Before experimental philosophy stormed the philosophical world, empirically minded social choice theorists were cross-checking the conclusions of political theorists against the expressed attitudes of survey respondents. Yet, despite a growing receptivity to the use of empirical methods in addressing philosophical questions, the influence of this important body of work on political philosophers has been relatively restricted. Empirical Social Choice sets out to change that.

Three of the six chapters (1, 2 and 6) of Wulf Gaertner and Erik Schokkaert's slim volume are dedicated to an overview of the methods of empirical social choice, as distinguished from the more axiomatic approaches of formal social choice, welfare economics and political philosophy. In these studies respondents (often undergraduates) are typically presented with a story that encapsulates a distribution problem. Questions are then used to elicit the respondents’ intuitions as to how best to resolve that problem. The normativity here is important, as the questions are intended to uncover participants’ implicit notions of the just or good. For example, one of Gaertner's studies asked students if it would be better to allocate a sum of money to helping a handicapped person become more independent or to furthering the talents of an intellectually gifted child. Rawls's difference principle would have us prefer a policy which would benefit the least well off. Variants of the question increased the number of gifted children that would benefit, and the study was repeated, across time and in different counties, to check for evidence of attitudinal change or universality. In the Gaertner (Reference Gaertner, Gaertner and Klemisch-Ahlert1992) study 92.3% of respondents at the University of Osnabrück chose to allocate the funds to the handicapped person, although this support dipped as the number of gifted children who benefited increased, as respondents became more knowledgeable about economic efficiency, and according to the country in which the survey was conducted (in the Baltic countries only 65.7% of respondents chose to allocate funds to the handicapped person). Prima facie Gaertner's findings suggest some accordance between the expressed attitudes of the respondents and Rawls's difference principle. I'll note here that it seems possible that some respondents might have chosen to support the gifted child with an eye toward the benefits incurred by those who are less well off in the future. A society which systematically neglects its most gifted members might atrophy. Below I'll take a closer look at the gap between the empirical findings and the theories upon which those findings are supposed to cast light.

As mentioned, Empirical Social Choice is principally an exercise in methodological advocacy. Chapter 2 includes a manifesto of sorts, presenting arguments intended to underscore the relevance of empirical social choice to political theorizing. Many of these arguments echo those which have been advanced in favour of experimental philosophy (Knobe and Nichols Reference Knobe and Nichols2008, ch. 1). Of these, the most interesting is Gaertner and Schokkaert's appeal to the role that such studies might play in a process of reflective equilibrium. Tension between the results of a survey and our theories may prompt revision where accordance can be the basis for confidence. Unless one holds that the normative convictions of the folk are largely matters of prejudice, unmoored from the findings of theoretical reflection, this vast body of findings should be of interest to political philosophers.

The remainder of the text collects and summarizes studies which relate, in various ways, to problems of distributive justice. A linear reading of chapters 3–5 is exhausting: the authors tear through 30+ studies with only minimal attempts to connect the studies to each other or to the philosophical literature. However, as the summaries are well organized and indexed, the text is quite suitable as a reference book. I will not attempt to provide more than a bare-bones overview of these chapters, opting instead to focus on Gaertner and Schokkaert's treatment of one of the reviewed studies.

Chapter 3, titled ‘Traditional Questions in Social Choice’, begins with studies that appear to indicate that respondents vary their distributional strategies according to whether the goods in question are construed as needed or merely wanted. I'll discuss one of these studies in considerable detail below. In section 3.3 Gaertner and Schokkaert present a series of interesting studies which consider, following Harsanyi, two positions from which questions about justice might be addressed. For Rawls, the veil of ignorance guarantees impartiality by depriving parties of any information which might distort their judgements of fairness. Nevertheless, deliberators adopt an attitude of self-concern, as they expect to live within their chosen institutions. This involved stance might be contrasted with the viewpoint of the umpire or outside evaluator, who does not otherwise participate in the society for which she is making policy. Amiel et al. (Reference Amiel, Cowell and Gaertner2009) asked respondents to distribute a basic good in a hypothetical society. Questions differed only in the stance they asked the respondent to adopt. Respondents routinely chose more equitable distributions when adopting the umpire role. Gaertner and Schokkaert suggest that these studies diminish the attractiveness of the veil of ignorance as an approach to thinking about justice. In one of the few non-questionnaire studies presented in the book, Section 3.4 looks at attempts to simulate a Rawlsian Original Position. Frohlich et al. (Reference Frohlich, Oppenheimer and Eavey1987) put participants behind a veil of ignorance, asking them to arrive at a consensus as to a distribution scheme according to which an eventual allocation would be made. Contrary to Rawls, only a small percentage of the groups chose the maximin principle, with the vast majority choosing utilitarianism with a floor, which strove to maximize the total allocation while guaranteeing every participant a minimum. Finally, section 3.5 looks at respondents’ propensity to accept the Pareto principle, which holds that some allocation should be preferred if it increases the standing of one individual without harming anybody else. While assumed to be largely uncontroversial, Beckman et al. (Reference Beckman, Formby, Smith and Zheng2002) found that 28.8% of respondents, across nations, opposed Pareto improvements when the recipient occupied a higher income position, although this opposition dipped to 6.4% in the USA.

It is standardly accepted that welfarist social choice is impossible (Arrow) and would otherwise be undesirable (Dworkin); reasonable social preferences can't and shouldn't be derived from individual preferences alone, without additional information. Chapter 4, titled ‘New questions: fairness in economic environments’, looks at questionnaire studies that bear on the issue of the kinds of additional information which may be relevant to social choice. For example, it's long been recognized that welfarist approaches are insensitive to desert or personal responsibility (Feinberg Reference Feinberg1970). In section 4.1 Gaertner and Schokkaert review several studies (Konow Reference Konow1996, Reference Konow2001; Gaertner and Schwettmann Reference Gaertner and Schwettmann2007) concerning the role that responsibility plays in the determination of a just allocation. Not surprisingly, respondents were largely comfortable with the unequal distribution of a good so long as the distribution corresponded to factors which were under the subject's control (e.g. effort). When productivity differences were due to exogenous factors, such as a physical disability, respondents tended to distribute the goods equally. Section 4.2 looks at the so-called ‘claims problem’, where respondents are tasked with distributing a good when the resources fall short of what was promised. For example, Herrero et al. (Reference Herrero, Moreno-Ternero and Ponti2010) asked respondents to divide money obtained from a bank's liquidation among its creditors when that amount falls short of the creditors’ original deposits. Respondents by in large divided the money in proportion to the size of their original deposit (rather than, say, equally), although other studies show that egalitarian solutions become more popular as the inequality in claims increases or the amount to be divided decreases. Section 4.3 looks at the effect that framing issues can have on justice evaluations: respondents in Gamliel and Peer's study (Reference Gamliel and Peer2006) were more tempted by non-egalitarian principles when formally identical allocation problems were framed in terms of the distribution of benefits rather than harms.

Chapter 5 focuses on studies concerning distribution and rationing problems in the special case of healthcare. For example, 631 Swedish politicians were asked by Lindholm et al. (Reference Lindholm, Emmelin and Rosen1997) whether they would rather fund a programme which had a higher overall efficiency in reducing the mortality rate of a given disease or a programme which more efficiently targeted the subpopulation which is more deeply affected by that disease. So long as the reduction in overall efficiency was quite low (less than 10%), a majority of respondents chose the programme which benefited the subpopulation. Given this chapter's relatively narrow reach I will not make any further attempt to summarize the findings of these studies.

A persistent difficulty concerns the applicability of the studies to the evaluation of more formal theories of justice. While, as the authors repeatedly emphasize, the book is intended to prompt conversations regarding the significance of the findings to larger theoretical issues, the interpretive suggestions are often so quick and superficial that they barely suffice as a foil against which meaningful conversation can take place. Chapter 3 begins with an overview of Yaari and Bar-Hillel's landmark study (Reference Yaari and Bar-Hillel1984), where random undergraduates were asked to justly distribute hypothetical goods – grapefruit and avocados – between two recipients according to various possible distribution schemes, including utilitarianism and the maximin. On one variant of the question, the goods were posited to satisfy biological needs of the intended recipients. I'll reproduce this question in full:

In another variant of the question, the situation was recast so that, while formally identical, the focus was on tastes, not needs. Where Jones is willing to pay $1.00 per pound for grapefruit but is uninterested in avocados, Smith is willing to pay $0.50 for either fruit. The price each recipient is willing to pay presumably corresponds to how much utility each expects to receive from the fruit.

On the needs question Yaari and Bar-Hillel found that 82% of respondents chose the maximin over overall utility. That is, the vast majority gave Jones 8 grapefruits and Smith 4 grapefruits and all 12 avocados, a solution which was responsive to the recipient's different metabolic needs and yielded equal amount of vitamins for both recipients. However, only 28% of undergraduates chose this solution on the tastes question; 35% opted to maximize overall utility by distributing each fruit to the highest bidder—12 grapefruits to Jones and 12 avocados to Smith (24% chose a middling position).

Gaertner and Schokkaert draw a number of inferences. Firstly, the study affirms that our moral intuitions are sensitive to the distinction between needs and tastes. Secondly, and much more radically, they make the strong claim that welfarism in general ‘seems to be decisively rejected’ by the study's respondents. Third, the study speaks against the broad application of the Rawlsian maximin principle. As the first claim seems correct, I'll focus on the second and third.

Welfarism, at least as classically understood, is the view that the goodness of a state of affairs is strictly a function of individual subjective utility. Utilitarianism, which seeks to maximize the total amount to utility, is one version of welfarism. One may argue against welfarism by suggesting that non-utility information (e.g. liberty) nevertheless bears on our considerations of the good. So how does the intuitive distinction between needs and tastes speak against welfarism? Gaertner and Schokkaert don't say; or rather they make the question begging assertion that questions concerning needs and tastes are ‘formally identical from a welfarist perspective, but differ from each other with respect to situational features’ (31). Perhaps they are referring to the arguments of Scanlon (Reference Scanlon1975), who contrasts ‘urgency’ and utility. However, Sen (Reference Sen and McMurrin1980) and others have argued that the welfarist need not see urgency as independent of utility. Goodin, for example, argues that the value of needs is instrumental, and so cannot exceed the value of the wants they serve (Goodin Reference Goodin1985: 621). Obviously, needs should be distinguished from tastes. But the respondents’ propensity to treat needs differently may simply reflect the recognition that vitamins serve a wider variety of subjective preferences than do fruits which merely taste good.

Assume for the moment that the study does in fact provide prima facie evidence for non-welfarist positions such as Rawls's theory of justice. Gaertner and Schokkaert think that the study also calls into question the applicability of Rawls's difference principle: ‘It is interesting to see that the more egalitarian maximin criterion loses its attractiveness if we move from a situation in terms of needs to a situation in terms of tastes, while the more efficiency-oriented utilitarian criterion gains popularity’ (36).

While Yaari and Bar-Hillel abstract the maximin principle from Rawls's overall theory of justice it will be helpful to reframe those findings within Rawls's larger theoretical apparatus. The distinction between needs and tastes appears to track Rawls's distinction between primary social goods and other sorts of goods. Construed as such, a difficulty immediately presents itself. Rawls begins with the assumption that since individuals are the authors of their own final aims, a theory of justice must concern itself only with those goods which are particularly useful in the pursuit of those various life plans. These means include primary goods such as income and wealth, defined by Rawls as those ‘various things which he would prefer more of rather than less’ regardless of the particular content of a given life plan (Rawls Reference Rawls2005: 92). Because Yaari and Bar-Hillel stipulate that vitamin F is required by Jones and Smith, we can assume that their various life plans would be compromised without it. But there are other ways vitamin F seems importantly different from a Rawlsian primary good. In general, vitamins are subject to threshold effects in a way that income or wealth are not. Perhaps without a minimum intake Jones and Smith would be stricken, unable to vigorously pursue their respective interests. Would respondents still apply the maximin principle if they discovered that, in doing so, neither Smith nor Jones would be able to get out of bed? Another difference between vitamins and income: where an increase in wealth tends to better enable the pursuit of one's aims, vitamins do not typically work this way (indeed, too much of a given vitamin tends to be toxic). The thought experiment nevertheless induces the respondents to think about vitamins as we might normally think about income (‘the more the better’). But if one is able to consume vast quantities of vitamin F, how, exactly, does this tend to promote the realization of various life plans? Due to the strangeness of this vitamin and its obscure relation to possible life plans, the survey respondents appear to know less (or worse, as the respondents are asked to set aside their understanding of how vitamins normally work) than those deliberating about income from behind the veil of ignorance, seriously blunting the study's scope of applicability to Rawls's theory. Moreover, so much of the Rawlsian framework is left out (e.g. the link between needed goods and a life plan) that different respondents might be working with different conceptions of how the vitamin is needed. The results, while purporting to buttress the difference principle, might just be a reflection of how undergraduates proceed in a confrontation with the uncanny.

Even more puzzling is the significance of the findings regarding ‘tastes’. In this case, as mentioned, respondents tend to distribute fruit to the highest bidder. From within a Rawlsian framework, ‘tastes’ might refer to goods which aren't basic but rather connected to the life plans of specific individuals. Thus, the fact that Jones is unwilling to attach any price to avocados might be construed as a reflection of the fact that, somehow, avocados play little role in the realization of his final aims. But in this case, the fact that the respondent doesn't apply the maximin principle is very much in line with what Rawls would have us expect. Why should non-basic goods be distributed to people for whom those goods are of little value? But while Gaertner and Schokkaert appear to take the findings as a challenge to the difference principle, Rawls rejects the possibility of a universal regulative principle (Rawls Reference Rawls2005: 29); when it comes to the realization of our life plans, something like bounded utilitarianism might in fact capture the rationality of our actions. There is no incompatibility between the findings of the study and Rawls's deployment of the difference principle.

The complaint, here, is not with Yaari and Bar-Hillel's study per se. This often-cited study more or less launched a new and important research programme which anticipated the methods of Experimental Philosophy by more than decade. These remarks are rather intended to illustrate a shortcoming with Gaertner and Schokkaert's book: there remains consistent and considerable gap between the empirical findings and the sophistication of the theories the findings purport to be about. The authors typically dedicate a sentence or two to the theoretical implications of a given study before moving on. They present these studies with the intention of provoking these sorts of discussions: ‘A good questionnaire will reflect as much as possible the subtleties of present theoretical modeling. Surprising empirical results may then stimulate further theoretical thinking’ (199). But without being more explicit about the theories the questionnaires are intended to test, the authors sometimes misconstrue what even counts as a surprising result. While careful and lucid in the presentation of the studies, readers may be left unsatisfied by Gaertner and Schokkaert's cursory attempts to articulate the relevance of these empirical findings.