A model for how complexity aversion works
Xavier Gabaix’s paper introduces a tractable model of complexity aversion centered on the concept of “first-order complexity aversion”: when people recognize they are making mistakes due to situational complexity, they experience a psychological cost—a form of “dread”—that is proportional to the expected size of their error rather than the square of the error.
This model extends traditional utility functions by incorporating a complexity aversion cost that applies even when cognitive effort is zero, distinguishing it from mere information processing costs. The key innovation is that this dread cost persists regardless of attention level, making complexity reduction valuable even for highly behavioral consumers who invest minimal thinking effort.
Findings
- Simple pricing schemes
The model explains why firms like Whole Foods maintain uniform delivery pricing across all time slots despite varying marginal costs. When complexity aversion is sufficiently high, constant pricing becomes optimal to avoid burdening consumers with complex price systems. This result emerges when the marginal complexity aversion cost exceeds the marginal allocative benefits of variable pricing. - Simple tax systems
Traditional Ramsey taxation theory recommends different tax rates for each good based on demand elasticity, creating potentially thousands of different rates. With complexity-averse agents, the optimal tax system becomes dramatically simpler—featuring uniform tax rates when complexity aversion is large enough compared to the dispersion of elasticities across goods. - Aversion to indexation
While basic economic theory suggests most contracts should be indexed to aggregate risks like inflation or GDP to reduce consumer risk, complexity aversion explains why contracts typically remain non-indexed. The model shows that indexation becomes optimal only when risk-reduction effects are large enough to overcome the complexity costs of understanding indexed terms. - Inflation costs and monetary policy
Perhaps most significantly, complexity aversion generates first-order costs of inflation proportional to the absolute inflation rate, rather than traditional second-order costs proportional to the squared rate. This finding has profound implications for monetary policy, suggesting central banks should target exactly zero inflation in most circumstances. A rough calibration indicates that 4% annual inflation imposes complexity costs of approximately 2% of consumption—substantially larger than traditional estimates. This provides a behavioral foundation for commonly observed central bank inflation targets around 2%. - First-order vs. second-order complexity aversion
The paper demonstrates that first-order complexity aversion is essential for generating complete simplicity in economic mechanisms. Second-order complexity aversion, similar to regret theory, fails to robustly produce simple contracts and pricing schemes because the complexity costs become negligible compared to first-order allocative gains.
A Theory of Complexity Aversion
Author: Xavier Gabaix
From: Harvard University
Experimental and empirical evidence on complexity aversion
Two new papers, one by Konstantinos Georgalos and Nathan Nabil and another by Geoffroy de Clippel, Paola Moscariello, Pietro Ortoleva, and Kareen Rozen advance the understanding of complexity aversion from both descriptive (what models fit best) and explanatory (why the phenomenon occurs) perspectives. These working papers complement each other effectively: Georgalos and Nabil provide the empirical foundation showing that heuristic models better describe behavior under complexity, while de Clippel et al. provide the mechanistic understanding of why complexity aversion occurs through cognitive uncertainty and ambiguity-like caution.
Differences
- Georgalos and Nabil focus specifically on testing competing models of complexity aversion, particularly comparing heuristic-based approaches against expectation-based utility models. Their core question is whether people use simple rules of thumb (heuristics) or modified utility calculations when facing complex choices.
- de Clippel et al. take a broader approach, investigating the general phenomenon of complexity aversion across diverse domains and exploring its underlying psychological mechanisms. They examine whether people systematically undervalue complex options and what drives this behavior.
Data, experimental design, and statistical techniques
- Georgalos and Nabil
The re-analyze existing data from Moffatt et al. (2015) involving 80 subjects in lottery choice experiments. The authors use a two-phase experimental design with 27 tasks each. Subjects chose between lotteries with equal expected values but different complexity and risk levels. Complexity is defined as the number of outcomes in a lottery (simple: 3 outcomes, complex: 9 outcomes, very complex: 27 outcomes). Finally, they employ Hierarchical Bayesian econometric techniques for simultaneous estimation of individual-level parameters and group-level distributions. - de Clippel et al.
They conduct three new large-scale experiments with 2,245 subjects across diverse tasks. The authors test complexity aversion in multiple domains: belief updating, visual perception, and compound risk. They use Multiple Price Lists (MPL) to elicit dollar valuations of different bets, comparing complex options to 50/50 lottery benchmarks. Finally, they developed a novel constrained regression approach to avoid spurious correlations when both complexity aversion and ambiguity aversion are defined using the same benchmark values.
Findings
- Georgalos and Nabil
The study finds strong evidence that heuristic-based models outperform expectation-based models for the majority of subjects. The most prevalent heuristics are risk-averse strategies that protect against worst-case scenarios: Minimax (choosing the gamble with highest minimum outcome), Least Likely (avoiding gambles with high probability of worst outcomes), and Priority Heuristic (lexicographic comparisons). The authors provide empirical validation for adaptive toolbox models of cognition, and challenge the dominance of expectation-based models even when modified for complexity aversion. - de Clippel et al.
The study demonstrates systematic undervaluation of complex options across domains. The authors argue that complexity aversion stems from cognitive uncertainty interacting with ambiguity aversion. They suggest individuals dislike the cognitive uncertainty they experience with complexity in ways reminiscent of ambiguity aversion.
Practical implications
- Georgalos and Nabil
Suggest that financial products and decision environments should account for individuals’ reliance on heuristics rather than sophisticated calculations. - de Clippel et al.
Provide insights for policy design and product development, suggesting that simpler options may be more effective than complex alternatives even when the latter appear superior. The research supports the principle that complexity reduction can improve decision-making outcomes.
Testing Models of Complexity Aversion
Authors: Konstantinos Georgalos, Nathan Nabil
From: Lancaster University Management School
Caution in the Face of Complexity
Authors: Geoffroy de Clippel, Paola Moscariello, Pietro Ortoleva, Kareen Rozen
From: Brown University, Princeton University
