Geng Integrity Review: Influence of English Proficiency on Employment Quality of Undergraduates — An Empirical Analysis Based on a National Graduate Employment Survey

Academic fraud report · Geng Detector

Summary

Verdict: CLEAR (no academic misconduct detected). This review examined a 2025 empirical study by Ruan Qian, Yan Siyu, and Yue Changjun, published in the Fudan Education Forum (DOI: 10.13397/j.cnki.fef.2025.06.012), which investigates how undergraduate English ability affects employment quality using a national graduate employment survey. Five diagnostic passes were applied: image inspection (not applicable, as the paper contains only text-based tables, no Western blots or microscopy images); random-number-generator-style data checks (descriptive statistics such as CET-4 mean=457, SD=63.92, min=230, max=673, and non-integer standard errors like 0.021, 0.100, and 0.094 show authentic dispersion typical of survey data); statistical logic (OLS on log wages, O-Logit, MNL, and KHB mediation are appropriate; sample sizes consistent at N=2591 with N=2450 in one column due to documented item-missingness; total=direct+indirect effects mathematically exact); and timeline/methodology (2023 survey data, policy document from 2025-04-08 cited 2025-04-20, paper received 2025-06-17 — fully consistent). One minor rhetorical weakness: H1 predicted full significance, but some coefficients (e.g., oral fluency on wages, 0.033, SE 0.025, p>0.1) are non-significant; the authors partially reconcile this in the discussion. This is a writing-style observation, not data fabrication. No action recommended.

Verdict

CLEAR — No evidence of academic misconduct was found across five diagnostic checks. The paper demonstrates authentic survey data characteristics, statistically appropriate methods, internally consistent numerical results, and a coherent publication timeline.

Key findings

Image integrity: Not applicable. The article contains only text-based descriptive and regression tables (Tables 1–6); no Western blots, microscopy, flow cytometry, or pixel-level figures are present.
Data authenticity: Descriptive statistics exhibit natural dispersion (CET-4 mean=457, median=451, SD=63.92, range 230–673). Standard errors are non-integer floats (e.g., 0.021, 0.100, 0.094), inconsistent with fabricated data.
Statistical logic: Method selection matches data types — OLS on log starting salary, O-Logit for ordinal job satisfaction, MNL for job-search channels, and KHB for mediation. Sample sizes are consistent (N=2591 across most tables; N=2450 in one column of Table 5 with documented missingness). KHB decomposition is internally consistent (e.g., M1 general oral proficiency: total effect 0.427 = direct 0.387 + indirect 0.040).
Timeline coherence: 2023 survey data → 2025-04-08 policy cited on 2025-04-20 → paper received 2025-06-17 — no anachronisms.
Sample construction rigor: Authors explicitly exclude further-study, undetermined, flexible-employment, and English-major graduates (who take TEM-4/8 instead of CET-4/6), reflecting attention to construct validity over maximizing N.
Honest reporting of null results: Table 3 column (3) shows the oral-fluency coefficient on starting salary is 0.033 (SE 0.025, p>0.1), non-significant — explicitly reported rather than suppressed.

Evidence highlights

CET-4 descriptive statistics: mean=457, median=451, SD=63.92, min=230, max=673 — distribution shape consistent with the real CET-4 population.
Standard errors in regression tables are non-trivial floats (0.021, 0.094, 0.100, 0.025), ruling out hand-computed or templated fabrication.
Table 5 column showing reduced N (2450) is footnoted as due to item-level missingness, not silent case deletion.
KHB mediation in Table 6 satisfies total = direct + indirect exactly across all reported decompositions.
Reference list includes a 2025-04-08 policy document accessed 2025-04-20, predating the 2025-06-17 submission date.

Notes

One minor rhetorical issue: Hypothesis H1 posits a "significant effect" of English proficiency on both starting salary and satisfaction, yet the empirical results are partial — oral fluency is non-significant on wages (coefficient 0.033, SE 0.025, p>0.1) and the CET-4 certificate is non-significant on satisfaction. The authors reframe this in the conclusion as "differential effects across dimensions of English ability," which is a standard academic reconciliation rather than evidence of p-hacking or selective reporting.
Limitations of this automated review: it cannot verify microdata against the underlying national graduate employment survey, nor confirm the chain of custody for the analyzed sample. Definitive determinations of academic integrity require institutional investigation.
Recommended follow-up: none. The paper may serve as a positive methodological exemplar for KHB-based mediation analysis in education economics.

*Disclaimer: This report was AI-assisted and is intended for academic discussion only.*