Effect of class size on student achievement at public secondary schools in Azerbaijan

Abstract

This research investigates the impact of class size on student performance in Azerbaijani public secondary schools, utilizing State Examination Center (SEC) final exam scores as a key indicator. Data from the Ministry of Science and Education is analyzed, covering approximately 80,000 9th and 11th-grade students in 4,400 schools. The primary objective is to understand the relationship between class size and academic outcomes. Through correlation and regression analyses, this study explores the dynamics linking class size with student performance. The results consistently reveal a positive correlation between class size and student performance, even after controlling for different school categories. These findings hold significant implications for educational policy, providing valuable insights into the complex interplay of factors influencing student outcomes. The study contributes to the ongoing discourse on class size and student achievement, underscoring that while class size is a pivotal factor, it exists within a broader educational context where numerous variables shape academic success.

Keywords:

• Class size
• student achievement
• regression analysis
• Azerbaijan

REVIEWING EDITOR:

• Sammy King Fai Hui, Curriculum & Instruction, The Education University of Hong Kong, Hong Kong

SUBJECTS:

• Assessment & Testing
• Classroom Management & Organisation
• Teaching & Learning

Introduction

This paper analyzes the effects of class size on student academic achievement in secondary schools in Azerbaijan as measured by State Examination Center (SEC) final exam results. By examining a contentious factor in education, the authors aim to contribute to academic discussions about class size’s impact on student outcomes and provide policymakers with insights for informed decisions on class size adjustments.

The prevailing understanding of the influence of class size on student academic achievement largely hinges on perspectives from teachers, parents, and societal stereotypes. While these viewpoints collectively assert that class size significantly shapes student performance, the empirical foundation to substantiate this notion remains rather limited. Commonly held beliefs are rooted in public perception, with teachers, parents, and the broader community asserting that smaller class sizes are conducive to enhanced learning outcomes. Despite the absence of substantial empirical evidence, these viewpoints persist and contribute to the ongoing discourse surrounding class size’s impact on student achievement.

In the Azerbaijani educational context, pertinent insights from the Ministry of Science and Education reveal a concerning aspect – classrooms accommodating more than 35 students. This situation presents unique challenges for educators, who struggle to allocate sufficient time for vital tasks such as reviewing assignments, offering personalized feedback, and facilitating effective interaction. A conventional classroom setting allows a mere 2 minutes per student during a 45-minute class period, with five minutes designated for administrative aspects. In contrast, classrooms with over 35 students pose a distinct dilemma, as education experts contend that individual students may receive less than 1 minute of direct instructional attention. Additionally, an alarming trend emerges where schools originally designed for 1200 students now accommodate up to 2000 students, leading to an escalation in class sizes. This issue resonates deeply with parental concerns about diminished student-teacher engagement within these expanded classrooms.

Addressing this gap, the Cabinet of Ministers’ decision on ‘Class size standards/rules in public secondary schools’ (2020) presents specific guidelines regarding student numbers per classroom across primary, lower secondary, and upper secondary levels. These guidelines establish a baseline of 20 students per classroom, allowing for potential classroom division if student numbers exceed 30, contingent upon classroom dimensions. Regulations also govern changes in student enrollment during the initial academic term to align with these standards. Further specificity pertains to specialized upper secondary groups, which are expected to maintain student counts between 15 and 20, as stipulated by the same decision.

This study’s primary contribution lies in its exploration of an uncharted territory – the potential impact of class size on student academic achievement within Azerbaijan’s educational milieu. At present, a comprehensive investigation of this nature has not been undertaken in Azerbaijan. Our research transcends the limitations of existing opinions and stereotypes, delving into a systematic analysis that evaluates the interplay between class size and student outcomes on a significant scale.

By encompassing a vast spectrum of 4000 schools across major Azerbaijani cities and regions, our research serves as a pioneering effort within the field of education. Focusing on core subjects including Azerbaijani, Russian, English, and Mathematics – pivotal subjects for 9th-grade students – adds to the depth and relevance of our study. Through centralized standardized testing conducted by the SEC and involving uniform questions for all schools, our research methodology ensures a robust analysis that informs educational policies and practices.

In essence, this study bridges a crucial gap in Azerbaijani educational research, opening avenues for nuanced policy considerations and potential enhancements in teaching practices by shedding light on the unexplored effects of class size on student achievement.

We study the effect of class size on student academic achievement with the following research questions and hypotheses:

RQ: Does class size within Azerbaijani secondary schools impact student achievement as measured by SEC final exam results?

H0: Class size in secondary schools influences student achievement.

H1: Class size in secondary schools does not affect student achievement.

We hypothesize that small class size could potentially yield a positive effect on student academic achievement.

Literature review

When discussing school factors that affect student achievement, class size is one of the most argued points. Numerous studies have studied the relationship between class size and student achievement across different countries worldwide. In general, smaller classes are associated with better student achievement. However, there are studies, which reveal little to no relationship between class size and student achievement. Some researchers even prove the contrary. Despite all the studies, no satisfactory conclusion has been reached so far about this relationship.

Previous research has extensively investigated the relationship between class size and student achievement, and several consistent themes emerge from these studies. The works of Bourke (1986), Finn and Achilles (1990), and the series of STAR (Student/Teacher Ratio) Project studies have contributed to the discourse by suggesting that smaller class sizes can be beneficial for student outcomes. These studies have pointed out potential advantages, such as increased opportunities for individualized attention and personalized instruction, which align with the qualitative insights highlighted by Pedder (2006). This body of research underscores the perception that smaller class sizes can positively influence student engagement, motivation, and achievement.

Early research by Glass and Smith (1978), Bourke (1986), and Finn and Achilles (1990) established a foundational understanding of the benefits associated with smaller class sizes. These studies emphasized the positive correlation between reduced class sizes and enhanced student achievement. Subsequent investigations through the STAR Project further supported these findings. Qualitative studies, as highlighted by Pedder (2006), underscored teachers’ perceptions of smaller classes as facilitating individualized attention, supplementary learning materials, and improved student engagement and motivation.

A conducive learning environment, characterized by enhanced interaction and individualized instruction, was noted by Blatchford et al. (2002). Incorporation of teaching strategies such as group work, pair work, open-ended questions, and consistent feedback further contributed to student achievement (Harfitt, 2013).

European studies consistently associated smaller class sizes with improved student achievement. Research in France and Denmark identified a positive albeit modest relationship between class size and academic performance. Swedish experiments demonstrated measurable improvements in test scores for mathematics and Swedish due to reduced class sizes (Jepsen, 2015). Danish investigations indicated a more pronounced impact of class size on primary school students, with diminishing effects in higher grades (Nandrup, 2015).

Konstantopoulos and Sun (2013) study highlighted the nuanced relationship between class size and teacher effectiveness. While a positive link existed, its realization depended on teachers’ adept use of smaller class opportunities.

Contrasting studies from Zimbabwe and Bolivia emphasized challenges in large classes, including reduced individual attention and difficulties in implementing supplementary materials (Urquiola, 2006; Wadesango et al., 2016). Turkic school research by Koc and Celik, (2015) highlighted a negative relationship between class size and student achievement, emphasizing the need for effective class size management.

While many studies support the benefits of smaller class sizes, inconsistencies persist. Some researchers, including Nye et al. (1999), Hanushek (1986, 1989), and Slavin (1989), suggest limited or non-linear relationships between class size and student achievement. Maasoumi et al. (2005) further indicated that reductions beyond a certain threshold might have minimal impact on test scores, with mixed results in smaller class reductions.

Cross-country studies by Altinok and Kingdon (2011) indicated limited evidence of a positive class size effect on standardized test scores. Norwegian research (Leuven et al., 2008) found no significant impact of class size on 9th-grade student achievement, while Ukrainian findings (Coupé et al., 2015) underscored the relative insignificance of class size on secondary students’ academic performance.

Woessmann (2016) offered an international perspective and emphasized the contextual nature of class size effects. Their study emphasized that while reducing class size can yield positive results, the quality of instruction remains a critical factor. Consequently, policy decisions should consider both class size and teacher education.

The study conducted through analysis of statistical data derived from yearly school censuses and national educational plans/strategies across many countries in Sub-Saharan Africa (SSA) formed the foundation for a strong evaluation of secondary school teaching workloads and class sizes where the authors reveal the shortcomings of using the student-teacher ratio as the main indicator for assessing efficiency and quality in educational services within this crucial sector of education provision, hence not confirming relationship between the two variables (Bennell, 2022).

The authors utilized Norway’s strict class size regulation and matched individual and school data from 1982 to 2011 to investigate the long-term effects of reduced class size on income and education. In contrast to findings in the US and Sweden, they found no significant overall impact on these outcomes. They also examined various external conditions like teacher quality, school choice, and labor market conditions to explore potential factors influencing the class size effect (Falch et al., 2017).

In conclusion, the research landscape surrounding class size and student achievement is characterized by diverse findings and perspectives. The impact of class size on student outcomes is intricate and interwoven with multiple factors, including teacher practices, individual student characteristics, and classroom processes. While some inconsistencies persist, the cumulative body of research highlights the need for a comprehensive understanding of the class size-effect relationship. While some inconsistencies persist, the cumulative body of research highlights the need for a comprehensive understanding of the class size-effect relationship.

In the Azerbaijani context, this research gap is further pronounced, as there is a lack of studies exploring the potential impact of class size on student achievement. Therefore, this study aims to address this research gap and provide insights into the relationship between class size and student outcomes within the Azerbaijani secondary education system. By examining this relationship and considering the unique contextual factors at play, this study seeks to contribute to the broader academic discourse on class size effects and inform educational policymakers and practitioners in Azerbaijan.

Methodology and data collection

The main purpose of this research is to understand the effect, if any, of class size on student academic performance. To answer this question, we run regression analysis where class size is our independent variable while academic performance is dependent variable. As mentioned earlier, for consistency reasons we used the SEC exam as a measure of academic performance. These examinations were conducted at all schools in Azerbaijan at the end of the 9th and 11th grades, where for each subject, students across the country take the exam simultaneously. Therefore, for our study, we considered all 9th and 11th grade students in all public schools in Azerbaijan. For 9th grade students, exam is conducted on two subjects, Azerbaijani or Russian, depending on the medium of instruction for the student and mathematics. On the other hand, the SEC exam is conducted on three subjects for 11th grade students, Azerbaijani or Russian, Mathematics and English.

For our analysis, we used the secondary data provided by the Ministry of Science and Education of the Republic of Azerbaijan in 2021. This data is consistently collected and stored by the Ministry for keeping track of the school and student performances, calculate all the useful school related statistics and for helping with policy making decisions. The main reason for using secondary data in our analysis is its preciseness compared to self-reported data. That is, having collected proper, clean, and reliable data on all the 9th and 11th grade students in all the public schools in Azerbaijan removes the sampling bias from our study and makes our analysis and results more valid and valuable. For each 9th student, we collected data on (i) their school and class in the school, (ii) their gender (iii) their Azerbaijani/Russian language score, and (iv) their Mathematics score. For each 11th student, we collected data on (i) their school and class in the school, (ii) their gender (iii) their Azerbaijani/Russian language score, (iv) their Mathematics score. Reason for collecting each of this information is as follows: we collect the school and class information (bullet point (i)) for every 9th and 11th grade student to identify the class size in all 9th and 11th grade classes in all public schools in Azerbaijan and to understand the size of the class each student is in. We collected the language and mathematics exam scores as a measure of performance of the students. These two sets of variables will help us to measure the effect of the class size on student performances on different types of subjects (language and mathematics). Moreover, we collect gender information of the students to understand whether there is a difference between the class size effect between boys and girls. Overall, we collected data for approximately 80,000 students from the 9th and 11th grades of almost 4,400 public schools in Azerbaijan.

The scores of the students on SEC exams are on different scales for 9th and 11th grade students: for 9th grade exams, the scale ranges from 0 to 100, whereas for 11th grade exams, the scale ranges from 2 to 5. One can question the interesting scale of 2 to 5 for 11th grade students. However, this is a common practice in public schools in Azerbaijan, where 5 is an outstanding grade, 4 is a good grade, 3 is a satisfactory grade, and 2 is a failing grade. We want to stress this difference in grading scales in SEC exams in the 9th and 11th grades so that the reader is not surprised by the difference in the numbers and coefficients in the analysis section.

We first run correlation analysis between class size and academic performance of the students in SEC exam subjects to understand whether there is a strong and statistically significant relation between these variables using hypothesis testing for significance of correlation coefficient. Once we observe a strong relation between these variables, we check for causality relation. That is, for variables who do not show any significant relation, there is no need to check for causality. To check for causality relation on whether the class size has effect on academic performance of the students, we run regression analysis where class size is our independent variable and results of language and mathematics exams are our dependent variables. We run separate regression for each class and subject as we would like to differentiate effects on language type of subjects and quantitative subjects across different grades. Moreover, we re-run the same analysis for boys and girls separately, to see whether there is a difference between the effect level for two different groups.

Analysis and results

The primary objective of this study is to investigate the potential impact of class size on the academic achievement of secondary school students. To ensure the consistency of our analysis, we have chosen to utilize the centralized State Exam Centre (SEC) final examination results, which are conducted nationwide at the conclusion of the 9th and 11th grades. In consideration of potential variations in the influence of class size on performance across different subjects, we have opted to examine each subject separately. It is worth noting that the examinations for the 9th and 11th grades differ slightly in their structure and grading scales, necessitating a separate analysis for each. To commence our analysis, we will begin with the 9th-grade examination results.

As mentioned earlier, in the 9th grade exam, there are only two subjects: Azerbaijani/Russian language and mathematics. First, we run correlation analysis on class size and the average exam results in each subject to understand whether there is a strong significant relation between these variables. We use hypothesis testing for significance of correlation coefficient to assess the strength of the relation. The results are summarized in Table 1.

Table 1. Results of correlation analysis between (i) class size, (ii) Azerbaijani/Russian language grades and (iii) Mathematics grades.

	1	2	3
1. Class size	–
2. Average of Azerbaijani/Russian	0.346***	–
3. Average of Mathematics	0.304***	0.895***	–

***p<.01.

As shown in Table 1, the correlation between class size and results in both subjects is approximately 30–35%. Although this number does not seem very high, the relationship is highly significant (even at 1% level of significance). Moreover, the correlation between the two exam results was strong and highly significant as well, it is almost 90%.

Once we showed the strong relation between class size and exam results, we would like to understand the causality relation between the variables. That is, we will analyze the level at which class size has an impact on exam results. To do so, we run simple linear regressions where class size is independent variable, and each exam result is dependent variable. Since there are only two variables (one independent variable), simple linear regression seems like a plausible tool to be used to conduct this analysis. We run two separate regressions (one for each subject). Note that exam results in both subjects are on a scale of 0–100. The findings from our regression analysis are presented in Table 2.

Table 2. Results of simple regression of class size on (i) Azerbaijani/Russian language and (ii) Mathematics grades.

(i)	B	SE	t	R^2	F
Model				0.12	864.8***
Azerbaijani/Russian	0.59	0.02	29.4***
(ii)	B	SE	t	R^2	F
Model				0.1	646***
Mathematics	0.48	0.02	25.4***

***p<.01.

As the results in Table 2 show, for both subjects, class size is a significant predictor of exam performance (even at 1% level of significance). That is, for both subjects, the higher the class size, the better students perform: increasing class size by 1 person, is expected to increase average grade by 0.59 points for Azerbaijani/Russian language and by 0.48 points for Mathematics. This may seem to be a surprising result, as one would expect students to perform better in an environment where there are fewer students, less distraction, and more one-to-one interactions with the instructor. However, in Azerbaijan, most public schools in rural regions and small villages have a lower number of students in the class, and teaching quality in those schools may not be as good as the teaching quality in most big city schools. Therefore, it is natural to observe a lower performance among students from schools in rural regions and small villages. Moreover, some quality-wise popular public schools in big cities have a high demand and fill their classes up to the maximum limit set by the Ministry of Science and Education. That is, in the most quality-wise popular schools in the country, the class size is around 39 students. Therefore, it is natural to observe higher performance from the students in those quality-wise popular schools as those schools are the ones with the best instructors and provide best teaching quality.

Although these two cases with small class-sized schools in rural regions and large class-sized schools in big cities somehow explain our results of positive relationships between class size and academic performance, these two extreme cases may distort our results. To determine whether this is true, we measured the class size effect when these two extreme case schools were removed. That is, we remove the schools with very small class sizes, which are the schools in rural regions, and the schools with very large class sizes, which are popular high-performing schools, and re-run our regression analysis only for the remaining schools that have class size of 15–30 students. Table 3 presents our results.

Table 3. Results of simple regression of class size on (i) Azerbaijani/Russian language and (ii) Mathematics grades for medium size classes.

(i)	B	SE	t	R^2	F
Model				0.09	369.18***
Azerbaijani/Russian	1.04	0.05	19.2***
(ii)	B	SE	t	R^2	F
Model				0.07	285.7***
Mathematics	0.89	0.05	16.9***

***p<.01.

Although we eliminated schools with very small and very large class sizes, our results for both subjects are the same: class size is a significant predictor of exam performance. As before, we found that, for both subjects, the higher the class size, the better the students performed. However, differently from the case where we consider all schools, with the schools with medium size increasing class size by 1 person is expected to increase average grades by 1.04 points for Azerbaijani/Russian language and by 0.89 points for Mathematics. These numbers were higher (almost double) than those in the case where we considered all schools. Once again, this relation may seem strange, especially after we eliminated very small and very large sized schools from our analysis. However, since the behavior of the people remains the same, even if we eliminate the extreme cases, we will still obtain the same results. That is, parents would want their children to study in good popular schools. Therefore, demand for those good schools is always high, while demand for relatively low-quality schools is low. Therefore, highly demanded, and thus schools with high class sizes will have better students compared to the less demanded schools, which end up having relatively smaller class size.

Another interesting question that has been widely analyzed in the literature is the difference in the effect of class size on the performance of students across different genders. To answer this question, we separated students based on gender and re-ran our correlation and regression analysis separately for each group. The results of the correlation analysis for male and female students are separately presented in Table 4.

Table 4. Results of correlation analysis for (i) males and (ii) females.

(i)	1	2	3
1. Class size	–
2. Average of Azerbaijani/Russian	0.259***	–
3. Average of Mathematics	0.218***	0.895***	–
(i)	1	2	3
1. Class size	–
2. Average of Azerbaijani/Russian	0.299***	–
3. Average of Mathematics	0.278***	0.861***	–

***p<.01.

As we can see from Table 4, the correlation between class size and results in both subjects is approximately 22–26% for males and 28–30% for females. Although this number does not seem very high, the relationship is highly significant (even at 1% level of significance). It should be noted that the correlation was slightly higher for female students than for male students. As before, the relationship between different exam results was highly significant.

Once again, after showing the existence of significant relation between class size and exam results, we would like to understand the causality relation between the variables. As before, to do so we run simple linear regressions for male and female students separately, where class size is independent variable, and each exam result is dependent variable). The results are presented in Tables 5 and 6, respectively.

Table 5. Results of simple regression of class size on (i) Azerbaijani/Russian language and (ii) Mathematics grades for male students.

(i)	B	SE	t	R^2	F
Model				0.07	457.9***
Azerbaijani/Russian	0.72	0.03	21.4***
(ii)	B	SE	t	R^2	F
Model				0.05	317.7***
Mathematics	0.58	0.03	17.8***

***p<.01.

Table 6. Results of simple regression of class size on (i) Azerbaijani/Russian language and (ii) Mathematics grades for female students.

(i)	B	SE	t	R^2	F
Model				0.09	615.6***
Azerbaijani/Russian	1.04	0.04	24.8***
(ii)	B	SE	t	R^2	F
Model				0.08	524.7***
Mathematics	0.88	0.04	22.9***

***p<.01.

As our results in Tables 5 and 6 show, for both subjects, class size is a significant predictor of exam performance for both male and female students. As before for both subjects, the higher the class size, the better both male and female students perform: increasing class size by 1 male student, is expected to increase average grade of male students by 0.72 points for Azerbaijani/Russian language and by 0.58 points for Mathematics, and increasing class size by 1 female student, is expected to increase average grade of female students by 1.04 points for Azerbaijani/Russian language and by 0.88 points for Mathematics. That is, increasing class size increases the performance of female students more than that of male students in both subjects. The reasoning for positive relation is like what we mentioned earlier: good urban schools are usually more crowded compared to relatively weak suburban schools. Also, we note that, the impact of single student increase in class size is higher for female students than male students. One explanation can be that the female students are relatively more competitive and thus adding more students increases the competition level and thus the hard work students put into their studies. Similar reasoning may be true for male students as well, but in a relatively lower level.

Grade 11 exams

As mentioned earlier, in the 11th grade exam, there are three subjects: in addition to Azerbaijani/Russian language and Mathematics, there is also English exam. As we did for 9th grade exam, we first run correlation analysis on class size and the average exam results in each subject to understand whether there is a strong significant relation between these variables. We use hypothesis testing for significance of correlation coefficient to assess the strength of the relation. Table 7 summarizes the results.

Table 7. Results of correlation analysis between class size, Azerbaijani/Russian language grades, Mathematics grades and English language grades.

	1	2	3	4
1. Class size	–
2. Average of Azerbaijani/Russian	0.32***	–
3. Average of Mathematics	0.337***	0.889***	–
4. Average of English	0.367***	0.853***	0.864***	–

***p<.01.

As shown in Table 1, the correlation between class size and results in all three subjects was approximately 32–37%. Although this number does not seem very high, the relationship is highly significant (even at 1% level of significance). Moreover, the correlation between the exam results of any two participants was also very strong and highly significant.

Once we showed the strong relation between class size and exam results, we would like to understand the causality relation between the variables. As before, to do so we run simple linear regressions where class size is independent variable, and each exam result is dependent variable. We run two separate regressions (one for each subject). Note that the exam results for all three subjects were in the range of 2–5. The findings are presented in Table 8.

Table 8. Results of simple regression of class size on (i) Azerbaijani/Russian language, (ii) Mathematics and (iii) English language grades.

(i)	B	SE	t	R^2	F
Model				0.05	277***
Azerbaijani/Russian	0.02	0.001	16.6***
(ii)	B	SE	t	R^2	F
Model				0.06	296.9***
Mathematics	0.02	0.001	17.2***
(iii)	B	SE	t	R^2	F
Model				0.09	489.7***
English	0.024	0.001	22.1***

***p<.01.

As the results in Table 8 show, for both subjects, class size is a significant predictor of exam performance. Moreover, we found that for all three subjects, the higher the class size, the better students perform: increasing class size by 5 people, is expected to increase average grade by 0.1 points for Azerbaijani/Russian language, by 0.1 points for Mathematics and by 0.12 points for English language. As we explained earlier, most public schools in rural regions and small villages of Azerbaijan have a lower number of students in the class, and teaching quality in those schools may not be as good as the teaching quality in most big city popular public schools that usually have high demand and fill their classes up to the maximum. Therefore, as it was for 9th grade students, it is natural to observe higher performance from the students in those popular schools as those popular schools are the ones with the best instructors and provide best teaching quality. Also note that although the magnitude of increase in exam results is lower compared to that for 9th grade students, the scale in which two scores are measured are different. Thus, one should not compare the magnitudes of the increases in the exam results.

As before, to see whether schools with very high and very low number of students reason for this positive relation are, we remove these schools and re-run our regression analysis only for the remaining schools that have class size of 15–30 students. Table 9 presents the results of the study.

Table 9. Results of simple regression of class size on (i) Azerbaijani/Russian language, (ii) Mathematics and (iii) English language grades for medium size classes.

(i)	B	SE	t	R^2	F
Model				0.06	174***
Azerbaijani/Russian	0.033	0.003	16.2***
(ii)	B	SE	t	R^2	F
Model				0.06	180.6***
Mathematics	0.04	0.003	13.4***
(iii)	B	SE	t	R^2	F
Model				0.07	210.9***
English	0.04	0.003	14.5***

***p<.01.

Although we eliminated schools with very small and very large class sizes, our results for all three subjects are the same as before: the class size is significant predictor of exam performance. As before, we found that for all three subjects, the higher the class size, the better students perform. When we only consider schools with medium size, increasing class size by 5 people is expected to increase average grades by 0.165 points for Azerbaijani/Russian language, by 0.2 points for Mathematics and by 0.2 points for English language. These numbers are higher (almost double) than the number in the case where we considered all schools. Once again, this relation may seem strange, especially after we eliminated very small and very large sized schools from our analysis. However, since the behavior of the people remains the same, even if we eliminate the extreme cases, we will still obtain the same results. That is, parents would want their children to study in good popular schools. Therefore, demand for those good schools is always high, while demand for relatively low-quality schools is low. Therefore, highly demanded, and thus schools with high class sizes will have better students compared to the less demanded schools, which end up having relatively smaller class size.

Next, we consider male and female students separately and try to understand the impact of class size for two groups separately. We re-run our correlation and regression analysis separately on each group. Our results for correlation analysis are presented in Table 10.

Table 10. Results of correlation analysis for (i) males and (ii) females.

(i)	1	2	3	4
1. Class size	–
2. Average of Azerbaijani/Russian	0.11***	–
3. Average of Mathematics	0.134***	0.86***	–
4. Average of English	0.185***	0.78***	0.797***	–
(ii)	1	2	3	4
1. Class size	–
2. Average of Azerbaijani/Russian	0.23***	–
3. Average of Mathematics	0.234***	0.836***	–
4. Average of English	0.29***	0.78***	0.78**	–

**p<.05; ***p<.01.

As we can see from Table 10, correlation between class size and results in all three subjects is around 11–18% for males and 23–29% for females. As it was for 9th graders, this correlation is slightly higher for female students than male students. Moreover, the relation between class size and exam results is highly significant both for male and female students (even at 1% level of significance).

Once more, having established a significant relationship between class size and exam results, our next aim is to explore the causality between these variables. To achieve this, we conduct separate simple linear regression analyses for male and female students, with class size as the independent variable and each exam result as the dependent variable. Results for male students are presented in Table 11 and for female students in Table 12.

Table 11. Results of simple regression of class size on (i) Azerbaijani/Russian language, (ii) Mathematics and (iii) English language grades for male students.

(i)	B	SE	t	R^2	F
Model				0.02	60.5***
Azerbaijani/Russian	0.013	0.001	7.78***
(ii)	B	SE	t	R^2	F
Model				0.02	90.7***
Mathematics	0.02	0.002	9.53***
(iii)	B	SE	t	R^2	F
Model				0.034	175.5***
English	0.024	0.002	13.2***

***p<.01.

Table 12. Results of simple regression of class size on (i) Azerbaijani/Russian language, (ii) Mathematics and (iii) English language grades for female students.

(i)	B	SE	t	R^2	F
Model				0.054	277***
Azerbaijani/Russian	0.034	0.002	16.6***
(ii)	B	SE	t	R^2	F
Model				0.055	278.3***
Mathematics	0.036	0.002	16.7***
(iii)	B	SE	t	R^2	F
Model				0.084	440***
English	0.046	0.002	20.9***

***p<.01.

As indicated by the results presented in Tables 6 and 10, it becomes evident that the class size significantly influences the performance of both male and female students across all three subjects. Consistently, for all these subjects, a larger class size is associated with improved performance for both male and female students. Specifically, an increment of 5 male students in the class is projected to raise the average grade by 0.065 points for Azerbaijani/Russian language, 0.1 points for Mathematics, and 0.12 points for English language. Conversely, an increase of 5 female students in the class is anticipated to yield a more substantial improvement, with a 0.17-point increase for Azerbaijani/Russian language, 0.18 points for Mathematics, and 0.23 points for English language. Thus, it is evident that augmenting class size has a disproportionately positive impact on the performance of female students, surpassing the improvement observed in male students across all three subjects. As we explained for the 9th graders, the reasoning for positive relation is like what we mentioned earlier: good urban schools are usually more crowded compared to relatively weak suburban schools. Also, we note that, the impact of single student increase in class size is higher for female students than male students. Once again, one explanation can be that the female students are relatively more competitive and thus adding more students increases the competition level and thus the hard work students put into their studies. Similar reasoning may be true for male students as well, but in a relatively lower level.

Discussion

When parents make choices about their children’s education, various factors come into play, including teacher competence, school environment, and class size. While class size may not always be the primary consideration, it remains an important factor. This study aims to explore the relationship between class size and student achievement, aiming to determine if class size can serve as a predictor of student success. We also seek to identify whether certain student demographics, such as gender, are more influenced by larger class sizes. These are the central questions addressed in this paper.

We evaluate student achievement using SEC (State Examination Center) exams, which encompass subjects such as Azerbaijani/Russian language and mathematics for 9th-grade students, and English language for 11th-grade students.

As evident in Tables 1 and 7, we observe a robust correlation between class size and student performance in all subjects for both grade levels, with correlation coefficients ranging from 30 to 40%. Significance tests confirm the strength of these correlations, indicating that these relationships are highly statistically significant across all grades and subjects. In other words, a clear and meaningful connection exists between class size and student achievement. Notably, this correlation is consistently positive across all courses, which may appear surprising initially. However, similar findings have been documented in schools in Israel and France (Shafrir et al., 2016).

Our subsequent analysis delves into whether class size serves as a significant predictor of student performance. Individual simple linear regression analyses were conducted for each grade and subject. The results presented in Tables 2 and 8 indicate a consistent trend: across all grades and subjects, class size emerges as a statistically significant predictor of exam performance.

The positive relationship between class size and student performance, observed in our study, may be contextualized. Smaller class sizes are often characteristic of public schools in rural areas and small villages in Azerbaijan, which may not consistently offer the same level of teaching quality as their urban counterparts. This can lead to lower academic performance among students in these regions.

Conversely, certain popular public schools in larger cities tend to maximize their class sizes to meet the Ministry of Science and Education’s limits. As a result, these schools, which typically have highly qualified instructors and top-tier teaching quality, often yield higher student performance. Considering these scenarios, we excluded schools at the extremes of class size and re-conducted regression analysis for the remaining schools, where average class sizes ranged from 15 to 30 students.

To our surprise, the results remained consistent. Across all grade levels and subjects, class size continued to emerge as a significant predictor of student performance. Moreover, after excluding schools at the extremes, increasing class size by one student in the remaining schools led to a more substantial improvement in student performance compared to our analysis involving all schools. Several factors may explain these results, including collaborative learning environments, a competitive environment, and teacher quality.

One compelling reason for the observed positive relationship between class size and student performance may be the competitive environment fostered by larger classes. Ranking as a top student often serves as a strong motivator for diligent students. However, achieving this status within larger classes can pose a more significant challenge, incentivizing students to exert greater effort in their studies and create a positive feedback loop.

These reasons may be more pronounced among female students than their male counterparts, who tend to form study groups and engage in collaborative learning more often. Additionally, female students often exhibit a stronger drive to excel academically. Our gender-segregated analysis supports this, showing a notably stronger correlation between class size and student performance among female students, underscoring the gender-specific nature of these dynamics.

It is important to state that in our study, we observe a correlation between class size and student performance, which means there is a statistical association between these two variables. Specifically, we find a positive correlation, indicating that as class size increases, student performance, as measured by SEC final exam scores, tends to improve. However, it is crucial to emphasize that correlation does not imply causation which underscores the need for additional investigations to identify the underlying factors responsible for the observed correlation. Through regression analyses and a deeper examination of the dataset, we aim to uncover the potential causative factors that explain the relationship between class size and student performance.

Incorporating the recent article ‘Class Size and Student Achievement: A Modern Meta-Analysis’ by Opatrny et al. (2023), provides a valuable contemporary perspective to our discussion on the relationship between class size and student achievement. The study contributes to the existing body of literature by summarizing and synthesizing the results of numerous studies, allowing for a more comprehensive assessment of the class size-student achievement connection. Their findings reaffirms the idea that the empirical evidence does not consistently demonstrate a strong relationship between smaller class sizes and improved academic outcomes.

Similarly, in our analysis, while we observe a positive correlation, we cannot conclude that larger class sizes directly cause better student performance. It is equally plausible that other factors, which we need to explore further, may be influencing both class size and student performance.

By integrating the insights from this modern meta-analysis into our discussion, we can further emphasize the consensus among contemporary research that class size, while important, may not be the sole determinant of student performance. This addition highlights the ongoing and evolving nature of the discourse surrounding class size and its influence on student achievement in the context of educational policy. By distinguishing between correlation and causation, our study maintains a rigorous and nuanced approach to understanding the dynamics at play in Azerbaijani public secondary schools. This methodological clarity allows us to draw more accurate and meaningful conclusions about the factors that truly impact student achievement, which, in turn, can inform educational policy discussions and decisions.

Limitations

As a measure of student achievement, we did not rely on the grades assigned by teachers, as we were uncertain about the objectivity of these assessments. Instead, we employed standardized test scores for this study. However, it is essential to acknowledge that standardized test results, such as the SEC results utilized in our study, also have limitations, as they focus primarily on specific skills of students (Geiser & Santelices, 2007). Other significant factors, such as the socioeconomic status of the students, can influence their academic performance. Regrettably, we were unable to collect the necessary data to conduct an analysis while controlling for socioeconomic status.

Conclusion

In conclusion, this study presents both theoretical insights and practical implications that challenge traditional assumptions regarding class size and its impact on student achievement. The unexpected positive correlation between class size and performance prompts a reevaluation of prevailing educational paradigms and highlights the potential benefits of collaborative learning and competition within larger class settings. This calls for a reconsideration of pedagogical strategies and class size reduction policies. Moreover, the persistence of this correlation, even after excluding extreme cases, underscores the need for comprehensive interventions that consider diverse classroom contexts.

Furthermore, as policymakers and educators navigate the complex landscape of educational reform, these findings offer a fresh perspective and call for a more nuanced approach to improving student outcomes, taking into account the dynamics of student interaction, gender disparities, and the broader socioeconomic context in which education is delivered. It is worth noting that this study utilized standardized test scores as a measure of student achievement, recognizing their limitations in capturing the full range of student abilities and competencies. Future research could explore additional factors, including socioeconomic status, to gain a more comprehensive understanding of student performance. This study not only advances our theoretical understanding of class size effects but also provides practical guidance for shaping effective educational policies and practices, aiming to foster optimal learning environments for all students.

Acknowledgement

Authors gratefully acknowledge the financial support of the Institute of Education of the Republic of Azerbaijan.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Institute of Education of the Republic of Azerbaijan.

Notes on contributors

Azar Abizada

Azar Abizada is an Associate Professor of Economics at School of Business, ADA University. He holds PhD in Economics from University of Rochester, NY, USA. The main areas of interest of the author are College admissions, School Choice, Mechanism Design and Education Economics.

Sevinj Seyidova

Sevinj Seyidova is a Research Fellow at The Institute of Education of the Republic of Azerbaijan and PhD candidate at Baku State University.