92
Depending on the direction of price determination as revealed in the
causality analysis, the coefficient estimates associated with these non
price parameters may vary in sign but possess the same absolute value
relative to the structural coefficient estimate. For a non-price
coefficient b^ in either the retail or wholesale structural model, a
recursive upward price causality relationship between adjacent retail
and wholesale market levels produces a corresponding coefficient in the
margin model with the same sign and absolute value as b^ found in the
retail model. Alternatively, for a coefficient b^ in either the retail
or wholesale structural model, a recursive downward price causality
relationship between adjacent market levels produces a corresponding
coefficient in the margin model with the same sign and absolute value as
b^ found in the retail model. Alternatively, for a coefficient b^ in
either the retail or wholesale structural model, a recursive downward
price causality relationship between adjacent market levels produces a
corresponding coefficient in the margin model with the opposite sign and
same absolute value as b^ as found in the wholesale model. Given this
method of derivation, the margins obtained from recursive structural
estimates allow inferences to be made about the effects on the margin
due to changes in exogenous variables from only one side of the
margin. This is due to the margin being expressed in terms of only the
causal price.
Reduced and Final Form Margins
Margin models can be derived from the reduced or final form coeffi
cient estimates of the structural price models. If the structural
models are recursive or simultaneous and do not contain lagged prices,