Dear Mr. FathiAzar,

Many thanks for your comments. The authors do agree with you in that only enforcing stringent regulations would not improve the market and insurance penetration and other incentives such as private-public cooperations (including Iran Catastrophe Insurance Pool) need to be implemented. We will add more contents in the conclusion section and elaborate more on solutions in this regard as far as possible.

Dear Dr Behnamfar,

Many thanks for your detailed comments. As mentioned in your Comment 2, the first audience of this sponsored resaerch are government decisionmakers and then insurance firms. Luckily, the results of the resaerch have already used by Central Insurance of Iran in designing laws related to Iran Public catastrophe Pool and also shared with and well received by public and private insurance companies. This paper tends to share the results with technical audience and other resaerchers. Table 1 rates are based on current market rates in Iran and extracted from aggregator insurance websites. We agree that these rates are an underestimation of actual risk. Earthquake Model of Middle East (EMME) has been chosen for several reasons such as participation of regional scientist in its preparation, its global credibilty, use of more recent seismological data, and its compatibility to GEM's OpenQuake platform. This is the same model used by Kohrangi et al (2021) as well. Regarding Figure 1, we think values are reasonable and comparable to GEM 2023 global hazard map and some local studies we checked. We agree to add more informatin to this section (seismic sources, GMPEs used, logic trees and soil model) will help the reader to better understand the process. We agree that phrasing in 259 might be misleading. We will rephrase that. Sharp contrast between hazard levels (Figure 1) is caused by the method of interpolation. In the risk assessment, the value of hazard is calculated in the centroid of exposure grids, so the impact of interpolation error will be minimised. The statistics we used is based on 2016 census which indicate that about half of built area in Isfahan province is either masonry and adobe. Kohrangi et al (2021) has reported 44% which is very close to our exposure split. Our AAL values for the COUNTY of Isfahan is less than what Kohrangi et al (2021) calculate for the CITY of Isfahan. Since we almost share the same hazard and exposure models, this might probably stem from different vulnerabiliy curves used in two studies. We agree to add more explanation at the end of numerical results for more clarity. Again, we would like to thank you for your constructive comments.

Dear reviewer,

Many thanks for your time and constructive comments. Please find our answers below in italic:

- I suggest the authors discuss why these five provisional capitals were chosen. An additional discussion on the diversity of the zones and construction types found in this city in section 3 would benefit the readership. A paragraph was added at the end of section 3.1.1. to explain why we selected these cities. We also deleted a line that might confuse the reader regarding the construction types at this stage of the paper. We added more information in the 4.2 section where we use a portfolio of dwelling sin these cities to portraits the difference between two approaches.

- I suggest the authors briefly overview all of the data sources before section 3.1. In this segment, it's not clear to the reader where the data to construct the portfolio of 1,500 residential dwellings was obtained. The data sources we used for seismic hazard modelling are introduced in hazard, vulnerability, and exposure section separately to be more relevant to the context.

- Figures 1-3 and maps elsewhere, I suggest removing the abbreviated names of the neighbouring countries or using standard ISO-alpha-3 country codes, e.g., SAU would be KSA and ARE would be UAE. We double checked the 3-digit ISO name of neighbouring countries and they were actually correct (Saudi Arabia: SAU and United Arab Emirates: ARE).

- I suggest the authors add some discussions on the limitations of their study and future research directions in the conclusion section. A paragraph added to the end of the conclusion section to discuss the role of data availability in enhancement of the results.

The authors would like to thanks the reviewer for his detailed and useful comments provided. More information on the criteria for selecting a test portfolio will be added to Chapter 3 to remove potential ambiguities. The rationale behind selecting such portfolio is to include possible variabilities both in terms of seismotectonic nature of cities in Iran and different types of building for which premium ratios exist both in the local market (premiums calculated based on aggregator sites) and our seismic cat model. The replacement cost of buildings is the average cost of reconstruction per unit area for a conventional engineered (complying with local construction codes) which was estimated about USD 300 per sqm at the time of writing the paper. We will add a new chapter for discussing the results as well as relevant topics as the effect of cat pools, parametric insurance and public/private collaboration in post-disaster construction. Regarding the GDP, we are not sure if we understood the point raised. To us, solvency capital should be calculated based on risk perception of insurance firms and 1-in-200 year probability could be valid for the Iranian market as well.

Dear reviewer,



Many thanks for reviewing our paper and providing constructive comments. Please find below answers to your comments in bold fonts:

1. Line 11: It is suggested to write "fire insurance policy" instead of "fire policy": Thanks, we will use your suggested term for better clarity.
2. Line 18: It is mostly "risk management practice" rather than "...techniques).: Will be replaced.
3. A brief summary of the quantitative findings is suggested in the ABSTRACT.> We will add more results to the abstract.
4. Line 30: It also hit big cities like Sarpol Zahab but you may say "if the epicenter was within or very close to big cities.....": The population of Sar-e Pol-e Zahab city at the time of 2017 earthquake was 46,000 which is equivalent to small cities.
5. Line 111: Please explain briefly about "life" and "non-life" insurance.: An explanation will be added.
6. Starting from line 119: Needs more explanation on how Solvency II has been carried out. The paragraph describes the first Pillar...how about the other two pillars?: A brief description will be added on other two pillars However, the topic of the paper is more related to Solvency Capital Requirement (SCR) which is mainly Pillar One.
7. Line 152: Not clear what is meant by "motor".: Motor is the line of business for automobiles.
8. Line 168: 1500 residential dwellings. Please indicate that these are "Housing Units" and not buildings.: Clarification will be added.
9. Please justify better why only 1500 housing units (HU) have been taken into account (can this represent the solvency as intended... please justify better). How the dataset has been distributed for each big city and what is the overall share for each city from 1500 HU. Why your calculation does not consider the complete building pool since you are utilizing Openquake? Because it seems more useful to use the complete building pool to understand the national shortcomings and to devise a better policy and to assign more proper premiums for the selected policy. Or this may be a scope for your future research... please mention it in such case. If there is as purpose for such choice please clarify (for example: it may be just for a sake of comparing two solvency schemes just in a relative sense)> 1500 dwelling are selected because this is a reasonable amount of risk assumed by a medium insurance company in Iran. Our intention would be to examine how an average insurance company in the Iranian market could accumulate loss over time event by providing earthquake coverage for a small portfolio of buildings. The risk ratio has been assumed to be constant withing the cities (both because of the resolution of our model -Shahrestan and risk ratios extracted from the market. That is why spatial variability does not play a role here.
10. Table 1 & Table 2 and text: premium rate... what is the unit? Rate (ratio of two USD values) is dimensionless.
11. Line 217: instead of "national level" please indicate "country level" conforming with your formula....and spell-check "country" in Eq 5: We will harmonise that throughout the paper.
12. Line 290: Is this 55% housing units or buildings... it seems housing units is correct... please also recheck the paper for this issue.: That is true. But in large numbers the ratio would be valid for buildings as well. We will modify this.
13. Line 301: It is Fig. 4> We will correct this.
14. 4: please indicate the reference in the caption.: We will add the reference to caption.
15. Line 314: incomplete sentence.: We will correct this.
16. Line 323: change to: Figure 1 and Figure 2: we will correct this.
17. There is a mixture of "one-in 200" (text), "1-in-100" (Figure 6), "1-in-200' (text and Figures... it seems "1-in-200 years" is correct. Please check and correct all the text and Figures (caption and above legends).: We will harmonise that to avoid misunderstanding.
18. Figure 6 – to cite in the text.> Will be added.
19. Line361: Is this 300 USD per m² (square meters): Yes, it is. It is based on the USD equivalent of the average quality "construction" in Iran at the time of conducting the research.
20. Table 3 and Line 387: Please indicate number of housing units involved... otherwise it seems too low for big cities .: Will be added.
21. Line 400: 1-in200 years> Will be modified.

Please recheck referencing and references... Line 297, "Mansouri and...[38]" please use proper referencing scheme... it is actually GEM-EMME WP4 project accomplished in IIEES in 2013 (relevant to section 3.2.3 first paragraph). "Mansouri, B., and Amini-Hosseini, K., 2013, “Global Earthquake Risk Model (GEM) - Earthquake Model of the Middle East Region (EMME) - WP4: Seismic Risk Assessment”, Final Report, 2013"> the reference will be corrected.

Dear reviewer,

We would like to thank you for your precious comments to which we answered in bold font in the below:

In table A1 is shown the level of risk associated with different building typologies in Iran, according to the study of Ghafory-Ashtiany M. (1991). However, there is no description of hazard scale used. May be useful to know the correspondence between such hazard levels and intensity measure (e.g., PGA ranges) used to define it.: These are seismic zone codes based on Standard 2800 (design of buildings to seismic loads in Iran). More description will be added for better clarity.

Table 1 shows averaged earthquake insurance premiums for different building typologies in five provincial capital cities of Iran. To allow an easier understanding of the table contents, the values reported should be further commented. For instance, the cities of Tehran, Tabriz, Kerman presents a premium rate for masonry buildings equal to 1.1, while the cities of Esfahan and Ahvaz a premium rate of 0.78. Is such difference in the premium rate due only to the different hazard level? Is it due also to the different construction features of masonry buildings in the area (e.g., Adobe and Traditional or Confined Masonry, as reported in table A1)? Also, if the difference is only due to the different hazard level, why are the premium rates for masonry buildings the same in the city of Kerman and Tabriz while the rates for other typologies are different (e.g., 0.50 in Tabriz and 0.37 in Kerman for steel buildings)? Please, provide additional comments on it.: The premium ratio values reported in this table are those used currently by the insurance market in Iran. Although they are based on the 1991 study of Ghafory-Ashtiany but they have evolved over time by the market. The authors are aware that some of the values might not sound reasonable, but they preferred to keep them as they are as the current practice in the market.

Event-based stochastic modelling is adopted in this study to quantify seismic risk. Despite this study focuses on the comparison of solvency capital calculation methods, a brief description of the risk assessment procedure adopted in the study could be useful. Please, consider briefly describing how  hazard, exposure and vulnerability are incorporated in the process to generate event loss tables and how OpenQuake platform performs seismic risk calculation. At least, references to documents reporting such descriptions should be provided: We tried to put more weight on the main topic of the paper Solvency, as you mentioned. There are many details involved in the modelling of risk that might overshadow the main topic. However, more information on the damage and financial loss calculation will be added for better clarity in sub-chapter 3.2 (Modelling the earthquake risk in Iran).

In section 3.2.2 the exposure modelling is described. As no information on building’s construction year is provided in 2016 census data, all dwellings built between 2011 and 2016 are assumed constructed with modern material such as steel and RC. Are such dwellings equally divided into the two building typologies (RC and steel)? Such distribution should be specified as the two typologies could have different seismic performance (e.g., as shown in figure 4).: That is true. Because the 2016 census data lacked the year of construction we used the prevailing distribution construction years of 2011 census and made assumptions to complete the "construction year" field. As you mentioned, we assumed that any number of dwellings reduced in each county for each construction type in 2016 belongs to the oldest year built group and whatever added belongs to either RC and Steel (because the number of Masonry dwellings was reduced in almost all counties between 2011 and 2016). As we new the construction type of the post-2011 buildings we distribute them to the corresponding type. It means we did not divide them equally between RC and Steel types.

In section 3.2.3 a better description of vulnerability classes is needed. Nine vulnerability classes are identified by the adopted vulnerability model (Mansouri and Amini-Hosseini). Two classes for masonry buildings are defined. How do these two classes differ? For instance, do they differ in terms of number of storyes? Do they differ in terms of quality construction? Moreover, how the model characterize the quality construction? In other words, what is the meaning of “medium-quality construction” according to the model? Are there also other vulnerability classes for buildings characterized by low-quality and high-quality construction? A better description is needed. Moreover, in line 279 the authors claim that the buildings vintage is used as proxy for the quality of construction. Please, provide an example of how the building’s age is used as proxy for the quality of construction.: The only difference between two classes of masonry buildings is the period of construction (before 1976 and after 1976 according the paper) which is a proxy of construction quality. The height of the class is assumed to be low-rise in both cases. More information on the adoption of construction period as a proxy of quality will be added to the section.

In line 281, it is stated that an auxiliary population dataset with a 30-arc-second resolution is used to disaggregate the county-level building exposure data. First, a brief description of the downscaling procedure of exposure data adopted for such disaggregation should be provided. Also, please add a comment for justify why a finer resolution for exposure modelling is needed for losses calculation.: The requested information will be added to the section including the method of downscaling and why it is necessary (due to the variation of seismic hazard across the county).

The authors should provide a definition of “country” level adopted in this study. Indeed, in figures 3,5 and 6 the “country border” seems representing the national border, while in Tables 1 and 2 the country level seems to be smaller than the province level but still different to the city level. A precise definition of the scale is important also to understand input data used (e.g., exposure data provided by census).: That could be a spelling issue for "county" and "country". We will double check that.

In figure 3 the residential building value is reported. How is it calculated? Which database is used to derive such value? Is the value adopted differ only based on buildings construction material or other parameters (such as quality of construction) are considered for its evaluation? Also, is this value assumed constants in the entire country, regardless the building location (e.g., province, city)? Please, provide additional information about residential building values adopted. Furthermore, to be consistent with comment in lines 285 – 292, maps in figure 3 might be also shown in terms of number of buildings instead of in terms of exposure value.: Three class of age (according to the period of construction) have been considered for major types of construction (RC, Steel and Masonry) and the value of construction for each time at the present time has been depreciated to estimate the value of dwelling for each type and each age class.

Vulnerability curves adopted for losses calculation are described in section 3.2.3. However, it is not clear the translation of physical damage into monetary losses. In other words, once the damage ratio is given, how are economic losses calculated? Is it function of the replacement cost for the building? Is the building surface also considered for losses calculation? Even if the value of replacement costs is presented in line 360, I would be better to introduce it before showing maps with expected losses (Figure 5 and Figure 6).: Each pixel of the study has an ELT calculated by multiplication of damage ratios of each event for each of the building classes by number of dwellings in each construction class and average built area of each dwelling and its corresponding value. The economic loss of each pixel is calculated by dividing the sum of all events losses by number of years simulated.



Economic losses shown in Figure 5 may do not allow an exhaustive comprehension of seismic risk in Iran. In other words, in location where AAL is high it is not easy to understand if it is high due to the exposure (i.e., the presence of many buildings exposed to earthquakes) or due to the high seismic hazard as well as to the high vulnerability of residential buildings in the area. Please consider adding a figure showing the value of losses/m2. It could be also useful to confirm comment reported in lines 317 – 330.: This is true. That is why we use exposure maps, degree of vulnerability of different building types, and seismicity of each region to interpret the resulted risk. The normalisation of losses by exposure will result in AAL ratio. A table of AAL ratios will be added in the appendix.

The assumptions made for the application presented in section 4.2 could be oversimplified. Despite the main aim of this study is to compared solvency capitals calculated with different approaches, the assumption that 100 buildings are covered by earthquake policies in each of the selected cities in the country, regardless their residential population, may lead solvency capital values (shown in table 3) too unrepresentative of real cases. In fact, 100 buildings could correspond to the 100% of residential buildings in a city and to the 1% of residential buildings in another city, depending on how populated they are. Thus, it would be more appropriate to define a fixed percentages of buildings covered by earthquake policies in each city and estimate the number of buildings covered based on the total number of residential buildings in the city. Moreover, differences in the diffusion of a given typologies in each area of the country should be considered. Instead of assuming the same percentage of masonry, RC and steel buildings in each city, it would be more appropriate to derive the percentage of occurrence of such typologies in each city from the exposure model (Figure 3) and to adopt such percentages for a better exposure/vulnerability characterization at city level. Therefore, the authors may consider adopting more appropriate assumption for that application.>The selection of a rather small portfolio of 1500 buildings is reasonable for a medium size insurance company in Iran. Here, we did not intend to portray different levels of risk, since we already did that by creating the risk maps. The main objective of this example is to consider different seismicity zones in Iran in the calculation. As seen, results show that event a small portfolio of flats can impose significant amount of risk to a medium size firm in Iran.

As this study may be hard to understand for those who are not experts in the field of earthquake insurance, please consider the following suggestions:

 In line 196 CRESTA zones are introduced. Please, provide a brief description of the CRESTA zones.> Will be added.

Likewise, the Weighted Total Value Insured (WTIV) the Total Insured Value (TIV) are mentioned in in lines 195 and 196. Please consider providing their definition and how they are derived.: More information will be added as far as possible.

In line 235 the event loss table (ELT) is introduced. What is the information provided in the ELT? Please, provide a briefly description on its contents.More information will be added as far as possible.

Additionally, it is recommended to implement the following modifications (technical corrections):

The acronym “VaR” is presented in line 199. However, it is already used before (e.g., line 157). Please, add the specification for the acronym at its first mention.

In line 317 replace “figure 4” with “figure 5”. Likewise, replace “figure 5” with “figure 6” in line 331. Please, check the numbering of all figures.

The description of the figure in line 331 (one-in-200-year losses) is not in line with the figure caption 6 caption (Earthquake 1-in-100 loss). Please, modify it. Will be modified.

Please, correct the following typing errors:

Line 102: replace “Christchurch quakes” with “Christchurch earthquake”. We meant "two" earthquakes happened in 2011 in NZ.

Line 112: use the square brackets as in the line 107.Will be modified.

Line 297: modify the reference “Mansouri and Amini-Hosseini [38]” using the proper reference scheme.Will be modified.