Beyond the Detour: Modeling Traffic System Shocks After the Francis Scott Key Bridge Failure

Response to Reviewer 1 Comments

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

2. Questions for General Evaluation

Reviewer’s Evaluation

Response and Revisions

Is the content succinctly described and contextualized with respect to previous and present theoretical background and empirical research (if applicable) on the topic?

Can be improved

Are the research design, questions, hypotheses and methods clearly stated?

Must be improved

Are the arguments and discussion of findings coherent, balanced and compelling?

Can be improved

For empirical research, are the results clearly presented?

Must be improved

Are the results clearly presented?

Must be improved

Is the article adequately referenced?

Can be improved

Is the article adequately referenced? Can be improved

Are the conclusions thoroughly

supported by the results presented  Can be improved

in the article or referenced in

 secondary literature?

3. Point-by-point response to Comments and Suggestions for Authors

Comments 1: The Introduction would be strengthened by the addition of more references to reflect current state of the art. Reference [1] relates specifically to secondary incidents. Suggest it is replaced with a more suitable reference. Also, not all references [9] to [12] relate to traffic disruptions.

Response 1: Thank you for pointing this out. We agree that the references could be strengthened to better reflect the state-of-the-art concerning primary infrastructure failures.

·         We have replaced the previous reference [1] with two new, more relevant citations that discuss system-wide shockwaves from infrastructure failures and cascading disruptions in transport networks.

·         We have also thoroughly reviewed and updated the references that were previously numbered [9] through [12]. The new citations now specifically point to studies using Fixed Effects, Mixed-Effects, and Difference-in-Differences models in the context of transportation research, ensuring their relevance to our methodology.

These changes can be found in the Introduction, paragraphs 3 and 6 of the revised manuscript.

Comments 2: Reference to real-time data (line 40), (line 63) and real-time traffic management (section 2.1) is potentially mis-leading as the reported study is retrospective (post hoc).

Response 2: Agree. While the data source (ClearGuide) operates in real-time, our analysis is indeed retrospective. To eliminate any ambiguity, we have revised the manuscript to clarify this distinction.

·         We have changed the phrasing to emphasize that the study uses "high-frequency data" sourced from a real-time platform, rather than claiming the analysis itself is real-time.

·         We have also clarified that the study’s goal is to use this detailed data to inform future real-time traffic management strategies.

·         These revisions are located in the Introduction (paragraph 6) and throughout Section 2.1.

Comments 3: Delete description of Clear Guide as a "powerful tool" (line 43). Revise text to relate to tools like Clear Guide. Likewise, avoid describing Clear Guide a "leading" - (line 162)

Response 3: Agree. The use of subjective and promotional language is not appropriate. We have revised the manuscript to adopt a more neutral, academic tone.

•          We have removed the phrases "powerful tool" and "leading" when describing the ClearGuide platform.

•          The text has been rephrased to focus on the objective capabilities and widespread use of such transportation analytics platforms by public agencies.

•          These changes can be found in Section 3.1, paragraphs 2 and 3.

Comments 4: References [7] and [8] - explain in what way this study adds to findings of these references. Newspaper reports in the days following the disruption suggested that alternative routes could carry diverted traffic. Were these reports confirmed or disputed by more in-depth studies. Ultimately, it is essential that authors explain clearly how their analysis adds to knowledge.

Response 4: Agree. The original references [7] and [8] in the first draft were preliminary assessments. Our study provides a more rigorous, in-depth academic analysis. We have removed these preliminary references in the revised draft and instead focused the Introduction on the methodological gaps in the academic literature. Our paper builds on preliminary observations by applying a multi-faceted econometric framework (FE, ME, DiD, Stratified models) to quantify impacts, assess adaptation over distinct seasonal periods, and isolate causal effects—a level of analysis that goes far beyond initial reports.

The core contribution—a comparative analysis of these models on a single event using high-frequency data—is now more clearly stated in the final two paragraphs of the Introduction.

Comments 5: The paper would benefit from a clear and concise description of the geography and road network of the affected area.

Response 5: beyond initial reports.

A concise description of the local geography has been added to the Methods section (Section 3.2, "Data Structure and Processing"). This paragraph now explains the role of the Key Bridge in the I-695 Baltimore Beltway and identifies the two primary alternative routes (the I-95 Fort McHenry Tunnel and the I-895 Harbor Tunnel). Further, we have updated Figure 1 to include a "star" icon indicating the location of the collapsed bridge, providing a clear visual anchor for the reader.

Comments 6: Objectives should not make reference to methods as a general rule. Delete reference to specific statistical methods from Reference 3.

Response 6: Thank you for pointing this out. We have revised the third research objective to focus on the intended outcome rather than the methodology.

·         The third objective now reads: "Quantify the impacts attributable to the bridge collapse by developing and implementing a multi-pronged comparative framework."

·         This change, located in the Introduction (line 41), preserves the mention of the framework's development while removing the explicit list of statistical models, in line with your suggestion.

Comments 7: Building on comment (5) explain how corridors were defined. It is unclear whether a corridor is the same as a route. Provide details of the length and characteristics of corridors.

Response 7: Thank you for pointing this out this lack of clarity. We have revised the third research objective to focus on the intended outcome rather than the methodology.

·         In Section 3.2 ("Data Structure and Processing"), we have clarified that the 30 major "corridors" are defined roadway segments from the Iteris ClearGuide platform, selected for their network importance around the collapse site.

·         While we use the terms "corridor" and "route" interchangeably, we have added a sentence in the Limitations section (5.2) to acknowledge that a more detailed analysis could incorporate specific corridor characteristics (e.g., length, number of lanes, baseline volume), which could be a direction for future work.

Comments 8: Unclear how References [20] and [30] support the selection of analysis of morning and evening peak periods. Is this even necessary as it is pretty standard practice in traffic studies?

.

Response 8: You are correct; citing this as standard practice is sufficient and the previous citations were not the strongest support. We have revised this.

·         We have removed the previous references and simply stated that focusing on weekday AM (7:00–9:00) and PM (16:00–18:00) peak periods is "standard practice" in transportation analysis. This can be found in Section 3.2, paragraph 3.

Comments 9: The authors claim that they are presenting an "integrated", multi-faceted analytical framework (Section 3.4) which seems to over-embellish what looks like a hierarchical approach which progressively targets the most affected routes. Figure 2 is not a roadmap.

Response 9: Thank you for this critical feedback. We acknowledge that our language may have been overly enthusiastic. Our intent was to show that the models are complementary, not just hierarchical.

·         We have toned down the language in Section 3.4 ("A Framework for Synthesized Analysis"). We now describe the approach as a "comprehensive framework" where models are used for "triangulation," with each offering a unique perspective (average change, spatial variation, causal impact, and extreme conditions).

·         We have also revised the description of Figure 2, now calling it a "graphical representation of our analysis strategy" rather than a "roadmap," which is more accurate.

Comments 10: Explain all notation in Equations 1 to 4. Explain also how time invariant characteristics unique to corridor controlled? Indeed, how were these measured? Table 2 (Results) includes an interaction term within the FE model specification which is not included in Equation 1.

Response 10: Thank you for catching these important omissions and inconsistencies. We have made several corrections.

·         We have updated Equation (1) for the Fixed Effects model to include the interaction term ( ) that was already present in the results in Table 2. This corrects the inconsistency.

·         We have added detailed explanations for all terms and subscripts for all three equations directly below each one, defining TTI, Post, Peak, Proximity, Treatment, the interaction term, the fixed effects, time effects, and random effects, and the error term (in Section 3.3.1., 3.3.2. and 3.3.3.).

·         We have clarified that in a Fixed Effects model, the time-invariant characteristics (e.g., corridor geometry, number of lanes) are controlled for by the model's design and do not need to be explicitly measured. This is a fundamental strength of the FE approach, and this explanation has been added to Section 3.3.1.

Comments 11: Remove the word "staggering" (line 271).

Response 11: Agreed. This word has been removed and replaced with more objective language to describe the increase in TTI for hotspot routes. The revised sentence can be found in Section 4.3.3., paragraph 1.

Comments 12: Wednesday PM in the Fall and Winter have a higher percentage change in TTI compared to Fridays which contradicts discussion of Results.

Response 12: Thank you for spotting this contradiction between our text and the figures. You are correct. We have revised the discussion in Section 4.1 to accurately reflect what the data in Figure 3 shows.

·         The text now correctly states that while the initial shock was acute on Friday, the pattern evolved, and in the Fall and Winter periods, "the most significant congestion shifted to mid-week, with Wednesday PM peaks consistently showing a greater TTI increase than Fridays"

Comments 13: Echoing point 5 above, in figure 5 show where the bridge collapse takes place.

Response 13: An excellent suggestion for improving clarity. As mentioned in our response to Comment 5, we have updated Figure 5 in the revised manuscript. A distinct

·         "star" icon has been added to all maps in the figure to clearly mark the location of the Francis Scott Key Bridge collapse.

Comments 14: Figure 5 - It's difficult to detect much change in mapped TTI (a), (c) and (e) (also (b), (d) and (f).

Response 14: We appreciate this feedback on the visual clarity of Figure 5. While the maps are intended to show the spatial patterns of congestion hotspots (e.g., the specific corridors that turn red), we acknowledge that subtle changes between seasons are difficult to discern. We have added a sentence in the discussion of Section 4.2 to guide the reader, emphasizing that the maps illustrate the transition from "widespread shock to persistent bottlenecks" and that the quantitative analysis in the subsequent tables provides the precise magnitudes of these changes.

Comments 15: The phrase "marginally significant" for a p-value of 7.9% is not appropriate (line 336) to use in an academic paper. This result should be considered as not statistically significant.

Response 15: We agree that the term "marginally significant" is ambiguous and should be avoided. We have revised the language to be more precise and objective.

·         In Section 4.3.2, we now state that the effect "approaches conventional levels of statistical significance (p=0.079)". This phrasing avoids making a definitive claim of significance while still noting that the p-value is low, allowing readers to interpret the finding in context.

Comments 16: (line 370) clarify that Thway is short hand for Thruway

Response 16: Thank you for catching this. It is an oversight on our part. We have clarified this in the text. The first time the abbreviation appears in the main body text (Section 4.2, paragraph 1), we have written out the full name "Harbor Tunnel Thruway". In the conclusion, we have spelled out "Thruway" as well to ensure clarity.

Comments 17: Explain what traffic management or demand reduction measures were implemented during any of the time periods studied...These measures could have a significant impact on results

Response 17: This is a very important point and a key limitation of our study. Acknowledging this is crucial for proper interpretation of the results.

·         We have added a new paragraph to the Limitations section (5.2) explicitly stating that our analysis does not model the specific traffic management interventions (e.g., signal retiming, public information campaigns) implemented by authorities.

·         We acknowledge that these interventions could be a confounding factor and that disentangling their effects from organic network adaptation is a valuable area for future research.

Comments 18: Inherent biases and their implications (line 404) should be considered more fully.

Response 18: We agree that this point deserves more consideration. We have expanded the discussion of data bias in the Limitations section (5.2).

·         The text now clarifies that GPS and connected vehicle probe data may carry biases (e.g., underrepresentation of certain vehicle types or demographics) and that future work could strengthen the analysis by

·         fusing this data with other sources, such as public transit ridership data or freight-specific telematics, to create a more holistic and less biased view.

Comments 19: Explain "parallel trends assumption" (line 406)

Response 19: Thank you for noting the need for a clearer explanation. A robust DiD model relies on this assumption, and we have expanded the text to define it and explain our validation process.

·         In Section 3.3.3 ("Difference-in-Differences Model"), we have added a clear definition: "A critical assumption for the DiD model is that the treatment and control groups would have followed parallel trends in the absence of the treatment".

·         We then explicitly state how we validated this assumption by examining traffic patterns in the four weeks preceding the collapse and have added a new

Figure 2 that graphically demonstrates that the trends were indeed parallel, providing strong support for the model's validity.

Comments 20: Describing the research as providing a "comprehensive blueprint" seems to overstate the paper's contribution. Stating that the approach followed moves beyond a single analytical lens is open to challenge. Is this really the case. If so it should be justified with reference to past studies.

Response 20: This is a fair critique. We have revised the language in the Conclusion to be more measured and better supported.

·         We have removed the phrase "comprehensive blueprint" and replaced it with "a structured framework for analyzing major infrastructure failures".

·         To better justify our claim of moving "beyond a single analytical lens," we have strengthened the narrative throughout the paper—particularly in the

·         Introduction (Section 1), Literature Review (Section 2.4), and Framework description (Section 3.4)—by emphasizing that while past studies have used these econometric models individually, their systematic comparative application to a single catastrophic event is a key contribution of our work. This comparative aspect reveals nuanced insights that any single method alone might miss

Response to Reviewer 2 Comments

1. Summary

Thank you very much for your time and for providing these insightful comments. Your feedback has been invaluable in helping us clarify the paper's core contributions and strengthen its methodological rigor. We have carefully considered each point and have made revisions accordingly. Please find our detailed point-by-point responses below.

2. Questions for General Evaluation

Reviewer’s Evaluation

Response and Revisions

Is the content succinctly described and contextualized with respect to previous and present theoretical background and empirical research (if applicable) on the topic?

Can be improved

Are the research design, questions, hypotheses and methods clearly stated?

Can be improved

Are the arguments and discussion of findings coherent, balanced and compelling?

Can be improved

For empirical research, are the results clearly presented?

Can be improved

Are the results clearly presented?

Can be improved

Is the article adequately referenced?

Can be improved

Is the article adequately referenced? Can be improved

Are the conclusions thoroughly

supported by the results presented   Can be improved

in the article or referenced in

 secondary literature?

3. Point-by-point response to Comments and Suggestions for Authors

Comments 1: I suggest to include the key motivation and the research gap of this paper. And also highlight the main contribution of the paper.

Response 1: Thank you for this suggestion. We agree that a clearer articulation of the motivation, gap, and contribution is essential. To address this, we have substantially revised the Introduction.

·         The revised Introduction now establishes the key motivation by highlighting how previous studies have often relied on a single modeling approach or retrospective data.

·         Based on this, we define a clear research gap: a lack of studies that systematically apply and compare multiple advanced econometric models to a single, catastrophic event using high-frequency data.

·         Finally, we highlight the paper's main contributions in the last three paragraphs of the Introduction: 1) developing and applying a comparative analytical framework that contrasts the performance of multiple models; 2) providing the first comprehensive, multi-dimensional quantification of the traffic impacts from the Francis Scott Key Bridge collapse; and 3) deriving practical, targeted policy insights that go beyond system-wide averages to address critical "hotspots".

Comments 2: The treatment/control split is based on Euclidean distance, I think it is too simple and might be confounded by socioeconomic factors. I think the additional robustness check is required.

Response 2: Thank you for this critical point regarding our treatment/control group definition. We acknowledge that Euclidean distance is a simplification that does not capture the full complexity of network topology or socioeconomic factors.

·         To clarify our rationale, we have added a new paragraph to the Methods section (Section 3.2). This paragraph justifies the use of Euclidean distance as a strong and intuitive first-order proxy for identifying the group of corridors most directly and immediately impacted by the abrupt, physical infrastructure failure. We argue that initial rerouting decisions are overwhelmingly dictated by proximity to the severed artery.

·         While a full robustness check is beyond the scope of the current paper, we have acknowledged this point in the Limitations section (5.2). We state that future work could employ more sophisticated network-based distance metrics or control for socioeconomic variables to build upon our findings.

Comments 3: I suggest to include the recent work of the network reliability analysis due to bridge failure. For instance “Graph neural network surrogate for seismic reliability analysis of highway bridge systems” and “End-to-end heterogeneous graph neural networks for traffic assignment”

Response 3: Thank you for this timely and valuable suggestion to include state-of-the-art literature. We have significantly enhanced the Literature Review section to incorporate this.

·         We have added a new subsection, 2.3 "Emerging Approaches and Remaining Gaps," which is dedicated to discussing recent advancements in network analysis.

·         This new section specifically discusses the use of Graph Neural Networks (GNNs) as a state-of-the-art approach, referencing how they are used for system reliability and traffic assignment, in line with the examples you provided.

·         We also use this section to contrast the predictive power of GNNs with the high interpretability of the econometric models used in our study, thereby better positioning our contribution.

Comments 4: The “immediate” period compares one month before and after the collapse. This is vulnerable to short-term confounders

Response 4: Thank you for raising this important point about potential confounders.

·         Our rationale for the one-month window is that it is a standard approach to capture the event's immediate shock while minimizing the influence of longer-term confounders like seasonal variation.

·         More importantly, we clarify in the revised manuscript that our Difference-in-Differences (DiD) model design inherently controls for unobserved, short-term confounders that would have affected both the treatment and control groups over that period (e.g., regional gas prices, weather events).

·         We also acknowledge in the Limitations section (5.2) that our analysis does not explicitly model corridor-specific traffic management interventions implemented by authorities, which could be a confounding factor and represents a valuable area for future research.       

Comments 5: The findings are highly case-specific, with little discussion of generalizability.

Response 5: This is an excellent point. We have revised the manuscript to better discuss the generalizability of our findings.

·         We acknowledge in the Conclusion that while the specific quantitative TTI values are case-specific, the methodological framework and the observed patterns are highly generalizable to other large-scale infrastructure disruptions.

·         We elaborate that these patterns—such as 1) the heightened vulnerability of PM peak hours, 2) the process of network adaptation over time, and 3) the creation of persistent, localized "hotspots" despite overall system recovery—are foundational dynamics likely to be observed in other metropolitan areas facing similar crises.

·         Furthermore, our Policy Implications (Section 5.1) are framed as general principles (e.g., dynamic peak management, hotspot mitigation, investment in redundancy) that are broadly applicable for building more resilient and sustainable transportation systems elsewhere

Response to Reviewer 3 Comments

1. Summary

Thank you for your thorough and encouraging review. We are grateful for your positive feedback on the multifaceted framework and the overall organization of the paper. Your specific, constructive suggestions have been instrumental in enhancing the manuscript's readability, methodological rigor, and clarity. We have carefully addressed each point and have incorporated the corresponding revisions into the manuscript.

2. Questions for General Evaluation

Reviewer’s Evaluation

Response and Revisions

Is the content succinctly described and contextualized with respect to previous and present theoretical background and empirical research (if applicable) on the topic?

Yes

Are the research design, questions, hypotheses and methods clearly stated?

Can be improved

Are the arguments and discussion of findings coherent, balanced and compelling?

Can be improved

For empirical research, are the results clearly presented?

Can be improved

Are the results clearly presented?

Can be improved

Is the article adequately referenced?

Can be improved

Is the article adequately referenced? Can be improved

Are the conclusions thoroughly

supported by the results presented   Yes

in the article or referenced in

 secondary literature?

3. Point-by-point response to Comments and Suggestions for Authors

Comments 1: Introduction and Abstract

Although there are some grammatical issues that need to be fixed, the abstract successfully explains the main findings... For instance, it is unclear what is meant by the sentence "With integrating proposed framework could be expected to valuable insights…" ... Think about changing it to something like this: "It is anticipated that incorporating the suggested framework will yield insightful information." ... Additionally, look for any additional uncomfortable wording. ... Make sure your language is consistent (for example, always use "Mixed Effects (ME/LMM)").

Response 1: Thank you for these specific suggestions on improving the language. We agree completely.

·         We have revised the abstract to correct the grammatical issues. The sentence you highlighted has been changed to your suggested phrasing:

·         "It is anticipated that incorporating the suggested framework will yield insightful information...". We have also proofread the entire abstract and introduction to correct other instances of awkward phrasing.

·         The phrase "abruptly determine anomalies" has been changed to "quickly detect anomalies" for better flow.

·         We have ensured consistent terminology throughout the paper. For instance, we now consistently use "Mixed-Effects (ME) model" when referring to the method, while also clarifying its connection to LMMs in the literature review.

Comments 2: Methodological Detail

While the methodological approach makes a significant contribution, there are a few details that need more explanation. Parallel pre-collapse patterns are a crucial premise for the Difference-in-Differences (DiD) model. Although the book claims that this assumption was "validated by inspecting pre-collapse data," trust would be increased if supporting documentation was included, such as a plot of trends or a formal test. ... Additionally, describe the standard error estimation process (e.g., grouped by route?). More information on how to deal with outliers... would be beneficial... Provide real sample sizes..

Response 2: We agree that providing more methodological detail is crucial for transparency and reproducibility.

·         Parallel Trends Validation: As you rightly pointed out, showing is better than telling. We have added a new Figure 2 in the revised manuscript that explicitly plots the pre-collapse TTI trends for both the "Proximate" (treatment) and "Distant" (control) groups. The plot visually confirms that the trends were parallel, providing strong evidence to support this critical assumption for the DiD model. This is now referenced in Section 3.3.3 .

·         Standard Errors: We have clarified in the text that robust standard errors clustered at the corridor level were used to account for potential heteroskedasticity and auto-correlation within each route's repeated measurements. This detail is crucial for the validity of our significance tests.

·         Outlier Handling: In Section 3.2, we have clarified that smoothing outliers beyond ±2.5 standard deviations from a rolling median is a standard approach to reduce the influence of data errors (e.g., faulty sensor readings) without removing entire observation periods.

• Sample Size: We have added the total number of observations used in the analyses to provide readers with a clear sense of the data's scale.

Comments 3: Interpretation of the Results

When analyzing relevance in the DiD data, caution is required. The Immediate PM estimate (+0.733), for example, is only slightly significant (p≈0.079). This may be overstated when the text refers to it as having a "unambiguously causal impact." ... Likewise, explain the persistence of the Winter AM DiD impact (+0.074, p<0.001), which is a negligible effect (even if statistically significant). The practical significance could be better understood by readers if confidence intervals or effect sizes, rather than merely p-values, were included.

Response 3:     This is a very important point regarding the nuanced interpretation of statistical results. We have revised the text to be more cautious and to better distinguish between statistical and practical significance.

·         "Marginally Significant" Result: We have removed any language suggesting an "unambiguous" causal impact for the Immediate PM DiD estimate. The revised text in

·         Section 4.3.2 now states that the effect is notable in its magnitude and "approaches conventional levels of statistical significance (p=0.079)", allowing the reader to interpret the finding with appropriate caution.

·         Practical vs. Statistical Significance: For the Winter AM DiD result, we have added a sentence explaining that while the effect is highly statistically significant (p < 0.001), its practical magnitude is modest. This suggests that while a lasting causal impact is detectable, the system had largely adapted by the winter period.

·         Confidence Intervals: Our regression tables (e.g., Table 2) already include confidence intervals. We have revised the main text to refer to these intervals more directly when discussing the magnitude and uncertainty of the model estimates, rather than focusing solely on p-values.

·          

Comments 4: Structure and Clarity

Although the work is typically well-structured, there are some extremely dense portions. To make lengthy paragraphs easier to read, think about segmenting them. ... The methodology and outcomes may benefit from a similar usage of subsections or numbered phases to increase clarity. ... It could be beneficial to restate in brief the meaning of each model's coefficients.

Response 4: Thank you for this excellent feedback on improving the structure.

·         We have broken down several longer paragraphs in the Introduction, Literature Review, and Methods sections into shorter, more focused paragraphs to improve readability.

·         In Section 3.3 ("Multi-Faceted Analytical Framework"), we have already structured the presentation of the models using numbered subsections (3.3.1, 3.3.2, etc.), one for each model, to clearly delineate them.

·         We have now added detailed explanations of the key coefficients directly below each equation. For example, under Equation 3 (DiD Model), we explicitly define the interaction term (β3) as the DiD estimator that captures the causal effect.

Comments 5: Tables and Figures

Although the figures are generally instructive, the following recommendations could make them better

Response 5: We appreciate these specific recommendations for improving our tables and figures.

·         Figure 3 (Framework): The caption has been expanded to briefly outline the workflow, explaining that the framework integrates four complementary models to provide a multi-layered understanding of the impact. Labels have been checked for readability.

·         Figures 4 & 5 (TTI Changes): We have reviewed all axes and legends to ensure they are clear, readable, and include units (% change). The use of different lines/colors already distinguishes the time periods.

·         Figure 6 (Spatial Maps): We have ensured all maps are high-resolution. The caption has been expanded to briefly explain what each panel (a-f) represents . We have also added a "star" icon to mark the location of the bridge collapse for better context, per the suggestion from another reviewer.

·         Tables: We have added a footnote to Table 4 to explain the significance notation (* p < 0.10, ** p < 0.05, *** p < 0.01). We have also reviewed and clarified column headings for better interpretation (e.g., Table 5 now reads "Post-Collapse TTI Change (Coefficient)").

Comments 6: Discussion and Conclusions

One recommendation is to use caution when using causal language in the Conclusion... Perhaps use a somewhat more measured language... Lastly, to make the policy proposals more actionable and succinct, think about condensing some of the text to important bullet points, as you did in Section 5.1.

Response 6: We agree that the language should be precise and measured, especially concerning causality.

·         We have reviewed the Abstract and Conclusion and softened the causal language. Instead of definitive statements, we now frame the findings as providing "strong evidence" or highlighting associations, which is more appropriate for a quasi-experimental study.

·         The policy recommendations in Section 5.1 are already presented in a succinct, bulleted format to enhance their clarity and actionability.

Comments 7: References

The majority of the references are current and relevant. But there are a few problems to resolve... Eliminate any blank or missing information (for example, Reference 21 shows "Empty – Fixed effects- Transportation")... Make sure all formatting follows the journal's style...

Response 7: Thank you for catching these errors in the reference list.

·         We have removed the empty/placeholder reference and have conducted a thorough review of the entire reference list to ensure that every entry is complete and correctly formatted according to the journal's style guidelines.

·         We have also cross-checked to ensure that all sources cited in the text appear in the reference list and vice-versa.