Attention,Assortment,Affiliation: An Agenda for Wellbeing Economics

Introduction

Over the past couple of decades, wellbeing economics has transitioned from a marginal subfield into a central component of empirical social science and policy discourse. A decade or so ago, the primary intellectual task was still to establish the legitimacy of subjective wellbeing (SWB) as a measurable and meaningful objective for individuals and societies. Building on our recommended questions, governments routinely collect SWB data (Dolan & Metcalfe, 2012; Office for National Statistics [ONS], 2011), and SWB is now a recognized outcome for the economic appraisal of new policies and programs in the UK. As the field has developed, the concern is no longer whether SWB should be measured, but rather what aspects of human experience matter most, how they interact, and how they can be impacted by policy interventions (Diener et al., 2009).

This paper reflects on the science of SWB since I wrote Happiness by Design (HBD, Dolan, 2014) and argues that the next phase of wellbeing economics requires a shift in emphasis. Current approaches remain limited in three important respects: (1) they often rely on measures that may not align with what people actually attend to in their daily lives; (2) they tend to focus on average effects that obscure substantial heterogeneity; (3) they underplay the central role of social connection, especially in the form of shared experiences. The contribution of this paper is to organize these challenges around three interrelated themes: attention, assortment, and affiliation.

First, I argue that SWB measures should better reflect the allocation of attention in everyday life. Survey responses, especially off-the-cuff global evaluations, may capture something meaningful, but they do not necessarily correspond to the positive and negative experiences that dominate people’s lives. A better approach to measurement should therefore align more closely with what individuals notice, feel, and engage with over time. Such issues have been core to my own research, impacted greatly by Daniel Kahneman, who was a huge influence on my life in general for over two decades (and he wrote the foreword to HBD).

Second, the paper emphasizes the importance of assortment, or heterogeneity in the production of SWB. There is no single “recipe” for happiness. The same circumstances, such as income, employment, relationships, health, can have very different effects depending on personal characteristics, preferences, expectations, the attention individuals afford them, and even on baseline happiness levels. Yet much of the empirical literature continues to focus on average treatment effects. A more informative science of SWB must instead seek to understand for whom, and under what conditions, different factors matter.

Third, the paper highlights affiliation, or social connection, as the attribute that comes closest to being a universal determinant of SWB. There’s growing evidence that connection is regulated in the brain in ways that resemble our most basic needs, like thirst and hunger. It’s that important. I have long argued that “having someone to talk to about stuff” is about the best proxy for SWB. While the importance of close relationships is well established, less attention has been paid to the role of shared experiences, such as those found in music, sport, and collective rituals, in generating feelings of connection, even among strangers. These forms of affiliation, often characterized by synchrony and shared attention, represent a powerful pathway to improving SWB.

Taken together, these three themes point toward a more mature research and policy agenda for wellbeing economics. Rather than focusing narrowly on average effects or relying predominantly on evaluative measures, future work should aim to capture SWB as it is experienced over time, account explicitly for heterogeneity in responses, and treat social connection as a central component of SWB. A significant shift in “what works” by way of policy might result from the “new” wellbeing economics.

Measurement—Attention

SWB is a broad term for how we think and feel about life and our experiences. Consequently, there are several ways in which it might be tapped into. For the most part, this paper will be focused on self-reports, on how people respond survey questions that ask them about SWB on scales ranging from, say, 0 to 10. The empirical study of SWB owes an intellectual debt to Kahneman’s distinction between the remembering self and the experiencing self (Kahneman & Riis, 2005). The former is best captured by global evaluations of life satisfaction although, as I pointed out to Danny on several occasions, evaluations can include anticipations and expectations as well as memories and reflections. You can look forward to me saying more about the effects of looking forward below.

The research using life satisfaction has proliferated over the last decade or so and the measure is increasingly accepted as a legitimate metric through which to make inferences about how changes in circumstances and public policy affect SWB. Even Kahneman, who was initially committed to a more experience-based account of SWB, warmed to the idea of using life satisfaction as a guide to welfare. I remain less convinced. First, it takes someone less than two seconds to answer a life satisfaction question, and so they are unlikely to be answering the question posed because being asked to evaluate your life is quite challenging and would take a lot longer than a couple of seconds to assess. Psychologists have long known that we will look to find an easier question to answer than the one being asked and it might be that the shortcuts people use are still good guides to SWB but that has not yet been established.

I maintain that questions about hedonic and eudemonic sentiments are more tractable and the responses are easier to interpret. Kahneman et al. (2004) operationalized experiences through the Day Reconstruction Method (DRM), which provided a scalable approach to capturing hedonic feelings of joy, worry, stress etc. across daily activities. It asks people to divide the previous day into a series of episodes, like in a film, and to say what they were doing, who they were with and how they felt. In White and Dolan (2009), we extend the DRM to capture eudemonic sentiments of meaning, purpose and pointlessness. This paved the way for the pleasure-purpose principle articulated in HBD, which suggests that the happiest lives are the ones that find the right balance between pleasure and purpose. (The UK subtitle was “Finding pleasure and purpose in everyday life”)

Ultimately, SWB is determined by what we pay attention to. As elaborated on in HBD, attention functions as a scarce resource that determines what is attended to (inputs) in ways (the production function) that contribute toward overall SWB (the output). The DRM focuses our attention on our activities. Kahneman et al. show that time spent with your boss is one of the least happy activities of the day. By adding meaning into the mix, we have shown that time with the boss is not so bad. Both papers speak to the misery of commuting. The US subtitle of Happiness by Design was “Change what you do, not how you think,” and is consistent with an approach to SWB that is rooted in how we use our time.

This is not to say that thoughts are unimportant to happiness. Killingsworth and Gilbert (2010) suggest that we are happier when we pay attention to our experiences (whatever they might be) than allowing our minds to wander elsewhere. As someone who has sat through way too many professorial meetings, and whose mind has wandered to infinitely better places to be, I know that this cannot be universally true. We have shown that Killingsworth and Gilbert are broadly right in the sense that intrusive thoughts and happiness are negatively related, but also that mind wandering from miserable experiences can enhance happiness (Schneider et al., 2024). As with everything in life, context matters.

Evaluating your life and experiencing your life are psychologically and empirically distinct processes, and so, unsurprisingly, the research that has compared them show different life events and conditions matter depending on the measure. For example, using longitudinal data, Luhmann et al. (2012) show that life satisfaction is affected more by income, employment status and marital status, and emotions are more impacted by stressors, health symptoms and, quite obviously, time use. So, the measure matters. Context matters too, but, in general I would rather focus policy on helping people to feel better than on improving the way they think their lives are going. We routinely pay attention to our feelings whereas evaluative judgments are activated only occasionally, and so improved experiences are much more likely to show up in every sense than improvements in evaluations will.

Experience-based measures can provide more direct insights into the effects of policies that shape daily activities, environments, and interactions, such as commuting, workplace design, public services, and social settings. Experiential data should therefore play a more prominent role in policy areas where the aim is to improve my how people spend their time and how they feel while doing so. This is not to say that evaluative measures are redundant. They could be used to serve as high-level indicators of overall SWB. The challenge for policymakers is not necessarily to choose between these measures, but to use them in complementary ways that better reflect both the structure and the lived experience of people’s lives.

For the purposes of policy appraisal, however, we should be using measures that capture the duration of benefits as well as their intensity. Life goes better when we feel better and for longer, and both need to be expressed in a single metric if we are to choose between competing uses of resources that improve lives in different combinations of intensity and duration. Experience-based measures, especially when used alongside time use data, are much better suited to this task than life satisfaction, which does not have an obvious time frame or “flow” associated with it. Duration weighted-SWB will allow us to measure the area under the wellbeing curve, along the lines of quality-adjusted life years (QALYs) used in the appraisal of health interventions.

Heterogeneity—Assortment

People are different, thankfully. We would never have developed as we have if we were all the same. Just as companies have different production processes, humans have different production functions for converting inputs into happiness. And yet so much of the happiness literature talks as if there is a—the—“recipe” for happiness (Caunt et al., 2013). Much of this happens by default as by design. Economists are fond of using regression models, which identify mean effects about how income, say, relates to happiness. Inferences are then made about how much money “is enough”—for each of us. But how much money affects happiness will be determined not only by the attention production function of happiness but also by how much money matters in the first place.

As I discuss in Happy Ever After (Dolan, 2019), the powerful social narratives about what we should aspire to (wealth, marriage, children etc.) can be good for some people but can become constraining “narrative traps” and even sources of misery for others. If we look at the variance of the effects of marriage, it is generally the case that getting hitched is more a health and happiness gamble for her than it is for him. We should view this variance as an important signal rather than as a source of noise. Moreover, social comparison mechanisms will moderate the relationship between myriad life circumstances and SWB. We can all think of people for whom their place in the distribution (of income, say) matters a lot to them and others for whom such comparisons matter much less.

The effects of income provide a particularly clear example of how we should look beyond averages and focus on differential effects across the distribution of SWB. Kahneman and Deaton (2010) reported that daily moods increase with income only up to a threshold, beyond which they plateau. In contrast, Killingsworth (2021) found a continuous positive relationship between income and moods. An adversarial collaboration (of which more later) demonstrated that both explanations are right depending on baseline SWB (Killingsworth et al., 2023). Daily moods plateau in income for those in the lowest quintile of SWB but continues to increase in income for everyone else. Other research (e.g., Binder & Coad, 2011) show that income, employment, and health have larger marginal effects for individuals at lower levels of SWB.

Nowhere is a recognition of assortment more important than in our understanding of how people react to situational factors. Research in behavioral genetics has gone so far as to be able to identify genetic markers for the degree of environmental sensitivity, with some people predisposed to react more strongly than others to both positive and negative environments (Pluess & Belsky, 2013). This work inspired us to identify environmentally more sensitive “rough oceans” and less sensitive “calm seas” in an app-based study with over 200 LSE students, which asked several happiness questions every day plus DRMs over three weeks. Consistent with sensitivity being a trait, we identify people whose happiness bounces around compared to those with more stable responses.

These findings have important implications for policy design and evaluation. Policies that appear effective on average may deliver large benefits to some groups while having negligible, or even negative, effects for others. Rather than asking how much a given intervention increases average SWB, policymakers should ask for whom the intervention works, under what conditions, and with what degree of variability. This implies embedding heterogeneity directly into appraisal frameworks, for example by routinely reporting effects across subgroups defined by baseline SWB, preferences, or environmental sensitivity.

In practical terms, this also calls for a shift toward more targeted and adaptive interventions. Policies should be tailored to those most likely to benefit, such as individuals with lower baseline wellbeing or greater sensitivity to environmental conditions. Advances in data collection and analysis, including high-frequency and individual-level data, make such segmentation increasingly feasible. Further, evaluation strategies should move beyond retrospective identification of heterogeneous effects and instead incorporate them by design. Experimental and quasi-experimental studies can be structured to test differential responses across groups, allowing policymakers to learn not only whether an intervention works, but how its effectiveness depends on individual characteristics and context.

Connection—Affiliation

What floats my boat might sink yours, but if there is one form of ballast that is universally good for us, it is social connection. Recent research in neurobiology suggests that our basic needs are regulated by related systems that help maintain internal balance, a process known as homeostasis. Traditionally, thirst and hunger were viewed as purely physical drives, while loneliness was considered a psychological state. But all three can be understood as signals of imbalance. Through processes linked to interoception, the brain monitors bodily and social conditions and generates feelings that motivate action. When the body lacks hydration or energy, we feel thirst or hunger. In a similar way, loneliness arises when social connection is insufficient, producing an uncomfortable state that encourages us to reconnect with others. While loneliness does not rely on precisely the same mechanisms as thirst or hunger, it can be understood as a biologically grounded signal serving a comparable regulatory function.

It has long been known that affiliation through social relationships is amongst the strongest predictors of SWB (Helliwell & Putnam, 2004). Using panel data, Bartolini and Sarracino (2014) show that social capital is a robust predictor of long-term changes in life satisfaction, often outperforming GDP growth. At the more micro level, time-use data reveal that social activities, such as conversation, shared leisure, and communal experiences, are among the most enjoyable and meaningful daily activities (Kahneman et al., 2004; White & Dolan, 2009). In addition to strong social bonds, weak ties have been shown to contribute meaningfully to daily happiness, highlighting the importance of casual, everyday interactions alongside close relationships (Sandstrom & Dunn, 2014).

At the same time, there is growing recognition of the harms from loneliness and social isolation. These phenomena have been linked to mental health problems, reduced productivity, and increased healthcare costs (Cacioppo & Cacioppo, 2018). The policy responses to the COVID-19 pandemic underscored the importance of social connection. Restrictions on social contact led to widespread declines in SWB (Fancourt et al., 2021). These effects persisted even when economic conditions were stabilized, suggesting that social connection is not easily substitutable by material resources. Given all that we knew about the harms of social isolation, it shocked, saddened, and surprised me that there was not greater recognition of how life experiences and life expectancies were being curtailed and cut short by the social distancing measures.

Who should we be with and in what ways to satisfy our basic needs most effectively? As alluded to above, much of the evidence on the effects of connection on SWB comes from analyzing the properties of social networks—the networks of people “to talk to about stuff.” In parallel, experimental and observational studies suggest that engaging in activities alongside others can produce immediate increases in SWB. Concepts such as “collective effervescence” and “emotional synchrony” have been used to describe the heightened sense of connection that can emerge from sharing in the same experiences at the same time as other people for example, music, sport, and ritual (Paez et al., 2015). These effects are often observed even among individuals who do not have pre-existing relationships, suggesting that shared experiences can generate forms of connection that are not reducible to established social ties.

Social networks and shared experiences will often go hand in hand—the time we spend with those to whom we can talk about stuff will frequently be spent sharing in experiences of music, sport, and ritual. But they are not perfect complements, and they could nourish our need for connection differently, just as protein, carbohydrates, and fats nourish us differently, and in complex combinations. Since time and attention are scarce resources, we should ideally know when to connect to our network and when to share an experience (with those inside or outside of our network). It is fair to say that we know a lot more about how best to consume foods than we do about how best to connect with other people.

I should also mention the happiness hit of volunteering at this point because it is a well-established causal determinant of happiness. In our own recent research (Dolan et al., 2025), we used the heavily oversubscribed volunteering program during the pandemic to show that those who are selected are significantly happier than those not chosen. We need to move away from harmful notions of a hierarchy of charity, which seek to cleanse altruism of self-interest (Laffan & Dolan, 2023) and instead tap into the personal benefits of pro-sociality. Of interest to wellbeing economists is the finding that the program’s gains in SWB when expressed in monetary terms are about 140 times greater than the costs of the program.

Discussion

This paper has outlined three elements that must be given greater regard in a more mature science of wellbeing: (1) the role of attention when measuring happiness; (2) the assortment of causes of happiness; and (3) types of human affiliation. The future research agenda should seek to address these issues simultaneously by expanding the use of experience sampling methods (ESM) to capture SWB as it unfolds in real time—in private and in social settings. Any happiness question suffers from reactivity effects—respondents must interrupt the flow of their experience to respond—and so future research should look to wearable technologies to measure physiological reactions as well as self-reports (Gloor et al., 2009). I was saying long before HBD that I would love to find out how happy someone was without asking them, and the technologies are emerging that would allow us to do that, including large language models online (Greyling & Rossouw, 2025; Iacus et al., 2015).

All approaches must capture not only between-person differences but also within person variation. This requires large sample, high-frequency datasets to observe how different people experience the same circumstances, conditions, and environments in different ways. To get at causality, future research should develop interventions that deliberately shift attentional focus for example, through digital nudges and social prompts, and assess their impact on SWB. This should not only consider the impact of the “the event” but also how people feel before (in anticipation) and after (memories) of the experience itself. The value of family holidays, for example, is probably best explained by the hope-tinted spectacles of anticipation and the rose-tinted spectacles of fond memories rather than the actual (and typically painful) experience of childcare somewhere else (Nawijn et al., 2010). Most public policies affect people for some considerable time, and we need to do more to capture their full long-run effects.

Our own research is focusing on measuring the happiness associated with live events. They are a critical part of happiness, especially at a time when demand for them is ever increasing and when we are simultaneously retreating further into the digital and virtual worlds. We began in September 2025 with the “Happiness Experiment” with the DJ Fatboy Slim (Norman Cook). We collaborated with the wearable company Polar to measure heart rates and heart rate variability in 200 people at two specially curated gigs. So far as we are aware, this is the largest study of heart rates at the same time in the same place in response to the same stimulus. We have demonstrated that physiological synchrony across strangers is not only measurable but that it can be predicted to rise and fall at certain times, peaking when the bass is low in anticipation of “the drop” (when the bass kicks in). We are developing future experiments to test the robustness of this finding and to establish the extent to which people feel connected to other people at the same time as they are physiologically connected.

If attention, assortment, and affiliation are central to wellbeing, then more explicitly policy-related research into SWB must move beyond average treatment effects and instead design environments that shape what people attend to, for whom interventions work, and how connection is facilitated. This implies a shift from a “one-size-fits-all” models and toward more adaptive, segmented approaches. The science of wellbeing should seek to identify which groups benefit most from which interventions and under what conditions. This will require embedding heterogeneity directly into policy appraisal frameworks, moving beyond mean SWB impacts toward distributional and interactional effects across different population segments.

There is also a strong case for treating social connection as a form of infrastructure. Just as physical infrastructure facilitates the movement of goods and people, “social infrastructure” can be understood as the set of environments, institutions, and opportunities that make meaningful interaction between people more likely. This includes not only traditional community assets such as parks, libraries, and local high streets, but also organized activities and shared experiences. Live events, sports participation, cultural programs, volunteering schemes etc., can bring people together in time as well as in space. They can also connect people from across the socio-economic spectrum, not only fostering greater social cohesion but also “buy-in” from all sections of society to policies designed to enhance connection.

Policies should therefore prioritize investments that lower the barriers to social interaction and increase the frequency and quality of shared experiences. This might involve funding community spaces and local events, supporting grassroots organizations and clubs, designing urban environments that encourage interaction, and embedding opportunities for connection within public services such as health, education, and employment support. Treating connection as infrastructure therefore implies a rebalancing of public spending toward interventions that create and sustain the conditions under which not only social relationships but also “lighter touch” shared experiences can flourish. Such investments are central to human welfare, with demonstrable returns in terms of mental health, productivity, and healthcare savings.

A final point on how we conduct the research, drawing on my latest book, Beliefism (Dolan, 2025) and again inspired by Kahneman’s influence. Wellbeing economics contains several strong voices, who can sometimes talk past one another. Bringing together researchers with differing theoretical priors and empirical strategies to agree, in advance, on hypotheses, data, and methods offers a disciplined way of adjudicating between accounts—or, as in the income example, of showing how they can both be right under different conditions. More broadly, adversarial collaboration (a term coined by Kahneman though he was not the first to use the approach) aligns closely with the themes of attention, assortment, and affiliation. It forces greater attention to what is being measured and why; it foregrounds heterogeneity by design, and it is itself a form of intellectual affiliation, requiring trust, transparency and shared purpose among scholars. Understanding what makes life go well requires not just better data, but better ways of disagreeing. I hope you can agree with me on that, at least.