Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Empirical investigation of factors influencing faculty followership’s

239 views

Published on

Factors Influence Followership Perceptions

Published in: Education
  • Be the first to comment

  • Be the first to like this

Empirical investigation of factors influencing faculty followership’s

  1. 1. 1 Empirical Investigation of Factors Influencing Faculty Followership’s Perception of Institutional Leader in Malaysian Institutions of Higher Learning Hairuddin Bin Mohd Ali Institute of Education International Islamic University Malaysia Gombak 50728, kuala Lumpur, Malaysia E-mail address: hairuddin@iab.edu.my and Mohammed Borhandden Musah Institute of Education International Islamic University Malaysia Gombak 50728, kuala Lumpur, Malaysia E-mail address:mbmusaus@yahoo.com Abstract This study aims at testing psychometric properties of the 32-item of Romance of Leadership Scale (RLS). The study is equally aims at investigating factors influencing faculty followership’s perception of institutional leader in selected Malaysian institutions of higher learning. Online survey questionnaires were conveniently sent to 2560 faculties of 16 public universities sampled, using guidelines of 95% confidence interval and margin of error at ± 02% to collect the data. A confirmatory factor analysis was performed to test factorial validity and psychometric properties of the three constructs underlie the faculty followership’s perception of institutional leader. It is anticipated that findings of the study will provide leaders of the institutions of higher learning with proper understanding of factors influencing faculty followership’s perceptions of institutional leaders, which is critical to organizational performance and faculty retention. Furthermore, it is expected that the findings will contribute to the initiatives of empirical studies on followership especially in higher education sector. The study as being the first of its kind in Malaysian context to empirically evaluate psychometric properties of the scales underlie faculty followership’s perception of institutional leader and investigate factors influencing faculty followership’s perception of institutional leader using well experienced participants rather than inexperience participants of the previous studies which sampled university students, high school and company employees would be a remarkable contribution to the followership literature. Keywords: faculty, romance of leadership scale, followership, Malaysia, psychometric properties, CFA, institutions of higher learning. INTRODUCTION Organisational research has witnessed predominantly studies on leadership in isolation of followership. This narrow focus has been overturned when Kelley (1992) discloses the missing element in organisational research. This is due to the fact that according to social constructive approach leadership is defined as an experience undergone by followers (Meindl, 1993:97). In other words, individuals are actively involved in constructing leadership rather than leadership being simply what a leader does.
  2. 2. 2 Practically, followers represent if not all, about 99% of the institutional human capital. It is empirically proven that while leaders in most of the organisations contribute to organisational success not more than 20 percent on average, followers are instrumental to the remaining 80 percent (Kelley, 1992). Unfortunately, for decades, organisational literature has been tightly concentrated on investigating leadership theory, traits, styles, leader’s viewpoint and related issues rather than taking into equal significant scale, the follower perspective (Kelley, 1992; Vartanian, Seeley and Trouyet, 2003; Rusher, 2005; Kelley, 2008; Vrba, 2008). Even in a situation where research does investigate the concept of followers as a subject matter, the main purpose is to understand leadership better (Kelley, 2008). In other words, followership is traditionally ignored in the early organisational literature, while the two components are inseparable. Two independent components (Kelley, 1992) but intertwine with common focus “followers and leaders both orbit around the purpose; followers do not orbit around the leader” (Chaleff, 2009:13). This study focuses on evaluating psychometric properties of the romance of leadership scales (RLS) in the context of Malaysian culture. It is also investigates factors influence faculty followership’s perception of institutional leader in selected Malaysian institutions of higher learning. LITERATURE REVIEW The Development of Romance of Leadership Scale This particular RLS was established by Meindl and Ehrlich (1988) with the aim of predicting leadership influence from the followers’ perceptions and experiences. More specifically, RLS defines leadership as the fulfilment of followers’ preconceived ideas as to the qualities and behaviours a leader should exhibit (Eden and Leviatan, 1975). Instruments of RLC were meant to measure the degree to which an individual holds a romantic view of leadership. Recently, RLC is widely used in organisational literature after the wakeup turn from heavily dependency on leadership in the organisational literature. Since the establishment of RLC, researchers have been using the instrument to control for follower bias when studying follower attributions. Furthermore, the validity and reliability of RLC is also cross-culturally tested and evaluated across a dichotomous cultures. An experimental study which sampled 304 undergraduate students found two factor structures and one factor contained two items (Awamleh and Gardner, 1999). Similar findings (three factor-structure) were reached in Chong and Wolf’s (2009) study which sampled 452 management students in New Zealand. In addition, Cho and Meindl (n.d.) found a three factor structure of the same version of the instruments. However, the findings of a study which used RLS on election in the state of California found a two-factor solution (with only 11 items) for the RLS, with the reverse coded items loading on a separate factor (Bligh, Kohles and Pillai, 2005). Given the forgoing findings, it would be imperative to test the factor structure of the RLS in the context of Malaysian culture within the settings of institutions of higher learning. As such, this study tests the following hypotheses:
  3. 3. 3 Hypothesis 1:Components of RLS are best represented by three-factor model in the context of Malaysian institutions of higher learning. Hypothesis 2: Factors of RLS are valid and reliable. Followers’ Perceptions of Organisational Leader Back to four decades or so, organisational literature has started to pay considerable attention to the vital role the followers play in the organisational performance. This approach has led to the emergence of studies of followership styles, factors influencing followership styles, followers’ perceptions of organisational leaders, followers work experience, followers traits and values and so on to mention a few (Kelley, 1992; Vartanian, Seeley and Trouyet, 2003; Rusher, 2005; Kelley, Chaleff and Blumen, 2008; Vrba, 2008; Chaleff, 2009). This study investigates followers’ perceptions with particular reference to organisational leader. It is commonly known that followers judge organisational leaders based on performance reports rather than leaders’ potentials. This is often take place within the context of discussion between followers about their organisational achievements (Calder, 1977). Empirical studies suggest existence of differences in terms of followers’ personal values and work experience associated with a relationship with a leader in a given context, which is deemed to have an impact on followers’ work performance and behaviour (Ehrhart and Klein, 2001; Miller, Butler and Cosentino, 2004; Bjugstad, 2006). Chong et al. (2009) however argued that work performance and behaviour are outcomes. Given a stand as such, researchers suggested constructive values, work experience and age formed followers’ behaviours in relation to their leaders (Chong et al., 2009; Ehrhart et al., 2001). This study however tests only work experience and age for whether or not these variables influence followers’ perception with the organisational leader interchangeability. As such, the study posed the following hypothesises: Hypothesis 3: Older followers perceive greater influence of organisational leader compared to the perception of younger followers of the universities sampled. Hypothesis 4: Experienced followers perceive organisational leader as having less influence as compared with the perception of less experienced followers in the context of the universities sampled. Method Instrumentation The process of investigating followers’ perception of organisation leader in the contect of institutions of higher learning used an online survey questionnaire, which sampled 16 universities across Malaysia. Apart from demographical information of the respondents, the questionnaire comprised 32 items of RLS (Meindl and Ehrlich, 1988; Chong et al., 2009). The study adapted the instruments and slight modifications were made to suit the context of this study especially to the Item 27,
  4. 4. 4 which in the original version states “the president of US can do very little to shape the course of our country” and was modified to “the Prime Minister of Malaysia can do very little to shape the course of our country”. The pool of the questionnaire of the 32 items is built-up of close-ended statements to determine the phenomenon that the researchers aimed to investigate. Furthermore, it comprised two major sections. Section one, formed 7 items, which requested respondents to indicate general demographic information. Thus, the items asked and collected characteristic information such as gender, age group, race group, marital status, work experience, monthly income and university affiliated to. Section two, comprised 32 items of RLS used in the study. A 7-point Likert type scale was employed to collect the data. The scale is interpreted as: 1=very strongly agree, 2=strongly agree, 3=agree, 4=neither agree nor disagree, 5=disagree, 6=strongly disagree and 7=very strongly agree. Since items of RLS contained 15 negatively worded statements, reverse coding procedure was performed. During the reverse coding, a value of 7 =very strongly disagree, 6 =strongly disagree, 5 =disagree, 4 =neither disagree nor agree, 3 = agree 2=strongly agree and 1 =very strongly agree were selected. This recoding was used instead of 1 =very strongly agree, 2 =strongly agree, 3 =agree, 4 =neither agree nor disagree, 5 =disagree, 6=strongly disagree and 7 =very strongly disagree. This reverse coding procedure covered all 15 negative statements of RLS scale. Participants A total of 16 institutions of higher learning were sampled to represent the population of this study. 150 respondents were drawn from each to represent population of academic staff in their respective universities. Participants were drawn from universities such as International Islamic University Malaysia (IIUM), University Malaya (UM), University Putra Malaysia (UPM), University Kebangsaan Malaysia (UKM), Universiti Teknologi Malaysia (UTM), Universiti Utara Malaysia (UUM), Universiti Sains Malaysia (USM), Univerisiti Teknologi Mara (UiTM), Universiti Pendidkan Sultan Idris (UPSI), Universiti Malaysia Sabah(UMS), Universiti Tun Hussein Onn Malaysia (UTHOM), Universiti Pertahanan Nasional Malaysia (UPNM), Universiti Teknikal Malaysia Melaka (UTMM) and Universiti Malaysia Sarawak (UNIMAS) academic staff. The sample size (2400) was determined through the confidence interval of 95% and margin of error at ± 3.5% as suggested by (Ferguson, 1981; Vockell and Asher, 1998). Respondents were randomly selected and voluntarily participated through online in answering the survey questionnaires. Surveys questionnaires were e-mailed to the respondents through their staff directory e-mail addresses. They were asked to express their level of agreement or disagreement to the survey instruments. Of the 2400 questionnaires e-mailed, 436 questionnaires were successfully returned. However, two incomplete questionnaires were discarded, which resulted into a total of 434 questionnaires being retained and analysed. This indicated a ratio of (18.08%) of response rate which is acceptable in online survey. Data Analysis Data screening processes Since instruments RLS were adopted and used in different cultural context, two statistical analyses were performed. First, principal component analysis (PCA) using SPSS version 17.0 software to test factors underlie the constructs understudy. This
  5. 5. 5 was then followed by application of structural equation modelling (SEM) through AMOS version 18.0 software, using confirmatory factor analysis (CFA) to test confirm constructs (Arbuckle, 2008). Finally, assessment of construct validity through average variance extracted (AVE) and construct reliability through composite reliability index (CRI) were evaluated. Results of the Study Respondents’ demographic information The accuracy of the dataset is often assessed through code and value cleaning using descriptive frequencies. The results of the frequencies indicated that (n =168, 38.5%) of the respondents were male academic staff, while (n =266, 61.3%) were female academic staff. With reference to respondents’ age group, the analysis showed that the majority of respondents were as young academic staff (age between 29-39 years) n =296, 68.2 %). In addition, respondents who were between 40-69 years who identified as old academic staff were small in terms of academic staff participation (n = 138, 31.8%). The analysis demonstrated that the majority (n= 366, 84.3%) of the respondents, who participated in the study were identified as experienced academic staff (6-10 years teaching experience). Followed by participants, who identified themselves as inexperience academic staff (1-5 years) n =68, 15.7%). Pertaining to academic staff monthly income, the results of Table 1 showed that the majority (n=227, 52.3%) of academic staff surveyed were earning a monthly income above RM 6000. Table 1 also identified n=67, 15.4% of the respondents as academic staff who earn a monthly income ranged between RM 3001 – RM 4000. Furthermore, n=56, 12.9% of the respondents were identified as academic staff who earn RM 4001 – 5000 and RM 5001 respectively. In addition, the least category of respondents according to the scale of respondents’ monthly income, was the category of academic staff who earn between RM 2000-3000 (n=28, 6.5%). Regarding respondents’ race group, the analysis showed that the majority of the participants (n = 363, 83.6 %) were Malay academic staff. Followed by foreign academic staff (n =32, 7.4 %). Chinese academic staff (n =21, 4.8%) participated in the study. In addition only (n=18, 4.1%) Indian academic staff have participated in this study. Regarding respondents’ marital status, the data yielded that majority (n = 364, 83.9 %) of academic staff surveyed were married. This was followed by academic staff who were classified themselves as unmarried (n =51, 11.8%). Some (n =10, 2.3 %) of the participants were classified as themselves as divorcees and only (n =9, 2.1%) of them were identified as widowed. Finally, responses of academic staff surveyed indicate that the majority (n=60, 13.8%) of the respondents participated in this study were from UiTM. Table 1 identifies n=47, 10.8% the percentages of academic staff participated in the study from IIUM. Furthermore, n=41, 9.4% was accounted for UKM academic staff participated in the study. Similarly, n=33, 7.6% of the respondents were accounted for academic staff from UPM. Interestingly, moderate percentages (n=31, 7.1%) and (n=31, 7.1%) of respondent participations were captured from both UTMM and UMSR academic staff population respectively. Moreover, the table revealed that (n=30, 6.9%) of the respondents participated in the study were academic staff of UPSI. The table also identified that (n=26, 6.0) of respondents were from USIM. In addition, both (n=22, 5.1%) and (22, 5.1%) were identified as academic staff from
  6. 6. 6 UPNM and UTM respectively. Nonetheless, (n=20, 4.6%) of academic staff affiliated themselves to USM. The analysis further presented both (n=18, 4.1%) and (n=18, 4.1%) as academic staff from UTHOM and UM. Also, (n=12, 2.8) were identified as academic staff from UMS who responded to the survey. Finally, academic staff of UMK were classified as the least on the participation scale (n=11, 2.5%). Table 1 depicts the detailed results. Table 4 Frequency and Percentages of Respondents' Demographic Variables Variable Frequency (N) % Gender Male 168 38.5 Female 266 61.3 Total 434 100 Age Group Young 296 68.2 Old 138 31.8 Total 434 100 Race Group Malay 363 83.6 Chinese 21 4.8 Indian 18 4.1 Others 32 7.4 Total 434 100 Marital Status Unmarried 51 11.8 Married 364 83.9 Divorcee 10 2.3 Widowed 9 2.1 Widowed 7 1.2 Total 434 100 Working Experience Less experienced 68 15.7 experienced 366 84.3 Total 434 100 Monthly Income RM 2000-3000 28 6.5 RM 3001-4000 67 15.4 RM 4001-5000 56 12.9 RM 5001-6000 56 12.9 RM above 6001 227 52.3 Total 434 100 University UM 18 4.1 UKM 41 9.4 UPM 33 7.6 UiTM 60 13.8 UUM 12 2.8 USM 20 4.6 UTM 22 5.1 IIUM 47 10.8 UPSI 30 6.9 UMS 12 2.8 UTHOM 18 4.1 UMK 11 2.5 UPNM 22 5.1 UTMM 31 7.1 UMSR 31 7.1 Total 434 100
  7. 7. 7 Principal Components Analysis (PCA) Prior to PCA, reliability analysis was performed, to assess internal consistency of the instruments. The results of reliability analysis revealed overall Cronbach’s alpha of .86. This indicated a substantial internal consistency between individual items, thus the items have positive covariance meanwhile the alpha is very close to desired cutoff. Furthermore, this finding yielded that there is a very high reliability level of internal consistency of the items, which indicated that the instruments were suitable therefore; its results would be reliable in association with internal consistency build up. A PCA was conducted for the constructs. The analyses involved an iterative process to reach the final solution, whereby items that did not contribute significantly and practically to the extracted factors were automatically discarded. Furthermore, factors with eigenvalues of 1 or greater were considered as good factors, and therefore retained. Given a rule of thumb as such, a number of factors were extracted from the pool of items. The correlation matrix table yielded more than 2 correlations were greater than 0.30. The measures of sampling adequacy (MSA) requirement of (0.50 or greater) were also satisfied. Thus, the anti-image correlation ranged between .59 and .96. Furthermore, all communalities were greater than .50 (ranged between .53 to .76), which indicates fulfilment of this requirement. Moreover, the analysis revealed 3 interpretable factors with eigenvalues greater than one. The extracted factors accounted for 51.17% of variance explained in the constructs analysed. Interestingly, the degree of inter-correlation among the items also reached satisfied level. Bartlett’s test of Sphericity was statistically significant χ2 (496) = 9020.179, ρ≤.001, KMO = .80. Table 2 displays results the details. Table 2 Interchangeability and Influence Scale; Factor Loading, Measures of Sampling Adequacy and reliability. Construct Eigenval ues %Varianc e Indicator Loadi ng MSA α Interchangeabilit y of Leader 8.27 25.84 20. When an organisation is performing poorly, the first place one should look to is its leaders. .87 .71 .87 21. The process by which leaders are selected is extremely important. .86 .68 19. It’s probably a good idea to find something out about the quality of top level leaders before investing in an organisation. .86 .72 28. Leadership qualities are among the most highly prized personal traits I can think of. .77 .72 6. Sooner or later, unqualified leadership at the top will show up in decreased organisational performance. .76 .69 8. High-versus low quality leadership has a bigger impact on a organisation than a favorable versus unfavorable organisational environment. .74 .61 5.The great amount of time and energy devoted to choosing a leader is justified, because of the important influence that person is likely to have. .73 53 1. When it comes right down to it, the quality of leadership is the single most important influence on the functioning of an organization. .71 .63 24. When the top leaders are good, the organisation does well; when the top leaders are bad, the organisation does poorly. .70 52 4. Anybody who occupies the top level leadership positions in an organisation has the power to make or break the organisation. .67 51
  8. 8. 8 9. It is impossible for an organisation to do well unless it has high-quality leadership at the top. .62 .70 17. The connection between leadership and overall organisational performance is often a weak one. .61 .53 14. With a truly excellent leader, there is almost nothing that an organisation can’t accomplish. .56 53 12. An organisation is only as successful or as unsuccessful as its leaders. .53 65 25. There’s nothing as critical to the “bottom line” performance of an organisation as the quality of its top level leaders. .52 .66 Influence of a leader 5.65 17.65 31. One leader is as good or as otherwise as the next. .75 .79 .82 27. The Prime Minister of the Malaysia can do very little to shape the course of our country. .74 .70 16. Top level leaders make life and death decisions about their organisations. .72 .73 11. Many times, it doesn’t matter who is making decisions at the top, the fate of an organisational is not in the hands of its leaders. .71 .75 26. In many cases, candidates for a given leadership position are pretty much interchangeable with one another. .66 .66 22. So what if the organisation is doing well; people who occupy the top level leadership positions rarely deserve their high salaries. .65 .75 23. In comparison to external forces such as the economy, government regulations, etc., an organisation’s leaders can have only a small impact on its performance. .63 .73 18. Many times, organisational leaders are nothing more than figureheads like the King and Queen of England. .61 .70 2. The majority of business failures and poor organisational performances are due to factors that are beyond the control of even the best leaders. .60 .69 30. There are many factors influencing an organisation’s performance that simply cannot be controlled by even the best of leaders. .58 .71 10. When faced with the same situation, even different top-level leaders would end up making the same decisions. .54 .70 7. Luck has a lot to do with whether or not organisational leaders are successful in making their organisations successful. .53 .65 13. You might as well toss a coin when trying to choose a leader. .52 .60 Parallel Analysis (PA) A further critical analysis on how many factors are accurately extracted through the application of PCA was performed. This analysis is critical especially when the analysis involves composing multiple scale items into a single score. This is because determination of the number factors to retain is the most crucial issue researchers encounter when using PCA. One of the effective methods proposed in determining the significance of principal components, is the parallel analysis (PA), which consistently proven the accuracy of determining the threshold for significant components, variable loadings and analytical statistics when decomposing a correlation matrix. PA involves comparing the size of eigenvalues obtained from PCA analysis with those obtained from a randomly generated data of the sample size (Horn, 1965; Pallant, 2007; Franklin, Gibson, Robertson, Pohlmann and Fralish, 1995). The PA threshold denotes if eigenvalues of extracted factors through application of PCA are greater than the corresponding eigenvalues of PA, the factors are retained. However, if the otherwise is true, then the factors fail to fulfil the criteria are rejected (Franklin et al., 2007; Horn, 1965; Pallant, 2007).
  9. 9. 9 A Monte Carlo PCA for PA software was used to perform PA analysis. As such, the number of variables (in this case 32), the number of subjects (in this case 434) and the number of replications (specifically 100) were calculated. The analysis revealed that the eigenvalues of the first two factors extracted through PCA were greater than their corresponding eigenvalues of the PA, and were thus significant at p = 0.05. Therefore, these factors were retained for further interpretation. However, the remaining third factor was rejected, since its eigenvalues of PA was greater than the initially eigenvalue extracted through PCA (Franklin et al., 2007; Horn, 1965; Pallant, 2007). These results did not support Hypothesis 1 with the finding that components of RLS are best represented by three-factor model in the context of Malaysian institutions of higher learning Table 3 depicts the details. Table 3 Comparison of PCA and PA Eigenvalues for Data from RLS Scale Component no. Actual eigenvalue from PCA Criterion eigenvalue from PA Decision 1 8.25 1.54 Accept 2 5.65 1.47 Accept 3 1.40 1.41 Rejected Construct Reliability and Validity of the Measures A construct validity analysis was performed to assess psychometric properties of the items indexed factors that constituted interchangeability of leader and influence of a leader. Two rigorous tests (composite reliability index (CRI) and average variance extracted (AVE)) were conducted to assess construct validity, convergent validity and composite reliability of the instruments. The first test conducted was a CRI, which indicates how well each structure has been described by the observed variables was performed to establish more accurate reliability values. This test was conducted for the fact that researchers (Raykov, 1998; Zinbarg, Revelle, Yovel and Li, 2006; Gordon, 2008) cautioned heavily dependency on Cronbach’s Alpha to proclaim reliability of a construct due to its limitations. According to Ranganathan, Dhaliwal and Teo (2004), while the traditional reliability measure of Cronbach’s Alpha assumes equal weight for the items measuring the construct in which it is influenced by the number of items in the construct, the CRI estimates rests on the actual readings to compute the factor scores which is better indicator of internal consistency reliability. Among issues of which researchers pinpointed in relying on coefficient alpha per se are (1) “it does not make allowances for correlated error of measurements nor does it treat indicators influenced by more than one latent variable” Bollen (1989: 221), (2) it may over or underestimate reliability Raykov (1997, 1998) and (3) it is somehow regarded as a lower bound estimate of reliability. Furthermore, Fornell and Larcker (1981) argue that the CRI is a more advanced criterion of estimation compared to coefficient alpha. As such, according to (Raykov and Shrout, 2002), a method that would allow researchers to evaluate factor indicator scores more accurately is needed. Given the foregoing reasons, the CRI test was performed to rigorously examine the reliability of factor loadings extracted through PCA applications. The conventional cutoff value for CRI is 0.70 or greater (Fornell and Larcker, 1981; Bagozzi and Burnkrant, 1985). According to Fornell and Larcker (1981) and Hair et
  10. 10. 10 al. (2010), the evidence of construct validity is established if the CRI of each factor is 0.70 or greater and AVE of each factor is 0.50 or greater. The second test performed is AVE, which measures the amount of variance that captured by the construct in association with the amount of variance due to the measurement error (Fornell and Larcker, 1981). In other words, AVE represents the percentage of the total variance of a measure represented or extracted by the variance due to the construct, as opposed to being due to error. The AVE is a method that evaluates convergent validity and discriminant validity of a given construct (Fornell and Larcker, 1981; Bagozzi, 1981). It is according Fornell and Larcker (1981), calculated as the square root of the average communality. As such, it is proposed that a benchmark of equal to or greater than .50 should be attained in order to establish evidence of construct and convergent validity of a given construct (Fornell and Larcker, 1981; Bagozzi, 1981; Hair et al., 2010). It is worth noting that only convergent and construct validity were calculated in this section. Interchangeability of Leader and the Influence of a Leader factors The AVE analysis of interchangeability and influence factors revealed an adequate estimation of AVE for interchangeability 0.51 and Influence 0.50). The estimates (0.51 and 0.50) had fulfilled the recommended value of AVE (Fornell and Larcker, 1981). This finding demonstrates evidence of convergent validity for the two factors. Moreover, the construct validity of the factors was further evaluated through application of CRI method. Results of CRI revealed substantial evidence of construct validity (0.93 through 0.89). These results fulfilled Hair et al.’s (2010) guidelines where CRI .70 percent. Therefore, evidence of construct and convergent validity were established. These results confirmed Hypothesis 2 with the finding that factors of RLS are valid and reliable. Table 4 shows the details. Table 4 Construct Reliability and Validity of Interchangeability and Influence Factors Construct α AVE CRI Interchangeability of Leader 0.87 0.51 0.93 Influence of a Leader 0.82 0.50 0.89 Note: Composite reliability Index (CRI) formula = )² /( )²+ Average variance extracted (AVE) formula = )²/( )² + Construct Validity Since the test of psychometric properties of the constructs was one of the principal aim this of study, a more rigorous structural equation modelling-based approach of CFA is needed to validate the two factors extracted through application of PCA. According to Byrne (2010), CFA would be the best choice to validate instruments that have been fully developed and their factor structures were validated. RLS inventory as widely used to meet this criterion. To assess the validity of the 28-item measurement models, the analysis relied on a number of inferential fit indices, which included the (a) Chi-square (χ2 ) which is a minimum value of the discrepancy between the observed data and the hypothesised model, thus it determines whether a relationship exists between two categorical variables (CMIN), (b) Degrees of Freedom (df), measures the amount for mathematical information available to estimate model parameters, (c) Comparative
  11. 11. 11 Fit Index (CFI), which conceptualises goodness of fit by comparing an existing model with a null model, which assumes that the latent variables in the model are not correlated, (d) Turker-Lewis Index (TLI) compares a proposed model against a null model and (e) Root Mean Square Error of Approximation (RMSEA), which corrects the tendency of the model χ2 statistics to reject any specified model with a sufficient large sample. Byrne (2010) argues that the appropriate values for CFI range from 0-1. Conventionally, CFI should be .90 or greater as criterion value indicative of satisfactory model fit (Chueng and Rensvold, 2002). The value of RMSEA of .08 or less shows a reasonable error of estimation. According to Brown and Cudeck (1993), RMSEA ≤0.05 indicates close fit, RMSEA ≥0.05 to ≤0.08 indicates fair fit and RMSEA ≥0.08 to 0.10 shows poor fit. This study uses RMSEA with its point estimate and associated confidence interval (CI) through method of lower and upper bound of the CIs in combination with 0.05 and 0.10 as the cut-off values used to validate the extent to which the hypothesised models fit the data. When the RMSEA with its CIs are used, a given model is rejected if the lower bound of the CI is greater than the value of 0.05. Similarly, a given model is rejected if the upper bound of the CI is greater than the value of 0.10 (Chen, Curran, Bollen, Kirby and Paxton, 2008). More recently, Hu and Bentler as cited in Byrne (2010) suggested that values below .06 as indicative of good fit. Validity of the Measures of Interchangeability of Leader A confirmatory factor analysis (CFA) with maximum likelihood was used to assess interchangeability factor. CFA allows an overall assessment of the validity of indicators of the factor. The following measures; Chi-square (χ2 ), Comparative Fit Index (CFI), Turker-Lewin Index (TLI) and Root Mean Square Error of Approximation (RMSEA) as the cutoff values used to validate the item measured this factor. Also, RMSEA RMSEA with its point estimate and associated confidence interval (CI) through method of lower and upper bound were checked. The results of the analysis demonstrated poor fit statistics; χ2 (90) = 626.480, CFI = .86, TLT = .84, RMSEA = 0.117 and CMIN/DF = 6.96. Since the study sought a better fit, post hoc model modifications were checked to develop a better fit and a more parsimonious model. The model was re-estimated and three inter-correlations among 6 errors were freed based on the suggestions of modification indices (MIs). More specifically, the following connections were established; error 3 (v14) and error 4 (v17), error 4 (v17) and error 10 (v8) and error 6 (v4) and error 11 (v6). These connections were allowed to co-vary to reduce the total amount to 626.480 χ2 , and therefore ameliorate the fit indices. These connections were supported methodologically through the use of AMOS and theoretically owning to the fact that the two elements of the measurement errors were correlated showing commonalities among pairs of observed variables. Prior to that three items were eliminated doe low and cross loading issues. As such the model improved χ2 (51) = 197.773, CFI = .95, TLT = .94, RMSEA = 0.08 and CMIN/DF = 3.87. Moreover, the RMSEA with its CI of the lower and upper bound also fell within the desired zone; LO .04 and HI .09 providing additional evidence of model acceptance (Chen et al., 2008). Figure 1, depicts the details.
  12. 12. 12 Figure 1: Revised Model of Interchangeability factor The parameter estimates of the interchangeability factor were free from offending values. All path coefficients were statistically significant at .005 levels, and were of practical important, since the smallest value of the standardised path coefficient was 0.50. Moreover, the squared multiple correlations (SMC), which indicate how well the observed variables serve as measures of the latent variables, and provide evidence of the reliability of the indicators were also investigated. It is worth noting that the values of SMC of the interchangeability model had fulfilled the requirement (.25) or greater for all indicators. The values ranged from .26 (v14) to .80 (v20). This provided substantive values to explain the variance in the 12 indicator variables of interchangeability construct. Influence of a leader A second CFA was performed for influence of a leader factor. The results showed that the overall fit of the model was χ2 (65) = 654.839, p = 0.001, which was statistically significant, indicating an inadequate fit between the covariance matrix of the observed data and the implied covariance matrix of the model. The RMSEA fell beyond the range of acceptable value (RMSEA .05 to .08). other fit indices also did
  13. 13. 13 not satisfy the requirements of good fit CFI = .73, TLI = .68, L0 = .13, HI = .15 and CMIN/DF =10.07. All these indices indicate poor model fit. Since the study sought a better fit, post hoc model modifications were checked to develop a better fit and a more parsimonious model. As such, the model was re-estimated and involved two modifications. First, three items were eliminated due to low loading issues. Second, six inter-correlations among 12 errors were freed based on the suggestions of modification indices (MIs). More specifically, the following connections were established; error 2 (v7) and error 4 (v30), error 2 (v7) and error 7 (v23), error 2 (v7) and error 9 (v26), error 6 (v18) and error 11 (v16), error 7 (v23) and error 11 (v16) and error 10 (v11) and error 11 (v15). These connections were allowed to co-vary to reduce the total amount to 654.839 χ2 , and therefore ameliorate the fit indices. These connections were supported methodologically through the use of AMOS and theoretically owning to the fact that the two elements of the measurement errors were correlated showing commonalities among pairs of observed variables. As such the model improved χ2 (29) = 126.793, CFI = .93, TLT = .90, RMSEA = 0.08 and CMIN/DF = 4.37. Moreover, the RMSEA with its CI of the lower and upper bound also fell within the desired zone; LO .05 and HI .08 providing additional evidence of model acceptance (Chen et al., 2008). Figure 2 presents detailed results. Figure 2: The Revised Model of Influence of a Leader In addition to this, the parameter estimates were also examined and were found to be statistically significant and practically important as shown in Figure 4.5. The loadings ranged from .41 (v7) to .74 (v31). They were free from any offending estimates and showed logical direction. Also, the squared multiple correlations
  14. 14. 14 (SMC), which indicate how well the observed variables serve as measures of the latent variables, and provide evidence of the reliability of the indicators were also investigated. The values of SMC scaled from 0 to 1 (Schumacker and Lomax, 2004). It is worth noting that the values of SMC of the influence of a leader model (other than v7) had fulfilled the requirement of (.25) or greater for all indicators. The values ranged from .25 (v181) to .55 (v31). This provided reasonable values to explain the variance in the 10 observed variables of influence of a leader model. A t-test analysis was conducted to investigate the perceptions of followers towards their organisational leader across their work experience. The results indicate that the experienced followers hold a stand that interchageability of organisational leader will not have much effect on organisational performance, as compared with the perceptions of less experienced followers in the institutions of higher learning sampled M=49.29, D=9.97 for experienced followers and M=47.73, D=11.27 for inexperienced followers respectively. However, this difference in followers’ perceptions is hardly claimed in isolation of performing independent samples t-test, which either confirm or otherwise the degree of significance in the difference. In other words, independent samples t-test table must be examined. The F statistics which determines whether the equality of variance assumption is satisfied or otherwise was statistically insignificant F = 0.064, p ≥ 0.05, therefore, as the variances are not significantly different, we can accept the equal variances assumption and use the bottom line values, since Hypothesis 4 is directional. The test of t statistic revealed that the difference obtained in the mean scores t (407) = 3.454, p ≤ 0.001 is significant. This result confirmed Hypothesis 4 with the finding that experienced followers perceive organisational leader as having less influence as compared with the perception of less experienced followers in the context of the universities sampled. Finally, the study tested the level of influence of a leader as perceived by followers according to their age. The results indicate that younger followers in the institutions higher learning sampled believe that organisational leader exhibits great influence on organisational performance compared to the perceptions of older followers M=32.2, D=17.61 for young followers and M=27.42, D=10.46 for old followers respectively. However, this difference in followers’ perceptions is hardly claimed unless independent samples t-test, which either confirm or otherwise the degree of significance in the difference is tested. In other words, independent samples t-test table must be examined. The F statistics which determines whether the equality of variance assumption is satisfied or otherwise was statistically insignificant F = 0.26.29, p ≥ 0.05, therefore, as the variances are not significantly different, we can accept the equal variances assumption and use the bottom line values, since Hypothesis 3 is directional. The test of t statistic revealed that the difference obtained in the mean scores t (407) = 4.46, p ≤ 0.001 is significant. This result rejected Hypothesis 3 with the finding that younger followers perceived greater influence of organisational leader compared to the perception of older followers of the universities sampled.
  15. 15. 15 Discussion The findings of the present study have contributed to the literature of leadership in several ways. First, the results replicated the psychometric adequacy of RLS instruments, especially this is the first study that sample academic staff in validating the scales. The measures seem sufficient to represent the ongoing concerns of RLS instruments in different cultural context. As well as providing and evidence that influence and interchangeability of leader are two distinct factors. This finding corresponded with the findings of (Awamleh and Gardner, 1999; Felfe, 2005; Shyns et al., 2007; Chong, 2009). Second, the results confirmed that the two factor structure of the RLS (Awamleh and Gardner, 1999; Bligh, Kohles and Pillai, 2005). The results also revealed adequacy of two factor model, which can be used as a means to predict followers’ perceptions of organisational leader on ameliorating academic staff performance. However, this two factor solution contradicts with the study of Chong and Wolf (2009) and Cho and Meindl (n.d.), which found three factor solution in relation to RLS underlying factors. Thirdly, the results of the study suggest that with age, older followers feel that organisational leader has less influence on the organisational performance. This finding suggests more friendly leader-follower relations that leads to organisational stability. Finally, the findings of the study indicate the existence of relationship between the influence of a leader and the work experience in the organisational context. This finding is consistent with findings of Miller et al. (2004) and Chong and Wolf (2009) who found that performance of more experienced followers was independent of their perceived relationship with their leaders. Limitation and the Suggestions for Future Study This study used only online survey questionnaire to sample universities involved in this study, future studies should consider self-administered questionnaire as well. The study sampled only academic staff, but future study should involve administrative staff as well. It also could be a good idea for a mixed methods to be employed in future studies.
  16. 16. 16 REFERENCES Bagozzi, R. P. & Burnkrant, R. E. (1985). Attitude organisation and the attitude behaviour relation: a reply to Dilon and Kumar. Journal of Personality and Social Psychology, 49, 47-57. Bollen, K. A. (1989). Structural equation models with latent variables. New York: John Wiley and Sons. Brown, M. W. & Cudeck, R. (1993). ‘Alternative ways of assessing model fit’, in Bollen, K. A. & Long, J. S. (Eds), Testing Structural Equation Models, Sage Newbury Park, CA. Cheung, G. W. & Rensvold, R. B. (2002). Evaluation goodness-of-fit indexes for testing measurement invariance. Structural Equation Modelling, 9, (2), 233- 255. Chen, F. Curran, P.J., Bollen, K.A., Kirby & J. Paxton, P. (2008). An empirical evaluation of the use of fixed cutoff points in RMSEA tests statistic in structural equation models. Sociological Methods and Research, 36, (4), 462-494. Calder, B.J. (1977). An attribution theory of leadership. in Staw, B.M. & Salancik, G.R. (Eds), new directions in organizational behaviour. Chicago: St Clair. Chaleff, I. (2009). The courageous follower: Stand up to and for our leaders (3rd ed.). San Francisco: Berrett-Koehler Publisher’s Inc. Cho, G. S. & Meindl, J. R. (undated). The romance of leadership and work ethics: Occidental and oriental variations. Unpublished manuscript. Eden, D. & Leviatan, U. (1975). Implicit leadership theory as a determinant of the factor structure underlying supervisory behavior scales. Journal of Applied Psychology, 60(6), 736-741. Ehrhart, M.G. and Klein, K.J. (2001). Predicting follower’s preferences for charismatic leadership: the influence of follower values and personality, Leadership Quarterly, 12, pp. 153-79. Franklin, S. B., Gibson, D. J., Robertson, P. A., Pohlmann, J. T. & Fralish, J. S. (1995). Parallel analysis: a method for determining significant principal components. Journal of Vegetation Science, 6, 99-106. Fornell, C. & Larcker, D. F. (1981). Evaluating structural equation models with unobservable variables and measurement error, Journal of Marketing Research, 18 (1), 39-50. Gordon, R. (2008). A note of using alpha and stratified alpha to estimate reliability of a test composed of item parcels. British Journal of Mathematics and Statistical Psychology, 61, (2), 515-525. Hair, F., Anderson, E., Tatham, L. & Black, C. (2010). Multivariate data analysis (7th edition). New Jersey: Pearson Prentice Hall. Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179-185. Kelley, R. E. (1992). The power of followership: How to create leaders people want to follow and followers who lead themselves. New York: Bantam Doubleday Dell Publishing Group Inc. Kelley, R. E. (2008). “Rethinking followership”. In Riggio, R. E., Chaleff, I. & Lipman- Bluman (eds.), The art of followership: How great followers create leaders and organisations (pp. 5-15). San Francisco: Jossey Boss.
  17. 17. 17 Meindl, J. R. & Ehrlich, S. B. (1988). Developing a Romance of Leadership scale. Proceedings of the Eastern Academy of Management, 30, 133 – 135. Meindl, J. R. (1993). Reinventing leadership: A radical, social psychological approach. In J. K. Murnighan (Ed.), Social Psychology in Organizations (pp. 89-118). Englewood Cliffs, NJ: Prentice Hall. Miller, R.L., Butler, J. and Cosentino, C.J. (2004). Followership effectiveness: an extension of Fiedler’s contingency model. Leadership & Organization Development Journal, 25(3), 4. Pallant, J. (2007). SPSS survival manual: a step by step guide to data analysis using SPSS for windows third edition. England: Open University Press, McGraw-Hill Education. Riggio, R. E., Chaleff, I. & Blumen, J. L. (2008). The art of followership: How great followers create great leaders and organisations. Singapore: Warren Bennis Signature Books. Rusher, T. R. (2005). A case study of followership in the charter schools of Washington County, Michigan. Unpublished doctoral dissertation, Capella University. U.S.A. Raykov, T. (1998). Coefficient alpha and composite reliability with interrelated non- homogeneous items. Applied Psychological Measurement, 22, 375-385. Ranganathan, C., Dholiwal, S. J. &Teo, S. H. T. (2004). Assimilation and diffusion of web technologies in supply chain management: an examination of key drivers and performance impact. International Journal of Electronic Commerce, 9, (1), 127-161. Raykov, T. (1998). Coefficient alpha and composite reliability with interrelated non- homogeneous items. Applied Psychological Measurement, 22, 375-385. Raykov, T. & Shrout, P. E. (2002). Reliability of scales with general structure: point and internal estimation using a structural equation modelling approach. Structural Equation Modelling, 9, (2), 195-212. Vartanian, L., Seeley, T. & Trouyet, V. (2003). Leadership rethought. Paper presented at Alliant International University’s Second Leadership Conference. San Diego, California. Verba, A. M. (2008). Relationship between followership behaviour and perception of effective leadership characteristics in adult teachers. Unpublished doctoral dissertation, Capella University. U.S.A. Vockell, E. L. & Asher, J. W. (1998). Educational research (2nd ed.). New Jersey: Prentice Hall Inc. Zinbarg, R. Revelle, W., Yovel, I. & Li, W. (2006). Cronbach’s Alpha, Revelle’s beta and McDonald’s Omega H: their relations with one another and two alternative conceptualisations of reliability. Psycometrica, 70, (1), 123-133.
  18. 18. 18
  19. 19. 19

×