Are maximizers really unhappy? The measurement of maximizing tendency

Recent research suggesting that people who maximize are less happy than those who satisfice has received considerable fanfare. The current study investigates whether this conclusion reflects the construct itself or rather how it is measured. We developed an alternative measure of maximizing tendency that is theory-based, has good psychometric properties, and predicts behavioral outcomes. In contrast to the existing maximization measure, our new measure did not correlate with life (dis)satisfaction, nor with most maladaptive personality and decision-making traits. We conclude that the interpretation of maximizers as unhappy may be due to poor measurement of the construct. We present a more reliable and valid measure for future researchers to use.


Introduction
The term "satisfice" was coined by Herbert Simon (1955Simon ( , 1956 to describe a decision-making strategy that strives for adequacy, rather than optimality. People generally satisfice, according to Simon (1958), "because they have not the wits to maximize" (p. 62). More recently, the tendency to satisfice (maximize) has been conceptualized as an individual difference or trait (Schwartz, Ward, Monterosso, Lyubomirsky, White, & Lehman, 2002). Schwartz and his colleagues developed the Maximization Scale, which contains 13 self-report items aimed at assessing the general tendency to seek optimality (e.g., "I treat relationships like clothing: I expect to try a lot on before I get the perfect fit."). Low scores on the measure are intended to reflect a general tendency to satisfice. Findings based on the use of this scale have appeared in top academic journals (e.g., Bruine de Bruin, Parker, & Fischhoff, 2007;Iyengar, Wells, & Schwartz, 2006;Schwartz et al., 2002), and inferences drawn about maximizers have been reported in major media (e.g., Flora, 2004;Schwartz, 2004a) and a best-selling book (Schwartz, 2004b). In general, maximizers are characterized as an unhappy group of people, debilitated by unrealistically high expectations in the face of an excess of options.
Given that strong inferences have been drawn about the mental well-being of maximizers versus satisficers -based on responses to the Maximization Scale -it is especially important that we have confidence in the construct validity of the measure. Highhouse and Diab (2006) suggested that the maximization scale falls short of commonly accepted psychometric standards. The purpose of the present research, therefore, was to better understand the nature of the construct measured by the Maximization Scale, and to propose an alternative measure of the maximization construct. Our findings suggest that maximizers may not be so unhappy after all.

What is maximizing?
To have a viable construct, one must first have a theory of the attribute (Guion, 1998). This means that the maximizer attribute or construct must first be defined and then distinguished conceptually from related attributes with different labels. In defining the maximizer construct, Schwartz et al. (2002) drew from Simon's (1955;1956) conceptualization of maximizing as seeking only the best option and not settling for anything less. In other words, maximizing involves spending more resources to find a better option than the best one found so far. The opposite end of the continuum, satisficing, involves settling for "good enough" options. Maximizing, therefore, involves a trade-off between spending resources and achieving a more optimal solution. The maximizing/satisficing concept can be seen as a special case of the reflectionimpulsivity dimension (Messer, 1976;van Merriënboer & Jelsma, 1988). Maximizing tendency involves spending more resources in an effort to make incrementally better decisions. Although Simon viewed satisficing as a universal behavioral tendency, Schwartz et al. (2002) envisioned a trait, wherein people vary in their tendency to maximize or satisfice. Moreover, Schwartz et al. describe this trait in bipolar terms, such that people can not be both maximizers and satisficers. This theory of the maximizer attribute, therefore, implies unidimensionality.
Although Schwartz et al. (2002) do not present a clear nomological net surrounding the maximizer construct, they did correlate their measure with measures of other constructs. Results showed that maximizers reported significantly less life satisfaction, happiness, optimism, and self-esteem and significantly more regret, perfectionism, and depression than did satisficers. In his book, The Paradox of Choice, Schwartz (2004b) argued that the proliferation of choice in everyday life adversely affects maximizers' subjective well-being, and that being a maximizer plays a "causal role in people's unhappiness" (p. 86). If we examine, however, Schwartz et al.'s (2002) own interpretation of Simon's work, maximization represents an optimization goal. That is, maximization is pursuit of the best option. At the trait level, this implies a general tendency to pursue the identification of the optimal alternative. With this in mind, many of the items in the Maximization Scale seem to diverge from this definition. For example, one item deals with fantasizing about living a different life, another deals with difficulty in writing letters to friends, and one even assesses preferences for ranking things like movies. It is no wonder, therefore, that even Schwartz et al.'s (2002) original data suggest that the Maximization Scale is multidimensional.
As noted earlier, an important characteristic of the maximization attribute is unidimensionality. As such, all thirteen items of the Maximization Scale should load onto one factor. It should also be internally consistent (Allen & Yen, 1979). Highhouse and Diab (2006) administered the Maximization Scale to 319 undergraduates and found little evidence for unidimensionality, as a principal components analysis yielded a 5-factor solution, with the first rotated factor accounting for only 14% of the variance in responses. They also found little evidence for internal consistency (α = .58), with over 50% of the item-total correlations falling below even the most liberal psychometric (>.25) standard.

The present study
The primary purpose of this study was to examine whether the pessimistic conclusions about maximizer well-being maintain when the construct is measured in a way that is more faithful to the theory of the attribute.
We developed an alternative measure of maximizing tendency that focuses on the general tendency to pursue the identification of the optimal alternative. Items were generated based on this narrow definition (e.g., "I am uncomfortable making decisions before I know all of my options"), and the scale was designed to be internally consistent. We expected that this theory-based measure of maximization tendency would relate less with maladaptive traits, and more with behavior-based criteria, than the original Maximization Scale. Specifically, we examined the correlation with measures of indecisiveness, avoidance, regret, neuroticism, and life (dis)satisfaction. In addition, we designed behavior-based criterion measures, which included situational dilemmas and behavior reports about the amount of resources spent to obtain more information before making a decision. These criteria reflect the behaviors we would expect to follow from one's tendency to maximize or satisfice.

Participants
Participants were recruited from several sections of an introductory undergraduate psychology course at a large Midwestern public university. A total of 210 students participated in this study, 191 of whom provided valid data for all of the major study variables. The mean age was 19 (SD = 1.4), and 65% were female.

Measures
Maximizing Scale (MS). Schwartz et al.'s (2002) 13item measure was used as one measure of maximization.
Maximizing Tendency Scale (MTS). We began with three items from the Maximization Scale, and added six items that reflected the narrow definition of maximization as presented in the Introduction. High scores are intended to reflect a general tendency to maximize, whereas low scores are intended to reflect a general tendency to satisfice. Items are presented in Appendix A.
Personality, decision-making, and life satisfaction measures. These scales use a 5-point Likert-type response format.
Indecisiveness was measured using the 22-item Indecisiveness Scale of Germeijs and De Boeck (2002;e.g., "It is hard for me to come to a decision").
Avoidant decision making was measured using 5 items from Scott and Bruce's (1995) Decision Making Style measure (e.g., "I put off making many decisions because thinking about them makes me uneasy").
Neuroticism was measured using the 20 neuroticism items from the IPIP neuroticism scale (Goldberg et al., 2006; e.g, "I often feel blue").
Situational dilemmas. We created five situational dilemmas to pose to the participants. Each posed a dilemma that required a choice between three possible alternatives, ranging from pragmatic satisficing, to more time-or labor-intensive options that would provide more complete information for optimizing the quality of the decision. Participants were asked which response they would be most likely and least likely to exhibit. All 5 items are provided in Appendix B.
If a participant selected the maximizing option as "most likely," this counted +1 toward his or her score on that particular situational dilemma. If a participant selected the least maximizing (satisficing) option, it counted as −1. The remaining option was treated as a 0. For the "least likely" question, if a participant selected the least maximizing (satisficing) option for the "least likely" answer, this counted +1 toward his or her score on that particular situational dilemma. If a participant selected the maximizing option, it counted as −1 toward the score for that dilemma. The remaining option was treated as a 0. These option scores were then summed for each individual dilemma, yielding a 5-point scale (+2 to −2), consistent with the scoring of dilemmas of a similar format from situational judgment tests (e.g., Oswald, Schmidt, Kim, Ramsay, & Gillespie, 2004). 1

Behavior reports.
Participants were asked several questions about the extent of their behavior to seek and select the best possible option in typical past life events, such as the purchase of a vehicle, selecting which college 1 In order to verify the scoring of response options, we correlated each of the three response option scores with an overall situational dilemma score (the mean of the 5 items). This was done separately for "most likely" and "least likely" responses, yielding six "option-total" correlations for each item. Although the total score was not sufficiently reliable to use as a scale score, this analysis nevertheless provided a check of whether each response option was indeed coded correctly. We expected that the option considered most maximizing (always "a") would be most positively correlated with the total score when endorsed as "most likely," and most negatively correlated with the total when endorsed as "least likely." The option considered least maximizing (most satisficing, always "c") would be expected to exhibit the opposite trend, and the middle option ("b") was expected to fall between the two options as endorsed for the "most likely" and "least likely" responses. These correlations had the expected pattern for all of the items. to go to, or which apartment to rent. Seven items were piloted, and 5 demonstrated sufficient reliability to combine into one score (α = .72). The retained items are provided in Appendix C, along with their results. The open-ended items were transformed using the natural logarithm. Log transformed values were standardized prior to computing the scale score in order to ensure each item contributed equally to the overall score. Higher scores on this Behavior Reports Scale reflected more information seeking and hence more maximizing. Lower scores indicated less information seeking and thus more satisficing.

Results
Descriptive statistics and intercorrelations among all measures are provided in Table 1. Coefficient alphas are presented in the main diagonal. Some noticeable differences emerged between Schwartz et al.'s (2002) Maximization Scale (MS) and our revised Maximizing Tendency Scale (MTS). First, the MTS demonstrated substantially greater internal consistency reliability. Although the MS had higher internal consistency here than reported by Highhouse and Diab (2006), almost half of the items had corrected item-total correlations lower than the most liberal conventional (>.25) standard. Coefficient alpha reliability for our new MTS is .80. The elimination of any item would decrease the internal consistency of the measure, and the corrected item-total correlations range from .35 to .63.
Second, The MS showed much stronger relations with maladaptive personality and decision-making constructs. The MS correlated positively with Indecisiveness, Avoidance, Regret and Neuroticism, and negatively with Life Satisfaction. The only scale that our MTS correlated with was Regret, but this correlation was lower than observed for the MS (r = .27 for MTS; r = .45 for MS). These differences in the magnitude of the correlations using the MTS versus the MS were all statistically significant (p < .05).
Finally, the MTS showed more, and stronger, relations with the behavior reports and situational dilemma criterion measures. The MS correlated significantly with three of the six measures (mean significant r = .18), while the MTS correlated significantly with 5 of them (mean significant r = .24). The correlation for the MTS exceeded the MS in every instance. This difference is clearest for the 5-item Behavior Reports Scale, where the MS correlated .08 (n.s.) and the MTS correlated .21 (p < .05). 2 2 We conducted an analysis on just the three items common to the MS and MTS. Alpha was .65 (vs. an alpha of .80 for the MTS). The pattern of correlations, shown below, is similar to the MTS. Although the 3item scale dubbed as "high standards" by Nenkov et al. (2008) seems to function well, we believe that the MTS combines bandwidth, reliability, and content validity. As such, it gets more at measuring maximizing per

Discussion
The title of this article poses the question: Are maximizers really unhappy? Our results suggest that -aside from being more prone to experiencing regret -maximizers are just as happy as satisficers, and are no more likely to be indecisive, avoidant, or neurotic. In light of the shortcomings of Schwartz et al.'s (2002) Maximization Scale, we developed a theory-based measure of maximization tendency. This new scale had strong measurement properties, did not generally correlate with maladaptive traits, and allowed for construct valid inferences. We found that Schwartz et al.'s measure had poorer measurement properties, correlated with maladaptive traits, and was less predictive of behavioral outcomes (i.e., as assessed via situational dilemmas and behavior reports).
Although the Maximization Scale does not appear to be measuring what it is intended to measure, it is capturing reliable variance. This is evidenced by research showing, for example, that people who score high on it obtain jobs with higher salaries (Iyengar et al., 2006). Also, despite the poor internal consistency for a 13-item scale, the Maximization Scale correlates with a number of attitudes and traits. What, therefore, is this scale measuring? The most that we can say is that it is a broad band-width measure that taps elements of avoidance, indecision, and emotional instability. This is not the same thing as saying that people who maximize have these dysfunctional characteristics. Our study calls into question the validity of inferences based on scores from the Maximization Scale.
The renewed focus on hedonic experience in judgment and decision making research suggests that one's state of well-being is an important criterion for evaluating good decision making (e.g., Kahneman, Diener, & Schwarz, 1999). It is especially important therefore to critically evaluate research that suggests any one method of decision making leads to unhappiness. In contrast to Iyen-gar et al.'s (2006) conclusion that maximizers get higher paying jobs but are less happy, Crossley and Highhouse (2005) found that job seekers who reported engaging in a careful and deliberate decision process were more satisfied with their jobs than people who used a more haphazard or intuitive approach. Similarly, Kmett, Arkes, and Jones (1999) found that graduating high school students who were forced to use a decision aid that made them carefully consider college attributes were more satisfied with their ultimate choice than students in the control group. We need a better understanding of decision styles and their consequences for choice satisfaction. This article presents a valid measure of maximizing tendency that will be useful in this endeavor. Corrected item-total correlations are provided in parentheses.

Appendix B: Situational dilemma items
Please read each of the following scenarios and the corresponding behaviors. Try to clearly visualize each scenario as you read it. Then, indicate which behavior you would be MOST LIKELY to do and the behavior you would be LEAST LIKELY to do by indicating the appropriate response option (i.e., a, b, or c).
1. Imagine you are at the car dealership and you have found a car that you really want at the right price, except that it is not in your ideal color. Getting the ideal color requires waiting a month for it to come into this dealership, or driving far away to another dealership and renegotiating a deal.
(a) You buy the car anyway because you need to buy a car soon.
(b) You wait until the color that you want becomes available.
(c) You go to more dealers to see if they have the color that you want.
Which behavior are you MOST LIKELY to do?
Which behavior are you LEAST LIKELY to do?
2. You go shopping for clothes because you have a formal event coming up this weekend. You walk into a store and find something that you like. You try it on, and it fits well. You can also afford to buy it.
(a) You buy the clothes because you need them for a coming event, and you feel satisfied with the decision you made. (b) You buy the clothes because you need them for a coming event, but you wonder whether you made the right decision later. (c) You check out more stores to see if you might like something else better, for this was the first store you walked into.
Which behavior are you MOST LIKELY to do?
Which behavior are you LEAST LIKELY to do?
3. You are currently working. Although you are satisfied with your job, you feel that you can find a better one.
(a) You stay in your current job because you like it. (b) You don't actively look for other jobs, but you make sure you pay attention to anything that comes your way. (c) You actively look for other jobs because you feel that there must be a better one out there.
Which behavior are you MOST LIKELY to do?
Which behavior are you LEAST LIKELY to do? 4. You have to find housing for next year pretty soon. You are living in an apartment that you like. However, there is another apartment that you like more and you really want to live there. Although there is a good chance that the apartment will be available, you won't know for sure until after the deadline for signing the lease for your current apartment.
(a) You sign the lease for your current apartment because you don't want to risk moving for no good reason. (b) You wait to see if you can get the apartment that you really want even though you might not get it. (c) You try to find other housing options similar to where you really want to live.
Which behavior are you MOST LIKELY to do?
Which behavior are you LEAST LIKELY to do? 5. You apply for graduate school. You apply to 5 programs that you think are good programs. You would be more than happy to attend any of the 5 schools, but you do have a number 1 choice school. You get into 2 programs, you haven't heard from 3 other programs, one of which is your number 1 choice school. The schools that accepted you may not award you funding if you don't provide an answer soon.
(a) You accept one of the available offers, and you feel satisfied with your decision because all the 5 programs are good.
(b) You accept one of the available offers, but wonder whether you made the right decision later.
(c) You wait until you find out about the programs that you haven't heard from yet.
Which behavior are you MOST LIKELY to do?
Which behavior are you LEAST LIKELY to do?
Note: One response (always "a") was the least maximizing (satisficing) and another response (always "c") was the most maximizing.