Académique Documents
Professionnel Documents
Culture Documents
Laurian J. Unnevehr
Fatoumata C. Gouzou
The retail demand for characteristics of honey is estimat- has experienced a sharp reduction in federal sup-
ed using retail scanner data. The results show that con- port.1 With the loss of federal programs, honey
sumers pay substantial premiums for honey based on producers must rely on payment for pollination ser-
form, container, brand, and floral source. The highest vices, which are likely to be small,2 or on the con-
price premiums are associated with unique monofloral sumer market. For the latter market, honey pro-
sources, followed by other unusual, but less pure, floral ducers may want to consider product differen-
sources. The results demonstrate that consumers will pay tiation to increase the total value of sales. To do
for unique characteristics of honeys associated with par- this, producers will need information about current
ticular floral sources. Thus, honey producers may be able consumer demand for product characteristics. New
to further promote honey products or develop new prod- products are frequently combinations of character-
ucts on the basis of floral source differentiation. 1998 istics in existing products; thus, an understanding
John Wiley & Sons, Inc. of current demand for characteristics may give in-
sights into the potential demand for new combina-
tions.3
Recent changes in U.S. agricultural policy will This article demonstrates how retail scanner data
bring greater market orientation in many commodi- can be used to estimate demand for product charac-
ty markets. Honey is one agricultural product that teristics. Scanner data can provide a new source of
information about market demand for characteris-
Requests for reprints should be sent to L. Unnevehr, Dept. of Agricul- tics. In the case of honey, an important question is
tural Economics, 305 Mumford Hall, MC-710, University of Illinois, whether consumers are already attuned to differ-
1301 W. Gregory Dr., Urbana, IL 61801. ences among honeys and willing to pay premiums
L.J. Unnevehr is an Associate Professor in the Department of Agricultural and Consumer Economics, University
of Illinois at UrbanaChampaign.
F.C. Gouzou is a graduate student in the Department of Agricultural and Consumer Economics, University
of Illinois at UrbanaChampaign.
Support for this research was provided by the Value-Added Program of the Illinois State Experiment Station. The authors thank Profs. May
Berenbaum and Gene Robinson, Department of Entomology, UIUC, for their collaboration and insights, and Mary Humann of the National
Honey Board for sharing information.
49
Unnevehr and Gouzou
for specialty characteristics, such as unique floral of the product multiplied by the marginal implicit
sources. The retail scanner data used here provide price of the characteristic.
new and interesting insights into consumer demand Because the yield of most product characteristics
for specialty characteristics of honey that may be is constant for each unit of product, dXoj /dqi 5 Xij
useful to the honey industry for informing product 5 constant is assumed. Furthermore, the marginal
development and promotion efforts. implicit price is also assumed to be constant and is
represented by Pij. Therefore, Eq. (1) for a partic-
ular product, F, becomes
Consumer Demand for Quality Characteristics
M
50
Honey Characteristics
Note: The total sample contains 1295 items. Number of items and percent of sample refer to obser-
vations from which the dummy variable 5 1.
a
The flavor dummy variables contain the following flavors identified in the data:
DFLAVB1: Clover includes all flavor names with clover as one of the words: alfalfa and clover,
Canadian clover, clover, clover blossom, clover wildflower, Dakota clover, Dakota sweet clover,
gold clover, Montana clover, mountain clover, New England clover, Pennsylvania clover, Texas
clover, and white clover.
DFLAVB2: Orange blossom includes all flavor names with orange as one of the words: CA orange,
orange, orange blossom, and Santa Barbara orange.
DFLAVB3: Other monofloral includes all flavors that appear to be pure monoflorals but are not
clover or orange: acacia, alfalfa, apple, avocado, basswood, blackberry, blueberry, blueberry
blossom, buckwheat, CA safflower, cranberry, eucalyptus, fireweed, gallberry, HA passion fruit,
huckleberry, lemon, mesquite, MI thistle, mint, raspberry, safflower, sage, strawberry, sunflower,
thistle, tupelo, tupelo blossom, and vetch blossom.
DFLAVB4: Other floral includes any flavor not in the above categories: apple and cranberry, as-
sorted, cactus, desert blossom, desert valley flower, fireweed wildflower, forest haujillo, multi-
flower, ozark wildflower, Texas wildflower, tupelo wildflower, western wildflower, wildberry, wild
huckleberry, and wildflower.
tainer type, flavor, and form, as well as total unit The total retail market for honey reflected in these
sales, dollar sales, and average unit price (e.g., data was worth $127.9 million in 1994 to 1995.
$1.61 for the item above). There are a total of 1295 Honey is marketed under a large number of brands.
observations or individual honey items sold in these About 28% of total dollar volume is sold with gener-
data.b Table I reports the distribution of product ic or store-controlled labels. The remaining 72% of
characteristics in the data by percent of items and dollar volume is marketed by 253 name brands.
percent of dollar volume. However, there are a large number of brands with
b
very small market shares. Forty-eight brands ac-
Retail scanner data are costly to obtain, because they are collected
and marketed by private firms. In purchasing our data, we chose to fo-
count for 64% of total dollar volume; the remaining
cus on product characteristics rather than on seasonal or regional vari- 205 brands account for only 8% of dollar volume.
ation in demand. While such variation may exist, we wanted to estimate This indicates that the market for honey is charac-
average demand for characteristics. Therefore, we requested data that terized by a high degree of brand competition.
are summarized on an annual national basis but provide the most de- Honey products are often differentiated by fla-
tail possible regarding product characteristics. vor, which generally indicates floral source. The
51
Unnevehr and Gouzou
market is dominated by regular or unspecified very unusual form (e.g., powder) or container (e.g.,
flavor (45.4% of dollar volume) and clover flavor wine glass) or to which other foods have been added
(47%). The market for specialty flavors is a small as flavoring (e.g., chocolate) were deleted because
proportion of the total dollar volume (7.1%) and is these are not directly comparable to the character-
worth $9.7 million in retail sales. This market is istics identified for the majority of items, and the
highly differentiated, and 78 individual flavors are number of unusual observations was too small in
sold (Table I). each category to provide a valid statistical measure.
For the analysis of price premiums below, honey The dependent variable in the regressions is the
flavors are organized into five groups: regular, unit retail price in dollars. The independent vari-
clover, orange blossom, unique monofloral sources, ables include dummy variables that summarize in-
and other floral sources. Unique monoflorals are formation about form, flavor, container, and
honeys from unusual monofloral sources (e.g., the whether sold under a store or generic label, and
acacia tree) other than the relatively common continuous variables for size in ounces and size
clover or orange blossom sources. The category of squared (to capture any price discounts associated
other floral sources includes flavors whose names with large containers). The coefficients describe
indicate less purity than those in the unique mono- how the price differs for a particular attribute,
florals (e.g., western wildflower). Sorting of the fla- holding all other attributes constant. The intercept
vor data was carried out in consultation with ento- is the average price for honey that is regular flavor,
mologists familiar with potential floral sources. liquid form, glass container, and not marketed un-
Other market characteristics noted in the data are der a store controlled or generic label.
the container type and form. About 40% of the It is possible that not all of the dependent variable
honey items are sold in plastic containers, including observations are truly independent, because items
honey bears and squeeze bottles, and the remainder sold under the same brand are likely to have prices
in glass (Table I). About 2% of the items are sold that are not independently determined. This may
with the comb in; 7.3% are creamed or spun honeys account for some of the heteroskedasticity in the
in which the sugar has been crystallized. residuals, which is discussed below. Because there
Honey color is classified by the USDA from white are 253 name brands and numerous store brands
to dark amber.13 (See Hoff and Willett14 for a de- that are not differentiated in the data, we still have
scription of wholesale honey grades.) In the whole- a considerable number of truly independent obser-
sale market, darker honeys are discounted vations. Most of our characteristics, even unique
(USDA)13; thus, the designation of light color monoflorals, were observed across many different
should carry a price premium. Although the whole- brands.
sale price differences vary widely across years, Table II reports estimated coefficients for honey
darker honeys consistently have lower prices than characteristics in dollars per unit. The characteris-
lighter honeys. Darker honeys are often sold for use tics included in the regression explain 77% of the
in processed products, and many honeys sold at re- variation in retail prices. Because multicollinearity
tail are blended to achieve a uniform amber color. is often a problem in hedonic price estimation, the
Unfortunately, color is mentioned very infrequently Belsley et al.15 diagnostics were generated. The
on the label data reported to A.C. Nielsen12, so it highest condition index was 15.7; thus, the esti-
was difficult to include this important characteristic mates are not degraded by multicollinearity.
in our analysis of retail prices. Whites test for heteroskedasticity was significant16
because the variance of the residuals differed by
flavor categories. The variance of the residuals was
Price Premiums for Honey Characteristics higher for observations in the unique monofloral
and other floral flavor categories, in comparison
Price premiums are estimated based on Eq. (2) with regular, clover, and orange blossom observa-
above. The regression estimates use 1192 observa- tions. Unobserved characteristics cause greater
tions from the original data set. Items that have a price variation in these categories. Because the
52
Honey Characteristics
Note: See Table I for definitions of variables. The standard errors have been corrected using the het-
eroskedasticity consistent covariance matrix. The dependent variable is average price per item in dollars.
The intercept reflects the average price for honey that is liquid, regular flavor, in a glass container, and
sold under a name brand. The regression was estimated using 1192 observations from the original sample
of 1295. Observations with very unusual containers, form, or flavors added were dropped because they
are not comparable to other items.
Adjusted R2; 0.77. Whites test; 105.19. Highest condition index; 15.69.
F test for Categorical Variables:
Container dummies 5 0, F (3,1180) 5 9.05.
Form dummies 5 0, F (2,1180) 5 49.15.
Flavor dummies 5 0, F (4,1180) 5 19.96
All F ratios are significant at the 1% level.
standard errors from ordinary least squares are bi- zero (Table II). Thus, prices do vary according to
ased and inconsistent in this case, these were cor- form, container, and flavor. Consumers pay lower
rected using Whites heteroskedasticity consistent prices for plastic containers in comparison to glass,
covariance matrix.16 Significance levels in Table II and the price discounts are similar for the three dif-
are based on the corrected standard errors. ferent kinds of plastic containers.
The regression estimates show that several charac- In addition to brand marketing and packaging,
teristics of honey are significant determinants of the characteristics intrinsic to honey are also signif-
price. The regression coefficient estimates show icant determinants of price. There are positive
that consumers pay $0.07 for each ounce. The size price premiums for honey sold in different forms
squared coefficient is not significant, indicating that (i.e., with the comb in or creamed). With respect to
larger containers do not carry any significant dis- flavor, the F test indicates that consumers pay sig-
count in the honey market. In comparison to name nificant price premiums for monofloral flavors over
brands, consumers pay lower prices for generic or regular flavors. Clover and orange blossom carry
store labels, as would be expected. similar price premiums of $0.08 to $0.10/item high-
The F tests for the container, form, and flavor er than regular flavor items. The price premium for
variables reject the hypothesis that the dummy unique monoflorals is $0.55/item. It is interesting to
variables in each of these categories are equal to note that this premium is higher than the $0.39/item
53
Unnevehr and Gouzou
associated with other florals that are not from one ducers of common floral sources that are not cur-
source. Thus, the distinction between pure and rently marketed separately (e.g., soy honey) might
mixed floral sources in labeling is reflected in mar- benefit from new knowledge about honey character-
ket prices. istics that could be promoted to consumers. A po-
tential limitation to further honey product differen-
tiation could be the higher cost of producing and
Conclusions guaranteeing floral source purity. Thus, another
important area for research would be reduction of
The results show that honey products are highly these production and marketing costs.
differentiated and that consumers are willing to pay The demand analysis reported above shows that
substantial premiums for particular honey charac- scanner data can provide a new source of informa-
teristics, especially unique floral sources. These flo- tion about product characteristic demand at the re-
ral sources carry a 65% higher price premium over tail level. The characteristics that are identified as
the average item price for regular honey. Thus, the currently commanding substantial price premiums
potential exists to market different floral sources of provide insights into directions for future promo-
honey to consumers based on their distinctive char- tion efforts. In addition, new product development
acteristics. The small market share of these floral can be focused on characteristics that have the po-
sources may indicate the need to promote their spe- tential to command higher premiums in the market.
cial flavor characteristics to a larger group of con- Such ex ante economic analysis can inform multi-
sumers. disciplinary efforts for product development and
The results also provide insights into where pro- promotion.
duction research might have market payoff. Pro-
References
1. F. Hoff, Honey: Background for the 1995 Farm Legisla- sumers Pay for Less Fat? Journal of Agricultural and Re-
tion, Agricultural Economic Report 708, U.S. Depart- source Economics, 18, 288 (1993).
ment of Agriculture, Economic Research Service, Washing- 11. R.W. Cotterill, Scanner Data: New Opportunities for De-
ton, DC, 1995. mand and Competitive Strategy Analysis, Agricultural
2. M.K. Muth and W.N. Thurman, Why Support the Price and Resource Economics Review, 23, 125 (1994).
of Honey? Choices, Second Quarter, 19 (1995). 12. A.C. Nelson Co., Special Data Break-Out for Honey Sold
3. B. Senauer, E. Asp, and J. Kinsey, Food Trends and The at Retail, obtained through Charlotte Yanik, Account As-
Changing Consumer, Eagan Press, St. Paul, MN 1991. sociate, Schaumburg, IL March 1996.
4. Z. Griliches, Ed., Price Indexes and Quality Change, Har- 13. USDA, United States Standards for Grades of Extracted
vard University Press, Cambridge, MA, 1971. Honey, Effective May 23, 1985, Agricultural Marketing
5. K. Lancaster, A New Approach to Consumer Theory, Service, Washington, DC, 1985.
Journal of Political Economy, 74, 132 (1966). 14. F. Hoff and L. Schertz Willett, The U.S. Beekeeping In-
6. S. Rosen, Hedonic Prices and Implicit Markets: Product dustry, Agricultural Economic Report 680, U.S. Depart-
Differentiation in Pure Competition, Journal of Political ment of Agriculture, Economic Research Service, Washing-
Economy, 82, 34 (1974). ton, DC, 1994.
7. R. Lucas, Hedonic Price Functions, Economic Inquiry, 15. D.A. Belsley, E. Kuh, and R.E. Welsch, Regression Diag-
13, 157 (1975). nostic: Identifying Influential Data and Sources of
8. G.W. Ladd and V. Suvannunt, A Model of Consumer Collinearity, Wiley, New York, 1980.
Goods Characteristics, American Journal of Agricultural 16. H. White, A Heteroskedasticity-Consistent Covariance
Economics, 58, 504 (1976). Matrix Estimator and a Direct Test of Heteroskedasticity,
9. J.A. Espinosa and B.K. Goodwin, Hedonic Price Estima- Econometrica, 48, 817 (1980).
tion for Kansas Wheat Characteristics, Western Journal 17. W.M. Hanneman, Quality and Demand Analysis, in New
of Agricultural Economics, 16, 72 (1991). Directions in Econometric Modeling and Forecasting in U.S.
10. L.J. Unnevehr and S. Bard, Beef Quality: Will Con- Agriculture, G.C. Rausser, Ed., Elsevier, New York, 1982.
54