Food Values

Each time a new food technology is developed, a new brand advertised, or a food policy debated, there is an associated spike in research eliciting consumer preferences for the particular product or attribute in question. Examples include numerous papers on genetically modified food, hormone use in meat production, food irradiation, nutrition, ecofriendly products, etc. When studying future food policies and technologies, an important question is whether researchers must begin anew and treat each new product or product attribute as independent. Do consumers have some stable set of meta-preferences or food-specific values that drive preferences for food product attributes?

Economists have traditionally refrained from drawing a sharp distinction between values and preferences. The neoclassical view is that consumers make choices based on their preferences. In turn, observed choices can be used to infer people's preferences. As pointed out by McFadden (2001) and others, such consumer theory is virtually tautological: preferences explain choices and given price and income constraints, choices explain preferences. A critical underlying assumption in this framework is that preferences are complete and stable. Research, however, suggests that preferences, the construct that defines choice, may not be as stable as posited by this theory. It appears that people's choices and thus preferences can be affected by the frame and context of the decision task (e.g., Fishhoff 1993; Kahneman and Tversky 2000) and that people use trial-and-error learning to “discover” preferences (e.g., Shogren, List, and Hayes 2000).

In demand theory it is assumed that when a consumer evaluates two consumption bundles, x and y, they can identify which bundle is preferred to the other or whether they are indifferent. The consumer then takes this preference ordering, expressed via a utility function, and maximizes utility subject to a budget constraint. As described by Varian (1982, 1983), if a consumer prefers x to y in one time period but then in another prefers y to x, then the consumer has violated the axioms of revealed preference theory (assuming prices are unchanged). Such a violation implies behavior inconsistent with maximization of a stable utility function. This means, for example, that preferences measured in one time period are no longer of use in predicting demand or calculating welfare changes in another time period. The importance of such preference instability can be noted, for example, by the significant amount of attention devoted to determining whether there was a structural change in meat demand and whether the weak axiom of reveal preferences (WARP) had been violated (e.g., Chalfant and Alson 1988; Eales and Unnevehr 1993; Moschini and Meilke 1989). When interest lies in determining preferences for new or unfamiliar goods, it is quite possible to observe the kinds of preference reversals discussed above.

These observations have led some social scientists (e.g., Hechter 1993), including some notable economists (e.g., Becker 1976), to draw a distinction between potentially malleable preferences (the construct defining choices between market goods) and more stable values. Becker (1976) refers to these values as “ underlying preferences.” He argues, “The preferences that are assumed to be stable do not refer to market goods and services … but to underlying objects of choice. … These underlying preferences are defined over fundamental aspects of life, such as health, prestige, sensual pleasure, benevolence, or envy, that do not always bear a stable relation to market goods and services.” For example, the assumption is not that people's preference ranking of nongenetically modified food over genetically modified food is stable, but rather the more stable preferences are for the outcomes that are expected to disseminate from eating nongenetically modified instead of genetically modified food.

Although values are generally conceptualized as relating to abstract end states of existence, a prudent question is whether there might exist a set of intermediary values that relate specifically to people's food choices, values that are perhaps more stable than people's preference ranking of a specific set of foods or food attributes. The concept of such intermediary values is, in fact, well ingrained in the conceptual framework related to means-end chain theory (Gutman 1982), a framework used by psychologists and marketers that links people's beliefs about concrete product attributes with the abstract values.

A primary purpose of this article is to identify consumers' food value systems. That is, we aim to identify people's beliefs regarding the preferability of competing outcomes resulting from food purchase and consumption along a continuum of relative importance. One of the primary impediments in empirically studying an issue like food values is measurement. Economists have historically shied away from directly measuring constructs that are not directly observable (e.g., Colander 2007). A primary concern is that people do not always use measurement scales identically, causing difficulties in interpersonal and cross-cultural comparisons (e.g., Steenkamp and Baumgartner 1998). Thus, a second primary purpose of this article is to apply recent advances in the field of best-worst scaling (sometimes referred to as maximum-difference scaling) to value measurement. This approach permits the measurement of food values on a ratio scale by observing people's choices of which values are most and least important. Finally, this article investigates the validity of measured values by investigating their relationship to stated and revealed preferences for organic food.

Background and Development of Food Values

In his seminal work on the issue, Rokeach (1973) defined a value as “an enduring belief that a specific mode of conduct or end-state of existence is personally or socially preferable to an opposite or converse mode of conduct or end-state of existence. A value system is an enduring organization of beliefs concerning preferable modes of conduct or end-states of existence along a continuum of relative importance.” He developed a set of eighteen terminal values—values such as “a comfortable life,” “inner harmony,” “pleasure,” and “ wisdom” that relate to desirable end states of existence and another set of eighteen instrumental values—values such as “ambitious,” “honest,” and “self-control” that relate to desirable modes of conduct that lead to particular end states. Rokeach proposed a method to measure the relative importance of these values that involved a straightforward ranking task. He and others have shown that such values have considerable power in explaining variations in wide range of individual behaviors.

More recent work, such as that conducted by Schwartz (1992), has focused on other value classification systems. Schwartz classified values into the following ten categories: achievement, benevolence, conformity, hedonism, power, security, self-direction, stimulation, tradition, and universalism. Some of these value categories do not have a direct relation to food (i.e., power and achievement), but many are of direct relevance. For example, people for whom benevolence is an important value might be concerned about how their food consumption affects others in the supply chain; people for whom hedonism is important may care a great deal about taste and visual appeal; people concerned about security will not likely want to consume foods that will cause illness and financial harm; and people who desire stimulation (i.e., excitement and novelty) will not likely be as concerned about new technologies, etc.

Values research has gained widespread use in marketing research due to the development of means-end chain theory (e.g., Gutman 1982), which argues that a set of core underlying values motivate consumers' purchasing decisions. Buying a particular product or product attribute is conceptualized as a means to obtain some desirable end state. Means-end chain theory, along with the empirical approach of measuring means-end chains called laddering, is widely used by firms for market segmentation and in developing advertising and promotional strategies (e.g., Reynolds and Gutman 1988; Gengler, Mulvey, and Oglethorpe 1999). With the laddering approach consumers are shown a variety of competing products and asked to identify why they might choose one product over another. The typical answer involves the identification of a product attribute such as package size, nutritional content, etc. Then, a monitor asks a question such as, “Why is that important to you?” Once an answer is provided, consumers are further encouraged to state more abstract reasons for why an attribute is important to them, until they respond with a terminal value of the sort described by Rokeach (1973), or until the consumer cannot state a more abstract reason. Common answers across consumers are classified into means-end chains. Examples of the use of means-end chain analysis related to food product marketing and food policy can be found in the work of Bredahl (1999, 2001) on genetically modified food and Makatouni (2002) and Padel and Foster (2005) on organic food.

Measuring consumers' values and means-end chains represents an explicit attempt to identify why consumers prefer a particular product over another. Some have argued that economists should refrain from attempting to determine why consumers choose one product over another (e.g., Gul and Pesendorfer 2005). However, regulatory agencies must often determine the criteria or standards that foods must meet to be called organic, grass-fed, natural, etc. Such decisions implicitly involve judgment calls about the underlying values that each of these claims aim to support. For example, one set of criteria might be optimal if consumers primarily buy organic because they believe it safer than traditional food. A different set of criteria would likely be optimal if consumers desired organics because they believe such foods improve the environment. As argued by Lusk et al. (2006), in the context of country of origin labeling, knowing why consumers prefer a particular attribute is critical in knowing whether a market failure exists and thus whether public policy is needed. In the case of organic food, for example, if such products are bought for private-safety reasons, government intervention in the market may be unnecessary, but if bought to help eliminate environmental externalities, intervention may be warranted. Of course, firm advertising would benefit substantially from knowing why consumers buy certain products, such that advertisements and promotional campaigns can be tailored to the value being served.

Although means-end chain analyses typically focus on linking product attributes to end states of existence, in many cases these are too abstract to be of practical relevance for food policy or advertising. However, the typical laddering approach elicits a number of intermediary values along the chain to the final end states of existence. It is here that we argue for the development of a set of food-specific intermediary values. Selecting the particular items to include as a potential set of food-specific values is no easy task; many previous articles have asked consumers to rank their concern for a variety of food technologies (e.g., Hwang, Roe, and Teisl 2005), food safety issues (e.g., Finn and Louviere 1992), or food attributes (e.g., Baker 1999). These approaches, while useful in their own right, are too specific to be broadly applied when explaining consumers' choices across a variety of foods. Further, such studies often measure people's preferences for technologies or concepts that are relatively novel or unfamiliar and for which people do not likely have well-formed preferences.

The approach taken in this article was to review the extant literature on consumers' willingness to pay for food products and studies conducting food-related means-end chain analyses to identify a set of food values that are likely to be relatively stable across time. The idea is not to identify food attributes per se, but rather more abstract attributes, consequences and “end states” of food consumption that are potentially applicable in explaining choices between a wide range of food products. The list of food values that we compiled along with descriptions shown to consumers in our empirical analysis is shown in table 1.

Some of the items in table 1 could be considered product attributes (e.g., convenience, nutrition, taste, etc.). However, in each case, the value in table 1 is a more abstract construct potentially representing numerous physical product attributes. For example, several studies have investigated people's preferences for fat content and vitamins in food; however, all of these studies are related to the concept of nutrition. Although consumers may not have well-defined preferences over Vitamin A content versus content of Vitamin B12, they are much more likely to know whether nutrition, generally, is more or less important than, say, price. That is, people's relative value for “nutrition” is likely to be much more stable and well-reasoned than people's relative preferences for Vitamin A versus Vitamin B12 or Omega 3 fatty acid versus conjugated linoleic acid. We purposely refrained from selecting terms such as “food quality,” which were too abstract to be operationally meaningful. Some of the values, such as origin, do relate to specific food attributes, but also have broader consequences with outcomes related to the community and with values related to ethnocentrism that are not likely captured by other values. As with Rokeach's (1973) terminal values, the values in table 1 can be classified as either personal (i.e., self-centered) or social (i.e., society-centered). The values of tradition, origin, fairness, and environmental impact would fall into the latter category.

The values shown in table 1 are reasonably comprehensive in covering the breadth of issues likely to motivate consumer food choice, and they undoubtedly capture the major values describing food choice. Still, we are under no illusion that other researchers might have generated a slightly different list had they undertaken the same task. Our response is the same as Rokeach's (1973): “The overall procedure employed in selecting the two lists is admittedly an intuitive one, and there is no reason to think that others working independently would have come up with precisely the same list of … values. It would be interesting to see which values others might produce working independently.”

In the following we describe the approach we use to measure the relative importance consumers place on the values in table 1 and identify whether the measured values have explanatory power in explaining food choice.

Data and Methods

To determine the relative importance consumers place on the values in table 1, we utilized recent advances in best-worst scaling. The method was originally introduced by Finn and Louviere (1992), and the theoretical properties of probabilistic, best-worst choice models were recently elucidated by Marley and Louviere (2005). The method is rapidly gaining popularity in business-marketing research (e.g., Sawtooth Software 2007) and has been recently been applied to health care issues (Flynn et al. 2007) and to measure ethical beliefs in an international context (Auger, Devinney, and Louviere 2007). Lee, Soutar, and Louviere (2007) have even applied the best-worst approach to measuring people's overall life values. Their results suggested that the best-worst approach provides a better measure of people's values than some frequently used rating approaches.

The best-worst method works as follows: consumers are shown a set of items and are asked to indicate which is best (or most important) and which is worst (or least important). Consumers make several repeated choices where the set of items varies across questions. Responses to the questions can be used to measure each item's position on a continuum of the construct of interest—“importance” in this case. Figure 1 provides an example of one of the best-worst questions used in our application.

Best-worst scaling has several advantages over other methods of measurement. A major difficulty with rating-based methods (where a person responds on a scale of 1 to 5, with 1 being not important and 5 being very important) is that people are not forced to make trade-offs between the relative importance of issues. Indeed, it is common for people to say all issues are “important.” Another problem with rating-based methods is that different people use the scale differently, with a 5 for one person possibly representing a 4 for another. At best these measures provide measurement on an ordinal scale. Best-worst scaling avoids these pitfalls. By having people choose the best and worst options, people are forced to decide which issues are more or less important, and unlike rating scales, there is only one way for people to respond to the question (with a choice).

Best-worst scaling is an extension of Thurstone's (1927) paired comparison method, which has a long history in psychological research. In the paired comparison method people are simply presented two items and are asked which is best (or more important). Best-worst scaling provides much more information than paired-comparisons, so much so, that it is possible to construct individual-level scales and provide a measurement of the underlying construct on a ratio scale.

As shown in table 1, there are eleven food values of interest. The challenge was how to allocate these eleven values to different choice sets in a survey. We used a 2¹¹ main-effects orthogonal experimental design to assign each of the eleven values to choice sets.¹ The resulting design consisted of twelve choice sets. Thus, each respondent in our study answered twelve best-worst questions. Of the twelve choice sets, six contained four values, three contained six values, and three contained eight values. Each of the food values appeared six times across the twelve choice sets. In summary, each survey contained all twelve choice sets, and consumers were asked to indicate the most important and least important issue for each set as shown in figure 1.²

Econometric Analysis

When responding to each best-worst question, consumers can be conceptualized as choosing the two items that maximize the difference between two items on an underlying scale of importance. If a choice set has J items, or values in this case, then there are J(J − 1) possible best-worst combinations a person could choose. The particular pair of items chosen by the consumer as best and worst, then, represents a choice out of all J(J − 1) possible pairs that maximizes the difference in importance.⁴

Table 2. Characteristics of Survey Respondents

Variable	Definition	Mean	Standard Deviation
Gender	1 if female; 0 if male	0.351	0.479
Age	Age in years	55.604	15.263
Degree	1 if obtained college degree; 0 otherwise	0.611	0.489
Child	1 if child under age of 12 living in household; 0 otherwise	0.166	0.373
Income	Annual household income in $1,000s	74.044	37.567
West	1 if resides in West U.S. census region; 0 otherwise	0.201	0.402
Midwest	1 if resides in Midwest U.S. census region; 0 otherwise	0.339	0.475
South	1 if resides in South U.S. census region; 0 otherwise	0.282	0.451
Northeast	1 if resides in Northeast U.S. census region; 0 otherwise	0.178	0.384

• ^a Note: Number of observations is 176.

Formally, let λ_j represent the location of value j on the underlying scale of importance, and let the true or latent unobserved level of importance for individual i be given by I_ij = λ_j + ɛ_ij, where ɛ_ij is a random error term. The probability that the consumer chooses, say, item j and item k, as the best and worst, respectively, out of a choice set with J items, is the probability that the difference in I_ij and I_ik is greater than all other J(J − 1) − 1 possible differences in the choice set. If the ɛ_ij are distributed i.i.d. type I extreme value, then this probability takes the familiar multinomial-logit (MNL) form:

(1)

The parameters λ_j can be estimated by maximization of the log-likelihood function based on the probability statement in (1). That is, the dependent variable takes the value of 1 for the pair of values chosen by the consumer as best and worst, and a 0 for the reaming J(J − 1) − 1 pairs of items in the choice set that were not choicen as best and worst. The estimated λ_j represents the importance of value j relative to some value that was normalized to zero to prevent the “dummy variable trap.”

One of the primary objectives of this article is to determine how values affect preferences for food, and as such, we need information on each individual's values. Unfortunately, the MNL described in equation (1) assumes that all individuals in the sample place the same level of importance on each value (i.e., there is no i subscript on λ_j). To overcome this weakness of the MNL, we also estimated a random parameters model (RPL). In particular, let the importance parameter for individual i and issue j be specified as , where and σ_j are the mean and standard deviation of λ_j in the population, and μ_i is a random term normally distributed with mean zero and unit standard deviation. Such a specification implies that the importance of food value j is assumed to be distributed according to a normal distribution with mean and standard deviation σ_j. Substituting into equation (1) yields a probability statement that depends on the random term in μ_ij. Rather than attempting to explicitly integrate over these random terms, following Train (2003), the model was estimated via simulation. In particular, the parameters were estimated by maximizing a simulated log-likelihood function, evaluated at 200 pseudorandom Halton draws for μ_ij. The random draws are individual-specific, which takes into consideration the fact that each person answered twelve best-worst questions. Train (2003) provides more information on computational details for the RPL.

One of the underlying assumptions of the RPL is that the variance of ɛ_ij is assumed to equal one. It is possible, however, for the scale to differ by people or alternatives (Louviere 2001). The RPL model outlined above can be also interpreted as an error-component model where each alternative (or food value) has a separate variance (e.g., Train 2003), and as such the model accommodates differences in potential scale over alternatives. Given the normalization of the variance of ɛ_ij to one, however, it is possible that the mean estimates of λ_j remain confounded with differences in scale, and as such, it is important to recognize that the estimated population parameters reflect both differences in mean and scale. Fortunately, McFadden and Train (2000) have shown that the RPL can approximate any underlying random utility model, and as such, the RPL can produce goods estimate of predicted probabilities, even if the there is some misspecification in the choice model.

Given the potential confound with scale, we calculate a “share of preference” for each food value, which is the forecasted probability that each food value is picked as most important:

(2)

These shares of preferences must sum to one across all eleven food values. Equation (2) reports the importance of the value j on a ratio scale, meaning that if one value has a share value twice that of another value, it can accurately be said that the former value is twice as important as the latter. It is important to note that the calculated share of preference for a food value will reflect both true importance of the value as well as relative uncertainty in the importance people place on the value; however, the share of preference calculations convey a key message—the probability that an issue is picked as more important than another.

As shown by Train (2003) and Huber and Train (2001), once the parameters from the RPL are estimated, so-called individual-specific estimates can be obtained by using the estimated parameters as a prior and using each person's actual choices to form an individual-specific posterior estimate. These posterior estimates of food values for each individual can be compared with each person's stated and revealed preferences for organic food to determine if relationships among the variables exist. It is important to note that these Bayesian calculations do not produce each person's value parameters per se, but rather as discussed thoroughly by Train (2003), they are the means of the conditional distribution (conditional on the person's actual choice), which are not necessarily the same as person's actual coefficients. As Train (2003) shows, however, the difference between these two statistics becomes small when people face even ten choice situations and he further shows that such individual-level Bayesian estimates improve the ability to predict the choices that each person actually makes. Thus, while we use the term individual-specific values, the estimates are, more precisely, the mean of the parameter distribution conditioned on each individual's actual choices.

Results

Table 3 shows the relative importance of each of the eleven food values as estimated by the MNL and the RPL models. The importance of each food value was estimated relative to origin (the least important value according to the estimated models). Results reveal that safety, on average, is the most important food value and significantly more important than origin. Nutrition, taste, and price are the next most important food values on average, whereas environment, fairness, and tradition are the least-important food values. On average natural, convenience and appearance were of intermediary importance but were significantly more important, on average, than origin. Despite the averages, results reveal large standard deviations for the importance measures (note: the MNL restricts the standard deviations to equal zero), implying that there is significant heterogeneity in the sample with regard to relative importance of food attributes. Indeed, despite the fact that tradition and origin were the least important food values, on average, there were some people in the sample who picked these issues as most important every time the issue appeared in one of the twelve best-worst questions.

One interesting observation is that origin was found to rank last in average importance, but some previous studies related to beef have found origin of production to be very important to consumers (e.g., Menneke et al. 2007; Loureiro and Umberger 2007). Given recent attention to the issue of origin in policy circles, a brief discussion on the discrepancy in these findings is in order. Previous research such as that conducted by Loureiro and Umberger (2005) showed that U.S. consumers believe that U.S. beef is safer than foreign beef. As such, origin is likely to serve as a proxy for food safety and may even serve as a proxy for other attributes such as fairness, tradition, and taste. In previous country-of-origin studies conducted by Menneke et al. (2007) and Loureiro and Umberger (2007), it is unclear whether the measured preferences for origin were a result of uncontrolled beliefs about differences in food safety, tradition, fairness, environmental impact, etc. across origins.

Our approach explicitly measures preferences for food safety relative to origin. Because our study shows food safety as the most important food value, it is possible that if origin is a proxy for food safety it too would also be ranked highly on a list of importance if safety preferences were not fully disentangled from preferences from origin. There are also differences in the way importance is calculated in previous conjoint studies and in the present best-worst scaling approach. In the conjoint studies importance is measured by calculating the utility difference in the best and worst levels within an attribute, but in our study no specific attribute levels are specified, and the explicit attribute levels used in the previous conjoint studies may or may not be the levels that endogenously come to mind for the consumer. Finally, these previous studies dealt specifically with beef steaks, whereas our study deals with food in general and meta-preferences.

One of the difficulties in evaluating the importance of each value that results from the MNL and RPL models is that the estimates themselves have no natural interpretation. To remedy this problem, the last two columns in table 3 report the shares of preference for each value as calculated using equation (2). According to the RPL model, about 34.2% of people would rate safety as the most important food value. Interestingly, price has the next highest share of preference, with 21.5% of people on average believing price to be the most important value. Although nutrition and taste are important to consumers, food safety is about twice as important as these two values. Less than 3% of people would pick any of the remaining food values (natural, convenience, appearance, environment, fairness, tradition, or origin) as most important. Nevertheless, it appears that natural and the environment are one-and-a-half to two times as important as convenience and fairness, and more than five times as important as origin.

Table 4 reports the correlation between people's food values, as determined by the individual-specific estimates stemming from the RPL. None of the value estimates exhibit correlations above 0.5, indicating that each of the values represents a unique construct. Still, there are groups of values that appear to be related. For example, safety and taste exhibit negative correlations, meaning people who believe safety is important are less likely to believe taste is important. One interesting finding is that the correlation between safety and naturalness is small and negative. Many have hypothesized that concern for modern food technologies is a result of food safety concerns, but the low correlation between these two food values suggests otherwise. Interestingly, people who place a high importance on fairness tend to place low values on the more self-centered values of taste, price, convenience, and appearance.

The preceding results are useful in identifying the relative importance of food values in this somewhat small sample of individuals, but an important question is whether these values exhibit explanatory power over these people's choices. To address this issue, consumers were asked to state how much more they were willing to pay for organic bread over nonorganic bread. Table 5 reports Pearson and Spearman-rank correlation coefficients between each person's measured food value as determined by the RPL individual-specific shares of preference and the person's willingness-to-pay premium for organic bread.⁶ Overall, results are consistent with a priori expectations. People for whom price is more important stated lower willingness to pay for organic food than people for whom price is less important. There is a relatively large correlation between importance of naturalness and willingness to pay for organic bread, suggesting that naturalness (defined as extent to which food is produced without modern technologies) is a key motivating factor behind people's preferences for organic food. Interestingly, one of the factors often attributed to organics—safety—was not significantly related to stated willingness to pay a premium for organic food, suggesting safety consideration may not be a driving force behind growth of the organic food market. As expected, table 5 shows that the relative importance of the environment was positively correlated with willingness to pay a premium for the organic food.

Because responses to contingent valuation questions can be strongly influenced by the frame of the question, including the hypothetical nature of the inquiry, we also asked a revealed preference question by asking people whether they had previously purchased organic food. Across the entire sample, 64% of respondents indicated they had previously purchased organic food. Table 6 reveals significant differences in mean shares of preferences of those people who have and have not previously purchased organic food. Overall, results are very similar to the correlations with stated preferences reported in table 5. For example the mean share of preference for price was 26.2% for nonpurchasers but 17.7% for purchasers. That is, price was a less important attribute for those people who purchased organic food than for those who had not previously purchased organic food. Purchases of organic believed naturalness and the environment to be significantly more important than nonpurchasers of organic.

Conclusions

Agricultural economists have devoted a significant amount of attention in recent years to estimating people's preferences for one food product attribute over another. Such studies often focus on measuring preferences for goods and attributes for which people have little prior knowledge or experience and unfortunately, this means that the measured preferences may be less stable than is often postulated in theoretical models of consumer decision-making. This article sought to identify a set of food values or meta-attributes for which people may have more well-defined preferences. Implicit in the approach is an attempt to crack open the consumer's “black-box” and provide some insight into why she/he chooses one product or attribute over another.

Based on previous literature on human values and food preferences, a list of eleven food values was compiled. The relative importance that people placed on each of the food values was measured using the relatively new approach of best-worst scaling, which involves people stating the most and least important issue out of a set of competing issues. Results reveal that the values of safety, nutrition, taste, and price were among the most important to consumers on average; however, we also found significant heterogeneity across consumers in terms of the relative importance placed on food values.

We found that measured values were significantly related to stated and revealed preferences for organic food. People for whom naturalness, fairness and the environment were more important were more likely to have previously bought organic food and to state a higher willingness to pay for organic food. The opposite was true for people that placed a relatively high level of importance on price. Overall, the relative importance people placed on naturalness, a value that was defined as the extent to which food is produced without modern technologies, seemed to have the largest influence on stated and revealed preferences for organic food.

The research reported in this article represents a first attempt at measuring people's values for food. More work is needed. First, if interest lies in identifying the relative importance of food values in the population, work such as that presented here should be repeated with a larger and more representative sample of consumers. Second, this study only compared the relationship between food values and preferences for organic food. One benefit of measuring food values is that they should have significant explanatory power in explaining choices between a wide range of food products, and future research will reveal whether this is indeed the case. Third, additional work along the lines of that outlined above could help predict consumers' responses to food policy and marketing changes. Small-scale laboratory research can be used to determine the link consumers make between specific food attributes, such as use of biotechnology, fat content, meat tenderness, etc. and food values. Then, knowledge of the distribution of food values in the population can aid in identifying how consumers, in aggregate, will respond to changes in food composition. Fourth, a benefit of measuring food values is that these constructs should be more stable over time; however, even concepts such as “culture,” which are thought to be relatively impermeable are known to change. This suggests that there may be some benefit in tracking a panel of consumers over time, measuring how and when food values change. Finally, measuring differences in food values across countries may have significant utility in explaining international differences in food consumption patterns and responses to new food technologies.