90
quarterly data have been redefined such that TMCIt and the prices which
may appear in each represent simple three-month averages of the
monthly data. The parameters OL^, OL^, L31t, and L21t represent three-
month totals of the monthly data.
Margin Models
Estimation of the monthly and quarterly price dependent demand
models provides a set of structural coefficients for each model. These
estimated coefficients can be utilized to derive a system of expressions
describing the price spreads (margins) between adjacent market levels
for each size class. Exact specification of these margin models is
conditional on the lead/lag structures identified by the causality
analysis. The interpretation of the derived estimates for prices in the
margin models strictly depends on the nature of price determination and
the identification of the causal market level(s). Non-price parameter
estimates in the margin models will vary in sign depending on which
prices are identified as exogenous or endogenous in the structural
models. In addition, the derivation of the margin estimates is condi
tional on whether structural, reduced form, or final form estimates are
utilized. A brief discussion of margins follows derived from structural
and reduced or final form estimates.
Structural Margins
This discussion of margin models assumes symmetric price response,
no price lags, and recursive price determination arbitrarily stated to
be upward between wholesale and retail prices. Given these conditions,
the margin between wholesale and retail price can be written simply as