Deploying partial least squares to investigate the influence of managerial assumptions on corporate social responsibility in the hotel industry

Purpose – This paper aims to investigate the interplay between managerial assumptions and institutional corporate social responsibility, and determines how such fit affects performance. Design/methodology/approach – The authors developed and tested a model using survey methodology. The authors’ data from 210 hotels located in Qatar and the UAEwere analysed using the partial least squares (PLS) approach. Findings – The results reveal that firms with entrepreneurial, political and professional frame of reference (FoR) engage in institutional corporate social responsibility (CSR) practices. In addition, the entrepreneurial and professional FoR enhances the institutional CSR – organisational performance link. Research limitations/implications – The findings will help managers to determine the effect of their FoR on their adoption of institutional CSR, thereby increasing the effectiveness and efficiency of their CSR strategy. As the study is exploratory in nature, several limitations have been highlighted and discussed. Originality/value – To the authors’ knowledge, this is one of the few papers that inspect the relationship between managerial assumptions and institutional CSR and establishes their effect on performance.


Introduction
For over four decades, attention for corporate social responsibility (hereafter CSR) has grown dramatically in both academic and practitioner communities (Wang et al., 2016).Several researchers recognise CSR as a key factor to success and a tool to generate various business benefits (Aguinis and Glavas, 2012;Campbell, 2007).However, other scholars found no connection between CSR and firm's economic performance (Story and Neves, 2015).This inconsistency in the literature is caused by poor monitoring over CSR antecedents (Du et al., 2013).In particular, the important role of managerial characteristics in shpaing the CSR activities and practices (Christensen et al., 2014).
Whilst the advocate of institutional approach pointed to the important role of organisational and managerial factors in influencing the firm's choice of environmental strategy (Sharma, 2000), our knowledge of managerial interpretation on CSR strategy remains embryonic (Du et al., 2013).This knowledge scarcity leads several scholars to call for a new focus on how managerial characteristics might shape CSR activities (Du et al., 2013).This paper responds to these calls by investigating the interplay between managerial frame of reference (FoR) and institutional CSR practices.
The logic behind considering such relationship is the fact that, whilst CSR is a corporate practice, it is the leaders who define the range of organisational reality and limit the repertoire of possible options (Christensen et al., 2014).Therefore, in exploring the CSR antecedent, it becomes important to understand CSR's institutional environment, mainly its management role, which shapes the organisational strategies (Christensen et al., 2014;Du et al., 2013).The lack of research on the interface between organisational leadership and CSR is noteworthy.
Relatedly, complying with the cognitive theory, which proposes that an alignment between top management and the organisation's strategic choices leads to better performance (Story and Neves, 2015), this paper seeks to investigate the CSR practice effect on performance.
We believe that examining such relationships makes several important contributions.First, to the best of our knowledge, this study is the first to examine the effects of managerial assumptions on CSR practices.From a theoretical viewpoint, examining such relationships will enhance our knowledge on one potential antecedent of a firm's CSR, further our understanding of the CSR formulation and provide a basis for understanding the consequences for the performance of each CSR practice.
Secondly, whilst the study of CSR antecedents and effect is significant in its own right, regardless of organisation's type or location, most previous studies on CSR were conducted in developed economies, and parley investigated the hotel industry (Pratyameteetham and Atthirawong, 2017), which makes it difficult to develop a cogent theory and management practice in the field (Sweeney et al., 2011).Accordingly, this paper extends the work on this topic to a new setting, and with this setting choice, this study aims to not only replicate previous research in a new and less-researched context, but also extend the related research by testing a new relationship.
Finally, this study is one of few papers in the tourism field that applies partial least square (PLS) and examines unobserved heterogeneity using PLS-SEM, which is considered a key area of concern according to several researchers (Sarstedt et al., 2011).From a managerial standpoint, our results identify a match between managerial assumptions and CSR practice, which should help firms to examine the internal logic of their CSR-related profiling carefully and manage their CSR practice accordingly.
The remainder of this study is organised as follows.The following section reviews previous literature on managerial assumptions and CSR practices, and an appropriate classification of each variable in the theoretical framework is selected.Then, we present its model and hypotheses, and we test the model using a sample of 210 hotels from Qatar and the UAE.After discussing the results and implications of the PLS-path analysis, the paper concludes with a discussion of the findings, implications for management and future avenues for research.

Literature review
The first two sections here review previously researched managerial assumptions and CSR practices.In particular, we justify our use of the FoR framework and the institutional CSR framework.The last two sections explore how FoR might impact CSR practice choices and how the latter affects performance.

Managerial assumptions
Management researchers have identified the importance of managerial assumptions in the formulation of organisational strategies and their role in explaining the diversity of strategic behaviour (Child, 1972;Hambrick and Mason, 1984;Kaplan, 2011).This line of work suggests that executive cognitions, values and perceptions all play a pivotal role in shaping organisational strategic choices (Abu Farha and Elbanna, 2018).Researchers have found that managerial values and cognitive base influence their vision, selectivity and reading of the surrounding environment, thereby "directly entering in the strategic choice" (Hambrick and Mason, 1984, p. 195).
Whilst there were several conceptualisations and terms in managerial assumptions, this study adopts the FoR typology proposed by Shrivastava and Mitroff (1983) as a determinant of managerial assumptions due to its rigorous basis and because its constructs have been clearly classified and operationalised.Shrivastava and Mitroff (1983, p. 163) described FoR as "the fundamental assumptions on which organisational inquiry into problems is based upon".They include methodological, epistemic, ontological and ideological assumptions that enable organisational members to make consensual meaning out of social events (Shrivastava and Mitroff, 1983).In other words, FoR is the tool that enables organisations to make sense of events and pave the way for their "mindscapes" (Shrivastava and Mitroff, 1984).Organisational FoR is intended to be a collective term, in the sense that it is shared by most of an organisation's employees.Shrivastava andMitroff (1983, 1984) identified four types of organisational FoRpolitical FoR, entrepreneurial FoR, bureaucratic FoR and professional FoR (Table I)based on six constructs: (1) cognitive elements: the fundamental units of informationideas, notions, concepts and assumptions that are often taken for granted but cannot take place without personal inquiry or concept formation; (2) cognitive operators: the methods by which vast amounts of data are obtained, ordered and rearranged to create meaning; (3) reality tests: the processes that "anchor organizational inquiry and resulting actions in specific rules and regulations, personal values, social ideologies, customs, or scientific knowledge and thereby give them 'truth' or 'reality' status" (Shrivastava and Mitroff, 1983, p. 165); (4) domain of inquiry: the scope and boundaries of the inquiry process that firms consider and address; (5) degree of articulation: this is not an element of FoR, but rather pertains to the "degree to which the assumptions embodied in the other four elements have been articulated and codified" (Shrivastava and Mitroff, 1983, p. 166); and (6) metaphors: symbolic constructions of the firm's world in meaningful ways.
Institutional CSR.CSR is defined as "the broad array of strategies and operating practices that a company develops in its efforts to deal with and create relationships with its numerous stakeholders and the natural environment" (Waddock, 2004, p. 10).Currently, CSR has become influential in the business strategy, where most companies adopt a policy to address CSR and produce an annual report detailing its activities.
Even though existing literature distinguishes between technical CSR practices, which are related to a company's primary stakeholders, and institutional CSR practices, which are linked to the secondary stakeholders (Godfrey et al., 2009), in this paper, we focus on the institutional CSR practices for several reasons.First, the institutional CSR practices are dominant and essential for the company performance, which explains why companies are increasingly considering various practices to support local communities and reduce their eco-footprint (Aguinis and Glavas, 2012).Second, the primary stakeholders tend to have more power in the decision-making process, which leads to short-term impact and often responsive, mainly related to enhance a company's revenue.Whereas, secondary stakeholders lack power over the business control (Mitchell et al., 1997), which more likely to lead to voluntary decision-making by managers.

Effect of managerial assumptions on CSR practice
The importance of managerial assumptions in influencing an organisation's choice of environmental strategy has been acknowledged by the institutional approach (Sharma, 2000).This approach argues that social responsibility strategies are affected by the way managers interpret the external environment (Delmas and Toffel, 2004).Their logic is that institutional forces define the range of organisational reality and limit the repertoire of possible options.In this regard, Jennings and Zandbergen (1995) recognised the influence of managerial assumptions in determining the organisational environmental strategies and advised researchers to incorporate it in future studies.In addition, Du et al. (2013) found that leadership style has a direct influence on CSR practices.
Based on the previous discussion, which established that managerial assumptions play a vital role in strategic decision-making, it is expected that institutional CSR will be affected by the FoR of the dominant coalition.

Linking CSR practice and performance
Whilst prior research on the organisational outcome of institutional CSR has been inconsistent (Aguinis and Glavas, 2012), most of the work on this topic has shown that CSR practices produce various positive business paybacks (Du et al., 2013), which include developing stronger relationships with stakeholders, becoming more attractive to professional future employees, developing a more optimistic public image and generating community support (Story and Neves, 2015).Therefore, we would expect a positive relationship to exist between an organisation's FoR and CSR practices.
This paper followed Lindgreen et al. (2008) and proposes the parsimonious conceptual framework presented in Figure 1.The proposed framework integrates the paper's three sets of variables: the organisation's FoR, CSR practices and firm's performance.The framework illustrates the thesis of this study, namely, that an organisation's FoR affects its choice of CSR practice.In addition, it assumes that a fit between the institutional CSR practices and organisational FoR will improve the outcomes of the company's CSR activities.These two types of relationships are appealing if one considers the empirical and theoretical work that emphasises the role played by managers' assumptions in determining the organisation's strategic direction (Pels, 2010;White et al., 2003) and the effect that CSR practice has on firm performance.

Measures
The conceptual model of this study was examined using a structured survey method.The research variables were based on earlier measurements.The organisational FoR was developed based on the results from a questionnaire originally reported by Shrivastava and Mitroff (1983); 28 questions reflecting the four components of FoR were developed.As for institutional CSR, the study adopted Du et al. (2013) questionnaire, which contained 12 questions.Finally, performance is considered as a multi-dimensional phenomenon; therefore, we measured it using four aspects of performance measures: profitability measures, which were based on Sweeney et al. (2011), and stakeholder relationship, corporate reputation and visibility measures, which were adopted from Du et al. (2013).These indicators are extensively recognised as being amongst the most important indicators of performance because they reflect measures used in the broader marketing literature (Du et al., 2013).
As the paper used measures from previous work that were based on an extensive literature review, construct validity of the instruments is justified.As for the face validity, four marketing practitioners evaluated each item, and the questionnaire was pre-tested with a group that consists 10 Arabic executive students in the MBA program.Similar to those targeted to participate in the research.Those executives hold the same characteristics to those who are planned to participate in the research.Only few minor wording changes were recommended, but in general, the questionnaire and terms used were understandable and interpreted appropriately.Therefore, it can be interpreted that that the instrument used established the content and face validity.

Data collection and sample
In an attempt to answer the study questions, this paper focuses on the hotel industry, as this industry started giving extra focus on environmental issues and adopting more of a socially responsible behaviour (Holcomb et al., 2007).The reasons for such trend ranges from increased economic profits, employee organisational commitment, public scrutiny, improved investor relations, the good of society as a whole, up to and including the general view that CSR is the "right thing to do" (Lee and Park, 2009;Henderson, 2007).
The responses were collected from managers of hotels located in Qatar and the UAE because this setting reflects the Gulf region's environmental conditions.Despite the increasing economic and political importance of this region, empirical data on its organisational and management practices are still lacking, and any attempt to analyse psychological and practical aspects of its firms are valuable (Abu Farha and Elbanna, 2018).The sample contained 210 hotels operating in Qatar and the UAE, and the data were gathered by four experienced research assistants, using a non-probabilistic sampling method, the drop-off/pick up method of questionnaire administration (Tuncalp, 1999).After identifying the hotels and the manager or person in charge of the marketing function, phone calls were made to set up appointments.One of the researchers visited the hotels that agreed to participate, and the respondents were given the choice to complete the questionnaire immediately or to take a few days to complete it and have the researcher come back to pick it up.
The demographics of the study sample were as follows: 59 per cent of the hotels were established less than 20 years before the time of this study, and 64 per cent were four-and five-star hotels.Of these hotels, 35 per cent were wholly domestically owned, 21 per cent were completely foreign-owned and the rest were jointly owned.In terms of the respondents' characteristics, 85 per cent were male, 74 per cent were aged between 30 and 50 years and 81 per cent were in upper management positions.

Data analysis
The data were analysed using the PLS approach, which offers three advantages that coincide with the goals and characteristics of this study: PLS is recommended for models that emphasise theory development (Chin, 2010); it has an inherent ability to compute a cause-and-effect relationship model (Hair et al., 2012); and it puts minimum demand on measurement scales such as the sample size and conditions for normality (Henseler et al., 2009), making it particularly suitable for this study.
Non-response bias, common method bias and recalling information bias.The data were collected from a single respondent in each hotel, a limitation that could cause different types of bias such as retrospective rationalisation and incomplete recall.To ensure the absence of possible bias, the paper followed Podsakoff et al. (2003) recommendation to include a common method factor in the PLS model.We compared the standardised regression weights of the full measurement model with and without a common latent factor; the results indicated the differences between regression weights in all paths of the two models were less than 0.2.This display that any potential common method bias is small (Zhao et al., 2011), and we can safely claim that any relationship reported in this study indicates substantive, rather than artificial effects.

Data analysis
The PLS analysis was applied using SmartPLS 3.0 software (Ringle et al., 2005).Following Henseler et al. (2009), the test was performed on a three-tier procedure; the measurement model, the structural model and multi-group analysis (MGA).

Measurement model results
As recommended by Hair et al. (2012), we evaluated the measurement model reliability and validity using the values of the indicator coefficients, composite reliability (CR) and the average variance extracted (AVE) for each construct.Table II presents the outer loadings and the model's r c, Cronbach's a AVE.Item reliability in the model presented was deemed adequate because all reflective factor loadings exceeded the cut-off point of 0.5 recommended by Chin (2010).All other loadings below the cut-off point were deleted (19 items).As for the CR index, all constructs exceeded the minimum value of 0.70 (Hair et al., 2012), with performance showing the highest (0.942) and political FoR presenting the lowest (0.782).These results indicate that the model passed the construct reliability test.As for Managerial assumptions model's convergent validity, the last column in Table II shows that all the AVE values are above the cut-off point of 0.50 recommended by Fornell and Larcker (1981).
To confirm that discriminant validity was established, two tests were conducted; AVE (proposed by Fornell and Larcker, 1981) and the heterotrait-monotrait ratio of correlations (HTMT) (developed by Henseler et al., 2015).Table III shows that the AVE square-root (on the diagonal line) is higher than all correlations below, which means that each latent variable has more variance with its indicator than with any other latent variable (Henseler  , 2009).As for the HTMT, using the absolute HTMT 0.85 criterion, Table IV shows that all ratios are below 0.85, and the upper interval up for all relation is below 1, which indicates that discriminant validity is not an issue in our research.In sum, the results show satisfactory discriminant validity on the construct and on the item level.

Structural model results
The relationships between the latent constructs were tested by calculating 5,000 bootstrapping subsamples.Table V shows the psychometric properties of the structural, which includes the coefficient of determination (R 2 ), Stone-Geisser's (Q 2 ) and communality for all constructs.
As advised by Henseler et al. (2016), we tested the model standardised root-mean square residual (SRMR) as an indicator of the model fit criterion.The model reported an SRMR value of 0.069, which is considered adequate.Regarding R 2 , we obtained scores of 0.379 for institutional CSR and 0.176 for performance, which, according to Henseler et al. (2009), are considered moderate scores that are acceptable in value.The Q 2 , which shows the model's capacity to predict, showed values of 0.189 and 0.112, which are above zero; this means that they have great predictive relevance (Henseler et al., 2009).The last column presents the communality values, which show how much variance each construct shares with all the other variables included in the model.All the constructs reported communality above Managerial assumptions the critical point of 0.5 (Hair et al., 2012), thus indicating that they have sufficient explanations.
Table VI reports the effect size, f 2 , and the variance inflation factor (VIF).The f 2 values of the model are above the critical point of zero (Henseler et al., 2009) and range from small to medium (Cohen (1988).The third column shows that the VIF values are far below 10, indicating the absence of multicollinearity.The proposed model's significant path coefficient and moderate R 2 , Q 2 and f 2 indicate the strong explanatory power, high predictive relevance and applicability of the chosen variables; therefore, it is appropriate now to provide an analysis of the model's findings.

JHTT 9,3
The empirical results shown in Figure 2 indicate all the significant paths after running the bootstrap 5,000 times and deleting all the non-significant relationships.
Analysis of heterogeneity in the sample Hair et al. (2016) recommended testing for and dealing with unobserved heterogeneity, as, if not carefully considered, it may create a misleading interpretation.This paper applies FIMIX-PLS of SmartPLS (Ringle et al., 2005) to inspect any differences across observations.The FIMIX-PLS method has been described as one of the best-known approaches for uncovering unobserved heterogeneity in PLS-path modelling (Sarstedt et al., 2011).
Following Hair et al. (2016), we calculated the number of segments by dividing the number of sample size by the minimum sample size (50), which resulted in four segments.Table VII demonstrates the goodness of fit for the different solutions.Sarstedt et al. (2011) recommended using a combination of Akaike's information criterion (AIC3), consistent AIC (CAIC) and normed entropy (EN) as the most adequate.Based on these criteria, our results suggested that the two-segment solution was the most appropriate choice, as this segment size has the minimum values for CAIC and AIC3, and the highest EN value (above the cutoff point of 0.5), which indicates that the segments are well separated (Ringle et al., 2005).
Next, we assigned each case to one of the two subgroups, according to its maximum probability of subgroup membership.We applied different approaches to conduct the multigroup comparison; the results of the subgroup-specific PLS analysis and the significance of the differences between them is presented in Table VII.A quick comparison of those two segments, one can find similarities and some substantial differences in the strength of the path coefficients, and in some cases, small sign changes.Nevertheless, the finding from the FIMIX-PLS analysis reveals that the R 2 values, as well as the path coefficients, and CR of the model differ from one segment to another, but they look more persuasive.
Comparing the two segments, we find that segment 1 (larger segment, size of 67 per cent) may be characterised as incorporating a larger and older hotel that has a clear procedure and system in decision-making (i.e. more professional FoR type).Segment 2 includes those hotels that are new and not part of bigger chain (i.e. higher entrepreneur FoR).In segment 1, the strongest relationships in the structural model exist between professional FoR and CSR (0.339), and the CSR effect on performance (0.363) is lower than Segment 2. As for the other path coefficients, three out of five paths were found to be significant, and close in strength to the original model.In Segment 2, the strongest relation is found between entrepreneurial FoR and CSR (0.469); there were no sign changes, but two paths went down in strength (bureaucratic and professional FoR effect on CSR) where both are much higher in Segment 1.
Whilst the R 2 for CSR showed an increase in for both segments, it did increase only in Segment 1 for performance.For CSR, R 2 in Segment 1 increased by 68 per cent, whilst it increased 3 per cent only in Segment 2. As for performance, Segment 1 R 2 went up by 77 per cent but went down over 75 per cent in Segment 2 compared to the original model.The twosubgroup solution based on FIMIX-PLS provided a better fit than the global model, especially for CSR (Table VIII).

Managerial assumptions
Finally, we sought to identify whether any other factors explain the differences between the two segments.We ran non-parametric Mann-Whitney U-tests on such variables as hotel size, number of employees or age.The results imply that the segments can be separated meaningfully on the basis of hotels size and age.Segment 1 contained hotels with older age and bigger in size; Segment 2 contains hotels which are younger in age and smaller.Other factors did not significantly explain the differences between the two groups.In summary, the four relationships were found to be significant, which establishes the existence of heterogeneity in the original sample.The FIMIX-PLS analysis also achieved a considerably improved model fit, according to the higher R 2 values.

Findings
The results show that several relationships exist between the managerial assumptions and institutional CSR.The first relationship shows that hotels with greater entrepreneurial managers have more institutional CSR practices.This might be an interesting finding because such managers tend to be directed towards their personal goals.However, if we consider the managers to be "action oriented", with a preference for personal sources of information (Shrivastava and Mitroff, 1983), this will push them to keep close touch with the surrounding environment, and thus become involved with CSR practices to achieve their growth and expansion goals.
The second positive relationship was found between professional FoR and institutional CSR.These types of managers are characterised by the long-term perspective and broad domain of inquiry.This broad scan and thinking of the environment should stimulate organisational learning and foster institutional CSR practices.In addition, as professional FoR managers tend to view their origination as interdependent, rather than isolated, from its surrounding environment, thereby advancing the interests of secondary stakeholders by investing and supporting social acts.
The third link was found between political FoR managers and institutional CSR.This significant relationship is expected if one considers that political FoR managers emphasise the coalition and surroundings to which they belong.In other words, these types of managers place considerable importance on their close environment (Shrivastava and Mitroff, 1983), which fosters their support in their local societies.
In terms of the organisational outcome of the institutional CSR, our findings support previous research results that found a positive link between institutional CSR and various performance benefits.Such a positive effect is explained by the fact that consumers respond to CSR through favourable evaluations of the company and its products and through increased loyalty (Maignan et al., 1999).

Discussion
Despite the growing body of research documenting the important role of CSR on the global corporate agenda, our knowledge on the antecedents of CSR such as the interface with managerial assumptions remain incipient (Christensen et al., 2014).This paper sought to fill this gap by investigating the way in which managerial FoR affects the firm institutional CSR activities and the firm outcomes of CSR.
The findings of this study show that the organisational FoR and institutional CSR practices are inter-related, indicating that the adoption of institutional CSR is based partially on their perceptions and readings of the environment.The results show that firms with entrepreneurial, political and professional FoR are likely to engage in institutional CSR activities.
The framework presented in this paper highlights the differential roles that managerial FoR types play in institutional CSR activities and makes several contributions to enrich the marketing and tourism literature.First, although previous approaches have inspected several antecedents to CSR, no study examines the managerial assumption effect.This paper tries to bridge the gap between management and CSR literature to provide empirical evidence on the FoR-institutional CSR link.
Second, the results support and reinforce the literature for the positive impact of CSR on organisational performance.Whilst our findings are by no means novel within this line of work, they expand the current body of knowledge by demonstrating the impact of CSR on organisational performance in emerging economies is similar to that in Western developed economies.
Finally, from a methodological perspective, this study investigated the unobserved heterogeneity in some Arabic firms using the FIMIX-PLS analysis and identified two types of groups.This issue is a critical area of concern that has rarely been addressed in previous studies (Sarstedt et al., 2011;Henseler et al., 2009).

Research implications
From a managerial standpoint, the results obtained in the present research offer a major implication for those managers seeking an opportunity to generate a competitive advantage by adopting institutional CSR practices.The results suggest that institutional CSR is linked to managerial FoR.Hence, managers should carefully examine the internal logic of their CSR-related profiling and manage it accordingly.That is, if managers acknowledge how Managerial assumptions their FoR impacts the adoption of institutional CSR, it could increase the effectiveness and efficiency of their CSR strategy.
In addition, the findings suggest that institutional CSR practices are contingent on managerial FoRs.Consequently, a better understanding of the right match will also provide managers with a basis for better recommendations for improving these activities.Finally, our results suggest that improving performance is not tied to the use of CSR practices but rather to the right configuration of these activities with managerial interpretations.As the results show, organisational performance is maximised when the right FoR exists at the execution of CSR activities.
In this regard, Du et al. (2013) suggested that building an effective CSR strategy demands an understanding of how it relates to the managerial and leadership styles, and it might be valid for ensuring the survival of that organisation.

Limitations and future research
As this study is exploratory in nature, several limitations should be highlighted.First, given the novelty of the approach adopted in this study, we drew conclusions about association, but not causation.Second, this paper was linked to theorising (Weick, 1995), and by extension, contributed to theory building rather than theory testing; therefore, we did not suggest any hypotheses.Third, the study was restricted to Qatar and the UAE and involved only a small sample, which may reduce the generalisability of its findings and conclusions.
Despite the limitations highlighted, we believe that the findings of this study open avenues for potential future research.First, future work might want to explore in detail why such relationships exist.This exploration can be done for the whole model, including all relationships, or by taking one cluster, investigating its occurrence and examining the dynamics that affect its relationships.
Furthermore, this study included firms from the hotel industry.An interesting branch of research development might be to conduct a study within a range of industries within tourism.Moreover, this study was restricted to the Qatari and UAE markets; thus, a replication of this revision to other countries and/or markets, with a comparison of results, would increase the validity and application of these findings and allow for greater insights into its generalisability to be shared.

Table VI .
Effect size (f 2 ) and VIF results