Are our preferences for risk dependent on relative choices?

Investment portfolios
Experiment Type
Better understand financial decision-making
Increase long-term savings
Focus Areas
Behavioral Concepts
Framing Anchoring

What Happened

It worked. The analysis showed that people’s preferences for and tolerance of risk is informed by cues and influences in the surrounding context.

Lessons Learned

The analysis provides evidence that people’s preferences for and tolerance of risk is not static but instead informed by cues and influences in the surrounding context.


When people enroll in a retirement plan, inevitably they will be asked to choose a “style” of investing that reflects their comfort with investment risk. Most commonly, this experience comes in the form of selecting from set of portfolios labeled as “Conservative,” “Moderate,” or “Growth.” Their selection is linked with an underlying blend of equity, bonds, and other investments that are intended to match their preferences.

Figuring out our own preference for risk is not straightforward and, unfortunately, people often end up selecting a set of investments that are not ideal. Young investors construct a portfolio that is too cautious.

Investors over-correct the portfolio to adjust for recent changes in the market. In the end, people rely on a range of heuristics and mental shortcuts to estimate how much risk they want in their investments, which can lead to suboptimal decision making about their long-term savings.

This project specifically explores what role the choice architecture related to the presentation of investment portfolios plays in creating a mismatch between risk and individual tolerance for risk.

Key Insights

We hypothesized that people’s investment choices and risk preferences are, in part, shaped by the relative comparisons that they could make. Most people don’t know how to evaluate choices independently; they need to evaluate them within the context of alternatives.

Relative comparison plays an important role in how we as consumers evaluate and make financial decisions.

Just as one example, prior research demonstrates that changing the choice set to include a “decoy” can shift people’s preferences. People are more likely to purchase options that are similar, but superior to the decoy choice even when they were more purchase a different option before the decoy choice was added.


In this experiment, we randomly assigned 605 respondents into three different “investment portfolio” conditions. In each condition, they were shown a set of four investment portfolios and asked to select the portfolio that they would be most likely to invest in if they had the chance to do so.

The portfolio sets varied for each condition based on the distribution of risk and return across the four investment portfolios:


We started our analysis by coding each portfolio 1-4, in terms of risk and reward within a specific choice set. So while the “Balanced” portfolio was coded as a “3” in the low-risk choice set, it was coded as a “1” in the high-risk choice set. If people were sensitive to changes in risk, we would expect the average risk of selected portfolios to change as the overall distribution of risk skewed in one direction of the other.

We see that respondents made investment selections that reflects a sensitivity to risk -- the average risk of the selected portfolios significantly decreases between the low-, even-, and high-risk choice sets (p<0.001). For example, participants shown the low-risk choice set would select the portfolio with highest risk and return in the choice set, indicating a preference for more aggressive investments.

We then used the average responses from all three conditions to create an “expected” distribution of selections across all six portfolios. Then, we predicted how many people would select each of the portfolios when presented a particular choice set. If people selected investment portfolios independent of the choice set, the observed distribution would match the expected distribution.

We saw that this is not the case. Although respondents did shift their selections in ways that were sensitive to risk, they did not shift as much as expected. Respondents shown the low-risk choice set selected, on average, significantly less risky portfolios than expected(p=0.001). Respondents shown the high-risk choice set selected significantly riskier portfolios than expected (p<0.001).

These differences may have been driven by the naming conventions used to label each of the investment portfolios. Even still, the analysis provides evidence that people’s preferences for and tolerance of risk is not static but instead informed by cues and influences in the surrounding context.