[Paper Review] The evolving liaisons between the transaction networks of Bitcoin and its price dynamics

Alexandre Bovet, Carlo Campajola+5 more|arXiv (Cornell University)|Jul 8, 2019

Blockchain Technology Applications and Security8 citations

TL;DR

This study investigates causal relationships between Bitcoin's transaction network topology and its price dynamics using four network representations—address and user networks—at daily and weekly frequencies over nine years. It reveals that price drops are preceded by increased heterogeneity in transaction activity, indicating herding behavior, with significant Granger causality from network structure to price movements, especially post-Mt. Gox collapse.

ABSTRACT

Cryptocurrencies are distributed systems that allow exchanges of native tokens among participants, or the exchange of such tokens for fiat currencies in markets external to these public ledgers. The availability of their complete historical bookkeeping opens up the possibility of understanding the relationship between aggregated users' behaviour and the cryptocurrency pricing in exchange markets. This paper analyses the properties of the transaction network of Bitcoin. We consider four different representations of it, over a period of nine years since the Bitcoin creation and involving 16 million users and 283 million transactions. By analysing these networks, we show the existence of causal relationships between Bitcoin price movements and changes of its transaction network topology. Our results reveal the interplay between structural quantities, indicative of the collective behaviour of Bitcoin users, and price movements, showing that, during price drops, the system is characterised by a larger heterogeneity of nodes activity.

Motivation & Objective

• To understand how structural changes in Bitcoin's transaction networks relate to price movements over time.
• To identify causal relationships between network topology metrics and Bitcoin price dynamics.
• To analyze the impact of major events—particularly the Mt. Gox bankruptcy—on network and price evolution.
• To explore whether network heterogeneity signals collective user behavior such as herding during market volatility.
• To compare four distinct network representations (address/user, daily/weekly) for robustness in detecting structural-price linkages.

Proposed method

• Four network representations are constructed: address network (AN) and user network (UN), each at daily and weekly time scales, using Bitcoin's full transaction history.
• Network topology is analyzed via out-degree distribution moments, which quantify heterogeneity in user transaction activity.
• Granger causality tests are applied to detect directional relationships between network metrics and price, including standard, lagged, and tail-event variants.
• Tail-event Granger causality uses conditional quantiles (10th and 90th percentiles) computed via rolling windows and non-parametric estimation (Davis method) to detect extreme event dependencies.
• A Daniell kernel with bandwidth M is used in spectral density-based test statistics to assess causality at different time scales.
• False Discovery Rate (FDR) correction at 5% is applied to control for multiple hypothesis testing in causality analysis.

Experimental results

Research questions

• RQ1 Does the structural evolution of Bitcoin’s transaction network causally influence its price movements?
• RQ2 How do changes in network heterogeneity—measured by out-degree distribution moments—relate to price drops or rallies?
• RQ3 Are there detectable causal links between extreme network events (e.g., high activity spikes) and extreme price movements?
• RQ4 How did the Mt. Gox bankruptcy in 2014 alter the causal relationship between network structure and price?
• RQ5 Do different network representations (address vs. user, daily vs. weekly) yield consistent causal patterns?

Key findings

• A significant causal relationship exists from network structure to Bitcoin price, particularly during periods of price decline, indicating that structural changes precede and may drive price movements.
• The moments of the out-degree distribution in the transaction network show increased heterogeneity during price drops, suggesting herding behavior among users.
• Post-Mt. Gox, the system underwent a clear regime shift, with stronger and more persistent causal links between network topology and price dynamics.
• Tail-event Granger causality reveals that extreme network activity (e.g., high out-degree events) predicts extreme price movements, especially on the downside.
• The daily network representation shows stronger causal signals than the weekly one, likely due to higher temporal resolution capturing short-term feedback loops.
• The user network representation reveals more robust causal relationships than the address network, implying that inferred user-level behavior better reflects collective economic decision-making.