Presented at AFSP (Lyon, June 30 - July 2, 2026, ST 03 "Characterising and measuring public policies for comparison"), this talk proposes a layered theory of the social construction of target publics, measured on millions of Canadian and Quebec parliamentary utterances. At the core, the Schneider–Ingram typology; around it, public-policy domains, parties and position, MPs, and attention peaks. Each layer is tested separately, and each one displays its own share of stability and plasticity.
Alternation changes imputation

On the same publics, the government-to-opposition switch flips the affective regime by 23 to 32 pp, but does not displace publics to other domains (r = 0.96 between profiles) and barely touches the proof burden (≈ 2 pp). Plasticity is real, but bounded to affect.

Some groups must have their worth proven

The probative gradient runs between groups (3 % for the middle class, 17 % for low-income people) and between parties, not between government and opposition (only +2 pp). Proof surges within deliberative peaks, where informational theory expects it, not where the negativity bias predicted.

A structured fabric that has its artisans

Net of party × position × year cell averages, MPs hold individual regimes of their own (σ = 7.2 pp on negativity, vs 2.6 under permutation, p < 0.002; σ = 3.6 pp on proof). Their thematic portfolios self-correlate at 0.945 across career halves, against 0.869 with party colleagues.

A repertoire of indicators beyond budgets

Beyond budgets and administrative outputs, the discursive fabric of target publics becomes measurable: 17 publics × 21 CAP domains × two regimes (affective + probative), at the sentence level, on millions of utterances. A replicable pipeline, comparable across time, parliaments, and languages.

The making of target publics - AFSP 2026
PDF Open
Scroll Navigate between slides Click Interact with data Fullscreen for a better experience
Use arrow keys or click to navigate through the slides.
AFSP Congress 2026 French Political Science Association, Lyon