A PLS-SEM APPROACH TO FACTORS AFFECTING VOLUNTARY LABOUR TURNOVER INTENTION. (2024)

Byline: SANMUKHIYA CHINTAMANEE

Keywords: Provide Minimum 3 keywords separated by semicolon (Brand Loyalty; Customer Satisfaction; Switching Cost)

1.0. INTRODUCTION:

Labour turnover would continue to be serious issue in hotels in Mauritius and this would continue to involve both direct and indirect costs. Labour turnover incurs direct costs in terms of selection, recruitment and training costs whereas indirect costs include the loss of customers, reduced productivity, low morale and extra work for remaining workers (Gima, Desaa, and Ramayah, 2015). This study postulates that Herzberg's Two-Factor theory may explain and improve job retention among employees in any organization especially the hotel industry.

This paper examines the attitudes of 122 employees within a particular department in a hotel in Mauritius. The effects of Herzberg's Two-Factor theory on employees' attitudes towards work and ultimately on their intention to resign were examined.

Prospects from rival hotels were also included in the proposed research framework. The PLS-SEM approach was deemed most appropriate for data analysis. Structural equation modelling (SEM) deals with latent variables that combine the logic of factor analysis and path modelling (Maruyama, 1997). These latent variables act as building blocks based on which the models are assessed and tested (Tipping and Bishop, 1999). However a serious limitation of SEM is that it cannot confirm a model but instead it can only not disconfirm the model.

Objectives of Study

This study proposes an innovative way to predict labour turnover intentions. It seeks to:

* Provide an innovative theoretical framework and research model that link up the Herzberg's Two-Factor theory and labour turnover intention.

* Examine the indirect impacts of Herzberg's two factor theory on labour turnover intention via employees' attitudes towards work.

* Contribute to existing theory on labour turnover intention through the inclusion of the latent variable 'offers from rival hotels'. This study uses this latent variable to capture those indicators that represent the perceived alternative job opportunities at other hotels.

* Examine whether attitudes towards work mediate the relationship between labour turnover intention and its predictors (hygiene factors, motivating factors, offer from rival hotels).

* Examine whether age (i) affects the relationship between job offers from rival hotels and labour turnover intention and (ii) whether age changes the relationship between these two factors.

* Examine whether the impacts of hygiene factors, motivating factors and perceived offers from rival hotels on labour turnover intentions differ by gender.

* Discuss resulting practical implications and inform policy making so as to halt exit of effective performers.

* Inform future researchers about the avenues to be explored.

2.0 Literature Review

Herzberg postulates that hygiene and motivation factors are important determinants of job satisfaction and dissatisfaction (Herzberg, 1959, 1966, 1968). Satisfaction and dissatisfaction are clearly defined by Locke (1976). Herzberg argues that the opposite of job satisfaction is not job dissatisfaction whereby the former is measured on a scale of no satisfaction to high satisfaction. Similarly job dissatisfaction operates separately within a scale of no dissatisfaction to high dissatisfaction. This two factor theory suggests that dissatisfaction is captured through hygiene factors (categorised as dissatisfiers) which are extrinsic to the job and satisfaction are captured through motivating factors (categorised as satisfiers) which are intrinsic to the job.

Workers would be affected by hygiene factors such as inadequate salary, poor working conditions, job insecurity, unfair distribution of tasks, unfriendly supervisor, conflicts with co-workers and work overload whereas examples of motivation factors would include pride, the work itself, opportunity to learn new things, creative/challenging work, autonomy, generous holidays, achievements, personal growth, and recognition. This theory postulates that meeting the hygiene factors or lower-level needs of employees would only prevent them from being dissatisfied but would not motivate them to work harder. Better performance may be attained only if motivation factors or higher level needs are met. Thus no increase in productivity may be expected unless much effort is exerted in improving motivation factors.

There is empirical evidence that supports Herzberg's two factor theory such that the sources of dissatisfaction at work are different from the sources of work satisfaction (Hur, 2018) although the classification of elements into hygiene factors and motivators may differ across cultures or countries (Andersson, 2017; Matei and Abrudana, 2016).

However the distinction between hygiene factors and motivation factors may not be so distinct in reality. Some factors which were originally categorised as hygiene factors may end up being classified as motivating factors and vice versa once the data collected is analysed such as welfare programs (vacation trips and banquets) and provision of accommodations. As reported in Tan (2013), some hygiene factors such as work group interactions, co-workers' support and relationships between co-workers, contribute to job satisfaction and the Herzberg's framework need to be updated to accommodate these. Alternatively he proposes a combination of the two groups (hygiene and motivation factors). Several other studies have also reported overlaps between hygiene and motivating factors (Abdullah et al., 2007; Kovach, 1987; Lin and Lin, 2011; Ruthankoon and Ogunlana, 2003; Savery, 1996; Spillane, 1973).

Despite various criticisms, the two-factor theory remains a very popular approach for investigating into the core factors that bring out satisfaction and dissatisfaction in a job across various sectors (Andersson, 2017; Chu and Kuo, 2015; Holmberg, Caro, and Sobis, 2018; Thant and Chang, 2021), and these are important aspects to increase overall productivity and profitability. Also given that this Herzberg's theory aligns with the objectives of this study, the rest of this paper is based on this theoretical framework.

Voluntary turnover implies self-initiated terminations and not organization-initiated terminations (Mobley, Griffeth, Hand, and Meglino, 1979) and it has been categorized as follows: the thought of quitting the organization; the intention to search for a new job; and the intention to quit the current position within the organization. Also lower labour turnover intention does not imply that employees are satisfied. Perceived alternative job opportunities may also impact on an employee's intention to resign. The study conducted by A1/2ivkovic, Fosic, and Vorkapic (2023) argues that perceived job opportunities act as a moderator. They conclude that employees who believe that they either have no better opportunities or fewer better opportunities, tend to have lower intention to resign.

Those with high turnover intentions were reported to have increased job opportunities outside their companies and some could be characterized as having severe deviant behaviours (Sender, Morf, and Feierabend, 2021; Wossen and Alemu, 2018). Similar findings were reported for example, by Alpar (2020) for Information Technology outsourcing in Vietnam and Zhou et al. (2022) among rontline nurses in China. Perceived alternative job availabilities may be gauged through the unemployment rate which describes the tightness of the labour market (Ehrenberg and Smith, 1982).

Labour turnover intention is low when many are being laid off and job opportunities are few whereas the intention to quit is high when job opportunities are higher compared to the number of job seekers. However Martin (2011) argues that an employee looks for alternatives a natural response to the dissatisfaction they experience at work. Data on 'offers from rival hotels' was collected in an attempt to measure perceived alternative job opportunities for the current study.

3.0 Methodology

In this study, the PLS-SEM approach of structural modelling is adopted instead of CB-SEM. PLS-SEM is preferred in few situations where analysis cannot be carried out in CB-SEM, for instance, for prediction, exploratory research, formatively measured constructs, continuous moderators and use of latent variables in subsequent analysis are to be used (J. Hair, Hollingsworth, Randolph, and Chong, 2017). The use of PLS-SEM with small sample sizes may generate greater statistical power which is suitable for prediction and theory development when compared to CBSEM. A small size of 100 is enough to achieve acceptable levels of statistical power in PLS analysis (Reinartz, Haenlein, and Henseler, 2009). Thus this completed research consists of 122 observations with no missing data and is deemed sufficient for the purpose of this study.

Alternatively to determine the adequate sample size, one needs to specify the significance level, effect size and the desired power should be set at least at 0.80 (Cohen, 1992). Higher statistical power implies that PLS-SEM is more likely to detect a statistically significant relationship when it exists in the population (Hair, Matthews, Matthews and Sarstedt, 2017) and minimise both Type I and Type II errors. Hence in this paper, statistical power is computed before any conclusion is drawn.

The user-friendly Smart PLS version 3 was used for all analyses. Three formative constructs, one reflective construct, a single item endogenous construct and three overall reflective indicators were measured by 16 indicators which were derived after reviewing extensively the existing literature on labour turnover. These are displayed in Table 1. With a scale from 1 to 5, respondents had to assess each statement on the questionnaire [(1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree and 5 = strongly agree) or (1 = very unlikely, 2 = quite unlikely, 3 = I do not know, 4 = quite likely and 5 = very likely)]. Single item construct can be used by PLS-SEM. This study abides by the ten times rule (Joseph F. Hair, Hult, Ringle, and Sarstedt, 2017).

The sample size is more than ten times bigger than the largest number of formative indicators within a single latent variable in the measurement model. Also the sample size is more than ten times the largest number of structural paths aimed at a specific latent variable in the structural model.

Linear combinations of formative indicators form the formative constructs and a change in the value of the indicator cause a change in the value of the construct (Sarstedt, Hair, Ringle, Thiele, and Gudergan, 2016). Content validity for formative indicators are examined through content specification where the researcher clearly defines the domain of the content that formative indicators intend to measure and the questionnaire must contain all items that capture at least the major aspects of the formatively measured construct (Bollen and Lennox, 1991). Quality of such constructs may be assessed through 3 steps. The first step is convergence validity examined through redundancy analysis, the second step is examining collinearity among formative indicators and the third step is examining statistical significance and relevance of those indicators.

In structural equation modeling, internal consistency reliability is more appropriately measured by composite reliability instead of Cronbach alpha because the former accounts for the indicators' differential weights (Dijkstra and Henseler, 2015). Convergent validity is measured through the average variance extracted (AVE) for each reflectively measured construct. Outer loadings for each indicator forming each latent variable are used to calculate AVE. Convergence validity is achieved when AVE exceeds 0.5 (Fornell and Larcker, 1981). It implies that more than 50% of the indicator's variance is captured by the latent variable (Joseph F. Hair et al., 2017). Discriminant validity refers to the extent to which constructs differ from one another. The standard measures are the Fornell-Larcker criterion, cross-loadings and the Heterotrait Monotrait ratio (HTMT).

These apply to reflectively measured latent variables and cannot be applied to formatively measured latent variables (Jorg Henseler, Ringle, and Sarstedt, 2015). However Klein and Rai (2009) suggest examining the cross loadings of formative indicators.

Table 1: Latent variables and corresponding items on questionnaires

###Latent Variables###Indicator###Item on questionnaire

###Hygiene factors###HF1###I do not face conflicts at work

###HF2###I have no job security

###HF3###Pay rise depends on performance at work

Formatively Attitudes towards###ATW1###I can satisfy my basic needs with my pay

masured###work###ATW2###My job given me a sense of pride

contruct###ATW3###My chance of being promoted is fair

###Job offer from rival###OFRH1###Rival hotels offers higher pay to their employees

###hotels###OFRH2###Possible of being promoted is higher at rival hotels

###OFRH3###Rival hotels offers higher training opportunity to

###their employees

Reflectively Motivating factors###FM1###Promotion depands on tenure

masured###FM2###Promotion is determined through recognition from superiors

contruct###FM3###Promotion depends on feedback from superviors

Overall###Hygiene factors###FH4###I have good working conditions

reflective###Attitudes towards work###ATW4###I am satisfy with my current job

indicator###Job offer from rival###OFRH4###Rival hotels offers better monetary and

###hotels###non-monetary packages

Ultimate###Labour turnover###LT1###How likely are you to quit your current job?

endogenous###intention

variable

Convergent Validity in the case of formative constructs may be assessed through redundancy analysis (Wong 2013). It refers to the extent to which a formative construct correlates with a reflective construct whereby both capture the same concept. The reflectively measured construct may be measured through a single item although Sarstedt et al (2016) recommends the use of multi-item scales to prevent low levels of predictive validity. In formative constructs, the relevance of indicator weights and their statistical significance must be assessed. All indicator weights above 0.1 with the 'correct' sign (proposed by theory or empirical evidence) which are statistically significant are retained within the formative constructs (Andreev et al 2009).

If the confidence interval of a weight includes zero, this implies that the weight is statistically insignificant and hence may be removed from the measurement model (Sarstedt and Ringle 2017). However there are some exceptions where a nonsignificant weight is retained, for instance, if the indicator's loading is at least equal to 0.5 (Hair et al 2017) as its absolute contribution to the construct may be substantial (Cenfetelli and Bassellier 2009).

Assessing indicator collinearity implies computing the variance inflation factor (VIF) of each item in the formative constructs. Although a VIF value below the value of 3.3 implies that multicollinearity among the indicators is not an issue (Diamantopoulos and Siguaw 2006), a VIF value below 5 is generally accepted. High collinearity results in high standard errors which hinder the possibility of finding statistically significant weights and incorrect weights with reversed signs (Joe F Hair, Ringle, and Sarstedt, 2011).

Path coefficients are similar to the standardised beta coefficients of OLS regressions and may be interpreted accordingly (J. Hair et al., 2017; Joseph F Hair, Black, Babin, and Anderson, 2010). A larger path coefficient relative to another indicates a bigger effect on the endogenous construct. These were taken into consideration before analyses and interpretations were done for this paper.

Predictive abilities of the model are also examined in this paper. The coefficient of determination, R2 explains and predicts endogenous latent constructs in the structural model. The pseudo f-test, f 2 evaluates the predictor's incremental explanation of the endogenous construct where small, medium and large effect sizes are denoted by 0.02, 0.15 and 0.35, respectively (Cohen, 1988, 1992). Predictive relevance is assessed through the cross-validated redundancy (Q2) proposed by Stone (1974) and Geisser (1975) as it includes the structural model during its prediction of eliminated data points (Joseph F Hair et al., 2017). It measures how well the path model can predict observed original values. Also q2 is used to assess the relative effect of the structural model for forecasting the observed measures of the endogenous latent variable (Ringle, Sarstedt and Straub 2012).

These are obtained through the Blindfolding procedure with an omission distance of 7 as recommended by most researchers (Hair et al 2012). It should be noted that the blindfolding procedure applies usually to reflective endogenous constructs or to single item constructs (Joe F Hair et al., 2011) but may be used for formatively endogenous constructs. Also the standardised root mean square residual (SRMR) was developed to identify model misspecification which occurs when SRMR exceeds 0.08 (Joe F Hair et al., 2011).

4.0 Analysis

The model includes a mediator, 'attitudes towards work' (ATW) which receives inputs from three constructs: hygiene (HF), motivating factors (MF) and offers from rival hotel (OFRH) and converts these into outputs represented by the endogenous variable, labour turnover intention (LTI). The aim is to explain how current employment in terms of the Herzberg's two factor theory along with OFRH translates into employees' willingness to resign. A continuous moderator, age is also used to create an interaction term with OFRH such that age affects the strength of relationship between an employee's intention to resign and job prospects at rival hotels. Finally a categorical moderating effect, gender is used through the multi-group analysis to evaluate whether the effects of the constructs on labour turnover intention differ significantly between men and women.

4.1 The Measurement Model

4.1.1 Model Specification

The standardised root mean square residual (SRMR) is used as an absolute goodness of fit measure and is equal to 0.051 which is less than the value of 0.08 as recommended by some researchers (Jorg Henseler et al., 2015; Hu and Bentler, 1998). The model is stable and stops after 3 iterations only. Usually PLS-SEM does not suffer from convergence problems. The data is normal and hence the bootstrap procedure does not produce inflated standard errors. Nevertheless it should be noted that PLS-SEM is very robust even with highly skewed data.

4.1.2 Formative Constructs

As depicted in Figure 1, convergent validity for each formative construct is ensured with path coefficients exceeding the ideal magnitude of 0.8 (Hair et al 2013) with the exception of the path coefficient for hygiene factor and attitudes towards work which are slightly below 0.8 but still exhibiting the desired strength of the path coefficient linking both reflectively and formatively measured constructs that should be at a minimum of 0.7 (Joe F Hair et al., 2011). These overall single-item reflective indicators were specified during the research design stage and included in the questionnaire. Single items were used to avoid discouraging voluntary participants and increasing their fatigues through lengthy and time consuming questionnaires which would have eventually resulted in lower response, increased missing values or at most refusal of participation in the survey.

The relevance and statistical significance of each indicator weight were examined. All weights carry the correct sign, none of them are below 0.1 and none of their confidence intervals contains zero (Table 2), implying that all indicators may be retained for further analysis. The highest VIF value of the items on the questionnaire does not exceed 3.3 and thus, there is no critical collinearity issue among indicators.

Table 2: Examining collinearity, relevance and significance of weights of formative constructs

Latent Factor Indicators###VIF###Weights###Confidence Intervals

Attributes###ATW1###2.108###0.5###{0.317, 0.674}

towards work###ATW2###1.761###0.442###{0.311, 0.568}

###ATW3###1.937###0.024###{0.057, 0.379}

Hygiene###HF1###1.285###0.371###{0.185, 0.557}

factor###HF2###1.595###0.543###{0.376, 0.706}

###HF3###1.569###0.319###{0.144, 0.506}

Offers from###OFRH1###1.549###0.290###{0.163, 0.414}

rival hotels###OFRH2###1.355###0.232###{0.11, 0.372}

###OFRH3###1.843###0.658###{0.49, 0.806}

Similarly there is no evidence of severe collinearity of the constructs HF, MF and OFRH as predictors of ATW. The same applies for the constructs ATW, NF, MF, OFRH, age and moderating effect as predictors of LTI. As displayed in Table 3, all VIF values are below the critical value of 5. It may therefore be stated that formatively measured constructs in this research are not highly correlated.

Table 3: VIF values in the structural model

###Attributes###Confidence###Labour###Confidence

###towards work###intervals###turnover intention###Intervals

Age###1.1568

Attributes towards

work###4.9409

Hygiene factor###3.2012###3.6972

Moderating Effect###1.2039

Motivating factors###3.7402###4.6636

Offers from rival

hotels###4.2034###4.9711

4.1.3 Reflective construct

All indicator reliability values are above the minimum requirement of 0.4 (Wong 2013). The composite reliability value is higher than 0.7 and hence shows high level of internal consistency reliability. Convergent validity is attained as the latent variable's average variance extracted (AVE) exceeds the minimum threshold of 0.5.

Table 4: Results summary for reflective construct

Latent###Indicators###Loadings###Indicators###Composite###AVE

variable###reliability###reliability

Motivating###MF1###0.919###0.845###0.934###0.835

Factors###MF2###0.907###0.823

###MF3###0.899###0.808

Discriminant validity was computed where 'motivating factors' is the sole reflectively measured construct in the model. Following J. Hair et al. (2017), the HTMT criterion is reported here instead of cross loadings and the Fornell-Larcker criterion because these two measures considerably overstate the presence of discriminant validity. As shown in Table 4, the HTMT values are below the required threshold of 0.85. Also all the Fornell-Larcker values are below the square of the AVE for 'motivating factors' and all items have the highest loadings within their respective constructs.

Table 5: Discriminant analysis: HTMT criterion

###Age###Moderating###Motivating###Labour turnover

###Effect###factors###intention

Age###--###--###--###--

Moderating Effect###0.014###--###--###--

Motivating factors###0.295###0.333###--###--

Labour turnover intention###0.586###0.163###0.147###--

The assessment of the reflective and formative measurement models shows that all constructs exhibit good levels of reliability and validity and hence the structural model may be assessed next.

4.2 Predictive Capabilities

The bootstrapping procedure was conducted with 5000 bootstrap samples as recommended by Joseph F Hair et al. (2017). The predictive capabilities of the theoretical structural model are assessed through the coefficient of determination R2, Cohen's effect size f 2, cross-validated redundancy (Q2) and path coefficients.

Table 6: Structural model assessment, hypothesis testing and predictive accuracy

###Path###t-###p-###f2###q2

###coefficients values values

H1 Hygiene factor a+' Attitude towards work###0.232###2.619###0.009###0.10###0.015

H2###Motivating factors a+' Attitude towards work###0.394###4.633###0.000###0.25###0.072

H3###Offers from rival hotels a+' Attitude towards work###-0.347###3.747###0.000###0.17###0.048

H4###Hygiene factors a+' Labour turnover intention###0.121###0.771###0.441###0.007###-0.005

H5###Motivating factors a+' Labour turnover intention###-0.107###0.750###0.453###0.004###-0.007

H6###Offers from rival hotels a+' Labour turnover intention 0.308###2.009###0.045###0.033###0.056

H7###Age a+' Labour turnover intention###-0.635###8.926###0.000###0.60###0.543

H8###Age*OFRH a+' Labour turnover intention###0.204###2.630###0.009###0.056###0.03

H9###Attitudes towards work a+' Labour turnover intention###0.428###2.678###0.007###0.053###0.10

When motivating factors and offers from rival hotels were included into the model to explain employees' attitudes towards work, their q2 were small. However the inclusion of hygiene factors had a medium impact on ATW.

4.2.1 R-Squared

Together the three predictors (hygiene factors, motivating factors and offer from rival hotels) explain around 83% of the variations in employees' attitudes towards work which according to Jorg Henseler, Ringle, and Sinkovics (2009) demonstrates substantial predictive accuracy. However hygiene factors, motivating factors, offers from rival hotels, attitudes towards work, age and the moderating effect explain only around 42% of the variations in labour turnover intention. According to Chin (1998), it may be argued that the whole model seems to have moderate predictive accuracy and provides a satisfactory level of explanation of the factors behind an employee's intention to resign.

4.2.2 F-Squared Effect Sizes

The f 2 effect size gauges the extent to a predictor matters at the structural level (Chin, 1998). The f 2 values for both endogenous variables (ATW and LTI) are given in Table 3. Motivating factors have a medium effect size of 0.25 whereas offers from rival hotels and hygiene factors have small effects (0.17 and 0.10, respectively) on attitudes towards work. Age has a large effect size (0.60) but offers from rival hotels and attitudes towards work have small effect sizes (0.033 and 0.053, respectively). Motivating and hygiene factors have no effect (0.004 and 0.007, respectively) on labour turnover intention. According to the criteria for interaction terms provided by Kenny (2016), it may be argued that the moderating effect has a large effect (0.056) on the labour turnover intention. Thus it may be deduced that age has a substantive impact on labour turnover intention. Herzberg's two factor theory has no impact on labour turnover intention but instead its components have small and medium effects on ATW.

4.2.3 Cross-Validated Redundancy and Q-Squared

Q 2 measures the extent to which observed values are recreated by the model and Q>0 indicates predictive relevance (Chin, 1998). The cross-validated redundancy Q 2 for the endogenous construct, ATW is 0.567 and that of LTI is 0.349. These are above zero, thus upholding the proposed research model's predictive accuracy.

4.2.4 Path Coefficients

The path coefficients are displayed in Table 4. When assessing the relative importance of the exogenous constructs on the endogenous construct ATW, it may be deduced that motivating factors ([beta]=0.394) are most important, followed by offers from rival hotels ([beta]=-0.347) and hygiene factors ([beta]=0.232). It must be noted that offers from rival hotels exert a negative impact on employees' attitudes towards their current employment. All these results are significant at 1% level. On the other hand, Herzberg's two factor theory, that is, motivating factors ([beta] =-0.107) and hygiene factors ([beta] = 0.121) have no bearing on labour turnover intention as they are statistically insignificant. Instead age has the greatest influence ([beta] =-0.635) on an employee's intention to resign, followed by attitudes towards work ([beta] = 0.428), offer from rival hotels ([beta] = 0.308) and the moderating effect ([beta] = 0.204).

These results are statistically significant at 1% level except for OFRH which is significant at 5% level. Given the path coefficients and the guidelines provided by Reinartz et al. (2009) and Cohen (1988), it may be concluded that both motivating factors and offer from rival hotels exert medium effects on ATW whereas hygiene factors exert a weak effect on ATW. Age has a strong but negative effect on labour turnover intention. Mobley et al. (1979) had also reported an inverse relationship between age and labour turnover in their study. Medium effects on LTI are exerted by ATW and OFRH whereas the moderating effect on LTI is weak.

4.3 Moderating Effect

Better offer from rival hotels has a positive influence on labour turnover intention and this effect is strengthened by age. The red line in Figure 4 represents the relationship for an average age level between OFRH and LTI. The high age value (green line) has a sharper slope than the low age value (blue line) with a shallower slope. All three lines indicate a positive relationship between OFRH and LTI. Better offers from rival hotels are associated with higher labour turnover intention. However older workers are relatively more willing to resign if better prospects are available at rival hotels. This may be due to attractive packages offered elsewhere to experienced workers. Here age may be acting as a proxy for experience.

4.4 Mediating Analysis

Both the direct and indirect effects of HF on LTI are statistically insignificant. Thus employees' attitudes towards work do not function as a mediator in the relationship between hygiene factors and labour turnover intention. This is a case of no effect nonmediation which suggests that the proposed theoretical framework is unsound and implies that some other unobserved variable may translate lack of hygiene factors into voluntary resignation.

Table 7: Mediating analysis

Indirect Path###Indirect###t-###p-###Confidence###Direct###Remark

###effect###values###values###interval###Path

###coefficient

HF a+' ATW a+' LTI###0.099###1.655###0.098###[0.01, 0.236]###HF a+' LTI###No mediation

MF a+' ATW a+' LTI###0.169###2.446###0.015###[0.044, 0.316]###MF a+' LTI###Full mediation

OFRH a+' ATW a+' LTI###-0.149###2.183###0.029###[-0.292, -0.033]###OFRH a+' LTI*###Partial mediation

The direct effect of MF on LTI is statistically insignificant whereas its indirect effect is significant at 5% level. Hence employees' attitudes towards work fully mediate the relationship between motivating factors and labour turnover intention. This is a case of indirect-only mediation which implies that ATW fully conforms with the hypothesized theoretical framework. This case may be termed as the best-case scenario.

Both the direct and indirect effects of OFRH on LTI are significant at 5% level. Here employees' attitudes towards work partially mediate the relationship between offers from rival hotels and labour turnover intention. Since the coefficients of the direct and indirect paths have opposite signs, here employees' attitudes towards work act as a suppressor and considerably reduces the magnitude of total effect of OFRH on LTI. This is a case of competitive mediation.

4.7 Multi-Group Analysis

A multi-group analysis was conducted to identify any difference in PLS path coefficients across gender. According to Henseler (2007), p-values below 0.05 and above 0.95 indicate a statistically significant difference between group specific path coefficients for the relationship under study. Table 8 displays all results. The results were not significant except for the relationship between hygiene factors and labour turnover intention (p = 0.991). Male employees (-0.182) had lower path coefficient relative to that of women (0.645). vHence it may be concluded that lack of hygiene factors may strongly encourage more women than men to resign. Similar findings were reported by Marsh and Mannari (1977) for Japanese manufacturing workers.

Table 8: Multi-group analysis

###Path Coefficients-diff###p-Value###Path Coefficients Path Coefficients

###(male - female)###(male vs female) Original (female) Original (male)

Age -> LTI###0.1162###0.2088###-0.691###-0.575

Moderating effect -> LTI###0.1182###0.2408###0.119###0.238

ATW -> LTI###0.2194###0.2667###0.205###0.424

HF -> ATW###0.1441###0.7671###0.334###0.190

HF -> LTI###0.8267###0.9914###0.645###-0.182

MF -> LTI###0.1488###0.2047###0.311###0.460

MF -> LTI###0.5459###0.0401###-0.421###0.125

OFRH -> LTI###0.0565###0.6165###-0.291###-0.347

OFRH -> LTI###0.0437###0.5624###0.314###0.270

5. Conclusion

This paper aims to examine whether the Herzberg's Two-Factor theory combined with offers from rival hotels (OFRH) affects an employee's intention to resign through a mediator 'attitudes towards work' (ATW). It also investigates whether there is an interaction between age and OFRH such that age would change the relationship between labour turnover intention and offers from rival hotels. A multi-group analysis was done to capture multi-group effects across gender.

The use of PLS-SEM is justified as it addresses the main concern of social and business research which is about combining prediction and explanation to model estimation (Joseph F Hair, Matthews, Matthews, and Sarstedt, 2017). Human resource managers want to avoid the mistake of not identifying a good employee and subsequently not retaining him/her (Type I error). Also human resource managers want to prevent wrongful classification of poor performers as effective performers and subsequently overspending in retaining the former (Type II error). False positives (Type I errors) are considered to be more problematic than false negatives (Type II errors) in business and social research (Cohen, 1988). This explains why researchers use a minimum of 95% confidence interval and a statistical power of 80 percent (Goodhue, Lewis, and Thompson, 2012) while conducting analysis.

The model used in this study (SRMR<0.08) indicates good fit. Given that the effect size f 2 assesses the contribution of an exogenous latent variable to the R2 of an endogenous latent variable, it may be concluded that hygiene factors exert a medium effect on attitudes towards work whereas motivating factors and offers from rival hotels exert small effects on attitude towards work. Attitudes towards work fully mediate the relationships between the three predictors (hygiene factor, motivating factor and offer from rival hotels) and labour turnover intention. An employee's age has a large effect on labour turnover intention while the moderating effect of age on offer from rival hotels has a small effect on labour turnover intention. All other endogenous constructs have no effect on this endogenous construct.

In short with a statistical power level of 100%, it may be concluded that neither motivation factors nor hygiene factors affect labour turnover intention directly. However unlike motivation factors, hygiene factors have no indirect impact on labour turnover intention. Age is the main factor that influences an individual's decision to resign voluntarily and this effect is accentuated in the case of older employees with better job prospects at rival hotels. The impacts of better job offers from rival hotels and attitudes towards work on labour turnover intention are however small. Women have a greater tendency than men to exhibit high intention to quit when hygiene factors are inadequate.

Employers should investigate into the most frequent sources of dissatisfaction and satisfaction to enable employees to work to their full potential. They should strive to provide their effective performers an employment environment that improves their attitudes towards work and lowers their job search. Unnecessary workloads and stress should be avoided and relatively attractive salaries should be offered to high performers and experienced workers. These measures will add to job satisfaction and reduce job dissatisfaction.

6. Limitations of Study

Future research is required to identify the determinants of voluntary labour turnover because other variables that were excluded from this study (such as organisational commitment and organisational justice) may predict labour turnover better. Also similar to the study conducted by (Kmieciak (2021)), this research was based on a cross-sectional data with a small sample size whereby data were collected among a hom*ogenous group of workers. However it may be argued that the results may be generalised to other employees in other hotels with similar working environments.

7. Suggestions for Future Studies

Lack of co-worker support deteriorates the well-being of employees and increases stress with significant role overload (Chiaburu and Harrison, 2008) which may eventually stimulate voluntary labour turnover. The conservation of resources theory states that employees would tend to reduce activities which are unjustified and excessively energy-draining. Without adequate co-worker support, the employee has to put in more effort to perform his/her tasks and becomes less satisfied. Similarly psychological withdrawal behaviours such as horizontal and vertical knowledge withholding could explain the reasons behind resignations. Also employee burnout reduces job satisfaction, increases negative attitude towards the job and adds to labour turnover (Yu, Ariza-Montes, Hernandez-Perlines, Vega-Munoz, and Han, 2020). Thus future studies may examine the impacts of these factors on voluntary labour turnover.

In order to avoid misinterpreting correlational data as causal data, future studies may compare competing models that have not been falsified instead of just testing a model (Hox and Bechger, 1998).

Acknowledgement: I thank Miss Roopnah Moushmi for collecting this data.

REFERENCES

Abdullah, M., Shuib, M., Muhammad, Z., Khalid, H., Md Nor, N., and Jauhar, J. (2007). Employee Organisational Commitment In Smes: Evidence From The Manufacturing Sector In Malaysia. International Review of Business Research Papers, 3.

Alpar, P. (2020). Turnover intentions of employees of information technology outsourcing suppliers in Vietnam. International Journal of Human Resources Development and Management, 20(1), 43-60. doi: 10.1504/IJHRDM.2020.105109

Andersson, S. (2017). Assessing Job Satisfaction Using Herzberg's Two-Factor Theory: A Qualitative Study Between US and Japanese Insurance Employees IAFOR Journal of Business and Management, 2(1), 22-35. doi: 10.22492/ijbm.2.1.02

Chiaburu, D. S., and Harrison, D. A. (2008). Do Peers Make the Place? Conceptual Synthesis and Meta-Analysis of Coworker Effects on Perceptions, Attitudes, OCBs, and Performance. Journal of Applied Psychology, 93, 93(5), 1082-1103. doi: https://doi.org/10.1037/0021-9010.93.5.1082

Chin, W. W. (1998). The Partial Least Squares Approach to Structural Equation Modeling. In G. A. Marcoulides (Ed.), Modern Methods for Business Research. New Jersey, London: Lawrence Erlbaum Associates.

Chu, H. C., and Kuo, T. (2015). Testing Herzberg's two-factor theory in educational settings in Taiwan. Journal of Human Resource and Adult Learning, 11(1), 54-65.

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (Second ed.). New York: Lawrence Erlbaum Associates.

Cohen, J. (1992). Statistical Power Analysis. Current Directions in Psychological Science, 1(3), 98-101.

Dijkstra, T. K., and Henseler, J. (2015). Consistent Partial Least Squares Path Modeling. MIS Quarterly, 39(2), 297-316. doi: 10.25300/MISQ/2015/39.2.02

Ehrenberg, R. G., and Smith, R. S. (1982). Modern Labor Economics. International Journal of Manpower, 34, 584-615.

Fornell, C., and Larcker, D. F. (1981). Evaluating Structural Equation Models with Unobservable Variables and Measurement Error. Journal of Marketing Research, 18(1), 39-50. doi: 10.2307/3151312

Geisser, S. (1975). The Predictive Sample Reuse Method with Applications. Journal of the American Statistical Association, 70(350), 320-328. doi: 10.2307/2285815

Gima, G. C. W., Desaa, N. M., and Ramayah, T. (2015). Competitive Psychological Climate and Turnover Intention With the Mediating Role of Affective Commitment. Procedia - Social and Behavioral Sciences, 172, 658 - 665. doi: 10.1016/j. sbspro.2015.01.416

Goodhue, D. L., Lewis, W., and Thompson, R. (2012). Does PLS Have Advantages for Small Sample Size or Non-Normal Data? MIS Quarterly, 36(3), 981-1001. doi: 10.2307/41703490

Hair, J., Hollingsworth, C. L., Randolph, A. B., and Chong, A. Y. L. (2017). An updated and expanded assessment of PLS-SEM in information systems research Industrial Management and Data Systems, 117(3), 442-458. doi: 10.1108/IMDS-04-2016-0130

Hair, J. F., Black, W. C., Babin, B. J., and Anderson, R. E. (2010). Multivariate Data Analysis. Upper Saddle River, NJ Prentice Hall.

Hair, J. F., Hult, T. G. M., Ringle, C. M., and Sarstedt, M. (2017). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) (Second ed.). New York: SAGE Publications, Inc.

Hair, J. F., Matthews, L. M., Matthews, R. L., and Sarstedt, M. (2017). PLS-SEM or CB-SEM: Updated Guidelines on Which Method to Use International Journal of Multivariate Data Analysis, 1(2), 107-122.

Hair, J. F., Ringle, C. M., and Sarstedt, M. (2011). PLS-SEM: Indeed a silver bullet. Journal of Marketing Theory and Practice, 19, 139-151. doi: 10.2753/MTP1069-6679190202

Henseler, J., Ringle, C. M., and Sarstedt, M. (2015). A New Criterion for Assessing Discriminant Validity in Variance-Based Structural Equation Modeling. Journal of the Academy of Marketing Science, 43(1), 115-135. doi: 10.1007/s11747-014-0403-8

Henseler, J., Ringle, C. M., and Sinkovics, R. R. (2009). The Use of Partial Least Squares Path Modeling in International Marketing. Advances in International Marketing, 20, 277-319. doi: 10.1108/S1474-7979(2009)0000020014

Herzberg, F. (1959). Motivation to work. New York: Wiley.

Herzberg, F. (1966). Work And The Nature Of Man. Cleveland: World Pub. Co.

Herzberg, F. (1968). One more time: How do you motivate employees. Harvard Business Review.

Holmberg, C., Caro, J., and Sobis, I. (2018). Job satisfaction among Swedish mental health nursing personnel: Revisiting the two-factor theory. International Journal of Mental Health Nursing, 27(2), 581-592. doi: 10.1111/inm.12339

Hox, J. J., and Bechger, T. M. (1998). An introduction to structural equation modeling. Family Science Review, 11, 354-373.

Hu, L.-t., and Bentler, P. M. (1998). Fit Indices in Covariance Structure Modeling: Sensitivity to Underparameterized Model Misspecification. Psychological Methods, 3(4), 424-453.

Hur, Y. (2018). Testing Herzberg's two-factor theory of motivation in the public sector: Is it applicable to public managers? Public Organization Review, 18(3), 329-343. doi: 10.1007/s11115-017-0379-1

Klein, R., and Rai, A. (2009). Interfirm Strategic Information Flows in Logistics Supply Chain Relationships. MIS Quarterly, 33(4), 735-762.

Kmieciak, R. (2021). Co-Worker Support, Voluntary Turnover Intention and Knowledge Withholding Among IT Specialists: The Mediating Role of Affective Organizational Commitment. Baltic Journal of Management, ahead - of - print (ahead-of-print). doi: https://doi.org/10.1108/BJM-03-2021-0085

Kovach, K. A. (1987). What motivates employees? Workers and supervisors give different answers Business Horizons, 30(5), 58-65.

Lin, S.-C., and Lin, J. S.-J. (2011). Impacts of coworkers' relationships on organizational commitment-and intervening effects of job satisfaction. African Journal of Business Management, 5, 3396-3409.

Locke, E. A. (1976). The nature and causes of job satisfaction. In M. D. Dunnette (Ed.), Handbook of Industrial and Organizational Psychology (Vol. 1, pp. 1297-1343). Chicago: Rand McNally.

Marsh, R., and Mannari, H. (1977). Organizational commitment and turnover: A predictive study. Administrative Science Quarterly, 22, 57-75.

Martin, M. J. (2011). Influence of Human Resource Practices on Employee Intention to Quit. (Doctor of Philosophy), Virginia Polytechnic Institute and State University, Blacksburg, VA.

Maruyama, G. M. (1997). Basics of Structural Equation Modeling (First ed.). CA: SAGE Publications, Inc.

Matei, M.-C., and Abrudana, M.-M. (2016). Adapting Herzberg's Two Factor Theory to the Cultural Context of Romania. Procedia - Social and Behavioral Sciences 221 (2016) 95 - 104, 221, 95-104. doi: 10.1016/j.sbspro.2016.05.094

Mobley, W., Griffeth, R., Hand, H., and Meglino, B. (1979). A Review and Conceptual Analysis of the Employee Turnover Process. Psychological Bulletin, 86, 493. doi: 10.1037/0033-2909.86.3.493

Reinartz, W. J., Haenlein, M., and Henseler, J. (2009). An Empirical Comparison of the Efficacy of Covariance-Based and Variance-Based SEM. International Journal of Research in Marketing, 26(4), 332-344. doi: 10.1016/j.ijresmar.2009.08.001

Ruthankoon, R., and Ogunlana, S. (2003). Testing Herzberg's Two-Factor Theory in the Thai construction industry. Engineering, Construction and Architectural Management, 10(5), 333-341. doi: 10.1108/09699980310502946

Sarstedt, M., Hair, J. F., Ringle, C. M., Thiele, K. O., and Gudergan, S. P. (2016). Estimation Issues with PLS and CBSEM:Where the Bias Lies! Journal of Business Research, 69, 3998-4010. doi: 10.1016/j.jbusres.2016.06.007

Savery, L. K. (1996). The congruence between the importance of job satisfaction and the perceived level of achievement. Journal of Management Development, 15(6), 18-27. doi: 10.1108/02621719610120110

Sender, A., Morf, M., and Feierabend, A. (2021). Aiming to leave and aiming to harm: the role of turnover intentions and job opportunities for minor and serious deviance. Journal of Business and Psychology, 36, 449-460. doi: 10.1007/s10869-020-09685-5

Spillane, R. (1973). Intrinsic and extrinsic job satisfaction and labour turnover: A questionnaire study of Australian managers. Occupational Psychology, 47(1-2), 71-74.

Stone, M. (1974). Cross-validation Choice and Assessment of Statistical Predictions. Journal of the Royal Statistical Society Series B, 36(2), 111-147.

Tan, S. K. (2013). Herzberg's Two-Factor Theory on Work Motivation: Does it Works for Todays Environment? Global Journal of Commerce and Management Perspective, 2, 18-22.

Thant, Z. M., and Chang, Y. (2021). Determinants of Public Employee Job Satisfaction in Myanmar: Focus on Herzberg's Two Factor Theory. Public Organization Review, 21, 157-175. doi: 10.1007/s11115-020-00481-6

Tipping, M. E., and Bishop, C. M. (1999). Probabilistic Principal Component Analysis. Journal of the Royal Statistical Society, Series B (Statistical Methodology), 61(3), 611-622.

Wossen, S. A., and Alemu, M. A. (2018). The influence of job role, job opportunity, satisfaction and stress on employees turnover intention. In three government Bureau of Dire Dawa Administration. European Journal of Business and Management, 10, 76-82.

Yu, J., Ariza-Montes, A., Hernandez-Perlines, F., Vega-Munoz, A., and Han, H. (2020). Hotels' Eco-Friendly Physical Environment as Nature-Based Solutions For Decreas-ing Burnout And Increasing Job Satisfaction And Performance. International Journal of Environmental Research and Public Health, 17(6357), 1-17. doi: https://-doi.org/10.3390/ijerph17176357

Zhou, L., Kachie, A. D. T., Xu, X., Quansah, P. E., Epallem, T. M., Ampon-Wireko, S., and Nkrumah, E. N. K. (2022). COVID-19: The effects of perceived organizational justice, job engagement, and perceived job alternatives on turnover intention among frontline nurses. Frontiers in Psychology, 13, 1-18. doi: 10.3389/fpsyg.2022.920274

A1/2ivkovic, A., Fosic, I., and Vorkapic, A. P. (2023). Perceived alternative job opportuni-ties and turnover intention as preliminary steps in employee withdrawal. Strategic Management, XX(20xx), 0xx-0xx. doi: 10.5937/StraMan2300042Z%20