Learning from the lab: Redesigning retirement savings calculators to nudge users to increase contributions

Background

Retirement savings projection tools and calculators are everywhere – seemingly every financial service provider and financial institution has their own version. The variation between these many different tools is staggering, both in how they are structured and in how they communicate the results. While there has been a lot of work focused on understanding the factors that motivate contributions to retirement savings, there seems to be a gap in understanding the best way to structure these calculators.

To better understand what the best way is to structure a retirement savings calculator, we conducted a series of studies exploring their different elements and features. We were specifically interested in the usability of these calculators and what design choices would be most likely to encourage users to engage in positive behaviors like increasing contributions to their account.

Key Insights

Experiment 1: Uncertainty and spending in retirement

The first experiment that we conducted focused on a kind of retirement projection that displays both savings accumulation and decumulation overtime. We were interested in exploring two distinct questions related to this projection

1. Does graphically displaying decumulation change how a user thinks about their retirement savings?

2. Does introducing uncertainty by showing multiple estimates for how savings might grow over time change how users think about their future financial wellbeing?

Experiment 2: Framing Spending in Retirement and Graphics

The second experiment that we conducted used decumulation and focused on the type of graphic (line chart or bar chart) and the framing of retirement spending. Once again we were interested in how graphics and framing changed users' perceptions of retirement readiness and willingness to contribute to their retirement.

Experiment

Experiment 1: Uncertainty and spending in retirement

For the first experiment, we designed a 2x2 experiment where users were randomly assigned into 1 of 4 conditions:

• Accumulation with median only: Users saw a graph that did not show decumulation over time and only showed the median estimate.

• Accumulation with multiple estimates: Users saw the same graph as in the first condition but with two estimates for future savings amounts.

• Decumulation with median only: Users saw a graph that showed both accumulation and decumulation over time with only the median estimates.

• Decumulation with multiple estimates: Users saw a graph as in the first decumulation condition but with two estimates for future savings amounts.

Participants reviewed the retirement projections for two different hypothetical cases – one more prepared for retirement and one less prepared. They were then to assess the future financial wellbeing of hypothetical cases, their confidence in their assessment, how accurate and useable the graph they thought was, and what they would recommend the hypothetical person to do.

Experiment 2: Framing Spending in Retirement and Graphics

For the second experiment, we designed a 2x2 experiment where users were randomly assigned into 1 of 4 conditions:

• Years of retirement covered by savings with bar chart: Users were told how many years their retirement was expected to cover based on an assumed annual spend and the bar chart compared the years covered with the years expected in retirement.

• Years of retirement covered by savings with line chart: Users were told how many years their retirement was expected to cover based on an assumed annual spend and the line chart showed when the user would run out of money.

• Proportion of annual expenses covered by savings with bar chart: Users were told how much they would be able to spend in retirement based on an assumed length of retirement and the bar chart compared the amount covered with the amount expected to be needed in retirement.

• Proportion of annual expenses covered by savings with line chart: Users were told how much they would be able to spend in retirement based on an assumed length of retirement and the line chart showed the decrease in balance over time.

Study participants were asked to assess their own retirement status using a projection calculator as well the status of a theoretical individual’s retirement savings.

Results

Experiment 1: Uncertainty and spending in retirement

In the first experiment, we found that how the results from a retirement projection are displayed to users significantly changes how they interact with and understand those projections. Rather than consistently moving participants’ assessments of financial wellbeing in one direction, showing decumulation had differential effects: participants rated the more prepared case as even higher and the less prepared case as even lower (p= 0.001 and 0.096, respectively). Introducing multiple estimates did not have a discernable effect.

The results of the study suggest that, perhaps unsurprisingly, people anchor on visualizations. We also saw that participants who were shown the full graph that included both accumulation and decumulation were significantly more confident in their assessments as well.

Interestingly, the relationship between the projection and people’s recommendations on behavior was not straightforward. People were much more likely to recommend the less prepared case to increase their contributions, regardless of condition. A simple comparison finds that respondents were significantly more likely to recommend increasing contributions when shown decumulation (p=0.035). However, controlling for other factors and looking at the interaction between the two conditions suggests this relationship is likely significantly influenced by other factors.

Experiment 2: Framing Spending in Retirement and Graphics

In the second experiment, we found that the graphics (bar chart or line chart) made little difference in perceptions about retirement readiness. But framing played a large part! When the projections were framed in terms of years of retirement covered, users felt much less ready. Participants were not able to stretch out their dollars to make them last longer in the years of retirement framing conditions, so it seemed as though they were more likely to run out of money. Participants in the proportion of annual expenses covered condition were much less concerned about the state of their savings, since they were primed by the framing to think their money could be stretched to last through their retirement. They didn’t seem particularly concerned that the money available each year would not be enough to meet their needs or have an enjoyable life.