Asymmetric relationship between attribute performance and customer ...

International Journal of Hospitality Management 29 (2010) 559–569

Contents lists available at ScienceDirect

International Journal of Hospitality Management journal homepage: www.elsevier.com/locate/ijhosman

Asymmetric relationship between attribute performance and customer satisfaction: A new perspective Lisa Slevitch a,*, Haemoon Oh b,1 a b

School of Hotel and Restaurant Administration, Oklahoma State University, 210X HESW, Stillwater, OK 74078-6173, United States Department of Hospitality and Tourism Management, University of Massachusetts, Flint 203C, United States

A R T I C L E I N F O

A B S T R A C T

Keywords: Asymmetrical effects Attribute-level performance Customer satisfaction Performance optimization Two-factor theory

The objectives of this study were two-fold. First, this inquiry attempted to provide additional support to the studies conceptualizing the relationship between attribute-level performance and overall satisfaction as non-linear or asymmetric. Second, the study aimed to provide an explanation to the observed asymmetry, thus addressing the gap in the previous research in the area. Asymmetric response of customer satisfaction to different types of attribute performance was tested and interactions between attributes were examined as an explanation for the observed asymmetry. Results of the study confirmed the non-linear nature of the customer satisfaction function. Moderating effects of attribute type explained the asymmetrical relationships between attribute performance and customer satisfaction, thereby providing theoretical rationalization to the observed, but often ignored, phenomenon. ! 2009 Elsevier Ltd. All rights reserved.

1. Introduction Customer satisfaction (CS) is a central issue in the hospitality field due to its imperative role in organizational performance and, ultimately, in the survival of hospitality companies. Research on CS has been proportionate to its growing managerial importance (Fournier and Mick, 1999; Oh and Parks, 1997). Much of the existing CS literature focuses on understanding the process of performance evaluation and identifying sources of CS (Knutson et al., 2003). While some definitional and technical issues are to be further resolved, most researchers agree that CS is best evaluated on a multi-attribute scale. That is, CS depends on a number of determinants at an attribute level and measuring CS through attribute-level performance captures the multifaceted nature of consumption experience. The multi-attribute approach prevailing in the area of CS evaluation provides several advantages (Oliver, 1997; Yi, 1990). It allows a higher level of specificity and diagnostic usefulness than merely measuring overall satisfaction. An attribute level of satisfaction measurement captures CS with a specific aspect or dimension and can be aggregated into an overall satisfaction score, thereby providing specific constructive feedback to management for action strategies (Szymanski and Henard, 2001).

* Corresponding author. Tel.: +1 405 744 7643; fax: +1 405 744 6299. E-mail addresses: [email protected] (L. Slevitch), [email protected] (H. Oh). 1 Tel.: +1 413 545 2061. 0278-4319/$ – see front matter ! 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijhm.2009.09.004

Most multi-attribute CS studies have conceptualized the relationship between attribute-level performance and satisfaction as linear or symmetric (Oliver, 1997). In these studies, an implicit assumption was that adequate performance at an attribute level would lead to a certain level of satisfaction, which was counterproportionate to the level of dissatisfaction if the attribute performance was equally inadequate. In other words, under the linear relationship assumption, unreliable performance of a hotel service staff leads to dissatisfaction, whereas reliable performance leads to satisfaction with the same magnitude of positive or negative impact. Yet, this symmetric view has been criticized and shown as not always holding true (Cadotte and Turgeon, 1988; Hui et al., 2004; Johnston, 1995; Maddox, 1981; Mittal et al., 1998; Swan and Combs, 1976). Several studies indicate that an asymmetrical relationship exists between attribute-level performances and CS (Ammar et al., 2008; Backhous and Bauer, 2000; Hui et al., 2004; Oliva et al., 1992; Matzler and Sauerwein, 2002). For example, stained linens in a hotel room can cause serious dissatisfaction, but will not cause the same degree of satisfaction when presented clean. A complimentary chocolate can produce high satisfaction but, when not offered, will not entail any negative effect on CS. It is commonly agreed that accurate understanding of CS is imperative but has yet to be accomplished (Babin and Darden, 1996; Oliver, 1997). Despite the abundance of research on CS, complexities beyond the linear relationship between CS and attribute performance remain largely unexplained. More importantly, in spite of growing number of studies supporting

560

L. Slevitch, H. Oh / International Journal of Hospitality Management 29 (2010) 559–569

asymmetric responses of CS to attribute-level factors, no explanation to the observed asymmetry has been provided. This study aim to provide such explanation and thus enrich the existing theoretical knowledge on CS. Substantial empirical evidence indicates that the relationship between attribute performance and CS is inconsistent and varies in its strength and direction (Anderson and Mittal, 2000; Johnston et al., 1990; Johnston, 1995; Mersha and Adhlaka, 1992). When there is such an unexpected, inconsistent relationship between a theoretical predictor (attribute performance) and criterion (satisfaction), some unspecified moderator effects are likely to exist, especially in cases where a relationship holds in one setting but not in another (Baron and Kenny, 1986). Little is known about potential moderating effects of the attributes as determinants of satisfaction. However, examination of moderator effects is important because it would allow a more accurate understanding of the satisfaction process by distinguishing between the effects that attributes produce and the way they interact in the consumers’ judgments. Accounting for asymmetry and moderator effects is essential from a practical standpoint as well. Attribute-level performances should be optimized, not simply maximized, to increase overall satisfaction. Both academics and practitioners have reported that equally investing in greater performance along all service attributes in order to increase satisfaction would not be effective and not justify additional investments (Cadotte and Turgeon, 1988; Hui et al., 2004; Kano, 1984; Matzler et al., 2004; Matzler and Sauerwein, 2002; Mittal et al., 1998; Ting and Chen, 2002). The existing system of measuring CS however is not tailored for performance optimization. Rather, it strives for maximization. It is based on the assumption that the nature of the relationship between attribute-level performance and satisfaction is linear or symmetric and, therefore, it is assumed that the better the performance on a certain attribute, the more it contributes to overall satisfaction. Though strong support of non-linear nature of CS function exists, no explanation to the observed effects was provided (Deng et al., 2008; Kano, 1984; Matzler and Sauerwein, 2002). This study addressed this gap by introducing interactions among attributes as the cause of this phenomenon. Additionally, this study examined attribute performance optimization in the lodging settings and looked at the complexity of the relationship between factors that play significant roles in customers’ minds during assessment process. 2. Background 2.1. Conceptualization of attributes Marketing literature defines attributes as dimensions of a product or service. All products and services are viewed as a bundle of attributes or features that influence consumer choice (Kotler et al., 2003). When consumers evaluate an offering, their opinion is extensively affected by the performance of attributes associated with the offering (Oliver, 1997). Many attempts have been made to categorize the attributes’ dimensions in a way that would be applicable across industries. One of the most commonly used views on product or service attributes describes attributes in the ring model proposed by Levitt (1983). The innermost ring of this model defines core attributes of the offering, the ‘‘musts’’ that encompass consumer expectations for what the basic offering should constitute. The outer rings, delights or satisfier (or ‘‘facilitators’’), support and enhance core attributes. This outer group of attributes constitutes embellishments to the consumer’s standard set.

Similarly to Levitt (1983), Kano (1984) divided attributes into three major categories: basic factors (minimum requirements that are fully expected by customers) and performance and excitement factors (attributes with high impact on CS if delivered). One major limitation of these similar conceptualizations is instability of attributes’ positions within the frameworks. Competitive actions often cause the content of the outer rings to migrate into the center over time, as all competitors replicate satisfiers/ delights or excitement factors, thus raising consumer standards and expectations for what a basic offering should constitute (Oliver, 1997). Some researchers tried to develop alternate views to the traditional ring model. Gronroos (1984) and Lehtinen and Lehtinen (1991) proposed to categorize attributes into two groups: process or technical quality and outcome or functional quality. Process quality attributes capture how customers receive services. Outcome quality attributes refer to what customers receive as a result of obtaining services. Though having received support from many scholars, this categorization has been criticized because it has remained largely on a theoretical level without any strong empirical evidence validating these dimensions (Ekinci, 2002). Swan and Combs (1976) applied a different approach and categorized attributes into two categories: instrumental and expressive. Instrumental attributes refer to the physical performance of an offering, such as the level of bed comfort; expressive attributes derive from psychological performance, such as personnel friendliness. Swan and Combs further suggest that expressive attributes tend to have a greater positive effect on satisfaction when performed well. Low performance on instrumental attributes leads to a more pronounced negative effect resulting in dissatisfaction. Though providing some advances, Swan and Comb’s study had limitations. For example, it focused on products only and may not be generalizable to services. Chowdhary and Prakash (2005) took a slightly different approach and classified attributes as vantage or qualifying depending upon the state and nature of competition. According to them, attributes can be divided into the vantage factors that enable offerings to qualify for competition or the qualifying factors that put companies above competition by motivating customers to consume the offering as opposed to that of the competitor. Though this classification provided a distinct view on the division of attributes, it appeared analogous to the core/ facilitating attribute topology in that, when the performance of the qualifying factors was below the acceptable level, consumers felt dissatisfied regardless of the performance of the vantage factors. To summarize, all described attribute conceptualizations have limitations. Only Levitt’s (1983) ring model and the classifications by Gronroos (1984) and Lehtinen and Lehtinen (1991) provide theoretical explanations as to why attributes should be categorized in the way they were proposed. However, because the latter classifications lack supporting empirical evidence, Levitt’s categorization appears to have a stronger position than the other topologies and is employed in this study. 2.2. Attribute-level performance and CS The attribute-based approach involving importance-performance analysis prevails in CS literature (Matzler and Sauerwein, 2002). In this approach, a list of key product or service attributes is generated first and, then, consumers are asked to rate an offering with reference to how each attribute on the list is delivered. Additionally, consumers are asked to rate attributes in terms of their importance. An overall attribute score is then computed as


the sum of individual attribute scores weighed by the level of their importance (Oliver, 1997). The attribute-based approach like this is popular because the overall CS measurement approach tends to mask specific product or service problems (Oliva et al., 1992). However, the attributebased evaluation reduces such problems and provides researchers and practitioners with a higher level of specificity and diagnostic feedback about the company’s performance. In spite of its popularity, the attribute-level approach has weaknesses. Incorporating importance into the assessment process involves ambiguities to some degree (Oliver, 1997). Often when consumers evaluate attributes, it is not clear whether the attribute is important for its presence or its absence. Consumers may also become acclimated to the point of indifference to some attributes that may be important but are normatively incorporated into all offerings in the market and, thus, are becoming an industry standard. For example, do many consumers ponder if there is a towel in their hotel room? This last example points to one potential limitation of the attribute-level approach: it does not explain what happens when an attribute becomes less important for either differentiation or competition because all firms offer this feature to survive in the market. Oliver (1997, p. 37) refers to this situation as a paradox because once an attribute becomes an industry norm, it will become even more important or critical as a dissatisfaction driver for an offering failing to deliver this feature. The paradox situation observed by Oliver (1997) induced several researchers to question the prevailing assumption that the link between attribute-level performance and satisfaction is symmetrical. Apparently, the above illustration for the paradox situation implies that, depending on its market demand status, an attribute may be related to either satisfaction or dissatisfaction to a different magnitude, which gives rise to a plausible asymmetric relationship between certain attributes and CS. Although the question of the asymmetric attribute–CS relationship has led to several studies showing that the relationship between the performance and CS constructs could be non-linear (Mittal et al., 1998; Oliva et al., 1992), additional theoretical and empirical accounts are desirable, especially in the hospitality industry. 2.3. A non-linear relationship between attribute-level performance and CS The speculation of the asymmetric relationship attributes may have with satisfaction and dissatisfaction stems from the twofactor theory originally developed in the area of employee satisfaction by Herzberg (1967). Herzberg argued that two groups of factors, job content and job context, had separate and distinct influences on employee satisfaction. Job content factors (motivators) such as employee empowerment act only to increase satisfaction, whereas job context (hygiene factors) factors such as compensation only increase dissatisfaction with no or little effect on satisfaction. Based on the observed asymmetry of positive and negative effects of hygiene factors and motivators, Herzberg (1967) did not deduce that CS function was non-linear but moved in a different direction concluding that satisfaction was a uni-polar concept, implying that the opposite of satisfaction is no satisfaction, not dissatisfaction and that dissatisfaction is not the opposite of satisfaction. Several studies tested Herzberg’s (1967) theory and showed that consumers’ responses varied across satisfaction or dissatisfaction suggesting that things that satisfied customers were different from things that dissatisfied customers (Bitner et al., 1990; Cadotte and Turgeon, 1988; Johnston, 1995; Maddox, 1981; Mersha and Adhlaka, 1992; Swan and Combs, 1976). Particularly noteworthy was a study by Kano (1984) that in addition to demonstrating that

561

attributes can affect CS in asymmetric way, clearly pointed on nonlinear nature of CS function (Matzler et al., 2004). In congruence with Kano (1984), Mittal et al. (1998) suggested that satisfaction was affected asymmetrically by attribute-level performance: ‘‘although positive and negative performances on an attribute are two sides of the same coin, each side of the coin buys different amount of overall satisfaction’’ (p. 45). The authors also raised a very important question: ‘‘was the observed asymmetry related to the specific qualities of an attribute?’’ While leaving this question unanswered, the authors made references to the study by Kahn and Meyer (1991) that attributes could be utility-preserving, also known as core attributes (Levitt, 1983; Kotler et al., 2003) that seemed to have a stronger association with dissatisfaction, or utility-enhancing or facilitating attributes that had a higher potential for creating satisfaction. Although these studies provided strong evidence that attributelevel performance had asymmetric effects on CS, they offered no cogent theoretical explanation about the observed effects. This is an important gap in consumer behavior research because it is essential to understand the nature of such asymmetry and to answer why customers are very particular about inadequate performance of some factors but show indifference when these features performed adequately. Hui et al. (2004) offered some interesting theoretical explanations by examining interaction effects between variables as a possible cause of the observed asymmetry. Similar to Levitt (1983) and Chowdhary and Prakash (2005), Hui et al. (2004) categorized service attributes into two categories: outcome quality attributes, which were analogous to facilitating or vantage factors that promoted a product or service ahead of competitors and process quality attributes, which were equivalent to core or qualifying factors in that they were necessary to stay in a competitive pool. They examined if interactive effects between these two types of attributes would follow the pattern predicted by either the twofactor theory or the fairness heuristic theory in which fairness and trust in the provider should moderate the effect of attributes on repatronage intentions. Hui et al. (2004) found support for the twofactor view by showing a significant interactive effect between qualifying and vantage factors, but refuted the interaction pattern predicted by the fairness heuristic theory. Although notable, Hui et al.’s (2004) study has shortcomings. The authors considered only two service attributes in their experiment: the sales representative’s behavior (core attribute) as either favorable or unfavorable and speed of delivery (facilitating attribute) as either adequate or inadequate. Additionally, their study focused narrowly on service quality perceptions as the dependent variable. It is questionable whether their findings will be applicable to the entire spectrum of CS and related issues. The summary of the studies pointing on asymmetrical effects is presented in Table 1. 3. Research hypotheses In this study, service attributes were categorized into two groups, facilitating and core (Levitt, 1983), similar to vantage and qualifying attributes of service quality as conceptualized by Hui et al. (2004) and Chowdhary and Prakash (2005). This decision was primarily based on the fact that the core/facilitating categorization was most commonly used in the marketing area, specifically in the CS literature (Kotler et al., 2003). The core attributes are the attributes that are important to customers but have less pronounced positive effects on CS than facilitating attributes when they perform beyond the acceptable level (Chowdhary and Prakash, 2005; Oliver, 1997). Customers get acclimated to these attributes and perceive them as given. That is why their presence alone no longer attracts the consumer, but only

562


Table 1 Empirical studies demonstrating asymmetric attribute-level effects on CS. Authors

Attribute categorization

Methods

Asymmetrical effects

Major findings

Herzberg (1967)

Hygiene factors and motivators

Critical incident technique

Observed

Swan and Combs (1976) Maddox (1981)

Instrumental and expressive factors Instrumental and expressive factors

Critical incident technique Critical incident technique

Observed Observed

Kano (1984)

Regression analysis

Observed

Analysis of complaints and complements Critical incident technique

Observed

Bitner et al. (1990)

Basic, excitement, and performance factors Satisfiers, dissatisfiers, criticals, and neutrals Satisfiers, dissatisfiers, and neutrals

Two groups of factors affecting satisfaction asymmetrically Partial support of two-factor theory Partial support of two-factor theory of CS Three-factor model of service quality and CS; non-linearity of CS function Support of two-factor theory

Mersha and Adhlaka (1992)

Causes of good and bad service

Observed

Johnston (1995) Mittal et al. (1998)

Satisfiers and dissatisfiers N/A

Attributes rank order based on importance Critical incident technique Regression analysis

Backhous and Bauer (2000)

Value-enhancing and minimum requirements factors


Hui et al. (2004)

Outcome quality and process quality factors

Regression analysis

Observed and partially explained

Chowdhary and Prakash (2005)

Vantage and qualifying factors


Observed

Cadotte and Turgeon (1988)

qualifies for competition, as discussed earlier. To examine whether such tendency holds true in the lodging industry, we propose: H1. When both groups of attributes perform above the acceptable level, core attribute performance has a weaker positive effect on CS than facilitating attribute performance. On the other hand, core attributes have strong negative effects when their performance is unfavorable and their absence or lacking quality would alienate customers, because this group of attributes has a narrow zone of tolerance (Johnston, 1995). As suggested by the two-factor theory, inadequate core attribute performance tends to be strongly associated with dissatisfaction; however, when core attributes perform adequately, they have a decreasing or no effect on CS (Hui et al., 2004). As applied to lodging products, therefore: H2. Core attributes performing below the acceptable level have a greater impact on CS (i.e., dissatisfaction in this case) than core attributes performing above the acceptable level have on CS (i.e., satisfaction in this case). Customers’ evaluations are also substantially influenced by facilitating attributes as differentiators of a particular offering and, at the same time, consumers are more tolerant about inadequate performance of facilitating attributes in some situations (Cadotte and Turgeon, 1988; Oliver, 1997). Facilitating attributes tend to correspond to satisfaction when performing well, but have no or a decreased effect on CS when performing poorly. Thus, we predict: H3. Facilitating attributes performing above the acceptable level have a greater impact on CS (i.e., satisfaction) than facilitating attributes performing below the acceptable level have on CS (i.e., dissatisfaction). Several studies suggested that facilitating attributes tended to have a stronger positive effect on CS when core attributes performed above the expected level (Cadotte and Turgeon, 1988; Chowdhary and Prakash, 2005). When the strength or direction of relationships between constructs changes under

Observed

Observed Observed and examined Observed

Critical incidents associated with three factors Different causes of good and bad service identified Support of two-factor theory Negative performance has a greater impact on CS than positive performance Value-enhancing and minimum requirements factors have distinct effects on CS Process quality factors moderate relationships between outcome quality factors and CS Support of two-factor theory

certain conditions, it is a strong indication of the presence of an interaction between the constructs, or moderator effects (Baron and Kenny, 1986). To this end, we proposed the following hypothesis to test the moderator effect of core attributes on the relationship between facilitating attributes and CS: H4. Core attribute performance moderates the effect of facilitating attribute performance on CS. Specifically, facilitating attribute performance has stronger effects on CS when core attribute performance is above the acceptable level than when core attribute performance is below the acceptable level. 4. Methods The sample of the study comprised of the 5567 faculty and permanent full-time staff of a large U.S. university (no students included). Such sample was considered acceptable for purposes of this experimental study because usage of homogeneous samples tend to reduce the likelihood of extraneous variables adversely affecting the study results (Greenberg et al., 1987; Reynolds et al., 2002) and because such samples tend to provide a stronger test of theories (Calder et al., 1981). Additionally, demographic and travel patterns of the sample were similar to the general AHLA hotel guest profile (American Hotel & Lodging Association, 2006). The study employed quasi-experimental design (random assignment of treatments) and consisted of two phases: preliminary and main. During the preliminary phase, a pilot test with a random sample of 170 members of the main study population was conducted electronically to determine the set of attributes that would represent core and facilitating factors in lodging experience. Of the 170 contacted, 45 valid responses were obtained yielding 26% response rate. Based on the pilot study, the four attributes that were most consistently identified as core (i.e., ‘‘absolute musts’’) were room cleanliness, bed/pillow comfort, property safety, and responsiveness of essential personnel. Facilitating attributes (i.e., ‘‘nice to have even if not necessary’’) included personalized services (e.g., customized bathroom amenities, firmness of pillows, and

563

L. Slevitch, H. Oh / International Journal of Hospitality Management 29 (2010) 559–569 Table 2 Scenarios for treatment groups. Attributes

Attributes condition

Provided information

Scenarioa

Core

Excellent (high)

Clean room/bathroom Comfortable bed/pillows Safe and secure property Responsive essential personnel Moderately clean room/bathroom Adequately comfortable bed/pillows Partial safety features employed Moderately responsive essential personnel Dirty room/bathroom Uncomfortable bed/pillows Safety features not employed Essential personnel not responsive

1, 2, 3

Personalized services (customized firmness of pillows/bathroom amenities, breakfast-on-the-go) Complimentary snacks Good ambiance of public facilities Jacuzzi and hot tub Limited personalized services (firmness of pillows) Some complimentary snacks Adequate ambiance of public facilities Partially functioning jacuzzi/hot tub No personalized services Damaged complimentary snacks Unpleasant ambience of public facilities No jacuzzi/hot tub

1, 4, 7

Acceptable (medium)

Poor (low)

Facilitating

Excellent (high)

Acceptable (medium)

Poor (low)

4, 5, 6

7, 8, 9

2, 5, 8

3, 6, 9

a 1, excellent core, excellent facilitating; 2, excellent core, acceptable facilitating; 3, excellent core, poor facilitating; 4, acceptable core, excellent facilitating; 5, acceptable core, acceptable facilitating; 6, acceptable core, poor facilitating; 7, poor core, excellent facilitating; 8, poor core, acceptable facilitating; 9, poor core, poor facilitating. Participants were randomly assigned to each scenario resulting in 9 cells with 120 responses.

breakfast-on-the-go), complimentary snacks, ambience of public facilities, and hot tub/sauna. The main phase of data collection was also performed electronically by assigning subjects to a role-play scenario describing a hotel stay experience (see Appendix A). This between-subjects experiment involved a 3 (core attribute performance: excellent, acceptable, poor) ! 3 (facilitating attribute performance: excellent, acceptable, poor) factorial manipulation (see Table 2). A pre-test was conducted electronically with another sample of 170 members of the main study population to assure the proposed manipulation effects and the reliability of the measures. The response rate was 42% or 72 usable replies. Factor analysis performed on the eight (i.e., four core and four facilitating) performance attributes supported the division of the attributes into core and facilitating as it was originally proposed. Core factor eigenvalue was 5.038, with 58.11% of explained variance; facilitating factor eigenvalue was 2.234, with 29.80% of explained variance. The Kaiser–Mayer–Olkin (KMO) test was performed to examine sampling adequacy for factor analysis. The obtained KMO value was .889 and exceeded a recommended acceptable level of .7 (George and Mallery, 2001). A reliability test for the satisfaction measurement items showed that the proposed multiple-item scale was reliable, as evidenced in Cronbach’s alpha of .981. ANOVA was performed to examine if experimental manipulations would produce different levels of responses in different scenarios. The ANOVA results indicated significant differences (p < .05) in the attribute performance scores for both facilitating and core attribute manipulation levels: excellent, acceptable, and poor (see Table 3). Therefore, experimental manipulations were considered satisfactory. The 9 pre-tested scenarios were randomly assigned to respondents yielding 1080 usable responses (19.5% response rate) and resulting in 120 responses in each scenario cell. All nine scenarios contained the same set of questions that were split into two sections. The first section included questions evaluating attribute performances described in the scenario as well as questions evaluating the level of satisfaction. The second section

included a set of questions surveying demographic profiles of the respondents and a question about how the respondent found about the study (see Appendix A). The eight manipulated attribute performance (i.e., four core and four facilitating) were evaluated on a seven-point ‘‘poor-excellent’’ Table 3 ANOVA results for different attribute manipulation levels (n = 72). Attribute

Level

Mean

SD

F

Room cleanliness

Excellent Acceptable Poor

6.33 4.67 1.45

.86 1.04 .66

193.19* H, M, La

Bed/pillow comfort


6.54 5.41 1.62

.66 1.10 .97

184.95* H, M, La

Property safety


6.33 4.66 1.58

.82 1.57 .93

103.97* H, M, La

Responsiveness of essential personnel

Excellent

6.41

1.01

159.55* H, M, La

Acceptable Poor

4.34 1.58

1.05 .72

Personalized services


6.04 3.23 1.67

1.26 1.45 1.09

72.78* H, M, La

Complimentary snacks


5.87 3.79 1.54

1.29 1.38 .83

78.93* H, M, La

Ambience of public facilities

Excellent

6.79

.41

Acceptable Poor

3.20 2.20

1.31 1.10


6.62 3.66 1.33

.49 1.60 .82

Hot tub/sauna

133.81* H, M, La

145.05* H, M, La

a Pair-wise mean comparison for three attribute performance levels, excellent (H), acceptable (M), and poor (L) showed significant mean differences (p < .05). * p-Value