Our a priori expectation is that the own-effects of food safety information are negative, while
the cross-effects are less obvious. Inspecting the estimated parameters on food safety information
indicates that only three out of eighteen parameters are precisely estimated in mode 1. These results
suggest that the own- and cross-effects of beef safety information at t are negative. We refrain from
drawing inferences on the own- and cross-effects of pork and poultry safety information, because
these food safety coefficient estimates are not statistically different from zero.
The estimated food safety coefficients in mode 2 follow the similar pattern. In terms of the
effects of contemporary food safety information, only the own- and cross-effects of beef safety are
negative and statistically significant. The effects of pork and poultry safety information are not
precisely estimated. Interestingly, according to the results in mode 2, pork safety information at
t - 1 has statistically significant direct negative effects on demands for beef, pork, and poultry at
t. Because relatively few articles contained in the pork media index were exclusively about pork
safety, it is not clear from these results whether concerns over pork safety spill over into the demand
for beef and poultry.
If adverse health information reduces the degree of habit persistence of red meat, then the low
carb-high protein movement popular during the past a few years may have increased the degree
of habits for red meat. To exploit this hypothesis, a dummy variable, D98, that is equal to zero
prior to 1998 and one thereafter is included in the model. It interacts with the ai0’s in (7) and
with the bi’s in (6). The former interaction is intended to capture changes in the intercepts, while
the latter is designed to approximate any potential shift in the degree of habit persistence. At
quarterly frequencies, both mode 1 and 2 point an increase in the degree of beef habit around
1998. However, no statistically significant changes in pork and poultry habits are found. If the
low carb-high protein fad was the sole rationale for the estimated increase in beef habit, we would
expect that the degree of habit persistence for pork should also increase. For instance, the Atkins
Diet recommends consumption of bacon. It is conjectured that improvement in beef quality partly
due to the USDA certification program in the 1990s may have contributed to the increase in the
degree of beef habit.
Fourth-differencing appears to be sufficient to render the ∆4eit's white noise. Graphical inspec-
tion of the error terms in mode 1 and 2 shows that the assumption of homoscedastic innovation may
be difficult to maintain. Table (2) reports the standard Lagrange Multiplier tests for the absence
14