Debunking the Myth of the “Teacher Pay Gap,” Again


Just in time for the start of the school year, the Economic Policy Institute (EPI) has returned to an old and familiar claim: Public school teachers are underpaid, and the pay gap is widening. The claim is perhaps too old and too familiar. EPI’s new report, authored by Sylvia Allegretto and Lawrence Mishel, is a repeat of its 2008 report, and as such, it ignores the important work on teacher compensation that has been done since. Allegretto and Mishel fail to address or even acknowledge major methodological advances in the literature in recent years that cast serious doubt on their conclusions.


We start with pensions because they matter more than any other issue. The difference between the EPI authors’ pension valuation and the correct valuation is huge.

Allegretto and Mishel calculate the value of the pension benefits that teachers earn in a given year based on how much their employers contributed to their retirement plans in that year, using data from the Bureau of Labor Statistics’ Employer Costs for Employee Compensation (ECEC) survey. This was the same method the authors used in 2008. For a 401(k)-type plan, that approach works fine: the employer’s cost is nothing more than the amount contributed to workers’ accounts.

Most teachers, however, receive traditional defined benefit (DB) pensions, in which the employer promises not a contribution today but a fixed, guaranteed benefit in retirement. For such pensions, the conflation of contributions with benefits doesn’t work. In fact, the employer’s pension contribution in a given year bears no legal or mathematical relation to the pension benefits that teachers accrue in that year. Sometimes, governments facing unfunded liabilities will contribute more than is needed to cover accruing benefits. Other times, governments will shortchange the plan or even skip a contribution entirely. Nevertheless, teachers earn the same pension benefits in all of those years based on a formula written into law, and governments are legally obligated to pay when the bill comes due. Whether a government contributes nothing or an infinite amount, the benefits accruing to employees do not change.

Moreover, the way that state and local governments calculate their pension contributions means that two employees receiving exactly the same benefits could be assigned very different pension compensations under Allegretto and Mishel’s methodology. The DB plans funded by state and local governments, unlike private sector DB plans or DB plans for public employees in other countries, base employer contributions on how much a government assumes its plan’s investments will earn over time. For instance, the Connecticut teachers’ retirement plan assumes an 8 percent annual return, while the Indiana teachers’ plan assumes a much lower 6.75 percent return. Even if these two plans promise exactly the same benefits, Connecticut would contribute about one-third less toward its teacher pensions than Indiana would. Allegretto and Mishel’s ECEC data would then classify Connecticut teachers as receiving lower pension benefits, even though this difference is entirely due to how the two states chose to finance the same benefits over time. Connecticut makes lower contributions but takes more investment risk; Indiana does the opposite. But neither investment strategy tells us the benefits earned by teachers in those states.

This problem has a small effect when comparing teachers in one state to teachers in another. It has a massive effect when comparing teachers’ pensions to those of private-sector employees. While state and local governments may base their contributions on the assumption of roughly 8 percent investment returns, private-sector pensions, as well as most public pensions in other countries, are required to value their liabilities using a much lower bond yield to capture the fact that pension benefits are guaranteed against market or default risk. The sponsors of private plans must therefore contribute much more for every dollar of promised benefits than governments contribute to teacher pension plans that value liabilities using an 8 percent assumed return on portfolios heavily weighted with stocks, hedge funds, or private equity.

Virtually all professional economists agree that calculating the value of guaranteed pension benefits using the assumed return on a portfolio of risky assets “understate[s] their pension liabilities and the costs of providing pensions to public-sector workers.” We addressed this problem by adjusting the actuarial costs to a common discount rate that accurately reflects the low-risk nature of public pension benefits. When we did that in our 2011 analysis, we found that pension costs for teachers were worth not the 11 percent of wages reported by the ECEC but a remarkable 32 percent of wages. That figure reflects how much a worker with a 401(k) would need to save to receive the same retirement benefits with the same level of risk as those paid to the average public school teacher.

After we pointed out that ECEC-style data on annual contributions to public pension plans are a dramatic underestimate of pension costs, most of the pay comparison literature followed our lead in applying some kind of correction. Only one non-EPI study we know of continued to use the unadjusted ECEC, but that study’s authors explicitly noted the limitation.

The new EPI study on teachers never acknowledges the ECEC’s undervaluation of pension benefits and does not cite any of the above papers. Its readers would never know that a controversy over pension valuation even exists.

Retiree Health Benefits

While pensions are undervalued in the EPI report, the retiree health coverage provided to most teachers is ignored entirely. Most teachers earn the right to health benefits in retirement, which can provide full coverage from retirement through Medicare at age 65; they often receive supplementary benefits thereafter. Actuarial valuations show that in 2015, Texas teachers accrued future retiree health benefits equal to an additional 5.9 percent of their wages. For Illinois teachers, it was about 7.9 percent of wages. In New York City, it was 15.7 percent. In the private sector, by contrast, retiree health coverage is much less generous—and rapidly disappearing altogether.

Because governments finance retiree health benefits on a “pay-as-you-go” basis—meaning that benefits are paid to retirees as they come due—the ECEC does not include any employer contributions for retiree health costs. Nevertheless, the benefits are real.

We pointed out this shortcoming of the ECEC in 2011, noting that “BLS data do not report on retiree health coverage for private sector or state and local government employees as these plans are generally unfunded, meaning there are no current employer contributions to measure.” Later in 2011, the Center for State and Local Government Excellence study cited above made the same point: “Clearly,” it said, “retiree health should be added” to benefits calculated from the ECEC data. The lack of retiree health benefits in the ECEC data has been known for years and has been corrected for in most of the recent literature. But, as with pension valuation, Allegreto and Mishel fail to mention the issue.


Allegreto and Mishel contend that teachers receive wages that are 17 percent lower than “comparable” workers. However, the basic controls they use, such as age and education, do not make worker skills comparable across occupations. Had Allegretto and Mishel included more occupational variables in their regression, readers would have discovered that virtually every occupation is either “underpaid” or “overpaid” by their reasoning. In fact, by simply replacing teachers with a different occupation, EPI could publish additional reports with titles like “The Food Service Pay Penalty” and “The Architect Pay Premium.”

Which is more likely: that the labor market has failed to provide the right level of average compensation to almost every occupation, or that workers simply differ in important ways beyond their age and the number of years they spent in school? Roughly half of public school teachers have education degrees, for example. Should an education degree be worth as much in the marketplace as an engineering degree? That is Allegretto and Mishel’s logic, despite strong evidence that education majors lag behind other college graduates in terms of marketable skills. None of this evidence is mentioned in their report.

A better test of whether current teachers are underpaid would be evidence that many have left the profession for better-paying jobs elsewhere. But despite all the anecdotes about teachers being lured away from public schools to lucrative private-sector work, the available evidence shows that, if anything, teachers earn less when they leave teaching for another job. Allegretto and Mishel do not address this evidence despite asserting that wage increases are crucial for retaining teachers.

Work Time

As they did in the 2008 report, Allegreto and Mishel rely on the weekly wages reported by public school teachers in the Current Population Survey, leading to confusion about whether the wage data refer to annual salary divided by 52 weeks or by some smaller number of weeks that reflects teachers’ shorter work year. Their new report rehashes a decade-old debate over that technical issue, which is related to their 2008 claim that “all of the data available show that teachers work at least as many hours each work week as comparable college graduates.” That statement is no longer true, thanks to recent analyses of time-use data that Allegreto and Mishel fail to cite.

The American Time Use Survey (ATUS) captures work time throughout the year, as reported by employees themselves whenever and wherever work may occur. Multiple studies have used the ATUS to generate more accurate measures of how much time teachers spend working. In our own study, we found that teachers report working approximately two hours fewer than other professionals worked during a typical work week and about 83 percent as many hours over the full calendar year. Those reported hours included time teachers spent working at night or on weekends.

It is impossible to know whether Allegretto and Mishel’s weekly wage method implicitly captures a similar work year, and there is no reason to continue arguing about it. Using the recent ATUS results, researchers can adjust annual teacher salaries for their shorter work year simply by dividing by 0.83 (or a similar ratio if another specification is preferred). Why not take advantage of these new findings instead of persisting with an out-of-date methodology?

“Wider Than Ever”?

Even if Allegretto and Mishel undervalue teacher compensation each year, can we still conclude that teacher compensation has been falling relative to alternative jobs? Not necessarily, for two reasons. First, teachers’ unobserved skills may have fallen farther behind the skills of allegedly “comparable” workers during the same time period. This could happen for any number of reasons, but one of the most obvious is teachers’ tendency to increase their formal educational credentials without necessarily increasing their teaching skills. Between 1992 and 2014, the percentage of teachers with more than a bachelor’s degree increased from 46 percent to 56 percent, based on data from the Current Population Survey. In terms of formal educational credentials, teachers are one of the most highly educated occupations in the United States. At the same time, however, the average SAT scores for prospective teachers as reported by the College Board are middling and have risen only slightly since the 1980s.

Moreover, there is little evidence that postgraduate degrees for teachers either instill additional teaching skills or allow schools to identify individuals who are better teachers. According to the Urban Institute’s Matthew Chingos, “the fact that teachers with master’s degrees are no more effective in the classroom, on average, than their colleagues without advanced degrees is one of the most consistent findings in education research.”

This again highlights the problem with Allegretto and Mishel’s wage analysis. They state that “the small fraction of the most cognitively skilled college students who elect to become teachers has declined for decades.” In other words, teacher skills are falling. Yet, in terms of formal educational credentials—the key variable that they use to benchmark teachers’ wages against “comparable” workers—teacher quality has never been higher! As teachers increase their educational credentials without necessarily increasing their skills, the Allegretto-Mishel model will show declining relative pay even if teachers are keeping pace with workers whose skills are truly the same.

A second reason that the teacher pay gap may not be “wider than ever” is that pension benefits have been increasing in ways not fully captured by the EPI report’s faulty methodology. In a review of teacher pension benefits, Robert Clark and Lee Craig write, “The main story of the past quarter century has been the increased generosity of teacher retirement plans. Normal retirement ages have been reduced, generosity parameters increased, and the number of years in the salary averaging period have been reduced. As a result, replacement rates rose by 5 percentage points, or almost 10 percent, between 1982 and 2006.”

Compounding the rising generosity of pension benefit formulas is the decline of interest rates on low-risk investments, which raises the cost of providing teachers with a fixed, guaranteed pension benefit. If state and local pensions were paying mind to interest rates—as they should, and as corporate and overseas public employee plans are required to do—contributions would have risen significantly as the yield on 20-year U.S. Treasuries dropped 3.7 percentage points between 2000 and 2016.

Thus, Allegretto and Mishel have not simply neglected to establish that a teacher compensation penalty exists; they have not even established that teacher compensation is falling behind that of competitive occupations.

The Real Policy Questions

The compensation of the average public school teacher relative to alternate occupations is, in a sense, a secondary question. Three million Americans have willingly chosen to pursue teaching at the pay rates offered by school systems, indicating that the pay, benefits, work conditions, and job satisfaction offered in the teaching profession are at least as attractive as what these individuals might receive in other jobs.

The larger question is whether higher teacher pay and benefits would boost teacher quality and student outcomes sufficiently to justify the additional costs. As Raj Chetty and colleagues found, better teachers are associated with better outcomes for students. But in school systems where pay is almost wholly determined by educational credentials and job tenure, neither of which is strongly associated with performance, it is unclear how much raising average pay would boost teacher quality—or whether it would boost it at all.

For instance, even when schools are offered more attractive candidates with specialized majors in the quantitative fields or with higher GPAs from more competitive colleges, school administrators often opt for seemingly less-qualified applicants who took the traditional ed school route. Moreover, if higher salaries attracted larger numbers of less-qualified applicants, then even fewer good new teachers might end up in the classroom. Raising salaries for all teachers irrespective of performance or specialty is unlikely to be the most effective or cost-efficient way to boost teacher quality and student outcomes.


Even in 2008, many researchers doubted EPI’s claims that public school teachers were dramatically underpaid. Since that time, the reasons to doubt this conclusion have only multiplied. If Allegretto and Mishel had incorporated recent methodological advancements involving pensions, retiree health benefits, wages, and work time, then their report would have been a genuine contribution to the state of knowledge on teacher-pay trends. Unfortunately, their methodology remains stuck in 2008, and their readers remain in the dark about an issue that deserves careful treatment.

Andrew G. Biggs is a resident scholar at the American Enterprise Institute. Jason Richwine is a public policy analyst based in Washington, D.C.

Last Updated


Notify Me When Education Next

Posts a Big Story

Business + Editorial Office

Program on Education Policy and Governance
Harvard Kennedy School
79 JFK Street, Cambridge, MA 02138
Phone (617) 496-5488
Fax (617) 496-4428

For subscription service to the printed journal
Phone (617) 496-5488

Copyright © 2024 President & Fellows of Harvard College