Insurer Resilience in an Era of Climate Change and Extreme Weather: An Econometric Analysis

Round 1

Reviewer 1 Report

The authors made some serious effort in considering the reviewer’s comments and improving the readability of the paper. Although I still think that the paper would greatly benefit from using data on a more granular level as proposed in my previous report, I understand the authors' reasoning to follow a different path. To sum up, I think that the revised version of the paper is acceptable for publication in Climate.

Author Response

We thank Reviewer #1 for accepting our answer. Certainly the analysis he suggested can find space in the future for further work and another manuscript.

Reviewer 2 Report

Thank you for taking into account my first comments, the improvments in the core paper help clearly to understand the global methodology.

I still have two comments :

In the introduction, I do not necessarly agree that outcome risk is harder to evaluate than hazard risk. The probability of occurence of hurricanes in the current and future climates is a tricky question.

I also do not clearly understand why the dichotomy between event risk and outcome risk helps to explore the link between anthropogenic climate change and extreme weather.

Author Response

We thank the reviewer for further comments on our manuscript. Considering these comments we revised the text as follows (text highlighted in blue in the text). 

As for comment (1) the opinion of the reviewer may be correct in some situations but not necessarily considering the generality of cases since the prediction of any outcome is linked to the prediction of hazards. So, as the reviewer said, predicting hurricane is tricky, then predicting any hurricane related loss is tricky too. The reviewer is correct particularly when outcome prediction is weakly related to hazard or when a purely statistical model is used on outcome-only data (e.g. using machine learning models). In other words, because of the choice of the model that may not need hazard data, then the model may predict more easily outcomes. 

 

As for comment (2), the dichotomy between hazard risk and loss risk (causally related of course considering the mechanisms) can serve to investigate the uncertainty between their linkage as well as the uncertainty related to how climate change is affecting extreme events. Extreme events, as any other natural processes, are governed by their own stochastic dynamics that may be affected by external factors like climate change. CC can alter frequency and magnitude of hurricanes and losses but the question is: is the altered frequency and intensity of losses truly associated to climate change via altered hurricanes or it is related to other factors (with immutable hurricane dynamics)? Other factors in this space can be for instance building practices and/or other financial policies leading to losses somehow. 
The linkage between losses and loss-drivers is complicated in its own causality by climate change that, hypothetically, may affect one, or none, and the latter one (losses) are dependent on other features (such as the ones we described).  

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

see attached file

Comments for author File: Comments.pdf

Reviewer 2 Report

[A brief summary]

This paper represents unique results for one of the major issues in the insurance market that is troubling us recently. The authors shows that there are no significant correlation between ROE of insurance business in US and hurricane activities based on the insurance statistics about hurricane losses in US. In addition, they depict the insurance market have high resilience against hurricane losses based on historical data, statistics and their multiple-linear regression model. 

The authors well understand the point of view of (re)insurance companies and issues that should be addressed regarding to artificial climate change and extreme weather, and their discussion includes some insights about how to deal with these issues. However, I feel there are some problems about their logic, methodology and the structure of this paper as follows. I cannot recommend the acceptance of the present paper in this form for publication in the journal.

[Broad Comments]

The important explanations of data and methodologies are written in Appendix. These should be written in main part of this paper. I recommend you to reconstruct your manuscripts following common structure of science paper (introduction, data and methodology, result, discussion, conclusions) to make your paper more understandable.

 

The authors discuss the impact for insurance market by using return of equity (ROE). However, although, I am not an expert in the finance and acconting of insurance company, I feel ROE is inadequate for the purpose to reveal the impact of hurricane activity to insurance business. You should explain the definition of ROE, because insurance company often use adjusted ROE whose equity means sum of equity and policyholder surplus. On the basis of adjusted ROE, hurricane losses affect the both of numerator and denominator of ROE. When huge hurricane loss event occurs, equity must decrease because policyholder surplus would be withdrawn. Profit will be stay or decreasing because the withdrawn monetary value is recorded as income. In addition to this, profits of insurance company are not only came from hurricane related insurance such as fire insurance. If you want to prove no correlation between hurricane activity and finance of insurance business, I feel you should be more careful to explain the relationship by using basic financial metrics and breakdown of financial statements.

The authors should describe significant levels or p-values and method of statistical test when you discuss statistical significance of correlation. In the appendix B, the authors discuss the correlation between some index and ROE based on  statistical values, but they didn't mention about significant level. In addition to this, the Table 2 which may include the statistical values are not attached in the document (Table 1 also cannot be found).

I recommend you to refer to natural catastrophe (NatCAT) model (e.g. https://onlinelibrary.wiley.com/doi/book/10.1002/9781118906057). (Re)insurance companies have took a large effort to evaluate natural catastrophe risks by developing and using NatCAT model. NatCAT model have been broadly used in the practical insurance business which includes risk management purpose evaluating uncertainty of natural catastrophes. I believe it will be good for readers that the limitations of NatCAT model from the viewpoint of artificial climate change are shown in the paper.

I also recommend you to refer to the activity of Task Force on Climate-related Financial Disclosures (TCFD). A lot of (re)insurance companies declared the support for the TCFD Recommendations and they will start to disclose the expected potential impact of climate change to their finance following the guideline of TCFD. I believe it deeply relates with your study.

[Specific comments]

P.13-14, line 256-262: If you discuss time series variation of hurricane losses, you should mention about the relationship between hurricane activity and Atlantic Multi-decadal Oscilation (AMO). A large number of papers demonstrate that hurricane activity is linked to the AMO (e.g. Goldenberg et al. 2001).

Goldenberg et al.(2001). The Recent Increase in Atlantic Hurricane Activity: Causes and Implications, Science, 293, 474-479.

P.20, line 937 and P.25 line 512:  Original names of some acronyms such as OLS and ROI are not shown in the document.

Fig.1-3: Data sources are not shown clearly. You should show citations or URL. If these are not open data, how to get the data should be explained.

Fig.3: The data period is different from the title of this figure. The figure seems to include 1950-2009.

Reviewer 3 Report

As a cat modeler in a reinsurance company, i found this paper really innovative and focusing on a practical point of view. Maybe the appendixes are a bit too long when comparing to the paper itself. 

Page 4 line 69 : a few exemples could be interresting after the sentence "perhaps in ways that are not yet well understood"

Page 6 line 110 : a part of the sentence is missing

Page 9 lines 162 to 172 : this paragraph is not very clear and could be improved

I suggest a last global reading of the paper because some of the sentences have a few words missing.

But despite these remarks, this paper is worth publishing in Climate