Defense Flashcards
I had some trouble understanding how the different accounts of clients in Paper 3 are linked. Do these clients have a trading account and a checking account at the same bank? Is spending via credit card or directly from the account? Some further clarification would be helpful.
Clients that we observe indeed have a trading account and a checking account at the same bank and we merge data of the two. Regarding credit card transactions, a problem that we were facing was that these are settled at the end of the month. Therefore, it was not possible for us to trace back these expenditures to specific dates and therefore credit card transactions are excluded from our analyses. I made some adjustments to the writing in an attempt to explain this more clearly in the updated version.
How do you know that the consumption effect is driven by emotion regulation theory?
I say that our results are consistent with emotion regulation theory. It provides an empirical challenge to measure whether people actually feel negative emotions when they receive dividends from scandalous companies.
One option would of course be to simply ask them, however, this creates several problems. First, it could be the case that the social desirability bias or demand effects could influence the responses. When we ask people whether they feel bad for receiving scandalous dividends, they might think that the socially appropriate response would be yes, even though this does not actually reflect their emotions. Another problem would be that it is very difficult to accurately remember decisions in hindsight, so we would have to be able to ask people about their emotions immediately after they received the dividend and people would have to respond to us immediately, which is very difficult to execute.
So beside self-report, we can measure negative emotions in terms of behavior. Based on the literature that we cite, the behavior that we expect when people indeed feel bad for receiving scandalous dividends would be increased consumption from this negative dividend income and this is exactly what we observe. Therefore, the claim that I make is that our results are consistent with emotion regulation theory.
In future research, I think that it would be very interesting to show that people indeed have a negative emotional reaction when they receive a scandalous dividend. There are other ways than self-report, I could for example imagine a study using neuroimaging, or recording behavior such as facial expressions in a laboratory study.
Is it empirically true that ESG funds have higher fees than regular funds? There have been reports from ESMA that seem to contradict this. This concerns more the motivation of the study, but I think it is important in terms of external validity.
I have read the ESMA report and find it quite convincing. In contrast to the ESMA report, Baker et al. (2022) show that among index funds in their sample period from 2019 – 2022, expense ratios of ESG funds were 5.9 basis points higher than expense ratios of non-ESG funds. Aragon et al. (2023) show that north-American university endowment funds that invest sustainably pay 7.7 basis points higher fees. When speculating about these mixed results, I believe that it could be the case that there is a fee premium in the American market, but not in the European market. It could also be the case that there is a fee premium present for some financial product categories, such as index funds, but not for others. Further, as we show in our study, it is important to not only look at average fees, but to also separately consider products marketed to different customer segments. For example, it would be interesting to see whether there is a sustainability premium in product categories that are typically marketed to unsophisticated retail investors (e.g. dominated bonds (Egan, 2019)).
External validity is always the Achilles Heel of experiments. In that sense, I would appreciate more discussion of the limitations of the results. The findings are impressive but where do they leave room for alternatives? As an example, might peer information (in paper 2) have a different effect, when real investment decisions are concerned?
This is a valid concern, given some recent evidence on the limited effects of peer information on behavior in the long-term, for example in retirement saving (see Bauer, Eberhardt, and Smeets (2022) who find no effect and Beshears et al (2015), who even find a negative effect). I analyzed portfolio holding data of participants over the months following the experiment. Specifically, I explore, whether the proportion of sustainable funds over conventional funds in participants’ portfolios changes between October 2022 (prior to the launch of the survey) and January 2023. I apply two different definitions of sustainability in investors’ fund holdings. First, I make use of the definition of the Sustainable Finance Disclosure Regulation (SFDR). Specifically, I define a fund as “sustainable”, if it falls under article 8 or article 9 according to the SFDR and as “conventional” otherwise. Second, I use the globe rating of the platform Morningstar. Sustainably-minded investors have been shown to invest a larger share into those funds that have five globes (Hartzmark & Sussman, 2019). Therefore, I define a fund as “sustainable”, if it has five globes according to the Morningstar sustainability rating and as “conventional” otherwise. Table 3.5 of the updated manuscript shows the outcome of two sets of OLS regressions with the changes in the proportion of sustainable funds as dependent variables. All coefficients of the treatment dummies are not significantly different from 0. Therefore, as anticipated by the committee member, peer information does not affect sustainable investment behavior outside of the decision environment of the experiment. I made this clear in the interpretation of the findings for the final version of the dissertation.
How much more money are investors paying? You only look at whether or not they reject, why not look at the exact amount?
I would like to start answering this question by saying that our main interest lied with behavior of the financial advisors, which is why we only briefly talk about the clients in the paper. The fact that our research question concerns the behavior of advisors lead to us to make experimental design choices that caused us to have an unrepresentative sample of clients. For example, one criterion for the clients that we hired for the experiment was that half should have submitted sustainability preferences and half should not have submitted sustainability preferences. Also, half of the clients should be male and the other half should be female. Therefore, any outcome of the analyses of client behavior in this experiment should be taken with a lot of caution and the information about client behavior outside of the experiment may be limited.
Therefore, when talking about this analysis, you will see that I write that in the experiment, giving advice to sustainable investment clients is a realistic opportunity for advisors to earn higher fees.
How do we define Article 8 and Article 9 funds?
- Article 6 covers funds which do not integrate any kind of sustainability into the investment process and could include stocks currently excluded by ESG funds such as tobacco companies or thermal coal producers. While these will be allowed to continue to be sold in the EU, provided they are clearly labelled as non-sustainable, they may face considerable marketing difficulties when matched against more sustainable funds.
- Article 8, also known as ‘environmental and socially promoting’, applies “… where a financial product promotes, among other characteristics, environmental or social characteristics, or a combination of those characteristics, provided that the companies in which the investments are made follow good governance practices.” Check, whether funds promote, among other characteristics, environmental or social characteristics, or a combination of those characteristics.
- Article 9, also known as ‘products targeting sustainable investments’, covers products targeting bespoke sustainable investments and applies “… where a financial product has sustainable investment as its objective and an index has been designated as a reference benchmark.” This is quite easy: Check the objective.
Some data is removed and it is not so clear why. Would you say this is good academic practice?
We had quite some discussions on all the decisions to restrict the sample.
For example, we only include firms that have at least one TVL score change per quarter. With that, we follow Chen et al (2023), who argue that this restriction is required to measure individual responses to ESG-related news coverage. By explaining these restrictions more clearly, I hope to make explicit that the coefficient estimates that we report are only relevant to dividend income from firms who receive regular ESG-related news coverage. Similarly, the coefficients that we report are relevant for investors who receive dividends.
Another reason why we removed dividend investors is that it could be the case that they indeed receive dividends, but we just cannot observe them, because they flow into another account.
Please outline more clearly the implications for policy, perhaps one per chapter.
Starting with chapter 2, I believe that such far-reaching regulation changes, like the amendment to the MiFID II regulation should be properly evaluated with scientific methods. In chapter 2, I contribute to such an evaluation by providing data with some hopefully relevant pricing implications that may not have been considered by the High-Level Expert Group on Sustainable Finance that initiated the regulation change. Specifically, our findings highlight potential unanticipated consequences of the regulation, as investors that submit sustainability preferences and cannot signal high financial literacy may bear the burden of high fees. In equilibrium, sustainable investors are already expected to receive lower financial returns (Pástor et al., 2022). In addition, the potential impact of dominant sustainable investing in public markets has recently come under scrutiny (Berk & van Binsbergen, 2021; Hartzmark & Shue, 2023; Kölbel et al., 2020). When combined with higher fees, the attractiveness of sustainable investments is put at risk in the long-term.
Moving on to chapter 3, I hope that the results are also informative for the European Commission’s action plan to increase sustainable investments. The results suggest that timing of provided information is important to have an effect on sustainable investment behavior, but also that not too much hope should be put in the provision of additional information to drastically increase the sustainability of investors portfolios.
Finally, I think that the implications for policy from chapter 4 are less clear. I think what can be taken away is that people are very reluctant to sell to extreme ESG-related events, even though they seem to realize the information and even react to them to some extent by spending dividends more quickly.
Do you not think that competition will drive out the effect?
Financial advisors in our experiment do not really face competition. Arguably they face a bit of competition, because they compete against clients’ alternative option to simply select their own investment. So they cannot simply set the fee as high as they want. But they do not compete against other advisors, so I have to speculate here.
There are two reasons why I believe that competition is unlikely to drive out the effect. First, competition in the market for financial advice is not perfect, because there are barriers to switching. There are search costs, clients have to go through an initial
Zooming out from the market for financial advice, we see for example in the paper by Hortacsu and Syverson that equivalent financial products i.e. an S&P500 index fund are traded at vastly different prices and are marketed to different client groups. They argue that this is driven by investors with high information search costs, i.e. investors with low financial literacy. Our results show that it is especially these clients who are charged the premium.
Therefore I would speculate that even in the presence of competitive market forces, the fee difference would prevail.
Is the Wald Test correct here?
We put both regressors into the same model, that is negative dividend inflows and non-negative dividend inflows. Comparing the size of coefficients within the same model can be done with the Wald test, as far as I am aware.
Considering them in two separate models would have created an endogeneity problem, as we would have moved the variables from the equation into the error term. I remember a conversation with Prof. Rodriguez, during which he gave me his opinion that endogeneity would be worse than the multicollinearity problem.
An approach using Seemingly Unrelated Estimations would have allowed us to consider two regressions separately, the error terms of which would have been allowed to be correlated. The problem in our case was that all equations would have had the same dependent variable and, in this technique, and the dependent variables would still have to be exogenous in this model specification.
What about misclassifications of (non-)negative dividends in the sense that they are both paid out on the same day?
This is indeed a shortcoming of our approach. We see consumption on day i, but it is not possible in this specification to know whether this consumption response is due to a negative or a non-negative dividend, if they are paid out on the same day. Looking at our data, there are indeed some days on which both are paid out on the same day. However, these would downward-bias our estimates of the true difference between the coefficients.
Another problem would arise if non-negative dividends would systematically be paid out together more or less often, compared to negative dividends. However, I have seen no indication for that in our data.
An option could have been to simply remove the dividend days, on which more than one dividend was paid out, however, this would have given rise to new biases.
How would you defend your statement: Financial advice will not eliminate the disparity between household sustainability preferences and their manifestation in investment behavior.
The outcome of a survey with German investors that I report in the introduction, according to which a majority report having no or very little knowledge about sustainable investing. Importantly, a large fraction of those who had an ESG consultation with their financial advisor after the MiFID amendment reports having learned nothing or little from this session. This suggests that financial advisors have limited ability to reduce this knowledge gap.
Further, the findings from my dissertation suggest that new information provided to investors has a limited effect on sustainable investment behavior. Also, the fact that financial advisors charge sustainable investors with low or unknown financial literacy higher fees may reduce the attractiveness of giving a sustainable investment mandate, or even to communicate sustainable investment preferences to financial advisors.
I would also talk about the field in general. Investors who submit a preference for sustainability to their financial advisor should have trust that recommended financial products have a positive impact. However, it is difficult to measure the impact of sustainable investment products. Many sustainable funds use divestment strategies, which have been shown to have limited, no, or even a negative impact (e.g. Kölbel et al. (2020); Hartzmark and Shue (2023) or Berk and van Binsbergen (2021)).
How would you defend your statement: Behavioral interventions alone will not solve the most pressing policy problems, such as climate change or retirement savings.
The stance that is taken by literature on behavioral interventions is that changing the behavior of the individual, in the absence of systemic reforms, can solve policy problems. And there has been some promising early work on the efficacy of behavioral interventions. For example, in the climate change domain, a study by Hunt Alcott shows that sending letters to people in which they are informed about their energy consumption relative to their neighbors increases average energy conservation levels. In the retirement savings domain, for or example a switch from having to opt-into retirement savings plans to an opt-out frame has increased retirement savings. As a result, a lot of public policy focuses on the individual, rather that the system in which this individual operates. For example, Obama was a big fan of behavioral interventions and had his own “Nudge Unit” to inform public policy.
However, there has now been increasing discussion casting doubt on the long-term efficacy of behavioral interventions alone to solve these policy problems. There was a recent discussion paper by George Loewenstein and Nick Chater, in which they argued against relying too much on behavioral interventions, which is remarkable given that they were highly influential in this field of research. In the discussion paper, they argue for a refocus on system change, for example in the form of a higher carbon tax to set the incentives right to fight climate, or in the form of mandating employers and employees to allocate more to retirement saving and the prohibit withdrawals from retirement savings.
That being said, I think that scientifically evaluating the effectiveness of behavioral intervention is important and using this evidence in policy is important. However, I do not think that such interventions alone will be sufficient in to completely solve climate change and the “retirement savings crisis”, as it was termed by noble price winner Richard Thaler.
Why do cumulative MPCs sometimes go down? How do we interpret Figure 4.1?
MPC coefficients can also be negative, if on a certain day people consume less per euro of income, compared to the reference time frame, i.e. days outside of the event window.
How do you get to the extra EUR 12.81 of spending on day zero?
I think that we looked at the average size of dividends that were paid out and multiplied that with our coefficient difference in Table 4.1, column 1.
