What Types are There?

From unobserved to observed preference heterogeneity: a revealed preference methodology

“…We start by briefly recapturing the Cosaert (2019) partitioning procedure, which is based on the original work of Crawford and Pendakur (2013). Next, we show how to recover preference heterogeneity contained within various observable characteristics.…”

“…Operationally, it boils down to partitioning the sample into subsets so that each individual subset satisfies the generalized axiom of revealed preference (GARP); the minimum number of preference types then corresponds to the partitioning with minimum cardinality (for more details, see Crawford and Pendakur 2013). More recently, Cosaert (2019) addresses the computational complexity of the original Crawford and Pendakur procedure. 1 He introduced a graph-theoretical approach to identify the number of types that is based on the weak axiom of revealed preferences (WARP) (instead of GARP).…”

“…1 He introduced a graph-theoretical approach to identify the number of types that is based on the weak axiom of revealed preferences (WARP) (instead of GARP). 2 Both Crawford and Pendakur (2013) and Cosaert (2019) focus on identifying the number of consumer types with similar preferences by analysing only the observed choice behaviour (prices and quantities), without directly including information on the consumers' observable characteristics. 3 We extend this earlier work by explicitly including these observed consumer characteristics in the preference identification analysis.…”

“…3 We extend this earlier work by explicitly including these observed consumer characteristics in the preference identification analysis. Building on the Cosaert (2019) partitioning procedure, we introduce a novel method that identifies the consumer characteristics that are most informative in describing the observed heterogeneity in choice behaviour. In particular, our method calculates a measure that quantifies the contribution of every observable characteristic to the non-parametrically identified preference heterogeneity.…”

Revealed preference theory: An algorithmic outlook

“…Crawford and Pendakur (2013) also consider this problem in the context of preference heterogeneity, and provide algorithms for computing upper and lower bounds on the number of 'dictators'. Cosaert (2017) links this to the problem of computing the chromatic number of a graph. Furthermore, Cosaert formulates an integer program to partition the observations into sets, so that the observed characteristics within each set are as homogenous as possible.…”