The name is absent

made in the household intertemporal consumption decision process, and are contained in the error
term eit. The parameters of good 1 (beef) appear in both equations with a10 normalized to 1. The
model is linear conditional upon the discount factor. Because a linear model is easier to handle,
we decide not to estimate β but fix it at 0.98.

In addition, there is another justification for preserving the linearity of the regression. Since
seasonally unadjusted data are used for estimation, seasonality has to be taken care of. If the
transformed variables in (9) are better characterized as nonstationary seasonal processes, seasonal
differencing is more appropriate. At quarterly frequencies, this amounts to fourth-differencing the
transformed variables in (9) which is feasible because of the linearity of the regression.

Under uncertainty and assuming rational expectations, the error term ∆4eit = eit - eit-4 that
contains the innovations is orthogonal to variables in the information set (Ω_t) as of period t. That
is, when evaluated at the true parameter values,

E[(∆4e2t ∆4e3t)'∣Ωt] =0.                                     (10)

The instrumental variables in the information set include choice variables dated t - 1 and earlier,
demand shfiters and prices dated t and earlier. Furthermore, eit is serially uncorrelated since it is
in the information set at period t + 1. The moment conditions used by the GMM estimation of
parameters can be summarized as

E [(∆4e2t∆4z21 ∆4e3t∆4z31)'] = 0                               (11)

where ∆4 z2t and ∆4 z3t are the corresponding vectors of fourth-differenced instruments.

Note that every variable in (9) is divided by an expenditure at t. Therefore, all of the trans-
formed variables are endogenous and have to be instrumented. To imitate the variables in (9) as
closely as possible, the list of instruments includes food safety variables dated t and t - 1 and
quantities demanded at t - 2, all of which were first divided by expenditures dated t - 2 and then
fourth-differenced. The beef/pork and beef/poultry equations were estimated jointly imposing
equality of the beef parameters across equations and bij = bji .

4.3 Results and discussion

There are many ways in which the newspaper indices discussed eariler could be used to represent
consumers’ perception of meat safety. Because there is no clear a priori reason for preferring one

12

<<   9   10   11   12   13   14   15   >>

More intriguing information