The best educators have several things in common, the least of which is their passion for their area of expertise. Shachar Kariv meets this criterion. One of the world’s top game and decision theorists, Kariv is the Benjamin N. Ward Professor of Economics at the University of California at Berkeley, where he served as department chair from 2014 to 2017. He is also chief scientist of TrueProfile, an award-winning behavioral economics research and technology firm. When you talk to him about his research, client preferences, and the TrueProfile gamified, decision science-based client profiling platform, his passion is difficult to contain.
I spoke to Kariv shortly after it was announced that FPA had partnered with TrueProfile and Berkeley Executive Education at the University of California to launch a multi-year focus on research and inventions emerging from the field of behavioral economics. Through this partnership, a wide range of learning opportunities will be created to help financial planners understand their clients.
1. You were recently quoted saying, “There’s very little economics in behavioral economics.” What do you mean by that?
Behavioral economics is the marriage of standard economics and cognitive psychology. In many cases, psychology alone is being mislabeled as behavioral economics, and as a result, there are a lot of new experts emerging who ignore the power of the economics toolkit. My only guidance is that in this fusion food we call behavioral economics, you need to look at the label and understand what you are eating or being fed; the economics part is important and essential, because it brings a level of rigor and precision, and that can be lost if you are only using psychology.
2. Your research suggests that a client’s risk tolerance includes risk aversion and loss aversion, and we should also pay attention to ambiguity aversion. What is ambiguity aversion?
I think the first two concepts are in fact not well defined, so let’s define risk aversion and loss aversion precisely, then we’ll define ambiguity aversion.
Suppose I give you the option between gaining $100 for sure or participating in a game with a 50 percent chance you get $0, and a 50 percent chance you get some amount higher than $200. Let’s call this higher amount X. What is the X amount that will make you indifferent between playing the game and taking $100 for sure? Different people will have a different level of X where they switch from taking the sure $100 to taking risk on the coin flip. The higher this X, the higher your risk aversion or the more compensation you need for taking risk. As a profession, to really understand the client, we must find a client’s risk aversion X. For illustration, let’s say your X was $300.
Now, loss aversion will be taking this exact exercise, but instead of looking at it in the domain of gaining money, let’s look at it in the domain of losing money. Why? We want to see if the client makes a different decision in the domain of losses than they do in the domain of gains. So now, you again have two options. Both are bad, (sorry!) but you have to choose one of them. The sure option is you will lose $100 for sure—ouch. The “risky” decision is to play the game where 50 percent of the time you lose nothing, and 50 percent of the time you’ll lose $300. What do you decide?
If your answer to the second question is taking the $100 loss for sure, you have some loss aversion. As we professionalize the industry, we are obligated to use the highest standards and to measure precisely, and there’s no way of recovering an actual risk tolerance or loss aversion parameter from questionnaires.
Now, to illustrate ambiguity aversion, consider two scenarios—both are bad, but which is worse?
Suppose you need a medical procedure, and you ask your doctor how likely is it that the procedure will be successful. In scenario one the doctor tells you, “70 percent success rate.” In scenario two, the doctor tells you, “You have between a 60 and 80 percent success rate.” Most people are going to dislike the second scenario, because many people don’t like the uncertainty in probabilities. So, ambiguity aversion is the aversion to unknown probabilities or probability ranges/distributions.
Financial decision-making doesn’t only involve risk, it can also involve ambiguity.
We need to bring the science of understanding the client to the same high level of the science of building portfolios and to do so, we need to leave the old tools behind or use these questionnaires purely as a discussion guide.
3. In 2007 you published research on using graphical representations to determine individuals’ preferences. What did that research reveal, and what impact are those findings having on the future of financial advice?
First, what are preferences, actually? We need to understand what we’re trying to measure here. I claim that all decisions in life—large and small, financial or not—are governed by what I call the three fundamental trade-offs: risk versus return, today versus tomorrow, and self versus others. Preferences are basically the attitude with which people are solving these trade-offs.
There are three different ways to know about different preferences. The first one is a survey or a questionnaire. Let me be blunt here. It’s rubbish. Asking someone about how risk averse he is, is exactly like asking someone, do you have high cholesterol? I ate a big steak yesterday, so probably, who knows? People cannot answer what their preferences are in dimensions like this.
Another way to know about preferences is by having data on people’s decisions, often called naturally occurring data. Amazon is very good at understanding people’s basic preferences toward consumer goods, right? But unfortunately, we don’t have data like this when it comes to financial decisions, because financial decisions are made very infrequently and are multi-dimensional. And when I’m changing my portfolio, you don’t know if it is my preferences or the preferences of the adviser, so not only are financial decisions very infrequent, I don’t know whose preferences they reflect.
The third way is through simulations that have important properties and structure represented in graphical representations. This has been a large part of my research since I was a graduate student. The decision games have a lot of economic theory behind them, and by the way that you play the game I can understand your preferences.
In a very simple gamified environment, you make trade-offs between risk versus return. How you make those trade-offs reveals your risk aversion, your loss aversion, and your ambiguity aversion—they are mathematical measures about your preferences that inform portfolio construction.
4. What’s the difference between “stated preferences” and “revealed preferences”?
Stated preferences are simply I ask you what your preferences are, and you state your preferences to me. For example, from 1 to 10, how much do you like dolphins? (1 you hate dolphins, 10 you love dolphins). Think about the number in your head.
Now I’m going to ask you something about your actions, not about your preferences: how much have you donated to charities for dolphins? Your answer is likely to be zero. Your stated preferences are such that you like dolphins, but the revealed preferences from your actions are that you cannot care less about dolphins because if you really cared, you would have donated money, time, or anything.
With stated preferences, you ask people to communicate their preferences, which they don’t actually know and they cannot articulate. With revealed preferences, you use their actions to uncover their preferences.
5. Why should planners care about this difference?
Because understanding the client is the single biggest responsibility and source of our advantage and future as a profession. Oh—and, clients’ actions speak louder than words.
6. Planners have many options for tools for assessing clients’ risk preferences. What makes TrueProfile different?
You will learn more about your clients and have richer engagement by looking at the client’s decisions, not their words. The information we need from clients is trapped inside their decisions. We put you in a game environment where you make these trade-offs, and from the decisions that you make, we uncover your preferences.
The interface is very simple, but under the hood there is something very, very sophisticated. We know the formulae to recover your preferences. We have a statistical theory to analyze the data, and if you make mistakes, meaning you don’t have coherent preferences, we will be able to see this. We can flag to the adviser, watch out, it will be very hard to satisfy this person’s risk attitudes because he doesn’t have coherent, well-defined risk attitudes. That’s the key.
7. How do you see financial industry regulators’ views of client risk tolerance questionnaires evolving?
We’ve been talking to [regulators] in the U.S., Singapore, Australia, England, and I believe that regulators fully understand that there is a problem. They understand that when they say, “You need to know your clients” (and all the regulators say it in a variety of ways), that knowing your client is not something well-defined. They also clearly understand that clients can’t tell you how risk tolerant they are—the dispute resolution system is filled with examples of this.
The regulators will not tell the industry what to do; the industry will continue to have to put clients first. I’m sure there will be more pressure on this from regulators and the public, because financial well-being is becoming a much larger part of overall well-being, and the methods that support financial advice and financial planning are under growing focus. It’s hard not to operate at the highest standard of such an important dimension of planning and advice giving.
8. What is the current focus of your academic work?
I was trained as an economic theorist doing decision theory and game theory, and my work has always been this combination of theory, experimental work, and work with naturally occurring data.
My most recent work is with a philosopher from Yale on distributional preferences. It’s the preferences with which you trade off your own well-being and the well-being of a random American. We built models of these preferences theoretically. We’re taking our models and testing them with a very large sample of Americans and with very specific groups that are of interest to us. For example, we published a paper in the Proceedings of the National Academy of Sciences about the altruistic preferences of medical students. You want to think people who choose to go to medical school are more altruistic, that they care more about random others than the rest of the population.
We measured this, and the results showed a wide range of levels of altruism among physicians. The paper has created a lot of discussion in the medical profession. Medical schools are talking with us, asking, how can we measure people’s altruistic preferences at the start of admission, and how can we fix them to be more altruistic?
9. Have you ever been surprised by one of your research findings?
No, I’m never surprised by anything. You need to go into a research question without any biases. If it surprised you, it means that you actually had conjectured before seeing the data one way or another, and you shouldn’t have.
10. What role will naturally occurring or big data have on client understanding?
The key is to turn big data into smart data. The power of combining preference data and naturally occurring data is the “smarts,” an area where we spend a lot of time. You can’t glean the most important lessons from my naturally occurring data unless you understand how I make the fundamental trade-offs in life. You might be able to suggest some rules of thumb, but you need to know my preferences to make the most sense of all of the surrounding data you have about me.
Carly Schulaka is editor of the Journal. Contact her at HERE.