capita values are mid-quarter total U.S. population from the Department of Commerce, Bureau of
Economic Analysis(BEA).
Figure 2 illustrates the real quarterly per capita expenditures on beef, pork and poultry. Sea-
sonality is apparent in all three series. Beef expenditure peaks in the second and third quarter
corresponding with the summer grilling season. For poultry, expenditure is the highest in the
fourth quarter partly because of the high demand around Thanksgiving. Pork expenditure also
reaches its highest level in the fourth quarter, partly because demand for hams is strong during the
winter holiday season.
Apart from seasonality, meat expenditures appear to trend differently before and after 1998.
This is most evident for beef but less obvious for poultry and pork. In 1998, beef expenditure
reversed its nearly two-decade downward trend and continued to rise until the end of the sample.
One probable explanation is that the high protein-low carbohydrate diet fad may have shifted
consumer preferences toward meat products.
4 Empirical Specification and Results
4.1 Direct translog preferences
For this study, a flexible direct translog utility function augmented to include habits is used to
describe the household preferences over consumption goods 1 (beef), 2 (pork) and 3 (poultry)
3 13 3 3
ut = У? ailnxit + 2 ΣΣbijlnxitlnxjt + bilnxitlnxit-1 (6)
where ai, bij and bi are parameters to be estimated. Christensen and Manser (1977) estimated
U.S. consumer preferences for meats with a static direct translog utility function. The last term
on the right-hand-side of equation (6) is responsible for capturing any potential intertemporal non-
separability in the preferences. Meghir and Weber (1996) and Carrasco, Labeaga and L´pez-Salido
(2005)1 used the same preferences structure to test for intertemporal nonseparable preferences with
U.S. and Spanish micro data, respectively. Following the discussion in section 2, habit persistence
implies bi > 0 while durability or inventory adjustment2 implies bi < 0. If bi = 0 ∀i time separable
1Note that the parameter estimates interpreted by Carrasco, Labeaga and L´pez-Salido as evidence of habit
formation are actually evidence of inventory adjustment. This is because their results imply negative marginal effect
of past consumption on the marginal utility of current consumption—an indication of inventory adjustment.
2 In this paper, the terms inventory adjustment and durability are used interchangeably.
10