B School for Public Policy
Regulating Robo-Advisors, Wharton Business Radio Interview
With big data and automation becoming more common, so too has the “robo advisor”, any automated service that ranks or matches consumers to financial products on a personalized basis. Tom Baker, Professor of Law and Health Sciences at the University of Pennsylvania School of Law, joins host Dan Loney of Knowledge@Wharton to discuss his recent B-School for Public Policy seminar about research he’s been doing on the regulation of robo-advisors, particularly within the financial services industry.
Knowledge@Wharton: We have seen quite a significant growth in robo-advisors in the last several years. These are automated services that can match consumers to financial products using personal data. And these robo-advisors promote the fact that they can do this in a more effective and lower cost manner than traditional advisors. And within this new arena have come concerns of fraud. These concerns also extend to other parts of robo-advising, whether they be in the insurance world, banking and elsewhere. Tom Baker is a Professor of Law and Health Sciences here at the University of Pennsylvania. He has been doing work that looks at robo-advising, and he joins us to discuss that right now. Nice seeing you again, Tom. Thanks for coming in.
Tom Baker: It’s my pleasure. Thank you.
Knowledge@Wharton: How much of a recognition of potential concern is there by regulators right now?
Baker: Well, I think right now regulators are just trying to get their hands around what this automated advice is. And in the area where they’re most thinking about it is the investment area. For most people, when they think about a robo-advisor–if they think about it at all, let’s be clear–they’re thinking about Betterment, Wealthfront, Vanguard’s Personal Advisor–one of these services where you can go on the web, take a risk questionnaire, and then do your investing with the advice of this robo-advisor.
But the same technology can be used for matching people to insurance and to banking products. And I would say in the insurance and banking area, regulators are not yet really focusing on the potential for automated advice.
Knowledge@Wharton: When you look at the investment circle, then, how much of the market do robo-advisors really have right now?
Baker: If you’re talking about fully automated robo-advisors, where people don’t interact with a person, it’s a very small percentage of the market. But if you count programs that help human advisors pick, then it’s a much bigger part of the market. You know, most of your big investment houses have their algorithms. They have their formulas, their methods that they use. And those are not really different in kind from the fully automated, on-line, do-it-yourself model.
Knowledge@Wharton: But is it your expectation that we will see it continue to grow? Obviously, as you said, it’s really kind of a new idea in things like banking and insurance, but will we see it continue to grow on a steady path over the next few years just because of the digital nature of our society these days?
Baker: Absolutely. Although what’s interesting is in the investment area, people like me thought that we’d be seeing a lot more completely automated. But what you’re really seeing the growth in is in the BlackRocks, the Vanguards, the Schwabs–where they’ve got a person that you can talk to, but that it’s basically an algorithm that’s picking the investments and doing the work behind the scenes. People seem to still want to talk to a person.
Knowledge@Wharton: And that’s interesting, because we’ve seen stories in the last year or two about whether or not we’re going to see the loss of investment counselors at a lot of these firms because of the nature of robo-advising and the algorithm–which seemingly all businesses rely on at this point–that they were basically going to take some of this business away from the actual human beings being there.
Baker: Right. I think what we’re going to see is a change in what we think of as the “person” who’s the financial advisor. It’s not going to be so much the economics/Wharton finance-major, and it might be more a person trained in social work–someone who’s really good at helping people make a plan and sticking to it–because the investment work is really being done by the programmers who are writing the algorithms for the robo-advising part.
Knowledge@Wharton: In looking at the other areas, such as the banking sector, it’s not really a surprise. Because if robo-advising is used in investments, it is natural tie to also see it in banking. But with the insurance industry, robo-advising is more of a new process. But then again, insurance is based on data points and algorithms.
Baker: Believe it or not, the Affordable Care Act played a role in the insurance robo-advising. The Affordable Care Act set up these insurance exchanges, and the insurance exchanges are kind of a robo-advisor–to the extent they are really decision-support tools, rather than just sorts and ledgers. So your web brokers, your insurance exchanges–these are places where they’re using tools to help match people to a financial product. Those tools are not as advanced, in general, as they could be.
But they’re moving in that direction. And believe it or not, banking is the trailing part of the financial services sector here.
Baker: Yes. There are not good mortgage comparison tools, for instance. There are starting to be some credit card comparison tools and some small business loan comparison tools, but part of the thing about banking is in banking, they give you the money, and then you pay it back. So there’s more of a concern about automating all the way through the transaction. Whereas in investment automated, you give them your money, so there’s less of a concern about fraud on the part of the consumer. Of course there’s plenty to be concerned about fraud on the other end. But with banking, they give you the money, so they’re very concerned about automating the process because of consumer fraud. And some of that we saw in the financial crisis. It wasn’t automated, but there was plenty of people ending up with money who shouldn’t have.
Knowledge@Wharton: Now the concern about fraud in general, with all of the different hacking that we have seen go on in a variety of different sectors, I think most Americans are concerned about fraud at this point. The element of fraud that we were mentioning before we went on the air is not necessarily the people, but it’s the actual production of the program where you could potentially see that fraud, which could catch up a lot of people.
Baker: Right. To date, the growth of robo-advisors has been by companies that are using the kinds of financial advice ideas developed by people like our Wharton Finance Department–these are good, consumer-focused, broad-diversification, passive investing, largely strategies–that are good. But as robo-advisors grow, there’s going to be room for people to be advertising robo-advisors that can beat the market.
Robo-advisors, so far, they’re more of an index approach where you’re trying to match the market due to smart tax-loss harvesting and things like that. The more we get into the direction of people trying to buy robos to beat the market, the more room there is going to be for fraud.
Knowledge@Wharton: And there are other areas–things like Zillow.com, the real estate website. You can go on there and search for a variety of different potential houses that you may want to buy or rent on the Zillow website. But there’s also the element of the mortgage part of it, which they have incorporated, as well.
Baker: Yes. They have sort of a rudimentary–but it’s the best one I’ve found out there–of a mortgage comparison tool. And it’s more suggestive of what you could do, because robo-advisors at their best are individualized, so they learn about you, and then they match you with the products that are available in the market. And right now, the Zillow Mortgage Comparison tool just asks you to tell them whether you want a fixed mortgage or some kind of variable rate mortgage. But what robo-advisors in the mortgage space could do is ask you questions that would then allow them to match you with a product that maybe you’re not even thinking about.
Knowledge@Wharton: How much monitoring is done from a regulatory side right now, and how much more, potentially, do you think needs to occur in the years to come, as robo-advising kind of continues to develop here?
Baker: The SEC and FINRA both have robo-advisors on their radar screens. And as part of their market conduct exams, they are looking into them. In the insurance and banking world–in the insurance world, the National Association of Insurance Commissioners has set up a technology task force that’s starting to look at them. But right now, if you’re a web broker, you’re just regulated like you’re a broker, and there’s not something special about the automated nature of it. And I don’t believe that insurance regulators really have the capacity yet to evaluate whether these algorithms are doing what they’re supposed to be doing.
Knowledge@Wharton: But again, it’s something that they have to start to look into. And I’ll throw out another example, a story that came out earlier today. Speaking of the SEC, Jay Clayton made statements about initial coin offerings because of the concern of cryptocurrency. He is–at least on his part of the agency–he is looking forward to the potential of where that may develop down the road. It’s the same way for a lot of these agencies, whether it be at the federal level or maybe even at the state level, where you’re talking about state insurance formulas that are watching out for the consumer in Pennsylvania or New Jersey or Delaware or wherever.
Baker: Right. And my mantra for this is that–what I think is important–is that the people who are regulating the investment robo-advisors, the people who are regulating the banking robo-advisors, and the people who are regulating the insurance robo-advisors recognize that they have the same technology issues that they need to get their hands around, and that they should be cooperating with each other. What we’re seeing is that, in their own silos, regulatory agencies are starting to think about this as part of the larger fintech question. And similarly, in their own silos, the U.K., the EU, the United States, Singapore, Hong Kong are all starting to think about it, but we’re not seeing enough learning across those domains. And that’s what we’re trying to promote here at Penn.
Knowledge@Wharton:But because of the fact that you’re working off of kind of the same sort of algorithm, same sort of programs, then realistically you could almost have one office that is looking at it, with branches of it looking at insurance, looking at the banking sector, looking at investments–because of that, correct?
Baker: You certainly could. And that would take a lot of heavy lifting from a regulatory change perspective. What you could do that would be more manageable, I think, would be recognizing that are best practices that you can learn about, about how to verify that an algorithm is doing what the creators say that it’s doing, and that that knowledge could be shared across agencies.
Knowledge@Wharton: So you would love to be able to see, I guess, a set of minimum standards basically set into place, correct?
Baker: Correct. Yes.
Knowledge@Wharton: What would that encompass, in your mind?
Baker: Well, we have to step back for a minute and recognize there are four pieces to a robo-advisor. There’s access to the data that you need in order to personalize the recommendations. So we need to have best practices around access to data. And that’s why the investment robo-advisors are so far ahead, because the investment data about the products is all public information.
Baker:So you need access to data. Two, there are the algorithms. You know, what are the ranking methods in looking at them? And there you need basically to involve computer science and other people, other than lawyers and finance people, about how to teach regulators in order to make sure that their algorithms are doing what they’re doing.
And then the third is what people call “choice architecture,” which is the decision environment. You could have great data, a perfect algorithm, but you could have an environment that leads people to make bad choices. A co-author of mine did an experiment using health insurance, where they switched gold and bronze as the color for plans. And they kept everything else the same. And people preferred the “gold plans” that were really the “bronze plans,” because it was gold. So we need to know about choice architecture.
And then the final thing is privacy and security.
Knowledge@Wharton: What you just talked about brings up a good point, and something we should discuss is the fact that leaving this, in many cases, in the hands of the consumer may end up being a hard thing because of the fact that the consumer—and we see this to a degree coming through what we’re seeing now in 401(k)s—is that people don’t really manage and look at their 401(k)s a lot. It would probably be very similar when you’re talking about these investments or banking or insurance–that the people don’t investigate enough into what they’re actually getting.
Baker: Right. So that’s why we need to have these. The good thing about a robo-advisor is that if people choose to use it, and if it’s well designed, it will provide good advice. But that’s not enough. We need to set up that decision environment or the choice architecture so that people make the good choice.
So for instance, now here at Penn, my savings for my retirement—if I don’t make a choice, they’re put into a target retirement fund, which is the kind of thing that our Finance Department here at Wharton would say is a good idea. And I can choose to do something else. So that’s an example of a choice architecture. So with robo-advisors, we need to have competent algorithms. We need to have good choice architecture. And then the last thing is this privacy and security that’s not unique to robo advisors, but the more of our financial life we put online, the more important it is that we have both good privacy protections and security.
Knowledge@Wharton: Right. Which means that you have to have the buy-in. I guess you would, anyway, from the IT community to be able to be a part of this process so that we don’t have the potential breakdown on this, correct?
Baker:Absolutely. And here, for this sort of thing, we need to look broader even than financial services—to our health care colleagues. They are obviously thinking about this, too. And so one of the points that we make in our research here is that when it comes to privacy and security, it’s not sure that we have a comparative advantage over at the Law School, that really this is going to be more of an IT focus.
Knowledge@Wharton: The tricky thing I see here is because of the fact that the digital society that we’re in, and all of the different technology that we have now coming to our finger tips–that always is moving ahead. And it’s moving ahead almost at an hourly basis or daily basis—at worst, weekly basis. So we already know that there are issues on the regulatory side of keeping track of certain things that have been in play for twenty years, and really the government or regulators are catching up on. So should we feel confident that we can have the platforms that we need to have and the protections we need to have going forward because of some of the issues we’ve seen pop up in the past?
Baker: You know, I do feel pretty confident about this—in part because people are, to some degree, resistant to automated advice. As I said before, the robo-advisors that are taking off have people as an interface. And so what that means is the trajectory from working with a person who can cheat you in a lot of ways—to a more fully automated but perhaps, in some ways, more vulnerable system–is not going to take place over night. And so I do feel that regulators are focusing on this. I would like them to focus on it more. And I would like Penn to be a place where they come to for advice about how to do that. But I feel like we’re going to be okay here.
Knowledge@Wharton: Tom Baker is a professor here at the University of Pennsylvania at the Law School. Going back to something I said a second ago, though, even though the consumer has to be more aware, there is a level of security that they have to feel with all of this potential technology and the robo-advising. And I think for the most part, consumers do go in with that, correct?
Baker: I think they do. Although it’s interesting. I taught a fintech seminar for the last couple of semesters, where we look at hot topics in relation to fintech. And one of the things we look at is robo-advisors. And my students, who are selected into this seminar, who are therefore sort of tech-savvy people, when it came to robo-advisors, they said, “You know, I’d really rather talk to a person.” And I would show them all this research about how people make mistakes and how advisors sometimes cheat consumers, but they want to talk to a person. So I think that people feel relatively confident about turning over their data and information to a large organization whose brand they trust. But I think they’re still somewhat questioning, “Can I really trust the algorithm to make the right choice for me?”
Knowledge@Wharton: Well, let me ask you this: Where do we stand–we talked about how with the investment sector, there is that human component in many cases. Do we have that component in some of the other areas–like you mentioned, when we’re talking about health insurance, or we’re talking–obviously banking, I think we do. But health insurance may be the one, really, to focus on if we do or do not have that.
Baker: One of the things that was surprising to people like me, when the health insurance exchanges were set up, was how few people actually bought the insurance in a fully automated way. Most people used the call center. They used one of these navigators, or sometimes an insurance broker, so that in the insurance space, people are still using people. And so one of the things that I focused on in my own work is designing tools that these human advisors can use, so that they can help the person make the right choice. Because making the right health insurance choice or making many kinds of insurance choices is really complicated. And actually an algorithm can do a better job than a person.
Knowledge@Wharton: We have heard the stories about people who were just confused by what they potentially needed within their own health coverage. And you mentioned between the different plans, it made it that much more difficult to truly understand what they needed, because of the conditions that they already have or maybe need to think of down the road.
Baker: Also, health insurance plans are more different from each other than mortgages. The thing is–and this is where, I think, places like Wharton are so important–is that we need good models of what makes a good insurance plan. We need good models of what makes a good banking product for someone. Similar to how we’ve had, we do have good models of what is a good investment strategy for an ordinary person.
One of the exciting things about robo-advisors is that there is going to be a faster path from academic research into what is a good financial product to the dissemination, because if you’ve got people and advisors relying on robo-advisors, you’ve then got a method for disseminating that.
It’s not like Jeremy Siegel needs to develop an idea, and then we need to convince a million financial advisors to do it.
Baker:If BlackRock, Schwab, Vanguard–you know, ten companies–have a lot of the market and are using automated tools, that idea gets out fast.
Knowledge@Wharton: Are they making that investment? Are they looking down that path already?
Baker: Certainly your big investment houses have made a big investment in what they call “hybrid robo-advisors,” which are primarily automated systems with a human interface that allow people to ask questions and so forth. And that’s a big investment area.
Knowledge@Wharton: But are most of those situations the robo side first, and then the human side as the backup?
Baker: They are actually marketed simply as a low-cost investment advisor. And the way that it’s able to be low-cost is that there’s the back end. And that’s different and has been more successful from a gaining market share perspective than those companies that have advertised as being automated.
Knowledge@Wharton: Tom, great having you here today.
Baker: My pleasure.
Knowledge@Wharton: I look forward to talking to you about this again, because I know this is going to continue to develop in the years to come. Thank you very much.
Baker: Thank you.
Knowledge@Wharton: Thanks. Tom Baker, from the University of Pennsylvania Law School.