Assigment Food is a basic requirement of human life.pdf

CHAPTER 1: INTRODUCTION

Food is a basic requirement of human life; man’s hunt for food has evolved from

gathering and hunting to cultivating plants and domesticating animals for food. Today food is

produced and marketed on a very large scale. The major steps in modern day food marketing

include transportation, processing, and distribution of processed product to the retail outlet. The

increased length of the food marketing channel has added vulnerable points where food could be

exposed to contamination. Consumers are faced with choosing among an assortment of

competitively priced food products. A safe and nutritious food supply has contributed

significantly to human wellness, but food recalls can disrupt the trust and confidence of

consumers. To avoid food recalls quality control and hygiene assurance is very important. To

address the food safety issue government has developed food safety regulations, and to

implement these regulations various agencies were created.

1.1 Food Safety Agencies

Food safety has been an important issue for a long period of time. In the past, people

ensured that food was safe by improving manufacturing, handling, cooking and preserving

techniques. During the industrial revolution food began to be processed, packaged and marketed.

The introduction of refrigeration techniques dramatically increased the shelf life for food items.

The use of chemicals, increased handling, and the lack of regulation created numerous food

safety concerns. In 1862 president Abraham Lincoln founded the United States Department of

Agriculture (USDA). Later the USDA became the parent institution of several federal institutions

charged with ensuring that food in the U.S. is safe. The expansion of railroads fueled the

transportation of livestock and processed meat. The importation of livestock was identified as a

source of diseased livestock by the USDA secretary Isaac Newton in 1865. Upon the request of

the USDA secretary congress introduced a law to quarantine imported animals. In 1884 president

Chester Arthur established the Bureau of Animal Industry (BAI) within USDA, BAI’s focus was

on preventing the utilization of diseased animals for human food. In 1890 the initial Meat

Inspection Act was expanded to the export of salted pork and bacon, the act was amended further

in 1891 to cover all live cattle for export (USDA).

In 1905 author Upton Sinclair published a novel titled “The Jungle” depicting the

exploitation of meat industry workers and the threat it posed to meat consumers. Sinclair’s novel

led to a rise in public awareness for unsafe processing conditions. As a result of increased

consumer concern, the Food and Drug Act and Meat Inspection Act were passed in 1906. This

was an early example of how information can change consumers’ attitudes, which led to increase

government regulation to ensure food safety. In 1927 the Food and Drug Act was reorganized to

establish the Food, Drug, and Insecticide Administration and was housed under the USDA. In

1931 it was renamed the Food and Drug Administration (FDA) and transferred to the Federal

Security Agency, which became the Department of Health and Human Services. Since its

inception FDA has been an important institution charged with ensuring the safety of food in the

United States. Today FDA is responsible for assuring safety and security of human and

veterinary drugs, biological products, medical devices, cosmetics and a large portion of the

nation’s food supply. FDA regulates $1trillion worth of food products each year. FDA is charged

with ensuring safety of all the food products except for meat, poultry and some egg products

(Food and Drug Administration).

In 1950 and 1960 the growth of the interstate highway system and development in

refrigerated transportation of meat products by trucks and trains allowed the meat packaging

industry to grow at a much faster pace. While the focus was on visible contamination of meat

products like the use of diseased animals and faulty handling practices, and with the increase in

meat packaging concern over invisible hazards like chemicals used for treatment grew among

consumers. In 1977 the Food Safety and Quality Service (FSQS) was established. The FSQS was

charged with the responsibility of inspecting meat and poultry products, later it was renamed as

the Food Safety and Inspection Service (FSIS). In 1981, FSIS was one of the important

institutions under USDA which regulated food safety. In 1993 an outbreak of E.Coli caused the

death of four people and sickened 400 in the United States. The FSIS took steps to address this

food safety issue in 1996 by introducing a regulation called Hazard Analysis and Critical Control

Point (HACCP). HACCP’s focal point was on prevention and reduction of microbes that could

cause illness with raw meat products. FSIS has been a very important branch working under

USDA. As of today, FSIS is responsible for setting up appropriate food safety standards for meat

and poultry products, and performing inspection to make sure the standards are met by the

industry (Food Security and Inspection Service). Another agency helping to reduce food safety

risk in the United States is the Center for Disease Prevention and Control (CDC). Initially CDC

was established to fight malaria, over the years the role of CDC has grown. CDC is primarily

responsible for the prevention and control of infectious and chronic diseases, injuries, workplace

hazards, disabilities, and environmental health threats. However the CDC also monitors health

surveillance and helps prevent disease outbreaks including bioterrorism.

While agencies like the USDA have been charged with taking care of food safety, the

Department of Homeland Security (DHS) is charged with protecting the nation’s food supply

against any intentional attack. In response to the September 11, 2001 attacks the Department of

Homeland Security was created with the primary responsibility of protecting U.S. territories

from terrorist attacks and responding to natural disasters on U.S. territories. Since September 11,

2001 DHS has improved its coordination with other agencies and is also ensuring safety against

any biological terrorism or contamination attack (Department of Homeland Security). In 2002 it

enforced two provisions in the Bio-terrorism Act, the registration of facilities that manufacture,

process, or hold food for import into the U.S., and the prior notice of those shipments presented

for entry. According to a 2006 CRS report for congress, the appropriations and user fees for

agriculture-related homeland security activities in USDA and DHS had more than tripled from a

$225 million “pre-September 11” baseline in 2002 fiscal year to $797 million in 2006 fiscal year

(Monke, 2006). A presidential directive to protect critical infrastructure from 1998 did not

include agriculture and food, agriculture was added to the list in 2003. Also there have been

several bills introduced in Congress to authorize the funding or else improve the level of

preparedness and coordination of response to an agro-terrorist attack (Monke, 2006).

1.2 Major Food Recalls

In 1994, a Salmonella contamination of ice-cream was detected. The outbreak resulted in

the sickening of 740 people in 30 states. In 1997, over 2.6 million pounds of contaminated

strawberries were recalled after thousands of students across several states reported illnesses

from eating frozen strawberries in their school lunches. The strawberries were found to be

contaminated with Hepatitis A (Klein et al., 2009). In February 2008, USDA officials announced

that California based Westland/Hallmark Meat Company had recalled 143 million pounds of

ground beef. It was the largest meat recall in U.S. history. In July 2008 FDA issued a warning for

the salmonella contamination in the jalapeno peppers and Serrano peppers. In 2006, Spinach

contaminated with E.Coli O157:H7 caused several deaths and numerous illnesses. In 2005 and

2006 an outbreak of Salmonella contaminated tomatoes in four states, sickened hundreds of

people across the country. In March 2009 another major recall was announced, Peanut Butter

Corporation of America (PCA) recalled its 3,918 separate products made with PCA peanut

butter. The salmonella contaminated PCA products caused nine deaths and 714 illnesses in the

country, and cost government and food companies $1 billion (Flynn, 2009). In July 2009 after

FDA issued a warning against E.Coli O157:H7 contamination in Nestle cookie dough, Nestle

recalled the product from the market, CDC reported 76 persons from 31 states had been infected

due the outbreak. Table 1.1 presents the notable food recalls and food-borne illness outbreaks in

U.S. history.

Table 1.1 Notable Food Recalls and Foodborne Illness Outbreaks in U.S. History

Year

Food Recalls and Outbreak

1990

Salmonella Javiana contamination in Tomatoes, 174 ilnesses

1991

Cantaloupe contamination by Salmonella Poona, 400 illnesses

1994

Salomnella contamination in Ice Cream, sickened 740 people

1996

Raspberries contamination by Cyclospora cayatenensis, 1500 illnesses

1997

25 million pound of ground beef recalled for listeria contamination

1997

Hepititis A contamination in sliced frozen strawberries, 2.6 million pound strawberries

recalled

1999

35 million pounds of frozen, ready-to-eat meat products recalled

2002

27.4 million pounds of poultry product recalled

2002

ConAgra recalled 18.6 million pounds of ground beef

2006

E.Coli contamination in Spinach, 198 people sickened across 25 states

2007

21.7 million pounds of frozen meat patties were recalled due to e.Coli contamination

2008

143 million pounds of ground beef recalled

2009

salomnella contamination in PCA peanut butter products found, 714 illnesses

2009

Nestle cookie dough was recalled due to E.Coli contamination, 76 illnesses

Source: FDA, CDC, Klein et al. (2009)

In recent years consumers in the United States have witnessed many food safety events

one after another. The food borne disease outbreaks have been noticed from high risk

commodities to the healthiest and daily food products. A Center for Science in the Public Interest

(CSPI) report used data since 1999 from Center for Disease Prevention and Control (CDC) to

identify the top ten riskiest foods (Table 1.2) regulated by Food and Drug Administration (FDA)

(Klein et al., 2009).

Table 1.2 Top Ten FDA Regulated Riskiest Food since 1999

Food Porduct

Number of

Outbreaks

Reported Cases of

Illness

Leafy Greens

363

13,568

Eggs

352

11,163

Tuna

268

2,341

Oysters

132

3,409

Potatoes

108

3,659

Cheese

2,761

Ice Cream

2,594

Tomatoes

3,292

Sprouts

2,022

Berries

3,397

The top ten FDA regulated riskiest foods report noted that salmonella and E.Coli were the

two pathogens commonly associated with these outbreaks, salmonella alone accounted for 33

percent of outbreaks related foods regulated by FDA (Klein et al., 2009).

1.3 Media Coverage and Effect

In recent years, an increase in the number of food recalls and the rise of the modern news

media has made a notable impact on the food industry. Recent food recalls in the United States

have raised consumers’ concerns about food-borne illness and decreased their confidence in the

safety of the U.S. food supply. According to a Food Marketing Institute study, more than 80

percent of consumers expressed confidence in the safety of food they purchased in grocery

stores, but this percentage fell to 66 percent in 2007(Anonymous, 2008).

Even though efforts have been made to improve the quality of the food supply chain, food

safety has been increasingly perceived as an important health risk by consumers. However, in

recent years, consumers’ trust in the food supply has been eroded as the number of food recalls

increased 135 percent from 240 recalls in 2006 to 565 recalls in 2008 (Food Industry

Report, 4/14/09). Media has become an influential factor in altering consumers’ perceptions and

attitudes regarding food safety. The increased competition among major media sources like

television and newspaper has forced the various media outlets to compete and to follow

sensational news stories regarding food safety events (De Jonge et al., 2010).

With increased media coverage, consumers’ perception of the risks associated with the

food supply has increased. Today, consumers want to know if their food is safe or not. Therefore,

various studies have taken notice of the media’s impact on consumer confidence and have

studied the effects of media coverage on consumer confidence (De Jonge et al., 2010; Tansel,

1993; Kinsey et al., 2009). Numerous surveys and studies have been conducted to measure

consumers’ confidence in the safety of food (Degeneffe et al., 2009; Stinson et al., 2008), but

very few have tried to find linkages between the media and consumer confidence. Previous

researches have shown that media coverage of food related risks has a negative impact on

consumer risk perceptions (Frewer et al., 2002; Siegrist and Cvetkovich, 2001; Liu et al., 2004).

One example of research that found linkages between the media and consumer

confidence was a study published in 2009 (Kinsey et al., 2009). In the 2009 study, the authors

constructed two continuous food safety tracking (CFST) indices that measured consumer

confidence in food safety and food defense and consumer perceptions regarding how prepared

the food system is in dealing with food safety events. The indices were constructed by

aggregating frequency counts of individual responses from an ongoing weekly survey. A media

tracking index (MTI) was also constructed (Kinsey et al., 2009). The article found that changes

in media coverage significantly affected consumer confidence in the U.S. food supply.

The media coverage of a food safety event has the potential to magnify the impact of the

event on the industry experiencing the crisis. The media coverage impacts the industry by

inducing change in the perceived public risk and in political implications. The increased

perceived risk leads to a decline in demand, giving firms associated with the food safety event an

additional incentive to avoid the these events (Swinnen et al., 2005). These studies (Kinsey et

al., 2009) have shown that food recalls and food safety events have the potential of disrupting

consumer’s life and shaking their confidence in safety of the U.S. food supply. To better assist

and prepare consumers for these kinds of situations, it is necessary to gain a better understanding

of consumer attitudes and concerns (Degeneffe et al., 2006). The media coverage of a food safety

crisis affects the demand for the associated food product by increasing the perceived risk of

consuming the food product (Swinnen et al., 2005). The strong public outcry resulting from these

events may force governments to adopt regulations governing the import and export of the

commodity, thus affecting the trade of the commodity (Buzby, 2001). In the October 4, 2009

edition of New York Times reporter Michael Moss introduced readers to Stephanie Smith, a

children's dance instructor from Minnesota who was partially paralyzed from E. coli O157:H7,

after eating hamburgers produced by Cargill. This story was quickly popular in the media;

Stephanie sued Cargill for $100 million (Flynn, 2009).

One may argue that all consumers do not respond in the same way to the information

coming from mass media. People vary greatly in their attention to mass media, some actively

seek information, while others acquire it without much effort (McCombs and Shaw, 1972). To

address this issue, the research in the following chapters uses segmentation analysis to group

consumers into different segments according to their responses. The identification of consumer

segments will allow development of specific strategies by focusing on the needs of specific

segments of consumers.

Continuously tracking consumer confidence in the safety of the U.S. food supply chain

will allow this research to follow the trends in consumer confidence and to determine whether

there is a need for change in food safety regulations or practices. Additionally, the media tracking

will allow this research to test the influence of media stories on consumer confidence and will

provide information regarding the power of various types of media to communicate risks. Using

consumer segmentation will permit an estimation of the influence of media stories on consumer

confidence across different consumer segments.

1.4 Problem Statement

Between 1988 and 1992, a total of 2,423 outbreaks of food-borne disease were reported,

which caused a reported 77,373 illness cases (Bean et al., 1996). During the period of 19931997,

a total of 2,751 outbreaks of food-borne disease were reported, these outbreaks caused a reported

86,058 illness cases (Olsen et al., 2000). The number of reported cases for food-borne illness

increased to 128,370 cases and the number of outbreaks of food-borne diseases increased to

6,647 outbreaks during the period of 1998-2002 (Lynch et al., 2006). According to the Center for

Disease Control and Prevention (CDC), each year 76 million people contract food-borne

illnesses, 325,000 are hospitalized, and 5,000 die. Out of these 76 million food-borne illness

cases, known pathogens account for an estimated 14 million illnesses, 60,000 hospitalizations,

and 1,800 deaths. Among these cases, three pathogens, Salmonella, Listeria, and Toxoplasma, are

responsible for 1,500 deaths each year, more than 75% of those caused by known pathogens.

While unknown agents account for the remaining 62 million illnesses, 265,000 hospitalizations,

and 3,200 deaths (Mead et al., 1999). According to the USDA Economic Research Service

(ERS). Food-borne illnesses cost the U.S. economy at least $6.9 billion a year (ERS-USDA,

2005). In today’s world, food related risks extend beyond natural contamination or contamination

during production/processing. Incidents like the September 11, 2001 terrorist attacks and the

London subway bombing have raised issues regarding safety of United States food supply chain.

A survey of U.S. consumers found that 89% of consumers considered food safety more important

than safe drinking water, crime prevention, health and nutrition, and the environment (2004).

Since most American citizens are unfamiliar with food safety protocols and the risks

associated with food borne illnesses, food safety issues are unlikely to be noticed by these

individuals. Therefore, most consumers are expected to acquire their information and knowledge

about food safety events and food recalls from mass media (Kinsey et al., 2009). Consumer

confidence perceived food safety risks affected by food safety events and media. Although how

media and other factors play a role in affecting consumer perception are not well understood.

The present study uses the most recent data from the Continuous Food Safety Tracking (CFST)

survey to conduct an individual-level analysis (rather than aggregate) of the media agendasetting

effects on consumer confidence in the U.S food supply. The use of individual-level analysis will

allow this research to analyze the effects of socioeconomic factors on consumer confidence in

the U.S. food supply.

1.4.1 Specific Objectives

• To estimate media agenda-setting effects on consumer confidence in the safety U.S.

food supply chain.

• To analyze how the agenda-setting effect varies across consumer segments.

1.5 Organization of Study

The remaining chapters in this study follow the following order. The second chapter

surveys past research on the topics covered by this study. Chapter three describes the data

collection procedure, survey design, variables measurement and modeling and methods applied

for the analysis. The fourth chapter presents the discussion and interpretation of the results

obtained in the analysis. The last chapter focuses on the conclusions and suggestions from the

study.

CHAPTER 2: LITERATURE REVIEW

Researchers have tried to get a better understanding of the factors that determine

consumer confidence in order to develop and apply better risk management and communication

strategies. There are various determinants which shape consumer confidence. De Jonge et al.

(2007) stated that consumer confidence in the safety of food consists of two dimensions -

optimism and pessimism. Trust and consumer confidence in the safety of product groups act as

the basis for optimism about the safety of food, while pessimism is affected by individual

difference variables like food allergies and trait worry. The results from the study indicated that

to a significant extent optimism and pessimism about the safety of food developed from

consumer trust in regulators and actors in the food chain and the perceived safety of meat and

fish rather than other product categories. The study also found that notion of optimism and

pessimism are distinct and they are influenced by different determinants (De Jonge et al., 2007).

The article also found that consumers’ recall of food safety incidents affected the consumers’

level of optimism and pessimism differently. Consumers who recalled the food safety incidents

were not less optimistic relative to consumers who did not recall food safety incidents, but the

consumers who recalled food safety incidents were more pessimistic than the consumers who did

not recall food safety incidents (De Jonge et al., 2007).

Marsh et al (2004) investigated the impact of meat product recall events on demand of

beef, pork, poultry, and other consumption goods in the United States. The Food Safety

Inspection Service’s meat recall events and the newspaper reports over the period 1982–1998

were used to develop beef, pork and poultry recall indices. Findings from the study indicated that

the Food Safety Inspection Service’s announcement of meat recall events significantly impacted

the demand of beef, pork, poultry and other consumption goods in the United States (Marsh et

al., 2004).

Studies have tried to differentiate consumer attitudes towards food safety based on type

of the food safety issues concerned. Brewer and Prestat (2002) surveyed consumer attitudes

about the safety of the food supply in general, and related these general concern levels with the

groups of specific items of concern, regulatory issues and prioritization of food safety finding

areas, and compared them with results from a 1994 study. The study factored the consumer

responses in six factors and later used MANOVA and univariate ANOVA to analyze the effect of

general concern levels on the specific food safety areas. The study found that consumer priorities

for various food safety concerns did not follow the same trend as for the general food safety

concerns. The consumers were less concerned about chemical and health issues and were more

concerned about regulatory issues (Brewer and Prestat, 2002). Authors Brewer and Rojas

(2008) conducted a similar study. They found that the consumer concerns regarding

microbiological issues was substantially higher than past studies, also concern over regulatory

issues increased dramatically compared to the past studies (Brewer and Rojas, 2008).

According to Ajzen and Fishbein (1980), information plays a vital role in altering

consumers’ beliefs, attitudes, and choices. A person’s understanding about food safety events, the

government and industries protocols for managing food safety risks, and individual perceptions

of the consequences are estimated to be alleviated by the person’s belief and attitudes about food

safety (Kinsey et al., 2009; Ajzen and Fishbein, 1980). Studies that have applied this concept

have found both direct and indirect effects on consumer confidence in the food supply. A study

by Han and Harrison (2007) investigated the linkages between consumer beliefs and attitudes

regarding the risks and benefits of genetically modified foods and consumer purchase intentions

(Han and Harrison, 2007). Another study by Moon and Balasubramanian (2004) demonstrated

that trust, sense of outrage, and socio-demographic factors play an important role in shaping

public attitudes (Moon and Balasubramanian, 2004).

Nayga (1996) studied the socio-demographic factors that affect the concern level of main

meal planners in households. Although he did not study whether the sources food safety

information had any effect on the consumer concern level, he did suggest that the source of food

safety information might affect consumer concern (Nayga, 1996). The notion that the media

frames the way people think about certain issues, and in doing so, influences the public’s

attitudes about said issues is referred to as the media agenda setting effect (Kinsey et al., 2009;

McCombs and Shaw, 1972). Mass media plays a very influential part in framing peoples’

thinking. “Media may not tell us what to think, but they tell us what to think about” (McCombs

and Shaw, 1972).

Mass media information and reports affect how individuals frame their thinking. Even

sound information about a food recall event will add to consumers’ knowledge. Most individuals

learn about risk through media; this may cause “risk amplification,” where media work as a

amplification station (Kasperson et al., 1988). According to Zucker (1978), if the individual has

less direct experience and/or knowledge about an issue, he or she is more likely to rely on mass

media for information about the issue and more likely to be influenced by the agenda setting

effect (Zucker, 1978). Kornelis et al. (2007) conducted a nationally representative survey in the

Netherlands to examine the preferences of consumers for the different information sources when

they have a question about the food safety. The empirical results from the study indicated that

two-thirds of the consumers were selective in their use of the information sources and prefer

either the institutional or the social sources (Kornelis et al., 2007).

The media coverage of food safety events has also had an impact on trade and export

markets and an effect on the political atmosphere. An article by Swinnen et al. (2005) provided

an empirical framework from the two food safety crises in Europe. The article concluded that the

food safety crises provoked strong consumer responses, had considerable impacts on export

markets, and led to important political implications (Swinnen et al., 2005). The study by Piggott

and Marsh (2004) developed an empirical framework to investigate if food safety information

surrounding beef, pork and poultry had an impact on the consumption of meat in the United

States. The study used LexisNexis academic version tool to search the top fifty newspapers in

the country for any news about the food recall events with certain keywords in it. The data series

collected was used to create a food safety index that can measure the impact of information on

the consumption of meat in the United States. The study found that undesirable publicity in

relation to food safety concerns do have statistically important own- and cross commodity

impacts on demand for meat in the U.S. The study also found that average impact of these effects

have been economically small over the last several decades (Piggott and Marsh, 2004).

In a 2000 study Verbeke et al. used probit analysis to investigate the impact of BSE

(Bovine spongiform encephalopathy) and television communication on fresh meat consumption

in Belgium. The study focused primarily on assessing the impact of television coverage on the

fresh meat consumption. It found that television coverage had a highly negative impact on

decision making toward fresh red meat consumption. The study also revealed that the likelihood

of reducing the fresh meat consumption was increased with the presence of younger children in

the household and also with the increase in age of consumer. It was also found that younger

people’s decision were more susceptive to media coverage relative to the older age groups

(Verbeke et al., 2000). The food safety event and food recalls has also affected stock markets, a

study by Wang et al. (2002) investigated the effects of five food recalls on the stock returns of

the two affected companies. The results from the study indicated that around the event time

period, first recalls had significant negative effects on the daily stock returns for both the

companies. The study also found that the value of firms plummeted initially, but not in the

subsequent events, which mean that the risks related to recalls were expected by the investors

once the market became sensitized to the food safety issues (Wang et al., 2002).

Studies have shown that people accept negative information presented by media more

quickly than the positive information (Siegrist and Cvetkovich, 2001; Verbeke and Ward, 2001).

Liu et al. (2004) conducted a case study of milk contamination to demonstrate the demand

adjustment process to a temporarily unfavorable shock. The results from the study indicated that

effects of positive and negative information to adjustment of consumption and risk perception

were asymmetric over time, it also indicated that positive media had a lag period and positive

media coverage could help reduce the loss of consumption (Liu et al., 2004).

Swinnen et al. (2005) noted two kinds of media, quality media and the popular media.

Popular media results from competing media outlets that are intensely covering popular events

like food safety recalls. It is characterized by intense coverage in the early periods, followed by a

rapid loss of interest. The competition and selectivity of reporting leads to bias in the treatment

of the situation and a development of a mostly negative tone (Swinnen et al., 2005). Studies have

been conducted to estimate the impact of negative information and positive information (Verbeke

and Ward, 2001; Smith et al., 1988). In a study Ten Eyck (2000) investigated how the food safety

issues were marginalized by reporters, as the mass media coverage tends to cluster around crisis

situations. The study collected media stories from 1986-1997 to study the effect of the

information, in addition to it the article also investigated two food safety issues- mad cow disease

and the Alar event. The study suggested that media coverage tended to cluster around the food

safety crisis (Ten Eyck, 2000).

Verbeke and Ward (2001) showed that TV coverage of health risk related to meat

consumption had a negative impact on meat consumption. The study also showed that the higher

negative TV coverage may have outweighed the industries’ advertising efforts to increase the

consumption. In 1988, a study by Smith et al. (1988) sought to estimate lost sales following a

food contamination incident of heptachlor contamination of fresh fluid milk in Oahu, Hawaii.

The study found that the media coverage following the milk contamination incident had a

significant effect on milk purchases, and it also found that the negative media coverage had

outweighed the positive media coverage (Smith et al., 1988).

Currently, less than two percent of the U.S. population is engaged in agricultural

production, and the average consumer has little knowledge of the agricultural and food

production system. As a result, consumers often rely on mass media for relevant information

about food safety (Kalaitzandonakes et al., 2004). It has been argued that mass media can play an

important role in building or undermining consumer confidence in the safety of foods,

particularly because consumers have limited ability to assess food safety prior to consumption

(Verbeke et al., 1999).

Media coverage of food safety issues has primarily been studied in relation to specific

food incidents and food products (De Jonge et al., 2010; Verbeke et al., 1999). De Jonge et al.

(2010) addressed how daily media reporting on the totality of food safety events may accumulate

to affect consumer confidence in the safety of food. This was accomplished by monitoring

actual newspaper coverage about food safety issues in parallel to evidence from consumer

recollections of the food safety incidents (De Jonge et al., 2010). The results from a 2008 study

by Stinson et al. showed that United States residents are low in confidence level that the nation’s

food supply is safe from natural or accidental contamination. The results of study also showed

that consumers’ concern over food defense has grown and that the public was holding the

government increasingly responsible for food defense and food safety (Stinson et al., 2008).

These studies demonstrated that information-processing strategies substantially mediated the

relationship between the local news media and the public’s perception of the food safety, with

elaborative processing being more influential than active reflection in people's learning from the

news media (Fleming et al., 2006).

Credibility of the information sources also matters in determining consumer attitudes

towards food safety. A study by Bruhn and Schutz (2007) found that the Science magazines were

considered highly reliable by more persons than the food or news magazines. Also Television

was considered reliable by fewer people than print media (Bruhn and Schutz, 2007).

In addition to media coverage, the socioeconomic characteristics also affect consumers’

confidence. Generally, women are likely to be less confident about the safety of food relative to

men (De Jonge et al., 2004). A study by Verbeke and Viaene (1999) surveyed meat consumers in

Belgium for their attitudes toward meat. The study found that male consumers attach more

importance to attribute of food safety than female consumers (Verbeke and Viaene, 1999). Also,

a person with more education is less worried about food safety issues (Dosman et al., 2002). A

study by Herrmann et al. (1998) investigated the public reaction to the 1989 Alar crisis in the

wake of extensive media coverage. The Alar crisis was the result of a report that apples became

hazardous for consumption after being treated with Alar, a chemical. The study found that

awareness of the crisis was greater among older adults, those with higher formal education, and

those who reported frequent television news viewing (Herrmann et al., 1998).

Burger (1998) investigated gender differences in the attitudes about the safety of fish by

interviewing 197 men and 94 women. Significant differences were found among genders,

women believed that it was less safe to eat fish relative to men (Burger, 1998). In the study men

significantly perceived small fish as safer than large fish, but women did not. Also People

consistently believed it was safer to eat fish they caught themselves or bought in a fish market

than those from a supermarket (Burger, 1998). A study by Fein et al (1995) used the data

collected from the two national telephone surveys conducted in 1988 and 1993 to describe

consumer perception of food-borne illness. The study further used1993 data to assess the

relationship between the perception that a food-borne illness had recently been experienced and

awareness, concern, knowledge, and behavior related to food safety. The study found that people

from 18 to 39 years of age in both surveys were more likely to believe that they had experienced

a food-borne illness relative to those in other age groups. The study also found that people with

at least some college education were more likely to believe they had experienced food-borne

illness than those with less education (Fein et al., 1995).

Studies have been conducted to see how the socio-demographic factors influenced the

perception among the main meal planners of households regarding food safety (Nayga, 1996;

Lin, 1995; Wilcock et al., 2004). Nayga (1996) found that consumers with a high level of

education, a high income level, or were female tended to be highly concerned about food safety

issues. The study by Lin (1995) examined how the main meal planner’s socioeconomic or

demographic characteristics influenced his or her beliefs about the importance of food safety in

food shopping.

For a long period of time, private industry has been using a segmentation approach to

understand consumer group preferences. The segmentation approach is recognized as being

better at explaining or predicting consumer behavior than conventional demographics.

Convincing consumers to adopt better risk protection practices can be accomplished more easily

by segmenting consumers since the food safety messages can be customized specific to each

consumer segment (Rimal and Real, 2006). Klontz et al. (1995) conducted a telephone survey of

1,620 respondents to assess the prevalence of selected self-reported food consumption and

preparation behaviors associated with increased risks of food-borne illness and the demographic

characteristics related to such behaviors. Study found that the persons who were female, were at

least 40 years old, and had a high-school education or less consistently reported safer food

consumption and preparation behaviors (Klontz et al., 1995).

Various studies have surveyed consumers and grouped them into segments for effective

targeting based on their preferences regarding food safety (Baker and Crosbie, 1993; Kennedy et

al., 2008). Kennedy et al. (2008) surveyed U.S. consumers and found five distinct consumers

segments with differences in their food safety preferences. The study found that

sociodemographic characteristics like education, income, person with allergy in the household,

and person under the age of six living in the household, varied significantly for each of the

consumer segment. Baker and Crosbie (1993) used cluster analysis to construct consumer

segments based on the structure of their individual preferences regarding food safety. The study found

substantial differences among all three consumer segments in paying for certified food produce.

The traditional segmentation approach segments or groups individuals based on their past

choices or behavior (Baker and Crosbie, 1993). A study examined knowledge levels about food

safety practices, food safety and food science for consumers in Ireland. This study used

hierarchical cluster analysis to segment the population in four different segments, the study also

proposed target specific promotions for food safety promoters based on these segments

(McCarthy et al., 2007).

In a study by McGuirk et al (1990), the groups of consumers reporting similar food safety

concerns and shopping behaviors were identified using cluster analysis. The study found

significant differences in the perceptions and reactions regarding the food safety hazards, these

differences were used in study to derive important implications for food marketing strategies and

food safety policies (McGuirk et al., 1990). In a 1990 study by Funk and Phillips, benefit

segmentation procedure was used to segment the market for table eggs in Ontario, Canada. Study

followed the agglomerative hierarchical cluster analysis to identify four market segments. Once

identified, profiles of these four segments were developed using beliefs about eggs, attitudes

toward eggs, lifestyle factors, health and nutrition consciousness, media habits, consumption

habits, and demographics. Based on the profiles developed for the market segments, the study

suggested promotional programs and marketing strategies (Funk and Phillips, 1990).

In contrast to this approach, predictive segmentation segments individuals based on how

they might respond in the future. A study by Degeneffe et al. (2006) used predictive

segmentation to analyze the attitudes of U.S. consumers’ regarding terrorism by segmenting the

consumers into six segments. The study demonstrated the value of consumer segmentation in

that it can be stretched beyond the traditional marketing applications (Degeneffe et al., 2006).

Another study by Degeneffe et al. (2009) used predictive segmentation to segment U.S.

consumers for food communication strategies. The study supported the development of a

communication strategy that anticipates the reactions of U.S. consumers in the event of another

terrorist attack. Both of these studies identified six consumer segments, based on the pattern of

response given by consumers for independent (attitude/value statements) and dependent

(concerns/expectations relating to terrorism) measures; both also used demographic profiles to

identify the segments (Degeneffe et al., 2009).

CHAPTER 3: DATA AND METHODOLOGY

3.1 Consumer Survey Design

The survey design was developed after earlier surveys conducted by The Food Industry

Center at the University of Minnesota and LSU AgCenter with funding from the National Center

for Food Protection and Defense (Stinson et al., 2008; Degeneffe et al., 2009). The survey asked

questions about consumers’ attitudes towards terrorism in general and about food defense and

food safety, and after defining the difference between the two terms to the respondents. These

surveys and the current continuous survey are administered via the internet with respondents

selected from Taylor Nelson Sofres’ (TNS) national online panel of more than two million U.S.

consumers. Respondents are contacted by TNS and invited to come to a website to complete a

survey. The sample of respondents is selected in such a way that it comprises a nationally

representative cross section of consumers by geographic region, income, household size, and age

of respondent. A six point Likert scale is used to indicate the strength of positive and negative

attitudes for each question. Data collection started on May 8, 2008. This thesis uses consumer

survey data collected over 67 weeks, from May 2008 to August 2009.

3.2 Media Tracking

To assess the impact of media on consumer confidence a media index is needed.

Different approaches have been tried to create a media index to analyze the effect of media,

Tansel (1993) used dummy variables to measure the effect of anti-smoking campaigns on the

cigarette demand in Turkey (Tansel, 1993). A study by Smith et al (1988) used the actual

newspaper article counts, marked as positive or negative to analyze the impact of media

coverage on a specific event (Smith et al., 1988). Chang and Kinnucan (1991) used a cumulative

number to create a media index to analyze the impact of cholesterol information on the

consumption trends of fats and oils (Chang and Kinnucan, 1991). A study by Burton and Young

(1996) used a media index for BSE and incorporated it in the AIDS model for meat demand. The

study used an indicator created by the count of newspaper articles that mentioned BSE. The

article used the indicator in two ways, the number of articles per quarter to measure the transitory

effect on meat expenditures and as the cumulative number of articles as a modifier for long run

relationships. The study found that the media coverage for BSE had a significant effect on the

allocation of consumer expenditure among the meats (Burton and Young, 1996). Similar to the

study by Burton and Young (1996), Verbeke and Ward (2001) developed a media index as a

measure of television coverage and negative press related to fresh meat issues. The study

observes media stories from TV coverage and keeps track of positive and negative stories

separately. However, the study found most of the media coverage was based on the negative

stories, also the correlation between negative stories and difference was 0.98, making it

impractical to weigh the negative and positive stories separately (Verbeke and Ward, 2001). The

study also included a five period lag for TV stories, thus effectively extending interval to a

period of 6 months for the negative press.

A study by Kinsey et al (2009) developed a media index. The index included several

media sources which were weighted according to consumer responses, this research incorporates

the media tracking index created by Kinsey et al (2009). A food safety media tracking index

(MTI) was constructed during the same 67 week period by investigators at the Louisiana State

University Agcenter. The MTI is constructed from article counts associated with food safety

events from selected newspapers and/or television news programs in the United States. The reach

of media intensity is not fully reflected by article counts as media exposure varies by the media

type and the nature of the event. These shortcomings are addressed by constructing a media

index. The media index incorporates the respondents’ use of selected media types and normalizes

article/transcript counts across media types. The formula used for normalizing media counts is

X - Min(X ) Zkt =

kt k

(1)

Max (Xk ) - Min(Xk )

where Zk is the standardized score for media source k during week t, Xkt is the article/transcript

count for media source k during week t, and Min(Xk) and Max(Xk) are the minimum and

maximum counts for the kth media source over the sample period (Arundel et al., 2002; Kinsey et

al., 2009). The X’s are the article or transcript counts of news stories containing at least one of

the following key words: food safety, food defense, food terrorism, agricultural terrorism or

agterrorism, food poisoning, food contamination, food borne illnesses, food-borne diseases, and

food recall. The media sources included for keyword searches were: national and local

newspapers, network and cable TV, radio, news magazines, and the internet.

The next step in construction of the media tracking index involves aggregation of

standardized scores using the following formula:

MTIt = ∑ wkZkt, (2)

Where MTI is the media tracking index value for week t and wk is the weight assigned to the kth

media source where ∑wk = 1 and 0 ≤wk ≤1. Each respondent in the survey was asked to indicate

which of the selected media outlets they considered their primary source of news. Frequency

counts from these questions were used as estimates for the weights in equation 2.

3.3 Factor Analysis

Factor analysis helps in defining the structure of the interrelationships among variables in

the analysis. Factor analysis yields groups of variables, which are highly inter-correlated, and

describes an underlying dimension in the data. In this analysis, rather than using every attitudinal

question as a dependent variable, we identify factors of these attitudinal questions by use of

factor analysis. Each factor identified in the study represents a dimension indicating consumer

confidence.

To reduce the ambiguity and achieve simpler and more meaningful factors, factor rotation

was performed. In factor rotation, the reference axes of the factors are turned about their origins.

The factor rotation can be performed in two ways, orthogonal factor rotation, where reference

axes are maintained at 90 degrees, or oblique factor rotation, where reference axes do not need to

be at 90 degrees. A VARIMAX orthogonal factor rotation method was used in this analysis.

VARIMAX factor rotation is a more popular and widely used factor rotation method, and it gives

clearer separation among factors. This method maximizes the sum of variances of the required

loadings in the factor matrix (Hair et al., 2008). A raw VARIMAX procedure tends to give equal

weights to variables with lower and higher communality. Thus Kaiser

Normalization was performed before using VARIMAX rotation. Kaiser Normalization, founded

by James Kaiser, divides each factor loading within a factor structure by the square root of the

communality of that factor. The Kaiser normalization is removed after rotation by multiplying

each loading by the square root of communality of the variable in each row (Harris, 2001).

3.4 Segmentation

Market segmentation has been a very important concept in marketing since it was

introduced for the first time by Smith. Smith recognized that “success in planning successful

marketing activities requires precise utilization of both product differentiation and market

segmentation as components of marketing strategy” (Smith, 1956). Various methods can be used

for market segmentation, but at first they can be classified in two groups a priori and post-hoc

methods. A segmentation approach is a priori when the numbers of segment are determined in

advance, while in post-hoc method the numbers of segment are determined on the basis of data

analysis. The other way to classify is depending upon the statistical method used predictive or

descriptive. The methods used for market segmentation include Contingency tables,

Discriminant analysis, clustering procedures, out of these methods cluster analysis is widely

popular among researchers (Wedel and Kamakura, 2000). As mentioned before the purpose of

segmentation is to identify groups of individuals with common attitudes and values.

Conventionally, consumers have been placed in the segments based on their past response or

behaviors, but the predictive segmentation approach uses past responses and behaviors to

examine how consumers might respond, to the same situations or changes in situation. In 1994,

Baker and Crosbie (1994) developed market segments based on consumer preferences for food

safety and other product attributes. Study used conjoint analysis to identify preferences and

cluster analysis was used to identify the market segments (Baker and Crosbie, 1994). A study by

Verbeke and Vackier (2003) based on cross-sectional data collected in Belgium investigated the

consumer profile and effects of consumer involvement in fresh meat as a product category

(Verbeke and Vackier, 2003). A principal component factor analysis was performed followed by

hierarchical cluster analysis procedure to assess the consumer profile towards fresh meat. Results

identified four segments significantly different in their approach to involvement in the fresh meat

(Verbeke and Vackier, 2003).

3.4.1 Cluster Analysis

The purpose of cluster analysis is to place objects into groups or clusters, such that

objects in a given cluster tend to be similar to each other in some sense, and objects in different

clusters tend to be dissimilar. In cluster analysis, the clusters are created on the basis of distance

or proximity. The factor mean scores of the factors obtained in the earlier procedure were used to

define the clusters.

Among the various methods for cluster analysis, the two are most commonly used are

hierarchical cluster analysis and partition cluster analysis. Hierarchical cluster analysis uses a

hierarchy or a dendogram (treelike structure) to identify the clusters in the dataset. This

procedure produces N-1 number of clusters, with N being the number of observations. There are

some pros and cons for hierarchical cluster analysis. Hierarchical cluster analysis is simple, since

there are various measures of similarities available for clustering, and it generates different

clustering solutions. However, hierarchical cluster analysis may lead to early undesirable

combinations and confusing results. Also, outliers may have a major impact on clustering

procedure.

The advantages for partition clustering are that it is less vulnerable to outliers in the data

and that large data sets can be clustered in less time. The disadvantages for the partition

clustering method are that the number of clusters needs to be specified beforehand and that the

initial seed points need to be specified. Even specification of seed points does not guarantee an

optimal clustering solution.

Therefore, in this study we use both steps for clustering the dataset, hierarchical and

partition clustering procedures. Using STATA, hierarchical cluster analysis with Ward’s

minimum variance method was used to calculate the distance between observations. In Ward’s

minimum variance method, the distance between two clusters is the sum of squares between the

two clusters summed over all the variables.

From the clusters generated in hierarchical clustering, the mean value for each cluster was

calculated. These mean values were used as initial points for a K-means clustering procedure. In

the K-means clustering procedures, each observation was assigned to the nearest seed to form the

temporary cluster. The seeds of these temporary clusters were then replaced by the means of

temporary clusters. Euclidean distance was used as the means to cluster the observations in

partition clustering. Euclidean distance is the measure of the length of a straight line drawn

between two objects when represented graphically (Hair et al., 2008).

3.5 Model and Analysis Procedure

This study proposes that the media coverage for food safety/ defense events will affect

consumers’ confidence in the safety of the U.S. food supply chain and the preparedness of the

U.S. food supply chain. There are several different factors affecting consumers’ confidence in the

food supply chain (Figure 3.1).

Figure 3.1 Theoretical Framework of Consumer Confidence in the United States Food

Supply Chain

Figure 3.1 presents choice process model introduced by Engel et al in 1978 and was also

used by Han and Harrison in 2007. The choice process model is used as conceptual basis for

model specification in this study. Figure 3.1 shows the theoretical framework of the study. It

shows socioeconomic and demographic factors influence the information and beliefs of the

consumers. Information sources play role in determining consumer beliefs regarding food safety.

Information and belief both affect consumer attitude regarding food safety. The perceived risk

and beliefs about food safety directly affects consumers’ confidence in the food safety.

Socioeconomic/Demographic Factors

Education, Income, Age, Occupation,

etc.

Beliefs

Knowledge of Food

Supply Chain

Perceived Risk and Consequences

of Specific Event on Health

Attitude

Confidence in Food Supply Chain

Trust in the Preparedness of Food

Supply Chain

Information

Use of Media

Media Coverage of Food Safety or

Security Events as measured by

Media Tracking Index

Adapted from Han and Harrison 2007, and

Engel et al. , 1978

The Kinsey et al. (2009) study identified two primary indicators of consumer’s

confidence. The first measures consumer’s current confidence in the safety of U.S. food system,

and the second measures their belief regarding how better prepared the food system is regarding

food safety relative to a year ago. This was accomplished using factor analysis separate

attitudinal questions in the survey into two sets of questions (appendix E). All the questions

included in these two sets use a likert scale that ranges from 1 to 6. The first set of questions

measures level of concern about food safety, or inversely their confidence in the safety of food (1

being Not At All Concerned to 6 being Extremely Concerned). In order to measure the consumer

confidence, the scale for these four questions is reversed (1 being Extremely Concerned to 6

being Not At All Concerned). Responses for these four questions are aggregated to obtain a new

aggregated variable to measure respondents’ confidence in the safety of our food, and it is scaled

from 4 to 24. For the ease of interpretation and calculation these 21 categories were consolidated

into 7 categories, three categories were summed together to form a single category e.g.

categories 4, 5, and 6 were added together to form category 1 and so on to create 7 new

categories.

The second set of questions obtained from factor analysis measures respondents’ attitudes

regarding how prepared we are for food safety/defense events compared to a year ago.

Responses for questions in the second set were aggregated together to obtain a new aggregated

variable to measure respondents attitudes regarding how prepared we are for food safety/defense

events compared to one year ago, and it is scaled from 2 to 12.

In order to measure the effect on the consumer confidence, the data for each section of the

figure 3.1 was collected, through a consumer survey and media tracking survey. In order to use

the collected data to measure the consumer confidence in the food safety/ defense events

regression model was developed.

CSFTCi = f(MTI, MEDIASOURCE, DEMOGRAPHICS)

Where, where CFSTC is the aggregated variable measuring consumer confidence in food

safety, i represent each segment for which the model was run, MTI represents the Media

Tracking Index, MEDIASOURCE is the different primary media sources used by consumers.

A similar model was developed to measure consumers attitudes towards preparedness of food

supply chain in different consumer segments.

CSFTPi = f(MTI, MEDIASOURCE, DEMOGRAPHICS)

Where, CFSTP is the aggregated variable measuring consumer attitudes regarding

preparedness of U.S. food system to deal with food safety/defense events, i represent the each

segment for which model has been calculated. MTI represents the Media Tracking Index,

MEDIASOURCE is the different primary media sources used by consumers.

3.6 Ordered Probit Analysis

Since the dependent variables are ordinal, an ordered probit model is used for the

analysis. The ordinal regression model is commonly presented as a latent variable model with a

structural equation specified as, yi* = xiβ + εi , where yi* is a latent variable ranging from -∞ to

∞. This model is derived from a measurement model in which yi* is mapped to an observed

variable y which is thought of as providing incomplete information about an underlying y*

according to the measurement equation (Long, 1997).

The use of an ordered probit model provides two primary advantages over the OLS

model. First, the ordered probit model provides a solution to the problem of heteroskedasticity,

which occurs when a regression model is used to analyze a categorical dependent variable; and

second, maximum likelihood estimates are, under general conditions, consistent, asymptotically

efficient, and asymptotically normal (Hamath et al., 1997).

The probability of observing y = m, of CFSTC taking a value 1 to 7, or of CFSTP taking

a value 2 to 12 given x is,

Pr(y = m | x) = 𝐹(τm − xβ) − 𝐹(τm − 1 − xβ)

where F is the Cumulative Distribution Function for εi(Long, 1997).

This research uses several explanatory variables in the model, like demographic variables

and media source variables; but the variable of primary interest is the Media Tracking Index

(MTI). As described above, the media tracking index is constructed from daily article counts and

is a continuous variable.

To be consistent with the previous study done by authors of this article, and according to

theory of media agenda setting, we hypothesize that media coverage has a negative effect on

consumer’s confidence. Higher media coverage is expected to induce decline in consumers’

confidence in the safety of our food.

CHAPTER 4: RESULTS AND DISCUSSION

This chapter discusses the results of the Consumer Food Safety Tracking (CFST) survey and

the media tracking. The analysis used 12,236 observations from the Consumer Food Safety

Tracking (CFST) survey, collected over a period of 67 weeks. The media tracking data was

collected over the same period of time.

4.1 Consumer Demographics

The summary statistics of the demographic distribution of the respondents for Continuous

Food Safety Tracking (CFST) survey are presented in Table 4.1 below. To collect a

representative sample of the U.S., the country was divided into nine geographic regions; the

survey data was collected from these nine geographic regions. The highest number of responses

came from the Middle Atlantic, East North Central, South Atlantic, and Pacific regions of the

country. More than 56% of our respondents had household age of 50 years or more. The number

of respondents for the survey was evenly divided across the household income category. Over

33% of the respondents had a household income of over $75,000 or more, while 26% of the

respondents had a household income of $30,000 or less.

Continuing with the household size demographic characteristic from Table 4.1, 27% of our

respondents had a household size of one member and 40% had two members. Only seven percent

of our respondents had a household size of five or more members. The survey respondents were

highly dominated by one race; around 90% of the survey respondents were white. The survey

response was weighted toward female respondents, in that nearly 80% were female. Around 28%

of the respondents had some college education and no degree; almost 25% of the respondents

had a bachelor’s degree; and14% of the respondents had a post graduate degree.

Table 4.1 Frequency Distribution of the Demographic Characteristics of Survey

Respondents

Charecteristic Category (n=12,236) Frequency Percent

Geographic New England 657 5.37

regions Middle Atlantic 1,842 15.05

East North Central 2,022 16.53

West North Central 900 7.36

South Atlantic 2,300 18.8

East South Central 671 5.48

West South Central 1,237 10.11

Mountain 871 7.12

Pacific 1,736 14.19

Household Age Under 30 Years 1,068 8.73

30 through 39 Years 1,810 14.79

40 through 49 Years 2,352 19.22

50 through 59 Years 2,811 22.97

60 Years and Over 4,195 34.28

Household Size 1 Member 3,393 27.73

2 Members 4,875 39.84

3 Members 1,714 14.01

4 Members 1,340 10.95

5 or More Members 914 7.47

Race No Answer 123 1.01

White 10,953 89.51

Black/African-American 576 4.71

Asian or Pacific Islander 304 2.48

American Indian, Aleut Eskimo 60 0.49

Other 220 1.8

Marital Status No Answer 30 0.25

Now Married 7,071 57.79

Never Married 2,163 17.68

Divorced, Widowed, Separated 2,972 24.29

Gender Male 2,504 20.46

Female 9,732 79.54

Education Grade School 18 0.15

Some High School 190 1.55

Graduated High School 2,477 20.24

Some College-no degree 3,432 28.05

Graduated College –Associate’s Degree (2 years) 1,269 10.37

Graduated College- Bachelor’s Degree (4 years) 3,147 25.72

Post Graduate Degree 1,703 13.92

When respondents were asked to denote their primary source of news, over 55% of the respondents

listed television, 24% listed the internet, and 14% listed newspapers as their primary source of

news.

Household

Under $30,000

3,218

26.3

Income

$30,000 - $49,999

2,492

20.37

$50,000 - $74,999

2,468

20.17

$75,000 and Over

4,058

33.16

Primary source

Television

6,782

55.43

of news

Newspapers

1,748

14.29

Magazines

Radio

Internet

Local Church

Other (Specify)

600

2,951

0.37

4.9

24.12

0.27

0.63

4.2 Factor Analysis Results

This section presents the results from the factor analysis conducted by using the consumer

concern questions and attitude towards preparedness of food system questions. The factor

analysis was performed with the aim of identifying common dimension among several concern

and attitudinal questions. Further factor rotation was performed which yielded two very clear

factors with no variable with ambiguous loadings or cross loadings. To identify factors easily for

each variable in the factor analysis, only factor loadings higher than 0.6 are displayed in the

Table 4.2. The last column in the table shows the unique variance for each specific variable.

Unique variance is associated uniquely with each variable; it is independent from any correlation

with the other variables in the analysis.

Table 4.2 Rotated factor loadings (pattern matrix)

and unique variances

Variable

Factor1

Factor2

Uniqueness

q2_4

0.8048

0.3497

q5_4

0.7306

0.4662

0.8935

0.1887

0.7745

0.3723

q10

0.8933

0.1963

q11

0.8625

0.2366

(blanks represent abs(loading)<.6)

The first factor contains four attitudinal questions measuring each respondent’s current

level of concern about food safety. Therefore, the first factor can be characterized as measuring

consumers’ current confidence in the safety of the U.S. food system. The second factor set

contains two attitudinal questions measuring each respondent’s attitudes on how prepared the

U.S. is for food safety/defense events compared to one year ago.

4.3 Segmentation

This section reports results obtained from the segmentation procedure. The segmentation

procedure involved factor analysis for attitude statements followed by hierarchical and partition

clustering.

4.3.1 Factor Analysis for Characteristic Statements

Factor analysis was performed with aim of identifying a common dimension among

several attitudinal questions (Table 4.3). The factor analysis was performed using twenty-nine

attitudinal variables as input, and it yielded seven factors, with some variables having cross

loadings over two or more factors. To obtain factors that were cleaner, VARIMAX orthogonal

factor rotation was performed. Using factor rotation, the cross loadings, which had been present

for certain variables, were eliminated step-wise from the factor analysis. In the end, twenty

attitudinal variables were used to obtain a seven factor solution, with factor loadings of 0.5 or

higher. To ease interpretation, statement 16 among characteristic statements was reversed on its

scoring scale. According to the description of the attitudinal questions, each factor was given a

name.

Table 4.3 Characteristic Statements Factor Analysis

Charecteristic

Factor1

Factor2

Factor3

Factor4

Factor5

Factor6

Factor7

Uniqueness

Most people are inherently good

It is important to question

authority

0.7335

0.7580

0.3894

0.4359

I must admit that I like to show off I

follow the latest trends and fashions

0.7310

0.6569

0.4397

0.4267

Just as the Bible says, the world was

literally created in six days

0.8048

0.3364

There is far too much sex on television

today

0.6399

0.4835

I strive to win the admiration of

others

0.5958

0.4731

Freedom of action and thought is very

important to me

0.7307

0.4106

Table 4.3 Contd.

(blanks represent abs (loading)<.5)

I am frightened by diseases I have

recently heard about

I believe most of the health threats

0.7192

0.3594

in the news are overblown

(Reversed)

0.6453

0.4951

I maintain a healthy and balanced

diet

0.7041

0.3878

The danger of catching a serious

illness is increasing

I like to learn about things even if

0.6359

0.3635

they may never be of any use to

0.6255

0.5261

I am optimistic about the future I

believe that future events are

predestined

I contribute regularly to a

0.7248

0.6526

0.4418

0.4219

retirement plan e.g. IRA, 401-K, etc.

0.7504

0.3603

I have one or more life insurance

policies

0.8198

0.2979

I have set a weekly/monthly budget,

and stick to it

0.8021

0.3263

With respect to danger, I like to live

a bit on the edge

0.7362

0.3482

I tend to seek adventure in my life

0.6641

0.4117

4.3.2 Cluster Analysis

This study uses a two-step hierarchical clustering analysis, followed by partition

clustering. The factor scores obtained for the seven factors in the factor analysis previously

performed were used as input scores.

In the next step, K-means partition clustering method was performed using the

FASTCLUS procedure in SAS. The seeds from the hierarchical cluster analysis were used as the

input for the SAS FASTCLUS procedure. To perform the K-means clustering, the observations

were clustered based on Euclidean distances. The observations were assigned to their nearest

seed to create temporary clusters, and these seeds were then replaced by the temporary clusters.

Various cluster solutions ranging from two to eight segments were analyzed and tested for

usefulness and interpretation. In the end, it was determined that the cluster solution with eight

segments yielded the most useful and interpretable segments.

4.3.3 Comparison of Segments

The average factor scores were used to label the segments obtained from the cluster

analysis. Segment 1 was labeled as Non Differentiators because it had a positive factor score on

all the factors. Segment 2 was labeled as Predestinarians/Disciplined because it had a high

average factor score for factor three Predestination and factor seven Disciplined Life. Segment 3

was labeled as Afraid because it had a relatively high average factor score on factor four fearful.

Segment 4 was labeled Adventurists because it only had a positive average factor score for factor

one Adventurous. Segment 5 was labeled as Freedom Seeker because of its relatively higher

average factor score for factor two personal freedom. Segment 6 was labeled as Life

Planner/Freedom Seeker because of its higher average factor scores for factor two personal

freedom and factor six planned life. Segment 7 was labeled as Life Planner because it had a

relatively high positive factor score on factor six planned life. Segment 8 was labeled as

Predestinarians/Optimists since it had a relatively higher average factor score for factor three

predestination and factor five optimistic view.

Table 4.4 Cluster Analysis Summary: Average Factor Score for Consumer Attitude

Dimension

Factor1

Factor2

Factor3

Factor4

Factor5

Factor6

Factor7

Segment

Show off,

Adventurous

Personal

Freedom

Predestinati

Fearful

Optimistic

View

Planned

life

Disciplined

Life

Non-Differentiators

1.506

0.173

0.540

0.335

0.378

0.250

0.425

Prdestinarians/disci

plined

-0.635

-0.650

0.706

-0.618

0.293

0.414

0.662

Afraid

-0.561

0.579

0.246

0.982

-0.858

-0.478

0.664

Adventurists

0.383

-1.260

-0.379

-0.004

-0.628

-0.346

-0.276

Freedom Seeker

0.025

0.591

-0.914

-0.770

0.300

-0.809

0.407

Life

Planner/freedom

seeker

-0.022

0.668

0.278

-0.770

-0.800

0.792

-0.614

Life Planner

-0.270

0.106

-0.780

0.526

0.494

0.939

-0.157

Predestinarians/opti

mists

-0.378

0.178

0.669

0.249

0.625

-0.721

-0.970

Segment1-Non Differentiators

Characteristics for this segment included a younger household, a majority of the people

being either 30 to 39 years old or 30 years or less, and a majority of the people having a

household income of $75,000 or higher. The number of people who never married, i.e. single,

was relatively higher than the general population, and people from this segment used television

as their primary media source.

Segment2-Predestinarians/Disciplined

People in this segment believed in predestination of future events. Additionally, they

followed a disciplined life style, maintained a healthy diet, and set up a weekly budget.

Demographic characteristics for people in this group were a higher household age, a majority

being over 60 years old, also a higher representation in lower middle income category of $30,000

to $50,000 compared to the general population, a majority of the population being currently

married, a relatively higher percentage of females.

Segment3-Afraid

People in this segment could be defined as afraid of food scare events and of catching

illnesses. The major demographic characteristics for this segment were a household age of 60

years or higher, an income level of less than $30,000 per year, a high school education or some

college, a relatively higher number of divorcees, and a higher number of females compared to

the general population. The primary media source for this segment was television.

Segment4- Trendy and Adventurists

The primary characteristics of this segment were that they followed the latest fashion

trends and that they were adventurous. The major demographic characteristics for this segment

were that the majority of the population was younger, with a household age of 30 to 39 years or

30 years or less; that the number of people with a low household income level, i.e. $30,000 or

less, was relatively higher; and that the majority of the people used television as their primary

media source.

Segment5-Freedom Seekers

The primary tendency of this segment was to pursue the freedom of expression in that

people in this segment liked to question authority and liked to learn new things. Other

characteristics were that the population tended to be older; that most people had only one

member in the household, compared to the general population; that over half of the population

had graduated from college or had a post graduate degree; that they were never married or were

divorced/widowed; that they had a relatively higher percentage of males; and that they were

using internet and newspaper as primary sources of information.

Segment6-Life Planners/Freedom Seekers

People in this segment exhibited characteristics of both Life Planner and Freedom Seeker.

People in this segment considered freedom of expression very important, liked to learn new

things, contributed regularly to retirement plans, and bought more than one insurance policy.

This segment could be characterized as people who were mostly middle aged, i.e. in categories

30 to 39 years and 40 to 49 years, and who had a household income of $75,000 or more. This

segment had a higher percentage of males relative to general population and a higher percentage

using the internet as their primary media source.

Segment7-Life Planners

This segment was characterized by a desire to secure their future in life; people invest in

insurance policies to prepare for hard times in the future. The prominent characteristics of this

segment were that they were late middle aged, i.e. household age was 40 to 49 years or 50 to 59

years; that they had a high household income of $75,000 or higher; that they had higher

education bachelor’s or post graduate degree; that they were married; and that they had a

relatively higher percentage individuals using newspapers as their primary media source.

Segment8-Predestinarians/Optimists

This segment tended to believe in predestination of future events, to believe that most people

were inherently good by nature, and to have an optimistic view about life. The demographic

characteristics of this segment were that they were generally older, over 60 years; that they had a

low income level, less than $30,000 a year; and that they had high school or some college

education. This segment had a higher percentage of females compared to the general population

and used television as their primary media source.

4.4 Ordered Probit Results

An ordered probit model was used for this research because it allowed for the calculation

of predicted probabilities for each category of ordered dependent variable and the marginal

effects. Since the ordered probit model is a non-linear model, the estimated coefficients were not

the marginal effects. Thus, the estimated coefficients and the marginal effects had to be

calculated and discussed separately. The results for the general ordered probit model and each of

the ordered probit models by segment are presented in this section. Chi square tests were used to

test the significance of each model, and z-tests were used to test the significance of each

coefficient associated with each of the models.

4.4.1 Consumer Confidence in Food System

As discussed previously, the four concern questions were grouped into one factor. These

four questions, which had responses that individually ranged from 1-6, were aggregated together

creating a response range of 4-24. To expedite the analysis, the response range was divided into

groups of three (4-6, 7-9, etc), resulting in a new aggregated variable that ranged from 1 to 7,

with one being not at all concerned and seven being extremely concerned. We reversed the rating

for this new aggregated variable so that we would obtain a confidence scale.

Several alternative specifications of the model were estimated, relating CFSTC to

different combinations of explanatory variables. The final model used to estimate CFSTC is

specified as:

CFSTCi = β1MTI + β2Age + β3Region2 + β4Region3 + β5Region4 + β6Region5 + β7Region6 +

β8Region7 + β9Region8 + β10Region9 + β11Mediasource2 + β12Mediasource3 +

β13Mediasource4 + β14Mediasource5 + β15Mediasource6 + β16Mediasource7 + β17Hage2 +

β18Hage3 + β19Hage4 + β20Hage5 + β21Hincome2 + β22Hincome3 + β23Hincome4 + β24Hsize2

+ β25Hsize3 + β26Hsize4 + β27Hsize5 + β28Race2 + β29Race3 + β30Race4 + β31Race5 +

β32Education1 + β33Education2 + β34Education4 + β35Education5 + β36Education6 +

β37Education7 + β38Maritalstatus2 + β39Maritalstatus3 + β40Gender2

where CFSTC is the aggregated variable measuring consumer confidence in food safety, i

represents each segment for which the model was run, and β1 through β40 are the estimated

coefficients representing the change in CFSTC given a unit change in the associated explanatory

variables, holding all other variables constant.

Tables 4.5A and 4.5B show the results from the ordered probit model, which used the

aggregated variable measuring respondent’s confidence in the safety of the food system as the

dependent variable. The media tracking index (MTI) and the age variable were the only

continuous variables in the model. All other independent variables were categorical; therefore, a

dummy variable was created for each category. To create the dummy variables, one category

from each of the variables was used as the reference category and was left out of the model.

Tables 4.5A and 4.5B compare the coefficients for the general ordered probit model and each of

the ordered probit models by segment. The results are discussed according to each segment.

Table 4.5A Coefficients for Ordered Probit model for Consumer Confidence for each

Segmenta

Variables

General

NonDifferentia

tors

Predestinarian s/

Disciplined

Afraid

Media Tracking Index

-0.171***

0.069

-0.043

-0.076

Age

-0.002

-0.003

-0.010*

-0.004

Regionc

Middle Atlantic

-0.021

0.032

-0.008

0.094

East North Central

0.088*

0.009

0.141

0.252

West North Central

0.159***

0.271

0.294*

0.056

South Atlantic

-0.042

-0.095

0.019

0.213

East South Central

-0.151***

-0.040

-0.119

0.055

West South Central

0.089*

-0.065

0.120

0.118

Mountain

0.124**

0.204

0.149

0.186

Pacific

0.176***

0.011

0.213

0.231

Primary Media Sourcec Newspaper

0.221***

0.095

0.180**

0.188*

Magazines

0.374**

0.385

-0.159

0.929

Radio

0.237***

0.119

0.280**

0.110

Internet

0.091***

0.098

0.074

0.165**

Local Church

-0.115

0.986*

-0.128

0.370

Other (Specify)

0.405***

-1.187*

0.611**

-0.799*

Household Agec

30 through 39 Years

-0.183***

-0.359***

-0.075

-0.124

40 through 49 Years

-0.292***

-0.416***

-0.041

-0.377*

50 through 59 Years

-0.348***

-0.430**

0.015

-0.339

60 Years and Over

-0.357***

-0.432

0.092

-0.254

Household Incomec $30,000 -

$49,999

0.077***

0.184*

-0.012

0.016

$50,000 - $74,999

0.084***

0.143

0.181**

0.157*

$75,000 and Over

0.143***

0.222**

0.094

0.269***

Household Sizec 2

Members

-0.089***

-0.162*

-0.080

-0.119

3 Members

-0.136***

-0.215**

0.045

-0.224*

4 Members

-0.147***

-0.321***

-0.026

-0.298**

5 or More Members

-0.145***

-0.113

-0.152

-0.238

Racec

Black/African-American

-0.194***

-0.175

-0.094

0.046

Asian or Pacific Islander

-0.295***

-0.222*

-0.538**

0.051

American Indian, Aleut Eskimo

0.069

0.345

-0.031

0.149

Other

-0.090

-0.475**

0.075*

0.174

Educationc Grade

School

0.699***

0.043

2.181***

-3.979

Some High School

-0.029

-0.223

0.032

0.364*

Some College-no degree

0.066**

0.076

0.111

0.067

Graduated College –Associate’s Degree (2

years)

0.133***

0.034

0.179*

0.082

Graduated College- Bachelor’s Degree (4

years)

0.283***

0.192**

0.284***

0.159

Post Graduate Degree

0.446***

0.414***

0.374***

0.059

Marital Statusc Never

Married

0.106***

-0.024

0.089

0.165

Divorced, Widowed, Separated

-0.002

-0.070

0.003

-0.144

Table 4.5A Contd.

Genderc

Female

-0.272***

-0.139*

-0.109

-0.240***

Log Likelihood

-20239.0

-2057.8

-2519.6

-1636.6

Number of observation

12236

1375

1492

1335

LR chi2(40)

1125.76

127.97

110.32

85.94

Prob. > chi2

Pseudo R2

0.0271

0.0302

0.0214

0.0256

a The coding for questions measuring respondents current level of concern is reversed in order to measure respondents

confidence in the safety of our food.

*: significant at 0.10 level, **: significant at 0.05 level, ***: significant at 0.01 level

C The selected reference variables are: region1- New England; mediause1- Television; householdage1- under 30 year;

householdincome1- under $30,000; housemembers1- 1, race1 – white, education1- Graduated High school, marital status -

Currently Married.

Table 4.5B Coefficients for Ordered Probit model for Consumer Confidence for each

Segmenta

Life

Variables AdventuristsTrendy and

Fseekerreedom s PlannerFreedom

s/ PlannerLife s Predes/

Optimistsstinarian seekers

Media Tracking Index -0.355** -0.376** -0.217 -0.257* -0.049

Age

-0.007

0.000

-0.003

-0.002

0.004

Regionc

Middle Atlantic

-0.010

-0.213

-0.017

0.057

-0.110

East North Central

0.112

-0.111

-0.061

0.184*

0.044

West North Central

0.138

-0.028

0.286

0.118

0.155

South Atlantic

-0.068

-0.308**

-0.158

0.138

-0.123

East South Central

-0.247

-0.278

-0.078

-0.116

-0.213

West South Central

0.017

-0.013

0.117

0.118

0.098

Mountain

0.247*

-0.073

0.149

0.215

-0.031

Pacific

0.162

-0.003

0.106

0.213*

0.159

Primary Media Sourcec Newspaper

0.067

0.308***

0.199**

0.244***

0.158*

Magazines

0.558**

-0.011

-0.881

0.432

0.221

Radio

-0.041

0.110

0.282**

0.390***

0.268*

Internet

0.035

0.022

0.145**

0.095

0.054

Local Church

-0.469

-0.499

-0.612

No Obs.

-0.347

Other (Specify)

0.307

0.499*

0.857**

0.574*

0.248

Household Agec

30 through 39 Years

-0.165*

-0.091

0.014

-0.066

-0.197

40 through 49 Years

-0.200

-0.229

-0.031

-0.064

-0.429**

50 through 59 Years

-0.163

-0.407**

-0.104

-0.199

-0.426**

60 Years and Over

-0.121

-0.438

-0.089

-0.219

-0.431*

Household Incomec $30,000

- $49,999

0.012

0.061

-0.059

0.013

0.149**

$50,000 - $74,999

-0.047

-0.024

-0.036

0.023

0.061

$75,000 and Over

-0.042

0.137*

0.057

0.079

0.182**

Household Sizec 2

Members

-0.130*

-0.059

-0.069

-0.146

-0.038

3 Members

-0.042

-0.149

-0.162

-0.211*

-0.084

4 Members

-0.243**

-0.096

-0.177

-0.123

0.128

5 or More Members

-0.204*

-0.249*

-0.081

-0.136

0.007

Table 4.5B Contd.

Racec

Black/African-American

***

0.203

***

0.325

***

2709.1

2138.3

2877.7

-0.218**

0.031

-0.311**

-0.080

-0.239**

Asian or Pacific Islander

-0.255**

-0.156

-0.170

-0.273

-0.519**

American Indian, Aleut Eskimo

-0.312

0.186

0.174

0.136

-0.515

Other

0.062

0.009

-0.373*

0.008

-0.293

Marital Statusc Never

Married

-0.066

0.048

0.214*

0.021

0.354***

Divorced, Widowed, Separated

-0.099

0.001

0.027

0.006

0.158**

Genderc

- - -

Female -0.243*** -0.195***

0.259

Log Likelihood -2995.7 -2426.9

Number of observation

1815

1564

1274

1808

1573

LR chi2(40)

135.8

161.62

137.25

166.99

127.87

Prob. > chi2

Pseudo R2

0.0222

0.029

0.0311

0.0282

0.0257

a The coding for questions measuring respondents current level of concern is reversed in order to measure respondents

confidence in the safety of our food.

*: significant at 0.10 level, **: significant at 0.05 level, ***: significant at 0.01 level

C The selected reference variables are: region1- New England; mediause1- Television; householdage1- under 30 year;

householdincome1- under $30,000; housemembers1- 1, race1 – white, education3- Graduated High school, marital status -

Currently Married.

4.4.1.1 Overall/General Model

The log likelihood statistic indicates that the model was significant at greater than the 99

percent level of confidence. The continuous variable for MTI was significant and negative,

which means that a larger MTI value decreased consumer confidence in food safety as measured

Education Grade

School

0.522

0.489

No Obs.

-5.118

Some High School

-0.102

-0.168

0.274

0.102

-0.079

Some College-no degree

-0.077

0.101

-0.033

0.113

0.147**

Graduated College –Associate’s Degree (2 years)

-0.088

0.281**

0.055

0.151

0.231**

Graduated College- Bachelor’s Degree (4 years)

0.104

0.296***

0.226**

0.422***

0.379***

Post Graduate Degree

0.315***

0.489***

0.476***

0.502***

0.344***

by the ordered probit model’s index function. In other words, a higher MTI increased a person’s

concern level. Thus, greater media coverage about food safety events reduced consumer

confidence in food safety. These results were consistent with the hypothesis of the media agenda-

setting effect, as described earlier in the paper.

The coefficient for the East South Central region had a negative sign and was significant,

while the coefficients for the East North Central, West South Central, West North Central and

Pacific regions were significant and positive (Table 4.5A). This meant that confidence in food

safety decreased if a person lived in the East South Central region of the U.S. relative to the

omitted New England region. On the other hand, a person living in the East North Central, West

South Central, West North Central or Pacific regions of the country was indicated to have higher

confidence relative to the New England region.

The coefficients for all media sources were significant and positive, except for the

variable for local church (Table 4.5A). If a person used newspapers, radio, internet, magazines or

other sources as their primary media source, the person’s confidence in food safety increased

relative to a person who used television (the reference category) as their primary media source.

This suggests that individuals who rely on television as their primary media source had generally

less confidence in the safety of the food system over the sample period. Moreover, greater than

50 percent of the respondents in the sample indicated that television was their primary news

source, implying that television coverage of food safety events was an important driver of the

public’s opinion regarding food safety.

The coefficients for household age were negative and significant in the model. The age

variable indicated that people aged greater than or equal to 30 were generally less confident

about food safety, relative to the reference category of people aged less than 30. The results

suggest that people older than 30 were more concerned about food safety.

The dummy variables for the household income categories were significant and positive

(Table 4.5A). The positive direction of the categories indicated that a person with higher

household income, relative to the reference category of household incomes under $30,000 would

be more confident about food safety. These results suggested that households with higher

incomes were less concerned about food safety, perhaps because the higher income allowed them

a wider variety of food choices relative to lower income households.

The variables measuring the number of household members were found to be negative

and significant (Table 4.5A). These results were in accordance with the expectation that people

with larger families were more concerned about food safety. Along with the number of members

in his or her household, a person’s education had a significant negative effect on his or her

confidence in food safety (Table 4.5A). Specifically, having higher education increased a

person’s confidence in food safety, relative to a person who graduated from high school. These

results were in line with the expectation that a person with higher education would be

knowledgeable and concerned about the safety of his or her food. Finally, the variable gender

showed significance in the model and had a negative direction (Table 4.5). The results showed

that women were generally less confident about food safety, relative to men.

Figure 4.1 presents predicted probabilities for the continuous variable MTI over

consumer confidence for the overall model. As mentioned before, to ease visualization and

understanding, the seven categories of dependent variables were consolidated into three

supercategories, with the top two categories representing the highest confidence super-category,

the middle three categories representing the middle super-category, and the bottom two

categories representing the lowest confidence super-category. The graph with all seven categories

is presented in the appendix A.

Figure 4.1 shows the decrease in probability of a person being in the highest confidence

or middle confidence super-category as the MTI increased. This result was in line with our

hypothesis that as the media coverage increased; a person’s confidence in food safety went

down. Also, the probability of a person being in highest confidence category was very low.

In Figure 4.2, the predicted probabilities for the different media sources are plotted. In the

overall model, all the media sources except Local Church were significant relative to television.

One should note that the probability of being in the lowest confidence categories was lower for

these media sources.

Figure 4.1 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Overall Model

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidnce

Figure 4.2 Probability curve showing the effect of Media Source on Consumers Confidence

about Food Safety for the Overall Model

4.4.1.2 Non Differentiators

For the segment Non-Differentiators, the number of observations was 1375, and the log

likelihood estimate indicated that the model was significant at greater than 99 percent level of

confidence. The Media Tracking index was not significant in the model (Table 4.5A).

The coefficients for the media sources Local Church and Other were significant. Local

Church had a positive sign for its coefficient in model, while the coefficient for Other media

sources was negative. This indicated that, relative to a person who used television as his or her

primary media source, a person who used the local church as the primary media source had

higher confidence in food safety, and a person who specified his or her primary media source as

“other” had lower confidence.

The household age variables also showed a significant impact on consumer confidence.

The coefficients for all household age categories except the last category, 60 years and over, had

a significant and negative effect. This suggested that, relative to the base category of 30 years or

lower, if a person was in the age categories of 30 through 39 years, 40 through 49 years or 50

through 59 years, he or she would be less confident in food safety.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Newspaper

Magazine

Radio

Internet

Local Church

Other

ry Media S

ourc

Lowest Confidence

Middle

Highest Confidence

The coefficients for the household income categories $30,000-$49,000 and $75,000 and

over were positive and significant. The base category here was a household income of $30,000 or

lower. This result suggested that if a person was in a higher income category, it increased his or

her confidence about the safety of food.

If household size was two, three, or four members, the probability of a person within this

household having confidence in safety of food decreased, relative to the person with a household

size of one member. Additionally, a person’s confidence in the safety of food increased if the

person had earned a bachelor’s degree or a post graduate degree, relative to a person who had

only graduated from high school. Lastly, females were more concerned about food safety than

males were.

Figure 4.3 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Non-differentiators. There was very little variation in the

probability for consumer confidence to change due to a change in MTI, although there was a

high probability for a person to be low in confidence about food safety.

The predicted probabilities for the various media source within the segment

NonDifferentiators are shown in Figure 4.4. This segment had a relatively higher percentage of

the population using the television as their primary media source than the general population did.

One should note that the people who specified a source other than the listed sources had a higher

probability of being low in their confidence than those people who specified one of the listed

sources.

Figure 4.3 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Non-Differentiators

Figure 4.4 Probability curve showing the effect of Media Source on Consumers Confidence

about Food Safety for the Segment Non-Differentiators

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Probability curve showing the effect of MTI on Consumers Confidence

about Food Safety for segment Non

Differentiators

Lowest Confidence

Middle

Highest Confidence

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

4.4.1.3 Predestinarians/Disciplined

The segment Predestinarians/Disciplined had 1492 observations, and the ordered probit

model for this segment was significant at a greater than 99 percent level of confidence. The

Media Tracking Index did not have a significant effect on consumers’ confidence in the safety of

food (Table 4.5A).

Only the East North Central region had a significant and positive effect. This implied

that, relative to the New England region, a person from the East North Central region was more

confident in the safety of food. Additionally, the media sources Newspaper, Radio and Other had

a significant and positive effect. This meant that, relative to the television users, the people who

listed newspaper, radio or other specific media as their primary media source were more

confident in the safety of food. Furthermore, a person with a household income between $50,000

and $75,000 was more confident in the safety of food, compared to the base category of people

with household incomes lower than $30,000.

The level of education also had some effect on consumers’ confidence regarding food

safety in this segment. The coefficients for Grade School, 2-year college degree, 4-year college

degree, and post graduate degree were all positive and significant. This meant that, relative to a

person who graduated from high school, people with education levels of grade school, 2-year

college degree, 4-year college degree, or post graduate degree were more confident about the

safety of food.

Figure 4.5 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the Predestinarians/Disciplined segment. For all three categories, there

was not much variation in the probability of a person being confident. However, as opposed to

the first segment, Non-Differentiators, the probability of a person being in the highest confidence

category or the middle confidence category was much higher.

Figure 4.5 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Predestinarians/Disciplined

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

Figure 4.6 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment Predestinarians/Disciplined

The population sample in the segment Predestinarians/Disciplined indicated relatively

higher use of newspapers as their primary media source, compared to the general population.

Figure 4.6 shows that the probability of people with newspaper, radio and other specific sources

of being in middle and highest confidence categories were higher.

4.4.1.4 Afraid

The ordered probit model for the third segment, Afraid, was significant at a greater than 99

percent level of confidence as indicated by the log likelihood statistic. Like the previous

segment, the coefficient for the Media Tracking Index was not significant in this segment.

The coefficients of the dummy variables Newspaper, Internet, and Other specified sources

were significant. Newspaper and Internet had a positive sign, while Other specified source had a

negative sign. Therefore, relative to people who used the television as their primary media

source, people who used the newspaper and the internet as their primary media source were more

confident about the safety of food, and people who used other specified sources as their primary

media source were less confident.

The dummy variable coefficient for the household age category 40 through 49 years had a

significant and negative effect. This can be explained as, relative to the household age category

30 or lower, a person in the category 40 through 49 was less confident about the safety of food.

Additionally, people in the household income categories of $50,000 - $75,000 and $75,000 and

over were more confident in the safety of food, relative to people in the household income

category of $30,000 and lower.

Household size also had a significant and negative effect on consumers’ confidence in the

safety of food. People with household sizes of three members and four members were less

confident about the safety of food than people with a household size of one member. Relative to

people who graduated from high school, people with some high school education were more

confident about the safety of food. Females were less confident about the safety of food than

males.

Figure 4.7 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Afraid. As expected from the segment’s name, the

probability of a person being in the lowest confidence category was very high. Although the

probability of person being in the lowest confidence category increased slightly with an increase

in MTI, there was not much variation. The probability curve shows that the probability of people

in this segment being low in confidence was very high and was not affected by changes in media

coverage.

Figure 4.7 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Afraid

Figure 4.8 presents the probability curves for media sources within the segment Afraid. A

majority of the people within this segment indicated the television as their primary media source.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

Additionally, most of the people who indicated another specified media source as their primary

source had a very high probability of being in the lowest confidence category.

Figure 4.8 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment AfraidAfraid

4.4.1.5 Trendy and Adventurists

The log likelihood statistic for the Trendy and Adventurists segment indicated that the model

was significant at a greater than the 99 percent level of confidence. The continuous variable for

Media Tracking Index (MTI) was significant and negative, which meant that a larger MTI value

decreased consumer confidence in food safety, as measured by the ordered probit model’s index

function.

Amongst all regions, only the Mountain region had a coefficient that was significant and

positive, implying that a person residing in the Mountain region tended to be more confident

compared to a person residing in the New England region. Amongst primary media sources,

Magazines had a significant and positive effect. Therefore, if a person indicated that his or her

primary media source was magazines, he or she tended to be more confident in food safety than a

person who used the television primarily. Additionally, a person within the household age

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

category of 30 through 39 years was less confident about the safety of food, relative to a person

within the household age category of 30 years or less.

Household size had a negative and significant effect in the model. A person who had two,

four, or five members within his or her household was less confident in the safety of food,

relative to a person who had only a single member in his or her household. A person who had a

post graduate degree was more confident in the safety of food, relative to a person who

graduated from college, and females were less confident in food safety than males.

Figure 4.9 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the Trendy and Adventurist segment. MTI had a negative and

significant effect on consumer confidence regarding food safety in this segment. One should note

that the probability curve for being in the highest confidence and middle confidence categories

was going down with an increase in MTI, while the curve for lowest confidence was going up.

This indicated that people who were in the segment Trendy and Adventurists became less

confident about food safety as the media coverage increased.

The population of consumers with television as their primary media source was higher in the

segment Trendy and Adventurist, relative to the general population. For all the media sources,

the probability of a person being in the middle or highest confidence category was higher (Figure

4.10).

Figure 4.9 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Trendy and Adventurists

Figure 4.10 Probability Curve Showing the effect of Media Source on Consumers

Confidence about Food Safety for the Segment Trendy and Adventurists

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

4.4.1.6 Freedom Seekers

The model for the segment Freedom Seeker was significant at a greater than 99 percent level

of confidence. The continuous variable for Media Tracking Index (MTI) was significant and

negative, meaning that an increase in the Media Tracking Index decreased consumer confidence

in food safety (Table 4.6).

Only the South Atlantic region had a negative and significant effect. This meant that if a

person resided in the South Atlantic region, he would be less confident in food safety, compared

to a person residing in the New England region. If a person used the newspaper and other

specified media sources as his or her primary media source for information, he or she would be

more confident about food safety, relative to a person using the television as his or her primary

media source.

A person within the household age category 50 through 59 years was less confident in food

safety, compared to a person within the younger household age category of 30 years or less. If a

person’s household income was over $75,000 dollars, he or she was more confident about the

safety of food, relative to a person with a household income of lower than $30,000. A household

size of five or more members had a negative and significant impact on consumer confidence

regarding food safety. Therefore, a person with a household size of five or more members would

be less confident in food safety, as compared to a person who had only a single person in his or

her household.

For the segment Freedom Seeker, education had a positive and significant effect. If a person

had earned a 2-year college degree, 4-year college degree, or post graduate degree, he or she was

more confident about food safety, relative to a person who graduated from high school. Females

in this segment were less confident about food safety than males were.

Figure 4.11 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Freedom Seeker. MTI had a significant and negative effect

on consumer confidence regarding food safety in this segment. As displayed by the figure, the

probability of a person being less confident about food safety was low when MTI was 0, but the

probability of being less confident increased with the increase in MTI. On the other hand, the

probability of being highly confident or medium confident decreased with increases in media

coverage.

A higher percentage of the population within the segment Freedom Seeker used the

internet or newspapers as their primary media source, compared to general population. The

probability of a person within the segment Freedom Seeker being in the lowest confidence

category was lower for all media sources, except for the local church (Figure 4.12).

Figure 4.11 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Freedom Seeker

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Medi

a Tracking Index

Lowest Confidence

Middle

Highest Confidence

Figure 4.12 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment Freedom Seeker

4.4.1.7 Life Planners/Freedom Seekers

The log likelihood statistic indicated that the model for the segment Life Planner/Freedom

Seeker was significant at a greater than the 99 percent level of confidence. The explanatory

variable Media Tracking Index was not significant for this segment (Table 4.6).

In this segment, the media sources Newspaper, Radio, Internet and Other specified sources

had a positive and significant effect in the model. Consequently, people who indicated

Newspaper, Radio, Internet and Other specified sources as their primary media source for

information were more confident about food safety, compared to people who indicated television

as their primary media source.

The higher education categories of a 4-year college degree and a post graduate degree had a

significant and positive effect. If a person who had earned a 4-year college degree or post

graduate degree, he or she was more confident about food safety, as compared to a person who

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

had graduated high school. Additionally, in this segment, people who said that they were never

married were more confident about food safety, relative to people who indicated that they were

currently married. Females in this segment were also less confident about food safety than males

were.

Figure 4.13 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Life Planner/Freedom Seeker. Although MTI did not have

a significant effect, the change in the probability curve was evident with increases in media

coverage for each category.

Figure 4.13 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Life Planner/Freedom Seeker

There were a higher percentage of internet users in the segment Life Planner/Freedom

Seeker, as compared to the general population. As can be seen in Figure 4.14, people who relied

on newspapers, the radio, the internet and other specific sources had a higher probability of

being in the middle or highest confidence categories..

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

Figure 4.14 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment Life Planner/Freedom Seeker

4.4.1.8 Life Planners

The model for the Life Planner segment was significant at a greater than the 99 percent level

of confidence, based on the log likelihood statistic. The continuous variable for MTI was

significant and negative, meaning that as the Media Tracking Index increased, consumer

confidence in food safety decreased (Table 4.6).

For the segment Life Planner, the East North Central and Pacific regions had a positive and

significant effect. A person residing in the East North Central or Pacific regions was more

confident about food safety, relative to a person living in the New England region. Additionally,

the media sources Newspaper, Radio and Other specified sources had a positive and significant

effect. People who used these sources as their primary media source for information were more

confident about food safety, compared to people who indicated the television as their primary

media source.

The higher education categories 4-year college degree and post graduate degree had a

significant and positive effect in the model. A person who had earned either of these two degrees

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Newspaper

Magazine

Radio

Internet

Local Church

Other

Prim

ary Media S

ourc

Lowest Confidence

Middle

Highest Confidence

was more confident about food safety, as compared to a person who had graduated from high

school. Females in this segment were less confident about food safety as compared to males.

Figure 4.15 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Life Planner. MTI had a significant and negative effect for

this segment. When the media coverage was zero, there was a higher probability of a person

being confident about food safety, but as the media coverage increased, people became less

confident.

Figure 4.15 Probability Curve Showing the Effect of MTI on Consumers Confidence about

Food Safety for the Segment Life Planner

In Figure 4.16, the predicted probabilities for the different media sources for the segment

Life Planner were plotted. As can be observed, there was a higher probability of a person being

in the middle confidence than in lowest confidence category, for all media sources. For this

segment, there were a higher percentage of consumers whose primary media source was

newspapers, compared to the general population.

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

Figure 4.16 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment Life Planner

4.4.1.9 Predestinarians/Optimists

The log likelihood statistic indicated that the model for the segment

Predestinarians/Optimists was significant at a greater than the 99 percent level of confidence.

The continuous variable for MTI was not significant.

For this segment, Newspaper and Radio had a significant and positive effect. A person who

used these sources as his or her primary media source was more confident about food safety,

relative to a person who relied on the television.

Household age showed a significant impact on consumer confidence. The coefficients for the

household age categories 40 through 49, 50 through 59 years, and 60 or over had were negative.

A person in one of these household age categories was less confident about food safety than a

person with a household age of 30 or less.

For the segment Predestinarians/Optimists, higher education had a positive effect. A person

with some college education, an Associate’s degree, a Bachelor’s degree or a post graduate

degree was more confident in food safety, compared to a person who graduated from high

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Newspaper

Magazine

Radio

Internet

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

school. Additionally, people who indicated that they were never married or that they were

divorced or widowed were more confident about food safety than people who were currently

married. As was generally the case, females in this segment were less confident about food safety

than males were.

Figure 4.17 presents the predicted probabilities for the continuous variable MTI over

consumer confidence for the segment Predestinarians/Optimists. MTI did not have a significant

effect on this segment. However, there was not much change in the probability of a person’s

confidence level with a change in media coverage the probability of being in the lowest

confidence category was always higher than the probability of being in the other two.

Figure 4.17 Probability Curve Showing the effect of MTI on Consumers Confidence about

Food Safety for the Segment Predestinarians/Optimists

Figure 4.18 shows the predicted probabilities for each media source on consumer

confidence for the segment Predestinarians/Optimists. No trends in the data were evident. Most

of the population in this segment recorded television as their primary media source.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Lowest Confidence

Middle

Highest Confidence

Figure 4.18 Probability Curve Showing the Effect of Media Source on Consumers

Confidence about Food Safety for the Segment Predestinarians/Optimists

4.4.1.10 Marginal Effects

By definition, marginal effects across all categories of dependent variable must sum to

zero - since the probabilities must sum to one (Cranfield and Magnusson, 2003). When

interpreting marginal effects for a continuous variable, all other things equal, a unit change in the

explanatory variable will result in an increase or decrease in the predicted probability equal to

the size of the marginal effect. In the case of a categorical dummy variable, the marginal effect is

the change in the predicted probability based on whether the observation falls in that category or

not. Since while calculating marginal effects all the remaining variables assume their average

values, the marginal effect shows the change in the predicted probability for each category for an

average respondent (Long, 1997). Due to the large number of explanatory variables for each

segment and ease of understanding, the marginal effects for each segment are explained together

by comparing among segments. Marginal effects for the variables which are significant in the

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Lowest Confidence

Middle

Highest Confidence

model are explained below, and the tables showing marginal effects for each segment are

included in the appendix B.

The explanatory continuous variable MTI was significant in the overall model and in the

segments Trendy and Adventurists, Freedom Seekers, and Life Planners. The lower categories of

the dependent variable have a positive sign for the marginal effects, while the higher categories

have negative signs across all the segments. The marginal effects for MTI imply that increase in

media coverage of the food safety events decreases the probability that a subject’s response will

fall in the higher categories for the dependent variable. This is consistent with the finding that

MTI negatively impacts consumer confidence in food safety. For example, if a person is in the

segment Trendy and Adventurists lists Category 5 of the dependent variable, an additional unit

increase in the media coverage will result in decreasing that person’s probability of being

confident and of being in Category 5 by 0.0587, holding all other variables constant at their

mean.

The continuous variable Age was only significant in the segment

Predestinarians/Disciplined. The marginal effects indicate that the probability of a person to be

confident decreases with the increase in age, while the probability of being low in confidence

about food safety increases.

The region East North Central had a significant effect in the Overall model and in the

segment Life Planner. For both the models the marginal effects indicate that when a person is

living in this region, the probability of him/her being low in confidence decreases while the

probability of being high in confidence about safety of the food increases relative to a person

living in New England region of the country. The region West North Central had a significant

effect in the Overall model and in the segment Predestinarians/Disciplined. The marginal effects

could be interpreted as when a person resides in these regions the probability of him/her being

low in confidence decreases while the probability of being high in confidence increases relative

to a person living in New England region of the country. For example the probability of being

Not Confident at all declines by 0.0537 for person living in West North Central region relative to

a person living in the New England region within the segment Predestinarians/Disciplined,

holding all other variables at their means.

The region South Atlantic had a significant effect in the segment Freedom Seeker. The

marginal effects for this region suggest that if a person resides in this region the probability of

being him/her being completely confident decreases and being low in confidence increases

relative to a person residing in the New England region. The regions East South Central and

West South Central are only significant for the Overall model. The marginal effects for these two

regions were interpreted as, if a person resides in the East South Central region the probability of

being him/her being high in confidence decreases and being low in confidence increases relative

to a person residing in the New England region. While if the person lives in the West South Central

region the probability of being him/her being high in confidence increases and being low in

confidence decreases relative to a person residing in the New England region.

The Mountain region had a significant effect in the Overall model and in the segment

Trendy and Adventurist. For both the models the marginal effects indicate that when a person is

living in this region, the probability of him/her being low in confidence decreases while the

probability of being high in confidence increases relative to a person living in the New England

region of the country. The Pacific region had a significant effect in the Overall model and in the

segment Life Planner, and had a similar effect as of Mountain region. For both the models the

marginal effects indicate that when a person is living in this region, the probability of him/her

being low in confidence decreases while the probability of being high in confidence increases

relative to a person living in New England region of the country.

Newspaper had significant effect for overall model and segments

Predestinarians/Disciplined, Afraid, Freedom Seeker, Life Planner/Freedom Seeker, Life Planner,

and Predestinarians/Optimists. Marginal effects for these segments can be interpreted as if a

person used Newspaper as his/her primary media source the probability of him/her being low in

confidence decreases while the probability of being high in confidence increases relative to a

person who uses Television as primary media source. Magazine had a significant effect for

overall model and segment Trendy and Adventurists, similar to a person with newspaper as

primary media source, if a person uses magazine as his/her primary media source the probability

of him/her being low in confidence decreases while the probability of being high in confidence

increases relative to a person who uses Television as primary media source. For instance, the of

probability of being completely confident increases by 0.0194 for a person who uses Newspaper

as his/her primary media source compared to a person who uses television as his/her primary

media source within the segment Freedom Seeker, holding all other variables at their means.

Radio also had a similar effect for overall model and segments

Predestinarians/Disciplined, Life Planner/Freedom Seeker, Life Planner, and

Predestinarians/Optimists. If a person is using Radio as his/her primary media source the

probability of him/her being low in confidence decreases while the probability of being high in

confidence increases relative to a person who uses Television as primary media source. Same

with the Internet for the overall model and in the segments Afraid and Life Planner/Freedom

Seeker, if a person is using Internet as his/her primary media source the probability of him/her

being low in confidence decreases while the probability of being high in confidence increases

relative to a person who uses Television as primary media source.

Local church was only significant for the segment Non-Differentiators, and according to

the marginal effects if a person is using Radio as his/her primary media source the probability of

him/her being low in confidence decreases while the probability of being high in confidence

increases relative to a person who uses Television as primary media source. The persons who

used Other sources than listed in questionnaire had a significant effect in for overall model and

all segments except Trendy and Adventurists, and Predestinarians/Optimists. For these segments

except for the Afraid segment the marginal effects can be interpreted as, if a person is using a

source other than listed in questionnaire as his/her primary media source the probability of

him/her being low in confidence decreases while the probability of being high in confidence

increases relative to a person who uses Television as primary media source. While for the Afraid

segment if a person is using a source other than listed in questionnaire as his/her primary media

source the probability of him/her being low in confidence increases while the probability of

being high in confidence decreases relative to a person who uses Television as primary media

source.

The household age category 30 through 39 years had a significant effect in overall model

and in the segments Non-Differentiators, and Trendy and Adventurists. Marginal effects for these

segments can be interpreted as if a person is within household age category 30 through 39 years

the probability of him/her being low in confidence increases while the probability of being high

in confidence decreases relative to a person who is within household age category under 30

years. The household age category 40 through 49 years had a significant effect in overall model

and in the segments Non-Differentiators, Afraid, and Predestinarians/Optimists. So if a is within

household age category 40 through 49 years the probability of him/her being low in confidence

increases while the probability of being high in confidence decreases relative to a person who is

within household age category under 30 years . Similar are the marginal effects interpretation for

the household age categories 50 through 59 years and over 60 years. Category 50 through 59

years was significant for overall model and the segments Non-Differentiators, Freedom Seeker,

and Predestinarians/Optimists. Category over 60 years had significant effect for overall model

and segment Predestinarians/Optimists. For example in the segment Trendy and Adventurists, the

probability of being in category 5 of confidence scale declines by 0.0266 for household age

category 30 through 39 years than household age category under 30 years, holding all other

variables at their means.

The household income category $30,000 to $49,999 had a significant effect in the overall

model and for the segments Non-Differentiators and Predestinarians/Optimists. According to the

marginal effects, if a person was within the household income category $30,000 to $49,999, the

probability of him or her being low in confidence decreased, while the probability of that person

being high in confidence increased relative to a person who was within the household income

category under $30,000. For the category $50,000 to $74,999, the overall model and the segments

Predestinarians/Disciplined and Afraid had a significant effect. For the category

$75,000 and over the overall model and the models for the segments Non-Differentiators, Afraid,

Freedom Seeker, and Predestinarians/Optimists had a significant effect. The marginal effects for

these two categories can be interpreted the same way as the category $30,000 to $49,999. For

example in the segment Predestinarians/Optimists, the probability of being in category 4 of

consumer confidence scale increases by 0.0217 for household income $30,000 to $49,999 than

household income under $30,000, holding all other variables at their means.

The household size of two members had a significant effect in the overall model and the

segments Non-Differentiators, and Trendy and Adventurists. Marginal effects for these segments

could be interpreted as if a person has household size of two members the probability of him/her

being low in confidence increases, while the probability of being high in confidence decreases

relative to a person with a single person in household. For household size of three members, the

overall model and the segments Non-Differentiators, Afraid, and Life planner have the

significant effect. For household size of four members overall model, segments

NonDifferentiators, Afraid, and Trendy and Adventurists had a significant effect, and for the

household size of five or more members the overall model and segments Trendy and

Adventurists and Freedom Seeker have a significant effect. The marginal effects for household

size with three, four, and five or more members could be explained in a same way as marginal

effects for household size with two members. For instance in the segment Life Planner, the

probability of being Not confident is 0.0614 higher for household size of three members than

household size of single member, holding all other variables at their means.

The education grade school had a significant effect for overall model and the segment

Predestinarians/Disciplined. From the marginal effects it could be interpreted, a person with the

grade school level education increases his/her probability of being highly confident, and

decreases the probability of being low in confidence relative to a person who graduated from the

high school. For education level some high school, only segment Afraid had significant effect.

For the education level some high school-no college the overall model and the segment

Predestinarians/Optimists had a significant effect. For education level Associate’s degree the

overall model, segments Predestinarians/Disciplined, Freedom Seeker, and

Predestinarians/Optimists had a significant effect. For Bachelor’s degree the overall model,

NonDifferentiators, Predestinarians/Disciplined, Freedom Seeker, Life Planner/Freedom Seeker,

Life Planner, and Predestinarians/Optimist had a significant effect. For the Post graduate degree

except the segment Afraid every segment had a significant effect. The marginal effect for all

these education levels could be explained in a similar way as of grade school education. For

example in the segment Freedom Seeker, the probability of being in category 5 of confidence

scale is 0.045 higher for Bachelor’s degree than high school graduation, holding all other

variables at their means.

A person who is never married and is in the segment Life planner/Freedom Seeker and

Predestinarians/Optimists increases his/her probability of being highly confident and decreases

the probability of being low in confidence relative to a person who is currently married.

Also for all the segments being females increases probability of being less confident and

increases the probability of being high in confidence relative to males in the respective segments.

For instance in the segment Afraid, the probability of being in lowest confidence category

increases by 0.0953 for females than males, holding all other variables at their means.

4.4.2 Consumers Attitude towards Preparedness of Food System

This section discusses the ordered probit results for consumers’ attitude toward

preparedness of food system. As described earlier CSFTP is a aggregated variable of two

preparedness questions and is scaled from 2 to 12. Several alternative specifications of the model

was estimated, relating CFSTP to different combination of explanatory variables. The final

model used to estimate CFSTP is specified as:

CFSTPi = β1MTI + β2Age + β3Region2 + β4Region3 + β5Region4 + β6Region5 + β7Region6

+ β8Region7 + β9Region8 + β10Region9 + β11Mediasource2 + β12Mediasource3 +

β13Mediasource4 + β14Mediasource5 + β15Mediasource6 + β16Mediasource7 + β17Hage2 +

β18Hage3 + β19Hage4 + β20Hage5 + β21Hincome2 + β22Hincome3 + β23Hincome4 +

β24Hsize2 + β25Hsize3 + β26Hsize4 + β27Hsize5 + β28Race2 + β29Race3 + β30Race4 +

β31Race5 + β32Education1 + β33Education2 + β34Education4 + β35Education5 +

β36Education6 + β37Education7 + β38Maritalstatus2 + β39Maritalstatus3 + β40Gender2

where, CFSTP is the aggregated variable measuring consumer attitudes regarding preparedness

of U.S. food system to deal with food safety/defense events, i represent the each segment for

which model has been calculated. β1 through β40 are the estimated coefficients representing the

estimated change in CFSTP given a unit change in associated explanatory variable, holding all

other variables constant.

Table 4.6A and 4.6B shows the results for the ordered probit model with the dependent

variable being the aggregated variable measuring respondent’s attitudes regarding how prepared

we are for food safety/defense events compared to one year ago. The model has used the same

independent variables as in the consumer confidence model and the log likelihood statistic

indicated that the model for each segment except for segments Life Planner/Freedom Seeker and

Predestinarians/Optimists were significant at greater than 99 percent confidence, whereas model

for segment Life Planner/Freedom Seeker was significant at 95 percent confidence level. The

model for the segment Predestinarians/Optimists was not significant.

Table 4.6A Ordered Probit Model for each Segment to Measure Consumers Attitude

toward Preparedness of the Food Supply Chain

Variables

General

NonDifferentiators

Predestinarians

/ Disciplined

Afraid

MTI

-0.282***

-0.467***

-0.159

Age

0.001

-0.010*

0.006

-0.004

Regionc

Middle Atlantic

0.057

0.009

-0.111

0.315**

East North Central

0.094**

-0.057

-0.002

0.366**

West North Central

0.163***

-0.116

0.136

0.338**

South Atlantic

0.065

-0.076

-0.107

0.380***

East South Central

0.026

-0.257

-0.107

0.285*

West South Central

0.159***

0.012

0.153

0.195

Mountain

-0.043

-0.052

-0.413**

0.142

Pacific

0.090*

0.015

-0.021

0.216

Gender

Female

0.097

2644.82

Primary Media Sourcec Newspaper

0.071 **

0.102

0.204 ***

0.007

Magazines

-0.003

1.873***

-0.002

-1.113

Radio

-0.049

-0.310**

0.108

-0.042

Internet

-0.106***

-0.058

-0.057

-0.152**

Local Church

-0.038

0.038

-0.472

-0.514

Other (Specify)

-0.225*

-0.328

-0.501*

-0.484

Household Agec

30 through 39 Years

-0.166***

-0.211*

-0.280*

-0.226

40 through 49 Years

-0.242***

-0.014

-0.452***

-0.374*

50 through 59 Years

-0.286***

0.046

-0.314

-0.278

60 Years and Over

-0.300***

0.092

-0.390

-0.326

Household Incomec $30,000

- $49,999

-0.001

-0.025

0.043

-0.049

$50,000 - $74,999

0.000

-0.031

0.028

-0.082

$75,000 and Over

-0.007

-0.121

0.128

-0.041

Household Sizec 2

Members

0.011

-0.070

0.082

-0.041

3 Members

0.069*

-0.038

0.203*

0.004

4 Members

0.049

-0.001

0.139

-0.196

5 or More Members

0.109**

0.127

0.068

0.019

Racec

Black/African-American

0.157***

0.038

-0.020

0.477***

Asian or Pacific Islander

0.275***

0.155

0.089

0.460*

American Indian, Aleut Eskimo

-0.014

-0.123

0.087

-0.831*

Other

0.058

0.489**

-0.114

0.214

years)

Table 4.6A Contd.

Graduated College- Bachelor’s Degree (4 years)

-0.122***

-0.168*

-0.081

-0.201**

Post Graduate Degree

-0.162***

0.006

-0.161

-0.290**

Marital Statusc Never

Married

-0.010

-0.202**

0.034

0.142

Divorced, Widowed, Separated

-0.012

-0.121

-0.023

-0.013

-0.152*** -0.258*** -0.068

Log Likelihood -26152.27 -2976.68 -3198.15

Numeber of observation

12236

1375

1492

1335

LR chi2(40)

267

85.93

67.83

76.21

Prob > chi2

0.0039

0.0005

Pseudo R2

0.0051

0.0142

0.0105

0.0142

Education Grade

School

0.213

1.508*

0.215

0.399

Some High School

-0.025

0.147

-0.039

-0.306*

Some College-no degree

-0.114***

-0.006

-0.103

-0.163**

Graduated College –Associate’s Degree (2

-0.089**

0.052

-0.167

-0.236**

*: significant at 0.10 level, **: significant at 0.05 level, ***: significant at 0.01 level

C The selected reference variables are: region1- New England; mediause1- Television; householdage1- under 30 year;

householdincome1- under $30,000; housemembers1- 1, race1 – white, education1- High school, marital status - Currently

Married.

Table 4.6B Ordered Probit model for each Segment for Consumers Attitude toward

Preparedness of the Food Supply Chain

Life

#Predestinar

Trendy and Freedom Planners/ Life

Variables ians/

Adventurists Seekers Freedom Planners

Optimists

Seekers

-0.210

0.012**

Regionc

Middle Atlantic

0.009

-0.044

0.284 *

-0.078

0.112

East North Central

0.111

-0.042

0.234

-0.021

0.142

West North Central

0.183

0.080

0.384**

0.044

0.201

South Atlantic

0.037

0.019

0.263

-0.096

0.108

East South Central

-0.053

-0.165

0.347*

0.093

0.070

West South Central

0.022

0.278**

0.340*

0.165

0.169

Mountain

-0.130

-0.024

0.125

0.039

0.006

Pacific

0.099

0.025

0.256

-0.075

0.265

Primary Media Sourcec Newspaper

0.087

-0.008

0.017

0.087

-0.080

Magazines

0.240

0.377

-0.525

-0.694**

-0.223

Radio

0.002

0.016

-0.132

-0.107

-0.045

Internet

0.083

-0.167***

-0.107

-0.141**

-0.080

Local Church

0.594

0.173

-5.837

No Obs.

0.476

Other (Specify)

-0.120

-0.333

0.825**

-0.352

-0.173

Household Agec

30 through 39 Years

-0.069

-0.144

0.253*

-0.226

-0.173

40 through 49 Years

-0.217*

-0.207

0.154

-0.165

-0.122

50 through 59 Years

-0.427**

-0.424**

0.363

-0.176

-0.182

60 Years and Over

-0.569**

-0.391

0.697**

-0.291

-0.178

Household Incomec $30,000

- $49,999

-0.104

-0.044

0.058

0.042

-0.061

$50,000 - $74,999

-0.088

-0.045

0.041

-0.107

0.061

$75,000 and Over

-0.178**

0.010

-0.026

-0.058

0.002

MTI

0.393

0.269

0.496

***

0.024

Age

0.007

0.021

***

0.000

0.003

0.237

***

0.058

0.148

3278.81

2689.12

3701.63

Table 4.6B Contd.

Household Sizec 2

Members

0.022

-0.023

0.098

0.134

0.016

3 Members

0.121

0.138

0.198*

0.124

-0.003

4 Members

0.094

-0.092

0.213*

0.137

0.038

5 or More Members

0.062

0.186

0.110

0.284**

0.140

Racec

Black/African-American

0.068

0.178

0.013

0.141

0.155

Asian or Pacific Islander

0.307**

0.181

0.172

0.299*

0.114

American Indian, Aleut Eskimo

0.331

-0.356

0.299

0.429

0.259

Other

0.099

0.006

0.034

-0.080

-0.122

Educationc Grade

School

-0.631

0.640

No Obs.

0.227

Some High School

-0.294*

0.307

0.347

0.249

0.225

Some College-no degree

-0.227***

-0.035

0.058

-0.219**

-0.022

Graduated College –Associate’s Degree

(2 years)

-0.163*

0.015

-0.024

-0.157

0.006

Graduated College- Bachelor’s Degree (4

years)

-0.275***

0.094

0.011

-0.225**

0.017

Post Graduate Degree

-0.273***

0.014

-0.019

-0.246***

-0.115

Marital Statusc Never

Married

0.041

-0.054

0.084

-0.068

0.056

Divorced, Widowed, Separated

0.058

-0.112

0.207**

0.044

-0.017

Genderc

Female -0.175*** -0.234

Log Likelihood -3699.24 -3376.81

Numeber of observation

1815

1564

1274

1808

1573

LR chi2(40)

72.75

68.58

60.33

68.26

34.98

Prob > chi2

0.0012

0.0033

0.0158

0.0019

0.6956

Pseudo R2

0.0097

0.0103

0.0111

0.0091

0.0052

*: significant at 0.10 level, **: significant at 0.05 level, ***: significant at 0.01 level

C The selected reference variables are: region1- New England; mediause1- Television; householdage1- under 30 year;

householdincome1- under $30,000; housemembers1- 1, race1 – white, education3- High school, marital status -

Currenly Married.

4.4.2.1 Overall/General Model

The overall model for variable CFSTP measuring respondent’s attitudes regarding how

prepared we are for food safety/defense events compared to one year ago was significant at

greater than 99 percent of confidence level. The continuous explanatory variable Media Tracking

Index (MTI) was negative and significant at 1 percent significance level. This indicated that with

increase in the media coverage, respondent’s attitude that the United States food system was not

prepared to deal with food safety events grows (Table 4.6A).

In the model East North Central, West North Central, West South Central and Pacific

regions had a positive direction and were significant. This meant that relative to the base region

of New England respondents in these four regions believed that our food system was better

prepared to deal with the food safety events as compared a year before. In the model media

source Newspaper had a positive and significant impact, while Internet and Other specified

media source had a significant and negative impact on the consumers’ attitude regarding the

preparedness of food system in dealing with the food safety events. This suggested that relative

to the consumers with Television as primary media source the consumers who listed Newspaper

as their primary media source believed that our food system was better prepared to deal with the

food safety events than it was a year ago. However consumers with Internet and Other specified

information sources believed that our food system was not better prepared to deal with the food

safety events than it was a year ago, relative to the consumers who rely on television as their

primary media source.

Consumers through all household age categories relative to the household age category of

Under 30 years had a significant and negative impact on consumer’s attitude regarding the

preparedness of the food system in dealing with food safety events. This could be interpreted as,

relative to the consumers in household age category of Under 30 years, the consumers in all the

higher household age categories believed that our food system was not as prepared to deal with

food safety events as it was a year ago.

The household size of three members and five or more members had a negative and

significant effect on consumer’s attitude regarding the preparedness of our food system to handle

the food safety/defense events, meaning relative to a consumer with the household size of single

member the consumers in these two categories believe that our food system was not prepared to

deal with the food safety/defense events than it was a year ago. Consumer’s education also

affected their attitude regarding preparedness of food system, in the model relative to the

consumers who graduated from high school the consumer’s with some college, Associate degree,

Bachelor’s degree or post graduate degree believed that our food system was not prepared to deal

with the food safety/defense events than it was a year ago. Furthermore relative to the male

respondents female respondents believed that our food system was not better prepared to deal

with food safety events than it was a year ago.

The predicted probabilities for effect of Media Tracking Index on the consumers attitudes

regarding how prepared we are for food safety/defense events compared to one year ago are

plotted in figure 4.19. The probability of a person believing that United States food system is not

prepared to deal with food safety events grows with the increase in media coverage.

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

Figure 4.19 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Overall

Model

Over half of the total respondents had indicated that they used television as their primary

information source. The predicted probabilities relative to the television are plotted in the figure

4.20. As can be observed the predicted probabilities were almost the same for all media sources

except for the other specified media source relative to television.

Figure 4.20 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Overall Model

4.4.2.2 Non-Differentiators

The ordered probit model for segment Non-Differentiators was significant at greater than

99 percent confidence level, also the continuous variable MTI was significant at greater than 99

percent confidence level and had a negative direction. This could be explained as, if the MTI

value increases the consumer attitude regarding the preparedness of the food system to deal with

0.1

0.2

0.3

0.4

0.5

0.6

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

food safety events increasingly changes that our food system is not better prepared to deal with

the food safety events. Another continuous variable age had a significant and negative effect,

thus as the respondents age increases the attitude that our food system is better prepared for food

safety events, changes to that it is not better prepared (Table 4.6A).

Relative to television as the primary media source, Magazines and Radio had a

significant effect; Magazine had a positive effect, while Radio had a negative effect on

consumer’s attitude regarding the preparedness of the food system in dealing with food safety

events. So consumers in the segment Non-Differentiators with Magazines as primary media

source believed that our food system is better prepared to deal with food safety events than it was

a year ago, relative to the consumers who used television as their primary media source, whereas

consumers with radio as primary media source believed it’s not prepared to deal with food

safety/defense events.

Consumers within household age category 30 through 39 years relative to consumers

within household age under 30 years believed that our food system is not prepared to deal with

food safety events than it was a year ago.

Relative to consumers who have graduated from high school, consumers who had grade

school education believed that our food system is better prepared to deal with the food safety

events than it was a year ago, while consumers who had Bachelor’s degree were unlikely to

believe that.

Consumers who were never married relative to consumers who were currently married

believed that our food system is not better prepared to deal with food safety events. Similarly

relative to male respondents females believed that our food system is not better prepared to deal

with food safety events than it was a year ago.

In figure 4.21 the predicted probabilities showing the effect of Media Tracking Index on

consumers’ attitudes regarding how prepared we are for food safety/defense events compared to

one year ago for the segment Non-Differentiators are plotted for the three super categories. A

steep increase can be noticed for the probability of being Not Prepared, and a decrease can be

noticed in the other two categories. The increase in media coverage increases a person’s

probability believing that our food system is not prepared for food safety/defense events. The

figures with multiple outcomes are included in appendix C.

Figure 4.21 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Non-Differentiators

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Med

ia Tracking

Ind

Not Prepared

Middle

Better Prepared

Figure 4.22 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the

NonDifferentiators

Predicted probabilities showing effect of media source on respondents attitudes regarding

preparedness of our food system for food safety events within segment Non-Differentiators are

shown in the figure 4.22. This segment had relatively a little bit higher population using

television as media source than general population. The media sources magazine and radio had a

significant effect in the model.

4.4.2.3 Predestinarians/Disciplined

The log likelihood statistic for the segment Predestinarians/Disciplined indicated that the

model was significant at greater than the 99 percent level of confidence. The continuous variable

for MTI was not significant in the model (Table 4.6A).

Only mountain region had a significant effect in the model relative to the New England

region. The negative direction of the coefficient for Mountain region indicated that people

residing in Mountain region of country believed that our food system is not prepared to deal with

the food safety events than past year. Relative to the consumers with Television as the primary

media source, consumers who listed Newspaper as their primary media source believed that our

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Local Church

Other

Sour

ary Media

Not Prepared

Middle

Better Prepared

food system is better prepared to deal with food safety events than it was a year ago, whereas

consumers who used Other specified sources as their primary source of information do not

believe that our food system is prepared to deal with food safety/defense events.

Consumers within the household age categories 30 through 39 years and 40 through 49

years relative to the consumers within household age category Under 30 years believed that our

food system is not prepared to deal with the food safety/defense events than it was a year ago.

The consumers with the household size of three members relative to the consumers with a single

member household size believed that our food system is better prepared to deal with food safety

events than it was a year ago.

Figure 4.23 displays the predicted probabilities showing the effect of Media Tracking

Index on respondent s attitudes regarding the preparedness of our food system in dealing with the

food safety/defense events for the segment Predestinarians/Disciplined.

Figure 4.23 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Predestinarians/Disciplined

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

Figure 4.24 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Predestinarians/Disciplined

Predicted probabilities for different media sources have been plotted in figure 4.24 for the

segment Predestinarians/Disciplined. Population in the segment Predestinarians/Disciplined

indicated relatively higher use of Newspaper as primary media source than the general

population. It can be observed, the probability for person with local church and Other specified

sources as their primary media source believed that our food system is not prepared for food

safety/defense events is higher than other media sources.

4.4.2.4 Afraid

The ordered probit model for the segment Afraid was significant at greater than 99

percent confidence level. However; the explanatory variable MTI was not significant in the

model (Table 4.6A).

The coefficients for geographic regions of country Middle Atlantic, East North Central,

West North Central, South Atlantic, and East South Central were significant and positive in the

model. Denoting relative to the consumers residing in the New England region of country

0.1

0.2

0.3

0.4

0.5

0.6

Newspaper

Magazine

Radio

Internet

Local Church

Other

Pri

mary Media

Sou

rce

Not Prepared

Middle

Better Prepared

consumers living in these five regions of the country believed that our food system is better

prepared to deal with the food safety/defense events than it was a year ago.

Only the coefficient for the media source Internet was significant, relative to the

consumers who use television as their primary media source consumers with Internet as primary

media source believed that our food system is not better prepared to deal with the food safety

events than it was a year ago. Similarly consumers within household age category 40 through 49

years relative to consumers within household age under 30 years believed that our food system is

not better prepared to deal with the food safety/defense events than it was a year ago.

Relative to the consumers who had graduated from high school, the consumers with some

high school education, some college, Associates degree, Bachelor’s degree, and post graduate

degree believed that our food system is not better prepared to deal with the food safety/defense

events than it was a year ago.

Similar to the probability of consumer confidence in the food safety for the Afraid

segment, the probability of a person in the Afraid segment believing that our food system is not

prepared for dealing with food safety/defense events was very high and does not vary much with

change in media coverage (Figure 4.25).

Figure 4.25 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for Segment Afraid

Also for all the media sources the probability of a person believing food system is not

prepared for food safety/defense events is very high (Figure 4.26). The primary media source for

this segment is television.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

Figure 4.26 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Afraid

4.4.2.5 Trendy and Adventurists

The log likelihood statistic for the segment Trendy and Adventurists indicated that the

model was significant at greater than 99 percent level of confidence. The continuous variable for

MTI was not significant in the model (Table 4.6B).

The continuous variable age had a significant and positive effect in the model, with the

increase in age consumers believed that our food system is better prepared to deal with food

safety events than it was a year ago.

Consumers through all household age categories except for the 30 through 39 years, relative

to the household age category of Under 30 years, had a significant and negative impact on

consumer’s attitude regarding the preparedness of the food system in dealing with the food

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

safety/defense events. This can be explained as, relative to the consumers in household age

category of under 30 years, consumers in all the higher household age categories of 40 through 49

years, 50 through 59 years, and over 60 years believed that our food system is not better prepared

to deal with the food safety/defense events than it was a year ago. Similarly relative to consumers

with the household income less than $30,000, the consumers with household income over $75,000

believed that our food system is not better prepared to deal with food safety events than it was a

year ago.

In the segment Trendy and Adventurists relative to consumers who had graduated from

high school, consumers with some high school education, some college, Associates degree,

Bachelor’s degree, and post graduate degree believed that United States food system is not better

prepared to deal with food safety events than it was a year ago. Also relative to male consumers

females in the segment believed that United States food system is not better prepared to deal with

food safety events than it was a year ago.

The Media Tracking Index (MTI) did not have a significant effect in the model, but it can

be noted that the probability of a person believing our food system is not prepared for food

safety/defense events below the middle category (Figure 4.27).

Majority of the people in the segment used television as their primary media source;

however none of the media sources are significant relative to television (Figure 4.28).

Figure 4.27 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Trendy and Adventurists

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

Figure 4.28 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Trendy and Adventurists

4.4.2.6 Freedom Seekers

The ordered probit model for the segment Freedom Seeker was significant at greater than

99 percent confidence level. The continuous explanatory variable Media Tracking Index (MTI)

was negative and significant. This indicated that with the increase in media coverage,

respondent’s attitude that the United States food system is not better prepared to deal with food

safety/defense events that it was a year ago grows (Table 4.6B).

Only West South Central region had a significant effect in the model relative to the New

England region. The positive direction of the coefficient for West South Central region indicated

that people residing in West South Central region of country believed that our food system is not

better prepared to deal with the food safety/defense events than past year. Also only the

coefficient for media source Internet was significant, relative to the consumers who used

television as their primary media source, the consumers with Internet as their primary media

source believed that our food system is not better prepared to deal with the food safety/defense

events than it was a year ago.

Relative to the consumers within household age category under 30 years, the consumers

within household age category 50 through 59 years believed that our food system is not better

prepared to deal with the food safety/defense events than it was a year ago. Similar to household

age, relative to the male respondents females in the segment believed that the United States food

system is not better prepared to deal with the food safety events than it was a year ago.

For the segment Freedom Seeker, the continuous variable Media Tracking Index had a

significant effect in the model. The predicted probability of a person believing our food system is

not prepared for the food safety/defense events increases with the increase in media coverage

(Figure 4.29)

Figure 4.29 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Freedom Seeker

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

Figure 4.30 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Freedom Seeker

Percentage of the consumers who used Internet and Newspaper as their primary media

source within this segment was higher with respect to the general population. Internet had a

significant effect in the model and as can be noticed the probability of the consumers who

believed that our food system is not prepared for the food safety/defense with internet as primary

media source is more than 0.5.

4.4.2.7 Life Planners/Freedom Seekers

The model for the segment Life Planner/Freedom Seeker was significant at greater than

95 percent level of confidence. The important explanatory continuous variable MTI did not had a

significant effect for the segment Life Planner/Freedom Seeker. Another continuous variable age

had a significant and negative effect, thus as the respondents age increases the attitude that our

food system is better prepared for the food safety/defense events, changes to that it is not better

prepared (Table 4.6B).

In the model Middle Atlantic, West North Central, East South Central and West South

Central regions had a positive direction and were significant. This means that relative to the base

region of New England respondents in these four regions believed that our food system is better

prepared to deal with the food safety/defense events as compared a year before. Among the

primary media sources only the consumers who used Other specified media source believed that

countries food system is better prepared to deal with food safety events as compared a year

before.

Consumers within household age 30 through 39 years and over 60 years believed that the

United States food system is better prepared to deal with the food safety events as compared a

year before. In the same way the consumers with household size of three members and four

members relative to the consumer with single member household size believed that our food

system is better prepared to deal with the food safety/defense events than it was a year ago. Also

consumers who were divorced, widowed, or separated believed that our food system is better

prepared to deal with the food safety/defense events than it was a year ago.

The Figure 4.31 presented predicted probabilities for the continuous variable MTI over

consumer attitudes regarding preparedness of our food system to deal with food safety events for

the segment Predestinarians/optimists. MTI did not have a significant effect on this segment,

though the probability of a person in the segment believing our food system is not prepared is

always higher.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

Figure 4.31 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Life Planners/Freedom Seekers

Figure 4.32 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Life Planners/Freedom Seekers

4.4.2.8 Life Planners

The log likelihood statistic for the segment Life Planner indicated that the model was

significant at greater than the 99 percent level of confidence. The continuous explanatory

variable Media Tracking Index (MTI) was negative and significant. This indicated that with the

increase in the media coverage respondent’s attitude that United States food system is not better

prepared to deal with the food safety/defense events compared to a year before grows (Table

4.6B).

In the segment Life Planner consumers who indicated Internet and Magazines as their

primary media source had a significant and negative impact on consumer’s attitude regarding the

preparedness of food system in dealing with the food safety/defense events. This suggested that

0.2

0.4

0.6

0.8

1.2

Newspaper

Magazine

Radio

Internet

Local Church

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

100

consumers with Internet and Magazines as their primary information sources believed that our

food system is not better prepared to deal with food safety events than it was a year ago, relative

to the consumers who rely on television as their primary media source.

Consumers with household size of five or more members believed that our food system is

not better prepared to deal with the food safety/defense events than it was a year ago, relative to

the consumers with household size of single member. In the segment consumers’ education also

affected their attitude regarding preparedness of food system, in the model relative to the

consumers who graduated from high school, the consumer’s with some college, Bachelor’s

degree and post graduate degree believed that our food system is not better prepared to deal with

the food safety/defense events than it was a year ago. As well relative to the male consumers

female consumers in the segment believed that our food system is not better prepared to deal

with the food safety events than it was a year ago.

Media Tracking Index had a significant effect in the segment Life Planner, the variability

in the probability of a person being in one of the categories can be easily noticed in the

probability curve plotted in the figure 4.33

101

Figure 4.33 Probability Curve Showing the Effect of MTI on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Life Planners

The percentage of consumers with newspaper as primary media source was higher for the

segment Life Planner than the general population. Although the probability of a person with

newspaper as primary media source believing food system is not prepared was slightly higher

than being in other categories, for other sources its much higher.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Media Tracking Index

Not Prepared

Middle

Better Prepared

102

Figure 4.34 Probability Curve Showing the Effect of Media Source on Consumers Attitudes

Regarding Preparedness of U.S. Food System to Deal with Food Safety for the Segment

Life Planners

4.4.2.9 Predestinarians/Optimists

The log likelihood statistic for the segment Predestinarians/Disciplined indicates that the

model is not significant at all.

4.4.2.10 Marginal Effects

Similar to the results presented before, due to the large number of explanatory variables

for each segment and ease of understanding, the marginal effects for each segment are explained

together by comparing among the segments. Marginal effects for the variables which were

significant in the model are explained below, and the tables showing marginal effects for each

segment are included in the appendix D.

The MTI was significant in the overall model and in the segments Non-Differentiators,

Freedom Seeker, Life Planner. The lower categories of the dependent variable had a positive

sign, while the higher categories had negative signs across all the segments. The marginal effects

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Newspaper

Magazine

Radio

Internet

Other

Primary Media Source

Not Prepared

Middle

Better Prepared

103

for the MTI imply that increase in media coverage of food safety events decreases the

probabilities that a subject’s response falls in the higher categories for the dependent variable.

This is consistent with the finding that the MTI negatively impacts consumers’ attitude towards

preparedness of our food system in dealing with the food safety/defense events. For example, if a

person in the segment Freedom Seeker indicates category 10 of the dependent variable, an

additional unit increase in media coverage will result in decreasing that person’s probability of

believing that our food system is prepared to deal with food safety event and of being in the

category 10 by 0.0158, holding all the other variables constant at their mean.

The continuous variable age was only significant in the segment Non-Differentiators,

Trendy and Adventurists, Life Planner/Freedom Seeker. For the segments Non-differentiators

and Life Planner/Freedom Seeker the marginal effects indicates that the probability of a person

to believe that our food system is prepared to deal with the food safety events decreases with the

increase in age, while the probability of believing that our food system is not prepared to deal

with the food safety events increases with the increase in age. While for the segment Trendy and

Adventurists the probability of believing that our food system is prepared to deal with the food

safety events increases with increase in age, and probability of that our food system is not

prepared to deal with the food safety events decreases with increase in age, holding all other

variables constant at their mean.

The Middle Atlantic region had a significant effect in the segments Afraid and Life

Planner/Freedom Seeker. The marginal effects for both of these segments can be interpreted as,

the consumers’ attitude regarding preparedness of food system that our food system is prepared

for the food safety/events increases if the person is living in either of these regions compared to a

person living in New England region. For East North Central region the overall model and the

104

segment Afraid had a significant effect. For West North Central the overall model and the

segments Afraid, Life Planner/Freedom Seeker had a significant effect. For South Atlantic region

only segment Afraid had a significant effect. For East South Central region the segments Afraid

and Life Planner/Freedom Seeker had a significant effect. For West South Central region the

overall model and the segments Freedom Seeker, Life Planner/Freedom Seeker had a significant

effect. For the Pacific region only the overall model had significant effect. Marginal effects for

all of these regions could be interpreted in a same way as of Middle Atlantic region. The

Mountain region was significant in only the segment Predestinarians/Disciplined, the marginal

effects for the segment are explained as, if a person is living in the Mountain region his

probability of believing that our food system is not better prepared to deal with the food

safety/defense events increases, relative to a person living in New England region. For instance

in the overall model, the probability of being in category 8 of preparedness scale is 0.0119 higher

for Pacific region than New England region, holding all other variables at their means.

The media source newspaper had a significant effect for the overall model and the

segment Predestinarians/Disciplined; the marginal effects signify that if a person is using

newspaper as his/her primary media source the probability of him/her believing that our food

system is better prepared to deal with the food safety/defense events increases relative to a

person who uses television as his/her primary media source. Magazines had a significant effect

for segment Non-Differentiators and Life Planner, the marginal effects signify that if a person is

using magazines as his/her primary media source the probability of him/her believing that our

food system is not prepared to deal with the food safety/defense events increases relative to a

person who uses television as his/her primary media source. Radio was significant only for the

segment Non-Differentiators; the marginal effects for the segment could be explained in a same

105

way as the Magazines. Also internet was significant for the overall model and the segments

Afraid, Freedom Seeker, Life Planner, the marginal effects for them can be interpreted in a same

way as for magazines. For example in the segment Afraid, the probability of being in category 6

of preparedness scale is 0.0143 lower for media source Internet than Television, holding all other

variables at their means.

Household age category 30 through 39 years had a significant effect for the overall

model, and the segments Non-Differentiators, Predestinarians/Disciplined, and Life

Planner/Freedom Seeker. The marginal effects for the overall model and the segments

NonDifferentiators, Predestinarians/Disciplined could be interpreted as, if a person is within

household age category 30 through 39 years the probability of him/her believing that our food

system is not prepared to deal with the food safety/defense events increases relative to a person

within household age category under 30 years , while for segment Life Planner/Freedom Seeker

the marginal effects can be interpreted as if a person is within household age category 30 through

39 years the probability of him/her believing that our food system is better prepared to deal with

the food safety/defense events increases relative to a person within household age category under

30 years. For instance in the segment Life Planner/Freedom Seeker, the probability of being in

category 10 of preparedness scale is 0.0114 higher for household age category 30 through 39

years than under 30 years, holding all other variables at their means.

Household age category 40 through 49 years had a significant effect for the overall

model, and the segments Predestinarians/Disciplined, Afraid, and Trendy and Adventurists. The

marginal effects for the overall model and the segments can be interpreted as, if a person is

within household age category 40 through 49 years the probability of him/her believing that our

106

food system is not prepared to deal with the food safety/defense events increases relative to a

person within household age category under 30 years . Household age category 50 through 59

years had a significant effect for the overall model, and the segments Trendy and

Adventurist, and Freedom Seeker. The marginal effects for the overall model and the segments

can be interpreted as, if a person is within household age category 50 through 59 years the

probability of him/her believing that our food system is not prepared to deal with the food

safety/defense events increases relative to a person within household age category under 30

years.

Household age category over 60 years had a significant effect for the overall model, and

the segments Trendy and Adventurist, and Life Planner/Freedom Seeker. The marginal effects for

the overall model and the segment Trendy and Adventurists can be interpreted as, if a person is

within household age category over 60 years the probability of him/her believing that our food

system is not prepared to deal with the food safety/defense events increases relative to a person

within household age category under 30 years, while if the person is in the segment Life

Planner/Freedom Seeker the probability of him/her believing that our food system is better

prepared to deal with the food safety/defense events increases relative to a person within the

household age category under 30 years.

Only household income category over $75,000 had a significant effect for the segment

Trendy and Adventurist. The marginal effects can be explained as, if a person is with household

income of over $75,000 his/her probability of believing that our food system is better prepared to

deal with the food safety/defense events decreases relative to a person with household income

under $30,000. For example, the probability of being in the outcome 5 is 0.0136 higher for

107

household income over $75,000 than household income under $30,000, holding all other

variables at their means.

The household size of three members had a significant effect in the overall model and the

segments Predestinarians/Disciplined and Life Planner/Freedom Seeker. The marginal effects

indicate that, if a person is with the household size of three members his/her probability of

believing that our food system is better prepared to deal with the food safety/defense events

increases relative to a person with single member in household. The household size of four

members had a significant effect for Life Planner/Freedom Seeker and can be interpreted in a

same way as household size of three members. For example in the segment

Predestinarians/Disciplined, the probability of being in the outcome 3 of the preparedness scale

is 0.0128 lower for household size of three members than household size of single member,

holding all other variables at their means.

The household size of five or more members had a significant effect in the overall model

and segment Life Planner. The marginal effects for the overall model and the segment Life

Planner indicate that if a person is with the household size of five or more members his/her

probability of believing that our food system is better prepared to deal with the food

safety/defense events increases relative to a person with single member in household.

The education level grade school had only significant effect in the segment

NonDifferentiators. The marginal effects indicate that, if a person had grade school level

education his/her probability of believing that our food system is better prepared to deal with the

food safety/defense events increases relative to a person who graduated from high school. As

oppose to it, if a person has some high school education and is in the segment Afraid and Trendy

and

108

Adventurists, his/her probability of believing that our food system is not prepared to deal with

the food safety/defense events increases relative to a person who graduated from high school.

The category some college education was significant in the overall model, and the segments

Afraid, Trendy and Adventurists and Life Planner, the marginal effects can be interpreted in a

same way as category some high school education. Education category Associate’s degree had

significant effect for the overall model and the segment Afraid, and Trendy and adventurists, it

can be also be interpreted in a same way as some high school education. Education category

Bachelor’s degree had significant effect for the overall model and the segment Afraid, Trendy

and adventurists, and Life planner it can be also be interpreted in a same way as some high

school education. Education category post graduate degree had significant effect for the overall

model and the segment Afraid, Trendy and adventurists, and Life planner it can be also be

interpreted in a same way as some high school education. For instance in the segment Life

Planner, the probability of being in outcome 9 is 0.0124 lower for Post graduate degree than high

school graduation, holding all other variables at their means.

Females had a significant effect for the overall model and the segments Non-

Differentiators, Trendy and Adventurists, Freedom Seeker, and Life Planner. The probability of

females believing that our food system is not prepared to deal with the food safety/defense

events increases relative to the males in these segments. For example in the segment Freedom

Seeker, the probability of being in outcome 7 is 0.0188 lower for females than males, holding all

other variables at their means.

109

CHAPTER 5: SUMMARY AND CONCLUSION

5.1 Introduction

The increasing incidents of food recall events are affecting consumers’ confidence in the

safety of food and changing their attitude towards preparedness of the food system. Previous

research has shown that food safety/defense events affect consumer confidence in food safety

and consumer attitude towards preparedness of the food system (Kinsey et al., 2009). The loss of

consumer confidence has resulted in reduction of demand for food products and economic loss

(De Jonge et al., 2004). These food safety incidents and terrorist attacks have raised questions

about preparedness of our food system (Degeneffe et al., 2006). For a long time consumers look

toward government and food industry to make sure the food they consume is safe (FSIS).

Government has established several agencies to make sure the food products distributed for

consumption are safe. Also food industry has been stepping up and trying to make sure food is

safe by improving handling, manufacturing and distributing practices.

Even though food industry and various government agencies are making an effort to

control food safety incidents, with the increase in trade, food products and knowledge of new

pathogens the number of food scare events are on rise (FSIS). These food scare events are

catching eye of new age media. The mass media industry today is much bigger than the mass

media industry a decade ago and has a greater outreach and impact, and as the previous studies

have suggested media plays an important role in altering consumer confidence over the food

safety issues (Kalaitzandonakes et al., 2004). The main purpose of this study was to measure the

impact of mass media coverage on consumers’ confidence in food safety and attitude toward

preparedness of the U.S. food supply chain. The study also investigated the role of different mass

110

media outlets and consumer demographics in altering consumer confidence and attitude toward

preparedness of the U.S. food supply chain.

The specific objectives of this study were 1) to estimate media agenda-setting effects on

consumer confidence in the safety of the U.S. food supply chain, and 2) to analyze how the

agenda-setting effect varies across consumer segments. To accomplish these objectives, survey

data were collected using a nationwide Continuous Food Safety Tracking (CFST) survey

developed by researchers at University of Minnesota and Louisiana State University AgCenter.

The survey was conducted online by an online consumer participant’s pool of Taylor Sofres

Nelson Company. The CFST survey data was collected on a weekly basis. To estimate the effect

of mass media the data for media article count was collected on a daily basis at Louisiana State

University AgCenter using academic version of LexisNexis.

The study used 67 weeks of survey data with 12,236 observations collected from May 5th,

2008 to August 2009. Cluster analysis was used to segment individuals in eight different

segments based on their responses to characteristic statements in the CFST survey. The Kinsey et

al. (2009) study identified two primary indicators of consumer’s confidence. The first measures

consumer’s current confidence in the safety of U.S. food system, and the second measures their

belief regarding how better prepared the food system is regarding food safety relative to a year

ago. These two indicators used as ordered dependent variables in the study were created after

factor analyzing six questions from the survey. The first dependent variable created by using four

questions measures consumers confidence in safety of the food. The second variable created by

using the two remaining questions measuring consumers’ attitude towards preparedness of U.S.

food supply chain than it was a year before. An ordered probit regression analysis was used to

measure the effect of mass media coverage and other explanatory variables including media

111

sources on the consumers’ confidence in the food safety and attitude towards preparedness of our

food system.

5.2 Results

In the Continuous Food Safety Tracking (CFST) survey the country was divided into nine

geographic regions. The survey data was collected from these nine geographic regions. More

than fifty six percent of our respondents had household age of 50 years or more. Over thirty three

percent of the respondents had a household income of over $75,000 or more. Twenty seven

percent of our respondents had a household size of one member and forty percent had two

members. Around ninety percent of the survey respondents were white. The survey response was

weighted toward female respondents, in that nearly eighty percent were female. Around twenty

eight percent of the respondents had some college education and no degree. Over fifty five

percent of the respondents listed television as their primary media source.

This study followed a two-step approach, hierarchical cluster analysis followed by

partition clustering. Factor analysis was used to create seven factors from the consumer

characteristic statements. The factor scores obtained for the seven factors in the factor analysis

were used as input scores in the cluster analysis. The eight identified segments were named as

Non Differentiators, Predestinarians/Disciplined, Afraid, Trendy and Adventurists, Freedom

Seeker, Life Planner/Freedom Seeker, Life Planner, Predestinarians/Optimists.

Two separate models were developed to measure the change in consumer confidence and

consumer attitudes regarding the preparedness of the United States food supply across each

identified segment. The study found that media coverage had a significant and negative effect on

consumer confidence for the overall model, and in segments titled Trendy and Adventurists,

112

Freedom Seeker, and Life Planner in national food supply chain. In other words, an increase in

mass media coverage about food safety events leads to a decline in consumer confidence in the

U.S. food supply chain.

Over fourteen percent of respondents listed newspaper as their primary media source.

Newspapers had a significant and negative effect on consumer confidence for the overall model

and the segments titled Predestinarians/Disciplined, Afraid, Freedom Seeker, Life

Planner/Freedom Seeker, Life Planner, and Predestinarians/Optimists. Indicating that if a person

is using Newspapers as his/her primary media source the probability of him/her having low

confidence decreases, while the probability of being high in confidence increases relative to a

person who uses Television as their primary media source. Around twenty five percent of the

respondents listed Internet as their primary media source. Internet was significant and had a

negative effect for the overall model and in the segments Afraid and Life Planner/Freedom

Seeker, if a person is using Internet as his/her primary media source the probability of him/her

being low in confidence decreases while the probability of being high in confidence increases

relative to a person who uses Television as primary media source.

The results for respondents’ primary media sources were interpreted relative to television

coverage. Television is an audio visual media information source. Unlike print media the visual

transmission for television is expected to be more effective. The results indicate that using

television as primary media source increases the probability of falling in the lower confidence

range. Respondents for the segments identified in the study had different major primary media

sources which were used to characterize the respondents in the segment. In the segment titled

Afraid, the majority of people indicated television as their primary media source and throughout

the survey period they remained in the lower confidence range. While the segment titled

113

Freedom Seekers and Life Planner/Freedom Seekers have a higher percentage of people

indicating use of internet and newspapers as their primary media source. Use of internet and

newspapers as primary media source decreases the probability of falling in the lower confidence

range.

For the model to measure consumers’ attitudes towards the preparedness of food system,

MTI was significant in the overall model and in the segments Non-Differentiators, Freedom

Seeker, and Life Planners. In other words, an increase in mass media coverage about food safety

events strengthens consumer belief that the nation is not prepared for safety of its food supply

system. As mentioned before, over fourteen percent of the respondents listed newspaper as their

primary media source. The media source newspaper had a significant and positive effect relative

to television, for the overall model and the segment Predestinarians/Disciplined model. If a

person is using newspaper as his/her primary media source, the probability of him/her believing

that our food system is better prepared to deal with the food safety/defense events increases

relative to a person who uses television as his/her primary media source. Around a quarter of the

respondents listed Internet as their primary media source. Internet was significant and had a

positive effect for the overall model and the segments Afraid, Freedom Seeker, Life Planner.

Meaning that if a person is using Internet as his/her primary media source the probability of

him/her believing that our food system is better prepared to deal with the food safety/defense

events decreases relative to a person who uses television as his/her primary media source.

Most of the media sources have a similar effect on consumers’ attitude towards preparedness of

the food system, like they have on consumer confidence. Similar to consumer confidence

television also negatively affects consumers’ attitude towards preparedness of U.S. food supply

chain. In the segment titled Afraid television is the primary media source for most of the

114

respondents in the segment. The results indicate that using television as primary media source

increases the probability of believing that the U.S. food supply is not prepared against terrorist

attack or accidental contamination. Segment titled Freedom Seeker was characterized with most

of the respondents in the segment using internet and newspapers. The results show that using

internet and newspapers as primary media sources for information increases respondents

probability of believing that U.S. food supply is better prepared to deal with terrorist attack or

accidental contamination.

The results of the study imply that increase in media coverage has a negative effect on

consumer confidence and consumer attitude towards preparedness of U.S. food system.

Therefore the results support the hypothesis of an agenda setting effect. Also media sources like

television increases the probability of a person being in low confidence range relative to other

sources of media and believe that the U.S. food system is not prepared to deal with terrorist

attack or accidental contamination.

5.3 Implications

The findings of this study are important and helpful for government agencies and private

companies to understand the magnitude of consumer response to mass media, and for adjusting

their response to food safety incidents and determining the economic downturn in the sale of

their products and for how long into the future. The consumer segments developed in the study

can be used for integrating better risk communication strategies directed toward a specific

consumer segment.

5.4 Limitations and Future Research

The sample of respondents obtained for this study is not representative the sample is

skewed towards female respondents. Around ninety percent of respondents are white. Some

115

studies have indicated that media coverage has a residual effect on consumer confidence and

have recommended use of time lag to capture the accurate effect of media coverage on consumer

confidence(Verbeke and Ward, 2001). This study did not use time lag to capture the effect of

media coverage on consumers’ confidence. Future research will focus on determining the time

lag to accurately measure effect of media coverage. Also this study did not differentiate between

negative media coverage and positive media coverage. Future research can also explore what is

the dollar value economic effect of mass media coverage of food safety incidents on the product

sales and economy.

116

Students also viewed