A Public, Open, and Independently-Curated Database of Happiness Coefficients

We present a nascent database of happiness coefficients. This is a synthesis of evidence on the size of improvements to human life experience that can be expected from changing objective, policy-amenable circumstances. The wealth of data on people’s self-reported satisfaction with life in a wide variety of circumstances, from around the world, including respondents undergoing a diversity of changes and life events and subject to a variety of public policies and policy changes, has provided a rich base of knowledge about what makes life good. This growing research literature has in recent years been met with interest from central governments looking for accountable but more human-centred approaches to measuring progress, as well as for communicating objectives, making policy, and allocating resources. Meanwhile, frameworks for benefit-cost accounting using inference from life satisfaction data have been devised. In some cases central government finance departments and treasuries are incorporating this approach into their formal methodology for budgeting. The body of causal inference about these effects is still somewhat diffuse. Collating, reviewing, and synthesizing such evidence should be led initially by academia and ultimately by a broad academic, civil society, and government collaboration. We report on the assembly of a database of summary estimates for Canada, supplemented where needed by evidence from around the world. The categorized domains of individual experience and circumstances include Education, Environment, Work, Finances, Health, Social Capital, and Crime. The paper also explains the context for and limitations of the use of a database of happiness coefficients.


3 Introduction: Precedent Databases
As governments strive to establish new and updated frameworks for the evaluation and planning of programs, policies, and budgets based on modern evidence about human wellbeing, some new institutions will be needed ( Barrington-Leigh, 2021). This paper (1) suggests some principles for the curation of growing knowledge about what makes for a good life, and happy societies; (2) provides a fledgling sample of what a database of such research findings might look like; and (3) articulates some important limitations on how such knowledge can be used in policy making.
Throughout this paper, the term "happiness" can be taken as a short hand to mean the set of subjective wellbeing measures that are used to gauge overall quality of life, most prominent and important among them being respondents' own numerical evaluation of their level of satisfaction with life (SWL), obtained through a single survey question. The focus on response data to this one specific subgroup of subjective well-being questions, known as "evaluative" subjective well-being or as "cognitive evaluations of life," is motivated by extensive evidence that it best captures the impacts of enduring lived circumstances, while subjective well-being questions focused more on affective states are better suited to capture day-to-day influences (e.g., Abdel-Khalek, 2006;Helliwell et al., 2022). It is also the measure recommended to government statistical agencies for use as an overarching measure of well-being (OECD, 2013;Stone & Christopher, 2014).
Reviews of econometric studies of happiness have in several instances compiled summary effect sizes into tabular form, in which the existing evidence on several different influences on life satisfaction are brought together (Clark et al., 2019, online Annexes 2-5). A more comprehensive but less synthetic approach is embodied as part of the World Data Base of Happiness (Veenhoven, n.d.). Its "Correlational Findings" section reports estimates of effects on happiness from a vast number of studies. Frijters et al. (2020) describe a process by which the UK might maintain an authoritative list of the best available estimates for any given influence on wellbeing. Frijters et al.'s description may represent an overly-frequentist conception of filtering and aggregating evidence, but they emphasize the importance of moving towards transparency in however the updating of the database is carried out. Barrington-Leigh (2021) similarly advocates for a process to debate and distill knowledge about the relationship between policy-influenced variables and human experience, in an accountable and ongoing process, but suggests that this be led initially by the analytic community, rather than initially by government, with a transition towards more independence over time.

Principles for Curation
In what follows, we dub the database of such knowledge a Database of Happiness Coefficients (DoHC) and, for the purposes of discussion, the public body tasked with curating it a Wellbeing Knowledge Centre (WKC). We propose the following principles for a WKC and DoHC to support policy making by government: 1. The DoHC should be curated independently or at arms-length from government. This is to ensure that all findings wll be made available to the public as well as to government agencies. While the knowledge embodied in the DoHC will never be sufficient to dictate policies (see Sect. 4), it must be available to the public and to public and private organizations in order to help to push government to adopt a more evidence-and humanoriented policy making approach. 2. The WKC must strive for maximum transparency of its methods, including the criteria for selection and integration of studies. Collating, reviewing, and synthesizing evidence for the DoHC should be a collaborative undertaking, with engagement from all interested stakeholders. Initially, this task should likely be led by academia but ultimately it should become a broad academic, civil society, and government collaboration. This will ensure that the evidence remains (1) robust, (2) inclusive of evidence, such as government program trials, which is high quality but may not be published, and (3) likely to be both used and usable by interested parties. A possible productive service for the WKC would be to host an ongoing open review process of all research sources used to build the DoHC. 3. The WKC must always embrace an openness to revision of the database. Core to the DoHC are evaluations of the degree of confidence in causal inferences behind each coefficient. Future evidence will continuously revise and deepen the DoHC. 4. The DoHC should be designed to inform calculations about the expected distribution of wellbeing. Most frequentist statistical models in use in this field focus on estimating mean values of wellbeing, or use strong distributional assumptions, but these may ultimately prove inadequate to inform policy choices, which will be based on the full predicted distributions of wellbeing outcomes. That is, policy makers will want to consider both the univariate distribution of (for instance) SWL, as well as its variation along standard dimensions of disadvantage, oppression, and inequality. 5. The DoHC should target content to support the needs of planners and decision makers.
Explanatory variables (predictive factors) in academic models are often chosen based on (a) their hypothesized importance in accounting for variance in happiness, or (b) simply on their availability, or (c) on being able to show or argue that their variation constitutes or contains a natural experiment of some kind. In order to be useful to decision makers and community planners, estimates will instead need increasingly to focus on the effects caused by objective, policy-amenable outcomes. One may therefore expect initially many knowledge gaps in the DoHC. The WKC may need to help direct resources to fill those gaps. In order to be accessible and useful, evidence in the DoHC also needs to be available in a format or formats tailored to the needs and capacities of relevant analysts and policy makers. It will need to be effectively communicated to ensure both awareness and timely attention. 6. The DoHC should be constructed so as to allow for hierarchically-sourced evidence and be able to privilege locally-contextualized evidence. A national WKC will incorporate evidence from around the world and liaise with other national or international curators of DoHCs, or possibly to lead in the curation of an international one. In any case, locally-contextualized evidence should be given appropriate priority, and at all geographic scales. A municipal or community government will need to lean heavily on evidence about wellbeing gathered from beyond its jurisdiction, but at the same time will want to emphasize local experience.
In practice, large government departments may inevitably maintain their own version of the DoHC internally, but it is expected that internal government studies and experience will eventually make it into the public domain, so that novel information should ultimately all flow into the public DoHC.

3
The WKC will most likely need to commission studies to summarize the knowledge in a given field, and incorporate the synthesized findings into the DoHC. The What Works Centre for Wellbeing in the U.K. is already playing this role of commissioning reviews (e.g., What Works Centre for Wellbeing, 2018).

Seed DoHC for Canada
In the interest of seeding an effort of building a DoHC for Canada, and in order to communicate the concept, we report the construction of a small DoHC.

Methods
Briefly, the following procedure was carried out to arrive at our database entries. First, a search of EconLit, EconPapers, Scopus, and JSTOR for publications in economics and psychology led to a set of 189 academic articles and working papers related to life satisfaction in Canada.
Secondly, these papers were retrieved and sorted by topic. Features such as survey data used, sample size, age of respondents, geographic scope, temporality (cross-section or longitudinal), and the subjective wellbeing measure in use were all tagged.
Thirdly, studies with large sample sizes, relevant SWB measures, nationwide scope and/or longitudinal data were preferentially chosen. Within each, we identified estimates derived through well-defined methodologies and evaluated the confidence in their effect and causality. These features were recorded in the database, along with any free-form comments or clarifications.
Our database is similar in intent to that of the What Works Centre for Wellbeing (2018), and different from that of the World Database of Happiness (Veenhoven, n.d.), in that it aims to synthesize a literature and interpret the relevance, confidence, and causal identification of available studies rather than to comprehensively enumerate them all. This process will always require judgment, but (Bayesian) statistical procedures needed to achieve the principles described above, especially the sixth, in a reproducible way will still need to be developed.

Findings
Our database is available online at https:// lifes atisf action. ca/ dohc and included in full (as at the time of writing) at the end of this manuscript. Table 1 shows a few sample values from the database, which also includes commentary on the persistence of effects over time, the degree of confidence in effect and causality, the data source and type, and of course the relevant citation(s).

The Role of Happiness Policy in Context
In any discussion of the life satisfaction approach to benefit-cost accounting, it is important to keep the context in mind. There is a lot that any DoHC or wellbeing policy approach will never be able to do, and DoHCs do not have the potential to diminish policy making towards a deterministic or technocratic exercise. This section describes three important limitations (discussed in more detail in Barrington-Leigh 2021) to what can be expected from a DoHC.

Distributions
First, as mentioned above, the knowledge base around predicting policy effects on wellbeing should in principle be designed to predict distributions of outcomes, not just averages. Having a good understanding of wellbeing impacts means one can disaggregate the overall effects of a given policy or budget based on different demographic groups or subpopulations and, importantly, intersectional groups. Many governments, when carrying out evaluations or projections, already disaggregate outcomes in this way. Using a new or more encompassing measure of wellbeing as an objective does not change the need nor challenge of understanding distributional outcomes.
Moreover, those distributional outcomes are fundamental to decision making. While the early literature on cost/benefit accounting for policy-making (e.g., Happiness Research Institute, 2020; Frijters and Krekel, 2021) emphasizes scalar objectives and descision criteria, in reality decision makers are sensitive to non-scalar considerations. For instance, Fig. 1a, b show hypothetical distributions of current and projected future life satisfaction.
The prospective policy appears to increase wellbeing from 6.6 to 7.3, according to its mean, yet the distribution shows that some people are worse off afterwards than before. Panel (c) disaggregates the anticipated outcome into a demographic subgroup (shown in orange) and the rest of the population (shown in blue). The relative lack of thriving of the subgroup may be a considerable concern for policy makers for ethical (equity) or political reasons. In any case, nothing about a life satisfaction approach nor the information in a DoHC will resolve the question of how to value different parts of a distribution in coming to an overall decision. These kinds of considerations do not happen automatically with a wellbeing approach, just as they do not happen automatically when using traditional welfare measures like family income.

Dynamics
There is a second reason that a DoHC does not act as a policy oracle. Policy makers may disagree about whether reducing a given disparity is best carried out through strong government intervention and redistribution, or more through removing barriers and allowing for people to change their own situation. However, this question is not just about ethics and principle, but also about the dynamics of how people behave and invest over their life course, and indeed how all kinds of possible and typical government investments pay off over time. Those are questions to which a wellbeing approach assumes you already know the answer. That is, the DoHC is likely to specialize, especially early on, in answering the question, "Given a set of objective conditions at some (future) point in time, how happy would someone be?" In order to project the outcome of a policy change or budget allocation today, one will need to predict future objective conditions driven by the policy change. This information is all outside of the DoHC's contribution (for more explanation, see Barrington-Leigh, 2021, 2022. If anything, though, the policy synergies made possible by having an overarching, well-understood measure of wellbeing may make it much more desirable and valuable for governments to have sophisticated and detailed models of the return to human and non-human investments over the life course.

Precautionary Approach
Despite the limitations above, the most ambitious and attractive promise of a wellbeing approach is that it offers a way to add up all the effects of taxation, legislation, and expenditure, along with extant conditions, to come up with a reasonable prediction of the distribution of outcomes for a prospective policy. This system, which boasts accountability to measurable outcomes and a growing evidence base, can provide cost/benefit or cost effectiveness guidance to a decision maker who has a way to handle distributional questions. However, there is another dimension in which this vision has its limits: one cannot feasibly apply the wellbeing approach to all questions about future public investments. In particular, when considering questions about some investments with far-future Hypothetical distributions of life satisfaction responses described in the text. a is before a policy change; b is a predicted outcome after the policy takes effect, and c a disaggregation of the resulting distribution for a demographic subgroup payoffs, the uncertainty in predictions of objective outcomes will lead to a large amount of uncertainty about the implications for future human wellbeing. This uncertainty can overwhelm any decision-making clarity for decisions about alternative uses and benefits of a resource in the short term. That is, for long-run, unfamiliar, unpredictable, complex, and uncertain dynamics, the calculations described in the previous sections may not provide precise enough answers for making decisions in the same way that shorter-run decisions can be made. They will not always be able, therefore, to direct us when making choices between short-term outcomes and long-run outcomes.
This limitation is, again, nothing to do with switching to a more evidence-informed metric for human wellbeing. It is instead an existing challenge that is unchanged by the availability of a DoHC except in that it comes into sharper focus. When one has a more explicit measure of human wellbeing, the question of whether policy is simply meant to maximise it is starker than when pursuing vague, proxy objectives like economic growth, which no one would argue is a singular goal of optimal policy. The implication of this limitation is that some other principle, i.e., beyond wellbeing maximisation, is needed to make long-run decisions whose ramifications are particularly speculative or far-off. Barrington-Leigh (2021, section 6.1) again describes the alternative, or solution, in more detail, and associates these long-run quandaries with the idea of sustainability. A "precautionary approach" is typical language for how to handle such uncertainty when the costs and benefits for human wellbeing are not sufficiently understood or precise.

Conclusion
The availability of a DoHC with sufficient coverage and precision to be useful for informing government decision-making has become an imminent reality. The UK Treasury (2021) already has explicit guidance in place for this kind of quantitative evaluation. Canada's new Quality of Life framework (Department of Finance, 2021) is perfectly suited to benefit from it also. On the way there, however, are significant capacity gaps and institutional transitions. A close relationship with academic researchers will be necessary in the beginning to construct this important database of human knowledge. The nascent DoHC in this paper may serve as an example for researchers and government agencies to begin thinking about how to shape, organize, and curate such information in an open and transparent and geographically hierarchical way.
As this idea permeates government agencies, a few cautions or points of advice are in order, and described above. To reiterate, (1) quantitative wellbeing approaches do not release governments from the duty of judging questions of distribution and equity; nor do they diminish the role of politics and debate in this task; (2) a DoHC does not predict the future; it only tells us how a given future may map onto experienced wellbeing; great efforts are needed in bolstering governments' abilities to model returns to investments, in particular investments in people which bear fruit throughout the life course; and (3) many questions of long-run sustainability cannot be sufficiently handled through quantitative optimization of wellbeing and should instead be debated and settled using an alternative framing principle, such as the goal of more arbitrary conservation.