transactions, suggesting that packers lost money in about half of the transactions. This
suggests that market power, if any, is expected to be small.
An Indirect Test of the Encompassing Model
This section tests outlines the procedures to test our encompassing model. The test is
based on a single-equation regression of price spread on number of bidders and market-
level concentration. The test whether the hypothesis that an aggregate model (i.e. the
NEIO model) is consistent with the data against the hypothesis that the disaggregate
model (i.e. the structural auction model) is consistent with the data. This is a rather
indirect test. A direct test of our theory using equation (2.13) would require estimating a
bid-shading factor using the auction model, and use this estimate as an explanatory
variable in our NEIO like regression in the second step. Notice that estimation of bid
shading using equation (2.4a) requires data with at least two bidders in every transaction.
However, our experimental data contained numerous transactions with only one bidder,
precluding a meaningful estimation of markups using the structural auction model. This
is limitation of the structural auction model.
The test of an aggregate model against a disaggregate model is nonnested
because, in principle, neither of the two models can be obtained from the other by
imposing restrictions on parameters of either model. The encompassing test considered
here consists of artificially nesting the two candidate models within a single model, and
then carry out hypotheses tests.
The candidate models are single-equation regressions of a packer margin (i.e.
price spread between wholesale beef price and bid price) on a set of explanatory
variables. The encompassing model (M3) nests models (M1) and model (M2). Model
19