Migration, Corruption, and Economic Drivers: Institutional Insights from the Balkan Route

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article analyses the determinants of migrants’ route choice along the “Balkan corridor.” It applies logistic regression and machine learning approaches on IOM flow monitoring survey data and governance indicators to empirically argue that corruption acts as a dual push–pull factor shaping route choice and offers policy implications for origin, transit, and destination contexts. The article is relevant and adds value to the topic of migration; however, several significant concerns should be addressed before publication.

 

1. The theoretical part extensively discusses an adapted migration “gravity” model that incorporates corruption [Section 2.1, lines 109-120] and also references de Haas's aspirations-capabilities framework [Literature review, lines 83-86; see also Section 2.1, lines 114-117]. Although the article frames these theories, it fails to formalise the mechanisms and expectations involved in applying the model. It does not operationalise or estimate key elements related to the gravity model, such as i) bilateral flows between origin-destination pairs as functions of size (population/GDP), distance, and other bilateral characteristics; ii) network effects (including smuggling), time-varying (border) policies, and/or contingent shocks (at least in robustness checks), which are crucial drivers of transit decisions. Moreover, gravity-like models focus on spatial variation rather than temporal dynamics, so causal claims and policy implications concerning corruption should be interpreted with caution or require explicit temporal validation.
   
   
2. The IOM flow monitoring surveys are among the best available route-based sources but do not represent a probability sample; they report locations, timing, and potential selection, and often apply route or region weights to improve representativeness. It remains unclear whether route weights were applied and how site selection might bias the comparison between Bulgaria and Greece, especially considering that the IOM data tracks other entry points in the Western Balkan transits (such as Albania and North Macedonia). The article should either justify the binary approach or adopt multinomial models in line with the IOM's framing of route options. the Western Balkans flow.
   
   
3. The governance clustering is also likely to introduce more noise and potential bias. Although the article acknowledges the measurement risk with WGI and the (substantial) 44% “Unknown” origin‑cluster [Lines 195-205; 348-357; 423-431; 470-476], it should address these limitations by justifying validation metrics (why k=4 clusters), running sensitivity tests with alternative indicators (e.g., V‑Dem/TI‑CPI), or using a design that is robust to missing information.
   
   
4. The analyses show weak predictive performance and some inconsistencies. While the article acknowledges that all models exhibit poor predictive ability (AUC ≈ 0.51–0.61; currently labelled as “moderate,” although values between 0.5–0.7 are considered, as per standards accepted guidelines, as poor, not moderate). The implications merit more careful consideration. Such weak performance may indicate that the theoretical framework is fundamentally misspecified; key variables such as smuggling networks, border enforcement, or temporal shocks are omitted; the binary outcome oversimplifies multiple decision-making processes; and/or individual-level survey data may be insufficient for understanding route dynamics. Please also double-check where the models show "perfect Recall (=1.000)"; this may typically indicate that the model predicts the majority class for all cases, which signifies underfitting or a complete lack of discrimination, rather than overfitting [Table 4, and Lines 286-288].
   
   
5. Policy claims overstate evidence from the analysis. Each recommendation is reasonable but requires a stronger connection to observed associations, recognition of uncertainty, caution regarding external validity, or references to implementation evidence. This contextualisation will mitigate generalisations from a model with limited discrimination.
   
   
6. Some other minor issues also need to be addressed:
   
   
   1. The transit-stage policy paragraph about gender-sensitive support contains a stray “[?]” [Line 396], probably a missing reference.
   2. The Data Availability Statement section [482-484] states that the Displacement Tracking Matrix data are openly available, adding a direct link to the specific FMS wave(s), and any code repository for reproducibility is necessary. For example, what was the exact question that the respondents were answering in coding the binary outcome [Y = 1 if entered via Bulgaria, 0 if entered via Greece, Line 163] from the IOM survey questions?
   3. Lines [58-61] seem to be an incomplete or fragmented sentence: “By weaving micro-level survey insights with macro-level governance clustering and advanced quantitative tools, including modifications to the gravity model that foreground corruption’s disruptive influence.”
   4. The article states that Bulgaria accounts for 38.8% of entries and has an ascendant trend (Lines 56–62; 132–137; lines 151–159; 195–205; lines 302–306. Please verify and cite the exact IOM source for the figures and the “ascendance” trend for Bulgaria, as IOM/Frontex data indicates substantial regional declines in 2024–2025.

    

Author Response

Thank you for your detailed and constructive feedback. We appreciate the opportunity to address these points, which will strengthen the manuscript. In uploaded document, we respond to each comment in turn, drawing directly from the content
and analyses already presented in the paper.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors' article entitled Migration, Corruption, and Economic Drivers: Institutional
Insights from the Balkan Route is quite original since it deals with analyzing the factors influencing the migrants’ decisions to enter Europe via Bulgaria or Greece along the Balkan route. Using IOM survey data (2022–2025, N=5,536), an adapted gravity model, logistic regression, and machine learning techniques add robustness in terms of empirical analysis. Lso, governance clustering via WGI adds a quite important degree of novelty. 

However, it is appropriate to address a series of comments that would improve the article for publication in World.

1. The Unknown cluster comprises approx. 44% of origins (n=2,446), which undermines the explanatory power of Hypothesis 3. Please give more details on matching procedures and robustness checks.
2. All models show modest fit (AUC ≤0.612). Explain in detail this issue and justify subsetting to N=4,429 for ML. In would be highly recommended to add diagnostics (e.g., confusion matrices for all models / calibration plots) and use alternatives (e.g., balanced sampling / ensemble methods) to prove that your models are ok.

3. The interaction term (Age × Economic Reasons) is insignificant. Why did you include it? If no essential reasons, please drop it and rerun models. It is highly recommended to include multicollinearity tests (VIF values), heteroskedasticity tests (Breusch-Pagan) and endogeneity tests (IV for push factors). 

4. The literature review is comprehensive, but redundant. Subsections 2.1-2.4 should be condensed! Add recent Balkan-specific studies (post-2023) for better novelty. Verify the citations for consistency and completeness. 

5. Add standard errors for ML metrics for Table 4 and add odds ratios for Table 3! Fig. 1 (ROC) should include confidence intervals. 

6. Conclusion and contributions should be rephrased in order to be better lonked with empirical results!

Author Response

Thank you for your detailed and constructive feedback. We appreciate the opportunity to address these points, which will strengthen the manuscript. In uploaded document, we respond to each comment in turn, drawing directly from the content
and analyses already presented in the paper.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for your thoughtful and thorough revisions and response.

Reviewer 2 Report

Comments and Suggestions for Authors

I recommend the publication in its current version.