Policy change: If market liberalization reduces the costs associated with grain distribution, this
is expected to decrease grain prices in deficit markets and increase grain prices in the surplus
markets. The coefficient on DLIB is thus expected to be positive in surplus markets and negative
in deficit markets. DLIB takes the value of 0 for the pre-liberalization period, values ranging from
zero to 1 during a transition period and the value of 1 when the full effects of liberalization have
been felt. This specification assumes that the effect of liberalization is a gradual process having
its full effect after certain period. A grid-search procedure is used to determine the length of the
transition period, choosing from a range of 6-month periods between 6 and 36 months. A
transition period of 12 months was found to maximize the value of the likelihood function.
Rainfall: Rainfall may have complex lagged effects that differ over the short and long run. The
overall effect of good rain conditions is to reduce grain price levels. The rainfall variable used in
the model is the unweighted 3-month moving average of rainfall, lagged one month, for the
markets considered in the analysis. This allows prices in the current period, for example, to be
affected by cumulative rainfall conditions over the previous one to three months.
Food aid: Food aid may affect market prices both by reducing the amount of grain that recipient
households may otherwise have purchased in the market (thus reducing demand), and by potential
sales of food aid onto markets, thus increasing market supplies). The food aid data used in this
analysis is disaggregated region-specific monthly food aid distribution from the National Disaster
Prevention and Preparedness Committee (DPPC). After the food aid reached the recipients it is
also assumed that it takes some time before food aid recipients potentially dispose of some part
of the food aid they received on the market or to change their decision whether to buy from the
market. To take into account these lagged effects, a three- month moving average of food aid
volume lagged by one month is used in the SURE model estimation. The effect of food aid on
market price levels is expected to be negative.
The specification of (2) may also be used to consistently estimate the effects of exogenous factors
on price spreads, through their effect on equilibrium price levels. In this approach, any two price
equations in the system (2) can be used to construct the price spread (PSij) for any two markets
i and j, through subtraction:
P,t — Pjt = PSj t
By subtracting the right-hand sides of any two price equations in (2), the price spread between
these two markets can be specified as follows (assumption of weakly integrated markets):
11
P = (μi-μj) + Σ>ni-^nj)Dn + FA1DDi - j + (RfiFι - rjRFj + YiPi,t_ 1 - γj 1 + zij,t (3)
where PSijt is the marketing margin between ith and jth market at time t, the other coefficients and
variables are as defined before, and zijt is the disturbance term. The coefficients on each variable
are derived directly from estimation of (2). The standard errors for the coefficients in (3) are also
derived from (2) as:
-7-