Overcoming Deterministic Limits to Robustness Tests of Decision-Making Given Incomplete Information: The State Contingent Analysis Approach

Abstract

Incomplete information may result in multiple factors combining to jointly affect the consequences of decision-making. The typical response to incomplete information has been tests of robustness and a fixed decisions’ capacity to withstand a wide variety of future conditions. But what of reversed contexts, where the revealed future alters decision-making via experience, learning and innovation such that the decision itself changes? In this paper we contrast a commonly applied expected value robustness metric to state contingent analysis which allows for learning and innovation. State contingent analysis views robustness as how decision-makers achieve profits across all future states by reallocating resources ex post to maximize payoffs and/or minimize losses via outputs that are conditionally specific. Consequently, the state-contingent approach enables researchers to identify the benefits and constraints of resource reallocation—rather than fixed decision-making—over plausible scenarios. Within SCA, scenarios can thus be uncoupled from the historical averages to explore rare events, even if never before experienced, including thin- and fat-tailed probability distribution outcomes and their impact on decision-making, innovation and future solutions. A case study assessment of water resource management in a large river basin provides the basis for our comparison. We find that expected value models mask innovation and adaptation reactions by decision-makers in response to external stimuli (e.g., increased droughts) and under-represent water reallocation outcomes. Conversely, state contingent models represent and report decision-maker reactions that can be more readily interpreted and linked to stimuli including policy interventions, expanding the study of complex human-water systems.