long-run impacts of promotion and research are not too different in this model than in models with longer
lag lengths, and this model has superior performance from a theoretical standpoint (e.g., expected sign for
variables).
The export demand for U.S. cotton is specified as a partially reduced-form equation given by (16)
in which exports of U.S. cotton are modeled as a function of the price of domestic cotton, prices of
substitute fibers, and demand and supply shifters of foreign-produced textile products.20 The relevant
information provided by the export demand equation is the price elasticities of export demand with
respect to the price of U.S. cotton and the world cotton price.
The model was estimated in linear form21 using the following variables (the corresponding
variable name from (16) is included in parentheses):
Mi,t = monthly dummy variables (Mi=1 for ith month, 0 otherwise) for
i=1,...,11 where December is the reference month
EXPORTSt (Qfdx) = U.S. exports of raw cotton (thousands of bales)
PCOTTONt (Pcd) = U.S. real raw fiber equivalent price of cotton (cents/lb)
PPOLYt (Zr) = U.S. real raw fiber equivalent price of polyester (cents/lb)
FTEXWAGEt (Wm) = real foreign manufacturing wage ($/hour)
WPCOTTONt (Pcf) = real A Index of the world cotton price (cents/lb)
DECIt (Wm) = U.S. real energy cost index, used as a proxy for foreign energy costs
FGDPt (Zx) = foreign real GDP for OECD countries other than U.S. (billions of $)
ROWSTKt (Zx) = foreign cotton stocks (pounds)
EXPORTSt-1 (Qfdx) = lagged U.S. exports of raw cotton (thousands of bales)
rho = first-order autocorrelation parameter value
The foreign supply and demand shifters included in the model include prices of competing fibers
in the foreign fiber market (PPOLYt, WPCOTTONt), prices of other factors affecting foreign textile
manufacturing demand for U.S. cotton (FTEXWAGEt, DECIt, ROWSTKt), and foreign income (FGDPt).
20In addition, we estimated a variety of specifications including promotion and nonagricultural research. However,
neither seems to have a significant effect on export demand in our preferred model. Thus, in this paper, we focus on
the effects of the CRPP on domestic demand and exclude the impacts on exports.
21Other specifications were also estimated, including double-log and semi-log models, but the linear model cannot be
rejected based on the results from any of the models that we estimated. In other words, the linear model is at least as
good as any alternative models that we tried. Therefore, we chose to use the linear model because it is relatively
simple.
18