reseņas educativas (Spanish)
resenhas educativas (Portuguese)
This review has been accessed times since November 17, 2005
Allegretto, S. A., Corcoran, S. P., & Mishel, L. (2004). How Does Teacher Pay Compare? Methodological Challenges and Answers. Washington, D.C.: Economic Policy Institute.
Pp. 58
$9.95 ISBN 1-932066-14-4Reviewed by Paula Razquin
RAND CorporationNovember 17, 2005
The aim of the book How Does Teacher Pay Compare? Methodological Challenges and Answers is to reevaluate previous evidence regarding teacher relative salaries by discussing methodological and data limitations and to provide new evidence that contradicts and seemingly improves upon what has been found in the past. Previous evidence showed that teachers get higher salaries than people in other occupations and concluded that salaries are competitive enough to attract high quality candidates into teaching. The authors argue that previous studies completely ignored some data issues and, therefore, arrived at incorrect conclusions about the sign and size of the salary difference between teachers and people holding other jobs. The authors will examine, though a thorough methodological discussion, ways in which results may differ, i.e., when using different salary measures and data sources. In this review, I will summarize the debate and evidence that the authors respond to, revisit their methodological discussion and findings, and emphasize the book’s strengths and weaknesses.
The teacher salary debate: A comparative concern
Behind Allegretto, Corcoran, and Mishel’s book lies a debate that has important implications for education policy: Are teachers well paid? Or are they underpaid? Are teacher salaries enough to attract and retain teachers of the quality that the nation desires? Most educators believe that teachers should be paid as much as necessary to attract and retain qualified people and to motivate teachers to perform well. But how can we determine whether teachers are paid as necessary? One way is to look at teacher salary schedules and examine the incentives structure associated with education, age, professional development, subject taught, job responsibilities, and in some instances, teachers’ merit. Although they are very informative, salary schedules are of little value for issues of teacher recruitment and retention.
Comparing teacher salaries with salaries in other professions, on the other hand, allows us to examine the incentives that teachers get relative to what they would get in other occupations and helps us understand more about people’s decisions to work as a teacher. We can use salary schedules and complement them with other data to compare average salaries across occupations. The problem with average-salary comparisons is that they do not control for variations in other factors that may influence salaries and, therefore, may give the wrong measure about the true salary difference for teachers. For instance, a portion of the salary difference between teachers and other professionals may be due to differences in human capital investment rather than actual differences in the rewards for different occupations. In other words, comparisons based on averaged salaries are “unadjusted” comparisons that ignore variations in other factors that are also associated with the salary level and are not a good measure of teacher comparable pay.
One way to solve the problem of comparing teachers’ and other professionals’ salaries is to recur to a third approach—one that gives us “adjusted” salary differences. Using individual-level data on salaries and other personal characteristics, we can estimate the salary difference between teachers and other professionals, holding other characteristics constant. As we shall see in the following section, Allegretto, Corcoran, and Mishel’s book offers a methodological and data discussion that tells us about the limitations of previous evidence on teacher comparable pay (or teacher salary differential, as is referred to the economic and sociological literature).
Allegretto, Corcoran, and Mishel’s discussion and findings
In the introduction, “The debate over teacher pay,” the authors explain the importance of having a measure of teacher comparable salaries and outline the goals of the book. There are two forces that are shaping a renewed interest in teacher comparable pay. More and more interesting employment opportunities have been opening up for women, and recruiting teachers, especially highly qualified ones, has become a difficult endeavor. Yet, while recruiting and retaining highly qualified teachers may have become a difficult task, it has also become indispensable for improving student academic achievement.In fact, Title II of the No Child Left Behind Act of 2001 (NCLB) makes preparing, training, and recruiting high-quality teachers, particularly math and science teachers and those in high-need schools, one of its statutory purposes. Hence, it is within this context of a decline in the supply of teachers, together with an increase in the demand for qualified teachers under NCLB, that having good measures of teacher comparable salaries becomes fundamental for teacher recruitment and retention policies. This is what the authors will set out to do in the book.
In chapter I, “The evidence on teacher pay and teacher quality,” Allegretto, Corcoran, and Mishel trace the empirical evidence on teacher comparable salaries. The authors show that studies differ in their methodology and data, and as a consequence, their results differ too. In each of the following chapters, the impact of these methodological issues on the results will be articulated and examined in more depth, and the authors will finish by showing that their methodological approach yields evidence that contradicts what other authors have found to be the case for the United States. Due to the dialogical nature of the book, it is worth summarizing the evidence that the authors respond to before examining their discussion and results.
A 2003 issue of the journal Education Next: A Journal of Opinion and Research was dedicated to examining the “widely held assumption” that teachers are underpaid. Three economists presented their findings: Peter Temin from the Massachusetts Institute of Technology, Richard Vedder from Ohio University and a senior fellow at the Independent Institute, and Michael Podgursky, chairman of the Department of Economics at the University of Missouri-Columbia. Evidence is mixed on whether teachers are underpaid or overpaid and so are the policy conclusions the authors derive from their findings:
- Temin (2003) found that female teachers aged 35 to 44 earn less than women in other occupations. Using Current Population Survey (CPS) data from the Bureau of Labor Statistics (BLS) from 1979 to 1999, he found that, in 1999, the salary disadvantage was of about 10 percent for teachers with just college education (mostly elementary teachers) and about 40 percent for those with some graduate education (mostly high school teachers). From the late 1970s to the late 1980s, female teachers with college education enjoyed a salary premium, but the advantage started to diminish in the early 1990s and turned into a salary disadvantage by the end of that decade. On the other hand, since the late 1970s, teachers with graduate education have worked under a salary disadvantage, which almost doubled between 1979 and 1999. Because he found that teachers are underpaid compared to people in other occupations, Temin suggests that salaries will need to rise to a competitive level if we want to recruit and retain good teachers.
- Like Temin, Vedder (2003) also used CPS data but for the year 2001. Unlike the other author, Vedder found that, in 2001, teachers earned about the same, on a weekly basis, as four other professional occupations (accountants, biological and life scientists, registered nurses, and editors and reporters) but more than two other professional groups (social workers and artists). Vedder argues that the weekly salary differential is a poor measure of the actual teacher salary difference, because teachers have longer vacations and therefore work fewer days per year than other professionals. To obtain a better comparison, he looked at hourly wages using data from the National Compensation Survey (NCS) of the BLS. On an hourly basis, teachers’ pay is higher than the pay for other professional workers that require as much professional training as teachers. It is important to note that in this case the author is referring to average salaries and not the actual teacher salary differential.
Vedder tells us that fringe benefits are one of the most important factors in salary comparisons between teachers and other professionals. Teachers are, to a large extent, overpaid not only because they make more on a weekly and hourly basis but also because teachers typically receive more generous benefit packages than other professionals. His calculations show that teachers’ benefits represent more than 26 percent of teacher salaries, while benefits in all other domestic industries represent about 19 percent of salaries, and 17 percent for occupations in the private sector. When he accounts for this advantage, teachers’ average hourly compensation plus benefits exceeds the average for all professional workers by roughly 10 to 15 percent.
- Podgusrky (2003) also used BLS data and found that, on an hourly basis, teachers’ average salaries are higher than salaries in occupations such as accountants and auditors, computer programmers, and engineers, architects, and surveyors in non-private practice. Only engineers, architects, and surveyors in the private sector and attorneys have higher hourly salaries, on average, than teachers. On an annual basis, thus, Podgursky found that teachers with bachelor’s degrees or higher education who were 30 years old and younger seem to earn about 10 percent less than other college graduates of similar age, as indicated by comparisons of median earnings from the March CPS, 1999 through 2001 data pooled. Podgursky argues that this teacher-nonteacher gap is explained by the shorter workday and work year for teachers.
Allegretto, Corcoran, and Mishel show us that one problem with estimating teacher comparable pay is its great sensitivity to data measurement and sources. In each of the four following chapters (II, III, IV, and V), they discuss four methodological challenges and show us how their own results differ from previous research once these challenges are addressed. They review (a) the pay interval used (annual, weekly, or hourly pay), (b) the reference group chosen to compare teacher relative salaries (all other non-teaching occupations, all college graduates, or specific professions) and how such group is determined, (c) the salary measure used (basic wages or total compensations, including fringe benefits), and (d) data sources used.
a) The pay interval used (Chapter II)
In this chapter, the authors tell us that annual salary is the most common pay interval used for salary comparisons among professionals—annual salary comparisons are available from the Current Population Survey (CPS), a national monthly survey of about 50,000 households conducted by the Bureau of Labor Statistics (BLS) for more than 50 years. Because teachers work fewer weeks and fewer weekly hours than other professionals, using annual salaries as the comparison basis overstates the salary difference between the two groups. Using weekly salaries instead of annual salaries would solve this problem; however, we would be assuming that teachers who desire to work during the summer are able to do so at the same weekly salary rate as their contract teaching salary. If these assumptions were not true, then we would be understating the teacher salary differential.
Using the hourly wage rate would also solve the problem of teachers and other professionals not working the same amount of weeks per year. Moreover, it will solve the problem of both groups showing different hours worked per week. Hourly wages are available in another BLS survey called the National Compensation Survey (NCS) of establishments and occupations. Allegretto, Corcoran, and Mishel found some serious inconsistencies in the measurement of weeks and weekly hours worked between teachers and other professions, inconsistencies that deter them from using the NCS in comparisons of teacher salaries across occupations.
With the above considerations in mind, the authors use weekly earnings from the BLS-CPS survey to give us a measure of teacher comparable salaries for the period 1979-2003. They work with a sample of male and female full-time workers between 18 and 64 years old, working 35 hours per week or more, and they focus on elementary and secondary school teachers only. Results from a regression model that controls for education, experience, and other characteristics show that, in 1979, female teachers earned 2.9 percent more than other women, male teachers earned 22.1 percent less than men in other professions, while teachers in a pooled sample of women and men earned 6.8 percent less than other workers. In the period studied, both groups of teachers saw a decline in their salary disadvantage, and, in 2003, female and male teachers respectively earn 8.9 and 23.1 percent less than other professionals.
b) The occupational group chosen to compare teacher salaries (Chapter III)
The authors argue that much of the variance in salaries across occupations is due to occupational differences in skill requirements of the job and the market value of those skills. Previous studies have failed to address this type of variance and have, instead, selected occupational groups somehow arbitrarily or, at most, those based on education degrees similar to those required for teaching (i.e., other college graduates). Using NCS data, Allegretto, Corcoran, and Mishel selected occupations that had similar skill requirements to teachers along ten dimensions: knowledge, supervision received, guidelines, work complexity, nature or scope and effect of the work, personal contacts, purpose of the contacts, physical demands, work environment, and supervisory duties. They also assigned a market value to those skills. Based on the these two factors, they identified 16 occupations that were “similar” to teaching: accountants and auditors, underwriters, personnel-training and labor relations specialists, inspectors and compliance officers (except construction), architects, forestry and conservation scientists, registered nurses, occupational therapists, physical therapists, trade and industrial teachers, vocational and educational counselors, archivists and curators, clergy, technical writers, editors and reporters, and computer programmers. The authors then looked at the average weekly and hourly wages and found that teachers have lower weekly salaries than people in the other 16 occupations. Furthermore, they found that teachers work more hours per week, and not less, as it is commonly believed, than these 16 groups, and, therefore, their hourly wages are on average about 14 percent less. Between 1996 and 2002, the salary disadvantage for teachers increased by about 10 percent.
c) The salary measure used and the fringe benefits bias (Chapter IV)
Evidence of teacher comparable pay depends on the salary measure we use; for example, whether we look at basic wages, total wages, or total compensation packages, including fringe benefits. The authors show us that the teacher salary comparisons used before were based on CPS data and were for basic wages. We could use the same CPS data to look at total wages (W-2 wages, which include basic wages and extra salaries, such as overtime pay) or at total compensation differences across occupations (by adding fringe benefits and payroll taxes to the total W-2 wages).
The size and attractiveness of teacher fringe benefits is another matter of dispute in the debate on teacher comparable salaries. The authors tell us that previous studies have shown that teachers’ fringe benefits are larger than those of other occupations, yet none of those studies has really attempted to use available data to estimate the fringe benefits’ bias—the bias resulting from estimating teacher salary differentials without accounting for fringe benefits. In Allegretto, Corcoran, and Mishel’s book, the fringe benefits bias is estimated as the total compensation differential (wages plus fringe benefits and payroll taxes) minus the wage differential (either basic wages or total W-2 wages that include overtime pay). For this, they used data on benefits costs from another BLS survey (ECEC, or the Employer Costs for Employee Compensation).
The authors found, first, that teachers indeed have better benefit packages than other professionals, but this is because they receive health benefits for a full-year schedule when in fact their salaries are for less than a full year. However, because some teachers do not participate in the Social Security system, their payroll taxes are less than those of other professionals. As a result, the authors conclude that previous studies that used the basic-wages measure overstate teacher comparable pay. On the other hand, the total-wages measure understates the salary difference between teachers and other professionals.
d) Data sources (Chapter V)
This chapter offers an interesting comparison of the two data sources most commonly used for teacher salary comparisons—the National Compensation Survey (NCS) and the Current Population Survey (CPS), both from the BLS. After reviewing both sources and their salary measures, the authors conclude that the NCS is inappropriate and produces non-intuitive evidence of hourly pay differences between teachers and other professionals (“non-intuitive" meaning that teachers earn less than other professional workers). The authors claim that inconsistencies in the measurement of annual weeks worked and weekly hours worked between teachers and people in other occupations creates biased estimates of the relative hourly wage of teachers. The authors advocate for the use of weekly wages from the CPS survey. Specifically, they show that, when compared to the CPS survey, the NCS data: (i) overstates average weekly wages for half of all occupations, including teachers, and understates weekly wages for the other half; (ii) understates the hours worked per week for the majority of occupations, including teachers; and (iii) overstates the hourly differential between teachers and other professionals.
In addition to NCS and CPS differences, the authors show that the NCS measure of weeks worked per year is inconsistent across occupations, and therefore weekly wages and annual wages used for estimates of teacher comparable pay are biased. For occupations that have year-round schedules, NCS weekly salaries correspond to actual weeks paid, including holidays and vacation. However, for occupations that have less than a year-round schedule, like teachers, the NCS weekly salary measure corresponds to weeks worked rather than weeks paid. Because of this, they suggest that we use annual salary instead of weekly salary estimates.
As with average weeks worked per year, the NCS’ measure of average hours worked per week is inconsistent. For occupations on a full-time schedule, mean hours worked more closely corresponds to a traditional 40 hours per week schedule and this includes lunchtime and work breaks. On the contrary, for teachers, average hours worked in the NCS survey corresponds to contracted hours of work, excluding lunchtime, breaks, and hours that teachers work outside of school.
The book’s strengths and weaknesses
Overall, an examination of the methodological issues behind the empirical evidence regarding teacher comparable pay is timely, even past due. This is the major strength of the book. Instead of just describing their data and presenting results that disagree with previous findings, the authors introduce us to some of the challenges involved in estimating teacher comparable salaries and give us tools to assess the quality of seemingly contradictory evidence. Future empirical research on teacher comparable salaries will not be able to ignore the book’s contribution.
One challenge the book faces is that the authors fail to discuss other important model specification issues involved in estimating teacher comparable pay or salary differential. One issue is the selection bias resulting from working with a sample of people that have not been randomly assigned to the teacher and non-teacher groups but, rather, have self-selected into those groups. Similarly, selection bias is found when one works with a sample of wage earners that is the result of previous non-random decisions to participate in the labor market. If selectivity bias is not properly modeled, it can lead to inconsistent estimates of the teacher comparable pay. In his work for the UK, Dolton (1996) suggested that when modeling salary differential estimations, we should discuss self-selection issues and the simultaneity of the occupational and labor market participation decisions. These are issues that deserve more attention in the book, particularly when the authors present their regression results in Chapter II.
In addition to the data measurement and model specification issues, there are other relevant features in presenting results on teacher comparable salaries, which might have important policy implications for issues of teacher recruitment and retention. The authors present us with a very aggregated picture of salary differences between teachers and other professionals but held themselves back from showing, for example, regional or state variations or variations in non-pecuniary conditions of work for different occupations. Country-level estimates may omit variations in how “teacher friendly” some states and school districts are. In other words, states may differ in their ability and willingness to pay for public services, as well as the teacher salary they are willing to offer and the state’s trade-offs between class size and teacher salary. The authors are apparently working on obtaining regional estimates of teacher comparable salaries and will present them in future publications.
Finally, there is one minor weakness in the way the four discussion chapters are developed that it makes them harder to read: when the authors discuss their main results in Chapter II, they present adjusted salary differences without showing us their regression model. We are led to believe that the following chapters will also give us adjusted salary differences, but in some of them only average salary differences are discussed.
References
Dolton, P. (1996). Modelling the Labour Market for Teachers: Some Lessons from the UK. Education Economics, 4(2), 187-205.
Podgursky, M. (2003). Fringe Benefits. Education Next, 3(3), 71-76.
Temin, P. (2003). Low Pay, Low Quality. Education Next, 3(3), 8-13.
Vedder, R. (2003). Comparable Worth. Education Next, 3(3), 14-19.
About the reviewer
Paula Razquin is a RAND/Spencer Postdoctoral Fellow in Education Policy. She received her Ph.D. in International Comparative Education from Stanford University in April 2003, and also holds an M.A. in Sociology and an M.A. in International Educational Administration and Policy Analysis from the same university. Before coming to RAND, she was the Director of the Master’s Program in Social Sciences in Education and an acting Assistant Professor in Education in International and Comparative Perspective at the Stanford University School of Education. Dr. Razquin has written about teacher relative salaries and teacher incentives in developing countries, teachers’ strikes in Latin America, and education and the labor market, higher education, and women’s education in Argentina. She has consulted for he World Bank, USAID, the Inter-American Development Bank, the Organization of American States, and federal, state, and local government agencies.
Copyright is retained by the first or sole author, who grants right of first publication to the Education Review.
Editors: Gene V Glass, Kate Corby, Gustavo Fischman
~ ER home | Reseņas Educativas | Resenhas Educativas ~
~ overview | reviews | editors | submit | guidelines | announcements | search ~