The Effect of IPCC Reports and Regulatory Announcements on the Stock Market

Round 1

Reviewer 1 Report

Brief Summary:

The paper seeks to understand how U.S. public companies react to IPCC reports on the effects of climate change, and whether those reactions differ from their reactions to climate related regulatory actions. They focus on a limited sample of firms within the S&P 500 index that are denoted as being in carbon intensive industries and posit that firms with such designations should have stronger reactions than peers from less intensive industries. The authors argue that climate risk is likely to be underpriced in the market, and if it is underpriced then those firms that contribute most to climate change are likely to have the most severe underpricing of risk. The results show some abnormal reactions to IPCC reports but they do not seem to be correlated with the carbon intensiveness of the firm, nor does there appear to be any long term abnormal return related to either regulatory changes or IPCC reports.

 

I would start by saying that while I am sympathetic to this line of research in general – in fact, I think there is probably a lack of literature surrounding this issue – I am not sure that the current paper does much to address that lack and add to the literature. I will enumerate my biggest concerns below and then follow with more minor issues.

 

Major Comments:

1. My biggest concern is with the sample selection. Why only use firms in the S&P 500? This severely limits both the validity of the statistical testing (because of sample size) and the value of the results (the S&P 500 index is not necessarily generalizable to the industry as a whole, and more importantly the companies that are both in the S&P and in the “industries” measured are likely not representative of the industries of interest).
   1. The sample size issue is exacerbated by the authors attempt to limit survival bias and only keep companies in the sample that meet the sample criterion (includes being in the S&P among other things) for the entire sample period (nearly 30 years). This is only 268 firms in total.
      1. Why not keep all S&P firms at least and use the surviving sample as a robustness test? That will give you significantly more data. Although this is certainly not the ideal option.
   2. The value issue is potentially more hurtful. Why not use the entire market as in Ramiah et al 2013?
      1. If it is because not every firm will relate their GHG related emissions, that does not seem to be a sufficient argument given that the authors don’t actually use the individual firms own GHG scores but rather match firms to a sector based measure provided by a third party.
      2. This could also be done for the whole market as these super sectors match up fairly closely to 2 or 3 digit SIC codes. (this would also solve the power issue above). I would argue that carbon intensive industries are relatively identifiable without a third party verification, and that SIC codes would be sufficient to identify this cross-section.

• If it is a data access issue then I am sympathetic, but unfortunately that doesn’t alleviate the issues raised above, even though it makes the solution intractable.

3. Additionally it is not clear to me just how big the sample is for each test. In section 3.2 the authors state that there are (154 firms with 17 events, 120 firms with 15 events, and 2 firms with 12 events). Are these all included in the same sample? Or are you testing them as sub samples? It is not clear from the writing exactly what constitutes the sample at each point in time.
4. Finally both a and b above exacerbate the issue of generalizability, and here I will focus specifically on the issue of “industry”. Looking at Table 1 we see that 6 of the 19 industry portfolios contain <= 5 firms and 9 have <10. This is problematic because the paper seeks to draw conclusions about the industry reaction and uses industry abnormal return as its central focus. The described industries are hardly small – for instance Telecommunications has only 3 firms in the portfolio – and I find it rather difficult to accept results from such a limited portfolio can be applied the industry as a whole.

 

2. I am also concerned with the event choice, but mostly with why the authors keep changing the events that they use in successive tests. When looking at the IPCC reports the authors go back and forth between using 17 events (report announcements and sub-events that are news related) and 5 events (just the 5 report announcements).
   1. First it’s not clear to me why all of these events and sub-events are useful or necessary, especially because they are likely to be strongly correlated with each other in terms of the information contained in the report.
      1. For instance, if climate risk is incorrectly priced in equity and the IPCC report provides new information sufficient to correct this mispricing then it should be the case that the first information about the report provides such information and the mispricing is corrected at that point.
      2. Alternatively, if the first information or any of the information that precedes the full report does not contain the full and necessary information to correct the mispricing then we would only expect to see a correction at the point of time when the full report is released.

• The current methodology would only seem appropriate if the information in each sub-event is independent of the last and we should therefore expect markets to react to each subsequent information event as well as the previous ones. If this is the case then unless the mispricing is very large the corrections would not likely be significant for multiple events, and perhaps some sort of cumulative effect is more appropriate (that is quasi-cumulative over all the sub events).

1. Presuming that they are all necessary, why are they not all utilized in all the tests presented? Even in the CAR analysis which requires the sub events to be at least 10 days apart (to satisfy the -10,+10 window), it would appear from the appendix that all the sub-events are at least that far apart.

 

 

3. I have a number of issues with the presentation of the tables and the discussion of the results which cumulatively present a major issue and I will thus group these issues together. However, a good number of them will be addressed by following more closely to the table design in Ramiah et al 2013 which is a very similar study and in which the tables and analysis are much more clearly presented.
   1. In tables 2, 3, 4, and 6 the authors only report Positive and Negative abnormal results but no column for no abnormal result, despite no result being the most common outcome.
      1. As a corollary it is misleading and somewhat pointless to report the proportion of negative abnormal returns only in reference to total abnormal returns and not total returns. It conveys a reaction to these events that is much greater than what the results actually suggest.
   2. It would be very helpful to have table description to help the reader understand exactly what is being presented, as it was not always clear to me.
      1. For instance in Table 5 it is not clear to me what is being presented. For IPCC report 1 are the 4 food and beverage CARS’s simply the industry CAR in different but overlapping windows? How is this independent information? This cannot be reported as 4 CAR’s surrounding the first event. They are the same thing.
         1. Additionally this really strikes me as data mining, given that the CAR’s for different industries don’t cover the same window. Essentially this means that for none of the other overlapping window is there an abnormal return. That implies that what the authors have captured is simply noise.
      2. Rather than telling the reader how many abnormal returns each industry had (like Table 1 etc.), I would rather see a table that tells me which industries had abnormal returns after which event (like Table 3 from Ramiah et al 2013). That is much more informative about what is actually going on.

 

 

4. I have a few concerns about the hypotheses as well, but this may just be an issue of writing and understanding.
   1. These are all directional hypotheses and predict that returns will be different in a specific way rather than just that they will be different. Are the authors thus using 1-sided significance tests? If so this needs to be stated as the barrier for statistical significance is lower in one sided tests than for two sided tests.
   2. I would like to see the null hypothesis written out, especially because these are directional predictions. For instance, in the analysis Hypothesis 1 is rejected, but rejected in favor of what? The null or the opposing directional prediction?
      1. As a corollary, I am not sure that the evidence is sufficient to reject Hypothesis 1. Especially given that there are no significant CAR’s for the regulatory announcements, but there are significant CAR’s for the IPCC announcements. This would (to me) seem to suggest that there is a stronger reaction for IPCC events – in line with this hypothesis.

 

 

5. The writing in general and the grammar in specific needs significant work, it was often difficult to understand what the authors were trying to say on the first pass. I would suggest a professional English editor here, but perhaps also just a careful re-reading of the paper.
   1. For instance, in the abstract the authors claim that there is no violation of the effective market hypothesis, rather than the efficient market hypothesis as it is referenced in the rest of the paper.

 

 

 

Minor Comments:

 

1. The literature review I think needs to be re-worked. As written it does not tell me how this paper contributes to the literature on the pricing of climate risk, rather it tells me how some of the rest of the literature guides the writing of this paper. In other words, I would want to see how this paper fits and adds to the literature and that is not what is currently provided. In addition, there is a good deal of new and very interesting papers dealing with the pricing of climate risk that could be added here.
2. In the analysis section, I am hoping that the formula for return is simply mistyped and missing either an ln function (and even then the Taylor series approximation is only valid for small changes so may not be appropriate) or the price change in the numerator.

 

3. I have to take issue with the suggestion (at multiple points in the paper) that short term profits could be made by recognition of this analysis. It is not at all clear to me that this would be possible given that the results show that there is little cross sectional correlation with carbon intensity (and that is the only cross sectional test applied) and return. The authors would need to explain how exactly an investor could use these results to identify profitable trades ex-ante. Especially since in the majority of industry/event dates the average outcome seems to be no abnormal return.

Author Response

Please find our reply in the attached file. Thank you very much for your valuable comments and critical remarks!

Author Response File: Author Response.pdf

Reviewer 2 Report

The topic of the paper is innovative and the processing of the paper is in accordance with the current state of knowledge in the scientific community. The abstract contains the main required components. The paper is written in accordance with traditional IMRD (introduction, methodology, results, discussion) structure of scientific papers. The literature review is not at the required level. It would be useful to enrich the literature and to include references from journals indexed in reputable databases in the last two years. Discussion in its traditional structure should be enriched by a constructive comparison of results of own research with the contemporary state of knowledge formulated so far. The scientific value of the paper should be highlighted by the critical construction of framework conditions of research outcome´s application as well as detected barriers and future perspective of research. It might be interesting to extend the research by a study of the probability of financial problems in companies as a result of negative abnormal returns due to the IPCC Reports and Regulatory Announcements.

 

Author Response

We express our gratitude to the reviewer for very interesting proposals and critical comments.

We have extended the literature review and added findings from papers published in the last two years. The limitations and future prospects of the study are presented in the Discussions section. We think that the proposal by the reviewer –to study a probability of financial problems in companies – is very interesting, but it hardly can be investigated on the base of event studies approach (that captures only short term market reaction), so probably it deserves a separate study.

Reviewer 3 Report

I think assessing the effect of climate change news on stock prices is an important area of research, and looking at the effect of IPCC announcements is a smart way of doing it.

I think exposition can be improved in many ways:

1)      Sometimes the choice of terminology is wrong. To give two examples from page 1: the abstract talks about “Effective markets hypothesis” to mean “Efficient markets hypothesis”; the first paragraph uses the verb “to endorse” in the wrong way

2)      The introduction should describe the findings, not only the research questions

3)      The 3 hypotheses should not be in the literature review

4)      The authors should be more clear about what tables actually mean, in particular Table 2.

Author Response

We express our gratitude to the reviewer for very interesting proposals and critical comments.

 

I think assessing the effect of climate change news on stock prices is an important area of research, and looking at the effect of IPCC announcements is a smart way of doing it.

I think exposition can be improved in many ways:

• Sometimes the choice of terminology is wrong. To give two examples from page 1: the abstract talks about “Effective markets hypothesis” to mean “Efficient markets hypothesis”; the first paragraph uses the verb “to endorse” in the wrong way

Sorry for these mistakes. We looked carefully through the text and made necessary corrections. Moreover, the native language speaker has proofread the paper and confirmed that the terminology was correct now.

• The introduction should describe the findings, not only the research questions

We re-worked the introduction by adding new contributions and specifying the contributions that have been already defined. However, as findings are briefly presented in the Abstract, and more specifically in the results section, we consider excessive also place findings in the introduction. As a matter of fact, findings are rarely presented in introduction sections in top journals.

• The 3 hypotheses should not be in the literature review

As hypotheses are based on the literature review, we renamed the section to Literature Review and Hypotheses Development. It seems logical to present them together.

4)      The authors should be more clear about what tables actually mean, in particular Table 2.

We have re-worked all the tables (as we had increased the sample) and presented the results in a better manner.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The edits made have improved the paper substantially.