Rewarding and Employing Teachers Based on Their Value-Added
The following is an excerpt from What Lies Ahead for America’s Children and Their Schools, a new book edited by Chester E. Finn, Jr., and Richard Sousa for Hoover Institution Press. This excerpt comes from a chapter called “Covering the Costs” by Caroline M. Hoxby.
In recent years, researchers have demonstrated that individual teachers differ substantially in their value-added—their propensity to raise students’ achievement. For our purposes, the key take-aways from the research are as follows:
1. A teacher who is in the top 10 percent of the current distribution of value-added raises student achievement by several times what a teacher in the bottom 10 percent does. 
2. If all US teachers had value-added equal to what the current top 10 percent has, the average American student would achieve at the level of students whose parents have incomes in the top 10 percent of the family income distribution. This is approximately equivalent to the level at which the average student in Singapore achieves. 
3. Even after a single year of teaching, we can predict a teacher’s value-added sufficiently well that the retention decision can be made at that point. Two years of teaching experience adds information, but subsequent years of teaching add little to our predictions. 
4. After the first few years of teaching (during which most teachers’ value-added rises, apparently through on-the-job learning), a teacher’s value-added remains fairly stable. 
5. Teachers’ measured value-added based on students’ test scores is highly correlated with their measured value-added based on long-term outcomes such as students’ earnings, employment, and college attainment. 
6. Teachers’ value-added on students’ non-cognitive outcomes, such as truancy, attendance, and disciplinary incidents, can be measured and gives therefore a better fix on how teachers affect long-term outcomes. 
7. Teachers with master’s degrees do not have systemically higher value-added. If anything, the evidence suggests that master’s degrees are associated with lower value-added. 
8. There are almost no credentials that predict a teacher’s value-added, with the exception that individuals who attended a more selective college tend to have higher value-added. This is presumably due not only to their higher aptitude and better education but also to the fact that selective colleges admit students in a fairly holistic way that rewards leadership, motivation, and similar traits. 
9. There is evidence that teachers improve their value-added when exposed to other teachers who have high value-added and when they are offered pay based on their value-added. 
Suppose that we wanted to have a pay system that attained the goal that all teachers (except the brand new, untried ones) would have value-added in the range we currently see among the top 10 percent of teachers (hereafter: “high value-added teachers”). A well-designed pay system would achieve this both through selection (getting the right individuals to become and stay teachers) and through incentives (giving teachers incentives to raise their value-added). What would such a system look like? First, within teaching, pay would be aligned with value-added. This would provide the right incentives for improvement. Second, to ensure that high value-added individuals select into teaching, their pay would have to be competitive with alternative jobs that they could obtain. Since the evidence suggests that people who can be high value-added teachers are also better at other jobs, we should assume that their alternative jobs will be those that a baccalaureate degree holder with above-average aptitude, college quality, and motivation could obtain. Note that these are not the alternative jobs that most current teachers now take if they leave teaching. This is because most current teachers have aptitude and college quality well below average among baccalaureate degree holders.
Consider some private sector occupations that are filled by people with baccalaureate or more education (but not a professional degree or PhD): accountants, compensation and benefits managers, computer programmers, editors, landscape architects and surveyors, property managers, occupational therapists, regional planners, public relations specialists, and buyers for major retail stores. These people are all paid based on their productivity and rewarded by private sector employers for their intelligence, motivation, and ability to work with other people. Benjamin Scafidi, David Sjoquist, and Todd Stinebrickner show that only about 5 percent of people who leave teaching ultimately take a job along these lines.  Sara Champion, Annalisa Mastri, and Kathryn Shaw show that the teachers who leave for such jobs are those whose value-added in teaching is unusually high. 
Let us say, then, that any high value-added teacher’s compensation—pay plus benefits—must be equal to the average for full-time workers in these occupations.  Their average annual compensation was $89,989 in the 2009–10 school year (the most recent year for which full financial data on US schools are available).  Suppose that, by employing value-added-based pay and creating the right selection and incentives, US schools were able to attain a teaching workforce almost entirely composed of high value-added teachers. Would they have sufficient money to pay compensation of $89,989 to all teachers except untried, new ones? The answer is “yes” with current budgets. In the 2009–10 school year, public schools paid $275.3 billion in compensation (wages, salaries, payments for benefits) to instructors and employed 3,123,957 full-time equivalent classroom teachers.  This is $88,132 per teacher—enough to pay 95 percent of teachers the competitive compensation of $89,989 and pay the remaining 5 percent of teachers (presumably the untried ones) compensation of about $53,000. If, by employing only high value-added teachers, schools could eliminate instructional support staff (people who are not certified instructors but who help with instructional improvement, curriculum development, staff training, and the like), the average teacher might be paid $94,177 out of current budgets. 
Keep in mind that no adjustment has been made for the fact that teachers’ contract hours are only about 1,200 hours per year (about 0.6 of full-time) and that teaching jobs have work hours, work locations, and vacation timing that allow them to avoid child care costs that the typical full-time worker must bear. Thus, it is generous to assume that they need to be compensated like someone who works full time. I have made such generous assumptions because high value-added teachers may actually work a good many more hours than their contract hours.
One might wonder how it is possible that US public schools could, within their current budgets, pay teachers in a manner that is so competitive with private sector rewards. The main explanation is that although high value-added teachers are currently underpaid, low value-added teachers who have high seniority, master’s degrees, and other paper credentials are systemically overpaid relative to their alternative jobs. They have no incentive to leave teaching, therefore. They also have no incentive to improve their value-added. Low value-added teachers absorb so much of the total compensation budget that little is left for high value-added starting teachers, who are not only underpaid if they do teach but who tend to leave teaching as a result. A second, less important explanation is that a non-trivial share of instructional compensation goes to people who are not classroom teachers but who provide some form of instructional “assistance.” Randomized controlled trials suggest that such assistants add little or no value in terms of student achievement.  As a logical matter, this may be because principals often steer incompetent teachers into non-classroom instructional jobs. That is, such jobs act as “holding tanks” for ineffective teachers who have not been dismissed—sometimes because of formal barriers to dismissal but more often because there is no norm of dismissal for mere incompetence.
Importantly, a system in which teachers are compensated in a manner that is fully competitive with their private sector alternatives would be highly sustainable. This is a major but often underappreciated benefit of education reforms that move schools in the direction of managing themselves as private organizations do. This is in contrast to well-meaning but starry-eyed educational methods that depend on school personnel being so altruistic or specially attuned to teaching that they are rare or ignore the fact that they are undercompensated.
Caroline Hoxby is professor of economics at Stanford University.
Notes: 1. This calculation is based on the persistent teacher effects computed by Douglas Staiger and Jonah Rockoff, “Searching for Effective Teachers with Imperfect Information,” Journal of Economic Perspectives 24, no. 3 (Summer 2010), 97–118. As they point out, however, their value-added calculations are very similar to those of numerous other researchers who use different data but like methodology. It is important to use shrinkage methods to compute the persistent teacher effect, as opposed to the annual teacher effect which also contains noise from student composition. 2. This calculation is based on National Assessment of Educational Progress individual student data, equated to TIMSS (Trends in International Mathematics and Science Study) data for the latter statement. 3. See Staiger and Rockoff, “Searching for Effective Teachers.” 4. See ibid. for a summary of the numerous studies that have demonstrated this point. 5. See Raj Chetty, John Friedman, and Jonah Rockoff, “The Long-term Impacts of Teachers: Teacher Value-added and Student Outcomes in Adulthood,” NBER Working Paper 17699, 2011. 6. See Clement Kirabo Jackson, “Non-Cognitive Ability, Test Scores, and Teacher Quality: Evidence from 9th Grade Teachers in North Carolina,” NBER Working Paper 18624, 2012. 7. See Thomas J. Kane, Jonah E. Rockoff, and Douglas O. Staiger, “What Does Certification Tell Us About Teacher Effectiveness? Evidence from New York City,” Economics of Education Review 27, no. 6 (December 2008), 615–631. 8. See Kane, Rockoff, and Staiger, “What Does Certification Tell Us?” and Staiger and Rockoff, “Searching for Effective Teachers.” 9. On the former point, see C. Kirabo Jackson and Elias Bruegmann, “Teaching Students and Teaching Each Other: The Importance of Peer Learning for Teachers,” American Economic Journal: Applied Economics 1, no. 4 (2009), 85–108. On the latter point, see Sally Hudson, “The Effects of Performance-Based Teacher Pay on Student Achievement,” SIEPR Discussion Paper 09-023, 2010, Stanford Institute for Economic Policy Research. In addition, there is some evidence that a very well-designed evaluation and feedback system may help teachers improve their value-added: see Eric Taylor and John Tyler, “The Effect of Evaluation on Teacher Performance,” American Economic Review 102, no. 7 (2012), 3628–51. 10. See Benjamin Scafidi, David Sjoquist, and Todd Stinebrickner, “Do Teachers Really Leave for Higher Paying Jobs in Alternative Occupations?” The B.E. Journal of Economic Analysis & Policy 6, no. 1 (2006), 1–44. 11. See Sara Champion, Annalisa Mastri, and Kathryn Shaw, “The Teachers Who Leave: Pulled by Opportunity or Pushed by Accountability?” Stanford University working paper, 2011, available at http://www.rand.org/content/dam/rand/www/external/labor/seminars/allca/Champion-Mastri-Shaw-2011.pdf. 12. Full-time is defined by the US Bureau of Labor Statistics as a nominal 2,080 hours per year, but in fact few of the workers in question are paid on a truly hourly basis or “punch the clock,” so their exact hours of work are not relevant or even accurately recorded. Full-time is thus what is usual among workers who define themselves as “full-time” on the jobs in question. They are at the office more than forty hours in a typical non-vacation week, but they have paid time off in various forms. 13. Author’s calculations based on Occupational Employment Statistics, US Bureau of Labor Statistics, 2013. 14. Author’s calculations based on the 2009–10 Common Core of Data (US Department of Education, 2013) and the 2009–10 (2010 fiscal year) Public Elementary-Secondary Education Finance Data (US Department of Commerce, 2013). 15. Instructional support staff is defined in Exhibit 1a: School-Level Personnel by the United States Department of Education, Institute of Education Sciences, National Center for Education Statistics, 2013, http://nces.ed.gov/pubs/web/9619ex1a.asp. 16. See Alan Krueger, “Experimental Estimates of Education Production Functions,” Quarterly Journal of Economics 114, no. 2 (May 1999), 497–532.
Reprinted from What Lies Ahead for America’s Children and Their Schools, edited by Chester E. Finn Jr. and Richard Sousa, with the permission of the publisher, Hoover Institution Press. Copyright © 2014 by the Board of Trustees of the Leland Stanford Junior University.