LSE — Philosophy of Economics | 2024
Can well-being be measured for public policy? This essay confronts a longstanding philosophical objection — that well-being is too heterogeneous to admit of a common metric — and argues against it. Drawing on philosophy of language and psychometrics, the essay defends the Differential Realization View (DR View): the claim that well-being has invariant semantic content across contexts, even though it is realised differently for different people. This provides the philosophical foundations for using subjective well-being measures — specifically life satisfaction scores — in policy evaluation.
This philosophy essay provides the conceptual foundations for the MSc dissertation on Prioritarian Social Welfare Functions. The dissertation uses life satisfaction (WELLBYs) as the welfare metric — a practice that requires the philosophical defence mounted here: that life satisfaction can be compared across persons despite heterogeneity in how well-being is realised. Read the dissertation analysis →
The philosophical debate about whether well-being is measurable for policy has two canonical poles. Hausman (2006) represents scepticism: preference satisfaction is the only valid welfare measure, subjective reports are unreliable, and comparisons across persons are illegitimate. Alexandrova (2016) represents optimism: well-being science is making real progress, and context-specific measures can be valid even if no universal concept exists.
The essay's contribution is to take Alexandrova's optimism seriously while rejecting her Contextualism — the specific semantic thesis that explains how it is possible. A better semantic framework, the essay argues, is the Differential Realization View.
Both Contextualism and the Differential Realization View accept that well-being looks different across persons and contexts. They disagree on the semantic explanation for this variation.
The semantic content of "well-being" varies by context. "Well-being" means something different when applied to a child, an elderly person, and a cancer patient. There is no single, context-independent concept — only a family of related, context-specific concepts.
There is a single, invariant core concept of well-being — something like: the degree to which one's life goes well for oneself. This concept has the same semantic content across all contexts. What varies is how that concept is realised: the substantive conditions that constitute well-being for different persons in different circumstances.
Alexandrova motivates Contextualism by analogy with epistemic Contextualism (Cohen 1986; DeRose 1992): just as "knows" means something different in high-stakes versus low-stakes contexts, "well-being" means something different in different life circumstances. The essay argues this analogy fails for a crucial reason.
Epistemic Contextualism holds that whether an agent "knows" a proposition varies with the standards operative in the attributor's context, not the agent's — and crucially, the object of knowledge is the same across contexts. Well-being, by contrast, is irreducibly agent-relative: what counts as going well for a person depends on that person's nature, projects, and circumstances. This is a metaphysical feature of well-being, not evidence that its concept is context-variable. The analogy misfires: well-being really is context-sensitive, but that is a feature of its realization, not its concept.
If Contextualism is correct and "well-being" is semantically different for different people, then welfare comparisons across persons involve comparing different things — a form of semantic equivocation. This is not merely inconvenient: it undermines the very possibility of aggregative welfare economics and distributional analysis.
Policy decisions routinely require comparing welfare across individuals — which group is most affected by a lockdown? Which income quartile benefits most from a transfer programme? These comparisons presuppose that the welfare of different people is being measured on the same conceptual scale. If Contextualism is true, this presupposition is false. The DR View, by maintaining semantic invariance, preserves the conceptual coherence of interpersonal welfare comparisons — while acknowledging that different people's well-being is realised differently, and that empirical measurement must be sensitive to this.
The DR View has direct implications for welfare measurement in public policy. If there is an invariant core concept of well-being — even if it is realised differently for different persons — then it is philosophically legitimate to use a single subjective measure as a proxy for that concept, provided the measure is appropriately validated.
The DR View permits acknowledging that two people with identical life satisfaction scores may have very different lives — different sources of meaning, different levels of need-fulfilment, different objective circumstances. This heterogeneity in realisation is not an objection to measurement: it is exactly what psychometricians call construct validity. A measure of well-being can be valid even if the construct is realised heterogeneously, as long as respondents are accessing the same underlying concept when they report their satisfaction.
The Hausman sceptic is wrong to conclude that heterogeneity implies incommensurability. The Contextualist is wrong to conclude that it implies semantic variation. The DR View offers the correct diagnosis: well-being is the same concept for everyone; it just looks different, because we are different.