This example is highly stylized, and the outcome depends on predispositions within the relevant population. We could easily imagine a group for which a misdemeanor conviction would be decisive in joint or separate evaluation; we could even imagine a group for which such a conviction would be especially weighty in joint evaluation.12 Shrewd political consultants should be alert to the possibility that one or another mode of evaluation is in their candidate’s interest. I will return to this point, because it raises the specter of manipulation (and poses a challenge for some accounts of rationality). The only point is that in choosing among candidates, as among products, people’s preferences may shift as between the two modes.
Global Evaluation and Real Life
It should be emphasized that joint evaluation is not global evaluation, understood as evaluation of all relevant options with all relevant characteristics. In the cases just given, joint evaluation has two distinctive features: First, the two options vary along only two dimensions. Second, there are only two options. In real life, any option has a large assortment of characteristics, some of which will be evaluable in separate evaluation and some of which will not. And in real life, there are usually numerous options, not merely two.
For these reasons, experiments involving joint and separate evaluation should be seen as mere approximations of actual decisions, illuminating their characteristics. In well-functioning markets, we might be inclined to assume that something like global evaluation is involved. But that inclination should be resisted. Recall the baseball card study, involving an actual market, in which experienced traders show a kind of reversal. Some options are, in fact, evaluated separately, or close to it; people focus on purchasing a product or not doing so. It may take effort, and exposure to a kind of friction, to focus as well on alternatives; people may avoid “comparison friction.”
The point helps explain why global evaluation is usually a mere thought experiment. People cannot easily hold in their minds an entire option set, even if we could agree on how to define it (microwave ovens, books about time travel, hybrid cars). To be sure, many choices involve multiple evaluation rather than joint evaluation, especially when multiple options are squarely faced in front of people (e.g., at supermarkets or drug stores).
It is also important to see that in the cases given, the selection of one other option in joint evaluation is hardly random. It is specifically designed to make one important characteristic more evaluable than it was before or evaluable when it was not evaluable before. For dictionaries, CD changers, politicians, and baseball cards, we could imagine, in joint evaluation, a wide range of options that would accentuate, and make evaluable, one or another characteristic. As we shall see, there are ample opportunities for manipulation here. We could easily imagine tests not of joint evaluation, but of evaluation of three, four, or forty options.13 The fact that joint evaluation is not global evaluation, or close to it, also has normative implications, to which I will return.
Note as well that in actual life, there is a continuum between joint and separate evaluation, rather than a sharp dichotomy.14 In some cases, people explore an option entirely on its own, but in other cases, another option, or two or more, are in some sense in the background. People might go to a store to purchase a cell phone with a clear understanding that other cell phones are available. In some cases of joint evaluation, two or more options are simultaneously visible, but in others, the chooser has to do some work to make a comparison; they are not quite side by side. Efforts to reduce comparison friction can be helpful insofar as they eliminate the need for that work. The general point is that the line between joint and separate evaluation might not be so crisp.
Is Less More?
Preference reversals between joint and separate evaluations can be found in multiple domains.15 It is possible to predict when they will occur and when they will not. As a first approximation: if an option has some characteristic X that is (1) difficult to evaluate in separate evaluation, (2) much easier to evaluate in joint evaluation, (3) dominated by characteristic Y in separate evaluation (solely because of the problem of evaluability), and (4) deemed to be more important than characteristic Y in joint evaluation, then we will see a preference reversal. Sometimes conditions 1 through 4 arise without any conscious efforts by sellers. But sometimes they are deliberately engineered by those whose economic, political, or legal interests are at stake.
There is a relationship here to extremeness aversion, which also produces a kind of preference reversal and which also can be manipulated.16 For example, suppose that I prefer option A, a small piece of chocolate cake, to option B, a medium-sized piece of chocolate cake, in a pairwise comparison. But if I am offered three options—those two sizes and also a large piece—I switch to the medium-sized piece and thus prefer option B to option A. The reason is a kind of heuristic, sometimes understood as a compromise effect: choose the middle option. Extremeness aversion also makes people susceptible to manipulation. A seller might deliberately add to options A and B some not-so-attractive option C, so as to shift people from option A to option B. They might exploit the compromise effect so as to encourage choosers to select the more expensive option. Politicians might do the same thing.
With respect to joint and separate evaluation, the normative questions—my central focus here—remain insufficiently understood. When is joint evaluation preferable to separate evaluation and vice versa? By what criteria? Might both modes of evaluation be subject to characteristic problems?
Resisting Temptation
Let us stipulate that for consumption choices, the question is what will improve consumer welfare—and let us bracket the hardest questions about what, exactly, that means. Even with that bracketing, it is tempting to conclude that joint evaluation is better simply because it makes one or more relevant considerations easier to evaluate. That is indeed an advantage. Consider the cases given previously: in all of them, an important variable was discounted or ignored simply because people did not know how to evaluate it. Joint evaluation supplied relevant information, and the variable received the attention that people thought that it deserved. The baseball card example is the most straightforward. In joint evaluation, people could see that it is better to have more than less.
But this conclusion is much too simple. First, joint evaluation might make salient a difference that does not much matter in an actual experience.17 For dictionaries, more words are better than fewer (within limits), but how much better? It is possible that for most users a ten-thousand-word dictionary is as good as a twenty-thousand-word dictionary. It is also possible that a torn cover is a constant annoyance. Second, life generally is lived in separate evaluation. (I will raise complications about this point in due course.) It might greatly matter whether a CD changer holds five or twenty CDs. And in terms of listening experience, the difference between 0.003 percent and 0.1 percent might not matter at all. The bare numbers do not tell us. To be sure, a gain of four thousand jobs is significant, and in the case of congressional candidates, joint evaluation does seem better. But if a misdemeanor conviction is predictive of corruption or of misconduct that would cause real harm, then joint evaluation might lead to some kind of mistake.
We can therefore see some choices in joint evaluation as reflecting a disparity between “decision utility” and “experience utility”18 or (more simply) a “hedonic forecasting error,”19 in which people make inaccurate predictions about the effects of their choices on their subjective welfare. There is a specific reason: joint evaluation often places a bright spotlight on a characteristic that does not much matter.20 Here is an intuitive example. Jones is deciding between two houses:
House A: very large, with a long commute to work
House B: large, with a short commute to work
In separate evaluation, Jones might well be willing to pay more for house B. After all, it is large, and the commute is short. In joint evaluation, Jones might well favor house A. A very large house seems a lot better than a large house, and p
erhaps the commute will not loom particularly large in his decision. This is a hedonic forecasting error in the sense that Jones is undervaluing the day-to-day inconvenience of a long commute. He is inattentive to that inconvenience because he is focused on something more immediate: the concrete difference in size.
Do these points apply outside of the context of consumer choices? Suppose that the goal is to make good personnel decisions. It is reasonable to suppose that joint evaluation is better: if an employer is presented with two or more options, it might focus on attributes that greatly matter, and so the quality of the decision would be higher in joint evaluation. Consider the following:
Potential employee A: strong record, good experience, recommended by a friend, attended the same college as the CEO
Potential employee B: exceptional record, exceptional experience, recommended by a stranger, attended a superb college with no alumni at the company
We could easily imagine a preference for potential employee A in separate evaluation. But in joint evaluation, potential employee B looks better—and might well be chosen. Here too, however, it is important to be careful. On certain assumptions about what matters to employers, separate evaluation could produce better outcomes. Perhaps the chooser cares greatly about a recommendation from a friend, and in joint evaluation that turns out to be decisive. Or perhaps the chooser cares greatly about cultivating the CEO, and hiring someone who went to the same college would seem to do the trick. If we stipulate that potential employee B is better, then joint evaluation might produce mistakes.
In this light, we can raise a question about the view that in separate evaluation, people make emotional judgments, and that in joint evaluation, their judgments are more deliberative.21 To be sure, it is possible to devise situations in which that is true, leading to the reasonable suggestion, supported by data, that “emotions play too strong a role in separate decision making.”22 But the opposite may also be true. Consider the following, understood as options for dating:
Person A: great personality, fun, always kind, attractive
Person B: good personality, incredibly fun, usually but not always kind, devastatingly attractive
I have not collected data, but we can imagine people who would prefer person A in separate evaluation but person B in joint evaluation. Is separate evaluation more emotional and less deliberative? The same question raises a concern about the closely related view that in separate evaluation, people focus on what they “want,” whereas in joint evaluation, people focus on what they “should” want (or do).23 The view fits with some of the data. Consider this example:
Option A: improve air quality in your city
Option B: get a new cell phone
It is possible that in separate evaluation, people would be willing to pay more for a new cell phone, but that in joint evaluation, they would be willing to pay more for air quality.24 We could easily devise situations in which separate evaluation triggers a desire, while joint evaluation focuses people in a way that produces a response connected with their normative judgments. But we could devise situations that yield exactly the opposite outcomes. Here is a candidate:
Option A: contribute to pay down the national debt
Option B: take your romantic partner to a romantic dinner
Among some populations, option A would produce a higher figure in separate evaluation, and option B would do so in joint evaluation. Everything depends on the antecedent distribution of preferences and on what becomes the focus in joint evaluation—a point to which I now turn.
How to Win Friends and Influence People
In either separate evaluation or joint evaluation, sellers (and others) have identifiable routes by which to influence or to manipulate choosers. The appropriate design should be clear.
Using Separate Evaluation
In separate evaluation, sellers should show choosers a characteristic that they can easily evaluate (if it is good) and show them a characteristic that they cannot easily evaluate (if it is not so good). In these circumstances, the option will seem attractive even if it has a serious problem. In fact, sellers should choose separate evaluation if they can whenever this presentation of options is feasible.
I have emphasized that in real markets, options do not simply have two characteristics; they have an assortment of them. Recall that much of the experimental evidence is a radical simplification. But the essential point remains. In fact, it is fortified. Sellers (and others) can choose among a range of easily evaluable options (appealing ones) and display a range of others that are difficult or impossible to assess (not-so-appealing ones). It is well known that some product attributes are shrouded, in the sense that they are hidden from view, either because of selective attention on the part of choosers or because of deliberative action on the part of sellers.25 A problem with evaluability belongs in the same category; it is a close cousin. Nothing is literally shrouded, but choosers cannot make much sense of the information they are given. In real markets, they might be able to find out. In fact, that might be easy. But because of comparison friction—defined as people’s unwillingness to obtain comparative information, even when it is available—a problem of evaluability in separate evaluation might persist.
Using Joint Evaluation
In joint evaluation, by contrast, sellers should allow an easy comparison along a dimension that is self-evidently important to choosers, even if the difference along that dimension matters little or not at all to experience or to what actually matters. The trick is to highlight a characteristic on which the product in question looks good, or better, or great. It is best, of course, if in reality, that characteristic does matter a great deal. But even if it does not, use of joint evaluation, putting a spotlight on the characteristic whose appeal is heightened by way of comparison, is a good strategy. The point holds for consumer products and political candidates, but it should work for options of all kinds.
An imaginable example: A new computer is introduced with a reduction in size and a superb screen; it is far lighter and it has a far better screen than that of existing models. When the existing models are compared with the new one, the difference in weight and between the screens is easy to see. They stand out. Assume, however, that consumers have no problem with the existing weights and screens; they find them inferior only in joint evaluation. Assume finally that the new computer has to sacrifice along an important dimension to have the spectacular screen—say, its battery time is much reduced or its keyboard is much less comfortable to use. It is easy to imagine that in joint evaluation, people will choose (and purchase) the new computer, but in separate evaluation, they will have a much better experience with the old one.26 A more general way to put this point is to suggest that in joint evaluation, people might overweight attributes that are simple to evaluate or that trigger a strong response, including a visceral one.27
Net Welfare Gains
These points should be sufficient to show that there is no abstract answer to the question whether joint evaluation or separate evaluation produces better decisions on the part of consumers. Neither is ideal, and both might lead to mistakes. Both modes of evaluation have characteristic problems and pathologies. With separate evaluation, the most important of these is incomplete information, producing a failure of evaluability. With joint evaluation, the most important is a focus on one characteristic, producing excessive salience.
In the baseball card case, joint evaluation is obviously better; more is better than less. But in the dictionary and CD cases, we need more information to know whether joint evaluation is better. The same is probably true for the case of the congressional candidates. In the context of consumption choices, the question is whether the ability to evaluate a characteristic of an option and the salience of that characteristic produces decisions with net welfare gains for choosers in joint evaluation. If the factor that is evaluable in joint evaluation does not matter much, and if the factor that is downplayed or ignored in joint evaluation is actually important, then separate e
valuation is likely to be better.
There is a psychological wrinkle here. It is tempting to think that life is lived in separate evaluation, which helps explain why and when joint evaluation leads to mistakes.28 But is life really lived in separate evaluation? The answer depends on the context and on the person. Suppose that a consumer buys a dictionary with ten thousand words or a computer with a terrific keyboard but a less-than-ideal screen. If that consumer does not think about dictionaries with more words or computers with ideal screens, her purchasing decision will be shown to be correct. If her attention is not spent on issues of comparison, she will live in separate evaluation. But suppose that she does think about those products, at least some of the time, and focuses on her dictionary with limited words or her computer with a screen that, while excellent, pales by comparison with a screen that she rejected. If so, she is living, to a greater or lesser extent, in joint evaluation.
People vary in their propensity to engage in product comparisons. The universe of comparison products defines the frame of reference by which many of us define our experiences—which helps explain why product improvements may impose serious welfare losses on people who had been enjoying their goods a great deal.
Law and Policy
I have emphasized that preference reversals between joint and separate evaluation have been found in the domains of law and politics as well. The domain for analysis here is very large, and research remains in an early state. One of my goals here is to vindicate the suggestion with which I began: the choice between joint and separate evaluation depends on the relevant task and the underlying goals of the enterprise. For that reason, we might well end up rejecting both modes of evaluation. We might simply block some grounds for action (such as racial prejudice). We might seek to rely on global evaluation. We might want to use some kind of algorithm. If, for example, optimal deterrence is the goal, neither joint nor separate evaluation is likely to do a great deal of good.
How Change Happens Page 22