How are teachers teaching? A nonparametric approach

Abstract

This paper examines which configuration of teaching activities maximizes student performance. For this purpose a nonparametric efficiency model is formulated that accounts for (1) self-selection of students and teachers in better schools and (2) complementary teaching activities. The analysis distinguishes both individual teaching (i.e., a personal teaching style adapted to the individual needs of the student) and collective teaching (i.e., a similar style for all students in a class). Moreover, we examine to which group of students the teacher is adapting his/her teaching style. The model is applied on the Trends in International Mathematics and Science Study 2003 data. The main results show that high test scores are associated with teaching styles that emphasize problem solving and homework. In addition, teachers seem to adapt their optimal teaching style to the 70% least performing students.

Keywords:

• nonparametric
• teaching style
• TIMSS 2003
• problem solving
• performance

Acknowledgement

We would like to thank two anonymous referees for useful comments. The usual caveat applies.

Notes

1. Test scores on math are considered as a proxy for student performance. Clearly, this is only one student outcome. A more complete list should include motivation, general training in citizenship, well-being at school, etc.

2. The nonparametric efficiency model is rooted in the Data Envelopment Analysis (DEA; Charnes, Cooper, and Rhodes 1978) literature. It is a linear programming technique that allows us to estimate the most efficient set of inputs (i.e., teaching activities) to obtain the outputs (i.e., test scores).

3. The indication of problem solving, as provided by the teachers in the TIMSS data, does not provide insights in the level of problem solving. At one extreme, problem solving could indicate one identical problem for all students in the class. At another extreme, problem solving could indicate a different problem suited to the level of each student.

4. In the current setting, constant returns to scale are a straightforward assumption as the set of lecturing activities sums for all students to t.

5. Note that the difference between the first and second best could be exploited for policy purposes. In particular, it is possible to detect students who have, due to their background characteristics, a different optimal teaching style than the ‘average’ student in the class. These students could be provided with resources foreseen for student counseling (e.g., additional training on math, speech therapy, or pedagogical training).

6. For example, this could represent teachers in apprenticeship or short replacements who do not observe information on student background.

7. Individual results are available upon request.

8. We note that a sensitivity analysis showed that the results are very similar if we use other deviation than 10%.

9. As the grid is determined on 0.05 (i.e., α increases by 0.05) some discontinuity arises in the figure. Therefore, we also present the polynomial trend line of order six. The trend line hides the discontinuity of the grid and shows a clear trend in the results.