.
As policymakers increasingly propose and enact meaning-
ful changes to policies that teachers’ unions have long pro-
tected, implement reforms that unions have historically op-
posed, and limit unions’ ability to garner political resources,
the groundwork will be laid for more fruitful avenues of re-
search on not only the impacts of unionization, but the im-
pacts of unions’ activities and strength on relevant outcomes.
In particular, recent changes to state laws provide multiple
opportunities for further study. State law represents the pri-
mary authority not only for labor relations but for education
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of
and what we need to learn, Economics of Education Review (2015)
itself, and heavily influence the flexibility with which dis-
tricts can set local, context-specific regulations and contracts.
Recent state-level changes to teacher-related policy provide
important sources of variation that scholars may use to im-
prove the methodological rigor of studies of union effects
on district operation. Future studies may follow in the foot-
steps of earlier work (e.g., Hoxby, 1996; Lovenheim, 2009;
Lott & Kenny, 2013) that has already exploited differences in
statewide laws concerning unions over time to introduce pre-
sumably exogenous variation in local conditions that might
affect union strength and district level outcomes.
Another point of entry into research on teachers’ unions
might occur in the ways scholars measure union activity and
contract strength. We have shown how scholars have recently
exploited cross-sectional variation in teacher contracts to
expand counterfactual scenarios beyond binary union/non-
union distinctions or simple union size measures to a range
of union influence and strength (e.g., Goldhaber et al., 2013;
Goldhaber, Lavery, & Theobald, 2014; Moe, 2009; Strunk,
2011; Strunk & Grissom, 2010; Strunk & McEachin, 2011).
The provisions of these contracts may themselves be chang-
ing in response to perceived and actual challenges to ar-
eas that unions have long prioritized. Further longitudinal
study of contract restrictiveness and the provisions included
in contracts will enable greater identification of the causal
effects of policy change and contract strength on educa-
tional outcomes. In addition, as more and more researchers
employ increasingly sophisticated ways to measure union
strength, it will become possible for scholars to pursue cross-
state analyses that capitalize both on these technical im-
provements and state-level variation in policy and policy
shifts.
States that have seen the enactment or the attempted
enactment of policy reforms intended to curb union power
and/or remove union-advocated protections also offer fertile
ground for qualitative explorations of the teachers’ unions’
role in these policies. In addition, in instances where there
are multiple states changing specific types of laws in similar
ways (e.g., the tying of teacher evaluation to student achieve-
ment, tying tenure/promotion to teacher evaluations, and the
implementation of Right-to-Work laws), researchers have the
opportunity to assess if indicators of union power (e.g., money
spent in state elections, the degree to which previous state
education codes favor teachers’ unions’ preferences, and col-
lective bargaining agreement restrictiveness or the inclusion
of specific regulations in CBAs) are associated with the uptake
of such policies or change as a result of the enactment of such
policies.
In this dynamic policymaking context, we conclude not
simply with the usual call for more research, but for more
research of a certain kind. The variation in union strength
identified in recent literature, the new policy experiments oc-
curring in states across the country, and the sheer availability
of large administrative datasets that link individual students
to teachers in school across the country now allow a new field
of highly focused questions that link educational outcomes
to rules, regulations and conditions directly attributable to
union efforts. We do not expect the results of this work to
end all controversy on the topic, but perhaps the debate may
shift from one that assigns blame to one that identifies par-
ticular, evidence-based solutions.
teachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16 15
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
References
Abood, et al. v. Detroit Board of Education 431 U.S. 209 (97 S.Ct.
1782, 52 L.Ed.2d 261). Downloaded from http://www.law.cornell.edu/supremecourt/text/431/209 on September 18, 2014.
Argys, L. M., & Rees, D. I. (1995). Unionization and school productivity: A
reexaminiation. Research in Labor Economics, 14, 49–68.Bascia, N. (1994). Unions in teachers’ professional lives: Social, practical and
intellectual concerns. New York: Teachers College Press.Baugh, W. H., & Stone, J. A. (1982). Teachers, unions and wages in the
1970s: Unionism now pays. Industrial and Labor Relations Review, 35(3),368–376.
Brunner, E. J., & Squires, T. (2013). The bargaining power of teachers’ unions
and the allocation of school resources. Journal of Urban Economics, 76,15–27.
Bryant, J. (2014) ‘Stupid, absurd, non-defensible’: New NEA president LilyEskelsen García on the problem with Arne Duncan, standardized tests
and the war on teachers Salon.com http://www.salon.com/2014/07/30/stupid_absurd_non_defensible_new_nea_president_lily_eskelsen_
garcia_on_the_problem_with_arne_duncan_standardized_tests_and_
the_war_on_teachers/ Retrieved 8/21/14.Carini, R. M., Powell, B., & Steelman, L. C. (2000). Do teacher unions hinder ed-
ucational performance?: Lessons learned from state SAT and ACT scores.Harvard Educational Review, 70(4), 437–467.
Chambers, J. G. (1977). The impact of collective bargaining for teachers onresource allocation in public school districts. Journal of Urban Economics,
4(3), 324–339.
Chingos, M. M., & Peterson, P. E. (2011). It’s easier to pick a good teacherthan to train one: Familiar and new results on the correlates of teacher
effectiveness. Economics of Education Review, 30(3), 449–465.Clotfelter, C. T., Ladd, H. F., & Vigdor, J. L. (2006). Teacher-student matching
and the assessment of teacher effectiveness. Journal of Human Resources,41(4), 778–820.
Cowen, J. M. (2009). Teacher unions and teacher compensation: New evi-
dence for the impact of bargaining. Journal of Education Finance, 35(2),172–193.
Duplantis, M. M., Chandler, T. D., & Geske, T. G. (1995). The growth and impactof teachers’ unions in states without collective bargaining legislation.
Economics of Education Review, 14(2), 167–178.Easton, T. (1988). Bargaining and the determinants of teacher salaries. Indus-
trial and Labor Relations Review, 41(2), 263–278.Eberts, R. (1983). How unions affect management decisions: Evidence from
public schools. Journal of Labor Research, 4(3), 239–247.
Eberts, R. W., & Stone, J. A. (1984). Unions and public schools: The effect ofcollective bargaining on American education. Lexington, MA: Lexington
Books.Eberts, R. W., & Stone, J. A. (1986). Teacher unions and the cost of public
education. Economic Inquiry, 24, 631–643.Eberts, R. W., & Stone, J. A. (1987). Teacher unions and the productivity of
public schools. Industrial & Labor Relations Review, 40(3), 354–363.
Estlund, C. & Forbath, W.E. (2014, July 2). The war on workers: The supremecourt ruling on Harris v. Quinn is a blow for unions. The New York Times.
Fensterwald, J. (2014, September 16). Lawsuit challenges teachers’ compul-sory dues. EdSource.org. Downloaded from http://edsource.org/2014/
lawsuit-challenges-teachers-compulsory-dues/67486#.VBsx_fldWSpon September 18, 2014.
Figlio, D. N. (2002). Can public schools buy better-qualified teachers? Indus-
trial and Labor Relations Review, 55(4), 686–699.Fowles, J., & Cowen, J. (2014). In the union now understanding public sector
union membership . Administration & Society.Freeman, R. B. (1986). Unionism comes to the public sector. Journal of Eco-
nomic Literature, 24, 41–69.Freeman, R. B., & Han, E. (2012). The war against public sector collective
bargaining in the US. Journal of Industrial Relations, 54(3), 386–408.
Gallagher, D. G. (1979). Teacher negotiations, school district expenditures,and taxation levels. Educational Administration Quarterly, 15(1), 67–82.
Goldhaber, D. (2006). Are teachers unions good for students? In J. Hannaway,& A. Rotherham (Eds.), Collective bargaining in education (pp. 141–158).
Cambridge, MA: Harvard Education Press.Goldhaber, D., DeArmond, M., Player, D., & Choi, H.-J. (2008). Why do so few
public school districts use merit pay? Journal of Education Finance, 33(3),
262–289.Goldhaber, D., Lavery, L., & Theobald, R. (2014). My end of the bargain are
there cross-district effects in teacher contract provisions? ILR Review,67(4), 1274–1305.
Goldhaber, D., Lavery, L., Theobald, R., D’Entremont, D., & Fang, Y. (2013,April-June). Teacher collective bargaining in Washington: Assessing
the internal validity of partial independence item response measures
of contract restrictiveness. Sage Open, 1–16. http://classic.sgo.sagepub.com/content/3/2/2158244013489694.full.pdf+html.
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of t
and what we need to learn, Economics of Education Review (2015)
Grissom, J. A., & Strunk, K. O. (2012). How should school districtsshape teacher salary schedules? Linking student performance to pay
structure in traditional salary schemes. Educational Policy, 26(5),663–695.
Hartney, M., & Flavin, P. (2011). From the schoolhouse to the statehouse:Teacher union political activism and US state education reform policy.
State Politics & Policy Quarterly, 11(3), 251–268.
Hess, F. M., & Leal, D. L. (2005). School house politics: Expenditures, inter-ests and competition in school board elections. In W. G. Howell (Ed.),
Beseiged: School boards and the future of education politics (pp. 228–253).Washington, DC: Brookings Institution Press.
Hill, P. T. (2006). The costs of collective bargaining agreements and relateddistrict policies. In J. Hannaway, & A. J. Rotherham (Eds.), Collective bar-
gaining in education (pp. 89–109). Cambridge, MA: Harvard EducationPress.
Hoxby, C. M. (1996). How teacher unions effect education production. Quar-
terly Journal of Economics, 111, 671–718.Hoxby, C. M., & Leigh, A. (2004). Pulled away or pushed out? Explaining
the decline of teacher aptitude in the United States. American EconomicReview, 94(2), 236–240.
Kaspar, H. (1970). The effects of collective bargaining on public schoolteachers’ salaries. Industrial and Labor Relations Review, 113(3), 671–
718.
Kerchner, C. T. (2003). The modern guild: The prospects for organizing aroundquality in public education. In D. Lipsky, & J. Brock (Eds.), Transforming
union: Industrial relations research association series. Urbana, IL: Univer-sity of Illinois Press.
Kerchner, C. T. (2004). Organizing around quality: The union struggle toorganize mind workers. In R. Henderson, W. Urban, & P. Wolman (Eds.),
Teacher unions and educational policy: Retrenchment and reform (pp. 187–
222). Boston: Elsevier/JAI.Kerchner, C. T., & Koppich, J. E. (1993). A union of professionals: Labor relations
and educational reform. New York: Teachers College Press.Kerchner, C. T., & Koppich, J. E. (1999). Organizing the other
half of teaching. In L. Darling-Hammond, & G. Sykes (Eds.),Teaching as the learning profession. San Francisco: Iossey-
Bass.
Kerchner, C. T., & Koppich, J. E. (2000). Organizing around quality. In T. Love-less (Ed.), Conflicting missions? Teacher unions and educational reform.
Washington, DC: Brookings Institution Press.Kerchner, C. T., Koppich, J. E., & Weeres, J. G. (1997). United mind workers:
Teachers and unions in the knowledge society. San Francisco: Iossey-Bass.Kerchner, C. T., Koppich, J. E., & Weeres, J. G. (1998). Taking charge of quality:
Teachers, unions and educational reform. San Francisco: Iossey-Bass.
Kleiner, M. M., & Petree, D. L. (1988). Unionism and licensing of public schoolteachers: Impact on wages and educational output. In R. B. Freeman,
& C. Ichniowski (Eds.), When public sector workers organize. Chicago, IL:University of Chicago Press.
Kurth, M. M. (1987). Teachers’ unions and excellence in education: An anal-ysis of the decline in SAT scores. Journal of Labor Research, 8(4), 351–
367.
Lott, J., & Kenny, L. W. (2013). State teacher union strength and studentachievement. Economics of Education Review, 35, 93–103.
Lovenheim, M. F. (2009). The effect of teachers’ unions on education produc-tion: Evidence from union election certifications in three midwestern
states. Journal of Labor Economics, 27(4), 525–587.Manna, P. (2006). Teachers unions and no child left behind. In J. Hannaway,
& A. Rotherham (Eds.), Collective bargaining in education (pp. 159–180).Cambridge, MA: Harvard Education Press.
Medina, J. (2014, January 31). Fight over effective teachers shifts to court-
room. New York Times.Milkman, M. (1997). Teachers’ unions, productivity, and minority student
achievement. Journal of Labor Research, 18(1), 137–150.Moe, T. M. (2001a). A union by any other name. In P. E. Peterson (Ed.), Choice
and competition in American education. New York: Rowman and Little-field.
Moe, T. M. (2003a). Politics, control and the future of school accountability.
In P. E. Peterson, & M. West (Eds.), Leave no child behind? The politics andpractices of school accountability. Washington, DC: Brookings Institution
Press.Moe, T. M. (2003b). The politics of the status quo. In P. E. Peterson (Ed.), Our
schools and our future: Are we still at risk?. Stanford, CA: Hoover Press.Moe, T. M. (2005). Bottom-up structure: Collective bargaining, transfer rights,
and the plight of the disadvantaged. Department of Political Science. Stan-
ford University.Moe, T. M. (2006a). Political control and the power of the agent. Journal of
Law, Economics and Organization, 22(1), 1–29.Moe, T. M. (2006b). Union power and the education of children. In J. Hannway,
& A. Rotherham (Eds.), Collective bargaining in education (pp. 229–256).Cambridge, MA: Harvard Education Press.
eachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006
16 J.M. Cowen, K.O. Strunk / Economics of Education Review 000 (2015) 1–16
ARTICLE IN PRESSJID: ECOEDU [m3Gdc;March 23, 2015;17:12]
Moe, T. M. (2006c). The union label on the ballot box. Education Next, 6(3),58–66.
Moe, T. M. (2009). Collective bargaining and the performance of publicschools. American Journal of Political Science, 53(1), 156–174.
Moe, T. M. (2011). Special interest: Teachers unions and America’s public schools.Washington, DC: Brookings Institution.
Moe, T. M. (2014). Teacher unions and American education reform: The
power of vested interests. In J. A. Jenkis, & S. M. Milkis (Eds.), The pol-itics of major policy reform in postwar America. Cambridge: Cambridge
University Press.Moe, T. M. (Forthcoming). Vested interests and political institutions. Political
Science Quarterly, 130(1).National Council on Teacher Quality (2014a) State Policy Database,
www.nctq.org; Retrieved 9/24/14.National Council on Teacher Quality (2014b) 2013 State Teacher Policy Year-
book National Summary. www.nctq.org Retrieved 9/24/14.
Nelson, F.H., & Rosen, M. (1996). Are teacher unions hurting Americaneducation? A state-by-state analysis of the impact of collective bar-
gaining among teachers on student performance. Institute for Wis-consin’s Future. http://files.eric.ed.gov/fulltext/ED404746.pdf accessed
3.09.15.Register, C. A., & Grimes, P. W. (1991). Collective bargaining, teachers, and
student achievement. Journal of Labor Research, 12(2), 99–109.
Rivkin, S. G., Hanushek, E. A., & Kain, J. F. (2005). Teachers, schools, andacademic achievement. Econometrica, 73(2), 417–458.
Rockoff, J. E. (2004). The impact of individual teachers on studentachievement: Evidence from panel data. American Economic Review,
247–252.Rose, H., & Sonstelie, J. (2010). School board politics, school district size and
the bargaining power of teachers’ unions. Journal of Urban Economics, 67,
438–450.Sanes, M., & Schmitt, J. (March 2014). Regulation of public sector collective
bargaining in the states. Washington D.C.: Center for Economic and PolicyResearch.
Strunk, K. O. (2011). Are teachers’ unions really to blame? Collective bargain-ing agreements and their relationships with district resource allocation
and student performance in California. Education Finance and Policy, 6(3),
354–398.
Please cite this article as: J.M. Cowen, K.O. Strunk, The impact of
and what we need to learn, Economics of Education Review (2015)
Strunk, K. O. (2012). Policy poison or promise?: Exploring the dual nature ofCalifornia school district collective bargaining agreements. Educational
Administration Quarterly, 48(3), 506–547.Strunk, K. O. (2014). The role of collective bargaining agreements in the
implementation of teacher quality reforms: Perils and possibilities. InR. Hess, & M. McShane (Eds.), Teacher quality 2.0: Will today’s reforms
hold back tomorrow’s schools? Harvard Education Press.
Strunk, K. O., & Grissom, J. A. (2010). Do strong unions shape district policies?Collective bargaining, teacher contract restrictiveness, and the political
power of teachers’ unions. Educational Evaluation and Policy Analysis,32(3), 389–406.
Strunk, K. O., & McEachin, A. (2011). Accountability under constraint: Therelationship between collective bargaining agreements and schools’ and
districts’ performance under no child left behind. American EducationalResearch Journal, 48(4), 871–903.
Strunk, K. O., & Reardon, S. (2010). Measuring union strength: A partial in-
dependence item response approach to measuring the restrictiveness ofteachers’ union contracts. Journal of Educational and Behavioral Statistics,
35(6), 629–670.Vergara v. State of California, et al. Superior Court of the State of California
County of Los Angeles Case No.: BC484642.West, K. L., & Mykerezi, E. (2011). Teachers’ unions and compensation: The
impact of collective bargaining on salary schedules and performance pay
schemes. Economics of Education Review, 30, 99–108.Winkler, A. M., Scull, J., & Zeehandelaar, D. (2012). How strong are U.S. teacher
unions?. Thomas B. Fordham Institute.Winters, J. V. (2011). Teacher salaries and teacher unions: A spatial econo-
metric approach. Industrial and Labor Relations Review, 747–764.Woodbury, S. A. (1985). The scope of bargaining and bargaining outcomes in
the public schools. Industrial and Labor Relations Review, 38(2), 195–210.
Wright, et al. v. New York (2014). Supreme Court for the State of NewYork Complaint, 7/28/14, http://nylawyer.nylj.com/adgifs/decisions14/
072914summons.pdf accessed 3.09.15.Zigarelli, M. A. (1996). Dispute resolution mechanisms and teacher bargain-
ing outcomes. Journal of Labor Research, 17(1), 135–148.Zwerling, H. L., & Thomason, T. (1995). Collective bargaining and the
determinants of teachers’ salaries. Journal of Labor Research, 16(4),
467–484.
teachers’ unions on educational outcomes: What we know
, http://dx.doi.org/10.1016/j.econedurev.2015.02.006