Assessing the psychometric properties of the revised and abbreviated self-leadership questionnaires

Orientation: Self-leadership is considered to be vital for improved performance in the South African working context. Limited research has been done on the psychometric properties of the Revised and Abbreviated Self-Leadership Questionnaires on a sample of working adults. Research purpose: The aim of this study was to assess the psychometric properties of the Revised (RSLQ) and Abbreviated (ASLQ) versions of the Self-Leadership Questionnaires for a sample of working adults in South Africa. Motivation for the study: Researchers have not previously published psychometric properties of the RSLQ and ASLQ using the original conceptualisation and based on a sample from the South African working context. Research design, approach and method: The RSLQ and ASLQ were administered to a non-probability sample of 405 working adults in South Africa. Confirmatory factor analysis was employed to estimate the goodness-of-fit associated with competing conceptualisations of both the Revised and Abbreviated Self-Leadership Questionnaires. Main findings: The authors found acceptable goodness-of-fit for both the RSLQ and ASLQ versions of the Self-Leadership Questionnaires. More specifically, both the composite (representing self-leadership) and calculated scores (representing subscales of self-leadership) indicated acceptable levels of reliability. However, the RSLQ is best conceptualised as consisting of a strong general factor. Practical/managerial implications: The study found that both the RSLQ and the ASLQ have good measurement properties, indicating that these questionnaires can be applied in a variety of settings. Contribution/value-add: Research has indicated the value of investigating the psychometric properties of the Revised and Abbreviated Self-Leadership Questionnaires for a sample of employees in the South African working context. In this study, a bifactor model provided evidence that the RSLQ consists of a strong general factor (i.e. self-leadership). The ASLQ also seem to measure a single dimension that is very reliable. Utilising the ASLQ will save time in research and also when applied in the working context.


Introduction Orientation
According to Houghton, Dawley and DiLiello (2012), in these present times of economic uncertainty and fierce competition, many firms are shifting away from a traditional top-heavy leadership paradigm to embrace a new model of leadership that involves empowering employees at all organisational levels to greater responsibility for their own work-related behaviours and actions (Houghton et al., 2012).Ugurluoglu, Saygili, Ozer and Santas (2013) are of the opinion that, under today's conditions, the most appropriate leader is the self-leader who also leads others towards self-leadership.Self-leadership (the process of influencing oneself to perform more effectively) has attracted a significant amount of attention over the past two decades (Neck & Houghton, 2006), as is evident in the dozens of academic articles written on this issue during this period (e.g.Alves, Lovelace, (RSLQ; Houghton & Neck, 2002), however, more empirical studies have been conducted (e.g.Sahin, 2008;Ugurluoglu et al., 2013).Most of these studies were conducted in the USA, Europe and Asia, whilst only a single study by Mahembe, Engelbrecht and De Kock (2013) was conducted in South Africa.This study was applied to young adults studying full time at a South African university, not to working adults.It is, therefore, questionable whether the second-order structure that Mahembe et al. (2013) provided in their research would be replicable on a sample of working adults in the South African working context.Furthermore, the application of confirmatory factor analysis on the second-order factors of the RSLQ could confirm the construct validity of the RSLQ, specifically for research participants in a South African working context (Mahembe et al., 2013).
Another relevant study, by Houghton et al. (2012), focused on the development and validation of a nine-item abbreviated version (ASLQ) of the 35-item RSLQ.The applicability of this version, however, has not been tested on a South African working population before.

Purpose
The current study focuses on the assessment of the psychometric properties of the revised (RSLQ) and abbreviated (ASLQ) versions of the Self-Leadership Questionnaire for a working population in South Africa.Self-leadership should have a wider socio-economic relevance in developing countries such as South Africa, because it is a competency that could prove critical in transformation on an individual, group, organisational and societal level (Mahembe et al., 2013).Yet, not much South African-related research on self-leadership (especially regarding its measurement within the working context) has been done (Van Zyl, 2009).Furthermore, no research focusing on the applicability of the ASLQ within any South African context has been done previously.Against this background, the primary research aim of this study was to assess the psychometric properties of the RSLQ and ASLQ for a sample of working adults.
More specifically, the following will be evaluated (based on a working population in the South African context): • The goodness-of-fit associated with the various conceptualisations of the RSLQ (35 items) using a sample of working adults.• The degree to which the RSLQ consists of a strong general factor, using a sample of working adults.• The goodness-of-fit associated with the various conceptualisations of the ASLQ (nine items) using a sample of working adults.

Contribution to the field
This study will contribute by closing the gap in research on the psychometric properties of the RSLQ for a sample of South African working adults.In addition, the presence of a strong general factor has previously been investigated using a higher-order factor structure.However, the current study will employ a bifactor model, which is a more appropriate approach to determine the presence of a strong general factor.Furthermore, a need has been identified for research on the psychometric properties of the ASLQ for a similar sample (South African working adults), because the shorter version will save time when researchers want to assess self-leadership as part of a more extensive study including other variables.
Both these instruments can help to identify individuals with poor self-leadership skills in the South African working context.By doing so, support can be provided to such individuals to improve their self-influencing (selfleadership) skills which, in turn, could help them to perform more effectively.

Self-leadership: Conceptual overview
Self-leadership is a process of behavioural and cognitive self-evaluation and self-influence by which people achieve the self-direction and self-motivation needed to shape their behaviours in positive ways in order to enhance their overall performance (Houghton et al., 2012, p. 217).More specifically, self-leadership involves strategies and normative actions that will help to enhance performance (Houghton et al., 2012;Ugurluoglu et al., 2013).Self-leadership strategies and actions are usually divided into behaviour-focused strategies, natural reward strategies and constructive thought strategies (Norris, 2008).
Behaviour-focused self-leadership strategies pertain to self-observation, goal setting, self-reward, self-punishment and cues (Neck & Houghton, 2006).Self-observation refers to individuals' awareness of how, when and why they engage in specific behaviours (D'Intino et al., 2007).According to Ugurluoglu et al. (2013), self-observation entails that people escalate their awareness of why and when they display certain behaviours.Such awareness could help people to overcome the display of effective but unproductive behaviour (Ugurluoglu et al., 2013).With accurate information regarding current behaviour and performance levels, individuals can more successfully set effective behaviour-altering goals for themselves (Manz & Neck, 2004).Rewards set by an individual, along with self-set goals, can aid significantly in energising the efforts necessary to accomplish the goals (D'Intino et al., 2007).Self-rewards could be as simple as mentally praising oneself for a job well done or something more tangible, such as treating oneself to a new outfit or a night at the movies (Houghton et al., 2012).Self-punishment is an additional strategy for the selfinfluence of behaviour.However, habitual self-punishment and guilt have, in general, received mixed to negative support as a self-leadership strategy (Stewart, Courtright & Manz, 2011).Finally, behavioural rehearsal (cues) prior to actual performance can promote refinement, improvement and corrective adjustments for greater individual effectiveness (Stewart et al., 2011).
Natural reward strategies are designed to enhance the intrinsic motivation that is vital for performance (Mahembe et al., 2013).These types of strategy help individuals to build pleasant and enjoyable features into their activities so that the tasks themselves become naturally rewarding (Norris, 2008).Houghton et al. (2012, p. 218) add that individuals can employ natural rewards also by shifting the cognitive focus to the intrinsically rewarding aspects of the job.Focusing on building pleasant and enjoyable features in work tasks will create feelings of intrinsic motivation (D'Intino et al., 2007).
Constructive thought strategies involve visualising successful performance, engaging in positive self-talk and examining individuals' beliefs and assumptions in order to align cognitions with desired behaviour (Ho & Nesbit, 2013).Through a process of identifying and altering distorted beliefs, individuals can minimise dysfunctional thinking processes and engage in more rational and effective cognitive processes (Houghton et al., 2012).Positive self-talk and evaluation of existing habits and ways of thinking (for instance, I am an opportunity thinker) can enhance constructive thought patterns (Van Zyl, 2008).
Assessment of self-leadership Manz (1993) developed a set of initial items to capture elements of both self-management and self-leadership.
Self-management (Manz, 1986;Stewart et al., 2011) is conceptualised as strategies for getting oneself to complete difficult but necessary tasks and is assessed by scales capturing self-observation, cueing strategies, self-goal setting, self-reward, self-punishment and practice (Stewart et al., 2011).In contrast, self-leadership is conceptualised as more intrinsically motivated and includes scales like building natural rewards into work, choosing pleasant surroundings, building naturally rewarding activities into work, focusing on pleasant aspects of work and focusing on natural rewards rather than on external rewards (meaning rewards outside the work itself; Manz & Simms, 1991;Neck, 1996;Stewart et al., 2011, p. 191).Mahembe et al. (2013) indicated that, at about the same time, Cox (1993) developed and tested a 34-item unpublished Self-Leadership Questionnaire (SLQ), with eight factors, labelled as self-problem-solving initiative, self-efficacy, teamwork, self-reward, self-goal setting, natural rewards, opportunity thought and self-observation or evaluation (Mahembe et al., 2013, p. 4).Anderson and Prussia (1997) continued to work on the Manz scale and subjected it to content validation (Stewart et al., 2011).Subject matter experts placed the original 90 items into three categories: behaviour-focused strategies, natural reward strategies and constructive thought strategies (Stewart et al., 2011, p. 191).Houghton et al. (2012) indicated that the 50-item instrument of Anderson and Prussia is a first step in developing a self-leadership scale, but it was plagued by some inherent reliability and validity problems and, therefore, required additional refinement.Houghton and Neck (2002) developed the RSLQ by eliminating or rewriting ineffective questions from Anderson and Prussia's (1997) SLQ and adding items from Cox's (1993) instrument (Houghton et al., 2012).Houghton et al. (2012) argue that, although the RSLQ demonstrates reasonably good reliability and validity across a number of empirical studies (Houghton, Neck & Singh, 2004;Neck & Houghton, 2006), additional research is needed to further assess the reliability and validity of the RSLQ (Houghton et al., 2012, p. 222).Houghton et al. developed and validated a nineitem abbreviated version (ASLQ) of the 35-item RSLQ.
A confirmatory factor analysis indicated that the ASLQ is a reliable and valid measure that has inherited the nomological network of associations from the original version of the RSLQ (Houghton et al., 2012, p. 216).
A South African study by Mahembe et al. (2013) focused on a confirmatory factor analytic study of the RSLQ.The findings indicated that the RSLQ demonstrated sufficient factorial or construct validity.The second-order measurement model confirmed that the eight self-leadership factors contributed to an overall self-leadership approach (Mahembe et al., 2013).Although this research indicated strong psychometric properties, the sample of young adults studying at a South African university cannot be generalised to working adults in the South African context.
No South African study to date has attempted to investigate the psychometric properties of the ASLQ.It is therefore important to also investigate the psychometric properties of the ASLQ, because using the longer version (RSLQ) could pose a challenge to researchers when self-leadership is being examined together with other variables.It may however be applied on its own when trying to identify employees with poor self-leadership skills (in the workplace).Houghton et al. (2012, p. 222) put it as follows: 'overall survey length can quickly become unwieldy, leading to rater fatigue, inaccuracy and missing survey data'.

Research approach
In order to execute the research, the current study followed a cross-sectional design with a survey data collection technique.
More specifically, the current study investigated competing measurement models representing the three different conceptualisations as suggested by the developers of the RSLQ.The three competing measurement models were: (1) a hierarchical model of self-leadership, (2) a unidimensional model and (3) a three uncorrelated factors model of selfleadership.In addition, the current study investigated a fourth competing measurement model representing a bifactor structure.This latter model will provide more direct evidence as to whether or not the RSLQ measures a strong general factor (i.e.self-leadership).
A similar approach (i.e.competing measurement models) was followed to investigate the psychometric properties associated with two conceptualisations of the ASLQ: (1) a three uncorrelated factors model of self-leadership and (2) a unidimensional structure associated with self-leadership.

Participants
A total of 405 working adults participated in the study.Women were in the majority (72%).Most of the participants came from a designated group (71%) and were in the age group of 26-35 years (81%).Most of the respondents were married (64%).Accidental sampling was used within different organisations in the financial services sector.

Measuring process
Permission for the research was granted by the research committee of the Faculty of Economic and Management Sciences and all relevant ethical issues were cleared.The participants were briefed about the aim of the study, their right to voluntary participation and the anonymity of the information they would provide.Participants received instructions on how to complete the RSLQ.They also completed a biographical questionnaire.

Measuring instruments
Self-leadership was measured using the RSLQ, which consists of 35 items.The questionnaire covers three dimensions (behaviour-focused strategies, natural reward strategies and constructive thought pattern strategies).The three dimensions have the following subscales: • Behaviour-focused strategies: self-goal setting (five items), self-reward (three items), self-punishment (four items), self-observation (four items) and self-cueing (two items).• Natural reward strategies: focusing thoughts on natural rewards (five items).• Constructive thought pattern strategies: visualising successful performance (five items), self-talk (three items) and evaluating beliefs and assumptions (four items).
The ASLQ sovers three dimensions, namely behaviour awareness and volition, task motivation and constructive cognition.The three dimensions have the following subscales: • Behaviour awareness and volition: self-goal-setting (two items) and self-observation (one item).• Task motivation: visualising successful performance (two items) and self-reward (one item).• Constructive cognition: evaluating beliefs and assumptions (two items) and self-talk (one item).
Participants also completed a biographical questionnaire providing information related to age group, gender, home language and marital status.

Analysis
The current study employed LISREL 8.80 (2006) to estimate the goodness-of-fit of each of the competing measurement models associated with both the RSLQ and ASLQ.A test of multivariate normality was performed to determine whether the data violated the assumption of normality.The results suggested that the data deviated from normality with regard to skewness and kurtosis.Hence, the robust maximum likelihood method of estimation was used to estimate the various models (Brown, 2006).Several fit indices were used as well, including the Satorra-Bentler scaled chi-square, root mean square error of approximation (RMSEA), standardised root mean square residual (SRMR), comparative fit index (CFI) and the goodness-of-fit index (GFI).Values close to 0.95 for GFI and CFI are considered indicative of good model fit.Hu and Bentler (1999) suggest that values close to 0.06 are indicative of acceptable fit for RMSEA, whilst values smaller than 0.08 are acceptable for SRMR.In addition, Akaike's information criterion (AIC) is used in the comparison of competing measurement models with smaller values representing a better fit of the proposed model (Byrne, 2006).
Cronbach's alpha (α) was used to estimate the reliability of the dimensions of the constructs being investigated in the current study.Reliability estimates of 0.7 and higher are indicative of good reliability.However, estimates as low as 0.6 may be acceptable when conducting exploratory research (Hair, Black, Babin, Anderson & Tatham, 2006, pp. 137, 778).
It should be noted that, in order to facilitate direct comparison with the results obtained by developers of the RSLQ, the current study used (1) the same indicators and (2) the same conceptualisations of the three competing measurement models (hierarchical model of self-leadership, one-factor model and three uncorrelated factors model).The current study used the same instructions to create item parcels (i.e.composite scores) for eight of nine subscales.In addition, the developers used the three items with the highest factor loadings (items 26, 32, 35) for the natural reward subscale (Houghton & Neck, 2002, p. 678).

Revised Self-Leadership Questionnaire
From Table 1, it is clear that the three measurement models proposed by the developers of the RSLQ show different levels of fit with regard to CFI, RMSEA and SRMR.The three uncorrelated factors model has poor fit.In contrast, both the unidimensional and higher-order factor models exhibit acceptable levels of fit when looking at CFI, RMSEA and SRMR.The bifactor model also seems to fit the data well when considering the value of CFI.Together, both the unidimensional and bifactor model results may lend support to the possibility that the RSLQ measures a general construct (i.e.self-leadership).
The developers of the RSLQ concluded that the 'behaviour focused, natural rewards, and constructive thoughts factors have a higher-order factor; namely self-leadership' (Houghton & Neck, 2002, pp. 679, 681, 685).Mahembe et al. (2013, p. 3) claim that the higher-order factor model could provide evidence that the RSLQ consists of a strong global or general factor.However, this model is not the only or most suitable conceptualisation in determining the presence of a strong general factor.This second-order model states that the target trait (self-leadership) is a second-order dimension that explains why the three primary dimensions (behaviour-focused strategies, natural reward strategies and constructive thought pattern strategies) are correlated.
Unfortunately, there is no direct relationship between the item and the target trait (self-leadership), but the relationship between self-leadership and each item or parcel is mediated by the primary factor.Hence, it models an indirect effect.A more suitable approach to determine the presence of a strong general factor is using a bifactor model (Chen, West & Sousa, 2006;Reise, 2012;Reise, Moore & Haviland, 2010).The latter model allows researchers to determine directly the degree to which items reflect a common trait (e.g.self-leadership) and the degree to which they reflect subtraits (i.e.sub-dimensions of the RSLQ; Reise et al., 2010, p. 546).
Table 2 shows that the majority of items have higher loadings on the general factor than on the group factors (i.e.dimensions of the RSLQ).This is indicative of a strong general factor (Reise et al., 2010).As per the suggestion of Reise et al. (2010), an omega hierarchical reliability coefficient of the general factor of the bifactor model was also obtained, namely 0.90.
The reliability estimates associated with each of the three dimensions of the RLSQ are reported in Table 3.It is clear that all the dimensions have acceptable reliabilities that exceed 0.70.
Table 4 provides the reliability estimates associated with each of the eight subscales of the RSLQ.It is evident that both the self-cueing and self-goal setting subscales are the most reliable subscales of the RSLQ.

Abbreviated Self-Leadership Questionnaire
According to Table 5, the unidimensional model of the ASLQ is the better fitting model as evident from the small value of AIC.In addition, all the fit indices associated with this model (CFI, RMSEA and SRMR) are indicative of an acceptable fit.In contrast, the three uncorrelated factors model does not fit the data well.
It is clear from Table 6 that all three the dimensions associated with the ASLQ have acceptable estimates of reliability that exceed 0.70.The behaviour awareness and volition dimension seem to be the most reliable of the three.

Psychometric properties of the Revised Self-Leadership Questionnaire
When comparing the goodness-of-fit statistics obtained in the current study with those reported by Houghton and Neck  2002), the results are encouraging.Houghton and Neck found that the unidimensional model did not fit the data well.However, the current study found that the unidimensional model did fit the data well.Where the developers concluded that the higher-order factor model was the best fitting model, the current study found that both the unidimensional and higher-order models were the best fitting when using the conceptualisations suggested by the developers.Both these results may suggest that the RSLQ measures a strong general factor (i.e.self-leadership).However, only the bifactor model results would be able to support this assumption.
The reliability estimates reported by the developers are fairly similar or higher than those obtained by the current study with regard to the following six subscales: self-goal setting (0.84 vs 0.91), self-observation (0.82 vs 0.82), self-cueing (0.91 vs 0.91), focusing thoughts on natural rewards (0.74 vs 0.80), evaluating beliefs and assumptions (0.78 vs 0.87) and visualising successful performance (0.85 vs 0.86).However, the current study obtained lower reliabilities with regard to the following three remaining subscales: self-reward (0.93 vs 0.77), self-punishment (0.86 vs 0.71) and self-talk (0.92 vs 0.71).
A recent South African study reported on the psychometric properties of the RSLQ within a sample of young adults (Mahembe et al., 2013).They tested a higher-order factor model consisting of the eight subscales.The results obtained by the current study seem to be fairly similar to those reported by Mahembe et al. (2013).Specifically, the values of CFI are almost exactly the same for both the three uncorrelated factors model (0.99 vs 0.98) and the higher-order factor model (0.99 vs 0.98).In contrast, the values obtained in the current study are slightly higher (but still acceptable) with regard to the RMSEA in both the three-factor model (0.04 vs 0.07) and the higher-order factor model (0.04 vs 0.07).
The reliability estimates reported by Mahembe et al. (2013) are fairly similar or lower than those obtained by the current study with regard to the following six subscales: self-goal setting (0.84 vs 0.91), self-observation (0.82 vs 0.82), selfcueing (0.82 vs 0.91), focusing thoughts on natural rewards (0.74 vs 0.80), evaluating beliefs and assumptions (0.76 vs 0.87) and visualising successful performance (0.82 vs 0.86).
However, the current study obtained lower reliabilities with regard to the following two remaining subscales: self-reward (0.90 vs 0.77) and self-talk (0.87 vs 0.71).Unfortunately, Mahembe et al. did not include the self-punishment subscale in their study.
In short, the psychometric properties of the RSLQ seem to suggest an instrument that is best conceptualised as measuring a single factor that is also very reliable.
Presence of a strong general factor: Although the results of the current study point to a well-fitting single factor model for self-leadership when using the RSLQ, this is not sufficient to claim that this instrument measures a strong general factor.In order to determine the presence of a strong general factor, the current study also employed a bifactor model to investigate this possibility.In contrast to the approach followed by Mahembe et al. (2013), the bifactor model is a more suitable conceptualisation to investigate the presence of a strong general factor.The current study indeed found evidence of a strong general factor.However, where Mahembe et al. tested a higher-order factor structure with the eight subscales (instead of the nine subscales available), both the developers and the current study used the three dimensions associated with the RSLQ.The current study therefore concludes that, when using the RSLQ, it is better to use a single composite score representing self-leadership.However, given the high value associated with omega hierarchical, the group factors (i.e. the dimensions of the RSLQ) do not seem to have any psychometric value.They do not contribute any additional variance already accounted for by the general factor.Hence, the use of subscores representing the dimensions associated with the RSLQ may be overly optimistic.

Psychometric properties of the Abbreviated Self-Leadership Questionnaire
The current study found that the one-factor model, associated with the ASLQ, fitted the data well.As was evident from the results reported for the RSLQ, the three uncorrelated factors model also did not seem to be a good representation of the ASLQ.In contrast, Houghton et al. (2012) found acceptable levels of fit associated with the three uncorrelated factors model.The developers of the ASLQ reported a reliability estimate for the total scale (α) of 0.73 (Houghton et al., 2012, p. 226).In contrast, the current study found a much higher reliability estimate (α = 0.89).In addition, the reliability estimates associated with each of the three subscales obtained in the current study are fairly similar (task motivation = 0.71) or slightly better (constructive cognition = 0.78, behaviour awareness and volition = 0.85).
It can therefore be concluded that it is better to conceptualise the ASLQ as measuring a single factor (i.e.self-leadership).
Although the sub-dimensions did exhibit acceptable estimates of reliability, it seems as if their use in future research may not add more theoretical support than what may already be gained from treating self-leadership as a single, strong general factor.

Practical implications
The authors recommend that practitioners and researchers use both the RSLQ and ASLQ as a research and development tool.When these questionnaires are applied as a research tool, they can be considered as a valid and reliable measure applicable in the South African working context.
Both the RSLQ and ASLQ can also help in the working context in identifying employees with poor self-leadership skills.In this way, the relevant development actions can be identified and implemented which, in turn, could contribute to improving the situation.Keeping in mind low productivity outputs amongst South African employees as indicated by Van Zyl (2009), improved self-leadership skills in the workplace can influence people to perform more effectively.
Applying the shorter version (ASLQ) will save time not only in research (especially when extensive research with many variables is conducted), but also when applied in the working context (especially considering that employees usually have limited time to complete questionnaires).

Limitations and recommendations
Although the current study provided some insight into the psychometric properties of self-leadership, the following suggestions are put forward in order to improve on the measurement of the self-leadership construct.Firstly, further investigation is needed into the factorial invariance of both the ASLQ and RSLQ amongst different language and racial groups in a South African context.It was not possible to validate both these measures in individual language and ethnic groups due to sample size constraints.Hence, future researchers should obtain samples with large enough groups representing the various language and ethnic groups to determine the measurement equivalence of the selfleadership construct.Secondly, there are more advanced statistical techniques, such as Rasch analysis, that should be used to supplement the results obtained in the current study, especially with regard to unidimensionality.The Rasch model is the preferred technique to determine the unidimensionality of a construct (such as self-leadership).

Conclusion
The current study concludes that both the ASLQ and RSLQ have sound psychometric properties and can be applied within the South African working situation.When applying the longer version (RSLQ), researchers could use a composite score (representing self-leadership) as obtained from the bifactor model.The latter suggests the presence of a strong general factor.The shorter version (ASLQ), will save time when utilised in research and when applied in the work situation.In summary, the current study concludes that both the ASLQ and RSLQ are suitable for use within a South African context amongst working adults.