Mobile Edition
Print
Get the Mag
Weekly UpdatesConsumer willingness-to-pay (WTP) for certain food quality attributes is an important indicator of consumer response to food labels and a determinant of the anticipated change in demand under a food labeling program. Estimated WTP is used as an input or proxy for demand change in welfare analyses of food policy and to provide useful information for food labeling programs (e.g., Lusk and Anderson 2004; Lubben 2005). Therefore, understanding and accurately determining consumer WTP for food quality attributes is important for policy makers as well as food producers and processors.
Economists and market researchers have used contingent valuation (CV), choice experiments (CEs), and experimental auctions (EAs) or combinations of the three methods extensively to elicit consumer preferences for food labels on attributes, such as tenderness of beef, country of origin of meat and vegetables, organic foods, foods containing genetically modified organisms (GMO), and numerous other product attributes (Fox et al. 1994; Fox 1995; Huffman et al. 1996; Hossain et al. 2003; Loureiro and Umberger 2003, 2005; Hu et al. 2004; Krystallis and Chryssohoidis 2005). One concern in these studies is that different food quality attributes are assumed to be independent of attributes that are not provided to respondents in the survey or experiment. For example, to estimate consumer WTP for country of origin labeling (COOL) by using CV or EA, respondents are asked to choose between food with or without COOL or bid for products with or without a label; no other information on product attributes such as safety, nutrition level, quality, etc., is provided. In CE respondents make choices from alternatives that differ in numerous quality attributes but lack information on other attributes typically available in real-world shopping.
Hallman et al. (2003) show that research questions reinforced consumer attention to the information of interest. However, in consumer WTP studies, only limited information on food quality attributes can be provided to respondents. A clear view of how additional labeled food quality attributes affect consumer WTP will help economists to better understand welfare analysis of food labeling policy by using WTP from published studies as well as enhance producers' abilities to make more appropriate decisions on labeling programs on the basis of WTP estimates.
The "missing attributes" issues have been analyzed in the conjoint analysis literature. However, most studies have investigated the effects of missing attributes for cases in which partial profiles existed--an attribute was not eliminated from all the options; only levels of the attribute were missing for some profiles. As a result, consumers could infer the missing attribute levels by determining the systematic pattern of the profile or by using cues (e.g., price) provided in the survey (Huber and McCann 1982; Johnson and Levin 1985; Broniarczyk and Alba 1994). Kardes, Posavac, and Cronley (2004) conducted a comprehensive review of the process and context of consumer inference under incomplete information. In more recent work Islam, Louviere, and Burke (2007) showed that a missing attribute in CEs affected the systematic component and increased the variance of the random component of the consumer utility function. Hensher, Rose, and Greene (2005a) determined that the CE respondents' ignorance of certain attributes resulted in significantly lower WTP estimates for travel time saving. However, including or excluding an attribute in choice sets was not systematically related to heterogeneity across respondents. Hensher (2006b) studied the effect of complexity of CEs on WTP for travel time saving. His results indicated that the mean-weighted average WTP for time saving was not significantly influenced by the design dimensionality if all five of the design dimensions (i.e., number of choice sets, attributes, alternatives, attribute levels, and range of attribute levels) were controlled. However, he showed that the weighted-average WTP for time saving had a positive relationship with the number of attributes in a CE if other dimensions were not controlled.
In this study, we estimate marginal effects of additional label information on consumer WTP for food quality attributes. We take a different approach from previous studies. A utility function and WTP are estimated for each CE in the surveys as if those surveys were conducted independently by the researcher. This avoids the monotonic effects of the number of attributes on WTP when using interactions between an attribute and the numbers of attributes in pooled data. The overlapped CEs in the surveys enable us to conduct both within-subject and between-subject comparisons of the influence of additional attributes on the means and variances of estimated WTP. In addition we design our survey using two sets of attributes: one set includes a cue attribute, and the other does not. The effects of additional attributes on WTP have not been tested in those two distinct information contexts. In particular we develop a conceptual argument for why and, using survey data, illustrate how marginal WTP for food product attributes in CE, CV, or EA studies depends on the particular food attributes presented to respondents on the label. This information is important because it demonstrates that WTP estimates in CE, CV, and EA studies depend on the set of product attribute information provided to respondents.
Model Development: Changes in WTP with Additional Attributes
The theoretical model in equation (1) shows that as the number of attributes changes in a consumer's utility function, consumer WTP for a specific attribute may also change. Assuming a linear random utility function, consumer utility (U) could be defined by
(1) [U.sub.ij] = [[alpha].sub.i] x [p.sub.ij] + [T.summation over (k = 1)] [[beta].sub.ik] x [x.sub.ijk] + [[epsilon].sub.ij]
where [p.sub.ij] is the price of alternative j for person i, [x.sub.ijk] is the kth attribute of alternative j for person i, [[alpha].sub.i] is marginal utility of price for person i, [[beta].sub.ik] is marginal utility of the kth attribute, and [[epsilon].sub.ij] is a stochastic disturbance of alternative j for person i. T is the number of attributes of alternative j.
Consumer i's WTP for the kth attribute is the amount of money that she/he would be willing to pay to stay at her/his previous utility level when the kth attribute changes. Now, assume that the kth attribute in alternative j improves from level zero (without attribute k) to level one (with attribute k); the WTP of consumer i to accept this change is the price premium she/he would pay for this change such that the following equality holds
(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where the superscript 0 of attribute x indicates that the attribute is unavailable in the product, and superscript 1 indicates that the attribute is available in the product.
Solving (2) for WTP, we get
(3) [WTP.sub.k] = [[beta].sub.ik]/[[alpha].sub.i] ([x.sup.1.sub.ijk] - [x.sup.0.sub.ijk]).
As a result, for a linear utility function consumer WTP for the kth attribute is the negative ratio of the parameter of the kth attribute to the parameter of price: [WTP.sup.k] = [[beta].sub.ik]/[[alpha].sub.i].
To test the effect of additional attributes on consumer WTP for attribute k in equation (1), we assume alternative j has M-T additional attributes (M > T). In this case consumer i's random utility function can be expressed as
(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].
Equation (4) implies that with more attributes added to the consumer utility function, the marginal utility of price and attributes changes from [[alpha].sub.i] to [[alpha].sup.*.sub.i] and from [[beta].sub.ik] to [[beta].sup.*.sub.ik], respectively. With these changes consumer WTP for attribute k will change from [WTP.sup.k] = - [[beta].sub.ik]/[[alpha].sub.i] to [WTP.sup.k*] = -[[beta].sup.*.sub.ik]/[[alpha].sup.*.sub.i].
We hypothesize that the difference between [WTP.sup.k] and [WTP.sup.k*] can be anticipated depending on the types of product attributes provided in the initial description and the descriptors added to the product label. For example, when a cue attribute is present on a product label, as more information on other product attributes is provided, the cue attribute may lose its power of signaling product quality or being a proxy for other attributes. In this case, we expect that WTP for the cue attribute would decrease because of the substitute effect of additional attributes. In a case in which a product is not described using a cue attribute, adding more product descriptors may also change WTP. If consumers maximize their utilities subject to their budget constraints by choosing the optimal attribute bundles (Lancaster 1972), the utility function in equation (1) implies that consumers allocate their budgets to T product attributes, whereas the utility function in equation (4) indicates that consumers would allocate their budgets to T + M product attributes. The change in the consumer utility function from (1) to (4) simply means that consumers reallocate limited budgets to a larger attribute set. Other potential reasons underlying the change in WTP include potential diminishing marginal utility of additional product attributes and consumers' diminishing ability to discern value differences among a large number of attributes as well as substitution or complementary effects between attributes (Lusk 2003).
If consumer WTP for one attribute [x.sub.ijk] is not affected by the other product attributes, consumer WTP for [x.sub.ijk] will not change, such that [absolute value of ([WTP.sup.k] = - [[beta].sub.ik]/[[alpha].sub.i]) - ([WTP.sup.k*] = - [[beta].sup.*.sub.ik]/[[alpha].sup.*.sub.i])] < [xi], where [xi] is a small positive number. If consumer WTP for attribute [x.sub.ijk] is not independent of other product attributes, [absolute value of [WTP.sup.k] - [WTP.sup.k*] > [xi].
FEED