Examining the Relationships among Antecedents of Guests’ Behavioural Intentions in Ghana’s Hospitality Industry: A Structural Equation Modelling Approach

This study empirically examines the critical antecedents of behavioural intentions and the structural interrelationships that exist among the antecedents in the hotel industry in Ghana. The study was a cross-sectional survey of 700 respondents using structured questionnaire personally administered. A usable 359 questionnaire were obtained, representing 51.3% response rate and analysed using Structural Equation Modelling approach. The findings indicate that the proposed model has high goodness-of-fit indices and explains 89.5 and 91% of the two behavioural intention variables loyalty and Positive Word of Mouth Communication (PWOMC) respectively. It also found that loyalty could be influenced through PWOMC, customer satisfaction, perceived service quality, perceived value and perceived quality of ambient factors, whereas PWOMC was influenced by satisfaction, perceived value and perceived quality of Ambient factors. Theoretically, the study fills the dearth of conceptual models in understanding the critical determinants of BI in the hotel sector in developing country context. It also provides important implications for marketing management in hotel industry. Limitations of the study are noted and recommendations for future research have been suggested. This study contributes to the body of knowledge in the area of consumer loyalty in the hospitality industry.


INTRODUCTION
Achieving long term relationship with profitable customers has been a major concern for most modern business organisations that operate with customercentric philosophy (Kotler and Keller, 2006;Gro¨nroos, 1994;Narver and Slater, 1990). The quest to retain customers is principally informed by the fierce competition in many service industry contexts (Lovelock and Wirtz, 2007). Customers are increasingly searching for better services that meet and exceed their requirements and expectations from service providers for whom they want to remain loyal.
Many service providers in hospitality industries in many developing countries are in keen competition and are increasingly faced with the challenge of developing effective marketing strategies towards meeting customer Perceived Service Quality (PSQ), achieving Customer Satisfaction (CS) and increasing customer Perceived Value (PV) in an attempt to influence customer Behaviour Intentions (BI).
Since the past two decades, the demand for hospitality services in Ghana has been increasing at a faster rate and as a result many hotels have been established towards meeting the demand for accommodation from both local and foreign customers (Sasu, 2011). The growth in competition in the hotel/restaurant sector requires that firms improve upon PSQ, CS and increase PV in order to increase market share and keep loyal customers (Abdullah and Rozario, 2009). As a result, hotel managers need to understand the critical factors that affect customer behavioural intentions that could significantly influence profitability (ibid).
Given that PSQ, CS and PV are strongly linked to customer BI in general (Abdullah and Rozario, 2009;Udo et al., 2011;Nimako, 2012;Sureshchandar et al., 2002;Wang and Shieh, 2006) and that these antecedents of BI may differ from one service context to another (Brady and Robertson, 2001;Clemes et al., 2011), it becomes critically important for service providers in hotel/restaurant in general and in Ghana in particular to understand the critical determinants that significantly drive BI. Moreover, apart from the fact that few empirical studies have generally been done on service quality in the hospitality industry in Ghana (Amoako et al., 2012), very little is known regarding other critical drivers of BI in Ghana's hotel/restaurant sector in particular. Although there are a few international empirical studies on the accommodation sector, those studies focused on the hotel industry in developed country contexts (Bitner, 1992;Choi and Chu, 2001;Clemes et al., 2011;Lockyer, 2005;Pan, 2002;Tzeng et al., 2002). This calls for more empirical research in developing country contexts to further our understanding of the determinants of BI in general and in Ghana's hotel/restaurant sector in particular. This is expected to provide empirical basis for effective marketing management strategies and to contribute to the theoretical debate in the marketing literature on the antecedents of BI and their interrelationships particularly in developing country context. Therefore, the main purpose of the study is to examine the structural relationships that exist among the antecedents of BI in the hotel sector of Ghana's hospitality industry. This study is guided by the following objectives: This was to give more meaning to tourism infrastructural development in the country. The tourism industry has been a major source of foreign exchange with its contribution of 7% to GDP (www.ghana.gov.gh). The average growth rate of the industry is estimated at 4.1% over the next two decades (Ghana Investment Promotion Center, 2010). The tourism industry includes services that are rendered by hotels, motels, resorts, restaurants, places of tourist attraction, among others. Over the past two decades, there has been rising demand for hotels and restaurants services in Ghana as more and more visitors are attracted to the country through tourism, business and educational exchange programmes, invitation and exploration for foreign investment opportunity, especially in the wake of the oil find in Ghana. As a result of these trends, there has been influx of many foreign and local investments into hotel and restaurant services resulting in the establishing of many more hotels across the country.
According to Bokpin and Nyarko (2009), in terms of governance structure of the hospitality industry, 60% of the 3-star hotels are wholly owned Ghanaian companies, 20% are joint ownership with majority Ghanaian shareholding and other 20% are joint ownership with majority foreign shareholding. For 40% of 4-star hotels the companies are joint ownership with majority foreign shareholding and 40% of other types of hotels are jointly owned with majority Ghanaian shareholding. There were no 3 or 4 star hotels wholly owned by Ghanaian companies within the hospitality industry.
In Ghana, the Ghana Hotels Association (GHA) is one key organisation that championed the course of hotel operators in the country. The GHA promotes and regulates the operations of members to ensure effective delivery of hotel services to clients and customers. Membership of the GHA is made up of hotels, motels and guest houses that have been certified and licensed by the Ghana Tourist Board to offer accommodation, catering and other tourism services in Ghana. The GHA has over 1000 registered members, out of which 817 are hotel only as of September 2012. The distribution of hotels according to regions in Ghana is presented in Table 1.
Behaviour Intentions (BI): BI has been studied extensively and used to predict actual behavior because many previous studies in learning theory have found intention to be a good predictor of actual behaviour (Cronin et al., 2000;Ajzen, 1991). BI refers to indicators which signal whether customers will remain with or defect from the company (Zeithaml and Bitner, 1996). BI may be indicated by customer loyalty, Word-Mouth-Communication (WOMC), switching behaviour, complaining behaviour, among others. Thus, in the marketing literature favourable and unfavourable dimensions of BI have been examined (Gera, 2011;Lobo et al., 2007). This study focuses on loyalty and WOMC dimensions of BI.
Loyalty is perceived as a consumer's response to a company's management strategy and therefore described as an outcome (Caruana, 2004). Loyalty as an outcome of organisational strategy has been conceputalised to consist of behavioural, attitudinal, emotional/affective dimensions (Barroso and Picón, 2012;Gremler and Brown, 1996). Behavioural loyalty considers loyalty as repeated purchase towards a brand or service (Buttle and Burton, 2002;Oliver, 1997), placing emphasis on actual past behaviour experience. This approach is limited since there is no clear distinction as to whether repeated transaction is as a result of customer's willingness to maintain future continual relationship or it is as a result of merely convenience, monetary incentives, safety, necessity or laziness (McKercher et al., 2012;Pritchard and Howard, 1997). Since loyal customers and habitual users share a common feature of repeated transaction, habitual usage alone does not necessarily constitute loyalty (McKercher et al., 2012). Attitudinal loyalty considers loyalty as future attitude of the consumer to repurchase (repurchase intentions). This may include attitudes such as becoming less likely to switch service provider, willingness to engage in Positive Word-of-Mouth Communication (PWOMC) or recommend the service provider or brand (Barroso and Martin, 1999;Butcher et al., 2001), consideration of the supplier as first choice (Mattila, 2001) and strength of preference (Mitra and Lynch, 1995). Affective or emotional loyalty has been described as a desire to maintain a relationship on the basis of a generally positive feeling towards established ties and purchasing experience (Oliver, 1999). It includes a sense of attachment and strong social bonds towards the service or service provider (Fournier, 1998), which could reduce the customer's likelihood to switch the service provider to a competitor. Social bonds are personal ties that focus on service dimensions to develop buyer-seller relationships through interpersonal interactions, friendships and identifications (Berry, 1995;Lin et al., 2003;Smith, 1998).
From the above review, it is argued that while loyal consumers may engage in positive word-of-mouth communication, it does not necessarily mean that consumers who communicate positively about a service or service provider also intend to be loyal (attitudinal loyalty) to the service provider. This is because PWOMC about a service or service provider could be done by any customer who feels satisfied but may not intend to maintain a continual relationship with a service provider. Therefore, the conceptual framework for this study, treats PWOMC and loyalty as two separate constructs of BI and hopes to test this assumption in the present study.
Moreover, most research in marketing conceputalises loyalty as multi-dimensional, although agreement on whether it has two or three dimensions is lacking (Jones and Taylor, 2007). Attitudinal loyalty has been found to be stronger than behavioural (McKercher et al., 2012). The authors noted that, "…attitudinal loyalty is a much more powerful indicator than behavior (PWOMC). In particular, metrics that reflect personal attachment such as expressions of trust and preference are more meaningful than external measures such as recommendations and positive word of mouth delivered to third parties." (McKercher et al., 2012). Therefore, our conceptual framework for loyalty includes attitudinal loyalty measures such as intentions to repurchase and make the service provider a first choice, in addition for switching intention which is an indication of both attitudinal and affective loyalty.

Drivers of Behavioural Intention (BI):
There are both convergence and divergence in the literature regarding the drivers of behaviour intentions in general and in the hospitality industry in particular. There is convergence in the marketing literature that BI is driven principally by three key constructs, namely, satisfaction, service quality and perceived value (Cronin et al., 2000;Gera, 2011). According to Reichheld (1996), CS is regarded as a necessary antecedent of customer loyalty, which in turn drives profitability and performance (Heskett et al., 1997;Reichheld, 1996). CS, PSQ, PV tend to generate many post-purchase behaviour among customers (Cronin et al., 2000;Gera, 2011;Oliver, 1980;Oliver, 1993a;. Thus, based on marketing literature, it is expected that CS, PSQ and PV would influence behavioural intentions (PWOMC and loyalty) and that PWOM could positively affect loyalty intentions, therefore, the following hypotheses are proposed: H1 : CS will have significantly positive influence on loyalty. H2 : PSQ will have significantly positive influence on loyalty. H3 : PV will have significantly positive influence on loyalty H4 : CS will have significantly positive influence on PWOMC. H5 : PSQ will have significantly positive influence on PWOMC. H6 : PV will have significantly positive influence on PWOMC. H7 : PWOMC will have significantly positive influence on loyalty.
Though there are some convergence in the marketing literature on the drivers of BI, previous studies have recommended the inclusion of other relevant dimensions or constructs that could drive BI in order to enrich theory and marketing practice for a more comprehensive conceputalisation of the BI phenomenon (Cronin et al., 2000;Gera, 2011). This study attempts to fill this gap by including variables such as Perceived Relationship Quality (PRQ) and Ambient Factors (AMF) and test their influence on loyalty and PWOMC.
Perceived Relationship Quality (PRQ) has long been identified in the service marketing literature as a functional quality construct that indicates how customers are served by the staff of the service provider (Parasuraman et al., 1988). This is very significant in hotel/restaurant services since there could be frequent face-to-face interaction between staff and customers that could induce trust, reliability and commitment in customers for a service provider (Bansal et al., 2004;Pather and Usabuwera, 2010). While relationship quality has been conceptualised as a higher order construct with many other distinct but related components, this study focuses on customer perceived relationship quality as one construct measuring how customers perceive the overall quality of their relationship with service provider. The customers' perception of the relationship quality between them and the service provider could influence the PWOMC and loyalty intentions in hotel service context. Therefore, it is hypothesized that: H8 : PRQ will significantly and positively influence loyalty. H9 : PRQ will significantly and positively influence PWOMC.
Ambient Factors (AMF) are part of the physical environment factors such the temperature, comfort, background music, lighting systems, general appeal of the service environment, cleanliness, neatness and smell of hotel rooms and environment, among others (Lovelock and Wirtz, 2007). Physical environment factors are generally termed Tangibles (Parasuraman et al., 1988). Tangibles concerns the visible appearances of all representations of the service provider to the outside world seen in such things as employees' uniforms, firm support materials and appealing nature of buildings and other physical facilities (Parasuraman et al., 1988). In the hospitality and tourism services, Tangibles includes ambient factors, which has been found to have significant influence on both consumers' and employees' behaviour (Bitner, 1992). AMFs could induce PWOMC in existing customers about the service provider and in particular, positively affecting the loyalty intentions of customers. It is expected that when hotel customers perceive ambient factors to be positive and good, it will positively influence them to engage in PWOMC and induce strong intentions to be loyal to a hotel service provider. Therefore, it is hypothesized that: H10 : AMFs will have a significantly positive influence on PWOMC. H11 : AMFs will have a significantly positive influence on customer loyalty.

Interrelationship among drivers:
In spite of the agreement in the marketing literature regarding the drivers of BI, there are however considerable disagreements in perspectives and findings regarding the interrelationships that exist among the key constructs of the BI. This has resulted and several divergent models of BI. For example, value-based models suggest that PV is directly influenced by CS and PSQ and that PV is directly linked to favourable outcomes such as loyalty (Cronin et al., 2000;Sweeney and Soutar, 2001). Satisfaction-based models also claim that, contrary to the value-based models, satisfaction is directly linked to outcome measures such as loyalty (Andreassen and Lindestad, 1998;Bolton and Lemon, 1999). Moreover, while some authors postulate an indirect relationship between PSQ, CS, BI (Taylor, 1997;, others provide evidence that direct relationships exist between the drivers of BI (Cronin et al., 2000;Gera, 2011). Since different industry contexts could induce different interrelationships, there is still void in the marketing literature regarding the interrelationships among the drivers of BI in the hospitality industry. This study attempts to fill this gap by proposing and testing a theoretical framework for understanding the interrelationships among the drivers of BI in the Ghanaian hospitality industry as shown in Fig. 1.

Drivers of satisfaction:
CS is a central concept in marketing literature because of its importance as a key element of business strategy (Anderson et al., 1994). Kotler and Keller (2006) opine that, ''Satisfaction is a person's feeling of pleasure or disappointment resulting from comparing a product's performance (outcome) in relation to his or her expectation.'' This study conceptualises CS as overall performance of a product/service or the overall performance of a product/service provider (Cronin and Taylor, 1992). It treats overall performance as cumulative because it is based on service experiences over a time period (Anderson et al., 1994) which is more useful in terms of its diagnostic and predictive value. Moreover, this study conceptualises satisfaction as both cognitive and affective dimensions as supported in previous studies (Edvardsson, 2005;Kotler and Keller, 2006;Mano and Oliver, 1993).
Many previous studies have established that service quality strongly drives satisfaction in general (Gro¨nroos, 2001;Kotler and Keller, 2006;Nimako, 2012) and in the tourism and hospitality industry in  (Clemes et al., 2011;Gera, 2011). The Perceived Service Quality (PSQ) concept was first coined by Gro¨nroos (1982) and has been found to have both cognitive and affective dimensions and an important determinant of customer satisfaction (Edvardsson, 2005;Gro¨nroos, 2001). Customer Perceived Quality (CPQ) is defined as the confirmation (or disconfirmation) of a consumer's expectations of service compared with the customer's perception of the service actually received (Gro¨nroos, 1982).
Moreover, previous studies have established the relationship between CS and perceived value (Cronin et al., 2000;Zeithaml, 1988). Perceived Value (PV) refers to "a customer's overall assessment of utility of a product based on the perceptions of what is received and what is given" (Zeithaml, 1988). It is the difference between the costs or sacrifices the consumer makes in exchange for the benefits of the service received from a service provider. While some studies have found that CS could directly affect PV in some service contexts (Andreassen and Lindestad, 1998;Bolton and Lemon, 1999), many previous studies, especially in the tourism and hospitality service context, have established that PV could positively and directly influence CS, especially where overall, cumulative satisfaction is implied (Chitty et al., 2007;Cronin et al., 2000;Moliner et al., 2007). This is because logically, where customer perceived value is high, customers tend to be content, happy and overall satisfied with the service and/or a service provider.
Furthermore, Perceived Relationship Quality (PRQ) has been found to directly affect CS (Nimako, 2012). It is expected that where there is effective and cordial relationship between the staff of a service provider in a hotel/restaurant services, and customers would likely be happy and overall feel satisfied. Ambient Factors (AMF) could influence CS through service quality (Bitner, 1992), but may not have a direct influence on CS. Based on the above discussion, the proposed model for this study conceptualises PSQ, PV and PRQ as direct antecedents of CS. These gaps are crucial to explore in order to establish the extent to which PSQ, PV and PRQ could affect CS for management strategy in the hotel/restaurant sector in Ghana. Therefore, the study seeks to test the hypotheses that: H12 : PSQ will significantly and positively influence CS. H13 : PV will significantly and positively influence CS. H14 : PRQ will significantly and positively influence CS.
Drivers of PSQ: Furthermore, Perceived Service Quality (PSQ) is expected to be influenced by PRQ and AMF simply because the customers' PSQ, overall perception of the quality of service, could be a product of the PRQ and perceived quality of AMFs. Therefore, the study proposes that: H15 : PRQ will significantly and positively influence PSQ. H16 : AMF will significantly and positively influence PSQ.

Drivers of PV:
Finally, PV is expected to be influenced by PSQ, PRQ and AMF. PRQ could be an important component of the customer's value of a relationship with a service provider (Zeithaml, 1988). In hotel services, customers could derive value or some benefits from the PSQ, PRQ and AMF and therefore, these factors could influence consumers' PV. AMFs could also induce positive perception of a firm's service quality (Bitner, 1992) and be an important component of the core service product in hotels that are intended to provide the full value proposition of a hotel service provider. Based on this, the study tests the hypotheses that: H17: PSQ will significantly and positively influence PV. H18: PRQ will significantly and positively influence PV. H19: AMF will significantly and positively influence PV.
Proposed research model: Based on the literature reviewed and hypotheses, the conceptual framework and hypothesised relationships that will be empirically tested are depicted in Fig. 1.

RESEARCH METHODOLOGY
Sampling and data collection: The population consisted of individual customers of hotel operators in Ghana. A convenient sample size of 700 respondents was chosen for the study. In order to collect data of high quality that reflect customers' opinion and improve representativeness of the sample, a survey was conducted from 10 hostels with restaurant facility in Kumasi, the second largest city in Ghana in April 2012. Out of the 700 questionnaire administered, a usable 359 were obtained, representing 51.3% response rate.

Research instrument:
A self-administered, structured questionnaire was developed and pre-tested to a sample of 20 customers. Adjustments were made based on the pre-test to get a more effective instrument. After that the questionnaire was finally administered to customers in the selected hotels through personal contact by researchers for nearly two weeks. We used informed consent form to seek permission from the hotel authorities for the study and assured respondents of anonymity and confidentiality of their responses. Since high predictive validity was a major concern, a fivepoint Likert scale was used, as recommended in previous work (Danaher and Haddrell, 1996;Delvin et al., 1993;Rust and Oliver, 1994), to measure variables for the research constructs: Loyalty, PWOMC, PSQ, CS, PV, PRQ and AMF. The Likert scale ranged from strongly disagree to strongly agree, coded 1 to 5 respectively. In all, the measurement items for the seven constructs had 20 items that were derived from previous studies and modified within the context of the GHS as shown in Table 2. The questionnaire also contained respondents' demographic data: gender, age, education, income and marital status.

ANALYSIS OF RESULTS
Respondents' characteristics: For the characteristics of the respondents, in terms of gender, 66.6% of the respondents were males and 33.4% were females. 10.6% of the respondents were below 25 years, 51.5% of them were within the ages of 25-35 years, 29.8% were between 36 and 45 years and 8.1% were 46 years and above. This implies that majority of them were in the economically active population. All respondents were educated with about 76% of them having tertiary level of education, while 9.5 and 10% had Senior High School (SHS) and post-SHS education respectively. In terms of income, 59.3% of respondents earned monthly income up to GH¢500 of which 9.5% earned below GH¢100 and 47% earned between GH¢100 and GH¢500. While 30.6% earned between GH¢500 and GH¢1000, about 10% earned monthly income above GH¢1000. This indicates that most of them earned considerably low incomes. In terms of marital status, 48.7% of the respondents were married, 49.3% were single (not married) and 1.9% of them were in other marital category.
Analysis for proposed model: Data was analysed using SPSS version 16.0 and Amos version 18.0 to perform Confirmatory Factor Analysis (CFA) and Structural Equation Modelling (SEM) to test the hypothesized relationships among the constructs in the proposed model (Fig. 1). Since the study involved testing the relationships among series of separate, but interdependent, multi-dimensional constructs simultaneously, SEM method was the most suitable method to adopt. SEM approach involves several methods such as covariance structure analysis, latent variable analysis, confirmatory factor analysis, path analysis and linear structural relations analysis and can estimate the interdependent, multiple regression equation simultaneously among different constructs (Hair et al., 2006). In SEM, first, the reliability and validity of the constructs are assessed, followed by assessment of model fitness and then the path coefficients of the hypothesized relationships.
Reliability: Reliability refers to the extent to which a measuring instrument yields consistent results under similar conditions (Hair et al., 2006). For good construct reliability, all the items should be derived from empirical studies with strong theoretical background (Hair et al., 2006) and should have high Overall, the employers of this hotel have cordial relationship with me. 2 Self-developed PRQ2 Overall, the quality of my relationship with this hotel is cordial. Ambient factors AMF1 The hotel room is not clean. 3 Self-developed based on Lovelock and Wirtz (2007)  AMF2 The music inside the restaurant is pleasant. AMF3 The rooms in this hotel have good smell.   Table 4 shows the goodness-of-fit indices for the proposed model against the benchmark factor loadings greater than or equal to 0.5 and high Composite Reliability (CR) value greater than or equal to 0.7 (Robinson et al., 1991;DeVellis, 2003). From Table 2, all constructs had strong theoretical background. From Table 3, the Cronbach alphas indicate values greater than 0.70, implying acceptable level of reliability for each construct, except AMF, which was 0.689 that is close to 0.70. Construct item reliability values are shown by the factor loadings or standardised estimates presented in Table 3. It indicates that all the items have high factor loading above 0.5 implying that the individual items explain well the variances of the construct they represent.
Construct validity: Construct validity are assessed through convergent validity and discriminant validity (Hair et al., 2006). Convergent validity refers to how indicators together explain a construct and shows the extent to which each measure correlates with other measures of the same latent construct (Hair et al., 2006). Convergent validity could be accessed through item reliability, composite reliability and the average variance extracted (Fornell and Larcker, 1981). As already demonstrated for item reliability, in Table 3, the factor loadings of items to their respective constructs are strong providing evidence to support the convergent validity of the items measured (Anderson and Gerbing, 1988). The Composite Reliability (CR), which is a measure of internal consistency comparable to coefficient alpha (Fornell and Larcker, 1981), is in excess of 0.70 except PRQ (0.629) which is close to 0.70. Thus, CR for each construct is at acceptable level. Finally, we use the Average Variance Extracted (AVE), which measures the amount of variance captured by the construct in relation to the amount of variance attributable to measurement error. Convergent validity is judged to be adequate when AVE equals or exceeds 0.50. It is estimated as the square root of the variance extracted. As shown in Table 2, all the AVE values in the diagonal are greater than 0.5. Therefore, taken together, the evidence from the high composite reliability values, high factor loadings, combined with high AVE estimates provide strong evidence in support of convergent validity.

Discriminant validity:
Discriminant validity refers to the extent to which the measure of a construct does not correlate with measures of other constructs and thus measures the extent to which constructs are distinct. At the construct level, discriminant validity is considered adequate when the variance shared between a construct and any other constructs (covariance) in the model is less than the variance which that construct shares with its measures (Fornell et al., 1982). The variance shared by any two constructs is obtained by squaring the correlation between the two constructs. The variance shared between a construct and its measures corresponds to AVE. For discriminant validity to be judged adequate the AVE for a given construct should be greater than the correlations between that construct and all other constructs. As indicated in Table 3, the AVE estimates in the diagonal are greater than the covariance below the diagonal (inter-construct correlations). Therefore, discriminant validity appears satisfactory at the construct level in the case of all constructs. This indicates that each construct shared more variance with its items than it does with other constructs. Since the results show good discriminant validity for the constructs, the constructs in the proposed research model are deemed to be adequate.
Model goodness-of-fit: In using SEM, the structural model is expected to show a good model fit index before proceeding to examine the psychometric properties of the model. The usual method is the use of the chi-square method or the ratio of the chi-square to its degree of freedom, with a value less than 3 indicating acceptable fit (Hair et al., 2006;Patrick, 1997). However, due to the fact that the chi-square of the default model could be affected by large sample size greater than 250, many researchers recommend a combination of several goodness-of-fit indices for judging the fitness of a structural model (Hair et al., 2006). Several benchmarks for good-fit indices have been suggested by many scholars (Bagozzi and Yi, 1988;Hair et al., 2006;Patrick, 1997) as shown in Table 3. Hair et al. (2006) advise that to provide strong evidence of good model fit, a combination of at least one absolute goodness-of-fit measure, one absolute badness-of-fit index, one incremental fit measure and one comparative fit index should be used.
In this study, as shown in Table 4, the results show a significant Chi-square value which is to be expected due to the large sample size above 250 ( n = 359) as noted in previous work (Hair et al., 2006). However, the ratio of the chi-square to its degree of freedom is good (3.051). All other fit-indices are better than their corresponding recommended values (AGFI = 0.842, RMSR = 0.045, RMSEA = 0.076, NFI = 0.931, CFI = 0.952, TLI = 0.940, PGFI = 0.637, PCFI = 0.757, PNFI = 0.745), except GFI (0.886) which is very close to 0.90. Therefore, there is good fit for the model. Thus, we proceed to examine the regression co-efficients for the estimated structural model (Table 5).
Assessing hypothesized relationships: Table 5 and Fig. 2 and 3 provide a summary of the results of hypotheses testing for analysing the path co-efficients.
For analysis of drivers of BI, the results in Table 5 and Fig. 2 depict the standardised regression estimates for the two behavioural intention variables: loyalty and PWOMC. First, for loyalty, the results show that all the hypotheses were supported by the data for loyalty except PRQ. Specifically, it indicates that CS significantly influences loyalty by 72.5% (β = -0.725,  p<0.05), supporting hypothesis H1. However, the negative sign indicates that lower satisfaction leads to lower loyalty. PSQ significantly influences loyalty by 56.4% (β = 0.564, p<0.05), supporting H2. PV significantly affects loyalty positively by 60.8% (β = 0.608, p<0.001), supporting hypothesis H3. Moreover, loyalty is negatively affected by AMFs by 19.2% (β = -0.192, p<0.05), providing support for H10. Furthermore, loyalty is positively and directly influenced by PWOMC by 88.8% (β = 0.888, p<0.001), supporting H7. Finally, the influence of PRQ on loyalty is not significant (β = -0.197, p>0.05). Thus, hypothesis H8 is not supported. Among the drivers of loyalty, PWOMC appears to have the greatest influence, accounting for 88.8% of the variance in loyalty, followed by CS, PV, PSQ and AMF respectively. In all, the six drivers together explain 89.5% of loyalty (R 2 = 0.895). The results also indicate that, generally, the proposed model helps predict customer loyalty by 89.5% in the research context.

DISCUSSION AND IMPLICATIONS TO THEORY AND PRACTICE
The principal objective of the research is to examine the structural relationships that exist among the antecedents of BI in the hotel sector of Ghana's hospitality. The findings of this study make several contributions to marketing theory and hotel marketing management.
Theoretical implications: One major contribution of the study is that it has validated a theoretical model predicting two behavioural intentions-Loyalty and PWOMC. The proposed model found five significant determinants of loyalty (CS, PSQ, PV, AMF and PWOMC) and three critical drivers of PWOMC (CS, PV and AMFs) in hotel sector in Ghana. It has also determined the extent to which the drivers of the two behavioural intentions are interrelated in hotel services in a developing country context. Since few of the international empirical studies in the hotel industry were conducted in developed country contexts (Bitner, 1992;Choi and Chu, 2001;Clemes et al., 2011;Lockyer, 2005;Pan, 2002;Tzeng et al., 2002), this study fills the dearth of empirical models in the hotel service context, especially in developing country context. It presents a strong validated model of behavioural intentions that is capable of explaining 89.5 and 91% of the critical factors that influence loyalty and PWOMC respectively in the hotel sector in Ghana's hospitality.
Theoretically, the validated model adds to the convergence in the general marketing literature and in particular the hospitality service literature that, loyalty could be influenced by satisfaction, perceived service quality and perceived value (Clemes et al., 2011;Cronin et al., 2000;Gera, 2011;Nimako, 2012). The findings support satisfaction-based models that postulate that satisfaction is directly linked to outcome measures such as loyalty (Andreassen and Lindestad, 1998;Bolton and Lemon, 1999). The findings are also consistent with much of the service marketing literature that postulates a direct relationship among perceived service quality, satisfaction and behavioural intentions (Cronin et al., 2000;Gera, 2011;Nimako, 2012).
Another important contribution is that the validated model extends the literature on the determinants of loyalty by confirming that positive word-of-mouth communication and perceived quality of Ambient Factors could directly influence loyalty, while perceived relationship quality may not directly affect loyalty. Aside loyalty, the present study adds to the existing literature by its findings that PWOMC could be significantly influenced by satisfaction, perceived value and perceived quality of Ambient Factors, in which satisfaction exerts the strongest influence. Thus, while loyalty is influenced by PWOMC, satisfaction, perceived service quality, perceived value and perceived quality of Ambient factors, the PWOMC is influenced by satisfaction, perceived value and Ambient factors, but not perceived service quality and perceived relationship quality. These findings suggest that loyalty and PWOMC could be separated as two distinct constructs in the marketing literature, in which the latter positively influences the former (by 88.8% in the present study) and is the strongest of the drivers of loyalty. This finding contributes to the academic debate in the marketing literature on whether PWOMC and loyalty should be viewed as synonymous constructs or as distinct. The explanation is that while loyal consumers may engage in PWOMC as noted in previous studies (Velazquez et al., 2011;Veloutsou et al., 2005), it does not necessarily mean that consumers who communicate positively about a service provider also intend to be loyal to service provider. Thus, positive word-of-mouth communication could be done by customers who feel satisfied but may not necessarily intend to maintain a continual relationship with a service provider. This is consistent with much of the general hospitality and tourism services literature. This theoretical contribution is in support of the current Service-Dominant logic (S-D Logic) perspective of marketing (Vargo and Lusch, 2008). The S-D Logic emphasises that to customer are co-creators and coproducers of value and that marketers need to understand the critical factors that drive the valuecreation process in order to satisfy customers in the services cape. Thus, the study provides marketers the knowledge of critical factors that drive BI in the hotel industry.
Again, the finding that PSQ and PV collectively affect CS by 93% and that there is significant interrelationship among the determinants of CS is another unique contribution of the study. First it suggests that SQ and CS could be separated as two distinct constructs in the marketing literature, in which SQ positively influences CS. This finding contributes to the academic debate in the marketing literature on whether SQ and CS should be viewed as synonymous constructs or as distinct (Parasuraman et al., 1988). Second, it suggests CS could be explained better by the collective influence of PSQ and PV (Cronin et al., 2000) than the individual influence of PSQ on CS (71%) and PV on CS (35%). Generally, this is consistent with previous finding that CS could be influenced by PSQ and PV constructs individually (Cronin et al., 2000;Clemens et al., 2011), although, the specific finding that PSQ exerts a greater influence on CS than PV controverts the findings of Cronin et al. (2000). Third, it suggests that PSQ could positively influence CS through PV and the strength of this indirect relationship is 0.187. This finding is consistent with previous work that PV could mediate the relationship between PSQ and CS (Taylor, 1997;.

Managerial implications:
This study offers several implications and recommendations to management of hotels in particular and the hospitality industry in general. First, it found that PWOM has the strongest influence on customer loyalty. This implies that the more customers speak favourably about a hotel service provider, the more other guests are influenced to be loyal and have the intention of a continual patronage and relationship with the service provider. It is recommended that management should put in effective strategies to manage PWOMC among hotel guests. Managing PWOMC includes both inducing it and reducing the effects of negative reputation about the service provider. Thus, generally, hotel managers should also attempt to implement effective public relations strategies to promote favourable guest perception, image and reputation of the service provider and manage negative information from hotel stakeholders about the service provider. Aside managing general corporate reputation which is likely to influence PWOMC, the study also found that PWOMC could also be influenced by overall satisfaction, perceived value and ambient factors. This implies that hotel guests are likely to carry PWOMC to other social groups and potential guests about their satisfaction of the service provider, perception of value and experience of the uniqueness of ambient factors of hotel service. Thus, hotel managers should keep increasing these factors in hotel guests to enhance customer loyalty.
Moreover, the study found that loyalty is significantly influenced by overall customer satisfaction. Satisfaction was also found to be influenced by two important factors: perceived service quality and perceived value. The importance of customer satisfaction in determining guest loyalty has long been recognised in the general marketing literature and in particular hotel and tourism literature (Cronin et al., 2000). This implies that managers should keep track of customers' satisfaction through customer feedback and attempt to increase satisfaction through the service quality factors and value-added services to hotel guest. Satisfaction should be the focus of hotel marketing management. Managers should ensure that to hotel guest are consistently satisfied and their complaints handled with maximum effectiveness to ensure service recovery.
Furthermore, the findings that loyalty could be influenced by perceived service quality and perceived relationship quality have important implications for hotel management. Like satisfaction, service quality is an important requirement for competitive strategy. While service quality in the hotel industry may have several dimensions, this study found that perceived service quality is significantly influenced by two important factors, namely, perceived relationship quality and ambient factors. In hotel service quality, ambient factors include quality issues relating to physical environment factors such as the temperature, comfort, background music, lighting systems, general appeal of the service environment, cleanliness, neatness and smell of hotel rooms and environment, among others (Lovelock and Wirtz, 2007). Perceived service quality, on the other hand, includes quality issues relating to how customers are served and their relationship with service provider managed. Thus, in order to significantly influence guests' loyalty, hotel management should ensure that staff are well trained to deliver quality in areas such as empathy, responsiveness, assurance, respectfulness, reliability, among others that have been variously described as interactive quality (Lehtinen and Lehtinen, 1991) and functional quality (Parasuraman et al., 1988).The quality of such frequent face-to-face interaction between staff and customers could induce loyalty as proven in this study, trust, reliability and commitment in customers for a service provider (Bansal et al., 2004;Pather and Usabuwera, 2010).
In addition, hotel managers should endeavour to implement effective customer relationship management systems and technologies for collecting and analyzing quality data on hotel guests. This should be used to facilitate effective tracking, communication and followups on hotel guests. It could also help recognize the unique needs and requirements of guests so that management could attempt to meet such needs and requirements anytime guests return to the hotel.
Finally, the findings that perceived value significantly influences loyalty and that perceived value is also influenced by perceived service quality, perceived relationship quality and ambient factors imply that hotel management should understand the concept of perceived value in hotel management and how to influence it. While the concept of perceived value has generated much academic debate in the marketing and tourism literature, the key element in the perceived value concept is that it constitutes "a customer's overall assessment of utility of a product based on the perceptions of what is received and what is given" (Zeithaml, 1988). It includes whatever customers sacrifice such as money, time and efforts for the services received. In this light, it becomes critically important for hotel management to ensure that essential hotel services are consistently available at affordable cost to target customers. Moreover, managers should understand what activities and added services are of higher priority to different customer segments in order to focus more attention on them. This is consistent with the S-D Logic that pre-supposes that value is created by customers at the consumption stage and that marketers do not necessarily create value but only propose and deliver what is of value to customers; this could be a source of competitive advantage to a firm.

LIMITATIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH
The findings of the study should be interpreted within its limitations. First, it is limited in terms of its external validity. Thus, while the findings are consistent with much of the empirical literature, these findings could typically be generalised to the specific context of Ghana's hotel service sector. Similar studies should be done in similar industry contexts in different countries to compare the results before global generalisations could be made. The critical antecedents of behavioural intentions and their interrelationships identified in this present study are not exhaustive in the hospitality and tourism services contexts as have long been noted in the hospitality service literature (Clemes et al., 2011;Cronin et al., 2000;Gera, 2011). Therefore, future research should explore other critical factors that could affect BI in hotel services in developing country contexts such as image of service providers, sales promotion activities, specific customer service and perceived customer experience, among others.

CONCLUSION
In conclusion, the purpose of the study was to empirically examine determinants of BI and the interrelationship that exist between these constructs in the hotel industry in Ghana. The present study proposed and validated a model of behavioural intentions, in which the determinants of loyalty and PWOMC were investigated. Given that the proposed model has high goodness-of-fit indices and explains 89.5 and 91% of the two behavioural intention variables-loyalty and PWOMC respectively, it promises a valid model for predicting consumer loyalty and PWOMC. It concludes that loyalty is influenced by PWOMC, satisfaction, perceived service quality, perceived value and perceived quality of Ambient factors, but not influenced by perceived relationship quality, whereas PWOMC is influenced by satisfaction, perceived value and Ambient factors, but not perceived service quality and perceived relationship quality. The study fills the dearth of conceptual models in understanding the critical determinants that influence BI in the hotel sector in developing country context and in particular in Ghana to aid hotel managers' strategic management decision making. The limitations of the present study are noted, the implications to theory and business management have been discussed and recommendations for future research have been suggested.