Defaults Are Not the Same by Default

Disney World is the land of magic and fairy tales, but even there you cannot escape science.

When ordering something to eat, one of us (Jon) noticed that the default choices in kids’ meals were all geared toward healthier options. (When you research decision-making for a living, it’s hard not to observe choice architecture everywhere, even on vacation.)

The menu swapped soda for juice and french fries for fruits and vegetables. Indeed, a recent study shows that this change in Disney World’s policy lead to the consumption of 21 percent less calories, 44 percent less fat, and 43 percent less sodium. These defaults are helping “the happiest place on Earth” become a healthier one.

Defaults are one of applied behavioral science’s biggest success stories. There are two reasons underlying their widespread adoption: first, defaults can be very simple, even consisting of just the one-word difference between, “If you want to be an organ donor, please check here,” (opt-in) and, “If you don’t want to be an organ donor, please check here” (opt-out). Second, defaults are surprisingly effective in a variety of contexts, in retirement planning decisions, health decisions, as well as consumer decisions.

Despite, or perhaps because of, the widespread use and success of defaults, a few important questions have remained in the background: How have defaults been implemented? Does it matter how they are implemented?

Knowing when and why defaults work highlights the importance of actively, rather than passively, considering and applying choice-architecture tools.

There are many ways researchers, policymakers, and other practitioners have attempted to use defaults. As alluded to above, defaults can be implemented in a variety of domains, such as in consumer settings (CFL versus incandescent lightbulbs) or health domains (organ donations). Defaults can also vary in how easy it is to opt out, ranging from a click on a website to requesting several forms under Austria’s organ donation law.

The second question revolves around how to see the effectiveness and widespread adoption of defaults in context: defaults are only one of many tools available in a choice architects toolbox. For example, while citizens could be defaulted into health insurance plans, they could also be asked to select their health insurance plan from a smaller, curated set. Similarly, employees could be defaulted into retirement savings plans when joining a company, or, alternatively, they could be given a limited time window in which to sign up. Policymakers thus have an array of options to choose from, beyond defaults, when determining how to use choice architecture to attain desired outcomes.

What matters, then, is understanding how effective defaults are as a choice architecture tool, as well as how different kinds of implementations alter a default’s effectiveness. This was the aim of a meta-analysis of all prior default studies, which we recently published in Behavioural Public Policy. Meta-analysis helps provide a summary statistic that reflects how strong a default is, on average, in prior work. Because there is variation in the effectiveness of defaults across different studies, we can exploit this variation to help us understand when defaults are more likely to be effective.

In total, we found 58 default studies with a total sample size of 73,675 participants. The studies came from a wide variety of contexts, topics, fields, and countries. One thing became apparent in our analysis: on average, defaults are a strong choice architecture tool, shifting decisions by 0.63 to 0.68 standard deviations. What this means is that in decisions where there are two possible options, the option that is preselected is on average chosen 27 percent more often than the option that is not preselected. That means that the average default study was about two times more effective in changing behaviors as other strong behavioral interventions that shift decisions by 0.2 to 0.3 standard deviations—one of them being, for example, Opower’s social norm intervention on energy savings, another widely popular choice architecture tool. So, on the one hand, defaults work!

On the other hand, there were also substantial differences in the effectiveness of defaults. In some studies, a default was far more effective than in other studies; and in others yet, defaults did not alter participants’ decisions. This is an important caveat, which highlights that choice architects should not blindly apply defaults to all situations, but instead be more careful in when and how they implement defaults.

We wondered what factors make defaults particularly more likely to be effective. To do so, we drew on a theoretical framework which highlights that defaults operate through three channels: first, defaults work because they reflect an implicit endorsement from the choice architect—your company’s HR department, your city’s policy office, your credit card company, your child’s school. Second, defaults work because staying with the defaulted choice is easier than switching away from it. Third, defaults work because they endowdecision makers with an option, meaning they’re less likely to want to give it up, now that it’s theirs. As a result, we hypothesized that default designs that trigger more of these channels (also called the three Es: endorsement, ease, and endowment) would be more effective.

In our analysis, we find partial support for this idea. That is, we find that studies that were designed to trigger endorsement (defaults that are seen as conveying what the choice architect thinks the decision maker should do) or endowment (defaults that are seen as reflecting the status quo) were more likely to be effective.

In addition, we find that defaults in consumer domains tend to be more effective, and that defaults in pro-environmental domains (such as green energy defaults) tend to be less effective. What this highlights is that the intensity and the distribution of decision makers’ underlying preferences—what it is that they care about and want—plays an important role in how effective defaults are. When decision makers care less about a particular choice, a default may be more persuasive in swaying their decision. Likewise, when preferences within a population are more varied, such that some people may have preferences that align with the default, but many people may not, then a default may be less effective.

When decision makers care less about a particular choice, a default may be more persuasive in swaying their decision.

One domain that people tend to care about deeply —and which tends to be divisive—is their environmental attitudes. As a result, someone who holds more pro-environmental attitudes may be more likely to stick with a default that offsets the carbon emissions arising from their flight, while someone who holds anti-environmentalism attitudes may be more likely to switch away from the default. In addition, environmental attitudes tend to vary broadly throughout the population, as research on the acknowledgement of human-caused climate change, or lack thereof, shows. As a result, both the strong intensity with which people hold environmental attitudes and their broad distribution in the population make it less likely that defaults will be effective.

In contrast, a domain that people tend to care less deeply about—and which tends to be less divisive—is which search engine they use. While there are many search engines available,  like DuckDuckGo or Qwant, more than 75 percent of searchers currently go through Google. This metric is accounted for in part because Google is the default search engine on a number of browsers, including the company-owned Chrome, but also Firefox and Safari—a default setting that prompted Google to pay Mozilla and Apple billions of dollars last year. Because people don’t care very deeply about which search engine they use, a default setting is likely to be more effective.

To help understand how to best design defaults, using the three Es and taking into account intensity and distribution of preferences, we put together a checklist of questions that policymakers and other practitioners could ask themselves during the next choice-architecture design meeting. We note that these questions are not exhaustive but highlight specific aspects to pay attention to when designing defaults.

Our research exploring when and why defaults work highlights the importance of actively, rather than passively, considering and applying choice-architecture tools. It also shows the benefits of understanding how they work. Such ideas may help us predict how well a default could operate in a given setting and figure out how to design defaults that work better. In addition, defaults may not always be the most effective solution. They represent just one of many tools in the choice architect’s toolbox. To better explore when defaults should be used over other tools, choice architects should also evaluate the effectiveness of defaults versus other possible interventions.

When introducing defaults into complex real-world environments, choice architects thus need to be mindful that defaults are not the same by default.

Defaults Are Not the Same by Default