+a.1996.the american customer satisfaction index

8/12/2019 +A.1996.The American Customer Satisfaction Index

1/13

Claes Fornell, Michael D. Johnson, Eugene W. Anderson,Jaesung Cha, Barbara Everitt Bryant

The American Customer SatisfactionIndex: Nature Purpose and FindingsThe American Customer Satisfaction Index (ACSI) is a new type of market-based performance measure for firms,industries, economic s ectors, and national econom ies. The a uthors discuss the nature and purpos e of ACSI andexplain the theory underlying the ACSI model, the nation-wide survey methodology used to collect the data, andthe econometric approach employed to estimate the indices. They also illustrate the use of ACSI in conductingbenchmarking studies, both cross-sectionally and over time. The authors find customer satisfaction to be greaterfor goods than for services and, inturn, greater for services than for government agencies, as well as find causefor concern in the observation that customer satisfaction in the United States is declining, primarily because ofdecreasing satisfaction with services. The authors estimate the model tor the seven major economic sectors forwhich data are collected. Highlights of the findings include that (1) customization is more important than reliabilityin determining customer satisfaction, (2) customer expectations play a greater role in sectors in which variance inproduction and consumption is relatively low, and (3) customer satisfaction is more quality-driven than value- or

price-driven. The authors conclude with a discussion of the implications of ACSI for public policymakers, managers,consumers, and marketing in general.

T he economy The economy is changing. The centralfeature of the old economy was the mass productionand consumption of commodities. The modem econ-omy is based on production and consumption of increas-ingly differentiated goods and services.How should we measure economic perfonnance in this

new wo rld? As the economy chang es, theories and measuresmust change, too. In particular, it seems clear that conven-tional "output." or "quantity," measures of economic perfor-man ce, such as productivity, are not only extrem ely difficultto compu te in a differentiated marke tplace, but also that theyprobably tell us less than they used to. For example, theUnited States leads the world in productivity, but incomeshave stagnated. Italy has had one of the most dramatic pro-ductivity increases of any country, but it has not translatedinto strong economic growth. The current trend towarddownsizing in U.S. firms may increase productivity in the

Claes Fornell is the Donald C. Cook Professor of Business Administrationand Director of the National Quality Researcti Center, University of Michi-gan Business S chool, MichaelD Johnson is Professor of Marketing and acore faculty member of the National Quality Research Center, Universityof Michigan Business School, EugeneW Anderson is Associate Professorof Marketing and a core faculty member of the National Quality ResearchCenter, University of Michigan Business School, Barbara Everitt Bryantand Jaesung Cha are Research Scientists at the National QualityResearch Center, University of Micfiigan Business School, The authorsgratefully acknowledge the financial support of the University of MichiganBusiness School, ihe American Society for Quality Control (ASQC), andthe ACSI's corporate sponsors. The authofs also thank Jack West ofASQC for his contributions in spearheading the development and deploy-ment of theACSI.This research has benefited from the constructive con-tributions of the Editor and three anonymous Mreviewers.

short term, but the downsized firm s' future fmancia perfor-mance will suffer if repeat business is dependent on labor-intensive customized service (Anderson, Fomell, and Ru.st1996).Hence, in the new economy, producing morehoweverefficientlyis not necessarily better. There is a pressingneed to augment current approaches to evaluating the fman-cial health of individual firms, let alone the wealth ofnations. Moreover, as the economy continues to evolve, "thegap between the two economies^the one that govemmentmeasures and the one businesses and economists are strug-gling to understandis widening" [Fortune 1993, p. 108).To understand more fully the modem economy, and thefirms that compete in it, we must measure the qua lityof eco-nomic output, as well as its quantityWe introduce the American Customer Satisfaction Index(ACSI), which represents a new type of customer-basedmeasuretnent system for evaluating-and enhancingthepertbrmance of firms, industries, economic sectors, andnational economies. It is designed to be representative of theeconomy as a whole and covers more than 200 firms, with1994 sales in excess of $2.7 trillion competing in over 40industries in the seven major consumer sectors of the econ-omy. On an annual basis, the ACSI system estimates a finn-level customer satisfaction index for each company in thesample and weights these firm-level indices to calculateindustry, sector, and national indices.

The American Customer Satisfaction Index measuresthe quality of the goods and services as experienced bythe customers that consume them. An individual t1rm"sACSI represents its served market 'sits customers'overall evaluation of total purchase and consumptionexperience, bolh actual and anticipated (Anderson. For-


2/13

nell, and Lehmann 1994; Fomell 1992; Johnson and For-nell 1991). Analogously, an industry ACSI represents anindu stry s cu stom ers overall evaluation of its marketoffering, a sector ACSI is an overall evaluation of that sec-tor, and the national A CSI gau ges the na tion s total con-sumption experience. Hence, ACSI represents a cumula-tive evaluation of a firm s market offering, rather than aperso n s evaluation of a specific transaction. A lthoughtransaction-specific satisfaction measures may providespecific diagnostic information about a particular productor service encounter, overall customer satisfaction is amore fund amental indicator of the firm s past, current,and future perfomiance (Anderson, Fomell, and Lehmann1994).The balance of this article is structured into five sec-tions.First, we explain the nature of ACSI: the theory under-lying the ACSI, the nation-wide survey methodology used tocollect the data, and the econometric approach employed toestimate ACSI. Second, we discuss the use of ACSI in con-ducting benchmarking studies, both cross-sectionally andover time. Third, to demonstrate the m odel s general applic-

ability and usefulness, we estimate the model for the sevenmajor economic sectors for which data are collected. Fourth,we discuss systematic cross-sector variation in each of theseareas. Fifth, we conclude with a discussion of tbe implica-tions of ACSI for public policymakers, managers, and indi-vidual consumers, as well as for marketing in general.

The ACSI Model and MethodologyTbe concept behind ACSI, namely, a measure of overall cutomer satisfaction that is uniform and comparable, requirea methodology with two fundamental pro perties. First, thmethodology must recognize that ACSI and the other constructs in the model represent different types of customeevaluations that cannot be measured directly. AccordinglACSI uses a multiple indicator approach to measure overacustomer satisfaction as a latent variable. The result is latent variable score or index that is general enough to bcomparable across firms, industries, sectors, and nations.

Second, as an overall measure of customer satisfactioACSI must be m easured in a way that not only accounts foconsumption experience, but also is forward-looking. Tthis end, ACSI is embedded in the system of cause aneffect relationships shown in Figure 1, which makes it thcenterpiece in a chain of relationships running from thantecedents of overall customer satisfactionexpectationperceived quality, and valueto the consequences of oveall customer satisfactionvoice and loyalty. As was indcated, the primary objective in estimating this system model is to explain customer loyalty. It is through th

Fo r a mo re extensive and detailed description of the ACmethodology, please see American Society for Quality Contr(1995).F I G U R E 1The American Custonner Satisfaction Index ACS I) Model

PERCEIVEDQUALITY

CUSTOMERCOMPLAINTS

OVERALLCUSTOMER

SATISFACTIONACSI)

PERCEIVEDVALUE

CUSTOMERLOYALTY

CUSTOMEREXPECTATIONS


3/13

design that ACSI captures the served mark et's evaluation ofthe firm's oflering in a manner that is both backward- andforward-looking. Moreover, m odeling ACSI as part of sucha system serves to validate the index from a nomologicalstandpoint. Nomological validity, a form of construct valid-ity, is the degree to which a construct behaves as predictedwithin a system ot related con structs called a nomologic lne t (Cronbac h and Meehl 1955). To the extent that themodel predictions are supported, the validity of the ACSI issupported. CSi ntecedentsAs is shown in Figure 1, overall customer satisfaction(ACSI) has three antecedents: perceived quality, perceivedvalue, and customer expectations. The first determinant ofoverall customer satisfaction is perceived quality or perfor-mance, which is the served market's evaluation of recentconsumption experience, and is expected to have a directand positive effect on overall customer satisfaction. Thisprediction is intuitive and fundamental to all economicactivity. To operationalize the perceived quality construct,we draw on the quality literature to delineate two primarycomponents of consumption experience: (I) customization,that is, the degree to which the firm's offering is customizedto meet heterogeneous customer needs, and (2) reliability,that is, the degree to which the firm's offering is reliable,standardized, and free from deficiencies.

The second determinant of overall customer satisfac-tion is perceived value, or the perceived level of productquality relative to the price paid. Adding perceived valueincorporates price information into the model andincreases the comparability of the results across firms,industries, and sectors. Using value judgments to measureperformance also controls for differences in income andbudget constraints across respondents (Lancaster 1971),which enables us to compare high- and low-priced prod-ucts and services. For perceived quality, we expect a posi-tive association between perceived value increases andcustomer satisfaction.The third determinant of overall customer satisfaction isthe served market's expectations. Tbe served market'sexpectations represent both the served market's prior con-sumption experience with the firm's offeringincludingnonexperiential information available through sources suchas advertising and word-of-mouthand a forecast of thesupplier's ability to deliver quality in the future. As such, theexpectations construct is both backward- and forward-look-ing. It captures all previous quality experiences and infor-mation from t - 1, t - 2 t - m. Hence, it naturally has adirect and positive association with a cumulative evaluationof the firm's performance, such as overall customer satisfac-tion. At the same time, the served market's expectations attime t forecast a firm's ability to satisfy its market in futureperiods t + 1, t + 2, ..., t + n. This role of expectations isimportant because the nature of the ongoing relationshipbetween a firm and Its customer base is such that expectedfuture quality is critical to overall customer satisfaction.This predictive role of expectations also suggests that itshould have a positive effect on ov erall custome r satisfaction(Anderson, Fomell, and Lehmann 1994).

Finally, customer expectations should be positivelyrelated to perceived quality and, consequently, to perceivedvalue. Customer knowledge should be such that expecta-tions accurately mirror current quality. Hence, we expect theserved market to have expectations that are largely rationaland that reflect customers' ability to learn from experienceand predict the levels of quality and value they receive(Howard 1977). CSi ConsequencesFollowing Hirschman's (1970) exit-voice theory, the imme-diate consequences of increa.sed customer satisfaction aredecreased customer comp laints and increased customer loy-alty (Fom ell atid W emerfelt 1987). When dissatisfied, cu.s-tomers have the option of exiting (e.g., going to a competi-tor) or voicing their complaints in an attempt to receive ret-ribution. An increase in overall customer satisfaction shoulddecrease the incidence of complaints. Increased overall cus-tomer satisfaction should also increase customer loyalty.Loyalty is the ultimate dependent variable in the modelbecause of its value as a proxy for profitability (Reichheldand Sasser 1990).The final relationship in the model is between customercomplaints and customer loyalty. Although there are nodirect measures of the efficacy of a firm's customer serviceand complaint-handling systems, the direction and size ofthis relationship reflect on these systems (Fomell 1992).When the relationship is positive, the implication is that thefirm is successful in tuming com plaining custom ers intoloyal customers. When negative, the fimi's complaint han-dling has managed to make a bad situation even worseithas contributed further to customer defection. CSi MethodologyThe American Customer Satisfaction Index is designed to berepresentative of the nation's economy as a whole. Accord-ingly, in selecting the companies to measure, each of theseven major economic sectors (one-digit standard industrialclassification [SIC[ code level) with reachable end-userswere included in the design: (I) Man ufacturing/Non-durables, (2) Manufacturing/Durables. (3) Transportation/Com munications/Utilities, (4) Retail, (5) Finance/Insurance.(6) Services, and (7) Public Administration/Govemmem.Within each sector, the major industry groups (two-digit SICcodes) were included on the basis of relative contribution tothe gross domestic product. Within each industry group, sev-eral representative industries (four-digit SIC codes) wereincluded on the basis of total sales. Finally, within eachindustry the largest companies were selected, such that cov-erage included the majority of each selected industry's sales.For each firm, approximately 250 interviews were con-ducted with the firm's current customers. Interviews camefrom 48 replicate national probability samples of house-holds in the continental United States with telephones (95%of households). Prospective respondents (selected withoutsubstitution from the househo ld by the nearest birthda ymethod) were screened to identify purchasers of specificgoods or services within defined purchase and consumptiontime periods. These periods vary from three years for thepurcha se of a major dura ble, to within the past mo nth for


4/13

TABLE 1Measurement Variables Used in the ACS ModelMeasurement Variable Latent Variable

.Overall expectation of quality (prepurchase)2. Expectation regarding customization, or how well the product fits the customer's personalrequirements (prepurchase)3. Expectation regarding reliability, or how often things would go wrong (prepurchase)4. Overall evaluation of quality experience (postpurchase)5. Evaluation of customization experience, or how well the product fit the customer'spersona requirements (postpurchase)6. Evaluation of reliability experience, or how often things have gone wrong (postpurchase)7. Rating of quality given price8. Rating of price given quality9. Overall satisfaction10. Expectancy disconfirmation (performance that falls short of or exceeds expectations)

11 . Performance versus the customer's idea product or service in the category

12. Has the customer complained either formally or informally about the product or service?13. Repurchase likelihood rating14 . Price tolerance (increase) given repurchase15. Price tolerance (decrease) to induce repurchase

Customer expectationsCustomer expectationsCustomer expectationsPerceived qualityPerceived qualityPerceived qualityPerceived valuePerceived va ueACSIACSIACSI

Customer complaintsCustomer loyaltyCustomer loyaltyCustomer loya ty

frequently purchased cons um er goods and services, to cur-rently having a bank account or insurance policy in the per-son s own name.Once a respondent was identified as a customer, theinterviewer proceeded with the customer satisfaction ques-tionnaire. Each questionnaire contains the same 17 struc-tured questions and 8 demographic questions. Lead-in word-ing and examples were tailored to specific goods and ser-vices. In Table I. we describe the 15 measurement variablesfrom the ACSI survey that are used in the model estimationand identify the associated latent variable.Customer expectations were measured by askingrespondents to think back and remember the level of qualitythey expected on the basis of their knowledge and experi-ence with a good or service.- Three expectation measureswere collected: (I) overall expectations, (2) expectationsregarding customization, and (3) expectations regardingreliability. Customers then rated their recent experience withthe good or service by using three measures: (1) overall per-

ceived quality, (2) perceived customization, and (3) per-ceived reliability. Two questions then tapped perceivedvalue, quality relative to price, and price relative to quality.Overall customer satisfaction (ACSI) was operational-ized through three survey measures: (I) an overall rating ofsatisfaction, (2) the degree to which performance falls shortof or exceeds expectations, and (3) a rating of performaneerelative to the cu stom er s ideal good or service In the cate-gory. Whereas the latter are commonly used as antecedentsin models of transaction-specific satisfaction (Oliver 1980;

^Although such post hoc m easures of expectations are Imperfect,the cost of obtaining expectations prior lo purchase is prohibitivein a study of this magnitude.

Yi 1991), their use as reflective indicators of overall cutomer satisfaction is consistent with the cumulative nature oACSI, because each measure represents a qualitatively dferent benchmark customers use in making cumulative evauations, such as overall customer satisfaction (ACSI). Morover, the latent variable methodology employed to estimaoverall customer satisfaction only extracts shared variancor that portion of each measure that is common to all threquestions and related to the ACSI construct s position in thmo del s chain of cause and effect. T hus, satisfaction is nconfounded by either disconfirmation or comparison to aideal. Only the psychological distance between perfonnancand expectations, and between performance and the cutom er s ideal po int, was used to estimate overall custo msatisfaction (ACSI).Customer complaints were measured by whether a cutomer had complained either formally (as in writing or bphone to a manufacturer) or informally (as to service pesonnel or a retailer). In addition, there were two measures

customer loyalty. The first was repurchase likelihood. Thsecond measure was constructed from two survey variablethe degree to which a firm could raise its prices as a pecentage before the customer would definitely not choose buy from that firm again the next time (given that the cutomer has indicated he or she is likely to repurchase) and tdegree to which a firm could lower its prices as a percentabefore the customer would definitely choose again from thfirm the next time (given that the customer has indicated hor she is unlikely to repurchase).Scales and Modet stimationThe frequency distribution of satisfaction and quality ratinis always negatively skewed in competitive markets (Fome


5/13

1995). To reduce the statistical problems of extreme skew-ness, the ACSI uses 10-point (versus 5- or 7-point) ratingscales to enable customers to make better discriminations(Andrews 1984). The use of multiple indicators also reducesskewn ess (Fomell 1992). A version of partial least squares(PLS) is used to estimate the mode] (Wold ]989). Partialleast squares is an iterative procedure for estimating causalmodels, which does not impose distributional assumptionson the data, and accommodates continuous as well as cate-gorical variables. Because of the model structure, PLS esti-mates weights for the survey measures that maximize theirability to explain customer loyalty a.s the ultimate endoge-nous or dependent variable. The estimated weights are usedto construct index values (transformed to a 0- to lOO-pointscale) for ACSI and the other model constructs.

The ACSI values lor individual industries and sectors, aswell as the overall economy are computed by aggregatingfinm-ievel results. An industry-level ACS] is an agg regate offinn results weighted by firm sales (such as Philip Morrissales for Miller Beer in the beverages and beer industry andexcluding other corporate Philip Morris sales). A sectorACSI is an aggregate of industry results weighted by indus-try sales. The overall ACSI is an average of the sector resultsweighted by each sector's contribution to the gross domes-tic product, A forma] statement of the ACSI model and itsmeasurement is provided in the Appendix..h volution of National IndicesThe first truly national custom er satisfaction index w as theSwedish Customer Satisfaction Barometer, or SCSB (Fomell1992). Developed in 1989, the SCSB includes 31 majorSwedish industries. In Germany, the Deutsche Kunden-barometer, or DK, was introduced in 1992 and as of 1994also includes 3 industries (Meyer 1994). The AmericanCustom er Satisfaction Index was first introduced in the fall of1994.with information on 40 indu stries and seven major sec-tors of the U.S. economy (National Quality Research Center1995).Recently, New Zealand and Taiwan also have startedindices for customer satisfaction. The European Union alsohas recommended that it be done in its member countries.The original Swedish Barometer used perceived value(quality given price and price given quality) and a singlemeasure of customer expectations. It has since been respec-ified to conform to the ACSI, By drawing on the quality lit-erature and its distinction between customization and relia-bility, the ACSI approach introduces a perceived qualityindex to go with the original value index. The addition ofthis quality construct has two advantages. Distinguishingbetween quality and value provides information regardingthe degree to which satisfaction is price- versus quality-dri-ven. The quality construct also provides information regard-ing the relative importance of customization and reliabilityin detemiining quality.

Other improvements incorporated into the ACSI focuson the measurement properties of the model. The three mea-sures of perceived quality (.see Table I) fumish the ACSImodel with three corresponding measures for operationaliz-ing customer expectations (compared to the single expecta-tions measure used in the original SCSB model). An addi-tional question is asked in the ACSI to improve the mea-

surement of price tolerance, ln the original Swedish model,customers were only asked how much more they would pay,given that they were likely to repurchase. In the ACSImode , customers are also asked how much the price wouldneed to be reduced for them to repurchase, given that theyare unlikely to repurchase. This provides a more completeprice tolerance measure that improves the measurementproperties of the loyalty construct.Finally, the ACSI requires a different approach to surveysampling because of the complexity of the U.S. economy, [nthe SCSB, a firm is surveyed for customer satisfaction witha single representative brand. This is impractical in theUnited States, where manufacturers such as the domestic"big three" automobile companies and General Electric oflera wide range of popular models under more than one brandname. As was described, customers in the ACSI are surveyedat the brand or model level, and the various brand or modelobservations are combined to estimate a firm-level model.

ACSI as a Benchmark CrossSectionally and Over TimeIn Figure 2, we provide a breakdown of the industry, sector,and overall ACSI results for the baseline ACSI releasedOctober, 1994. The overall ACSI for the 1994 baseline was74.5. What do the numbers shown in Figure 2 mean? Aslatent variables, the ACSI measures are comparable (albeitnot blindly) across firms, industries, sectors, and nations.For example, the ACSI averages approximately 80 forgoods, 75 for services and retailers, and 64 for public andgovemment agencies. This implies that customers are gen-erally more satisfied with goods than with services and areleast satisfied with public administration and govemmentagencies.The American Customer Satisfaction index measureslend themselves well to benchmarking over both time andcontext. For example, the 1994 ACSI indices provide a base-line for determining whether the marketplace is becomingmore or less satisfied with the goods and services providedby individual firms, entire ind ustries, and different sectors ofthe economy, as well as with the economic life of the nationas a whole. With regard to the last item, it is interesting toobserve the decline in ACSI at the national level, which isshown in Figure 3. The American Customer SatisfactionIndex falls from the baseline level of 74.5 in 1994 to a lowof 73.0 in the first quarter of 1996, The dec line is driven pri-marily by decreasing customer satisfaction with services. Tothe extent that long-term profitability depends on customerloyalty and the efficiencies gained from long-term buyer-seller relationships, this drop in satisfaction with servicesshould be seen as a waming signal about the long-termfinancial prospects for the finns affected. More important,because services are a large and growing portion of theeconomy, such a decline may reflect a weakening of theeconomy in general and a lowering of living standards withregard to consumption quality.

Tracking ACSI over time can also yield interestinginsights at the firm level. One of the biggest winners in thefirst release of the firm-level results of the index was theU.S, Postal Service, which rose 13.0% to 69 in 1995 {Far-


6/13

FIGURE 21994 Baseiine ACSI ResultsACSINational Score = 74.5

ManufacturingNondurables81.6

82 Apparel79 AthleticShoes83 Beer81 Cigarettes87 CannedFood87 Chocolate86 Milk, IceCream84 BakedGoods82 Cold Meats,Cheese83 Cereal78 Gasoline72 Newspapers84 PersonalCare86 Soft Drinks

ManufacturingDurables79.2

79 A utomobiles83 ConsumerElectronics85 HouseholdAppliances78 PersonalComputers,Printers

TransportationCommunicationsUtilities75.4

1172 Airlines77 Broadcasting,Television75 ElectricService81 ExpressDelivery79 Phone/Local82 Phone/LongDistance60 U.S. PostalService

77776976

Retail75.7

DepartmentStoresDiscountStoresRestaurants,Fast FoodsSupermarkets

FinanceInsurance75.4

74 Banks/Commercial77 Insurance/Life81 Insurance/Property

Services74.4

74 Hospitals75 Hotels77 MotionPictures

PublicAdministrationGovernment64.311GarbageService:74 CentralCity74 SuburbanPolice:61 CentralCity65 Suburban55 Internal

RevenueService

FiGURE 3Nationai ACSi Over Time

73.7 7 7 73.7ACSI

Baseline Q4 1994 Q l 1995 Q2 1995 Q l 1995 i 1995 Q 1996 Q2 1996Quarter

tu 1995). The postal service is engaged in a massive effortto improve the quality of their services, and it posted recordprofits last year. Large declines in the index were seen atcompanies that recently engaged in substantial downsizing,such as GTE (down 5.3% to 72) and Kmart {down 5.4% to70).Firm, industry, sector, and national ACSI scores also canbe compared cross-sectionally within a given time period.For example, we can determine how well a particular firm isdoing relative to the best firms in its own industry, the bestfirms in other industries in that sector, or the hest in nation

as a whole. Industries and sectors can be compared with oneanother in a similar fashion.

The American Customer Satisfaction Index also can bcompared to the findings of Sweden's SCSB and GermanyDK. For examp le, the pattern of sector differences found the ACSI is consistent with that observed in the SCSB anthe DK. That is, goods score higher than services, and public administration is always the lowest scoring sector However, viewed against the SCSB and DK, the ACSI scores arelatively high, especially in the goods and services sector

Although it is relatively straightforward to ask whicfirms, industries, sectors, and nations are relatively moeffective at providing satisfying goods and/or services, it a different question to ask whether a particular firm, indutry, sector, or nation is "performing well." To do so, it is neessary to put the AC S index numbe rs in context. Froresearch conducted using SCSB data, it is known that certaifactors make it more or less difficult to achieve high cutomer satisfaction and are likely to lead to higher or lowcustomer satisfaction for different types of industries. Fonell and Johnson {1993) find customer satisfaction higher industries with a significant level of competition and diffeentiation. Anderson (1994) finds satisfaction higher whecompetition, differentiation, involvement, or experience high or when switching costs, difficulty of standardizatioor ease of evaluating quality is low. Both studies find satifaction higher for goods than for services or retailers. Thresearch enables us to conjecture that the observed highsatisfaction in the United States is due to the greater degreof competition found in most U.S. industries.


7/13

The question remains, however, once structural differ-ences are taken into account, is a particular ACSI score"go od" or "bad "? This question pertains to differences in theACSI that arc attributable to conduct, as opposed to struc-ture. Such henchmarking judgments are relatively straight-forward within a particular industry, in which firms can becompared against one another. In such cases, industry struc-ture is held constan t, such that differences in firm perfor-mance may be attributed to the firm's conduct.Firms that do particularly well relative to their competi-tion include Southwest Airlines, with an ACSI of 76 relativeto an industry average of 69, and Wal-Mart, with a score of80 relative to an Industry average of 74. Both fimis havedeveloped difficult-to-imitate strategies and resources thathave put them far ahead of the competition in their respec-tive industries. Firms that trail the competition in theirindustry include Hyundai, with an ACSI of 68 relative to theautomobile industry average of 80, and A&R with a score of69 relative to the supennarket industry's average of 74, andare each increasingly vulnerable to aggressive competitors.

Within a sector, though different industries are likely toshare similar characteristics, benchmarking should be donewith deference to differences in industry structure , such asthe degree of com petition. As is shown in Figure 2, the ACSIfor long-distance telephone service (82) is greater than localtelephone service (79). Consequently, il is possihle to saythat customer satisfaction with long distance service ishigher. However, it is not possible to say whether industrypertbmiance is good or bad without first taking into accountdifferences in satisfaction due to structural characteristics. Inother words, greater competition is likely driving the ACSIhigher in long-distance service, but it is not appropriate toevaluate the conduct of the industry without first taking thisstmctural difference into account. For example, long-dis-tance service's score of8 may be weak, given the industry'sstmctural characteristics, whereas the local telephone indus-try's score of 79 may be high, given its situation. If so. thenthough long-distance service would be providing higher cus-tomer satisfaction, it might be encouraged to do better Atthe same time, local telephone's performance would be "bet-ter," because the served market is more satisfied than wouldbe expected on the basis of its industry characteristics.Hence, benchmarking requires further "handicapping" toaccount tor differences in both. Development of a deeperunderstanding of how to handicap ACSI .scores for hench-marking purposes is a promising avenue for further research.

General pp licab ility andUsefulness of CSIUsing individual respondents as observations, we heredescribe the results of estimating the ACSI model for theseven measured sectors of the economy.^ In contrast to the

'The samples were as follows: (I) Manulacturing/Nondurablcs(n = 12,075; 26.8% of sample), (2) Manufacturing/Durables(n = 7.S2S: 17.4%), (3) Transportation/Communications/Utilitiestn = 10.10 ; 22.4% ). (4) Retail (n = 7.243; 16.1%). (5)Finance/Insurance (n = 3,236; 7.2%), (ft) Other Services(n = 3.32S; 1.4 ). and (7) Public Administratioti/Governmenl(n = I.1S3; 2.6%).

baseline results, in which the model is estimated for eachfirm and the results are aggregated to industry and sectorindices and to an overall national ACSi, here each sector istreated as a .subsegment of the overall ACSI population ofrespondents. In particular, we discuss the general applicabil-ity of the model and several key findings regarding (I) therelative importance of customization and reliability, (2) thepredictive nature of expectations, and (3) the relative impor-tance of price and quality. Throughout, we use standardizedvariables (correlations) to evaluate the measurement portionof the model and fit measures, whereas we use unstandard-ized v ariables (covariances) as input to estimate effect sizes.Jackknifing is used to obtain standard errors for each of themodel parameters. Wherever model estimates (loadings andeffects) are compared or contrasted across sectors in the dis-cussion of results, the differences are significant {p .05).

General applicability of the model. Overall, we expectthe ACSI model to he generally applicable to multiple .sec-tors. The model and measures are designed to provide thisgenerality. This prediction is examined through several indi-cators. The first is whether the estimated path coefficientsare significant in the predicted d irections. We find themodel's path coefficients to be significant and in the pre-dicted direction for 54 of 56 possible cases.-^The second indicator of the model's performance is it'sability to explain important latent variables in the model,especially overall customer satisfaction (ACSI) and loyalty.We find ihai the estimated model explains a substantial pro-portion of the variance in both constructs. For overall cus-tomer satisfaction (ACSI), R- measures range from .70 forsector 1 to .80 for sector 5 (average of .75 ). For custom erloyalty, R- ranges from .26 for sector 6 to .47 for sector 5(average of .36).The third and fourth indicators capture the fit of the mea-surement variable (MV) and latent variable (LV) portions ofthe model: the proportion of available covariance in the MVsexplained, and the proportion of available covariance in theLVs explained. The measurement variable loadings for theACSI model (not shown) are all relatively large and positive.The percent of MV covariance explained ranges from 84%for sector I to 89% for sectors 3 and 5 (average of .87). Thepercent of LV covariance explained ranges from 92% forsector 2 to 95% for sector 5 (average of 94%). Each model,therefore, explains well over 90% of the LV covarianceavailable in a model that specifies 9 of 15 possible LV rela-tionships. This suggests that there are no major relationships

in our data that the ACSI model fails to capture.Customization versus reliabilityThe measurement load-ings suggest that customization is more central to cus-tomers' expectations and perceptions of quality than relia-bility. The average loadings for the expectations constmctwere .81, .85, and .68 for expected quality, customization,and reliability, respectively. The average loadings for theperceived quality construct were .907, .906, and .77 for

The tw(j relationships thai were not correctly predicted were forsector5 (Finance/Insurance), in which there is a negative, albeitnotsignificant, eftect of expectations on overall cuslomer satisfaction.and seclor 7 (Pubiic Administration/Government), in which theeffect ol expectations on value is positive but mil significant (.035).


8/13

quality, customization, and reliability, respectively. Consis-tent with the nature of production and consumption in ser-vice-oriented sectors, customization is more central to qual-ity for sectors 3, 5, 6, and 7 (average of .909) than for themanufactured goods sectors, and 2 (average of .899). Forall .sectors, the loadings for the customization measures aresignificantly higher than the loadings for the reliability mea-sures (for both the expectations and perceived quality con-structs). This implies that squeezing more variance out of amanufacturing or service delivery process may not increaseperceived quality and customer satisfaction as much as tai-loring goods and services to meet customer or market seg-ment needs.

ti predictive nature of customer expectations. As wasexpected, we find that customer expectations are largelyrational in that expectations predict quaiity, value, and cus-tomer satisfaction. This is consistent with previous researchusing the SC SB, which shows that the served market sexpectations are relatively stable and accurate, especially inthe aggregate (Anderson, Fom ell, and Lehmann 1994; John-son, Anderson, and Fomell 1995).With regard to the impact of expectations on quality andvalue, it is useful to look at the sum of the two effects([expectations on quality] + [expectations on value]). Thisjoint effect is greatest in four sectors: Manufacturing/Non-durables = .68; Transportation/Communications/Utilities =.71; Retail = .81 ; and Public Administration/Govemm ent =.67. The joint effect is smaller in Man ufacturing /Durab les.Finance/Insurance, and Services (.45, .58, and .59, respec-tively). These findings are compatible with the argumentthat expectations are less predictive when variance in con-sumption and production factors are high (Anderson 1994).On the production side, if a particular good or service is dif-ficult to standardize or quaiity is relatively unambiguous,variance in consumption experience is greater and expecta-tions should have less influence. Similariy, on the consump-tion side, if customers are more likely to perceive variancein productionperhaps because of involvement or expertisegained through experiencethen expectations should,again, have less infiuence. For example, customer expecta-tions should be better predictors of quality, value, and satis-faction in those sectors in which customers make frequentand relatively routine purchase and consumption decisions(Howard 1977). When such interactions are less frequent,customers have less direct knowledge and their expectationsshould be weaker predictors of perceived quality and value.The findings for the direct association between expec-tations and satisfaction are similar (see Figure 4). Theimpact of expectations is larger in two sectors (sector I:Manufacturing/Nondurahles = .06, and sector 4: Retail =.07), in which variance in production and consumption fac-tors is relatively low. The effect is not as high in the otherfour competitive market sectors. The largest effect ofexpectations on satisfaction is for sector 7: Public Admin-istration/Govemment at .09. This may reveal the impactthat a negative image can have on satisfaction because ofthe halo surrounding certain publicly provided services(such as the Internal Revenue Service). This interpretationis consistent with research in which negative factors and

FIGURE 4The Direct Effect of Expectations on CustomerSatisfactionPublic Ad m ini Mitral ion/G overn m ent

Olher ServicesFinance/Insurance

RetailTransportation/Com miinication/Utilities

Manutacluring/DurabiesManufacturing/Nondurahles

-.0 2 .02 .06 .1Expectations Eftect

framing effects have a larger impact on evaluations than dtheir positive counterparts (Kahneman and Tversky 1979)The observed negative effect for Finance/Insurance -.01is not significant.An examination of the total effect of expectations on satisfactionthe direct effect of expectations on satisfactionplus the effect of expectations on overall customer satisfaction through quality and valueyields similar resultsExpectations have the greatest impact in the Public Administration/Govemment (.59) and Retail (.59) sectors, followed by Transportation/Communications/Utilities (.53)Manufacturing/Nondurables (.49). and Services (.47). Th

total effect of expectations is lowest for the ManufacturingDurables (.36) and Finance (.41) sectors. For the latter, current quality experiences may be relatively more salientsuch as the perfomiance of an automobile or a mutuafundand may take precedence over previous quality exp eriences in determining overall customer satisfactionwhereas long-term reputation effects may play a greater rolin sectors in which expectations have a greater impact.Price versus quatity driven satisfaction. The impact oquality on overall customer satisfaction is greater than thaof value in each of the seven sectors. The average direceffect of q uality on overall custome r satisfaction is .55

whereas the direct effect of value on overall customer satisfaction is .36. The total effect of qu ality on satisfaction thdirect effect plus the effect through valueaverages .76This difference is consistent with the notion that thougvalue may be more central to the formation of customersinitial preferences and choice, quality, in contrast, is morcentral to the consumption experience itself.An important question is how does the relative importance of price versus quality vary across sectors? If pricerather than quality, is driving overall customer satisfactionthe effect of a one-po int chang e in value on overall custo mesatisfaction should be high relative to the total effect of one-point change in quality on overall customer satisfactionA price- versus quality-driven satisfaction ratio is calculate


9/13

FIGURE 5Prjce Versus Quality Driven Satisfaction RatioPublic Admini^lrali^^n/Govcrnment

Other ServicesFinance/Insurance

Rcla i lTr i insponai inn/Communiuat ionAJl i l i t ies

Manufactur ing/DurablesManufaciur ing/Nondurables

3 4 5 6Price Driven Sat i Taction

as the impact of a one-point change in perceived value onoverall customer satisfaction divided hy the total effect of aone point change in perceived quality on overall customersatisfaction (the direct effect of quality on satisfaction plusthe indirect effect of quality on satisfaction through value).As is shown in Figure 5, relative price-driven satisfac-tion is greatest in sectors 1 and 3 , in which the ratio ofeffects equals .53 and .56, respectively, compared to anoverall average of .47 across sectors, For sector I (Manu-facturing/Nondurables) the result is consistent with the shifttoward price-based competition in this industry, which wasobserved throughout tbe 1980s (Buzzell, Quelch, andSalmon 1990). in which price competition is fostered by theavailability of low-priced house and generic brands and dis-count retailing. Price-driven satisfaction is highest for sector3 (Transportation/Communications/Utilities), in whichcompetition is relatively commodity-based and price plays acorrespondingly important role.Relative price-driven satisfaction is lowest for Manufac-turing/Durables (.43), Services (.43), Retail (.44), and Gov-emmenl Agencies (.40), which implies thut quality is rela-tively more central to market behavior in these sectors. Forthe first two, this finding is consistent with the high involve-ment and customized nature of the products involved. ForRetail, the effects on both quality and value are relatively

low. This may be due to the location-driven nature of thissector. The ratio for Public Administration/Govenimentmost likely reflects the take it or leave it nature of pricingin this sector.Although our observations are based on aggregationover the wide variety of segments, finns, and industrieswithin a given sector, overall it appears that a ratio of theimpact of value to the impact of quality on satisfaction pro-vides a useful measure of price- versus quality-driven satis-faction with strong face validity. In further research, the effi-cacy of the price-driven satisfaction ratio also should bedemonstrated through as.sociation with other constructs inthe database, as well as through the relatitmships between

those constructs. For example, in industries in which overall

customer satisfaction is relatively price-driven we mightexpect the effect of overall customer satisfaction on loyaltyto be relatively low. At the sector level, we find this to be thecase for Manulacturing/Nondurables. yet the Transporta-tion/Communications/Utility sector is average in terms ofhow sensitive loyalty is to overall customer satisfaction.Thus, the price-driven satisfaction ratio also should be asso-ciated with degree of loyalty.With the notable exception of Manufacturing/Non-durables, we Imd loyalty to be lower in sectors in whichoverall customer satisfaction is relatively price-driven. Inpart, this is driven by relatively high price tolerance becauseof the low-priced nature of goods sold in that sector. How-ever, repurchase likelihood also is highest in this sector, aswell as in the Transportation/Communications/Utilities sec-tor. Clearly, the mixed nature of these findings suggests thatadditional research on these subjects must carefully controlfor category characteristics and industrial organization fac-tors, such as switching costs and concentration, that canaffect loyalty and the overall customer satisfaction-loyaltyrelationship.

Discussion and ImplicationsThe American Custom er Satisfaction Index represents a sig-nificant step forward in the evolution of national satisfactionindicators. It provides an independent and uniform means ofassessing the quality of what is consumed and prt)duced inthe economy. It is a much needed missing link in what weneed to understand about the health of the economy and theindividual firms ihat com pete in it. For exam ple, if quality inthe United States is declining as a consequence of decliningoverall customer satisfaction with the service sector, thisshould be cause for concern.For pijblic policymakers, ACSI has the potential to be auseful tool for evaluating and enhancing the health of thenation's economy, both in terms of national competitivenessand the welfare of its citizens. In as.sessing the health of theeconomy, it can provide an important complement to con-ventional measures of the quantity of goods and servicesproducedsuch as productivity and price indicesand canbalance these measures against the quality of goods and ser-vices produced. For legislative efforts, ACSI can be useful inpredicting and monitoring the effects of puhMc policy deci-sions on issues as diverse as deregulation, taxes, interestrates, price ceilings, and subsidies. In terms of balance oftrade, ACSI can provide an early warning as to whether anindustry is vulnerable to competitive encroachment underconditions of free trade.

For managers and investors, ACSI provides an importantmeasure of the fimi's past and current perfomiance, as wellas future financial health. The ACSI provides a means ofmeasuring on e of a firm's most fundamental revenue-gen er-ating assets: its customers. Higher customer satisfactionshould increase loyalty, reduce price elasticities, insulatecurrent market share from competitors, lower transactioncosts, reduce failure costs and the costs of attracting newcustomers, and help build a linn's reputation in the market-place (Anderson, F om ell, and Lehmann 1994). As such,ACSI provides a leading indicator of the firm's future finan-


10/13

cial health. By establishing a standard measure of qualitywith clear links to long-term pertormance, ACSI may evenhelp instill more of a long-term perspective in both manage-ment and investors.The empirical evidence that ACSI is a leading indicatorof financial performance is becoming increasingly persua-sive.This is true for accounting profits, as well a.s for share-holder value. Specifically, it has been shown that both the

ACSI (Ittner and Larcker 1996) and its Swedish counterpart(And erson, Fom ell, and Lehmann 1994} have a po.sitiveassociation with return on investment. In terms of marketvalue. Ittner and Larcker (1996) estimate that a one-unitchange in ACSI is associated with a $654 million increase inthe market value of equity above and beyond the account-ing-book value of assets and liabilities. Stock trading strate-gies based on either the ACSI or SCSB have delivered port-folio returns well above market returns (Fomell. Ittner, andLarcker 1993, 1996). Also, recent results suggesi that thepublic release ofACSIscores causes a significant stock mar-ket reactionpositive market adjusted returns for high-scor-ing firms and negative adjusted returns for low-scoring firm s(Fomeli, Ittner, and Larcker 1996).The American Customer Satisfaction Index also hasimplications for managers formulating competitive strategy.One of its key benefits is that ACSI represents a uniform andcomparable system of measurement that allows for system-atic bencbmarking over time and across firms. In addition, itcan be useful in analyzing tbe strengths and weaknesses oftbe firm or its competitors. For example, declining overallcustomer satisfaction is likely to be symptomatic of deeperproblems facing a firm. Because ACSI provides a measure ofthe effectiveness with which a firm is defending current cus-tomers, firms with low ACSI scores are particularly vulnera-ble and provide expansion opportunities for more competentorganizations.For customers, ACSI provides information that is notonly useful in making purchase de cisions, but also likely tolead to improvements in tbe quality of the goods and ser-vices they consume, as well as in their overall standard ofliving. The independence, uniformity, and methodologyunderlying ACSI means that it provides information to buy-ers not found in ad hoc methodologies employed in productratings by popular magazines and commercial marketresearch. Moreover, tbe mere existence of such a measurewould likely lead to improvements in tbe quality of goodsand services. In a monopoly situation, ACSI may belp policethe market. In addition, ACSI should have particularlyimportant implications for both the quality of tbese servicesand tbe prices customers pay for tbem. In competitive situa-tions, too, ACSI should encourage quality competition andlead to greater customer satisfaction over time. The ultimateoutcome should be an improvement in the quality of eco-nomic life.

To summ arize, ACSI represents a new means of evaluat-ing and enhancing performance for the modem firm and themodem economy. It provides a complement to conventionalmeasures, such as productivity and price indices, that treatquality as a residual. In doing so , it has the potential to moveto center stage tbe quality goods and servicesas experi-enced by tbe customers of those goods and servicesof

firms, industries, and nations seeking to maintain and/ostrengthen their positions in tbe increasingly competitiveconomic environment that is unfolding as we move into thtwenty-first century. Because marketing scholars and practitioners have long recognized that customer satisfaction is aimportant and central concept, as well as an important goaof all business activity, the role of marketing in this newworld should be self-evident.

AppendixThe formal expression of the model depicted in Figure can be written as a series of equations such that the systematic part of the predictor relationships is tbe conditionaexpectation of the dependent variables for given values opredictors. The general equation is thus specified as stochastic:

where n ' = (n b 12 Hm) and ^' = (^|, ^2 ^n) are vectors of unobserved (latent) endogenous and exogenous varables,respectively, B (m x m ) is a matrix of coefficient parameters for tl, and f (m x n) is a matrix of coefficient parameters for ^. This implies that E[ri^'] = E[^^'l = E(^l = 0w h e r e i ; - n - E l l i n - ^ l .The corresponding equation tbat relates tbe latent varables in the model is

where

0P21 0

0

00P3200

000

P43

0000

P54

00000 Hs

Y i iY2173 100

, = customer expectations,Til = perceived quality,TI2 = perceived value,TI = ACSI,TI4 = customer com plaints, andri5 = customer loyalty.Tbe general equations relating the latent variables to tbmeasurement variables are

y =and

5,where y' = (yi. y2. - yp) and x' = (X|, K2 ..., x^) are tbe mesured endogenous and exogenous var iables , respect ively . A(p X m) and A^ (q x n) are the corres pon ding regre ssiomatr ices . By impl icat ion f rom PLS est imat ion (Fomel l anBookstein 1982) , we bave E|el = EI6] = Elne ' l = E l ^ 5 ' | = Tbe corresponding equat ions in ACSI are

3=

W , |W j ,W3 I

Sz53


11/13

and

y?

0 0 0W | 0W22 0 w 3

^ 3

^ 2 5

E.IE4

Ey 4

whereX| = customer expectations about overall quality,X2= customer expectations about reliability,Xj = customer expectations about customization,yi =overall quality,yi =reliability,yi = customizati(m,y4 = price given quality,y^ = quality given price,y^ = overall customer satisfaction,yy = confirmation of expectations,yg = distance to ideal product (service),yg = formal or informal complaint behavior,

yio = repurchase intention, andyii = price tolerance (reservation price).Tlie general form of the ACSI is as follows:

where ^ is the latent variable for overall customer satisfac-tion, and [.],Min{.\ and Ma.x\.\denote the expected, theminimum, and the maximum value of tbe variable, respec-tively. The minimum and tbe maximum values are deter-mined by those of the corresponding measurementvariables:

and

Max[xJ

where Xj's are the measurement variables of tbe latent over-all customer satisfaction, Wj s are tbe weights, and n is tbenumber of measurement variables. In calculating the ACSI,unstandardized weights must be used if unstandardizedmeasurement variables are used.

In ACSI, there are three indicators for overall customersatisfaction, which range from 1to 10. Then, tbe calculationis simplified to

ACSI = X 100

ACSI= X 00 where tbe Wj's are the unstandardized weights.

R F R N SAmerican Societyfor Quality Control (1995),American Customer

Satisfaction Index: Methodology Report. Milwaukee, WI:American SocietyofQuality Control.Anderson, Eugene W. (1994), "Cross-Category Variation In Cus-tomer Satisfaction andRetention,"Marketing Letters.5(Janu-ary). 19-30., Claes Fomell, and Donald R. Lehmann (1994), "CustomerSatisfaction, Market Share, and Profitability: Findings FromSweden." Joumat of Marketing.58(January), 53-66 ., . andRoland T. Ru.st (1996). "Customer Satisfac-tion, Productivity, and Profitability: Differences BetweenGoods and Services," w orking paper. N ational QualityResearch Center, Ann Arbor.Ml.

Andrews. Frank M. (1984). "Construct ValidityandError Compo -nentsof Survey Measures:AStructural M odeling Approach ,"Puhlic Opinion Quarterly.48, 409^2.

Bu/.^ell. Robert D.. John A. Quelch,andWaiterJ.Salmon (1990),"The Costly Bargain of Trade Promotion," Hurvard BusinessReview. Reprint No. 9020 .6S (October/November). 3.^^5.

Cronbach, LeeJ. andPaulE.Meehl (1955), "Construct ValidityinPsychological Tests," Psychological Bulletin,52(4), 2K1-302.Fornell, Claes (1992), "A National Customer Satisfaction Barome-ter:TheSwedish Experience,"Journal of M arketing,56(Janu-

ary),6 - 2 1 .

(1995), "TheQualityofEconomic Output: Empirical Gen-eralizations About Its Distribution andAssociation to MarketShare,"M arketing Science, 14 (3). G2O3-II.and Fred L. Bookstein (1982). TwoStructural EquationModels: LISREL and PLSApplied to Con.sumer Exit-VoiceTheory," Joumai of Marketing Research, 14 (November),440-52., Christopher D. Ittner, and David F. Larcker (1995),"Understanding and Using the American Customer SatisfactionIndex (ACSI): Assessing the Financial ImpaciofQuality Initia-tives,"Proceedings of the Juran institute s Conference on Man-agingfor otalQuality, forthcoming., , and (1996), The Valuation Conse-quencesof Customer Satisfaction," National Quality ResearchCenter Working Paper. National Quality Research Center. AnnArbor,Ml.and Michael D . Johnson (199 3), "Differentiation asa Basisfor Explaining Customer Satisfaction Across Industries," Jour-na lofEconomic Psychology, 14 (4). 681-96.and Birger Wemerfell (1987), "Defensive Marketing Strat-egy byCustomer C omplaint Management,"Journal of Market-ing Research.24(November). 33746.Fortune(1993), "Why the Economic Data Mislead Us," (March8),108-14.(1995), "Americans Can't Gei NoSatisfaction." (Decem-ber II), 178-94.


12/13

Hirschman. Albert O. (1970),Exit. Voice and Loyalty Responsesto Decline inFirms. Organizations and States. Cambtidge,MA: Harvard University Press.Howard, John A. (1977),C onsumer Behavior: ApplicationofThe-ory.New York: McGraw-Hill.Inner, Chri.stopher D. and D avid F. Larcker (1996), "Me asuringtheImpactofQua lity InitiativesonFirm Financial Performance,"inAdvances in the Management of Organizational Quality.Vol.1, Soumeh Ghoshand Donald Fedor, eds. Greenwich, CT:JAIPress, 1-37.lohnson, Michael D.,Etigene W.Anderson, andClaes Fomell(1995), "RationalandAdaptive Performance Expectationsin aCustomer Satisfaction Framework," Journal of ConsumerResearch 2\ (March) , 28^0.and Claes Fomell (1991). AFramework forComparingCustomer Satisfaction Across Individuals andProduct Cate-gories," Joumal of Economic Psychology 12 (2), 267-8 6.Kahneman. Daniel andAmos Tversky (19 79), "Prospect Th eory:An Analysisof Decision Under Risk." Econometrica 47 (2),263-91.

Lancaster. Kelvin (1971). Consumer Demand:ANewApproachNew York: Columbia University Press.Meyer, Anton (1994). DasDeutsche Kundenbarometer 199Munchen. Germany: Ludwig-Maximilians-UniversitaMiinchen.National Quality Research Center (1995),American Customer Satisfaction Index: Methodology Report. University of MichiganBusiness School. Milwaukee, WI: American Society forQuality Control.Oliver, Richard L. (1980), ACognitive Modelofthe A ntecedentand ConsequencesofSatisfaction Decisions," Joumal of Marketing Research 17(November), 46 0-69.Reichheld, Fredrick F. andWEarl Sasser (1990 ), "Zero Defections: Quality ComesToServices," Harvard Business Review68 (September/October). 105-11.Wold, Herman (1989), Theoretical Empiricism:AGeneral Rationale for Scientific Model Building.New York: Paragon House.Yi,Youjae (19 91). "A Critical ReviewofCustomer Satisfaction,"inReview of Marketing 1990 Valerie ZeithamI ed. ChicagoAmerican Marketing Association, 68-123.

Structural Equa tions Modeling ha s never been easier QS5.3revolutionizespath analysis.Path analysis, confirmatory factoranalysis (CFA),andLisrel-typemod-els have become dominant in thefieldof Marketing. You can now easilyper-form these analyses with DiagrammerEQS' new modeling tool.

Use path diagramasmodel input outputEQS5.3allowsyou todraw presen-tation-quality path diagrams liketheone above and automatically runs yourmodel accordingto thediagram.Youneed noprior knowledge of com-mand languageor of EQS tobuilda model. EQSalso providesawiderange ofstatistical anddata explora-tion tools, allowing researchers tobet-ter understand their dataand formu-lateanoptimal model.

Multivariate Software Inc.4924 Balboa Blvd., #368EncincCA 91316(800) 301-4456e-mail: [email protected]: http://www.mvsoft.com

Performanceand JobSatisfaction Model

E6

Modeling eatures Diagram draw ing facility' Automatic EQ S command generator Comprehensive data manager Normal theory estimatesEstimates with elliptical corrections' Saton-a-Bentler scaled .statistics Asym ptotic distribution free' Multi-.sample analysisAnaiysis of means & covahanc es Polyserial/polychoric correlation R b lb S T function for faster convergence' Effect deco mposition

Built-in simulation facilityCut and paste data from spreadsheet Missing value diagnosis and imputations Multivariate LM test to improve model Basic stati.stics(t tests. ANOVA,descriptive, frequency, cross tabs, varianceand covariance m atrix, multiple andstepwise regression, factor analysis)

'Datae xplora tion plots (line, area,histogram, quantile-quantile,normalprobability, scatter, matrix, box.Pie,bar,error har charts)


13/13

+a.1996.the american customer satisfaction index

Documents