Household unit values, measurement units, and the current-share Stone index in LA/AIDS estimation

Household surveys often replace unavailable transaction prices with unit values. This article examines linear approximate Almost Ideal Demand System (LA/AIDS) applications in which raw or incompletely standardized household unit values enter a contemporaneous current-share Stone index. Building on the established non-commensurability of that index, it develops a household-level representation of the measurement-unit channel, isolates its transmission through the real-expenditure deflator, and quantifies the resulting sensitivity. An exact-AIDS Monte Carlo design with 500 replications per scenario shows that fixed commodity-specific unit scales generate elasticity differences through the Stone deflator; these grow with scale magnitude or dispersion and vanish in a homothetic null check. Corrected Stone and Törnqvist leave numerically negligible fixed-scale differences, with maximum mean absolute differences below 7×10−8. The Malawi application compares paired LA/AIDS specifications within a common sample while holding market-price regressors fixed. Standardized unit values are substantially more comparable with separately collected kilogram-based market-survey prices. The mean absolute centred raw–standardized index difference is 0.628 under current-share Stone, versus 0.493, 0.402, and 0.339 under corrected Stone, Törnqvist, and base-share indices; the mean absolute difference in conditional expenditure elasticities is 0.307. The evidence quantifies specification sensitivity for conditional purchased-food responses in this LA/AIDS implementation.