Factors of consumers choices a reaveled preferences analysis for 3 in 1 coffee

Introduction

Researchproblem

Inrecentyears,coffeeisanimportantproductofindustrysector,agriculturesector,andservicesectori n Vi etnam.Ac c o rd i n g t o histori ca ldataofInternationalCoffeeOrganization(ICO),besideBrazil,Vi etnamisoneofthekeycountries incoffeeproductionandconsumptionintheworldwith19.4percentoftotalcoffeeproduction,and4. 5percentofdomesticconsumption.A n n u a l yieldofVietnamcoffeeproductionincreasesb y21timesintheperiod1990–

2014(ICO,2015).Thevolumeofcoffee exporti n Vietnamcontributesmorethan2 4 percentofGD Pin2012(VietnamMinistryofIndustryandTrade,2012).

According to ACNielsen's 2015 report, the 3-in-1 coffee market in Vietnam is dominated by five main manufacturers: Vinacafe, Nestlé, Trung Nguyen, Fes Vietnam, and Tran Quang, which collectively accounted for 88% of the total volume share in 2015, down from nearly 99% in 2014 The market share breakdown includes Vinacafe at 38%, Nestlé at 19%, Trung Nguyen at 14.6%, Fes Vietnam at 4.2%, and Tran Quang at 12.2% In terms of value, 3-in-1 coffee products represent 83% of the total instant coffee market The primary packaging types are bags, boxes, and sachets, which hold a 99.9% volume share This indicates a strong demand for 3-in-1 coffee compared to other instant coffee products, leading to intense competition among manufacturers for increased market share.

Duetothehighdemandofconsumers,especiallyyoungconsumers,manymanufacturershav ediversifiedtheir3in1coffeeproductsintermofbrands, prices,segments,packages,packs i z e , promotion,mainingredients.Forexample,fivemainmanuf acturersincludingVinacafe,Nestlé,TrungNguyen,Fes

Vietnam, led by Tran Quang, offers 27 varieties of 3-in-1 coffee, differing in key ingredients, packaging, and brands The competitive landscape among these manufacturers is evident in pricing and promotional strategies While price increases can boost profit margins, they may also lead to a decline in sales volume To counteract this, manufacturers can implement promotional activities to raise consumer awareness of their products and brand names, utilizing various marketing forms to engage potential customers.

Page1 as weightpromotion,additionalsachets,oragiftofrelatedproduct,forexample,spoon,plasticc u p , orgl asscup.Thus,twoimportantquestionsareraisedthat:

Researchobjective

According to Batsell and Louviere (1991), experimental methods, which combine econometric and psychometric analysis, are widely used to research consumer preferences These methods identify key factors influencing preferences, generate hypothetical profiles, collect consumer choices, and analyze choice data Data for these methods are gathered through two main survey techniques: revealed preference and stated preference methods, both of which are essential for understanding consumer behavior For instance, Durevall (2007) found that the long-term demand for coffee in Sweden is less affected by price changes due to consumer preferences and population structure Additionally, Wolfe et al (2011) highlighted that the interaction of product attributes also significantly influences consumer preferences, alongside price Thus, this study aims to achieve three research objectives.

(1) Identifyingthekey determinantsofconsumer’schoicesfor3in1coffee,

Several factors, alongside price, significantly influence consumer choices in the instant coffee market Identifying these factors enhances the understanding of consumer preferences, which is essential for suggesting market implications and developing strategies for manufacturers Additionally, analyzing the relationship between price and consumer utility is crucial for assessing how price changes affect consumer choice probabilities for each product This insight allows manufacturers to evaluate their competitive advantages and disadvantages in the 3-in-1 coffee market, ultimately aiming for higher profits compared to their competitors.

Page2 considerstheimpactofdiscount,promotion,andpricechangeonconsumer’schoices.Since3 i n 1 coffeemarketisoligopolyinVietnam,anychangeofcoffeeattributesofonebrandwillhavesign ificantimpactoni t s q u a n t i t y sol d T h e r e f o r e , manufacturerscouldoptimizet h e i r marketin gactivitiesforcapturingmoremarketshares.

Scopeofstudy

Thisstudyisapracticalresearchwhichreliesonthebasisofrandomutilitytheory.Thedatacollectio nofthisstudyisconductedinsuper- marketsinHoChiMinhCityin2016byapplyingr e v e a l e d preferencemethodwiththeadditionofs everalhypotheticalchoicescenarios.Duetot h e limitation offinanceand timespan,asmall sampleof197respondentswhoare3in1coffeeconsumersisc o l l e c t e d E a c h respondentisassume dtof a c e all1 9 surveyedalternativesi n a c t u a l choicescenario,sothedatasetofthisstudyistrea tedaspaneldataset.

Thecontributionofthisstudyistoinvestigatetheassociationbetweenconsumer’schoicesa n d c offeeattributessuchasprice,mainingredientsofcoffee,packaging,andmanufacturersb y usingt herevealedpreferencemethod.Fromthat,producerscouldunderstandmoreaboutt h e s i g n i f i c a n c e ofseveralattributes,whichmayh a v e heavycontributiontoc o n s u m e r ’sc h o i c e s Moreover,therelationshipbetweenpricesandconsumer’sutilitycouldhelpproducerst o e v a l u a t e t h e t r u s t ofconsumerst o t h e i r b r a n d s whenmarketc h a n g e s i n t e r m ofp r i c e s , p r o m o t i o n s , anddiscounts.

Thesisstructure

The remainingofthisstudyincludesfourchapters.Chapter2presentsliteraturereview,whichc o m p r i s e s theoreticalreviewandempiricalreview.Chapter3presentsresearchmethodology,whichd e s c r i b e s thequestionnaired e s i g n , s u r v e y p r o c e s s , a n d empiricalmodel.C h a p t e r 4 prese ntsthedatadescription,regressionr e s u l t , a n d discussion.C h a p t e r 5 summarizestheconclusio n,theimplication,thelimitation,andthefurtherresearchdirectionofthisstudy.

Literaturereview

Randomutilitytheory

Probabilisticchoicetheoriesweretheimportantpartsinpsychology.Theyweredevelopedtoe x p l a i n theinconsistencyandnon- transitionofindividuals’preferencesinexperimentalobservations(Luce& S u p p e s , 1 9 6 5 ) T hei n c o n s i s t e n c y a n d non- transitionofindividuals’preferencescouldbereflectedthroughchoicesituationswhenindividualsdo notchoosesamealternativesindifferentchoicesituationsordifferentsetsofalternatives.

In 1927, Thurstone introduced the "law of comparative judgment," which focuses on comparing physical stimulus intensities and qualitative judgments, such as evaluating the excellence of specimens on an educational scale This evaluation process, known as the discriminal process, highlights how individuals respond differently to various stimuli based on their specific demands Even when judged by the same individual, the discriminal processes for different stimuli vary The difference between two alternatives is quantified by a scale termed discriminal difference For specific specimens, these discriminal processes are distributed according to a standard deviation called discriminal dispersion Thurstone noted that each specimen chosen by individuals is characterized by two components: a scale value and a discriminal dispersion, both of which can be determined Furthermore, the scale value consists of an origin with a specific unit of measurement and an unknown correlation between the discriminal deviations of two different stimuli, which Thurstone assumed to be constant across the entire series of stimuli.

The “law of comparative judgment” serves as a foundation for understanding individual choice determinants, leading to the development of constant utility and random utility approaches by Luce and Suppes in 1965 The constant utility approach, introduced by Luce in 1959, assumes fixed utilities for alternatives, indicating that individuals may not always select the option with the highest utility In contrast, the random utility approach, first proposed by Marschak in 1960, incorporates an economic perspective based on Thurstone’s 1927 principle, positing that individuals choose the alternative with the highest utility, which is modeled as a function of attributes combined with a random component.

AlthoughMarschak(1960)interpretedThurstone’s“lawofcomparativejudgment”intheecon omicsfield,McFadden(1974a)introducedthegeneralprocedureinordertoapplyrandomu t i l i t y t h e o r y f o r analyzingqualitativechoicebehavior.Fromt h a t , t h i s t h e o r y i s widelyreviewedi n t h e r e s e a r c h ofD a n g a n z o (1979),H e n s h e r a n d Button( 2 0 0 0 ) , T r a i n (2009).AccordingtoMcFadden(1974a),conditionallogitanalysisisa ppropriateeconomicanalysismethodforconsumer’schoicesbehavior.Theresearchofconsumer

Consumers evaluate observable attributes of alternatives when making decisions, with each option providing a stimulus that economists refer to as utility It is assumed that respondents act rationally, selecting the alternative that offers the highest perceived utility Furthermore, perceived utility is influenced by two main components: a deterministic component and a stochastic component.

Thegeneralhypothesisoftherandomutilitytheoryisthatindividualsarerationaldecision- makers,andtheytrytomaximize theirutilitywhenfacingachoicebetweenmultiple(mutuallyexclusive)alternatives.Inotherword,they comparetheutilitiesofalternatives,whicht h e y faceandchoosethealternativewiththehighestutil ity.Luce(1959)introducedanimportantaxiomofrandomutilityapproach,whichexpressedthat thepresenceorabsenceofadditionalalternatived i d noti n f l u e n c e therelativeoddsofchosenalte rnativeovert h e secondone.M c F a d d e n (1974a)formalizedthisaxiomintothreebelowassu mptions:

(1) Independenceofirrelevantalternatives(IIA):therelativeratioofchoiceprobabilityofonealt ernativeoverc h o i c e probabilityofa n o t h e r alternativeisaffectedi n e q u a l l y proporti onbythepresenceoftheotheralternatives.

(2) Positivity:c h o i c e probabilitiesofa l l alternativesi n a l l possiblealternatives e t s areposi tive.

AccordingtoMcFadden(1974a),theperceivedutilityUcomprisestwocomponents:thesyste maticutilityVandtheerrorterm .Thesystematicutilityrepresentstheutility,whichisperceive dbydecisionmakersinthesamepurchasingcontext.Alternativesandattributesareknowna s important componentst o describedifferentp u r c h a s i n g contexts.Theerrort e r m representstheunknownd eviationofu t i l i t y p e r c e i v e d b y d e c i s i o n makersf r o m t h e utility.Specifically,theerrorte rmcapturestheeffectsofallunobservablefactors.T h e relationshipbetweentheperceivedutilit yU ,thesystematicutilityV ,andtheerrort e r m  isexpressedbythebelowequation:

Researchers are unable to directly observe the utility (U) of decision-makers; however, they can analyze the characteristics of these decision-makers and the attributes of the alternatives they face This analysis allows them to estimate choice probabilities According to McFadden (1981), choice probabilities must meet two criteria: they should be non-negative and sum to one, and they must depend on both the observable attributes of the alternatives and the characteristics of the decision-makers Additionally, researchers lack knowledge of the error term in their estimations.

  whichcapturesunobservablefactorsthataffectutilityofdecisionmakersandarenoti n c l u d e d inthesystematicu t i l i t yV Thus,researcherst r e a t t h e errort e r masr a n d o m c omponent.The distributionoftheerrortermmainlyd e p e n d s onresearcher’sconsiderationofthe deterministiccomponentV.

Supposethat,decisionmakerstrytomaximizetheirutility,andtheirutilityisdescribedbyt h e uti lityfunction:

Page6 wheres aremeasuredattributes,xischosenalternativefromthealternativesetbydecisionmakers. Moreover,utilitycomponentVi sthefunctionofmeasuredattributesS :

(1974a)suggestedtwolemmasinwhichthevalueoftheerrortermisindependentlyidenticallyd i s t r i b u t e d withWeibull(Gnedenko,extremevalue)distributioninthefirstlemma,andwithGumbel distribution(ExtremeValueTypeI)inthesecondlemma.Undertheconditionofthes e c o n d lem ma,McFadden(1974a)provedthatthechoiceprobabilityofonealternativeequalst h e proportionof exponentialfunctionofutilityofthisalternativeovertheexponentialfunctionofut il it yofre ma ininga lternativesi n thealternativese t.McFadden’s(1974a)fi ndingaboutc h o i c e probability isexpressedby thefollowingequation:

(2.4) whereBisalternativeset;x,ya r ealternativesofalternativesetB Then,therelativeoddsofchoicesis expressedbythefollowingequation: logP i   V i

P j V j where P i and P j are probabilitiesofchoosingalternativeia n dj; V i and V j alternativei a n dj. areutilitiesof

Basedone q u a t i o n (2.3),i t i s notedthatu t i l i t y componentVd e p e n d sonmeasuredattr ibutes,andtheunknownbetasofthatequationshouldbeestimatedinordertocalculatetheutilitiesa n d c h o i c e probabilitiesofalternatives.Thebetasareestimatedb y t h e maximumlikelihoodestimation ( M c F a d d e n , 1974a).M c F a d d e n (1974a)s u p p o s e d t h a t a l l measuredattributeswhi chareincluded inthedeterministicutilityV areindependentwiththeunobservedcomponento f u t i l i t y .Moreover,e a c h respondent’sc h o i c e isindependentwiththeothers.Basedonthoseass umptions,thechoiceprobabilityofalternativei forr e s p o n d e n t n is: y n i y n

 (P ni ) , (2.6) i where y ni  1ifrespondent n choosesalternativeia n d y ni  0ifrespondent n d oesnot choosealternativei ,and y ni  0forallother alternatives.Then,choiceprobabilityofeach respondentinthesampleofN respondentsis:

L() N (P ni ) , (2.7) n1i whereisthevectorcomprisingallcoefficientsintheequation(2.3).Thus,log- likelihoodfunctionisexpressedasthefollowingequation:

 P ni (x ni  x ni x nj ) n1i distribution functions of unobserved factors, which are expressed by the function

() isalwaysnegativewitheveryvaluesof.Itmeans d 2  thatlog- likelihoodisgloballyconcave.Thus,thereisexistenceofacriticalpointofthatmaximizeslog- likelihoodvalueandthatcriticalpointofsatisfiestheequation: dLL() d 0 (2.11)

Inaddition,thepredicted probabilitiesthatarecalculatedfrom are closettotheobservedchoices(McFadden,1974a).

McFadden(1974a)alsopointedoutt he advantagesanddisadvantagesofthisestimationmethod ofchoiceprobabilitiesbasedontheformula(2.4).Intermofadvantages,thisestimationcouldinterp retthechoiceprobabilitiesintermoftherelativesystematicutilityofalternatives.Moreover,t h i s e stimationcouldestimatet h e effectofp r e s e n c e ofn e w alternatives.Specifically,thechoicepro babilityofoldalternativewillproportionallyequallydecreasebyt h e choice probabilityofnewalternative.Inaddition,t h i s estimationc o u l d estimatethechoicealternativeeff ectswithoutreplication,andpredictthechoicebehaviorfromextrapolationofobservedalternativesets

.Ontheotherhand,McFadden(1974a)pointedoutt h a t thelimitationsofthisestimationr e l a t e d t o t h e independenceofirrelevantalternativesaxiom.B a s e d ont h a t axiom,t h e alternativesetsm a y i n c l u d e alternatives,whicha r e closesubstitutes.

Randomutilitymodelforms

AccordingtoTrain(2009),thevariationofrandomutilitymodelformsisderivedunderthediffer ent f().Train(2009)supposedthatasampleofrespondentsfacethesameobservedutilityV,however, thevaluesofunobservedfactorsaredifferentamongrespondents.Thus,thefunctionf()reflectsthedi stributionofunobservedutilityamongrespondentswhofacethe sameobservedutilitywithinasample.

Inthissection,severalpopularformsofrandomutilitymodel,whicharewidelyusedinmar ketresearch,aredescribed.Theyincludelogitmodel,GeneralizedExtremeValue(GEV)

 model,probitmodel,a n d mixedlogitmodel.W i t h e a c h f o r m ofr a n d o m u t i l i t y model,t h e situ ationinwhichtheformisappliedwillbediscussed.

First,logitmodelisconsideredasthemostpopularandwidelyusedmodel.Basedontheassumption ofindependenceirrelevantalternatives(IIA),Luce(1959)derivedtheoriginalformulaofl o g i t model.T h e n , thatformulawasprovedt o b e c o n s i s t e n t witht h e utilitymaximizationbyMarsc hak(1960).Finally,McFadden(1974a)showedacompleteeconomicanalysisoflogitmodelbyprovin gthatlogitmodel forchoiceprobabilitiesworksundertheassumptionofextremevaluedistributionofunobservedfac tors.Thatdistributionisusuallyc a l l e d asGumbeldistributionorExtremeV a l u e TypeI.B a s e d onthatassumption,t h e distributionofeachunobservedfactorforwhenadecisionmakerface sJalternativesise x p r e s s e d bythefollowingequation:

If logit modelisappliedfortwo alternativesja n di ,and nja n d  nia r e identicallydistributed extremevalue,thenthedifferencebetweenthem *

)followslogistic distribution: nj i nji nj ni

Equation(2.14)isappliedforbinarylogitmodelincaseoftwoalternatives.Theextensionoflogitmod elwhenrespondentsfacemanyalternativeswaswidelyknownasmultinomiallogitmodelorcondition allogitmodel.

Second,G e n e r a l i z e d ExtremeV a l u e models,whicha r e knowna s G E V m o de l s, a r e theg eneralizationofstandardlogitmodel.GEVmodelscomprisemathematicalformulationthatdescribe sdifferentcharacteristicfunctions.ThekeycharacteristicofGEVmodelsisthatthedistributionofu n o b s e r v e d u t i l i t y ofa l l alternativesfollowsageneralizedextremevalue.Moreover,thatdi stributionallowsthecorrelationsamongalternatives.Thedisappearanceof allcorrelationsamongalternativeswilltransformGEVmodelstostandardlogitmodel.The

GEV familyincludesseveralmodelssuchasnested(ortwo-level)logitmodel,andthree- leveln e s t e d logitmodel.

The probit model necessitates a normal distribution of all unobserved utility components, a requirement that may not be suitable in many cases, particularly regarding the estimated coefficients of price variables, which should encompass both positive and negative values However, the probit model addresses three significant limitations of the logit model Firstly, it accounts for random taste variation, allowing estimated coefficients to vary among respondents rather than being fixed, as seen in logit regression results Secondly, the probit model does not assume independence from irrelevant alternatives, enabling researchers to identify appropriate substitution patterns with specific data, leading to more accurate parameter estimation and interpretation Lastly, the probit model is applicable to panel data, where each respondent can make choices among different alternatives over various time periods or choice scenarios.

Fourth,mixedl o g i t model,whichwasi n t r o d u c e d b y Mc Fa dd e n a n d Train(2000),i s a m odelwithhighlevelofflexibility.Inmixedlogitmodel,choiceprobabilityofalternativesise x p r e s s e d asthefollowingequation: e  x ni

P ni   x f( )d , j where f( )isadensityfunctionof.Inthespecialcase,if f()describesfixedestimated parameters,then f()1for

b,and f()0forb.Therefore,equation(2.15)will becomechoiceprobabilitiesforstandardlogitmodel: e bx ni

Randomutilitymodelforbeverageorfood

Schiffmana n d Kanuk( 2 0 0 0 ) d e f i n e d perceptionas“theprocessb y w h i c h a n individual o b s e r v e s , selects,organizesandreactstoenvironmentalstimuliinameaningfulway”.

Product and consumer characteristics significantly influence consumer preferences According to Issanchou (1996), these characteristics can be categorized into two groups: intrinsic and extrinsic Intrinsic characteristics pertain to the sensory attributes of products, including appearance, texture, taste, aftertaste, odor, and aroma In contrast, extrinsic characteristics relate to external factors such as personal attributes (age, gender, income, education) and situational factors (price, brand familiarity, environmental attributes, and product availability) Cardello (1996) suggested that consumer preferences can be explored through these two connections.

(1)i n t r i n s i c characteristicsandconsumer’spreferences,and(2)extrinsiccharacteristic sandconsumer’spre fe re nce s.

Consumers often compare various product attributes when faced with an abundance of options, leading to trade-off problems in their choices These trade-offs can be analyzed using conjoint analysis, a multivariate technique that assesses how purchasers navigate decision-making when confronted with hypothetical multi-attribute alternatives According to Louviere and Hensher (1983), the first choice of an alternative reflects the highest utility for consumers Understanding how consumers evaluate multi-attribute alternatives and make choices is a valuable approach for both basic and applied consumer research.

Due to the limited existing profiles of available products, hypothetical product profiles are created to explore the relationship between product attributes and consumer preferences These hypothetical profiles combine various attributes with defined ranges and levels, ensuring they are broad enough to capture potential consumer preferences while remaining narrow enough to maintain estimation efficiency For instance, a study by Mtimet and Albisu (2006) focused on four attributes of designation of origin (DO) wine: origin, price, wine aging, and grape variety Using methods proposed by Street, Burgess, and Louviere (2005), they developed 27 choice sets, each featuring four levels for each attribute, demonstrating the effective application of this approach.

(2006)suggestedprice,regionoforigin,brandname,andawardasimportant labelinformationt oconsumer’swinec h o i c e s 2 0 choicetaskswereg e n e r a t e d b y combiningvariousl e v e l s ofaboveattributes.

Besidehypotheticalp r o f i l e s ,“notbuy”c h o i c e isu su a l l y addedt o consumer’sc h o i c e tasks.H owever,inmanycases,hypotheticalprofilesbringouthigher utilityforconsumersthan“notbuy”choiceor“non-purchase”alternative(Mtimet&Albisu,2006).

Intermofmethodology,thebasicprocessbasedonrandomutilitytheoryanditassociateddiscretec hoicemodelssuchasstandardmultinomiallogit(MNL)model(McFadden,1974a),n e s t e d mu ltinomiallogit(NMNL)model(McFadden, 1978), andmultinomialprobit(MNP)model(Da ganzo,1979).Byapplyingrandomutilitytheory,MtimetandAlbisu(2006)foundoutthat:

(1)designationoforigin, wineaging(+),andgrapevarietyalsoplayanimportantrolei n consumer’schoices,and(2)pricehasopt imalpointatwhichrespondent’sutilityishighest.Inaddition,bycombiningrandomutilitytheoryandsim ulation,Lockshinetal. (2006)claimedt h a t thecontributionofprice,regionoforigin,brandname,andawardtoconsum er’schoicesi s quite complex.Moreover,priceisalsoimportantfactor,whichmatterspurchaseprobability.

Besideattributesofproducts,social- demographiccharacteristicsandfrequencyofproductconsumptionofconsumersa r e a l s o c o n s i d e r e d inmanyr e s e a r c h e s Lockshine t a l

(2006)notedthatfrequencyofwineconsumptionisdividedintofivelevels:almosteveryday,twoort h r e e timesperweek,onetimeperweek,twoorthreetimespermonth,andonetimeorlessp e r mont h.Basedonthisexperience,MtimetandAlbisu(2006)dividedDOwineconsumersi n t o twoseg mentsduetothefrequencyofconsumptionincludingfrequentconsumption(everyday,ortwoorthreeti mesperweek),andoccasionalconsumption.Withdifferentgroupsofconsumers,t h e i r perceivedu t i l i t y isd i f f e r e n t (Mtimet& A l b i s u , 2 0 0 6 ) Specifically,u n d e r determinedpricelevel,occa sionalconsumersenjoyhigherutilitythanfrequentconsumersdo.However,theopposite trendisobservedwhenpricelevelis higherthandeterminedpricelevel.

Theinvestigationofcoffee’sattribute

When researching consumer choices in coffee, it is crucial to identify the significant factors that influence these decisions Coffee attributes can be categorized into two main groups based on perceived value: emotional and functional attributes According to Issanchou (1996), coffee attributes are further divided into intrinsic and extrinsic attributes Intrinsic attributes pertain to the product's characteristics, such as price, promotion, and sensory qualities, while extrinsic attributes relate to external factors, including consumer characteristics and purchase contexts This section will primarily focus on the intrinsic attributes of coffee.

A study by Hanspal (2010) revealed that consumers prioritize coffee quality, taste, flavor, certification marks, health impact, and price, with quality, taste, and flavor being the most significant factors This finding aligns with Geel, Kinnear, and Kock (2005), who identified flavor, liking, and brand familiarity as key components influencing consumer ratings of instant coffee Additionally, Watanabe, Suzuki, and Kaiser (1998) noted that consumers focus more on the sensory experience of drinking coffee rather than its health benefits Furthermore, fluctuations in coffee prices have minimal impact on consumer choices, reinforcing the importance of flavor and brand familiarity in their purchasing decisions.

First,i t i s recognizedt h a t s e n s o r y c h a r a c t e r i s t i c s haves t r o n g e r associationwith consumer’schoicesthanotherfactors,especiallywiththecaseofcoffeeconsumers.Coffeeisa s p e c i a l foodp r o d u c t becausei t b r i n g s outmoreemotionalvaluet h a n functionalvaluet o

Consumers derive enjoyment from sensory characteristics of products, which can be categorized into five areas: appearance, odor, taste, sound, and mouthfeel (Seo, Lee, & Hwang, 2009; Lazim & Suriani, 2009) Specifically, regarding odor, Mayer and Grosch (2001) noted that roasted coffee should possess typical qualities such as caramel-like, roasty, sulfurous, and smoky aromas Additionally, Ross, Peck, and Weller (2006) found that aroma and bitterness in coffee are significantly influenced by storage conditions, particularly temperature, and that consumers often consider these factors when selecting different types of coffee.

Price and promotion significantly influence consumer choices; however, research by Srivastava (2007) indicates that price cuts and sales promotions may not enhance brand loyalty While discounts can attract new consumers trying coffee for the first time, their long-term effectiveness is limited Manufacturers often struggle to maintain discounts due to financial constraints, and over time, consumers reassess a brand's value, leading to a stronger influence of brand loyalty on their choices This loyalty is shaped by past consumption behaviors, which directly affect current purchasing decisions (Faria, 2003).

Third,besideabovef a c t o rs, brandisal so recognizedasim port antdeterminantofconsumer

Consumer choices regarding coffee are significantly influenced by brand reputation, as factors such as basic ingredients, pricing, and perceived benefits are often similar across brands (Srivastava, 2007) The emergence of new brands, a decrease in consumer boredom thresholds, improved quality standards, and slow innovation among existing brands contribute to the decline of sole brand loyalty over time Consequently, new brands often serve as substitutes for established ones in the market Nowadays, consumers typically evaluate multiple brands before making a decision, leading to a shift in their preferred choice sets Thus, identifying these choice sets is a crucial aspect of consumer research.

Consumer’ssocial-demographiccharacteristics

Research by Hanspal (2010) highlights the connection between social-demographic characteristics and coffee preferences, identifying key factors such as education level, occupation, age, gender, and income Notably, individuals with higher education tend to explore a variety of coffee brands, while those with professional degrees often exhibit brand loyalty Additionally, Watanabe, Suzuki, and Kaiser (1998) suggest that married individuals, those with higher education, small families, or more leisure time are more inclined to consume coffee Furthermore, coffee consumption is notably higher among private sector employees, students, and experts, with age also influencing preferences, particularly among younger respondents.

36yearshavemoredemandoncoffee.However,thatdemandreduceswiththei n c r e a s e ofage. Ontheotherhand,intermofgender,empiricalresultshowedthatmentendt o consumemorecoffe ethanwomendo.

The frequency of coffee consumption significantly influences consumer choices, as individuals often select different pack sizes based on how often they drink coffee Research by Geel, Kinnear, and Kock (2005) utilized data from trained panelists and consumers, revealing that many consumers drink instant coffee at least once a day, with a majority being over 25 years old Through cluster analysis and internal preference mapping, the study identified four distinct consumer groups: "pure coffee lovers," "instant coffee blend lovers," "not serious coffee drinkers," and "general coffee drinkers." Additionally, the authors detailed the preferences associated with each consumer group.

Fourgroupsofcoffeeconsumersinclude“purecoffeelovers”,“coffeeblendlovers”,“notseriousco ffeedrinkers”,and“generalcoffeedrinkers”(Geel,Kinnear,&Kock,2005).“Purec o f f e e lo vers”areusuallyolderconsumersandtheircoffeepreferencesarestable.Inparticular,t h e y p a y moreattentiononcoffeeattributessuchasappearance,aroma,bitterness,andalittleofmouth - feel.Inaddition,t h e y preferpurecoffeea nd a r e insensitivet o highpriceofpurec o f f e e B esides,“coffeeblendlovers”aredescribedaspeoplewhohavelimitedincomeand

Page16 highlevelofsensitivitytopricechanges.Theyusuallylikeanykindsofcoffee,whichhavel e s s intenseofcoffeeflavorandhighintenseofsweetness,forexample,instantcoffee.“Notseriouscoff eedrinkers”donotlikepurecoffeeareneutralbetweencoffeeblendsandotherhotdrinksuchast ea,chocolate.Moreover,theyareusuallyin15-

24yearsoldinwhichtheya r e i n bui l di ng processofc o f f e e preferencesandbrand/ manufacturerknowle dge, a nd specifyingtheirpreferredcoffeeattributes.

Researchmethodology

Revealedpreferencemethod

Experimentalanalysisofchoicebehaviorhasbeenwidelyinterestedfromthe1970sinordert o dev elopmodel,analysisprocedure, andexperimentaldesign.Twopopularmethodsusingforcoll ectingchoice dataarerevealedpreference andstatedpreferencemethods AccordingtoLouviereetal. (2000),revealedpreferencemethodbasedonthestrategythat“theworldasiti s ” , whilestatedpref erencemethodbasedonthestrategythat“theworldasitcouldbe”.Inthissection,r e v e a l e d preferenc emethodh a s beenreviewedi n t e r m o f advantagesa n d d i s a d v a n t a g e s Moreove r,s e v e r a l importantissuesi n t h e d a t a collectionp r o c e s s s h o u l d b e clarifiedsuchas: (1)howtodeterminethealternatives,(2)howtoidentifyconsumer’schoices e t ,

The primary advantage of the revealed preference method is its reliance on actual consumer behavior observed at existing prices, making it simpler to analyze than hypothetical decisions, which often introduce bias (McFadden, 1974) Economic entities aim to maximize their expected utility by comparing alternatives and focusing on the actual risks to their welfare (McFadden, 1974a) Additionally, as a nonparametric method, revealed preference can be effectively applied even with a small number of observations and does not depend on assumptions regarding functional forms (Louviere et al., 2000).

However,itisrecognizedt h a t revealedpreferencemethodalsosuffersseverald i s a d v a n t a g e s Revealedpreferencemethodismainlyappliedtomeasuremarketvalueorusevalue(Ada mowicze t a l , 1994).Moreover,b e c a u s e revealedpreferencemethodb a s e d one x i s t i n g productsandspecificconsumptionsituation,consumersonlyfaceasmallnumberofcombinationo fattributes(Adamowiczetal.,1997).Asaresult,revealedpreferencemethod cannotbea p p l i e d forprototypeproductsord e v e l o p i n g productsbecauseitisdifficultt o

Page 18 emphasizes the importance of stimulating actual purchasing scenarios to observe consumer choices directly (Adamowicz et al., 1994) Ben-Akiva et al (1994) argue that the revealed preference method fails to estimate demand for new products, which is better addressed by the stated preference method However, consumer purchasing behavior occurs in real-world settings without researcher control, making it susceptible to unexpected external influences For example, consumer choices are often heavily influenced by the respondents' market perceptions (Caldas & Black, 1997).

The dataset for the revealed preference method is derived from actual consumer purchasing behavior Previous studies indicate that variations in research subjects and respondents result in different methods for collecting choices For non-market value, choices are typically gathered through diaries that track alternatives chosen by respondents, while for market value, data is collected through direct observations of purchasing scenarios Since 3-in-1 coffee is a commercial product, participants in this study should be identified as consumers who purchase these products at retail locations However, it is important to note that the consumer may not always be the decision-maker, necessitating additional questions in the questionnaire to determine whether the respondent is indeed the decision-maker.

During the data collection process of the revealed preference method, several key issues arise Firstly, it's essential to define a specific purchasing context that remains consistent for all consumers, considering factors like location, temperature, and the number of available alternatives Ideally, variations in purchasing context among consumers should be minimal, with all transactions occurring in similar environments such as supermarkets or convenience stores Secondly, to ensure the accuracy of purchasing behavior, consumer actions are observed discreetly According to Adamowicz (1994), the revealed preference method involves individuals selecting one alternative from all available options, with the order and number of choices also recorded Lastly, a significant challenge lies in identifying the consumer's choice set, which is influenced by the researcher's arguments within a specific research context.

Page19 bethesubsetofallavailableproductsinthemarket,orallavailableproductsthattheyfaceatt h e purc hasingplace,orallavailableproductsinthemarket.Basedonthatidentification,thed a t a s e t wouldb e t r e a t e d a s paneldatai n whichdecisionmakersfaceallalternativestheirc h o i c e set,andeachalternativeisanobservation.

Attributesofcoffee

In the literature on market valuation, attributes are categorized into three main groups: monetary, non-monetary, and environmental attributes These attributes significantly influence consumer choice preferences For 3-in-1 coffee, monetary attributes include price and key ingredients like coffee content, sugar, and non-dairy creamer, which are crucial to consumer decision-making Non-monetary attributes focus on convenience, encompassing factors such as the number of packs, weight, and packaging ease, all of which are vital for instant coffee products Lastly, other characteristics like promotions, brand identity, and product appearance also play a significant role in shaping consumer choices.

(1993),keyattributesofalternativescouldbeidentifiedthroughf o c u s groupsprocessandpre- surveyprocess.Besideoursuggestedattributes,potentialattributesa r e a l s o notedasimportanto n e s Forexample,p o t e n t i a l attributesofpromotion,environmentprotectionlevel,andconvenienc elevelalsomatterconsumer’schoices.However,t h e numberofattributesshouldbeminimizedasm uchaspossibleinordertoavoidcomplexc h o i c e t a s k forc o n s u m e r s , a n d r e f l e c t t h e d ifferenceamongvariousb r a n d s AccordingtoM c F a d d e n (1986),t h e numberofattributes, whicha r e t h e outcomeofs c r e e n i n g process,s h o u l d belessthanten.

(1998),attributesofalternativesareselectedbyt h e r e s p o n s e s ofexpertsorwhohaveusedpro ductswithhighleveloffrequency.Theseattributesa r e modifiedbyusingfocusgroups.Focusgrou pswilldiscusstofinalizethelistofattributes.Moreover,levelsofattributesareidentifiedbyaskingrespo ndentsthehighestandlowestlevelsofattributes atwhichtheyagreetobuythatproduct.Thehighestandlowestlevels ofattributesm a y bedifferentfromcurrentlevelofexistingproducts.Therangeofattributesre flectboth willingnesstopayandwillingnesstoacceptforattributechange.Thefollowingdiscussionwillf o c u s moreontheidentificationofattributesandlevelsofthem.

Table 3.1 outlines the suggested attributes of 3-in-1 coffee, finalized through a preliminary survey conducted in two steps Initially, a small sample of approximately 30 coffee drinkers provided insights into the factors they consider when selecting a 3-in-1 coffee Subsequently, a screening process was employed to evaluate whether these attributes pertain to specific alternatives or consumer preferences It is important to note that the final attributes must be tailored to each brand or manufacturer, as well as to the different types of 3-in-1 coffee, indicating that attributes can vary significantly among brands and products.

Weightofalternative gram/alternativeNumberofsachetsperalternativesachet/alte rnativeP a c k a g i n g (paper,plastic)

Choiceset

AccordingtoHowardandSheth(1969),choicesetisasetofproducts,whichareconsideredforpurc hasing byconsumers.Thechoicesetofeachrespondent isusuallyidentifiedbyaskingr e s p o n d e n t “Beforechoosingoneormoreproduct(s),wha talternativesyouconsiders?”.Thenumberofavailableproductsinthemarketislarge,andtheconsider ationofallproductsbefore

FesVietnam 1.6 6.7 2.2 8.6 choosingiscomplicated.Choicesetisthesubsetofallavailableproductsinthemarketplace,a n d c h o i c e s e t ist h e wayi n whichconsumersr e d u c e t h e complexityo f c h o i c e process.Howev er,insomecases,respondentssaidthattheydonotcareaboutotheralternativesbecauset h e y justlike aspecificalternative.Theassumptionofthatcaseisthatconsumershavefinalizeda finalproductaftert heexperienceprocessforalongtime.

The size of the choice set for each respondent in market research is influenced by various factors, including the number of available products, the significance of the product category, consumer education level, brand loyalty, age, and income (Gruca, 1989; Howard & Sheth, 1969) For instance, Jarvis and Wilcox (1973) noted that the importance of a product category, such as instant coffee, could negatively impact the size of the choice set In Vietnam, however, consumer preferences for 3-in-1 coffee have developed over time due to extensive usage and experimentation with different brands The popularity and affordability of 3-in-1 coffee make it accessible to a wide audience, leading to the assumption that each respondent's choice set includes all 19 alternatives available in the market.

3in1coffeeproductsareconsideredinthisstudybelongtofourmainmanufacturersinHoC h i M i n h City.T a b l e 3 2 presentsthevolumes h a r e a n d valueoft h o s e fourmainmanufacturers, whichincludeVinacafe,Nestlé,TrungNguyen,andFesVietnam.Itisrealizedt h a t theorderinter mofvolumeshareandvalueshareofthosemanufacturersarethesame.Vinacafeleadsi n bothvol umes h a r e a n d values h a r e i n 2 0 1 4 a n d 2015.Theremainingmanufacturersa r e followed b y N e s t l é , T r u n g Nguyen,a n d FesVietnam.F r o m t h a t , i t isclaimedthatthedifferencea mongproductsintheaspectofpriceissmall.Thatsupportstheremarkthat3in1coffeemarketismor ecompetitiveandistheoligopolymarket.

Table3.3presentsthelistof19kindsof3in1coffee.Theyarenotallkindsofavailable3 i n 1 co ffeein themarketi n HoChiMinhCity.However,accordingt othestatisticsofAC

Nielsen(2015),thevolumeshareofthese3in1coffees isover0.1percent.Allbrandsthathavevolumes h a r e , whichi s l e s s than0 1 % , a r e e x c l u d e d f r o m t h i s l i s t Iti s recognizedthatmanufacturershavediversifiedtheirproductsi n t e r m ofb r a n d s , packaging,a n d weightofs a c h e t T h u s , consumersmayfa c e a productwithvariou sp a c k a g i n g Moreover,T a b l e A 1 presentsasummaryofall19alternativesandtheirattributessu chmainingredients,packaging,a n d manufacturer.

No.Manufacturer Brand Product Packaging Weightofsachet

1 Vinacafe VinacafeGoldoriginal Box–20sachets 20grams

Chất Chất–Hanoi Box–15sachets 21grams

4 Chất–Saigon Box–10sachets 29grams

5 Wake-up Wake-up–SaiGoncafé Bag–24sachets 19grams

6 Wake-up–SaiGoncafé Bag–40sachets 19grams

8 Nescafered Nescafered–Richaroma Box–15sachets

9 Nescafered–Richaroma Bag–46sachets 17gramsNestlé

Nescafegreen–Richtaste Box–15sachets 17grams Nescafegreen–Richtaste Box–20sachets 17grams

12 Nescafegreen–Richtaste Bag–46sachets 17grams

18 FesVietnam Maccoffee CaféPho Box–10sachets 22grams

Questionnaire

The study's questionnaire, detailed in the Appendix, is structured into three main sections The first section captures the respondent's actual purchasing scenario, including the alternatives chosen, the decision maker's identity, and their relationship with the respondent The second section gathers social-demographic data and coffee usage information from the respondents, which encompasses details such as name, gender, age, religion, marital status, residence, education level, occupation, and income, alongside their coffee consumption habits.

The questionnaire begins by assessing the frequency of coffee consumption and the number of sachets used per occasion It then identifies potential substitute alternatives by asking respondents, "If your actual alternative is not available at the supermarket, which alternative would you choose?" Respondents are presented with five to seven hypothetical scenarios, where they evaluate their choices among two or more alternatives, including their actual choice and potential substitutes, while considering factors such as discounts, weight promotions, and price increases.

Surveyprocess

This study utilizes revealed preference methods to gather data by observing consumers' actual purchasing behaviors and the information provided on product labels in the market Surveys are conducted outside supermarkets after the purchasing context, collecting data categorized into three groups The first group includes consumer choices and prices of all available alternatives in the supermarket, noting that if consumers select multiple options, they are counted as multiple responses This group captures both actual and hypothetical choices The second group consists of social-demographic information from respondents, including name, gender, age, religion, marital status, education level, residence, occupation, and income The third group focuses on the usage of 3-in-1 coffee, detailing the frequency of consumption and the number of sachets used per serving.

In Ho Chi Minh City, the primary places to purchase 3-in-1 coffee include supermarkets, traditional markets, and grocery or convenience stores Each of these venues offers different purchasing contexts and pricing According to Table A.2, there is a fluctuation in 3-in-1 coffee prices across various supermarkets Supermarkets are deemed the most suitable option for this study for several reasons They offer a wider variety of 3-in-1 coffee brands compared to traditional markets and grocery stores, allowing consumers to explore multiple options in one location, which can influence their purchasing decisions Additionally, supermarkets provide better preservation conditions for products, ensuring stability and uniformity in product quality, unlike grocery or convenience stores.

Identifying the decision-maker is crucial in this survey, which focuses on individuals who have purchased one or more types of 3-in-1 coffee from a list of 19 alternatives Respondents were selected based on their purchases observed at the checkout counter The survey includes a question to determine if the buyer is also the decision-maker If the buyer is not the decision-maker, demographic and coffee usage information about the decision-maker will be collected through the buyer It is assumed that the buyer can represent the decision-maker in hypothetical scenarios, as they typically have a close relationship and understand the decision-maker's preferences Additionally, the rationale of buyers is consistent with that of other buyers who are decision-makers, suggesting that buyers who are not decision-makers can still effectively represent the decision-maker in this study.

Modelspecification

(3.1) where:U isutility; 1,2,3, ,k arecoefficients;X 1,2,3, ,ka r e attributesofalternativesincluding priceofalternative,bitterness,sweetness,fat,weightofalternative,numberofsachets,paperp a c k a g i n g , Vinacafe,TrungNguyen,Nestlé,squaredprice,andsquaredbitterness;

Typeofvariable Variable Denotation Unit Description

Dependent variable Choice choice 1=yes,0=no

Fat fat %w/ wW e i g h t ofalternative weight_alt gram Independent variable

Paperpackaging pap_pack 1=yes,0=no

Plasticpackaging pla_pack 1=yes,0=no

2=5-