Feedback Proposals 2021
Introduction
These are comments that I type along the way when I am reading your proposal. Most of my comments are questions where I think you can clarify something or be more precise. Sometimes I have questions. Some of these are because they are not in my area of expertise but I tried to ask questions that I think are important to resolve. Either to clarify in the text for your thesis so that your assessors do not get confused either or because the answer to the question can help you to think about your empirical approach a bit more.
In general, this means that my comments will focus on the weaknesses in your argument. Don’t take this personal, I have 10 proposals to read and I don’t have the time to write down everything I do like about your proposal and there is a lot to like. There will also be a number of typos. I did not carefully reread everything. Sorry! H
onestly, I am a bit worried about the quality of the proposals. The main problem that I see is that (with one exception) the literature review is fairly superficial. The main problem I see is that most of you have good idea of what analysis you are going to run or which data you are going to use but it is not always clear why you should run this analysis with that data. If you go through the rubric you will see that the main focus is that every section should be connected: the introduction should make it clear what the literature review will be about; the method section should follow from the literature review and prepare the results section; the results section should answer the precise question that was raised in the literature review and the introduction.
That connection is lost in a lot of proposals because it feels that you all already have the analysis fixed and than reverse engineer the literature review to fit that analysis. The goal should be that the literature review helps you to do the best possible analysis. This means that the analysis should be influenced by the literature review which does not happen if you are already set on the analysis.
The good thing about fixing the analysis is that I am confident that you all have a feasible project and that is why your marks are not terrible. However, I fear that the proposals are not setting you up for a first class thesis. If your analysis is poorly motivated it is going to be hard to explain results that are unexpected. Unexpected results are fine and will not effect your marks. However, if you cannot explain these unexpected results, this will effect your mark.
More specific issues
A lot of you do not clearly explain what you mean by one of your main constructs of interest. Within your research field there might be different definitions and operationalisations of your main constructs. It should be clear from the introduction what you mean by the theoretical constructs that you are interested in. At least make sure that you have something like a working definition for every construct that is in your title1.
One common mistake in the motivation that I still see is that you motivate your study by saying that your topic is important. However, that is not enough, others have studied this topic before you. I want to know what is different about your particular study that makes it worthwhile to read it.
I did not talk a lot about control variables in my comments because you are just following the literature but the literature is really, really bad with control variables. I also know that nobody will listen to this but I am going to put it here anyway.
- It’s not correct that you need to control for factors that influence your outcome variable. You need to control2 for all causal factors that have an impact on both your treatment variable and your outcome variable.
- You should not interpret your control variables. Unless, you put in all the right control variables that have an impact on the control variable that you want to interpret and the outcome variable (see the previous point), you cannot really interpret your control variable.
Another pet peeve of mine is that most of the literature ignores is that it’s actually hard to look at performance effects (stock returns, accounting profit, you name it) when your independent variable is a choice that a firm can make. If you assume that there are different costs and benefits to the choice for each firm, and you assume that firms make the best choice for their own firm than you would expect to see differences in the choice but no differences in performances (after adjusting for all the necessary other variables that drive performance). When your study investigates performance at least think about what your data will look like when you make the same assumptions.
One more pet peeve, that is also common in the literature. I do not think it really make sense to hide your ‘best’ research design in the robustness test. If you have a valid instrumental variable or if you can use propensity score matching to get around some endogeneity concerns than it should be your main analysis and you should be able to defend why you use this approach. However, these techniques are not magic bullets and require some assumptions which you should be able to defend. I am much more comfortable with proposals that say that they are going to use a regular regression where you explain the endogeneity issues and how they might effect your estimate than proposals that say that they are going to use an IV approach without an explanation of why the instrumental variable is valid or even what the instrumental variable is.
Alain Noel
- In the introduction, I am missing the link between investments, customer concentration, and CEO options. It is not clear why you are investigating these relations.
- I think what you want to say in the literature review is that customer concentration can be an indicator of operational and financial risks (i.e. a major customers leaving has a big effect) or it is an indicator of an operational strength of the company (the customer cannot easily leave). You provide mixed evidence for the two effects and it would be good if you could explain those mixed findings. Can this be explained by different samples in the studies, different definitions of customer concentration, different industries? This could give you some idea of when you would expect the increase in risk and when you would expect the decrease in risk with more concentration.
- You can then make the link with investments by first explaining which investments are more likely to happen in more risky firms.
- Finally, you expect option incentives can make CEO be willing to make more risk. However, I think you need to dig a bit deeper in that argument. For instance, if the CEO is paid a salary in a risky firm and no options, they would also be willing to take more risk. The reason they might be risk averse is because their income depends on equity compensation. To me this argument would suggest that the appropriate measure is not the option compensation but a ratio or difference between equity compensation and equity compensation.
- It was not 100% clear to me why option compensation would have a moderating effect and not just an overall effect on risk taking. In addition, have you thought about the potential effect of the board using more option compensation for the CEO in situations where they want them to take more risk (i.e. with customer concentration)?
- From section 4, it was not clear to me how you would measure industries that need continuous R&D.
Sherin Abisekaraj
- Your citation style does not conform to any style I know.
- The motivation/introduction sets up a comparison between FOMO and risk attitude. It reads to me that you are saying that some financial decisions are currently explained by some people been willing to take on more risk but it could also be explained by FOMO (e.g. gender effects). It’s not clear to me why financial literacy is important to that research question from the introduction.
- There is a subtle difference between FOMO in the crypto/investing context and in the BNPL context. In the crypto context, the fear is to miss out on the investment itself. With BNPL, the fear is to miss out on a purchase and that purchase is enabled by BNPL. I am not saying that the comparison is not valid but I think it is useful to make that distinction. Another distinction is that BNPL is a short term debt while crypto and meme stock investing are speculative investments. They have different payoff structures which you might have to take into account when you use one literature to motivate studying the other. Again, it does not mean that these comparisons are not relevant but it does mean that not all results can directly be translated from one context into the other.
- The literature review has a tendency of being one study after another. I think it would help if you could focus the specific research question that you are interested in. Is it ‘does FOMO and not risk attitude explain gender differences in BNPL decisions?’ or ‘is FOMO a predictor of BNPL decisions?’ or something else (I have no strong preference for a research question, my issue is that it is not immediately clear from reading the proposal). Once you have a good idea of what the actual research question is that you have, you can then structure the literature review around that research question.
Alasdair Philip
- You are mixing “we” and “I”, if this is deliberate (i.e. we = the reader and you, while I = you) this can work. However, I got the impression that it was more being inconsistent.
- My main struggle is that I am not sure what the exact comparison is that you want to make and whether it is possible. For the method section that struggle means that I am not sure that you are presenting a like for like comparison between meme coins and non meme coins.
- A first step would be to more clearly define what you mean by the different characteristics. What do you mean by fundamental value? Do you mean potential future monetary value? Do you mean the underlying technology? What do you mean by safe have status? For instance if Bitcoin behaves different from gold, to mean that does not necessary mean that it cannot be a safe have. It could be that Bitcoin protects against other downside risk than gold. Or is do I have the wrong definition here?
- Are all characteristics relevant for your research question? For instance, you spend a lot of time on Bitcoin as a safe haven but I would not expect that safe haven status is really what you will investigate for the meme stocks. Once you decide on which characteristics you want to investigate, you can then structure your literature review around these characteristics. From the method section, I gather that you want to investigate some drivers of returns (e.g. momentum and attention). The literature review should make a better case for why you want to investigate these drivers and not others.
- Empirically, does it make sense to compare the return and risk profile of assets that have a different order of magnitude in terms of trading volume and transaction costs (e.g. MSCI global index vs Bitcoin vs Dogecoin). Maybe it does but you will need to make a stronger case in the method section of your thesis.
- You do not describe what the variables are in the correlation matrices 3 and 4. I assume these represent correlation between daily returns but I was not sure.
Ben Chamberlain
My basic issue is that in the introduction and part of the literature review, you consider social media as a medium to communicate information and you contrast it with more traditional modes of corporate communication. However, empirically you seem to be using social media activity more as a measure for attention by retail investors for a certain stock. To me it would make more sense to structure the literature review and the introduction around the difference between retail investors and sophisticated investors. I would then also make more sense why you are at the same time studying short interest. That is social media attention is a measure for the proportion of retail investors while short interest is an indicator for more sophisticated investors.
As a sidenote, the Elon Musk tweet is interesting to make the distinction between the attention effect of social media and the information effect of social media. We know that the tweet by Musk was not true. He did not have the financing to take Tesla private and he got (sort of) punished for the tweet by the SEC.
I think of sentiment not necessarily as containing information about the fundamentals of the stock but about the perception of the fundamentals by retail investors. While sentiment could be an indicator of fundamental information about the stock it could also be an indicator of short-term momentum (retail investors like the stock, so they are going to buy/hold).
Related to the basic issue: On p9-10 you write: “If stock prices efficiently reflect firm’s fundamentals over time, the change in short interest should be informative about future returns”. This can be debated, the short interest is not a fundamental of the firm. It’s information about the position of supposedly sophisticated investors. That means that the proposed change in returns is not driven by a change in fundamentals but by an adjustment of less sophisticated investors to what sophisticated investors are doing. The larger point is that I feel that some of you arguments seem to rely on the argument that fundamental information is immediately reflected in prices, while some parts of the arguments rely on the existence of less sophisticated investors who may not be able to process all that fundamental information. I think a more integrated framework would help to make the arguments more coherent.
There are quite a number of grammatical mistakes and typos. Make sure that you give yourself enough time and drafts for your thesis. Typos and grammatical mistakes are easy ways to make your assessors grumpy.
Jun Heng Chou
The main issue I have is that it is not clear from the outset why exactly you want to interview the different stakeholders. There are roughly two reasons why you would want to do the interviews: (1) you want to know what the best accounting standards for crypto assets would look like, (2) you want to understand the standard setting process and the different perspectives of the stakeholders. Reason (1) assumes that the stakeholders will largely agree on the correct standard while reason (2) assumes that the stakeholders will disagree and will lobby for their preferred perspective.
You lean strongly towards reason (1). However, I think if that is what you are trying to do you need to make a better case of what you mean by the correct accounting standard. If we could go 10 years into the future, how would we know whether an accounting standard is good or not? This should also help you to be more specific about which standards are appropriate for which crypto assets. You could also make the distinction between financial institutions and other companies because that distinction is already been made for certain other financial assets in the current standards. In general, you want to bring out the comparison between different crypto assets and comparable “regular” assets more explicitly. You for instance say that some crypto coins could be treated as a foreign currency and follow the foreign currency standards. I think you would benefit from making these comparisons much more explicit. The comparisons don’t have to be correct but then at least you can compare them to the interviews with the different stakeholders.
Being more explicit about what a good standard is will also help you to reconcile why different stakeholders might have different opinions. They might have a different definition in mind of what a good standard is.
I think your literature review would also benefit from linking more with the empirical literature. For instance, Yermack (2015) says that Bitcoin (at least at the time) is not really being used as a currency and does not show the usage of currency more that of a speculative asset. These type of findings would probably inform the views of stakeholders about what the right standard is and whether a new standard is necessary.
Nafe Hamid
In the introduction, try to be more specific about who the audience is of the audit report and the KAM disclosures. I assume that you generally mean equity investors. It would also good to think more about the incentives of the client firm. Do they have incentives for more/less (detailed) KAM disclosure? Is this different for different clients.
Similarity is not well defined in the introduction and the literature review. The measure of similarity specifically measures the similarity in KAM between year 1 and year 2 of the same client. It would be good to see this made explicit in the introduction and more discussion in the literature review of why this is the type of similarity that users are interested in. For instance, I would expect that the KAMs should be quite similar from one year to an other because the firm’s internal controls, IT systems, operational risks, and so on will probably not change a lot from one year to another.
I know you are planning to test them in different regressions but it would still be interesting to think about what the results can actually tell you about auditor effects versus client effects. For instance, imagine that a firm is operating in a more risky industry, has a lot of leverage, with an aggressive reporting strategy. If these type of firms have more similarity, this will show up as a client effect in the regression for hypothesis 2. However, it is also very likely that the auditors will charge higher fees for these clients and thus this will show up as auditor characteristics in the regression for hypothesis 1. This effect would be driven by client characteristics. In general, if the audit variables are driven by the characteristics of their clients, the hypothesis 1 regression might be hard to interpret. In general, a more careful consideration of the research design would be useful.
Thomas Higgins
The first paragraph in the introduction illustrates my main issue quite well. You discuss long term changes in the energy and gas market but at the end you say that you are studying day-ahead auction prices which are obviously driven by short term factors. You don’t make it clear how this analysis can lead to an understanding of the influence of these long term trends. Similarly, in the method section, you mention that the prediction of wind are highly unreliable and shortly after that you say that you are going to see whether these predictions have an effect on prices. If you are right that wind predictions are unreliable they should not be associated with prices. In general, the study seems to be motivated by the data analysis but it is not clear what you can learn about the real world. The most straight forward application of your study would be for traders in these markets but than I would expect to see more literature on these traders and their strategies.
The second paragraph on p.4 was hard to follow. You use a lot of methodological jargon (e.g. “long-memory approach”) which makes it hard to assess what the implications are of the studies that you cite.
On p5, you say that shocks to demand and supply distort gas prices. Why do you call that a distortion? Is that not what is supposed to happen to prices when supply or demand changes?
On p6, paragraph 2. The second sentence runs over 5 lines and is really hard to follow. You have a number of these runaway sentences.
Your study is partly motivated by rising energy prices for consumers in Australia but the discussion on p9 seems to suggest that gas prices are kept artificially low in Australia compared to exports. As a result, it seems that is unlikely that the rising energy prices are caused by Australian policies but rather by global price increases.
From p9 onwards you discuss a lot of details about the Australian gas market. I think the details fit better in the method section because they help to explain how you construct your measures. These details could also be useful to explain why the Australian market is useful to investigate your research question. The research question at the is point is unfortunately not really clear and I think that is what makes you focus on these institutional details so much because you feel that you need to discuss them all. It is better if you focus on the details that are relevant to the actual question you are interested in.
Yuyang Zhou
I had a hard time following the proposal. It looks like a first draft that has not received sufficient feedback from your supervisors. Make sure that you give yourself enough drafts and enough time when you write your thesis.
One of the issues is that you seem to describe inflation, expectations and stock prices as forces that are independent from any economic behavior by companies, investors or consumers. This is not uncommon in macro economics and not necessary bad but it does make it harder to follow the arguments. The typical approach in macro economics is to start from a formal model of how all these variables are supposed to correlate with each other in equilibrium. An overall framework is missing from your literature review and it makes it very hard to interpret all the (sometimes contradictory) findings that you list. The absence of an overarching framework or model also makes it hard to judge what your contribution is. The impression I have at the moment is that the type of data you will work with is quite noisy and there are a lot of contradictory findings in the literature. This probably means that your results will be consistent with some earlier studies and not with other but that does not really tell us anything. We already know that there are contradictory findings in the literature.
Section 2.1, second paragraph. You are already talking about the methodology. This is not the literature review.
Section 2.2., you suddenly introduce consumption-to-wealth ratio and term structure shock without any explanation to the reader. There are other examples of this. What are these shocks? Why are they relevant to the research question? How do they fit into the broader framework?
What you have above 2.3 starting from “Theoretically, …” must come much earlier, before all the results. When further clarified, this is what your overarching framework should look like. It should also provide the structure for the literature review.
Measurement and Methodology: Why do you need the cointegrating coefficient? What does that measure? Why is it good that your measure of inflation expectation is independent from targeted inflation? Is the whole idea of targeted inflation not that it influences expectations and therefor actual inflation? Again, maybe this will make sense if you explain the theory better and it becomes clear what the actual question is that you want to answer.
Liwei Liu
You don’t need to give all the details of the method in the introduction unless the goal of the study is to improve on the method of previous studies.
There are quite a bit of grammatical mistakes and typos. Try to give yourself enough time and drafts for the thesis so that you can avoid those.
In the introduction, you say that you have three different measures of attention but you don’t say attention to what exactly. This makes it harder for the reader to understand why for instance temperature would be a measure of attention. Try to be more specific about who’s attention you are talking about and what they are paying attention to.
For the underreaction hypothesis is the logic that low carbon companies will operationally outperform high carbon companies but the market underestimates this effect? As a result, if the stock price of low and high carbon companies is the same at t=0, over time the stock price of the low carbon company will improve because it’s profit has increased more. The logic of the risk premium hypothesis is different. There, investors are expected to foresee the risk of high carbon companies and will require a higher return. That is at t=0, they are willing to pay less for high carbon companies and thus if they perform equally well in the future, there is a higher return for the high carbon company. There are a number of additional assumptions in these two hypotheses which may or may not be relevant. The underreaction hypothesis basically means that investors are not good at incorporating information about carbon exposure while the risk premium hypothesis assumes that they are. I also think that both effects can coexist and partly cancel each other out. That does not necessarily mean that you have to change anything but being as clear as possible about the hypotheses can help you interpret your results better. For instance, if you find no support for either hypothesis. It could be that they cancel each other out in the aggregate.
Chris Harding
Do not overplay the strength of your measure to avoid endogeneity issues. Every measure that captures differences between the individuals in a company will be subject to the possibility that those individuals were selected by the company and that the effect you find are driven by differences between the companies and not by differences between those individuals. Now, for the effects that you are interested in (i.e. corporate culture), I think this is less of a problem. That is whether the a company with a strong environmental culture hires more environmentally aware board members and CEOs or whether the CEOs and board drive that culture has very similar implications. However, you can make the same argument for the studies that use gender as a proxy.
The emphasis on measurement in the intro feels at times a bit repetitive. I think you could the mechanisms by which these people can have an impact (or the selection effect I explain above) deserve more attention. It can also help you to think about additional test.
I personally find the discussion about shareholder versus stakeholder perspective a distraction. Both perspectives agree that a firm should undertake environmentally beneficial activities that improve shareholder value. These are the environmental activities that you focus on. The distinction is more meaningful if you want to investigate whether environmental conscious CEOs engage in activities that are good for the environment but bad for the long term value of the firm.
I am times distracted by what you call corporate culture. I think the management accounting literature would reflect to what you are trying to measure more as tone at the top. You can also have a look at the leadership literature. I see corporate culture more as the result of the tone at the top but also long standing norms in the company which are probably influenced by the internal systems and hierarchies. The individual CEO or board will probably have less influence on corporate culture in their first year in the job. It might be worthwhile thing a bit more about for which CEOs and board members the effect is going to be more pronounced.
In general, I think I would like to see more precision in your final thesis. Some of this is language (corporate culture versus tone at the top) and depends somewhat on subjective preferences. However, I still feel that the description of for instance the main variables can be more precise. For instance, you say we use ratings from ASSET4 without really describing how these ratings are being constructed. Some more examples of the types of questions could already go a long way. Similarly, I think I can guess how your measure for corporate culture is going to be constructed but you explain it more precise without giving so much room to the reader to fill in the details.
As an aside about the corporate culture measure. I think it would make conceptual sense to construct a measure for the CEO and board separately. Basically, average the measure for the board members for the latter one. Your discussion about the role of the board and CEO seem to suggest that it is either the board or the CEO who drives the culture. If you construct the culture member as the average over the CEO and the board members, the CEO is seen as equally important as every single board member and you measure is going to downplay the CEO effect. You can construct your overall measure than as the average of the board measure and the CEO measure but you could also look at the separate CEO effect and board effect. Some descriptives on the correlation between the board measure and the CEO measure would also be interesting.
Joseph Monisse
You do not have a lot of references from finance journals in the reference list. You should look into those type of journals more for high quality research into the topics that you are interested in. The journals you want to look for are “Journal of Finance”, “Review of Finance”, “Journal of Financial Economics”, “Journal of Financial Studies”, “Review of Finance”, “Review of Corporate Finance Studies”, “Review of Asset Pricing Studies”, “Review of Financial Studies”. You need to be exposed to the established statistical methods on how to measure cost of capital and asset price returns. You cannot rely on blogs and your undergraduate classes. My lecture notes have some advice on how to look for the appropriate literature.
In section 1 Introduction, you immediately start with ESG policies. You don’t explain what ESG stand for and what these policies could be. You do not have to explain your statistical methodology in the introduction if it is standard and if you do not have a methodological contribution. Furthermore, you say that your study will focus on ESG related practices but as far as I can establish from the methodology you are not going to measure ESG practices in any specific way. You just assume that over time ESG practices have become more important. The hidden assumption is then that the shift in ESG practices is the only important shift over time, which is very hard to defend.
In the sections 2.1 - 2.3, you seem to suggest that adopting ESG practices can decrease firm risk. For instance, they might be more likely to be bailed out by the government. However, in the section 2.4, you say that adopting ESG practices is more risky for smaller firms. I think you need to think more carefully through the mechanisms here. Are you talking about different types of risks? Are you talking about the fact that adopting these practices have relatively higher costs for smaller companies? Or are you setting up two mutually exclusive hypotheses? Section 2.4 also seems to switch back and forth between reporting of ESG practices and the actual practices. What do you want to focus on? What is driving the effects in your theory.
I do not think you can do a study about what ESG regulation should look like. For one, not all ESG disclosure is based on regulation. A lot of it is based on private demand for that information.
I think what you are trying to do with market concentration is driving you too far from the ESG research question. There is quite a bit of evidence that the shift towards more concentration and superstar firms preceded the importance of ESG practices. It is going to be hard to show that concentration is the result of ESG regulation or disclosure requirements.