Paper: arXiv 2310.16841
Authors: Yi Jiang, Shohei Shimizu
Abstract
While economic theory explains the linkages among the financial markets of different countries, empirical studies mainly verify the linkages through Granger causality, without considering latent variables or instantaneous effects. Their findings are inconsistent regarding the existence of causal linkages among financial markets, which might be attributed to differences in the focused markets, data periods, and methods applied. Our study adopts causal discovery methods including VAR-LiNGAM and LPCMCI with domain knowledge to explore the linkages among financial markets in Japan and the United States (US) for the post Covid-19 pandemic period under divergent monetary policy directions. The VAR-LiNGAM results reveal that the previous day’s US market influences the following day’s Japanese market for both stocks and bonds, and the bond markets of the previous day impact the following day’s foreign exchange (FX) market directly and the following day’s Japanese stock market indirectly. The LPCMCI results indicate the existence of potential latent confounders. Our results demonstrate that VAR-LiNGAM uniquely identifies the directed acyclic graph (DAG), and thus provides informative insight into the causal relationship when the assumptions are considered valid. Our study contributes to a better understanding of the linkages among financial markets in the analyzed data period by supporting the existence of linkages between Japan and the US for the same financial markets and among FX, stock, and bond markets, thus highlighting the importance of leveraging causal discovery methods in the financial domain.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced causal discovery methods like VAR-LiNGAM and LPCMCI, involving DAGs and latent variable modeling, while using real financial market data from a specific period and presenting statistical results.
Research Flowchart
flowchart TD
A["Research Goal:<br>Explore linkages among financial markets in<br>Japan & US for post-COVID period"] --> B{"Methodology"}
subgraph B ["Causal Discovery Methods"]
B1["VAR-LiNGAM"]
B2["LPCMCI"]
end
C["Data: Post-COVID FX, Stock, & Bond Markets of Japan & US"] --> B
B --> D{"Computational Processes"}
subgraph D ["DAG Estimation & Analysis"]
D1["Identifies Causal Structure"]
D2["Detects Latent Confounders"]
end
D --> E["Key Findings & Outcomes"]
E --> E1["US Market → Japanese Market"]
E --> E2["Bond Market → FX & Stock Market"]
E --> E3["VAR-LiNGAM uniquely identifies DAG"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The study applies causal discovery methods, specifically VAR‑LiNGAM and LPCMCI with domain knowledge, to investigate financial market linkages between Japan and the United States for the post‑COVID‑19 period.
Source sentence
Our study adopts causal discovery methods including VAR-LiNGAM and LPCMCI with domain knowledge to explore the linkages among financial markets in Japan and the United States (US) for the post Covid-19 pandemic period under divergent monetary policy directions.
Assumes 4
- Time‑series are stationary (or made stationary by first‑order differencing) as required by VAR and VAR‑LiNGAM. statistical
Source sentence
Second, we adopted the augmented Dickey–Fuller (ADF) test to check whether these time series are stationary, which is an assumption of VAR and VAR-LiNGAM. Therefore, we took their first-order difference and confirmed that the null hypotheses of ADF test are rejected this time, with p-values close to zero.
- Residuals of the VAR model are non‑Gaussian, satisfying the non‑Gaussian disturbance assumption of VAR‑LiNGAM. model
Source sentence
We used the VAR model with all six variables and tested the normality of the residuals of the VAR model by Jarque–Bera test. VAR-LiNGAM In verifying the assumptions for VAR-LiNGAM, we confirmed the non-Gaussian distri-bution of the disturbance terms in Section 5.2.
- Relationships between variables are linear, as checked via scatterplots. model
Source sentence
We additionally verified the linearity of the relationships between the variables using scatterplots. The scatterplots for each pair of variables (Appendix E) indicate a linear relationship between some of the variable pairs.
- Causal sufficiency is not guaranteed; latent confounders may exist, addressed by LPCMCI. model
Source sentence
The LPCMCI results indicate the existence of potential latent confounders. Furthermore, we leverage the LPCMCI model to validate the results when a latent variable exists, as the causal sufficiency assumption is not guaranteed by previous research or related theories.
Data
- USD/JPY exchange rate market-prices unknown
Source sentence
For the US Dollar (USD)/JPY rate (USD), we used data from one of the three Japanese mega-banks, Mizuho Bank.
- Nikkei 225 closing price market-prices public
Source sentence
The closing price of the Nikkei 225 index (Close Nikkei), which represents the Japanese stock market, is obtained from the download center of Nikkei.
- S&P 500 closing price market-prices public
Source sentence
Data on the closing prices of Standard & Poor’s (S&P) 500 (Close SP), a weighted index of 500 listed companies in the US representing the US stock market, and the closing yields of the US 10-year Treasury note (Close US10Y ) were downloaded from the open-source website MarketWatch.
- US 10‑year Treasury note yield market-prices public
Source sentence
Data on the closing prices of Standard & Poor’s (S&P) 500 (Close SP), a weighted index of 500 listed companies in the US representing the US stock market, and the closing yields of the US 10-year Treasury note (Close US10Y ) were downloaded from the open-source website MarketWatch.
- Japanese Government Bond (JGB) futures closing price market-prices public
Source sentence
Data for the Japanese bond market, including JGB futures closing prices (Close JGBF) and 10-year JGB closing yields (Close JGB), were obtained from invest-ing.com, another open-source website.
- Japanese Government Bond (JGB) 10‑year yield market-prices public
Source sentence
Data for the Japanese bond market, including JGB futures closing prices (Close JGBF) and 10-year JGB closing yields (Close JGB), were obtained from invest-ing.com, another open-source website.
Costs
Not applicable
Finds 5
- VAR‑LiNGAM shows that the previous day’s US market (stock and bond) influences the following day’s Japanese market (stock and bond). statistical
Source sentence
The VAR-LiNGAM results reveal that the previous day's US market influences the following day's Japanese market for both stocks and bonds, and the bond markets of the previous day impact the following day's foreign exchange (FX) market directly and the following day's Japanese stock market indirectly.
- The previous day’s US bond market directly impacts the next day’s FX market and indirectly the next day’s Japanese stock market. statistical
Source sentence
The VAR-LiNGAM results reveal that the previous day's US market influences the following day's Japanese market for both stocks and bonds, and the bond markets of the previous day impact the following day's foreign exchange (FX) market directly and the following day's Japanese stock market indirectly.
- LPCMCI indicates the presence of potential latent confounders in the causal structure. statistical
Source sentence
The LPCMCI results indicate the existence of potential latent confounders.
- The bond markets of the previous day act as information sources for the following day’s FX market, supporting the interest‑rate parity theory. theoretical
Source sentence
Regarding the linkages among different markets, the VAR-LiNGAM results show that the bond markets of the previous day are the information sources for the following day’s FX market, which supports the IRP theory.
- The FX market directly impacts the same‑day Japanese stock market. statistical
Source sentence
At the same time, the FX market directly impacts the same day’s Japanese stock market.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 28 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.