Van Dyne and Pierce (2004) observed that psychological ownership is a person’s possessive feeling towards an organisation that affects that organisational member’s attitude and behaviour. As mentioned earlier, psychological ownership refers to the feeling an individual experiences that a target, or a piece of that target, is theirs (i.e. ‘this is MINE’) (Pierce et al., 2001, p. 299). The basis of psychological ownership is the individual’s sense of possession of an object (Jussila et al., 2015). Psychological ownership is cognitive and affective, because it indicates an individual’s awareness, thoughts and beliefs concerning the possessed target and the personal meaning and emotion attached to it (Pierce et al., 2001). Psychological ownership can be directed to numerous objects (or targets) such as an organisation, a job or a work project, causing the target(s) to turn into an extension of the self and become closely associated with the individual’s identity (Pierce, Kostova, & Dirks, 2003).

The literature on psychological ownership identifies the sense of individual possession as rooted in three human needs: (1) efficacy and effectance – the individual feels efficacious and in control of the target or object, (2) self-identity – the target becomes part of the individual’s self-identity when they establish perceived ownership, (3) belongingness (having a place to ‘dwell’) – attachment to a target makes an individual feel ‘at home’ (Pierce et al., 2001, 2003). These needs can be seen as the motivational factors that serve to set individuals on their path towards psychological ownership (Jussila et al., 2015).

Individuals require psychological ownership resulting from innate motives relating to efficacy and effectance, self-identity and belongingness (having a place to dwell). Several factors influencing psychological ownership include the strength of the motives, personality, a strong sense of self and personal values (Pierce et al., 2003). The strength of motives is an indication of the likelihood that a feeling of psychological ownership will develop. Scholars have suggested that Boomers, Gen X and Millennials all have strong motives relating to their organisations because they have certain expectations about the development of their careers (Srinivasan, 2012).

Personality is another factor that can influence the development of psychological ownership. Personality traits determine the way individuals pursue an ownership relationship with objects. Boomers, for example, will develop a feeling of psychological ownership when they are personally involved in their organisation and when they learn new skills in a creative way (Gursoy, Chi, & Karadag, 2013). Gen X prefers to have control and power over an object rather than a close relationship with it because they are more committed to their careers than to their organisation. Millennials regard technological factors and a well-established lifestyle as important (Durocher et al., 2016). It can thus be assumed that Millennials will develop a sense of psychological ownership if the organisation can satisfy their technological and lifestyle requirements.

The third factor influencing psychological ownership is a strong sense of self (i.e. high self-esteem). Pierce et al. (2003) suggested that individuals with a strong need to achieve self-actualisation may pursue intrinsic targets. In contrast, individuals with a weaker self-concept are more prone to pursue materialistic targets. Boomers and Millennials value status and they desire extrinsic rewards for their loyalty and commitment towards their organisation, whereas Gen X enjoy intrinsic rewards that are less tangible (Cogin, 2012). Therefore, Gen X can be described as a group of individuals with a high self-concept, resulting in a higher tendency to develop a sense of psychological ownership.

The influence of the last factor, which is personal values, varies depending on the cohort involved. Boomers’ personal values are of immense importance to them because they believe that their values form part of their core identity and that personal values and work ethics must be aligned (Wong, Gardiner, Lang, & Coulon, 2008). Millennials were born at a time when deviation was allowed in personal values, resulting in a perception that personal values do not have to match work ethics or affect the route to success (Van der Walt et al., 2016). Lu and Gursoy (2016) observed that Gen X tend to leave the organisation when they are not satisfied. Similarly, Millennials tend to leave organisations to find a fun environment with various personal growth opportunities (Calk & Patrick, 2017). However, Boomers tend to remain at one organisation as they value loyalty towards the organisation (Cogin, 2012; Rani & Samuel, 2019). Personal values make particular objects more-or-less esteemed. From a self-concept perspective, individuals may strive to increase their feelings of self-worth by psychologically possessing items that are the most significant to them. Individuals are likely to feel ownership over those objects (such as their organisation in the case of Boomers) that are considered to be the most important according to their values (Pierce et al., 2003).

Psychological ownership is also influenced by the idea of the ‘place for psychological ownership’ (Jussila et al., 2015). This entails that some individuals might experience objects more closely than others, especially if these objects have more emotional value for these individuals. Thus, targets to which individuals, in general, attach more meaning are more likely to become targets of psychological ownership. Research conducted by Calk and Patrick (2017) is of relevance here. They established that Millennials need to see the meaning and added value of their contributions to the work before they take on full responsibility. Thus, when Millennials identify meaning in their work, the piece of work is likely to become a target of ownership.

Although personal characteristics could have an influence on the development of psychological ownership, psychological ownership per se has an influence on personal attitudes and behaviours (Pierce & Jussila, 2010). For example, if an employee perceives that a change concerning a target (for which the employee feels ownership) is self-initiated, evolutionary and additive, the employee will be positively oriented towards that change. The converse is also noticeable: employees will have a negative orientation towards change when they perceive it as imposed upon them or threatening, revolutionary and subtractive (Jussila et al., 2015). In a study conducted by Gursoy, Maier and Chi (2008), it was identified that Boomers tend to resist change and react negatively towards imposed change. Durocher et al. (2016) noticed that Gen X and Millennials, on the other hand, expect and demand change within their organisation. Management and change specialists need to take this generational difference into account when they consider changes in an organisation. As Boomers might experience changes negatively, it will result in a loss of psychological ownership and work-related outcomes being negatively affected.

As explained earlier, psychological ownership is a cognitive phenomenon reflecting an individual’s beliefs about what is ‘mine’. Nonetheless, Pierce and Jussila (2010) observed that ‘mine’ could also be ‘ours’. Millennials tend to be team-oriented and focused on ‘collective action’ (Calk & Patrick, 2017; Weeks et al., 2019). Thus, one can conclude that increasing Millennials’ teamwork can increase the psychological ownership not only of an individual but also of the individuals within a group, and this increase can simultaneously extend to include other individuals or groups (Weeks et al., 2019; Wong et al., 2008).

Organisations can thus benefit from managing psychological ownership because the possession of psychological ownership can motivate employees to perform at high levels and even inspire them to engage in extra-role behaviours and also organisational citizenship behaviours (Man & Farquharson, 2015; Olckers & Du Plessis, 2012). Knowing how generational cohorts differ in their work values and being able to identify how psychological ownership can improve these values might result in better work-related outcomes such as commitment, job satisfaction and the intention to stay with the organisation. As observed by Olckers and Du Plessis (2012), the intangible psychological phenomenon is vital and frequently ignored by human resource practitioners and managers who generally retain talent by employing tangible means.

To make meaningful inferences and comparisons about psychological ownership across the generational cohorts, it is essential to make sure that the psychological ownership scale measures the same trait across all three generational cohorts. Through the use of measurement equivalence, valid and meaningful comparisons can be made between generational cohorts’ psychological ownership. Moreover, valid and meaningful interpretations and analyses of these differences will be possible.

Measurement equivalence, which indicates that the same construct is measured, is important when studying differences or changes amongst groups (Yuan & Chan, 2016). Measurement equivalence entails that the measurements in different groups are comparable and have the same meanings across different groups (Chen, Sousa, & West, 2005). Multigroup confirmatory factor analysis (MGCFA) is one of the statistical techniques that could be used for establishing measurement equivalence and involves a sequence of progressively stricter statistical tests that build on each other. ‘Only when one form of equivalence has been satisfied can the subsequent test be performed’ (Budruk, 2010, p. 29). Firstly, configural invariance should be established. Configural invariance is supported if the data across the different generation cohorts fits a specified model (Yuan & Chan, 2016). Once configural invariance is established, metric invariance can be tested. Metric invariance testing examines whether factor loadings across groups are equal. If metric invariance is established, it will mean that the different generational cohorts respond to the scale items in the same way. Finally, if metric invariance is satisfied, scalar invariance could be tested. Scalar invariance implies that scores from the different groups have the same factor loading and origin (intercept) (Chen et al., 2005). Only after configural, metric and scalar invariance is established can one directly evaluate generational differences concerning psychological ownership.

In this study, structural equation modelling (SEM) was used for employing a MGCFA approach, to determine how psychological ownership was measured within the sample. Secondly, measurement invariance was employed to assess whether psychological ownership was measured or experienced similarly or differently across the three generational cohorts. The study tested for the three levels of measurement invariance, namely configural invariance (similar factor structures), metric invariance (similar factor loadings) and scalar invariance (similar intercepts).

The study adopted a quantitative approach with a cross-sectional design. According to Spector (2019), there is a renewed interest in cross-sectional studies because they provide insightful information at a single point in time.

The non-probability, purposive convenience sample comprised 945 skilled individuals employed in various public and private organisations in Gauteng, South Africa. The reason for choosing employed individuals was that psychological ownership towards an organisation can only develop in individuals who are employed (Pierce et al., 2001, 2003; Van Dyne & Pierce, 2004). The participants were grouped into one of the three-generational cohorts based on their year of birth: Boomers (n = 282), Gen X (n = 486) and Millennials (n = 177) (Becton et al., 2014; Smeak, 2020). The representation of numbers in this study was as expected because Millennials were starting to enter the workforce. In contrast, Boomers were retiring, leaving Gen X as the dominant generational cohort in current workforce demographics.

Table 1 presents the demographic profile for the sample across the three generational cohorts.

As indicated in Table 1, the black ethnic group was the dominant group in all the generations. Females tended to be the dominant gender in both the Gen X and Millennial groups. Most of the respondents in the sample had post-graduate degrees and occupied positions at different operational levels in various industries in South Africa. Boomers made up the dominant group at senior management level (53.84%) whereas Gen X made up the dominant group at middle management level (84%). This was expected as these generational cohorts had been part of the workforce for longer than the Millennials, they tended to have more experience and occupied senior positions in organisations.

The data collection process consisted of two methods for distribution purposes. The questionnaire was electronically distributed to participants via email using a Qualtrics platform, and a hard copy of the questionnaire was made available to participants who were not able to complete the questionnaire electronically. Each anonymous questionnaire included a cover letter inviting the participants to participate voluntarily in the study and highlighting their rights and responsibilities. Participants were assured that their responses would remain confidential and would be used for research purposes only. Permission to conduct the research was obtained from the research institution’s research ethics committee.

Olckers (2013) developed the South African Psychological Ownership Questionnaire (SAPOS) for use by South African organisations to measure employees’ psychological ownership. Olckers (2013) confirmed the validity of a 35-item four-factor model of the SAPOS comprising the following four dimensions, namely identity, responsibility, autonomy and territoriality. Identity refers to the personal cognitive connection between an individual and a target and reflects the individual’s perception of oneness with the target. Responsibility refers to the investment of time, effort and care on the target of ownership in order to maintain, enhance and protect the target. Autonomy refers to the extent to which employees could exercise control over their day-to-day activities, other employees and their personal space. Territoriality refers to a negative behavioural expression of an individual’s feelings of ownership towards physical or social objects. Although the SAPOS originally included territoriality as a prevention-oriented dimension, it was excluded in several other studies. The reason for this being the suggestion of several researchers that territoriality should be positioned as a behavioural outcome of psychological ownership rather than as a dimension of psychological ownership (Brown, Crossley, & Robinson, 2014; Brown, Lawrence, & Robinson, 2005; Pierce & Jussila, 2010). Several studies (Aggenbag, 2015; Brits, 2016) have, however, confirmed a three-factor structure of the SAPOS comprising identity, autonomy and responsibility (Olckers & Van Zyl, 2017). Brits (2016) refined the SAPOS and confirmed a shortened SAPOS scale with internal reliability of 0.79 for the overall scale. This shortened version of the SAPOS was used in this study. Each item in the SAPOS was measured by using a six-point rating scale ranging from 1 = strongly disagree to 6 = strongly agree. An example item of each of the dimensions is as follows: I have considerable opportunity for independence and freedom in how I do my work (Autonomy); I feel a strong linkage between my organisation and me (Identity) and I accept ownership for the results of my decisions and actions (Responsibility). A biographical instrument was also administrated determining the age, sex, qualification and operational level of the employees in their organisation.

Model fit was determined by using SEM, employing an MGCFA approach. MPlus version 8.2 (Muthén & Muthén, 2019) was used to process the data. Firstly, descriptive statistics (means, standard deviations, skewness and kurtosis) were employed to determine the distribution of the data and consistency in measurement. The internal consistency of the measuring instrument at both the lower (Cronbach’s alpha) and upper (composite reliability) level limits were reported. Composite reliability coefficients (ρ > 0.80) were calculated as it is deemed more appropriate for latent variables (Raykov, 2009). The minimum acceptable criterion for Cronbach’s alpha was set at α = 0.70, although values of α > 0.80 are preferable (Pallant, 2007). The guidelines suggested by Hair, Black, Babin and Andersen (2014) was followed to test for convergent and discriminant validity. Convergent validity is present when (1) the standardised factor loadings of the items is ≥ 0.50, preferably ≥ 0.70, (2) the average variance extracted (AVE) is ≥ 0.50 and (3) construct validity (CR) is ≥ 0.70. Discriminant validity was determined by the heterotrait monotrait (HTMT) ratio of correlation as was suggested by Henseler, Ringle and Sarstedt (2015). Values close to 1.00 indicate a lack of discriminant validity. The HTMT criterion should be compared with a predefined threshold. There is a lack of discriminant validity if the value of the HTMT criterion is higher that the suggested threshold of 0.85 or 0.90 (Henseler et al., 2015).

Secondly, an MGCFA approach with a comparative fit modelling strategy was employed. The maximal likelihood (ML) estimation method was used to estimate fit for all the models. Maximal likelihood was used because the outcome measures were continuous and the data were normally distributed (skewness < 2 and kurtosis < 7; Ryu, 2011). The first phase of the analyses aimed to establish a baseline model for the entire sample by employing a traditional confirmatory factor analysis (CFA) approach. Given the shared variance reported in previous studies, an oblique rotation was employed (Browne, 2001). The second phase of the analyses aimed to systematically compare the model fit for each generational cohort against the baseline model. The following cut-off criteria as suggested by Kline (2016) were employed to assess model fit: the root mean square error of approximation (RMSEA) ≤ 0.08, standardised root means square residual (SRMR) ≤ 0.10, Tucker-Lewis index (TLI) and the comparative fit index (CFI) ≥ 0.90 for good fit.

Finally, we used measurement invariance to assess whether psychological ownership was measured or experienced similarly or differently across three generations (Boomers, Gen X and Millennials). Measurement invariance was assessed by applying increasingly restrictive equality constraints on the three groups through: (1) configural invariance (similar factor structures), (2) metric invariance (similar factor loadings) and (3) scalar invariance (similar intercepts). Metric and scalar invariance are established when the change in model fit between the configural and the metric invariance models and metric and scalar invariance models are equal or smaller than the following tolerable changes if the sample is greater than 300 (N > 300): CFI (Δ < 0.01), RMSEA (Δ < 0.015) SRMR (Δ < 0.03) for configural versus metric and SRMR (Δ < 0.01) for metric versus scalar (Cieciuch & Davidov, 2015). The least restrictive configural model was used as the baseline model for comparing the increasingly constrained metric and scalar models.

Secondary data were used in this study. Data were collected by Olckers as part of her PhD studies. Ethical approval was granted by the University of Pretoria to collect data. Ethical approval was granted to Booysen to use the data as secondary data in her Master’s studies.

Table 2 provides an overview of the descriptive statistics (means, standard deviations, skewness and kurtosis) and Cronbach’s alpha and composite reliability levels amongst the latent variables across the three different generational cohorts. The results indicated that the data for the constructs were normally distributed (skewness < 2 and kurtosis < 7 [Ryu, 2011]). Furthermore, all constructs across all three generational cohorts were measured as reliable and both Cronbach’s alpha (α ≥ 0.70) and composite reliability (ρ > 0.80) scores were found to be higher than the suggested cut-offs. The SAPOS displayed convergent validity because all factor loadings across the generations were significant and above 0.50 (see Table 4), the AVE values of each construct was above 0.50 and all CR values where > 0.80 (see Table 2). Although the AVE values of responsibility for the Gen X group (AVE = 0.47) and autonomy for the Millennial group (AVE = 0.49) were slightly lower than the suggested cut-off (AVE ≥ 0.05), the composite reliabilities were higher than 0.80 and therefore, according to Fornell and Larcker (1981) the convergent validity of the construct is still adequate. Discriminant validity was supported because results of the HTMT showed values below the suggested thresholds of 0.85 and 0.90. The HTMT results ranged from 0.16 to 0.69 across the generations.

An MGCFA approach was employed to establish factorial validity across the three generational cohorts. A second-order three-factor model (with psychological ownership as second-order factor and responsibility, autonomy and identity as first-order factors) SAPOS model was assessed within the overall data set for each individual generational cohort (Boomers, Gen X and Millennials). No items were omitted from the analysis across groups, and the observations (measured items) were treated as indicators for the different latent variables within each measurement model. In order to ensure parsimony and model identification, as well as to enhance model fit, error terms on items 1 and 6 (identity) were allowed to correlate because they were very similar in wording (Wang & Wang, 2012). Table 3 indicates that this originally hypothesised second-order three-factor CFA model for promotive psychological ownership met the criteria for acceptable model fit suggested by Kline (2016) across the baseline (parent) model (χ² = 485.73; df = 115; CFI = 0.95; TLI = 0.94; RMSEA = 0.05; SRMR = 0.03), the Boomers model (χ² = 266.9; df = 115; TLI = 0.94; CFI = 0.93; RMSEA = 0.07; SRMR = 0.05), the Gen X model (χ² = 336.92; df = 115; TLI = 0.94; CFI = 0.93; RMSEA = 0.06; SRMR = 0.04) and the Millennials model (χ² = 211.31; df = 115; TLI = 0.93; CFI = 0.93; RMSEA = 0.07; SRMR = 0.06).

Table 4 shows the results for the standardised factor loadings, the associative standard error and the item uniqueness of this CFA model. All items loaded statistically significantly (p < 0.01) and sufficiently on the respective latent factors (> 0.50; Wang & Wang, 2012) across the baseline model and the three generational cohorts’ models (ranging from λ = 0.58 to 0.86).

Table 5 provides an overview of the standardised latent variable path coefficients for the CFA model and of the explained variance in respect of each generational cohort. The results showed that identity, responsibility and autonomy significantly contributed to psychological ownership across all generational cohorts. The standardised beta coefficients ranged from 0.45 to 0.93. The variance declared by identity in respect of the baseline model was 79% (p < 0.01), whereas it ranged from 58% to 86% (p < 0.01) between the three generational cohorts. Similarly, the variance declared by responsibility in respect of the baseline model was 31% (p < 0.01) but ranged from 20% to 31% (p < 0.01) in respect of the three different generational cohorts. Furthermore, autonomy declared 63% (p < 0.01) of the total variance in psychological ownership, but it ranged from 61% to 70% (p < 0.01) across the groups.

The CFA model produced statistically significant factor intercorrelations across all generational cohorts (baseline: M_r = 0.55; Boomers: M_r = 0.55; Gen X: M_r = 0.55; Millennials: M_r = 0.45). These results are displayed in Table 6.

Measurement invariance was assessed across the three generational cohorts, namely between the Boomers (n = 282), Gen X (n = 486) and Millennials (n = 177) models. The first model to be tested was the configural model (M1). Goodness-of-fit results indicated in Table 7 showed adequate fit of the model (χ² = 864.36, df = 348, CFI = 0.93, TLI = 0.92, RMSEA = 0.07, SRMR = 0.05) to the data, and thus the presence of configural invariance. This means that the same number of factors (identity, responsibility and autonomy) and factor loading pattern is the same across the three generational cohorts. The factor structure across the generational groups is similar, although not identical.

To test for metric invariance (factorial invariance), all factor loadings were constrained to be equal across the three generation groups. The chi-square difference test between the configural (M1) versus metric model (M2) was significant (Δχ² [Δdf = 28] = 93.45, p < 0.001). Chi-square is, however, affected by large sample size (Byrne & Van de Vijver, 2010). Given the fact that the chi-square difference test was based on a large sample size (N = 945) and the change in CFI (Δ < 0.01) and RMSEA (Δ < 0.015) was smaller than the suggested cut-off (Cieciuch & Davidov, 2015; Platania, Morando, & Santisi, 2020), the assumption could thus be made that there was no appreciable difference between the three generational cohorts regarding the factor loadings of the measured variables. In addition, although the change in SRMR (Δ SRMR = 0.04), was larger than the suggested 0.03 cut-off criteria as were suggested by Cieciuch and Davidov (2015), these authors, however, suggested that CFI should be used as the main criteria because both RMSEA and SRMR are more likely to be affected by sample size and model complexity. This result indicates that the factor loadings of the measured variables were invariant across the three generation groups.

To test for scalar (intercept) invariance, all intercepts were constrained to be equal across the three generation groups, in addition to the factor loadings of the latent variables. Although the chi-square difference test between the scalar invariance model (M3) and the metric model (M2) was significant (Δχ² [Δdf = 28] = 73.29, p < 0.001), the large sample size should be taken into consideration. The change in CFI (0.92 vs. 0.93) was equal to the suggested 0.01 criteria (Δ CFI = 0.01). Also, the change in RMSEA (Δ RMSEA = 0.00) was smaller than the suggested 0.015 cut-off criteria. The change in SRMR (Δ SRMR = 0.04), however, was larger than the suggested 0.03 cut-off criteria (Cieciuch & Davidov, 2015; Platania et al., 2020) and as indicated previously, could have been affected by sample size and model complexity. These results indicated that both the factor loadings and item intercepts of the measured variables were invariant across the three generation groups. The factor means could thus be compared across the groups.

The factor means across the three generation groups were compared using one-way analysis of variance. There were significant differences across groups: F = 2.84, df = 6, p = 0.009, η_p² = 0.009. Significant differences were found between Gen X and the Boomers. Statistically significant differences were established between levels of psychological ownership and specifically with regard to identity and autonomy. Results indicated that the Boomers (4.96, s.d. = 0.61) experienced higher levels of psychological ownership compared with Gen X (4.76, s.d. = 0.69). Boomers’ levels of identity (4.77, s.d. = 0.84) and autonomy (4.77, s.d. = 0.82) were also significantly higher compared with Gen X’s levels of identity (4.50, s.d. = 0.96) and autonomy (4.55, s.d. = 0.99).