Weighted Signed Networks Reveal Interactions between US Foreign Exchange Rates

TL;DR

This paper uses network analysis to study how US foreign exchange rates interact, revealing trends in global financial connections and simpler fluctuation patterns over time.

Contribution

A new method for selecting time scales in DCCA analysis and a fluctuation propagation algorithm for exchange rate networks.

Findings

Negative cross-correlations between exchange rates have become increasingly rare over the past 30 years.

Positive cross-correlations have increased, reflecting greater global financial interconnectivity.

Fluctuation propagation in exchange rate networks has become simpler and more predictable over time.

Abstract

Correlations between exchange rates are valuable for illuminating the dynamics of international trade and the financial dynamics of countries. This paper explores the changing interactions of the US foreign exchange market based on detrended cross-correlation analysis. First, we propose an objective way to choose a time scale parameter appropriate for comparing different samples by maximizing the summed magnitude of all DCCA coefficients. We then build weighted signed networks under this optimized time scale, which can clearly display the complex relationships between different exchange rates. Our study shows negative cross-correlations have become pyramidally rare in the past three decades. Both the number and strength of positive cross-correlations have grown, paralleling the increase in global interconnectivity. The balanced strong triads are identified subsequently after the network centrality analysis. Generally, while the strong development links revealed by foreign exchange have begun to spread to Asia since 2010, Europe is still the center of world finance, with the euro and Danish krone consistently maintaining the closest balanced development relationship. Finally, we propose a fluctuation propagation algorithm to investigate the propagation pattern of fluctuations in the inferred exchange rate networks. The results show that, over time, fluctuation propagation patterns have become simpler and more predictable.

Figures