New York closes a few hours before Tokyo opens. FlowState measures how much New York’s close helps predict Tokyo’s next move, using Transfer Entropy (Schreiber, 2000). Each reading covers the past year of trading and gets a grade, from Strong to No credible lead.
FlowState is a research and data product. It shows how closely Tokyo follows New York over time. It is not a daily trading signal.
Insights
The direction holds. In every period with a detectable link since 2017, information has flowed from New York to Tokyo.
The strength does not. Before August 2024, the link was graded Strong on 68% of trading days. Since then, on 17%.
Why FlowState
It sees direction. Correlation tells you two markets move together. It can’t tell you which one moves first. FlowState can.
It says when there’s nothing to see. When the data doesn’t support a link, FlowState says so: “No credible lead.” A weak period is labeled as one.
It takes the yen out. When the yen moves, both markets can move with it. FlowState strips out that shared push, so the reading reflects New York’s own influence on Tokyo.
It shows the full record. Every reading since 2017 is one click away, with major shocks marked, from the 2020 COVID crash to the 2024 yen carry unwind.
Use Cases
Testing a model’s assumptions. Many cross-market models assume Tokyo takes its cue from New York. FlowState shows when that held and when it broke down.
Spotting a regime change. The link weakened after August 2024 and has stayed weak. FlowState shows shifts like this as they build, not years later.
Reviewing a shock. See how tightly Tokyo was tied to New York in the year leading up to a crash or a policy surprise.
Methodology
FlowState asks one question: does knowing New York’s close help predict Tokyo’s next move, beyond what Tokyo’s own recent history already tells you?
It answers with Transfer Entropy, a measure from information theory that detects the direction of information flow without assuming markets behave in any particular way. (Dimpfl & Peter, 2018; Kuang, 2021) Each reading is compared against shuffled versions of the same data, then graded by five reliability checks.
Validated on the Nikkei 225 and S&P 500 pair, with DAX and S&P 500 under active research. Other pairs, such as HSI, FTSE, and ASX, are available on request, subject to validation.
Information flow among stock markets leads to the co-movement of stock prices and even the financial risk contagion globally. Measuring information flow on multi time-scales helps to develop optimal portfolio strategy for market participants and establish targeted financial regulation for policy makers, but is a challenge for traditional econometrics. This study aims to construct the information flow networks on multi time-scales among international stock markets. First, we use the data of 31 stock market indices during 2007 to 2018, and decompose each of stock index series into short and mid-long time-scale components by empirical mode decomposition (EMD). Then, we extend a novel concept in information science, called transfer entropy (TE), to measure information flow and characterize causality structure between two stock markets, and find that information flows inside of EU stock markets only cluster on short time-scale but not on mid-long time-scale. Moreover, both out-node and in-node centrality measurements for developed markets in the entropy-based networks are much higher on short time-scale, indicating that these markets are more dominant but more vulnerable to short-term risk contagion. Finally, based on the minimum spanning tree (MST) model, we extract the most probable risk contagion paths from entropy-based networks. The results show that outflow MSTs on short time-scale follow a star-like structure to precipitate the risk contagion in the international stock markets, especially during crisis period (2007∼ 2010). Overall, we emphasize the financial risk contagion on short time-scale.
2018
Elsevier
Analyzing volatility transmission using group transfer entropy
We analyze the transmission of volatility between the oil, stock, gold, and currency markets using transfer entropy. Our approach, which we denote by group transfer entropy, measures the overall sensitivity of a process to lagged realizations of a group of other processes. We supplement our measure by a suitable block bootstrap approach which allows to conduct inference. Our empirical analysis uses daily data of OVX, VIX, GVZ, and EVZ from 2008 to 2017. The results show that oil and stock market volatility are the most affected by past volatility changes in the other markets, which highlights the importance for investors to hedge against this predictable risk transmission. Furthermore, we show that group transfer entropy can unveil nonlinear dynamics beyond the standard vector autoregressive methodology and provides a useful basis for forecast averaging.
An information theoretic measure is derived that quantifies the statistical coherence between systems evolving in time. The standard time delayed mutual information fails to distinguish information that is actually exchanged from shared information due to common history and input signals. In our new approach, these influences are excluded by appropriate conditioning of transition probabilities. The resulting transfer entropy is able to distinguish effectively driving and responding elements and to detect asymmetry in the interaction of subsystems.