equations containing equation (2.9b) and a supply equation was not well identified since
there was no variable in the demand equation that was not in the supply equation.
Following Paul’s (2001) suggestion, equation (2.9b) was estimated alone assuming
several values for cattle supply elasticity (0.2, 0.4, 0.8 and 1). Specifically the following
equation was estimated:
(2.21)
p' = ao + pf [1 - (1 '‘"HH ] + aiHHI, (Yf /SHOW, ) + v,
εsf
where p,r is the average price of boxed beef in week ,, p,f is the average cattle dressed
price, SHOW, is the total inventory of cattle in the show list a0, a1, a2 and Θ are
parameter to be estimated, and v,, is a error term.
To account for possible measurement error and endogeneity that leads to
inconsistent OLS because E[u, x,] ≠ 0, e quation (2.21) is estimated by nonlinear two-
stage least squares (N2SLS) using the MODEL Procedure, SAS 9.1 (SAS Institute, 2002-
03). The N2SLS estimator is consistent and asymptotically efficient when endogenous
variables are correlated with error terms (Zellner and Theil, 1962).
Results
Maximum likelihood estimates of the encompassing model represented by equation
(2.13) are shown in table 2. The estimates of interest are the coefficients of the
Herfindahl index (ω^17 = 11.95), and the coefficient for indicator variables for one bidder
(G14 = 0.47), two bidders (G15 = 0.38), and three bidders (G16 = 0.38). These
coefficients are significant at the 10% level, except the coefficient for the indicator
variable when for one bidder. Theory predicts that the price spread between beef price
and cattle price should decrease as the number of firms or/and the number of bidders
25