6. Limitations and suggestions
This study has some limitations that need careful consideration. First, the respondents in this study came from five-star hotels in Beijing, China. Since the results of this study are based on a sample of Chinese five-star hotel employees, the selection of a single service setting and a single country may raise concerns regarding the issue of generalisability. Thus, the findings of this study may not generalise to other hospitality contexts, cultures or times, and research in other settings, geographical areas or times might yield different results (Trochim & Donnelly, 2008). The results of this study should be validated in different service settings, various star-rated hotels and other countries. Future research needs to replicate this study in different settings or use cross-cultural data to establish the generalisability of our findings.
Second, the cross-sectional design of the study did not allow us to test causal relationships among the variables. Future research tracking changes in variables over time would strengthen the ability to make causal inferences. Moreover, although the directions of the relationships investigated in our study were proposed based on the theoretical basis, opposite directions may be plausible as well. This possibility, however, does not negate the contribution of this study to the literature. More investigation in future studies would verify the causal relationships among the variables. Longitudinal study designs are needed to examine the proposed processes.
Third, this study used perceptual self-reported measures, which may have generated exaggerated relationships among variables. Because the data for this study were collected using a self-report questionnaire, respondents of the survey may have ‘faked good’ under the influence of social desirability, and the data that underwent statistical analysis could have contained SDRB and CMV (Podsakoff et al., 2003). Nevertheless, to alleviate this bias, future research needs to obtain more objective and potentially less biased measures of variables through, for example, surveying respondents’ superiors or co-workers (Beal & Weiss, 2003).