High school seniors in schools with merit-pay programs for teachers score slightly higher on standardized tests than their counterparts at schools that offer no special salary incentives, according to a national study by the University of Florida.
Interest in teacher pay-for-performance programs has grown in recent years, and some jurisdictions, such as Florida, have even begun to mandate them. But David N. Figlio and Lawrence W. Kenny, both economists at the university in Gainesville, said their study may be the first in the country to find a link between financial incentives for individual teachers and better student achievement. It鈥檚 scheduled to be published in a forthcoming issue of the peer-reviewed Journal of Public Economics.
鈥淏asically, the data haven鈥檛 existed to really allow you to do much with this topic,鈥� Mr. Kenny said in an interview.
To gather the data for their study, the researchers conducted elaborate analyses to patch together information on achievement and incentive practices from several surveys and tests. They eventually collected data on 502 public and private schools across the country, many of which had incentive programs in place for teachers in 2000 and were very likely, the researchers determined through separate analyses, to have also had them in place seven years earlier.
In addition, all of the schools had taken part in a federal longitudinal study that tested 12th graders in 1993 in four subjects: reading, mathematics, science, and history. Matching the 1993 test results from those schools to data from surveys on their teacher-incentive practices in 2000, the researchers found that students in the pay-for-performance schools scored an average of 1 to 2 percentage points higher on the exams overall than peers at schools where teacher salaries were more uniform.
The incentives seemed to be most effective, the study also found, in schools that were more selective in doling out pay rewards and those that served high proportions of low-income families.
鈥淚 think that incentive programs can work and do work, and we were getting these results under whatever incentive mechanisms people were using in 1992,鈥� said Mr. Kenny.
Limitations Conceded
The authors acknowledged, however, that the data sets they used were not ideal. For one, seven years had elapsed between the 1993 testing wave and the 2000 survey. Short of conducting an experiment, they added, there鈥檚 also no way to tell for sure if the schools with incentives were qualitatively better to begin with than other schools in the survey.
鈥淏ut we did go through a lot of testing to try to minimize the extent to which that is a problem,鈥� Mr. Kenny said.
For instance, the researchers adjusted the data to account for achievement and socioeconomic differences among the student bodies at the schools they studied. They also looked at whether schools were unionized, had nonfinancial incentives in place for teachers, or were located in jurisdictions that had begun allowing charter schools.
They also analyzed the data separately for private and public schools and saw that the pay-for-performance effects held up in both cases. (About 20 percent of the schools in the 1993 study sample were private, a slightly higher percentage than national data showed was representative at the time.)
However, the effects tended to dissipate and even disappear altogether in schools that tended to share the wealth, spreading out pay rewards to larger percentages of teachers.
鈥淒oling out merit pay to most teachers leaves them with little incentive to do a better job,鈥� the researchers wrote. 鈥淥ur evidence 鈥� suggests that there is a relation between test scores and merit pay targeted to a few, but no association between student performance and indiscriminate merit pay.鈥�
Douglas N. Harris, a professor of education policy studies at the University of Wisconsin-Madison, called the research worthwhile, despite the shortcomings in the data.
鈥淭he next step, short of an experiment, is to do a before-and-after study,鈥� in which the same schools would be compared with and without incentives in place, he said. He and other experts said the data will also improve as teacher-incentive programs become more widespread.
He cautioned, though, that a broader question looms. 鈥淎ccountability systems and incentives will almost certainly make test scores go up,鈥� he said. 鈥淲hat we don鈥檛 know in the long run is whether students are learning more.鈥�
