Study on the Functional Improvement of Economic Damage Assessment for the Integrated Assessment Model

Round  1

Reviewer 1 Report

See attached file.

Comments for author File: Comments.pdf

Author Response

1．In Figures 1, 2, and 8 the blue series represents what the authors refer to as the “true damages” to China from climate change. True damages based on what data? This needs to be made clear.

Response: True damage data is based on the historical data of Statistical Yearbook of Meteorological Disasters. And the information is added in the data section. We adjust the structure of the paper, so the location of the figures has changed and it may be more reasonable.

2. Throughout the paper, the authors seem to imply that floods, droughts, and extreme temperature events within their sample period occur only as a result of climate change, which is not true. These extreme meteorological events have been occurring for millennia. The difference is that they may now be occurring at a higher frequency as a result of climate change. The authors seem to make no effort to account for the historical mean frequency of such events. But without accounting for the baseline historical frequency, and instead treating every extreme event as if it is a direct result of climate change, this would over-estimate climate damages.

Response: Thanks a lot for your kind comments. The question you mentioned has been considered. It is very reasonable to separate the loss of extreme climate events caused by climate change from the total meteorological loss. But this is very difficult to achieve. Because part of the loss in the base period is caused by climate change, and climate change is cyclical, the choice of the base period will vary greatly. And we tried to use the longest data series to do our research; however, we are short of the historical economic loss data. If the damage function adopted in the IAM, as the frequency and intensity of extreme climate change events be higher and strong, the result of the damage function in IAM would show what this trend could affect the GDP.

3.Section 3.1, lines 324-332: This is not a very credible exposition. Where does the data for A* come from? What is the estimation procedure used? I presume it is ordinary least squares (OLS). But in any case, an R-squared of 0.047 is terribly low, indicating this model has a very bad fit. And the authors do not report whether the coefficient estimates for B0, C0, or E0 are statistically significant (that is, statistically different from zero).

Response: A* is the proportion of economic loss of climate change to GDP. And the economic data is based on the historical data of Statistical Yearbook of Meteorological Disasters. The R-squared of 0.047 is the result without considering extreme weather events. If extreme climate factors are introduced, R square will be greatly improved, and the result can be seen in Table 1. We are sorry for our bad organization of the paper, we adjusted the structure and hope it would be clear. The coefficient estimates for B0, C0, or E0 are statistically significant, and we add the information in the paper.

4. Section 3.1, Equations 3-5, and lines 341-347: Here also, where is this data coming from for your dependent variables, AC, AD, and AF? How are the frequency and intensity variables measured? Related to my comment above, would it be better to use the deviation from historical baseline frequency and intensity? In Table 1, the R-squared values are much better, but the authors still do not indicate whether the coefficient estimates are statistically.

    Response: The dependent variables are the climate change factors. We use the data from the GCM models. With reference to the method of defining extreme climate indicators with the ETCCDI relative threshold, the method of extracting extreme temperature events used in this paper is to arrange the temperature sequence of each month of year in ascending order, with the value at the 95th (5th) percentile of the temperature sequence as the extreme. We stated these in the data section. And we add the explanations for these data resource in the model part.

5.significant or not. It is customary to report standard errors along with the coefficients, along with some indication of statistical significance.

Response: Thanks a lot for your kind comments. We make it more clearly in our paper.

6. Regarding Equations 3-5, the authors do not explain the estimation procedure used, so it is assumed they use simple OLS. But it is not clear to me that these three processes (floods, droughts, and temperatures) are completely independent of each other. Might it be better to estimate these three equations simultaneously using a seemingly unrelated regressions (SUR) procedure?

Response: Thanks a lot for your kind comments and nice suggestions. The method we used is the simple OLS. We think seemingly unrelated regressions (SUR) procedure would be better for our model. Since the result of OLS is also very good, we would not like to make change. Anyway, it is a nice suggestion.

7. Lines 375-376: The authors state that the correlation coefficient between the RICE model predicted damages and their measure of “true” damages is -0.0966. This fails the eye test based on Figure 8. Surely there must be some mistake. In particular, if you look at the data from 2004-2011, the correlation coefficient would be clearly positive and much closer to 1. I have a hard time believing that the divergence thereafter would cause the correlation coefficient to fall all the way to -0.0966.

Response: Thanks a lot for your kind comments. We recalculated the correlation coefficient between the RICE model predicted damages, and the result shows that it is indeed -0.0966.

8. Is Table 3 really necessary? Most researchers have a basic enough understanding of elementary statistics to know that a p-value of<0.01 is highly significant.

Response: Thanks a lot for your kind comments. We delete the table 3.

9. The Editorial Comments.

Response: Thanks a lot for your nice comments. We revised the problems one by one.

Reviewer 2 Report

This paper suggests an approach to improve the damage functions used in integrated assessment models (IAM) by linking them closer to knowledge from studies of damages of climate change. The authors use an example with observations of economic damages from extreme events in China and show that the correlation between estimated and observed estimates increase from about 10 percent when using the impact functions in the IAM-model RICE to more than 99 percent when using the approach proposed in this paper.

I think that this paper provides important contributions to the general question raised by the authors, which is how can impacts of climate change be better and more adequately represented in those IAMs that link impacts of very rough climate indicators to GDP. They point out the need to relate impacts to changes in precipitation and the challenges in representing geographical variations, which are needed to draw on studies and observations on economic damages of extreme events. I fully agree that it is necessary to find ways to do so to claim that the IAMs represent impacts of climate change that adequately reflect how climate change will affect GDP. The paper represents one possible way to get further, although there are weaknesses that should be mentioned, such as using historical data from a period with unknown climatic changes that can be related to an increase in climate forcing from human actions.

I do not agree that the proposed damage functions apply in the IAMs, however. This is because the paper does not say anything about how the observed damages affect GDP. IAM focus on GDP for several reasons, and primarily to asses the social cost of carbon, which relates directly to the balance the present cost of mitigation with the future benefits. A second reason for the attention to GDP is that there are few alternative measures that enables policy makers to evaluate the need to curb climate change, and it is up to the policy makers to agree on the necessary global actions, in the end.

In other words, it is important to have knowledge about the economic costs of climate impacts better represented in IAMs. However, but the main challenge is not to generalize cost estimates to apply in the damage functions, as suggested in this paper, but to find ways to link these cost estimates to GDP. There are several, and rather apparent, reasons to why this is challenging. Let me just mention an observation. In my country, where the population is about 0.35 percent of China’s population, there was a study on how a wind storm with the greatest economic damages ever, as reported from damages to public property (infrastructure) and private losses (insurances), from an extreme event affected GDP. There were no signs of any impact. In the next step, the impact on the regional GDP in the most affected region (one of 20 sub-regions) was checked. Again, there was no a significant impact, but there were signs of a change (did not meet a 90 percent confidence interval), which was positive. A possible explanation is that the damages gave rise to economic activities in terms of repair and rescuing, which stimulates GDP.

Therefore, if IAMs aim at focusing on the impact of GDP, which I agree with the authors is important, I do not think the problem is solved by integrating damage functions based on cost estimates, as suggested in this paper. I must add that I do not know what the cost estimates used in the paper relate to, however, because this is not explained. I think, however, that the paper provides important contributions, as a first step, to provide more relevant damage functions in the IAMs, but the authors must explain this clearly, meaning that the angling of the paper must be changed. They may mention the problem in representing damages in the IAMs they focus on in the submitted paper, but add that there is a long way to go before these estimates can be used in the damage functions in these models, and perhaps it is impossible. National aggregates of damages based on observations may be useful in many other contexts, however.

My main comment is, therefore, that the study is useful and worth publishing, but the paper has to be revised. In a revised version, I also suggest a brief overview of different IAMs. There is some confusion in the present paper, which is about the IAMs that focus on optimal climate policy, such as RICE, but refer also to IAMs underlying the IPCC-reports, where I think they refer to the IAMs used to generate the RCPs. These IAMs aim at answering completely different questions (projection of drivers), and do not include impacts of climate change. There are also other IAMs that do not fit into any of these categories – but where the cost estimates provided in this study might be useful because they aim at an economic assessment of GDP base on sectoral activities, e.g. CGE models.

Finally, there is a need to go through the language. The first sentence in the intro does not make any sense. Other changes (just from the introduction): lines 44-45, 54 (reference), 106-107; 109 (also – also).

Author Response

1. The paper represents one possible way to get further, although there are weaknesses that should be mentioned, such as using historical data from a period with unknown climatic changes that can be related to an increase in climate forcing from human actions.

Response: Thank you for your kind comments. It indeed exist the weaknesses of the data we used. But we are lack of the economic loss data. If the economic data can be found or estimated longer, we would like to make the research work further. Anyway, this is sure a weakness.

2. I do not agree that the proposed damage functions apply in the IAMs, however. This is because the paper does not say anything about how the observed damages affect GDP. IAM focus on GDP for several reasons, and primarily to asses the social cost of carbon, which relates directly to the balance the present cost of mitigation with the future benefits. A second reason for the attention to GDP is that there are few alternative measures that enables policy makers to evaluate the need to curb climate change, and it is up to the policy makers to agree on the necessary global actions, in the end. In other words, it is important to have knowledge about the economic costs of climate impacts better represented in IAMs. However, but the main challenge is not to generalize cost estimates to apply in the damage functions, as suggested in this paper, but to find ways to link these cost estimates to GDP.

Response: Thank you for your nice comments. We cannot agree with you more on the two reasons for which the IAM focus on GDP. In fact, the damage function can tell the policy makers clear that if we adopt the mitigation action earlier, how much GDP loss can be avoided, which is the benefit. However, in our paper, we do not show this. And it is sure that we will go further about our research on this way. In our plane, the damage function would be added in the GDP produce function. And the economic loss would affect the investment and consumption in the next period.   

3. There are several, and rather apparent, reasons to why this is challenging. Let me just mention an observation. In my country, where the population is about 0.35 percent of China’s population, there was a study on how a wind storm with the greatest economic damages ever, as reported from damages to public property (infrastructure) and private losses (insurances), from an extreme event affected GDP. There were no signs of any impact. In the next step, the impact on the regional GDP in the most affected region (one of 20 sub-regions) was checked. Again, there was no a significant impact, but there were signs of a change (did not meet a 90 percent confidence interval), which was positive. A possible explanation is that the damages gave rise to economic activities in terms of repair and rescuing, which stimulates GDP.

Response: Thank you for your nice and interesting comments. By understanding your idea, I conform that there would be a long way to go to make it clear that how and how much climate change affect the GDP for the complex factors and relationships. However, it still can be explained at some extent, for China, the economic loss usually comes from the damage of agriculture. Farmers have to cut down their budget of consumption for the less income and the government should cut down the investment of other projects. In total, the country’s investment for the next period will be less. And the total capital for production would be less, thus the less GDP.

4. Therefore, if IAMs aim at focusing on the impact of GDP, which I agree with the authors is important, I do not think the problem is solved by integrating damage functions based on cost estimates, as suggested in this paper. I must add that I do not know what the cost estimates used in the paper relate to, however, because this is not explained. I think, however, that the paper provides important contributions, as a first step, to provide more relevant damage functions in the IAMs, but the authors must explain this clearly, meaning that the angling of the paper must be changed. They may mention the problem in representing damages in the IAMs they focus on in the submitted paper, but add that there is a long way to go before these estimates can be used in the damage functions in these models, and perhaps it is impossible. National aggregates of damages based on observations may be useful in many other contexts, however.

Response: Thank you for your kind comments. We make a short discussion about the difficult in the conclusion part. And thanks again for your nice opinions.

5. My main comment is, therefore, that the study is useful and worth publishing, but the paper has to be revised. In a revised version, I also suggest a brief overview of different IAMs. There is some confusion in the present paper, which is about the IAMs that focus on optimal climate policy, such as RICE, but refer also to IAMs underlying the IPCC-reports, where I think they refer to the IAMs used to generate the RCPs. These IAMs aim at answering completely different questions (projection of drivers), and do not include impacts of climate change. There are also other IAMs that do not fit into any of these categories – but where the cost estimates provided in this study might be useful because they aim at an economic assessment of GDP base on sectoral activities, e.g. CGE models.

Response: Thank you for your nice suggestions. We add a brief overview of IAM in the introduction part.

6. Finally, there is a need to go through the language. The first sentence in the intro does not make any sense. Other changes (just from the introduction): lines 44-45, 54 (reference), 106-107; 109 (also – also).

Response: Thank you for your nice suggestions. We used the English edition service to go through the language, hoping the paper to be better understanding.

Reviewer 3 Report

General comments

The paper in general is interesting but does not appear to be adequately articulated.  I suggest some integration as reported in the following text and in the attached PDF file of the revised paper.

Specific comments for sections

·         In the Introduction Section the objective that the authors are pursuing in the paper is not clearly expressed. In fact, it is not clear if the function of the IAM damage is improved and how? What is the relationship between the average indicators of climate change and the extreme ones of climate change? Is it not clear if and how do the authors introduce regional impacts on climate change into the IAM damage function? Is it not clear, if global climate change is not in line with regional climate change trends, how can IAM be improved? Etc.

·         The section "2. Materials and Methods" should be reviewed by the authors, because, part of it, could be allocated in the section Methods that is absent in this paper; for example the sub section "2.2 Construction of damage function", could be inserted in the Methods section recalling in it the traditional IAM damage function that can be instrumental to highlight the novelty elements introduced by the authors in this paper; in fact the subsection "2.2.2 Improvement mechanism of IAM damage function" at the moment, should be integrated  recalling of the traditional IAM damage function.

·         It is suggested to introduce a section Methods in which to move parts of the sub-sections recalled before and in the attached PDF file, in which it is necessary to also recall a brief description of the RICE Model, which the authors simulate to compare results of their approach but which do not describe never even briefly.

·         The Results section and in particular the sub section 3.1 Construction of the multi-climate factor IAM damage function must be reorganized, as part of the Methods section and part of the results are present in it.

·         • The Conclusion Sections clarifies the objectives that the authors should have introduced in the Introduction section, but can be further improved by highlighting the advantages of the new approach proposed by authors with respect to the traditional IAM damage function.

Final comments

The paper, in general, is interesting, could have significant repercussions from the scientific point of view, but at the moment, presents some critical issues related to the poor organization of its structure. In order to be considered acceptable for publication, it must be reorganized and integrated into the various sections, some of which are absent and of all references to the various models / approaches used, which must be briefly described and not referred to a bibliographic reference, this is not acceptable from the scientific point of view.

Comments for author File: Comments.pdf

Author Response

   Thanks a lot for your kind and helpful suggestions for the paper. We revised the paper according to your nice comments one by one.

General comments

  •In the Introduction Section the objective that the authors are pursuing in the paper is not clearly expressed. In fact, it is not clear if the function of the IAM damage is improved and how? What is the relationship between the average indicators of climate change and the extreme ones of climate change? Is it not clear if and how do the authors introduce regional impacts on climate change into the IAM damage function? Is it not clear, if global climate change is not in line with regional climate change trends, how can IAM be improved? Etc.

      Response: Thank you very much for your suggestions and comments on the introduction, which is very helpful to the logical combing of this article. According to the questions mentioned above, we have reorganized the introduction and made some supplementary explanations about the lack of explanation of the logic of the paper.  The detail of the correction can be seen in the paper.  

  •The section "2. Materials and Methods" should be reviewed by the authors, because, part of it, could be allocated in the section Methods that is absent in this paper; for example the sub section "2.2 Construction of damage function", could be inserted in the Methods section recalling in it the traditional IAM damage function that can be instrumental to highlight the novelty elements introduced by the authors in this paper; in fact the subsection "2.2.2 Improvement mechanism of IAM damage function" at the moment, should be integrated  recalling of the traditional IAM damage function. It is suggested to introduce a section Methods in which to move parts of the sub-sections recalled before and in the attached PDF file, in which it is necessary to also recall a brief description of the RICE Model, which the authors simulate to compare results of their approach but which do not describe never even briefly. The Results section and in particular the sub section 3.1 Construction of the multi-climate factor IAM damage function must be reorganized, as part of the Methods section and part of the results are present in it.

Response: Thank you very much for your kind and nice comments and suggestions. We have made major adjustments in this section by adding the method section and incorporating the model of Part 3 into the method section. A description of the methods in the literature is added. The review of traditional IAM damage function is supplemented and illustrated with our innovative work. The description of RICE model is added. And the details can be seen in the paper.

  • The Conclusion Sections clarifies the objectives that the authors should have introduced in the Introduction section, but can be further improved by highlighting the advantages of the new approach proposed by authors with respect to the traditional IAM damage function.

Response: Thank you very much for the comments of the judges. In the conclusion section, we have adopted your suggestions and made corresponding supplements and modifications.

Round  2

Reviewer 1 Report

I do not find the authors' empirical methodology to be very convincing. However, the main thing that must be corrected before I would sign off on this paper is the following.

In my initial report I stated the following:

Throughout the paper, the authors seem to imply that floods, droughts, and extreme temperature events within their sample period occur only as a result of climate change, which is not true. These extreme meteorological events have been occurring for millennia. The difference is that they may now be occurring at a higher frequency as a result of climate change. The authors seem to make no effort to account for the historical mean frequency of such events. But without accounting for the baseline historical frequency, and instead treating every extreme event as if it is a direct result of climate change, this would over-estimate climate damages."

The authors responded that there is simply no way for them to do this. In that case, they are NOT measuring "the direct economic losses caused by climate change" as they claim throughout the paper. Rather, they are measuring the direct economic losses caused by meteorological events, only some of which are a result of climate change. Therefore, the authors must adjust their language throughout the paper to make this clear. 

To the extent that accurately measuring the economic losses from extreme meteorological events is important for measuring the impacts of climate change, the exercise is still relevant for IAM analysis. But the authors should not claim to be measuring something they are not actually measuring.

Author Response

General Comments:

Throughout the paper, the authors seem to imply that floods, droughts, and extreme temperature events within their sample period occur only as a result of climate change, which is not true. These extreme meteorological events have been occurring for millennia. The difference is that they may now be occurring at a higher frequency as a result of climate change. The authors seem to make no effort to account for the historical mean frequency of such events. But without accounting for the baseline historical frequency, and instead treating every extreme event as if it is a direct result of climate change, this would over-estimate climate damages."The authors responded that there is simply no way for them to do this. In that case, they are NOT measuring "the direct economic losses caused by climate change" as they claim throughout the paper. Rather, they are measuring the direct economic losses caused by meteorological events, only some of which are a result of climate change. Therefore, the authors must adjust their language throughout the paper to make this clear.  

Response: Thank you for your comments. We adopt your suggestion, and make the expression clearer. First, we change our title from “Study on the Functional Improvement of Economic Damage Caused by climate change for the Integrated Assessment Model” to “Study on the Functional Improvement of Economic Damage Assessment for the Integrated Assessment Model”. Second, we make explanation in the introduction about the relationship between measuring the direct economic losses caused by meteorological events and economic losses caused by climate change. Finally, we restated our work from “economic losses caused by climate change” to “economic losses caused by meteorological events” in all figures and result explanation.

We think that maybe the most important thing is to calculate the difference of economic loss with the different mitigation targets, especially for IAMs. Thus from this point view, our work may make some kind of contributions.

Anyway, thank you for your nice comments.Reviewer 2 Report

The paper has been improved, and the contributions of the approach in this paper are more adequately described in the context of all the challenges in estimating damage functions. The (new) description of IAMs is a bit misleading, however. They distinguish between three categories, which is OK, but category 2 (DICE, RICE, FUND etc.) is the only one that integrates impacts of climate change, which is what is relevance here. This is not mentioned, which makes the new paragraph a bit confusing. A small thing perhaps: PAGE was used in Stern Reveiw, but the model is developed by Chris Hope. 

Author Response

General Comments:

The (new) description of IAMs is a bit misleading, however. They distinguish between three categories, which is OK, but category 2 (DICE, RICE, FUND etc.) is the only one that integrates impacts of climate change, which is what is relevance here. This is not mentioned, which makes the new paragraph a bit confusing. A small thing perhaps: PAGE was used in Stern Reveiw, but the model is developed by Chris Hope. 

Response: Thank you very much for your kind comments and suggestions. It is sure very important to point out that the category 2 (DICE, RICE, FUND etc.) is the only one that integrates impacts of climate change. And we add this information in our paper, which can be seen in the line 58-60. And for the PAGE model, we correct our misunderstanding expressions. Thanks again for your nice remind.

Reviewer 3 Report

The authors reviewed the paper based on the suggested suggestions and now the level of the paper has improved considerably.

I recommend only a general reading of the paper and a general check of the English language.

Author Response

General Comments:

I recommend only a general reading of the paper and a general check of the English language.

Response: Thank you very much for your kind suggestions. We have checked the English language though out the paper and made some corrections.