Grade inflation effects of capacity expansion in higher education: a longitudinal study in undergraduate teacher education programs from 2003 to 2022

Türkiye’s higher education system has rapidly expanded, growing to 205 universities and about 7 million students, with 128 universities founded after 2006. This expansion includes substantial increases in teacher training programs and enrollments (faculties of education increased from 48 to 99; enrollment from ~139,000 to >300,000). While expansion may boost human capital and regional development, concerns have emerged about maintaining educational quality amid limited physical resources, insufficient faculty growth, and rapid privatization. In teacher education, additional pressures include rising numbers of graduates outpacing public-school hiring and potentially lower-achieving students entering education faculties due to capacity increases. This study addresses potential grade inflation in teacher education, asking: (RQ1) whether the share of graduates with high honors (GPA ≥3.50) and honors (3.00–3.49) changed over time, (RQ2) whether grade inflation exists, and (RQ3) whether gender, admission scores, teaching field, program establishment year, and student–faculty ratios affect graduation GPAs. Using longitudinal GPA data from 173,232 graduates (2003–2022), the study evaluates whether concerns about grade inflation in Turkish teacher education are warranted.

The background synthesizes definitions and mechanisms of grade inflation: awarding higher grades without commensurate achievement, increasing the frequency of high grades and reducing their informational value. Prior work (largely US-focused) shows widespread grade inflation, its links to student evaluations of teaching, career orientation, rising tuition, and institutional competition for resources. Grade inflation can weaken academic standards, reduce the signaling value of grades, lower student motivation and study time, and complicate distinctions among students. Instructor incentives (seeking favorable evaluations, reduced workload from avoiding disputes) may encourage lenient grading. Institutional policies can counter inflation (e.g., Princeton’s cap on A grades), and clearer grading criteria may help. Literature notes mixed findings on determinants such as admission selectivity and disciplinary differences; some studies find lower grades in science/math fields relative to non-STEM. The study positions three issues: rising shares of high-honors graduates, artificial increases in GPAs, and contributing factors affecting GPAs.

Design: Secondary research using anonymized institutional graduate databases as data sources. Data were collected from faculties of education/educational sciences in Türkiye. Sample and data selection: Teacher training programs span 32 fields; programs were narrowed to those training for basic education levels (science, primary school mathematics, English, preschool, guidance and psychological counseling, classroom, social studies, and Turkish) in 70 faculties, excluding programs with aptitude-test admissions (physical education, music, art), low-enrollment secondary/high school fields (e.g., physics, chemistry, mathematics, history, Turkish language and literature), rare language programs (German, Arabic, French, Japanese), and special education programs that were redesigned. Programs were ranked annually (1999–2022 intake; graduates 2003–2022) by the last graduate’s GPA and categorized (very high, high, medium, low, very low). One program covering all grade levels per year was selected. Final dataset covers 32 education faculties and 173,232 graduates (65.5% female) from 2003–2022. Standardization: Due to a grading scale change (pre-2006 out of 100; post-2006 out of 4), GPAs were converted using a semantic letter-grade mapping to a 4.0 scale (AA=4.00, BA=3.50, …, FF=0.00). General Weighted GPA ranges from 2.00 to 4.00. Analytical approach: For longitudinal analysis addressing RQ1 and RQ2, a “real” random effects estimator (REE) per Greene (2005) was employed to account for unit- and time-specific error components and heterogeneity. Covariates included student characteristic (percentage female), university characteristics (percentile of the lowest-scoring admitted student, program establishment year/faculty age, public vs. private status), and year dummies (N−1) to control for year effects. For RQ3, ANOVA and t tests examined effects of: gender; admission score (high, medium, low); teaching field (science, primary school mathematics, English, preschool, guidance and psychological counseling, classroom, social studies, Turkish); program establishment (younger: 1–5 years; medium: 6–10 years; experienced: ≥11 years); and student–faculty ratio (lower ≤35; intermediate 36–50; higher ≥50). Analyses used SPSS v25 and Excel. Model summary (Table 2): Random effects results indicated significant coefficients for ln(% female) = 0.37 (SE 0.073, p<0.001), ln(percentile of last admitted student) = −0.41 (SE 0.059, p<0.001), faculty age = −0.13 (SE 0.182, p<0.001); non-profit private status nonsignificant (0.03, SE 1.983). σi=0.095, σε=0.084, ρi=0.796, within-R²=0.635; N=173,232 across 32 universities.

- GPA inflation over time: Mean graduation GPA rose from 2.83 to 3.34 between 2003 and 2022 (≈18% increase). Peaks occurred in 2012 and during the COVID-19 period (2021–2022). - Distribution shifts (2003 vs. 2022): High honors (GPA ≥3.50) increased from 4.6% to 36.5%; Honors (3.00–3.49) from 37.4% to 56.5%; Good (2.50–2.99) decreased from 47.8% to 6.9%; Intermediate (2.00–2.49) from 10.2% to 0.1%. - Random effects determinants of high honors: • Gender composition: A 1% increase in the proportion of female students was associated with a 0.37% increase in high-honors graduates (p<0.001). • Admission selectivity: A 10% decrease in admission score percentile (i.e., lower selectivity) increased the share of high-honors graduates by 4.1% (p<0.001). • Program age: A 10% increase in program age decreased high-honors share by 1.3% (p<0.001), implying newer programs produce more high-honors graduates. • Public vs. private: No significant effect. - Field differences (ANOVA, P<0.001): Lowest mean GPA in science teaching (M=2.93, SD=0.36); highest in guidance and psychological counseling (M=3.24, SD=0.31). - Program experience (ANOVA, P<0.001): Experienced (≥11 years) programs had lowest GPAs (M=2.95, SD=0.30); younger (1–5 years) had highest (M=3.11, SD=0.26); medium (6–10 years) M=3.04 (SD=0.27). - Student–faculty ratio (ANOVA, P<0.001): Programs with higher ratios (≥50 students per faculty) had the highest GPAs (M=3.49, SD=0.10); those with lower ratios (≤35) had lower GPAs (M=3.09, SD=0.13). - Temporal context: 2012 spike linked to capacity expansion and cohort size anomalies; 2021–2022 spikes linked to pandemic-related assessment changes and relaxed attendance requirements.

Findings indicate substantial grade inflation in Turkish teacher education from 2003–2022, consistent with international literature documenting rising grades. After controlling for observable factors via REE, a persistent upward shift remained, supporting the presence of grade inflation rather than improvements in student performance. Peaks in 2012 align with rapid capacity expansion and a temporary drop in high school graduates without a corresponding reduction in university quotas, facilitating entry of lower-achieving students and lenient grading practices. COVID-19 disruptions in 2020–2021 engendered remote instruction, reduced attendance requirements, and simplified assessments, coinciding with higher GPAs in 2021–2022. Institutional determinants align with proposed mechanisms: higher student–faculty ratios are associated with higher GPAs, suggesting pressures toward leniency when instructional loads rise. Newer programs exhibit higher GPAs than established ones, possibly reflecting less mature standards or oversight. A negative relationship between admission selectivity and high-honors shares suggests selective institutions may impose stricter grading (or adopt policies to counter inflation). Field-level differences mirror prior findings, with lower GPAs in science-related fields and higher GPAs in social science-oriented areas. Gender composition positively relates to high honors, echoing evidence of higher female academic attainment. Collectively, results address the research questions: the proportion of high-honors and honors graduates has increased markedly (RQ1); grade inflation is present (RQ2); and gender, admission selectivity, program age, teaching field, and student–faculty ratios significantly affect GPAs (RQ3), with the student–faculty ratio emerging as a key driver.

The study provides large-scale longitudinal evidence of pronounced grade inflation in Turkish undergraduate teacher education from 2003 to 2022. Mean GPAs rose by approximately 0.51 points (2.83 to 3.34), and the share of high-honors graduates increased from 4.6% to 36.5%. Determinants associated with higher GPAs include greater proportions of female students, lower admission selectivity, newer program age, higher student–faculty ratios, and social science-oriented fields; public/private status had no significant effect. Peaks in 2012 and 2021–2022 correspond to system expansion and pandemic-related instructional shifts, respectively. Policy implications include monitoring and managing student–faculty ratios, reinforcing grading standards (especially in newer programs), and reviewing assessment practices to protect the informational value of grades. Future research should incorporate additional institutional and instructional variables, examine interactions among determinants, and investigate causal mechanisms (e.g., policy changes, evaluation practices) to design effective anti-inflation interventions.

The analysis relied on secondary, anonymized graduate databases and focused on a subset of teacher education programs, excluding several fields (e.g., aptitude-based admissions, low-enrollment or rare language programs, and redesigned special education) which may limit generalizability. Historical conversion of GPAs from a 100-point to a 4.0 scale, while standardized via a letter-grade mapping, may introduce measurement noise. Due to dataset constraints, only selected factors were modeled; additional variables and interaction effects could not be fully examined.