[Paper Review] What shakes the FX tree? Understanding currency dominance, dependence and dynamics

Neil F. Johnson, Mark McDonald+3 more|RePEc: Research Papers in Economics|Mar 1, 2005

Complex Systems and Time Series Analysis4 citations

TL;DR

This paper applies Minimum Spanning Trees (MSTs) to foreign exchange (FX) market data to analyze currency correlations and dynamics, revealing that certain currencies act as dominant or dependent nodes over time. The study finds that FX market linkages are highly persistent—over 50% of MST connections survive for more than two years—indicating a stable, evolving 'ecology of clusters' with measurable leadership structures.

ABSTRACT

There is intense interest in understanding the stochastic and dynamical properties of the global Foreign Exchange (FX) market, whose daily transactions exceed one trillion US dollars. This is a formidable task since the FX market is characterized by a web of fluctuating exchange rates, with subtle inter-dependencies which may change in time. In practice, traders talk of particular currencies being 'in play' during a particular period of time -- yet there is no established machinery for detecting such important information. Here we apply the construction of Minimum Spanning Trees (MSTs) to the FX market, and show that the MST can capture important features of the global FX dynamics. Moreover, we show that the MST can help identify momentarily dominant and dependent currencies.

Motivation & Objective

• To understand the time-dependent structure of currency correlations in the global FX market, which is highly active and complex.
• To address the lack of systematic tools for detecting when specific currencies are 'in play'—i.e., dominant or dependent in market dynamics.
• To investigate whether the FX market exhibits persistent, robust network structures despite rapid fluctuations.
• To explore whether MSTs can serve as a diagnostic tool for identifying leadership and clustering patterns in currency networks.
• To examine the implications of network stability for modeling and predicting FX market behavior.

Proposed method

• Constructs Minimum Spanning Trees (MSTs) from daily exchange rate returns of major FX currency pairs over a two-year period.
• Uses Pearson correlation coefficients between exchange rate returns as edge weights to build the correlation network.
• Applies the Kruskal algorithm to generate MSTs that minimize total edge weight while ensuring all nodes are connected.
• Analyzes temporal evolution of MSTs using multi-step survival ratios to assess link persistence over time.
• Compares two definitions of survival ratio: one restrictive (requiring continuous presence across steps) and one generous (requiring only start and end presence).
• Visualizes and interprets MST structures to identify regional clustering and dominant currencies (leaders) within the network.

Experimental results

Research questions

• RQ1 Which currencies act as dominant or dependent nodes in the FX market at any given time, and how do these roles shift over time?
• RQ2 To what extent are the connections in the FX correlation network stable over time, and what fraction of links persist beyond two years?
• RQ3 How do the structural features of real FX MSTs differ from those generated by randomized data?
• RQ4 Can the MST framework detect transient clusters of correlated currencies and their dynamic evolution?
• RQ5 What is the impact of external shocks (e.g., news events) on the stability and topology of the FX network?

Key findings

• The MST constructed from real FX data exhibits clear regional clustering, unlike MSTs from randomized data, indicating genuine underlying structure.
• Over 50% of MST connections survive for more than two years, with the restrictive survival ratio still showing 54 out of 109 links surviving, indicating extreme persistence in currency correlations.
• The MST reveals a dynamic 'ecology of clusters' where clusters form, evolve, and dissipate over time, suggesting the FX market is not static but self-organizing.
• The network structure is robust to short-term fluctuations, with long-lived links indicating deep-seated interdependencies among major currencies.
• The MST framework successfully identifies momentary dominant and dependent currencies, offering a data-driven method to detect when a currency is 'in play'.
• The two definitions of multi-step survival ratio form a 'corridor', confirming that even conservative estimates show strong link longevity, underscoring the stability of FX correlation networks.