Refinement and reassessment of the SERVQUAL scale. (multiple-item scale for measuring service quality)

In a previous article we presented SERVQUAL, a multiple-item scale for measuring service quality (Parasuraman, Zeithaml, and Berry 1988). In the present article, we discuss findings from a follow-up study in which we refined SERVQUAL and replicated it in five different customer samples. We also compare our findings with those of other researchers who have recently employed and evaluated SERVQUAL. On the basis of insights from this comparative discussion, we offer directions for future SERVQUAL research and applications.

In an article in the Spring 1988 issue of the Journal of Retailing, we discussed the development, testing, and potential applications of SERVQUAL, an instrument for measuring customer perceptions of service quality (Parasuraman, Zeithaml, and Berry - hereafter, PZB - 1988). To our knowledge, many proprietary service quality studies have productively used SERVQUAL. A number of published studies have also invoked the SERVQUAL framework (Crompton and Mackay 1989; Webster 1989; Woodside, Frey, and Daly 1989; Johnson, Dotson, and Dunlap 1988) and have assessed the scale's reliability and validity (Babakus and Boller 1991; Brensinger and Lambert 1990; Carman 1990; Finn and Lamb 1991).

The objectives of this paper are to: (i) describe a multi-sector study in which we refined our original SERVQUAL instrument and reexamined its reliability and validity; (ii) provide a comparative discussion of insights from our study and those from other SERVQUAL-replication studies; and (iii) offer directions for future research and SERVQUAL use.

In the following sections we first describe the study in which we replicated SERVQUAL. We then discuss refinements made to SERVQUAL and our rationale for them. Next, we examine the reliability, factor structure, and validity of the revised instrument. Finally, we discuss SERVQUAL-related issues emerging from the various replications and offer suggestions for future applications of the instrument.

THE STUDY

The findings discussed in this article are from a study in which customer assessments of service quality were measured for three types of services: telephone repair, retail banking, and insurance. With the exception of retail banking, these services are different from the services originally studied to develop SERVQUAL (PZB 1988). In the present study, SERVQUAL was used to measure the service quality of five nationally-known companies - one telephone company, two insurance companies and two banks. Following a pretest and refinement, to be discussed in the next section, a questionnaire containing the modified instrument was mailed to about 1,800 to 1,900 randomly chosen customers of each company. A reminder post card was sent two weeks after the questionnaire was mailed. The number of complete questionnaires returned ranged from 290 to 487 across the five companies (response rates ranged from 17 to 25 percent). The aggregate sample contained 1,936 customers for a combined response rate of 21 percent. Demographic profiles of the five customer samples were reviewed by managers in the respective companies and considered to be representative of their customer bases.

REFINEMENT OF SERVQUAL

The original version of SERVQUAL (PZB 1988) consists of two sections: a 22-item section to measure customers' service expectations of companies within a specific sector (e.g., banking) and a corresponding 22-item section to measure customers' perceptions of a particular company in that sector. This original version of SERVQUAL was pretested through a mail survey of a regional sample of 300 customers of the telephone company that participated in our study. Completed questionnaires were returned by 68 customers for a response rate of 23 percent.

Using the pretest data, means and standard deviations for the individual SERVQUAL items were computed, as were reliability coefficients (coefficient alphas) for the perception-minus-expectation gap scores making up the five SERVQUAL dimensions of tangibles, reliability, responsiveness, assurance, and empathy. Examination of these pretest results suggested several refinements as discussed next.

The means for a majority of the 22 items in the expectations section of SERVQUAL were above a 6" on the 7-point, "Strongly Disagree" to "Strongly Agree" scale (the overall mean expectation score was 6.22). These high mean values were not totally unanticipated because the items were intended to measure customers' normative expectations. For example, one expectation statement from the original instrument read: "Telephone companies should keep their records accurately." Recognizing that the "should" terminology in this and other expectation statements might be contributing to unrealistically high expectation scores, a slightly different wording was adopted. The revised wording focused on what customers would expect from companies delivering excellent service. To illustrate, the aforementioned statement was modified to read: "Excellent telephone companies will insist on error-free records." The instructions pertaining to the expectations items were also appropriately changed. The perceptions section of SERVQUAL remained unchanged except for minor wording changes in several items to make them parallel to the revised expectations items.

Of the 22 SERVQUAL items included in the pretest questionnaire, sixteen were worded positively and six were worded negatively. The pretest results indicated that the negatively worded items may be problematic for several reasons. First, the standard deviations for the negatively worded expectations items were consistently higher (mean standard deviation of 2.07) than for the positively worded expectations items (mean standard deviation of 0.77). The wider variation for the negatively worded items implied that respondents may have been confused by those items. Second, managers in the five companies who reviewed the preliminary questionnaire felt that negatively worded expectations statements were awkward and not as meaningful as the positively worded items. Third, the reliability coefficients (Cronbach's alphas), reported in the second column of Table 1, were consistently lower than in the original SERVQUAL study (Table 2, PZB 1988) for responsiveness and empathy-the two dimensions that included all the negatively worded items. For these reasons, negatively worded items were all changed to a positive format in the final questionnaire.

 
TABLE I 
  Reliability Coefficients (Alphas) for the SERVQUAL Dimensions 
                                           Final Study Results 
                 No. of   Pretest   Tel.   Ins.    Ins.    Bank   Bank 
  Dimension      Items    Results   Co.    Co. 1   Co. 2     1      2 
Tangibles          4       .60      .83     .80     .84    .85    .86 
Reliability        5       .85      .88     .92     .92    .92    .88 
Responsiveness     4       .61      .91     .92     .93    .92    .88 
Assurance          4       .81      .89     .87     .91    .90    .87 
Empathy            5       .66      .87     .85      89    .88    .87 

Two new items - one each under tangibles and assurance - were substituted for two original items to more fully capture the dimensions and to incorporate suggestions from managers who reviewed the pretest questionnaire. An original tangibles item, "The appearance of the physical facilities of telephone companies should be in keeping with the type of services provided" (an item that was felt to be confusing and also redundant with another item dealing with appearance of physical facilities), was replaced with, "Materials associated with the service (such as pamphlets or statements) will be visually appealing in an excellent telephone company. " The original SERVQUAL instrument had no item pertaining to the appearance of communication materials, an omission that managers believed should be corrected by adding the new item. Also, an original assurance item, "Telephone company employees should get adequate …