Choosing brands: fresh produce versus other products.

The inequalities in (6) provide basic hypotheses for our empirical analysis, but these hypotheses are not complete. If the quality effects dominate the prestige and design effects for all product categories, then the extra value of brands in electronics is the highest, and vice versa. However, the shares of the effects of these three dimensions in the overall brand values may significantly vary among the four product categories, which limits our ability to have complete hypotheses about the ranking of the perceived brand value among the products. Furthermore, fresh produce faces a greater uncertainty in the production process that makes it harder to maintain a standard of quality and design. A possible inconsistent quality of a brand will reduce the likelihood of repeated purchases, generate adverse images (word of mouth), and undermine the investment in brands (Heiman and Goldschmidt 2004). This process uncertainty will unavoidably challenge brands of fresh produce. Nevertheless, our analysis suggests that, relative to the base generic value, consumers have the lowest WTP for brands of fresh produce, followed by brands of packaged foods, and they have the highest WTP for brands of clothing and electronics.

Our analysis compares the WTP for brands across broad product categories, but each of these categories includes different products, and there is variation in WTP within a category. For example, refrigerators and toaster ovens are both electronics, but it is likely that brand premiums for refrigerators are higher because they generally last longer and represent a larger investment. Similarly, there are likely to be different brand premiums for different types of food products and different food categories (organic versus nonorganic). Because organic foods are credence goods, where quality depends on unobserved farmer behavior, and hence greater uncertainty, having brands may be more valuable. Some value of brands may be reduced with the introduction of organic certification programs, but in this case, the brand of the certifying agency acts as a brand in and of itself. Another issue for food items is geographic identification. In many cases, products differ by regions, and as a result there is growing usage of regional designated products (Champagne, Burgundy, and Napa Valley wines; Washington apples; etc.). However, within each region, there may still exist WTP for a brand (Sunkist versus a more generic Florida orange brand). In some cases, regional indicators may substitute for a brand. In our analysis, we elicit only consumer responses for brands in packaged food and fresh produce as broad food categories along with clothing and electronics. Future research should investigate WTP for brands within subcategories such as organics and the extent to which regional identification substitutes for brands.

Individuals with different socioeconomic backgrounds may have different attitudes toward brands. Retailers and brand managers may utilize sociodemographic information to create market segmentation (Gupta and Chintagunta 1994), choose retail locations (Ghosh and McLafferty 1987), forecast brand choices (Allenby and Rossi 1991; Chiang 1991; Kalyanam and Putler 1997; and Ainslie and Rossi 1998), etc.

Individuals' WTP for brands is likely to be related to informational gain that brands provide. As a source of information, brands will be substitutes to time and skills in the product quality assessment. They will provide extra informational value to (a) those lacking education, skills, experience, or product knowledge to detect quality and (b) those whose time is constrained or too valuable to intensively engage in pre-purchase product quality assessments. Zeithaml (1988) shows that women whose time constraint is more binding relied on brands when purchasing orange juice. Individuals with higher income have higher opportunity cost of search time, which leads them to have higher WTP for brands. Individuals who have sufficient status may not gain as much prestige from a brand as individuals without that status. Brands may be more valuable as status symbols to less educated individuals, as Fussell (1983) suggests for clothing.

Data

We conducted a survey on consumers' perception toward brands at one HEB store, two Albertson stores, one Wal-Mart Supercenter, and one local grocery store in College Station and Bryan, Texas, in fall 2006. We did not have permission to conduct a survey in other stores, including one Albertson store, three Kroger stores, and one HEB store in the study area. A total of 302 usable observations were collected among 305 in-person surveys conducted. As shown in table 1, the usable sample well represents the population of the study area based on the demographic information, including gender, age, race, household size, education, and income. Of the usable sample, 55 % are females compared with 48% of the population in College Station and 53% in Bryan. The average household size in the usable sample is 2.43 versus 2.25 for College Station and 2.49 for Bryan. The race distribution of the usable sample is similar to the population. Obviously, the sample is affected by who collected the data and where the collection was done. Because the data were collected by a Chinese-Canadian student in a college town, the percentage of Asian respondents is higher than otherwise. Similarly, the percentage with college education is higher than the U.S. Census Bureau data for the overall population because the Census Bureau does not account for students.

Respondents were asked to report their brand preference of four products (electronics, clothing, packaged food, and fresh fruits and vegetables), including the brand preference ranking from zero (do not buy brands at all) to 10 (always buy brands), and the choice of the WTP range as well as the best point estimate of WTP.

Individuals with a brand preference ranking of 8, 9, or 10 are considered to have a strong brand preference, and a WTP greater than zero is called positive WTP. Table 2 shows that (a) more than half exhibit strong brand preference for electronics but much fewer for food (less than 30%); (b) almost all respondents are willing to pay more for brands in the electronics product category (97%) but fewer for fresh produce (78%); (c) among people with strong brand preference in the associated product category, the average WTP of fresh produce and clothing is higher than that for electronics and packaged food; and (d) among people who are willing to pay more for brands, the average additional WTP for durable goods (electronics and clothing) is about 32%, while the average for food products is 26%. Furthermore, table 2 also shows some patterns of brand preference and WTP for brands in four product categories. For example, about 70% of respondents have a positive WTP for brand products and 7% of respondents state a strong brand preference toward all of the four categories. The WTP for brands of fresh produce is almost as high as in electronics and clothing among those who have a positive WTP for brands in all four of these product categories.

We asked respondents to choose the closest range of WTP for brands among six intervals (0-20%, 20-40%, 40-60%, 60-80%, 80-100%, and at least 100%). Based on their choices of WTP ranges, we estimate the empirical probability density function of the underlying WTP for brand products relative to generic ones using the maximum entropy density method. Adopting the methodology of Wu and Perloff (2007), we use a flexible functional form that nests many commonly used distributions,

(7) f([W.sup.*.sub.ik]) = exp (-[M.summation over (m=0)] [[lambda].sub.m] [([W.sup.*.sub.ik]).sup.m])

where [W.sup.*.sub.ik] is the underlying WTP of consumer i for brands of product category k that is unobservable to researchers, [[lambda].sub.m]'s are parameters to be estimated, and m = 0, 1, ..., M are polynomial orders. As figure 1 shows, the estimated density function based on the interval frequencies matches well with the histogram based on the perceived WTP ranges for each product category. Furthermore, the Pearson's [chi square] test for each product category also suggests a good fit. Based on the estimated density of WTP, we are able to conduct statistical tests on the mean difference of WTP between products (see table 3). The results indicate the highest WTP for brands of electronics, a substantial WTP for brands of clothing, followed by packaged food, and the lowest in fresh produce. However, the mean difference of WTP for food brands is not statistically significant.

[FIGURE 1 OMITTED]

We have not found empirical evidence of WTP for brands in the literature. To assess our results, we compare prices at one of the stores we surveyed, a Wal-Mart Supercenter, in Bryan, Texas. A typical example for electronics is the digital camera where the price of a Cannon Powershot with 6MP is over 50% more than similar products of other brands. We picked jeans as a typical case for clothing--the price of men's Levi-Strauss signature regular-fit is about 30% more than other brands. Regarding food items, Del Monte's fruit cans are 10% more than Dole's and 25% more than Great Value's, Wal-Mart's private label. Minute Maid's 1-gallon orange juice (pulp free) is 25% more than Tropicana's and 42% more than Great Value's. These antecedent evidences suggest that the stated WTP in the survey are comparable with actual brand premiums.