Education Expenditure and Sustainable Human Capital Formation: Evidence from OECD Countries
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
Hello Dear Author(s),
Please proceed with the following issues:
1. The variable range for “fiscbal” at 252.5 is probably implausible, likely a scaling, please confirm it again.
2. There is a need for more discussion of model fit R² of 0.21–0.41 indicates limited explanatory power.
Please at least mention it in the context.
3. The data period (1997–2020) is adequate for long-term analysis, however, the exclusion of post-pandemic years may introduce pre-pandemic bias.
Incorporating OECD data for 2021–2023 could enhance the robustness of the findings.
If it is not plausible to include the data at least please make a mention of limitations in the discussion section.
4. Some coefficients are large, like 14.882 for low-income pergdp, suggesting scaling issues. Please analyze it.
5. If it is plausible to include diagnostic tests like Breusch–Pagan, and Wooldridge tests, the context would be enhanced (not mandatory).
6. Discuss potential omitted-variable bias, for instance, government debt, education quality index.
No need to change the data, but an analysis of the reason why omitted could boost the depth analysis of the paper.
7. No mention of endogeneity and robustness tests like heteroskedasticity. At least include it to the limitations.
Without these, the causal interpretation is not appropriate. At least mention it in the limitations of the paper.
8. “Fiscal balance” is defined but not used in the models; please mention it in the context or remove it from table 1.
9. If it is plausible, please include in the literature about fiscal sustainability frameworks like IMF, World Bank
10. Also, in the literature review could be made a mention about education expenditure elasticity, to give more depth in the analysis.
11. There is no appendix of data availability or a link showing the raw or processed dataset used for the regressions (major issue).
12. There is a need for statistical hypotheses, for instance:
H1: There is a nonlinear (inverted U-shaped) relationship between income level and education expenditure as a share of GDP.
H2: Higher income inequality reduces education expenditure in high-income countries but increases it in low-income countries etc.
13. Also, please include a clear hypothesis about the main theory of the paper in the introduction, and if it has been confirmed in the discussion.
Good effort, please proceed.
Best Regards,
Author Response
Please check the attached file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors
The article is devoted to an empirical analysis of the factors determining the volume of public spending on education in OECD countries. The author examines the impact of income level, income inequality, fertility rate and population density on the share of education expenditure in GDP, using panel regression with fixed effects and the method of moment quantile regressions (MMQR) to identify heterogeneity of impacts in different groups of countries. The main conclusion is that there is a nonlinear relationship (inverted U-shaped shape) between income level and education expenditure, which indicates an increase in education expenditure at the initial stages of economic development and its decrease after reaching a certain income threshold.
However, some questions arise regarding the accuracy and relevance of the results obtained:
- The literature review is presented in a rather descriptive and superficial manner. The author summarizes well-known studies on the impact of income and inequality on educational spending, but does not demonstrate a deep critical assessment of previous results and does not sufficiently clearly justify the scientific niche of his own research. The text lacks an analysis of modern publications (after 2020), and the list of references is mostly classical in nature. There is also no systematization of hypotheses that would arise from the literature review, which reduces the analytical depth of the work. Therefore, a goal of the paper remains unclear.
- Although the use of fixed effects and MMQR is a correct approach for panel data, the article lacks an analysis of the problem of multicollinearity between variables, especially between income level and its square, as well as between income, inequality and fertility. This may distort the estimates of coefficients and explain their insignificance in some models. No tests for multicollinearity (e.g., VIF) are presented, which makes it impossible to assess the stability of the results.
- Many of the coefficients in estimation tables are statistically insignificant, which casts doubt on the strength of the conclusions drawn. However, the author does not discuss possible reasons for this non-significance –for example, sampling heterogeneity, autocorrelation, or collinearity. Nor are any diagnostic tests for heteroscedasticity or autocorrelation, which are essential for the correct interpretation of panel estimates, provided.
- The interpretation of results is often descriptive in nature and does not provide sufficient justification for the mechanisms of action of the observed relationships. For example, the author claims a U-shaped relationship in highly developed countries, but does not offer convincing theoretical explanations for why such a shape should arise.
- The conclusions and policy recommendations are rather general and are not supported by quantitative estimates of effects or testing the robustness of the results to alternative model specifications.
In general, the paper has an interesting topic and uses modern statistical tools, but it lacks analytical depth, theoretical reasoning, and methodological transparency. It is advisable to refine the article by expanding the theoretical basis, checking the stability of the results, and eliminating statistical shortcomings of the analysis.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
The empirical approach and the topic are relevant. Still, the paper needs to be strengthened with tighter theory-to-test alignment, corrected data handling, stronger identification/diagnostics, and clearer/result-consistent policy guidance.
Needs strengthening :
The research gap and objective are not clear
update references and enrich them
Theory for nonlinearity and interactions. The inverted-U (and the U in some subsamples) is reported but not theoretically pinned down.
Map each hypothesized sign to a concrete mechanism (public–private substitution, median-voter/redistribution, dependency structure, urbanization/economies of scale).
Consider reporting results by education level (primary/secondary vs tertiary) and funding source (public vs private) using OECD EAG sub-indicators, to align the econometrics with the sustainability claim.Variable definitions & plausibility checks.
In Table 2, fiscal balance shows a maximum of 252.5, which is not credible for “% of GDP.” Re-audit units, deflators, transformations, and outliers; document winsorization rules.
Similarly, ranges (0.058–2.509) look like logs of a scaled variable—spell out the scaling (e.g., log PPP in $10,000 units). These inconsistencies can materially affect turning-point calculations.
Some signs differ between FE (Table 4) and MMQR (Table 6)
Add a short subsection explicitly reconciling these patterns
Interaction model table (Table 5). The table appears garbled: duplicated columns, misplaced coefficients, and formatting glitches. Rebuild the table cleanly with clear column headers (All / High / Low), show marginal effects of income at representative inequality/fertility values, and provide interaction plots with 95% CIs.
Ensure Figures 1–3 show axis units, sources, and confidence bands (for coefficient profiles). Consider small multiples by income group
Add notes on clustering, FE specification (country & year dummies), and whether time trends are included. Harmonize significant stars and decimals.
Align in-text citations (e.g., Easterly & Rebelo, not “Revelo”) and verify years/DOIs; ensure consistency with MDPI style
Comments on the Quality of English Language
The English could be improved to more clearly express the research.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for Authors
The work has been significantly improved. However, there are still some wishes for the authors:
The interpretation of the results in some places goes beyond the limits of statistically significant coefficients. It is worth adjusting the text so that conclusions are drawn only for those changes whose effects are truly confirmed by the models. There is no explanation for the anomalously large and unstable standard errors in some models. It would be advisable to add a brief discussion of the possible reasons for such values ​​and the impact on the reliability of the conclusions.
The results section lacks structure. Some fragments repeat the same arguments in different subsections; it is important to reduce duplication and clearly distinguish key conclusions for different groups of countries.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
Thank you for your effort and time
Good luck
Comments on the Quality of English Language
The English could be improved to more clearly express the research.
Author Response
We sincerely appreciate your helpful comments on the clarity and quality of the English language in the manuscript. To address these issues, we have professionally edited the entire manuscript with MDPI Author Services. We sincerely hope these revisions will make the language of the manuscript clear and readable.
Author Response File: Author Response.pdf
Round 3
Reviewer 2 Report
Comments and Suggestions for Authors
Several parts of the manuscript interpret statistically insignificant coefficients as if they provided substantive evidence. This leads to causal or quasi-causal statements that are not supported by the empirical results. In a number of cases, the authors generalize findings from models where core variables (income, income², fertility, inequality) are insignificant. Such generalizations create conclusions that extend beyond what the data can confirm. Below I put the list of such misinterpretations and my proposals how to improve the text:
| Original | My proposal | Reason |
|---|---|---|
| “The lack of a significant association for high-income countries suggests that these pressures may be mitigated when demographic transitions are more advanced.” | “For high-income countries, the coefficient is not statistically significant; therefore, no robust conclusion can be drawn regarding the influence of fertility.” | Authors interpret an insignificant coefficient as evidence of an economic mechanism. |
| “…the relationship between income and education expenditure becomes a clearly inverted U shape once the interaction terms are included.” | “…the interaction specifications show some indications of a nonlinear pattern, although the evidence is not fully consistent across models and several coefficients are statistically insignificant.” | Claims a “clearly inverted U” when income and income² are insignificant in FE and unstable in interaction models. |
| “…in richer countries, the influence of fertility appears more limited.” | “…for richer countries, the estimated fertility effects are statistically insignificant, and thus no systematic influence can be confirmed.” | “Limited influence” incorrectly implies a weak but present effect; statistically the effect is absent. |
| “…rising inequality reinforces the tendency for the education expenditure share to decline at higher income levels…” | “…some interaction terms involving inequality show marginal significance in certain models, but the evidence is mixed and does not allow for a definitive conclusion.” | The effect is not consistently significant; the sentence overstates the empirical support. |
| “…income exerts a stronger and more pronounced inverted-U-shaped effect in high-income countries…” | “…in high-income countries, a nonlinear income effect is statistically significant only in middle and upper quantiles (Q50–Q90), while no significant effect is found in lower quantiles.” | Authors generalize MMQR results despite insignificance of income and income² at Q10–Q25. |
| “Fertility conditions play a more important role in lower-income OECD economies; conversely, in richer countries, the influence appears more limited.” | “Fertility effects are significant only in lower-income OECD economies; for richer countries the coefficients are generally insignificant.” | Again, the authors interpret insignificance as “limited influence,” which is misleading. |
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
