A theoretical model of semi-subsistence agricultural
households was developed which explicitly accounted for the ability of
households to store key food staples over the period between harvests. The model
yielded a simple inventory demand equation in
which carryout stocks are a linear function of
current consumption of the stored commodity and the difference between its
current and expected prices. These were interpreted as indicators of the
strength of arbitrage and food security motives for holding stocks.
Using panel data from three villages in southern
India, inventory demand equations for five groups
of stored food staples were econometrically estimated. In all villages, food
security motives generally dominated arbitrage motives in determining the level
of inventory demand. Empirically significant
arbitrage motives were found to exist only in the poorest of the three
villages.
Comparative statics analysis based on the theoretical
model indicated that stocks and expected revenue from future production will
have wealth (or profit) effects on current consumption, and that these affect
the response of marketed surplus through their effects on consumption and via
the price response of inventory demand.
Own-price elasticities of demand and marketed surplus for stored commodities were
computed. To do this, the parameters of commodity demand were econometrically estimated using a Rotterdam model. Theses were then combined with the
structural coefficients of inventory demand and
outside estimates of supply response. In most cases profit effects were
sizeable. Where the share of stocks in household wealth was quite large, profit
effects were strong enough to cause demand
elasticities to be positive. Computed marketed surplus elasticities were quite
variable, both within and across villages. In several instances these were found
to be negative.
Comparison of these elasticities with those computed
using traditional methods indicated that the traditional method yields larger
elasticities for both commodity demand and
marketed surplus. In several cases these differences were dramatic, the most
important being that all marketed surplus elasticities calculated using the
earlier method are positive.