132
The 21-25 Size Class
Only the price model for the retail market level is estimated.
This was because the retail/wholesale market level interface was recur
sive. The wholesale/ex-vessel market level interface was characterized
by simultaneity. This limitation on the models presented is warranted
due to the necessity of performing the asymmetry test, which does not
lend itself to non-recursive methods.
Retail structural estimates
Empirical estimates of the parameters and corresponding standard
errors for the retail price model are given as
R .342 + .851 R + .066 W + .210 W .105 W
C (.952) (.039)^ (.085) t (.127) (.088) t~
- .027 FW .0007 RDY .0069 TCFF + .0029 CPI
(.035) C (.0013) (.0042)2 C (.0016)2 t
where the values in parentheses are standard errors and the subscripts 1
and 2 denote significance at the .01 and .10 level of confidence. The
2
model explained approximately 99 (R -.9897) percent of the variation in
retail price. Four of the eleven estimates were statistically
significant at the .25 level or greater.
The significant positive parameter estimate for retail price lagged
one period (R^j) again indicates that factors which affect current
retail price (R£) have a further impact on price in the next period.
Specifically, a one cent change in retail price causes a .85 cent change
in retail price in the following period. The lagged dependent variable
was included to account for first order serial correlation in the error
terms. The inclusion of this term reduced the error terms to white
noise