COMPARING PERCEPTIONS OF SERVICE QUALITY IN BRAZIL AND UK

This paper investigates whether the SERVPERF model developed by Cronin and Taylor (1992) can be applied in the same industry (retail banks) in two different countries Brazil and UK. We find differences between the countries in the perception of the service provided by retail banks. We also find differences between the countries in the relationship between the service provided and (i) the overall perception of service quality, (ii) customer satisfaction and (iii) future purchase intentions. Factor analysis of the data identified different dimensional structures for Brazilian and British samples. The Brazilian sample presents almost the same structure of Parasuraman et alii (1988) SERVQUAL. The UK sample presented a different three-factor dimensional structure which shows that the dimensional structure does not travel well into other cultures.Also, as this study stands, the differences found between the two samples cannot yet be attributed to cultural differences between UK and Brazil. These differences may be related to the service provided by the banks.


INTRODUCTION
Service industries account for about 70% of the national income in most developed countries.Services are not only important in terms of creation of wealth but in terms of employment.According to Heskett (1987), 75% of non-farm jobs are in the service sector in the USA.Cronin and Taylor (1992) mention that 85% of all new jobs created since 1982 in the US have been in service industries.This trend is not limited to USA.Services industries employ 55% to 75% of the total work force in developed countries (UN, 1995).The importance of the service sector is not only increasing in developed countries, but also in less developed countries (LDCs).The main causes for this growth are tourism and banking sectors.
Service quality assessment has been the major issue for many researchers in the last twenty years.Few measurement instruments in the whole marketing area have received more attention thanSERVQUAL, developed by Parasuraman et alii (1988Parasuraman et alii ( , 1991))  objective of the present work is to observe the performance of an alternative service quality measurement scale alternative, SERVPERF (Cronin and Taylor 1994), in Brazil and UK.

LITERATURE REVIEW
One of the main schools of thought on service quality is the "Gap School", based on thedis confirmation of expectations paradigm (Oliver, 1977(Oliver, , 1980) ) which forms the conceptual basis of the SERVQUAL model developed by Zeithamlet alii (1990).For Parasuraman et alii (1988), "service quality, as perceived by customers, can be defined as the extent of discrepancy between customer's expectations or desires and their perceptions" (p.19).Parasuraman et alii (1988) found that customers assessed service quality through five dimensions: tangibles, reliability, responsiveness, assurance and empathy.
SERVQUAL has attracted criticism in a number of areas.Buttle (1996) and Smith (1995) provide helpful summaries of the major areas of perceived weakness.Problems are concentrated in two main areas.One of these problems is that, despite the large number of replication studies conducted (Carman, 1990, Babakus and Boller, 1992, Babakus and Mangold, 1992, Schneider et al, 1992, Gagliano and Hathcote, 1994, and Boshoff et al, 1995 ), researchers have consistently failed to replicate the five dimensional structure of service quality proposed by Zeithamlet alii (1990).In addition, the use of the disconfirmation paradigm as the theoretical basis for SERVQUAL has also attracted criticism (Carman, 1990, Cronin and Taylor, 1992, 1994Teas, 1993and Brown et al, 1993).Cronin and Taylor (1992), suggest that customers have expectations towards a performed service, but that expectations do not form "consumers' perceptions of service quality" (p.57).They suggest that perceived performance is the most appropriate measure of service quality and that the performance minus expectations construct is an inappropriate basis for the measurement of service quality (Cronin and Taylor, 1994, p. 125).Nevertheless, they agree that SERVQUAL adequately covers the domain of service quality, but included three more measures on future purchase intentions, overall quality and satisfaction.They modified the SERVQUAL model and thus, created their own model, SERVPERF, in which performance perceptions are used as measure of service quality.Cronin and Taylor (1994)   It is important to notice that all these studies have been conducted either in the USA or in Europe.Few studies have been conducted cross-nationally (Sjolander, 1992 andCollin-Dodds, 1996).The research discussed in this paper replicated the SERVPERF scale in Brazil and UK using one of the original industries, retail banking.The chief objective of the research was to examine the cross-cultural variation in service quality determinants.
Another objective was to establish whether service quality, customer satisfaction and purchase intention are linked and in which way, comparing with the findings of Cronin and Taylor (1992).

METHODOLOGY
The questionnaire used in this research was SERVPERF (Cronin & Taylor 1992).
This is basically the revised SERVQUAL measurement instrument (Parasuramanet alii, 1991) which wassubsequently modified by Cronin and Taylor (1992) to include repeat purchase intention, customer satisfaction and overall perception of service quality.Cronin and Taylor's questionnaire is basically a version of SERVQUAL in which only the measure of performance was retained.
The questionnaire consisted of twenty-two ( 22) items related to the measurement of service quality using a Likert seven point scale ranging from disagree strongly (1) to agree strongly (7).The questionnaire was translated into Brazilian Portuguese by a bilingual Brazilian translator.Back-translation by a bilingual British translator into English has ensured that the questionnaire is equivalent in both languages.
The selected cities (Rio de Janeiro -Brazil, and London -UK) were chosen for convenience and due to the significance of the markets (both cities have over 6 million inhabitants).The sample for this research consisted of customers of retail banks in Brazil and UK.The banks were chosen not only due to a comparable number of clients (both are rated 62 among the largest retail banks in their respective countries), but also due to a comparable use of technology.Both banks wished to remain anonymous, and will be referred to as UK Bank and Brazilian Bank.The sample chosen was a convenience one due to restrictions posed by both banks.According to Sekaran (1983), problems due to a non-random sample can be overcome by the use of matched samples in a cross-cultural study.Quotas were established in order to control some of the bias of a non-probabilistic sample.A hundred customers of each bank were selected at random from a sample of two UK Bank branches and three Brazilian Bank branches.The branches were selected by the banks themselves and represented small and large branches according to number of customers (UK) and number of cashiers (Brazil).
The interviews with customers occurred inside the branches, during July and August 1996, lasting on average five minutes each.A hundred usable questionnaires were collected from UK and 101 usable questionnaires were collected in Brazil.
The classification of socio-economic classes used is the one defined by the Central Statistical Office (CSO) in the General Household Survey and the same categories were applied when analysing the Brazilian sample in other to match the sampling criteria.Social classes in Brazil take intoconsideration also the ownership of consumer goods such as VCRs and TVs.Table 1 below shows the breakage of the sample used.The principal analysis made on the data was factor analysis.In this case, the underlying dimensions were expected to be the same as in SERVQUAL: tangibles, reliability, responsiveness, assurance and empathy.
In order to replicate the Cronin and Taylor study (1992), the method used for extraction of factors was the "oblique" method.The criteria used to establish the number of factors extracted was the Latent Root Criterion, or extracting as final factors any factor with eigenvalues over 1.

63
After the extraction of factors in both samples (UK and Brazil), correlation among variables and means related to gender, age, branch and social class were also analysed.All numbers retained for analysis had a < 0.05 or 5%.
In addition, comparison of means, correlation and multiple regression were used to identifyrelationships between service quality and future purchase intentions and customer satisfaction.

DATA ANALYSIS
The variables from SERVPERF instrument were named P variables, for "Performance" and will be referred as P1, P2…according to question number, e.g.P1 is associated with statement 1 in the questionnaire (Bank has modern looking equipment).A table linking variable codes and question texts can be seen in the Appendix.

COMPARISON OF MEAN SCORES
Among the three most highly rated variables, two are common to both banks: P15 (you feel safe in your transactions with UK Bank/ Brazilian Bank) and P16 (employees of UK Bank/ Brazilian Bank are consistently courteous with customers) seems to point that both banks give attention to the assurance level.The other most highly rated variable is different for UK and Brazil, but both are on the tangibles dimension: in UK, P3 (employees are neat appearing) and in Brazil, P1 (the bank has modern-looking equipment).It is interesting to note also that variables related to the reliability dimension which was considered by a Forum Corporation research as the most critical aspect for customers (Fojt, 1995) are not rated highly by customers of either bank.(See Appendix for a complete list of variable names and codes, and sample means).

FACTOR ANALYSIS RESULTS
Factor analysis was used in order to check Cronin and Taylor´s SERVPERF dimensional structure for the Brazilian and British samples.The method for extraction of factors used was the latent root criterion (eigenvalue).According to Hair et alii (1992), the use of eigenvalue to establish the number of factors is probably the most reliable method when having 20 to 50 variables to extract factors from.
The UK sample loaded only three factors while the Brazilian one loaded five factors in a way similar to SERVQUAL.
In the UK sample, Factor 1 encompasses all variables from the reliability, responsiveness and assurance dimensions.In addition, it also includes P3, P18 and P20.
Factor 2 includes two variables from the tangibles dimension (P1 and P2) and one from empathy (P19).Factor 3 comprises two variables from empathy (P21 and P22) and one from tangibles dimension (P4).In this case, it can be said that factor one is related to dependability of the service (reliability, responsiveness andassurance), factor 2 is more related to tangibility and factor 3 to empathy.

MULTIPLE REGRESSION
Multiple regression was used to examine the effect of the extracted factor scores (the independent variables) on three dependent variables: future purchase intentions (FRQUENCY), overall service quality (QUALSERV) and customer satisfaction (SATISFCN).The method used for multipleregression was stepwise.This was done not only due to the fact that was the method used by Cronin and Taylor (1992), but also because this method allows the independent variables to be entered one by one and eliminating unsuitable ones.The factor scores used were a simple mean score of the items loading on the factor in question.Scores on each variable were summed and then divided by the number of variables in the factor.In the Brazilian sample, for instance: Factor 1 = (P5 + P6 + P7 + P8 + P9 + P10 + P15) / 7 Factors were named Factor1, Factor2 and so on for computations sake.The beta values (B) indicates the value of each factor in the regression equation.
The equation for overall service quality is: 0.74 + 0.57 (dependable) + 0.35 (empathy).Factor 1 is the most powerful predictor of overall service quality.

c) Dependent variable: customer satisfaction
Once more, only factors 1 (Dependability) and 3 (Empathy) loaded satisfactorily into multiple regression model.Factor 2 (Tangibles) was not considered significant enough to be included, i.e.T significance = 0.3243.
The model presented a good goodness of fit, since R2 denotes 70%.This indicates that customer satisfaction is dependent on factors 1 (Dependability) and 3 (Empathy) and its value can be predicted from those factors.
The results above show that the regression equation for customer satisfaction is:-0.27+ 0.74 (dependable) + 0.36 (empathy).The numbers above indicate that Factor 1 is the best predictor of customer satisfaction.
As only factors 1 (Dependability) and 3 (Empathy) loaded satisfactorily into the models, this points to the fact that factors 1 (Dependability) and 3 (Empathy) seem to be best indicators of overall service quality and customer satisfaction since R2 for future purchase intention is 27%.This indicates that customer satisfaction is dependent on factors 2 and 4 and its value can be predicted from those factors.
The beta values (B) indicates the value of each factor in the regression equation.
Factor 4 (Assurance) is the most powerful predictor of overall service quality.

• Dependent variable: overall service quality
Only factors 2, 4 and 5 loaded satisfactorily into the model.Factors 1 and 3 were not considered significant enough to be included, i.e. factor 1 had a T significance of 0.1717 and factor 3 0.5854.The model displayed a good goodness of fit, since R2 is 69%.This indicates that the value of overall service quality can be predicted by the values of factors 2, 4 and 5.
The results above show that the regression equation for customer satisfaction is Y: -0.57 + 0.19 (tangibles) + 0.78 (assurance) + 019 (empathy).The numbers above indicate that Factor 4(Assurance) is the best predictor of customer satisfaction.
• Dependent variable: customer satisfaction Once more, factors 2, 4 and 5 loaded satisfactorily into the model.Factors 1 and 3 were not considered significant enough to be included, i.e. factor 1 had a T significance = 0.1287 and factor 3, 0.4131.The model presented a good goodness of fit, since R2 denotes 63%.
This indicates that customer satisfaction is dependent on factors 2, 4 and 5 and its value can be predicted from those factors.
The results above show that the regression equation for customer satisfaction is Y: -0.14 + 0.23 (tangibles) + 0.47 (assurance) + 0.41 (empathy).The numbers above indicate that both factors 4 (Assurance) and 5 (empathy) are the best predictors of customer satisfaction.
Overall, only factors 2 (Reliability), 4 (Assurance) and 5 (Empathy) are significantly related to the dependent variables.As the model displayed a good goodness of fit for service quality (R2 = 69%) and customer satisfaction (R2 = 63%), this indicates that the value of overall service quality and customer satisfaction can be predicted by the values of factors 2, 4 and 5.
Another study would be necessary to establish if Future Purchase Intention can be predicted by Service Quality and Customer Satisfaction.

DISCUSSION
This study shows that there are differences in the perception of service quality between the two countries.Not only the comparison of means could spot significant difference between the samples but also the factor analysis spotted differences.Comparing 60

61(
as measured by chi square statistic and the model's adjusted goodness of fit) than SERVQUAL.In 1994,Cronin and  Taylor asserted that "SERVPERF has greater construct validity based on the review of relevant literature and the fact that the SERVPERF measures also exhibit convergent and discriminant validity" (p.129).

Revista Eletrônica de Negócios Internacionais da ESPM
showed that SERVPERF model has a better fit

Table 2 -Mean Scores of the Most Highly Rated Variables
by the Brazilian sample.The other variable with a very low rating from the Brazilian sample was P13 (Employees of XYZ are never too busy to respond to your requests).

Table 5 -
R2, Factor Significance and Size -UK Sample

Table 6 -
R2, Factor Significance and Size -Brazilian Sample