Assessing the Compensation of Public School Teachers 19282337242

8/3/2019 Assessing the Compensation of Public School Teachers 19282337242

1/25

A Report

of The Heritage Center

for Data Analysis

A Report

of The Heritage Center

for Data Analysis

214 Massachusetts Avenue, NE Washington, DC 20002 (202) 546-4400 heritage.org

NOTE: Nothing written here is to be construed as necessarily reflecting the views of The Heritage Foundation

or as an attempt to aid or hinder the passage of any bill before Congress.

A FREE ENTERPRISE PRESCRIPTION:UNLEASHING ENTREPRENEURSTO CREATE JOBS

JAMES SHERK, KAREN A. CAMPBELL, PH.D., AND JOHN L. LIGON

CDA10-09 December 14, 2010

Assessingthe CompensAtionofpubliC-sChool teAChers

JAson riChwine, ph.D. AnD AnDrew g. biggs, ph.D.

CDA 11-03 november 1, 2011

In Partnership With


2/25


3/25

CDA 11-03 November 1, 2011

1

Executive Summary

The teaching proession is crucial to Americas soci-ety and economy, but public-school teachers shouldreceive compensation that is neither higher nor lowerthan market rates. Do teachers currently receive theproper level o compensation? Standard analyticalapproaches to this question compare teacher salariesto the salaries o similarly educated and experiencedprivate-sector workers, and then add the value oemployer contributions toward ringe benets. These

simple comparisons would indicate that public-schoolteachers are undercompensated. However, comparingteachers to non-teachers presents special challenges notaccounted or in the existing literature.

First, ormal educational attainment, such as adegree acquired or years o education completed, is nota good proxy or the earnings potential o school teach-ers. Public-school teachers earn less in wages on aver-age than non-teachers with the same level o education,but teacher skills generally lag behind those o otherworkers with similar paper qualications. We show

that: The wage gap between teachers and non-teachers

disappears when both groups are matched on anobjective measure o cognitive ability rather than onyears o education.

Public-school teachers earn higher wages than pri-vate-school teachers, even when the comparison islimited to secular schools with standard curriculums.

Workers who switch rom non-teaching jobs toteaching jobs receive a wage increase o roughly 9

percent. Teachers who change to non-teaching jobson the other hand, see their wages decrease by roughly 3 percent. This is the opposite o what one wouldexpect i teachers were underpaid.

Second, several o the most generous ringe benetsor public-school teachers oten go unrecognized:

Pension programs or public-school teachers aresignicantly more generous than the typical private

sector retirement plan, but this generosity is hiddenby public-sector accounting practices that allowlower employer contributions than a private-sectorplan promising the same retirement benets.

Most teachers accrue generous retiree health benetsas they work, but retiree health care is excluded romBureau o Labor Statistics benets data and thus requently overlooked. While rarely oered in the private sector, retiree health coverage or teachers isworth roughly an additional 10 percent o wages.

Job security or teachers is considerably greater than

in comparable proessions. Using a model to calculate the welare value o job security, we nd that jobsecurity or typical teachers is worth about an extra 1percent o wages, rising to 8.6 percent when considering that extra job security protects a premium paidin terms o salaries and benets.

We conclude that public-school teacher salaries arecomparable to those paid to similarly skilled privatesector workers, but that more generous ringe benetsor public-school teachers, including greater job secu

Assessingthe CompensAtionof

publiC-sChool teAChers

JasonRichwine, Ph.D., anD anDRewG. BiGGs, Ph.D.


4/25

THE HERITAGE CENTER FOR DATA ANALYSIS

2

rity, make total compensation 52 percent greater thanair market levels, equivalent to more than $120 billionovercharged to taxpayers each year. Teacher compensa-tion could thereore be reduced with only minor eectson recruitment and retention. Alternatively, teachers

who are more eective at raising student achievementmight be hired at comparable cost.

Jason Richwine, Ph.D., is Senior Policy Analyst inthe Center or Data Analysis at The Heritage FoundationandAndrew G. Biggs, Ph.D., is a Resident Scholar at the

American Enterprise Institute.


5/25

CDA 11-03 November 1, 2011

3

The compensation o public-school teachers is aperennial issue in American public policy, as the need tobalance budgets collides with the desire to recruit andretain quality teachers. Since the operation o publicschools is typically the largest local government expen-diture, decisions about education unding oten domi-nate local elections. Teachers unions or proessionalassociations operate in all 50 states, and they supportlocal candidates who do political battle with opponents

concerned about rising property taxes. In 13 states,teachers are allowed to strike to secure a more avorablecontractan option unavailable to police and reght-erswhich raises the political stakes considerably.1

Pensions and other teacher benets are oten undedat the state level, leading to periodic clashes betweenteachers unions and state governments over budgetpriorities. The recent recession has reduced tax receiptsand orced states to make painul decisions aboutspending cuts or tax increases, urther ampliying theseconficts.

One o the more prominent battles over teacher com-pensation has occurred in New Jersey, where GovernorChris Christie has required teachers and other public

workers to increase their health care contribution rom0 percent to 1.5 percent o salary.2 Reorms proposed by

Wisconsin Governor Scott Walker, which led to Democratic lawmakers feeing the state to deny the legislaturea quorum, reduced teachers benets and limited theirability to bargain collectively. Florida recently requiredstate employees to contribute 3 percent o their salaryto their pension plan, which had been unded exclusively by taxpayers. Florida teachers led a lawsuit in

response.3

Much o the debate over teacher compensation iscouched as a question about how much the government can aord at the present timethat is, how muchis it able to pay teachers? Reormers argue that reductions in wages and benets or teachers are a budgetary necessity, while teachers argue that savings shouldbe ound elsewhere. But a dierent question is otenignored in the education debate, one that is independent o any governments current scal situation: Howmuch should teachers be paid?

No one doubts the signicance o high-quality teachers to the school system and to the economy in generalbut even the most important public workers should

Assessingthe CompensAtionofpubliC-sChool teAChers

JasonRichwine, Ph.D.,anD anDRewG. BiGGs, Ph.D.

1. Allegheny Institute or Public Policy, Teacher Strikes in Pennsylvania, January 2011, at http://www.alleghenyinstitute.org/education/teacherstrikes.html (September 29, 2011).

2. Chris Megerian, N.J. Police, Fireghter Unions Sue to Stop Pension Reorm Laws, NJ.com, April 23, 2010, at http://www.nj.com/news/index.ss/2010/04/nj_police_reghter_unions_s.html (September 24, 2011).

3. Tami Luhby, Florida Teachers Sue State in Pension Dispute, CNN.com, June 20, 2011, at http://money.cnn.com/2011/06/20/news/economy/forida_teachers_pension_lawsuit/index.htm?iid=Lead(September 24, 2011).


6/25


4

be paid at a level commensurate with their skillsnomore, no less. Ideally, i a teachers skills are worth $X inthe private marketplace, that teacher should be paid $Xby the government. This system would ensure that thepublic gets qualied and competent teachers withoutoverpaying or the privilege. How close is the United

States to that ideal? There is much dispute about howto measure teachers skills and the total compensationthey receive or them.

This report is a comprehensive assessment o sala-ries, benets, and job security or public-school teach-ers, intended to resolve disputes over whether teachersas a group are overpaid or underpaid. We nd thatpublic-school teachers receive compensation about 52percent higher than their skills would otherwise garnerin the private sector, and we discuss the implications othis act or education policy in general.

SALARIES

We begin by introducing the human capital modelo wages, using it to compare teacher and non-teachersalaries with adjustments or skill dierences betweenthe two groups. We then show how the standard anal-ysis ocused on ormal educational attainment can bemisleading in the case o teachers, whose years o edu-cation may not be as valuable in the marketplace asor workers in other occupations. We also comparepublic-school teacher salaries to private-school-teachersalaries, and we conclude by examining how the aver-

age teachers wage changes when he or she leaves theproession.

Initial Wage Analysis. Public-school teachersreceive higher wages than the average worker, but itwould be simplistic to conclude that one group is over-paid relative to another group based only on averagesalaries. Because groups may have dierent character-istics that aect their productivity, proper comparisonsrequire controlling or worker skills.

Comparing Jobs versus Comparing Workers. Thereare two major ways that researchers attempt to account

or skill dierences between teachers and non-teachers.

One is to identiy occupations that have similar skilrequirements as teaching, and to then examine averagesalaries in those occupations as reported by the Bureauo Labor Statistics (BLS). For example, Jay Greene andMarcus Winters o the Manhattan Institute recentlyreported that public-school teachers nationwide earn

an hourly salary 11 percent higher than other proessional specialty and technical workers, and 36 percenthigher than white-collar workers in general.4 The majorstrength o this approach is that it uses compensationdata as reported by employers, which is generally morereliable than surveys o individual employees.

Use o the BLS occupational wage data has a number o drawbacks, however. First, the choice o occupations to compare with teaching is subjective. Evensystematic methods o identiying comparable occupations rely on assumptions about job characteristics that

are hard to veriy. Second, occupational comparisonscannot ully account or dierences in earnings-relatedworker characteristics, such as race, gender, marital status, and experience, which may be distributed dierently among seemingly comparable occupations. At theederal level, at least, it has been shown that the government hires and promotes employees who have lessexperience and education than private-sector workersin similar occupations.5

In contrast, the human capital model ocuses onthe education, experience, and other skills that employ

ees bring to a job. In this view, i two jobs are comparable in terms o other actors, such as risk and workconditions, the human capital that employees bring totheir jobs should account or dierences in their compensation. The Congressional Budget Oce (CBO) hasdeemed the human capital approach the dominant theory o wage determination in the eld o economics.6

Our preerred dataset or making worker-to-workercomparisons is the Annual Social and Economic Supplement o the Census Bureaus Current PopulationSurvey (CPS). The CPS is one o the best datasets oranalyzing salaries because o its large sample size andrich set o control variables. It has the drawback o rely

4. Jay P. Greene and Marcus A. Winters, How Much Are Public School Teachers Paid? Manhattan Institute Civic Report No. 50,January 2007, at http://www.manhattan-institute.org/html/cr_50.htm (September 24, 2011).

5. Melissa Famulari, Whats in a Name? Title Infation in the US Federal Government, University o Texas-Austin Working Paper,2002. See also Congressional Budget Oce, Reducing Grades o the General Schedule Workorce, September 1984.

6. Congressional Budget Oce, Comparing the Pay o Federal and Nonederal Law Enorcement Ocers, August 2005, at http://www.cbo.gov/tpdocs/66xx/doc6619/08-23-LawEnorcementPay.pd (September 24, 2011).


7/25


5

ing on sel-reports rather than employer data rom theBLS, but it allows us to control or a much larger set ohuman capital traits.7

Summer Vacation. It is well known that public-school teachers have shorter average work years thanmost other job holders. A teacher who receives a givensalary or nine months o work is clearly better com-pensated than someone who earns the same salary ora ull years work.

But just how lengthy summer vacations really are,and to what extent they benet teachers, are majorpoints o contention among education researchers.Some analysts have tried to avoid the issue by usingweekly salaries reported in the CPS during the schoolyear. Theoretically, this allows them to compare teacherand non-teacher earnings per week o paid work, with-out regard to summer vacation. In many cases, however,

weekly salaries in the CPS are simply annual salariesdivided by 52 weeks.8 Using weekly salaries withouturther adjustment or summer vacation will upwardlybias teacher compensation.

We treated teachers as ull-year workers or thepurposes o comparing salaries. We rst comparedthe annual CPS salaries o teachers and non-teacherswithout taking actual time worked during the year intoaccount. The salary variable that we used included onlysalaries that teachers received rom their primary job,thereby excluding earnings rom a potential second job

taken during summer vacations. We then included thevalue o summer vacation in the benets section, show-ing how dierent estimates o paid leave aected overallcompensation.

Method and Results. A representative sample o about100,000 American households participate in the CPSsupplement each year. Because a much smaller num-ber o respondents are teachers, we combined the past10 years o CPS data, rom 2001 through 2010. This

enlarged the sample size by an order o magnitude andgave our results greater precision.

We limited the analysis to respondents who workedull-time jobs in either the private sector or in the stateand local sector as teachers. (All ederal and non-teacher public workers were excluded.) Private workers wereincluded only i they were employed or all 52 weeksin the past year, while teachers were included whoreported working 35 weeks or more.9 These parametersallowed us to treat both teachers and non-teachers asull-year workers, with paid leave or each group to beadded in the benets section.

We employed ordinary least squares regression, withthe log o annual earnings as the outcome variable andthe ollowing control variables: usual hours worked perweek, experience,10 experience-squared, years o education, rm size (six categories), immigration status

state o residence, race, gender, marital status, and yearindicators to account or infation. We also includedinteraction terms: experience x education, experiencesquared x education, marital status x gender, and genderx race.

The explanatory variable most relevant or our purposes is one that indicates whether someone is a public-school teacher.11 By including a teacher indicatorvariable in the regression, we were able determine howthe average public-school teachers salary compares tothe salary o a private worker with the same earnings

related characteristics.Linear regression analysis can tell us how a one

unit change in a particular worker characteristic aectswages when all other characteristics are held constantFor example, Table 1 indicates that an extra year o education is associated with an 11.8 percent increase in theaverage workers wages, assuming other characteristicsstay the same. As the table shows, the eect on wages oteaching in public school, all else being equal, is -19.3

7. Sel-report errors are more common or special income categories (such as sel-employment and welare) than or the wage andsalary data that we use. See Daniel H. Weinberg, Income Data Quality Issues in the CPS, Monthly Labor Review (June 2006), pp.3845, at http://www.bls.gov/opub/mlr/2006/06/art4ull.pd(September 24, 2011).

8. Michael Podgursky and Ruttaya Tongrut, (Mis)-Measuring the Relative Pay o Teachers, Education Finance and Policy, Vol. 1, No.4 (Fall 2006), pp. 425440.

9. We also dropped workers with imputed wages, workers who reported less than $10,000 in annual earnings, and workers who alloutside the 18-to-64 age range.

10. Experience = age - years o education - six.

11. A public-school teacher is someone who is employed by a state or local government and lists his or her occupation as elementaryschool teacher or secondary education teacher. We opted not to include preschool, kindergarten, and special education teachersin our main analysis, but including them does not signicantly change the results.


8/25


6

percent. In other words, public-school teachers receivesalaries that are 19.3 percent lower than non-teacherswho have the same observable skills.

We should be careul not to draw strong conclusionsabout the wages o a single occupation rom a regres-sion such as this one. Unobserved ability dierencesand work conditions could still infuence the observedwage gap. I we added an indicator or architects to theregression, or example, we would nd that architectsreceive a wage premium over similarly skilled work-ers. Yet ew people would immediately conclude thatarchitects are overpaid. In the next several sections,we explore to what degree work conditions and unob-served ability dierences may be aecting the observedteacher wage penalty. We place particular emphasis onthe act that years o education is not a good measure oteacher quality, either within the teaching population orin comparing teachers to members o other proessions.

Problems with Education as a Measure o Teacher Quality. In standard wage regressions, such as theone we have presented, education is measured eitheras years o schooling or as level o degree obtained. Theimplicit assumption is that educations eect on utureearnings is consistent across elds o studya degree

in French literature is treated as equivalent to a degreein engineering or biochemistry. While that may seemlike a serious limitation on salary comparisons in general, it is usually not an issue when members o eachgroup work in a wide variety o occupations requiringdiverse educational backgrounds. Dierences in educational quality at any given educational level tend toaverage out in heterogeneous groups.

However, a problem exists when comparing the salaries o a single occupational group to the salaries ocomparably educated workers in the general popula

tion. A large proportion o teachers have bachelors ormasters degrees in education. One study estimated that72 percent o elementary school teachers and almosthal o secondary school teachers were educationmajors.12 In addition, more than two out o three teachers received their highest degree (typically a masters) inan education-related eld.13 Because elds o study orteachers are considerably less diverse than or the general population, relying solely on dierences in educational quantity (years o schooling) may mask importantdierences in educational quality between teachers andnon-teachers. In other words, the standard education

variable may not be an accurate measure o teacher skill

Consider the two main ways in which education isrelated to worker productivity. First, education indicates basic knowledge and experience. Someone whohas attended medical school, or example, will usuallybe better at diagnosing diseases than someone who hasonly a bachelors degree. For teachers, however, thereis little evidence that advanced degrees improve perormance at all. Careully constructed value-addedmodels tend to show that teacher eectiveness varies

Wage Regression Results: Teachers vs.Non-Teachers

Sources: Authors calculations based on data from the U.S. CensusBureau, 20012010 Current Population Surveys.

Table 1 CDA 11-03 heritage.org

Independent Variable Coefficient

Hours worked per week 1.4%

Experience in years 2.5Education in years 11.8

Foreign-born 8.2

Married 22.6

Black 18.7

Hispanic 10.3

Female 15.1

Public school teacher 19.3

Observations 495,038

Adjusted r-squared 0.41

12. These gures apply to teachers whose highest degree is a B.A. or B.S. Michael F. Thompson and Tanya J. Hall, Do Teachers HaveEducation Degrees? Matching Fields o Study to Popular Occupations o Bachelors Degree Graduates, In Context, Vol. 11, No. 1(JanuaryFebruary 2010), at http://www.incontext.indiana.edu/2010/jan-eb/article2.asp (September 24, 2011).

13. The gure in 2001 was more than our in ve. Low response rates to the NEAs survey may make their estimates imprecise.National Education Association, Status o the American Public School Teacher 2005-2006, March 2010, at http://www.nea.org/assets/docs/HE/2005-06StatusTextandAppendixA.pd(October 21, 2010).


9/25


7

considerably, but resume characteristics such as yearso schooling, certications, and experience beyond therst ew years o teaching show little to no relationshipto student achievement.14

A second way that education usually refects produc-tivity is that educational success can be a proxy or cer-tain personality traitsperseverance and intelligence,or examplethat are dicult to measure directly.Employers are oten impressed with a job applicantwho has completed years o challenging academicwork, even i the specic eld o study is not directlyrelevant to the job. This is sometimes reerred to as the

sheepskin eect, where holding a degree signals skillsor attributes distinct rom those acquired while obtain-ing the degree.

Elementary school teachers typically possess a degreein elementary or early childhood education. Secondary

school teachers, being more specialized, usually majorin an academic eld or in how to teach a particular eld,such as Math or School Teachers.15

Do these education degrees earned by teachers carrya strong sheepskin eect, as other degrees do? It isunlikely. Given the relative lack o rigor o educationcourses, many teachers have not aced as demandinga college curriculum as other graduates. More than 50years ago scholars were already noting the low grad-ing standards in university education departments.The Journal o Higher Education reported in 1960 that

32 percent o students in education courses receivedA grades, compared to just 16 percent in businesscourses.16

A hal century later, the situation is little changed.Economist Kevin Rask collected data on course gradesat a Northeastern liberal arts college rom 5,000 stu-dents who graduated between 2001 and 2009.17 Out othe 20 academic departments included in Rasks data,education awarded the highest grades.

Although Rasks data come rom only one college, hisresults are consistent with a larger study o three stateuniversities in the Midwest.18 Economist Corey Koederecently analyzed grade point averages (GPAs) at Indiana University, the University o Missouri, and MiamiUniversity o Ohio. He ound that education majors

had substantially higher GPAs than students majoringin the hard sciences, social sciences, or the humanitiesEducation majors at Indiana University, or examplehad an average GPA o 3.65, while math, science, andeconomics students averaged 2.88.

These results could mean that education majors arethe brightest and hardest-working college studentsHowever, as the ollowing sections show, it is morelikely that education courses are simply easier thansay, chemistry and math classes, which eature strictergrading.

Koedel suggests that grading standards are so lowin the education eld that distinguishing student perormance in education classes has become dicult toimpossible. He links his ndings to evidence that student eort decreases as grading standards decreaseimplying that education majors learn less than otherstudents as a consequence. I so, holding a degree ineducation should signal less knowledge than a degreein an alternate subject.

Aside rom the motivation and perseverance thaearning a degree may imply, succeeding in college also

suggests that a person possesses a certain raw intelligence useul or many dierent tasks. However, yearso education is a poor proxy or cognitive ability whencomparing teachers to non-teachers. Although teachersas a group score above the national average on intelligence tests, their scores all below the average or othercollege graduates. This implies that, to the extent thatcognitive ability aects earnings independently o education, ordinary wage regressions may overestimateteacher earnings relative to those o other proessions.

14. Stephen G. Rivkin, Eric A. Hanushek, and John F. Kain, Teachers, Schools, and Academic Achievement, Econometrica, Vol. 73,No. 2 (March 2005), at http://www.econ.ucsb.edu/~jon/Econ230C/HanushekRivkin.pd(September 24, 2011).

15. It is unlikely that the latter kind o teaching degree is as rigorous as the pure academic eld, but this is a matter o some dispute.National Science Board, Elementary and Secondary Education: Teacher Quality, Science and Engineering Indicators 2004,pp. 2430, at http://www.ns.gov/statistics/seind04/c1/c1s5.htm#c1s5l3 (September 24, 2011).

16. Robert M. Weiss and Glen R. Rasmussen, Grading Practices in Undergraduate Education Courses,The Journal o HigherEducation, Vol. 31, No. 3 (March 1960), pp. 143149.

17. Kevin Rask, Attrition in STEM Fields at a Liberal Arts College: The Importance o Grades and Pre-Collegiate Preerences,Economics o Education Review, Vol. 29, No. 6 (December 2010).

18. Corey Koedel, Grading Standards in Education Departments at Universities, University o Missouri Working Paper, June 2011,at http://i.bnet.com/blogs/education-major-study.pd (September 24, 2011).


10/25


8

Although the College Board is reluctant to say exact-ly what the SAT measures, it is essentially an IQ test.19In 2010, the College Board asked students taking theSAT about their intended college major. Students whoindicated that education was their intended majorearned a combined math and verbal score o 967, about

0.31 standard deviations below the average o 1,017,meaning the 38th percentile in a standard normal dis-tribution.20 In contrast, students intending to major inengineering had average combined SAT scores o 1,118.In a standard wage regression, however, individualswith bachelors degrees in education and engineeringare assumed to possess the same human capital andshould earn the same wages, all else being equal.

College graduates who take the Graduate RecordExamination (GRE) also indicate their intended eld ostudy when they sit or the test. During the past aca-

demic year, students who planned to study elementaryor secondary education in graduate school scored 0.13standard deviations below average on the GRE. I alleducation-related elds are countedincluding spe-cial education, early childhood education, and curricu-lum developmentthe dierence was 0.35 standarddeviations.21

Not all education majors go on to become teachers,nor do all teachers major in education, but even activeteachers exhibit low cognitive ability compared to othercollege graduates. A recent study examined scores on

the ACTan alternative to the SAT oten used in theMidwesto students who attended public collegesand universities in Missouri. Four-year graduates whobecame public-school teachers scored 0.23 standarddeviations below our-year graduates who did notbecome teachers.22

More broadly, the National Longitudinal Survey o Youth (NLSY) includes scores on the Armed ForcesQualication Test (AFQT), a cognitive test battery simi-

lar to a ull-scale IQ test.23 Teacher scores on the AFQTlag behind other ull-time workers with the same education levels by about 0.25 standard deviations.24 Thesedata indicate that, on average, teachers do not have thesame cognitive skills as other college graduates.

As both a direct measure o acquired knowledge andan indirect measure o innate ability, teacher educationdoes not compare well to education in other elds. Theresult is that years o education could be a highly misleading measure o teacher skill.

Wage Regression with IQ. Like the CPS, the NLSYis a rich dataset that includes the earnings-related variables needed to run wage regressions. Though its sample is small compared to the CPS, the NLSY providesthe opportunity to test the hypothesis that education isa misleading measure o teacher skill. We constructeda salary model with all o the usual controls except or

education, which we replaced with the AFQT score. Ithe teacher salary penalty remains large when an objective measure o cognitive ability is used in place o education, then we can reject our hypothesis.

As the name implies, the NLSY is a longitudinal survey that began interviewing young adults between theages o 13 and 21 in 1979. Each successive wave trackedthe growth and development o the original interviewees. In order to obtain the largest sample o NLSY teachers possible, we combined data rom the 1990 through1994 waves. These years contain wage data modeled

on the CPS collection procedure, and they come beoredropouts rom the sample became a large issue.

Using the log o wage-indexed hourly wages as thedependent variable, we ran regressions with and without education and AFQT as explanatory variables. Allo the regressions used the ollowing control variablesexperience, experience-squared, establishment size(our categories), rm size (two categories), immigrationstatus, region o residence, residence in a metropolitan

19. Meredith C. Frey and Douglas K. Detterman, Scholastic Assessment or g? The Relationship Between the Scholastic AssessmentTest and General Cognitive Ability, Psychological Science, Vol. 15, No. 6 (2004), pp. 373-378.

20. College Board, Total Group Prole Report, 2010, at http://proessionals.collegeboard.com/prodownload/2010-total-group-prole-report-cbs.pd(September 24, 2011).

21. Educational Testing Service, Guide to the Use o Scores 201011, at http://www.ets.org/s/gre/pd/gre_guide.pd (September 24,2011).

22. Michael Podgursky, Ryan Monroe, and Donald Watson, The Academic Quality o Public School Teachers: An Analysis o Entryand Exit Behavior, Economics o Education Review, Vol. 23 (2004), pp. 507518.

23. For the psychometric properties o the AFQT, see R. Darrell Bock and Elsie G. J. Moore,Advantage and Disadvantage: A Prole oAmerican Youth (Hillsdale, NJ: Lawrence Erlbaum Associates, 1986).

24. Based on the authors calculations using the sample o teachers described in the next section, Wage Regression with IQ.


11/25


9

area, race, gender, marital status, and an indicator vari-able or public-school teacher. We omitted state indica-tors, year indicators, and interaction terms because othe smaller sample size.

Table 2 shows how teacher salaries change depend-ing on whether education or AFQT is included in theregression. The rst row is the standard regressionbased on our CPS analysis in the previous section: Years

o education are controlled or, but AFQT is not. Thestandard regression shows a teacher salary penalty o12.6 percent.

The second row includes both education and AFQTin the same regression. The impact on teacher wages issmall: The penalty decreases by less than two percent-age points. The third row again includes AFQT but nowomits education. With this specication, the change isdramatic: The teaching penalty is gone, replaced by astatistically insignicant premium.

How to interpret these results? On the one hand,

the dierence in IQ between teachers and other collegegraduates by itselhas only a small eect on estimates othe teacher penalty. As the second row indicates, teach-ers with both the same education and AFQT score asother workers still receive 10.7 percent less in wages.

However, as we have shown, education is a mislead-ing measure o teacher skills in several ways. In addition

to the IQ dierence between teachers and non-teachersthe education major is among the least challengingelds o study, and years o education subsequentlyhave little to no eect on teacher quality. This suggeststhat eliminating education as a control variable and letting AFQT alone account or skills (as in the third row)

may provide the most accurate wage estimates.Replacing education with an objective measure o

skills eliminates the observed teacher penalty, indicating that non-teachers with the same education as a typical teacher will likely have more applicable skills. Weemphasize that a job is not necessarily less importantor less challenging when the credentials or it are easierto obtain. Indeed, eective teachers are highly valuableto society and the economy.25 Eective teaching doeshave skill requirements, including patience and empathy, which disqualiy many people rom the proession

Our point is that traditional skill measures do not allowor a air salary comparison o teachers to non-teachers

Public-School Teachers versus Private-SchoolTeachers.

A better assessment o teacher pay may come romchanging the comparison group. Rather than comparepublic-school teacher salaries with the salaries o allprivate workers, the salary comparison could compareteachers in the public sector with teachers in the pri-vate sector. This approach largely avoids the problemsdescribed in the previous section. We need not worry

about how the value o education changes by occupa-tion when everyone involved in the comparison is ateacher.

A teacher-to-teacher comparison also helps to eliminate intangible work-related actors rom the analysis. Ithere are certain aspects o teaching that are particularlyrustrating (or rewarding) relative to other occupationsa higher (or lower) salary or teachers may be requiredas a compensating dierential. By limiting both the reerence and comparison group to teachers, whatever salary dierences we observe are less likely to be driven bythese intangible actors.

Table 3 shows the results o a salary regression thatwas limited to elementary and secondary teachers. Thedata and model specication was the same as our earlier CPS analysis, except now the variable o interestis public school, which shows the salary premium associated with working or a public school rather than a

Teacher Wage Regressions withEducation and Armed ForcesQualication Test (AFQT)

* Significant at 95 percent level.

Note: All three regressions include controls for experience, experience-squared, establishment size, firm size, immigration status, region of resi-dence, residence in a metropolitan area, race, gender, and marital status.

Source: Authors calculations based on data from the U.S. Departmentof Labor, Bureau of Labor Statistics,19901994 waves of the 1979National Longitudinal Survey of Youth.


Row Key Controls Included Teacher WageEffect (%) R-squaredEducation AFQT

1 Yes No 12.6* 0.31

2 Yes Yes 10.7* 0.33

3 No Yes 0.6 0.29

25. Eric A. Hanushek, The Economic Value o Higher Teacher Quality, National Bureau o Economic Research Working PaperNo. 16606, December 2010.


12/25


10

private school. With all observable skills held constant,public-school teachers nationally earn 9.8 percent morein salaries than private-school teachers.

The change in relative salaries or public-schoolteachers is drastica 19.3 percent penalty when com-pared to private workers in general, as opposed to a9.8 percent premium when compared to private-schoolteachers. The choice o comparison group obviouslymakes a substantial dierence in the results, but whichchoice is preerable? We have already discussed theproblems with using private workers as the comparisongroup, and how they may be mitigated by using privateteachers. However, the use o private teachers as thecomparison group has its own problems.

Public and private schools do not necessarily per-orm the same unctions. While public schools mustprovide a general education to motivated and unmoti-vated students alike, private schools have more variedmissions. Elite private schools oten eature specialized

curriculums directed at select groups o students. Con-sequently, the skills needed or private-school teachingmay be somewhat dierent than the skills needed or

teaching in a public school. The religious mission inparticular o many private schools could also aect theattractiveness o teaching. For example, teachers in sectarian schools oten consider their work to be part otheir religious service, meaning they may accept belowmarket salaries.

I public schools ceased to exist, the demand or secular private schools with a standard curriculum wouldpresumably increase, changing the private labor market or teachers. For this reason, the salaries o privateschool teachers in general do not necessarily refect the

true labor market value o public-school teachers. Tobetter estimate the true market value, we can restrict thecomparison even urther. By excluding private schoolswith religious missions or non-standard curriculumsthe comparison o teacher salaries in the public andprivate sector becomes clearer.

Michael Podgursky, an economist at the University o MissouriColumbia, has used the Departmento Educations School and Stang Survey to perormthat analysis.26 Ater controlling or education, genderregion, and metro status, Podgursky ound that publicschool teachers earn higher salaries than teachers innon-sectarian and standard-curriculum private schoolsThe salary premium or public-school teachers rangedrom 9 percent to 28 percent depending on teacherexperience, with the least experienced receiving thehighest premiums.

Salaries o Teachers Who Leave the Proession Another method o comparison examines the salarieo teachers who leave the proession. I public-schoolteachers receive lower salaries than they could otherwise earn in the private market, we would expect alarge portion o those who leave teaching to take new

jobs that pay better.

Evidence rom Georgia and Missouri, however, indicates this is not the case. According to state data romthe 1990s, just 4 percent o Georgia elementary teachers who let their jobs or a non-education eld were

earning more than the minimum teaching wage a yearater their exit. The same gure or exiting high schoolteachers was 5 percent.27 In Missouri, women who quitteaching earned just 73 percent as much as their teach

Wage Regression Results: Public SchoolTeachers vs. Private School Teachers

Sources: Authors calculations based on data from the U.S. CensusBureau, 20012010 Current Population Surveys.


Independent Variable Coefficient

Hours worked per week 0.3%

Experience in years 2.3Education in years 8.7

Foreign-born 3.7

Married 5.5

Black 4.5

Hispanic 1.2

Female 3.5

Public school teacher 9.8

Observations 21,092

Adjusted r-squared 0.35

26. Michael Podgursky, Fringe Benets, EducationNext, Vol. 3, No. 3 (Summer 2003), at http://educationnext.org/ringebenets/(September 25, 2011).

27. Benjamin Scadi, David L. Sjoquist, and Todd R. Stinebrickner, Do Teachers Really Leave or Higher Paying Jobs in AlternativeOccupations?Advances in Economic Analysis and Policy, Vol. 6, No. 1 (2006), at http://www.gpp.org/pub/education/advances.pd(September 25, 2011).


13/25


11

ing wage in their new non-teaching jobs.Male teachers who quit saw no change intheir average salaries.28

We can broaden the scope o these resultsusing the Survey o Income and Program

Participation (SIPP), along with a regressiontechnique called xed eects. The regres-sion analyses we have presented so ar havebeen cross-sectional, meaning they comparedierent workers with similar skills at a sin-gle point in time. In contrast, a xed eectsanalysis ollows the same workers over timeas they switch into and out o the teachingproession.

The major benet o this approach is thatit automatically controls or dierences in

unobserved abilitiesintelligence, motivation, empa-thy, etc.because the individual teachers carry theseattributes with them rom job to job. I the same work-ernot merely a worker with similar skillsis paidless in a non-teaching job, it is dicult to argue that heor she was underpaid as a teacher.

The SIPP is a longitudinal dataset that ollows about50,000 households over three to our years. During thisperiod, many survey participants changed jobs, withsome switching between teaching and non-teachingcareers. To bolster the number o switchers into and out

o teaching, we combined the 2001, 2004, and 2008SIPP panels. We then examined the average percent-age change in infation-adjusted monthly wages expe-rienced by workers who switch rom a non-teaching

job to a teaching job (group abbreviation: NT), a teach-ing job to a non-teaching job (TN), one teaching jobto another teaching job (TT), and, as a control, a non-teaching job to another non-teaching job (NN).29

O course, workers oten change jobs because theyhave acquired new skills that justiy a higher salary, andthe xed eects regression allows us to include controlsor observable changes in age, education, marital sta-tus, region o residence, and residence in a metropoli-tan area. Worker characteristics that do not change overtime, even those that are not directly observable, areautomatically controlled or in this analysis.

Table 4 indicates that the control group that shitsrom non-teaching jobs to other non-teaching jobsexperiences a real wage increase o only 0.5 percent

Workers who switch rom non-teaching to teachinreceive a larger increase o 8.8 percent. Teachers whochange to non-teaching jobs, on the other hand, seetheir wages decrease by 3.1 percent. In other words, theeect on wages o switching into or out o a teaching

job is precisely the opposite o what one would expecti teachers were underpaid.

Given the small number o workers switchingbetween teaching and non-teaching, these data should

not be considered precise, but they at least cast strongdoubt on the notion that teachers are underpaid inwages.

Why does the xed-eects approach show a salarypremium or teachers, while the earlier cross-sectionalregression showed a penalty? The traditional controlsin a cross-sectional regression do not adequately measure teacher skills. By ollowing the same workers overtime, xed eects models capture worker characteristics that are not directly observable. Once those previously unmeasured characteristics enter the analysis, the

large teacher wage penalty appears to become a smallpremium.

The Bottom Line on Teacher Salaries. The claimthat public-school teachers endure a salary penalty isdubious. Although the standard wage regression sug

Wage Change After Job Switch

Notes: Full-time civilian workers only. Wage changes are adjusted for changes in age, education,marital status, region, and residence in a metropolitan area, using fixed effects regression analysis.See text for details.

Source:Authors calculations based on data from the U.S. Census Bureau, 2001, 2004, and 2008panels of the Survey of Income and Program Participation.


Label

Job Change WageChange (%)

SignificanceLevel NumberFirst Job Second Job

NN Non-teaching Non-teaching 0.5 99% 24,308NT Non-teaching Teaching 8.8 99% 148

TN Teaching Non-teaching 3.1 99% 145

TT Teaching Teaching 5.4 99% 325

28. Podgursky, Monroe, and Watson, The Academic Quality o Public School Teachers.

29. Teacher here reers to public-school teachers. A small number o private-school teachers will be included in the non-teachercategory.


14/25


12

gests that teachers receive about 19 percent lower salariesthan similarly skilled ull-time workers, the regressionis fawed. Important dierences exist between teachersand workers in other occupations that are obscured bytraditional control variables.

Years o education, in particular, is a misleadingmeasure o teacher skill, both within the teaching pro-ession and between teaching and non-teaching occu-pations. The eld o education is less challenging thanother academic concentrations, and teacher educationhas little measurable eect on classroom perormance.Furthermore, teachers have signicantly lower cog-nitive ability, on average, than non-teachers with thesame level o education. Removing education rom awage regression and replacing it with a measure o rawcognitive ability appears to erase the teacher salary pen-alty. Clearly, the standard regression masks important

acts about teacher skills.Teacher-to-teacher comparisons avoid the problem

o measuring skills across occupations. Public-schoolteachers receive signicantly higher salaries than pri-vate-school teachers, even more than private teachersat secular general-education schools.

Finally, workers who switch rom non-teaching toteaching jobs increase their wages on average, whileworkers who switch jobs in the other direction seetheir wages decrease. The totality o the evidence sug-gests that public-school teachers are not underpaid in

wages by private-sector standards, and they may evenbe overpaid.

BENEFITS

Evaluating the level o benets enjoyed by public-school teachers is more challenging than analyzingtheir salaries, or two main reasons. First, some ben-ets (such as pensions) are accrued during a personsworking lie but not collected until retirement. For ben-ets that are delivered in the present, such as healthcoverage, paid leave, or taxes paid on a workers behal,the employers contribution toward these benets isa good measure o what employees actually receive.However, or pensions, retirement health benets, andother deerred compensation, the current employercontribution is a poor indicator o the actual benetsto be received. The reason is that employers oten dis-agree about what current contributions are needed to

nance the same uture benet. To accurately calculatethe value o deerred benets, it is necessary to controboth or how much the employer contributes and howthat employer calculates the required contribution. Wemake this adjustment as needed.

A second challenge to measuring benets is that nocomprehensive dataset exists that would allow the kindo direct individual comparisons presented in the previous sections on salaries. Data must be pieced together rom separate sources. Our starting point is theEmployer Costs or Employee Compensation (ECEC)survey published by the Bureau o Labor Statistics, supplemented by additional data sources as needed.

The ECEC dataset does not report benets at theindividual level; rather, benets are reported on anaggregated basis by industry, rm size, proession, andother employer characteristics. The ECEC does lis

benets specically or public-school teachers, albeit atlower levels o detail than are available or the workorce as a whole.

For simplicity, benets are reported as a percentageo wages. Table 5 shows the main benet categories orpublic-school teachers in 2010. Total employer contributions toward benets are reported as equal to 41.2percent o teacher salaries. Subcategories include:

Paid leave (6.6 percent o wages)paid vacationholidays, and sick leave;

Insurance plans (16.1 percent)health, disabilitylie, and other insurance protections;

Retirement and savings (11.1 percent)denedbenet pensions and dened-contribution pensionsand

Legally required benets (7.4 percent)employertaxes toward Social Security, Medicare, unemployment and workmans compensation insuranceamong others.

We compared teachers to workers in establishmentso 100 or more employees.30 The latter group receives

benets equal to 41.3 percent o salaries, making teacher benets appear comparable. An alternate comparisongroup based on occupation would be ull time proessionals, and relatedthis group tends to receive loweraverage benets than the private-sector 100+ categorymaking the comparison shown in this section relativelyconservative.

30. Establishment size is the number o employees at a particular location; rm size is the total number o employees in anorganization.


15/25


13

Combining similar benet levels with standard esti-mates o the teacher salary dierential, which the priorsection showed are fawed, one could easily concludethat public-school teachers are substantially under-compensated relative to private-sector workers. How-ever, there are several important ways in which the BLSdata understate the total ringe benets paid to public-school teachers:

First, the BLS data on employer contributions to

pensions ail to adjust or the public sectors moreaggressive unding rules, which imply signicantlyhigher guaranteed benets per dollar o employercontributions compared to the private sector.

Second, BLS data exclude retiree health coverage.Most teachers and other public-sector employeesreceive this coverage, but most private-sector work-ers do not.

Third, the paid leave category in the BLS data doesnot account or the shorter working year enjoyed bymost teachers.

Ater a ull accounting, benets or teachers areshown to be signicantly more generous than those paidto employees o large private-sector establishments.31

Pensions. Most evaluations o public-sector and pri-vate-sector compensation compare employer retirement

plan contributions without accounting or importantdierences in how pensions are nanced. As a resulto this error, individuals who would receive exactlythe same benets in retirement could be credited withvery dierent levels o pension compensation whileworking.

Part o the conusion stems rom the dierent typeso retirement plans. A dened-contribution (DC) plansuch as a 401(k) or a 403(b), is an account owned andinvested by the worker himsel. Many employers willmake regular contributions to DC accounts as part otheir worker retirement plans, but no specic levelo benets is guaranteed. The employers obligationbegins and ends with its annual contribution. I everyworker had a DC pension, it would be easy to comparetheir retirement benetsjust look at their employersannual contributions to the accounts.

A dened-benet (DB) plan is essentially an annuitya regular, xed sum paid to workers ater theyretire. In order to have the money needed to pay or DBpensions, employers usually set aside unds each yearto be invested. But employers do not set aside $1 orevery $1 in pension benets they must eventually payInstead, they assume a certain rate o return on theirinvestments and then contribute an amount that theyexpect to grow to the proper level needed in the uture

While the annual employer contribution to Dplans is equivalent to the retirement benet workers

receive, the employer set-aside or DB pension undingis not the same as the benet. The higher the employerassumes the rate o return will be, the lower the annual set-asides or DB pensions need to be, even as theactual retirement benet to workers stays at the sameguaranteed level.

Public-sector pensions nance their benets witha more aggressive unding strategy than private-sectorplans. A unding strategy, in simple terms, encompasseshow much the employer contributes to und pensionbenets and how those contributions are invested. An

aggressive unding strategy implies lower contributionsinvested in higher risk assets, such as stocks, privateequity, and hedge unds. Public-sector pensions invesin assets with an expected return o around 8 percentwhich allows them to make lower contributions as longas the returns come as expected. I pension investments

31. Several prior studies have erroneously assumed that the BLS ECEC data account or all o the major benets that employeesreceive, without making the adjustments we discuss in the ollowing sections. See, or example, Sylvia A. Allegretto, Sean P.Corcoran, and Lawrence Mishel, How Does Teacher Pay Compare? Methodological Challenges and Answers (Washington, DC:Economic Policy Institute, 2004).

Average Benets as a Percentageof Wages, Initial Data

Public-School Teachers and Private Workers inEstablishments of 100 Employees or More, in 2010

Source: Authors calculations using data from U.S. Department ofLabor, Bureau of Labor Statistics. Note: The BLS benefits categorysupplemental pay is omitted, as these values are included in CPSwages.


Public-SchoolTeachers

PrivateWorkers

Total benefits 41.2% 41.3%

Paid leave 6.6% 11.4%

Insurance plans 16.1% 13.3%

Retirement and savings 11.1% 5.4%

Legally required benefits 7.4% 11.3%


16/25


14

all in value, then public employersmeaning, ulti-mately, taxpayersmust increase their contributions tothe pension unds. Because the risk is borne by taxpay-ers, the average public worker is eectively guaranteedan 8 percent return.

This guarantee makes comparisons o pension plans

dicult. As a recent paper published by the Center orState and Local Government Excellence noted, contri-butions to private sector 401(k) plans and public sectordened benet plans are not comparable. The pub-lic sector contribution guarantees a return o about 8percent, whereas no such guarantee exists or 401(k)s.Thus, the public sector contribution under-states publicsector compensation.32 (Emphasis in original.)

In DC plans, which are predominant in the privatesector, all the market risk is borne by the worker. I aworker wants a guaranteed benet in retirement com-

parable to that received by a public-school teacher, hemust invest his account in guaranteed assets: U.S. Trea-sury securities, yielding around 4 percent over 20 years.

These higher returns show themselves in higherbenets. According to the Public Plans Database com-piled by the Center or Retirement Research at BostonCollege, the average teacher pension plan required anemployee contribution o 5.1 percent o wages andpaid benets equal to 1.8 percent o nal wages multi-plied by the number o years o service.33 To illustrate,imagine a teacher who retired ater 30 years o service

with an annual salary o $40,000. Her annual DB pen-sion benet would be about $20,330.34

Now consider what she might receive in a private401(k) plan. According to BLS data in Table 5, theaverage employer pension contribution was about 5.4

percent o earnings. We assume that, like the teacherpension, employees contribute 5.1 percent o their ownpay. Following the practice o the CBO, we assume a 4percent interest rate on 401(k) contributions to matchthe guaranteed nature o DB pension benets. At retirement, the total account balance would equal $96,131

sucient to purchase a lietime infation-indexed annuity paying $4,450 per year.35

Thereore, assuming equal employee contributionsthe teacher can expect to receive retirement benets that are roughly 4.5 times higher than she wouldreceive rom a typical private-sector pension. This doesnot mean that every teacher receives such generousbenets. Since pension benets tend to accrue muchaster near the end o a teaching career, a teacher whochanges jobs early in her career may even lose money.36

On average, however, public-sector pensions are sim

ply ar more generous than private-sector plans.Compound interest is said to be the most powerul

orce in the universe. In the example above, the dierence between a 4 percent return and an 8 percenreturn, compounded over a working lietime, results ina much higher DC contribution rate needed to generatethe same benets as a DB plan. A similar dynamic takesplace with private-sector DB pension plans, though private plans are not allowed to discount their liabilities asaggressively as public-sector DB pensions.

The Teachers Pension and Annuity Fund o New Jer

sey illustrates the scale o the issue. The und reports thatthe average new retiree in 2009 claimed benets at age61 and received an annual benet o around $46,500. Tonance the accrual o these benets, the und requiredtotal annual contributions equal to approximately 8percent o worker wages.37 From this, one might iner

32. Alicia H. Munnell, Jean-Pierre Aubry, Josh Hurwitz, and Laura Quinby, Comparing Compensation: StateLocal Versus PrivateSector Workers, Center or State and Local Government Excellence, September 2011, at http://www.slge.org/index.asp?Type=B_BASIC&SEC={6B5D32FD-C99D-41F7-9691-4F1B1D11452B}&DE={43B3A1F9-3BCB-4EDB-A334-7F4B183BF3B2} (September 25,2011).

33. Public Plans Database, at http://pubplans.bc.edu/pls/htmldb/?p=198:3:945369477816020::NO (September 30, 2011). These gures areor plans where the employee participates in Social Security, to allow or a more straightorward comparison o benets.

34. This assumes that her wages grew 5.5 percent each year, the average rate o total wage growth assumed by teacher plans in thePublic Plans Database. This produces average earnings over her nal three years o employment o $37,648.

35. Based on rates published by the ederal governments Thrit Savings Plan.

36. Robert M. Costrell and Michael Podgursky, Golden Handcus, Education Next, Vol. 10, No. 1 (Winter 2010), athttp://educationnext.org/golden-handcus/(October 12, 2011).

37. These gures drawn rom Teachers Pension and Annuity Fund o New Jersey: Actuarial Valuation Report,Milliman, June 30,2009, at http://www.state.nj.us/treasury/pensions/pd/nancial/09tpavaluationreport.pd(September 25, 2011). New Jersey useda discount rate o 8.25 percent or its calculations. In addition to the contributions outlined above, New Jersey must makecontributions to nance ununded liabilities rom prior years. These are not considered part o current compensation.


17/25


15

that the implicit value o pension compensation is 8percent o wages. However, a private-sector DB pensionplan promising exactly the same benets would needto set aside a much larger amount, equal to almost 19percent o worker pay. A DC plan that sought to guar-antee the same level o retirement benets would need

to set aside an even greater amount, about 27 percent osalaries. These disparate contribution rates are drivennot by the generosity o the plans benets themselves,but simply by dierent accounting rules.

Note that the question here is not whether state andlocal pensions should assume a 4 percent riskless rateo return when deciding how to value their pensionobligations, although almost all economists believe thatthey should.38 When measuring the value o benetspaid as compensation to employees, that debate is irrel-evant. Even i state and local governments could some-

how generate higher pension benets with less riskand lower costs than the private sector, those higherbenets should not be excluded rom the tally o pub-lic-sector compensation. The savings rom the allegedgovernment eciency could just as well be devoted toother government outlays or tax reduction rather thanhigher employee compensation.

To properly compare pensions, we need to adjust ordierences in how pension contributions are calculated.The basic approach is to convert employer DB contribu-tions to the amount necessary to und the same benets

using a DC account.

39

This approach ensures that thesame dollar amount o uture retirement benets willbe converted to the same level o pension compensa-tion today, regardless o whether benets are paid roma public-sector DB plan, a private-sector DB plan, or aDC plan.

Data rom the Public Plans Database or 2009 showsthat state and local pensions that include teachers havean average cost o annual benet accruals (called the

normal cost in pension parlance) o 12.4 percent owages, when discounted at the 7.9 percent averageinterest rate assumed by teacher pensions in the data

base. This normal cost is split between employers (a 6.7percent average contribution rate) and employees (5.7percent).

We rst adjusted the 12.4 percent total normal costto account or dierences in discount rates. Our principal source was a March 2011 analysis prepared orthe Florida Retirement System (FRS) by the actuarialrm Milliman, in which the normal costs o nine pension plans under the FRS were calculated using a rangeo discount rates.40 Using these gures, we calculatedan adjustment actor which equaled the normal cost o

the average FRS plan using a 4 percent discount ratedivided by the normal cost assuming a 7.9 percent discount rate, the average interest rate assumed by teacherpension plans.41

The resulting quotient, 2.94, was multiplied by thereported 12.4 percent average normal cost or teacherpensions, resulting in a total normal cost o 36.5 percent o payroll. This is the contribution to a DC pensionaccount, compounded at the government bond rate oreturn, which would generate the same guaranteedpension benet at retirement as the typical teacher pen

sion plan. From this we subtracted the average employee contribution o 5.7 percent o payroll, or a total DBpension compensation o 30.8 percent o wages.42

In addition to DB pensions, teachers receive averageemployer contributions to DC pension plans o 1.2 percent o wages, or a total employer contribution toward

38. Congressional Budget Oce, The Underunding o State and Local Pension Plans, Economic and Budget Issue Brie, May 4, 2011,at http://www.cbo.gov/doc.cm?index=12084 (September 25, 2011), and Jerey R. Brown and David W. Wilcox, Discounting Stateand Local Pension Liabilities,American Economic Review, Vol. 99, No. 2 (May 2009).

39. Our approach is explained in more detail in Andrew G. Biggs and Jason Richwine, Comparing Federal and Private Sector

Compensation, American Enterprise Institute Working Paper, June 8, 2011, at http://www.aei.org/paper/100203 (September 25,2011).

40. Robert S. DuZebe, Study Refecting Impact to the FRS o Changing the Investment Return Assumption to One o the Following:7.5%, 7.0%, 6.0%, 5.0%, 4.0% and 3.0%, Milliman, March 11, 2011.

41. The use o a 4 percent riskless discount rate is common in the academic literature on public-sector pensions and approximatesmarket yields in the mid-2009 time rame in which the normal costs rom above were calculated. However, yields on Treasurysecurities are currently well below 4 percent, meaning it would be even more expensive to provide a given dollar value oguaranteed uture retirement benets. See Robert Novy-Marx and Joshua D. Rauh, Public Pension Promises: How Big Are Theyand What Are They Worth?Journal o Finance, orthcoming.

42. In Comparing Compensation, Munnell et al. use a very similar method in their analysis o state and local employee pensions, buthey assume a riskless discount rate that is higher than the one used in the academic literature.


18/25


16

retirement and savings o 32 percent. In contrast, thetypical private-sector retirement and savings compo-nent o compensation equals only 6.2 percent o wages,even ater adjusting private-sector DB contributions ina similar manner to public-sector DB plans above.

Retiree Health Benefts. BLS benets data do

not include inormation on health coverage providedto retirees. Most public-sector employees, includ-ing school teachers, become eligible to receive healthbenets in retirement. Economists Robert Clark andMelinda Morrill note that retiree health insurance planscover virtually all ull-time public sector employees.47Retiree health benets are extremely valuable or earlyretiring public employees, who otherwise would needto acquire health coverage in the more expensive indi-

vidual insurance market until they became eligible orMedicare benets at age 65.

In the private sector, retiree health coverage is lesscommon and less generous. Data rom the MedicalExpenditure Panel Survey indicate that roughly 18 percent o employees work or employers who oer retiree

health coverage, but in many cases that coverage hasbeen rozen or restricted. A Kaiser Family Foundationsurvey reports that 29 percent o companies that reportoering retiree health coverage do not oer it to newlyhired workers.48 Other companies have rozen urtheraccumulations o retiree health benets. Thus, signicantly ewer than 18 percent actually are truly eligibleor such coverage. In addition, as the Pew Foundationhas noted, o the large private rms that do oer ben

In Illinois, where public pension unding has beenespecially controversial, the states Teachers Retire-ment System has pointed out that the average teachers

pension is only around $43,000 per year.43 Whilesignicantly higher than a typical retirees pension,the number is misleadingly low. It is not an averageor teachers retiring today; rather, it includes teacherswho retired years or decades ago who, because salarieswere lower in the past, receive lower pensions than ateacher retiring today. The 2010 actuarial report orthe Teachers Retirement System o Illinois shows, orinstance, that retirees between the ages o 55 and 59receive average annual benets o $55,893.44

Even that gure is deceptive because it includes

benets paid to individuals who worked only part otheir careers in public schools. These retirees wouldreceive lower average benets, but they may haveretirement income drawn rom another job. The 2010

Comprehensive Annual Financial Report in Illinoisshows that the average benet paid to a 60-year-oldretired teacher with 35 to 39 years o servicea ull

working careerwas $67,452.45

Illinois teachers do not pay into, and do not receive,Social Security benets, meaning that comparisons toprivate-sector workers should include both privatepensions and Social Security benets. Advocates orteachers sometimes suggest that their inability to par-ticipate in Social Security is a disadvantage. However,Social Security pays middle-income and upper-incomeworkers a below-market rate o return, generatingbenets around one-third lower than workers couldreceive by investing in sae government bonds.46 In

contrast, public pensions pay employees more thanthree times the total bond yield. By and large, teachersand other public employees benet rom not partici-pating in Social Security.

Why Do Teacher Advocates Claim that Pension Benefits Are Modest?

43. Teachers Retirement System o the State o Illinois, Teacher Pensions are too Generous, at http://trs.illinois.gov/subsections/press/

TRSIssuesUpdate.htm#teacherpensions (October 21, 2011).44. Buck Consultants, Teachers Retirement System o the State o Illinois. Actuarial Valuation o Pension Benets as o June 30,

2010, December 2010.

45. Teachers Retirement System o the State o Illinois. Comprehensive Annual Financial Report. For the scal year ended June 30,2010, December 22, 2010.

46. Oce o the Chie Actuary, Social Security Administration, Moneys Worth Ratios Under the OASDI Program or HypotheticalWorkers, July 2010.

47. Robert Clark and Melinda S. Morrill, The Funding Status o Retiree Health Plans in the Public Sector, National Bureau oEconomic Research Working Paper No. 16450, October 2010.

48. Employer Health Benets: 2008 Summary o Findings, The Kaiser Family Foundation and Health Research and EducationalTrust, at http://ehbs.k.org/images/abstract/7791.pd(September 25, 2011).


19/25


17

ets, they tend to be considerably less generous thanthose oered by state governments.49 Many privateemployers have both tightened eligibility standards andincreased cost-sharing through new ormulas or explic-it global caps on employer subsidies.50

It is possible to estimate the value o retiree healthcoverage on a system-by-system basis through disclo-sures required by the Government Accounting Stan-dards Board (GASB). According to GASB, retiree healthbenetswhich are reerred to as Other Post Employ-ment Benets (OPEB)are a part o the compensa-tion that employees earn each year, even though thesebenets are not received until ater employment hasended.51 These reports, required by GASB Rules 43and 45, measure the value o accruing retiree healthbenets as a percentage o workers salaries, just as wemeasure other benets as a percentage o salaries in

the above sections. This value, known as the normalcost o accruing benets, can be thought o as thecost or OPEB being earned by employees in exchangeor [their] services now.52 It allows adjustments to BLSdata to include the value o retiree health coverage.

The generosity o retiree health coverage varies sig-nicantly. In some cases, retired teachers are merelyallowed to buy into health coverage provided to work-ing-age teachers. This provision still counts as a subsi-dy, since it allows retired teachers to purchase coverageat the price oered to younger working-age teachers.

Some employers provide explicit subsidies toward

health coverage, and in some cases all costs are coveredby employers, meaning retired teachers ace very ewout-o-pocket costs.

Unortunately, no national-level data on retireehealth care exist, meaning that we can illustrate benets only on a case-by-case basis. The situation is evenmore complicated or teachers, since retiree health issometimes provided by individual school districts evenwhen teachers participate in a statewide pension planTable 6 provides a selection o available retiree healthdisclosures. The normal costs or the sample averages

Retiree Health Costs

Retiree Health Coverage Costs for Teachers forSelect Areas, as Percent of Salaries

Sources: See footnote 53.


Milwaukee, WI 17.3%

Nor th Carolina 12.5%Louisiana 12.2%

Hawaii 11.2%

Boston, MA 9.6%

San Francisco, CA 8.9%

Illinois 8.3%

New Jersey 7.9%

Georgia 5.3%Washington 3.7%

Ohio 3.1%

Maine 2.1%

Connecticut 1.0%

49. The Pew Center on the States, Promises With a Price: Public Sector Retirement Benets, December 2007.

50. Retiree Health Benets Examined, Kaiser Family Foundation and Hewitt Associates, December 2006, at http://www.k.org/medicare/upload/7587.pd(March 25, 2011).

51. Governmental Accounting Standards Board, Other Postemployment Benets: A Plain-Language Summary o GASB StatementsNo. 43 and No. 45, at http://www.gasb.org/project_pages/opeb_summary.pd (September 29, 2011).

52. Caliornia Department o Education, Denitions o Key Terms, February 26, 2007, at http://www.cde.ca.gov/g/ac/co/documents/gasb45attha.doc (September 29, 2011).


20/25


18

to about 8 percent o wages, which we will use as arepresentative value in our calculations.53 However, itis worth reiterating that the generosity o retiree healthcoverage varies rom place to place. In analyzing anyspecic state, city, or school district, a local value shouldbe used.

The gures presented in Table 6 refect the cost tothe government, not the ull value to employees. Lack-ing retiree health coverage, a retired public-sectoremployee would purchase coverage in the individualmarket, where costs are on average 25 percent higherthan under group coverage.54 Thus, a stated cost toemployers o 8 percent o salaries or retiree health cov-erage would have a value to employees o 10 percent otheir salaries.

Given the data problems outlined above, it is impos-sible to value private-sector retiree health coverage pre-

cisely. The coverage value is likely quite small, but wecan develop a rough estimate by multiplying the retireehealth care value or public-school teachers (10 percento wages) by an assumed private coverage rate o 18 per-cent, then urther multiplying by 70 percent to accountor employers who report oering retiree health cover-age but have rozen their plans to new hires or whohave halted accruals o new benets. This produces aprivate-sector value o 1.3 percent o wages.

Paid Leave. The BLS reports that public-schoolteachers receive paid leave equal to 7 percent o sala-

ries on average. For private-sector workers in establishments o 100 employees or more, the comparablevalue is 11.4 percent o pay, counterintuitively implying much less paid leave or school teachers. Stranger still, BLS reports that state and local employees onaverage receive paid leave equal to 12.5 percent o salaries, implying that public-school teachers receive signicantly less paid leave than other public employees

53. The ollowing are the sources or each cell in Table 6, Retiree Health Coverage Costs or Teachers or Select Areas, as Percent oSalaries. Maine: State o Maine, State & Teachers Retiree Healthcare Plan, Actuarial Valuation or Fiscal Year 2008/09 GASB45 Inormation, June 30, 2008. Louisiana: Buck Consultants, GASB 43/45 Actuarial Valuation at 7/1/2009 o The State oLouisiana Post-Retirement Benet Plan or the Oce o Group Benets, April 2010. Milwaukee: Gabriel, Roeder, Smith &Company, Milwaukee Public Schools Retiree Health Care and Lie Insurance Programs: Actuarial Valuation as o July 1, 2009,June 25, 2010. Illinois: Gabriel, Roeder, Smith & Company, Teachers Retirement Insurance Program o the State o Illinois:GASB No. 43 Actuarial Valuation Report as o June 30, 2009. Washington: Oce o the State Actuary, Washington State 2008Other Post-Employment Benets (OPEB) Actuarial Valuation Report, August 2008. North Carolina: Aon Consulting, State oNorth Carolina Report o the Actuary on the Postemployment Medical Benets Valuation, December 31, 2005. Ohio: CavanaughConsulting, Report on the Retiree Health Care Valuation o the School Employees Retirement System o Ohio: Prepared as o June30, 2010. San Francisco: Memorandum to Mayor Gavin Newsom, Members o the Board o Supervisors, rom Ben Roseneld,

Controller, Report on Retiree (Post-Employment) Medical Benet Costs, December 15, 2010. Georgia: Cavanaugh Consulting,Georgia State Employees Post-Employment Health Benet Fund & Georgia School Personnel Post-Employment Health BenetFund, Report o the Actuary on the Retiree Medical Valuations: Prepared as o June 30, 2009. Boston: The Segal Company, TheCity o Boston Actuarial Valuation and Review o Other Post-Employment Benets (OPEB) as o June 30, 2009. Connecticut:Gabriel, Roeder, Smith & Company, The Report o The Actuarial Valuation as o June 30, 2006 o the Connecticut State TeachersRetirement System Retiree Health Care Plan. New Jersey: Aon Consulting, State o New Jersey Postemployment Benets OtherThan Pension Actuarial Valuation, July 20, 2010. Hawaii: Aon Consulting, State o Hawaii Postemployment Benets other thanPensions, July 1, 2007.

54. Melinda Beeuwkes Buntin, Jos S. Escarce, Kanika Kapur, Jill M. Yegian, and M. Susan Marquis, Trends and Variability inIndividual Insurance Products, Health Aairs, September 24, 2003, at http://content.healthaairs.org/content/early/2003/09/24/hltha.w3.449.ull.pd+html (September 29, 2011). A similar adjustment is made in Munnell et al., Comparing Compensation:StateLocal Versus Private Sector Workers.

Average Benets as a Percentageof Wages, Adjusted Data

Public-School Teachers and Private Workers inEstablishments of 100 Employees or More, in 2010

Source: Authors calculations based on data from the U.S. Depart-ment of Labor, Bureau of Labor Statistics, and pension/retiree healthcare plan disclosures.


Public-SchoolTeachers

PrivateWorkers

Total benefits 100.8% 43.5%

Paid leave 6.6% 11.4%

Insurance plans 16.1% 13.3%

Retirement and savings 32.0% 6.2%

Retiree health care 9.9% 1.3%

Legally required benefits 7.4% 11.3%

Work-year leave 28.8% 0.0%


21/25


19

despite the act that school teachers work a signicantlyshorter average work year.

The answer to this puzzle lies in a ootnote to theBLS data, which states that:

The NCS [National Compensation Survey] uses

contract hours or teachers in determining thework schedule. Contracts usually speciy thelength o the school day, the number o teachingand required nonteaching days, and the amounto time, i any, teachers are required to be in theschool beore and ater school hours. These hoursare used to construct the work schedule. Forexample, it is common or teacher contracts tospeciy that teachers will work 185 days per year.In these cases, the daily work schedule would bethe length o the school day plus any time teach-ers are required to be in school beore or ater

the school day, and the weekly work schedulewould be the daily schedule multiplied by 5 days(Monday through Friday). The number o weekswould be 37 (185 days 5 days per week). Thetime not worked during summer, Christmasbreak, and spring break would be excluded romthe work schedule and would not be consideredvacation or holiday.55

In other words, paid time o reported in BLS datarefects leave only during a 37-week work year andexcludes 15 weeks o leave during summers and holi-

days. To account or shorter work years, we created anadditional BLS category called Work Year Leave whichrepresents the value o a shorter work year. The base-line value o Work Year Leave equals 15 weeks dividedby 52 weeks, or 28.8 percent o salary.

The BLS assumption o a 185-day work year seemsreasonable.56 In some cases, teachers must work bothbeore and ater the ocial school year ends, and indeedsome work years are longer than 185 days. Washing-ton, D.C., public schools have 180 student days but195 total teacher days; likewise, Chicago public schools

have 193 teacher days. For the purposes o sensitiv-ity analysis, we also use a longer 195-day work year in

which work year leave equals 13 weeks divided by 52weeks, or 25 percent o salary.

Summary o Benefts. With modications to retirement income and paid leave, along with the inclusiono retiree health care, a uller view o non-wage compensation is presented in Table 7. The changes are arrom trivial. While the incomplete BLS data imply thatbenets or public-school teachers are worth about41.2 percent o their salaries, a more complete accounting puts the true value at 100.8 percent. Because oretiree health coverage and a small upward adjustmentto private-sector DB pensions, private-sector benetsrise to 43.5 percent o salaries.

Even with the lower value or work year leave premised on a 195-day work year, total teacher benetsequal 97 percent o salaries. While including the valueo longer paid leave is important, reasonable changes

in its value do not result in large changes to measuredcompensation. In act, benets equal 72 percent osalariestwo-thirds larger than or comparable private-sector workerseven when work year leave isexcluded entirely.

THE TEACHER COMPENSATION PREMIUM

We began both the salary and benets sectionby describing how the standard measures o teacher compensation are misleading and incomplete. Iaccepted at ace value, they would imply that public

school teachers are highly undercompensated compared to similar private-sector employees. The initialCurrent Population Survey regression suggested thateachers are underpaid in wages by 19.3 percent, andthe incomplete BLS data indicated that teachers receiveabout the same benets as a percentage o their salariesas non-teachers in large rms41.2 percent or teachers versus 41.3 percent or non-teachers.

However, once salaries and benets are properlymeasured, public-school teacher compensation is signicantly greater. We have shown that public-schoolteachers are not likely to be underpaid in wages, and

they may even enjoy a small wage premium. As a baseline, we assumed that public-school teacher salaries are

55. Richard Schumann, Work Schedules in the National Compensation Survey, Bureau o Labor Statistics, July 28, 2008,at http://www.bls.gov/opub/cwc/cm20080722ar01p1.htm (September 25, 2011).

56. See, or example, John Forsyth, Administrator Pay Vs. Teacher Pay, The School Administrator, December 2003,at http://www.aasa.org/SchoolAdministratorArticle.aspx?id=8932 (September 25, 2011).


22/25


20

in line with market ratesthat is, there isneither a salary penalty nor a premium. Withthat assumption, benets become the deci-sive actor. Including benets, total compen-sation is 40 percent higher or public-schoolteachers.57

Benets are so much higher or teachersthat they dwar the impact o salaries. Even iwe use the initial, fawed regression estimateo a 19.3 percent salary penalty, teacherswould still enjoy a 12.9 percent total com-pensation premium over similarly qualiedprivate-sector employees.58

Better data and urther research couldproduce improved wage and benet esti-mates, and total teacher compensation willvary rom state to state and district to district.

Nevertheless, the teacher-compensation pre-mium measured here is not a unction o anysingle adjustment or assumption. No reason-able set o assumptions will produce a teach-er compensation penalty, and job securityhas yet to be considered.

JOB SECURITY

Once hired, most public-school teachersace a short probationary period when theymay be red relatively easily. Ater that, teachers rarely

lose their jobs. Anecdotes abound about teachers whostay employed despite obvious misconduct or incom-petence. While perhaps an extreme case, New York Citymaintains the inamous rubber rooms where teach-ers whom the district is attempting to re are paid todo nothing, as the seemingly endless appeals processgrinds on.59

During the recent recession and state and local bud-get crunch, some public-school teachers were indeedlaid o. Employment in education by local government

declined by 2.9 percent between September 2008 and July 2011, according to BLS data. Nevertheless, these

job losses occurred in a period in which overall privatesector employment declined by 4.4 percent.

Relative to the workorce as a whole and to comparable occupations, public-school teachers do appear tobe unemployed rather inrequently. Chart 1 shows theaverage unemployment rates between 2005 and 2010or public and private teachers, along with several otheroccupations that have been deemed comparable interms o workload.60 The unemployment rate or public-school teachers is 2 percentage points lower than

57. Assume that both public-school teachers and private workers earn $X in salary. Then the ratio o teacher compensation to privatecompensation is (1+1.0077)X / (1+0.4345)X = 2.0077/1.4345 = 1.3996, or a 40 percent compensation premium. (Note that weuse an extra decimal place o precision in these calculations compared to what appears in the text. For example, we use 100.77percent o wages as the value o teacher benets, not the rounded 100.8 percent.)

58. Here, we assume that private-sector workers earn $X in salary, implying that public-school teachers receive (0.807)X because othe 19.3 percent salary penalty. Now the compensation ratio is (1+1.0077)(0.807X) / (1+0.4345)X = 1.129, or a 12.9 percentpremium.

59. Steven Brill, The Rubber Room: The Battle Over New York Citys Worst Teachers, The New Yorker, August 31, 2009,at http://www.newyorker.com/reporting/2009/08/31/090831a_act_brill (September 29, 2011).

60. The comparable occupations come rom Allegretto et al., How Does Teacher Pay Compare?p. 21.

Inspectors, testers, sorters, samplers, and weighers

Technical writers

Archivists, curators, and museum techniciansOther teachers and instructors

Editors

Counselors

Private school teachers

Non-teacher average

News analysts, reporters and correspondents

Insurance underwriters

Architects, except naval

Computer programmers

Accountants and auditors

Public school teachers

Select compliance officers

Registered nursesOccupational therapists

Physical therapists

Clergy

8.3%

7.8%

6.6%5.2%

4.8%

4.1%

4.1%

3.8%

3.7%

3.6%

3.5%

3.4%

3.1%

2.1%

1.8%

1.5%1.1%

1.1%

1.0%

heritage.orgChart 1 CDA 11-03

Average Unemployment Rates for OccupationsComparable to Teaching, 20052010

Source: U.S. Department of Labor, Bureau of Labor Statistics, Office of Publications &Special Studies, Employment & Earnings Online , at http://www.bls.gov/opub/ee/home.htm%5C(September 26, 2011).


23/25


21

or private-school teachers, and 1.7 percentage pointslower than the non-teaching average. To the degree thatormal educational qualications overstate the mar-ket value o public-school teachers, these gures mayunderstate the level o additional job security enjoyedby teachers.

Much o the observed publicprivate dierence inunemployment rates is attributable to the ability ounions to promote job security in the public sector,wrote economist Steven G. Allen in a National Bureauo Economic Research book on government employeeunions.61 Indeed, job security is one o the express goalso teachers unions. Job security is especially valuableto employees when it protects against losing a job thatalready oers higher pay than teachers could otherwisereceive. In other words, the value o job security partlydepends on the wage and salary premium we ound in

prior sections.The public-school system in Washington, D.C., now

oers salary increases and bonuses to the best-perorm-ing teachers in the city, those in the top 7 percent o per-ormance ratings. However, these pay increases comewith one catchteachers who accept the money mustgive up some o their job security. Remarkably, 20 per-cent o thebest teachers declined this oer, even whenthey could have increased their base salary by up to$44,000 per year, equal to a 50 percent salary increaseover the standard seniority-based pay scale.62 Clearly,

teacherslike other employeesvalue job security.

63

Putting an overall monetary value on job security is

not a straightorward task, and academic research hasdemonstrated

Documents

Assessing the Compensation of Public School Teachers 19282337242