Investing for retirement is one of the most important decisions a person will ever make, yet most people spend less than one hour making it. In that small window of time, the typical 401(k) investor eliminates about 160 million possible investment options by answering three questions: How much should I contribute to my retirement plan? Which funds should I contribute to? How much should I allocate to each fund?
Despite — or perhaps because of — the importance and complexity of investing for retirement, most people do little research about their choices, consult with no one other than family members and don’t make any change to their allocation when they have the opportunity to do so.
Further complicating such decision making, investments are tied to an inherently uncertain market, so most people can’t easily estimate what the value of their investments will be when it comes time to retire. Given a decision so rife with uncertainty, how can the average person ever be anything but woefully unprepared to make an informed choice about investing for retirement?
When financial-services firms sell investment products to consumers, they offer some guidance in the form of information brochures and questionnaires that assess individual risk attitude. If a consumer completes a questionnaire and the results indicate a low tolerance for risk, an adviser will typically direct her toward the range of choices that carry lower risk over time — and typically result in a lower return. Such tools shepherd clients into making retirement choices based on the attributes of investment products, much like consumers weigh the attributes of other products — a safe car, a fast computer, a trendy pair of shoes — before buying those that match their lifestyle.
But using questionnaires and brochures for guiding investing for retirement has significant limitations. First, psychologists have shown that people’s attitudes about risk vary over time and in different situations; pegging risk tolerance at a single point in time may not be a sound strategy on which to base investment decisions. Further, investment products are different from most consumer products because they aren’t consumed immediately, or even in the short term, and must grow in value over a long period of time. Yet financial-services firms’ advice to their clients remains focused on the current risk attributes of their products, not the likely outcomes of consumers’ choices.
Professor Eric Johnson worked with Daniel Goldstein of London Business School (and formerly a postdoctoral fellow at Columbia) and Nobel laureate William Sharpe of Stanford University to develop a new tool that can help firms and consumers shift the focus of retirement decision making. “The standard questionnaires treat people’s risk attitude as if it’s a personality characteristic,” Johnson says. “We turned that upside down. Instead of asking people about personality characteristics — Do you sleep at night if you’ve lost a thousand dollars? — let’s ask what kind of distribution of income people want at retirement.”
Johnson and his co-researchers created the Distribution Builder (DB), an online tool that allows consumers to simulate retirement-investment allocations and then see a probable retirement outcome — in other words, one likely total income at retirement — based on their choices.
The DB is a simple computer display. Its left axis represents different probabilities — each row represents the chance that the investment allocation will return a given percentage of pre-retirement income in the long term. With a budget of 100 markers, each representing one probable retirement outcome, the user arranges each marker on the grid to reflect her desired probability distribution — a probability range that she would be comfortable using to determine her level of income in retirement. The consumer is effectively choosing how to spread out her risk — more here, less there.
For example, a user who wanted to safely aim for a retirement income of 75 percent of her pre-retirement income — a commonly advised target — would allocate as many of her markers to the 75th percentile as the DB budget allowed and would cluster the remainder as close to that percentile as possible. A consumer who wanted to aim to have 90 percent of her current income at retirement, then, would allocate as many markers to the 90th percentile of the distribution as her budget allowed.
The DB mimics how such choices work in real life because any time a consumer opts to pay for upside potential — an allocation that could result in a large return on the initial investment — she also has to accept some downside risk — a relatively high risk of losing the initial investment. As the user moves markers toward the 100th percentile, the total cost of the allocation goes up, and as users move markers downward, the cost goes down. (View a demonstration of the DB here.)
Once all markers are placed, they disappear one by one until a single marker is left — representing one probable outcome of the user’s allocation decisions. This could be useful in a couple of ways. A consumer could allocate her markers exactly the same way with each use of the DB to see different probability outcomes to gain a sense of the range of probably best- and worst-case scenarios under that distribution. Or, a consumer could allocate her markers differently each time to assess various possible outcomes under vastly more risky or less risky allocation scenarios.
The researchers tested the DB to confirm its accuracy in predicting investment outcomes over the long term and confirmed that their subjects (adults who had been investing in retirement for at least five years) used the DB in a way that was consistent with how the subjects chose to allocate their actual resources.
The DB also has potential as a useful tool to assess individuals’ risk tolerance more accurately than other tools that are currently available. By conducting a gambling game with their subjects, the researchers were able to accurately predict how much risk subjects were likely to expose themselves to in the DB. Those results in turn helped the researchers develop a data set to compare with constant relative risk aversion (CRAA), a tool widely used by financial-services firms to estimate a normative range of risk that most consumers are willing to assume while investing for retirement. The researchers’ initial findings suggest that many consumers have lower risk tolerance than what the CRRA and similar tools predict.
“If all people do not have the same aversion to risk, following standard investment advice would go against their preferences,” Johnson says. “Why shouldn’t they be permitted to allocate funds into a selection that better reflects their comfort with risk?”
Johnson suggests that more experiences with the DB may increase tolerance for risk in retirement investment, since consumers can see that likely outcomes of riskier allocations may not be as risky as they perceive. “The Distribution Builder can function like a flight simulator, allowing those investing for retirement to explore the outcomes of their decisions with only virtual outcomes,” Johnson explains. “You can learn a lot from the crash, without the pain.”
Eric J. Johnson is the Norman Eig Professor of Business in the Marketing Division and director of the Columbia Center for Excellence in E-Business at Columbia Business School and codirector of the Center for Decision Sciences at Columbia University.