When Harm Helps: Building Choice Architecture for Investors

When I was 16 years old, my parents sat me down and said:

We know you’re at the age when you might start drinking. We are not encouraging you to drink—it will likely hurt your brain development and put you in danger of car crashes and premature pregnancy—but we know saying “don’t drink” probably won’t be effective.

Instead, we’re saying, “Please drink safely, and if you’re going to drink, do it in our house.”

My parents realized that hardline prohibition was ineffective and instead tried to realistically minimize my potential harm from drinking.

Today, as director of Behavioral Finance and Investing at Betterment, I apply this same pragmatic approach to help clients improve their financial outcomes when saving and investing. Not infrequently, clients engage in behaviors—or request the option to engage in behaviors—that are on balance likely to be harmful. So why would any behavioral designer allow customers to make these choices?

One reason is that my parents’ intuition was right. There’s ample evidence that hardline prohibition is less effective than a harm-reduction approach. Abstinence-only sex education is associated with higher teen-pregnancy rates; conversely, nicotine replacement therapies increase smoking quit rates between 50 and 70 percent. And while legal Prohibition did initially reduce public drinking (and possibly cirrhosis rates), it did so only temporarily and drove those who did drink to drink worse quality alcohol, increasing the severity of their health conditions.

But beyond the ineffectiveness of attempting to prohibit a behavior, there are three other compelling reasons for allowing clients the option to make harmful choices.

First, as a service provider, prohibiting behaviors means you are restricting your potential market to only those who place zero value on the prohibited service. Clients who believe they’d receive any value from the prohibited options must look elsewhere.

Second, it’s possible that by ungracefully repelling clients you’ll make them worse off. Clients who want to engage in the risky behaviors are likely the ones who would most benefit from guardrails and nudges.

Third, you won’t be able to generalize your findings. If you run randomized controlled trials or use other experimental methods to evaluate your interventions, prohibiting behaviors upfront means you’ve self-selected for a well-behaved sample.

So how can you design a system that allows for potentially harmful behavior but minimizes the harm such behavior causes? Here are six strategies I’ve found to be effective:

#1: Educate preemptively and just-in-time

There is much wisdom in the old saying “an ounce of prevention is worth a pound of cure.” The more that clients learn to avoid mistakes before they make them, the better. And it’s better to learn from small mistakes than big ones.

There are three ways of doing this: standard educational content, personal guidance, and relatively new-on-the-scene experience sampling. Whichever method you choose, timing matters. You should intervene both far in advance of the decision and right before the decision is made.

#2: Counterbias

Counterbiasing means employing a strong emotional force that counteracts a bias that leads to risky behaviors. For example, if a client wants to sell their stocks because they’re concerned about a market crash (recency bias), calculate the capital gains tax that would be due before they commit to it. Most people have an aversion to such unexpected costs, and some have an outright aversion to taxes. At Betterment, we’ve found that over 60 percent of allocation changes were aborted when we previewed a positive tax to the customer. 

#3: Make safe options sexy and dangerous options boring

Quite often, dangerous options are described in positive terms, such as “Advanced,” or “Pro,” or “Power User.” These are almost exclusively terms people enjoy using to describe themselves. If a feature or option increases the ability of a client to harm him or herself, don’t make it sound like a good thing. More mundane terms, such as “Detailed,” “Complex,” or “Unadvised,” can nudge clients in the direction of making choices that are aligned with their long-term financial goals.

From a visual design standpoint, it’s important to make advanced screens less attractive, interactive, and fluid. Designers call this approach “taps and text.” The more unique selections a person must make (taps) and the more real estate you take up with explanation (text), the less likely they are to reach a decision impulsively. A bit of friction goes a long way.

#4: Use customers’ time-inconsistent preferences to their advantage

With the luxury of temporal distance, clients often know the virtuous course of action they wish their future selves would take. They also know that in the heat of the moment, their self-control may be weak. Using simple self-control contracts can be an effective way to help them precommit to their cool-state plan.

For example, you can employ delayed gratification through cooling-off periods. A customer may indicate a desire to perform the unvirtuous action, but know it will not occur for a predefined period (say, three days). This delay has two benefits. First, the clients may forgo the first step, knowing that it won’t satisfy their immediate desire (“Three days is too long to wait!”). Second, they may perform the first step, but by the time the second step comes around, their ardor has faded and they no longer wish to proceed (“Never mind, I don’t need to anymore”). The fact that they’ve opted into this contract of their own accord is critical.

#5: Offer and emphasize alternatives with less potential for harm

Rather than allowing clients only one option—to sell everything—give them the alternative choice of taking just enough money off the table so that they feel safe. Use mental accounts to segregate risk into short-term and long-term buckets. Encourage clients to make moderate rather than extreme adjustments.

Importantly, set up a recovery plan. Half of the harm clients experience comes from not snapping back to where they should be after the storm has passed. Help them reinvest at the right levels, and do so as quickly as possible. Even better, make it part of the original decision, yoking the two choices together. 

#6: Give timely feedback

Recent research has shown it takes about two years to learn if you should continue being an active trader. A large reason why it might take so long is that most investors don’t receive high-quality feedback on their performance, and so (a bit incredibly) they don’t know what their returns truly are. Worryingly, the aggregate performance of day traders is actually negative—but clients might not know that.

When investors are given personalized, high-quality feedback on their trading performance, they dramatically reduce their trading, diversify their portfolios, earn higher risk-adjusted returns, and improve over time. This shows that when we give clients the ability to make choices that can make them worse off, we should help them learn from those choices as efficiently and effectively as possible.

Allowing clients the option to harm themselves may feel unnatural to many choice architects. However, we must remember that their fundamental motivations cannot be erased, only redirected. Being empathetic, flexible, and cooperative gives us a greater opportunity to help more clients. While we might have averages on our side, clients generally know more about themselves than we do.

Like my parents did for me, I let my toddler bump her head, scrape her knees, and touch hot things (with warnings) so that she learns for herself when, and why, it’s dangerous. We should allow our clients the right to do the same—as long as the right supports are in place.

Further Reading & Resources

  • Johnson, E. J., Shu, S. B., Dellaert, B. G., Fox, C., Goldstein, D. G., Häubl, G., ... & Wansink, B. (2012). Beyond nudges: Tools of a choice architecture. Marketing Letters23(2), 487-504. (Link)
  • Thaler, R. H., Sunstein, C. R., & Balz, J. P. (2014). Choice architecture. (Link)
  • De Bondt, W. F., Muradoglu, Y. G., Shefrin, H., & Staikouras, S. K. (2015). Behavioral finance: Quo vadis? (Link)
  • Kamenica, E. (2012). Behavioral economics and psychology of incentives. Annu. Rev. Econ.4(1), 427-452. (Link)