![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
Recent studies emphasize the importance of socio-emotional skills, but little is known about how everyday classroom practices impact development of these skills. Using data from the Czech Republic, we show that modern practices such as working in small groups improve these skills. Intrinsic motivation and self-confidence are particularly positively affected. Moreover, modern practices have no adverse effects on test scores. On the other hand, standard practices such as lecturing and memorizing have no impact on socio-emotional skills or test scores. Our results highlight that changing the composition of teaching practices slightly can have a substantial positive impact on socio-emotional skills.
Acknowledgments
We would like to thank Aline Bütikofer, Štěpán Jurajda, Barbara Pertold-Gebicka, Daniel Münich and Katrine Vellesen Løken and two anonymous referees for their valuable comments. We also thank the participants of the CSE conference, the Lisbon Workshop on Economics, Statistics and Econometrics of Education and the KU Workshop on Educations Economics for their comments. All errors remaining in this text are our own.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. They are also referred to as non-cognitive skills or personality traits.
2. In total, 59 countries participated in the 2007 wave.
3. Not all students are taught by five different teachers, 44% of students are taught by 4 different teachers and 13% by two or three teachers.
4. Section 4.2 compares our linear estimation with a range of alternative fixed-effects ordered logit model estimators. Those models use the original four point scale.
5. The results are qualitatively and quantitatively similar when using alternative definitions of intrinsic motivation (‘I find subject boring’, ‘I like subject’) and self-confidence (‘Subject is not my strength’).
6. Modern teaching practices are described slightly differently between math and science subjects in the questionnaire – ‘we work together in small groups’ for math and ‘we work in small groups on an experiment or investigation’ in science subjects. However, both descriptions represent the same activity, namely, working in a small group.
7. Our results are robust when the indices are created from simple means.
8. We restricted the original dataset of 24,225 observations for missing values in all outcome variables and teaching practices. Our dataset contains 93% of all observations. The results are robust to dropping missing values in all control variables.
9. In our sample, almost all classes (99%) remain the same across subjects. The fact that Czech children are exposed to the same peers in all their lessons reduces the chance that the results could be affected by students benefiting differently dependent on their exposure to better or worse peers (Hoxby Citation2000).
10. Using fixed-effects we limit our analysis to within student variation only and neglect across student variation. It could be argued that we explore only a small share of total variation. Table A7 summarizes both within and across student variation for all outcome variables. First, within student variation in socio-emotional skills is slightly higher than across student variation. Second, this is not true for test scores where across student variation is approximately double the size of within student variation. This documents that we use a sizeable part of the relevant variation for our analysis.
11. In grades 6 and 8, students sort into grammar schools. Grammar schools focus on preparing students to study at university. Students (approximately 8% of our sample) typically sort there based on their skills and motivation through admission exams. However, that selection is not subject-specific.
12. Table A3 estimates heterogeneity in effects across subjects. Math is the only subject where gains from standard practices are significantly higher and gains from the modern practices significantly lower for motivation than in other subjects. The coefficients for the other subjects are in accordance with our overall findings. This suggests that modern practices are especially important for socio-emotional skills in science subjects.
13. The results are robust to trimming 5% of the highest and lowest values for standard and modern practices (Table A5), minimizing the possibility that they are driven by extreme values (Figure A1).
14. We explore the impacts of teaching practices across the self-confidence distribution, since overconfidence has been shown to have negative consequences in various contexts (Dunning et al. Citation1990; Koellinger, Minniti, and Schade Citation2007; Berner and Graber Citation2008). Splitting the sample into two groups with higher and lower self-confidence shows that higher impacts are concentrated in the lower part of the self-confidence distribution (Table A6).
15. Table A2 documents the relationship between teaching practices, socio-emotional skills and test scores. Not surprisingly, higher test scores are associated with higher socio-emotional skills. The size of the coefficients for teaching practices remains almost identical when test scores are added into the socio-emotional skills regressions. It seems implausible that the effects of teaching practices are driven via effects on test scores. However, it is impossible to clearly disentangle the relationship between these variables using the TIMSS dataset.
16. A potential issue with measurement error in the teaching practice indices is investigated in Table A5. We create a variable called Consensus which measures agreement on teaching practices within a classroom. It is constructed as a class average of the difference between individual answers and the mean class value for the given teaching practices in a given subject. The final variable is an average across teaching practices and subjects (). Even though lower consensus on teaching practices is correlated with lower socio-emotional skills values, the coefficients are insignificant and do not change the value of the TP coefficients.
17. Similar findings were obtained for other functional forms. The results are available upon request.
18. Oster (Citation2013) builds on the previous work of Altonji, Elder, and Taber (Citation2005). The assumption of the proportionality of selection can be written as , where X is the treatment,
is a vector of observables and
of unobservables. Then, δ represents the strength of selection on unobservables to observables.
19. Bear in mind that motivation and self-confidence are self-reported and measured on a 4 point scale.
20. The results also hold when we look at all classes specializing in any subject (languages, sports, music and arts). These results are available upon request.
21. Estimates for the other outcomes variables are not included but show similar patterns. Dummy variables for teacher motivation, indicator variables for imputed values and dummy variables for each subject are not reported in . However, the results are consistent with the results presented in this section.
22. Complicated procedures for executing FE ordered models prevent us from using exactly the same clustering procedure appropriate for the TIMSS dataset. This could be the reason for a few of the dissimilarities in the significance levels across the models.