# 16-068/I (2016-08-30)

Author(s)
Paul van Bruggen, Erasmus University Rotterdam, the Netherlands
Keywords:
Nonparametric analysis, revealed preference, missing observations, GARP
JEL codes:
C14, D11

Varian (1988) introduced an important proposition regarding restrictions on consumption data if observations of the quantities of a good are missing. In this paper, a simple counterexample is presented to show that the original proof is incorrect, and a new proof is provided. The new proof is not based on choosing the missing quantities such that some bundles are revealed preferred to others, but rather on choosing the unknown quantities such that bundles are not revealed preferred to any bundle with a higher index.