The Case of Stated Preferences and Social Well-Being Indices

,


I. INTRODUCTION
Should policy goals be directly determined by citizens?This paper reviews new innovative techniques developed by economists that compose measurements of social well-being based on citizens' stated preferences (hereafter: SP).We suggest that these techniques can be useful for public policy because they provide a more direct understanding of what citizens care about, but that they also raise significant problems that we need to be aware of when employing them.In short, we import new economic techniques into political and philosophical contexts and examine them from a broader perspective.
The new techniques we discuss can be viewed as a novel implementation of instruments of direct democracy.We argue that, like many measures of direct democracy, they are prone to serious pitfalls-if not complemented cautiously by other measures, in particular, measures of public deliberation.We argue that attempts to implement the SP techniques should incorporate deliberative methods in multiple stages.This perspective is different from that usually embraced by economists who tend to regard this method as part of a 'social choice' tradition, one which is less focused on political-theoretical considerations.
Following the growing sense that GDP growth is no longer a sufficient indicator for social progress and well-being, recent years have witnessed the development and implementation of both local and international composite indices supposed to represent social well-being (Fleurbaey and Blanchet 2013;Boarini, Kolev, and McGregor 2014;Stiglitz, Sen, and Fitoussi 2009).Within the process of indices' structuring, various aspects of social well-being are selected (for example, education, health, life satisfaction, etc.), represented by indicators, and then combined within an index.This measurement of social well-being can then be adopted as a means for directing public policy (Bache and Reardon 2016).
This indexing process raises many questions concerning both the selection and combination of the indicators.The latter, the combination problem, often referred to as the 'index problem', is a pressing one if we refuse to be satisfied with a dashboard solution that leaves the separate indicators as they are instead of integrating them. 1 The index problem in our context asks: What relative weight should be given to each component that we believe social well-being encompasses? 2 This is a complex question.How can we determine whether health is more important than subjective happiness, and by what degree?Ascribing different weights to the indicators results in different indices and is therefore a process with political implications and significance (Decancq and Neumann 2016;Bache and Scott 2018).
SP methods for solving index problems have recently enjoyed particular attention from economists who have implemented and improved them (Benjamin, Heffetz, et al. 2014;Benjamin, Cooper, et al. 2017;Adler, Dolan, and Kavetsos 2017;Decancq, Van Ootegem, and Verhofstadt 2013;Decancq and Watson 2015).With regard to the index problem, there are instances of governments and international institutions adopting related mechanisms of consulting people about their preferences (Balestra, Boarini, and Tosetto 2018;Barrington-Leigh and Escande 2018, 910).These initiatives utilize up-to-date technologies and scientific techniques to collect SP data relatively easily and elegantly through internet platforms and computer software.
We can illustrate economists' SP exercises by reference to the wellestablished Human Development Index (HDI), which is composed of three indicators: GDP per capita, life expectancy, and schooling.The index problem focuses on the procedure through which the three indicators are combined.Fleurbaey and Blanchet (2013, 14-15) point to the arbitrariness of the existing methods of combination within this index.The SP approach, in this case, suggests going to the people, that is to say, to a representative group of people who are directly consulted, through a survey, about their individual preferences regarding the components.Under particular assumptions, a social scientist may aggregate the data and elicit weights for each of the three components on a social level, and in this way resolve the index problem (Decancq and Watson 2015).
From an economist's point of view this method is appealing.In this paper, however, we discuss the issue from broader political and philosophical perspectives and point to basic limitations in the SP method.These limitations, it will be argued, direct us to complement the method with various deliberation processes.Highlighting these limitations serves to enable the designing of a better process of aggregation and deliberation regarding the index problem.The paper points to the exact parts of the process where the two procedures of aggregation and deliberation should complement each other.
This leads us to an additional contribution our paper seeks to make.The problem of the relationship between aggregating citizens' preferences and using deliberation techniques is a persistent and well-known one in political theory, social choice theory, and public policy.3Should group decisions be determined by individual votes or by achieving some sort of consensus through discussion?If we want to integrate the two, what is the way in which we should do so?Usually, these questions are discussed theoretically from a general point of view, without looking at the specific content of the choice problems encountered by the social groups.Our paper shows that when looked upon from a careful examination of a particular case, we discover new specific difficulties in mere aggregation and new specific functions for deliberation procedures that a general discussion could not have identified.Our discussion thus contributes to the theoretical literature regarding aggregation and deliberation by demonstrating how it can benefit from taking into account particular types of political decisions.
The paper is structured as follows.The second section presents the SP approach in the context of the index problem, highlighting its advantages.The third discusses the prominent (unique and non-unique) limitations of SP methods in the social well-being indexing context from political and philosophical points of view.The fourth section discusses the potential role of deliberation in coping with these limitations and provides the basic trajectories that, in this context, allow for deliberation and preferences aggregation to complement each other to ensure a more fruitful direction of well-being guided public policy.The fifth section provides concluding remarks.

II. THE APPEAL OF SP METHODS IN DETERMINING THE INDEX PROBLEM
Scholars distinguish three categories of methods for solving the social well-being index problem: expert-based weights, data-based weights, and preference-based weights (Watson et al. 2019; see also Decancq and Lugo 2013;Balestra, Boarini, and Tosetto 2018).The latter attempts to elicit individuals' preferences regarding different potential elements of wellbeing and aggregate them, under some explicit presumptions, in order to establish social indexing.
In general, these preference-elicitation approaches to the index problem are considered attractive because they provide us with a practical aggregation/weighing apparatus.Moreover, they are usually perceived, justifiably or not, as non-paternalistic and respectful of people's own judgments (in these respects, they are often considered superior to expert-based/data-based methods).These features have been emphasized recently in the economic and philosophical literature as leading criteria for policy oriented indexing processes (Fleurbaey 2012;Hyabron and Tiberius 2015;Alkire 2016, 625-629).
There are a few possible methods for eliciting such preferences (see for example Watson et al. 2019).Most notably, one can distinguish observational data-'revealed preference' techniques that look at behaviour as evidence of preferences (Bargain et al. 2013;Decoster and Haan 2015)or use an external proxy to elicit preferences (for example, using SWB metrics as in Clark and Oswald 2002;Barrington-Leigh and Escande 2018). 4Alternatively, there are more direct techniques that use surveys to elicit preferences.Amongst the later, we could discern for instance willingness-to-pay surveys and also what are called here 'SP surveys' in which respondents directly rank/weight the alternative components.Thus, SP is but one method among various preference eliciting techniques and among various survey-based techniques (it is referred to here in this narrow sense).
Other preference-based methods and techniques have substantial advantages.For example, willingness-to-pay surveys have distinct strengths, such as their provision of a clear yardstick for comparisons and a potential for increased stability and consistency in the identified preferences (Streimikiene et al. 2019).However, while these techniques-and othersare suitable, SP methods have a special appeal in the social well-being index case, or so it seems.As we shall see, this is because this particular index problem might be interpreted in two distinct ways: as eliciting welfare/well-being or as determining social priorities.With the second interpretation, as a part of a political-democratic procedure, it may seem less appropriate to use monetized or other external benchmarks.For these reasons, recent research has tried to use surveys in which citizens are asked to directly rank well-being components one against the other in order to state their preferences explicitly (Benjamin, Heffetz, et al. 2014;Decancq, Van Ootegem, and Verhofstadt 2013;Adler, Dolan, and Kavetsos 2017;Decancq and Watson 2015).
Our focus of attention on these SP methods is thus not because they are 'better' in eliciting preferences, but because they seem the most straightforward in their emphasis on respecting peoples' attitudes (both as well-being subjects and as citizens with opinions).This feature makes it allegedly more resistant to the bulk of our criticisms.In any case, a great deal of the limitations we highlight here regarding SP procedures are relevant to the other methods of preference elicitation.
When investigating the SP method and its rationale, the following example may be of aid.Benjamin and co-authors (Benjamin, Heffetz, et al. 2014;Benjamin, Cooper, et al. 2017) advance one of the most thorough and comprehensive SP research programs, and so will be used here as a leading example.In Benjamin, Heffetz, et al. (2014), the authors composed a comprehensive list of 136 aspects of well-being, drawing from psychologists, philosophers, and economists, then asked 4,600 US survey respondents directly about their preferences, making pairwise comparisons between marginal changes in aspect α and aspect β.
By using a basic analogy between a social well-being index and a standard consumption-based-index, where well-being aspects substitute for commodities and services, the researchers elicit ordinal well-being functions from the preference data (analogous to ordinal utility functions).5These researchers are thus using well-developed economic aggregation techniques in approaching the social-well-being index problem (Benjamin, Heffetz, et al. 2014, 2703-2707).From the aggregated data they attempt to pool a combined well-being index that reflects the relative importance each well-being aspect receives from the relevant population.Thereby, Benjamin and co-authors are able to estimate the well-being index , 6 where  = ( + , … ,  ( ) represents the quantities of  fundamental well-being aspects.
According to the survey conducted in this research, for instance, American respondents see the happiness of their families (ranked 2nd) and their 'health' (ranked 3rd) as more important for their personal well-being than their physical safety and security (ranked 21st) or their material standard of living (ranked 98th).
Significantly, the researchers add to the survey another sort of question.While the first concerns personal well-being, this second concerns 'policy-vote scenarios', within which "the opening clause 'Imagine you are making a personal decision' is replaced with the clause 'Imagine that you and everyone else in your nation are voting on a national policy issue" (Benjamin, Heffetz, et al. 2014, 2710).
Note how the two kinds of investigation refer to different kinds of judgements and preferences: the first in letting people reflect upon their own well-being ('personal preferences'); the other as a 'direct-democratic' poll regarding what is best for society (eliciting 'citizens preferences').Whereas the former investigation is in common with economists' preference elicitation investigations, which focus solely on representing and aggregating personal well-being preferences (as in Decancq and Watson 2015),7 the latter is in common with real-world initiatives such as the OECD better life initiative, directed to aggregating social priorities. 8n advocate of the SP method may perceive this dual-goals investigation as an advantage of the method.Within our examination of the SP aggregation method and its limitations, however, it is necessary to attentively separate the two kinds of tasks (eliciting personal preferences and political priorities).In the next section, therefore, we will discuss the two SP exercises separately.
In any case, the research as a whole, with its two different directions, is explicitly directed to guiding politicians (Benjamin, Heffetz, et al. 2014, 2699).This is by way of providing an elegant preference aggregation/weighting mechanism and a solution to the index problem, when the aspects ranked higher should be granted more weight.SP's greatest contribution in this context, in the eyes of its advocates, is when competing with expert-based procedures: We believe it is more attractive to rely on what people's own stated preferences suggest about what they themselves care about than to paternalistically rely on the opinions and introspections of "experts" (such as researchers and policymakers) regarding which aspects to track and how to weight them.(Benjamin, Heffetz, et al. 2014, 2702) In the next section we examine this assertion from political and philosophical points of view and point to the essential (unique and nonunique) limitations of SP methods in this context.

III.I. Stated Preferences and Social Welfare
Referring to SP as the appropriate method for eliciting weights for index components conceals implicit assumptions about the very purpose of the index.These presumptions should be carefully considered when specifying the role of preference aggregation.It is important, therefore, to spell out the assumptions about the index itself and what exactly it is supposed to represent.
The two different sorts of exercises of SP that were noticed (reflecting upon personal well-being and expressing social/political priorities) are in accordance with two different presumed goals for the index: tracking welfare and determining normative social/political priorities.In this sub-section we focus on the first and turn to the other in the next sub-section.
In the first case, stated preferences are taken as the appropriate method to extract or even to 'track' welfare.Two basic assumptions about the index are actually hidden here.First, it is assumed that social wellbeing is equivalent to social welfare; in other words, welfarism is assumed-the view that what matters ultimately and exclusively is the aggregation (by a specific rule) of personal 'welfares' (Adler 2019, chapter 1).According to this presumption 'social well-being' (the concept the index is meant to represent) and 'social welfare' are equivalent.Second, it is assumed that 'preferences' are the right information in determining personal welfare.Both presumptions could be debated, and hence could be adopted or rejected by political institutions and the public.
Regarding the first assumption, it could be the case that the role ascribed to the social well-being index is not to track personal welfares alone, but to track also some other conception that goes beyond welfarism, such as 'social cohesion', 'social quality', and so forth (although this is not common among economists, this conception of social well-being is acceptable sometimes by sociologists and philosophers; see Abbott and Wallace 2012;Berger-Schmitt 2000;Wiland 2022).In these cases, it is not obvious that SP is the most relevant method.
Suppose that we do assume that the index should reflect aggregate personal 'welfares'; still, as with other cases of social welfare functions we need to reflect upon the (personal) welfare/well-being conception and measure (Adler 2019).Choosing the measure depends on our conception of personal welfare.As a standard in the literature, we have three theories of personal well-being, or of how well an individual's life is going (Adler 2019;Parfit 1986;Hausman 2011;Crisp 2017).Preference satisfaction is the view that personal well-being is a matter of preferences-realization; alternatively, the experientialist account is the view that personal well-being is a matter of mental experience, such as feelings, pleasures, satisfaction and so on; the third alternative, the objective account, posits particular objective qualities (such as accomplishments and friendships) as constituents of personal well-being, regardless of a person's attitudes.
When we ascribe a welfarist role to the index and take personal wellbeing to be conceptually linked to preferences realization we have good reasons to favour the preference elicitation method in order to inform the social well-being index.Note how for economists the two presumptions about the index-welfarism and preference satisfaction as the relevant standard-come as natural.The two presumptions are a part of microeconomics' raison d'être, and well-grounded in the economic paradigm (Hausman and McPherson 2006, chapter 8).Neither the public nor politicians, however, are obligated to take these presumptions as given.To use Manin's terminology, these presumptions are not "evident, simple, and luminous", and therefore, need to be deliberated in the strong sense of the term (Manin 1987, 347).
In case we deliberately favour the above conception of social well-being, it is still not obvious that we should prefer SP over other preference elicitation techniques.Thus, economists traditionally highlight difficulties concerning measurement and aggregation.A non-exhaustive list of problems would include interpersonal and intrapersonal comparisons (tackling preference heterogeneity-the possibility that different people have different preferences, a difference we would want to somehow reflect), stated preferences stability, and stated preference completeness.How far does the SP method allow these challenges to be overcome better than other techniques (Decancq and Nys 2021)?For example, willingness to pay surveys (alternative type of survey technique) and 'equivalent income' (a type of utility metrics), are regarded as capable of coping with the above problems by using a monetized numeraire (Fleurbaey and Schokkaert 2013;Fleurbaey 2016). 9enjamin , Heffetz, et al. (2014) acknowledge such difficulties and advise, therefore, utilizing other procedures such as SWB and money-metric approaches in particular, as complementary to their own.These methods complement the SP method by enabling the aggregation mechanism to be more precise (Benjamin, Heffetz, et al. 2014, 2732;Benjamin, Cooper, et al. 2017, 81).
This paper, however, highlights the possibility that part of the problems (among them completeness and stability) could be coped with through deliberation procedures.

III.II. Stated Preferences and Politics
We now turn to the other potential use of the method: employing SP for revealing policy priority preferences.This is an investigation in which the task is not to track personal welfare, but to represent the normative views of individuals regarding what is good for the society.Note how welfarism might constitute such a normative consideration in itself (i.e., one may presume that what is good for society is to promote aggregate welfare); however, this is not necessarily the case and there might be other goals to promote.For example, one could take sustainability as a granted weight in a social well-being index, even in cases in which it comes at the expense of current welfare (or at the expense of aggregated social welfare). 10ocusing on SP for this political purpose, we should notice the alternative different scales of the indices: local and universal (McGregor 2018).An index may allow, on the one hand, the specification of the particular manner by which different societies address social well-being.With this local index problem, every society acquires its own solution to the index problem.Alternatively, it may (attempt to) provide a universal yardstick by which we can compare the well-being of different societies and nations, or the same society or nation across different times.11 In accordance with the first goal-specifying a local index-some SP projects refer to the use of the method as an efficient tool for a policy maker when puzzling over criteria for redistribution and when deciding who is the worst-off within a society (Decancq, Fleurbaey, and Schokkaert 2015;Decancq and Schokkaert 2016). 12The second-universal-goal presents particular challenges for SP methods (for such an attempt, see Balestra, Boarini, and Tosetto 2018).In case we refer to the community of the whole world as one society, using SP in order to elicit the preferences of the world-wide population, the different (more controversial) preferences of the different sub-societies might end up marginalizing each other, marginalizing especially minorities. 13In any case, if we use SP to determine political priorities special attention should be given to the scope of the society in question.
Turning to another set of difficulties, the assumption that stated preferences is the right method to establish normative-political priorities is heavily debated in political theory.
When pointing to the weaknesses of relying on preferences in political contexts in general, we can build on a prominent political-theoretical literature that explores the general pitfalls and shortcomings of resolving political issues by aggregating individuals' preferences (see for example Elster 1983;Goodin 2003;Sunstein 1991).We shall briefly review five arguments raised in this literature that are relevant to our context and apply them to the SP method.
A significant first problem has to do with adaptive preferences (sometimes referred to as endogenous preferences).Preferences are adaptive when shaped by external circumstances with no intentional deliberation that may lead to their alteration (Bovens 1992;Elster 1983;Nussbaum 1999).We want to respect preferences because they generally reflect people's autonomy, but adaptive preferences are not like that-they are not autonomous.Preferences about well-being aspects, both personal and social, may be surprisingly adaptive.One mechanism behind this is that and Fitoussi 2009).It is also part of the broader and increasing phenomena of using indices and rankings in the domains of international relations and public policy (Cooley and Snyder 2015; Davis et al. 2012).In this context, an index is supposed to be capable of "being used to compare particular units of analysis, synchronically or over time" (Davis et al. 2012, 6).many times people do not see the value of things until they get to engage with them to a certain degree.Some well-being surveys ask about the importance of arts and culture, but while many people who were exposed to complex forms of artistic endeavour (like visual arts or literature) find them an indispensable part of the good life, people in environments with less exposure to those elements will not in general claim to value them highly.14Preferences regarding political rights may be adaptive as well.It is possible that people brought up in democracies find the participatory rights they enjoy a crucial part of social well-being while people who do not have them to begin with may find their contribution negligible, or even negative (see for example the recent case of Bhutan in Slater 2018).
A second problem is that stated preferences can avoid reflecting judgements about the common good and judgements about values.This can create a problem parallel in its structure to a collective action problem (Sunstein 1991). 15This issue can reiterate in the context of well-being surveys.This may happen, for example, for the following aspect in Benjamin, Heffetz, et al. (2014, 2715), which asks participants about: "The sense that you are making a difference, actively contributing to the well-being of other people and making the world a better place".It may be that people will not define engaging in altruistic behaviour as important for social well-being when they think that people will generally not reciprocate the same sentiment.
Relatedly, it should not be taken as obvious that preferences successfully reflect values (Haybron and Tiberius 2015;Christiano 2008).Hence, some scholars distinguish A's preferences-which may be whims that A finds no good reason to satisfy-from her values-a stronger and more distinct set of attitudes-in their attempt to define what A deliberately sees as good.One might answer that preferences could indeed reflect values, but we should consider, at the very least, complementing SP methods to bring preferences closer to reflecting people's values.Social-deliberation methods could be of aid in this endeavour.
A third problem is that of illegitimate preferences: the possibility that some preferences that people will claim to have regarding well-being aspects are immoral.For example, it is possible that some individuals will hold a preference for an ethnically homogeneous society.It seems plausible to argue that higher-order moral considerations render such preferences problematic (Kymlicka 2002, 26-32).Benjamin, Heffetz, et al. (2014, 2703) address this worry and suggest that the hypothetical choices in abstract scenarios may elicit more deliberated and thoughtful states of mind (more, that is, than in 'real life' scenarios) that will wash out 'emotional' preferences.Although this is possible, still, there is no guarantee that this tendency will take place.
The common solution to these types of problem in the general case is 'preference laundering'-we should not defer to the individual's actual preferences but to their 'purified' or 'laundered' preferences (Goodin 1986;Griffin 1986, chapter 9;Hausman 2011; see Infante, Lecouteux, and Sugden 2016 for a recent survey of theoretical models that make use of this notion).Our discussion of deliberation methods in the next section provides some practical tools to bring about such 'preference laundering'.
A fourth problem is that individual's preferences (both personal and political) can be extremely unstable, perhaps little more than some form of primitive impression (Schumpeter 2003, 256-264;see Freeder, Lenz, and Turney 2018 for some recent evidence in a political context).If preference a may be changed to b and back in a short period of time, it seems that the case for respecting preference a is undermined to a great extent.Moreover, in case it is unstable it cannot underpin a stable index.
Finally, it may be also the case that relying on preferences concerning social well-being priorities is a process that demands information and theorizing not available to most people.When trying to understand, for example, the place of inflation in individuals' well-being (an aspect in the list of Benjamin, Heffetz, et al. 2014), it is hard to see how individuals can evaluate in a meaningful way its impact on their lives or the lives of their fellow citizens, its trade-off with other well-being aspects, etc. Goodin (2003, 54-57), for instance, provides basic principles regarding the question of which kinds of preferences are suspected of being 'permissibly overridden' in a democratic framework.Among them, he points to 'uninformed preferences' of various kinds.
Another example for this last difficulty is with comparing between (or specifying the trade-offs between) objective social circumstances and subjective attitudes.What is the meaning of comparing the level of freedom of speech, or rate of unemployment to a subjective average of life satisfaction in a society (a policy-vote scenario in Benjamin, Heffetz, et al. 2014)?The task of attentively addressing the relation between the two is a complex one (and demonstrates a difficulty of parsing between instrumental and intrinsic valuations of goods). 16This kind of task within a survey violates a basic suggestion in methodological guidelines: to avoid giving the respondent too difficult a task to perform (Belson 1982, 389).It should be noted that the fact that respondents do provide their preferences regarding every two aspects does not guarantee their awareness of the issues behind the decision.There is a possibility that respondents use particular heuristics in order to avoid the complex decision-making behind the comparison.
The above problems can be mitigated, at least to some extent, by certain processes of deliberation (Manin 1987;Cohen 1996, 348-350;Sunstein 1991), both by experts and by the survey participants.

IV. EMBEDDING STATED PREFERENCES AND DELIBERATION
Focusing on the method of stated preference aggregation in the context of social well-being indexing allows explication of the concrete sorts of problematics that this aggregation procedure-taken as is-conceals.In this way it paves the way for complementing this procedure by various deliberation processes that could mitigate the problematic aspects-deliberation understood as a process through which individuals and groups justify decisions through giving reasons and exchanging information (Manin 1987;Cohen 1996;Gutmann and Thompson 2004).
Deliberation could be useful, in our case, as a process conducted by experts and by the survey participants.Deliberation processes are expected to allow for the SP aggregation process to become more focused and more appropriate for implementation.Deliberation processes have the dual potential to improve the inputs of the aggregation and to better the use of the output.
Thus, with the pitfalls and limitations of the SP methods borne in mind, the complementing deliberation procedures, so we suggest, should come at two stages of the process.
First, extensive deliberation processes should come prior to the SP exercise.Concrete dilemmas should be resolved.These preliminary processes could gain both from public deliberation and deliberation by experts.Among the deliberated issues, as clarified in the previous sections, it is necessary to specify the goal for the index.What concept exactly is it supposed to represent?Welfarism or something beyond?What is the scale of investigation (local or universal)?What is the purpose of the index (comparing between societies or between individuals within societies)?It is through resolving these issues that the exact role of aggregation is specified.
Secondly, group deliberation techniques should be better integrated into the SP exercise itself.Deliberation is directed in this stage of the process to strengthen self-reflection through conversation, to make preferences better reasoned and better informed, and to provide the opportunity to affect the judgments of others (Landa and Meirowitz 2009, 427).We should recall that in the political context of the exercises (eliciting 'citizen preferences'), the main theoretical consideration that motivated the development and implementation of the SP method was to respect individuals and avoid the imposition of external standards on them.The political philosophy literature, however, reminds us that deliberative methods implement these values in a way that aggregating votes (as SP methods do) cannot (Cohen 1998(Cohen , 2009;;Dryzek 2007).While a citizen who answered the SP survey can know that her voice is one of the many voices that affected the exercise, the exercise's results are still an opaque index that does not put forward any substantive justifications for the many policy implications it entails.Using discussion and debate as part of the process means that the citizens are actually presented with reasons underlying the policy choices and have richer opportunities to impact and challenge the policy decision.This implements a much more intuitive and convincing understanding of respecting individuals.
Moreover, deliberations will help us resolve many of the concrete issues we have identified with relying on stated preferences (in section III.II).Adaptive preferences, self-regarding instead of public-regarding preferences and values, illegitimate preferences, unstable preferences, and ill-informed preferences-could all, at least to some extent, be mitigated by deliberation between the survey participants.The process can stabilize the preferences (for a recent review of the empirical evidence see Kuyper 2018, 4-5).Being able to ask questions and to be exposed to arguments of other citizens will help participants to know what their fellowcitizens' preferences are, and so avoid collective action problems and the neglect of values they would actually accept and acknowledge if only they knew more about them (like the value some people would find in publicly funded art or the public representation of marginalized minorities).
Political choices about well-being priorities are not individual independent preferences, they make assumptions about others, and not giving participants the opportunity to explore these assumptions generates distortions.
Furthermore, we believe that deliberation processes might be helpful even in the context of pure well-being tracking (the personal context).In the deliberation process, participants can be presented with ideas that they could have not taken into account when answering the survey in isolation.They can also make their preferences more robust and reflective through discussion and be encouraged to also deliberate intrapersonally-the more formal events in which interpersonal deliberation is conducted can in themselves encourage participants to think more seriously about the questions they are presented with, and so provide more hardheaded answers to them.These more thorough and comprehensive answers will provide a better, more consistent understanding of what actually underlies the participants' well-being.
Think, for example, about someone who on a quick survey does not realize the importance certain economic opportunities or conditions provide to her, but after hearing more about their meaning, understands that they are a part of her own well-being.Deliberation, even in the tracking well-being case, may lead to 'corrections' in our index, and maybe also to 'laundering' preferences.
How should these participants' deliberations take place in practice?Mini-publics, in which the chosen participants are convened to discuss the posed problem, are the paradigmatic institution that the literature on deliberative democracy directs us to (Ryan and Smith 2014).However, if we want a very comprehensive picture of how well-being is conceived among a large group of citizens, these might be challenging to arrange in practice.We can suggest two alternatives that have been developed in the recent literature.First, we can use online deliberation.Well-tailored forums designed by research institutions and government agencies can provide a platform for high-quality long-term discussion that will allow a large group of citizens to construct meaningful images of what they think about when they think about a life that is going well (for a useful review of practical and theoretical considerations regarding such institutional online deliberation, see Strandberg and Grönlund 2018).Alternatively, or in addition, we can use deliberative procedures that inform us about what kind of differences in answers we can expect if deliberative elements were to be incorporated along the whole body of respondents, such as representative deliberative polling (Fishkin 1995).17In these deliberative processes, citizens would be able to suggest well-being aspects of their own, present their own reasons for prioritizing certain elements, enter into a long-term process allowing them to reconsider such reasons, and to change their choices about the policy priorities.As part of these deliberation efforts, more extensive discussions with experts can be conducted, allowing the weighing in of expert knowledge that will not be incorporated in the original SP procedure.
A common concern is that while deliberation is an attractive ideal, it does not work in practice-and that the deliberative procedure creates polarization in the real-world group (Sunstein 2002), conformity pressures that can distort some of the more eccentric voices, and, more broadly, a discussion that does not revolve around respectful reasonsexchange and mutual learning (this might seem like a particular problem with online deliberations).However, as some empirical results have demonstrated, such low-quality deliberation, whose fundamental problem is probably the lack of a disposition to engage with peers in a way that might lead to a change in one's opinion (Brennan and Landemore 2021, 223), is likely to occur only under specific conditions and contexts.In particular, such phenomena are likely to obstruct deliberation when individuals already have strong predispositions about the topic of deliberation, that they would seek to reinforce-for example, on topics already strongly politicized (Mercier and Landemore 2012).As defining well-being does not resonate right-out-of-the-gate with any salient public discourse, such patterns can be predicted to be less likely to occur in our context.Facilitated or moderated discussion, as those we are talking about will be, is also less likely to engender reduced quality (see Strandberg and Grönlund 2018, 369).Actually, a staple of successful democratic deliberation, the Icelandic constitutional process of 2010-2013 (Landemore 2015), revolved around giving citizens similar tasks of putting forward values they care about and deliberating on their relative importance. 18he use of multiple deliberative procedures fits a framework that asserts that the question of the nature of social well-being is too complex to allow the provision of reliable answers through immediate responses; the answers should be produced in a long-term, multi-layered process.These methods of preferences elicitation do their job if we have meaningful, well-formed preferences.But this is not the case with political preferences which require the processes of exploring and forming them.
Scientific scholarship has shown that deliberation processes have brought about changes in opinions and views, at least to some extent and in certain contexts (Ulbert and Risse 2005;Sulkin and Simon 2001;Fishkin and Luskin 2005).We stress again that the claim that deliberation will mitigate the political-theoretical problems of relying on stated preferences in this context is partly an empirical hypothesis that merits testing. 19It is not solely empirical, however.A part of the importance of deliberation cannot be proved by observation.The very legitimacy of the inputs is (by definition) strengthened through this process, as pointed out by Manin (1987, 352): "It is the process by which everyone's will is formed that confers its legitimacy on the outcome, rather than the sum of already formed wills".
Finally, acknowledging the problematic aspects of the SP aggregation (even in the deliberative mitigated cases), we could ascribe to the output of this aggregation procedure different roles within the broader decisionmaking process.As suggested in other contexts (Goodin and Dryzek 2006), the output, i.e., the index weights, could be used in various ways and not only for determining weights.It could also, for instance, inform an external process of determining weights and contribute to legitimizing particular indices.

V. CONCLUDING REMARKS
Stated preference methods in the context of the social well-being index problem carry significant strengths.However, they also have their limitations, especially from political and philosophical points of view.While the literature developing these techniques and the experiments implementing them have expanded in an accelerating manner, these limitations have not hitherto been systematically analysed.Their thorough analysis, we suggest here, allows us to address them in order to take better advantage of the SP innovations in the political context.
The supplemental suggestion advanced is that acknowledging the various strengths and pitfalls of the SP methodology clears the way for embedding deliberation procedures into the preferences aggregation processes with which it provides us.Within these mixed procedures, a significant role will be reserved for the SP, albeit a limited one.
From a theoretical point of view, this paper can serve also as a lesson for other cases of embedding aggregation and deliberation in practical realms.Our paper shows that the question of how to combine aggregation and deliberation needs to be considered in particular contexts, rather than solely through overarching theories.The detailed analysis of a particular case could point to the unique difficulties it raises, together with the more general ones identified by political scientists and philosophers.Understanding the particular aggregation problem aids in tailoring better deliberation procedures that can help to overcome the problems.
This process is especially significant in cases where scientists point to new sophisticated aggregation methods.In such cases, even when aggregation methods are compelling, their implementation should be modified by external context-dependent considerations.