Emergence of power laws in hierarchical dynamics on multi-level graphs

Armando Bazzani; Gregorio Berselli; Mirko Degli Esposti; Tommaso Rondini

arxiv: 2509.18782 · v2 · pith:6YCVM5Z7new · submitted 2025-09-23 · ⚛️ physics.soc-ph · physics.app-ph

Emergence of power laws in hierarchical dynamics on multi-level graphs

Tommaso Rondini , Gregorio Berselli , Mirko Degli Esposti , Armando Bazzani This is my paper

Pith reviewed 2026-05-18 14:56 UTC · model grok-4.3

classification ⚛️ physics.soc-ph physics.app-ph

keywords power-law distributionshierarchical networksrailway delaysqueue dynamicsLaplacian fluctuationspriority rulescomplex systems

0 comments

The pith

A hierarchical queue model with Laplacian fluctuations reproduces the power-law exponent of Italian local train delays.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how power-law delay distributions emerge in a multi-level railway network where high-speed and local trains follow simple priority rules. Stochastic fluctuations in travel times, modeled after empirical data, lead to local trains accumulating more delays than high-speed ones. A queue-based dynamical model calibrated on Italian data matches the observed power-law behavior and highlights how policy differences, such as delay compensation thresholds, create distinct cut-offs in the distributions for Italy but not for Germany. This demonstrates that complex statistical patterns can arise from basic hierarchical structures and rule-based interactions without needing elaborate network topologies.

Core claim

In a multi-level hierarchical network of train agents with assigned priority levels, introducing Laplacian-distributed stochastic fluctuations into scheduled travel times causes local trains to exhibit a higher incidence of large delays, resulting in power-law distributions whose exponent matches that observed in Italian railway data; the model also predicts policy-induced cut-offs at 30 minutes for high-speed trains and 60 minutes for local trains, which are absent in the German system lacking similar priority policies.

What carries the argument

The queue-based dynamical model on a multi-level graph where agents follow priority rules and experience Laplacian fluctuations in travel times.

If this is right

Local trains experience markedly higher delays than high-speed trains under the same fluctuations.
Delay distributions show cut-offs at 30 and 60 minutes corresponding to Italian refund criteria.
Such cut-offs do not appear in the German dataset where no comparable policies exist.
Operational policies like priority assignment influence the shape of the delay distributions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Power-law behaviors in other complex systems might similarly stem from hierarchical priority mechanisms rather than intricate interactions.
Adjusting the priority rules or fluctuation distributions could allow prediction of delay statistics in alternative transportation networks.
The approach might extend to modeling congestion in other multi-level systems like supply chains or communication networks.

Load-bearing premise

Stochastic fluctuations in scheduled travel times are assumed to follow a Laplacian distribution taken directly from empirical railway data.

What would settle it

If replacing the Laplacian fluctuations with a different distribution such as Gaussian causes the model to no longer reproduce the observed power-law exponent in Italian local train delays.

read the original abstract

Power-law distributions are widely recognized in complex systems physics as indicative of underlying complexity in interaction networks and critical macroscopic behavior. Previous studies, notably those of Newman and others, have emphasized the importance of network structure and dynamics in understanding the emergence of such statistical patterns and predicting extreme events. In this study, we investigate the emergence of power-law behavior in delay distributions within a multi-level hierarchical network of agents governed by simple priority rules. Using railway systems as a case study, we model the dynamics of high-speed and local trains agents assigned distinct priority levels-operating within a simplified hierarchical network framework. By introducing Laplacian-distributed stochastic fluctuations into scheduled travel times, derived from empirical data, we observe that local trains exhibit a markedly higher incidence of higher delays than high-speed trains. To account for this phenomenon, we propose a queue-based dynamical model, calibrated using Italian railway data, and validate our findings through comparative analysis with both Italian and German datasets. The model accurately reproduces the empirically observed power-law exponent associated with the Italian local train delays. Furthermore, we analyze the influence of operational policies, such as priority assignment and delay compensation thresholds, revealing distinct cut-offs in delay distributions at 30 and 60 minutes for high-speed and local trains, respectively-corresponding to refund eligibility criteria in Italy. Such cut-offs are absent in the German case, where no comparable priority-change policies are in effect. These results underscore the capacity of simple hierarchical structures and rule-based dynamics to generate complex statistical behaviors without necessitating intricate interaction networks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper shows a calibrated queue model on hierarchical train networks reproduces the Italian delay power-law exponent and links cut-offs to refund policies, but the fit uses data-derived fluctuations and tuning so the emergence claim needs robustness checks.

read the letter

The key point is that a hierarchical queue model with priority rules and calibrated parameters reproduces the power-law delay distribution seen in Italian local trains, and it ties the cut-offs in those distributions to refund policies that differ from Germany. What the paper does is apply a multi-level graph setup to railway agents, assigning distinct priorities to high-speed and local trains. They add Laplacian-distributed fluctuations to scheduled times based on real data, then simulate the queue dynamics. The result is that local trains accumulate more high delays, and the model matches the Italian exponent after calibration. They also show how policy thresholds create sharp cut-offs at 30 and 60 minutes for the two train types in Italy, which do not appear in the German data where such policies are not in place. This comparative part is solid and gives a clear mechanistic story for why the statistics look different across countries. The underlying idea that simple rules on hierarchical networks can generate power laws is not entirely new, but the specific mapping to train operations and policy effects adds a practical layer. The main limitation is that the fluctuations are taken from the empirical data, and the model has free parameters that are adjusted to fit the observed exponent. This makes the reproduction less of an independent test of the hierarchical mechanism and more of a demonstration that the queue dynamics can propagate the input tails when tuned appropriately. If the power law depends heavily on the choice of Laplacian or the calibration constants, then the claim of emergence from the structure alone needs more support, such as sensitivity tests. The citation pattern seems standard for complex systems work, referencing Newman and similar on power laws. For readers in physics of complex systems or applied network modeling, especially those dealing with transport delays, this offers a useful framework. It is not a major theoretical advance but provides a concrete case study with real data distinctions. The work shows clear thinking in connecting rules to statistics, so it is worth sending for peer review. Referees can check the details of the calibration and whether additional robustness results are needed.

Referee Report

3 major / 2 minor

Summary. The paper claims that a queue-based dynamical model on a multi-level hierarchical network, incorporating priority rules for high-speed versus local trains and Laplacian-distributed stochastic fluctuations derived from empirical railway data, reproduces the power-law exponent observed in Italian local train delays. It further asserts that the model accounts for distinct delay cut-offs at 30 and 60 minutes in the Italian system (tied to refund policies) that are absent in German data, demonstrating that simple hierarchical structures and rule-based dynamics suffice to generate complex statistical patterns.

Significance. If the reproduction of the exponent can be shown to arise independently from the hierarchical rules rather than from data-driven calibration, the result would be significant for complex systems physics: it would provide a parsimonious mechanism linking multi-level network structure and priority dynamics to power-law delay statistics, with potential applicability to other transportation or queuing systems. The explicit cross-validation against both Italian and German empirical patterns is a positive feature that strengthens the policy-related interpretation.

major comments (3)

[Abstract and model calibration] Abstract and calibration description: the model is calibrated using Italian railway data specifically to reproduce the observed power-law exponent for local train delays. This makes the reported match a post-hoc fit rather than an independent derivation from the hierarchical priority rules and queue dynamics, directly weakening the central emergence claim. The paper should test whether the exponent persists under fixed parameters or alternative input distributions.
[Stochastic fluctuations section] Stochastic fluctuations and input distribution: the Laplacian distribution is taken directly from the same empirical Italian data that drives the calibration. Without an explicit robustness check (e.g., replacing the Laplacian with a Gaussian or exponential while keeping network rules fixed), it remains unclear whether the heavy tails are generated by the multi-level hierarchy or simply propagated from the input.
[Validation and German comparison] Comparative validation: while the model reproduces Italian cut-offs at 30/60 min and their absence in Germany, the manuscript does not state whether the same parameter values are used for the German simulation or whether the priority-change policy is the sole difference. Quantitative fit metrics (e.g., Kolmogorov-Smirnov statistics or exponent confidence intervals) for the German case are needed to support the policy-distinction claim.

minor comments (2)

[Model definition] Clarify the exact mathematical definition of the queue update rules and delay-compensation thresholds with numbered equations; the current prose description leaves the priority assignment and threshold logic ambiguous.
[Results figures] Add error bars or bootstrap confidence intervals to the reported power-law exponents in figures and tables comparing model output to data.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their detailed and constructive report. We address each of the major comments below, indicating where we will revise the manuscript to incorporate the suggestions and where we provide additional clarification.

read point-by-point responses

Referee: [Abstract and model calibration] Abstract and calibration description: the model is calibrated using Italian railway data specifically to reproduce the observed power-law exponent for local train delays. This makes the reported match a post-hoc fit rather than an independent derivation from the hierarchical priority rules and queue dynamics, directly weakening the central emergence claim. The paper should test whether the exponent persists under fixed parameters or alternative input distributions.

Authors: We agree that the calibration to Italian data is a key aspect, and we will revise the manuscript to better emphasize that the power-law exponent arises from the hierarchical queue dynamics rather than being directly imposed. Specifically, we will add simulations using fixed parameters derived from the network structure alone and test alternative input distributions to demonstrate the robustness of the emergence. This will strengthen the claim that the multi-level hierarchy is responsible for the observed statistics. revision: yes
Referee: [Stochastic fluctuations section] Stochastic fluctuations and input distribution: the Laplacian distribution is taken directly from the same empirical Italian data that drives the calibration. Without an explicit robustness check (e.g., replacing the Laplacian with a Gaussian or exponential while keeping network rules fixed), it remains unclear whether the heavy tails are generated by the multi-level hierarchy or simply propagated from the input.

Authors: The choice of the Laplacian distribution is motivated by its superior fit to the empirical delay fluctuations in the Italian railway data. However, to address the concern, we will include a new robustness analysis in the revised version, where we replace the Laplacian with Gaussian and exponential distributions while keeping all network rules, priorities, and hierarchical structure fixed. We expect this to show that the power-law tails are indeed generated by the model dynamics. revision: yes
Referee: [Validation and German comparison] Comparative validation: while the model reproduces Italian cut-offs at 30/60 min and their absence in Germany, the manuscript does not state whether the same parameter values are used for the German simulation or whether the priority-change policy is the sole difference. Quantitative fit metrics (e.g., Kolmogorov-Smirnov statistics or exponent confidence intervals) for the German case are needed to support the policy-distinction claim.

Authors: We confirm that the German simulations use the same parameter values for the network topology, queue dynamics, and stochastic fluctuations as the Italian case, with the only modification being the removal of the priority-change policy linked to the refund thresholds. We will explicitly state this in the revised text. Additionally, we will add quantitative metrics, including Kolmogorov-Smirnov statistics comparing the simulated and empirical distributions, as well as confidence intervals for the fitted exponents in the German scenario. revision: yes

Circularity Check

2 steps flagged

Model calibration on Italian railway data to reproduce observed power-law exponent reduces claim to fitted reproduction

specific steps

fitted input called prediction [Abstract]
"By introducing Laplacian-distributed stochastic fluctuations into scheduled travel times, derived from empirical data, we observe that local trains exhibit a markedly higher incidence of higher delays than high-speed trains. To account for this phenomenon, we propose a queue-based dynamical model, calibrated using Italian railway data, and validate our findings through comparative analysis with both Italian and German datasets. The model accurately reproduces the empirically observed power-law exponent associated with the Italian local train delays."

The Laplacian input distribution is taken from the identical empirical railway dataset whose power-law exponent is later 'reproduced'; the queue model is then explicitly calibrated on Italian data to match that same exponent. The match is therefore statistically forced by the fitting step and input choice rather than derived from the multi-level hierarchy or priority rules alone.
fitted input called prediction [Abstract]
"we propose a queue-based dynamical model, calibrated using Italian railway data... The model accurately reproduces the empirically observed power-law exponent associated with the Italian local train delays."

Calibration of the model's two parameters on the target Italian dataset to reproduce the observed exponent means the reported agreement is a direct consequence of the fitting procedure, not an a-priori prediction from the hierarchical dynamics.

full rationale

The paper introduces Laplacian fluctuations derived directly from the same empirical Italian railway data used for calibration, then fits a two-parameter queue-based model on that data specifically to match the target delay power-law exponent. The reported 'accurate reproduction' therefore reduces by construction to a post-hoc fit of the input tails and parameters rather than an independent emergence from the hierarchical priority rules. Comparative validation on German data is mentioned but does not remove the load-bearing dependence on the Italian calibration for the central exponent claim. This matches the fitted-input-called-prediction pattern with partial self-definitional elements in the input distribution choice.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on empirical calibration of fluctuation and queue parameters plus the domain assumption of strict hierarchical priority rules; no new entities are postulated.

free parameters (2)

Laplacian fluctuation parameters
Parameters for stochastic travel-time variations are taken directly from empirical railway data to drive the model.
Queue model calibration constants
Parameters are adjusted to match the Italian delay exponent and cut-offs.

axioms (1)

domain assumption Trains follow simple priority rules within a multi-level hierarchical network structure.
This rule set is the foundational premise for the queue dynamics and power-law emergence.

pith-pipeline@v0.9.0 · 5809 in / 1369 out tokens · 54895 ms · 2026-05-18T14:56:11.311798+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

Springer, New York (2013)

Bak, P.: How Nature Works: The Science of Self-Organized Critica lity, 1st edn. Springer, New York (2013). https://doi.org/10.1007/978-1-4757-5426-1

work page doi:10.1007/978-1-4757-5426-1 2013
[2]

Reviews of modern physics 74(1), 47 (2002)

Albert, R., Barab´ asi, A.-L.: Statistical mechanics of complex net works. Reviews of modern physics 74(1), 47 (2002)

work page 2002
[3]

science 286(5439), 509–512 (1999)

Barab´ asi, A.-L., Albert, R.: Emergence of scaling in random netwo rks. science 286(5439), 509–512 (1999)

work page 1999
[4]

Phys ica A: Statistical 10 Mechanics and its Applications 369(1), 29–70 (2006)

Dhar, D.: Theoretical studies of self-organized criticality. Phys ica A: Statistical 10 Mechanics and its Applications 369(1), 29–70 (2006)

work page 2006
[5]

Cont emporary physics 46(5), 323–351 (2005)

Newman, M.E.: Power laws, pareto distributions and zipf’s law. Cont emporary physics 46(5), 323–351 (2005)

work page 2005
[6]

Physics reports 544(1), 1–122 (2014)

Boccaletti, S., Bianconi, G., Criado, R., Del Genio, C.I., G´ omez-Gar denes, J., Romance, M., Sendina-Nadal, I., Wang, Z., Zanin, M.: The structur e and dynamics of multilayer networks. Physics reports 544(1), 1–122 (2014)

work page 2014
[7]

S cientiﬁc reports 10(1), 18584 (2020)

B¨ uchel, B., Spanninger, T., Corman, F.: Empirical dynamics of railw ay delay propagation identiﬁed during the large-scale rastatt disruption. S cientiﬁc reports 10(1), 18584 (2020)

work page 2020
[8]

EPJ D ata Science 7(1), 35 (2018)

Monechi, B., Gravino, P., Di Clemente, R., Servedio, V.D.: Complex de lay dynam- ics on railway networks from universal laws to realistic modelling. EPJ D ata Science 7(1), 35 (2018)

work page 2018
[9]

nature 453(7196), 779–782 (2008)

Gonzalez, M.C., Hidalgo, C.A., Barabasi, A.-L.: Understanding individu al human mobility patterns. nature 453(7196), 779–782 (2008)

work page 2008
[10]

IEEE Transactions on Computational Social Systems 11(3), 3207–3217 (2023)

Lian, W., Wu, X., Zhou, M., L¨ u, J., Dong, H.: Bi-directional delay pr opagation analysis and modeling for high-speed railway networks under disturb ance. IEEE Transactions on Computational Social Systems 11(3), 3207–3217 (2023)

work page 2023
[11]

https:// www.viaggiatreno.it/

Ferrovie dello Stato Italiane: ViaggiaTreno - Oﬀerte e orari tre ni in Italia. https:// www.viaggiatreno.it/. Accessed: 2025-02-03 (2025)

work page 2025
[12]

Accessed: 2025-05 -14 (2025)

Br¨ ommel, P.: Deutsche Bahn Statistiken. Accessed: 2025-05 -14 (2025). https:// piebro.github.io/deutsche-bahn-statistics/

work page 2025
[13]

Jammalamadaka, S.R., Kozubowski, T.J.: New families of wrapped dis tribu- tions for modeling skew circular data. Communications in Statistics - T heory and Methods 33(9), 2059–2074 (2004) https://doi.org/10.1081/STA-200026570 https://doi.org/10.1081/STA-200026570 Appendix A Asymmetric Laplace Distribution as diﬀerence of two exponentials Starting from ( 1...

work page doi:10.1081/sta-200026570 2059

[1] [1]

Springer, New York (2013)

Bak, P.: How Nature Works: The Science of Self-Organized Critica lity, 1st edn. Springer, New York (2013). https://doi.org/10.1007/978-1-4757-5426-1

work page doi:10.1007/978-1-4757-5426-1 2013

[2] [2]

Reviews of modern physics 74(1), 47 (2002)

Albert, R., Barab´ asi, A.-L.: Statistical mechanics of complex net works. Reviews of modern physics 74(1), 47 (2002)

work page 2002

[3] [3]

science 286(5439), 509–512 (1999)

Barab´ asi, A.-L., Albert, R.: Emergence of scaling in random netwo rks. science 286(5439), 509–512 (1999)

work page 1999

[4] [4]

Phys ica A: Statistical 10 Mechanics and its Applications 369(1), 29–70 (2006)

Dhar, D.: Theoretical studies of self-organized criticality. Phys ica A: Statistical 10 Mechanics and its Applications 369(1), 29–70 (2006)

work page 2006

[5] [5]

Cont emporary physics 46(5), 323–351 (2005)

Newman, M.E.: Power laws, pareto distributions and zipf’s law. Cont emporary physics 46(5), 323–351 (2005)

work page 2005

[6] [6]

Physics reports 544(1), 1–122 (2014)

Boccaletti, S., Bianconi, G., Criado, R., Del Genio, C.I., G´ omez-Gar denes, J., Romance, M., Sendina-Nadal, I., Wang, Z., Zanin, M.: The structur e and dynamics of multilayer networks. Physics reports 544(1), 1–122 (2014)

work page 2014

[7] [7]

S cientiﬁc reports 10(1), 18584 (2020)

B¨ uchel, B., Spanninger, T., Corman, F.: Empirical dynamics of railw ay delay propagation identiﬁed during the large-scale rastatt disruption. S cientiﬁc reports 10(1), 18584 (2020)

work page 2020

[8] [8]

EPJ D ata Science 7(1), 35 (2018)

Monechi, B., Gravino, P., Di Clemente, R., Servedio, V.D.: Complex de lay dynam- ics on railway networks from universal laws to realistic modelling. EPJ D ata Science 7(1), 35 (2018)

work page 2018

[9] [9]

nature 453(7196), 779–782 (2008)

Gonzalez, M.C., Hidalgo, C.A., Barabasi, A.-L.: Understanding individu al human mobility patterns. nature 453(7196), 779–782 (2008)

work page 2008

[10] [10]

IEEE Transactions on Computational Social Systems 11(3), 3207–3217 (2023)

Lian, W., Wu, X., Zhou, M., L¨ u, J., Dong, H.: Bi-directional delay pr opagation analysis and modeling for high-speed railway networks under disturb ance. IEEE Transactions on Computational Social Systems 11(3), 3207–3217 (2023)

work page 2023

[11] [11]

https:// www.viaggiatreno.it/

Ferrovie dello Stato Italiane: ViaggiaTreno - Oﬀerte e orari tre ni in Italia. https:// www.viaggiatreno.it/. Accessed: 2025-02-03 (2025)

work page 2025

[12] [12]

Accessed: 2025-05 -14 (2025)

Br¨ ommel, P.: Deutsche Bahn Statistiken. Accessed: 2025-05 -14 (2025). https:// piebro.github.io/deutsche-bahn-statistics/

work page 2025

[13] [13]

Jammalamadaka, S.R., Kozubowski, T.J.: New families of wrapped dis tribu- tions for modeling skew circular data. Communications in Statistics - T heory and Methods 33(9), 2059–2074 (2004) https://doi.org/10.1081/STA-200026570 https://doi.org/10.1081/STA-200026570 Appendix A Asymmetric Laplace Distribution as diﬀerence of two exponentials Starting from ( 1...

work page doi:10.1081/sta-200026570 2059