Content uploaded by Fred Kuchler
Author content
All content in this area was uploaded by Fred Kuchler on Feb 13, 2017
Content may be subject to copyright.
Submitted Article
Demand for Whole-grain Bread Before and
After the Release of Dietary Guidelines
†
Lisa Mancino*, and Fred Kuchler
Lisa Mancino and Fred Kuchler are economists at the Economic Research Service,
U.S. Department of Agriculture, Washington, D.C.
*Correspondence may be sent to: Lmancino@ers.usda.gov
Submitted 8 April 2011; accepted 19 September 2011.
Abstract The federal government has issued Dietary Guidelines for
Americans seven times since 1980, but the 2005 whole-grain recommendation
was the first instance in which consumers were given a specific dietary target, that
whole grains should be at least half of their grain consumption. Anecdotal evidence
pointed to a unique result, an increase in demand for whole-grain foods.
Contemporaneous decreases in prices of whole-grain foods, relative to refined-grain
foods, however, confound the evidence. We show that for whole-grain bread, there
was an increase in retail demand even after accounting for price changes. Separate
models for higher- and lower-income consumers show that the demand shift
occurred among higher-income consumers, but not for lower-income consumers.
JEL codes: D12, I18.
Introduction
Since 1980, the U.S. Department of Health and Human Services and the
U.S. Department of Agriculture have worked together to develop and dis-
seminate the Dietary Guidelines for Americans, which provides information
and advice to help consumers choose a healthy eating pattern
1
. Though
updated every five years to reflect the most recent scientific evidence, the
guidelines have remained quite constant over time (Gifford, 2002). Indeed,
each new release of the Dietary Guidelines for Americans has encouraged
consumers to choose more fruits, vegetables, and grains while limiting
their intake of certain fats, sodium, and added sugars.
Despite these repeated and consistent recommendations, the average
American consumer does almost exactly the opposite by continuing to eat
too few fruits, vegetables, and whole grain foods, and taking in too much
†
The opinions expressed here are those of the authors and not necessarily those of the US Department of
Agriculture. We wish to thank the editor and reviewers for their comments. Any remaining errors are ours.
1
The 1995 Dietary Guidelines was the first mandated joint report, prepared according to the 1990
National Nutrition Monitoring and Related Research Act. Among the requirements, reports are
reviewed by a committee of experts, updated if necessary, and published every 5 years (National
Nutrition Monitoring and Related Research Act).
Published by Oxford University Press, on behalf of Agricultural and Applied Economics Association
2011.
Applied Economic Perspectives and Policy (2012) volume 34, number 1, pp. 76– 101.
doi:10.1093/aepp/ppr035
76
saturated fat, refined grains, added sugar and sodium (Krebs-Smith et al.
2010;Guenther et al. 2008). Given our persistent and generally poor diets,
some speculate (Rowe et al. 2011) that the Dietary Guidelines have had no
impact on consumers’ food choices
2
. From an economist’s perspective,
this suggests that the public resources devoted to this program might be
better directed elsewhere.
Some features of the 2005 Dietary Guidelines, however, represent a
marked departure from previous recommendations. For one, the 2005
Dietary Guidelines recommended that at least half of a person’s daily grain
intake come from whole grains. At the time, an explicit recommendation
about a healthier substitution was novel. The 2005 whole-grains recom-
mendation is also of interest because there is anecdotal evidence that the
recommendation had a noticeable effect on consumer behavior. An exami-
nation of at-home expenditures showed that whole-grain purchases
increased after the 2005 recommendations (Mancino, Kuchler, and Leibtag
2008). Similarly, analysis of consumption trends from the NPD Group
shows that, while whole-grain intake was steady from 1998 to 2005, it
increased 20% between 2005 and 2008 (Whole Grain Council 2009).
The problem with observing increased whole-grain expenditures, and
assuming that the Federal program induced these healthier food choices,
is that relative food prices changed at the same time (Mancino, Kuchler,
and Leibtag 2008). Thus, the observed shift could have been entirely the
result of prices (moving along the demand curve), not advice from the
Federal government (shifting the demand curve). For example,
Dharmasena, Capps, and Clauson (2011) investigated a similar question:
Did the 2000 Dietary Guidelines’ recommendation to reduce consumption
of sugar-sweetened beverages have an impact? While they found that calo-
ries from non-alcoholic beverages did shift downward after 2000, they did
not clearly distinguish whether this shift was attributable to changes in
consumer price sensitivity and other demand parameters, or was simply
the effect of a time trend. The policy problem raised by price-induced
dietary improvements is that prices are not a lever that policy-makers can
easily flip. Further, if such improvements in food choices were fully the
result of changing relative prices, better diets would not last any longer
than the price movement.
It is also possible that a major government study or recommendation on
the benefits of whole grains would be newsworthy and temporarily
increase demand for whole-grain foods: food safety issues have a long
history of short-lived influences on food choices through the volume of
reports in the news (see Piggott and Marsh 2004, for a summary of this liter-
ature). Chronic health issues have been investigated similarly, using the
flow of clinical articles discussing the linkage between dietary intake of
cholesterol, serum cholesterol, and health outcomes (Brown and Schrader
1990). The problem these studies point to is that if demand quickly reverts
to its pre-news state when newsworthiness declines, there is no impact on
diet quality.
In this paper we provide evidence that the increase in whole-grain con-
sumption following the recommendation to eat more whole grains was
2
Conceptually, it is possible that the Dietary Guidelines have been effective and that average diet
quality would be much worse in their absence. Rowe and others (2011) began with the idea that,
“What has been done till now isn’t working.”
Demand for whole-grain bread
77
not just a temporary response to prices; demand for whole-grains
increased after accounting for relative prices. While we cannot prove that
the whole-grain recommendation in the 2005 Dietary Guidelines was
responsible for consumers’ healthier choices, our evidence is consistent
with that possibility.
We estimate an Almost Ideal Demand System (AIDS) model (Deaton and
Muelbauer 1980) to test for distinct and sustained differences in whole-grain
demand both before and after the release of the 2005 Dietary Guidelines.
Estimating retail food demands allows us to distinguish between changes in
purchases that might be the result of the Dietary Guidelines and confounding
price changes. We show that demands in the two periods are distinctly differ-
ent. Simulating both demand systems using post-2005 prices and expendi-
tures fully controls for confounding effects of relative prices. The remaining
differences can be attributed to changes in demand.
We look at monthly purchases in the US over five years, from January
2003 through December 2007. The path of adjustment is not important for
our purposes; we focus on whether demand differs before and after
release of the Dietary Guidelines. We also compare the impact of the 2005
Dietary Guidelines to that of the 2000 Dietary Guidelines, which referenced
whole grains as well. While the 2005 recommendations were more specific,
asking consumers to increase by half their daily grain intake to whole-
grain, the 2000 Dietary Guidelines recommended that individuals “choose a
variety of grains daily, especially whole grains.” On the one hand, it is
possible that shoppers make consistent but temporary dietary changes in
response to dietary recommendations to eat whole-grain foods. As such,
the response to both dietary guidelines would be similar and short-lived.
On the other hand, making the whole-grain recommendations more spe-
cific in 2005 may have given consumers (and producers who might have
had to ramp up production of whole-grain bread to meet changing con-
sumer demands) a more concrete target. For this sensitivity analysis, we
use a parallel time frame by again tracking 5 years of monthly purchase
data between 1998 and 2002.
We focus solely on identifying changes in demand for bread. Here,
bread is characterized as three related products that are differentiated by
their grain attributes: refined-grain, whole-grain and multigrain bread
products. We concentrate on this single product category for two reasons:
first, compared to cereal, rice, pasta, barley, and flour, bread purchases
comprise the largest share of food-at-home grain purchases. In terms of
Figure 1 Monthly quantity share of grain product purchases 1998-2007
Applied Economic Perspectives and Policy
78
total pounds per month, they account for almost half of all grain pur-
chases (figure 1).
3
Second, substitutions among bread types should be
fairly straightforward for consumers. Switching to whole-grain bread may
require an adjustment – consumers may or may not agree that whole-
grain products taste as good as their refined-grain products. However, the
form in which bread is sold and consumed likely does not vary among
whole-grain, refined-grain, and multigrain breads.
As the bulk of USDA food and nutrition assistance programs seek to
improve diet quality among lower-income consumers, it is important to
understand how both information campaigns and prices influence lower-
income consumers’ eating patterns. Thus, we also run separate analyses
for higher- and lower-income households; this tests whether prices and
expenditure constraints appear to be relatively more of a barrier among
lower-income households for purchasing whole-grains compared to
higher-income households. Our analysis provides insight into the relative
merits of three possible policies for improving diet quality – dietary guid-
ance, pricing policies, and income support programs. The analysis also
explores how to increase the efficacy of policies aimed at improving diet
quality among lower-income consumers.
Estimation Approach and Data
We use the basic AIDS specification derived by Deaton and Muellbauer
(1980) to estimate the demand for refined-grain, whole-grain and multi-
grain bread,
4
but we modify the basic model in three ways. First, we
include regional dummy variables to account for spatial variation in
supply costs, as well as geographical variation in American diets. The
USDA’s Food Environment Atlas (http://ers.usda.gov/foodatlas) docu-
ments some of the ways in which food consumption varies geographically.
Our assumption is that even demand for staples such as bread may vary.
To account for geographic variation in bread demand, we calculated
prices, expenditures, and budget shares for 13 regions of the United States
(a description of which is provided below). Thus, our demand model
assumes regional fixed effects (Hausman and Leonard, 2005).
Second, we allow for systematic variation in share equations across
time. For many foods, consumption has a cyclical or annual seasonal var-
iation (Anande, Pick, and Gehlhar 2005). There may also be long-term
upward or downward trends. To directly test for significant differences
before and after the dietary guidelines, we make a third adjustment by
allowing our demand estimates to switch depending upon the time
period. A detailed explanation of our demand model and how it incorpo-
rates these three adjustments appears in the Appendix.
We used the Nielsen Homescan panel data set to estimate our demand
model. This is a nationwide panel of households that uses a scanning
device to scan the universal product codes (UPCs) on purchased products.
Participants scan their food purchases from all retail outlets at home after
they finish shopping. Data include records for each food item purchased
by each household. Using data from January 2003 through December
3
Barley purchases were excluded from the figure because they were too small to register.
4
We use the non-linear price index and impose the standard economic assumptions of homogeneity,
symmetry and adding up.
Demand for whole-grain bread
79
2007, we tested whether coefficients and resulting elasticity estimates
differ significantly before and after the release of the 2005 Dietary
Guidelines (released in January 2005). To gauge the impact of the 2005
guidelines relative to the 2000 guidelines, we also ran these same AIDS
equations to test for significant differences before and after the release of
the 2000 Dietary Guidelines (released in May, 2000). Using the 2003-07 data,
we also separately tested whether the before and after parameter differen-
ces are significant for lower- and higher-income households.
A strength of these data is that they have detailed information on the
types of foods purchased, the prices paid, and the exact day when each
household purchase was made. Thus, we were able to construct relatively
high frequency (monthly) purchase data. Also, the data come from a
nationwide sample of households, where selection is based on both demo-
graphic and geographic targets. The weighted proportion of households in
the sample matches the proportions of households in the U.S. Census
across 8 variables: household size, income, race, ethnicity (Hispanic or
not), female household head’s age and education, male household head’s
education, and the household head’s occupation. Each year, Nielsen recal-
culates household weights and adjusts the sample to match annual
updates to the census. The sample weights were used in the estimation
and analysis.
For each region, we constructed a monthly time series for the purchases
of three broad categories of bread, denoted here as refined-grain,
whole-grain and multigrain. Aggregating up to this level circumvents the
problems of zero purchases and censoring (a common problem when ana-
lyzing household purchase data that leads to inconsistent parameter esti-
mates). The demand models require expenditure data, which we created
from the Nielsen Homescan panel as the monthly sum of weighted expen-
ditures on each bread type for each region. Similarly, monthly quantities
were weighted sums across households for each month. Prices were calcu-
lated as unit values, dividing expenditures by quantity for each month
and region. Budget shares were calculated from monthly expenditures for
each bread type, divided by total expenditures for bread in each region.
Nielsen data does not contain specific codes for grain attributes.
However, the dataset includes a variety of abbreviations in its UPC
description field that can be combined to identify whole-grain products.
For example, any bread with a UPC description that indicated it was
whole grain, whole wheat, or whole multigrain was identified as whole-
grain bread.
Among the remaining breads not defined as whole-grains, we searched
their UPC descriptions for abbreviations indicating that it was labeled as
multigrain, brown, 100% wheat or stoneground. That is, we formed a sep-
arate group for non-whole-grain products that make an explicit claim
about the grain ingredients. We refer to these products as “multigrain
bread.” While these products may be manufactured from refined-grains
and offer no health benefits, there are at least two reasons why manufac-
turers might want to highlight the attributes of the (non-whole) grains. If
some consumers like the flavor, manufacturers can help consumers dis-
criminate among breads to find the products with the flavor they like best
and charge a price premium. Alternatively, the claim may suggest health-
fulness to some consumers. As long as the claim is true, it is legally per-
missible, even if it is meaningless from a health perspective. Some
Applied Economic Perspectives and Policy
80
manufacturers could use that ambiguity to their advantage by charging a
price premium. All remaining breads were classified as refined-grain.
Regions were defined by splitting the sample of households along two
geographic variables. Households in urban areas were grouped into one
of nine census divisions
5
; households outside of urban areas were
grouped into one of four non-urban census regions. This provided 13
observations each month. With 13 geographic markets over 60 months, we
were able to use a total of 780 bread-purchase observations over five
years, with 468 of those occurring after their release in January, 2005. We
also used 780 observations taken over 5 years for testing the impact of the
2000 Dietary Guidelines, with 403 observations occurring after their
release in May, 2000.
A benefit of using Nielsen Homescan panel data over point-of-sale (store
level) scanner data for this analysis is that the former includes household
demographic data. This allowed us to obtain separate price, expenditure,
and budget share estimates for lower-income and higher-income house-
holds. We classified each sampled household as higher- or lower-income
by comparing each household’s midpoint categorical income value condi-
tioned on household size to income cutoffs in annual poverty tables. Our
classification places households into categories depending on where house-
hold income is relative to 185% of the federal government’s poverty line
6
.
For the subset of households falling at or below the 185% poverty level
(which we will refer to as lower-income households), we calculated
monthly purchase quantities, expenditures, and prices for the 13 geo-
graphic divisions. We also did this for households whose income was
above 185% poverty level. For simplicity, we refer to this group as higher-
income households. For the full sample, lower-income sample, and higher-
income sample, we then used monthly prices, quantities, and expenditures
in each of the 13 geographic locations to estimate the demand for bread.
A potential drawback of using household level data for this analysis is
that prices should be exogeneous, but we used unit values (the total
amount spent divided by the total amount purchased), as if they were
prices. Unit values may reflect more than simply the variation in supply
cost. Unobserved differences, such as preferences for quality, access to
retail outlets and the ability to take advantage of volume discounts can
also affect unit costs, and thus bias our parameter estimates. However, the
way in which we constructed our price data should mitigate much of this
problem. First, our use of a regional, fixed-effects model will remove any
time invariant, unobserved differences. Also, bread is a relatively homoge-
nous product; especially since we only track UPC coded products that
may exclude many of the higher-priced breads from the bakery. We also
separated by grain type, which is likely one of the main sources of quality
differences. Finally, we ran separate estimates for higher- and lower-
income households.
Another source of endogeneity may arise from including total bread
expenditures as an explanatory variable. As pointed out by LaFrance (1991),
5
The nine urban census regions are Pacific, Mountain, West North Central, West South Central, East
North Central, East South Central, New England, Mid-Atlantic, and South Atlantic. Four non-urban
census regions are Western, North Central, Northeast and Southern.
6
This threshold is often used to categorize lower-income households, as it is the monetary cut-off for
certain food and nutrition assistance programs, such as the Supplemental Nutrition Assistance
Program (SNAP).
Demand for whole-grain bread
81
endogeneity in total category expenditures is often an issue when estimat-
ing conditional elasticities. Unfortunately, we do not have any strong
instrument(s) to circumvent this issue. Further, using a weak instrument
would not represent a marked improvement over our current approach.
Although our approach essentially assumes that total bread expenditures
are predetermined, which could be seen as inconsistent with utility maxi-
mization, our data suggest that total bread expenditures are in fact quite
constant over the examined time period and thus may be pre-determined
through dietary patterns and habit. However, this is not at odds with the
context of our study, as the question we explore is whether the mix of
bread purchased can be changed through dietary guidance.
Results
Summary statistics over both five-year periods are presented in table 1,
the top panel of which shows that between 1998 and 2002, total bread con-
sumption averaged 4.72 pounds per household per month, or approxi-
mately one loaf per week. In terms of expenditure shares, 86% was spent
on refined-grain breads, while whole-grain and multigrain breads each
accounted for 7%. Comparing purchased quantities after the release of the
2000 Dietary Guidelines, which advised consumers to choose a variety of
grains, with purchased quantities prior to its release shows that refined-
grain bread sales dropped almost 6%, whole-grain breads increased 5%,
and multigrain bread sales rose 10%.
The lower panel of table 1shows that from 2003-2007, total bread pur-
chases were lower than they were from 1998-2002, averaging 4.33 pounds
per household each month. Whole-grain bread accounted for 17% of total
expenditure shares, and multigrain bread accounted for another 8%.
Comparing purchased quantities of bread before and after the release of
the 2005 Dietary Guidelines shows a more marked pre- and post-difference
compared to the 2000 Dietary Guidelines. Refined-grain bread sales fell by
13%, multigrain bread sales increased by less than 3%, and whole-grain
bread purchases rose by almost 70%.
These summary statistics show that whole-grain purchases have been
trending upward, refined-grain purchases have been trending downward,
and trends for both grain types have become more pronounced since the
release of the 2005 Dietary Guidelines. However, without controlling for
changes in prices, expenditures and other drivers of demand, it is unclear
whether these changes mark an information-induced change in demand
or simply a response to changes in relative prices. As such, we estimated a
series of demand models to test if and how consumers have changed their
responses to prices and expenditures over this time period.
Table 2shows parameter estimates and associated standard errors for
the all-consumer demand model estimated with iterative seemingly unre-
lated regression
7
. The multigrain budget share equation was dropped for
the estimation, and its parameter estimates were recovered from adding
up, symmetry, and homogeneity. Panel A compares parameter estimates
before and after the release of the 2000 Dietary Guidelines. The SR-squared
7
We estimated the bread demand system using the Model procedure in the SAS 9.2 statistical software
program.
Applied Economic Perspectives and Policy
82
Table 1 Descriptive statistics for time series data in the bread demand models
Full sample January 1998-December
2002 (N5780)
Pre-2000 Dietary
Guidelines* (N5377)
Post-2000 Dietary
Guidelines* (N5403)
Variable Mean Standard Error Mean Standard Error Mean Standard Error Unit of measurement
Refined-grain bread
Quantity 4.13 0.02 4.27 0.03 4.00 0.03 Pounds per household per month
Price 1.05 0.00 1.02 0.01 1.08 0.01 $/pound
Share 0.86 0.00 0.87 0.00 0.85 0.00 % of bread expenditures
Whole-grain bread
Quantity 0.30 0.00 0.28 0.01 0.32 0.01 Pounds per household per month
Price 1.24 0.01 1.18 0.01 1.30 0.01 $/pound
Share 0.07 0.00 0.07 0.00 0.08 0.00 % of bread expenditures
Multigrain bread
Quantity 0.29 0.00 0.28 0.01 0.31 0.01 Pounds per household per month
Price 1.20 0.01 1.12 0.01 1.27 0.01 $/pound
Share 0.07 0.00 0.06 0.00 0.07 0.00 % of bread expenditures
Continued
Demand for whole-grain bread
83
Table 1 Continued
Full sample January 2003-December
2007 (N 5780)
Pre-2005 Dietary
Guidelines** (N 5312)
Post-2005 Dietary
Guidelines** (N 5468)
Variable Mean Standard Error Mean Standard Error Mean Standard Error Unit of measurement
Refined-grain bread
Quantity 3.39 0.02 3.68 0.02 3.20 0.01 Pounds per household per month
Price 1.22 0.01 1.15 0.01 1.27 0.01 $/pound
Share 0.75 0.00 0.80 0.00 0.72 0.00 % of bread expenditures
Whole-grain bread
Quantity 0.62 0.01 0.44 0.01 0.74 0.01 Pounds per household per month
Price 1.48 0.01 1.41 0.01 1.52 0.01 $/pound
Share 0.17 0.00 0.12 0.00 0.20 0.00 % of bread expenditures
Multigrain bread
Quantity 0.32 0.00 0.31 0.01 0.32 0.00 Pounds per household per month
Price 1.48 0.01 1.39 0.01 1.53 0.01 $/pound
Share 0.08 0.00 0.08 0.00 0.08 0.00 % of bread expenditures
* The 2000 Dietary Guidelines were released in May, 2000.
** The 2005 Dietary Guidelines were released in January, 2005.
Applied Economic Perspectives and Policy
84
Table 2 Parameter estimates from the bread demand model, full sample
Panel A (January 1998 through December 2002) Panel B (January 2003 through December 2007)
Prior to release of 2000
Dietary Guidelines
Post release of 2000
Dietary Guidelines
Prior to release of 2005
Dietary Guidelines
Post release of 2005
Dietary Guildelines
Parameter Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error
Intercepts
Refined grain 0.880 0.03 *** 0.829 0.02 *** 1.017 0.03 *** 1.002 0.05 ***
Whole grain 0.024 0.02 0.057 0.02 *** 20.097 0.02 *** 20.079 0.04 **
Multigrain 0.096 0.02 *** 0.113 0.02 *** 0.080 0.02 *** 0.077 0.03 ***
Price effects
Refined x refined grain 0.029 0.02 * 0.041 0.02 ** 0.014 0.02 ++ 20.082 0.03 **
Refined x whole grain 20.006 0.01 20.014 0.01 0.002 0.01 +++ 0.114 0.03 ***
Refined x multigrain 20.023 0.01 ** 20.026 0.01 ** 20.016 0.01 20.032 0.02 **
Whole x whole grain 0.015 0.01 0.034 0.01 *** 0.030 0.01 ** +++ 20.119 0.03 ***
Whole x multigrain 20.009 0.01 20.020 0.01 *** 20.033 0.01 *** +++ 0.005 0.01
Multi x multigrain 0.032 0.01 *** 0.046 0.01 *** 0.049 0.01 *** 0.026 0.01 **
Expenditure effects
Refined grain 20.007 0.02 0.011 0.01 20.050 0.01 *** 20.059 0.03 **
Whole grain 0.026 0.01 ** +0.000 0.01 0.051 0.01 *** 0.026 0.02
Multigrain 20.019 0.01 20.012 0.01 20.001 0.01 +0.033 0.01 **
Time trends
Refined grain 20.001 0.00 *** 20.001 0.00 *** 20.002 0.00 *** 20.002 0.00 ***
Continued
Demand for whole-grain bread
85
Table 2 Continued
Panel A (January 1998 through December 2002) Panel B (January 2003 through December 2007)
Prior to release of 2000
Dietary Guidelines
Post release of 2000
Dietary Guidelines
Prior to release of 2005
Dietary Guidelines
Post release of 2005
Dietary Guildelines
Parameter Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error
Whole grain 0.001 0.00 *** 0.001 0.00 *** 0.002 0.00 *** +++ 0.003 0.00 ***
Multigrain 0.000 0.00 ** 0.000 0.00 0.000 0.00 ** +++ 0.000 0.00 ***
Seasonal effects
Sine-Refined grain 20.004 0.00 * ++ 0.003 0.00 20.005 0.00 *** ++ 0.000 0.00
Sine-Whole grain 0.004 0.00 *** +++ 20.004 0.00 *** 0.005 0.00 *** ++ 0.000 0.00
Sine-Multigrain 20.001 0.00 0.001 0.00 0.000 0.00 0.000 0.00
Cosine-Refined grain 0.001 0.00 +0.007 0.00 *** 0.000 0.00 +0.004 0.00 ***
Cosine-Whole grain 0.001 0.00 ++ 20.004 0.00 ** 20.001 0.00 20.003 0.00 **
Cosine-Multigrain 20.002 0.00 20.003 0.00 ** 0.000 0.00 20.001 0.00 *
Restrictions
Homogeneity, refined grain 11.706 35.92 98.876 35.36 *** 305.991 45.71 *** 225.244 30.46
Homogeneity, whole grain 36.208 78.84 16.957 68.37 112.813 75.15 63.165 42.95
Homogeneity, multigrain 1.272 0.42 *** 0.024 0.14 2.804 1.12 ** 22.848 0.83 ***
Adding up 20.840 1.99 21.130 0.41 *** 9.669 3.45 *** 23.083 1.73 *
Coefficient is significantly different than zero at *p ,0.10, * *p ,0.05, and ***p ,0.01.
Difference in coefficients between the two time periods is significant at +p,.10, ++ p,.05, and +++p,.01.
Applied Economic Perspectives and Policy
86
value (Jitthavech 2010, page 521), which assesses the fit of the entire
system, was .498. Panel B compares parameter estimates before and after
the release of the 2005 Dietary Guidelines. In this later sample, the SR-
squared value increased to .750. The center column in both panels indi-
cates whether the parameter estimates differed significantly before and
after the release of these guidelines. For both the 1998-2002 and 2003-2007
data, Chow tests reject the null hypothesis with p ,.0001 that there was
no significant change in parameter estimates after the release of either the
2000 or 2005 Dietary Guidelines. Similarly, Likelihood Ratio Tests compar-
ing an unrestricted model that pools the time periods to models that split
the two samples into before and after the release of Dietary Guidelines indi-
cates that removing these splits would significantly reduce the models’ fit.
Comparing parameter estimates before and after the release of both sets
of the Dietary Guidelines provides insight into the specific nature of
changes in consumer demand. Among the intercept terms, we find no dif-
ferences before and after either set of guidelines. We also find few signifi-
cant differences among the expenditure terms. Among whole-grain price
effect variables, however, we find significant differences before and after
the release of the 2005 Dietary Guidelines – the cross-price effects of both
refined-grain and multigrain breads increased, while the own-price effect
decreased. The differences before and after the 2005 Dietary Guidelines
among time trend variables were also statistically significant. We find no
significant differences in these same variables before and after the 2000
Dietary Guidelines.
We next compare elasticities before and after Dietary Guidelines (table 3).
From a public health perspective, it would be preferable for whole-grain,
own-price elasticities to become more inelastic – regardless of its price,
people should be willing to buy it. Similarly, the cross-price elasticities
should become less significant – regardless of the price of substitutes like
refined- or multigrain-bread, people should choose whole-grain bread.
Since there were few significant differences after the release of the 2000
Dietary Guidelines, we conclude that these recommendations did not alter
demands. Thus, we only compare elasticity estimates before and after the
release of the 2005 Dietary Guidelines
8
. For each grain type, the center row
indicates whether the parameter estimates differed significantly before
and after the release of these guidelines.
After the release of the 2005 Dietary Guidelines, the cross-price elastic-
ity for refined-grain bread and whole-grain bread increased signifi-
cantly, indicating that consumers began to view the two bread types as
substitutes. Also, the own-price elasticity of whole grains became more
pronounced. Prior to the release of the 2005 Dietary Guidelines,a1%
decrease in the price of whole-grain bread would only bring about a
0.76% increase in whole-grain bread purchases. After the release of the
Dietary Guidelines, consumer demand for whole-grain bread became
much more elastic. A 1% drop in prices would bring about a 1.63%
increase in purchases.
The change in own- and cross-price elasticities after publication of the
2005 Dietary Guidelines underscores the importance of accounting for
8
We found only one significant difference in elasticity estimates for whole-grains after the release of the
2000 Dietary Guidelines compared to before – the cross-price relationship between refined-grain
bread and whole-grain bread becomes insignificant after the 2000 dietary recommendations.
Demand for whole-grain bread
87
price changes over this time period. Consumers became more sensitive
to changes in the price of whole grains, while the price of whole-grain
relative to refined-grain breads was decreasing. Thus, simply comparing
expenditures before and after 2005 would exaggerate the impact of these
recommendations. To better illustrate this, we simulate post-2005 bread
expenditure shares in absence of the 2005 Dietary Guidelines. Specifically,
we project post-2005 budget shares with post-2005 prices and expendi-
tures using pre-2005 demand parameter estimates from table 2, panel B.
The difference between the lines with and without the 2005 Dietary
Guidelines, the residual change, accounts for prices and expenditures. As
such, it is a more accurate depiction of the change in consumer demand
that can be attributed to publication of the 2005 Dietary Guidelines than is a
simple comparison of quantities purchased before and after release. After
accounting for price changes, we find there is a difference between the
predicted lines with and without the 2005 Dietary Guidelines: in their
absence, refined-grain budget shares would have been higher and whole-
grain shares would have been lower. At the end of the simulated period,
refined-grain budget shares were approximately 0.70 without the Dietary
Guidelines and 0.68 with the Dietary Guidelines, while whole-grain budget
shares were approximately 0.20 without the Dietary Guidelines and 0.24
with the Dietary Guidelines.
On average over the post-2005 period, and after accounting for the price
changes that also occurred after the 2005 Dietary Guidelines, consumers
Table 3 Testing for differences before and after release of the 2005 Dietary
Guidelines
Elasticity Estimates
(standard errors)
Grain
Type
Refined
grain price
Whole
grain price
Multi
grain price Expenditure
R
e
f
i
n
e
d
Before
2005 DG
20.93 *** 0.00 20.02 0.94 ***
(0.02) (0.02) (0.01) (0.02)
+++ +++
After
2005 DG
21.05 *** 0.17 *** 20.04 * 0.98 ***
(0.04) (0.04) (0.02) (0.11)
W
h
o
l
e
Before
2005 DG
20.38 *** 20.74 *** 20.33 *** 1.45 ***
(0.11) (0.12) (0.07) (0.11)
+++ +++ +++ ++
After
2005 DG
0.47 *** 21.62 *** 0.02 1.13 ***
(0.11) (0.14) (0.06) (0.11)
M
u
l
t
i
Before
2005 DG
20.18 20.39 *** 20.41 *** 0.98 ***
(0.12) (0.09) (0.11) (0.11)
++ ++ + +
After
2005 DG
20.68 *** 0.00 20.71 *** 1.38 ***
(0.17) (0.15) (0.13) (0.17)
Coefficient is significantly different than zero at *p ,0.10, **p ,0.05, and ***p ,0.01.
Difference in coefficients between the two time periods is significant at +p,.10, ++p,.05, and
+++p,.01.
Applied Economic Perspectives and Policy
88
reduced their expenditures and purchases of refined-grain bread by
approximately 3% and increased whole-grain purchases by 14%. This
works out to roughly one less 16-ounce loaf of refined-grain bread per
year, and one more whole-grain loaf of bread. Looking at the percentage
change in quantity over this time and using the own-price elasticities from
table 3, the change in demand after the 2005 recommendations was tanta-
mount to raising the price of refined-grain breads by 2% and lowering the
price of whole-grain breads by 12% over this same time period.
To assess how benefits from dietary guidance are distributed, we esti-
mate separate AIDS models for higher- and lower-income households
before and after the 2005 Dietary Guidelines. Summary statistics by income
group are reported in table 4. Descriptive statistics for the higher-income
households are similar to the pooled sample over this same time period.
In terms of total quantity, higher-income households purchased slightly
less bread than average and lower-income households purchased slightly
more. In terms of budget share, refined-grain bread purchases were higher
prior to the 2005 Dietary Guidelines for both higher- and lower-income
households. The whole-grain share increased more among higher-income
consumers, from 13-21%. For lower-income consumers, whole-grain
expenditure shares increased from 9-16%.
Using Chow and Likelihood Ratio tests, we again reject the null hypoth-
esis of no significant change after the 2005 Dietary Guidelines at p ,.001.
In terms of model fit, the SR-squared value was .655 for the higher income
sample and .67 for the lower income sample. When looking at differences
between parameter estimates (table 5), we find roughly the same story
Figure 2 Simulated monthly budget shares for refined-grain, whole-grain and multigrain
breads, with and without the influence of the 2005 Dietary Guidelines
Demand for whole-grain bread
89
Table 4 Descriptive statistics for time series data in bread demand models, by income group
Higher-income sample* January 2003–December
2007 (N 5780)
Pre-2005 Dietary
Guidelines (N 5312)
Post-2005 Dietary
Guidelines (N 5468)
Variable Mean Standard Error Mean Standard Error Mean Standard Error Unit of measurement
Refned-grain bread
Quantity 3.39 0.02 3.42 0.02 2.97 0.01 Pounds per household per month
Price 1.22 0.01 1.23 0.01 1.35 0.01 $/pound
Share 0.75 0.00 0.78 0.00 0.70 0.00 % of bread expenditures
Whole-grain bread
Quantity 0.62 0.01 0.48 0.01 0.79 0.01 Pounds per household per month
Price 1.48 0.01 1.43 0.01 1.56 0.01 $/pound
Share 0.17 0.00 0.13 0.00 0.21 0.00 % of bread expenditures
Multigrain bread
Quantity 0.32 0.00 0.34 0.01 0.34 0.00 Pounds per household per month
Price 1.48 0.01 1.43 0.01 1.56 0.01 $/pound
Share 0.08 0.00 0.09 0.00 0.09 0.00 % of bread expenditures
Applied Economic Perspectives and Policy
90
Refned-grain bread
Quantity 3.97 0.02 4.26 0.04 3.77 0.02 Pounds per household per month
Price 1.05 0.00 0.99 0.01 1.09 0.01 $/pound
Share 0.80 0.00 0.85 0.00 0.77 0.00 % of bread expenditures
Whole-grain bread
Quantity 0.49 0.01 0.34 0.01 0.59 0.01 Pounds per household per month
Price 1.40 0.01 1.37 0.01 1.42 0.01 $/pound
Share 0.13 0.00 0.09 0.00 0.16 0.00 % of bread expenditures
Multigrain bread
Quantity 0.25 0.00 0.25 0.01 0.26 0.00 Pounds per household per month
Price 1.37 0.01 1.29 0.01 1.42 0.01 $/pound
Share 0.07 0.00 0.06 0.00 0.07 0.00 % of bread expenditures
*Households above the 185% poverty threshold are classified as higher income.
**Households at or below the 185% poverty threshold are classified as lower income.
Demand for whole-grain bread
91
Table 5 Parameter estimates from bread demand models, by income group (2003– 2007)
Higher-income sample Lower-income sample
Prior to release of Dietary
Guidelines
Post release of 2005
Dietary Guildelines
Prior to release of Dietary
Guidelines
Post release of 2005
Dietary Guildelines
Parameter Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error Estimate
Standard
Error
Intercepts
Refined grain 0.909 0.03 *** 0.950 0.05 *** 1.201 0.04 *** +++ 1.020 0.05 ***
Whole grain 20.032 0.03 20.059 0.04 20.159 0.03 *** 20.093 0.04 **
Multigrain 0.122 0.02 *** 0.109 0.03 *** 20.042 0.02 * +++ 0.073 0.03 ***
Price effects
Refined x refined grain 0.056 0.02 *** ++ 20.033 0.04 0.010 0.02 0.029 0.03
Refined x whole grain 20.013 0.01 ++ 0.064 0.03 ** 0.008 0.02 0.026 0.02
Refined x multigrain 20.043 0.01 *** 20.031 0.02 * 20.018 0.01 * ++ 20.055 0.01 ***
Whole x whole grain 0.020 0.01 +++ 20.080 0.03 *** 0.004 0.01 20.031 0.02
Whole x multigrain 20.007 0.01 0.016 0.01 20.013 0.01 * 0.005 0.01
Multi x multigrain 0.050 0.01 *** ++ 0.015 0.01 0.030 0.01 *** +0.050 0.01 ***
Expenditure effects
Refined grain 20.016 0.02 20.034 0.03 20.076 0.02 *** 20.065 0.02 ***
Whole grain 0.029 0.01 ** 0.010 0.02 0.056 0.01 *** 0.042 0.02 **
Multigrain 20.013 0.01 0.023 0.02 0.020 0.01 ** 0.022 0.01 *
Applied Economic Perspectives and Policy
92
Time trends
Refined grain 20.002 0.00 *** ++ 20.002 0.00 *** 20.003 0.00 *** +++ 20.002 0.00 ***
Whole grain 0.002 0.00 *** +++ 0.003 0.00 *** 0.002 0.00 *** 0.002 0.00 ***
Multigrain 0.000 0.00 ++ 0.000 0.00 *** 0.001 0.00 *** +++ 0.000 0.00 ***
Seasonal effects
Sine-Refined grain 20.002 0.00 0.000 0.00 20.010 0.00 *** +++ 0.001 0.00
Sine-Whole grain 0.002 0.00 0.000 0.00 0.008 0.00 *** +++ 20.001 0.00
Sine-Multigrain 0.000 0.00 20.001 0.00 0.002 0.00 20.001 0.00
Cosine-Refined grain 0.001 0.00 0.006 0.00 *** 20.001 0.00 0.000 0.00
Cosine-Whole grain 20.003 0.00 * 20.004 0.00 *** 0.001 0.00 0.000 0.00
Cosine-Multigrain 0.001 0.00 ++ 20.002 0.00 ** 0.000 0.00 20.001 0.00
Restrictions
Homogeneity, refined grain 177.087 37.06 *** 29.497 29.28 2126.845 52.26 ** 213.426 39.37
Homogeneity, whole grain 215.130 71.05 26.803 41.08 2169.997 82.50 ** 25.881 62.15
Homogeneity, multigrain 1.137 0.64 * 20.884 0.45 * 0.578 1.55 20.660 1.13
Adding up 3.313 1.89 * 20.585 0.90 20.263 4.41 20.577 2.68
Coefficient is significantly different than zero at *p ,0.10, * *p ,0.05, and ***p ,0.01.
Difference in coefficients between the two time periods is significant at +p,.10, ++p,.05, and +++p,.01.
Households above the 185% poverty threshold are classified as higher income.
Households at or below the 185% poverty threshold are classified as lower income.
Demand for whole-grain bread
93
Table 6 Testing for differences before and after release of the 2005 Dietary Guidelines, by income group (2003–2007)
Elasticity Estimates
(standard errors)
Higher-income sample Lower-income sample
Grain
Type
Refined
grain price
Whole
grain price
Multi
grain price Expenditure
Refined
grain price
Whole
grain price
Multi
grain price Expenditure
Grain
Type
R
e
f
i
n
e
d
Before 2005
DG
20.91 *** 20.02 20.05 *** 0.98 *** Before 2005
DG
20.90 *** 0.01 20.02 * 0.91 *** R
e
f
i
n
e
d
(0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.01) (0.02)
+++ +++
After 2005
DG
21.01 *** 0.10 ** 20.04 0.95 *** After 2005
DG
20.89 *** 0.04 20.07 *** 0.92 ***
(0.05) (0.05) (0.03) (0.04) (0.03) (0.03) (0.02) (0.03)
W
h
o
l
e
Before 2005
DG
20.30 ** 20.85 *** 20.09 1.24 *** Before 2005
DG
20.51 *** 20.95 *** 20.16 ** 1.62 *** W
h
o
l
e
(0.12)(
0.12)(0.07)(0.11)(0.14)(0.15)(0.08)(0.15)
+++ +++ + ++ + +
After 2005
DG
0.27 ** 21.39 *** 0.07 1.05 *** After 2005
DG
20.07 21.21 *** 0.02 1.27 ***
(0.12)(0.14)(0.06)(0.12)(0.12)(0.15)(0.06)(0.13)
M
u
l
t
i
Before 2005
DG
20.34 *** 20.07 20.44 *** 0.86 *** Before 2005
DG
20.57 *** 20.19 * 20.54 *** 1.31 *** M
u
l
t
i
(0.12) (0.09) (0.11) (0.11) (0.13) (0.11) (0.11) (0.14)
++ + ++ +
After 2005
DG
20.52 *** 0.12 20.85 *** 1.25 *** After 2005
DG
21.09 *** 0.04 20.28 ** 1.33 ***
(0.19) (0.16) (0.14) (0.18) (0.17) (0.15) (0.12) (0.19)
Coefficient is significantly different than zero at *p ,0.10, * *p ,0.05, and ***p ,0.01.
Difference in coefficients between the two time periods is significant at +p,.10, ++p,.05, and +++p,.01.
Households above the 185 percent poverty threshold are classified as higher income.
Households at or below the 185 percent poverty threshold are classified as lower income.
Applied Economic Perspectives and Policy
94
among higher-income households as we did for the pooled sample. The
time trend variables differed between the two time periods, and the own-
price effect for whole grains became more significant. Among the lower-
income households, the most notable difference between the time periods
was the change in intercept values for refined-grain and multigrain breads
and the change in time trend variables for all three grain types.
Comparing pre- and post-elasticity estimates by income (table 6)
again shows that the findings among the pooled sample are mirrored
in the findings among the higher-income sample over this time period.
Whole-grain cross-price elasticity estimates became positive and more
significant, again indicating that consumers began to view whole-grain
breads as a substitute for refined-grain breads. Also, the own-price
elasticities became more pronounced. Before the release of the 2005
Dietary Guidelines, higher-income consumers’ demand for whole-grain
bread was inelastic. Afterwards, they became more responsive to
prices, indicating that any sort of price reduction would have a larger
impact on quantity consumed. The findings are similar among lower-
income households as well, but not as pronounced. While cross-price
elasticities are not significant after the release of the 2005 Dietary
Guidelines, whole-grain own-price sensitivity increased significantly
among this group as well.
To further illustrate these findings, we simulate post-2005 bread expen-
diture shares without the 2005 Dietary Guidelines by income group. These
simulations (figures 3and 4) reveal some differences masked in examining
the entire sample. For example, at the end of the simulated period, the
Figure 3 Simulated monthly budget shares for refined-grain, whole-grain and multigrain
breads, with and without the influence of the 2005 Dietary Guidelines, higher-income sample
Demand for whole-grain bread
95
whole-grain budget share was approximately 0.24 among the pooled
sample for the simulation with the 2005 Dietary Guidelines, but only 0.20
for the simulation without the Dietary Guidelines. Similarly, the presence of
the Dietary Guidelines accounted for a decline of approximately 0.02 in the
pooled sample’s refined-grain budget share. In comparison, the higher-
income sample shows larger impacts: whole-grains rose by 0.05 with the
Dietary Guidelines and refined-grains fell by 0.05. Thus, we find that the
residual effect of the 2005 Dietary Guidelines accounts for higher-income
consumers buying roughly 2 fewer loaves of refined-grain bread per year,
and 1.5 more loaves of whole-grain bread. And while the residual effect
for multigrain breads increased, the impact is negligible – less than one-
tenth of a loaf of bread per year. Looking at the percentage change in
quantity over this time and using the own-price elasticities from table 6,
the change in demand among higher-income consumers after the 2005 rec-
ommendations would have been equivalent to raising the price of refined-
grain breads by 3% and lowering the price of whole-grain breads by 14%
over this same time period.
Behavior of the lower-income sample was qualitatively and quantita-
tively different. For the lower-income sample, it was not obvious that
there was any difference in budget shares for whole-grains that could be
linked to the Dietary Guidelines, but rather a response to lower relative
prices during this timeframe. We find that the residual impact of the
Dietary Guidelines on whole-grain purchased quantities and expenditures
shares over this time was zero.
Figure 4 Simulated monthly budget shares for refined-grain, whole-grain and multigrain
breads, with and without the influence of the 2005 Dietary Guidelines, lower-income sample
Applied Economic Perspectives and Policy
96
The findings for refined-grain and multigrain bread however, are
harder to explain. The simulated results suggest that refined-grain budget
shares would have been lower without the 2005 Dietary Guidelines.
Looking at percentage changes in quantities over this timeframe, this
would have been comparable to a 3% reduction in the price of refined-
grain bread and no change in whole-grain bread. The interesting finding
among refined-grain budget shares may indicate some level of confusion
among lower-income shoppers. Properly identifying whole-grain breads is
most frequently cited as a barrier to increasing whole-grain consumption
(Marquart et al. 2006;Seal, Jones and Whitney 2006), and these results
indicate this problem may be more of a barrier among lower-income con-
sumers. And while multigrain bread budget shares decreased, this was
more than offset by an increase in refined-grain purchases.
Discussion
The prevalence of poor diet quality among Americans, combined with
rising rates of obesity, suggests that there could be large public health ben-
efits derived from educating consumers on how to limit caloric intake
while meeting nutritional needs. This has been a recurring theme in all
Dietary Guidelines since 1980 and is again echoed in the 2010 Dietary
Guidelines. The longstanding discrepancy between recommended and
actual dietary patterns could make one pessimistic about realizing a
marked improvement in overall diet quality from another release of the
Dietary Guidelines. However, the results of this study suggest that, for
whole grains, the 2005 Dietary Guidelines were able to nudge dietary pat-
terns in the direction the public health community wanted.
We find that after the 2005 Dietary Guidelines recommended half of all
grains consumed be whole grains, consumers’ demand for whole-grain
breads significantly changed; they increased their whole-grain bread pur-
chases and decreased their refined-grain purchases. Examined in isolation,
however, descriptive statistics on average quantities purchased before and
after the 2005 Dietary Guidelines suggest too large an impact. Bread prices
were rising over this period, but were converging – relative prices of
whole-grain breads were falling (Mancino, Kuchler and Leibtag 2008).
Falling relative prices of whole-grain bread accounted for much of the
movement towards whole-grain bread. Nevertheless, even after account-
ing for relative prices and other features of the demand for bread, there
was something left over that can reasonably be attributed to the publica-
tion of Dietary Guidelines. For the higher-income consumers, the increased
budget share for whole-grain bread translates into approximately two
loaves per year, switching from refined-grain to one-and-a-half loaves of
whole-grain.
One reason the 2005 whole-grain recommendation may have been able
to prompt consumers to purchase more whole-grain breads is that spe-
cific advice on healthy substitutes may be easier for consumers to act on
compared to more general guidance, such as limiting intake of certain
nutrients. A related reason may be that the required changes in shop-
ping and consumption patterns necessary to meet these recommenda-
tions was not terribly difficult – buying a loaf of whole-grain bread is
very much like buying a loaf of refined-grain bread. Additionally,
Demand for whole-grain bread
97
analysis of whole-grain new product introductions before and after the
2005 Dietary Guidelines (Mancino, Kuchler and Leibtag 2008) indicates
that this advice may have provided manufacturers with incentives to
reformulate their products in anticipation of increased consumer
demand for products with healthier attributes. While prices may theoret-
ically be exogenous for demand analysis, they may have been directly
affected by the pending release of the 2005 Dietary Guidelines. When
manufacturers ramped up whole-grain production from a niche market,
they might have enjoyed some scale economies. Combined, these find-
ings suggest that dietary recommendations can have a non-zero impact
when the message to consumers is clear, the change required to follow
recommendations is small, and supply-side factors are aligned to facili-
tate consumers’ response.
Unfortunately, our findings also indicate that benefits from dietary
guidance may not be equally reaching all consumers. Among lower-
income consumers, there was a slight uptick in whole-grain bread pur-
chases following the 2005 Dietary Guidelines, but our results indicate this
was only in response to declining whole-grain prices rather than any
long-run change in demand. Of course, it may be that the increased
demand for whole-grain bread among higher-income consumers brought
about some economies of scale for whole-grain bread, thus lowering its
market price for all consumers. This could then be viewed as an indirect,
spillover benefit of the dietary guidelines.
The lack of significant change in whole-grain demand parameters and
elasticity estimates among lower-income consumers could be explained by
health concerns being secondary to prices, taste and convenience. It may
be that more confusion exists over what constitutes a whole-grain food.
The significant change in some multigrain demand parameters after the
2005 Dietary Guidelines among lower-income shoppers may be an indica-
tion of this. Some lower-income households may also reside in areas with
limited access to affordable and nutritious foods, which could also reduce
the likelihood of choosing healthier substitutes.
Changing elasticities provide a compact way of describing the change
in consumer preferences that accompanied the 2005 Dietary Guidelines.
Moreover, elasticities provide a way of characterizing whether consumers
are moving toward goals the public health community set out in the
Dietary Guidelines. With regard to the latter, the public health community
is clearly averse to having diets display price elasticity. There is nothing
in the Dietary Guidelines suggesting that recommendations are in any
way conditional. The recommendations do not recommend consuming
whole-grains when it is convenient, say, when relative prices make
whole-grain bread attractive. The recommendations do not suggest that
consumers should purchase whole-grains when they judge their income
to be high enough. That is, economists can interpret the Dietary
Guidelines as recommending zero elasticity. As such, findings here point
to some success in convincing consumers to move toward a healthier
diet, but also point to elasticities generally rising. In effect, our findings
suggest that if relative prices reversed their trend, we would see higher-
income consumers abandoning most of the gains they had made in diet
quality, and lower-income consumers reverting back to their previous,
pre-2005 purchasing patterns.
Applied Economic Perspectives and Policy
98
References
Anande, C., D. Pick, and M. Gehlar. 2005. Testing and Incorporating Seasonal
Structures into Demand Models for Fruit. Agricultural Economics 33(supplement 3):
527– 532.
Brown, J. D., and L. F. Schrader. 1990. Cholesterol Information and Shell Egg
Consumption. American Journal of Agricultural Economics 72(3): 548– 555.
Deaton, A., and J. Muellbauer. 1980. An Almost Ideal Demand System. The
American Economic Review 70(3): 312– 326.
Dharmasena, S., O. Capps Jr, and A. Clauson. 2011. Ascertaining the Impact of the
2000 USDA Dietary Guidelines for Americans on the Intake of Calories,
Caffeine, Calcium, and Vitamin C from At-Home Consumption of Nonalcoholic
Beverages. Journal of Agricultural and Applied Economics 43(1): 13 – 27.
Gifford, K. D. 2002. Dietary Fats, Eating Guides, and Public Policy: History,
Critique, and Recommendations. The American Journal of Medicine 113(9, supple-
ment 2): 89–106.
Guenther, P. M., W. Y. Juan, M. Lino, H. Hiza, T. Fungwe, and R. Lucas. 2008. Diet
Quality of Low-Income and Higher-income Americans in 2003-2004 as
Measured by the Healthy Eating Index-2005. Nutrition Insight 42. Center for
Nutrition Policy and Promotion, U.S. Department of Agriculture, Washington,
D.C.
Hausman, J. A., and G. K. Leonard. 2005. Journal of Competition Law and Economics.
1(2): 279–301.
Jitthavech, J. 2010. A Goodness-of-Fit Measure for a System-of-Equations Model.
Songklanakarin Journal of Science and Technology 32(5): 519– 525.
Krebs-Smith, S. M., P. M. Guenther, A. F. Subar, S. I. Kirkpatrick, and D. W. Dodd.
2010. Americans do not Meet Federal Dietary Recommendations. Journal of
Nutrition 140:1832–1838.
LaFrance, J. 1991. When Is Expenditure “Exogenous” in Separable Demand
Models? Western Journal of Agricultural Economics 16(1): 49–62.
Mancino, L, F. Kuchler, and E. Leibtag. 2008. Getting Consumers to Eat More
Whole-grains: The Role of Policy, Information, and Manufacturers. Food Policy
33(6): 489–496.
Marquart, L., H. W. Chan, M. Orsted, K. Schmitz, E. Arndt, and D. Jacobs. 2006.
Gradual Incorporation of Whole-Grain Flour into Grain-Based Products. Cereal
Foods World 52(3): 118– 121.
National Nutrition Monitoring and Related Research Act. Public Law 101-445, Title
III, 7 U.S.C. 5301 et seq, 1990. Accessed at: http://uscode.house.gov/download/
pls/07C84.txt.
Piggott, N., and T. L. Marsh. 2004. Does Food Safety Information impact U.S. Meat
Demand? American Journal of Agricultural Economics 86(1): 154– 174.
Rowe, Sylvia, et al. 2011. Food Science Challenge: Translating the Dietary
Guidelines for Americans to Bring About Real Behavior Change. Journal of Food
Science 76(1): R29–R37.
Seal, C. J., A. R. Jones, and A. D. Whitney. 2006. Whole-grains Uncovered. British
Nutrition Foundation Nutrition Bulletin 31 (2), 129– 137.
U.S. Department of Health and Human Services and U.S. Department of
Agriculture. 2005. Dietary Guidelines for Americans 2005. Accessed at:
http://www.health.gov/dietaryguidelines/DietaryGuidelinesa2005/document/
pdf/DIETARY GUIDELINESA2005.pdf.
Whole Grains Council. 2009. “Are We There Yet? Measuring Progress on Making
at Least Half Our Grains Whole.” Conference Highlights from Make Half of
Your Grains Whole Conference, April 2009. Accessed at: http://www.
wholegrainscouncil.org/files/3.AreWeThereYet.pdf.
Demand for whole-grain bread
99
Appendix
We use the basic AIDS specification derived by Deaton and Muellbauer
(1980) using 3 products – refined-grain, whole-grain and multigrain
breads, indexed i¼1, 2, 3. Budget shares (
v
) are defined as v
it
¼p
it
Q
it
/
Y
t
, where p
it
is the price of product iat time t, Q
it
is the quantity pur-
chased at time t, and Y
t
is total expenditures over all products in the
demand system at time t.
Under the AIDS specification, the budget share for each good iacross
t¼1, ...,Ttime periods is specified as:
v
it =
a
i+
3
j=1
g
ij log pjt +
b
i(log Yt−log Pt),(1)
where the terms a,b, and gdenote parameters to be estimated. Each
budget share is a linear function of all prices and total expenditures on
the category of goods. Expenditures are deflated by a price index P, where
Pis defined as:
log Pt=
a
0+
3
k=1
a
klog pkt +.5
3
k=1
3
j=1
g
kj log pkt log pjt.(2)
We modify the basic AIDS model in three ways. First, we account for the
geographical variation in American diets, calculating prices, expenditures,
and budget shares for 13 regions (r) of the United States. Second, we
allow for systematic variation in share equations across time, accounting
for long-run trends and seasonality. We incorporate regional fixed effects
and time patterns by generalizing the intercept terms in budget share
equations to be functions of time and place:
Virt =
a
i+
a
ir +
f
icos(2
p
t/
t
)+
f
isin(2
p
t/
t
)+
k
it,(3)
where the parameters
a
ir signal regional fixed effects, f
i
and
w
i
reflect any
seasonal patterns, and k
i
indicates any long-term trends. Thus, our
demand model can be expressed as:
v
irt =Virt +
3
j=1
g
ij log pjrt +
b
i(log Yrt −log Prt).(4)
The geographic fixed effects, time trends and seasonal cycles are all
included in the price index, as are the share equations:
log Prt =
a
0+
3
k=1
Vkrt log pkrt +.5
3
k=1
3
j=1
g
kj log pkrt log pjrt.(5)
To test for differences before and after publication of the dietary guide-
lines, we make a third adjustment to allow our demand estimates to
switch depending upon the time period. Defining ˆ
tas the date that the
2005 Dietary Guidelines were released, we adjust our price index as
Applied Economic Perspectives and Policy
100
follows:
log Prt =
a
0+
3
k=1
Vkrt log pkrt +.5
3
k=1
3
j=1
g
kj log pkrt log pjrt
⎡
⎣⎤
⎦,if t,ˆ
t
log ˆ
Prt =ˆ
a
0+
3
k=1
ˆ
Vkrt log pkrt +.5
3
k=1
3
j=1
ˆ
g
kj log pkrt log pjrt
⎡
⎣⎤
⎦,if t≥ˆ
t.
(6)
This then necessitates the following change in how we estimate our share
equations in (4):
v
irt =(1−DG)∗Virt +
3
j=1
g
ij log pjrt +
b
i(log Yrt −log Prt)
⎡
⎣⎤
⎦+
(DG)∗ˆ
Virt +
3
j=1
ˆ
g
ij log pjrt +ˆ
b
i(log Yrt −log ˆ
Prt)
⎡
⎣⎤
⎦,
(7)
where DG =1if t≥ˆ
t,DG =0 otherwise, and ˆ
Virt =ˆ
a
ir +ˆ
f
icos(2
p
t/
t
)+
ˆ
w
isin(2
p
t/
t
)+ ˆ
k
it.This is equivalent to running two completely different
models before and after the release of the Dietary Guidelines. The benefit of
this specification, however, is it allows us to directly test for significant dif-
ferences between the two time periods.
The economic assumptions of homogeneity and symmetry are imposed.
Because our intercept terms vary regionally and across time, we aug-
mented the adding up restrictions by imposing the following restrictions:
3
i=1
a
i=
3
i=1
ˆ
a
i=1;
3
i=1
a
ir =
3
i=1
f
i=
3
i=1
w
i=
3
i=1
k
i=
3
i=1
ˆ
a
ir =
3
i=1
ˆ
f
i=
3
i=1
ˆ
w
i=
3
i=1
ˆ
k
i=0
(8)
.
Demand for whole-grain bread
101