Forecasting Youth Unemployment Through Educational and Demographic Indicators: A Panel Time-Series Approach

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

The article titled "Forecasting Youth Unemployment through Educational and Demographic Indicators: A Panel Time-Series Approach" addresses a timely and relevant topic in the field of educational policy and labor market dynamics: the relationship between tertiary education enrollment, demographic pressure, and youth unemployment in emerging economies. The subject is well-grounded in the academic literature and contributes by employing a panel time-series approach on a sample of countries from Central Asia and the Caucasus.

Below, I present my observations to support a possible revision of the article.

1. The study uses panel data across eight countries but does not test for cross-sectional dependence. Ignoring this assumption may affect the validity of FMOLS and DOLS estimators, which rely on independence across cross-sections.
2. The introduction presents a well-structured overview of the relevant literature and highlights key issues related to youth unemployment, education, and demographic dynamics, but the research gap is not explicitly and clearly formulated.
3. Although the authors acknowledge the need for more advanced tests (e.g., Pesaran and so on) in Section 4.2, these are not implemented. Given the small N and potential structural heterogeneity, the omission weakens the robustness of the cointegration results.
4. The GDP coefficient is significant under FMOLS but becomes insignificant under DOLS. While this is noted, the authors do not investigate the possible reasons.
5. The relationship between education and unemployment may vary depending on country characteristics (income level, education quality, digital infrastructure and so on).
6. Although briefly mentioned in the introduction, the paper does not report any causality or direction of effect testing between education and youth unemployment.
7. The coefficient interpretations for ENROL, GROWTH, and POPULATION are repeated across several sections (Sections 3.2–3.4).
8. Parts of the methodological section (detailed equations for unit root tests) can be condensed to improve readability.
9. The paper does not identify or compare individual countries in the sample. This limits the practical insight of the findings, especially since demographic pressure and educational effectiveness likely vary across contexts.
10. The figures included in the manuscript appear to be screenshots rather than properly formatted graphs. The axes lack proper labeling, legends are minimal or missing, and resolution is suboptimal.
11. The reference list combines various citation styles. The authors should ensure full compliance with the MDPI citation style.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

General comments

The manuscript makes a significant contribution to the literature on youth unemployment by combining advanced econometric techniques with policy-relevant insights. The methodology employs appropriate econometric techniques (Panel Unit Root Tests, Pedroni Cointegration Test, FMOLS, DOLS) to analyze panel data. The authors conducted robustness checks by comparing FMOLS and DOLS results, which strengthens the validity of their findings. The interpretation of coefficients and statistical significance is generally sound.

 

Some specific suggestions for the authors

The paper uses a sample of 8 countries, but their names are not explicitly stated. List them in Section 2.

 

Line 65: The study period is stated as "2000 to 2022," but in Section 2.1 (Data Description), it's stated as "2009-2023". Please ensure consistency in the study period mentioned throughout the paper.

 

Line 66-67: The introduction states that the study implements "an FMOLS and ARDL approach," but the methodology and results sections primarily focus on FMOLS and DOLS, with ARDL not being a central part of the analysis. Please clarify if ARDL was initially considered and then dropped, or if there's a misunderstanding. If ARDL was not used, this mention should be removed or corrected.

 

The mention of "social transfers" and the reference to Ardington, Case, and Hosegood (2007) [16] in lines 85-88 seems to introduce a variable that is not explicitly included in the model specification (Section 2.3) or the results. If social transfers were not modeled, this discussion should be either integrated more clearly as a limitation or removed to avoid confusion.

 

Model Specification (Section 2.3): The general panel model includes Eit as the error term. It would be good to explicitly state the assumed properties of this error term (e.g., independently and identically distributed, homoskedasticity).

 

Table 2: The table title indicates "Panel unit root tests" but the content refers to "Table 1 illustrates that most of the variables are not stationary at level" (line 167). Please correct the reference to Table 2 if it's indeed Table 2 being discussed.

 

Clarify the sample size (e.g., eight countries, 112 balanced panel observations in DOLS) earlier in the methodology section to provide context for readers.

 

In Table 5, the economic interpretation for POPULATION states "An increase of 100,000 young people raises unemployment by ≈ +1.11×10⁻⁶." However, in the text (lines 277-278), it states "a 0.19 point increase in unemployment per 100,000 additional youth." Please ensure consistency in these interpretations or clarify the units. It seems the 1.11E-06 coefficient is per thousand persons, so 100,000 persons would be 100 units of POPULATION, leading to 100 * 1.11E-06 = 1.11E-04, not 0.19. This discrepancy is significant and needs to be corrected.

 

The inconclusive Kao test result (line 379) suggests potential limitations in the panel’s cross-sectional properties. Consider applying second-generation cointegration tests (e.g., Westerlund’s error correction test) as recommended in Table 9 to address cross-sectional dependence.

 

The results are consistent across models, and the authors appropriately address potential limitations, such as the instability of the GDP growth coefficient. However, the inconclusive Kao test result and potential cross-sectional dependence could be further explored to strengthen the analysis.

 

Ensure all referenced figures (e.g., Figures 1–3) are included and properly formatted in the final manuscript. Provide clear captions and ensure figures are self-explanatory.

 

Figure 1 is quite compressed and difficult to read. The x-axis labels ("Country - Year") are extremely compressed and unreadable. Please consider rotating the labels or displaying them in a more sparse manner (e.g., showing every 5th or 10th year for selected countries) to improve readability.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

The manuscript is generally well-written, but there are instances of inconsistent terminology (e.g., "YOUTH_UNEMPL " vs. "YOUTHUNEMPL"). A thorough proofreading and standardization of terms would enhance readability. Additionally, some sentences are overly complex, which could be simplified to improve accessibility for a broader audience.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript presents a current and relevant topic - youth unemployment - through a rigorous econometric analysis. The study is well structured, with a reasoned introduction, an adequate methodology and a coherent interpretation of the results. The conclusions drawn are supported by the date and provide relevant contributions to the relationship between education, demography and youth unemployment. Recommendations for improvement:

• The English language used requires professional review. I recommend language editing for clarity and fluency.
• Although the analysis is rigorous, further clarifications on the limitations of the econometric approach could be useful, especially regarding the robustness of the results in the face of possible incompletely resolved endogeneities. Overall, the paper represents a valuable contribution and is recommended for publication, subject to minor revisions.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for your revised submission and the effort to address the comments provided in the initial peer review. After a detailed comparison between the initial and revised versions of the manuscript, I offer the following summary evaluation regarding the revisions made:

1. The inclusion of the Pesaran CD test in the revised methodology section and its results in Table 7 improve the econometric soundness of the FMOLS and DOLS estimations.
2. The introduction provides a good contextual overview and emphasizes the issue’s relevance, but the research gap remains somewhat implicit. I encourage the authors to state the gap more explicitly and clearly delineate how this study fills it.
3. The revised discussion recognizes the sensitivity of the GDP coefficient but does not empirically explore the underlying causes.
4. The manuscript now includes some narrative elements comparing countries, but no estimation technique is used to formally account for heterogeneity.
5. The addition of the Dumitrescu–Hurlin panel causality test (Table 9) improves the manuscript significantly.
6. Some redundant discussions have been trimmed, but repetition remains in the coefficient interpretation and FMOLS/DOLS explanation. 
7. There is more country-specific narrative in the revised version, but no individual results or figures are presented.
8. The figures included in the manuscript require substantial improvement. Several of them appear to be low-resolution screenshots rather than properly exported graphs from statistical software. Axis labels and units are either missing or extremely difficult to read. Figures lack clear labeling of both axes, and it is not always evident what each panel represents. Additionally, the bar chart comparing long-run coefficients does not display confidence intervals or standard errors. The legend is also insufficiently detailed, especially regarding the model selection criteria (SIC, HQC), which are not explained in the caption or main text.
9. There is a noticeable disconnect between the countries under study and the comparative references used. The discussion frequently draws parallels with findings from regions such as Sub-Saharan Africa and the Asia-Pacific region. Although international comparisons can enrich interpretation, they should be used with caution when the institutional, demographic, and economic contexts differ substantially. The discussion clearly justify why findings from other continents are relevant for the Central Asian and Caucasus context.
10. Table 10 aims to summarize policy implications but lacks clarity and precision. The language includes grammatical errors ("A increase") and the policy arguments are often generic or not well supported by the statistical findings. For instance, the claim that GDP growth reduces unemployment is inconsistent with the reported instability and insignificance of the GDP coefficient in the DOLS model. Similarly, the job quality is important but underdeveloped.
11. The reference list does not fully comply with the MDPI citation style.
12. The recommendations in Section 4.2 are relevant and well-justified, but their absence from the current analysis limits the study’s robustness. At least one advanced method (CIPS or Westerlund) should have been implemented rather than deferred entirely to future research.
13. The conclusion effectively summarizes the main findings and policy implications, but it overstates the robustness of the results, particularly regarding GDP, which showed instability across models.
14. The authors are encouraged to provide access to the dataset used to ensure replicability.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for the revised version of the manuscript. I appreciate the authors' efforts in addressing my previous comments. All concerns have been satisfactorily resolved.

However, I would like to note that Figure 1 still appears to be of very poor quality and remains nearly unreadable. I recommend that the authors provide a higher-resolution version of this figure to ensure clarity and readability for the readers.

Once this issue is corrected, I believe the manuscript will be ready for publication.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

Response to Authors

Thank you for your extensive revision of the manuscript and for the efforts made to improve the article. Upon reviewing the revised version, I clearly appreciate the progress achieved:

1. The inclusion of the Pesaran CD test is a welcome improvement.
2. The research gap is more clearly identified in the updated version.
3. The Dumitrescu–Hurlin test is well integrated into the methodology and convincingly presented in the results section.

At the same time, I note several aspects that still require further attention:

1. Redundancies in the interpretation of coefficients and the FMOLS/DOLS explanations persist.
2. Country-specific results are still missing and are necessary to support the qualitative discussion in Section 4.
3. The figures require significant technical improvements. Please refer to the previous recommendations.
4. Table 10 on policy implications still includes vague statements and some grammatical errors (e.g., "A increase"). Moreover, the link between the proposed policy measures and the empirical findings is not always clearly supported.
5. The reference list does not fully comply with MDPI citation style.
6. Advanced robustness methods have not been implemented, and the stated intention to use them in future research is not sufficient to compensate for their current absence.
7. The conclusion tends to overstate the robustness of the GDP effect, which is statistically insignificant in the DOLS model. I recommend tempering the related claims.

I encourage the authors to address these remaining issues in order to strengthen the manuscript further.

Author Response

Please see the attachment.

Author Response File: Author Response.docx