Recently, I was part of the Global Economic Symposium  held in Rio de Janeiro. This year’s theme was Growth through Education and Innovation; I presented as part of a panel entitled Effective Investments in Education.
My presentation focussed on the fact that a growing and compelling body of research shows that teacher effectiveness varies widely - even across classrooms in the same grade in the same school. Getting assigned to a bad teacher has not only immediate, but also long term, consequences for student learning, college completion and long-term income.
But most education systems do not know how many bad - or good - teachers they have, or where they are working, or whether their performance can be improved. They lack objective data on even the most basic elements of teachers' performance - whether they come late to class or not, how well they keep students engaged, whether they put effort into grading homework and preparing lessons, and above all, how much their students learn.
In the absence of real performance data, good and bad teachers are retained and promoted almost indiscriminately, on the basis of soft and subjective (you scratch my back, I'll scratch yours) performance reports and/or the number of formal degrees or training courses completed - all of which are poor proxies for actual teacher performance. As a result, school systems end up like Washington DC - where in 2007 only 8% of 8th graders were on grade-level in math, but 80% of all teachers were rated as "exceeding expectations".
Getting real data on teacher performance and using it to reward the best, improve the middle, and deal with the worst is the only true strategy for the really big, long-term improvements in education quality that many countries need. Without a solid system of teacher performance evaluation, it is impossible to get promotion and pay incentives right, and make the profession more attractive to high-talent individuals over time. Without a solid system of teacher performance evaluation, teacher training programs waste millions of dollars because they are unable to target the teachers who need it most and the type of help they need the most.
The ability of even low-income developing countries to put in place data-rich, robust and credible systems of teacher performance evaluation is higher than ever today, thanks to declining costs of technology and information. Video cameras can be placed in classrooms to collect regular footage of teachers at work that can be evaluated against clear rubrics; students can work on Khan-academy  style homework problems online, generating day-by-day data on how different teachers' 4th grade math classes are progressing - or how well a teacher deals with individual students who fall behind. Annual student assessments - or even biannual assessments as Secretary Costin introduced here in the Rio municipal schools, can provide real-time data on the value-added learning gains that different teachers are able to produce. All of these elements - plus feedback from peers, students, and school directors - can and should combine into rich and comprehensive, yet objective and comparable, evaluations of teacher performance.
It's cheaper than ever, but still not cheap. It's more feasible than ever, but still not easy, especially the political economy challenge of getting teachers' unions on board. We saw a recent strike in Chicago . There is currently union opposition in Peru, where the Ministry wants to introduce a well-designed, comprehensive system of teacher performance evaluation.
The good news is - there are a growing number of good examples - from Singapore, to Chile, to Ecuador, to Cincinnati, Ohio, to Washington DC. And when these systems succeed in identifying the best teachers so that their practice can be shared; identifying teachers who are struggling and need help; identifying those "bad apples" who should never be in a classroom, students gain more than from any other "solution" for improving education.