Research in the US has pointed out that the most important determinant of the quality of education is the quality of teachers but that students’ achievement is not linked to observable teacher characteristics such as qualification or experience. Using data from selected private schools in Uttar Pradesh, this column estimates the contribution or ‘value added’ of teachers to student scores in external examinations.
In recent years, there has been an increasing focus on the quality of education in India partially driven by the realisation that the rapid gains in school enrolment and attendance are not translating into gains in cognitive skills, as measured by test scores in reading, writing and math. These scores remain low compared to international benchmarks.1 Research over the past decade in the US confirms that the most important determinant of the quality of education is the quality of the teacher; however, it also suggests that the variation in students’ achievement cannot be predicted by most observable characteristics of teachers (including factors that are commonly considered to be proxies for quality such as experience, education and training) (Rivkin, Hanushek and Kain 2005, Rockoff 2004). Hence, a better way to assess the overall contribution of teachers to students’ achievement is to estimate the Teacher Value-Added (TVA).2 TVA is a measure of the extent to which a teacher is able to improve student learning (as measured by test scores in external examinations) during the period of time that they are responsible for teaching the concerned student.
Estimating ‘Teacher Value-Added’
We estimate TVA using administrative data from a private school system running multiple schools in a district in the state of Uttar Pradesh in India (Azam and Kingdon 2014). The data contains information on all 8,319 students who took the high-stake senior-secondary board examination (at the end of 12th grade) in 10 schools administered by the schooling system, during 2006-2010. The data provides students’ scores in all the subjects in the senior-secondary board examination, as well as information on subject-specific teachers (age, experience, training, education etc.) who taught the students during the two years of senior secondary (11th and 12th grade). In addition, the administrative records also contain subject-wise performance of each student in another high-stake examination - the secondary board examination (at the end of 10th grade).
We estimate the TVA of each teacher by looking at the variation in students’ scores in different subjects in the senior secondary board examination, controlling for the score in the same subject in the secondary board examination.3 In such an analysis, it is important to consider the fact that teachers are not randomly assigned to schools or to the classrooms within the school. For example, good teachers may prefer to work in good schools, or the principal of the school may assign good teachers to lower-ability students. In such cases, estimated TVA may also reflect unmeasurable student characteristics such as their ability and motivation. To address this issue, we estimate the variation in TVA across subject teachers for the same student (as the student is taught different subjects by different teachers), and hence, we can estimate the extent to which each teacher contributes to students’ scores. Since a teacher teaches the student for two years in senior secondary (between 10th grade and 12th grade), the TVA estimates the value added by a teacher in those two years.
We find that teachers matter a great deal as far as achievement of students is concerned, and there is a great deal of variation in TVA across teachers. If we rank teachers according to their TVA, the Standard deviation (SD) of that distribution is 0.366; this implies that, on average, if teacher quality is improved by 1 SD, it will add 0.366 of the SD to the score of the student4. This will move an average student (at 50th percentile) to the 65th percentile5. More importantly, being taught over a two-year period by a high-quality teacher (defined as the 75th percentile teacher) rather than a low-quality teacher (defined as 25th percentile teacher), adds 0.476 of the SD to the score.
The SD of TVA estimates is very much in line with the Slater, Davies and Burgess (2012) findings for the UK. They find that the SD of teacher effects for a two-year course is 0.358. The SD of TVA in our study and in the UK study is almost double of what is reported in the US literature; however, the US TVA estimates are for one year, while ours and Slater, Davies and Burgess (2012) TVA estimates are for a two-year period.
Difficulty to identify ´good´ teachers at the time of hiring
Although teachers matter, the estimated TVA is not explained by observed teacher characteristics such as experience, gender, training or education. Factors such as higher education degree and training that are typically rewarded may not be the ones that matter for teacher quality. It is the unobserved factors such as drive, passion, connection with the students, etc. that are likely to account for a majority of the variation in TVA across teachers.
Although there is potential for improving students’ achievement by improving average teacher quality, it is not so straightforward because good teachers are hard to identify beforehand based on observed teachers’ characteristics. In this scenario, evaluation of teachers based on their contribution to students’ achievement – that is, their ‘value added’ - may be optimal. Although one does not have TVA estimates at the time of hiring, teachers’ regularisation/ promotion decisions could be based on TVA. TVA can be estimated reliably only if teachers have spent some time on job and have taken at least a few classes.
In developing countries, large teacher-student matched administrative datasets (with information on both students and subject teachers and specifying which teachers teach which students) either do not exist or are not accessible to researchers. Our study, which is based on selected private schools, suggests both the importance of teacher quality and the unimportance of the traditional measures of teacher quality. Hence, we make a strong argument for the creation and accessibility of such datasets so that it becomes possible to assess TVA for a more general population of students and teachers.
- Two Indian states - Himachal Pradesh (HP) and Tamil Nadu (TN) - participated in the extended cycle of 2009 OECD PISA (Programme for International Student Assessment) survey of 15-year-olds’ knowledge and skills in reading, math and science. In the reading and math scores, out of the 74 regions participating in PISA 2009, HP and TN were second and third from the bottom respectively, beating only Kyrgyzstan. In science literacy, the results were even worse; while TN was 72nd of 74, HP came in last (Walker 2011).
- Even if a fairly comprehensive set of teachers characteristics is looked at, there is always the possibility that unobserved factors such as ability, motivation and effort of teachers matter a great deal for students’ achievement and may not be observed, or indeed be measurable at all.
- In the secondary board examination, students are required to take 7 subjects, while in the senior secondary board examination they take 5-7 subjects. The choice of subjects is restricted by what is offered by the school. Although, students are not required to take the same set of subjects in secondary and senior secondary, normally there is a substantial overlap in choices across the two examinations. In the case where a subject is taken up for senior secondary examination but not for secondary examination, it is dropped from the analysis. In our dataset, we have both 10th and 12th scores for at least four subjects for 99% of 12th grade students.
- Standard Deviation (SD) is a measure that is used to quantify the amount of variation or dispersion of a set of values.
- If all student scores are arranged in ascending order, then 65th percentile is the value of the score below which 65% of the scores lie.
- Azam, M and G Kingdon (2014), ‘Assessing Teacher Quality in India’, IZA Discussion Paper, 8622.
- Hanushek, EA and SG Rivkin (2012), "The Distribution of Teacher Quality and Implications for Policy", Annual Review of Economics, 4, 131-157.
- Muralidharan, K (2013), ‘Priorities for Primary Education Policy in India´s 12th Five Year Plan’, India Policy Forum 2012-13, Vol. 9, pp. 1-46.
- Rivkin, SG, EA Hanushek and JF Kain (2005), "Teachers, schools, and academic achievement", Econometrica, Vol. 73, pp. 417-458.
- Rockoff, JE (2004), "The impact of individual teachers on student achievement: Evidence from panel data",American Economic Review, 94, 247-252.
- Slater, H, NM Davies and S Burgess (2012), "Do teachers matter? Measuring the variation in teacher effectiveness in England", Oxford Bulletin of Economics and Statistics, 74(5), 629-45.
- Walker, M (2011), ‘PISA 2009 Plus Results: Performance of 15-year-olds in reading, mathematics, and science for 10 additional participants’, Australian Council for Educational Research.