External Shocks and Volatility Overflow among the Exchange Rate of the Yen, Nikkei, TOPIX and Sectoral Stock Indices
Round 1
Reviewer 1 Report
The paper is interesting and a good fit for JRFM. I have however a few comments that should improve the chances of the paper to be accepted at JRFM:
1) The motivation of the paper should be better: is there anything that is not well understood about Japanese financial markets and this paper aims at explaining?
2) The models are very simple. The author should also employ spillover models like in Diebold and Yilmaz.
3) The explanations and results should be put in a more general framework in the literature. Particularly, the explanations at page 8 are not up to the standards of a scientific journal as they too repetitive.
Author Response
Replies to respected Reviewers’ comments and questions
First of all, I would like to thank both reviewers for their time and very constructive comments and suggestions. I found all comments and suggestions beneficial for the improvement of my paper. Within ten days (a period required by the editors for revision), I adjusted the paper (as much as I could) according to the comments and suggestions received. The revised version is uploaded. The modified and added parts are marked with yellow color. The details are given below.
Reviewer 1
Comment: The motivation of the paper should be better: is there anything that is not well understood about Japanese financial markets and this paper aims at explaining?
Reply: To present the motivation better, I adjusted the lines 24-25, 39-45, and added additional information to the lines 55-80 in the Introduction.
Comment: The models are very simple. The author should also employ spillover models like in Diebold and Yilmaz.
Reply: I added more equations and explanations for the models in the Methodology section. I am specialized in GARCH type models, and the paper is written within the frame of a proposal for a research grant where the methodology is mentioned as GARCH type model. Receiving your suggestion, I checked Diebold and Yilmaz's methodology and found it very useful. Thank you very much! I mentioned it in Introduction and Conclusion sections, and I will consider it in my future research.
Comment: The explanations and results should be put in a more general framework in the literature. Particularly, the explanations at page 8 are not up to the standards of a scientific journal as they too repetitive.
Reply: I added additional information from different literature in the Introduction and Methodology sections. I also added some information about the differences between my research results and other literature in the Conclusion section. About explanations at page 8 (page 10 of the revised version), I tried to present that information in the form of a table and give short descriptions. The results were worse. So, as that part is the most important part of my findings, I kept them as in the original version. (I am very sorry. Your suggestion is very appropriate. However, I did not find a better version for that part)
Author Response File: 
Author Response.docx
Reviewer 2 Report
Report on the Manuscript ID jrfm-1436619:
External shocks and volatility overflow among the exchange rate of the yen, Nikkei, TOPIX and sectoral stock indices
Summary
The paper studies the changes in volatility overflow among the Japanese yen exchange rate, the Nikkei Stock Average, the Tokyo Stock Price Index (TOPIX), and the TOPIX sectoral indices using the EGARCH model. In addition, by testing volatility spillovers, the analysis revealed that volatility could spread indirectly across variables. Along these lines, external shocks could also indirectly affect the volatility of the variables.
Main comments
The article enriches the literature on the Japanese financial markets by both analyzing the effects of external shocks and volatility spillovers between equity and sectoral indices and among sectoral indices. However, the methodological and empirical structure of the paper needs to be improved.
Introduction
The introduction needs to be improved by:
- including a comprehensive literature review of methods to investigate volatility spillover, the term volatility spillover and the importance of volatility spillover (e.g., volatility spillover might be important when considering how quickly the market adapts to new information or for testing various hypotheses in finance and macroeconomics);
- clearly describing the problem of interest, the gaps in current work that merit attention, the specific research questions being answered, and the significance of the contributions in this paper.
Methodological Framework
The methodological part should be revised by:
- first introducing the testing procedure and then giving more details on Cheung and Ng's (1996) and Hong's (2001) tests for causality in mean and variance;
- providing more generality to the process r??=μ??+z??√h?? ;
- rewriting the conditional mean μ?? and variance h?? in a more general ARMA-GARCH model. For example, E[|z??|]=(2/π)1/2 if the underlying distribution follows the Normal distribution, but E[|z??|] is different for a standardized t-distribution.
Data Description and Empirical Findings
The empirical analysis should be extended through robustness checks.
Causality relationship
- The sampling period and the main time series used are the same as in Sultonov (2020). In this regard, in order to clearly differentiate the contribution of this research from the previous one, I would like the analysis to also consider the effects of the current Covid-19 pandemic.
- Ljung-Box Q(r) statistics for the first r autocorrelations of standardized residuals and standardized squared residuals are usually reported also for r > 5.
- The results in Hong (2001) reveal that non-uniform weighting is more powerful than uniform weighting in detecting volatility spillover between exchange rates. Therefore, it would be interesting to report the sensitivity analysis for different choices of k(.) and M.
- The two-way and one-way tests to identify the simultaneous causality and the direction of causality in mean and variance should be reported.
Effect of BR and USE
- In analyzing the effect of BR and USE (and hopefully Covid-19 lockdown) five days before and after these events (which, however, should be detected using structural breaks tests) are considered. How do the results change when a longer period is considered? Volatility shocks can be extremely persistent. For example, the average number of periods in which volatility reverts to its long-run level can be measured by the “half-life” of the volatility shock.
- At the same time, models can be estimated by using dummy variables in the conditional mean and variance equations, equal to unity from the break date forward, zero otherwise.
Minor comments
- The quality and resolution of the figures should be improved.
- The contributions and limitations of the paper, as well as future research directions, should be better highlighted.
- Please check the paper for grammar, typos, and missing words.
Author Response
Replies to respected Reviewers’ comments and questions
First of all, I would like to thank both reviewers for their time and very constructive comments and suggestions. I found all comments and suggestions beneficial for the improvement of my paper. Within ten days (a period required by the editors for revision), I adjusted the paper (as much as I could) according to the comments and suggestions received. The revised version is uploaded. The modified and added parts are marked with yellow color. The details are given below.
Reviewer 2
Comment: The article enriches the literature on the Japanese financial markets by both analyzing the effects of external shocks and volatility spillovers between equity and sectoral indices and among sectoral indices. However, the methodological and empirical structure of the paper needs to be improved.
Reply: Thank you very much for your kind comment! I incorporated additional information to the Introduction, Methodology, and Conclusion sections based on your suggestion.
Comment: The introduction needs to be improved by including a comprehensive literature review of methods to investigate volatility spillover, the term volatility spillover and the importance of volatility spillover (e.g., volatility spillover might be important when considering how quickly the market adapts to new information or for testing various hypotheses in finance and macroeconomics); clearly describing the problem of interest, the gaps in current work that merit attention, the specific research questions being answered, and the significance of the contributions in this paper.
Reply: I added additional information from different literature in the Introduction and Methodology sections. I also added some information about the differences between my research results and other literature in the Conclusion section.
Comment: The methodological part should be revised by: first introducing the testing procedure and then giving more details on Cheung and Ng's (1996) and Hong's (2001) tests for causality in mean and variance; providing more generality to the process r??=μ??+z??√h?? ; rewriting the conditional mean μ?? and variance h?? in a more general ARMA-GARCH model. For example, E[|z??|]=(2/π)1/2 if the underlying distribution follows the Normal distribution, but E[|z??|] is different for a standardized t-distribution.
Reply: I added information about Cheung and Ng's (1996) and Hong's (2001) tests in methodology and adjusted the equations. I adjusted the variance equation according to Hamori (2003) and mentioned distribution in the tables of estimations.
Comment: The empirical analysis should be extended through robustness checks.
Reply: I mentioned about robustness test of the results in the Methodology section and in Empirical findings. Also increased the Ljung-Box Q(r) statistics up to 10 lags.
Comment: The sampling period and the main time series used are the same as in Sultonov (2020). In this regard, in order to clearly differentiate the contribution of this research from the previous one, I would like the analysis to also consider the effects of the current Covid-19 pandemic.
Reply: Thank you very much for your suggestion! This paper is written within the frame of a research proposal prepared in 2017. That is why I have some limitations in the selection of the models and the variables. In this paper, I have to highlight the causality relationship and impact of Brexit and the US presidential elections (of 2016) for sectoral indexes. In the Conclusion section, I added your suggestion as a future research task related to this paper.
Comment: Ljung-Box Q(r) statistics for the first r autocorrelations of standardized residuals and standardized squared residuals are usually reported also for r > 5.
Reply: I increased the number of lags up to 10.
Comment: The results in Hong (2001) reveal that non-uniform weighting is more powerful than uniform weighting in detecting volatility spillover between exchange rates. Therefore, it would be interesting to report the sensitivity analysis for different choices of k(.) and M.
Reply: I increased the number of kernels. I added equations to the Methodology section and estimation results to tables in the Empirical findings section.
Comment: The two-way and one-way tests to identify the simultaneous causality and the direction of causality in mean and variance should be reported.
Reply: I added estimation results for contemporaneous correlation to tables in the Empirical findings section.
Comment: In analyzing the effect of BR and USE (and hopefully Covid-19 lockdown) five days before and after these events (which, however, should be detected using structural breaks tests) are considered. How do the results change when a longer period is considered? Volatility shocks can be extremely persistent. For example, the average number of periods in which volatility reverts to its long-run level can be measured by the “half-life” of the volatility shock. At the same time, models can be estimated by using dummy variables in the conditional mean and variance equations, equal to unity from the break date forward, zero otherwise.
Reply: I did calculations with dummies for 1 week, 2 weeks, 3 weeks, and the whole period after each event. The impact of dummies for 2 weeks was not significant. The impact of dummies for 3 weeks was significant only for Nikkei and TOPIX. The impact of dummies for the whole period after each event was significant for Nikkei and Electric appliances index, but the impact was negative, which is not consistent with the impact in the first week. That is why I refrained from using that part of calculations in this paper and focused only on the shock in the first days after each event.
Comment: The quality and resolution of the figures should be improved.
Reply: Thank you very much! To deal with the quality of the figures, I learned how to deal with figure quality settings of MS Word. In the past, I had many problems with the quality of the figures and did not know how to fix them. It seems MS Word has an automatic setting which decreases the quality of figures. Adjusting the setting, I improved the quality of all figures for the revised version of the paper.
Comment: The contributions and limitations of the paper, as well as future research directions, should be better highlighted.
Reply: I added additional information in the Conclusion section.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Report on the Manuscript ID jrfm-1436619:
External shocks and volatility overflow among the exchange rate of the yen, Nikkei, TOPIX and sectoral stock indices
I am not very satisfied with the revision of the paper. Most of the points raised in the previous report were either not addressed or addressed very superficially.
You stated that the modified and added parts are marked with yellow color. This is not always true. For example, the highlighted text “In this paper, we focus on the volatility overflow among the sectoral stock indices, among major financial indicators (JPY, Nikkei, TOPIX) and between sectoral stock indices and major financial indicators; the direct impact of external shocks on financial variables; and the possible indirect impact of external shocks on the financial variables through volatility spillover.” is identical to the text in the previous version of the paper.
Introduction
The literature review was not comprehensively addressed, as you have currently added only three new references.
In recent years, a considerable number of studies (not just for Japan) have been conducted on volatility spillovers across different assets or markets, also showing that the existence of volatility spillover is particularly important when considering the speed of market adjustment to new information. Along these lines, several articles have analyzed the effects of recent financial crises. However, these have been left out of the paper.
The motivations, aims, and contributions of the paper have not been sufficiently discussed and should be better documented in the Introduction.
You mention that the causality relationship among sectoral indices and between sectoral indices and major financial indicators, and the possible indirect impact of external shocks through volatility spillover are rarely addressed in the literature, citing only (Sultonov and Jehan 2018) and (Sultonov 2020). Are there only these two articles that address this issue?
Causality relationship
“This paper is written within the frame of a research proposal prepared in 2017. I have to highlight the causality relationship and impact of Brexit and the US presidential elections (of 2016) for sectoral indexes.” I believe this is not a sufficient reason for not performing a robustness analysis on the most recent Covid-19 period.
The sensitivity analysis for different choices of k(.) and M (see e.g. Hong (2001)) has been conducted only for (two) different kernels but not for different M.
The two-way and one-way tests to identify the simultaneous causality and the direction of causality in mean and variance have been performed only for variance.
In the empirical analysis, for the conditional mean, only AR(p) specifications are considered, resulting in coefficients that are not statistically significant in most cases. What about MA(q) or the more general ARMA(p,q) specifications? Your choice should be clearly motivated and reported in the paper.
If I understand correctly, the t-Student distribution is now used (this should be specified in the paper and in the caption of the Tables). However, the estimated degrees of freedom are not reported in the Tables.
Table 2 and Table 3 have different formatting styles for the reported coefficients. Also, there is no reference to Q2(k) in the captions of the Tables.
Overall, the results are discussed with little scientific rigor. For example, for Table 3: “The asymmetric effect of past information on the variance of all sectoral stock indices, the symmetric effect of past information on the variance of the electric appliances index and the effect of the previous periods’ variance on the variance of all sectoral stock indices are statistically significant at 1%–10% significance levels.” This is not always true, especially for the α coefficient. This applies to all the Tables reported in the paper, with overly standardized comments.
Although the Tables provide more information than the previous version of the paper, the comments in the text have not been changed, inexplicably not considering the new results obtained for each scenario.
Effect of BR and USE
In analyzing the effect of BR and USE, structural break tests have not been considered for detecting these events.
The “half-life” has not been considered for analyzing the persistence of volatility shocks. “I did calculations with dummies for 1 week, 2 weeks, 3 weeks, and the whole period after each event. The impact of dummies for 2 weeks was not significant. The impact of dummies for 3 weeks was significant only for Nikkei and TOPIX. The impact of dummies for the whole period after each event was significant for Nikkei and Electric appliances index, but the impact was negative, which is not consistent with the impact in the first week.” I think it would be helpful to leave evidence of these findings (not just for the five days before and after BR and USE) by reporting at least a general discussion in the paper.
“Owing to the flat log pseudolikelihood, the software (Stata 14.2) cannot estimate the mean and variance equations with dummy variables for the banks, information and communication and machinery indices.” Does this problem persist even when considering the impact of shocks for more than five days?
Author Response
Replies to the Respected Referee’s comments and suggestions
Thank you very much for your time and constructive comments, and suggestion. Following your comments and suggestions, we have adjusted the paper as much as we were able. Below the replies are given. The adjusted points in the paper are marked with green colour.
Comment: Most of the points raised in the previous report were either not addressed or addressed very superficially.
Reply: Following your comments and suggestions, we have adjusted the paper as much as we were able. The part which we were not able to do so far was acknowledged as a future research task in the Conclusion section of the first revision.
Comment: You stated that the modified and added parts are marked with yellow color. This is not always true. For example, the highlighted text “In this paper, we focus on the volatility overflow among the sectoral stock indices, among major financial indicators (JPY, Nikkei, TOPIX) and between sectoral stock indices and major financial indicators; the direct impact of external shocks on financial variables; and the possible indirect impact of external shocks on the financial variables through volatility spillover.” is identical to the text in the previous version of the paper.
Reply: In the first revision, for the comments and suggestions, which were partly covered by the original version of the paper, the related parts of the original paper were also marked with yellow color. To avoid misunderstanding, this time, in the second revision, we marked only the altered and added parts with green color (not related parts).
Comment: The literature review was not comprehensively addressed, as you have currently added only three new references.
Reply: As the related studies are covered by our previous research mentioned in the current paper, we decided not to mention all that literature repeatedly. In the second revision, we included additional literature about USE and BR (lines 80-83).
Comment: In recent years, a considerable number of studies (not just for Japan) have been conducted on volatility spillovers across different assets or markets, also showing that the existence of volatility spillover is particularly important when considering the speed of market adjustment to new information. Along these lines, several articles have analyzed the effects of recent financial crises. However, these have been left out of the paper.
Reply: You are right. However, we mentioned only the papers which we consider more related to our research topic.
Comment: The motivations, aims, and contributions of the paper have not been sufficiently discussed and should be better documented in the Introduction.
Reply: To make motivations clearer, we altered some parts of Abstract (lines 9, 16-18) and Introduction and added additional information in Introduction (lines 55-57).
Comment: You mention that the causality relationship among sectoral indices and between sectoral indices and major financial indicators, and the possible indirect impact of external shocks through volatility spillover are rarely addressed in the literature, citing only (Sultonov and Jehan 2018) and (Sultonov 2020). Are there only these two articles that address this issue?
Reply: Sultonov and Jehan (2018) and Sultonov (2020) are mentioned to show that this research is the continuation of them. In other words, we were suggested by many researchers, as a future research task, to cover sectoral indices and possible indirect impact in order to make the research more informative. To make it clear, we added additional sentences in the Introduction (lines 55-57) of the second revision.
Comment: “This paper is written within the frame of a research proposal prepared in 2017. I have to highlight the causality relationship and impact of Brexit and the US presidential elections (of 2016) for sectoral indexes.” I believe this is not a sufficient reason for not performing a robustness analysis on the most recent Covid-19 period.
Reply: The period assessed by this paper is February 10, 2016 to March 24, 2017. To avoid misunderstanding, we mentioned the time period in the Abstract of the second revision. Also, we mentioned in the Abstract of the second revision that for external shocks, we are going to focus on the impact of news about the results of the Brexit referendum (BR) and the United States presidential election (USE) in 2016 that might spread among the variables indirectly within a week.
Comment: The sensitivity analysis for different choices of k(.) and M (see e.g. Hong (2001)) has been conducted only for (two) different kernels but not for different M.
Reply: The estimations are done for (M=1,2….10). Only the largest value is reported. It is mentioned in the original form of the paper as “from lag 1 to lag 10” (line 234 of the second revision).
Comment: The two-way and one-way tests to identify the simultaneous causality and the direction of causality in mean and variance have been performed only for variance.
Reply: The early versions of this paper included causality in mean estimation, too. Only the causality in mean from information and communication index to banks index was detected. We were suggested to skip it. In the attached PDF form of this comment sheet, we put them, but we do not want to incorporate them in the paper.
Comment: In the empirical analysis, for the conditional mean, only AR(p) specifications are considered, resulting in coefficients that are not statistically significant in most cases. What about MA(q) or the more general ARMA(p,q) specifications? Your choice should be clearly motivated and reported in the paper.
Reply: We do not develop models. We only use the models developed by other researchers. In this paper, we followed Cheung and Ng’s (1996) test extended by Hong (2001) and Hamori (2003), as mentioned in the first revision (lines 73-74 of the second revision).
Comment: If I understand correctly, the t-Student distribution is now used (this should be specified in the paper and in the caption of the Tables). However, the estimated degrees of freedom are not reported in the Tables.
Reply: We reported the degree of freedom in the tables in the second revision and mentioned it in the caption of the tables.
Comment: Table 2 and Table 3 have different formatting styles for the reported coefficients. Also, there is no reference to Q2(k) in the captions of the Tables.
Reply: The number of lines in each row is different because of the space available. We changed and adjusted the shape of the coefficients (especially for the return coefficient; we changed a to b as it is given in the return equation). We also gave a reference about Q2 .
Comment: Overall, the results are discussed with little scientific rigor. For example, for Table 3: “The asymmetric effect of past information on the variance of all sectoral stock indices, the symmetric effect of past information on the variance of the electric appliances index and the effect of the previous periods’ variance on the variance of all sectoral stock indices are statistically significant at 1%–10% significance levels.” This is not always true, especially for the α coefficient. This applies to all the Tables reported in the paper, with overly standardized comments.
Reply: We re-checked the explanations. The comments really look “overly standardized”. The proofreading service suggested us to avoid repeated use of the phrase “statistically significant at 1, 5 or 10 percent significance level” .
Comment: Although the Tables provide more information than the previous version of the paper, the comments in the text have not been changed, inexplicably not considering the new results obtained for each scenario.
Reply: For causality in variance tables, we have mentioned about two first lines (different kernels) in Methodology. That is why the tables explanations do not repeat about the difference in kernels. About cross-correlation coefficients at lag 0 (contemporaneous correlation), we added explanations for each table (lines 238-239, 249-251, 262-263, 275-276 of the second revision).
Comment: In analyzing the effect of BR and USE, structural break tests have not been considered for detecting these events.
Reply: We tested the variables for a structural break with both unknown and known break dates. The test does not detect both events. As the test does not have important information, it is not incorporated.
Comment: The “half-life” has not been considered for analyzing the persistence of volatility shocks. “I did calculations with dummies for 1 week, 2 weeks, 3 weeks, and the whole period after each event. The impact of dummies for 2 weeks was not significant. The impact of dummies for 3 weeks was significant only for Nikkei and TOPIX. The impact of dummies for the whole period after each event was significant for Nikkei and Electric appliances index, but the impact was negative, which is not consistent with the impact in the first week.” I think it would be helpful to leave evidence of these findings (not just for the five days before and after BR and USE) by reporting at least a general discussion in the paper.
Reply: In this paper, we assess the impact of information about the results of the referendum and election in the first days after the announcement. We think that the news impact vanishes in a very short period, and the longer period impact could not be considered as the impact of information about the announcement of the results. That is why longer periods are skipped. To avoid misunderstanding, we noticed about the first week in Abstract (line 18) and Introduction (lines 85-86) of the second revision.
Comment: “Owing to the flat log pseudolikelihood, the software (Stata 14.2) cannot estimate the mean and variance equations with dummy variables for the banks, information and communication and machinery indices.” Does this problem persist even when considering the impact of shocks for more than five days?
Reply: For two weeks, software can’t estimate transportation index; for other indexes, the impact of BR and USE are not significant. As we focus only on the first days after each event (as it is mentioned in the paper many times), we use estimations only with a dummy for the first week.
Author Response File: 
Author Response.pdf
Round 3
Reviewer 2 Report
I have no further comments.
