Ritov, I., & Baron, J. (1992). Status-quo and omission bias. Journal of Risk and Uncertainty, 5, 49-61.
Paul owns shares in Company A. During the past year he considered switching to stock in Company B, but he decided against it. He now finds that he would have been better off by $1,200 if he had switched to the stock of Company B. George owned shares in Company B. During the past year he switched to stock in Company A. He now finds that he would have been better off by $1,200 if he had kept his stock in Company B.Subjects in Kahneman and Tversky's study predicted that George will feel worse than Paul. Landman (1988) has extended this finding to positive outcomes: people feel more elated when a positive outcome resulted from action rather than inaction. Anticipation of such positive or negative feelings could account for the status-quo bias or the omission bias. If the potential negative feelings are weighed more heavily than the potential positive feelings resulting from a decision made under uncertainty, then these biases would result. Kahneman and Miller (1986) explained this phenomenon in terms of norm theory: Outcomes are perceived as worse when subjects can easily imagine that a better outcome could have occurred. When an adverse outcome is caused by action, it is easy to imagine the outcome of inaction, so the emotional reaction is strong. But when an adverse outcome is caused by inaction, it is more difficult to imagine the outcome of action, so the reaction is not as strong. However, in all the above examples inaction was completely confounded with maintaining current state. It is possible that the change of state, rather than the protagonist's involvement in the action which brought about this change, is the cause of the intensified emotional reaction. The three experiments reported here were designed to test separately the two potential sources of the status-quo bias: preference for maintaining current state and preference for inaction. The first experiment found that a bad outcome (e.g., financial loss) of an event involving action on the part of the victim is perceived as worse than an equivalent outcome when no action was involved, even when action is associated with maintaining the status quo (e.g., objecting to a stock being sold). The second experiment examined decisions involving tradeoffs between two dimensions (e.g., deductible amount and co-insurance rate of insurance policies). The experiment manipulated orthogonally the options designated as the status quo (the first, the second, or neither) and the option associated with action (the first, the second, or both). It found an omission bias in choice, but not a consistent status-quo bias. The third experiment examined asked subjects to match two options in desirability by providing a value of one dimension of one option. Designating one of the options as the status quo had no effect on the value of the dimension chosen. Here, action is not involved, so this result supports the role of action in the status-quo bias. In the same experiment, choices between options favored the option associated with inaction, whether that option was the status quo or not. We conclude that at least part of the status-quo bias found elsewhere is the result of omission bias.
Frank owned shares in Company B. During the past year his investment manager asked him whether he would object to switching to stock in Company A. Frank did not object. Now he finds out that he would have been better off by $1,200 if he had kept his shares in Company B.If the change of state is the main source of intensified regret, than Frank should feel just as bad as George felt in the previous scenario. If, on the other hand, action is the primary factor in determining regret, then Frank is better off, since the outcome resulted from inaction on his part. However, there is another difference between this new scenario and the two former ones. In Frank's case the counterfactual alternative is in some way closer to the actual one: all he had to do was to say no to his investment manager's suggestion. According to norm theory, the undoing of the outcome is more easily imagined in Frank's case than in George's case, where one has to imagine George refraining from switching to the stock of Company A. If the ease of mentally undoing the event affects the feelings of regret, than we cannot directly compare Frank's case to the cases of Paul and George. To overcome this difficulty we introduce another scenario, in which a fourth person, Henry, faces the same situation as Frank, but decides to object to the switch:
Henry owns shares in Company A. During the past year his investment manager asked him whether he would object to switching to stock in Company B. Henry objected and got to keep his shares in Company A. Now he finds out that he would have been better off by $1,200, if he had switched to the stock of Company B.For both Frank and Henry the counterfactual alternative seems quite close (relative to Paul and George). However, Frank's state is changed through inaction, while Henry preserves his status quo state through action. Thus the comparison between subjects' prediction of Frank's and Henry's feelings will provide a proper testing of the `action' hypothesis. If action, rather than change of state amplifies feeling of regret, Henry should feel worse than Frank, although the bad outcome did not follow a change in his situation.
SQ-NO, the person's state did not change, and he committed no action; an example of this version is Paul's story, in the scenario described earlier; CH-ACT, the person's state was changed through his own action; George's scenario is an example of this version; CH-NO, the person's state was changed, without his committing an action; Frank's scenario is an example of this condition; and SQ-ACT, the person preserved the status quo state through action, as Henry did, in the above example.In addition to the case described above, four other cases were used. We will present here the CH-NO version of each of the cases:
I. Sam, upon arriving at the airport, was asked whether he would object if, for bureaucratic reasons he would be switched to another flight from the flight he was booked on for several weeks. Sam did not object. The plane crashed and he was injured. II. Frank, in the above example. III. Rose, a student at Penn, and her roommate enrolled in Section 1 of Biology 101. At the beginning of the term, for purely bureaucratic reasons, Rose was asked whether she would not mind being transferred to Section 2. She did not object. The term is over and Rose just learned she got a D in the course. Her roommate, who remained in Section 1, got a B. IV. Steve worked for Allied Van Lines. Last year a group of workers, including Steve, were transferred by the company to United Van Lines. Any of the transferred workers could have objected to the switch, and would have, in that case, kept his old job. A month ago Steve was permanently laid off by United Van Lines. V. Ruth took a cab to the airport. As traffic was heavy, the driver asked her whether she would object to his taking a certain shortcut road. Ruth did not object. However, traffic on the shortcut road was delayed because of an accident that had just occurred on that road and Ruth missed her flight.Procedure: Each subject received all four versions of each of the five cases. The instructions at the top of the questionnaire read as follows:
In each of the following cases, four people will experience similar events. The final outcome will be identical for all four, but the chain of events leading to this outcome will differ. You are asked to rank these four people in terms of how bad they feel. Write `1' next to the name of the person who feels worst, `2' nest to the name of the person who feels next worst, and so on. If you think that two (or more) people feel equally bad, write the same number next to both their names.Two different orderings of the versions (within each case) were used.
Table 1 Mean ranks (4=best) for the conditions of Experiment 1 (n=20) Case I II III IV V flight stock course job cab Version SQ-NO 3.55 3.35 3.10 3.35 3.50 CH-ACT 2.00 2.05 2.10 2.05 1.75 CH-NO 2.05 2.20 1.95 2.30 2.15 SQ-ACT 1.60 1.40 1.55 1.40 1.65The main finding of the present experiment concerns the comparison of a change in state which does not involve action (version CH-NO) and an action which is designed to keep the status quo (version SQ-ACT). In all cases, CH-NO is ranked higher than SQ-ACT. Fourteen subjects ranked CH-NO higher than SQ-ACT in most cases, but only 4 subjects gave the opposite ranking more often (p < .05). It seems that action, rather than change of state, is producing the intensified feelings one experiences when a bad outcome occurs. It is apparent also that the two new cases (CH-NO and SQ-ACT), in which a change must be accepted or rejected, are ranked lower than the original two cases (SQ-NO and CH-ACT) used by Kahneman and Tversky. Perhaps this result is explained by norm theory: the opportunity to accept or reject a change makes the chosen option seem closer to the forgone option. In these new cases, the decision must (presumably) be made at a particular time, because it is offered by another person. In the original cases, the decision can be made at any time.
Status-quo B: B is the status quo. Omission will lead to B, but action is required in order to choose A. Change to B: A is the status quo. Nevertheless, omission will lead to B (the status quo is about to be changed), and action is required in order to keep A.If subjects' choices of alternative A or B differ in those two conditions, they must be affected by preference for the status quo, rather than by preference for inaction. As an example of a test of preference for inaction, compare status-quo-B above to the following condition:
Change to A: B is the status quo. However, an omission will lead to A, and action is required in order to keep B.Since preference for the status quo alternative should lead to choosing B in both status-quo-B and change-to-A, a subject who makes a different choice in those two conditions can be affected only by the action/inaction distinction. For each problem, nine different versions were used. These are described in Table 2. The nine versions give rise to six independent comparisons in which preference for current state could be tested, and six independent comparisons in which preference for inaction could be tested. We hypothesize that comparisons of the latter kind will yield more differences than comparisons of the former kind.
Table 2 Conditions of Experiment 2. condition current action indicate status quo alternative action by 1. - A , B < , > 2. - B a (for action) 3. - A a 4. A A , B < , > 5. A B c (for change) 6. A A o (object to change) 7. B A , B < , > 8. B B o 9. B A cProcedure. Subjects first performed a matching task for each of the problems, in which they matched alternative B to alternative A by filling in a value for one of the dimensions of B. The values of the matched alternatives were used in all subsequent choice tasks. (Athough the options in the choices were equated for desirability, subjects did not seem to have difficulty in choosing. Indeed, Tversky, Sattath, and Slovic, 1988, have found that the preference order revealed by matching does not coincide with the one revealed by choice. Thus, using the values from the matching task in the choice task served only to prevent large discrepancies in desirability rather than to create indifferentiable options.) Subjects then went through conditions 1 to 9. In each conditions all five problems were presented. Stimuli were presented by computer. Subjects indicated their choice by pressing the corresponding key, as shown in Table 2. The ` < ' and ` > ' keys were used for choices in which neither option was the status quo. The inaction choice was always indicated by pressing the space bar. The letters `c' and `o' were used to indicate a change or an objection to a change, respectively.
Table 3 Results of Experiment 2 (n=50) OM pairs % in the # SQ pairs % in the displaying predicted displaying predicted conflicting direction conflicting direction choices choices (out of 6) (out of 6) Problem 1. 1.72 69 1.66 50 2. 1.50 69 1.64 56 3. 1.16 67 1.36 55 4. 1.20 73 1.80 43 5. 1.48 63 1.28 64 mean 1.49 72 1.55 51Table 3 displays the average across subjects of the total number of pairs (out of six pairs, for each problem) in which the action factor changed and in which conflicting choices were made, and the percentage of these pairs favoring omission (OM). OM was significantly greater than 50% in all five cases (t=2.35, p < .01; t=3.44, p < .01; t=2.46, p < .05; t=4.13, p < .01; t=2.04, p < .05, for cases 1 through 5, respectively) as well as overall (t=5.70, p < .001). OM was 67% for the conditions in which there was no status quo (conditions 2 and 3), which is significantly grater than 50% (t=3.17, p < .005), and OM was 71% for the conditions in which there was a status quo (conditions 5, 6, 8, and 9; t=4.85, p < .001). These two measures of OM did not differ (t=0.67). (This implies that the omission bias here is not simply a result of being reluctant to reverse a suggestion made by someone else.) Similarly we computed, for each subject, the number of pairs in which the status-quo factor changed and in which different choices were made, and the percentage of these pairs (denoted SQ) favoring the status quo. The averages of these measures are also displayed in Table 3. SQ did not differ significantly from 50% (t=.36, p=.722). SQ was significantly larger than 50% only in case 5 (t=.03, p=.97; t=1.18, p=.24; t=.88, p=.38; t=-1.52, p=.13; t=2.33, p=.02, for the five cases, respectively). Although SQ was 62% (significantly greater than 50%, t=2.13, p < .05) in the conditions in which there was no possibility of omitting action (conditions 4 and 7), it was only 49% in those conditions in which omission was possible (conditions 5, 6, 8, and 9). To compare the relative size of the status-quo bias and the omission bias, we computed for each subject the difference between OM and SQ. The average difference, across subjects, is 20%, which is significantly greater than zero (t=4.33, p < .001). It seems, then, that subjects did indeed prefer the inaction alternative, independent of whether inaction would lead to keeping the status quo or not, but we find little evidence for a status-quo bias.
Table 4 Conditions of Experiment 3. Condition task SQ dimension values indicate (version) from version action by 1. match B to A - 2. match A to B - 3. match B to A A 4. match A to B B 5. chose A or B - 1 < , > 6. chose A or B - 2 < , > 7. decide whether to object to change from A to B A 1 o(object) 8. decide whether to object to change from B to A B 2 o(object) Note: The column labeled `dimension values from version' indicates the version from which the dimension values were taken. '<' and `>' indicate the keys to be pressed to indicate the choice of A or B, respectively.
Table 5 Summary of results Experiment and measure Conditions compared Subjects favor ---------------------------------------------------------------- 1. Emotion Status-quo-omit vs. change-act Status-quo-omit Status-quo-act vs. change-omit Change-omit 2. Choice Omit vs. act Omit Status-quo vs. change No difference 3. Matching Status-quo vs. no-status-quo No difference Choice Status-quo-act vs. change-omit Change-omitOur findings are fully consistent with the norm-theory account as proposed by Kahneman and Miller (1986): omissions tend to be considered as the norm, and commissions tend to be compared to what would have happened if nothing had been done. Omissions therefore tend to be evaluated as `neutral' regardless of their outcome, while commissions are evaluated as positive if their outcomes are better and negative if their outcomes are worse than the presumed outcome of inaction. Our findings can also be reconciled with the standard interpretation of the status-quo bias in terms of loss aversion, but only if we assume that the reference point is omission rather than the status quo. By this account, negative consequences of the commission are weighed more heavily than positive consequences of the commission, regardless of the status quo. It is also possible, however, that loss aversion is itself limited to cases in which losses are caused by action. (The obvious exception here is our desire to rid ourselves of gambles with zero expected value, but this single phenomenon might be due to some other source of risk aversion.) Although the major accounts of the status-quo bias are not seriously affected by our findings, other possible accounts are brought into question. In particular, the bias cannot now be fully explained in terms of factors that cause an attachment to the status quo, e.g., familiarity or fear of the unknown. One such factor is adaptation to the status quo, which has been claimed to be a major factor in producing a reference point (Kahneman & Varey, 1989). On the other hand, these factors might still operate in situations in which subjects had more chance to acquaint themselves with the status quo than they did in our hypothetical scenarios. But note that some previous demonstrations of the status-quo bias do not give subjects a chance to adapt, even though real rather than hypothetical decisions are made (Knetsch et al., 1988). Spranca et al. propose other explanations of omission bias, which could be operating here. Subjects could hold themselves responsible for the negative consequences of commissions but not for those of omissions. This difference in responsibility, in turn, seems often to be supported by the belief that actors do not cause the outcomes of their omissions, for these outcomes would have occurred if the actor were absent or if the actor did not know of the possibility of affecting the outcome. Spranca et al., following Bennett (e.g., 1966, 1981), argue that this belief in non-causality is an illusion, for what matters is the alternative options known to be available. Preference for the omission option may also result from unwillingness to choose between the available options. In some cases, this could simply reflect a measure for reducing processing load. A `perfectly rational' utility maximizer should continually weigh other options against the status quo state, from possession of personal belongings to remaining in present employment. For most people this is an unbearable task. Foregoing choice by preferring the omission option is a way of reducing the number of decisions for which weighing and comparison of utilities is unavoidable. Note, however, that in these experiments, subjects have elevated this principle to the level of a rule of behavior. The effort of making the decision is required by the experiment and is therefore equal in all conditions. Our findings have some practical implications. First, they suggest that status-quo biases can be counteracted by changing the way in which options are presented to a decision maker. When both keeping and changing the status quo require action, people will be less inclined to err by favoring the status quo when it is worse. Some recognition of the role of omission bias is found in laws or rules that require specific actions to be renewed after some period of time, or in general `sunset' laws, but these laws may go too far in opposing the status quo (unless that is desired as a self-control strategy for the rule makers). A more neutral option would be to require a decision either way. A second implication concerns the measurement of values. Findings of differences between willingness-to-pay and willingness-to-accept, which prevent the measurement of values for public goods and other goods, might be reduced or eliminated by procedures that equate the degree of action, even though one of the options is in fact the status quo.