Working Papers
We develop a model of Collective Household Choice with Random Attention (CHC–RA), extending random attention frameworks to two-person environments with feasibility constraints. Each decision maker forms stochastic consideration sets according to item-specific salience, which are then expanded by cross-domain spillovers. Preferences applied to the expanded sets generate candidate pairs that are subject to a household feasibility filter. The induced stochastic choice function admits a unique representation, with default behavior arising from both inattention and infeasible allocations. We distinguish separable and non-separable menus, introduce lattice extensions, and provide axiomatic foundations ensuring representation and uniqueness. Applications to household public good provision show how coordination failures emerge from attentional frictions. Embedding a Caplin–Dean rational inattention stage with upward-link cost economies demonstrates that upward connections reduce marginal information costs and expand optimal consideration sets, thereby connecting collective household choice under limited attention to rational inattention models of stochastic choice.
This paper introduces the Ascending Random Boolean Choice (ARBC) model, a stochastic choice framework for multidimensional decision-making in lattice-structured environments. Building on the Random Utility Maximization (RUM) tradition, ARBC extends monotone comparative statics to unordered alternatives by incorporating Strictly Superextremal (SSE) relations, which capture complementarity in stochastic choice. The model is axiomatized through Monotonicity and a novel SSE-Crossing condition, generalizing the single-crossing property to lattice structures. We establish a combinatorial random utility representation and demonstrate its implications for stochastic monotone choice. The results provide a foundation for analyzing economic decisions involving interdependent attributes, with applications to consumer behavior, demand systems, and strategic decision-making.
This note examines the uniqueness of Single-Crossing Random Utility Models (SCRUM) in rationalizing stochastic choice data. While Apesteguia et.al. (2017) show that datasets satisfying Centrality and Monotonicity have a unique SCRUM representation, we demonstrate that they can also be rationalized by a continuum of Random Utility Models (RUMs) that do not satisfy single-crossing. This challenges the necessity of single-crossing preferences in stochastic monotone comparative statics and highlights probabilistic indeterminacy in RUM representations. Unlike single-peaked or single-dipped RUMs, which permit multiple representations, SCRUM uniquely identifies support preferences, offering a more parsimonious structure for stochastic choice analysis.
I develop the Validated Random Choice with Limited Attention (VRC-LA) model, which extends random consideration to interdependent environments in which externally supplied “validations” can cause agents to discard their initial shortlist and choose only among validated options. An axiomatic characterization yields a representation for marginal stochastic choice that features both upward and downward attentional links, generating violations of standard positivity and novel comparative statics under peer influence. Identification is fragile: validation can systematically purge superior items from observed choices, obscuring underlying strict preferences. I also study a sparsified variant—Sparsely Validated Random Choice with Limited Attention (SVRC-LA)—in which only a thin set of validations can activate; this delivers sharper testable restrictions and partial identification results. The framework clarifies when shared information enhances versus distorts decision quality and provides tools for analyzing conformity, peer effects, and correlated choice in networked settings.