SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

While previous studies have investigated country-of-origin effect from various angles, the extent to which Country-of-Origin Labelling (COOL) affects U.S. beef imports from specific countries remains unexplored. Using data from 1,079 consumers from the United States, we examined consumers’ willingness to pay (WTP) for Canadian and Australian beefsteaks. We also estimated WTP for bovine spongiform encephalopathy (BSE)–tested traceability-enabled, tenderness-assured, and natural beef. The results from both a mixed logit model and a latent class model (LCM) revealed unobserved taste heterogeneity and important differences in the WTP between the imported and domestic steak. The LCM, for instance, estimated the range of discount needed for consumers to switch from U.S. to Canadian steak as $1.09 to $35.12 per pound. This strongly suggested that U.S. consumers prefer domestic-originated beef to imported beef. In addition, consumers were found to be willing to pay significant amount for BSE-tested, traceability-enabled, and tenderness-assured beef.

Bien que des études antérieures aient examiné les répercussions de l’étiquetage du pays d’origine sous différents angles, les répercussions de cet étiquetage obligatoire sur les importations étatsuniennes de bœuf en provenance de pays spécifiques ne l’ont pas été. À l’aide de données tirées d’un échantillon de 1079 consommateurs étatsuniens, nous avons examiné le consentement à payer (CAP) pour du bifteck en provenance du Canada et de l’Australie. Nous avons également examiné le CAP des consommateurs pour du bœuf provenant d’un animal ayant subi un test de dépistage de l’ESB, traçable, de tendreté assurée et naturel. Les résultats obtenus à l’aide d’un modèle logit mixte et d’un modèle à classes latentes ont révélé une hétérogénéité non observée du goût et des écarts importants dans le CAP pour du bifteck provenant des États-Unis et de l’extérieur du pays. Le modèle à classes latentes, par exemple, a révélé que les écarts de rabais nécessaires pour que les consommateurs délaissent le bifteck américain pour le bifteck canadien variaient de 1,09 $à 35,12 $ la livre. Ces résultats montrent clairement que les consommateurs étatsuniens préfèrent le bœuf des États-Unis plutôt que le bœuf importé. Les résultats montrent également que les consommateurs sont prêts à payer plus cher pour du bœuf provenant d’un animal ayant subi un test de dépistage de l’ESB, traçable et de tendreté assurée.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

Previous studies on the U.S. Mandatory Country-of-Origin Labeling (COOL) indicated that consumers are generally in favor of the policy (Schupp and Gillespie 2001). Further, Loureiro and Umberger (2003, 2005, 2007) found that American consumers are willing to pay more for U.S.-labeled beef compared to unlabeled beef. However, in a meta-analysis that spans beyond agricultural products, Verlegh and Steenkamp (1999) found no significant country-of-origin effect on consumer purchasing intentions. In the light of a World Trade Organization litigation brought against the United States as the result of COOL, the extent to which consumers may be willing to pay for imported beef from specific countries is timely.

Additionally, we explored consumers’ preference for traceability-enabled, bovine spongiform encephalopathy (BSE)–tested, tenderness-assured, and natural beef. Although the primary focus is on the impact of COOL, realistically, steaks are often bundled with multiple attributes. Discussion with marketers and multiple studies signal the potential of these value-added attributes (Dickinson and Bailey 2002; Lusk et al 2003; Bailey et al 2005; Thilmany et al 2006; Tonsor et al 2009; Verbeke and Roosen 2009; Yang and Goddard 2011). None of the attributes has been a widespread success yet in the U.S. market. With a plethora of attributes created by marketers in response to consumers’ increasing attention to food safety and quality, Verbeke (2008) warned that information overload could result in rational ignorance—where consumers disregard information attached in a product (McCluskey and Swinnen 2004). By examining the willingness to pay (WTP) for these attributes, we can understand how they jointly affect consumer choices, which should be of interest to meat marketers.

BACKGROUND

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

The COOL provision of the 2002 and 2008 Farm Bills caused controversy among nations that export food products into U.S. markets. The final ruling by the USDA Agricultural Marketing Service went into effect on March 16, 2009, requires information regarding country-of-origin to be labeled on a number of fresh food including vegetables, fruits, and meat. On beef, the law mandates only products derived from cattle born, raised, and processed in the United States can be labeled as U.S. origin (USDA 2009). The law, in essence, differentiates imported beef from domestic beef at the retail level, which could have widespread consequences on demand of imported food. This prompted the governments of Canada and Mexico to challenge the legitimacy of COOL in accordance with the World Trade Organization's principle of national treatment (Suppan 2009).

The importance of the U.S. market for many beef exporting countries cannot be understated. The U.S. market accounts for about 30% total beef and veal sales for Canada, New Zealand, and Nicaragua. Cattle exports from Canada and Mexico were almost exclusively destined for the U.S. market (USDA 2010). Trade representatives of Canadian cattle and beef industry claimed the law is “devastating the Canadian livestock industry” and could result in a “glut of meat on store shelves in Canada” (Wyld 2009).

Proponents of COOL argue that consumers have a right to know where food comes from. With COOL, consumers can use label information to assess the quality and safety of the products. Some domestic producers also maintain that COOL may reduce search cost of those preferred domestic food products (Lusk et al 2006). Because origin of food products is a credence attribute, without COOL, supporters contended that consumers who wish to consume domestic food products could not do so, because they lack the necessary information regarding the origin of the product. Under these conditions, the absence of a COOL law could be made a case for market failure (Darby and Karni 1973; Caswell 1998).

Critics of COOL contested the role of COOL as a food safety measure. Ikenson (2004) contended the Food Safety and Inspection Service would not allow importation of any unsafe foods; COOL also exempts restaurants and smaller butcher shops, which diminishes the effectiveness of COOL's role as a food safety measure. Further, Krissoff et al (2004) noted that foods are rarely voluntarily labeled with sources of origin, which cast doubt on the true appeal of domestic origin to consumers; they argued, profit maximizing retailers, processors, and producers would voluntarily indicate products origin with labels if they deem the benefits exceed the cost.

Whether COOL is warranted depends heavily on consumers’ preference, as well as the extent that COOL might penalize imported food. By examining consumer preference for origin-differentiated beef, this study contributes to the debate on COOL.

PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

Previous studies suggested that consumers may use country-of-origin as an extrinsic cue in evaluation of the quality of the product (Hoffmann 2000; Northen 2000; Grunert 2005; Lusk et al 2006). In summary, country-of-origin may invoke consumers’ knowledge and beliefs regarding the place of production of the products. Additionally, in cases of repeated purchase on products without a strong brand, as with most fresh food, consumers may use the origin to re-identify the quality that they have found appealing.

Increased international competition from trade liberalization incentivized producers to use country-of-origin information to differentiate their products. Marette et al (2008) argued that with imperfect information and imperfect competition, domestic producers may gain from geographical-indication labels. When faced with the choice of familiar domestic products and unfamiliar imported products, domestic products inevitably emerge as the choice when the lack of knowledge or information regarding the quality of the imported products could induce uncertainty in consumers.

The country-of-origin effects gained research attention following introductions of mandatory origin-labeling law in the European Union, and more recently in the United States. Studies conducted on European consumers reveal consumers used country-of-origin to predict the eating quality and safety of beef (Becker 2000; Davidson et al 2003). In its U.S. counterpart, Schupp and Gillespie (2001) found a vast majority of the surveyed indicated support for mandatory labeling of origin on fresh and frozen beef sold in retail market. Further, 83% of the respondents rated U.S. beef higher quality and safer than imported beef. Multiple studies indicated European consumers are willing to pay more for domestic meat than imported meat (Alfnes and Rickertsen 2003; Alfnes 2004; Mørkbak et al 2010).

In a U.S. nation wide survey, Loureiro and Umberger (2007) found a positive WTP for beef labeled as U.S. products compare to unlabeled products. Further, they suggested that the WTP for USDA food-safety-inspection certifications is higher than U.S.-labeled beef, but the WTP for tenderness assurance and traceability is lower than U.S.-labeled beef. However, the difference in WTP for domestic versus imported beef is absent. In addition, the rankings of the attributes, which were estimated through a conditional logit framework,1 could be further scrutinized using estimators capable of discerning unobserved taste heterogeneity.

Previous studies point strongly to the connection between consumers’ perception and country-of-origin effect. We explore the differences in consumers’ perceptions of safety between domestic beef and imported beef from specific countries. In addition, this study expands Loureiro and Umberger's (2007) work in significant ways: we refine the scope of investigation to the difference in WTP between domestic-labeled steak and steak labeled as imported. Further, we investigated consumers’ relative preference of additional value-added attributes in the form of BSE-testing and natural beef. Using a mixed logit model (MLM) and a latent class model (LCM), we incorporated heterogeneous consumers’ preference in this analysis as well.

RESEARCH DESIGN

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

We conducted an online survey through TNS Global in May 2010. The sample was randomly selected through the vast panelist network of TNS Global. Respondents below age 17 were restricted from participation.2 We designed and tested the survey following general guidelines given in Dillman (2007). The survey is divided into two sections: the first part included questions pointed to consumers preference on beef adapted from related literature and demographic information: and the second section included a choice experiment to assess consumer WTP for imported beef and the aforementioned attributes. Consistent with previous literature (e.g., Tonsor et al 2009), the target responses were set as 1,000. The online survey closed with 1,079 responses. We did not pursue a mail survey after taking into account the challenges in targeting and obtaining a national sample. Nonetheless, Olsen (2009) suggested that Internet surveys are viable alternative to mail surveys in estimation of consumer WTP.

The design of the choice experiments was similar to that developed by Schroeder et al (2007) and Tonsor et al (2009). In this way, the results could be compared across studies. However the focus of the survey used in this study was narrowed to some specific interventions—BSE testing and traceability, rather than the food safety levels as used in the other studies. We used a fractional factorial design3 to generate the choice set in this study, which follows the same design as Aubeeluck (2010). The procedure produced 191 choice sets. To maintain a balance between respondent fatigue and degrees of freedom, these choice sets were distributed across 14 versions of the survey, 12 versions contained 14 choice sets, one version contained 13 choice sets, and one version contained 10 choice sets.4 We assigned approximately 77 individuals to complete each version of the survey. Each choice set presents choices of two steaks bundled with various attributes and prices (see the Appendix for a sample choice set); if neither steak appeals to them, the third choice of not buying (would-not-buy option) could be chosen.

We chose strip loin steak as the representative product for its well-defined and relatively homogenous properties. The choice profiles consisted of attributes from five categories: price, country-of-origin, production practices, tenderness, and food-safety assurance. Table 1 provides the description of these attributes. Four levels of prices were chosen ranging from $5.50 to $16.00, which reflected the low-end and high-end prices that could be observed in actual grocery store settings for steak at the time of this study.

Table 1.  Attributes levels and descriptions
CategoriesLevelsAbbreviationsDescriptions
Price ($/lb)  Refers to steak price in retail grocery store or butcher where the respondent typically shops
 5.50  
 9.00  
 12.50  
 16.00  
Country-of-origin  Refers to country in which the cattle were raised
 USA  
 CanadaCAN 
 AustraliaAUS 
Production practices  Refers to the method used in production
 Approved standards  Approved standards means production involved government-approved synthetic growth hormones and antibiotics
 NaturalNAT Natural means animal was raised without the use of synthetic growth hormones or antibiotics
Food safety assurance  Refers to the food safety assurance offered with the steak
 None  
 BSE-testedBSE BSE-tested means that cattle are tested for BSE prior to slaughtering process
 TraceableTRC Traceable means the product is fully traceable back to farm of origin from the point of purchase
 BSE-tested and traceableBSE_TRC BSE-tested and traceable were offered in combination
Tenderness  Refers to the softness in the steak's eating quality
 Not specified  Not specified means there are no guarantees on tenderness level of the steak
 Assured tenderTENDER Assured tender means the steak is guaranteed tender by testing the steak using a tenderness measuring instrument

In conjunction with domestic beef, Australian and Canadian beef were used as these two nations are the biggest volume exporters of beef to the United States. Canadian beef is noted for its similarity to U.S. beef in terms of breed, marbling, and feed. In contrast, Australian beef are typically grass-fed, which differs in eating quality to U.S. and Canadian, beef (Brester et al 2004; Mutondo and Henneberry 2007). While there may be notable difference in characteristics and eating quality between U.S., Canadian, and Australian steak, it is not clear how much typical consumers in the United States are aware of these differences especially given the lack of clear indication of origin prior to COOL.

All other attributes were determined by examining the related literature as well as discussing with beef experts and focus group members. The phrase natural steak refers to steak derived from cows raised without synthetic growth hormones and antibiotics, as opposed to approved standards, which means the cow is raised using government-approved growth hormones and antibiotics. In the choice experiment, steak may be assured tender or not specified. In the food-safety assurance category, a steak can be traceable, meaning that steak products on the market are traceable back to an animal from a specific farm/producer. A steak can be BSE-tested which suggests that the cattle where the steak is from was tested and verified free of BSE by the appropriate government agency. A steak can also be both BSE-tested and traceable. Notice that for these quality attributes, no specific agency was indicated as the organization who may issue the guarantees/assurances. This is to avoid consumers attaching specific values/disvalues associated with various agencies (Steiner et al 2010). Although consumer response to quality assurance issued by various organizations can be an interesting area of research, it is beyond the scope of this current study. All attributes were explained to the respondents in an information sheet (attached in the Appendix) before they were asked to complete the choice experiment. Readers may also refer to the informational sheet in the attached Appendix for a view of the choice sets given to survey our respondents.

Hensher et al (2005) noted omitting the would-not-buy alternative constrained decision makers into making a choice from the listed alternatives, which are effectively conditional choices and may not reflect all options available to decision makers in the real word. The inclusion of the would-not-buy option reflects a more realistic choice environment, where respondents are allowed to delay or decline to make a choice if the options presented are not appealing.

The validity of stated preference analysis, such as choice experiments, is debated for its potential downfall of hypothetical bias—where the lack of incentive-compatibility in the experimental nature of stated preference may lead to overstatement of WTP. Nonetheless, for new or hypothetical attributes such as the attributes examined in our study, the lack of reveal preference data necessitate the use of stated preference method. Other stated WTP elicitation methods, such as contingent valuation may be used, but a choice experiment is well-suited for a multiple-attributes setting as in this study (Adamowicz et al 1998). In an overview, Loomis (2011) concluded that no widely accepted methodology exists to control for hypothetical bias. Additionally, Lusk and Schroeder (2004) and List et al (2006) suggest that the marginal WTP for private goods produced by choice experiments is comparable to WTP measures from experimental auctions, which are revealed preference alternatives to choice experiments and are often used to investigate the behavior of a small group of consumers. Nevertheless, readers should be aware of the contentions around the WTP elicitation methods.

SUMMARY STATISTICS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

Table 2 presents the summary statistics. Eighty-three percent of the respondents identified themselves as the primary shopper in their household. The mean household annual income was a little over $52,000 and the median education level of the respondents was some college (including community college or technical training). Our sample corresponded closely to the U.S. population in terms of gender, education, and income, but it heavily over represented older consumers; the higher portion of older respondents could be due to the length of the survey deterring participation of younger age groups who may have more time constraints. Overrepresentation of the older population in online consumer surveys is not uncommon in the literature. For instance, Hu et al (2005) and Loureiro and Umberger (2007) reported mean age of higher than national average in their surveys. Nevertheless as with all surveys, readers should be cautious about the ability of the sample to represent the entire consumer population.

Table 2.  Sample descriptive statistics
   SampleU.S. Census
VariableGroupPercentmean/mediandata
  1. aNote: Median values.

Age15–19 0.93%56.62 36.8a   
 20–24 3.52%  
 25–29 2.22%  
 30–39 7.78%  
 40–4912.70%  
 50–6432.25%  
 65+40.59%  
GenderMale47.54% 49.20%
 Female52.46% 50.80%
Education<High school 1.11%14a  12a   
 High school23.08%  
 Some college39.39%  
 4 year degree24.28%  
 Graduate12.14%  
Household Income ($)<25k24.10%52.37k51.42k 
 25–40k23.54%  
 40–65k23.82%  
 65–80k 9.55%  
 80–100k 7.32%  
 100–120k 6.12%  
 >120k 5.56%  
Freq. grocery shoppingNever 1.85%  
 Sometimes14.74%  
 Frequently83.42%  

The survey also queried respondents’ frequency of beef consumption.5Table 3 reports the statistics. The vast majority (93.8%) of the respondents consumed beef. Further, the segments of beef consumers in the sample provided an approximated percentage of different cuts they consumed over the previous year.6 The most commonly consumed beef cut were ground and minced beef, which approximately accounted for 41% of the respondents’ total beef consumption. Steaks accounted for approximately 24% of the consumption.

Table 3.  Beef consumption patterns
Frequency of beef consumptionFreq.Percent
Never 67 6.21
Occasionally49245.60
Regularly52048.19
Beef CutMean percentStd. dev.MinMax
Ground and minced41.06 23.55 0100
Roasts15.50 12.61 0 90
Steaks24.18 19.02 0100
Sausages and other processed meat16.29 13.46 0100
Organ meat2.15 5.03 0 35
Other parts0.8116.2750100

PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

To assimilate consumers’ reaction to COOL, we elicit the sampled consumers’ preference for origin of beef. In this question, the respondents picked their most preferred country-of-origin for beef.7 The options were Australia, New Zealand, Canada, other countries, avoid imported beef, and neither like nor dislike imported beef. Figure 1 reports the result. While the majority (65.7%) indicated indifference toward imported and domestic beef, some of these respondents might pick this option to avoid sounding discriminatory. It is far reaching to conclude the majority of U.S. consumers to be equally likely to purchase imported and domestic steak based on these observations. Nonetheless, we expect these respondents to place less importance on the origin of beef. Consumers’ country-of-origin preference in beef is further explored with econometric analyses.

image

Figure 1. Stated country-of-origin preference for beef (N= 1,079)

Download figure to PowerPoint

More than one-quarter (27.5%) of the sample stated they do not prefer imported beef. Although a minority, this group may be large enough to induce the reluctance to practice voluntary origin labeling if retailers deem that the consequences of selling origin-labeled imported beef exceeds the benefits. After domestic beef, 4.4% of the sample preferred Canadian beef. Beef from Australia, New Zealand, and Argentina combined were preferred by 2.4% of the sample.

To address proponents claim that COOL could serve as a food safety cue, we dedicated a question8 to elicit respondents’ perceived safety levels on beef from various origins. Along with a no-opinion option, the respondents rated in five points Likert-scales (1 = very low perceived safety; 5 = very high perceived safety) for beef from unknown origin, Australia, Brazil, Canada, New Zealand, and the United States. Table 4 reports the result. As anticipated, the respondents perceived domestic beef as the safest. In contrast, unknown origin was perceived to be the most unsafe. Canadian beef ranked second despite multiple BSE cases reported over the last decade (Maynard and Wang 2010), follow by Australia, New Zealand, and Brazil. These rankings coincided with previous findings in Loureiro and Umberger (2005). More than 30% responded no opinion in regards to safety of imported beef, indicating limited experience and knowledge on imported beef. The pairwise t-test rejected the notion that the respondents perceived beef from other origins to be as safe as U.S. beef. Although a confident statement can be made that U.S. beef is perceived to be the safest in general, some consumers may still prefer imported beef. A taste-panel study by Sitz et al (2005), for instance, showed that a minority of consumers prefer Australian grass-fed steak. We address the aspect of taste heterogeneity econometrically in the next section.

Table 4.  Perceived beef products safety levels of various country-of-origin (N= 1,079)
Countries of originMeanaStd. dev.% No opinion
  1. Notes: a1 = very low; 5 = very high.

  2. bTests of differences in mean perceived safety of meat originated from the United States against other origins.

Unknown origin2.421.2836.05  
Australia3.241.1234.66  
Brazil2.831.0937.16  
Canada3.401.1030.49  
New Zealand3.211.1334.66  
United States3.811.0910.84  
Hypothesis testb t-test value p-value
 Ho: μus–μunknown origin= 018.320.000
 Ho: μus–μAustralia= 010.860.000
 Ho: μus–μBrazil= 015.800.000
 Ho: μus–μCanada= 0 9.340.000
 Ho: μus–μNew Zealand= 011.460.000

RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

MLM and LCM have been widely applied to capture unobserved preference heterogeneity in empirical research (Alfnes 2004; Scarpa and Del Giudice 2004; Hu et al 2005). Greene and Hensher (2003) provided an excellent exposition of both the models.

MLM assumes that the parameters associated with product attributes follow some parametric distribution, instead of fixed as in a conditional logit model (CLM). The distribution of random parameters captures taste heterogeneity. In addition, MLM is free of the independence of irrelevant alternatives assumption, thus reflects a more realistic substitution pattern than CLM (Train 2003; Hensher et al 2005).

In contrast, LCM assumes that individuals can be assigned into a set of Q classes each representing the cluster of individuals who behave in a particular way. The most notable difference between MLM and LCM is on the distributional assumption of the parameters associated with product attributes. Since LCM is semi-parametric, analysts are free from making potentially unreasonable distribution assumptions about the unobserved heterogeneity. However, Greene and Hensher (2003) argued that the extra flexibility in fully parametric MLM might compensate for having to make the distributional assumptions.

Random utility theory (McFadden 1974) furnishes MLM and LCM with an economic interpretation. The utility function of a consumer, i, faces alternatives, j, in choice set, t is denoted as:

  • image(1)

Vector xjt represents the attributes as described in alternative j in choice situation t. The model estimates the unknown parameter vector inline image. The error term inline image signals the randomness of the utility. Assuming utility maximizing behavior, the individual chooses alternative j if and only if the utility associated with alternative j is greater than other alternatives. McFadden (1974) showed that if the error term follows an independent and identically distributed (iid) maximum extreme value Type I distribution, the resulting choice probability is the conditional logit choice probability. It follows that the choice probability of individual i choosing alternative j in the tth choice set is represented as:

  • image(2)

The Mixed Logit Model

The MLM assumes that the unknown parameters β are random rather than fixed, thus allowing them to capture taste variation. Each random parameter β is assumed to be distributed as:

  • image(3)

Researchers are free to specify any appropriate probability distribution function on the random parameters, denoted as h (.). The random parameters β includes the mean value to be estimated θ, and an iid error-term v. The matrix Ω represents the covariance matrix of the parameters. The attributes can be specified to reflect correlation among each other. With correlated parameters specified, h (.) becomes a joint probability density function and the off-diagonal elements in the matrix Ω are nonzero reflecting the correlations.

The choice probability under an MLM with joint distribution assumed is denoted as:

  • image(4)

Equation (4) has no closed form solution, and requires approximation by simulation. Halton draws, which offers better coverage of density function and faster convergence, were utilized at 150 draws per iteration in the simulated maximum likelihood estimator (Train 2003).

Partitioning the utility function in Equation (1) into an observable component (Vijt) and an error component according to our specification of MLM yields

  • image(5)

Three components made up the deterministic part of the utility: first, the price scalar (cijt) along with its fixed parameter α; the price coefficient is specified as a fixed coefficient to avoid an unrealistic positive coefficient associated with price (Meijer and Rouwendal 2006; Olsen 2009). Second, the 8×1 vector xjt represents steak attributes with dummy variables. The variables in x correspond to attributes in the choice experiment as described in Table 1. The base cases are USA in origin labeling, approved standards in production practices, none in food-safety assurance, and not specified in tenderness, respectively.

Moreover, the random parameter β is specified to have normal distribution and correlated attributes, the model produced an 8×8 covariance matrix with nonzero off-diagonal elements reflecting the correlation. The last component captures the demographic-interaction effects inline image. The 4 × 1 demographic vector di interacts with the dummy variables CAN and AUS to capture the co-variation between demographic factors and country-of-origin preference.

The Latent Class Model

LCM assumes that individuals are implicitly assigned into Q classes (or segments). LCM choice probability of individual i choosing alternative j in choice situation t given class q is given as:

  • image(6)

As with the MLM, the scalar cijt represents the price and the 8×1 vector xijt represents observed characteristic of alternative j in choice situation t. Instead of just one set of parameters as in a CLM, LCM estimates Q sets of parameters (inline image, with each set describing the collective behavior of individuals found within that particular class. Following Greene and Hensher (2003), the class assignment probability of an individual i to class q in the LCM is given as:

  • image(7)

where zi is a set of observable characteristics of individual i, which are used to identify class memberships. In this application, gender, age, education level, and income level were chosen as the class determinants. The vector λq is the parameter associates with zi to be estimated. Note that only Q-1 sets of λq are produced, the Qth parameter is normalized to be zero for model identification purposes (Greene 2008, Chapter 21). From Hi,q, LCM also estimates the percentage of consumers in the sample belong to each class. LCM utilizes maximum likelihood procedure to produce parameter estimates.

The number of classes optimal to an LCM cannot be determined by a parametric statistical test (Swait 1994). Several information criteria are commonly used to determine the number of classes, they are: the minimum of the Akaike Information Criterion (AIC), the modified Akaike Information Criterion (AIC3), the Bayesian Information Criterion (BIC) and the maximum of the Akaike Likelihood Ratio Index (inline image (Ben-Akiva and Swait 1986; Kamakura and Russell 1989; Gupta and Chintagunta 1994; Swait 1994; Hu et al 2004).

Following Greene and Hensher (2003), the “testing down” approach was adopted where we started from a larger number of classes and gradually reducing to a smaller number of classes. The initial attempts on six or more classes failed computationally due to reaching singular covariance matrices. After comparing the information criteria in Table 5, we chose the five-class model as the final LCM specification as it achieved the best balance of parsimony and explanatory power.

Table 5.  Information criteria used in determining number of classes in the latent class model
Number of classesNumber of parameters (P)Log-likelihoodAIC inline image.AIC3BIC
  1. Notes: The sample size is 14,724 choices from 1,079 individuals (N). The restricted log-likelihood score is −16,175.97. The Akaike Information Criterion (AIC) is calculated as [−2(LL-P)]. The inline image(Akaike Likelihood Ratio Index) is calculated as [1−AIC/2LL (0)]. The AIC3 (Modified Akaike Information Criterion) is calculated as (−2LL + 3P). The Bayesian Information Criterion (BIC) is calculated as [−LL + P/2 inline imageln (N)].

565−10,286.620,703.20.360120,768.210,513.6
451−10,515.321,132.70.346821,183.710,693.4
335−11,101.622,277.20.311422,308.211,223.8
223−11,596.523,238.90.281723,261.911,676.8

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

We tested an array of specifications before the MLM and LCM were finalized and presented in Tables 6 and 7. Comparison of the McFadden R2 and log-likelihood scores reveal that both the MLM and LCM are superior in explanatory power than the CLM. The CLM recorded a McFadden R2 of 0.1535 compared to 0.3437 in the MLM and 0.3641 in the LCM. Thus, we can confidently reject the CLM in favor of the MLM and LCM.

Table 6.  Conditional logit model and mixed logit model parameter estimates
CategoriesAttributesConditional logit modelMixed logit model
Coef.S.E.Coef.S.E.
  1. Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

PricePrice−0.1625***0.0039−0.2567***0.0061
 Would-not-buy −0.8142***0.0577−1.6537***0.1228
Country-of-origin     
 AUS−1.7046***0.2105−3.3101***0.5537
 CAN−1.0031***0.2033−1.9477***0.4471
Food safety     
 BSE  0.9072*** 0.0430  1.4633*** 0.0798
 TRACE  0.9278*** 0.0430  1.5005*** 0.0818
 BSE_TRC  0.6803*** 0.0285  2.0664*** 0.0881
Tenderness     
 TENDER  0.6803*** 0.0285  1.0502*** 0.0502
Production practices     
 NAT0.0225 0.0290 0.0465 0.0489
Interaction terms     
 CAN*MALE  0.1916*** 0.0541  0.3061** 0.1241
 CAN*AGE−0.0139*** 0.0019 −0.0163*** 0.0042
 CAN*EDU  0.0554*** 0.0131  0.0895*** 0.0293
 CAN*INCOME0.0008 0.00090.0012 0.0020
 AUS*MALE  0.2295*** 0.0564  0.4178***0.1523
 AUS*AGE−0.0117*** 0.0019−0.0135***0.0051
 AUS*EDU  0.0659*** 0.0137  0.1262***0.0358
 AUS*INCOME  0.0039*** 0.00090.0029 0.0024
Log likelihood −13608.64  −10616.39 
McFadden R2 0.1535  0.3437  
AIC 27251.30 21338.80 
Table 7.  Latent class model parameter estimates
 Class 1 coef.Class 2 coef.Class 3 coef.Class 4 coef.Class 5 coef.
  1. Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Standard errors presented in parentheses.

PRICE−0.2547***−0.0847***−0.1860***−0.8526***−0.1240***
 (0.0346) (0.0159) (0.0154) (0.0457) (0.0102) 
WOULD-NOT-BUY1.6445***−1.0134***−0.5413***−5.3316***−4.1445***
 (0.4261) (0.2349) (0.1912) (0.3534) (0.2827) 
AUS−1.8228***−4.0782***−0.9112***−0.9252***−0.7591***
 (0.2728) (0.3090) (0.0824) (0.1491) (0.0769) 
CAN−1.6650***−2.8981***−0.5936***−0.6326***−0.5815***
 (0.2469) (0.2351) (0.0751) (0.1322) (0.0712) 
BSE 1.3496*** 0.3425**  1.6655*** 1.4151*** 0.9051***
 (0.3446) (0.1703) (0.1405) (0.1782) (0.0937) 
TRACE  1.4709***0.3105*   1.7450*** 1.5769*** 0.8422***
 (0.3280) (0.1643) (0.1395) (0.1834) (0.0911) 
BSE_TRC 1.6195*** 0.7634*** 2.3832*** 1.8915*** 1.3698***
 (0.3448) (0.1678) (0.1571) (0.1932) (0.0986) 
TENDER 0.8512*** 0.9217***  1.0855*** 0.9058*** 0.6497***
 (0.1866) (0.1210) (0.0712) (0.1168) (0.0616) 
NAT0.0354 0.0573  0.1475**0.0470 0.0341 
 (0.1706) (0.1068) (0.0679) (0.1187)(0.0640) 
Latent Segment Parameter Estimates H(.)   
Constant0.2526−0.0968  −1.0574  −4.1372***
 (0.8788)(0.9745) (0.8663) (0.9078)  
MALE−0.4628** −0.3925*  −0.1105  −0.2138  
 (0.2180) (0.2311) (0.2102)(0.2230)  
AGE 0.0155** 0.0369***  0.0223***  0.0404***
 (0.0074) (0.0082) (0.0074) (0.0088)  
EDU−0.0697  −0.1392** −0.0004  0.1168**
 (0.0589) (0.0601) (0.0516) (0.0521)  
INC−0.0061  −0.0034  0.0003−0.0026  
 (0.0039) (0.0038) (0.0033)(0.0035)  
Class probability     
 0.16  0.17  0.27  0.17  0.24  
Log likelihood−10287    
McFadden R20.3641     
AIC20703.2    

The diagonal values the Cholesky matrix (Table 8) identified the presence of taste heterogeneity within the tested attributes (Hensher et al 2005). These diagonal values revealed that significant taste heterogeneity in all eight coefficients specified as random parameters in the model. Multiple significant values in the off-diagonal elements of the Cholesky matrix suggest that significant correlation exist between the attributes, thus justifying the specification of joint distribution.

Table 8.  Cholesky matrix of correlated random parameters in the mixed logit model
 WOULD-NOT-BUYAUSCANBSETRACEBSE_TRCTENDERNAT
  1. Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

WOULD- NOT-BUY   3.1339***       
AUS−0.3008***  1.9661***      
CAN−0.2088***−1.3803***   0.5084***     
BSE−0.5729*** −0.1775 **   0.4010***  1.0753***    
TRACE−0.6625***  −0.1566*   −0.0711     −1.0322***  0.7268***   
BSE_TRC−0.6375*** −0.1259     0.0732  −1.5877*** 0.2739**0.3978***  
TENDER−0.2524*** −0.0186      −0.4610***−0.2832***−0.0114    −0.2168**  0.5493*** 
NAT−0.2576***−0.1479** −0.2820***−0.3020***−0.2409***0.3853***0.2177** 0.3663***

Given the presence of interaction terms and differences in scales across model, interpretation of individual coefficients is discouraged in MLM and LCM (Greene and Hensher 2003; Scarpa and Del Giudice 2004). Hence, we interpreted the results from both models in the context of the WTP estimates.

Results from the Mixed Logit Model

Consumers’ relative WTP for Australian and Canadian steak were calculated for nine selected consumer profiles based on their age, education, income, and gender. For brevity, we tied education to income as these factors tend to be positively correlated and the shopper's gender is assumed to be female.

The relative WTP follows interpretation of dummy variable, where the base case is the U.S. labeled steak. The WTP is calculated as a negative ratio, where the nominator is the combination of the estimated mean values of the coefficients associated with a particular country (θcountry) and its interaction effect (γ’country×d) and the denominator is the fixed price coefficient (αprince).

  • image(8)

The standard errors of the WTP estimates were produced using Krinsky and Robb (1986) simulation procedure with 2,000 replications (Hensher and Greene 2003). The result is presented in Table 9.

Table 9.  Willingness-to-pay estimations of selected profiles following mixed logit model
 Canadian steakAustralian steak
($/lb)S.E.($/lb)S.E.
  1. Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

Higher income, higher education    
Income =$80k, education = 16 yrs    
 Age = 35.3−3.89***0.53−5.99***0.66
 Age = 45.0−4.51***0.45−6.49***0.56
 Age = 56.62−5.24***0.41−7.10***0.52
Sample average income and education    
Income =$52.37k, education = 14 yrs    
 Age = 35.3−4.71***0.48−7.29***0.60
 Age = 45.0−5.33***0.39−7.80***0.49
 Age = 56.62−6.07***0.36−8.40***0.45
Lower income, lower education    
Income =$30k, education = 12 yrs    
 Age = 35.3−5.51***0.54−8.52***0.67
 Age = 45.0−6.13***0.47−9.03***0.59
 Age = 56.62−6.86***0.45−9.64***0.56

Not surprisingly, the results revealed that on average, imported steak is less preferred by consumers across all education, income, and age levels. The discounts (or negative WTP) calculated at the sample mean level of age (56.62 years), education (14 years), and income ($52.37K) were $6.07/lb and $8.40/lb on average for Canadian steak and Australian steak, respectively, when compared to steak from the United States. These estimates suggest that high-value imported beef are likely to encounter less favorable receptions from U.S. conusmers with the new mandatory COOL rule.

The magnitude of the discount indicated Canadian steak is preferred over Australian steak. We found that older consumers are less willing to pay for imported steak; similar observations of older consumers aversion toward imports were also reported in Alfnes (2004) and Loureiro and Umberger (2007). The magnitude of the discount decreased as education and income level of the shopper increased. For example, the average discount on the Canadian steak was $3.89 for a 35.3-year-old female shopper with household income of $80,000 and 16 years of education. The discount increases 41% to $5.51 for a same-aged female shopper with household income of $30,000 and 12 years of education.

The negative WTP for imported steak suggests that holding other factors constant, most consumers need to be compensated, either in price or in favorable attributes, for choosing Canadian or Australian strip loin steak over U.S. strip loin steak. One such strategy is to incorporate some additional quality features into imported steaks. Table 10 presents the marginal WTP of the non-country-of-origin attributes. The WTP is calculated as the negative ratio between the coefficient of an attribute to the price coefficient. On average, the marginal WTP for BSE-tested beef, traceable beef, or with both attributes combined were $5.70, $5.85, and $8.05, respectively; the WTP for these food-safety enhancements eclipse a large portion of the discount associated with country-of-origin for most consumers. In addition, the tenderness-assured steaks garner a premium of $4.08 on average. Although natural steak was not found to be associated with significant WTP, overall, the food-safety and eating-quality attributes might provide a viable way to differentiate imported steak from domestic products.

Table 10.  Marginal willingness-to-pay estimates from mixed logit model
 Coef. $/lbS.E.95% C.I.
  1. Note: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

WOULD-NOT-BUY−6.44***0.4403−7.30−5.58 
BSE 5.70***0.3306 5.056.35
TRACE 5.85***0.3307 5.206.50
BSE_TRC 8.05***0.3642 7.338.76
TENDER 4.08***0.2068 3.684.49
NAT 0.18  0.1884−0.190.55

Results from the Latent Class Model

The LCM provides a different perspective from the MLM. As noted, the model yielded five unique classes. We found that age, income, and education are significant in determining the latent class an individual belongs to (see Table 7). As with the MLM, coefficient estimates of the attribute variables from the LCM were best interpreted in the context of WTP. The average WTP for an attribute within a consumer class q is the negative ratio between an attribute coefficient in that class qattribute,q) and price coefficient in the same class qprince, q). The standard deviation of the WTP measure was simulated using the Krinsky and Robb (1986) procedure with 2,000 replications.

  • image(9)

As with the MLM, the LCM also showed wide-ranging taste heterogeneity for country-of-origin and other attributes. Of particular interest is the discount needed (or negative WTP) to switch from U.S. steak to imported steak. From Table 11, the discount needed for Australian steak ranged from as little as $1.09/lb to a prohibitive $49.48/lb across different classes, holding other factors constant. Similarly, the discount needed for Canadian steak, across all class membership, ranged from $0.74/lb to $35.12/lb. The higher values of the WTP range suggest that a significant portion of consumers are likely to avoid imported steak.

Table 11.  Willingness-to-pay estimates from the latent class model
 Class 1 Nonconsumer ($/lb)Class 2 Anti-imports consumers ($/lb)Class 3 Food-safety conscious consumers ($/lb)Class 4 Value-seekers ($/lb)Class 5 Strip loin lovers ($/lb)
  1. Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Standard errors presented in parentheses.

Would-not-buy option  6.67***−12.10***−2.91***−6.25***−33.60***
 (2.35)(2.03)(0.89)(0.24)(3.15)
AUS−7.28***−49.48***−4.92*** −1.09*** −6.18***
 (1.45)(9.16)(0.59)(0.17)(0.77)
CAN−6.64***−35.12***−3.22*** −0.74*** −4.74***
 (1.30)(6.50)(0.46)(0.16)(0.68)
BSE  5.33***4.07*  8.95***  1.66***  7.39***
 (1.41)(2.15)(0.86)(0.20)(0.90)
TRACE  5.83***3.65*  9.40***  1.85***  6.86***
 (1.46)(1.94)(0.86)(0.20)(0.87)
BSE_TRC  6.45***  9.17***  12.83***  2.22***  11.17***
 (1.49)(2.56)(1.08)(0.22)(1.11)
Tender  3.41***  11.16***  5.86***  1.06***  5.26***
 (0.87)(2.49)(0.53)(0.14)(0.64)
Natural0.120.70 0.78**0.050.29
 (0.69)(1.33)(0.37)(0.14)(0.52)
Class probability0.160.170.270.170.24

Overall, the marginal WTP estimations for BSE, TRACE, BSE_TRC, and TENDER revealed positive consumer interest in these attributes. Except for consumers in one class, natural beef was generally not regarded as an attractive attribute.

Of the five segments, only consumers in the first segment exibited postive WTP value for the would-not-buy coefficient that captured the utility/disutility yielded from not purchasing the steak. With the positive WTP value, these consumers disliked the strip loin steaks outlined in the choice experiment. These consumers could be vegetarians or did not prefer the particular cut of beef. For this reason, this class of consumers were labeled as non-steak consumers. They accounted for about 16% of the sample. Interestingly, even individuals who generally did not prefer the strip loin steak, if they were to make a choice, they would still choose an U.S. product with almost all other quality guarantees/assurances (except for natural). Estimates of the class membership determinant coefficients in Table 7 indicated that female and older consumers were more likely to be in this class.

The second segment accounted for 17% of the sample. These consumers were labeled as anti-imports consumers for displaying strong aversion toward imported steaks. The estimated discount needed for this group to switch from U.S.-origin steak to Australian and Canadian steak were $49.48/lb and $35.12/lb, respectively. Further, these consumers were found to be willing to pay more for tenderness than for BSE-tested and traceable steak; this implied, they valued eating-quality attributes more than food-safety attributes. The class determinant estimates revealed that female, older, or less educated consumers were more likely to be in this class.

The third group was categorized as food-safety conscious consumers. Even though they displayed moderate aversion toward imported steak, they had the largest WTP for food-safety attributes among all the groups. Interestingly, they were willing to pay a small premium for natural beef, which was insignificant in the CLM, the MLM, and the other classes in the LCM. This group constituted the largest segment, accounting for 27% of the sample. Older consumers were found to be more likely to be in this segment.

We observed the lowest discount on the imported steaks ($1.09 for Australian and $0.74 for Canadian) on value-seekers in segment 4. This segment accounted for 17% of the sample. Individuals in this segment exhibited the lowest WTP for all other attributes examined. Of all the segments, this group was the least likely to be affected by the COOL mandate. The class assignment estimates suggested older and higher educated consumers were more likely to belong to this class.

Consumers in the fifth segment were willing to pay a modest amount to avoid imported steak and for the non-COOL attributes. This group had the largest disutility associated with not buying the steak (−$33.60), as reflected by the negative WTP associated with the would-not-buy. Hence, this group is labeled as strip loin steak lovers. They accounted for 24% of the sample.

From the country-of-origin WTP within the LCM, only the value-seeking consumers in segment 4 appeared to be willing to make the trade-off between domestic and imported steaks with a modest WTP. The remaining 83% of the sample required at least $4.92/lb and $3.22/lb discounts for consuming Australian and Canadian steak. These findings reiterate the possibility of COOL exerting downward pressure on both the price and quantity demand for imported beef.

CONCLUSION AND DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

As a way to gauge the impact of the mandatory COOL provision, this study investigated the extent of consumers’ willingness to trade-off between U.S. labeled steak and imported Canadian and Australian steak. Raging debate on the necessity of COOL and limited understanding of consumers’ reaction to COOL motivated the research question. The notable contribution of this study is the inclusion of source-differentiated beef in the choice experiment, which enable a direct analysis of consumers’ preference and readiness to accept imported beef from the two biggest beef exporters to the United States.

Using MLM and LCM, we learned that imported beef is less preferred than domestic steak largely. Although significant taste heterogeneity exists in consumers’ preference for Australian and Canadian steak, these imported steaks are likely to feature less prominently in the mainstream U.S. market under COOL regime. Nonetheless, imported steak may be sought after by value-seeking customers or as niche products. We also found that import aversion was more prevalent in females, older, and less educated consumers. In addition, we found that consumers are willing to pay a premium for BSE-tested, traceable, and tenderness-assured beef. In particular, the potential for the food-safety attributes are stronger than tenderness assurance for most consumers.

Given the difference in the estimated WTP between domestic and imported beef, as shown in both the MLM and LCM, an immediate consideration is on COOL's ability to generate a premium for domestic beef. While the results provided an argument for such a premium, it is uncertain if such a marked WTP would be observed in a nonhypothetical setting. As much as our choice experiment attempts to simulate the decision process faced by consumers, grocery stores are unlikely to stock a single cut of beef from multiple countries at once. The decision concerning the choice of country-of-origin is likely to be determined upstream in supply chain. Nonetheless, consumers’ preferences are likely to influence those decisions.

In addition, consumers are unlikely to pay the reported large premium for domestic beef for a long period. The WTP estimates calculated in this study may not reflect a sustained premium over a long period because various factors, such as demand and supply elasticities, market power, trade, and other factors may influence WTP in the longer run (Chung et al 2009).

Echoing Brester et al (2004), we expect imported beef to be sold at a discount largely because domestic supply dominates the beef market. Even with COOL, sustaining a long-term price premium would still require producers’ collaboration on producing higher quality beef, maintaining the quality, and restricting supply (Carter et al 2006). For consumers to be willing to pay a premium, especially in repeated purchases, consumers must perceive higher quality for the food products (McCluskey and Loureiro 2003).

Recognizing that Australian beef and Canadian beef enter into the U.S. market in different forms, the outcomes for Australian and Canadian beef are likely to be starkly different. Most Australian beef is imported as boxed beef while Canada ships a large number of cattle to be processed in U.S. meatpacking plants. The final ruling of COOL stated that mixed-origin labels can be applied if imported cattle are commingled with cattle born and raised in the United States during the production day (USDA 2009). Given the less stringent requirement on mixed origin labels, most meat cuts derived from cattle of Canadian or Mexican origin are likely to be labeled as such. Logistic constraints prevent Australian live cattle from being imported and processed in the United States, therefore high-value Australian beef cuts are likely to be sold in niche markets where its grass-fed feature is sought after (Umberger et al 2002). Nonetheless, we expect mixed-origin labeled meat products to be more prevalent as the result of COOL, suggesting a worthwhile investigation for future research.

Footnotes
  • 1

    Loureiro and Umberger (2007) attempted mixed logit but found the model failed to detect significant unobserved heterogeneity.

  • 2

    The respondents were not limited to only meat consumers.

  • 3

    This analysis was generated using SAS™ software Version 9.2 for Microsoft Windows © 2010 SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA.

  • 4

    Previous choice experiments assigned a variety numbers of choice sets to each individual. Hu et al. (2005) asked each respondent to complete eight choice sets while Tonsor et al. (2009) assigned 21 choice scenarios to each respondent. Although there has been discussion in the literature on the impact of scenario complexity on choices, this is not the focus of this research. A total of 10–14 choice sets per person are in line with the past literature.

  • 5

    The question used in the survey was “How often do you buy beef? Is it (never, occasionally, or regularly)?”

  • 6

    The question used in the survey was “Please provide the approximate percentage of your beef consumption over the past year that would include the following beef products (ground or minced, roasts, steaks, sausage, organ meats, or other).”

  • 7

    The checkbox question used in the survey was “Do you prefer imported beef from New Zealand, Australia, Canada or other? (one answer only).” The options were “Imported beef from Australia, Imported beef from New Zealand, Imported beef from Canada, Imported beef from … (please identify), I avoid imported beef as much as possible, and I neither like nor dislike imported beef.”

  • 8

    The question used in the survey was “Whether you have ever knowingly purchased beef produced in another country or not, what is your perception of the level of food safety of beef by country-of-origin?”

ACKNOWLEDGMENTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

Funding supports from the Consumer and Marketing Demand Network and the Alberta Institute for American Studies are greatly appreciated. This research was supported by the University of Kentucky Agricultural Experiment Station and is published by permission of the Director as station number 13-04-033. We also acknowledge helpful comments from Michael Reed, the editor and the anonymous reviewer, in addition to editing assistance from Cindy Wiggers. Any errors are the responsibility of the authors.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix
  • Adamowicz, W., P. Boxall, M. Williams and J. Louviere. 1998. Stated preference approaches for measuring passive use values: Choice experiments and contingent valuation. American Journal of Agricultural Economics 80 (1): 6475.
  • Alfnes, F. 2004. Stated preferences for imported and hormone-treated beef: Application of a mixed logit model. European Review of Agricultural Economics 31 (1): 19–37.
  • Alfnes, F. and K. Rickertsen. 2003. European consumers’ willingness to pay for U.S. beef in experimental auction markets. American Journal of Agricultural Economics 85 (2): 396405.
    Direct Link:
  • Aubeeluck, A. 2010. A comparative analysis of consumer attitudes towards food safety, animal testing and traceability in the meat industry: Japan and Canada. M.S. thesis, University of Alberta.
  • Bailey, D. V., J. Robb and L. Checketts. 2005. Perspectives on traceability and BSE testing in the US beef industry. Choices 20 (4): 29397.
  • Becker, T. 2000. Consumer perception of fresh meat quality: A framework for analysis. British Food Journal 102 (3): 15876.
  • Ben-Akiva, M. and J. Swait. 1986. The Akaike likelihood ratio index. Transportation Science 20 (2): 13336.
  • Brester, G. W., J. M. Marsh and J. A. Atwood. 2004. Distributional impacts of country-of-origin labeling in the U.S. meat industry. Journal of Agricultural and Resource Economics 29 (2): 20627.
  • Carter, C., B. Krissoff and A. P. Zwane. 2006. Can country-of-origin labeling succeed as a marketing tool for produce? Lessons from three case studies. Canadian Journal of Agricultural Economics 54 (4): 51330.
  • Caswell, J. A. 1998. How labeling of safety and process attributes affects markets for food. Agricultural and Resource Economics Review 27 (2): 15158.
  • Chung, C., T. Zhang and D. S. Peel. 2009. Effects of country of origin labeling in the U.S. meat industry with imperfectly competitive processors. Agricultural and Resource Economics Review 38 (3): 40617.
  • Darby, M. R. and E. Karni. 1973. Free competition and the optimal amount of fraud. Journal of Law and Economics 16 (1): 6788.
  • Davidson, A., M. J. A. Schröder and J. A. Bower. 2003. The importance of origin as a quality attribute for beef: Results from a Scottish consumer survey. International Journal of Consumer Studies 27 (2): 9198.
  • Dickinson, D. L. and D. Bailey. 2002. Meat traceability: Are U.S. consumers willing to pay for it? Journal of Agricultural and Resource Economics 27 (2): 34864.
  • Dillman, D. A. 2007. Mail and Internet Surveys: The Tailored Design Method . New York : John Wiley & Sons Inc.
  • Greene, W. H. 2008. Econometric Analysis . New Jersey : Prentice Hall.
  • Greene, W. H. and D. A. Hensher. 2003. A latent class model for discrete choice analysis: contrasts with mixed logit. Transportation Research Part B: Methodological 37 (8): 68198.
  • Grunert, K. G. 2005. Food quality and safety: Consumer perception and demand. European Review of Agricultural Economics 32 (3): 36991.
  • Gupta, S. and P. K. Chintagunta. 1994. On using demographic variables to determine segment membership in logit mixture models. Journal of Marketing Research 31 (1): 12836.
  • Hensher, D. A. and W. H. Greene. 2003. The mixed logit model: The state of practice. Transportation 30 (2): 13376.
  • Hensher, D. A., J. M. Rose and W. H. Greene. 2005. Applied Choice Analysis: A Primer . New York : Cambridge University Press.
  • Hoffmann, R. 2000. Country of origin–a consumer perception perspective of fresh meat. British Food Journal 102 (3): 21129.
  • Hu, W., A. Hünnemeyer, M. Veeman, W. Adamowicz and L. Srivastava. 2004. Trading off health, environmental and genetic modification attributes in food. European Review of Agricultural Economics 31 (3): 389408.
  • Hu, W., M. Veeman and W. Adamowicz. 2005. Labelling genetically modified food: Heterogeneous consumer preferences and the value of information. Canadian Journal of Agricultural Economics 53 (1): 83102.
  • Ikenson, D. J. 2004. Uncool rules: Second thoughts on mandatory Country of Origin Labeling. The Cato Institute , http://www.cato.org/pub_display.php?pub_id=10681 (accessed July 7, 2010).
  • Kamakura, W. A. and G. J. Russell. 1989. A probabilistic choice model for market segmentation and elasticity structure. Journal of Marketing Research 26 (4): 37990.
  • Krinsky, I. and A. L. Robb. 1986. On approximating the statistical properties of elasticities. The Review of Economics and Statistics 68 (4): 71519.
  • Krissoff, B., F. Kuchler, K. Nelson, J. Perry and A. Somwaru. 2004. Country-of-origin labeling: Theory and observation. USDA, ERS, WRS-04-02 .
  • List, J. A., P. Sinha and M. H. Taylor. 2006. Using choice experiments to value non-market goods and services: Evidence from field experiments. The BE Journal of Economic Analysis & Policy 6 (2): 137.
  • Loomis, J. 2011. What's to know about hypothetical bias in stated preference valuation studies? Journal of Economic Surveys 25 (2): 36370.
  • Loureiro, M. L. and W. J. Umberger. 2003. Estimating consumer willingness to pay for country-of-origin labeling. Journal of Agricultural and Resource Economics 28 (2): 287301.
  • Loureiro, M. L. and W. J. Umberger. 2005. Assessing consumer preferences for country-of-origin labeling. Journal of Agricultural and Applied Economics 37 (1): 4963.
  • Loureiro, M. L. and W. J. Umberger. 2007. A choice experiment model for beef: What US consumer responses tell us about relative preferences for food safety, country-of-origin labeling and traceability. Food Policy 32 (4): 496514.
  • Lusk, J., J. Roosen and J. Fox. 2003. Demand for beef from cattle administered growth hormones or fed genetically modified corn: A comparison of consumers in France, Germany, the United Kingdom, and the United States. American Journal of Agricultural Economics 85 (1): 1629.
    Direct Link:
  • Lusk, J. L., J. Brown, T. Mark, I. Proseku, R. Thompson and J. Welsh. 2006. Consumer behavior, public policy, and country-of-origin labeling. Applied Economic Perspectives and Policy 28 (2): 28492.
  • Lusk, J. L. and T. C. Schroeder. 2004. Are choice experiments incentive compatible? A test with quality differentiated beef steaks. American Journal of Agricultural Economics 86 (2): 46782.
    Direct Link:
  • Marette, S., R. Clemens and B. Babcock. 2008. Recent international and regulatory decisions about geographical indications. Agribusiness 24 (4): 45372.
  • Maynard, L. and X. Wang. 2010. Context-dependent BSE impacts on Canadian fresh beef purchases. Journal of International Food & Agribusiness Marketing 23 (1): 3255.
  • McCluskey, J. J. and M. L. Loureiro. 2003. Consumer preferences and willingness to pay for food labeling: A discussion of empirical studies. Journal of Food Distribution Research 34 (3): 95102.
  • McCluskey, J. J. and J. F. M. Swinnen. 2004. Political economy of the media and consumer perceptions of biotechnology. American Journal of Agricultural Economics 86 (5): 123037.
  • McFadden, D. 1974. Conditional logit analysis of qualitative choice behavior. Frontiers in Econometrics 8: 10542.
  • Meijer, E. and J. Rouwendal. 2006. Measuring welfare effects in models with random coefficients. Journal of Applied Econometrics 21 (2): 22744.
  • Mørkbak, M. R., T. Christensen and D. Gyrd-Hansen. 2010. Consumer preferences for safety characteristics in pork. British Food Journal 112 (7): 77591.
  • Mutondo, J. E. and S. R. Henneberry. 2007. A source-differentiated analysis of U.S. meat demand. Journal of Agricultural and Resource Economics 32 (3): 51533.
  • Northen, J. R. 2000. Quality attributes and quality cues Effective communication in the UK meat supply chain. British Food Journal 102 (3): 23045.
  • Olsen, S. B. 2009. Choosing between internet and mail survey modes for choice experiment surveys considering non-market goods. Environmental and Resource Economics 44 (4): 591610.
  • Scarpa, R. and T. Del Giudice. 2004. Market segmentation via mixed logit: Extra-virgin olive oil in urban Italy. Journal of Agricultural & Food Industrial Organization 2 (1): 118.
  • Schroeder, T. C., G. T. Tonsor, J. M. E. Pennings and J. Mintert. 2007. Consumer food safety risk perceptions and attitudes: Impacts on beef consumption across countries. The BE Journal of Economic Analysis & Policy 7 (1): 127.
  • Schupp, A. and J. Gillespie. 2001. Consumer attitudes toward potential country-of-origin labeling of fresh or frozen beef. Journal of Food Distribution Research 32 (3): 3444.
  • Sitz, B. M., C. R. Calkins, D. M. Feuz, W. J. Umberger and K. M. Eskridge. 2005. Consumer sensory acceptance and value of domestic, Canadian, and Australian grass-fed beef steaks. Journal of Animal Science 83 (12): 286368.
  • Steiner, B., F. Gao, and J. Unterschultz. 2010. Alberta consumers’ valuation of extrinsic and intrinsic red meat attributes: A choice experimental approach. Canadian Journal of Agricultural Economics 58 (2): 17189.
  • Suppan, S. 2009. The Canadian and Mexican WTO challenges to U.S. Country of Origin Labeling. Institute for Agriculture and Trade Policy , http://www.tradeobservatory.org/library.cfm?refID=105552 (accessed August 10, 2010).
  • Swait, J. 1994. A structural equation model of latent segmentation and product choice for cross-sectional revealed preference choice data*1. Journal of Retailing and Consumer Services 1 (2): 7789.
  • Thilmany, D. D., W. J. Umberger and A. R. Ziehl. 2006. Strategic market planning for value-added natural beef products: A cluster analysis of Colorado consumers. Renewable Agriculture and Food Systems 21 (3): 192203.
  • Tonsor, G., T. Schroeder, J. Pennings and J. Mintert. 2009. Consumer valuation of beef steak food safety and quality assurances in Canada, Japan, Mexico, and the United States. Canadian Journal of Agricultural Economics 57: 395416.
  • Train, K. 2003. Discrete Choice Methods with Simulation . New York : Cambridge University Press.
  • Umberger, W. J., D. M. Feuz, C. R. Calkins and K. Killinger-Mann. 2002. US consumer preference and willingness–to–pay for domestic corn–fed beef versus international grass–fed beef measured through an experimental auction. Agribusiness 18 (4): 491504.
  • USDA, Agricultural Marketing Service . 2009. Mandatory country of origin labeling of beef, pork, lamb, chicken, goat meat, wild and farm-raised fish and shellfish, perishable agricultural commodities, peanuts, pecans, ginseng, and macadamia nuts; final rule. Federal Register 74 (10), http://www.ams.usda.gov/AMSv1.0/getfile?dDocName=STELPRDC5074925 (accessed August 9, 2010).
  • USDA, Foreign Agricultural Service . 2010. Production, supply and distribution online. http://www.fas.usda.gov/psdonline/psdQuery.aspx (accessed August 9, 2010).
  • Verbeke, W. 2008. Impact of communication on consumers’ food choices. Proceedings of the Nutrition Society 67 (3): 28188.
  • Verbeke, W. and J. Roosen. 2009. Market differentiation potential of country-of-origin, quality and traceability labeling. Estey Centre Journal of International Law and Trade Policy 10 (1): 2035.
  • Verlegh, P. W. J. and J. B. E. M. Steenkamp. 1999. A review and meta-analysis of country-of-origin research. Journal of Economic Psychology 20 (5): 52146.
  • Wyld, A. 2009. Canada launches complaint over U.S. meat labelling law. The Canadian Press , http://www.cbc.ca/money/story/2009/04/27/meat-label-wto-canada-0427.html (accessed April 27, 2009).
  • Yang, J. and E. Goddard. 2011. Canadian consumer responses to BSE with heterogeneous risk perceptions and risk attitudes. Canadian Journal of Agricultural Economics 59 (4): 493518.

Appendix

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. BACKGROUND
  5. PREVIOUS RESEARCH AND OBJECTIVES OF THIS STUDY
  6. RESEARCH DESIGN
  7. SUMMARY STATISTICS
  8. PREFERENCE AND PERCEPTIONS OF COUNTRY-OF-ORIGIN FOR BEEF
  9. RESEARCH METHOD: MIXED LOGIT, LATENT CLASS LOGIT, AND WTP ANALYSIS
  10. RESULTS
  11. CONCLUSION AND DISCUSSION
  12. ACKNOWLEDGMENTS
  13. REFERENCES
  14. Appendix

APPENDIX: INFORMATION SHEET AND A SAMPLE CHOICE SET

Stated Preference

In this final section of this survey, you are provided with 14 different pairs of alternative strip loin beefsteaks (also known as Kansas City strip and New York steak) that could be available for purchase in the retail grocery store or butcher where you typically shop that possess differing attributes. Steak prices vary from US $5.50/lb. to $16.00/lb. For each pair of steaks, please select the steak that you would purchase, or neither, if you would not purchase either steak. It is important that you make your selections like you would if you were actually facing these choices in your retail purchase decisions.

For your information in interpreting alternative steaks:

Country-of-origin refers to the country in which the cow/animal was raised and includes USA, Canada, and Australia.

Production practice is the method used to produce the cow/animal where:

Approved standards means the cow/animal was raised using scientifically determined safe and government-approved use of synthetic growth hormones and antibiotics (typical of cattle production methods used in Canada and USA)

Natural is the same as typical except the cow/animal was raised without the use of synthetic growth hormones or antibiotics

Tenderness refers to how tender the steak is to eat and includes

Assured tender means the steak is guaranteed tender by testing the steak using a tenderness-measuring instrument

Uncertain means there are no guarantees on tenderness level of the steak and the chances of being tender are the same as typical steaks you have purchased in the past

Food safety assurance refers the level of food safety assurance with the steak

None food safety means the steak meets current minimum government standards for food safety

Traceable means the product is traceable back to farm of origin from your point of purchase

BSE tested means that all animals are tested for BSE prior to meat being sold at your point of purchase

Table Choice set . 
Steak attributeABC
Price ($/lb.)$16.00$12.50I would not purchase any of these products
Country of originAustraliaCanada 
Production practiceNaturalApproved standards 
TendernessAssured tenderAssured tender 
Food safety assuranceTraceableBSE tested and traceable 
I would choose …