The second effect under examination is considered rather effective in directing individuals’ choices, but not yet well understood in terms of underlying psychological mechanisms (Dinner et al.,
2011; Jachimowicz et al.,
2019). Behaviorally, it is the tendency to stick with a preselected option, which leads to a shift in preferences at the expense of alternative options. This phenomenon is known as the default effect, which can be associated with, but not limited to, the status quo bias, that is the behavioral tendency to keep the current situation unchanged when facing a decision problem (Samuelson & Zeckhauser,
1988). As we will illustrate in the following sections, both effects seem to be based on reference-dependent judgments, but rather interestingly, we do not know what ought to be observed when a decision problem is both framed as a potential gain/loss and, simultaneously, one of the available options is presented as a preselected choice. In other words, currently, it is difficult to predict which one of the two effects would set the reference, if any, and how this would happen.
We believe that this problem may be of extreme relevance to (i) clarify which one of the two effects has the largest impact on choice behavior, (ii) enhancing our understanding of underlying psychological mechanisms, and therefore (iii) providing more detailed instructions to those who would apply these principles in practical contexts. To the best of our knowledge, no research has been explicitly devoted to answering this research question in decision-making under risk, except for one single preliminary study that has begun to uncover some interesting mechanisms regarding the effect of the default option in gambling decisions (Costa-Gomes & Gerasimou,
2020). For the sake of completeness, it should be mentioned that studies combining framing and default effects do exist, although they do not measure the proportion of risky choices, but only the acceptance rate of an option that is positively framed compared to an option that is negatively framed (that is the negated form of the positive statement; Johnson et al.,
2002).
Framing effect
In choice behavior, the simplest way to define the framing effect is a difference in preferences that occurs when individuals are confronted with two different, but logically equivalent descriptions of the same decision problem (Wallin et al.,
2016).
Considering the following example integrally reported from Tversky and Kahneman (
1981):
“Imagine that the United States is preparing to face an Asian disease that, given its exceptional severity, could cause the death of 600 people. Two alternative intervention programs are proposed to deal with this event. Assume that the exact scientific estimate of the consequences of the two programs is:”
Gain frame
“If Program A is adopted, 200 people will be saved.
If Program B is adopted, there is 1/3 probability that 600 people will be saved, and 2/3 probability that no one will be saved.
Which one of the two programs would you favor?”
Loss frame
“If Program C is adopted, 400 people will die.
If Program D is adopted, there is 1/3 probability that nobody will die, and 2/3 probability that 600 people will die.”
Which one of the two programs would you favor?”
In the original experiment (Tversky & Kahneman,
1981), the two versions of the problem were presented to two different experimental groups. Participant’s preferences were strongly polarized towards the sure option A in the gain frame (approximately 72%), and towards the risky option D in the loss frame (approximately 78%). When presented together, it is easy to notice that the two descriptions are the mirrored version of one another. Specifically, 200 people saved (A) implies that 400 people die (C), and the same applies to B (600 saved/no people saved) and D (no people die/600 people die). Therefore, the two versions are differently described but logically equivalent.
The difference in preferences, originally described by Tversky and Kahneman (
1981), violates the principle of invariance, which states that people’s choices should remain constant even when the surface description of a decision problem changes. Over the years, this phenomenon has attracted a considerable amount of attention since it challenges some core assumptions of the dominant normative decision theory.
This phenomenon has been originally explained by Prospect Theory (PT; Kahneman & Tversky,
1979; Tversky & Kahneman,
1981). In PT, the different options are called prospects, and the way the subjective utility of the outcomes is evaluated is similar to previous formal models, such as the Expected Utility Theory (EUT). In short, the psychological value of a prospect is the result of the utility of the outcomes multiplied by their probabilities. Importantly, both distributions of outcomes and probabilities are supposed to be represented non-linearly in the human mind. Specifically, the subjective pleasure or utility declines as the outcomes increase (Bernoulli,
1954; Kahneman & Deaton,
2010; Kahneman & Tversky,
1979). In psychology, this type of nonlinearity has been found in the way individuals rate monetary values on an affective basis (Giuliani et al.,
2021; Manippa et al.,
2021), and estimate prices (Giuliani et al.,
2017; Raposo et al.,
2018). For this reason, in the gain domain, a sure prospect of 200 may have more subjective utility than a risky prospect of 600 with a probability of 1/3, even though, mathematically, 1/3 of 600 is in fact 200.
Since in PT, the utility function is supposed to reverse in the loss domain, thus assuming negative values, a given prospect that has a higher positive value in the gain domain will have a higher negative value in the loss domain, thus leading people to avoid it. Hence, the S-shaped value function postulated by PT describes the tendency to be risk averse with potential gains and risk seeker with potential losses of the same amount.
It is worth noticing that in the above-reported example, the outcomes are potential losses, but they can appear to be potential gains when compared to a reference point determined by the words used to describe the outcomes. The use of the word “saved” (gain frame) implies that the reference is “600 people dead”, thus prospects fall in the gain domain and choices lean towards the safe option. In contrast, the use of the word “die” (loss frame) implies that the reference is “zero people dead”, thus prospects fall in the loss domain and choices lean towards the risky option.
Nonetheless, there are some alternative and conflicting interpretations that contribute to improving our understating of this type of framing effect and its underlying mechanisms. PT has been classified into the category of “value-first decision-making”, which has been questioned by “comparison-based decision-making without value computation”, a class of theories that rejects the core assumptions of EUT and PT (Vlaev et al.,
2011). For instance, within the family of fast and frugal heuristics (Drechsler et al.,
2014; Gigerenzer,
2004), the Priority heuristic (Brandstätter et al.,
2006) postulates that, when facing decision problems such as a binary choice between two alternative gambles, we search for pieces of information in a certain order (lexicographic), determining a hierarchy of reasons and stopping as soon as one reason reaches the aspiration level (“good enough”), otherwise we go ahead evaluating the next reason. Importantly, the aspiration level is based on a comparative judgment between the options and the first judgment focuses on the outcomes, without considering probabilities. In short, if the difference between minimum gains exceeds a certain portion of the maximum gain, the option with the
higher minimum gain is chosen. The same applies to losses, with the only difference that the more attractive option has the
lower minimum loss. Only if this condition is not met, probabilities of minimum gains/losses are considered.
Although this heuristic accounts for the
reflection effect, which is observed when real gains and losses are involved, it illustrates that value functions are not always necessary to explain choice behavior under risk and uncertainty. In the same vein, along the line of non-utility/value-based frameworks, the Fuzzy-trace theory (Brainerd & Reyna,
1990) and the Evaluative Polarity account (Wallin et al.,
2016) have been focused on linguistic and affective mechanisms as determinants of the framing effect.
Wallin et al. (
2016) have demonstrated that the difference between the perceived pleasantness of the options predicts the choice and that this relative comparison is only based upon the words used to describe the options, regardless of the positive or negative formulation of the problem. This account assumes that no domains are created based on reference points, and no mental operations on probabilities and outcomes are performed. Kühberger and Tanner (
2010) argued that fuzzy-trace theory would provide a similar account, assuming a direct comparison between options, in which, in the gain frame, the sure gain is preferred over the risky gain because the latter mentions the possibility that someone will not be saved, whereas the former does not (it only mentions that 200 people will be saved). In contrast, in the loss frame, the risky loss is preferred over the sure loss because the former mentions that someone will not die, whereas the latter does not (it only mentions that 400 people will die).
Analyzing these models, it becomes clear that they are not based on any independent assignment of value to the prospects before their comparison, but instead they are centered on the relative comparison between a simplified representation of the options.
The framing effect can occur in different contexts, demonstrating the robustness and pervasiveness of such phenomenon. For instance, in a classic study, McNeil et al. (
1982) found that patients with lung cancer found the risk of surgery to be acceptable when the rate of success was presented in terms of probability of living than in terms of the probability of dying. More recent research, analyzing real medical consultations, has confirmed that this bias is present in the way physicians propose either active surveillance or treatment through surgery or radiation to their cancer patients. In turn, the use of cancer survival or cancer mortality-related words seems to ultimately influence patients’ decisions (Fridman et al.,
2020).
Default effect
As stated above, the default effect is the tendency to adopt the preselected option, determining a choice bias toward the default while penalizing the alternative option. There are two paradigmatic and very straightforward examples illustrating how this phenomenon works. The most famous is likely the default effect on organ donation, reported by Johnson and Goldstein (
2003). They have demonstrated how the status quo of not being an organ donor, who can choose to become one (opt-in), leads to a lower rate of organ donations. In contrast, the status quo of being an organ donor, who can choose to withdraw from that status (opt-out), leads to a greater rate of organ donations.
Another well-known example has been reported by Madrian and Shea (
2001), demonstrating that when employees are automatically enrolled in a retirement plan (opt-out), it is 50% more likely that they will stay with that default, therefore having a retirement plan, compared to the group for which the default position was not to be enrolled in any retirement plan (opt-in). Finally, a further example can be found in Pichert and Katsikopoulos (
2008) who have demonstrated how defaults can be used to foster “environmentally friendly” choices in terms of energy usage.
These three examples seem to provide a clear, simple, and effective rule that can be useful in several contexts: whenever a choice needs to be encouraged to reach a given goal, present it as the default option, namely something people need to opt out of to make a different choice. This will maximize the percentage of people who passively “choose” the desired option.
Behind this apparent simplicity though, the default effect is more complex than it seems on the surface. In fact, it has been pointed out that the underlying mechanisms can be multiple (Dinner et al.,
2011) and that each default works differently based on contextual factors and on decision makers’ underlying preferences when no default is provided. In a recent metanalysis, Jachimowicz et al. (
2019) have pointed out that it is necessary to reach a better grasp of the reasons behind the effectiveness of the default effect to design a more effective choice architecture (see also Zlatev et al.,
2017). In other words, the default effect does not work the same way all the time as initially hypothesized. The discovery of these limitations constitutes a critique of the large proliferation of the phenomenon that has been probably both oversimplified and overgeneralized.
Theoretically, the default effect has been explained as the result of three different reasons why people may prefer not to make choices: (i) effort (ii) implied endorsement, and (iii) reference-dependent mechanisms (Johnson & Goldstein,
2003; McKenzie et al.,
2006).
The first explanation implies that the default is kept because opting out would require some form of physical or cognitive effort, as for instance filling out a form or doing some calculation. We speculate that a kind of default that may be based on the aforementioned mechanism is used in website cookies disclosure. They can be all accepted as a default or personalized by removing those that are not strictly necessary. However, it is certainly more comfortable and effortless to click on the “accept all” button, especially considering the fast pace that individuals usually surf the internet.
The implied endorsement works when individuals trust in an authoritative source that has provided the preselected choice, which can be represented by a policy maker, an expert, or an advisor (Tannenbaum et al.,
2017).
Finally, the reference-dependent account entails that the preselected option is coded as already chosen, becoming the status quo, thus acting like an instant endowment (Dinner et al.,
2011). The endowment effect is the tendency to consider an object as more valuable when it is owned than when it is not-owned (Thaler,
1980). The logic behind it is that if we ask a certain price to sell an object we own, but we are willing to pay less for the same object as buyers, it means that the reference point is set forth by the endowment, whereas giving up the object is framed as a loss. Therefore, we ask for more money to compensate for the negative value assigned to the transaction.
In a similar way, giving up the default option can be psychologically perceived as a potential loss, thus generating reluctance in opting-out behaviors. Dinner et al. (
2011) provide an alternative explanation of the psychological mechanisms responsible for this phenomenon, demonstrating that everything depends on the list of pros and cons generated in favor of the status quo. Specifically, positive aspects associated with the default and negative ones associated with alternatives (initial list) are generated before negative default and positive-alternative aspects (second list). Therefore, the initial list may result in a greater number of default pros and alternative cons compared to the last one (Query theory; Johnson et al.,
2007; Weber et al.,
2007).
This topic is part of a wider set of applied knowledge that has grown and developed in the last decade thanks to the diffusion of the Nudge theory (Hansen & Jespersen,
2013; Thaler & Sunstein,
2003,
2009). By definition, a nudge is
“… any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives” (Thaler & Sunstein,
2008). Therefore, the default effect can be considered a nudge. Importantly, every nudge can be used in two main ways: in the interest of the choice architect, namely the subject who designs a decisional environment with the goal of prompting a certain behavioral response, or in the best interest of the decision-maker who is the target of the design process (Gigerenzer,
2015). Naturally, as we illustrated above in the example of organ donors, the benefits are intended to be possibly extended to society as a whole.
It should be a crucial aspect of choice architecture to clearly define the goal of the actions taken to shape specific choice behaviors using mechanisms that are based on decision-makers automatic, and largely unaware, mental processes.
The joint effect of framing and defaults
Our study is devoted to investigating the default effect in a binary choice between a sure and a risky option both presented under two framing conditions: possible gains or possible losses. Moreover, to broaden our field of inquiry, we used two scenarios: a life-or-death decision and a financial decision. This setup uses a baseline in which no default is provided (no default), a condition in which the sure option is flagged (sure default), and a condition in which the risky option is flagged (risky default). Finally, we did not provide any information concerning the reason why a given option was flagged, but we asked participants to select what they believed to be the possible source of the default among four alternatives after the task. The presence of a baseline follows the recommendation of Jachimowicz et al. (
2019), who pointed out the importance of comparing default effects against an estimated distribution of preferences for a specific decision problem within a given population, whereas the presence of two scenarios would help to understand how generalizable framing and defaults can be.
In this experiment, we assumed there would be three different effects exerting their influence on the evaluation of options: framing effect, scenario, and default. Although the framing effect is generally considered rather robust (see e.g., Druckman,
2001a,
2001b; Kühberger,
1998), some replications have failed to report it (Bless et al.,
1998; Miller & Fagley,
1991). Nonetheless, we hypothesize that, in our experiment, a classic framing effect would influence the evaluation of the outcomes leading to risk aversion in the gain frame and risk propensity in the loss frame (H
1).
The framing effect has been studied in relation to several possible moderators (see for instance Maule & Villejoubert,
2007; Rettinger & Hastie,
2001; Kühberger,
1998). More specifically, one study has found an interesting difference between the types of outcomes, named arena of choice, comparing possible gains/losses of either human lives or money (Fagley & Miller,
1997; hereafter we will refer to them as F&M in the text). Precisely, they have reported the main effect of the arena of choice on the overall percentage of risky choices, with a higher risk-seeking tendency with human lives than with money. Hence, we would expect to replicate a similar pattern (H
2).
Regarding the default effect, it would presumably lead to an evaluation that favors the status quo, thus reducing the framing effect by leading individuals to be less risk seekers in the loss domain, when the default is on the sure option (H3), and more risk seekers in the gain domain, when the default is on the risky option (H4).
Finally, since our experiment leaves participants free to infer the origin of the flagged option, we will explore whether different beliefs about the default can influence its efficacy, by asking participants in the two default conditions why they think that a preselected option has been provided. This exploratory investigation would provide further information concerning the hypothesis of the implied endorsement as a potential antecedent of the default effect (Johnson & Goldstein,
2003; McKenzie et al.,
2006; Tannenbaum et al.,
2017).