BRAND PREFERENCES AMONG MEATS: AN APPLICATION OF PROBIT MODELS By OSCAR FERRARA A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005
Copyright 2005 by Oscar Ferrara
This document is dedicated to my parents, Oscar y Susana Ferrara.
iv ACKNOWLEDGMENTS I would like to express my gratitude to Dr. Ronald W. Ward, chairman of the supervisory committee, for his academic a nd personal support throughout this project. His valuable guidance and patience are deeply appreciated. I would also like to address very special thanks to Dr. Lisa House and Dr. Robert Degner for their generous assistance. I am also thankful to the faculty, st aff, and fellow students of the Food and Resource Economics Department of the Universi ty of Florida for maki ng me feel at home in Gainesville. Finally, my deepest gratitude goes to my wife Jaquelina and my children Sofia, Nicolas, and Marco for their understanding a nd constant encouragement throughout these years.
v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... ..x CHAPTER 1 INTRODUCTION........................................................................................................1 Problem Statement......................................................................................................10 Research Objectives....................................................................................................11 Research Hypotheses..................................................................................................11 Research Methodology...............................................................................................11 Data Description and Scope of the Thesis..................................................................12 Outline of Chapters.....................................................................................................13 2 LITERATURE REVIEW...........................................................................................15 Brand Preferences.......................................................................................................15 The U.S. Meat Industry..............................................................................................19 Examples of Probit Models........................................................................................21 3 INDUSTRY TRENDS AND DATA..........................................................................26 Major Industry Trends................................................................................................26 Brand Preferences.......................................................................................................31 Description and Definition of Demographic Variables..............................................33 4 BRAND PREFERENCE THEORY AND MODEL SPECIFICATION....................39 Brand Theory..............................................................................................................39 Conceptual Brand Preference Models........................................................................42 Meat Brand Model Specification................................................................................46 5 BRAND PREFERENCE PROBIT ESTIMATES......................................................51
vi Probit Brand Preference Models by Meat Type.........................................................53 Pooled Probit Brand Preference Model......................................................................57 6 BRAND PREFERENCES SIMULATIONS..............................................................60 Concept of Probabilities and Distribution..................................................................60 Probability of Brand Identif ication by Demographics................................................62 Probability of Brand Identific ation by Non-Demographics.......................................66 Structural Change in the Brand identity.....................................................................72 Ranking of Variables..................................................................................................74 Probabilities of Brand Prefer ence: Upper and Lower Limits.....................................77 7 SUMMARY AND CONCLUSIONS.........................................................................82 APPENDIX TSP PROGRAM: PROB IT MODELS AND SIMULATIONS.................87 LIST OF REFERENCES.................................................................................................105 BIOGRAPHICAL SKETCH...........................................................................................109
vii LIST OF TABLES Table page 3-1 Demographic variables from the NPD data set........................................................35 3-2 Distribution of States for each region......................................................................37 3-3 Correlation among demographic variables..............................................................38 5-1 Probit model estimates by meat categories..............................................................54 5-2 Pooled probit model estimates for brand preferences..............................................58 6-1 Distribution of meat purchas es by outlet (share by outlet)......................................66 6-2 Distribution of meat purchases by lo cation in the store (share by section)..............68
viii LIST OF FIGURES Figure page 1-1 Market value of U.S. meat products . Source: Datamonitor USA. March 2004.........2 1-2 Market volume of U.S. meat produc ts. Source: Datamonitor USA. March 2004.....2 1-3 Market Segmentation. Percenta ge share by volume in 2003. Source: Datamonitor USA. March 2004.................................................................................3 1-4 Consumption of meat products. Percen tage change from 1979 to 2002. Source: Livestock Marketing Information Center...................................................................4 3-1 Distribution of average consumption of meat products across time. Source: NPD panel data and the Livestock Marketing Information Center...................................27 3-2 Distribution of average price of meat products across time. Fish prices are from 1992 to 2002. Source: Livestock Marketing Information Center and the NPD panel data..................................................................................................................28 3-3 Variation on prices across time. Inde xed to years 1992. Source: NPD panel data..29 3-4 Distribution of total expenditures per capita of meat produc ts across the time. Fish expenditures from 1992 to 2002. Sour ce: Livestock Marketing Information Center and the NPD panel data................................................................................30 3-5 Expenditure shares on meat product s per capita in y ear 2002. Source: NPD panel data..................................................................................................................31 3-6 Change in brand identification for all meats across time. Source: NPD panel data........................................................................................................................... 33 6-1 Normal distribution and pr obability of brand purchase...........................................62 6-2 Change in probabilities of purchasing by brands according to country regions......63 6-3 Change in probabilities of purchasin g by brands according to occupation..............64 6-4 Change in probabilities of purchasin g by brands according to head of the household age...........................................................................................................65 6-5 Change in probabilities of purchasin g by brands accordi ng to store type................67
ix 6-6 Change in probabilities of purchasing by brands according to location within the store.......................................................................................................................... 69 6-7 Probabilities of purchasing by brands w ith prices indexed to 1.0 for the mean......70 6-8 Change in probabilities of purchas ing by brands according to seasons...................71 6-9 Brand identification across the time. Source: NPD panel data................................73 6-10 Change in brand identification for all meats across time. Source: NPD panel data........................................................................................................................... 74 6-11 Ranking of the Probability of buying branded products by meat category..............77 6-12 Upper and lower limits to the probabili ties of brand identity when purchasing each meatÂ…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…Â…..............................79 7-1 Ranking of variables aff ecting brand preferences....................................................85
x Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science BRAND PREFERENCES AMONG MEATS: AN APPLICATION OF PROBIT MODELS By Oscar Ferrara August 2005 Chair: Ronald W. Ward Major Department: Food and Resource Economics Brand identification among meats products has increased over the past years, revealing substantial differences in the br and structure across the competing products. Changes in market demand and consumersÂ’ preferences have encouraged the meat industry to differentiate its products, review production processes, reorganize its value chain, and use a mix of price and non-price marketing strate gies in order to increase market share. The objective of this research is to statistically estimate the impact of a series of demographic and non-demographic variables exp ected to affect the selection process and measure what variables most likely infl uence brand preferences among beef, fish, poultry, and pork products. Based on national and regional demographics, consumer surveys with information recorded across households over time were collected and processed by the National Panel Diar y Group Company (NPD) from 1992 to 2001, containing 775,976 observations. Each respondent identified if he or she made a meat
xi purchase based on brand preferences, thus giving a binary cla ssification for each purchase. Probit models are specified and es timated to reveal the likelihood of brand recognition for each type of meat category an d then the four meats are pooled into a single model to reveal the estimates fo r the entire meat industry. This methodology allows for comparison and ranking of factors positively or negatively affecting the purchasing process and determines if a respons e to a variable is unique to a particular meat or common to the four. Estimates presented considerable differen ces across the four meat categories and revealed that for the average consumer the likelihood of buying branded beef products is 28 percent; branded fish, 41 pe rcent; branded pork, 52 perc ent; and branded poultry is estimated to be 80 percent. These percen tages represent a year ly increase of 1.5 percentage points in the case of beef products and almost 1 percentage unit for pork products. Fish products showed very small increase and poultry pr oducts represent a decrease of almost 1 percentage po int over the time of this research. For all meat categories, results showed that brand preferences are most likely to be affected by type of store, location within th e store, region of the country, occupation, the household headÂ’s age, and time in the case of beef. Surprisingly, the rest of the demographic variables presented minor effect s on brand preferences in each of the four meat categories that are the subject of this research. The resulting ranking of variables reveals that changes in brands preferences have important connotations for market power and the underlying structure of the industry and will help industry analysts to develop different marketing strategies.
1 CHAPTER 1 INTRODUCTION Economic theory shows that increased indus trialization and specia lization within an economic sector lead to expanded consumer demand for high quality food products and marketing services. In the case of the U.S. meat industry, this trend has led to a wide range of improvements in quality, taste, c onvenience, and consistency. The result is a more efficient industry, but one with fewe r players that are able to satisfy the requirements of consumers. Today, the meat industry in the United St ates is becoming more concentrated. There are fewer small farms, more producer/processor alliances, and meat processors have joined food retailers, re sulting in fewer and larger businesses. This process of reorganization or consolidation began more than a half century ago in the meat processing and livestock industry. Its developmen t was driven by the need to maintain or gain a competitive edge and fulfill the expectations of the market. Two powerful economic forces drove this transfor mation: food demand and technology. Beyond concern about market power, the m eat industry impor tance as a lead enterprise in U.S. agriculture and key ec onomic engine in many rural areas is not debatable. According to the consulting co mpany Datamonitor USA, during the years 1999 to 2003, the meat production market, whic h comprises all sales of beef, poultry, pork, or fish products, has experienced st eady growth, representing a compound annual growth rate of 3.2 percent (Figure 1-1) w ith a production peak at $96.2 billion in 2002.
2 7.0 5.1 5.3 -4.2 19992000200120022003 Years 0 10 20 30 40 50 60 70 80 90 100 Market value ($ billions) 0.0 5.0 10.0 -5.0 Annual growth rate (%) All meats $ Billions% Growth $ Billions81.286.991.496.292.2 % Growth7.05.15.3-4.2 Figure 1-1: Market value of U.S. meat products. Source: Datamonitor USA. March 2004 In terms of market volume, the U.S. mark et grew from 24.9 million metric tons in 1999 to 26.4 million metric tons in 2003, repr esenting a compound annual growth rate of 1.4 percent (Figure 1-2). 0.7 -0.9 3.6 2.4 19992000200120022003Years0 5 10 15 20 25 30Market volume (metric tons million)0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 -0.5 -1.0 -1.5 -2.0Annual growth rate (%) All meatsMtr. tons% Growth Mtr. tons24.925.124.925.726.4 % Growth0.7-0.93.62.4 Figure 1-2: Market volume of U.S. meat products. Source: Datamonitor USA. March 2004 With respect to market segmentation of U.S. meat products, the poultry sector accounts for 34 percent of the market share by volume, followed by the beef sector at
3 32.7 percent, the pork sector at 24.9 percent, and the fish sector at 8.0 percent of the market share (Figure 1-3). Beef 32.4% Fish 8.1% Poultry 34.4% Pork 25.2% Figure 1-3: Market Segmentation. Per centage share by volume in 2003. Source: Datamonitor USA. March 2004 In general, the population is consumi ng more poultry and fish products and consuming less red meat, such as beef and pork. A recent report from the United States Department of Agriculture Economic Research Services clearly shows this trend: the total per capita consumption of meat has increased 11.9 percent from 1979 to 2002 while the consumption of beef and pork has decr eased 14.9 percent and 2.2 percent respectively during the same period. Poultry and fish consumption, however, have increased 72.7 percent and 25.3 percent since 1979, showing a solid trend which supports previous data presented in this chapter (Figure 1-4). The remarkable changes taking place in the demand for meat products over the past decad es have deeply affected the U.S. meat industry and encouraged all it s participants to review their production processes, reorganize their value chain, and use a mix of price and non price marketing strategies.
4 19792002% ChangeYears0 25 50 75 100 125 150 175 200 225 -25Consumption (Lbs. per Capita)Consum p tion Chan g e Rate (% ) All meatsBeefFishPoultryPorkTotal Beef7867-15 Fish131625 Poultry478173 Pork6463-2 Total 202.2226.311.9 Figure 1-4: Consumption of meat products . Percentage change from 1979 to 2002. Source: Livestock Marketing In formation Center. March 2004 It is clear that changes in relative prices, increases in health concerns, changes in demographics characteristics, lifestyles, and preferences, as well as technological improvements in quality and product pa cking, product preparat ion, storage, and distribution account for a large portion of this variability. Consequently, the meat industry has experienced a slow transition from a traditional commodity-selling perspective to a more contemporary marke ting approach in order to address those changes in market demand and consumersÂ’ preferences. Competition has become much more inte nse, triggering a permanent innovation inside the meat industry and transformi ng a non segmented industry of Â“cooperative goodsÂ” with little or no branding history into a more se gmented industry (e.g., poultry) with more brand advertising is used to influence market shares and where an increasing volume of information about product attri butes is demanded by consumers. Producers
5 and retailers are differentiating and increasing the value of th eir products (live animals or meat products) in order to build demand a nd increase financial profits for the firm. Advertising and promotions, which ofte n highlight production and marketing practices, are examples of the methods that companies and indi viduals have utilized to set their product apart in order to meet the needs and desires of specific consumer segments (Allen and Pierson, 1993). The fresh meat ma rket is a good example of this practice where some producers have learned to differe ntiate their products by marketing them base on their particular attr ibutes (e.g., Coleman Natural Beef, Harris Ranch Beef, etc). Purchase decisions are based on predicti ons of product performance. Consumers base their predictions in part on product cues an d are accurate to the extent that they have properly learned the relationship between the cues and performance. If consumers learn the relationship between produc t attributes and quality, th ey will differentiate among brands that possess different attributes and tr eat as commodities those brands that share the same attributes. Once the predictive rule is learned, it may be applied to any new brand that possesses the attribut es. In contrast, consumers who rely strictly on brand cues may ignore the underlying attributes and may in correctly differentiate physically identical brands (Van Osselaer and Alba, 2000). Advertising and promotion provide this information facilitating purchasing decisions, and in some cases they even ch ange the underlying preference function for a particular product or service. Depending on the characteristics of the product, potential buyers, and the scope and beneficiaries of th e marketing campaign, there are two types of advertising and promot ion of commodities: generic advertising, a cooperative effort among producers designed to collectively incr ease the primary demand (i.e., the size of
6 the pie) of a product by advertising the attr ibutes and characteri stics of the product, without influencing the market share of any producer (e.g., Beef Check off program, Florida citrus, Washi ngton apples); and brand advertising , is designed and funded by a specific firm with the intent of benefiting that firmÂ’s demand by differentiating the product from other suppliers in order to increas e the market share of the brand within the same industry (Ward, 1997). Commodities in general, and in our cas e meat products can be differentiated according to how consumers acquire informa tion about them and how this information affects consumersÂ’ perceptions towards the productsÂ’ performance. For search goods consumers have an active participation search ing out the product attributes which can be observed at the time of purchase (e.g., pr ice, color, size ). Alternatively, experienced goods are those where the attributes can be observed at the time of consumption (e.g., taste, tenderness, fat content) and influence c onsumersÂ’ satisfaction with a particular meat product. Finally, credence goods include attributes that are not explicitly observed in the product (e.g., safety: antibioti c and pesticide residues; h ealth: choleste rol levels; environmental: Â“greenÂ” production techniques/c onditions) but consumer credibility of the claims made about the product is influe nced by a high level of confidence on the effectiveness of the systems in charge of creating and monito ring the product (Codron, Sterns and Reardon, 2003.). Previous research has shown that the most important characterist ics of food are that it taste good and be guaranteed safe to eat, while the most important characteristics of meat products were that it look fresh, not have a lot of waste, be certified as USDA inspected, and be free of chemical residue s. New marketing programs should reinforce
7 these attributes, emphasizing the characteri stics of the product, adding value, and improving presentation in order to build a positive perception about branded products. Brands depict some level of differentia tion within a product category with the differentiation being achieved through both real and perceived attri butes. For some food goods, brands dominate the product category simp ly because one or more brands capture the entire market. Bananas are a good exam ple. Products requiring considerable value added through processing and/or packaging may need brand recognition in order to recoup the cost associated with the value-ad ded process. Even with brand selections, consumers often can judge the attributes th rough experimentation especially when the risk of trying the branded product is low. Orange juice is probably one of the best examples of a commodity with major br ands where experimentation is easy. Brand differentiation exists while brands ar e highly substitutable. In such cases, efforts to keep the brand recognition have to be extended considerably beyond brand loyalty, requiring continued reinforcement through advertising and promotions. Meats, defined to include beef, fi sh, pork and poultry, are unique among the food categories in that some of the meats are highly branded an d other much less so. Among these, beef has the least brand identification. Mu ch of the difference in meat brands can be tied to the level of value added such as the frozen f oods. The product has a longer shelf life and the packaging lends itself to brand identification. In contrast to many food categories, meats have experienced a high level of visibility with respect to food safety issues. Press coverage about Bovine Spongiform Encepha lopathy (BSE) also called Â“mad cow,Â” salmonella, or E-coli reveals potentially ma jor food safety problems associated with meats. If safety is a concern, experimenta tion is not an option and consumers look for
8 assurance about the safety. Quality assura nce through industry seals and government inspection may solve some of the problem, yet c onsumers may turn to brands if there is a history of food safety with a particular brand even with in a historically nonbranded category. Even with safe products, some attributes cannot be judged through experience. These credence attributes must be identifie d through means other th an consumption, and brands may capture shares of a market thr ough a consistent message about one or more credence attributes. For example, brands em phasizing low fat and/or U.S. produced beef may capture some brand loyalty even with the generally homogene ous nature of the product group. Also, the risk of a bad decision may be higher. Thus, brands within a generally common product category may evol ve. ConsumersÂ’ desire for convenience, variety, safety, quality, and consistency all proba bly contribute to some growth in brands within the meat group. How important are brands within the meat categories and what drives the brand preference? That issue, within the historically homogene ous sub-categories of beef, pork, poultry, and fish, is the focus of this thesis . Are consumers likely to buy branded or non branded meats and, if so, what are the majo r factors impacting th e brand preferences? Some of these industries have major programs in place to discuss the common a ttributes within the meat categories (i.e., the beef and pork prom otion checkoffs). Why, then, would one expect to see some branded growth ? Again, these preferences are the focus of this research and the answer has significant importance to the future structure of many aspects of each industry.
9 Branding of meat products may imply highe r product differentia tion and represents an important transition from the generic id ea of selling meat as a commodity to a campaign involving market segmentation and target marketing of specific consumer segments. According to Phillip Kotler, market segmentation may be described as an assumption that all consumers are unique an d the needs of individuals may not be satisfied with a mass marketing approach. Sim ilarly, target marketing may be defined as a market segment profiled by its dem ographic, economic, and psychographic characteristics so that marketing opport unities may be evaluated. The precise identification of a productÂ’s f unctional benefits and attributes , as well as the demographic and non demographic variables influencing consumer selection and decision-making process, would be useful in developing new marketing and merchandising strategies. This information will help the meat industry fulfill consumer needs and generate a much larger set of alternatives. The importance of brands w ithin the four meat types (i.e., beef, fish, pork, and poultry) differs considerably and likely has evolved over time. Consumer buying decisions based on brand identification have major implications for an industry and firmÂ’s marketing strategies. Brands may se gment the market and may or may not grow total demand for the product ca tegory. Brands may fail to pr ovide consumers with insight into the common important attributes to existing and potential new consumers. Yet brands may contribute to higher value products and consistency, as well as better product recognition. At this point, the major goal of this research is to determine the extent of brand buying within the four meat categories; then this research will de termine how much of
10 the brand buying is attributed to selected demographics, seasona l patterns, and other factors. The objective of the following chapters is to describe th e industryÂ’s background and develop an econometric model to analyze a nd simulate the resulting data in order to understand the driven factors behind brand identification within the meat industry. Problem Statement Changes in consumer demand and advan ces in technology have prompted the necessity to differentiate and add value to existing products. This research will address issues related to consumersÂ’ brand recogniti on and their effects on the meat purchasing process. The overall objective of this study is to describe, differentiate, and empirically measure the impact of the factors affecti ng consumer perceptions towards branded and non branded meat products, based on data collected during September 1992 to August 2000. Brand recognition is determined largel y by product attributes, the perceived performance of the brand on these attributes, and the importance that consumers attach to them. The importance of developing new products has prompted research that identifies consumer trends and establishes a rank of va riables which affect consumersÂ’ selection process. As a response to this increasing nece ssity, this study presum es that meat brands exist at a national level, that consumers know about these brands , and that there are perceived differences among brands. Empirical determination of purchasing decisions among meats will contribute to better understa nding of their impact on the industry, to establishing new advertising strategies ba sed on attributes, and to increasing market share. In order to do so, discrete choice models are specified to reveal the likelihood of brand recognition or degree of differentiation. The different demographics are separated
11 from other exogenous variables in order to show what is driving the preferences for each meat product. The resulting rank of the differe nt variables will help one to understand their impacts on purchasing decisions and to establish different marketing strategies. Research Objectives The objectives of this thesis are as follows: To describe and analyze the data base set To develop econometric models for meas uring brand identifi cation across meats To provide estimates of the effects of the principal variables influencing consumersÂ’ brand recognition To identify and provide a ranking of the va riables affecting the purchasing process of branded and non branded meat products To suggest potential improvements for m eat marketing and retail management based on empirical results Research Hypotheses Relative prices, demographics , seasonality, type of outle t, and location within the store have a significant impact on the d ecision-making process, when analyzing consumersÂ’ meat purchases throughout time Fresh meats and frozen products are the most important in terms of volume of sales, however there is a remarkable st atistical difference in terms of brand identification across type of outle ts and location within the store Brand identification among meat consumers continues to increase during the past years due to an increasing demand for qua lity products. Attributes such as tenderness, fat and cholesterol levels, c onvenience in preparation, and advertising are important factors, which affect cons umersÂ’ purchasing behavior and degree of brand awareness Research Methodology A comprehensive literature review of the U.S. meat industry and Brand theory will be conducted to formulate hypotheses and evalua te empirical findings. Topics related to industry concentration, livestock production, meat processing, and retailing, and brand
12 and marketing management will be discusse d to provide a theoretical background and facilitate a further analysis of the resulting data. In this study, the U.S. meat industry is analyzed from a marketing perspective and conclusions are presented based on panel data that provide secondary data consisting of time series observations on consumer purch asing behavior during a period between September 1992 and August 2001. The probabili ties of brand recognition among meats and the factors influencing this decision-ma king process are analyzed and ranked to measure the degree of identifica tion and preference. Four types of meat are considered in this study: beef, fish, poultry, and pork. Consumer preferences toward branded a nd non branded goods and services are based on the level of importance of the percei ved attributes of the product. For products such as meats, these attributes are define d subjectively by the consumer. Probit models based on discrete choice are presented to e xplain consumersÂ’ decisions. Simulations for each of the variables in the model will be pr esented and ranked in terms of the relative effect on the probability of brand identification. Data Description and Scope of the Thesis As mentioned before, the purpose of th is study is to develop and evaluate econometric models to measure consumer pr eferences towards branded and non branded meats. Data for this study were obtained from a mail survey conducted by the National Panel Diary Group Company (NPD) based on household panel reports. Based on a demographically representative panel, ever y month a questionnaire was sent to those households participating in the study. Each household was identified by a code and was excluded from the panel when it was no long er representative of the U.S. population. Household panel reports consist of eating diaries in which participating households
13 documented their purchasing rou tine for food during a designated period or Â“wave.Â” Each Â“waveÂ” was every two weeks (26 per year). This data set contains demographic and non-demographic information and was registered each month during a period of nine years from September 1992 through August 2001. The original data were classi fied by household, period, and product, with each entry representing one observation. The purchases of each type of meat were identified by a code: 557 (beef), 558 (fish) , 559 (poultry), 560 (por k). The full data set includes 775,976 observations and the approximate number of households per period is 2,000. Specifically, households reported quantitie s and expenditure levels across time of each meat product consumed, along with detailed information on household demographics (i.e., age, employment, edu cation and income levels, household size, geographic location, market size, etc.) and ot her variables including type of outlet, where within the store, seas onality, health, promoti ons, and advertising. The analysis will clearly illustrate the br and preferences of the different products by each household, including zero expenditures or no response. Finally, the scope of this research is limited to households within the U.S. Outline of Chapters The remaining chapters provide a detail ed description of the framework and methodology employed for this research, as well as examines and compares the statistical findings. Chapter 2 includes a literature revi ew on the meat industr y, brand preferences, and examples of Probit models. Topics re lated to meat demand, marketing of meat products, industry consolida tion, and brand preferences are outlined to support and facilitate the analysis of results. Chapter 3 provides a de scription of the meat industry in terms of statistical data and trends. Also in cluded in this chapter is a discussion on meat
14 brands and consumer statistics, which are presented to show the increasing demand for product differentiation. Chapter 4 includes an analysis and discussi on on brand theory as well as a full description of conceptual mode ls and probabilities. In Chapter 5, the final econometric models are estimated and the stat istical interpretations of the findings are presented. Chapter 6 describes the concept of probabilities and in cludes a series of simulations, which analyze changes in consumer preferences towards branded and non branded meats. Chapter 7 focuses on research conclusions and discusses the implication of the outcomes of the research.
15 CHAPTER 2 LITERATURE REVIEW Brand Preferences Branding is used to differentiate products and attract consumers to buy, the firmÂ’s products, conveying value and building brand loyalty. Several studies have examined brand preferences and consumersÂ’ attitude s toward meat products from different perspectives, trying to identif y the factors influencing this decision-making process and the main forces affecting the demand for meat products in the U.S. Changes in consumer lifestyles and the impacts of these change s on product selection are well recognized by processors and marketers of food. Convenience and nut rition have become more important and attributes such as low sodium, low fat, or read y to cook have become more prominent on labels, catching the attention of consumers. Consumer perceptions of these and other product characteristics will likely affect perceived quality (overall opinion) and the resulting preference for that particular food product. For example, Menkhaus, Colin, Whipple, and Field in 1993 concluded that he alth issues must be addressed by the meat industry, especially th e beef industry, by promoting the he alth benefits of meat labeling and developing new products focused on the he alth issues and convenience in order to provide meats with characteristics that enhance its quality perception and demand. As consumers have become more discri minating in their purchasing decisions (Barkema, Dranbenstott and Novack, 2001), th e beef and pork industries have responded by developing branded products. Quality, in f act, overweighs price in making purchasing
16 decisions for some food products and in the case of beef demand, changes in prices of competing meats such as poultry and pork alone cannot explain the shifts in the demand . Consumer buying behavior with respect to meat has exhibited marked and somewhat unexpected changes in recent years. Some studies have focused on the effects of changes in income levels and prices. A particular focus of attention has been the persistent appearance of negative cross price elasticity between beef and its substitutes in demand formulations. When pressure on real income forces reductions in total expenditures for meats, the impact of the reduced consumption will be felt by beef; pork consumption will decrease slightly; and consumption of chicken may actually increase (Menkhaus, St. Clair and Hallingbye, 1981). Another noticeable change in recent y ears is the growing proportion of income spent on Food Away From Home (FAFH). In f act, the share of food spending for FAFH rose from 33.8 percent in 1970 to 45 percen t in 1992. In contrast , the share of food spending allocated for Food At Home (FAH ) dropped from 66 percent in 1970 to 54 percent in 1992 (Nayga, 1995). These economic trends point to the increasing importance of FAFH consumption relative to FAH. Ot her studies like the one conducted by the Food Marketing Institute and Campbell Soup Compa ny in 2002 revealed that approximately 15 percent of total food dollars go to take-out purchases; 19 per cent of total food dollars are spent on food eaten in restau rants; and the remaining 66 percent is spent on food At Home. In essence, at least one in every three dollars spent on food is now going to convenience food marketers and Away From Home food outlets. Ailawadi, Nelsin and Gedenk in 2001 analy zed consumer choice between store and national brand promotions. They resear ched how consumer demographic and
17 psychographic traits affect consumer purchas es of store and national brand products. Psychographic traits were categorized as eco nomics (price, financial, quality), hedonic (shopping enjoyment, innovativeness, variety s eeking, impulsiveness, and motivation to conform), and costs (brand loya lty, store loyalty, planning, ti me pressure, thinking costs, and inventory build up). Economic psychogra phic traits such as price, financial constraints and cost are significant and have positive impacts on store brand usage, whereas brand loyalty is significant and ha s a negative impact on store brand usage (Parcell and Schroeder, 2003). Chakravarti and Janiszewski in 2004 examined the influence of generic advertising on primary demand and brand pref erences. They concluded firstly that generic advertising can increase or decr ease the perceived differentiation among competing brands and, thus, influence brand choice. Second, increas es in differentiation occur because generic advertising increases or decreases the weight consumers place on differentiating or non differentiating attributes . Generic advertisements that discussed a differentiating attribute decreased access to information about the non-differentiating attribute, which resulted in an increase in the importance of the differentiating attribute and increased price responsiveness. Rimal in 2005 addressed consumersÂ’ attit ude toward food labels and the influence of different aspects of meat labels on beef, poultry, and seafood consumption. This study found that consumer attitudes toward meat labels were influenced by consumersÂ’ perceived importance of nutriti on and ingredient information on the labels, consumersÂ’ opinion regarding the adequacy and enforcem ent of food safety regulations, and the responderÂ’s gender. The study also found th at those respondents who thought that
18 nutrition and ingredient information on food labe ls were very important also thought that meat labels helped them select beef and other meat products. Other studies, like the one conducted by Van Osselaer and Alba in 2000, focused on how consumers learn and predict product performance and how it affects purchase decisions and brand preferences. They studie d consumer learning of product cues as predictors of product quality with particul ar emphasis on the dist inction between brand and attribute cues, and whether consumers w ill routinely learn the determinants of product quality when attribute cues are freel y available and proce ssing is unconstrained. They suggested that learning can be suppres sed even under these relatively favorable conditions due to the learning phenomenon know n as Â“blockingÂ” where the learning of one predictive cue can Â“blockÂ” the learning of subsequently encountered predictive cues. They concluded that brand cues may block th e learning of quality-de termining attributes clues and that learning is both forward looking and competitive in the sense that cues compete against each other for predictive value. In 2000, Heilman, Bowman, and Wright examined how brand preferences and response to marketing activity evolve for c onsumers new to a market (those making their very first category purchases), based on the notion that choices made by new consumers entering a market are driven by two compe ting forces, consumersÂ’ desire to collect alternative information and their aversion to trying risky ones, and that these forces evolve in three stages of purchasing: an in formation collection stage that focuses initially on low-risk, big brand names; (2) a stage wh ere information is ex tended to lesser-known brands; and (3) a stage of information wh ere preferences are ba sed on brands that provides the greatest utility. They assumed that the rela tionship between search and
19 purchasing experience is an inverted U cu rve showing a negative, linear relationship between them where consumers with little category experience try a variety of alternatives before settling on a set that provides the greatest utility. The U.S. Meat Industry Food demand and technological innovation are transforming commodity-selling industries in order to address changes in cons umersÂ’ preferences and shifts in the demand for more differentiated meat products. During the past decades, the U.S. meat industry, beyond the farm gate, has undergone significant changes, showing a persistent trend toward consolidation. The U.S. market for meat products is dominated by a series of major US-based producers of meat: Ty son Foods, ConAgra Foods, Excel, Swift & Company and Smithfield Foods among others. St ructural change has been the principle characteristic of the meat industry in recent years, and the effect of this phenomenon on prices for consumers and producers, and the e ffect on the structure of the supply chain, have motivated a large number of studies. Firms have become larger and/or fewer in number. Vertical integration and/or coordination have been the primary method used by processors to increase efficiency in livestock marketing channels. Vertical inte gration refers to ownership across pricing points in a market channel. Vertical coordination may occu r with or without vertical integration. That is, different segments of the marketing channel may coordinate their efforts with or without the same firm owning bo th segments. In either case, the results are similar, producers and/or handlers act in ta ndem with processors, and processors gain control over at least a portion of the supply. Some argue that this leads to processing plantsÂ’ efficiency and enables them to prov ide better the types of products demanded by consumers (Bailey, 1993).
20 Menkhaus, et al. in 1981, studied the effect s of the structural changes including growing concentration, vert ical integration, and conglom erate merger, upon traditional agriculture and consumers. Firms combine in order to gain market share so they can either charge high prices to buyers and/or pay low input pri ces to suppliers or in some cases a firm will buy a larg e competitor to remove them from the market. Other studies, like the one conducted by MacDonald, Ollinger, Nelson, and Handy in 2000, concentrated their research on the differences between beef, pork and poultry slaughter and processing. One important diffe rence found by this study is that poultry commonly employs brand marketing, while beef does not, suggesting that poultry mergers may have been driven by improve ments in marketing performance. Another difference is that the beef industry is much more heterogeneous than the poultry industry in terms of plant size and productivity perf ormance, which may explain why poultry has moved so rapidly in terms of produ ct development and differentiation. TodayÂ’s time-pressed consumer is using hi s or her higher level of income to purchase more convenience, while looking for quality, variety, and value (Martinez and Stewart, 2003). Price reductions in poultry a nd pork products relative to beef and health concerns regarding the consumption of red m eat account for a large por tion of this trend. In response to shift in consumer preference s, different sectors of the food system are competing to identify and provide more pr ocessed and higher value-added products. Consumer food demand is shifting toward f ood products that are easy to prepare while also promising safe eating, improved nutrition, and greater consistency (Barkema, 2001). Many red meat processors now offer a va riety of convenient, fully cooked, or microwave-ready products. Moving away from selling meat as an unbranded commodity,
21 and again emulating poultry processors, so me beef and pork processors now are differentiating themselves from their co mpetitors by branding their products. These branded products are frequently prepackaged an d sold to retailers as Â“case-readyÂ” (Davis and Stewart, 2002). The likelihood of consum ers buying meats by brands is an evolving pattern that clearly differs by the type of brands. Examples of Probit Models Probit models extend the principles of gene ralized linear models to treat better the case of dichotomous dependent variables such as buying meats as branded or not branded products. Using the standard normal cumulati ve distribution func tion, a Probit model focuses on estimate the probability of occu rrence or nonoccurrence of an event or, in other terms, the probability that Y, the depende nt variable, equals 1 or 0, yes or no, voted or not voted, etc (Liao, 1994). These decisions are qualitative choices explained by a set of independent explanatory variables such as age, marital status, education level, income, etc., and a set of parameters (betas) that reflects the imp act of changes in x on the probability. Looking at all levels of possibl e interaction effects, Probit regression assumes that the categorical dependent refl ects an underlying quantitative or latent variable y* which is not observed and is based on utility theory or rational choice behavior (Long, 1997). Probit models are used in present research as a tool to measure, identify, and establish a ranking of the princi pal factors affecting consumer preferences and brand identific ation among meats. The objective of this part of the chapter is to review some Probit models that are relevant to the purpose of this research. Fo r example, Verbeke, Ward and Viane studied in 2000 the impact of television co mmunication and BSE (Bovine Spongiform Encephalopathy) on fresh meat consumption. Us ing discrete choice models (Probit), they
22 calculate the probability of changing meat consumption using cr oss-sectional data. Following the assumption that long-term shift in the demand for meat products may be the result of changing tastes and preferences and that short-term decreases are linked to negative media coverage, they found that the likelihood of cutting fresh meat consumption increases, particularly in house holds with the presen ce of young children and adults over 65 years old, as the attenti on to television messages increases. Discrete choice models were specified to explain cons umer decisions and the decrease of fresh meat consumption since the BSE-crisis ( DBSE ) as well as in the future ( DFUT ). In order to evaluate the impact of individual explan atory variables, a base model was set and applied to estimate the probabilities. The resulting estimates corroborated previous findings and expectations and led to a series of recommendations. Following the assumption that mass negative media coverage wi ll raise peopleÂ’s concerns about human health, the authors found that since th e BSE period and in the near future, frequent users were more likely to reduce consumption than the daily meat consumers. Also, consumers that paid high le vels of attention to television were more likely to have decreased meat consumption th an consumers with lower attention levels. The presence of young children in the househol d increases the probability of reducing fresh meat consumption because the family w ith children was more sensitive about health issues and potential harms from consuming meat . Aging was also found to be significant, showing that the probability of having decreased fresh meat consumption since the BSEcrisis was very high among young individuals whereas, rega rdless of the level of attention to the media, the probability increas es considerably for in dividuals older than 40 years. In addition, high versus low education of the head of the household almost doubles
23 the probability of reducing fres h meat consumption since the BSE-crisis, but in the near future education levels will have a negligib le influence on the intention to reduce fresh meat consumption. Findings from this research have several im plications for the m eat industry. First, because it shows a sharper reduction in meat consumption as negative mass communication frequency increases; second, it re veals the necessity of re-establishing a favorable meat image and; third, given th e reactions to negative press, future countermeasure strategies s hould be designed an d clearly targeted to communicate young people and specifically those who are form ing their consumption habits, about the positive characteristics and the advant ages of consuming meat products. In other research, Probit models were us ed to investigate the motives for mergers and acquisitions in the U.S. meat industr y. Nguyen and Ollinger in 2002, studied the changes in industry concentr ation and how merger and ac quisition (M&A) activity has caused concerns about abuses of market pow er. Results show that acquired meat and poultry plants were highly productive before me rgers, and that meat plants significantly improved productivity growth in the post-merg er periods, but poultr y plants did not. Two Â“efficiencyÂ” theories are cited in this study: the theory of Â“disciplinaryÂ” and Â“synergisticÂ” mergers. According to the first theory, acquiring firms makes profits by improving the labor productivity of the othe r company. If a merger is motivated by synergy, on the other hand, the acquiring firm targets only productive firms. After the merger, synergies between the firms improve the performance of the combined firm. Finally, if a merger is undertaken for purpos es of building monopoly power, target firmsÂ’ performance should not matter and the perf ormance of the combined firm is not
24 necessarily improved after a merger. Empirical results indicate that both initial plant size and productivity are positively related to ow nership change. Except for poultry industry products, Probit analyses prov ide strong evidence that plan t productivity growth is positively related to M&As. These findi ngs do not rule out monopolistic or monopsonistic pricing after an acquisition but do suggest that firms merge for synergetic purposes. In 2003, Parcell and Schroeder explored the effects of branded beef and pork products on retail prices and brand equity. They suggested that little information is known about the levels of differentiated pricing that brandi ng allows and how differentiated pricing varies across firms branding products. Assu ming that consumers base food purchasing decisions on the exte nded utility derived from the product under consideration and that decisions are made subject to budge t constraints and the attributes of the product, they used a characteristic demand model (Hedonic modeling) to measure the impact of implicit characteristics and mark et factors associated with the particular beef or pork product. Models were initially esti mated using Ordinary Least Squares and then re-estimated using the Multivariatet -errors robust estimati on to deal with the common concern of non normality residuals in hedonic models. Several implications developed from thei r research: (1) product size discounts are linear; (2) meat items on sale are sold at significant discounts to non sale items; and (3) specialty stores do not garner higher pri ces than supermarket/grocery stores, while warehouses/super center have premium price co mpared to supermarket/grocery stores. They found that beef and pork industries are branding differe ntiated products to attract consumers to buy a firmÂ’s products in order to command price premiums and to
25 build brand loyalty. Producers interested in branding products should focus on building consumer trust through quality and consistenc y. Having a third-part y verification process (e.g., ISO-9000 certified) in addition to cap turing economies of si ze through building a national branded program may be necessary. Th eir analysis also f ound that the level of brand premiums differed across beef and pork cu ts and concluded that brand equity value varied across beef cut with, as expected, steak having the largest premium. Ground beef had no brand premium. Another interesting fi nding is that the br and premium for an Angus steak relative to store brand is es timated to be $1.13/lb, which is one-half the premium consumers indicated they are willi ng to pay for Certifie d Angus BeefÂ® relative to generic steaks.
26 CHAPTER 3 INDUSTRY TRENDS AND DATA Major Industry Trends During the past three decades, the U.S. meat industry has experienced a profound transformation in all stages of its production chain. Today the industry is an efficient and technologically improved sector, es pecially the p oultry industry. The increasing demand for quality food products with specific attrib utes in terms of consistency, convenience, and safety has prompted an intense competition within the sector transforming an industry of commod ities into a more segmented and marketingoriented industry able to provi de consumers a vast set of al ternatives. The beef and pork industries have recognized this need and have shown clearly a str ong trend towards new marketing procedures, adding value to th eir products, reinforcing attributes, and improving presentation in order to build a positive perception about their products and regain their position in the market. In order to provide the read er with some insight into the meat industry and its trends, Figures 3-1 through 3-5 were cons tructed using data from the Livestock Information Marketing Center and estimations from the National Panel Diary (NPD) data set. These figures show statistics for the four types of meats, including total consumption, historic prices, and total expenditures from 1979 to 2002. It is very important to mention that due to the lack of reliable informati on, fish prices from year 1979 to 1991 are not included in this analysis. These data limitations were confirmed by the staff of the Fisheries Statistic Division (S T1) through a phone interview.
27 Figure 3-1 shows that over several decades there have been considerable shifts in consumption from red to white meat. Cha nging lifestyles are ca using consumers to demand more convenience foods, and white meat s are often perceived to be healthier than red meats (Medina and Ward, 1999). The poultry sector, and in less magnitude the pork sector, learned how to capitalize th e changes in consumersÂ’ preferences by conducting a deep transformation within their i ndustries, especially in areas re lated to their structure, product di fferentiation, and marketing. 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 200 2 Years0 10 20 30 40 50 60 70 80 90 100 Average consumption (lbs. per capita)BeefPorkPoultryFish Figure 3-1: Distributi on of average consumption of m eat products across time. Source: NPD panel data and the Livestoc k Marketing Information Center. Total per capita consumption of beef, pork, poultry and fish increased by 23 pounds over the period between 1979 to 2002 with the average person increasing his or her total meat consumption by more than 1.5 pounds a year since 1990 (USDA-ERS, 2004). Beef consumption has declined steadily over the last two decades, both in aggregate quantity and as a share of total U.S. meat consum ption (see Figures 1-4 and 1-5), showing a substantial long-term down ward trend of 15 percent decrease , especially af ter 1986 when
28 increasing concerns about health issues (cholesterol, fat levels, etc.) and safety issues (E. coli, BSE, etc.) among consumers originated a important shift in the demand for meat products. Poultry consumption, however, has presented a significant increase during the same period, showing a solid upward trend of 73 percent. This trend is due in part to the industryÂ’s emphasis on producing value-added, convenient products and the introduction of a technologically improved production pr ocess. Consumption of pork and fish products has remained close to their historic average, showing in th e case of pork, some short-term spikes such as the one between 1990 and 1992 during the BSE outbreak in Europe which, affected meat consumption worl dwide. Figure 3-2 shows price trends over the past three decades for beef, poultry, pork, and fish products. 1979 1985 1991 1997 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 200 2 Years0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Average price ($/lb)BeefPorkPoultryFish Figure 3-2: Distribution of average price of meat products across time. Fish prices are from 1992 to 2002. Source: Livestock Marketing Information Center and the NPD panel data. The sustained increase in beef prices might be attributed to the relatively tight supply of domestic beef cattle and to increasing levels of product differentiation observed along the beef production chain in response to the increasing dema nd for healthier and
29 safety products, as cited by Gregg Doud, ch ief economist of the National CattlemenÂ’s Beef Association as cited by Prewitt in 2003). Poultry prices show an increasing trend as a result of the st rong demand, reflecting concerns about health and safety issues, changes in lifestyles and consumersÂ’ attitudes a bout the other meats. Poultry products have become differentiated, supplying more bran d-based products. In Figure 3-2, average prices are being compared while recognizing th at the level of differentiation within each meat type is vastly different. Thus the aver age prices are also across levels of brand identification. Figure 3-3 shows the variation in prices across the time for each meat type. In order to understand the nature of the fluctuati ons and to compare prices, coefficients of variat ion were calculated (e.g, / CV ). 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 200 2 Years0.80 1.00 1.20 1.40 1.60 1.80 2.00 Price index (Year 1992 = 1.0) BeefFishPorkPoultry CVBeef = 6.8CVFish = 8.6 CVPork = 10.1CVPoultry = 12.5 Figure 3-3: Variation on pri ces across time. Indexed to years 1992. Source: NPD panel data. The biggest fluctuations corresponded to the poultry prices with 12.5 percent of variation across time, while pork, fish, and b eef products present smaller price variations during the same period with 10.1percent, 8. 6 percent and 6.8 pe rcent respectively.
30 Oligopolies, like the po ultry industry, with a large set of value-added products, almost always sell branded products at some pr emium over generic or unbranded products. Figure 3-4, shows a persistent upward trend in total expenditures for all meat types. 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 200 2 Years0 50 100 150 200 250 300 Total expenditures ($ per capita)BeefPorkPoultryFish Figure 3-4: Distribution of total expenditures per capita of meat products across the time. Fish expenditures from 1992 to 2002. Source: Livestock Marketing Information Center and the NPD panel data. Despite the fact that beef consumption has been reduced over the past decades, upward prices and increasing demand levels of all other meats, especially poultry, have impacted positively the overall level of expend itures for the entire industry. According to Medina and Ward (1999), within a two-w eek period 95 percent of U.S. households purchase some quantity of beef, fish, pork, or poultry with annual expenditures accounting for 2.2 percent of the typical household income. Expenditures per capita on beef, pork, a nd poultry products have increased 7.1 percent per year between 1979 and 2002, while during the period 1992 to 2002 expenditures on fish products have increase d 6.3 percent. Looking at the percentage change in expenditures during the period 1992 to 2002, househol ds expenditures in
31 poultry account for the largest increase with 33.8 percent, followed by pork with 30.5 percent, beef with 18.7 percent, and fish show ing the smallest increase with 12.4 percent. Beef 42.0% Fish 11.2% Poultry 15.3% Pork 31.5% Figure 3-5: Expenditure shares on meat pr oducts per capita in year 2002. Source: NPD panel data. In terms of share in expend itures per capita, Figure 3.5 cl early shows that beef lead the sector with more than 42 percent of the total expenditure which represents a value of $ 62 billions for the year 2002. Pork, poultry, and fish products present expenditures shares of 31.4 percent, 15.3 percent, and 11.1 percent representing in terms of dollar value $48.2 billions, $24.9 billions , and $13.3 billions respectively. Brand Preferences The U.S. meat industry has changed dur ing the past three decades from a commodity-selling orientation to a marketing-or iented industry. The pr inciple behind this transformation is to satisfy needs and prefer ences of different sectors of the population. These consumers have certain requirements that can only be satisfied through a specialized type of product and, in turn, improve the overall demand for the meat category.
32 New product development, product diffe rentiation, and product branding are some tools by which market segments are captured and maintained (Skaggs, Menkhaus, Torok, and Field,.1987). Through market segmentation, consumersÂ’ preferences are recognized and products are developed with characteristics oriented to influence consumersÂ’ predictions of product performan ce. According to Van Osselaer and Alba (2000), if consumers learn the relationship between product attributes and product performance (health, safety, convenience, etc) they will differentiate among brands that possess the expected attributes. The meat industry has learned how to ad apt its production line over time to meet the needs of consumers. Today, it is possibl e to find in retail stores branded meat products segmented on the basis of price by se lling un-graded, choice or prime meat; or on the basis of fat levels by selling lean products; or on the basis of convenience by selling easy to prepare meals or microwave re ady products; or on the basis of the origin by selling organic or natural products (Men khaus et al., 1993). Brands often reflect processed packaged meats bearing the mark s pointing to uniform characteristics or attributes, which in the long run can be tran slated into increasing levels of consumer loyalty and brand recognition. It is important to mention that fresh meat forms may have also brand identity (e.g., Certified Angus Beef). Within the industry, brand preferences among meat products ha ve shown different trends during the time frame of this research . In Figure 3-6, beef, fish, and pork products clearly present the largest incr ease in brand identification.
33 199219931994199519961997199819992000 0% 20% 40% 60% 80% 100% Percent of users indicating brand identification BeefFishPoultryPork Beef13%12%13%17%20%20%21%23%22% Fish42%43%45%46%50%47%47%46%47% Poultry80%77%78%80%81%80%80%78%76% Pork44%40%41%40%43%44%45%49%48% Figure 3-6: Change in brand identification for all meats across time. Source: NPD panel data. Beef brand identification in creased almost 9 percenta ge points over the period 1992 to 2000 while, fish and pork br and identification increased 5 and 4 percentage points respectively during the same period. Pou ltry products show a negative trend with reduction in brand identification of 4 percen tage points during the same period. Theses numbers show the importance of marketing prac tices and the vast pot ential for promotion initiatives (e.g., generic advertising programs) within the beef a nd pork industry, where most of the meat marketed is still nonbranded. Description and Definition of Demographic Variables From September 1992 to August 2001 a bout 775,976 observations on household consumption of beef, fish, pork, and poultry were collected by the National Panel Diary Group company (NPD, 2001), representing 9,964 participating households. Each household kept eating diaries, documenting th eir purchasing habits during a period of
34 two-week called a Â“waveÂ”. The minimum numb er of times a household reported is one and the maximum is 96 times. The respondent reported specifically whether during this period she or he purchased a meat product by br and, the type of store, where within the store, price, and some demographic inform ation such as income, household size, age, presence of children, employmen t level, education, etc. Each household and each type of meat has been identified by a code where 557 represents beef; 558, fi sh; 559, poultry; and 560 is for pork. Table 3-1 provides a full description of each of these demographic variables and their distribution in terms of number of households and percentage of participation during the time of the survey. About 67 percent of the total number of households had incomes below $50,000, while 33 percent of the households had annual incomes over $50,000. Household size was measured according to the number of members, and more than 35 percent of the households had four or more members follo wed of households with two members with 31 percent. Female heads of the househol d with ages between 40 and 65 represent 55 percent of the sample while almost 30 percent were less than 40 years old. All members of the household whit less than 18 years we re considered as children and almost 42 percent of the families had shown the pr esence of them. About 35 percent of the respondents worked full time or more than 35 hours per week, almost 19 percent worked part time or less than 34 hours per week a nd 47 percent responded th at they were not employed at the moment of the survey. Regarding the education level among the respondents, 51 percent of the female h eads had high school education, 41 percent graduated from college or had some college education, and only 6 percent responded that they had a post graduated education.
35 Table 3-1: Demographic variab les from the NPD data set. Description Variable Name / Range Values INC: income per household (dollars) INC1 INC4 INC3 INC2 0-24,999 25,000-49,999 over 75,000 50,000-74,999 3,366 (33.66%) 1,245 (12.45%) 2,049 (20.05%) 3,304 (33.04%) HWZ: household size (Number of members) HWZ1 HWZ4 HWZ3 HWZ2 1 2 3 4 or more 1,251 (12.51%) 3,117 (31.28%) 3,528 (35.40%) 2,068 (20.75%) AGF: age of female head (years) AGF1 AGF2 AGF4 AGF3 Under 25 Over 65 40 to 65 25 to 40 193 (2.10%) 1,396 (15.25%) 5,039 (55.04%) 2,526 (27.59%) CHD: presence of children under 18 (years) CHD2 CHD1 Yes None < 18 4,196 (42.11%) 5,768 (57.88%) EMF: employment female (hours per week) EMF1 EMF3 EMF2 Full time Part time Not employed 3,165 (34.57%) 4,249 (46.88%) 1,697 (18.53%) EDF: female head education level EDF1 EDF2 EDF4 EDF3 High school or less Post graduate College graduate Some college 4,701 (55.25%) 584 (6.37%) 1,540 (16.82%) 2,329 (25.44%) In terms of the occupation of the house hold head, there were twelve different classifications. They consisted of retired and unemployed that represented the largest group with 29 percent of the respondents, fo llowed by professionals and proprietors with 15 percent and 14 percent respec tively; craftsman with 13 percent and operative workers with almost 10 percent. In terms of market size, more than 48 percent of the respondents lived in large cities with a population of 1,000,000 or more. Approximately 30 percent lived in cities with sizes between 50,000 people and 1,000,000 people, and almost 22 percent lived in rural areas or ci ties with less than 50,000 people.
36 Table 3-1: Continued. OCC : occupation householder OCC1 OCC2 OCC3 OCC4 OCC5 OCC6 OCC7 OCC8 OCC9 OCC10 OCC11 OCC12 Professional Proprietor, Manager, etc Clerical Sales Craftsman Operative Military Service worker Farm (owner, laborer, etc) Student employed under 30 hrs Laborers Retires, Unemployed 1,505 (15.10%) 1,419 (14.24%) 400 (4.01%) 496 (4.90%) 1,312 (13.16%) 959 (9.62%) 2,910 (29.20%) 465 (4.66%) 113 (1.13%) 68 (0.60%) 208 (2.08%) 109 (9.66%) STA: Regions (based on census) STA1 STA2 STA3 STA4 STA9 STA6 STA7 STA8 STA5 New England Middle Atlantic East North Central Pacific South Atlantic East South Central West South Central Mountain West North Central 591 (6.18%) 1,704 (17.10%) 1,712 (17.18%) 625 (6.27%) 1,457 (14.62%) 648 (6.50%) 917 (9.20%) 487 (4.88%) 1,823 (18.29%) MSZ: Market Size (number of people) MSZ1 MSZ2 MSZ6 MSZ4 MSZ5 MSZ3 50,000-249,999 250,000-499,000 500,000-999,999 2,500,000 or more 1,000,0002,499,999 Non market size 704 (7.06%) 1,157 (11.61%) 1,059 (10.62%) 2,221 (22.29%) 2,479 (24.87%) 2,344 (23.52%) In relation to the geographi cal location of the respondent s, the country was divided into nine regions according to the 2000 U.S. census. About 42 percent of the respondents lived on the east coast (New England, Middle Atlantic and South A tlantic regions), 39 percent lived on the central states of the c ountry, almost 5 percent lived on the mountain region, and the last 13 percent lived on the we st coast. Table 3.2 s hows the dist ribution of the states for each region of the country according to the U.S. census.
37 Table 3-2: Distribution of States for each region. New England East North Central South Atlantic Mountain MaineOhioMarylandMontana New HampshireIndianaDelawareWyoming VermontIllinoisWashington D.C.Colorado MassachusettsMichiganVirginiaIdaho Rhode IslandWisconsinWest VirginiaNew Mexico ConnecticutNorth CarolinaNevada South CarolinaArizona FloridaUtah Georgia Mid-Atlantic West North Central East South Central Pacific New YorkMinnesotaKentuckyWashington New JerseyIowaTennesseeOregon PennsylvaniaMissouriAlabamaCalifornia NebraskaMississippi Kansas North Dakota South Dakota West South Central Arkansas Louisiana Oklahoma Texas One important aspect in the analysis of th e demographic variables is to measure the correlation or the degree of linear associa tion between two variab les (Gujarati, 2003). Given the large number of observations and participating households in the panel, minimum levels of correlation are expected. Table 3-2 presents the coefficients of correlation among demographic variables, and it is clear that th ere are no correlation problems between any of them.
38 Table 3-3: Correlation among demographic variables. DINCDHWZDAGFDCHDDEMFDEDFDOCCDSTADMSA DINC 1 0.143-0.002-0.078-0.1690.374-0.444-0.021-0.093 DHWZ0.143 1 -0.130-0.5540.1040.258-0.214-0.0110.005 DAGF-0.002-0.130 1 0.3530.3990.3130.298-0.021-0.014 DCHD-0.078-0.5540.353 1 -0.018-0.2100.249-0.013-0.003 DEMF-0.1690.1040.399-0.018 1 0.1440.2760.0230.016 DEDF0.3740.2580.313-0.2100.144 1 -0.2520.018-0.065 DOCC-0.444-0.2140.2980.2490.276-0.252 1 0.0040.030 DSTA-0.021-0.011-0.021-0.0130.0230.0180.004 1 -0.005 DMSA-0.0930.005-0.014-0.0030.016-0.0650.030-0.005 1
39 CHAPTER 4 BRAND PREFERENCE THEORY AND MODEL SPECIFICATION Brand Theory Â“A brand is a product that provides functiona l benefits plus added values that some consumers appreciate enough to buyÂ” (Jones, 1986). Branding has been around for centuries as a means to distinguish the goods of one producer from those of another. The American Marketing Associati on defines brand as a Â“ name, term, sign, symbol, or design, or a combination of them, intended to identify the goods and services of one seller or group of se llers and to differentia te them from those of the competition.Â” More sp ecifically, what di stinguishes a brand from its unbranded counterpart and gives it equity is the sum to tal of consumersÂ’ perceptions and feelings about the productÂ’s attributes. This sum to tal includes how the products perform, the reputation of the brand name, what the brand st ands for, and the company associated with the brand (Achenbaum, 1993). Added values are the most important part of a brand because they imply the set of differences that distinguish one product fr om another. Based on added values, the strongest brands are often the most distinctive in terms of the degree of benefits provided to consumers. Consumers buy any particular brand in the produc t field because the motivating benefits attached to the good or se rvice, but in the same way consumers might buy another brand based on the di scriminating benefits attached to it. Distin ctiveness is extremely important and highly desirable for a brand. By cr eating perceive differences
40 among products through branding, firms satisfy the needs of consumers, develop loyalty among consumers, reduce defection to comp etitors, ease introduction of new products, and generate profits, which represent the ultimate objective of any company (Jones, 1986). These differences may be rational and tangible--related to performance of the branded product--or more emoti onal or intangible--r elated to what th e brand represents for the consumer (Keller, 1998). Brands can also play a signi ficant role in signaling certa in productsÂ’ characteristics to consumers. Previous studies have classified products and th eir associated attributes or benefits into three major categories: s earch goods, experience goods, and credence goods. With search goods, productsÂ’ attributes can be evaluated by visual inspection (e.g., size, color, ingredient composition of the produc t, and weight). With experienced goods, product attributes cannot be assessed so easil y by inspection, and actual product trial and experience is necessary (e.g., durability, service quality, safety, taste, and easy to use). With credence goods, products attributes ma y be rarely learned (e.g., insurance coverage). Because of the difficulty in asse ssing and interpreting product attributes and benefits with experience goods and credence goo ds, brands may be particularly important to evaluate quality and other characteristics (Shocker, Srivastava and Ruekert, 1994) . In his 1986 book, Jones argue that bra nds are essential manifestation of oligopolistic competition, that they are a combination of functional and nonfunctional values and that the contribution of advertisi ng is mainly to encourage the use of a brand which helps build nonfunctional added values. In the eyes of consumers, added values play a role in almost all purchasing decisions and are often seen as the justification of the premium prices commonly ch arged for branded products.
41 Brands suggest a guarantee of homogeneity and product quality to buyers and more important, they provide a unique tool to differentiate one pro duct from another. However, it is important to mention that tastes differ so widely that no brand can be all things to all people. This is the main principle of bra nd differentiation and the core reason for which industries like poultry and in less degree pork has moved towards a deep marketing transformation, from a commodity selling i ndustry to a highly specialized sector. In recent years, numerous branded commod ities have appeared in the food market. A commodity can be defined as a very basic product that cannot be physically differentiated in the minds of consumers. Over the years a number of products that at one time were seem essentially as commodities have become highly differentiated as strong brands (Levitt, 1980). Examples of successf ully branded commodities are Chiquita (bananas), Perdue (chickens and turkeys), Qu aker (oatmeal), Dole (pineapples), etc. The main element for success in each of th e previous examples was that consumers became convinced that there was a meaningful difference within all the products offered in the category. Chiquita, for example, used a very effective marketing campaign that convinced consumers that its bananas were not a simple commodity and actually they were bananas of superior quality. One of the best examples of branding a commodity was the California Raisin Advisory Board, which created some badly needed brand personality and image for its product on the basis of some highly creative advertising (Keller, 1998). For retailers and othersÂ’ distribution cha nnels, brands are very important because they carry out a number of important functions . Brands can generate consumer interest, patronage, loyalty in a store, and more important higher ma rgins and profits. Retailers
42 such as Publix or Wal-Mart create their ow n brands by attaching uni que associations to the quality of their products such as variet y and pricing. These store brand products may come from third party manuf acturers or from the store itself and can be marketed by using the store name, by creating new names, or some combination of the two. A good example of this can be seen in the meat s ection of any major grocery store, such as Publix, where beef products are marketed as Publix Prime beef which has thin layers of fat, called marbling, or Publix Choice beef which has little or no marbling but a layer of pinkish fat on the outer edges. Chickens can be purchased whole, w ith weights that range from 3 to 10 pounds (1,350 to 4,540g); cut in parts such as breasts, thighs, legs, and wings; and even as ground meat. Some chicke n parts are available boneless and skinless. Chicken is usually sold fresh, but in some supermarkets it has been frozen and thawed (Publix Asset Management Company, 2005). It is obvious that brands play a central ro le in todayÂ’s meat food sector and Figure 3-6 clearly shows this market trend. At issu e then is how important are the brands for beef, fish, pork, and poultry and what drives the decision making process of purchasing by brand. Conceptual Brand Preference Models The primary focus of this research is to address brand awareness and whether or not consumers purchase meat products based on brand preferences. The data set includes 775,976 observations randomly collected and processed by the National Panel Diary Group Company (NPD) from September 1992 to August 2001. During a period or wave of two weeks, households provided a wide ra nge of information about meat consumption habits. Each entry consisted of a discrete response to the question if the meat product purchased was branded or non branded. Hence the adoption of limited dependent variable
43 models is appropriate since the binary response is yes or no given they purchased the product. Probit models are used to calculate the probabilities and the range of impacts that each factor or variable has on the likelihood of a consumer buying meats by brand. Given the assumption of normality and letting Â“BrandÂ” represents the binary brand identification, the model can be formulated as , exp 2 1 ) 1 (Brand Prob ) ( ) 1 Brand ( Prob2 /2dz X X F XX z i (4-1) where the probability that an individual will select a meat product is given by the values of the explanatory variable(s) Xi , which represent attributes and or demographic and nondemographic characteristics expected to have some impact on the likelihood of buying a branded meat product. F is the sta ndard cumulative normal distribution function (CDF) with zero mean and unit (=1) variance (i.e., NID (0, 1)) (Gujarati, 2003). The last term of the equation represents a range of c hoice alternatives and is measured by the area of the standard normal curve from iX to (Liao, 1994). In order to analyze the bina ry outcome, the Probit model assumes that there is a latent consumption decision or underlying response variable *y defined by the equation i i iX y *, (4-2) Where i represents the respondent, is a vector of structural coefficients that indicates the effect of th e independent demographic a nd non demographic variables
44 * oni iy X; i is the stochastic random error term for respondent, and the conditional expectations are expressed as E (i ) = 0 and iXequal to ) ( E* i iX y (Maddala, 1993). In practice, * iyis unobserved and i is symmetrically distributed with zero mean and has its cumulative distribution function (CDF) defined as F (i ).The latent variable is related to the observed binary vari able or dummy variable (iy) in a simple way: if the latent variable is greater than zero, the observed vari able is 1, that is, the consumer purchased based on the brand; otherwise it is 0 if did not: . otherwise 0 0 if 1* iy y (4-3) The interpretation of the binary outcome represents the probability for which the realized value is 0 or 1, although the theore tical value could be any intermediate number that representing the likelihood for buying a branded product. See Long (1997) for a detailed review. Since, the normal distribution is symmetr ic, from equations (4-2) and (4-3) we get the following expressions: ) 0 ( Prob ) 0 ( Prob ) 1 (y Prob* i i i iX y (4-4) ) ( Probi iX , (4-5) where the observed probability of brand preference iy is related to latent probability* iy , only if0*iy . To estimate the values of the coefficients in equations (4-4) and (4-5), both terms are corrected by the st andard error of the regression assuming that the value of is equal to 1
45 i iX Prob (4-6) i iX Prob. (4-7) Then the probability of buying a branded meat product can be defined as ) / ( ) 1 ( Prob i iX y (4-8) ) / ( 1 ) 0 ( Prob i iX y . (4-9) where(.) indicates the density function of the standard normal distribution for i . Therefore, the distribu tion of the error termi , will determine the link function of the generalized linear model (Liao, 1994). Since the observed va riables are just realizations of a binomial process with pr obabilities given by equations (4-8) and (4-9) and varying from trial to trial (depending oniX ) we can write the likelihood function as n i Brand i Brand ii iX X1 ) 1 ( ) (1 Likelihood . (4-10) Given the functional form of errors i , the parameters can be estimated through maximizing the value of the log likelihood function for each equation as follows: LogLikelihood = i i i n i iX Brand X Brand 1 ln ) 1 ( ln1. (4-11) Maximization of the likelihood function is used to estimate Probit parameters and is accomplished by nonlinear estimation methods. The principle of this maximization is to interact estimates with a set of iX parameters in order to get the highest probability of buying a meat product by brand (Maddala, 1993).
46 Several measures of goodness of fit are suggested for models with qualitative dependent variables. The classical R2 estimation is the relative predictive power of a model. It measures the percen tage of observations that ar e correctly predicted by the model or in other words, it measures the proportion of the to tal variation in y (the probability of buying meat by bra nd) that is explained by the simultaneous predictive power of all the explanatory variables (G reene, 2003). Estimates for each measure of goodness to fit are reported in the Pr obit model estimation section. Meat Brand Model Specification Probit models are extremely common in the social sciences. Capps and Branson (1998) studied consumersÂ’ char acteristics associated with the selection of lean meat products. Pelzer, Menkhaus, Whipple, Field, and Moore analyzed factors influencing consumer rankings of alternative beef pack aging. Verbeke, Ward and Viaene in 2000, explored the factors influencing meat consum ption decisions in Be lgium. Other studies included whether meat labels affect consum er attitudes and meat consumption patterns (Rimal, 2004); if mergers and acquisitions a ffect the productivity of U.S. meat products industries (Nguyen and Ollinger, 2002) and wh ether brand choices are affected by different marketing mixes in the shortand long-run. Probit models estimate the impact that independent variables have on the process of purchasing meat products by brand. Probit models are used also to calculate the probabilities of brand recognition under simu lated conditions (Liao, 1994). For this research, the final Probit model for all meats was estimated by combining the four meat types into a single model which also include d meat type, price, and time integers.
47 For convenience of interpre tation and based on the c onditional expectation E (i ) = 0, the dependent variables (* iy ) (unobserved) are denoted as Xin equation (4.12) and iy is denoted as BRNP in equations (4-13) and (4-14) respectively. In relation to the binary response, it equals 1 if the respondent purchased a meat product based on brand preference and equals zero if the purchase was not branded. The independent variables denoted as iX reflect demographic consumer charac teristics, seasonality, store type, where in the store, type of m eat, price, and time trend. All independent variables, except price and time, are binary or dichotomous variables. The response variable is qualitative in nature a nd can take only two values: yes or no, 1 or 0. Binary vari ables are usually treated as dummy variables, which are essentially devices to classify data into mutu ally exclusive categories. Each variable is defined in such a way that if the qualitative variable has m categories, it introduces only ( m 1) as a dummy variable XXZ XXXZiji i jj iji i jjj () .0 1 4 1 0 1 3 41 (4-12) As suggested in equation (4-12) the variablejiXrepresents the typical dummy term with four classes (i.e., income) andjZis the continuous variable. Depending on the characteristics of each variable, it might be represented by two, three, four, or more dummy variables. In the case of income, INC1, INC2, INC3, and INC4 are dummy variables for each income level ($0 $24,999, $25,000 $49,999, $50,000 $74,999, and over $75,000). (See table 3-1 for a full description of variables and thei r respective set of dummy terms). Although they are easy to incorporate in the regression model, one must
48 use dummy variables very carefully, specif ically to avoid a situation of perfect collinearity or perfect multicoll inearity where there is more than one exact relationship among the variables, also calle d the dummy variable trap . In the current analysis, the Probit model for brand preference am ong meat products is expressed as i i iX y *, . ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (4 3 1 66 66 4 3 1 62 4 3 1 59 12 11 1 48 48 5 4 1 43 5 4 1 39 6 5 1 34 9 8 1 26 4 3 1 23 12 11 1 12 3 2 1 10 2 1 10 4 3 1 6 4 3 1 3 4 3 1 0TT PCAT PCAT TT ZPRCP PCAT PCAT PCAT PCAT MTH MTH ZPRCP STWR STWR STYY STTY MSZ MSZ STA STA EDU EDU OCC OCC EMF EMF CHD CHD AGE AGE HWZ HWZ INC INC Xj j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j j (4-13) ) / ( ) 1 (BRNP Prob i iX . (4-14) ) / ( 1 ) 0 (BRNP Prob i iX . (4-15) Nine demographic variables are included: income level (INCj), household size (HWZj) , age (AGEj) of the respondent, pr esence of children unde r 18 years of age (CHDj) in the household, employment level (EMFj) of the respondent female head, occupation (OCCj) of the respondent, and education (EDFj) of the female head. GeographicÂ’s are measured with regional va riables (census regions) (STAj) , and market size (MSZj) . Each of the demographic variables is compared to the mean and the corresponding coefficients measure the difference or deviation of each variable with respect to the mean. Adjustments across time are captured with a time trend variable (TT) and season
49 dummies (MTHj) which are expressed in months a nd each month is compared to a consumer average for brand preference over the year. The store type (STTYj) represents where the meat purchase was made and each st ore is compared to the average of brand preference across outlets for all meats. Where in the store (STWRj) represents the location within the store and is compared to the av erage of brand preference across locations in the store. The data set presents four t ypes of meats: beef, fish, poultry, and pork (PCATj) ; and each type is compared to the average of brand preference across all meats. Interaction terms such as meat type x price (PCATj) (PRCPj) and meat type times trend (PCATj) (TT) are introduced in equation (4-13) to reveal the changes in consumer brand preferences across time and price duri ng the sample period. Each qualitative variable has been define d in such a way th at the sum of the coefficients for that particular variable is equal to zero in order to estimate better the effect of each qualitative variable on the probability of buying branded meat products. For example in the case of the type of store, the five outlet type coefficients are added to zero. Then each type of outlet coefficien t is estimated as follows: Supermarkets ( STTY1):39 0 ; Warehouse/Club store ( STTY2):40 0 ; Butcher/Meat market ( STTY3):41 0 ; Supercenters ( STTY4):42 0 ; and all other( STTY5): 42 41 40 39 0 . Hence 0 represents an overall average across outlets instead of just one of the four types of stores. Again this leads to the procedures set forth in both equations (4-12) and (4-13). The category in which no dummy variable is assigned is known as the base and all comparisons are made in relation to it. The intercept value (0 ) represents the mean value of the base category. The coefficients attached to the dumm y variables are known
50 as the differential intercept coefficients becau se, they tell by how much the value of the intercept that receives the value of 1 differ s from the intercept coefficient of the base category (Gujarati, 2003). Brand preferences are expected to be infl uenced first by the type of meat simply because of the unique attributes of each meat category and the underlying differences in the meat industry. Equation (4 -13) presents a rapid way to measure changes in the likelihood of brand preferences across meats for any combination of variables with respect to the average household. All variable estimates speci fied in this equation are reported in Table 5-1 along with the corresponding statistics.
51 CHAPTER 5 BRAND PREFERENCE PROBIT ESTIMATES As indicated earlier, brand preferences ar e binary and, hence, modeling theses preferences requires adopting models that de al with the binary issue. Probit models, based on the normal distribution, are a useful approach to measuring the driving factors of brand preference. Probit coefficient es timates for brand preference among meat consumers and their respective analyses are pr esented in this chapter. Using the data described in Table 3-1 and the specifications set forth in equations (4-14) and (4-15), discrete choice models were estimated to evaluate the changes in the probabilities of buying meat products by brand. In this research, the data set consisted in pooling time-series and cross sectional observations that were arranged where all obser vations of a cross-section were recorded chronologically. That is, the complete time se ries for the first group must be followed by the complete time series for the second group, etc . Note that the cross-sections are not balanced and in many cases there may be only one entry for a cross-section. Probit models were estimated separately for beef, fish, pork, and poultry and then a pooled Probit model combining of the four meat types into one data set was estimated. Both forms included a price i ndex variable and a time trend variable to measure changes in brand preferences for all meat types ove r the time and across price levels. Excepting price and the time trend, all other variables are binary and the coe fficients reflect the dummy variable estimates. Using the procedur e noted with equations (4-12), the sum of the coefficients for each dummy was restri cted to zero to avoid the well-known dummy
52 variable trap. The coefficients attached to the dummy variab les are known as the differential intercept coefficients because they tell by how much the value of the intercept changes (Gujarati, 2003) with each attribute wi thin a dummy variable . Hence, using this method to restrict the sum of the dummy coefficients to zero (i.e., k i i 10or 1 1 k i i k ) the intercept value (0 ) represents the mean value of the base category or in other words, the average household for whatever values are set for the conti nuous variable Z. This is a much more convenient approach given the large number of discrete right-hand side variables present in the models. In order to detect multicollinearity in the sample, the independent variables were tested and no substantial degrees of co rrelation were found among the independent variables. For all estimates, corresponding t-values are repo rted along with a number of supportive statistics. At the bottom part of each table, the number of observations used in the final estimation and the likelihood values are shown. The classical R2 estimation is the relative predictive power of a model (goodness-of -fit) given its limited interpretation in Probit models. Although in theory higher va lues of goodness-of-fit are preferred, it is very common to find low R2 values when using large pooled data sets. Generally, the different coefficients can be grouped in demographic effects and non-demographic. The demographic effects in clude income, household size, female age, presence of children, female employment, fe male education level, occupation, region of the country, and market size. Non-demographi c effects include store type, where in the store, months (seasonality), meat category, price, and time trends.
53 Probit Brand Preference Models by Meat Type Table 5-1 presents Probit estimates and supporting statistics for each meat category. Given the large sample size, the t-values can be compared to t-values tables of 1.96 for a two-tail 95 percent confidence leve l. The estimates in Table 5-1 are important in showing the positive and negative effects of each variable as well as the level of significance. Given a significant statistical estimate, a positive sign of a parameter suggest the likelihood of the response (brand id entification) increases with the level or presence of that particular explanatory vari able, with the other variables held constant. Conversely, a negative sign will suggest that the likelihood of the response decrease with the level or presence of that particular variable . However, these estimates are less useful in determining the magnitude of effect since the marginal responses to any of the variables depend on where they ar e on the normal distribution. For example, if a given set of variables selected will affect the probability of using a branded product. In order to facilitate the interpretation, c onsider for example the effect of income on brand preferences across all m eat categories. When comparing income estimates to an average household at the 95 percent level of confidence, they are not statistically significant when looking at income levels between $ 50,000 and $74,999 for beef and pork, and $25,000 to $49,999 for fish estimates. On the other hand, beef coefficients for incomes lower than $24,999 and in the rang e of $25,000 to $49,999 are statistically significant and positive. That is, they are significantly different from the average. The majority of the demographic coefficients are statistically significant when compared to the average household at the 95 percent level of confidence. However, there are some exceptions of particular relevance which mi ght require some consideration in further analysis.
54 Table 5-1: Probit model estimates by meat categories Beef Fish Poultry Pork Coef. t-value Coef. t-value Coef. t-value Coef. t-value Interception Store Type Where in Store Price Income Hwd. Size Female Age Children Employment Occupation Education Regions Market Size Months Time Trend Average STTY1 STTY2 STTY3 STTY4 STWR1 STWR2 STWR3 STWR4 PRCP INC1 INC2 INC3 HWZ1 HWZ2 HWZ3 AGF1 AGF2 AGF3 CHD EMF1 EMF2 OCC1 OCC2 OCC3 OCC4 OCC5 OCC6 OCC7 OCC8 OCC9 OCC10 OCC11 EDC1 EDC2 EDC3 STA1 STA2 STA3 STA4 STA5 STA6 STA7 STA8 MSZ1 MSZ2 MSZ3 MSZ4 MSZ5 MTH1 MTH2 MTH3 MTH4 MTH5 MTH6 MTH7 MTH8 MTH9 MTH10 MTH11 TT -0.9618 0.0486 0.1797 -0.2430 0.0331 -0.3298 0.0175 0.2002 0.3448 0.0534 0.0285 0.0393 0.0008 0.0158 -0.0350 0.0068 0.1625 -0.0318 -0.0531 -0.0127 0.0122 0.0075 -0.0613 0.0640 0.0690 0.1235 -0.0250 -0.0440 -0.0132 -0.4018 0.0333 0.1682 -0.0300 -0.0520 -0.0041 0.0326 0.0232 -0.0247 -0.0715 0.1428 0.0992 -0.2126 -0.0494 0.0163 0.0128 0.0374 0.0277 -0.0033 0.0147 -0.0082 0.0127 0.0980 -0.0200 -0.0089 -0.0132 0.0206 -0.0287 -0.0024 0.0003 -0.0271 0.0053 -59.6565 6.8696 14.0307 -17.1585 1.9778 -29.5365 0.6328 6.5155 23.6474 12.0083 5.4205 9.1213 0.1546 2.2652 -7.0790 1.3643 12.4401 -5.0680 -9.6516 -1.5067 3.0898 1.5995 -2.0169 7.8973 8.5174 10.5379 -2.1287 -5.4974 -1.4880 -15.9927 2.7911 7.3096 -1.6586 -11.0038 -0.7997 5.7099 2.4965 -3.8021 -11.7369 16.3271 17.0464 -21.1788 -6.2455 1.6100 1.6545 5.7841 4.0328 -0.6308 2.7899 -1.0111 1.5156 12.3676 -2.3781 -1.1016 -1.6246 2.5412 -3.5041 -0.2939 0.0419 -3.0250 55.9558 -0.1483 0.2021 0.4707 -0.7154 0.0606 -0.5923 -0.5930 0.2582 0.9512 -0.0963 0.0886 0.0108 -0.0600 0.0352 -0.1189 0.0431 0.0584 -0.0922 -0.0345 -0.0002 -0.0184 0.0459 0.0340 -0.0834 -0.1020 -0.0155 -0.0277 0.1425 0.0017 0.1399 -0.0116 -0.1513 0.0601 0.0328 0.1313 0.0132 -0.1794 0.0089 0.1899 -0.2074 0.1988 0.0884 0.0895 -0.0697 0.0218 0.0839 -0.2098 -0.0384 -0.0198 -0.0193 0.0351 0.0417 0.0102 0.0139 -0.0672 -0.0136 -0.0152 -0.0688 0.0020 0.0892 0.0008 -4.5963 12.9000 18.1299 -24.4355 1.3543 -38.9076 -18.8700 6.0512 58.0474 -8.3256 7.2516 1.1001 -5.5917 2.3116 -10.5874 3.6129 1.6906 -5.7721 -2.4606 -0.0089 -1.9466 3.9962 0.4629 -4.4675 -5.3829 -0.5165 -1.1059 7.0414 0.0722 2.2591 -0.4102 -2.6673 1.3178 3.1428 11.3601 1.0565 -8.7335 0.6316 12.0192 -8.7128 14.2248 3.5246 4.7031 -2.6310 1.1085 5.3421 -12.3558 -3.2648 -1.7041 -0.9950 1.9982 2.4126 0.5413 0.7082 -3.3650 -0.6889 -0.7824 -3.4252 0.1000 4.2383 3.6638 0.3642 0.1521 0.3744 -0.8948 0.4823 0.1248 -0.5560 -0.4990 0.8066 0.1125 -0.0960 0.0378 0.0203 0.0198 0.0245 -0.0035 0.0609 0.0252 -0.0210 0.0181 0.0002 0.0125 -0.3680 0.0283 0.0585 0.0833 -0.0270 0.0608 -0.0225 0.3741 -0.0950 -0.1578 -0.0628 -0.0583 0.0047 0.0052 0.0121 -0.0191 -0.0902 -0.1308 0.1314 -0.2421 0.0161 0.0885 -0.0719 0.0203 0.0163 -0.0122 0.0800 -0.0157 0.0017 0.0197 0.0000 0.0073 0.0192 -0.0480 -0.0119 0.0091 -0.0143 0.0529 -0.0005 19.6220 15.8588 21.5932 -51.0704 17.8172 11.5753 -25.8938 -14.9317 63.4342 21.7496 -14.8451 7.0907 3.3229 2.3144 3.9539 -0.5521 3.4713 3.1111 -2.9159 1.6962 0.0402 2.1329 -12.3125 2.9401 6.0083 5.4771 -2.0045 6.1933 -2.0421 12.5072 -6.8142 -6.1344 -2.9628 -10.1619 0.7370 0.7590 1.0422 -2.5858 -11.4794 -10.8070 17.5751 -21.8320 1.6353 6.2108 -7.0699 2.4412 1.8597 -1.9035 12.4224 -1.5870 0.1627 1.9279 0.0021 0.7234 1.8253 -4.7214 -1.1551 0.8821 -1.4177 4.9864 -4.2155 -0.0583 0.1558 0.1671 -0.4493 0.1983 -0.4623 0.1001 0.1954 0.1849 0.0474 -0.0266 0.0417 -0.0105 0.0108 0.0269 -0.0210 0.0267 -0.0607 0.0070 -0.0360 -0.0025 0.0025 0.0613 0.0452 0.0127 0.0949 -0.0175 -0.0355 0.0014 -0.2827 -0.0762 0.2774 -0.1091 0.0006 0.0549 -0.0030 0.0212 -0.0227 -0.1175 -0.0070 0.1257 -0.1045 0.0380 0.1320 0.0017 0.0470 -0.0495 -0.0148 0.0064 -0.0099 -0.0280 0.0239 0.0659 -0.0319 -0.0417 -0.0604 -0.0852 -0.0346 0.0002 0.0400 0.0026 -2.7401 16.0371 9.8758 -23.2270 7.8533 -36.5108 3.9368 5.2107 8.8782 6.0609 -3.6901 6.9864 -1.5532 1.1330 4.0369 -3.0459 1.2667 -6.3478 0.8351 -3.0672 -0.4552 0.3853 1.4713 3.9163 1.1259 5.4935 -1.0851 -3.2453 0.1175 -8.1687 -4.8848 8.9787 -4.6606 0.0951 7.5558 -0.3696 1.5968 -2.5934 -14.6908 -0.5910 15.3356 -8.5822 3.3469 9.4144 0.1614 5.2166 -5.1470 -2.0553 0.8807 -0.8746 -2.4128 2.1442 5.9784 -2.7956 -3.5125 -5.2185 -7.3048 -3.0589 0.0211 3.3859 21.0248 # Observ. # Pos. Obs. Scaled R2 R2 LogLikelihood Likelihood R. (NOB) (NPOS) (SRSQ) (RSQ) (LOGL) (LR) 354,795 64,209 0.0283 0.0304 -162,748.8 10,045.1 60,965 28,209 0.3511 0.3379 -30,769.2 22,637.5 221,304 174,608 0.0688 0.0700 -106,418 15,230.7 138,817 60,009 0.0346 0.0343 -92,531.7 4,824.1
55 Noted in Table 5-1 the last category coeffi cient within a dummy class is not shown since it is the negative sum of the categor ies estimated. In the case of household size, estimates for beef and poultry are not statistic ally significant when looking at 3-member household size while for pork it is not significant when ther e is just one person per household. Female age estimates are not statis tically significant for fish and pork, when comparing ages under 24 years and between 40 and 65 years old. Conversely, the estimates for female age between 25 and 40 years old are statistically significant and negative when looking at beef, fish and pork; and positive when l ooking at poultry. The effect of the presence of children is not stat istically significant on br and preference when looking at the beef, fish and poultry estimates ; but it is statisti cally significant and negative in the case of the estimate for por k. In the case of female employment, the coefficient for beef is not statistically significant for part-time employees; while for fish and pork the estimates are not statistica lly significant when looking at full time employees. In the case of poultry, both estimat es are not statistically significant. With respect to occupation levels they are not statistically significant in most of the fish estimates, as for example in the case of professionals, sales, craftsman, and farmers. In the case of pork, the estimate for clerical is not statistically significant as well as the ones for fish. For beef, estimates for laborers are not statistically significant. On the other hand, for proprietors, operatives, service work ers, and students estimates are statistically significant. Education estimates for college gr aduates are not statistically significant for fish, poultry and pork; and when looking at th e respondents with some college education, estimates are not statistically significant for beef and poultry. Where in the country (region) also has an effect across meat cat egories. For beef, estimates for households
56 located in the Mountain region ar e not statistically significant, while estimates for the rest of the country are significant. In the case of fish, estimates for households located in the Mid-Atlantic region are not statistically significant compar ed to the rest of the country. Poultry and pork estimates are not statistica lly significant for households located in New England, as well as households located in th e West South Central Region and West North Central for poultry and pork resp ectively. Estimates for market vary according to the type of meat and the number of people living in a particular market. For beef, markets with less than 249,000 people and markets betw een 1,000,000 and 2, 499,999 people are not statistically different from the average. For fish and pork, the smaller markets (less than 249,000 people) and the larger markets (more than 2,500,000 people) ar e not statistically significant. In the case of poultry, markets between 500,000 and 1,000,000 people, as well as markets between 1,000,000 and 2, 499,9 99 people are not statistically significant from the average. All other markets in all meat categories are significant. The time of the year (months) also called seasonality, pres ents estimates that seem to be of particular relevance for bra nd preference on pork products, with significant estimates across all meat cate gories. On the other side, the months of January and, particularly, October present estimates that are not statistically significant from the average for meat brand identity. The vast majo rity of the estimates for type of store, where in the store, and relative prices are sta tistically significant, and most of then are positive, indicating a strong effect on brand identity across the four meat types. The R2 values presented in Table 5-1, show that less than 3.5 percent of the variations in brand iden tity have been explained within the limits of the Probit models for beef, poultry and pork models; and less than 34 pe rcent in the case of the fish model. As
57 discussed before, higher values for goodness-to -fit are always preferred, although it is typical in cases of large cross sectio nal individual data to have low R2 values. The likelihood ratio test in each model shows that the combined effects of the independent variables are statistically different from zero. Pooled Probit Brand Preference Model In this section, a pooled Probit model is estimated with the four meat categories combined into one model using a dummy variab le to capture the m eat type effect. As specified earlier in equation (4-13), the presen ce of the price and time trend variables in the model allows one to measure the change s in the likelihood of brand identification among meat products for the average household or for any combination of demographics. Table 5-2 presents a descrip tion of each explanatory vari able with its corresponding coefficient and significance te st, as well as the supportive statistics. With only a few exceptions, all variables are statistically signi ficant, indicating the relevance of each of the variables in the model influencing the pr obability of buying meats by brand. Noted in Table 5-2 the last category coefficient within a dummy class is not sh own since it is the negative sum of the categories estimated. For ex ample, in the case of the type of store, supermarkets, warehouses, and Supercenters the estimates are statistic ally significant and positive indicating an important response to br and preferences relative to the average consumer, while in butchers or small stores th e response is negative. With respect to the location within the store, Ta ble 5-2 clearly shows a strong response to branded products in the freezer section. On the other side, the response to brand identification in the fresh case section and in the deli section is marked ly negative. Relative price also show its importance which means that there is a st rong relationship betw een prices and the likelihood of buying branded products.
58 Table 5-2: Pooled probit model estimates for brand preferences Variable Description Prob. Coef. t-value Variables Description Prob. Coef. t-value Intercept Store Type STTY1 STTY2 STTY3 STTY4 STTY5 Where STWR1 STWR2 STWR3 STWR4 STWR5 Price PRCP Income INC1 INC2 INC3 INC4 Hwd Size HWZ1 HWZ2 HWZ3 HWZ4 Female Age AGF1 AGF2 AGF3 AGF4 Children CHD Employm. EMF1 EMF2 EMF3 Occupation OCC1 OCC2 OCC3 OCC4 OCC5 OCC6 OCC7 OCC8 OCC9 OCC10 OCC11 OCC12 Avg. Household (binary) Supermarkets Warehouse/Club Butcher/market Supercenters All others store (binary) Fresh case Deli/food bar Gourmet Freezer All other (Index=1.0) Relative price (Annual $) 0 $24,999 $25 $49,999 $50$74,999 $75,000 plus (members) one two three four plus (Years of age) Under 24 yrs. 25 to 40 yrs. 40 to 65 yrs. Over 65 yrs. Children present (binary) Full time Part time Not employed (binary) Professional Proprietor Clerical Sales Craftsman Operative Military Service worker Farm Student Laborers Retired / Un-emp. -0.22120 0.14683 0.25950 -0.54842 0.19849 ----0.28816 -0.32420 0.00209 0.67134 ---0.05400 -0.01521 0.04010 0.00013 ---0.02101 -0.01523 0.00044 ---0.10134 -0.02644 -0.03096 ----0.01032 0.00159 0.01354 ----0.15859 0.04083 0.04145 0.09871 -0.02216 0.00364 -0.00821 -0.09359 -0.03301 0.07294 -0.03974 ----22.41473 32.09044 31.78620 -59.79004 17.05465 -48.13601 -25.23275 0.11777 90.16716 18.31284 -4.50572 14.49878 0.04165 4.70795 -4.79892 0.13663 11.18811 -6.27465 -8.31986 -1.87517 0.62298 4.44863 -8.48606 7.90268 8.03541 12.57888 -3.02236 0.71144 -1.43005 -6.32814 -4.37264 4.96123 -3.48388 Education EDU1 EDU2 EDU3 EDU4 Regions STA1 STA2 STA3 STA4 STA5 STA6 STA7 STA8 STA9 Market Size MSZ1 MSZ2 MSZ3 MSZ4 MSZ5 MSZ6 Months MTH1 MTH2 MTH3 MTH4 MTH5 MTH6 MTH7 MTH8 MTH9 MTH10 MTH11 MTH12 Category PCAT1 PCAT2 PCAT3 PCAT4 Meat x Prices CATPR1 CATPR2 CATPR3 CATPR4 Time Trend TT TPCAT1 TPCAT2 TPCAT3 TPCAT4 (level) High Sch./less Some College College Grad Post Grad (binary) New England Mid Atlantic E. N Central W. N Central S Atlantic ES Central WS Central Mountain Pacific (1000 pop) 50-249 250-499 500-999 1,000-2,499 2,500+ Non-size (binary) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec (binary) Beef Fish Poultry Pork (Indexed Price) Beef Price Fish Price Poultry Price Pork Price (Integers) 1-96 Beef Time Fish Time Poultry Time Pork Time -0.03405 0.02481 0.01540 ----0.00390 -0.01091 -0.06325 0.02467 0.12079 -0.17138 -0.00915 0.05268 ----0.00391 0.03523 -0.00857 -0.01390 0.02302 ----0.00921 0.00812 0.05873 0.01017 -0.00959 -0.02043 -0.02105 -0.03793 -0.01034 -0.00309 0.01108 ----0.89330 -0.12251 1.04462 ---0.00062 -0.00083 -0.00685 ---0.00199 0.00313 -0.00132 -.00223 ----11.30524 7.52299 4.21594 -0.64824 -2.70761 -15.93089 4.13970 31.74564 -28.06583 -1.77465 7.77187 -0.76484 8.29306 -1.89928 -4.17442 6.86667 -1.75305 1.51617 11.33161 1.90521 -1.81928 -3.79020 -3.94837 -7.09436 -1.95400 -0.58196 1.96363 -158.19035 -13.30549 166.95453 2.09833 -2.30709 -14.09945 28.67925 33.22278 -8.58871 -21.38148 Period: 1992:9 2001:8 t-table at 5 percent = 1.96 Number of observations = 775881 R-squared = 0.30833 Num. of positive observations = 327035 Log of likelihood function = -399674.11189 Scaled R-squared = 0.3158 Likelihood ratio test for 0 slopes = 257047.92180
59 It is evident that in larg e food centers consumers are mo re exposed to store brands or national branded products (i.e., SamÂ’s choi ce versus Tyson) than in small size stores. In addition, the location with in the store tells something about the relationship between product forms and the level of branding in that particular section. For example, in the freezer section of any food store, consumers ar e more likely to be exposed to two, three or may be four different brands of frozen chicken nuggets or strips (e.g., Publix, Perdue, Sanderson Farms, etc). Among the demographic variables, fe male age is a good example of the significance of the estimates. As seen in Table 5-2, as the female gets older the response to branded products decrease while among fema les younger than 24 year s old the level of response is considerably positive and increa ses. At lower levels of education, the likelihood of brand identificati on decreases considerably. On the other hand, as education level increases the likelihood of buying branded product also increase. Those estimates are very revealing in terms of life styles and might explain the tendency among young people of relate branded produc ts with fashion trends as result of heavy advertising campaigns. The corresponding st atistics show a R2 value of almost 32 percent which represents the percentage of variation explained by th e Probit model. As discussed before, higher values for goodness-of-fit are always preferred although, it is typical in cases of large pooled data to have low R2 values. In Chapter 6, coefficient estimates from Table 5-2 and their respective signs will be used to simula te the probabilities of brand identification among meat products.
60 CHAPTER 6 BRAND PREFERENCES SIMULATIONS Concept of Probabilities and Distribution The decision-making process of buying meat products by brand is based on consumersÂ’ perceptions about the productÂ’s at tributes and consumersÂ’ experience with the branded product. The models developed in early chapters provided empirical insight into the importance of meat brands using Probit models. Probit models are commonly used to estima te the probability of an event or how likely the event is to occur. As presented in Chapter 4, Probit models are developed first in terms of the regression of the latent vari able, equation (4-1). Th e latent variable is related to the observed, binary variable in a si mple way. If the latent variable is greater than some value, the observed variable is 1; otherwise it is 0, e quation (4-2). The model is linear in the latent variab le but the resulting probabiliti es are non linear since changes in any of the independent variables give points on a cumulative normal distribution varying form zero to one, equations (4-14) and (4-15) . By definition, the marginal response to any change among the right ha nd-side variables differs depending on the starting point on the cumula tive normal distribution. The predicted probabilities represent the li kelihood of brand iden tity for each meat relative to the average consumer, considering all the explanatory variables. While it is convenient to express the probability for one va riable with everything else set to the average consumer, there is nothing to prevent one from showing the marginal effects for any combination of the consumersÂ’ characteris tics. Using the average consumer is just a
61 matter of convenience. Figures 6-2 through 6-9 show the proportio n of variation in relation to the probability for brand identif ication of the average consumer for that particular variable. Based on these proportions la ter in this chapter, a detailed ranking of impacts will provide a simple and intuitive understanding of the importance of each variable in the likelihood of buying branded meat products. In the following section the probabilitie s of selecting branded meats are shown across the full range of variables incl uded in Table 5-1. Specifically, let be the cumulative normal distribution, then the pr obability of selecting the branded meat j for the average consumer is ) , (0 0 j where 0 is the intercept from Table 5-1 and j is the coefficient for beef, fish, pork, and poultry. Then let ) , , (0 j be the point in the cumulative normal after changing any variable included in the Probit model from the Table 5-1. The difference between and0 is the impact of a variable relative to the average consumer. Sinceo Â’s differ by the meat type, a us eful approach to compare the variable impacts is to show these differen ces as will be shown in the next several sections. For example, in Figure 6.1 a cumulative normal is shown for the average consumer and the defined time period. Note in Figure 6-1 that the probabilities shown by meat type are those estimated from the pr obit model for the aver age household in 2000. Beyond these points on the four cumulativ e normal distributions, the remaining distributions are drawn for illustration purposes for any Xj. As will be shown subsequently, the rate of change and, hen ce, the cumulative dist ribution depend on each explanatory variable. Next let th e variable change be either Xi (decline) or Xj (rise) and all other variables remain at the mean or aver age consumer. The potential differences across the four meats are easily seemed.
62 Figure 6-1: Normal distribution a nd probability of brand purchase. Probability of Brand Identification by Demographics In this part of the chapter, the impacts of the different dem ographic variables on brand identification are analyzed. In order to show which variab les are driving brand identification for each meat category, the diffe rent demographics will be ranked in terms of their relative effects on br and preferences. It is impo rtant to mention that some variables may be statistically significant but numerically unimportant when predicting probabilities . Based on Table 3-1 and equation (4.13) whic h define and specify the variables of the Probit model, household demographics can be grouped as follows: income, household size, age, presence of children, employment status, education, occupation, regions of the country, and market size. Acco rding to the estimates presented in Table 52, almost all demographic vari ables were statistically signif icant. However, numerically the ranges of impacts on the likelihoods of buying meat products by brand were very small in relative terms for some of the variables. With the exception of the region,
63 occupation, and female age, all demographics were in the range of Â± 2 percentage points from the average consumer in terms of the pr obability of brand iden tification. That Given this circumstance, only the simulations for those variables that show the more substantial impacts on the probability for brand identif ication will be presented in the following figures. Recall that for beef, fish, pork, and poultry the average probabilities are 28 percent, 41 percent, 52 percent, and 80 percent respectively as illustrated in Figure 6.1. N e w E n g l a n d M i d l e A t l a n t i c E . N o r t h C e n t r a l W . N o r t h C e n t r a l S o u t h A t l a n t i c E . S o u t h C e n t r a l W . S o u t h C e n t ra l M o u n t a i n P a c i f i c0.00 0.02 0.04 0.06 -0.02 -0.04 -0.06 Change in probability of brand preference by region Beef N e w E n g l a n d M i d l e A t l a n t i c E . N o r t h C e n t r a l W . N o r t h C e n t r a l S o u t h A t l a n t i c E . S o u t h C e n t r a l W . S o u t h C e n t ra l M o u n t a i n P a c i f i c0.00 0.02 0.04 0.06 -0.02 -0.04 -0.06 Change in probability of brand preference by region Fish N e w E n g l a n d M i d l e A t l a n t i c E . N o r t h C e n t r a l W . N o r t h C e n t r a l S o u t h A t l a n t i c E . S o u t h C e n t r a l W . S o u t h C e n t ra l M o u n t a i n P a c i f i c0.00 0.02 0.04 0.06 -0.02 -0.04 -0.06 Change in probability of brand preference by region Pork N e w E n g l a n d M i d l e A t l a n t i c E . N o r t h C e n t r a l W . N o r t h C e n t r a l S o u t h A t l a n t i c E . S o u t h C e n t r a l W . S o u t h C e n t ra l M o u n t a i n P a c i f i c0.00 0.02 0.04 0.06 -0.02 -0.04 -0.06 Change in probability of brand preference by region Poultry Figure 6-2: Change in probabilities of purch asing by brands according to country regions. Regions or where in the country, present substantial differences across the country with the same set demographic variables. As shown in Figure 6-2, for each type of meat the South Atlantic region (SAR) and the East South Central region (ESCR) show nearly opposite values in terms of the probability of br and identification. In all cases, the SAR is positive and close to 4 percentage points a bove the average while, the ESCR is negative
64 and 5 percentage points lower. All other regi ons of the country ha ve a variation range close to Â± 2 percentage points from the average probability for each type of meat. In Figure 6-3, occupation of the respondent presents a wide range of variations for all meat types. For beef, fish and pork the range in probability for brand identity is 10 percentage points, while for poultr y the range is about 8 units. Professional Service Worker Laborers Farm Craftsman/Forman Military Operative Propietor, Manager Clerical Student Retires, Unemployed Sales 0.000.020.040.06 -0.02 -0.04 -0.06 Change in probability of brand preference by occupation Beef Professional Service Worker Laborers Farm Craftsman/Forman Military Operative Propietor, Manager Clerical Student Retires, Unemployed Sales 0.000.020.040.0 6 -0.02 -0.04 -0.06 -0.08 Change in probability of brand preference by occupation Fish Professional Service Worker Laborers Farm Craftsman/Forman Military Operative Propietor, Manager Clerical Student Retires, Unemployed Sales 0.000.020.040.06 -0.02 -0.04 -0.06 -0.08 Change in probability of brand preference by occupation Pork Professional Service Worker Laborers Farm Craftsman/Forman Military Operative Propietor, Manager Clerical Student Retires, Unemployed Sales 0.000.020.040.0 6 -0.02 -0.04 -0.06 Change in probability of brand preference by occupation Poultry Figure 6-3: Change in probabilities of pur chasing by brands according to occupation. In all cases, professionals present the lowe st probability of buying meat products by brand, close to 6 percentage points belo w the average probability, while the group including retired and unemployed respondents and respondent working in sales show the largest probability, with a range between 3 and 4 percentage points above the average probability. Finally, it is clear that in some occupations such as craftsman/foreman, military, operative, proprietor/manager, a nd clerical, respondents show very little
65 difference from the average consumer. Th e negative impact among professionals was quite surprising. Age of the female also presents a signifi cant range of variations in terms of the probabilities of buying branded meat products. Figure 6-4 clearly shows that as females get older the probability of br and identification decreases. Fo r all meats, female under 25 years old presents a clear tendency for br anded products showing a probability range between 3 and 4 percentage points above the average probability. Under 25 years 25 to 40 years 40 to 65 years 65 years or more 0.000.020.04 -0.02 -0.04 Change in probability of brand preference by household age Beef Under 25 years 25 to 40 years 40 to 65 years 65 years or more 0.000.020.04 -0.02 -0.04 Change in probability of brand preference by household ag Fish Under 25 years 25 to 40 years 40 to 65 years 65 years or more 0.000.020.04 -0.02 -0.04 Change in probability of brand preference by household age Pork Under 25 years 25 to 40 years 40 to 65 years 65 years or more 0.000.020.04 -0.02 -0.04 Change in probability of brand preference by household a g Poultry Figure 6-4: Change in probabilities of purch asing by brands according to head of the household age. For the rest of the female ages, from 25 years and above, the probability of buying branded products is reduced to levels sl ightly below the average. Although all the demographic variables have a statistical significance in the pool ed Probit model (see Table 5-2), for marketing purposes those that have a numerical relevance, in particular
66 those that show a marked negative response to brand identification should be the focus of target marketing, promotions and advertisi ng. Professionals locate d in the East South Central region could be a prime target for focu sing marketing efforts if the objective is to gain more brand identity. Note, however, that for industries like beef or pork the goal is probably simple demand growth regardless of the brand. Probability of Brand Identifica tion by Non-Demographics This part of the chapter deals with th e impact of the different nondemographic variables on the probability of brand identifi cation with respect to the average consumer, using the same concept adopted in the previo us section. The resulting simulations are used to show the effect of variables such as type of outlet, where in the store, price, and seasonality on householdsÂ’ brand preferences for each meat type. Table 6-1 is presented to show the distri bution of meat purchases among outlets. It is clear that most of the meat consumed in the U.S. is bought in Supermarkets. Table 6-1: Distribution of meat purchases by outlet (share by outlet) Beef Fish Pork Poultry Supermarkets 81.65 73.44 86.26 83.05 Warehouse/Clubs 5.09 6.15 5.69 5.48 Butchers 5.53 8.67 2.47 4.58 Supercenters 1.63 1.23 1.20 1.47 Other stores 6.10 10.51 4.37 5.42 According to the NPD panel data used in this research, more than 81 percent of the beef, 73 percent of the fish, 86 percent of the poultry, and 83 percent of the pork were purchased in supermarkets. On the other hand, purchases made in supercenters represent no more than a one percentage point for ea ch meat type. In all other outlets the percentage of purchases did not exceed 6 perc ent for beef products, 9 percent for fish, 6 percent for poultry, and 5 percent in the case of pork products.
67 Among all the variables included in the Pr obit model, store t ype was the most important in terms of its impact on the probabi lity of brand identification. In Figure 6-5, the likelihood of buying meat products by bra nd is estimated across the five types of outlets (see Table 5-2). Butcher/Meat Mkts. All other outlets Supermarkets Supercenters Warehouses/Clubs 0.000.10 -0.10 -0.20 Change in probability of brand preference by store type Beef Butcher/Meat Mkts. All other outlets Supermarkets Supercenters Warehouses/Clubs 0.000.10 -0.10 -0.20 Change in probability of brand preference by store ty p Fish Butcher/Meat Mkts. All other outlets Supermarkets Supercenters Warehouses/Clubs 0.000.10 -0.10 -0.20 Change in probability of brand preference by store type Pork Butcher/Meat Mkts. All other outlets Supermarkets Supercenters Warehouses/Clubs 0.000.10 -0.10 -0.20 Change in probability of brand preference by store ty p Poultry Figure 6-5: Change in probabilities of pur chasing by brands according to store type. As expected, consumers buying in warehous es, supercenters, and supermarkets showed positive responses to brand identification while, butcher and meat markets customers indicated a strong nega tive response to brand iden tification. In warehouses, the probability of brand identification for beef, fi sh, and pork is approximately 10 percentage points above the average and 6 points for poultry. The probability of buying branded meat products in places like butcher and meat markets is well below the average levels for each meat category, reaching level between minus 15 and minus 20 points. These
68 results are of particular rele vance because they pr esent important aspect s about the nature of the industry in terms of product distribu tion and the structure of the industry. The location within the store is another important variable in terms of brand preferences, because by knowing from which pa rt of the store the purchase was made one can identify relatively easily what product form (i.e., frozen or fresh) and what brand attributes are preferred by consumers. Usi ng the data set provided by NPD, Table 6-2 was created to show the di stribution of purchases within outle tÂ’s sections. In the case of beef and pork most of the purchases were made in the fresh meat section representing 78 and 77 percent respectively. In the case of poultry and fish most of purchases were made in the fresh and freezer sections. For poultry prod ucts the percentages were 64 and almost 19 percent respectively, and fo r fish products of the percen tages were 38 and 32 percent respectively. In all other sections the purch ases for beef, fish, poultry, and pork did not exceed 4 percent, 8 percent, 3 perc ent, and 5 percent respectively. Table 6 -2: Distributi on of meat purchases by location in the store (share by section) Beef Fish Pork Poultry Fresh meat case 88.34 47.79 87.33 70.73 Deli/ Food bar 0.43 1.56 1.20 0.78 Gourmet section 0.29 0.71 0.58 0.29 Freezer section 4.35 38.69 3.40 23.33 Other stores 6.60 11.25 7.48 4.87 Figure 6-6 clearly shows that in the case of beef, fish and pork, the probability of buying by brand increased more than 20 percen tage points when the purchase was made in the freezer section, while for poultry products the probability is about 13 points above the average. On the other hand, when the purchase was made in the deli and food bar or in the fresh meat section the probabilities of brand identification for all meats were under the average. For beef and poultry products, the probabilities of buying by brand in these
69 sections were close to 10 a nd 8 percentage points below the average, while for fish and pork products the probabilities were roughly 12 percentage units below the average in both sections. Deli and Food bar Fresh Meat Case All other locations Gourmet Sections Freezer Section 0.000.100.20 -0.10 Change in probability of brand preference by where in store Beef Deli and Food bar Fresh Meat Case All other locations Gourmet Sections Freezer Section 0.000.100.20 -0.10 Change in probability of brand preference by where in sto r Fish Deli and Food bar Fresh Meat Case All other locations Gourmet Sections Freezer Section 0.000.100.20 -0.10 Change in probability of brand preference by where in store Pork Deli and Food bar Fresh Meat Case All other locations Gourmet Sections Freezer Section 0.000.100.20 -0.10 Change in probability of brand preference by where in sto r Poultry Figure 6-6: Change in probabilities of purch asing by brands according to location within the store. Figure 6-7 presents the effect of relative prices on the probability of buying meat products by brand. In equation (4 -13) price and meat categorie s were interacted in order to measure if the effect of prices on brand identification differed across the four meats. Note that in Table 5-2, these interactions are shown in the lower right portion of the table. In order to analyze the impact of prices on each meat category, a price index relative to the average price over the entire data set wa s used with the indexe d price taking a value of 1.0 for the average price.
70 0.10.30.50.220.127.116.11.18.104.22.168.22.214.171.124 0.00 0.02 0.04 -0.02 Change in probability of brand preference by relative prices Beef 0.10.30.50.126.96.36.199.188.8.131.52.32.52.72. 9 0.00 0.02 0.04 -0.02 Change in probability of brand preference by relative prices Fish 0.10.30.50.184.108.40.206.220.127.116.11.18.104.22.168 0.00 0.02 0.04 -0.02 Change in probability of brand preference by relative prices Pork 0.10.30.50.22.214.171.124.126.96.36.199.188.8.131.52 0.00 0.02 0.04 -0.02 Change in probability of brand preference by relative prices Poultry Figure 6-7: Probabili ties of purchasing by brands with pr ices indexed to 1.0 for the mean. The light colored bars in each meat pri ce represent the 95 percent confidence range. Indexing was necessary since pri ce levels differ in absolute value across the four types of meats. Consequently, prices above or belo w one indicate, in percentage points, the fluctuation of relative prices from the average. As seen in Figure 6-7, for all categories as prices increase the probability of buying branded meat increases. In the case of beef, fi sh and pork the range of variation from the average is between approximately 2 points below and around 4 percentage points above the average value of 28 percent, 41percent and 52 percent, respectively. For poultry products, the range of variation from the aver age is very small between 1 point below the 80 percent average probability and 2 percenta ge points above the average. Even though prices are a decisive factor on the level of expenditures on meat products, they have a small impact in the likelihood of buying brande d products even over a very wide range of
71 relative price levels. Figure 6-8 presents the effects of seasonality on the probability of purchasing meat products by brand. JanFebMarAprMayJunJulAugSepOctNovDec 0.00 0.02 0.04 -0.02 -0.04 Change in probability of brand preference by season Beef JanFebMarAprMayJunJulAugSepOctNovDec 0.00 0.02 0.04 -0.02 -0.04 Change in probability of brand preference by season Fish JanFebMarAprMayJunJulAugSepOctNovDec 0.00 0.02 0.04 -0.02 -0.04 Change in probability of brand preference by season Pork JanFebMarAprMayJunJulAugSepOctNovDec 0.00 0.02 0.04 -0.02 -0.04 Change in probability of brand preference by season Poultry Figure 6-8: Change in probabilities of pur chasing by brands according to seasons. Seasonality represents consumersÂ’ consum ption habits across the year and it is closely related to customs of the different population sectors and special dates during the year (Medina and Ward, 1999). The probability of buying branded meat products across the months of the year is quite small even during the major holidays such as Christmas, Thanksgiving, or July 4th, showing a spread of no more than 4 percent for all meats. However, fish and pork products seems to be a little more inclined to a positive brand identification during the month of March, which represents the end of the winter season in most parts of the country and consequently an increase of outdoor activities. During
72 the rest of the year, there are no signs of ma jor seasonality effect on the probability of buying branded meat product. Structural Change in the Brand identity In this section of the chapter an anal ysis of the structural change among householdsÂ’ brand preferences will be c onducted to compare the trends among meat products. During the time frame of this rese arch consumersÂ’ preferences for branded products have changed as well as the number of products present in the market. From Chapters 4 and 5, equation (4-13), and Table 51 provide the elements required to analyze these changes and calculate the probabil ities of brand recognition across time. Accordingly, the probabilities are estim ated by each time period for the average consumer as specified in the equations of the Probit Model. For the average consumer, the probabi lity of brand recognition among poultry products has changed very little across time remaining close to 80 percent. For fish products, this probability has changed around 3 pe rcentage points from 38.1 percent to 41 percent. In the case of beef and pork products this change has been larger than the other two, presenting an increase of more than 13 percentage poi nts over the same period with beef ranging from 14.6 percent to 27.6 percent, and almost 9 percentage points increase in branded pork products with an increase fr om 43.6 percent to 52.4 percent (Figure 6-9). These probabilities have a particular importa nce when analyze the structural changes across the meat industry. In the case of b eef and pork products, even though almost 72 percent of all beef purchases and almost 48 percent of all pork purchases were still without brand preferences, th is boost in brand identification among consumers clearly shows the efforts of those two sectors in tr ansforming their industrie s and migrates from a
73 commodity sale perspective to wards a more marketing orient ed industry, the potential for growth among branded products, and the importance of information programs. Figure 6-9 also presents a ra nking of the industry in term s of brand concentration, showing the poultry industry as the most branded or differen tiated sector while the beef industry is the less branded portion of the industry. On th e other hand, little room for branded growth is expected for poultry produc ts since poultry is al ready a highly branded commodity. 199219931994199519961997199819992000Years0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Users indicating brand identification (percent)BeefFishPorkPoultry Beef14.6%15.6%17.1%18.7%20.4%22.1%24%25.9%27.6%Fish38.1%38.3%38.6%38.9%39.2%39.6%39.9%40.2%40.4%Pork43.6%44.4%45.6%46.7%47.9%49.1%50.3%51.4%52.4%Poultry80%79.9%79.8%79.7%79.6%79.5%79.4%79.3%79.2% Figure 6-9: Brand identif ication across the time. Source: NPD panel data. Within the industry, brand preferences among meat products have shown different trends during the time frame of this research . For the average consumer, the probabilities for selecting meat products by brand are estimated across time, with September 1992 being the starting period of the data base. In Figure 6-10, beef a nd pork products clearly present the largest increase in brand identif ication. Beef brand id entification doubled in percentage over the period 1992 to 2000 with a y early average of 1.5 percentage points increase, while pork brand identification incr eased an average of one percent per year
74 during the same period. Fish pr oducts show a very small increas e of less than a quarter of percentage point in brand r ecognition while poultry products show a negative trend of around one percent redu ction since 1992. Figure 6-10 show again the importance of marketing practices and the vast potential for promotion initiatives (e.g., generi c advertising programs) within the beef and pork industry and that within the poultry sector, however, due to high levels of branding little room for increase in bra nd identification is expected. 91236912369123691236912369123691236912369 1992 1993 1994 1995 1996 1997 1998 1999 2000 Years0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 Preference Index (Year 1992=1.0)PoultryFishPorkBeef Figure 6-10: Change in brand identification for all meats across time. Source: NPD panel data. Ranking of Variables In order to compare the dr iving factors affecting cons umersÂ’ brand identification among meat products, the impacts of the di fferent exogenous variables were ranked according to the resulting in the probability of buying branded meats. Thus, based on equation (4-13) and coefficien t estimates from Table 5-2, a ranking of the impacts among the variables presented in the model was cons tructed in order to compare the changes in
75 preferences relative to the av erage consumer. Recall that for beef, fish, pork, and poultry the averages were 28 percent, 41 percent, 52 percent, and 80 per cent, respectively. Figure 6-11 represents the minimum and maximum range from the average probability for a brand purchase, as well the di fferences between the most negative effect and the most positive effect of each variab le. The differences were then sorted in descending absolute magnitude, thus giving a quick way to rank the impacts of each variable included in the probit model (Table 5.2). Reading the figures clockwise, it is clear that the factors having the most impact on every meat category are the nondemographic variables, where Â“in-the storeÂ” and the Â“type of storeÂ”. In the case of beef, fish, and pork products the probability range of buyi ng branded products according to the Â“location within the storeÂ” is ranked first with a 35 percenta ge point range. For poultry products, this variable is ranked s econd showing a range of 25 percentage points in the likelihood of buying poultry by brand and is well below the other three meats. With respect to the type-ofstore, for poultry products this vari able is ranked first showing a 25 percentage range in the probability of buying branded products which is less than the range for fish and pork (29 a nd 31 percentage respectively) and equal to beef products. Adjustments over time are of particular re levance for beef ranking third among the variables while for pork pr oducts adjustments over time are ranked fifth. For fish products and, in particular, for poultry products the time adjustments have little to no relevance. Figures 6-9 and 6-10 show the in creasing trend in the probability of buying branded beef and pork products across time. The last nondemographic variable ranked is price. As mentioned earlier, in order to analyze the impact of prices on each meat category, a price index relative to the average
76 price was used to compare the effect of this variable on th e probability of buying branded meat products. In the case of fish, price is ranked fifth while for all other meats it is ranked sixth. Generally one can conclude that while price does impact the likelihood of buying according to brand prefer ences, this variab le still is of considerable less importance among all meats. Also, the relative importance of price within each meat type was quite similar being either fifth or si xth in importance for all four meats. Among demographic variables, regions of the country, occupation of the respondent and female age are the most im portant in terms of the impact on brand selection with an average range of 10 per centage points across the regions, 9 units for occupations; and 5 percentage points over the female age range. For fish, poultry and pork regional differences and occupations are ranked in third and fourth places, respectively, while for beef the demographi cs are ranked in fourth and fifth places, respectively. Considering the age of the female responde nt, for fish and pork this variable is ranked fifth; age ranked in sixth place for poultry and seventh place for beef. As mentioned before, the probability of buying br anded meat decreases as the respondent age decreases and female respondent gets ol der. Beyond this point of the ranking, the impacts of the rest of the variables in terms of the likelihood of purchasing branded products is relatively small showing ranges in the neighborhood of Â± 2 percentage points from the average.
77 Where in Store Store Type Time Regions Occupation Price Age of Female Seasons Market Size Income Level Education Household Size Employment Children0.000.100.200.30 -0.10 -0.20 -0.30Minimum and Maximum Range from the average Probability for Brand Identity 0.35 0.25 0.14 0.10 0.09 0.05 0.05 0.03 0.02 0.02 0.02 0.01 0.01 0.01 0.000.100.200.300.40Range in the Probability for Brand IdentityBeef Where in Store Store Type Regions Occupation Price Age of Female Seasons Market Size Income Level Time Education Household Size Employment Children0.000.100.200.30 -0.10 -0.20 -0.30Minimum and Maximum Range from the average Probability for Brand Identity 0.38 0.29 0.11 0.10 0.06 0.06 0.04 0.03 0.03 0.02 0.02 0.01 0.01 0.01 0.000.100.200.300.40Range in the Probability for Brand IdentityFish Where in Store Store Type Regions Occupation Time Price Age of Female Seasons Market Size Income Level Education Household Size Employment Children0.000.100.200.30 -0.10 -0.20 -0.30Minimum and Maximum Range from the average Probability for Brand Identity 0.37 0.31 0.12 0.10 0.09 0.07 0.06 0.04 0.03 0.03 0.02 0.01 0.01 0.01 0.000.100.200.300.40Range in the Probability for Brand IdentityPork Store Type Where in Store Regions Occupation Age of Female Price Seasons Market Size Income Level Education Household Size Employment Time Children0.000.100.200.30 -0.10 -0.20 -0.30Minimum and Maximum Range from the average Probability for Brand Identity 0.25 0.24 0.08 0.07 0.04 0.04 0.03 0.02 0.02 0.02 0.01 0.01 0.01 0.01 0.000.100.200.300.40Range in the Probability for Brand IdentityPoultry Figure 6-11: Ranking of the Probability of buying branded products by meat category. Probabilities of Brand Preference: Upper and Lower Limits According to the analysis pr esented earlier in this chapte r, it is obvious that brand identification exists, that consumers are us ing brands as indicat ors of consistency, convenience, and reliability, and that some sectors (e.g., beef and pork) have shown a positive and increasing trend in the level of consumersÂ’ response towards brand identification (Figure 6-9). In addition it is k nown that almost al l of the variables
78 included in the Probit models are statistica lly significant and that their effects on the likelihood of brand preferences varied accordin g to the type of meat being considered. Under these circumstances, Figure 6-12 was prepared to illustrate the conditions under which the upper and lower limits of th e probability of brand preferences for each meat category are most likely to occur. Hence results from the simulation analysis presented earlier in this chapter were used to calculate the upper and lower limits of the model. The combination of va riables leading to upper limit for all meat categories was warehouse/clubs, freezer section, income level between $25,000 and $50,000, one member household, age of the respondent under 25 years, some college education, no children in the household, respondent working in sales, living in the South Atlantic region, market size between 250,000 and 499,000 people, and the month March. On the other hand, the combination for the lower li mit was butcher/meat market, deli/Food bar, professional, income level over $75,000, two member household, age of the respondent over 65 years, high school or less, presence of children in the hous ehold, respondent not employed at the moment of the survey, living in the East South central region, in a nonmarket-sized town and the month of A ugust. In this figure, the ranges in the probability of buying by brand are simulated us ing these combinations of demographics and nondemographic variables. The largest sp read between limits is for fish and pork products, with an 80 percentage point range of variation respectivel y, followed by beef and poultry with 76 points and 63 percentage points of variation, respectively. Even though the upper and lower limits for the beef sector represent the lowest values among all meats, more than 51 percent of the purchases were made above the average
79 representing the highest range compared to th e rest of the sectors and consequently the largest potential for relative growth in terms of brand identification. 79.42 87.81 92.42 98.74 3.66 7.39 11.92 35.45 Beef Fish Pork Poultry010203040506070809010080 52 41 28 Figure 6-12: Upper and lower limits to the probabilities of bra nd identity when purchasing each meat. Conversely, the poultry sector presents the highest values for upper and lower limits, but only 18 percent of the purchases we re made above the av erage showing little room for brand expansion primarily because it is already a hi ghly branded sector. Looking at the lower limits, the poultry sector presents a 35.4 per cent probability of brand selection, while for the b eef sector this value is just 3.66 percent, showing again the degree of branding among poultry products and the characteris tics of the industry. In addition, comparing the poultryÂ’s lower limit probability to the average probability of
80 brand preference for the beef industry, this lower limit is still considerably higher than seen for the lower range in beef. Figure 6-12 clearly shows th e differences in brand identification among the four meats and helps to identify marketing targets in order to achieve higher levels of brand purchases if that is a desira ble goal. Differences of stru cture among the i ndustry sectors are still affecting what is happe ning in terms of brand. In the case of beef, for example, even over 70 percent of purchases are still unbranded. From beef i ndustry perspective, and for that matter any of the sectors, the prevailing goal is to ach ieve growth through providing safe products that meets consum ers demand for convenience, variety and availability while remaining competitive. If branding facilitates these goals it should be pursued. This study shows the range of brandi ng for each meat sector but has not drawn conclusion about the relative merits of generic versus brand efforts. Sometimes growth/processor alliances are tied to some level of branding such as Â“brands by breedÂ”. Branding may enhance the brand demand for the breed or brand name but also reduce the grower/packer competitiv e structure where a producer may find fewer buyers for their meat. That is, some alliances to achieve a brand identity may have a cost in terms of the vertical competitive struct ure for initial animals in the distribution channel. The simulation analysis, the ranking of the variables and the upper/lower limits combination are very important in signaling the areas that could be m eaningfully targeted by brand advertising campaigns, and generic advertising programs. In the case of the brand advertising the objective will be to incr ease the market share of a particular brand, while generic promotions is ge nerally intended to expand the market of the entire sector.
81 The beef and pork promotion checkoffs ar e a good example of ge neric advertising. Finally, it is clear that by targeting nonde mographic and some de mographic elements such as regions, occupations and female age, it is possible to achie ve higher levels of brand identification when purchasing meat pr oducts. Again the merits of branding stills need to be explored.
82 CHAPTER 7 SUMMARY AND CONCLUSIONS Brand recognition is determined largel y by product attributes, the perceived performance of the brand on these attributes, and the importance that consumers attach to them. Over the past decades, changes in c onsumer demand and a dvances in technology have prompted the U.S. meat industry to differentiate the existing supply of meat products and increase levels of concentration within th e industry. New product forms able to satisfy the increasing demand for more convenient foods with fewer levels of fat and cholesterol and better quality and consistency are competing to attract the attention of consumers and influence the decision-maki ng process when purch asing branded or nonbranded meat products. Results from this rese arch provide the information necessary to analyze the factors influencing consumerÂ’s br and preferences, to establish a rank of the variables driving consumerÂ’s brand purchasi ng decisions for each meat category, and to introduce some recommendations to better assess the prevailing trends with branding in the meat industry. Data used in this study were based on household panel reports collected by the National Panel Diary Group Company. Househol ds completing the eating diary indicated if the purchase was based on a branded meat, thus giving a binary classification for each purchase. Using Probit model specification, demographi c and nondemographics were incorporated to estimate the impact of th e major factors influenc ing brand preferences. Thus, Probit models were estimated for each me at category to compare coefficients and
83 analyze the effect of the diffe rent variables on consumersÂ’ br and preferences. In addition, a pooled Probit model based on pooling all meat categories in one data set was estimated to measure changes in preferences over time a nd the price fluctuations . In both scenarios, the effects of the majority of the demographic variables and all of the nondemographics variables were statistically significant. Simulations analyses were performed to evaluate the impact of the different variables on the likel ihood of brand prefer ence relative to the average household when purchasing meat products. For the average consumer, the likelihood of buying branded beef products is 28 percent; branded fish, 41 percent; branded pork, 52 percent; and branded poultry is estimated to be 80 percen t. Coefficient estimates for most of the variables were statistically significant at the 95 percent level of confidence, but numerically the range of effects on the probability of buying or not buying branded products was for the vast majority quite sm all, below the 10 percentage point range. The resulting ranking of variab les showed that for all m eat categories the types of outlet and location within the store have the most important impact on the probability of brand preference. From Chapter 6 it is known that consumers buying in warehouses, supercenters, and supermarkets showed positiv e responses to brand identification, while butcher and meat market cu stomers indicated a strong negative response to brand identification. In relation to the location with in the store, for all meat categories, the purchase made in the freezer section showed positive response of more than 20 percentage points for beef, fish and pork while for poultry the response was about 12 percentage points. Results show ed that time has a positive and particular impact for beef and pork product. During the period between 1992 and 2001 the probabi lities of buying a
84 branded beef products have doubled showing a y early average increase of 1.5 percentage points while for pork the probabilities have incr eased an average of 2.3 percent indicating a clear trend in both sectors toward more br and identification when buying beef and pork. On the other hand, as prices increased the pr obability of brand pref erences also increased in all meat categories. While the effects of price variations were positive and statistically significant, their numerical values were st ill very low and had little effect on the probability of purchasing by brand. Among demographic variables, the region of the countr y, consumer occupation, and the age of the respondent showed the mo st numerical and statistically significant impact in the terms of the effects on the probability of brand preferences. Among the different occupations, the group formed by retired/unemployed and sales respondents indicated a surprising positive response to br and identification, while the group formed by professionals showed a strong negative re sponse to branded pr oducts. Occupations should be the focus of particul ar attention since they usuall y represent a ready target of promotions and marketing campaigns, especi ally professionals gi ven their purchasing power and negative response to the brands. Results also show ed that the effect of the female age in all meat categories was positiv e and numerically significant for consumers younger than 25 years while for females with ages above 25 years the probability of brand preference decreased as age increa ses. This positive response among young females might be explained by traditions a nd costumes carried over from their parents home or by changes in lifestyl es involving diets and the cons umption of particular lines of products.
85 In Figure 7-1, the reader can easily see the results of this research and fully compare the impact of each of the variable s present in the model according to the range of variation in the likelih ood of buying branded meat produc ts. Fallowing the results from Chapter 6, the variables are ranked relative to the average consumer and the ranges of probabilities are then shown. R a n g e i n P ro b a b i l i t y o f P u r c h a s i n g b y B r a n dR a n g e i n P r o b a b i l i t y o f P u r c h a s i n g b y B r a n d Figure 7-1: Ranking of variables affecting brand preferences Again, the type of store and the location within the store show the most significant impact in terms of brand id entity and must be the area of concentration for industry participants and marketing anal yst because they provide info rmation about the nature of the industry in te rms of product distribution, producer/p rocessor alliances, the structure of the industry, and also help to identify what product form (i.e ., frozen or fresh) and what
86 brand attributes are preferred by consumers. The extend of the difference with these two nondemographic variables is profound as clearly illustrated in this figure. Based on the results obtained in this research, the re levance of brands among the U.S. meat industry can be evaluated from the stand point of the perceived importance that consumers attach to the productÂ’s added values , attributes, consistency, and even health and safety features; and equally from the industry perspective th rough the ability to achieve levels of product differentiation, the degree of concentration in a sector, and the final value that the raw product acquires as it moves through the production and distribution chain. The percei ved characteristics of the product imply some degree of security and reliability which could easy e nhance the total demand for that particular product. However, it is very important to cl arify that there is no definitive evidence supporting the statement that br anded meat products are safe r to consume or have more quality than unbranded products. A major challenge for the meat industry is to satisfy the increasing demand for more convenient and value-added products cons istent with the changing preferences of consumers. Industry participants must unders tand the factors affecting consumersÂ’ buying decisions and their degree of recognition of the attributes of a branded or non-branded product in order to meet consumersÂ’ expectations.
87 APPENDIX TSP PROGRAM: PROBIT M ODELS AND SIMULATIONS OPTIONS MEMORY=1500; TITLE 'BRAND PREFERENCES FOR BEEF,POULTRY,FISH, AND PORK'; FREQ NONE; LIST ZVARZD FRSH SALE STTY STWR PCAT MNTH YEAR DINC DHWZ DAGF DAGM DCHD DEMF DEDF DEDM DOCC DSTA DMSA MEATCUT YRS BRAND; LIST ZVARZC FAMC PRCP PPER DADI POUNDS; ? out 'C:\ZBEEF2003\EATINGS\BRANDS\TSPPRG2\BRANDS2'; ? READ(format=free,FILE ='C:\ZBEEF2003\EATINGS\BRANDS\brand2.dat') zvarzd; ? READ(format=free,FILE ='C:\ZBEEF2003\EATINGS\BRAND S\brand2.dat') zvarzc; ? ? DOC STTY 'STORE TYPE 1-SUPERMKT,2-WAREHSE,3-BUTHCHER,4-LOCAL,5-CONVEN,6MDAY,7-SUPERC,8-OTHER' ? DOC PPER 'WAVE 274-1992:10, 372-2000:12' ? DOC PCAT 'MEAT CATEGORY 557-BEEF,559-POULTRY,558-FISH,560-PORK'; ? DOC MNTH 'MONTHS 1-12'; ? DOC YEAR 'YEAR 2(1992) 3(1993).. 0 (2000) SEE YRS'; ? DOC DINC 'INCOME 2 TO 33 SEE NOTE'; ? DOC DHWZ 'HOUSEHOLD SIZE MEMBERS'; ? DOC DAGF 'FEMALE HEAD AGE -YEAR GROUPS'; ? DOC DAGM 'MALE HEAD AGE YEAR GROUPS'; ? DOC DCHD 'Age and Presence of Children 8 GROUPS'; ? DOC DEMF 'Employment Stat us-Female Head -HOURS IF EMPLOYED'; ? DOC DEDF 'Education Female Head1= Grade School'; ? DOC DEDM 'Education Male Head'; ? DOC DOCC 'Occupation Householder'; ? DOC DSTA 'States and Regions'; ? DOC DADI 'Areas of Dominate Influence'; ? DOC DMSA 'Metropolitan Area Sizes'; ? DOC POUNDS 'Pounds purchased in a reported Wave'; ? DOC MEATCUT 'Type of meat depending on the category'; ? DOC YRS 'Years 1992...'; ? DOC BRAND 'Brand identification 0-no, 1-yes'; ? DOC FRSH 'FORM OF THE MEAT FRESH, ETC)'; ? DOC PRCP 'PRICE FOR THE MEAT PURCHASED'; ? DOC SALE 'WAS THE PURCHASE MADE ON SALE YES/NO'; ? DOC STWR 'WHERE IN THE STORE WAS THE MEAT PURCHASED'; ? OUT; ? DBLIST 'C:\ZBEEF2003\EATI NGS\BRANDS\TSPPRG\BRANDS2';
88 IN 'c:\ZBEEF2003\EATINGS\BRANDS\TSPPRG\BRANDS2'; HIST(DISCRETE) YEAR; ? INCOME CATEGORIES; ? 02= UNDER 7,500 19= 45,000-49,999 ; ? 05= 7500-9,999 21= 50,000-59,999 ; ? 07= 10,000-12,499 23= 60,000-69,999 ; ? 09= 12,500-14,999 24= 70,000-74,999 ; ? 11= 15,000-19,999 25= 75,000-99,999 ; ? 13= 20,000-24,999 27= 100,000 AND OVER; ? 15= 25,000-29,999 29= 100,000-149,999 ; ? 16= 30,000-34,999 31= 150,000-199,999 ; ? 17= 35,000-39,999 33= 200,000+ ; ? 18= 40,000-44,999 ? ? HOUSEHOLD SIZE; ? 1= One Member 6= Six Member ? 2= Two Member 7= Seven Member ? 3= Three Member 8= Eight Member ? 4= Four Member 9= Nine+ Member ? 5= Five Member ? Female Head Age ? 1= Under 25 6= 45-49 ? 2= 25-29 7= 50-54 ? 3= 30-34 8= 55-64 ? 4= 35-39 9= 65+ ? 5= 40-44 0= no such head ? Male Head Age ? 1= Under 25 6= 45-49 ? 2= 25-29 7= 50-54 ? 3= 30-34 8= 55-64 ? 4= 35-39 9= 65+ ? 5= 40-44 0= no such head ? Age and Presence of Children ? 1= Under 6 Only 5= Under 6 & 13-17 ? 2= 6-12 Only 6= 6-12 & 13-17 ? 3= 13-17 Only 7= All 3 Age Groups ? 4= Under 6 &6-12 8= None <18 ? Employment Status-Female Head ? 1= Under 30 Hours 9= Not Employed ? 2= 30-34 Hours 0= No Such Head ? 3= 35+ Hours ? Education Female Head ? 1= Grade School 5= Graduated College ? 2= Some High School 6= Post College Graduate ? 3= Graduated High School 0= No Such Head ? 4= Some College ? Education Male Head ? 1= Grade School 5= Graduated College ? 2= Some High School 6= Post College Graduate
89 ? 3= Graduated High School 0= No Such Head ? 4= Some College ? Occupation Householder ? 1= Professional 7= Military ? 2= Proprietor, Manager, O fficial 8= Service Worker ? 3= Clerical 9= Farm(Owner, Manager, Laborer) ? 4= Sales 0= Student Employed <30 Hours ? 5= Craftsman/Forman(Skilled) 10-= Laborers ? 6= Operative (Semi-skilled) 11= Retires, Unemployed ? STATE CODES ; ? 10Â’s= New England 60Â’s= East South Central ; ? 10= Maine 60= Kentucky ; ? 11= New Hampshire 61= Tennessee ; ? 12= Vermont 62= Alabama ; ? 13= Massachusetts 63= Mississippi ; ? 14= Rhode Island ; ? 15= Connecticut 70Â’s= West South Central ; ? 70= Arkansas ; ? 20Â’s= Middle Atlantic 71= Louisiana ; ? 20= New York 72= Oklahoma ; ? 21= New Jersey 73= Texas ; ? 22= Pennsylvania ; ? ; ? 30Â’s= East North Central ; ? 30= Ohio 80Â’s= Mountain ; ? 31= Indiana 80= Montana ; ? 32= Illinois 81= Wyoming ; ? 33= Michigan 82= Colorado ; ? 34= Wisconsin 83= Idaho ; ? ; ? 40Â’s= West North Cntrl 84= New Mexico ; ? 40= Minnesota 85= Nevada ; ? 41= Iowa 86= Arizona ; ? 42= Missouri 87= Utah ; ? 43= Nebraska ; ? 44= Kansas 90Â’s= Pacific ; ? 45= North Dakota 90= Washington ; ? 46= South Dakota 91= Oregon ; ? 92= California ; ? 50Â’s= South Atlantic ; ? 50= Maryland ; ? 51= Delaware Census Regions ; ? 52= Washington, D.C. East = 1,2 ; ? 53= Virginia Central = 3,4 ; ? 54= West Virginia South = 5,6,7 ; ? 55= North Carolina West = 8,9 ; ? 56= South Carolina ; ? 57= Florida ; ? 58= Georgia ; ? MSA MARKET SIZE ? 1= 50,000-249,999 5= 1,000,000-2,499,999 ? 3= 250,000-499,999 6= 2,500,000 + ? 4= 500,000-999,999 9= NON-MSA
90 LIST ZVARDUM FRSH SALE STTY STWR PCAT DINC DHWZ DAGF DAGM DCHD DEMF DEDF DEDM DOCC DSTA DMSA MEATCUT YRS BRAND; PROC XXXX; SELECT 1; DOT ZVARDUM; DD =( MISS(.)=0); SELECT DD=1; HIST(DISCRETE) .; MAT N.=NROW(@HISTVAL); DUMMY .; PRINT N.; ENDDOT; ZFRSH1 = (FRSH=1) + (FRSH^=1)*(-1); ? FRESH MEAT =1; ZSALE1 = (SALE=1) + (SALE^=1)*(1); ? PURCHASE WAS ON SALE =1; ?<<<<<<<>>>>>>>>>>>>>>; SELECT 1; ? STORE TYPE =1 SUPERMARKETS; ? STORE TYPE =2 Warehouse/Club Store; ? STORE TYPE =3 Butcher/Meat Market ? STORE TYPE =4 SuperCenters; ? STORE TYPE =5 ALL OTHERS; XSTTY = ( STTY=1) + (STTY=2)*2 + (STTY=3)*3 + (STTY=7)*4 +( (STTY^=1) & (STTY^=2) & (STTY^=3) & (STTY^=7) )*5; DUMMY XSTTY; ? NORMALIZING ON THE ALL OTHER TYPE 5; DOT 1-4; ZSTTY. = ( XSTTY. XSTTY5); ENDDOT; ?<<<<<<<<<<<<<<<<< WHERE IN THE STORE >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SELECT 1; ? WHERE IN THE STORE 1 = FRESH MEAT CASE (1); ? WHERE IN THE STORE 2 = D ELI AND FOOD BAR (2 & 3); ? WHERE IN THE STORE 3 = GOURMET SECTIONS (4); ? WHERE IN THE STORE 4 = FREEZER SECTION (5) ? WHERE IN THE STORE 5 = ALL OTHER INCLUDING NOT REPORTED ( 0 & >5); XSTWR = ( STWR=1) + ( (STWR=2 | STWR=3) )*2 + ( STWR=4)*3 + (STWR=5)*4 + (STWR=0 | STWR>5)*5; DUMMY XSTWR; ? NORMALIZING ON ALL OTHER LOCATIONS; DOT 1-4; ZSTWR. = XSTWR. XSTWR5; ENDDOT; ?<<<<<<<<<<<<<<<<<<<<<<< MEAT TYPE >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? BEEF =1 (557) FISH=2(558) POULTRY=3(559) PORK=4(560); DUMMY PCAT; HIST(DISCRETE) PCAT; HIST(DISCRETE) PCAT1-PCAT4; ? NORMALIZING ON PORK(4); DOT 1-3; ZPCAT. = PCAT.PCAT4; ENDDOT; SELECT 1;
91 ?<<<<<<<<<<<<<<<<<<<<<<< RELATIVE PRICES >>>>>>>>>>>>>>>>>>>>>>>>>>>; ZYEAR = int(1990 + YEAR + (YEAR=0)*10); HIST(DISCRETE) ZYEAR; DO J=557 TO 560; DO K=1992 TO 2000; SELECT PCAT=J & ZYEAR=K; PRICE=PRCP/POUNDS; MSD(PRINT) PRICE; ZPRCP = PRICE/@MEAN; ? PRICE RELATIVE TO THE AVERAGE PRICE FOR EACH MEAT AND YEAR; ENDDO; ENDDO; ? PRICE CHEAPER THAN THE AVERAGE SH OULD REDUCE THE BRAND PREFERENCE; SELECT 1; ?<<<<<<<<<<<<<<<<<<<<<<< INCOME DEMOGRAPHIC >>>>>>>>>>>>>>>>>>>>>>>>>; ? GREATING FOUR INCOME GROUPS; ? INCOME CATEGORIES; ? 02= UNDER 7,500 19= 45,000-49,999 ; ? 05= 7500-9,999 21= 50,000-59,999 ; ? 07= 10,000-12,499 23= 60,000-69,999 ; ? 09= 12,500-14,999 24= 70,000-74,999 ; ? 11= 15,000-19,999 25= 75,000-99,999 ; ? 13= 20,000-24,999 27= 100,000 AND OVER; ? 15= 25,000-29,999 29= 100,000-149,999 ; ? 16= 30,000-34,999 31= 150,000-199,999 ; ? 17= 35,000-39,999 33= 200,000+ ; ? 18= 40,000-44,999 XINC = (DINC<15)*1 + (DINC>=15 & DINC<=19)*2 + (DINC>=21 & DINC<= 24)*3 + (DINC>=25)*4; DUMMY XINC; ?NORMALIZING ON THE HIGHEST INCOME GROUP; DOT 1-3; ZINC. = XINC. XINC4; ENDDOT; ?<<<<<<<<<<<<<<<<<<<<<<< HOUSEHOLD SIZE >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; XHWZ =(DHWZ=1) + (DHWZ=2)*2 + (DHWZ=3)*3 + (DHWZ>=4)*4; HIST(DISCRETE) DHWZ XHWZ; DUMMY XHWZ; ? NORMALIZED ON THE LARGEST HOUSEHOLD SIZE; DOT 1-3; ZHWZ.=XHWZ. XHWZ4; ENDDOT; ?<<<<<<<<<<<<<<<<<<<< HOUSEHOLD FEMALE AGE >>>>>>>>>>>>>>>>>>>>>>>>>>; ? UNDER 25 =1 25/40=2 40/65=3 65+=4; XAGF = (DAGF=1) + ((DAGF>=2) & (DAGF<=4))*2 + ((DAGF>=5) & (DAGF<=8))*3 + (DAGF>=9)*4; XAGM = (DAGM=1) + ((DAGM>=2) & (DAGM <=4))*2 + ((DAGM>=5 ) & (DAGM<=8))*3 + (DAGM>=9)*4; WAGF = XAGF + (XAGF=0)*XAGM; ? IF NO FEMALE HEAD THEN USE THE MALE HEAD; HIST(DISCRETE) XAGF WAGF;
92 DUMMY WAGF; DOT 1-3; ZAGF. = WAGF. WAGF4; ENDDOT; ?<<<<<<<<<<<<<<<<<< PRESENCE OF CHILDREN >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ZCHD=( (DCHD=9)=0 ); ? CHILDREN UNDER 18 IN THE HOUSEHOLD; ?<<<<<<<<<<<<<<<<<< FEMALE EMPLOYMENT >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? 1 FEMALE FULL TIME OR HUSBAND EMPL OYED, 2=PARTIME, 3=NOT EMPLOYED; XEMF = (DEMF=3 | DEMF=0)*1 + (DEMF=1 | DEMF=2)*2 +( DEMF=9)*3; HIST(DISCRETE) XEMF; DUMMY XEMF; ? NORMALIZING ON NOT EMPLOYED; DOT 1-2; ZEMF.=XEMF. XEMF3; ENDDOT; ?<<<<<<<<<<<<<<<<<<<<< EDUCATION >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? 1= HIGH SCHOOL OR LESS, 2= SOME COLLEGE, 3= COLLEGE GRADUATE, 4=POST GRADUATE; ? IF NO FEMALE HEAD THEN USE THE MALE EDUCATION LEVEL SINCE THEY ARE THEN THE DECISION MAKER; XEDF = ( DEDF<=3)*1 + ( DEDF=4)*2 + ( DEDF=5)*3 + ( DEDF=6)*4; XEDM = ( DEDM<=3)*1 + ( DEDM=4)*2 + ( DEDM=5)*3 + ( DEDM=6)*4; WEDC = XEDF + (XEDF=0)*XEDM; DUMMY WEDC; DOT 1-3; ZEDC.=WEDC. WEDC4; ENDDOT; ?<<<<<<<<<<<<<<<<<<< OCCUPATION OF HEAD >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? 1= Professional 7= Military ? 2= Proprietor, Manager, Official 8= Service Worker ? 3= Clerical 9= Farm(Owner, Manager, Laborer) ? 4= Sales 0= Student Employed <30 Hours ? 5= Craftsman/Forman(Skilled) 10= Laborers ? 6= Operative (Semi-skilled) 11= Retires, Unemployed XOCC=DOCC +1; DUMMY XOCC; ? RANGE 1 TO 12; ? NORMALIZING ON THE RETIRED OR UNEMPLOYED; DOT 1-11; ZOCC. = XOCC. XOCC12; ENDDOT; ?<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? STATE CODES ? 10Â’s= New England ? 20Â’s= Middle Atlantic ? 30Â’s= East North Central ? 40Â’s= West North Cntrl ? 50Â’s= South Atlantic ? 60Â’s= East South Central ? 70Â’s= West South Central ? 80Â’s= Mountain ? 90Â’s= Pacific XSTA = INT(DSTA/10); HIST(DISCRETE) XSTA; DUMMY XSTA;
93 ? NORMALIZING ON TH E PACIFIC REGION; DOT 1-8; ZSTA. = XSTA. XSTA9; ENDDOT; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<< MARKET SIZE >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; ? MSA MARKET SIZE ? 1= 50,000-249,999 (1) 5= 1,000,000-2,499,999 (4) ? 3= 250,000-499,999 (2) 6= 2,500,000 + (5) ? 4= 500,000-999,999 (3) 9= NON-MSA (6) XMSA = (DMSA=1)*1 + (DMSA=3)*2 +(DMSA=4)*3 +(DMSA=5)*4 +(DMSA=6)*5 +(DMSA=9)*6; DUMMY XMSA; DOT 1-5; ZMSZ.= XMSA. XMSA6; ENDDOT; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<< SEASONAL DUMMIES >>>>>>>>>>>>>>>>>>>>>>>>>>>; DUMMY MNTH; DOT 1-11; ZMTH. = MNTH. MNTH12; ENDDOT; ?=========================; ? STARTING THE PROBIT MODELS ; ?=========================; SYMTAB; COMPRESS; SYMTAB; YBRN = BRAND; ? 1=YES TO BRAND, 0=NO TO BRAND; PROBIT YBRN C ZSTTY1-ZSTTY4 ZSTWR1-ZSTWR4 ZPCAT1-ZPCAT3 ZPRCP ZINC1-ZINC3 ZHWZ1-ZHWZ3 ZAGF1-ZAGF3 ZCHD ZEMF1-ZEMF2 ZOCC1-ZOCC11 ZEDC1-ZEDC3 ZSTA1-ZSTA8 ZMSZ1-ZMSZ5 ZMTH1-ZMTH11; ENDPROC XXXX; DO ZTTZ=1 TO 96; SELECT TT=ZTTZ; MSD(NOPRINT) YRS MNTH; PRINT ZTTZ @MEAN; ENDDO; MAT NR=NROW(MCOEF); MFORM(TYPE=GEN,NROW=70,NCOL=1) MCOEF2=0; DO J=1 TO 70; SET JJ=J+1; SET MCOEF2(J)=MCOEF_R(JJ,7); ENDDO; PRINT MCOEF2;
94 MFORM(TYPE=GEN,NROW=800 ,NCOL=3) MPROB=0; ?<<<<<<<<<<<<>>>>>>>>>>>>>; SET XB=0; SET X0=1; SET PROB=0; SET I=0; MSD(NOPRINT) TT; SET MTT= INT(@MEAN); PRINT MTT; DOT SZSTTY1-SZSTTY4 SZSTWR1-SZSTWR4 SZPCAT1-SZPCAT3 SZPRCP SZINC1-SZINC3 SZHWZ1-SZHWZ3 SZAGF1-SZAGF3 SZCHD SZEMF1-SZEMF2 SZOCC1-SZOCC11 SZEDC1 -SZEDC3 SZSTA1-SZSTA8 SZMSZ1-SZMSZ5 SZMTH1-SZMTH11 STT STPCAT1 STPCAT2 STPCAT3; SET .=0; ENDDOT; PROC INIT; DOT SZSTTY1-SZSTTY4 SZSTWR1-SZSTWR4 SZPCAT1-SZPCAT3 SZINC1-SZINC3 SZHWZ1-SZHWZ3 SZAGF1-SZAGF3 SZCHD SZEMF1-SZEMF2 SZOCC1-SZOCC11 SZEDC1-S ZEDC3 SZSTA1-SZSTA8 SZMSZ1-SZMSZ5 SZMTH1-SZMTH11 STT STPCAT1 STPCAT2 STPCAT3 SZCATPR1 SZCATPR2 SZCATPR3; SET .=0; SET SZPRCP=1; SET STT=MTT; ENDDOT; ENDPROC INIT; ?===========================================================================; PROC ZSIMZ; MAT BRDCOEFT=MCOEF2; SET I=I+1; SET STPCAT1=STT*SZPCAT1; SET STPCAT2=STT*SZPCAT2; SET STPCAT3=STT*SZPCAT3; SET SZCATPR1=SZPRCP*SZPCAT1; SET SZCATPR2=SZPRCP*SZPCAT2; SET SZCATPR3=SZPRCP*SZPCAT3; MMAKE ZVARZ X0 SZSTTY1-SZSTTY4 SZSTWR1-SZSTWR4 SZPRCP SZINC1-SZINC3 SZHWZ1-SZHWZ3 SZAGF1-SZAGF3 SZCHD SZEMF1-SZEMF2 SZOCC1-SZOCC11 SZEDC1-S ZEDC3 SZSTA1-SZSTA8 SZMSZ1-SZMSZ5 SZMTH1-SZMTH11 SZPCAT1-SZPCAT3 SZCATPR1 SZCATPR2 SZCATPR3 STT STPCAT1 STPCAT2 STPCAT3; MAT XB=ZVARZ'BRDCOEFT; SET XBB=XB(1); SET PROB = CNORM(XBB); SET MPROB(I,1)=SIMNUM; SET MPROB(I,2)=SIMVAR; SET MPROB(I,3)=PROB; ENDPROC ZSIMZ; msd zprcp; hist zprcp;
95 ?===========================================================================; ?========== STARTING THE SIMULATIONS TO GET THE PROBIT PROBABILITIES =======; ?=========================================================================== ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< AVERAGE #1 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=1; SET SIMVAR=0; ? ALL VARIALBES AT THE AVERAGE CONSUMER; INIT; ZSIMZ; <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< STORE TYPE #2 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=2; SET SIMVAR=1; INIT; SET SZSTTY1=1; ? SUPERMARKETS; ZSIMZ; SET SIMVAR=2; INIT; SET SZSTTY2=1; ? WAREHOUSE/CLUBS; ZSIMZ; SET SIMVAR=3; INIT; SET SZSTTY3=1; ? BUTCHER/MEAT MARKETS; ZSIMZ; SET SIMVAR=4; INIT; SET SZSTTY4=1; ? SUPERCENTERS; ZSIMZ; SET SIMVAR=5; INIT; SET SZSTTY1=-1; SET SZSTTY2=-1; SET SZSTTY3=-1; SET SZSTTY4=-1; ? ALL OTHER OUTLETS; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< WHERE IN THE STORE #3 >>>>>>>>>>>>>>>>>>>>; SET SIMNUM=3; SET SIMVAR=1; INIT; SET SZSTWR1=1; ? FRESH MEAT CASE; ZSIMZ; SET SIMVAR=2; INIT; SET SZSTWR2=1; ? DELI AND FOOD BAR; ZSIMZ; SET SIMVAR=3; INIT; SET SZSTWR3=1; ? GOURMET SECTIONS; ZSIMZ; SET SIMVAR=4; INIT; SET SZSTWR4=1; ? FREEZER SECTION; ZSIMZ; SET SIMVAR=5; INIT; SET SZSTWR1=-1; SET SZSTWR2=-1; SET SZSTWR3=-1; SET SZSTWR4=-1; ? ALL OTHER LO0CATIONS; ZSIMZ;
96 ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< TYPE OF MEAT #4 >>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=4; SET SIMVAR=1; INIT; SET SZPCAT1=1; ? BEEF; ZSIMZ; SET SIMVAR=2; INIT; SET SZPCAT2=1; ? FISH; ZSIMZ; SET SIMVAR=3; INIT; SET SZPCAT3=1; ? POULTRY ZSIMZ; SET SIMVAR=4; INIT; SET SZPCAT1=-1; SET SZPCAT2=-1; SET SZPCAT3=-1; ? PORK; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< RELATIVE PRICE #5 >>>>>>>>>>>>>>>>>>>>>>>; SELECT 1; SET SIMNUM=5; DO J=.50 TO 1.50 BY .25; SET SIMVAR=J; INIT; SET SZPRCP=J; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< INCOME #6 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=6; SET SIMVAR=1; INIT; SET SZINC1=1; ? UNDER $25,000 ZSIMZ; SET SIMVAR=2; INIT; SET SZINC2=1; ? $25,000 $50,000; ZSIMZ; SET SIMVAR=3; INIT; SET SZINC3=1; ? $50,000 $75,000 ZSIMZ; SET SIMVAR=4; INIT; SET SZINC1=-1; SET SZINC2=-1; SET SZINC3=-1; ? $75,000 PLUS; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< HOUSEHOLD SIZE #7 >>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=7; SET SIMVAR=1; INIT; SET SZHWD1=1; ? HOUSEHOLD SIZE =1; ZSIMZ; SET SIMVAR=2; INIT; SET SZHWD2=1; ? hOUSEHOLD SIZE =2; ZSIMZ;
97 SET SIMVAR=3; INIT; SET SZHWD3=1; ? HOUSEHOLD SIZE =3; ZSIMZ; SET SIMVAR=4; INIT; SET SZHWD1=-1; SET SZHWD2=-1; SET SZHWD3=-1; ? HOUSEHOLD SIZE >=4; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< AGE OF FEMALE #8.1 >>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=8.1; SET SIMVAR=1; INIT; SET SZAGF1=1; ? AGE UNDER 25 ZSIMZ; SET SIMVAR=2; INIT; SET SZAGF2=1; ? AGE 25/40 YEARS; ZSIMZ; SET SIMVAR=3; INIT; SET SZAGF3=1; ? AGE 40/65 YEARS; ZSIMZ; SET SIMVAR=4; INIT; SET SZAGF1=-1; SET SZAGF2=-1; SET SZAGF3=-1; ? AGE 65 PLUS; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<< EDUCATION OF FEMALE #8.2 >>>>>>>>>>>>>>>>>; SET SIMNUM=8.2; SET SIMVAR=1; INIT; SET SZEDC1=1; ? HIGH SCHOOL OR LESS ZSIMZ; SET SIMVAR=2; INIT; SET SZEDC2=1; ? SOME COLLEGE; ZSIMZ; SET SIMVAR=3; INIT; SET SZEDC3=1; ? COLLEGE GRADUATE ZSIMZ; SET SIMVAR=4; INIT; SET SZEDC1=-1; SET SZEDC2=-1; SET SZEDC3=-1; ? POST GRADUATE; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< CHILDREN UNDER 18 #9 >>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=9; SET SIMVAR=1; INIT; SET SZCHD=0; ? NO CHILDREN; ZSIMZ; SET SIMVAR=2; INIT; SET SZCHD=1; ? CHILDREN; ZSIMZ;
98 ?<<<<<<<<<<<<<<<<<<<<<< EMPLOYMENT OF HOUSEHOLD #10 >>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=10; SET SIMVAR=1; INIT; SET SZEMF1=1; ? FULL TIME EMPLOYMENT; ZSIMZ; SET SIMVAR=2; INIT; SET SZEMF2=1; ? PARTTIME EMPLOYMENT; ZSIMZ; SET SIMVAR=3; INIT; SET SZEMF1=-1; SET SZEMF2=-1; ? NOT EMPLOYED; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< OCCUPATION #11 >>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=11; SET SIMVAR=1; INIT; SET SZOCC1=1; ? ZSIMZ; SET SIMVAR=2; INIT; SET SZOCC2=1; ? ZSIMZ; SET SIMVAR=3; INIT; SET SZOCC3=1; ? ZSIMZ; SET SIMVAR=4; INIT; SET SZOCC4=1; ? ZSIMZ; SET SIMVAR=5; INIT; SET SZOCC5=1; ? ZSIMZ; SET SIMVAR=6; INIT; SET SZOCC6=1; ? ZSIMZ; SET SIMVAR=7; INIT; SET SZOCC7=1; ? ZSIMZ; SET SIMVAR=8; INIT; SET SZOCC8=1; ? ZSIMZ; SET SIMVAR=9; INIT; SET SZOCC9=1; ? ZSIMZ; SET SIMVAR=10; INIT; SET SZOCC10=1; ?
99 ZSIMZ; SET SIMVAR=11; INIT; SET SZOCC11=1; ? ZSIMZ; SET SIMVAR=12; INIT; SET SZOCC1=-1; SET SZOCC2=-1; SET SZO CC3=-1; SET SZOCC4=-1; SET SZOCC5=-1; SET SZOCC6=-1; SET SZOCC7=-1; SET SZOCC8=-1; SET SZOCC9=-1; SET SZOCC10=-1; SET SZOCC11=-1; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< SEASONS #12 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=12; SET SIMVAR=1; INIT; SET SZMTH1=1; ? ZSIMZ; SET SIMVAR=2; INIT; SET SZMTH2=1; ? ZSIMZ; SET SIMVAR=3; INIT; SET SZMTH3=1; ? ZSIMZ; SET SIMVAR=4; INIT; SET SZMTH4=1; ? ZSIMZ; SET SIMVAR=5; INIT; SET SZMTH5=1; ? ZSIMZ; SET SIMVAR=6; INIT; SET SZMTH6=1; ? ZSIMZ; SET SIMVAR=7; INIT; SET SZMTH7=1; ? ZSIMZ; SET SIMVAR=8; INIT; SET SZMTH8=1; ? ZSIMZ; SET SIMVAR=9; INIT; SET SZMTH9=1; ? ZSIMZ; SET SIMVAR=10; INIT; SET SZMTH10=1; ? ZSIMZ;
100 SET SIMVAR=11; INIT; SET SZMTH11=1; ? ZSIMZ; SET SIMVAR=12; INIT; SET SZMTH1=-1; SET SZMTH2=-1; SET SZMTH3=-1; SET SZMTH4=-1; SET SZMTH5=1; SET SZMTH6=-1; SET SZMTH7=-1; SET SZMT H8=-1; SET SZMTH9=-1; SET SZMTH10=-1; SET SZMTH11=-1; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< REGIONS #13 >>>>>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=13; SET SIMVAR=1; INIT; SET SZSTA1=1; ? ZSIMZ; SET SIMVAR=2; INIT; SET SZSTA2=1; ? ZSIMZ; SET SIMVAR=3; INIT; SET SZSTA3=1; ? ZSIMZ; SET SIMVAR=4; INIT; SET SZSTA4=1; ? ZSIMZ; SET SIMVAR=5; INIT; SET SZSTA5=1; ? ZSIMZ; SET SIMVAR=6; INIT; SET SZSTA6=1; ? ZSIMZ; SET SIMVAR=7; INIT; SET SZSTA7=1; ? ZSIMZ; SET SIMVAR=8; INIT; SET SZSTA8=1; ? ZSIMZ; SET SIMVAR=9; INIT; SET SZSTA1=-1; SET SZSTA2=-1; SET SZST A3=-1; SET SZSTA4=-1; SET SZSTA5=-1; SET SZSTA6=-1; SET SZSTA7=-1; SET SZSTA8=-1; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< MARKET SIZE #14 >>>>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=14; SET SIMVAR=1; INIT; SET SZMSZ1=1; ?
101 ZSIMZ; SET SIMVAR=2; INIT; SET SZMSZ2=1; ? ZSIMZ; SET SIMVAR=3; INIT; SET SZMSZ3=1; ? ZSIMZ; SET SIMVAR=4; INIT; SET SZMSZ4=1; ? ZSIMZ; SET SIMVAR=5; INIT; SET SZMSZ5=1; ? ZSIMZ; SET SIMVAR=6; INIT; SET SZMSZ1=-1; SET SZMSZ2=-1; SET SZMSZ3=-1; SET SZMSZ4=-1; SET SZMSZ5=-1; ZSIMZ; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< AVERAGE TREND #15 >>>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=15; DO J=1 TO 96; SET SIMVAR=J; INIT; SET STT=J; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< TREND FOR BEEF #16 >>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=16; DO J=1 TO 96; SET SIMVAR=J; INIT; SET STT=J; SET SZPCAT1=1; ? BEEF; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< TREND FOR FISH #17 >>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=17; DO J=1 TO 96; SET SIMVAR=J; INIT; SET STT=J; SET SZPCAT2=1; ? FISH; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< TREND FOR POULTRY #18 >>>>>>>>>>>>>>>>>>>; SET SIMNUM=18; DO J=1 TO 96; SET SIMVAR=J; INIT; SET STT=J; SET SZPCAT3=1; ? POULTRY; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< TREND FOR PORK #19 >>>>>>>>>>>>>>>>>>>>>>>; SET SIMNUM=19;
102 DO J=1 TO 96; SET SIMVAR=J; INIT; SET STT=J; SET SZPCAT1=-1; SET SZPCAT2=-1; SET SZPCAT3=-1; ? PORK; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< RELATIVE PRICE #20 >>>>>>>>>>>>>>>>>>>>>>; SELECT 1; SET SIMNUM=20.557; DO J=.50 TO 1.50 BY .25; SET SIMVAR=J; INIT; SET SZPRCP=J; SET SZPCAT1=1; SET SZPCAT2=0; SET SZPCAT3=0; ZSIMZ; ENDDO; SELECT 1; SET SIMNUM=20.558; DO J=.50 TO 1.50 BY .25; SET SIMVAR=J; INIT; SET SZPRCP=J; SET SZPCAT1=0; SET SZPCAT2=1; SET SZPCAT3=0; ZSIMZ; ENDDO; SELECT 1; SET SIMNUM=20.559; DO J=.50 TO 1.50 BY .25; SET SIMVAR=J; INIT; SET SZPRCP=J; SET SZPCAT1=0; SET SZPCAT2=0; SET SZPCAT3=1; ZSIMZ; ENDDO; SELECT 1; SET SIMNUM=20.560; DO J=.50 TO 1.50 BY .25; SET SIMVAR=J; INIT; SET SZPRCP=J; SET SZPCAT1=-1; SET SZPCAT2=-1; SET SZPCAT3=-1; ZSIMZ; ENDDO; ?<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>; SET SIMNUM=21; ?UPPER LIMIT; SET SIMVAR=1; INIT; SET SZSTTY1=0; SET SZSTTY2=1; SET SZSTTY3=0; set SZSTTY4=0; ? WAREHOUSE/CLUBS; SET SZSTWR1=0; SET SZSTWR2=0;SET SZSTWR3=0; SET SZSTWR4=1; ? FREEZER SECTION; SET SZINC1=0; SET SZINC2=1;SET SZINC3=0; ? $25,000 TO $50,000;
103 SET SZHWD1=1; SET SZHWD2=0; SET SZHWD3=0; ? ONE MEMBER; SET SZAGF1=1; SET SZAGF2=0; SET SZAGF3=0; ? UNDER 25 YEARS; SET SZEDC1=0; SET SZEDC2=1; SET SZEDC3=0; ? SOME COLLEGE; SET SZCHD=1; SET SZCHD=0; ? NO CHILDREN; SET SZEMF1=0; SET SZEMF2=1; ? PART TIME; SET SZOCC1=0; SET SZOCC2=0; SET SZOCC3 =0; SET SZOCC4=1; SET SZOCC5=0; SET SZOCC6=0; SET SZOCC7=0; SET SZOCC8=0; SET SZOCC9=0; SET SZOCC10=0; SET SZOCC11=0; ? SALES; SET SZSTA1=0; SET SZSTA2=0; SET SZSTA3 =0; SET SZSTA4=0; SET SZSTA5=1; SET SZSTA6=0; SET SZSTA7=0; SET SZSTA8=0; ? SOUTH ATLANTIC; SET SZMSZ1=0; SET SZMSZ2=1; SET SZMSZ3=0; SE T SZMSZ4=0; SET SZMSZ5=0; ? 250,000 TO 499,000; SET SZMTH1=0; SET SZMTH2=0; SET SZMTH3=1; SET SZMTH4=0; SET SZMTH5=0; SET SZMTH6=0; SET SZMTH7=0; SET SZMTH8=0; SET SZMTH9=0; SET SZMTH10=0; SET SZMTH11=0; ? MARCH; ZSIMZ; ?LOWER LIMIT; SET SIMVAR=2; INIT; SET SZSTTY1=0; SET SZSTTY2=0; SET SZSTTY3=1; set SZSTTY4=0; ? BUTCHER/MEAT MARKET; SET SZSTWR1=0; SET SZSTWR2=1;SET SZSTWR3=0; SET SZSTWR4=0; ? DELI AND FOOD BAR? SET SZINC1=0;SET SZINC2=0;SET SZ INC3=0; ? OVER $75,000; SET SZHWD1=0; SET SZHWD2=1; SET SZHWD3=0; ? TWO MEMBERS; SET SZAGF1=0; SET SZAGF2=0; SET SZAGF3=0; ? 65 OR MORE; SET SZEDC1=1; SET SZEDC2=0; SET SZEDC3=0; ? HIGH SCHOOL OR LESS; SET SZCHD=0; SET SZCHD=1; CHILDREN? SET SZEMF1=0; SET SZEMF2=0; ? NOT EMPLOYED; SET SZOCC1=1; SET SZOCC2=0; SET SZOCC3 =0; SET SZOCC4=0; SET SZOCC5=0; SET SZOCC6=0; SET SZOCC7=0; SET SZOCC8=0; SET SZOCC9=0; SET SZOCC10=0; SET SZOCC11=0; ? PROFESSIONAL; SET SZSTA1=0; SET SZSTA2=0; SET SZSTA3=0; SET SZSTA4=0; SET SZSTA5=0; SET SZSTA6=1; SET SZSTA7=0; SET SZSTA8=0; ? EAST SOUTH CENTRAL; SET SZMSZ1=0; SET SZMSZ2=1; SET SZMSZ3=0 ; SET SZMSZ4=0; SET SZMSZ5=0; ? NON MARKET SIZE;
104 SET SZMTH1=0; SET SZMTH2=0; SET SZMTH3=0 ; SET SZMTH4=0; SET SZMTH5=0; SET SZMTH6=0; SET SZMTH7=0; SET SZMTH8=1 ; SET SZMTH9=0; SET SZMTH10=0; SET SZMTH11=0; ? AUGUST; ZSIMZ; PRINT MPROB; Write (format=excel, file='C:\ZBEEF2003\EATI NGS\BRANDS\TSPPRG\PROMSIM#3.xls') MPROB; END;
105 LIST OF REFERENCES Achenbaum, Alvin A. "The Mismanagement of Brand Equity." Paper presented at the ARF Fifth Annual Advertising and Prom otion Workshop. New York, NY. March 1993. Ailawadi, Kusum L, Scott A Neslin, and Ka ren Gedenk. "Pursuing the Value-Conscious Consumer: Store Brands Versus National Brand Promotions." Journal of Marketing 65, no. 1 (2001): 71. Allen, John Wesley and Thomas R. Pierson. Food Marketing and Distribution : Implications for Agriculture, Food Industry Reports ; No. 9301. East Lansing, Mich.: Food Industry Institute, Michigan State University, 1993. Barkema, Alan, Mark Dranbenstott, and Nancy Novack;. "The New U.S. Meat Industry." Economic Review (Federal Rese rve Bank of Kansas City); 86, no. 2 (2001): 24. Capps, Oral Jr. and Robert E. Branson. "Consumer Characte ristics Associated with the Selection of Lean Meat Products." Agribusiness (1986-1998) 4, no. 6 (1988): 549. Chakravarti, Amitav and Chris Janiszewski. "The Influen ce of Generic Advertising on Brand Preferences." Journal of Consumer Research 30, no. 4 (2004): 487-502. Codron, Jean M., James A. Sterns and To mas Reardon. "Strategic Choices in Produce Marketing: Issues of Compatib le Use and Exclusion Cost." Journal of Food Distribution Research 34, no. 03 (2003): 1-12. Datamonitor. Meat Products in the Unite d States Industry Profile (2004) Datamonitor USA, March 2004 [cited Februa ry 07 2005]. Available from www.datamonitor.com . Davis, David E. and Hayden Stewart. "Changing Consumers Demands Create Opportunities for U.S. Food System." Food Review 25, no. Issue 1 (2002): 18-29. Economic Research Service, USDA. Per Capita Food Consumption Data System. Economic Research Services, United Stat es Department of Agriculture -, 2004 [cited February 08, 2005]. Available from http://www.ers.usda.gov/data/foodconsumption/ . Economic Research Service. Food Consumption: Household Food Expenditures USDAERS, 2005 [cited May 17 2005]. Available from http://www.ers.usda.gov/briefing/consumption/Expenditures.htm .
106 Greene, William H. Econometric Analysis. 5th ed. Upper Saddle River, NJ: Prentice Hall, 1997. Gujarati, Deodar N. Basic Econometrics. 4th ed. New York, NY: McGraw-Hill/Irwin, 2003. Hall, Bronwyn H. and Clint Cummins. Tsp International Reference Manual Version 4.4. Palo Alto, CA: TSP International, 1998. Heilman, Carrie M., Bowman, Douglas, Wri ght, Gordon P. "The Evolution of Brand Preferences and Choice Behaviors of Consumers New to a Market." Journal of Marketing Research 37, no. 2 (2000): 139-55. Jones, John Philip. What's in a Name? Advertising and the Concept of Brands. New York: Lexington Books, 1986. Keller, Kevin Lane. Strategic Brand Management : Buil ding, Measuring and Managing Brand Equity. Upper Saddle River, N.J.: Prentice Hall, 1998. Kotler, Philip. Marketing Management. 11 ed. New York, NY: Prentice Hall, 2002. Levitt, Theodore. "Marketing Success through Differentiation--of Anything." In Harvard Business Review, 83: Harvard Business School Publication Corp., 1980. Liao, Tim Futig. Interpreting Probability Models: Logit, Probit, and Other Generalized Linear Models. Thousand Oaks, CA: Sa ge Publications, 1994. Livestock Marketing Information Center. Retail Meat Prices and Production Livestock Marketing Information Center [cit ed March 21, 2005]. Available from http://www.lmic.info/index.shtml . Long, J. Scott. Regression Models for Categoric al and Limited Dependent Variables. Vol. 7, Advanced Quantitative Techniques in the Social Sciences Series. Thousand Oaks, CA: Sage Publications, 1997. MacDonald, James M., Michael Ollinger, Kenneth Nelson, and Charles Handy. "Consolidation in U.S. Meatpacking." In Agricultural Economic Research AER785, 42: USDA, Economic Research Service, 2000. Maddala, G.S. Limited-Dependent and Qualitative Variables in Econometrics. 2nd ed, Econometric Society Monographs: Cambridge University Press, 1983. Martinez, Stephen and Hayden Stewart. "From Supply Push to Demand Pull: Agribusiness Strategies for Today's Consumers." Amber Waves (2003): 1-8. McFadden, D. "Conditional Logit Analysis of Qualitative Choice Behavior." In P. Zarembka, (Ed.) Frontier of Econometric s, 105-42. New York: Academic Press, 1973.
107 Medina, Sara and Ronald W. Ward. "A Mode l of Retail Outlet Selection for Beef." International Food and Agri business Management Review 2, no. 2 (1999): 195219. Menkhaus, Dale J., Damien P. Colin, Glen D. Whipple, and Ray A. Field. "The Effects of Perceived Product Attributes on the Perception of Beef." Agribusiness 9, no. 1 (1993): 57. Menkhaus, Dale J., James S. St. Clair a nd A. Zahedi Ahmaddaud. "The Effect of Industry Structure on Price: A Case in the Beef Industry." Western Journal of Agricultural Economics 6, no. 2 (1981): 147-54. Morrison Paul, Catherine J. "Cost Economies a nd Market Power: The Case of the U.S. Meat Packing Industry." The Review of Economics and Statistics 83, no. 3 (2001): 531-40. Nayga, Rodolfo M. Jr. and Zafar Farooq. "C onsumption of Conveni ence Meat Products: Results from an Explanatory New Jersey Survey." Journal of Distribution Research (1995): 33-37. Nguyen, Sang V. and Michael Ollinger. "Mergers and Acquisitions and Productivity in the U.S. Meat Products Industries: Ev idence from the Micro Data." Paper presented at the 2001 American Agricultur al Economics Association, Chicago, IL, August 5-8, 2001. National Panel Diary Group (NPD). "The Natio nal Eating Survey." National Panel Diary Group, Inc. Chicago, Illinois. 2001. Parcell, Joe L. and Ted Schr oeder. "Determinants of Beef and Pork Equity." Paper presented at the NCR-134 Conference on Applied Commodity Price Analysis, Forecasting, and Market Risk Management , St. Louis, MO, April 21-22, 2003. Pelzer, Pierre M.L., Dale J. Menkhaus, Glen D. Whipple, Ray A. Field, and Shawn W. Moore. "Factors Influencing Consumer Rankings of Alternative Retail Beef Packaging." Agribusiness (1986-1998) 7, no. 3 (1991): 253. Prewitt, Milford. "Beefed-up Meat Prices Lead to Menu Sea Change." Nation's Restaurant News, December 8 2003, 1-3p. Publix Asset Management Company. Food and Nutrition Center: Meat and Poultry, 2005 [cited June 2, 2005]. Available from http://www.publix.com/wellness/notes/D isplay.do?id=Food_Guide&childId=Mea t_And_Poultry . Purcell, Wayne D. "The Case of Beef Demand: A Failure by the Discipline." Choices, 2nd Quarter 1989, 16-19.
108 Rimal, Arbindra P. "Factors Affecting Meat Preferences among American Consumers." Family Economics and Nutrition Review 14, no. 2 (2002): 36. Shocker, Allan D., Rajendra K. Srivastava, and Robert W. Ruekert. "Challenges and Opportunities Facing Brand Management: An Introduction to the Special Issue." Journal of Marketing Research (JMR) 31, no. 2 (1994): 149-59. Skaggs, Rhonda K., Dale J. Menkhaus, Steven J. Torok, and Ray A. Field. "Test Marketing of Branded, Low Fat, Fresh Beef." Agribusiness 3, no. 3 (1987): 257. Van Osselaer, Stijn J. and Joseph W. Alba. "Consumer Learning and Brand Equity." Journal of Consumer Research 27, no. 1 (2000): 16. Verbeke, Wim, Ronald W. Ward and Jacque s Viaene. "Probit Analysis of Fresh Meat Consumption in Belgium: Exploring B. S. E. And Televi sion Communication Impact." Agribusiness: An international Journal 16, no. 2 (2000): 215-34. Ward, Ronald W. "Advertising and Promotions." In Agro-Food Marketing, edited by Daniel. I. Padberg, 319-48: CAB International, 1997.
109 BIOGRAPHICAL SKETCH Oscar Ferrara was born and raised in Asunc ion, Paraguay. In 1992 he earned a B.S. degree in agricultural engineering at the Un iversidad Nacional de AsunciÃ³n in Paraguay. Oscar worked for almost 8 years in ar eas related to agri cultural pr oduction and agribusiness consulting in Paraguay. In D ecember 1999, Oscar and his family relocated to the U.S. and in August 2001 he earned a B.S. degree in applied economics from the University of Minnesota. In 2003 he started the Master of Science degree in food and resource economics at the University of Fl orida. His research interests are focused on agricultural marketing and agribusiness development.