You are on page 1of 18

Consumer research in the restaurant environment.

Part 2: Research design and analytical methods


JaksÏa Kivela
Associate Professor, Department of Hotel and Tourism Management,
Hong Kong Polytechnic University, Kowloon, People's Republic of China
John Reece
Lecturer in Psychology, Department of Psychology and Intellectual Disability Studies,
RMIT University, Bundoora, Australia
Robert Inbakaran
Lecturer in Tourism, Department of Hospitality, Tourism and Leisure,
RMIT University, Melbourne, Australia

Keywords Development of questionnaire (research


Customer satisfaction, 1. Operationalisation of the model instrument)
Restaurants, Hong Kong of dining satisfaction and return For the purpose of this study, a questionnaire
patronage
Abstract listing 28 restaurant attributes (see Appen-
In Part 1, a model of dining Measures of dining experiences, attribute dix) was developed as a result of pilot testing,
satisfaction and return patronage importance, customer expectations and sa- and was subsequently used to operationalise
was developed and described.
tisfaction have been diverse and dependent and test the model which was described in
Based on extensive review of the
relevant consumer behaviour lit- upon the domain of research, in both the Part 1 of the paper series. To develop the
erature the model was developed number of attributes and level of abstraction.
questionnaire a series of three semi-struc-
and underpinned by the disconfir- In the hospitality and tourism context, nu-
mation and expectancy theory. As tured face-to-face interviews (Delphi techni-
merous service quality research instruments
noted in the article, disconfirma- que) were conducted with:
have been used to measure customer satis-
tion theory is widely accepted as 1 Hong Kong Polytechnic University
an account of the process by faction. However, in the restaurant context
(HKPU) hotel and catering management
which customers develop feelings (excluding the fast-food category) there is a
of satisfaction or dissatisfaction, lack of survey instrument development, with faculty (n = 3 6 6);
that is, when customers compare
the exception of guest comment cards. 2 Maitre'ds in Category 2 restaurants
new dining experiences with some (n = 3 6 6); and
basis that they have developed Equally, service quality models such as
SERVQUAL and DINQUAL have not been 3 a systematic sample of diners (every 5th
from prior experiences. On the
other hand, the assumption that a able to identify the salient attributes that are customer who walked into the restaurant,
customer will weigh various res- particular to restaurant operations. It could n = 3 6 20),
taurant attributes is based on
expectancy theory. In the majority
be argued that restaurateurs and restaurant in various theme/ambience[1] restaurants in
of studies using disconfirmation marketers continue to be hampered by both the Tsim Sha Tsui area of Hong Kong (see
theory, expectations are formed the lack of appropriate customer feedback
according to customers' pre-ex-
Table I). The selection of restaurants for the
and problems related to the content validity
perience beliefs and standards pilot survey was based on a convenience
of guest comment cards, i.e. they measure
that they use to measure their sample because the researcher was familiar
purchase experience. These the- post-dining perceptions only. In addition, due
to high levels of fragmentation and diversi- with the operations. The purpose of the
ories bring together the social,
psychological and cultural con- fication in both the range of products, i.e. interviews was to elicit opinions from re-
cepts into four distinct groups of menus and menu items and market seg- spondents about the following questions:
variables: input variables both 1 What restaurant attributes did they nor-
internal and external, process
ments, it is difficult to generalise restaurant
customer satisfaction results. This is because mally associate theme/ambience restau-
variables and output variables
(Lowenberg et al., 1979; Finkel- restaurant industries have a distinct product rants with?
stein, 1989). This paper is a structure that is differentiated by price, 2 What were their most likely reasons for
continuation and explains: how dining out?
location, theme/ambience, service level, cui-
the model of dining satisfaction
and return patronage was opera- sine and style, which at the same time 3 What restaurant attributes were impor-
tionalised, that is, how the re- demand a wide variety of market segments tant to them when they dined out in
search instrument was developed; for the same products. It must also be noted theme/ambience restaurants?
how the sample size and survey that until this study, similar dining experi-
procedures were developed and The qualitative data from the first-round
conducted; and how the selection
ence studies have not been undertaken in
interviews were then collated for common-
of analytical procedures was con- Hong Kong. It was concluded that there was a
ceived. lack of appropriate research instruments ality and sorted under the following cate-
used for measuring dining satisfaction, so a gories:
new research instrument was developed for 1 food factors;
International Journal of 2 service factors;
Contemporary Hospitality this study.
Management 3 ambience factors;
11/6 [1999] 269±286 4 convenience factors; and
The current issue and full text archive of this journal is available at
# MCB University Press 5 a restaurant that offers . . . x (ambience/
[ISSN 0959-6119] http://www.emerald-library.com
service/product) factors.
[ 269 ]
JaksÏa Kivela, John Reece and Table I
Robert Inbakaran Restaurant categories
Consumer research in the
restaurant environment. Part Category 1 Category 2 Category 3 Category 4
2: Research design and
analytical methods Fine dining or gourmet Theme or ambience Family or popular Convenience or fast-food
International Journal of restaurants restaurants restaurants restaurants
Contemporary Hospitality Formal fine dining Formal authentic Informal Informal
Management
11/6 [1999] 269±286 (reconstructed atmosphere)
Informal panoramic Informal authentic
(reconstructed atmosphere)
Formal panoramic Informal entertainment ±
theme
Focus is on gastronomy; Presentation of Restaurants that offer Focus is on offering meals
sophisticated service; reconstructed theme or well-cooked and presented, to be consumed on site
elegant ambience; authenticity, i.e. ambience moderately priced meals and/or off-site for
spectacular views or and food style for singles, groups and customers who do not
location May be a fine dining families. Meals that can want to prepare meals for
Usually features authentic be substituted for home- themselves, and who
cuisine and decor made lunch/dinner meals. require a quick convenient
Menu items that suit the economic meal, with fast
taste of children and service
adults. Pleasant informal
dining atmosphere.
Comfortable seating, and
table service

Round two interviewees (the same respon- developed to operationalise the model and
dents as in round-one interviews, except that the following steps were employed in con-
the customer sample was substituted by structing the questionnaire:
another systematic selection of customers) . development of scales from a set of items
were then given the round-one attribute describing subjective perceptions (impor-
listings and were asked to elaborate, add tance), (expectations) and (expectations
additional items or disagree with the listings met) of restaurant attributes that were
in the context of three questions (above). related to the dining occasion, customer
Items from the round-two interviews were characteristics and customers' situational
then sorted out for commonality under the constraints; and
five categories and the same procedure for . validation of the scale.
third-round interviews and data analysis was
followed as for the second-round interviews. Validity
The qualitative analysis of the final round is Validity encapsulates two dimensions:
highlighted in Table II. Based on these 1 the discriminant validity of measured
findings, together with the findings of pre- expectations; and
vious research noted in Part 1, the question- 2 the dimensionality of customer satisfac-
naire for the study was developed. tion (construct validity).
Qualitative responses from the third-round
interviews and feedback that were similar Discriminant validity is established, for ex-
were compressed and identified with a gen- ample, when the between-factor item corre-
eric title. For example, dining out reasons lations are lower than the within-factor item
such as ``get together with friends'', ``office correlations. Miller (1977), who argued that
get together'', ``take family out'', ``regular multiple comparison standards for expecta-
weekly dine-out'' were compressed and iden- tions do exist first, raised the questions
tified as ``I am dining out because of a social associated with discriminant validity. In
occasion''. Responses such as ``it's conveni- accordance with this viewpoint, customer
ent'', ``too tired/busy to cook at home'', ``it's satisfaction researchers have measured dif-
more comfortable to eat out'', ``I don't know ferent expectations in their studies.
how to cook'', ``no need to wash-up after Churchill and Surprenant (1982), Tse and
meal'', ``convenient location'', were com- Wilton (1988) and Oh and Jeong (1996)
pressed as ``I am dining out because it is measured subjects' expectations on expected
convenient''. These were then used as the [predicted] product performance, whereas
basis for developing the questionnaire. Cadotte et al. (1987) measured both product
The questionnaire, items anchored to a norms and brand norms as comparison
five-point Likert (1967) scale, was then standards. Boulding et al. (1993) included
[ 270 ]
JaksÏa Kivela, John Reece and Table II
Robert Inbakaran Interview feedback results (items not in any particular order)
Consumer research in the
restaurant environment. Part Which restaurant attributes are
2: Research design and
analytical methods What attributes do you normally What are your most likely reasons for important to you when you dine
International Journal of associate theme/ambience dining out in theme/ambience out in theme/ambience
Contemporary Hospitality restaurants with? restaurants? restaurants?
Management
11/6 [1999] 269±286 Decor It's convenient Freshness of food
Theme of the restaurant Get together with friends Convenient location
Type and style of food Office get together Convenient parking
Foods which are not eaten at home Too tired/busy to cook at home Polite staff
Different eating experience Regular weekly dine-out Suitable for children also
Presentation of food For a meeting Quality of food
Nice tableware Birthday party Presentation of food
Being served and attended to On a date Friendly staff
Should be a happy place to go to Business lunch/dinner Food which is nutritious
Feels comfortable to eat there At the end of the month (just got paid) High standard of service
Nice view when dining To meet people Helpful staff
Variety of foods that are offered The restaurant was recommended Food which tastes good
Being a famous eating place To experience new foods Convenient parking
Being satisfied after dining It's more comfortable to eat out Good portion size
Offer experiences about different Entertaining a visitor from overseas Clean staff
foods and culture For a treat If the place is recommended
Should be a place of entertainment I was invited To experience different foods
Welcoming environment Somewhere to go Staff that can explain the menu
Take family out Clean food
It's part of my job Good wine list
I don't know how to cook Unusual foods
I like different foods Clean place
When there is a special promotion on
No need to wash-up after meal
Convenient location

measures of ideal expectations along the each of these research areas were brought
normative measures and found discriminant together to develop the customer satisfaction
validity between the two types of expecta- model described in Part 1. For example, the
tions. Elaborating further, the construct independent variables from the Delphi study
validity is important to consider because it as well as those from the previous customer
addresses the question of what the survey, satisfaction research identified in the litera-
and its measurement scale, is actually mea- ture review, namely supported Carman's
suring. On the other hand, content validity (1990) and Barsky's (1992) quality and dis-
focused on whether or not the items in the confirmation models respectively, are almost
survey adequately represented the construct identical as those representing the indepen-
under study. For example, the customer dent variables in this study. This also
satisfaction construct developed and imple- supports the inclusion of these specific terms
mented in this study was composed of three as well as their relationship with each other.
different sub-dimensions relating to a res- The model gives a customer satisfaction
taurant's products and services: score as a result of the interaction of these
1 expectations; terms. This customer satisfaction score
2 expectations met, i.e. perceived perfor- (EM 6 I) is content-valid therefore. This is
mance; and because it contains survey information from
3 importance. all three sub-dimensions. The degree of
content validity of the measurement scale
As noted in Part 1, this construct was derived may be assessed by evaluating how closely
from the disconfirmation paradigm and ex- the survey items represent the notion of
pectancy-value theories where Barsky (1992) customer satisfaction. The survey measures
argued that the disconfirmation paradigm is for ``expectations met'' (EM) and ``impor-
generally accepted as the construct that best tance'' (I) are also consistent with the find-
explains customer satisfaction. It was also ings from the literature reviewed. These
argued that the impact of attribute impor- terms, indicated separately on the survey
tance on consumer decision making is also questionnaire, validate the concept and
widely recognised. Salient elements from components of customer satisfaction. A
[ 271 ]
JaksÏa Kivela, John Reece and critical measure in this study is the result of Hong Kong Polytechnic University. The
Robert Inbakaran the comparison between customers indicat- English version was translated by a profes-
Consumer research in the ing their willingness to ``return'' and their sional translator into Cantonese, and the
restaurant environment. Part
2: Research design and satisfaction levels. This is based on the idea Cantonese version was then translated back
analytical methods which suggests that customers return to into English by another professional trans-
International Journal of restaurants because they are satisfied with lator. The researcher and both translators
Contemporary Hospitality agreed on the final Cantonese version, which
Management the restaurant ± nomological validity.
11/6 [1999] 269±286 Therefore, those customers that the model was adjusted for the Cantonese cultural
identified as ``satisfied'' were also willing to context while at the same time keeping the
``return'' (this will be highlighted by chi- conceptual meaning and the validity of the
square analysis and the logistic regression constructs intact. Both versions of the ques-
analysis in Part 3). It is argued, therefore, tionnaire were then pilot tested. As a result
that the measure of customer satisfaction of pilot testing the layout of the instrument
used in this study correlates in the theoreti- was modified from the four single (A4 size
cally predicted way with a measure of a paper) page format to one single page (A3 size
different ± nomologic, but related construct, paper) back-to-back. The language describing
for example the ``return'', a position sup- some of the items was also modified and the
ported by Tull and Hawkins' (1990) study. The sequence of some of the demographic ques-
content validity of the items (Kerlinger, 1986) tions was adjusted. The rest of the question-
was checked by a convenience panel of diners naire was not changed. The final version of
(n = 57), restaurant managers (n = 19), and the questionnaire is presented in the Appen-
hotel and catering educators (n = 16). The dix.
panel was asked to review the items and to
consider specifically: 2. Hypotheses
. with which of the five attribute sets (food,
service, convenience, ambience, and a Findings from previous research, the model
restaurant that offers . . . factors) each of dining satisfaction and return patronage,
item was associated; together with the objectives of the study
. the clarity of the items; and (presented in Part 1) provided the foundation
. the likelihood of the items being sensitive for the development of the following research
to the respondents. hypotheses for testing:
Ha1 The difference between pre-dining per-
As a result of panel analysis the question- ceptions (PrDp) and post-dining per-
naire was modified for clarity; the size and ceptions (PoDp) mean scores is
type of font and the sequence of questions statistically significant.
was adjusted, and the wording of some of the Ha2 There is a significant relationship be-
items was modified. Also, the number of tween the dining out frequency and
items for expectations, importance and ex- return patronage intentions.
pectations met was reduced from the original Ha3 There is a significant relationship be-
42 to 28 items shown in Table III. tween the dining out occasions and
The questionnaire was written in English return patronage intentions.
and was then translated into Cantonese using Ha4 There is a significant relationship be-
a double translation method. The research tween the customers who have dined in
instrument was translated by the Depart- the restaurant before and return
ment of Chinese and Bilingual Studies at the patronage.

Table III
Number of items for expectations, importance and expectations met
A restaurant that
Food Service Atmosphere Convenience (offers) . . .?
Presentation of food Friendly, polite, helpful staff Level of comfort in the Handling of telephone Offers a new dining
Menu item variety Attentive staff restaurant reservations experience
Nutritious food Staff greeting customers Level of noise in the Parking convenience Offers food of a consistent
Tastiness of food Efficient service restaurant standard
Freshness of food Staff are willing to serve View from the restaurant Feels comfortable to eat
Temperature of food Staff have food/beverage Cleanliness of the there
knowledge restaurant Offers service of a
Sympathetic handling of Dining privacy consistent standard
complaints Restaurant's temperature
Restaurant's appearance
Staff appearance

[ 272 ]
JaksÏa Kivela, John Reece and Ha5 There is a significant relationship be- dining satisfaction and return patronage
Robert Inbakaran tween the dining frequency and the model in Part 1.
Consumer research in the Dining satisfaction scores were derived by
restaurant environment. Part dining-out occasion.
2: Research design and Ha6 There is a significant relationship be- multiplying the ``expectations met'' scores
analytical methods tween the customers' age and return with the ``Importance'' scores. This provided
International Journal of patronage intentions. an indirect questioning technique for soli-
Contemporary Hospitality citing and measuring ``satisfaction'' percep-
Management Ha7 There is a significant relationship be-
11/6 [1999] 269±286 tween the customers' gender and return tions after the respondents had had their
patronage intentions. meal. Often, the direct questioning techni-
Ha8 There is a significant relationship be- ques encounter confounding (halo effect)
tween the customers' occupation and problems. The methodology for calculating
return patronage intentions. PrDp and PoDp is presented in Figure 1.
Ha9 There is a significant relationship be-
tween the customers' income and re-
turn patronage intentions. 4. Sampling design and survey
Ha10 There is a significant relationship be- methods
tween the customers' situational con- Because most convenience or fast-food op-
straints such as time and money and erations offer consumption on and off the
return patronage intentions. premises and predominantly trade on a low-
Ha11 Satisfaction with the dining experience cost-production high-volume-sales formula,
leads to return patronage. and because most of these operations are
Ha12 Customers' satisfaction with the quality either chained or franchised, this category
of food results in a greater probability was excluded from the study. Fast-food
of return than the satisfaction with operations are considerably different from
service. the more traditional restaurant categories in
Ha13 Customers' satisfaction with the quality terms of how they operate. Therefore, any
of food results in a greater probability comparisons between this category and
of return than the satisfaction with others would have been of little value. On the
ambience. other hand, fine dining operations accounted
Ha14 Occupation and first-timers/repeaters for a smaller percentage of the overall
are better predictors of return than the restaurant market in Hong Kong. It can be
occasional factors for dining out, and argued that Fine Dining or Gourmet (category
customers' income level. 1, see Table I) restaurants operate on a high-
Ha15 Customers' dining-out frequency has a cost-high-profit formula. They have a low
positive effect on return patronage in- share of the overall restaurant market in
tentions. Hong Kong and they generally target smaller
dining out groups who have greater disposa-
ble income and corporate customers. The
3. Research methods middle market was comprised of Theme/
Ambience restaurants and represented a
The research methods used in this study sizeable and the most volatile share of the
were based on gathering of primary data via restaurant market in Hong Kong. It is argued
a self-administered questionnaire. A review that customer loyalty, therefore, is critical to
of hospitality and consumer behaviour lit- these operations, which often have to survive
erature and secondary data was used in in a highly competitive environment. Taking
support. The purpose of the field survey these observations into consideration,
research was, first, to identify and address Theme/Ambience restaurants were selected
the deficiencies in the current restaurant for the survey in this study.
consumer behaviour literature; second, to One of the most salient issues in customer
identify and measure the strength and direc- satisfaction research is the nature of the
tion of the relationship between dining study setting used to model the construct.
satisfaction and return patronage; and, third, The debate in the literature has focused on
to collect a high-quality sample in order to two data generation methods:
test the dining satisfaction and return pa- 1 experimental designs; and
tronage model. The dining experience survey 2 field surveys.
developed for this study advances the find- Although experiments have been a popular
ings of previous customer satisfaction re- study design among many researchers
search and practices by focusing specifically (Churchill and Surprenant, 1982; Bolton and
on dining satisfaction (as a result of favour- Drew, 1991), other researchers have contin-
able dining experience) and return patronage ued to prefer field survey approaches (Lewis,
dimensions that were underpinned by the 1981; Auty, 1992; Almanza et al., 1994; Dube et
[ 273 ]
JaksÏa Kivela, John Reece and Figure 1
Robert Inbakaran Weighting matrix for calculating PrDp (E 6 I) and PoDp (EM 6 I)
Consumer research in the
restaurant environment. Part
2: Research design and
analytical methods
International Journal of
Contemporary Hospitality
Management
11/6 [1999] 269±286

al., 1994; Oh and Jeong, 1996; Qu, 1997; reactionary to extraneous variables that
Pettijohn et al., 1997). Although both experi- were difficult to control, as opposed to a
mental design and field survey methods have typical laboratory experimental design. This
respective strengths and limitations in study was based on field survey design and
producing generalisable study results, the its direction was to measure dining and post-
results of these studies using either of the dining experiences of restaurant customers.
methods have been generally accepted with-
out any critical considerations of the study Sample population
designs used. The study involved the administration of a
In general, the field survey approach has questionnaire (September-December 1997) to
been the dominant study design for customer a systematic selection of customers dining in
satisfaction research in hospitality applica- Theme/Ambience restaurants in Hong Kong.
tions. For example, Auty (1992) studied The sampling frame included subjects who
restaurant customers' expectation and satis- were residents of Hong Kong and who were
faction perceptions in various restaurants, 18 years of age or older. Completed ques-
while Almanza et al. (1994) collected data tionnaires with more than 15 per cent of
from a university staff and student cafeteria. items incomplete were excluded from the
Dube et al. (1994) examined dining satisfac- analysis. The surveys were conducted on-
tion and return patronage by eliciting re- site, the sampling target being all lunch and
sponses based on 16 dinner scenarios dinner customers. The selection of survey
(conjoint analysis) from a sample of diners in restaurants was based on a convenience
a fine dining restaurant. Oh and Jeong (1996) sample[2] and respondents were selected by a
collected data from fast-food customers in a systematic sampling method; every fourth
south-western city of the USA; and Qu's customer who walked into the
(1997) investigation collected data from a restaurant. The sample size for each restau-
number of Chinese restaurants in the state of rant was 25 per cent of the restaurant's total
Indiana. However, in all of these studies the seating capacity, as described in Table IV.
research instrument did not facilitate ``before According to the Survey of Wholesale,
and after'' measurement design. Experimen- Retail, and Import/Export Trade, Restau-
tal studies have not been widely used in rants and Hotels (1996-1997), by the end of
hospitality research. One of the major rea- July 1997, there were approximately 8,198
sons for this is that most hospitality opera- restaurants in Hong Kong, including 1,669
tions are multifunctional in nature, and the fast-food stores. Included in these figures
multifunctionality is more often than not were:
[ 274 ]
JaksÏa Kivela, John Reece and 1 street stalls, congee shops, noodle shops, published. A formal document to this effect
Robert Inbakaran and dai pai dongs (street restaurants); from the Hong Kong Polytechnic University
Consumer research in the 2 curry houses and small coffee shops; and was given to owners/managers noting that
restaurant environment. Part
2: Research design and 3 sandwich bars and teahouses. their respective commercial interests would
analytical methods be safeguarded. However, permission to
Excluded from these statistics were the 266
International Journal of identify the general location, style of cuisine
Contemporary Hospitality hotel restaurants. Items 1 to 3 above were
and seating capacity of the restaurants was
Management excluded from the survey, whilst the hotel
11/6 [1999] 269±286 granted by the various owners/managers.
restaurants were included in the survey.
This suggested that there were about 2,100 Survey arrangements
conventional restaurants in Hong Kong, of The surveys were conducted while the sub-
which approximately 900 were Theme/Ambi- jects were already in the restaurant, that is,
ence restaurants. (Hong Kong Hotels Asso- subjects had already made their decision to
ciation, 1995-1997; Hong Kong Restaurateurs dine at the restaurant. Every fourth diner in
Association, 1995-1997; Hong Kong Govern- these restaurants was the sampling target.
ment of the Special Administrative Region, The survey procedure is described below.
Department of Census and Statistics, 1995- . The researcher was advised by most
1997a, 1995-1997b, 1995-1997c.) Respective res- restaurant managers that is was a desir-
taurant managers and owners were formally able procedure to survey every fourth or
approached and asked if they wished to fifth customer. Their reason was that it
participate in this survey. A sample size of was better to be discreet so as to avoid
1,028 was set for the survey as noted in Table annoying other diners who were not
IV. Every fourth customer who walked into participating in the survey. This proce-
the restaurant was the target, until the dure was then adopted for all survey
sample size for that restaurant was reached. restaurants. It should be noted, however,
Similar to the pilot survey, approval to that this prolonged the survey schedule
conduct the survey was granted by the because not every fourth customer wanted
restaurant owners and managers on the to participate in the survey, in which case
strict undertaking that the names of the the next fourth customer was the sam-
restaurants would not be disclosed or pling target.

Table IV
Sampling frame
Total
Restaurant sites = 15 restaurant
Based on convenience samplea Total lunch sampleb Total dinner sampleb sample
A B C D
Restaurant, cuisine style Location Capacity % of capacity % of capacity
Shanghainese KLN 165 42 25 42 25 84
Mongolian KLN 160 40 25 40 25 80
(H) French KLN 160 40 25 40 25 80
Portuguese KLN 90 23 25 23 25 46
(H) Japanese KLN 80 20 25 20 25 40
Italian (Nouvelle) HKI 90 23 25 23 25 46
Middle Eastern HKI 75 19 25 19 25 38
Cantonese (Seafood) HKI 160 40 25 40 25 80
Contemporary (Californian) HKI 80 20 25 20 25 40
Cantonese (Dim Sum) HKI 215 54 25 54 25 108
Thai NT/I 130 33 25 33 25 66
Sichuan NT/I 190 48 25 48 25 96
American grill NT/I 125 32 25 32 25 64
Cantonese (Roast/BBQ goose/pork) NT/I 170 43 25 43 25 86
(H) Shanghainese NT/I 145 37 25 37 25 74
Total 2035 514 514 1028
(av. 136)
Notes: KLN = Tsim Sha Tsui/Tsim Sha Tsui East; HKI = Hong Kong Island; NT/I = Sha Tin
a
Excludes small coffee shops, sandwich bars, noodle shops, congee shops, fast-food restaurants, dai pai dongs, government/private/college
cafeterias; private clubs; fine dining and family/popular restaurants; H = Hotel restaurant; Total sample size [B and C] rounded off;
b
The sample population for the study was based on a systematic sampling procedure and every 4th customer entering the restaurant during the
survey was the target

[ 275 ]
JaksÏa Kivela, John Reece and . The survey sample for each restaurant 2 Does satisfaction with restaurant attri-
Robert Inbakaran was determined according to the seating butes increase the probability of return
Consumer research in the
restaurant environment. Part capacity of the restaurant, e.g. the sample patronage?
2: Research design and size for the restaurant was set at 25 per
analytical methods To operationalise the theoretical model, an-
cent of the seating capacity. Again, this
International Journal of swer to questions such as, What were your
was a constraint imposed by most restau-
Contemporary Hospitality expectations of each one of these items (res-
Management rant owners/managers, who argued that
taurant attributes) when you decided to dine
11/6 [1999] 269±286 this was a more than an appropriate
representation of their dining population. out in this restaurant?, and How important to
. The researcher, first to introduce himself you was each one of these items (restaurant
and second to seek their participation, attributes) when you decided to dine in this
approached all potential subjects. The restaurant?, give insights into customers'
researcher then explained the nature and initial reference point about restaurant at-
concept of the survey and what was tribute perceptions that they would use to
required from them. They were also make comparisons with the restaurant's
informed and instructed that the ques- actual performance after they had had their
tionnaire was in two sections, A and B. It meal. Also, answers to these two questions
was explained to them that section A had give insights into customers' pre-dining per-
to be completed before their meal and ceptions (Objective 2 in Part 1).
section B after their meal. At the end of The answer to the question Did the restau-
the survey the researcher then collected rant meet with your expectations for each of
the completed questionnaires. While some these items? (restaurant attributes), anchored
researchers suggest that there should be a to a scale of 1 = Has not met my expectations
sufficient time lag in measuring these and 5 = Has exceeded my expectations, pro-
constructs so as to minimise any con- vides an insight into the disconfirmatory
founding effects, the research to date effect between expectations and actual per-
supports the reliability of this sampling formance perceptions. The answer to the
method (Johnson et al., 1995). For this question Will you return to this restaurant?,
reason part B of the questionnaire was anchored to the scale 1 = I will definitely not
positioned on the back page. return and 2 = I will definitely return, reveals
. At the beginning of the survey process all the customers' post-purchase behavioural
subjects were informed that the survey intentions (Objective 3 in Part 1). More
was voluntary and without any obligation, importantly, however, it operationalises the
and that all responses and data would be most salient feature of the theoretical model ±
kept confidential by the researcher and how dining satisfaction influences return. As
would only be used for academic pur- noted in Figure 1, indirect weighted measure
poses. A consent form was presented to (expectations met scores 6 importance scores)
each subject seeking his or her consent, in was used to calculate customers' dining satis-
writing, to participate in the survey. faction scores (Objective 4 in Part 1).
However, only 16 subjects signed the
consent form. The vast majority of sub-
jects argued that the consent form was 5. Pilot study
unnecessary as they had already con-
sented to participate in the survey by the A self-administered, closed-ended question-
fact that they were willing to answer the naire was used to survey a systematic selec-
questionnaire. The researcher thought it tion of diners for piloting purposes. The
was more prudent to secure the necessary questionnaire comprised 174 variables di-
sample size rather than take issue on this vided into two major sections.
point with the respondents.
Section A. Pre-dining perceptions
Questionnaire 1 Rating on a five-point scale of 28 attri-
As suggested by Veal (1992), it is desirable to butes: Q1 ± ``expectations'', and Q2 ±
develop a research instrument that operatio- ``importance''.
nalises the research questions and research 2 Customer demographic profile; dining-out
objectives, and the theoretical model. To occasion; constraints.
recall from Part 1, the research questions in
this study were: Section B. Post-dining perceptions
1 Can a new conceptual model of dining 1 Dining-out frequency; whether or not they
satisfaction and return patronage based on had been to the restaurant before, and
disconfirmation theory be operationalised intention of returning to the restaurant.
that provides a more effective way of 2 Rating on a five-point scale of 28 attri-
predicting dining behaviour?; and butes: Q12 ± ``expectations met''.
[ 276 ]
JaksÏa Kivela, John Reece and A pilot survey was conducted in seven multi-item customer satisfaction measures
Robert Inbakaran Theme/Ambience restaurants (see Table V) in with alpha values of 0.75 to 0.96 and 0.90 to
Consumer research in the 0.95, respectively. In comparing these results
restaurant environment. Part Tsim Sha Tsui, Kowloon (KLN), Sha Tin,
2: Research design and New Territories (NT/I) and Hong Kong with those of single-item measures, it was
analytical methods Island (HKI) in order to provide valid and found that the multi-item measures out-
International Journal of reliable information on the questionnaire performed most of the single-item measures
Contemporary Hospitality in customer satisfaction instruments.
Management design and layout, wording and measure-
11/6 [1999] 269±286 ment scales. A reliability analysis (Cronbach's alpha)
was performed to test the reliability and
Reliability internal consistency of each of 28 attributes
Cronbach's (1951) alpha statistic is the most measured. The results showed that the alpha
frequently used indicator of instrument re- coefficients for all 28 attributes were high,
liability in survey research. A number of ranging from: Q1. 0.92; Q2. 0.84; to Q12. 0.95
studies have reported reliability indices of for the combined English and Cantonese
adopted scales; Westbrook and Oliver (1981) questionnaires. Alpha scores for the English
and Maddox (1985), for example, compared version of the questionnaire (n = 93) were:
the test-retest reliability of several single- Q1. 0.92; Q2. 0.85; and Q12. 0.95. Alpha scores
item scales such as the Delighted-Terrible (D- for the Cantonese version of the question-
T) content analytic and graphic scales. As a naire (n = 88) were: Q1. 0.95; Q2. 0.91; and Q12.
result they found the D-T scale to be the most 0.83. The reliability test for ``dining satisfac-
reliable among the examined scales, with an tion'' (S I 6 EM) was as follows: Q1. 0.92; Q2.
alpha value ranging from 0.65 to 0.85. How- 0.87; and Q12. 0.95. These were well above the
ever, most of the reliability estimates for minimum value of 0.5, which is considered
these repeated single-item scales were low to acceptable as an indication of reliability
(Hair et al., 1995).
moderate, suggesting that caution is needed
As a result of pilot testing, the layout of the
when using single-item measures.
instrument was modified from the four single
Reliability scores may also vary for multi-
(A4 size paper) page format, to one single
item instruments in consumer behaviour
page (A3 size paper) back-to-back format; the
research. Several researchers have studied
wording of some of the items was also
the reliability of selected multi-item scales:
changed and the order of some of the demo-
Oh and Jeong (1996) (alpha 0.90 to 0.66), and
graphic questions was rearranged.
Qu (1997) (alpha 0.70 to 0.85) for the Likert
scale. Westbrook and Oliver (1991) investi-
gated four different scales to assess their
reliability: Likert, semantic, differential,
6. Data analysis
graphic and verbal. The results of their study The Statistical Package for the Social
showed that the Likert and semantic-differ- Sciences, Windows Version 6.1. (SPSS 6.1)
ential scales performed equally well for was used to analyse the quantitative data

Table V
Sampling frame and sample selection methodology (PILOT)
Restaurant sites = 7 Total dinner
Based on a convenience selectiona Total lunch sampleb sampleb
Total pilot
Restaurant % of % of sample per
cuisine style Location Capacity capacity capacity restaurant
Szechuan KLN 75 19 25 19 25 38
Californian KLN 70 18 25 18 25 35
Thai NT/I 50 13 25 13 25 25
Cantones NT/I 75 19 25 19 25 38
Shanghainese H HKI 60 15 25 15 25 30
French HKI 60 15 25 15 25 30
Italian 55 14 25 14 25 28
Total 55 111 25 111 25 222
Notes: KLN = Tsim Sha Tsui/Tsim Sha Tsui East; HKI = Hong Kong Island; NT/I = Sha Tin
a
Excludes small coffee shops, sandwich bars, noodle shops, congee shops, fast-food restaurants, dai pai dongs,
government/private/college cafeterias; private clubs; fine dining and family/popular restaurants;
H = Hotel restaurant
b
The pilot study was based on a systematic sampling procedure and every 4th customer entering the restaurant
during the survey was the target

[ 277 ]
JaksÏa Kivela, John Reece and obtained from the research instrument. Six dining perceptions (PrDp) or their antici-
Robert Inbakaran types of procedures for the analysis of data pated level of dining satisfaction, and their
Consumer research in the were prepared in this study: reliability post-dining (PoDp) perceptions or their ac-
restaurant environment. Part
2: Research design and analysis; descriptive statistics; t-tests; cross- tual level of dining satisfaction.
analytical methods tabulation and chi-square analysis; factor A t-test procedure was computed for the
International Journal of analysis; and logistic regression analysis. pre-dining and post-dining sample, as ex-
Contemporary Hospitality Reliability analysis, according to Dillon et plained in sections 1 and 4 above, which
Management
11/6 [1999] 269±286 al. (1990, pp. 361-75), is when ``a measure is facilitated the testing of hypothesis Ha1, that
reliable if independent but comparable mea- the difference between PrDp and PoDp mean
sures of the same construct agree''. A reliable scores is statistically significant (at 0.05 and
instrument will, therefore, consistently yield 0.01 level).
similar reliability scores. However, reliabil- A cross-tabulation and chi-square analysis
ity depends on how much random error is was performed in order to test the
present. The sources of error are generally association/relationship of the joint fre-
transitory, that is, respondents' personal quency of two or more variables in the study.
characteristics influence the survey results The variables are usually divided into sub-
differently each time the survey is adminis- groups to reveal how a variable's value
tered, i.e. pre-dining and post-dining experi- changes from one sub-group to another,
ence perceptions in the model. In the context therefore allowing the study of each variable
of this study, it is important to note that data alone while holding the other variable con-
were reliable because a low sampling error stant. Unlike the one-way frequency analysis,
had greater reliability over samples, whereas cross-tabulation procedure also allows sta-
the greater the random error, the lower the tistical tests of significance that measure
reliability scores (Tabachnick and Fidell, whether a systematic relationship exists
1996; NorusÏis, 1994). Moreover, Lord and between two or more variables. Cross-tabu-
Novick (1968) suggested that the ``true score'' lation and the Pearson Chi-square statistic
represents the actual information, i.e. if the can be used to test the null hypotheses about
actual information (from the entire popula- a dichotomous variable. Furthermore, it is
tion) was obtained without error. The incon- often of interest to the researchers to test the
gruity between this score and the observed null hypothesis that the cross-tabulated
(survey information) attests to the level of variables are independent of each other.
error present, and deviation from this actual As with frequency tables, a visual presen-
score is the prime source of unreliability. tation of a cross-tabulation also simplifies the
Measurement error includes systematic and search for association, i.e. bar charts, to
random error sources. Only the random present the frequency or percentage for each
error is likely to vary with each survey. category of a variable. Although an exam-
Other measures, such as the internal consis- ination of the various row and column
tency, apply when there is a large number of percentages in a cross-tabulation is a useful
items in a survey. In this study, a Cronbach's first step in studying the relationship be-
alpha was performed to test the reliability tween the variables, row and column per-
and internal consistency of each 28 restau- centages do not allow for quantification or
rant attributes measured, for each attribute testing of that relationship. For this purpose,
category and for each factor (rotated factor it is useful to consider various indexes that
matrix) indicator (Tabachnick and Fidell, measure the extent of association, as well as
1996). statistical tests of the hypotheses that there is
A descriptive statistics procedure was re- or is not any association between the vari-
quired to develop profiles of the total sample. ables tested.
This typically includes frequencies, mean A statistic often used to test the null
scores and standard deviations. Frequencies hypotheses, i.e. that the row and column
were tabulated to identify the distributions of variables are independent, is the Pearson chi-
Hong Kong respondents' demographic vari- square. The calculated chi-square that is
ables, their occasional factors for dining out, computed is then compared to the critical
their dining-out frequency, whether or not points of the theoretical chi-square distribu-
the respondents were first-time diners or tion to produce an estimate of how likely (or
repeaters, and whether or not they would unlikely) this calculated value is if the two
return to the (surveyed) restaurant in the variables are in fact independent. The deci-
future. Means and standard deviations were sion to reject (or accept) the null hypothesis
computed to describe the respondents' per- is based on the probability, often known as
ceived expectations of restaurant attributes, observed significance level. If the probability
their perceived importance about the res- is small enough, usually less than 0.05 or 0.01,
taurant attributes, whether or not their then the hypothesis that the two variables
dining expectations had been met, their pre- are independent is rejected, i.e. there is a
[ 278 ]
JaksÏa Kivela, John Reece and relationship between variables. Therefore, A multivariate analysis (factor and regression
Robert Inbakaran the decision rule is: analysis) was used in this study to:
Consumer research in the create correlated variable composites
restaurant environment. Part
. calculated value > critical value = reject .

2: Research design and null hypothesis and conclude that there is from the original attributes ratings (i.e. to
analytical methods a relationship/association between the reduce original data to a manageable
International Journal of variables, or stage) and to obtain a relatively small
Contemporary Hospitality
Management . calculated value < critical value = accept number of variables that explain most of
11/6 [1999] 269±286 null hypothesis and conclude that the the variances among the attributes; and
variables are independent of each other
. to apply the derived factor scores in
(NorusÏis, 1994). subsequent regression analysis (i.e. to
predict return and non-return).
A chi-square test of significance was used in
this study to determine whether or not a Correlation, measures of sampling adequacy
(MSA), partial correlation among variables
relationship, other than chance, existed be-
and reliability alpha to ensure that the factor
tween the following variables investigated
analysis is appropriate to the data usually
(Objectives 5 to 7 in Part 1):
assess the appropriateness of factor analysis.
. ``expectations''; ``importance''; ``expecta-
The criteria for the number of factors to be
tions met''; ``pre-dining''; and ``post-din-
extracted is based on eigenvalue, percentage
ing'' perceptions and customers'
of variance, significance of factor loading,
characteristics;
and assessment of structure. Only the factors
. dining out frequency and return patron-
with eigenvalue equal to or greater than one
age intentions; are considered significant. In this study,
. customers' age and their return inten-
factors that were considered insignificant
tions; were disregarded. The solution that accounts
. customers' income and their return in- for at least 60 per cent of the total variance is
tentions; considered as a satisfactory solution. A
. customers' gender and their return inten- variable is considered of practical signifi-
tions; and cance and is included in a factor when its
. dining-out occasions and the return fre- factor loading is equal to or greater than 0.5
quencies. (Hair et al, 1995).
The following null hypotheses were tested: The data for the analysis in this study was
H01 There is no significant relationship 28 PoDp restaurant attribute items measured
between the dining out frequency and on a weighted matrix scale, as described in
return patronage intentions. Figure 1, ranging from ``very dissatisfied''
H02 The is no significant relationship be- (±7.5) to ``very satisfied'' (7.5). The principal
tween the dining out occasions and components and orthogonal (VARIMAX) ro-
tation methods were used in the factor
return patronage intentions.
analysis.
H03 There is no significant relationship
A logistic regression analysis was selected
between the customers who have dined
for this study because the logistic regression
in the restaurant before and return
analysis allows one to predict a discrete
patronage.
outcome such as a group membership from a
H04 There is no significant relationship
set of variables that may be continuous,
between the dining frequency and the
discrete, dichotomous or a mix. The discrete
dining out occasion.
outcome in logistic regression is either that
H05 There is no significant relationship
an event ``will happen'' or ``will not happen''
between the customers' age and return (i.e. return or non-return. It is difficult to deal
patronage intention. with a ``maybe return'' when actually trying
H06 There is no significant relationship to predict whether or not a customer will
between the customers' gender and return or not return) (Tabachnik and Fidell,
return patronage intentions. 1996). Stated in another way, for example, can
H07 There is no significant relationship return or non-return patronage be predicted
between the customers' occupation and as a consequence of dining satisfaction:
return patronage intentions. which resulted from the customer's overall
H08 There is no significant relationship favourable assessment of the dining experi-
between the customers' income and ence? The selection of logistic regression in
return patronage intentions. this study is based on a premise that `when
H010 There is no significant relationship the dependent variable has two dichotomous
between the customers' situational values, (i.e. will happen and will not happen)
constraints such as time and money and the assumptions necessary for hypothesis
return patronage intentions. testing in regression analysis are necessarily
[ 279 ]
JaksÏa Kivela, John Reece and violated. Therefore, the difficulty with mul- restaurant? outcome. The purpose of using
Robert Inbakaran tiple regression analysis [as opposed to logis- Will you return to this restaurant? as a
Consumer research in the tic regression analysis] for example, is that
restaurant environment. Part dependent variable was to identify the rela-
2: Research design and predicted values cannot be interpreted as tive importance of the dimensions derived
analytical methods probabilities. They are not constrained to fall from factor analysis in predicting respon-
International Journal of in the interval between 0 and 1, that is, will dents' return to the restaurant. Because the
Contemporary Hospitality
Management happen ± will not happen. (NorusÏis, 1994, p. 1). model produced by logistic regression is
11/6 [1999] 269±286 nonlinear, the equations used to describe the
Testing hypotheses about the coefficients outcomes (prediction) are more complex than
For large sample sizes such as in this study, those for multiple regression, i.e. it directly
the test that a coefficient is 0 can be based on estimates the probability of an event occur-
the Wald statistic that has a chi-square ring. The outcome variable (YÂ) is the prob-
distribution. For example, when a variable ability of having one outcome or another
has a single degree of freedom, the Wald based on a nonlinear function of the best
statistic is just the square of the ratio of the linear combination of predictors, i.e. with
coefficient to its standard error. For catego- two outcomes.
rical variables, the Wald statistic has degrees The regression model for Will you return to
of freedom equal to one less than the number this restaurant? was written as follows:
of categories. However, it should be noted
 ˆ 1
that the Wald statistic has an undesirable Y
1 ‡ eÿZ
property. When the absolute value of the
regression coefficient becomes large, the
Z ˆ A ‡ B0 ‡ B1 X1 ‡ B2 X2 ‡ B3 X3 ‡ B4 X4 ‡
estimated standard error is very large also.
This produces a small Wald statistic value, B5 X5 . . . Bp Xp ; where,
leading the researcher to fail to reject the
YÂ = Probability (``will return''/``will not
null hypothesis that the coefficient is 0, when
return'') ± dependent variable ± will you
in fact the null hypotheses should be rejected.
return to the restaurant?
Therefore, whenever there is a large coeffi-
A = Constant.
cient, the Wald statistic for hypothesis test-
Bi = Coefficients estimated from data.
ing should not be relied upon. Instead, the
Xi = Independent variables (factor dimen-
researcher should build a model (sequential
sions, dining-out occasions, dining-out
logistic regression) with and without that
frequency, occupation, age, income, and
variable and base the hypothesis test on the
customer characteristics).
change in the log likelihood (Hauck and
Z = Linear regression equation.
Donner, 1977).
Sequential logistic regression is similar to The probability of return patronage is esti-
sequential multiple regression and sequen- Â (will not return = 1 ± Y
mated as: Y Â (will
tial discriminant function analysis, in that return).
the researcher specifies the order of entry of
predictors into the model. For example, one
might start with a run predicting dining 7. Limitations of the study
satisfaction based on food and service
excellence. Then, in a second run, respon- The following limitations of the study should
dents' income and occupation are added to be addressed:
food and service excellence. The difference
. The sample was drawn from restaurants
between the two models is evaluated to in Hong Kong only and was represented
determine whether income and occupation by a systematic sample of 861 diners. The
significantly add to prediction above that findings were limited to the population
afforded by restaurant attributes. In stepwise studied in Theme/Ambience restaurants,
logistic regression, inclusion and removal of which limits generalisability to other
predictors from the equation are based solely restaurant categories, in particular the
on statistical criteria. Thus, a stepwise fast-food and convenience restaurant ca-
logistic regression is best seen as a screening tegory (Category 4 in Table I).
or hypothesis-generating technique, as it . The design of the study allowed the results
suffers from the same problems as statistical of the sample to be drawn exclusively
multiple regression and discriminant func- from respondents who dined in these
tion analysis. restaurants at the time only.
The purpose of logistic regression analysis . The findings of this study may not have
in this study was to explore how the PoDp sufficiently robust generalisability to
(dining satisfaction) dimensions derived Theme/Ambience restaurants in other
from the factor analysis related to the countries due to the study's different
dependent variable Will you return to this geographical and socio-economic setting.
[ 280 ]
JaksÏa Kivela, John Reece and However, the methodology described and Cadotte, E.R., Woodruff, R.B. and Jenkins, R.L.
Robert Inbakaran employed in this study was sufficiently (1987), ``Expectations and norms in models of
Consumer research in the vigorous and comprehensive to provide in- consumer satisfaction'', Journal of Marketing
restaurant environment. Part
2: Research design and sights and translatability about customers' Research, Vol. 30 No. 1, pp. 7-27.
analytical methods dining satisfaction perceptions and their Carman, J.M. (1990), ``Consumer perceptions of
International Journal of post-dining behavioural intentions as a re- service quality: an assessment of the
Contemporary Hospitality SERVQUAL dimensions'', Journal of
Management sult of their dining experience. It is suggested
Retailing, Vol. 66 No. 1, pp. 49-50.
11/6 [1999] 269±286 that the methodology and the theoretical
Churchill, G.A. and Surprenant, C. (1982), ``An
underpinning applied to support the model in
investigation into the determinants of custo-
this study can also be used for other restau-
mer satisfaction'', Journal of Marketing,
rant segments and in other geographical
Vol. 19 No. 4, pp. 491-504.
settings to measure customer satisfaction Cronbach, L.J. (1951), ``Coefficient alpha and the
and probability of return factors. internal structure of tests'', Psychometrika,
Vol. 16, pp. 297-334.
Notes Dillon, W.R., Madden, T.J. and Firtle, N.H. (1990),
1 The restaurant categorisation presented in ``Attitude scaling and concepts in reliability
Table I is an adaptation and elaboration of the and validity'', Marketing Research &
restaurant categories developed by Marketing Environment, 2nd ed., Irwin
Finkelstein (1989), and was used to guide the Homewood, IL, pp. 361-75.
sampling procedure in this study. Dube, L., Renaghan, L.M. and Miller, J.M. (1994),
2 Initially, the 15 survey sites were randomly ``Measuring customer satisfaction for strate-
selected from a listing of about 900 theme/ gic management'', The Cornell Hotel and
ambience restaurants. However, when the Restaurant Administration Quarterly, Vol. 35
respective restaurateurs were approached No. 1, pp. 39-47.
about conducting the survey in their restau- Finkelstein, J. (1989), Dining Out: A Sociology of
rants, nine refused the request. At a second Modern Manners, Polity Press, Cambridge,
attempt in random selection, 11 restaurants MA.
declined the request. This proved to be very Hair, J.F., Anderson, R. and Black, W.C. (1995),
time consuming which discouraged a third Multivariate Data Analysis with Readings,
attempt. The researcher then decided it was Prentice-Hall, Englewood Cliffs, NJ.
best to approach restaurateurs and restau- Hauck, W.W. and Donner, A. (1977), ``Wald's test
rant managers whom he knew to seek their as applied to hypotheses in logit analysis'',
permission to conduct the surveys in their Journal of the American Statistical
restaurants. A representative sample of Association, Vol. 72 No. 3, pp. 851-53.
theme/atmosphere restaurants was then se- Hong Kong Hotels Association (1995-1997), HKHA
lected for the surveys. Ethical issues were Annual Report, HKHA, Hong Kong.
addressed and the sampling procedure was Hong Kong Restaurateurs Association (1995-1997),
approved by the RMIT University and the HKRA Annual Report, HKRA, Hong Kong.
Hong Kong Polytechnic University's respec- Hong Kong Government of the Special Adminis-
tive ethics committee. trative Region, Department of Census and
Statistics (1996-1997a), Hong Kong Annual
References Digest of Statistics, Hong Kong SAR Govern-
Almanza, B.A., Jaffe, W. and Lin, L.C. (1994), ``Use ment Printer, Hong Kong.
of the service attribute matrix to measure Hong Kong Government of the Special Adminis-
consumer satisfaction'', Hospitality Research trative Region, Department of Census and
Journal, Vol. 17 No. 2, pp. 63-75. Statistics (1995-1997b), Hong Kong Monthly
Auty, S. (1992), ``Consumer choice and segmenta- Digest of Statistics, Hong Kong SAR Govern-
tion in the restaurant industry'', The Service ment Printer, Hong Kong.
Industries Journal, Vol. 12 No. 3, pp. 324-39. Hong Kong Government of the Special
Barsky, J.D. (1992), ``Consumer satisfaction in the Administrative Region, Department of Cen-
hotel industry: meaning and measurement'', sus and Statistics (1996-1997c), Survey of
Hospitality Research Journal, Vol. 16 No. 1, Wholesale, Retail, and Import/Export Trade,
pp. 50-73. Restaurants, and Hotels, Hong Kong SAR
Barsky, J.D. (1995), World-Class Customer Government Printer, Hong Kong.
Satisfaction, Irwin, Burr Ridge, IL. Johnson, M.D., Anderson, E.W. and Fornell, C.
Bolton, R.N. and Drew, J.H. (1991), ``A multistage (1995), ``Rational and adaptive performance
model of customers' assessments of service expectations in a customer satisfaction fra-
quality and value'', Journal of Consumer mework'', Journal of Consumer Research,
Research, Vol. 17 No. 2, pp. 375-84. Vol. 21 No. 2, pp. 695-706.
Boulding, W., Kalra, A., Staelin, R. and Zeithaml, Kerlinger, F.N. (1986), Foundations of Behavioral
V.A. (1993), ``A dynamic process model of Research, Holt, Reinhart and Winston, New
service quality: from expectations to beha- York, NY.
vioral intentions'', Journal of Marketing Lewis, R.C. (1981), ``Restaurant advertising:
Research, Vol. 30 No. 1, pp. 7-27. appeals and consumers intentions'', Journal

[ 281 ]
JaksÏa Kivela, John Reece and of Advertising Research, Vol. 21 No. 5, Tabachnick, B.G. and Fidell, L.S. (1996), Using
Robert Inbakaran pp. 60-74. Multivariate Statistics, 3rd ed., CSU, Harper-
Consumer research in the Likert, R. (1967), The Human Organization, Collins, Northridge, CA.
restaurant environment. Part
2: Research design and McGraw-Hill, New York, NY. Tse, D.K. and Wilton, P.C. (1988), ``Models of
analytical methods Lord, F.M. and Novick, M.R. (1968), Statistical consumer satisfaction formation: an exten-
International Journal of Theories of Mental Test Scores, Addison- sion'', Journal of Marketing Research, Vol. 25
Contemporary Hospitality Wesley, Reading, MA. No. 2, pp. 204-12.
Management Lowenberg, M.E., Todhunter, E.N., Wilson, E.D.,
11/6 [1999] 269±286 Tull, D.A. and Hawkins, D.I. (1990), ``Measure-
Savage, J.R. and Lubawski, J.L. (1979), Food ment techniques in marketing research'',
and People, Wiley, New York, NY. Marketing Research: Measurement and
Maddox, N. (1985), ``Measuring satisfaction with Method, 5th ed., Macmillan, New York, NY,
tourism'', Journal of Travel Research, Vol. 24 pp. 255-77.
No. 2, pp. 2-5. Veal, A.J. (1992), Research Methods for Leisure and
NorusÏis, M.J. (1994), SPSS Advanced Statistics 6.1,
Tourism: A Practical Guide, Longman,
SPSS Inc., Chicago, IL.
London.
Oh, H. and Jeong, M. (1996), ``Improving market-
Westbrook, R.A. and Oliver, R.L. (1991), ``Devel-
ers' predictive power of customer satisfaction
oping better measures of consumer satisfac-
on expectation-based target market levels'',
tion: some preliminary results'', in Monroe,
Hospitality Research Journal, Vol. 19 No. 4,
K.B. (Ed.), Advances in Consumer Research,
pp. 65-86.
Association for Consumer Research, Ann
Pettijohn, L.S., Pettijohn, C.E. and Luke, R.H.
(1997), ``An evaluation of fast food restaurant Arbor, MI, pp. 94-9.
satisfaction: determinants, competitive com-
parisons and impact on future patronage'', Further reading
Journal of Restaurant & Foodservice Market- Miller, J.A. (1977), ``Studying satisfaction, mod-
ing, Vol. 2 No. 3, pp. 3-20. ifying models, eliciting expectations, posing
This research project was
supported by the Hong Kong Qu, H. (1997), ``Determinant factors and choice problems, and making meaningful measure-
Polytechnic University and intention for Chinese restaurant dining: a ments'', in Hunt, H.K. (Ed.), Conceptualization
RMIT University in multivariate approach'', Journal of Restau- and Measurement of Consumer Satisfaction
Melbourne, Australia. The rant & Foodservice Marketing, Vol. 2 No. 2, and Dissatisfaction, Marketing Science Insti-
nature of support was non-
financial. pp. 35-49. tute, Cambridge, MA, pp. 72-92.

[ 282 ]
JaksÏa Kivela, John Reece and
Robert Inbakaran
Appendix
Consumer research in the
restaurant environment. Part
2: Research design and
analytical methods
International Journal of
Contemporary Hospitality
Management
11/6 [1999] 269±286

DINING EXPERIENCE SURVEY

[ 283 ]
JaksÏa Kivela, John Reece and
Robert Inbakaran
Consumer research in the
restaurant environment. Part
2: Research design and
analytical methods
International Journal of
Contemporary Hospitality
Management
11/6 [1999] 269±286

[ 284 ]
JaksÏa Kivela, John Reece and
Robert Inbakaran
Consumer research in the
restaurant environment. Part
2: Research design and
analytical methods
International Journal of
Contemporary Hospitality
Management
11/6 [1999] 269±286

[ 285 ]
JaksÏa Kivela, John Reece and
Robert Inbakaran
Consumer research in the
restaurant environment. Part
2: Research design and
analytical methods
International Journal of
Contemporary Hospitality
Management
11/6 [1999] 269±286

[ 286 ]

You might also like