Now that the supply curves have been empirically estimated for cotton and peanuts, the
estimated coefficients and various exogenous variable combinations are used to calculate the
annual shifts in the supply curves resulting from expenditures in research. The annual shifts in
the supply curves are simulated using the estimated coefficients under two scenarios: with no
research expenditures on that commodity and with the actual research expenditures. Comparing
the two simulated models allows estimation of the increase in production resulting from the
commodity-specific research expenditures.4
The calculations of producer and consumer surplus require not only annual shifts in the
supply curve (ht), but in the production curve (kt) as well. The values for (kt) are obtained using
the price elasticity of supply for each commodity from the estimated models. These estimates
are assumed to be the appropriate price elasticities of supply for each commodity in period
covered in this study, so they can be used to get (kt). In addition to the values for (ht) and (kt),
calculation of the changes in consumer and producer surplus also requires price elasticities of
demand. For cotton we use an estimated elasticity of demand from White and Wetzstein (1995)
and for peanuts one from Zhang, Fletcher and Carley (1992).
The total and separate changes in producer and consumer surplus were calculated and are
presented in tables 3 and 4 for cotton and peanuts, respectively. The surplus resulting from
agricultural research expenditures was positive and substantial for every year in the period. The
annual social benefits from investment in cotton research over the period 1963 through 1995
ranged from about 317 million dollars to 114 million dollars in 1982 terms. For peanuts, the
annual social benefits ranged from $110 million to $336 million in 1982 dollars over the period.
4 The no research expenditure simulation must be compared to a simulation of production with research
expenditures, not to actual production so as to remove any bias due to the estimation of the coefficients and/or the
specification of the models.
14