By Cora Pettipas
How a Coin Jar Estimation Game Threatens to Destroy the Disclosure Model
When something is repeated enough, it becomes assumed true. Cite an academic study along with the ‘fact’ and it becomes even more so. Even if the study has not been replicated independently, or has not been validly applied. Not all research is created equal, nor should studies be given equal consideration, especially by policy makers. There is a danger in this, especially when the studies are taken at face value with the author’s discussion points. This is how disclosure was vilified.
It has been argued in a recent publication that “(disclosure) can have unintended consequences, potentially exacerbating bias among brokers and harming investors it ostensibly intends to protect.” Wait, disclosure could be bad? This could have enormous consequences, as disclosure is the main means regulators use to deal with conflict of interest issues in the financial services industry. Conflicts of interest can reduce the flow of information to the client increasing agency costs leading to suboptimal decision making for investments. The assumption that disclosure is valuable and enhances client decision making is one of the basis of the CRMII changes to disclosing compensation. In the US, the Sarbanes-Oxley Act prescribes disclosure around conflicts of interest in securities analysis, and the SEC requires disclosure of compensation of recommendations, as well as the disclosure of the registrant firm’s non audit service fees.
The seemingly peripheral issue of disclosure is of the upmost importance in our regulatory environment. It could possibly influence future policy, like the fiduciary standard, commissions, and the ability for registrants to participate in other business activities. It could deprive investors of information that is seen as valuable to their decision making process. The world can unfortunately not be regulated into fairness, as long as people are gainfully employed, have families, or even strong personal preferences; there will always be conflicts of interest, they cannot be regulated away, as they will just pop up in more complicated forms. However, they can be disclosed, and then the recipient can make an informed decision. This article takes a closer look at the Cain, Loewenstein & Moore studies that attempt to debunk disclosure, as well as the stronger academic research of Church & Kuang (that unfortunately did not get the same attention) which concludes disclosure is not harmful but indeed helpful to client’s decision making process.
The So-Called Perverse Effect of Disclosure: The Cain, Loewenstein & Moore Studies
The study that has come to the attention of U.S. and Canadian Regulators that discounts the disclosure model, is dramatically entitled The Dirt on Coming Clean: Perverse Effects of Disclosing Conflicts of Interest by Cain, Loewenstein and Moore (2005). They later published a similar follow up study When Sunlight Fails to Disinfect: Understanding the Perverse Effects of Disclosing Conflicts of Interest (2010).
Despite their captivating titles, the Cain, Loewenstein and Moore studies are not worthy of the front row seat they currently have in regulators minds when it comes to policy making. They attempt to debunk disclosure by concluding two main points. First, that the presence of disclosure can bias an ‘advisor’s’ recommendations even more in their favor when a conflict of interest is present, a concept called moral licensing. The authors argue that conflict of interest increases advisor recommendation bias, but the bias is higher when disclosure is present because it ‘permits’ the bias. Second, the ‘estimator’ (the person who takes the advisor’s advice) will discount the advice, but not enough to compensate for the inflated recommendation that the advisor gives. The authors conclude from their findings that disclosure is harmful for clients because they do not discount properly the information they receive from the advisor with the conflict of interest. This is not an intuitive conclusion, or one that has been supported by policy decisions in the past, or other academic literature.
The Cain, Loewenstein and Moore studies state that disclosure can actually harm the people it is intended to protect by creating trust. Their writing suggests that because of this, conflict situations should be avoided at all costs. They argue that disclosure is harmful because the recipient does not process the disclosed information properly, and will make poor decisions. In fact, in their general discussion on policy, they state, “when a disclosure is made in person, the advice recipient may trust the advisors as a result of the disclosure.” Another major policy implication is disclosure creates obligation. For example, “if a (doctor) discloses that he gets $5,000 if a patient enrolls, the patient may feel pressured to enroll so as not to doubt the doctor’s integrity.” Therefore, when meeting in person, not only does the client not properly handle information given to them, but they may see someone disclosing a conflict of interest as more trusting and will feel pressured to make the transaction. These conclusions are a far stretch from what they studied and the experimental results they attained – let’s discuss why.
To access validity of any study, the devil is in the details. The 2005 study included 147 students in a lab setting. They were paid an average of ten dollars for participation. They were randomly assigned in the role of ‘estimator’ or ‘advisor.’ The goal was to accurately estimate how much money was in a jar of coins. The estimator had no information, the advisor was told it was a range between ten to thirty dollars. In the control condition, advisor’s interests were aligned with the estimator, as the advisor was paid on the accuracy of the estimator (but not having exact information themselves). In the other two conditions, the advisor was compensated on how high the approximation was from the estimator, and the estimator was paid on accuracy. In the test groups, one disclosed conflict of interest and the other did not.
The results indicated that the advisors gave higher values to the estimators under the direct conflict condition rather than the control condition. The values were higher still if the conflict was disclosed. The researcher’s conclusion was that disclosure gives advisors a moral license to mislead the client. What the researchers also found interesting was the way the estimators acted with the information under the different groups. Estimators discounted the values from advisors, but they did not discount them enough to compensate for the additional premium the advisors put on the coin value they recommended to estimators, and estimator guesses were more inaccurate in the disclosure condition. From this, researchers concluded that disclosure is actually harmful for the estimators.
The validity of generalizing the results of this study to financial services (or any other industry) are limited. First, this was a game, similar to the game of bluff, there was no explicit moral or ethical obligation for the advisor to act in the estimator’s best interests. In addition, there were no consequences for the advisor not to act in their own best interest. Second, the conflict of interest was direct, meaning the advisor profited at the estimator’s loss, this total non-alignment would not happen in industry, as reality is much more complicated than that, especially when factoring in reputation risks. Third, there was not an option for investors not to play, or ‘invest,’ which is a consequence of real world scenarios. Finally, the advisor had very little information, they were given a range of money in the jar, but not the actual value, so advice, biased or not, was a guess.
Cain, Loewenstein & Moore replicated their study in 2010, and attained similar results as there was not a significant structural difference in methodology. In the 2010 publication, there was a collection of four individual studies, but three were surveys. The first was an online survey of sixty-four people recruited from an alumni list with questions focusing on strategic exaggeration and restraint. The second was another online survey of people recruited from the University website focusing on how respondents perceived the ethical or moral licensing of influencing another’s estimation of jellybeans in a hypothetical situation. The third survey was given to ninety-two passengers on their way to Long Island regarding how likely they would be to accept conflicted advice from a doctor in one of eight hypothetical scenarios.
The last study is the most relevant for our purposes as it is an experimental study, and will be discussed, as policy making should focus more on how people behave, not how they perceive they will behave. Sixty-one students participated in a lab experiment with an average of ten dollars compensation. Students were again assigned as estimators or advisors. Estimators were to approximate the value of four houses in the University neighborhood (as opposed to the value of a jar of coins). Estimators and advisors were given information about the four homes, advisors were given additional information on comparable pricings in the neighbourhood. Estimators were compensated on accuracy of their guess. Advisors were compensated depending on the group, just like the first study. In the high undisclosed and high disclosed groups, advisors were paid on how high the estimator’s guesses were.
The results of the 2010 study were similar to the one in 2005, but with limited statistical significance. The advisors were incentivized to have the estimators guess high in the conflict of interest groups, and have them guess accurately in the control condition. On average, the high-undisclosed group gave ‘advice’ to estimators which was $31,351 higher than their personal guess, and the high-disclosed group gave ‘advice’ of $51,562 higher than their personal guess, a $20,301 difference, on average. In the conflict of interest non-disclosure group, estimators discounted advice (calculated by the advice received minus the estimator’s guess) by $11,216 and the conflict of interest disclosure group was $25,609, which was marginally statistically significant. In the conflict of interest disclosure group, the estimator’s guesses were less accurate than the conflict of interest non-disclosure group, but not statistically significant. However, the pay offs (which were structured by the researchers) for both groups differed, which was statistically significant. This, the authors argue, is the ‘perverse’ effect of disclosure.
The 2010 study had the same results as the 2005 study, but the criticisms of the former study were not adequately addressed. This study has limited generalizability to real world situations as the study was still in a game format. If only there was a study that regulators could use to address a more typical real world situation of investing. Luckily, there is.
The Alternative Value of Disclosure: The Church & Kuang Study
The most appropriate, but less publicized, study for applicability to our space was published in 2009 by Bryan Church and Jason Kuang called Conflicts of Interest, Disclosure, and (Costly) Sanctions: Experimental Evidence. This study had a similar structure to Cain, Loewenstein & Moore’s but had some important enhancements. 202 student participants were randomly placed in the role of an advisor and an investor, but the stimulus was a ‘risky’ investment. In this study, a risky investment is one with a potential outcome of receiving 50% to 150% of the investment principal back. The investor chooses how much to invest, if any, after looking at a recommendation from the advisor. The investor was incentivized to predict an unknown value accurately, and the advisor was incentivized to inflate the investor’s prediction (in the conflict of interest scenario). The two independent measures were disclosure and ability to sanction the advisor, one in a conflict of interest situation, and one in an aligned interest situation. The researches also controlled for risk tolerance of the investors. The ability to sanction the advisor had no benefit to the investor, and actually came at a modest cost.
The study’s results differed from Cain, Loewenstein & Moore in important ways. Church & Kuang found that there was no difference in advice from the conflict of interest groups, whether or not the disclosure was present. So the claims of Cain Loewenstein & Moore’s study that disclosure actually hurt investors as the advisors had moral license to inflate their recommendations when disclosure was present was disproved when a participant chooses freely to walk away. Church & Kuang rationalize that the moral license effect described by Cain, Loewenstein & Moore’s study was mitigated by the investor’s ability not to invest, motivating advisors to appear unbiased and give fairer values to the investor.
What is even more interesting, is that the bias in the conflict of interest groups was reduced when investors could make use of sanctions for what they considered poor advice suggesting that advisors expect investors to make use of sanctions when conflict of interest are disclosed. This has very important potential implications for regulation and investor protection, as investors had more accurate estimates of the investment value in the disclosure/sanctions available group than the other groups.
Another interesting finding from this study debunks the assumption there is manufactured trust from disclosure. The disclosure & available sanctions group had the highest percentage of investors opting out of investing. This also contradicts the theory by Cain, Loewenstein & Moore that investors feel more obligated to invest once an advisor has disclosed a conflict of interest because they appear more trustworthy. No investors opted out when conflicts of interest were not disclosed, indicating that investors are more sceptical of the advisors credibility when disclosure is present.
The findings from the above research are clearly contradictory in terms of the value of disclosure. Cain, Loewenstein & Moore’s 2005 study has been popularized as a case against disclosure. It would appear that the 2005 study is more well-known than the 2010 follow up study because the differences in the disclosure versus non-disclosure groups did not pass all statistical significance tests. The main body of research against disclosure stems from these studies. In contrast, the Church & Kuang 2009 study has more generalizability, and a more robust literature review. Their results (attributed to the investor being able to opt out) negate the moral license effect of exaggerated bias with disclosure for advisors. In fact, Church & Kuang give further guidance by showing that disclosure and available investor sanctions reduces the bias.
The implied tone in the Cain, Loewenstein & Moore literature is dangerous. It infers that clients cannot logically make decisions that benefit their circumstances. The message to policy makers would be to wrap investors up in a plastic bubble, make sure they are not exposed to potential risks (or opportunities) and eliminate any potential negative effects from conflicts of interest in industry by bringing in more costly and complicated legislation: to protect people from themselves. Unfortunately this is not the right direction, simplicity and transparency are best.
The alternative is to assess quality research like Church & Kuang, in addition to anecdotal qualitative experience, and ask ourselves what information really applies to our industry, how we can use it to roadmap smart policy and ultimately efficient markets. Transparency, not opacity, should be one of the guiding principles of market transactions and regulation. In order for policy to be efficient it needs to come from good information. There are more medical costs associated with obesity than smoking, but I do not see warning labels on potato chips (and am not suggesting there should be). What I am suggesting is that we question where the logic of policy making comes from, and the types of information and ‘facts’ regulators are using to come to their conclusions about our industry – such as ‘disclosure has perverse effects.’ As this article has demonstrated, it clearly does not.
Cora Pettipas FCSI, M.Sc., CFP, PhD (candidate) is currently the Vice President of NEMA. Cora is also the editor of Exempt Edge Magazine, the industry trade publication on the exempt market/private equity. She is also currently representing Canada on the editorial team of Financial Planet, and serving on the international BoK committee, for the Financial Planners Standards Board. She holds a Bachelor’s degree from McGill University, a Master’s Degree (Finance and Controlling) from Swiss Management Center University, as well as the following designations: CFP, FMA, CIM, and FCSI. Cora has had tenures with several financial institutions in the capacity of financial advisor, wealth management, and financial planning. Her most recent previous position was as a Professor at Mount Royal University, where she taught Finance and Financial Planning. Cora is also the founder and Co-Owner of Melodic Twilight, a successful business venture which is internet based and sells in over 20 countries. She has her work published and presented internationally.
 Go Beyond Disclosures to Stanch Conflicts of Interest by Brayden McCarthy published July 10 2015 in American Banker.
 The Dirt on Coming Clean: Perverse Effects of Disclosing Conflicts of Interest by Daylain M. Cain, George Loewnstein, and Don A. Moore. Published January 2005 by The University of Chicago.
 When Sunlight Fails to Disinfect: Understanding the Perverse Effects of Disclosing Conflicts of Interest by Daylain M. Cain, George Loewnstein, and Don A. Moore. Published January 2010 by the Journal of Consumer Research.
 Conflicts of Interest, Disclosure, and (Costly) Sanctions: Experimental Evidence. By Bryan Church and Jason Kuang. Published in 2009 by University of Chicago Press.